Single-cell NAD(H) levels predict clonal lymphocyte expansion dynamics

Abstract

Adaptive immunity requires the expansion of high affinity lymphocytes from a heterogenous pool. While current models explain this through signal transduction, we hypothesized that antigen affinity tunes discrete metabolic pathways to license clonal lymphocyte dynamics. Herein, we identify nicotinamide adenine dinucleotide (NAD) biosynthesis as a biochemical hub for the T cell receptor affinity-dependent metabolome. Through this central anabolic role, we found that NAD biosynthesis governs a quiescence exit checkpoint, pacing proliferation. Normalizing cellular NAD(H) likewise normalizes proliferation across affinities and enhancing NAD biosynthesis permits expansion of lower affinity clones. Furthermore, single-cell differences in NAD(H) could predict division potential for both T and B cells, prior to the first division, unmixing proliferative heterogeneity. We believe this supports a broader paradigm in which complex signaling networks converge on metabolic pathways to control single-cell behavior.

One-Sentence Summary:

Single-cell differences in NAD(H) are deterministic of cell cycle rates and underly heterogenous lymphocyte proliferation.

Introduction

Protective adaptive immune responses require the selective expansion of high affinity antigen-specific lymphocyte clones. The affinity of the antigen receptor (TCR or BCR) to cognate antigen determines the magnitude of the primary response as well as the rate at which naive lymphocytes exit quiescence (1–3). During this exit from quiescence, naive lymphocytes reprogram their transcriptome and proteome, while at the same time remodeling cellular metabolism to meet the demands of rapid proliferation (3–11). Multiple signaling and transcriptional effectors sensitive to activating signal strength are known to converge on metabolic reprogramming (e.g., Myc, mTOR, IRF4) (1, 2, 8, 11–29). Despite this, it remains unclear whether components of the resulting metabolome are similarly sensitive to affinity and work in concert with canonical signaling and gene networks to coordinate the rapid changes in cell cycle biology controlled by antigen receptors.

Herein, we test the hypothesis that TCR signaling relays antigen affinity to a discrete set of metabolic pathways responsible for supporting clonal dynamics. Using both unbiased and targeted approaches, we identify NAD biosynthesis as a biochemical hub for TCR affinity-driven metabolic remodeling. We find that NAD biosynthesis controls a quiescence exit checkpoint, in which TCR affinity tunes the magnitude of cellular NAD increase and the concurrent rate of cell cycle entry. We demonstrate that while mitogenic signaling sets maximal lymphocyte proliferative potential, cellular NAD levels and the biochemistry it facilitates control the rate cells divide within these bounds. Highlighting NAD biosynthesis as a central node for TCR-driven T cell expansion dynamics, we show that normalizing cellular NAD(H) levels is sufficient to normalize proliferation rates between low and high affinity ligand stimulated cells and enhancing NAD biosynthesis can permit the expansion of lower affinity clones in vivo. Finally, we find that this relationship between mitogenic signaling, NAD biosynthesis, and quiescence exit holds true across T and B lymphocytes. Altogether, these data illustrate how mitogenic signaling converges on cellular metabolism and that regulation of a single, central metabolic process that scales with signal strength ensures cell cycle demands are met. Accordingly, differences in this one parameter are sufficient to predict how individual lymphocytes behave, both for their metabolic and proliferative capacity. We believe this provides a broader model for how single-cell metabolic variation drives cellular heterogeneity amongst populations responding to the same signal.

Results

T cell receptor ligand affinity coordinates the kinetics of cell cycle progression and anabolic programming in CD8⁺ T cells

The strength of the activating TCR signal is known to dictate cell cycle entry kinetics and impact differentiation; however, the extent to which metabolic programming is similarly tuned by the TCR remains to be fully elucidated. We utilized the OT-I TCR transgenic model, in which all T cells encode a TCR with high affinity for the peptide SIINFEKL (N4), with well defined, lower affinity variants (e.g. SIIGFEKL, “G4”) (30–34). Consistent with previous findings, stimulation of OT-I T cells with the higher affinity N4 peptide resulted in more rapid cell cycle entry Myc expression, and proliferation compared to the lower affinity G4 peptide (fig. S1, A–E) (10, 19, 33, 35–38).

We then assayed whether the kinetics of downstream metabolic processes were similarly controlled by peptide affinity. At both 24 hours or immediately post-stimulation, N4-stimulated cells more rapidly reached higher basal glycolytic and respiration rates than G4-stimulated cells (Fig. 1A and B). Thus, antigen affinity coordinates cell cycle kinetics with the rate and magnitude of metabolic remodeling.

Figure 1: NAD biosynthesis is a core component of the TCR affinity-dependent metabolome.

(A) Seahorse analysis of 24 hour peptide stimulated (G4 or N4) OT-I T cells. (B) Seahorse analysis was run on OT-I T cells prior to and after stimulation with either 1μg/mL of N4 or G4 peptide (added via analyzer port). LC-MS analysis of 24 hour peptide stimulated (G4, T4, Q4, or N4) OT-I T cells. C) PCA plot generated from metabolite abundance. (D) Heatmap of named peaks plotted as log10 fold change of normalized values above G4. (E) Pathway Analysis using all metabolites in Cluster 1. Shown is the percentage of significantly matched pathways corresponding to each metabolite. (F) Cellular NAD quantification by cycling assay of OT-I T cells stimulated with N4 or G4 peptide 24 hours post-activation. (G) Schematic of NAD biosynthesis pathways. (H) Cellular NAD quantification by cycling assay in peptide stimulated OT-I T cells treated with FK866 or vehicle control, 24 hours post-activation. (I) Cellular NAD quantification by cycling assay in peptide stimulated OT-I treated with vehicle control or FK866 in the presence of Trp, NA, or NR. (J) Nampt transcript expression after 24 hours of stimulation with the noted concentrations of anti-CD3ε antibody, with or without anti-CD28 antibody (5μg/mL). (K) Flow cytometry analysis of NAD(P)H autofluorescence and CD69 in peptide stimulated (G4 or N4) OT-I T cells. Statistical significance was determined using Student’s T-test (A), 2-way ANOVA (B) for the interaction of affinity and time in both ECAR and OCR (p < 0.0001), (F) of peptide affinity (p < 0.0001) and concentration (p < 0.0001), (J) for anti-CD3e levels (p = 0.0082) and CD28 stimulation (p = 0.0007) was (I) F test of non-zero slope Trp (p = 0.42), NA (p = 0.81), NR (p = 0.0005). ****p < 0.0001

NAD is a central component of the TCR affinity-dependent metabolome

We next tested whether specific TCR-dependent metabolic pathways underly this correlation between affinity-driven quiescence exit and T cell metabolic activity. While the global metabolic changes during activation have been characterized, the extent to which specific metabolites are tuned in a scalable, affinity-dependent manner is undefined (20, 23, 39–43). To this end, we performed LC-MS on OT-I T cells stimulated for 24 hours with peptides of decreasing affinity – N4, Q4, T4, or G4 (Data file S1). We chose this time point because it preceded the first division in all conditions and allowed for optimal capture of metabolic changes associated with quiescence exit (fig. S1, B and F). Principal component analysis revealed that the global metabolite profile was sufficient to distinguish cells based on the affinity, confirming that metabolic rewiring is a central feature of affinity-driven changes during T cell activation (Fig. 1C). Corroborating this, we found that these metabolic signatures were sufficient to predict affinity with a 91.6% prediction accuracy using a random forest analysis (Table S1). However, hierarchical clustering revealed that not all metabolites were equally responsive to peptide affinity (Fig. 1D). Whereas some metabolites showed no significant relationship in their response to affinity (Clusters 3 and 4), one cluster significantly increased with affinity (Cluster 1) and another significantly decreased (Cluster 2) (Fig. 1D). We then built a random forest analysis model using specific clusters and found that Clusters 1 and 2 predicted affinity with 91.6% and 100% accuracy, respectively (Table S1). In contrast, Clusters 3 and 4 performed worse compared to the model containing all metabolites (Table S1). Thus, a discrete set of metabolites define affinity-driven T cell biology.

