Pyruvate dehydrogenase complex integrates the metabolome and epigenome in CD8+ memory T cell differentiation in vitro

Modulation of metabolic flux through pyruvate dehydrogenase complex (PDC) plays an important role in T cell activation and differentiation. PDC sits at the transition between glycolysis and the tricarboxylic acid cycle and is a major producer of acetyl-CoA, marking it as a potential metabolic and epigenetic node To understand the role of pyruvate dehydrogenase complex in T cell differentiation, we generated mice deficient in T cell pyruvate dehydrogenase E1A (Pdha) subunit using a CD4-cre recombinase-based strategy. Herein, we show that genetic ablation of PDC activity in T cells (TPdh−/−) leads to marked perturbations in glycolysis, the tricarboxylic acid cycle, and OXPHOS. TPdh−/− T cells became dependent upon substrate level phosphorylation via glycolysis, secondary to depressed OXPHOS. Due to the block of PDC activity, histone acetylation was also reduced, including H3K27, a critical site for CD8+ TM differentiation. Transcriptional and functional profiling revealed abnormal CD8+ TM differentiation in vitro. Collectively, our data indicate that PDC integrates the metabolome and epigenome in CD8+ memory T cell differentiation. Targeting this metabolic and epigenetic node can have widespread ramifications on cellular function.


INTRODUCTION
Pyruvate dehydrogenase complex (PDC) is a tripartite mitochondrial matrix enzyme which consists of pyruvate dehydrogenase (E1), dihydrolipoamide acetyltransferase (E2) and dihydrolipoamide dehydrogenase (E3). This enzyme complex is responsible for the oxidation of pyruvate to acetyl-CoA, the activated form of acetate, CO 2 and NADH, and serves as the link between glycolysis and the tricarboxylic acid cycle (TCA). The major metabolic fates of acetyl-CoA include the provision of carbon skeletons for the TCA cycle, and the biosynthesis of fatty acids and cholesterol.
In addition to its contributions to cellular energy, acetyl-CoA may also serve as a substrate for the post translational modi cation of histones. During histone acetylation, an acetyl group from acetyl-CoA is transferred to the primary amine in the ε-position of the lysine side chain. This epigenetic modi cation results in neutralization of positive electrostatic charge, ultimately affecting DNA access and transcription. Since one of the major sources of acetyl-CoA is glycolysis, it is not surprising to nd that histone acetylation is directly modulated by glycolytic ux and cellular metabotype 1 . Therefore, PDC may serve as a major node in T cells, integrating metabolism and epigenetics.
T cell activation and differentiation involve a series of coordinated steps involving metabolic, epigenetic and subsequently, transcriptional reprogramming [2][3][4] . Regarding metabolic reprogramming, activated T cells awake from their quiescent state of OXPHOS dependence to develop a Warburg-like metabotype.
Upon differentiation, this metabotype is retained (e.g., in ammatory T helper 1 cells) or recedes back to OXPHOS (e.g., CD8 + memory T cells (T M )). Following these rapid metabolic changes, T cells experience changes in chromatin accessibility and transcription, indicating that these processes are temporally linked and dependent 5 . As such, T cell activation and differentiation serves as an excellent in vitro model for studying the intersection between metabolic perturbations and epigenetics.
Despite its central position in metabolism, the role of PDC in integrating the metabolome and epigenome in T cells remains unclear. We hypothesized that ablation of PDC activity would have widespread metabolic and epigenetic consequences and lead to aberrant gene expression, ultimately impacting T cell differentiation. To understand the role of PDC in metabolism and its effects on the epigenome, we developed a mouse model of T cell Pdha1 (E1) de ciency using a cre recombinase-based strategy. In the present study, we de ned the effects of PDC de ciency on the metabolome and epigenome during CD8 + memory T cell differentiation in vitro. We speci cally chose this in vitro strategy to control for the contributions of extracellular metabolites (e.g., acetate) to T cell activation and differentiation 6 .

PDC de ciency in T cells: TPdh −/− mouse
To understand the effects of disruption of this critical metabolic node in T cells, we developed a model of T cell PDC de ciency by targeting Pdha using a CD4-cre recombinase. TPdh −/− mice display normal litter sizes, body weight and length and life span (data not shown). To con rm deletion of the Pdha locus, we performed qPCR on gDNA from splenic T cells. Pdha mRNA was not detected in TPdh −/− T cells (Fig. 1A).
