Metabolic support of tumor-infiltrating regulatory T cells by lactic acid

SUMMARY

Regulatory T (T_reg) cells, vital for maintaining immune homeostasis, also represent a major barrier to cancer immunity, as the tumor microenvironment (TME) promotes T_reg cell recruitment, differentiation, and activity^1,2. Tumor cells have deregulated metabolism leading to a metabolite-depleted, hypoxic, and acidic TME³, placing infiltrating effector T cells in competition with tumors for metabolites, impairing their function^4–6. Conversely, T_reg cells maintain high suppressive function within the TME^7,8. As previous studies suggested T_reg cells possess a distinct metabolic profile from effector T cells^9–11, we hypothesized the altered metabolic landscape of the TME and increased activity of intratumoral T_reg cells are linked. Here we show T_reg cells display broad heterogeneity in utilization of glucose metabolism within normal and transformed tissues and can engage an alternative metabolic pathway to maintain suppressive function and proliferation. Glucose uptake correlated with poorer suppressive function and long-term instability, and high glucose culture impaired T_reg cell function and stability. T_reg cells rather upregulate pathways in metabolism of the glycolytic byproduct lactic acid. T_reg cells withstood high lactate conditions, and lactate treatment prevented the destabilizing effects of high glucose, generating intermediates necessary for proliferation. T_reg cell-restricted deletion of MCT1, a lactate transporter, revealed lactate uptake is dispensable for peripheral T_reg cell function but required intratumorally, resulting in slowed tumor growth and increased response to immunotherapy. Thus T_reg cells are metabolically flexible: they can utilize ‘alternative’ metabolites in the TME to maintain suppressive identity. Further, our studies suggest tumors avoid destruction by not only depriving effector T cells of nutrients, but also metabolically supporting regulatory populations.

T_reg cells thrive under glucose restriction

Having previously shown heightened tumor cell metabolism correlates with decreased effector T cell function⁵, we sought to determine the relationship of heighted tumor metabolism with T_reg cell function. We measured the suppressor activity of T_reg cells isolated from Foxp3 reporter mice^12,13 bearing tumors of increasing glycolytic capacity, revealing glycolysis of tumor cells correlated with suppressive function of intratumoral T_reg cells (Fig. 1a). Expectedly, T_reg cells were enriched within tumors (Fig. 1b, Extended Data Fig. 1a,b) and were not metabolically suppressed but actively proliferating (Fig. 1c). Given the relationship between tumor glycolysis and T_reg cell function, we determined the glycolytic capacity of T_reg cells directly ex vivo and immediately after activation, a reprogramming event occurring downstream of the TCR in T cells¹⁴. TCR triggering induced less glycolysis in T_reg cells compared to naïve or antigen-experienced conventional T cells (Fig. 1d). This resistance to glycolysis persisted under strong activating stimuli (anti-CD3, CD28, and IL-2 for 48 h), and tumor-derived T_reg cells displayed even lower glycolytic activity (Extended Fig. 1c). Oligomycin treatment showed T_reg cells are resistant to engaging glycolysis even when oxidative metabolism is inhibited, suggesting glucose transport may be limiting in T_reg cells (Extended Data Fig. 1c). Indeed, T_reg cells had significant reductions in glucose uptake based on the fluorescent glucose tracer 2NBDG compared to T_conv (CD4⁺ Foxp3⁻) cells directly ex vivo (Fig. 1e, Extended Data Fig. 1d), regardless of CD44/CD62L expression (notably, the largest difference was observed between CD62L^hi populations) (Extended Data Fig. 1e). To directly compare activation-induced T_reg and T_conv glucose uptake, we adoptively transferred congenically marked OT-II CD4⁺ T cells from Foxp3^RFP mice into Vaccinia^OVA-infected hosts. OT-II T_reg cells took up significantly less 2NBDG than their Foxp3⁻ counterparts, despite sharing the same TCR and undergoing robust activation (Fig. 1f). While not selected on the OT-II TCR, activation by the same antigen in vivo under the same inflammatory context results in disparate metabolic phenotypes if Foxp3 is expressed.