Next, we sought to determine whether any specific metabolites might play a central role in coordinating the TCR affinity-dependent metabolic network and focused our analysis on Cluster 1 – metabolites that accumulate with increasing affinity. Targeted pathway analysis of Cluster 1 found metabolites matched significantly to 14 pathways (Table S2). Consistent with TCR affinity’s role in proliferation and supporting prior studies, Cluster 1 was enriched for metabolites involved in purine synthesis, the citric acid cycle and the malate-aspartate shuttle (Table S2) (44). We then ranked metabolites in Cluster 1 by the frequency with which they were involved across all identified pathways. This approach revealed that NAD was the most represented metabolite over all significantly enriched pathways (Fig. 1E). We further noted NMN (an NAD precursor) as well as metabolites such as orotate and lactate, which require NAD(H) as a coenzyme for their synthesis, were among the most affinity-sensitive metabolites identified (Fig. 1E). Together, these findings led us to hypothesize that NAD functions as a central, affinity-tuned metabolic hub that coordinates the broad biochemical changes of early T cell activation.

TCR signaling tunes the NAD salvage pathway to control single-cell NAD(H) levels

To corroborate our metabolomics findings that TCR affinity controls cellular NAD levels, we used an orthogonal, enzymatic assay to measure NAD. We found that cellular NAD levels are tuned by the strength of TCR stimulation in both a peptide affinity- and concentration-dependent manner (Fig. 1F). This held true when we titrated activating TCR stimuli in P14 transgenic T cells, WT polyclonal mouse CD8⁺ T cells, as well as primary human CD8⁺ T cells, suggesting this is indeed a generalizable phenomenon (fig. S2 A–C).

To further delineate which signals support the accumulation of cellular NAD, we provided CD8⁺ T cells with titrated amounts of TCR, CD28, and IL-2 stimulation. We found that while TCR engagement is absolutely required for enhanced NAD biosynthesis, CD28 stimulation is sufficient to further enhance cellular NAD levels (fig. S2C). Although exogenous IL-2 did not have a discernable impact when cells received high TCR stimulation, we observed that T cell NAD levels were sensitive to IL-2 in G4-stimulated cells, consistent with the role of IL-2 signaling in weakly stimulated T cells (fig. S2C and D) (21).

As NAD exists in either an oxidized (NAD+) or reduced (NADH) form, we next investigated whether this activation-induced NAD accumulation was a consequence of redox balance or an absolute increase in NAD. To assay this, we measured NAD and NADH in polyclonal CD8⁺ T cells activated for 24 hours with a range of anti-CD3e concentration. TCR stimulation dose-dependently induced both species, indicating that an absolute increase in total NAD(H) was required (fig. S2E).

NAD is made in cells by 3 primary pathways: De novo from tryptophan, the Preiss-Handler pathway from nicotinic acid (NA), or the NAD salvage pathway from nicotinamide (NAM) (Fig. 1G) (45). We found that only NAD salvage pathway metabolites displayed dose-dependency to TCR stimulation (fig. S2F). Consistent with this and prior reports, inhibition of NAMPT, the rate limiting enzyme of the NAD salvage pathway with FK866, prevented cellular NAD accumulation following activation (Fig. 1H) (46, 47). To test whether T cells rely solely on the NAD salvage pathway, we attempted to rescue FK866 treatment by supplementing media with tryptophan, NA, or nicotinamide riboside (NR), which bypasses FK866 by supporting NAMPT-independent NMN synthesis (Fig. 1H) (45, 48). Only the salvage pathway precursor NR rescued NAD levels in the presence of FK866 (Fig. 1I). Finally, as NAMPT expression is sensitive to T cell activation, we tested whether this is regulated in a signal strength-dependent manner (47). Both TCR and co-stimulation controlled Nampt transcript and protein levels (Fig. 1J, fig. S2G and H). Together, these data support a model in which activating signaling is integrated to promote NAD biosynthesis through the NAD salvage pathway.

Cellular NAD levels are controlled through both biosynthesis and consumption pathways (45, 48). In order to determine whether TCR signaling similarly tuned NAD consumption, we evaluated NAD decay rates post-FK866 treatment. Across all stimulation conditions, NAD followed a first order decay kinetic with a half-life of approximately 3.5 hours (fig. S2I and K). However, as the rate of NAD consumption is a function of concentration as well as half-life, and as NAD concentration increases with affinity, the rate of consumption likewise increases with affinity. In contrast, inhibition of major NAD consumption pathways (PARP’s, Sirtuins, and CD38) resulted in an increased NAD half-life (fig. S2J and K). Together, these data suggest that TCR signaling tunes cellular NAD levels through its impact on biosynthesis rather than consumption.

While we observed a direct relationship between TCR signal strength and NAD biosynthesis, the assays described thus far entail bulk analysis, limiting our ability to resolve how TCR-dependent processes such as anabolic state, cell size, and activation status might impact cellular NAD(H). Given both NAD and NADH changed proportionally in a TCR-dependent manner, we took advantage of the autofluorescent properties of NADH to interrogate single-cell differences in NAD(H) using flow cytometry. As NADPH has near identical autofluorescent properties as NADH, we assayed the specificity and sensitivity of this autofluorescence. The vast majority of NAD(P)H autofluorescence was derived from NAD salvage pathway activity and lost upon FK866 treatment (fig. S3A). We next tested the mitochondrial contribution to the NAD(P)H signature and found FCCP treatment significantly reduced NAD(P)H autofluorescence, but this effect was modest compared to FK866 (fig. S3A). This suggested that while redox state contributes to NAD(P)H autofluorescence, NAD biosynthesis is the dominant source. We next tested whether the per-cell NAD(P)H signal scaled with ligand concentration and/or affinity. NAD(P)H MFI dose-dependently increased with peptide concentrations and affinity, even when controlling for activation status (CD69 expression), size, or Myc expression (Fig. 1K and fig. S3, B and C). Altogether, these findings indicate that TCR signal strength tunes cellular NAD(H) in an analog manner, independent of activation status, cell size, or the expression of Myc.

Cellular NAD levels pace early cell cycle kinetics to control affinity-driven T cell expansion dynamics

Activated T cells have been shown to be sensitive to pharmacological NAMPT inhibition and conditional loss of Nampt disrupts thymocyte development, however the requirement for T cell Nampt expression during an adaptive immune response is unknown (46, 47, 49–53). To confirm that NAD biosynthesis plays a biologically relevant role in peripheral CD8⁺ T cell responses, we adoptively transferred either WT or Nampt-deficient OT-I T cells into WT recipients and infected those animals with Listeria monocytogenes expressing ovalbumin (Lm-Ova) (fig. S4A) (50). While WT OT-I T cells expanded as expected 8 days post-infection, Nampt-deficient cells failed to do so. Thus, NAD biosynthesis is required for in vivo T cell expansion (Fig. 2A). Suggesting that NAMPT is specifically required in activated T cells, we found an equivalent frequency of circulating OT-I T cells between naïve WT and Nampt-deficient donors (fig. S4B).