Similar to humans, Pdha is encoded on the X-chromosome 7 . Therefore, to determine the e cacy of our cre-recombinase, we studied both sexes for the presence of PDHA by immunoblot (Fig. 1B). PDHA was absent in both male and female mice, enabling us to use both sexes for subsequent experiments. Finally, we wanted to pyruvate oxidized was disrupted in activated TPdh −/− cells. To answer this question, splenic T cells from WT and TPdh −/− mice were isolated and activated for 24 hours with CD3/CD28 stimulation. Extracellular ux analysis was performed where glucose was removed and replaced by pyruvate (Fig. 1C).
While WT cells readily oxidized pyruvate, TPdh −/− cells were impaired, consistent with a block at the level of PDC.
TPdh −/− cells are dependent upon glycolysis PDC is an important gatekeeper in metabolism, linking glycolysis and the TCA cycle. Since inhibition of PDC activity by PDK promotes aerobic glycolysis in culture and in vivo 8,9 , we predicted a similar increase in T cells with PDC de ciency. To pro le glycolysis, we performed extracellular ux analysis following glucose injection on activated (24 hours) WT and TPdh −/− splenic T cells (i.e., glycolytic stress test, Fig. 2A). Following glucose injection, the extracellular acidi cation rate rose promptly with TPdh −/− cells peaking about 50 mpH/min higher than WT. The addition of oligomycin to quantify glycolytic reserve resulted in a minimal increase in TPdh −/− T cells, indicating that these cells were operating at their glycolytic maximum. To con rm increased utilization of glycolysis, we anticipated an accumulation of glycolytic intermediates. To identify these points of substrate accumulation, we conducted metabolomic analyses of glycolytic intermediates on activated T cells. While other glycolytic and pentose phosphate pathway intermediates were similar to WT ( Figure S1A and S1B), TPdh −/− T cells displayed elevated levels of glucose-6-phosphate and fructose 1,6 bisphosphate, the products of two key regulatory enzymes of glycolysis, hexokinase and phosphofructokinase, respectively (Fig. 2B). The accumulations observed at critical metabolic checkpoints are consistent with enhanced glucose metabolism 10 .
The transition from glycolysis to the TCA cycle occurs in the mitochondria via PDC to produce acetyl-CoA, a critical metabolite for the TCA cycle 11 . To con rm the interruption of glycolytic carbon transfer into the TCA cycle, we next examined the incorporation of 13 C carbon from [U-13 C] glucose into TCA cycle intermediates (Fig. 2C). In WT, approximately 40% of the citrate pool was labelled as M + 2, indicating a considerable glucose-derived contribution to citrate through PDC (Fig. 2D). Consistent with the genetic ablation of Pdha, the M + 2 isotopologues of TCA cycle intermediates citrate, fumarate, and malate were essentially absent, with unlabeled (M + 0) intermediates comprising the predominant isotopologue ( Fig. 2D and S1C). The same was true for M + 2 aspartate, an amino acid derived from oxaloacetate. Cycling of the TCA from glucose derived carbon was also signi cantly depressed as re ected by the M + 4 citrate isotopologue (Fig. 2D, lower right). To de ne functional glucose dependence of activated T cells, we studied proliferation by incubating stimulated T cells with increasing concentrations of 2deoxyglucose ( Fig. 2E). While WT displayed a dose dependent effect, TPdh −/− was found to have signi cant inhibition of proliferation at all doses of 2DG. Overall, our results not only con rm PDC de ciency, but also de ne a functional dependence of glycolysis on activated TPdh −/− cells.