Figure 1. Regulatory T cells possess a distinct metabolic profile from conventional T cells in normal and transformed tissues.

(a) (left) Representative histogram of T_conv responder cell proliferation dye dilution after 72hrs of co-culture with T_reg cells isolated from B16, Cl24 (melanoma), or MEER (HPV+ HNSCC) tumors. (right) Capacity of T_reg cells to suppress the proliferation of CellTrace Violet (CTV) labelled T_conv cells as a function of the glycolytic extracellular acidification rate (ECAR) of the tumors from which they were isolated (*p=0.047). (b) Percent T_reg cells (CD4⁺ Foxp3⁺) of total CD4⁺ T cells from lymph node (LN) and tumor infiltrating lymphocyte (TIL) preparations from C57BL/6 mice bearing B16 melanoma tumors (14 d post intradermal injection) (c) Percent proliferating (Ki67⁺) T_reg and T_conv cells (CD4⁺ Foxp3^-) from mice as in (b). (d) ECAR of in-Seahorse activated T_reg, CD8+, CD44+CD62L− and CD44-CD62L+ T_conv cells sorted from LN and spleen of Foxp3 reporter mice. Oligo = oligomycin, 2-DG = 2-deoxy-D-glucose, ΔECAR = max reading after αCD3 minus basal ECAR. (*p=0.022) (e) Flow cytogram depicting ex vivo 2NBDG uptake by T_reg and T_conv cells from the LN and B16 TIL of Foxp3 reporter mice. Representative plots gated on CD4+ cells. (**p=0.006) (f) Diagram of experimental procedure and quantification of ex vivo 2NBDG uptake by CD44⁺ OT-II T_reg and T_conv cells isolated 5 days after transfer into congenically mismatched hosts infection with Vaccinia-OVA. (g) Representative histogram and quantification of ex vivo GlucoseCy5 uptake by T_reg cells isolated from various tissues (*p=0.047). Results are representative of five (b), three (a,c,e, and g) or two (d,f) independent experiments. (*p< 0.05, **p <0.01, ***p < 0.001 ****p<0.0001) by unpaired two-tailed t test (a,b,d,f,g) or two-way ANOVA with Sidak’s multiple comparisons test (c,e). Data are mean values of biological replicates ±SEM.

To determine if low glucose uptake was a universal phenotype of T_reg cells, we interrogated the infiltrate of various tissues at normal and inflammatory states . 2NBDG is not only incompatible with YFP and GFP employed in Foxp3-driven Cre lines, but may have limited utility in some cells as a glucose tracer¹⁶. Thus, we synthesized a novel fluorescent glucose tracer (Cy5-linked 1amino-glucose; GlucoseCy5¹⁷), recapitulating our findings with 2NBDG while demonstrating superior specificity and sensitivity (Extended Data Fig. 1f–h). T_reg cell glucose uptake was quite heterogeneous in tissues of Foxp3^YFP-iCre mice, but notably low in B16 and imiquimod-inflamed skin (Fig.1g), suggesting inflammation may drive lower glucose uptake in T_reg cells. Conventional T cells possessed far less heterogeneity in glucose avidity when assayed from the same environments (Extended Data Fig. 1i).