Figure 2: Cellular NAD levels control affinity-driven T cell expansion dynamics.

(A) CD8⁺ T cells from Nampt^fl/fl x CD8α-Cre x OT-I mice or Cre controls were transferred into Lm-Ova infected CD45.1 recipients. The relative number of transferred cells was quantified at day 7 post-infection. WT or Nampt Tg x CD8a-Cre mice were infected with Lm-Ova and analyzed at day 7 post-infection. (B) Quantification of SIINFEKL-H2-Kb tetramer gMFI, normalized to TCRβ gMFI. (C) The number of SIINFEKL-H2-Kb tetramer⁺ cells. (D) Schematic of experimental design for E-I. CD8⁺ T cells were activated and treated with FK866 or vehicle control for 24 or 48 hours from the indicated time points. (E) Cell number and proliferation (48-hour treatment windows). (F) Cell cycle entry (24-hour treatment windows). CD8⁺ T cells were activated and treated with FK866 or vehicle control from either day 0–1 or 3–4. (G) NAD(P)H fluorescence lifetime in the mitochondria and cytoplasmic/nuclear compartments, plotting relative lifetime in FK866 versus vehicle controls. (H) Basal ECAR and OCR in activated CD8⁺ T cells treated with FK866 or vehicle control from day 0–1 or 1–2. Peptide (N4 or G4) stimulated OT-I T cells were treated with FK866 +/− NR at the indicated concentration. After 24 hours, (I) the frequency of Ki67⁺ cells and (J) Myc gMFI in G1 cells were quantified. K) Seahorse Mito Stress Test of WT CD8+ T treated with FK866 and NR at the indicated concentrations, 24 hours post-stimulation. Results from 2 (A) or 3 (B) independent experiments are shown. Statistical significance was calculated in (A-C and E-G) using multiple unpaired t-tests using Two-stage step-up (Benjamini, Krieger, and Yekutieli). (H) Multiple comparisons T test using Holm-Sidak correction was used for ECAR. Student’s T-test of fold change for OCR. EC 50 calculated for (I): N4 = 0.093μM NR, G4 = 0.197μM NR) and (J) N4 = 0.42μM NR, G4 = 0.43μM NR. **** p<0.0001

To begin testing our hypothesis that NAD biosynthesis acts as a metabolic conduit for TCR affinity, we asked whether globally enhancing NAD biosynthesis capacity in CD8⁺ T cells is sufficient to permit expansion of lower affinity clones during a polyclonal response. We therefore bred CD8α-Cre mice to Nampt overexpression transgenic animals (Nampt^Tg) (54). CD8⁺ T cells in these mice express higher basal levels of Nampt, uncoupling NAD salvage pathway regulation from TCR affinity (fig. S4C). In Nampt^Tg mice infected with Lm-Ova, responding clones displayed a lower normalized SINFEKL loaded H2-Kb tetramer MFI than WT controls, suggesting that the affinity of the polyclonal response was diminished (Fig. 2B). Despite Nampt^Tg animals displaying a lower affinity response, both transgenic and WT mice had the same number of SIINFEKL-reactive T cells, indicating that enhancing NAD biosynthesis is sufficient to permit equitable expansion of lower affinity clones (Fig. 2C).

We next sought to understand the mechanism for how NAD biosynthesis controls the complex cell biology downstream of the TCR. Given that cellular metabolic requirements change throughout T cell differentiation, we asked whether the NAD biosynthesis requirement was restricted to a specific stage of activation (1, 2, 39, 44). To this end, we assayed impact of prolonged inhibition of NAD biosynthesis during or after quiescence exit (Fig. 2D). Cells were most sensitive to NAMPT inhibition prior to cell cycle entry (days 0–2), however cells rapidly became insensitive to NAD biosynthesis inhibition following quiescence exit (days 2–4) (Fig. 2E and fig. S5, A and B). To better resolve these time-dependent requirements, we narrowed our treatment windows to 24-hours (Fig. 2F and fig. S5, C and D). We again observed that cells displayed a heightened sensitivity to FK866 treatment during quiescence exit, however this sensitivity was lost when most cells had entered the cell cycle (Fig. 2F). Suggesting that this was not simply due to differences in NAD levels over time, NAD levels fell to the limit of detection following FK866 treatment at both ends of the treatment window (day 0–1 and 4–5) (fig. S5E).

We next asked whether differential metabolic demands before and after quiescence exit explained this temporally restricted requirement for NAD biosynthesis. Using NAD(P)H fluorescence lifetime imaging microscopy (FLIM), we were able to attain information about the free (shorter lifetime) and protein-bound (longer lifetime) ratio of NAD(P)H in the cell. Loss of NAD biosynthesis had a greater impact on fluorescence lifetime in early activated T cells than later stage cells, indicating recently activated T cells have a heightened requirement for NAD biosynthesis to maintain proper NAD(H) utilization throughout the cell (Fig. 2G).

As NAD is a coenzyme both for mitochondrial respiration and glycolysis, we tested whether these two processes were differentially sensitive to NAD salvage pathway inhibition at the time of activation and shortly after. Day 0–1 FK866 treatment resulted in a marked impairment of glycolysis (Fig. 2H and fig. S5 F). In stark contrast, day 1–2 FK866 treatment yielded no detectable differences in glycolytic rate. Similarly, day 0–1 FK866 treatment almost completely abrogated mitochondrial respiration, whereas basal OCR was marginally affected after day 1–2 FK866 treatment, with spare respiratory capacity showing some sensitivity (Fig. 2H and fig. S5 G and H). These data suggest that T cells have a heightened requirement for NAD biosynthesis to support glycolysis and mitochondrial respiration as they enter the cell cycle, which is lost shortly after.

Our data demonstrate that the magnitude of TCR signaling generates a commensurate NAD biosynthesis response that is required as T cells exit quiescence. We therefore asked whether the degree of NAD biosynthesis downstream of a given TCR affinity directly regulated cell cycle rates in those cells. In the presence of FK866, NAM cannot be recycled back to NMN, leaving NMN synthesis reliant on the amount of NR supplemented in the media and enabling tight control of NAD biosynthesis capacity (Fig 1G and fig. S6A). For both N4 and G4 peptides, T cell expansion was proportionately controlled by the concentration of NR in the culture (fig. S6B). While we also observed that NAD biosynthesis capacity controlled T cell viability over time, there was no effect on cell survival within the first 24 hours (fig. S6C). This permitted us to assess the interaction of NAD with the cell cycle within this window. We found that NAD biosynthesis capacity sets the kinetics of G1-entry, G2 progression, and the time to the first division downstream of a given TCR signal strength (Fig. 2I and fig. S6 D, E and F). Furthermore, high and low affinity-stimulated T cells displayed differential sensitivity to NAD biosynthesis capacity over time. While N4- and G4-stimulated cells exhibited similar NAD-dependencies at the time of cell cycle entry, N4-stimulated cells displayed an increasing requirement for expansion later (Fig. 2I and J fig. S6 B and G). Combined, these data demonstrate that NAD biosynthesis capacity dictates the rate of quiescence exit and cell cycle progression to control expansion dynamics in activated T cells, within the bounds set by the corresponding strength of signaling.