Reprogramming of mitochondrial metabolism in TPdh −/−
The TCA cycle generates intermediates for sugars, amino acids, nucleic acids, and lipids, and provides reducing equivalents for oxidative phosphorylation [12][13][14] . Since the commitment of glucose derived carbon to the TCA cycle and aerobic metabolism was disrupted, we hypothesized that this would lead to metabolic adaptations 15,16 . To de ne these metabolic adaptations, we rst pro led TCA cycle intermediates via metabolomics in CD3/CD28 activated T cells. Notably, one TCA cycle intermediate was markedly reduced in our metabolomics study. Succinyl CoA, the product of α-ketoglutarate dehydrogenase and substrate for succinyl CoA synthetase 17 , was decreased along with a signi cant increase in GTP (Fig. 3A). These ndings suggest generation of GTP by substrate level phosphorylation via succinyl CoA synthetase 18,19 . Most other TCA cycle intermediates were similar between TPdh −/− and WT. Based on these results we hypothesized that alternative carbon sources may be contributing to anaplerosis ( Figure S2A and S2B). Glutamine is a critical amino acid for maintaining the TCA cycle in the setting of limited pyruvate availability 20 . To pro le the incorporation of glutamine carbon into the TCA cycle in TPdh −/− , we employed [U-13 C] glutamine (Fig. 3B). Glutamine is converted to glutamate and subsequently α-ketoglutarate, the substrate that generates succinyl-CoA and downstream metabolites succinate, fumarate and malate. The M + 5 isotopomer of glutamate was increased in TPdh −/− consistent with increased incorporation of 13 C carbon ( Fig. 3C and S2C). Monitoring M + 4 isotopologues downstream showed enrichment of glutamine carbon in fumarate, malate, and aspartate (Fig. 3C, Figure  S2C). Other evidence of alternative carbon sources included a depression of multiple anaplerotic amino acids by metabolomics in TPdh −/− , with phenylalanine, tyrosine, isoleucine and valine being the most signi cantly affected ( Figure S2D).
Anaplerosis not only helps regulate rates of biosynthesis by augmenting substrate availability, but may also contribute to cellular energy status 17 . Speci cally, glutamine oxidation can maintain the TCA cycle in the setting of compromised mitochondrial pyruvate transport 20 . To test whether increased incorporation of glutamine carbon translated to enhanced OXPHOS in TPdh −/− , we next performed extracellular ux analysis on T cells activated as above. Increased incorporation into the TCA by glutamine did not result in enhanced OXPHOS, but rather a depression (Fig. 3D), suggesting that this amino acid did not contribute to cellular energy status via OXPHOS in activated TPdh −/− T cells. Based on our observations, we next sought to de ne OXPHOS by extracellular ux analysis. In activated TPdh −/− T cells, basal respiration, ATP synthesis, maximal respiration and spare respiratory capacity were all depressed (Fig. 3E). Based on our stable isotope and extracellular ux analyses, we suggest that in activated T cells, a portion of glucose is completely oxidized in the mitochondria. In addition, this oxidation of glucose may help set the pace for OXPHOS. In PDC, a depression in OCR was also seen when the long chain fat palmitate was used as a substrate, indicating that mitochondrial fatty acid oxidation was also reduced ( Figure S2E).
Despite signi cant depressions in FAO and OXPHOS in TPdh −/− T cells, total cellular ATP was similar to WT, suggesting that substrate level phosphorylation was su cient to account for the de cit (Fig. 3F). We next asked whether metabolites provided by the extracellular environment could overcome the effects of PDC de ciency in T M differentiation. Acetate, a precursor to ketone bodies produced during infectious states, has been shown to be involved in the acetylation of metabolic enzymes (e.g., GAPDH) and histones 6,24 . To test whether replacement of acetyl-CoA by acetate supplementation (10 µM) could aid in the differentiation of TPdh −/− T M cells, we performed in vitro differentiation with IL-15 as above. Since, acetate alone did not produce changes in Ly6C expression (data not shown), we also decided to target the upregulation in glycolytic metabolism. To accomplish this, we employed a lactate dehydrogenase inhibitor (LDHi, 25 µM) to suppress the upregulation of glycolysis and aid in the adoption of a T M metabotype. With the addition of acetate and LDHi, we saw a slight improvement in Ly6C, suggesting phenotypic skewing towards the T M phenotype (Fig. 4D). Interestingly, this improvement in T M skewing in TPdh −/− was not due to changes in the spare respiratory capacity of OXPHOS (Fig. 4E). These results suggest that the provision of acetate alone was not su cient and that glycolytic suppression was also required. Furthermore, the acetyl-CoA derived from acetate was involved in mechanisms of differentiation outside of bioenergetics. Since acetyl-CoA production is impacted by PDC de ciency, and is an essential component of gene regulation by histone modi cation, we next