Glucose avidity predicts T_reg cell function

We next purified T_reg subsets of low and high glucose avidity and measured their function. Regardless of tissue, 2NBDG avid T_reg cells displayed significantly reduced suppressive capacity compared to 2NBDG^lo T_reg cells (Fig. 2a, Extended Data Fig 2a), not attributable to viability, as 2NBDG^hi T_reg cells were rather more viable after the assay (Extended Data Fig 2b,c), and retained their initial proclivities for glucose (Extended Data Figure 2b,d). Thus, within a tissue, avidity for glucose predicts poorly suppressive T_reg cells. Sequencing 2NBDG low and high LN and B16-infiltrating T_reg cells revealed reduced expression of Foxp3, Ikzf2 (Helios), Il2ra (CD25), Nrp1, and other T_reg signature genes in 2NBDG^hi T_reg cells, further confirmed by flow cytometry (Fig. 2b, Extended Data Fig. 2e). These data suggested while glucose avid T_reg cells are still Foxp3⁺, they harbor a ‘weaker’ T_reg cell signature. As 2NBDG^hi T_reg cells harbored some suppressor function, we measured cytokine production after a brief restimulation. 2NBDG^hi T_reg cells produced significantly more IL-10, suggesting distinct suppressive mechanisms between 2NBDG T_reg states, especially intriguing given the differences in mRNA (Fig. 2b,c). These differences may be driven by glycolysis which is tied to cytokine translation in conventional T cells¹⁸. To assess long-term stability and function of 2NBDG T_reg cell subsets under inflammation, Thy1.2⁺ Foxp3⁺ T_reg cells of high or low glucose avidity were transferred into Rag1-deficient animals, along with Thy1.1⁺ Foxp3⁻ T_conv cells to induce colitis. After 12 weeks, mice receiving glucose avid T_reg cells began to lose weight due to increased colitis compared to those receiving 2NBDG^lo T_reg cells, and transferred 2NBDG^hi T_reg cells lost Foxp3 expression (Fig. 2d,e, Extended Data Fig. 2f). We also investigated whether increased glucose uptake could be observed in T_reg cells from an unmanipulated autoimmune setting. T_reg cells isolated from pancreatic islets (known to be dysfunctional¹⁹) and draining lymph node of pre-diabetic NOD.Foxp3^GFP mice took up significantly more glucose than those from non-draining lymph node (ndLN) (Fig. 2f). Thus, glucose avidity in T_reg cells correlates with poor suppressive capacity, ultimately leading to instability.

Figure 2. Glucose avidity is associated with reduced T_reg cell functional identity.

(a) Flow cytogram of CD4+ T cells from B16-bearing Foxp3^Ametrine mice depicting sorting strategies for T_reg cells based on their glucose uptake. T_reg cells sorted based on 2NBDG uptake were assayed for their ability to suppress the proliferation of CellTrace Violet (CTV) labeled T_conv cells. Representative histogram shows 1:4 ratio of T_reg:T_conv. (**p=0.0056). (b) T_reg cell signature gene expression in RNAseq analysis of 2NBDG^hi or 2NBDG^lo T_reg cells sorted from lymph node (LN) or tumor (TIL) of B16 melanoma-bearing Foxp3 reporter mice as in (a). (c) Production of IL-10, IFN-γ, and IL-2 by sorted 2NBDG^lo and 2NBDG^hi T_reg cells stimulated overnight with PMA/ionomycin and stained intracellularly (*p=0.02). (d) Diagram and colon histology Rag1^−/− mice that received either 2NBDG^lo or 2NBDG^hi Thy1.2⁺ T_reg cells plus Thy1.1⁺ T_conv cells I.V and followed for 7 wks. (e) Percent of transferred 2NBDG^lo or 2NBDG^hi T_reg cells as in (d) expressing Foxp3 within the spleen (Spln), mesenteric lymph node (mLN), intraepithelial layer (IEL), and lamina propria (LP) (*p=0.02, **p=0.004). (f) Representative histogram and tabulation of GlucoseCy5 uptake by T_reg cells isolated from the non-draining lymph node (ndLN), pancreatic lymph node (pLN), and the islet of 10-12-week-old NOD.Foxp3^GFP-Cre mice (**p=0.005). (g) (left) Diagram of experimental procedure and (right) capacity of T_reg cells conditioned in 0,5, or 25mM glucose media to suppress the proliferation of CellTrace Violet (CTV) labelled T_conv cells. (1:4 *p=0.031, 1:8 *p=0.012 between 0 and 25 mM). (h) Glycolytic pathway gene expression as in (b). Results are representative of three (a,b,c,g,h), or two (d,e,f) independent experiments. Significance (*p< 0.05, **p <0.01, ***p < 0.001, ****p<0.0001) was determined by unpaired two-tailed t test (a,c,d,e,f), or two-way ANOVA with Tukey’s multiple comparison test (g). Data are mean values of biological replicates ±SEM.