As T cell quiescence exit is controlled by mTOR signaling and Myc expression, we asked whether NAD biosynthesis might modulate these key anabolic regulators. We observed that the phosphorylation of the mTOR substrate S6 as well as the kinetics of and amplitude of Myc expression were dose-dependently sensitive to NR supplementation, suggesting that NAD biosynthesis capacity is limiting for both factors (Fig. 2J and fig. S6H). We next tested whether cellular NAD levels also control T cell metabolic capacity. NAD biosynthesis capacity dose-dependently controlled the rates of both glycolysis and mitochondrial respiration (Fig. 2K).

NAD biosynthesis governs a quiescence exit checkpoint in activated CD8⁺ T cells

We next asked whether the cell cycle arrest exhibited following NAD salvage pathway inhibition represents a reversable checkpoint-like phenomena or an irreversible process such as senescence and/or cell death. We thus assayed whether cells unable to undergo NAD biosynthesis at the time of activation could be rescued and enter the cell cycle if the NAD salvage pathway was restored in a delayed manner (Fig. 3A, fig. S7A). No differences in T cell viability and activation prior to NR supplementation were observed in FK866-treated compared to vehicle-treated cells (fig. S6C, S7 B and C). Delayed NR supplementation resulted in a rapid recovery of G1-entry, with the frequency of G1 cells returning to that of untreated controls (Fig. 3B). Following this initial NR treatment period, “rescued” cells were then able to successfully proliferate, with nearly 100% of cells dividing 24 hours later (fig. S7D). These data indicate that NAD biosynthesis is necessary and sufficient for activated T cell quiescence exit and the cell cycle arrest observed upon NAD salvage pathway inhibition is reversible.

Figure 3: NAD biosynthesis governs a quiescence exit checkpoint.

(A) Schematic of “delayed NAD rescue” assay. (B) Peptide stimulated OT-I T cells were with FK866 for 24 hours and then acutely supplemented with 100μM NR (hereafter, the“delayed NAD rescue assay”). The frequency of Ki-67⁺ cells was quantified at the indicated time points post-NR treatment. Following 0.1μM or 100μM NR treatment of peptide (N4 or G4) stimulated OT-I T cells during the delayed NAD rescue assay, (C) the frequency of Ki67⁺ cells and (D) Myc gMFI in G1 cells was quantified. (E) Representative plots of Myc protein expression at 4 hours post-NR supplementation. (F) Seahorse Mito Stress Test on OT-I T cells following delayed NAD rescue assay, with maximal ECAR and OCR plotted. Untargeted LC-MS was performed on cells subjected to the delayed NAD rescue assay to quantify relative metabolite abundance pre- and post-NR supplementation (30, 60, 240 minutes). Peak areas were normalized to an internal standard. (G) Frequency of all detected metabolites that were NAD sensitive. (H) Schematic of glycolysis shown above graphs plotting the abundance of the bolded metabolites during the delayed NAD rescue assay. (I) Experimental schematic for J & K. The delayed NAD rescue assay was performed on OT-I T cells in either control media or media containing 0.03 mM glucose. After 24 hours, cultures were supplemented with NR and limiting glucose cultures were supplemented with 10 mM glucose and (J) cell cycle entry and (K) Myc gMFI in G1 cells were quantified. Dotted line denontes vehicle control levels at the time of NR addition. (L) Experimental schematic of M & N. The delayed NAD rescue assay was performed on OT-I T cells,treating with the indicated inhibitors 10 minutes before NR supplementation and (M) the frequency of Ki-67⁺ cells and (N) Myc gMFI in G1 cells were quantified. Significance was calculated using (B) multiple unpaired t-tests with Two-stage step-up (Benjamini, Krieger, and Yekutieli) (C and D) F-test comparing the slope of linear regression between 100μM NR and 0.1μM NR conditions. (F) 2-way ANOVA for dose and time interaction for max ECAR and max OCR. (G, M, and N) Multiple comparisons using Tukey’s correction. (J-K) Sidak’s multiple comparisons. All comparisons are significant except (M) 2DG and HA and (N) HA and FK866. ****p < 0.0001

Using this delayed NAD rescue approach, we tested if the cell cycle kinetics set by a given TCR affinity were dependent on the corresponding magnitude of NAD biosynthesis generated downstream. G4-stimulated cells were similarly able to exit quiescence and elevate Myc protein expression, however with a less rapid recovery kinetic than N4-stimulated cells (Fig. 3, C–E). Consistent with NAD biosynthesis being a rate-limiting factor for affinity-dependent cell cycle differences, limiting NAD biosynthesis capacity in N4-stimulated cells slowed cell cycle entry and Myc protein expression to rates comparable to G4-stimulated cells (Fig. 3, C–E).

T cell quiescence exit is known to require mTOR activity (21, 24, 25). We thus asked whether mTOR signaling was likewise essential for the NAD-dependent checkpoint. Like cell cycle entry and Myc expression, S6 as well as S6K phosphorylation were quickly restored after delayed NAD rescue, suggesting NAD biosynthesis is also necessary and sufficient for mTOR signaling during quiescence exit (fig. S7 E and F). Conversely, Torin treatment minimally impacted the rate of G1-entry following delayed NAD rescue, while only partially abrogating Myc protein expression (fig. S7, G–I). Consequently, while mTOR signaling is required for optimal proliferation and Myc expression, reengagement of mTOR activity does not fully explain how NAD biosynthesis controls quiescence exit.

We next hypothesized that once activation has occurred, NAD might regulate quiescence exit independent of mitogenic signaling. We thus asked whether NAD-dependent quiescence exit required sustained TCR and costimulatory signaling (fig. S7J). Regardless of whether T cells received continuous stimulation or were removed prior to rescue, we observed that NR supplementation restored proliferative capacity and Myc expression to comparable levels (fig. S7, K–M). These data demonstrate that mitogenic signals poise T cells for cell cycle entry and dictate the maximal rate of this response, however quiescence exit cannot occur until T cells overcome an NAD biosynthesis-dependent checkpoint.

NAD biosynthesis governs the rates of glycolysis and respiration to control quiescence exit

The coincidence of an NAD requirement for quiescence exit and sustaining energetics during early T cell activation led us to hypothesize these two processes were connected. Indeed, metabolic processes including glycolysis, the citric acid cycle, and oxidative phosphorylation are known to regulate cell cycle progression in T cells (1, 2, 39, 44). Having seen glycolysis and mitochondrial respiration require the NAD salvage pathway during early activation, we first tested whether T cells engage these metabolic processes during the delayed NAD rescue assay. We observed dose-dependent increases in basal and maximum ECAR and OCR following NR supplementation (Fig. 3F, fig. S8A).

As NAD(H) regulates many biochemical pathways beyond glycolysis and respiration, we aimed to more globally characterize the NAD(H)-dependent pathways engaged during T cell quiescence exit. To do so, we performed mass-spectrometry on OT-I T cells at 0, 30, 120 and 240 minutes after delayed NR treatment. Validating this approach, we saw that FK866 resulted in a striking depletion of cellular NAD levels that rapidly increased following NR supplementation (fig. S8B). In turn, this loss and recovery of cellular NAD had broad and differential effects on amino acids, nucleic acids, and central carbon metabolism (fig. S8B). A majority (63%) of all detected metabolites displayed significant sensitivity to the NAD salvage pathway, indicating that NAD biosynthesis plays a critical role maintaining the metabolic landscape during quiescence exit (Fig. 3G).