performed chromatin immunoprecipitation and sequencing (ChIPseq) studies by targeting acetylated histones. To visualize our genomic data, we constructed an MA plot ( Figure S4A). In the gure, we found that WT T M have nearly a log fold greater difference (purple) in bound sites when compared to TPdh −/− T M . This translated into a generally lower number of genomic reads for the top 50 genes and at all loci in general (Figs. 5C and 5D) for TPdh −/− . Although the number of acetylated sites was lower in TPdh −/− , the overall distribution of acetylated sites was similar between both groups ( Figure S4B). To con rm our ndings, we used ChIP PCR to probe several targets important for T M differentiation that were identi ed by our ChIPseq (Fig. 5E). Consistent with our ChIPseq results, TPdh −/− T M displayed a decreased ratio of acetylated target genes by ChIP PCR. These results were also consistent with our ATACseq results that showed limitations in chromatin accessibility for the aforementioned genes (Fig. 5F). Based on these ndings, we hypothesized that histone acetylation would be altered. To answer this question, we conducted a proteomic analysis for histones lysine acetylation. In general, TPdh −/− T M displayed decreased amounts of histone acetylation (Fig. 5G). In addition, histone H3 acetylation at lysine 27 (H3.1K27ac), an important marker associated with CD8 + memory T cell differentiation 28,29 , was also decreased. Overall, our results indicate perturbations in the epigenetic signature (i.e., activation and repression of loci) of TPdh −/− T M , which alters gene expression pro les and by extension differentiation.

DISCUSSION
Metabolites derived from intermediary metabolism play an important role in epigenetics and can mediate important health outcomes such as immunity. T cells undergo metabolic reprogramming following activation to develop a metabotype that is not only conducive to the bioenergetic and substrate needs of the cell, but also contributes to the epigenetic landscape. Herein, we studied the metabolic and epigenetic effects of disruption of PDC in T cells in vitro. PDC de ciency leads to widespread perturbations in glycolysis, mitochondrial metabolism, and the epigenome. The result is defects in T cell differentiation and changes in the response to extracellular metabolites. Our results indicate that glycolysis is a signi cant contributor to histone acetylation, and PDC serves as an important metabolic and epigenetic node in T cell differentiation.
Following engagement of the T cell receptor, pyruvate dehydrogenase kinase 1 (PDK1) becomes activated in T cells, leading to the phosphorylation and subsequent inhibition of PDC 30,31 . As a result, a smaller fraction of pyruvate (~ 40% by our stable isotope studies) is metabolized in the mitochondria, and T cells adopt a glycolytic metabotype. In our current model, TPdh −/− T cells lack a critical component of PDC, resulting in a de ciency of this enzyme complex. As a result of this block, pyruvate is not fully oxidized and subsequently, OXPHOS is downregulated (by ~ 48%). In response, TPdh −/− undergo metabolic rewiring and upregulate glycolysis as evidenced by our extracellular ux and metabolomic studies. Consequently, total cellular ATP levels are maintained via substrate level phosphorylation. This upregulation of aerobic glycolysis is dependent upon the regeneration of NAD + , a process which occurs in the cytoplasm via the conversion of pyruvate to lactate via lactate dehydrogenase (LDH) 32 . Indeed, our extracellular ux analyses support increased activity of LDH. Not only does this lead to an upregulation of glycolysis, but also a metabotype where glycolysis is operating at its maximum, unable to be pushed further. As a result, TPdh −/− T cells become functionally dependent upon glycolysis, as indicated by our proliferation studies with 2DG.
In aerobic organisms, the TCA cycle is a sequence of chemical reactions used to produce energy through the oxidation of acetyl-CoA derived from glycolysis, fatty acid oxidation or amino acid metabolism 33 . In our TPdh −/− T cell model, the TCA cycle undergoes metabolic rewiring involving anaplerosis due to a de ciency of acetyl-CoA from glycolysis. As a mechanism to maintain homeostasis, glutamine becomes essential in this case of loss of glycolytic carbon sources for the TCA cycle 34 . In our study, TPdh −/− T cells showed a depletion of multiple ketogenic amino acids (isoleucine, phenylalanine, tyrosine), as well as increased incorporation of glutamine into the TCA cycle as measured by stable isotopes. However, glutamine incorporation did not result in enhanced OXPHOS, indicating that its function lies beyond bioenergetics. One such important function may be the synthesis of aspartate from oxaloacetate seen in TPdh −/− . Aspartate synthesis in the setting of OXPHOS de ciency becomes an important pathway for producing DNA, RNA and protein in proliferating cells 35 . Furthermore, anaplerosis may also be enhanced by OXPHOS de ciency, leading to excessive anaplerosis 36 .