We next asked whether exposure to a high glucose environment would drive a less suppressive phenotype, regardless of glucose uptake status. Remarkably, conditioning T_reg cells in high glucose dampened suppressive function when later assayed in isoglycemic conditions (Fig. 2g, Extended Data Fig. 2g). We further interrogated the transcriptome of 2NBDG^lo T_reg cells to identify metabolic pathways that may support their function. 2NBDG^lo T_reg cells did not possess a transcriptome suggesting preferential lipid metabolism; rather, lipid metabolism genes were enriched in the 2NBDG^hi subset (Extended Data Fig. 2h). Both subsets expressed upstream enzymes in glycolysis, suggesting proclivity for glucose was not regulated at the level of transporter or enzyme expression (Fig. 2h). The 2NBDG^lo subpopulation, however, had enriched expression for the terminal steps of glycolysis: specifically lactate dehydrogenase (Ldha), and the monocarboxylate transporter MCT1 (encoded by Slc16a1) (Fig. 2h), confirmed by qPCR (Extended Data Fig 2i). While lactate is the end-product of glycolysis, it represents a significant fuel source for many cells²⁰. Lactate uptake is mediated through MCT1, where it is converted by LDH to pyruvate²¹. T_reg cells (but not T_conv cells) conditioned in low or glucose-deficient media increased expression of Ldha and Slc16a1 (Extended Data Fig. 2j). Thus, T_reg cells upregulate genes involved in lactate metabolism.

T_reg cells can metabolize lactic acid

Lactic acid is highly enriched in the TME²² (Fig. 3a) and known to be immunosuppressive^23,24, and, indeed, it curbed function of conventional T cells in vitro (Extended Data Fig. 3a,b). Conversely, both T_reg cell suppressor function and proliferation were resistant to tumor-equivalent concentrations of lactic acid (Extended Data Fig. 3a,c). Lactate can stimulate cells via GPR81²⁵ and enter as a metabolite via MCT1, so we asked if T_reg cells could consume lactic acid by measuring pH changes²⁶. Lymphocytes from Foxp3^YFP-Cre reporter mice were loaded with intracellular pH dye and incubated with lactic acid. T_reg cells, but not T_conv cells, took up lactic acid, evidenced by increased fluorescence (Fig. 3b). This assay revealed lactic acid uptake was also heterogeneous within various tissue-derived T_reg cells, akin to analysis of glucose uptake, notably low in tissues where T_reg cells were glucose avid, such as the liver (Fig. 1g, 3c). Indeed, most dye+ T_reg cells did not take up glucose (Extended Data Fig. 3d). In autoimmune NOD mice, islet T_reg cells took up less lactic acid than those from the ndLN or pLN (Fig. 3d). pH dye+ T_reg cells from LN and B16 TIL possessed increased CD44 and Nrp1 expression, while maintaining similar Foxp3 expression to dye negative cells (Fig. 3e). Thus T_reg cells are not only resistant to lactic acid but can take up this metabolite, and those that do display an activated T_reg cell signature and low glucose uptake.

Figure 3. T_reg cells metabolize lactic acid to support their proliferation and suppressor function.