Among the most striking NAD-dependent changes, we observed that glycolysis was markedly sensitive during the delayed NAD rescue assay (Fig. 3H). FK866 treatment resulted in a buildup of metabolites in glycolysis upstream of the GAPDH step and a depletion of metabolites immediately downstream. Reciprocally, this block was reversed following NR supplementation (Fig. 3H). These data suggest that NAD biosynthesis is required to support the rapid increase in glycolysis during early T cell activation via GAPDH. Whereas GADPH reduces NAD to NADH, the NADH oxidizing activity by LDH is understood to help recover NAD in highly glycolytic cells, where free NAD can become rate limiting. We observed that FK866 treatment had minimal effects on lactate and pyruvate levels, but that lactate quickly accumulated after NR supplementation with pyruvate trending down over time, confirming that glycolysis is rapidly engaged after NAD biosynthesis is restored (Fig. 3H). Our findings suggest that NAD(H) levels are limiting in early T cell activation and that NAD biosynthesis is required to sustain the high rate of GAPDH activity in these cells. While LDH activity may help recover free NAD to support this high glycolytic rate, it cannot restore redox balance sufficient for GAPDH in the absence of the NAD salvage pathway.

These findings led us to hypothesize that affinity-dependent NAD biosynthesis controls quiescence exit kinetics through its regulation of the rate of glycolysis and respiration. We first evaluated the impact glucose restriction has on cell cycle entry during T cell activation (fig. S8 C and D). We asked if delayed glucose supplementation could rescue quiescence exit in those conditions and in turn whether NAD biosynthesis governs the rate this occurs (Fig. 3I). We then tracked cell cycle entry over the following 6 hours (Fig. 3J). While delayed glucose supplementation permitted cell cycle entry and restored Myc levels, this was entirely dependent on the NAD biosynthesis capacity of the cells (Fig. 3, J and K). Thus, while glucose availability is necessary for cell cycle entry, it alone is not sufficient, as NAD biosynthesis is required for glucose utilization during quiescence exit.

As NAD(H) has roles beyond its activity as a coenzyme, we next sought to test the alternative hypothesis that rather than acting through central carbon metabolism, NAD biosynthesis primarily supports quiescence exit through its role as a substrate for NAD consuming enzymes (i.e. Sirtuins, PARP’s, and CD38) (45, 48). While inhibitory drugs against these pathways were sufficient to extend the NAD consumption half-life, they did not abrogate cell cycle entry following NR rescue (fig. S2J and S8E). We repeated this assay treating cells with 2DG, oligomycin, or heptelidic acid to inhibit glycolysis, ATP synthase, or GAPDH, respectively. All three inhibitors were sufficient to blunt cell cycle entry, with 2DG and heptelidic acid exhibiting the most potent effects (Fig. 3M). Similar to our cell cycle entry results, 2DG, oligomycin, and heptelidic acid all abrogated Myc protein expression after NR supplementation, with inhibition of glycolysis and GAPDH resulting in the most robust reduction in Myc protein expression (Fig. 3N).

These data indicate NAD biosynthesis licenses cells to exit quiescence through its role as a rate-limiting factor in central carbon metabolism. Once cells have cleared this checkpoint and have entered the cell cycle, their reliance on NAD biosynthesis starkly decreases, highlighting the distinct biochemical requirements at the time of cell cycle entry. Altogether, our findings support a model during early T cell activation in which sufficient carbon source availability and NAD biosynthesis are together required for quiescence exit to facilitate the energetic and synthetic demands of division. As the magnitude of NAD biosynthesis is proportional to ligand affinity, NAD functions as a metabolic conduit for TCR signal strength in controlling T cell expansion dynamics.

Single-cell differences in NAD(H) underly heterogeneity in lymphocyte proliferative responses

Thus far, our findings illustrate how TCR signal strength tunes cellular NAD levels in T cells to license the biochemical remodeling required for quiescence exit. While this was found at the population level, our NAD(P)H flow cytometry results show a broad distribution in NAD(H) levels amongst individual T cells responding to the same stimuli. Given proliferation rates are not uniform across populations of identically stimulated T cells, we hypothesized that single-cell differences in NAD(H) underly this heterogeneity. We first asked whether the NAD(H) heterogeneity observed was a readout for population level heterogeneity in TCR engagement or metabolic heterogeneity in cells receiving comparable signal. As Nur-77 expression is a highly sensitive reporter of TCR signal strength, we utilized an OT-I x Nur77-GFP mice to permit simultaneous assessment of single-cell differences in TCR signaling and NAD(P)H autofluorescence. As expected, Nur77-GFP expression was dependent on TCR affinity and antigen concentration (fig. S9A). We observed heterogeneity in the NAD(P)H signal even within populations of comparable Nur77-GFP expression (fig. S9B). These data suggest that while the median NAD biosynthesis capacity in a population of T cells scales with the strength of TCR stimulation, there exists a wide range of metabolic heterogeneity amongst individual cells, above and below this median.

To further test whether variation in single-cell NAD(H) levels contributes to activated T cell heterogeneity, we developed a flow cytometry sorting strategy that enabled us to unmix populations of cells based on NAD(P)H and then return them to culture. Using this approach, we ensured that all sorted cells were activated (CD69+) and had not undergone their first cell division (CTY) (Fig. 4A). Consistent with our findings that NAD levels support T cell metabolic potential, NAD(P)H high cells exhibited higher rates of glycolysis and oxygen consumption 24 hours post-sort (Fig. 4B). Additionally, we found that the magnitude of early cellular NAD(P)H was predictive of the rate that T cells exited quiescence and progressed through the cell cycle as early as 12 hours post-sort (Fig. 4C). Looking 24 hours post-sort, initial cellular NAD(P)H levels predicted both the rate T cells underwent their first division and their early proliferative potential, with the number of divisions proportional to NAD(P)H levels at the time of sort (Fig. 4D).

Figure 4: Single-cell differences in NAD(H) underly lymphocyte heterogeneity.

OT-I T cells were stimulated with peptide for 24hrs and then CD8+CD69+ cells were sorted by NAD(P)H autofluorescence. (A) Experimental schematic and example gating strategy. (B) Basal ECAR and OCR of NAD(P)H high and low cells one day post-sort. N4 stimulated OT-I T cells were sorted by quartiles of NAD(P)H autofluorescence and evaluated for (C) cell cycle 12 hours post-sort or (D) proliferative capacity 24 hours post-sort. (E) Schematic of NAD(P)H normalization sorting strategy. (F) OT-I T cells were stimulated with N4 or G4 peptide for 24 hours and were then sorted using the same gates for NAD(P)H high and low cells. Shown is the frequency of divided cells in each population, 24 hours post-sort. WT CD8⁺ T cells were stimulated and sorted by NAD(P)H autofluorescence at ~20 hours post-stimulation. Three days post-sort, cells were restimulated and evaluated for (G) proliferation (CTY) and IFNγ production (H) TNFα and IFNγ. (I) After 24 hours, N4 stimulated 45.1 OT-I T cells were sorted by NAD(P)H and adoptively transferred into Lm-Ova infected, congenically marked recipients on day 1 post-infection. At days 7 and 35, the number of transferred splenic OT-I T cells were quantified. WT CD4⁺ T cells or B cells were stimulated with anti-CD3e and anti-CD28 plate bound antibodies or LPS respectively. Cells were sorted 20 hours post-stimulation based on NAD(P)H autofluorescence. Proliferation was determined 1 day post-sort for (J) CD4⁺ T cells and (K) B cells. Statistical significance was determined using (B, I, J, and K) Student’s t-test, (D) F-test of non-zero slope of the linear regression of the rank assigned data, (F) Two-way ANOVA which determined that NAD(P)H levels contributed to 93.8% of the variance in sorted conditions. * p < 0.05 ****p < 0.0001