Since metabolism is intricately tied to T cell differentiation, it was not surprising to nd abnormalities in TPdh −/− CD8 + T cells. CD8 + T cells are highly energetic and have a requirement for intact OXPHOS. Unlike CD4 + T cells, activation of CD8 + T cells does not result in a complete shift to aerobic glycolysis 37 . In fact, OXPHOS levels increase and are an important source of ATP needed for cell proliferation. Therefore, impaired OXPHOS and enhanced glycolysis seen in PDC de ciency are more consistent with T E cells and may partially account for this distinct phenotype seen in TPdh −/− T M .
Beyond metabolic reprogramming, CD8 + T cell differentiation also involves epigenetic and subsequently, transcriptional reprogramming 2-4 . PDC de ciency leads to a de ciency of acetyl-CoA, an important substrate for histone modi cation 38 . Histone modi cation results in the activation and repression of key genetic loci involved in differentiation. The importance of acetyl-CoA derived from glycolysis in differentiation has also been reported in a number of cellular systems. For example, glycolysis-mediated changes in acetyl-CoA and histone acetylation control differentiation in embryonic stem cells 38 . In TPdh −/ − T M cells, histone acetylation was markedly depressed as shown in our proteomic and ChIP studies, suggesting that glycolysis is a signi cant source of acetyl-CoA in these cells. Therefore, the de cits seen in TPdh −/− differentiation are mediated by metabolic and epigenetic perturbations.
Interestingly, our ndings presented herein were in contrast to a recent paper utilizing genetic and pharmacologic inhibition of the mitochondrial pyruvate carrier (MPC) 39 . Wenes et al. described a metabolic-epigenetic axis that enables CD8 + memory T cell formation 39 . Histone H3 acetylation at lysine 27 (H3K27ac) is a marker of active chromatin regions associated with memory CD8 + T cell differentiation 28,29 . Inhibition of the MPC resulted in H3K27 acetylation, however the carbon source switched to glutamine, instead of glucose. In our study, not only was overall acetylation depressed, but H3K27 Page 9/24 acetylation was also absent in our histone proteomic study. Although we do not have a direct explanation for these ndings, it is worthwhile to point out the differences between these two models. Inhibition of the MPC via pharmacologic or genetic means may have different effects on metabolism when compared to PDC inhibition. For example, MPC1 +/− mice employ fatty acid oxidation (FAO) to meet their bioenergetic needs 40 . TPdh −/− T cells displayed depressed FAO. Beyond its effects on metabolism, MPC1 also engages in signaling transduction, interacting with mitoSTAT3 41 . Therefore, targeting different aspects of pyruvate metabolism may result in divergent phenotypes.
In summary, our data demonstrate that PDC de ciency leads to metabolic and epigenetic perturbations, affecting CD8 + memory T cell differentiation in mice. Based on our ndings, we propose that PDC occupies a major node in T cell intermediary metabolism by mediating both biochemical and epigenetic responses in activation and differentiation. Immunoblot studies For analysis, approximately 20 µg of protein was loaded on 4-20% Tris-glycine polyacrylamide gels and run at 150V for 1.5 h. Transfer to polyvinylidene di uoride membrane was done using the Trans-Blot Turbo Transfer System (Biorad, Hercules, CA). The membranes were blocked 1h room temperature in proprietary buffer (LI-COR Biosciences, Lincoln, NE). The membranes were probed with PDH and phospho-PDH (Abcam, San Francisco, CA) and b-actin (Sigma-Aldrich, St. Louis, MO). After washing the membranes three times (10 min each) with TBS 0.1% Tween 20, incubation with IRDye secondary antibodies was performed (LI-COR Bioscience, Lincoln, NE). Image capture and analyses were accomplished using an Odyssey Imager (LI-COR Bioscience, Lincoln, NE).