(a) Lactate concentration in spleen or B16 melanoma interstitial fluid (*p=0.041). (b) Lymphocytes from Foxp3^Cre mice were loaded with pHrodo Red (intracellular pH dye) and pulsed with 5mM lactic acid. Results are change of MFI from t=0 (**p=0.0012). (c) Lactic acid uptake determined as in (b) by T_reg cells infiltrating various tissues. (d) Lactic acid uptake from T_reg and T_conv cells isolated from islets, non-draining (ndLN), and pancreatic lymph node (pLN) of 10-12-week-old NOD.Foxp3^GFP-Cre mice (*p=0.017, **p=0.001). (e) Flow cytogram and tabulation depicting Nrp1 and CD44 expression in T_reg cells based on lactate-elicited pH change (*p=0.018, **p=0.0097). (f) Diagram showing incorporation of ¹³C derived from lactate into downstream metabolites. (g) Relative abundance determined by mass spectrometry of intracellular lactate, pyruvate, malate, and citrate in T_reg and T_conv cells activated overnight then pulsed with uniformly labeled ¹³C-lactate (m+n equal to the number of incorporated heavy carbons) (**p=0.0025). (h) Relative abundance of PEP derived from ¹³C-lactate as in (g) (*p=0.036, **p=0.0011). (i) Capacity of T_reg cells conditioned for 3 days in 25mM glucose media ±10mM lactic acid to suppress the proliferation of CellTrace Violet (CTV) labelled T_conv cells (*p=0.021). (j) As in (h) with the addition of PEPCK inhibitor 3MP (*p=0.018). (k) Ki67 expression in intratumoral T_reg cells of B16 tumor-bearing mice treated with 3MP or water for 3 days (*p=0.015). (l) IFN-γ expression by B16-infiltrating CD8⁺ T cells from mice treated as in (k) for 5 days. Results are representative of three (a,c,e,g,h,k), or two (b,d,i,j,l) independent experiments. Significance (*p< 0.05, **p <0.01, ***p < 0.001, ****p<0.0001) determined by unpaired two-tailed t test (a,b,d,e,i,j,k,l) or unpaired one-tail t test (l) or two-way ANOVA with Sidak's multiple comparisons test (g,h). Data are mean values of biological replicates ±SEM.

To identify how T_reg cells were utilizing lactic acid, we performed high resolution mass spectrometry on activated T_reg and T_conv cells pulsed with [U¹³C]-L-lactate (at pH 6.9) (Fig. 3f). We used LN-derived T_reg cells (which do readily take up lactate, Fig. 3b) to facilitate cellular input needed for MS analysis, confirming T_reg cells took up significantly more lactate than T_conv cells (Fig. 3g). T_reg cells converted ¹³C-lactate into pyruvate and subsequently into citrate and malate indicating mitochondrial import and entry to the TCA cycle (Fig. 3g). Further analysis of revealed T_reg cells also incorporated lactate-derived carbon into phosphoenolpyruvate (PEP) (Fig 3h), formed by phosphoenolpyruvate carboxykinase (PEPCK) when malate leaves the mitochondria and is converted to oxaloacetate. PEP can contribute to upstream glycolytic intermediates essential for proliferation, suggesting lactate may serve as a gluconeogenic fuel source, decreasing a T_reg cell’s need for glucose. As high glucose can inhibit T_reg cell suppressor function (Fig. 2g), we asked whether lactic acid could mitigate the deleterious effects of glucose. Indeed, conditioning T_reg cells in high glucose plus tumor-equivalent concentrations of lactic acid maintained suppressor function (Fig 3i).