We next asked whether differences in single-cell NAD(P)H could help explain the differences in proliferative potential between high and low affinity ligand stimulated cells and if normalizing NAD(P)H autofluorescence between low and high affinity cells would be sufficient to normalize responses (Fig. 4E). As expected, unsorted N4-stimulated OT-I T cells proliferated to a much greater extent than G4-stimulated cells (Fig. 4F). In contrast, when cellular NAD(P)H levels were normalized across activating conditions, G4-stimulated cells were capable of proliferating nearly to the same extent as N4-stimulated cells. While N4-stimulated cells retained a slight advantage after sorting, NAD(P)H levels explained 94% of the variance in proliferation across sorted populations (Fig. 4F). Furthermore, pre-division differences in single-cell NAD(P)H were deterministic of proliferative dynamics in naïve T cells, T cell stimulated with peptides of intermediate affinities, and when controlling for fluctuations in NAD(P)H autofluorescence by concurrently sorting high and low affinity stimulated cells (fig. S9 C and D). Altogether, our data indicate that T cell activation generates a continuum of single-cell NAD(H) levels, with increasing ligand affinity resulting in a broader distribution and a higher median. While low and high affinity stimulated T cells are phenotypically distinct at the population level, initial differences in single-cell NAD(P)H are sufficient to explain heterogeneity in proliferative potential across ligand affinities.

As effector function acquisition is tightly coupled to proliferative capacity, we next asked whether pre-division NAD(P)H was also predictive of cytokine production. Consistent with cell division kinetics, initial NAD(P)H levels predicted future functional capacity, with NAD(P)H^high cells producing more IFNγ and TNFα (Fig. 4 G and H, fig. S9E). Moreover, NAD(P)H^high sorted cells displayed greater overall polyfunctionality compared to NAD(P)H ^low cells (fig. S9F). Thus, early differences in single-cell metabolism are deterministic of the proliferative dynamics from a population of activated cells. This metabolic feature can further be used to isolate cells with discrete characteristics from an otherwise heterogenous pool.

The characteristics we observed in NAD(P)H^high cells (rapid proliferation, high glycolytic engagement, and early acquisition of effector function) were consistent with the major hallmarks of early effector cells, which quickly expand but poorly persist in vivo. We therefore asked whether early differences in single-cell NAD(P)H levels were similarly deterministic of in vivo T cell dynamics. We activated OT-I T cells with N4 peptide, sorted NAD(P)H^high and NAD(P)H^low cells, and adoptively transferred them into congenically marked recipients that had been infected with Lm-Ova one day prior to transfer. We found that while NAD(P)H^high and NAD(P)H^low cells expanded to equivalent levels at day 7 post-infection, NAD(P)H^high cells contracted to a greater extent by day 35 post-infection compared to NAD(P)H low cells (Fig. 4I, fig. S9G). Taken together with our in vitro findings, these data suggest that metabolic heterogeneity prior to the first division is not only deterministic of early expansion dynamics, but also long-term persistence in vivo.

Antigen receptor ligand affinity control of cell expansion dynamics is not unique to CD8⁺ T cells, but is a shared feature with CD4⁺ T cells and B lymphocytes. We therefore asked if the relationship we observed between CD8⁺ T cell ligand affinity, NAD biosynthesis, and cell cycle kinetics was conserved in these other lymphocyte populations. Like CD8⁺ T cells, antigen receptor signaling drove a dose-dependent increase of cellular NAD in both CD4⁺ T and B cells (fig. S10, A and D). As with BCR stimulation, increasing TLR stimulation drove a dose-dependent increase in cellular NAD, suggesting NAD biosynthesis plays a role in response to a broad range of mitogenic signals (fig. S10E). FK866 treatment abrogated G1-entry and Myc expression in CD4⁺ T and B cells, suggesting NAD biosynthesis is a general requirement for lymphocyte quiescence exit (fig. S10 B–C and F–G). Finally, we tested whether differences in single-cell, pre-division NAD(H) levels could help explain activated CD4⁺ T and B cell proliferative heterogeneity. Like CD8⁺ T cells, higher pre-division NAD(P)H levels predicted faster rates through the cell cycle and the time to the first division (Fig. 4, J and K).

Discussion

Heterogeneity underlies adaptive immune responses with responding clones varying in the number of divisions they undergo and the programs they adopt. Even the daughter cells of a single clone exhibit a range of proliferative and functional potential (55–59). Antigen receptor diversity determining a lymphocyte’s affinity for a cognate antigen is understood to be a primary driver of these phenomena (3–11). While the proximal signaling and downstream transcriptional responses that scale with affinity are well-characterized, the biochemical responses to affinity and their downstream mechanisms are largely undefined. Here, we identify NAD biosynthesis as a central biochemical mediator of affinity-driven lymphocyte responses. Moreover, we find that cell-to-cell variation in this single pathway is sufficient to explain large facets of inter- and intra-clonal heterogeneity.

In responding naïve lymphocytes, the strength of the activating signal determines the rate of quiescence exit and proliferation. Downstream of these mitogenic signals, cells are instructed to rewire their metabolism and meet the biochemical demands of rapid proliferation. The relationship between activating signal strength, cell cycle dynamics, and metabolic remodeling has been largely understood through the lens of signaling and gene regulation. Amongst these pathways, the most well described are mTOR, Myc, and IRF4. All three factors are sensitive to the magnitude of TCR signaling and each plays essential roles in patterning long-term proliferative and metabolic potential during early T cell activation (8, 11, 13, 14, 16–19, 21–25, 60). Whereas IRF4 is dispensable during early activation, Myc expression and mTOR signaling are essential for T cell quiescence exit and the anabolic program necessary to sustain growth (8, 17, 18, 21, 24, 25, 60). In large part, mTOR and Myc support T cell expansion and differentiation through their regulation of a broad range of key metabolic processes such as glucose import, amino acid uptake, as well as mitochondrial remodeling during the early stages of activation (1, 12, 15, 16, 18, 21–23, 25). Given that these anabolic mediators are tuned by antigen receptor signaling, we hypothesized that specific metabolic pathways would likewise be sensitive to the strength of activation signal.