Real time PCR
Extracted RNA (Pure link RNA mini kit, Thermo Fisher Scienti c) was reverse transcribed to cDNA iScript Kit ( BioRad). Reactions were cycled and quantitated with an ABI 7500 Fast Real Time PCR System (Applied Biosystems).

Metabolomics
Randomly selected mice were euthanized in a carbon dioxide chamber followed by cervical dislocation and spleens were extracted. Isolated splenic T cells were sent for metabolomic analyses by Clarus Analytics (SanDiego, CA).
Stable Isotope studies T cells were stimulated for 24 hours with immobilized anti-CD3 and anti-CD28. All labeling experiments were performed with 1 million cells/mL in RPMI. Glycose free or glutamine free media were replaced by their respective uniformly 13 C-labeled analog (i.e. [U-13 C]glucose or [U-13 C]glutamine; Cambridge Isotope Laboratories). Cells were cultured for 24 hours and then pelleted, and lysed in cold 50% methanol. Analyses were performed at the CRI Metabolomics Core, UT Southwestern. Lysates underwent three freeze-thaw cycles, followed by centrifugation to remove debris. The supernatants were evaporated, methoximated and derivatized by tert-butyl dimethylsilylation. Derivatized material (1 µL) was injected onto an Agilent 6970 gas chromatograph equipped with a fused silica capillary GC column (30 m length, 0.25 mm diameter) and networked to either an Agilent 5973 or 5975 Mass Selective Detector. The measured distribution of mass isotopologues was corrected for natural abundance of 13 C 42 .

Extracellular ux analysis
Oxygen consumption rate (OCR) and extracellular acidi cation rate (ECAR) were determined using a Seahorse XF96 analyzer (Agilent). T cells activated for 24 hours with anti CD3 and anti CD28 were attached with Cell-Tak (Corning) at 0.2 million cells/well in Seahorse Base Medium Minimal DMEM supplemented with 12mM glucose, 2mM glutamine and 1mM sodium pyruvate. Mitochondrial parameters were monitored using the Mitostress kit (Seahorse Biosciences) according to the manufacturer's standard protocol.

Flow cytometry
Single-cell suspension of tissues were prepared. Anti-CD4, CD8, Ly6C, CD62L, CD25 antibodies were purchased from BD Biosciences or ebioscience. Labeled tetramers (NIH tetramer core facility) were used to identify viral speci c T-cells. Data were acquired on CytoFLEX Flow Cytometer (Beckman Coulter) and analyzed using FlowJo software (Tree Star). Cells were loaded with 2µM CTV (ThermoFisher Scienti c) and proliferation was estimated on day 3 by FACS. Cells were stimulated with biotinylated anti CD3 and crosslinked with streptavidin (Sigma-Aldrich). Flow cytometry gating strategy is depicted in Figure S4.
In vitro differentiation Antibodies were purchased from BioXcell. OT-I cells were activated with OVA-peptide for 3 days. To differentiate into T E or T M , cells were cultured in the presence of IL-2 or IL-15 (10ng/mL) for 4 days, respectively 23 . Sodium acetate (25µM) and LDH inhibitor GSK2837808A (10µM), (Tocris, Bristol, UK) were added to differentiation media.
T cell killing assays Splenocytes (10 6 /mL) from OT1 mice were stimulated in RPMI + 10% FCS with 1µM OVA peptide for 3 days, washed and cultured with IL2 as above for 3 more days. Targets EL-4 cells loaded with Cell Trace Violet (C34557) at 2uM in PBS for 10 minutes then quenched and washed. Cells were pulsed with 1uM SIINFEKL peptide for 30-60 minutes at 37 degrees. Target cells were incubated with activated OT-1 cells, at varying ratios (1:5, 1:10, 1:20 in 200 µL media for 4 hours. Stained w/APC-anti-CD8 and live/dead for 30 minutes on ice. Acquired on Beckman CytoFLEX cytometer.

RNAseq
Poly-A selected RNA-seq libraries were constructed from 1 µg total RNA using the Illumina TruSeq RNA Sample Prep Kits, version 2. The resulting cDNA was fragmented using a Covaris E210. Library ampli cation was performed using 11 cycles to minimize the risk of over ampli cation. Unique barcode adapters were applied to each library. Libraries were quantitated by qPCR using the KAPA Library Quanti cation Kit (KAPA Biosystems) and pooled in an equimolar ratio. The pooled libraries were sequenced on a NovaSeq 6000 with version 1 chemistry. At least 90 million 150-base read pairs were generated for each individual library. Data was processed using RTA 3.4.4.