We then used the PEPCK inhibitor 3-mercaptopicolinic acid (3MP)²⁷ to dissect the individual contributions of lactate uptake and oxidation from upstream gluconeogenic reactions dependent on lactate-derived PEP. 3MP treatment of ¹³C-lactate pulsed T_reg cells reduced PEP accumulation without impacting incorporation into the TCA cycle (Fig 3j). In vitro 3MP treatment of T_reg cells in lactic acid significantly reduced their proliferation (Extended Data Fig. 3i). However, 3MP did not impact suppressive capacity of T_reg cells in vitro (Extended Data Fig. 3j). Treating tumor-bearing mice with 3MP decreased proliferation of tumor-resident, but not LN, T_reg cells, and increased intratumoral CD8 IFN-γ production (Fig. 3k,l). However, it did not overtly affect the intratumoral percentage of Foxp3⁺ cells, T_conv/CD8:T_reg cells ratios, overt tumor growth, or suppressor function when interrogated ex vivo (Extended Data Fig. 3f–h). Thus T_reg cells utilize lactic acid to not only feed the TCA cycle, but to generate PEP, critical for fueling T_reg cell proliferation within the tumor.

Lactic acid supports intratumoral T_reg cells

To determine the significance of lactic acid utilization by T_reg cells in vivo, we generated a constitutive, T_reg-specific deletion of MCT1 (Slc16a1^f/fFoxp3^Cre). qPCR confirmed Slc16a1 deletion, lack of compensation by Slc16a7/MCT2 or Slc16a3/MCT4 upregulation, and no deletion in CD4⁺Foxp3⁻ cells (Extended Data Fig. 4a,b). Expectedly, deletion of Slc16a1 resulted in loss of lactic acid uptake (Fig. 4a). These mice did not show any overt signs of autoimmunity or lymphoproliferation (Extended Data Fig. 4c,d). Rather, T_reg cells from the LN of Slc16a1^f/fFoxp3^YFPCre mice displayed similar suppressive capacity, proliferation, and expression of T_reg signature genes when assayed ex vivo (Extended Data Fig. 4e–h). MCT1 deficiency in T_reg cells revealed no significant differences in percentage of Foxp3⁺ cells infiltrating various tissues, although several compartments showed increases in glucose uptake at the steady-state (Extended Data Fig. 4i,j). Imiquimod treatment resulted in no increased inflammatory ear thickness, nor loss of T_reg cell markers in Slc16a1^f/fFoxp3^Cre mice (Extended Data Fig. 4k,l). We also determined long-term function and stability of MCT1-deficient T_reg cells in transfer colitis. Unlike 2NBDG^hi T_reg cells, MCT1-deficient T_reg cells did not show reduced stability (Extended Data Fig. 4m). However, they were not completely functional: at 7 weeks, mice receiving MCT1-deficient T_reg cells had more histologically severe colitis, although less severe than mice receiving 2NBDG^hi T_reg cells (Extended Data Fig. 4n,o). Thus, MCT1 deficiency appeared dispensable for T_reg cell survival and function, although under highly inflammatory conditions they may have decreased activity.

Figure 4. Tumor-infiltrating T_reg cells require lactate uptake to maintain their high suppressive function.