Through unbiased metabolomics analysis on OT-I T cells stimulated with peptides of varying affinity, we identified NAD biosynthesis as a central node of affinity-dependent metabolic changes during early activation. While NAD biosynthesis is understood to be required in activated T cells and can function as a marker of resting versus activated cells, the specific cellular functions supported by it and its connection to affinity-dependent biology were previously unexplored. Employing an in vitro system to manipulate cellular NAD levels, we identified the NAD salvage pathway as a mediator of quiescence exit and proliferation rates downstream of TCR stimulation. Collectively, our data indicate that the NAD salvage pathway regulates a quiescence checkpoint in naïve lymphocytes. Accompanying this NAD-dependent cell cycle entry, we observed a recovery in Myc protein expression and mTOR signaling. This suggests that these critical anabolic pathways are not only upstream of lymphocyte metabolic rewiring, but their sustained activity requires positive feedback from the successful deployment of this new biochemical program. Mechanistically, we found that NAD biosynthesis sets the rate of cell cycle entry through the control of GAPDH activity and central carbon metabolism downstream, as well as a majority of activation-dependent metabolic changes. Thus, NAD biosynthesis functions as a metabolic hub that scales with TCR signal strength to orchestrate the broader affinity-dependent metabolism and cell biology of T cell activation.

While affinity is a primary determinant of inter-clonal lymphocyte dynamics, significant heterogeneity remains within single clonal lineages. This has largely been explained through the action of extrinsic factors, such as cytokines, as well as through processes including asymmetric division. Asymmetric inheritance of factors and pathways such as Myc, mTOR, PI3K signaling, and the proteosome have all been shown to generate metabolic heterogeneity within a clonal lineage at the time of the first division (18, 22, 61–63). Given that we observe single-cell differences in NAD(P)H autofluorescence prior to the first division and that NAD biosynthesis capacity regulates subsequent proliferation rates, we asked whether NAD(P)H autofluorescence is similarly deterministic of future cell behavior. Indeed, we found that within a population of identically stimulated cells, NADH levels predicted proliferation rates, metabolic activity, the kinetics of effector function acquisition, and persistence following the contraction phase of an acute infection. Moreover, we found that across populations of cells disparately stimulated cells, selecting cells with similar NADH levels normalized the rate of quiescence exit and proliferation. Like CD8⁺ T cells, we found a similar role for NAD biosynthesis and single-cell variance in CD4⁺ T cells and B cells, suggesting a broader phenomenon in lymphocytes. Altogether, these findings highlight how early differences in cellular metabolism play a crucial role in dictating lymphocyte heterogeneity, through mechanisms that precede what has been previously described in the context of asymmetric division.

Combined, our data support a model in which NAD biosynthesis scales with the strength of mitogenic signaling to license the metabolic remodeling required for cell cycle entry and early proliferation. Our data suggest that NAD salvage pathway activity, rather than precursor availability, is rate limiting for lymphocyte NAD production, and thereby function. In this manner, cells with lower NAD(H) levels are biochemically constrained and unable to quickly exit quiescence relative to cells that generate more NAD. In populations receiving a low versus high affinity stimulus, there are high affinity cells that are limited by lower NAD biosynthesis and lower affinity cells that maximally generate NAD for their given stimulation strength. This metabolic constraint is therefore sufficient to normalize proliferative behavior between these two groups, despite differences in the initial mitogenic signal. Combined, we believe this model helps explain why cell-to-cell NAD(H) variation underlies large facets of inter- and intra-clonal heterogeneity.

These findings provide a metabolic framework for how ligand affinity and signal strength dictate clonal lymphocyte dynamics. We believe that this offers a comprehensive model for how mitogenic signaling coordinates cell cycle progression via a single metabolite. Our data illustrate how canonical signaling networks must work in concert with metabolic reprogramming to ensure that a newly deployed cellular program can be supported by a cell’s biochemical state. Thus, while signaling sets the limits of cellular function, single-cell differences in metabolism govern their behavior. This paradigm suggests that other aspects of cellular metabolism might similarly contribute to heterogeneity in other developmental processes and might be leveraged to rapidly identify cells with desired functional properties for cellular engineering or cell-based therapies.

Materials and Methods

Study design:

The major goal of this study was to identify the metabolic processes that facilitate affinity-dependent cell biology downstream of antigen receptor signaling and determine if this underlies inter- and intra-clonal heterogeneity found during lymphocyte responses. To this end, we performed unbiased mass-spectrometry to identify NAD biosynthesis as a central metabolic mediator of T cell affinity-dependent biology. We substantiated this using in vivo genetic tools as well as in vitro manipulation of NAD biosynthesis to dissect the cell biology and molecular mechanisms underlying this. We then employed an NAD(P)H sorting strategy to formally evaluate the role of single-cell variation in this pathway on lymphocyte proliferative and functional heterogeneity.

Mice:

Sex matched male and female C57BL/6 (Jax #000664), OT-I (Jax #003831), CD8-Cre (Jax #008766), B6-Ly5.1 (Jax #002014), and Myc-GFP (Jax #019075) mice (6–8 weeks old) were purchased from Jackson Laboratories and maintained in house. Namptfl/fl and Nampt^Tg mice were provided by J. Baur (50, 54). P14 (JAX: Tg(TcrLCMV)327Sdz) animals were a gift from E. John Wherry. Spleens from Nur77-GFP mice were a gift from B. Au Yeung (64). All mice were housed under specific pathogen-free conditions with a 12 hour on/off light cycle at the Children’s Hospital of Philadelphia. All experiments were performed in accordance with the Institutional Animal Care and Use Committee of the Children’s Hospital of Philadelphia (IAC 21–001325).

Lymphocyte cell culture:

Murine CD8+ T cells, CD4+ T cells, or B cells were isolated from the spleen and lymph nodes by negative selection using an EasySep Mouse Isolation kit (StemCell Technologies). Human T cells were obtained from the University of Pennsylvania Human Immunology Core. Unless otherwise noted, OT-I T-cells were stimulated with 1 μg/ml SIINFEKL (N4), SIIQFEKL (Q4), SIITFEKL (T4), or SIIGFEKL (G4) peptides (Anaspec), anti-CD28 (clone 37.51, BioLegend) and rIL-2 (BioLegend). P14 T cells were activated as OT-I T cells, using GP33 (KAVYNFATC) peptide (GenScript). C57BL/6 CD8+ T and CD4+ T-cells were stimulated with anti-CD3e (clone 145–2C11, BioLegend), anti-CD28, and rIL-2 (BioLegend). Human T cells were stimulated with anti-CD3e (clone OKT3, BioLegend), and rIL-2 (Biolegend). B-cells were stimulated with anti-IgM (Goat anti-mouse, Jackson Immunoresearch), CD40 (Clone FGK 4.5, BioXcell) or LPS (Sigma).

For all experiments using drug or metabolite supplementation, the following doses were used, unless otherwise noted: 100 nM FK866 (Daporinad), 100 μM NR (Elysium), Tryptophan (Sigma), Nicotinic Acid (Sigma), 1 μM rotenone (Sigma), 1 μM oligomycin (Sigma), 10 mM 2-deoxy-D-glucose, 0.2mM FCCP (Sigma), 10 μM Rucaparib (MedChemExpress), 50 μM EX527 (Selleck), 65 μM 78c (Tocris).