ATACseq
Tagmented DNA samples were ampli ed to add single indexed adapters using the Kapa HiFi PCR MasterMix (Roche). The nal libraries were twice puri ed using Ampure XP PCR Puri cation Beads (Agencourt). The libraries were pooled and then quantitated by qPCR. The pool balance was checked by performing a MiSeq run using a MiSeq Nano kit, version 2. The percentage of each library in the pool was determined from the demultiplexing and was used to rebalance the pool before sequencing.
The pooled libraries were sequenced on an SP ow cell on a NovaSeq 6000 using version 1.5 chemistry to achieve a minimum of 61 million 101 base read pairs. Raw sequence data were processed using RTA version 3.4.4. ATACseq reads were trimmed using Trimmomatic (v. 0.39) to ensure removal of adapter and transposase sequences. Adapter sequences compiled for Trimmomatic (Nextera-PE-PE.fa) were used for adapter trimming, with the option 'ILLUMINACLIP:NexteraPE-PE.fa:2:30:10:8:true'. ATACseq reads were aligned to the mouse GRCm38/mm10 reference genome sequence (EMSEMBL) using BWA MEM (v. 0.7.17). ATACseq reads that mapped unambiguously to one genomic site were retained by ltering out reads tagged with 'XA:Z:' or 'SA:Z:' in the SAM le generated by BWA. ATACseq reads that mapped to blacklisted regions of GRCm38/mm10 (ENCFF547MET.bed; https://www.encodeproject.org/ les/ENCFF547MET/), were removed due to di culty in accurately mapping reads to these genomic regions. MAC2 software (v. 2.2.7.1) was used to identify regions of open chromatin from the ATACseq data, with options '-g mm' to specify genome size and '-f BAMPE' to specify paired-end mapping of ATACseq data. ChIPseq Acetylated histones were immunorecipitated using anti-acetyllysine histone antibodies (ab1191, Abcam, Waltham, MA ) for ChIPseq or ChIP PCR. For ChIPSeq, libraries were constructed from 50 ng of ChIP DNA using Ovation Ultralow System V2 1-96 with 15 cycles of PCR ampli cation. The nal libraries were twice puri ed using Ampure XP PCR Puri cation Beads (Agencourt). The libraries were pooled and then quantitated by qPCR. The pool balance was checked by performing a MiSeq run using a MiSeq Nano kit, version 2. The percentage of each library in the pool was determined from the demultiplexing and was used to rebalance the pool before sequencing.
The pooled libraries were sequenced on a NovaSeq 6000 using version 1.5 chemistry to achieve a minimum of 37 million 51-base reads. The data were processed using RTA version 3.4.4.
First, quality control checks were performed using FastQC (0.11.9) (http://www.bioinformatics.babraham.ac.uk/projects/fastqc/) on raw sequence data in fastq format. The Phred scores for all samples were above 30, so we continued with the next steps. The fastq les were aligned to the mouse GRCm38/mm10 reference genome sequence (ENSEMBL) using BWA-mem (0.7.17).
An important issue with ChIP-seq data concerns the inclusion of multiple mapped reads where the same reads are mapped to multiple loci on the reference genome. Including multiple mapped reads increase the number of usable reads and the sensitivity of peak detection; however, the number of false positives may also increase. Thus, we ltered the output BAM les using Samtools (1.15.1 ) view to retain only uniquely mapping reads. Blacklisted regions are largely comprised of sequences like major satellite repeats. These regions will show aberrantly high signal in all samples, thereby skewing normalization and often adding meaningless peaks. Thus, the reads overlapping with backlisted regions were removed from bam les using Samtools view. The locations of the blacklisted regions were downloaded from the ENCODE project (https://www.encodeproject.org/ les/ENCFF547MET/). Next, we used MACS2 (2.2.6) to call broad peaks to identify areas in the genome that are enriched with aligned reads next to position of protein (histone) bound to DNA. Further, Diffbind (3.2, http://bioconductor.org/packages/release/bioc/html/DiffBind.html) was used to identify the differentially enriched peaks between wildtype and PDH-mutant samples. The peak pro les were annotated using ChIPseeker (1.28.3) in R (version 4.1.0). For visualization of peaks in UCSC Genome Browser, the bam les were rst sorted and indexed using Samtools sort and Samtools index respectively and the bedGraph les were converted to bigwig les using UCSC bedGraphToBigWig tool.

Histone proteomics
Posttranslational modi cations of histones were performed using previously published methods 43 .
Brie y, core histones were extracted puri ed using the "Histone Puri cation Mini Kit" (#40026, Active Motif, Carlsbad, CA) according to the manufacturer's instructions. Histones were acid-extracted, enriched on ion-exchange columns and desalted by perchloric acid precipitation. The puri ed histones were resuspended in HPLC-grade dH2O (1.0 µg/µL). Proteomic analyses was performed by the Mass Spectroscopy Section of NCI (Bethesda, MD).

Statistical analyses
All experiments were repeated 3 or more times and summary or representative data were presented as appropriate. All measurements were taken from discrete samples. Statistical analyses were performed using Prism (Graphpad Software). Summary statistics were generated for all data. Two-sided unpaired Student's t-test was used for comparing two groups where the populations followed a normal distribution, similar variance, and were sampled independently. P-value of < 0.05 was statistically signi cant. Means were represented by a single line with standard error of the mean for variation. Figure 1 Mouse model of T cell pyruvate dehydrogenase complex de ciency. A) Pdha DNA in splenic T cells from TPdh -/-. CD4 + cre-recombinase was used to target T cells for deletion of Pdha locus. qPCR for Pdha was performed. N = 4 mice/condition. B) Immunoblot for PDHA from TPdh -/-T cells. Total protein was extracted from splenic T cells. Immunoblots were probed for PDHA and normalized to actin (N = 3 / condition). C) Pyruvate oxidation in activated T cells. T cells were activated for 24 hours with CD3/CD28 and cultured in glucose free media supplemented with pyruvate as a carbon source. Extracellular ux analysis was performed. *** P < 0.001, **** P < 0.0001. Central line = mean, error bars = standard error of the mean. antibodies. Glycolytic stress test was performed. B) Metabolomics for glycolytic intermediates. Splenic WT and TPdh -/-T cells were stimulated for 24 hours as above. Cells were harvested and sent for metabolomic analysis. Metabolites were normalized to cellular protein levels. N = 10 mice/condition. C) D) Isotopologue labelling for citrate, malate, and fumarate. T cells were stimulated as above for 24 hours in the presence of 13 C-glucose (N = 3 /condition). E) Glycolytic dependence in proliferating cells. Splenic T cells from WT and TPdh-/-T cells were stimulated as above and incubated with increasing concentrations of 2-deoxyglucose (2DG). Proliferation was measured by Cell Trace Violet (CTV) dilution via ow cytometry (N = 3 mice/condition). Representative of 3 or more experiments. Error bars = SEM. * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001. Central line = mean, error bars = standard error of the mean. treatment, T cells were treated with acetate (10 mM) and a lactate dehydrogenase inhibitor (LDHi, 25 mM). Ly6C was determined by ow cytometry. E) Extracellular ux analysis of TM cells supplemented following treatment as in D). N = 5-6 mice/condition. Error bars = SEM. Flow cytometry and cell killing graphs are representative of multiple experiments. Experiments were repeated 3 or more times.

Figure 5
Genomic studies of T M cells. Splenic T cells were differentiated into T M cells using established protocols.
RNA was extracted and submitted for RNAseq (N = 5 mice/condition). A) Volcano plot demonstrating differentially expressed genes. B) Gene ontology (GO) for differentially expressed genes. Gene ratio is the percentage of total differentially expressed genes in the given GO term. C) ChIPseq for T M cells. Splenic T cells were differentiated as indicated above. Acetylated histones were immunopreciptated and DNA was sent for sequencing (N = 4 / condition). Top 50 genes detected by ChIPseq. D) Log 2 normalized reads for ChIPseq. E) ChIP PCR of select targets identi ed by ChIPseq. Acetylated histones were precipitate as above. Genes involved in T M differentiation which were also detected in ChIPseq were ampli ed using