Flow Cytometry:

All flow cytometry was performed on a CytoFLEX LX or CytoFLEX S cytometer (Beckman Coulter). Data was analyzed using FlowJo software (Treestar). Cell division was measured by labelling cells with CellTrace Violet or CellTrace Yellow (Invitrogen) before activation and evaluated for proliferation at the indicated time points. Staining was performed with combinations of the following antibodies: anti-CD8 (1:300, Clone 53–6.7, Thermo), TCRβ+ (1:300, Clone H57–597, Thermo), CD44+ (1:300, IM7, Thermo), Ki-67 (1:600, SolA15, Thermo), anti-cMyc (1:400, D84C12, Cell Signaling Technology), anti-S6 (1:75, 54D2, Cell Signaling Technology), and anti-phospho-S6 (1:75, D57.2.2E, Cell Signaling Technology). DNA content was measured by DAPI. Viability was assessed using eBioscience Fixable Viability Dye eFluor 780 was used (Thermo). NAD(P)H autofluorescence was determined by signal excited from the UV laser (375 nm) and detected at 450nm. Triplicate sample wells were run asynchronously to control for fluctuations in the autofluorescence signal. Absolute cell number in each sample was back calculated from recorded cytometer sample volumes following calibration. Cytokine was measured by adding brefeldin A (Invivogen) 30 minutes after the addition of PMA (20 ng/ml, Fisher) and ionomycin (20 ng/ml, Cayman); 4.5 hours post-restimulation, cells were fixed, permeabilized and stained with anti-IFNγ (1:100, XMG1.2, Thermo), anti-TNFα (1:100, MP6-XT22, Thermo), and anti-IL-2 (1:100, MQ1–17H12, Thermo). Tetramer staining was performed with SIINFEKL loaded H2-Kb tetramer (NIH tetramer core) for 1 hour at 4C prior to surface staining. The gMFI was calculated and normalized to TCRβ gMFI to calculate relative affinity.

Flow cytometric cell sorting:

Cell sorting of T cells and B cells was done on a Cytek Aurora CS cell sorter. For compensation calculation, FK866 treated CD8 T cells were used as a negative control for NAD(P)H autofluorescence. During the sort, the chamber was kept at 37C. To normalize NAD(P)H levels across N4 and G4/T4 populations, cells were analyzed together to generate gates and determine frequencies of NAD(P)H high and low cells. To distinguish populations of differentially stimulated cells, cells were stained with either high or low CTY concentrations and sorted separately (Fig. 4F) or using 45.1/2 congenic markers and mixed prior to sorting (fig. S9D).

Seahorse analysis:

Analysis was performed on cells at the noted time points post-activation. Cells were plated at 1 × 10^5 cells per well in a 96-well Seahorse assay plate, pretreated with Cell-Tak (Corning). Cells were equilibrated to 37 °C for 30 min before assay. OCR (pmoles/min) and ECAR (mpH/min) were measured as indicated upon cell treatment with oligomycin (1.5 μM, Cayman), FCCP (1.5 μM, Cayman) or Bam15 (2.5 μM, Cayman), rotenone (1μM, Cayman), and antimycin A (1 μM, Sigma), according to the manufacturer’s instructions.

Listeria monocytogenes infection:

Transgenic Listeria monocytogenes expressing ovalbumin (Lm-Ova) was a gift from H. Shen. For infection, Lm-Ova was grown to early log phase (OD600 nm 0.1) in brain heart infusion media (Becton Dickinson), washed and diluted with sterile saline, and then inoculated intravenously (10^4 colony-forming units per mouse).

qPCR:

RNA was extracted using the Zymo Quick-RNA MicroPrep Kit without DNase I treatment according to the manufacturer’s protocol. 1 μg of total RNA was reverse transcribed into cDNA using Thermo Maxima H Minus Reverse Transcriptase, 100 U per 20 uL reaction with 25 pmol each oligo (dT)18 and random hexamer primers. RT reaction was incubated for 10 min at 25 °C, 15 min at 50 °C, and then terminated by heating for 5 min at 85 °C. cDNA was diluted 1:10 before being used in RT-qPCR with PowerTrack SYBR Green Master Mix on a Bio-Rad CFX384 real-time qPCR system. Nampt primers: forward, ACCAATGGCCTTGGGGTTAATG; reverse TCCCCGCTGGTGTCCTATGT. Technical triplicates were quantified using 2-ΔCt normalized to Pol2Ra.

Western Blot:

2 million cells were lysed ice cold buffer containing: 50 mM Tris, pH 8.0, 1 mM EDTA, 150 mM NaCl, 1% NP-40, 0.5% Na-deoxycholate, and 0.1% SDS, 10 mM NaF, 1 mM Na3VO4, and proteinase inhibitor cocktail (Roche, no. 11836170001). Samples were separated by SDS-PAGE, followed by transfer using the Trans-Blot Turbo Transfer System. Membranes were blocked with 5% dry nonfat dairy milk in TBS-T (0.1%) for 30–60 min. Primary antibodies were incubated in TBS + 1% Casein overnight at 4°C with shaking. 3–5 washes of 3–10 min were performed before secondary antibodies were incubated in 5% milk in TBS-T for 30–60 min at room temperature with rocking. Washes were performed before membranes were visualized with using SuperSignal West Pico PLUS or Femto Chemiluminescent Substrate and the ChemiDoc MP Imaging System. Antibodies used: NAMPT (1:1000, A300–372A, Bethyl Laboratories), pS6 (1:5000, D57.2.2E Ser235/236, Cell Signaling Technology), S6 (1:5000, 5G10, Cell Signaling Technology), pS6K (1:1000, 108D2, Cell Signaling Technology), S6K (1:1000, 49D7, Cell Signaling Technology), anti-actin hFAB Rhodamine (1:2000, Bio-Rad).

NAD Enzymatic Cycling Assay:

Cells were washed with PBS, pelleted, and flash frozen. Lysates were prepared with 0.6 M perchloric acid at 4℃. Cleared acid extracts were collected after pelleting insoluble material at 4℃. NAD standards (4 ul) or of diluted tissue acid extracts were combined with 100 μl of cycling reaction reagent: 2% ethanol, 100 μg/ml alcohol dehydrogenase, 10 ug/ml diaphorase, 20 uM resazurin, 10uM flavin mononucleotide, 10 mM nicotinamide and 0.1% BSA in 100mM phosphate buffer, pH 8.0. After 20–30 minutes, resorufin accumulation was measured with excitation at 544 nm and emission at 590 nm.

NADH FLIM:

NADH FLIM was performed on a LSM 710 (Zeiss) using a femtosecond-pulsed two-photon laser (Coherent) as described (65). In brief, cells were mounted on the microscope in a four compartment CELLview cell culture dish with a thin glass bottom (cat. #627870, Greiner) in 500 μL Tyrodes solution, with temperature maintained at 37°C. NADH was excited at 730nm (<5 mW on the sample, 1 min acquisition time) and the autofluorescence was detected through a 680 short-pass and 460/50 bandpass emission filter. Time-correlated single photon counting was performed using an HPM-100–40 detector and the SPCM 9.81 software (both Becker and Hickl). FLIM images were analyzed by SPCImage 8.0 using a biexponential decay with Ƭ1 and Ƭ2 being fixed to 400 and 2500 ps for free and protein-bound NADH. The fixed lifetime components were used to reduce the number of photons required for a valid lifetime calculation. Nuclear NADH was quantified using ROIs, averaging a minimum of 5 cells per image section. Mitochondrial NADH was quantified using the threshold function of SPCImage and verified using ROIs for individual conditions.

Statistical analysis

Representative data are shown from experiments that were repeated at least 3 times. Each data point plotted is the mean of at least 3 replicates per well with error bars representing the standard deviation (SD). The sample size and number of independent experiments are indicated in the figure captions. All statistical tests are described in the figure captions and were performed in Prism (GraphPad). No data were excluded. P values of 0.05 or less were considered statistically significant.