(a) Lactate-elicited pHrodo fluorescence in Foxp3^Cre or Slc16a1^f/fFoxp3^Cre T_reg cells (*p=0.041). (b) B16 tumor growth and survival of Foxp3^Cre and Slc16a1^f/fFoxp3^Cre mice (*p=0.0235) (c) Percent of B16-infiltrating PD-1⁺Tim-3⁺ CD8⁺ T cells day 14 post tumor injection in mice as in (a) (*p=0.017). (d) Proliferation of tumor infiltrating T_conv and CD8⁺ T cells as in (c) (**p=0.009). (e) Tumor-infiltrating CD8⁺ and T_conv cells restimulated with PMA/ionomycin and stained for IFN-γ (*p=0.039). (f) Capacity of WT or Slc16a1-deficient tumor-infiltrating T_reg cells to suppress proliferation of CTV labeled T_conv cells. (g) Percent proliferating tumor infiltrating T_reg cells from mice as in (c). (*p=0.011). (h) Percent or MFI of WT and Slc16a1-deficient T_reg cells expressing Nrp1, PD-1, CD44 from B16 tumors (TIL) as in (c) (*p=0.029, **p=0.009). (i) GlucoseCy5 uptake by B16 tumor-infiltrating T_reg cells as in (c) (j) Pimonidazole staining of WT and Slc16a1-deficient T_reg cells from B16 tumors as in (c) (*p=0.029). (k) Tumor growth of B16 (melanoma, **p=0.0011) MC38 (adenocarcinoma, *p=0.023), and MEER (HNSCC, *p=0.0146) in tamoxifen treated Foxp3^Cre-ERT2 and Slc16a1^f/fFoxp3^Cre-ERT2 mice. Mice were given tamoxifen 5 consecutive days prior to tumor injection then 3x per week. (l) Survival of tamoxifen treated Foxp3^Cre-ERT2 and Slc16a1^f/fFoxp3^Cre-ERT2 mice injected with B16 and treated with IgG or anti-PD-1 antibodies 3x weekly (**p=0.002). Results are representative of four (b,c,d,f,g,h(Nrp1, CD44),i,j), three (a, k(B16)), or two (h(PD-1), k(MC38, MEER),l) independent experiments. Significance (*p< 0.05, **p <0.01, ***p < 0.001, ****p<0.0001) determined by the log-rank test for survival curves (b,l) or unpaired two-tailed t test (a,c,d,e,g,h,i,j) or two-way ANOVA (b,f,k). Data are mean values of biological replicates ±SEM.

However, we reasoned MCT1-mediated lactate uptake would be most important within lactate-rich tumor tissue. Inoculating Slc16a1^f/fFoxp3^Cre mice with B16 melanoma resulted in slowed tumor growth and prolonged survival (Fig. 4b). Characterizing the infiltrate when both genotypes harbored tumors of similar sizes (day 14) showed while CD8⁺ T cells in Slc16a1^f/fFoxp3^Cre mice exhibited higher coinhibitory markers PD-1 and Tim-3, indicative of terminal differentiation (Fig. 4c), CD8⁺ and T_conv cells were more proliferative and competent to produce IFN-γ, suggesting decreased suppressive function by MCT1-deficient T_reg cells (Fig. 4d, e). Indeed, intratumoral MCT1-deficient T_reg cells had reduced suppressive function ex vivo (Fig. 4f) and were less proliferative, resembling intratumoral T_reg cells from 3MP-treated mice (Fig. 4g, 3k). Analysis of 3MP-treated Slc16a1^f/fFoxp3^Cre mice suggested 3MP’s effects on T_reg cells was mediated through MCT1 (Extended Data Fig 3k,l). Characterization of intratumoral MCT1-deficient T_reg cells revealed decreases in CD44 and Nrp1 staining, concomitant with elevated PD-1 staining, potentially indicating dysfunctional T_reg cells²⁸ (Fig. 4h). Metabolically, MCT1-deficient, tumor-resident T_reg cells compensated by becoming glucose avid (Fig. 4i). We found, as a consequence, that MCT1-deficient T_reg cells would no longer persist within the lactate rich hypoxic areas of tumors, as measured by pimonidazole staining, despite overall tumor hypoxia remaining similar (Fig. 4j, Extended Data Fig. 4p).

To remove any unforeseen effects of MCT1 deletion on T_reg cell development, we interrupted MCT1 expression just prior to B16 melanoma injection using a tamoxifen-inducible, T_reg cell-specific model (Slc16a1^f/fFoxp3^Cre.ERT2) and showed similar immunologic and survival outcomes, including the development of a ‘fragile’²⁹ T_reg cell phenotype in which Foxp3⁺ cells begin expressing IFN-γ (Extended Data Fig. 5). Induced T_reg deletion of MCT1 also slowed tumor growth of MC38 adenocarcinoma and MEER HNSCC (Fig. 4k, Extended Data Fig. 5g,h). We thus reasoned loss of lactate uptake in T_reg cells produced an environment conducive to immunotherapy. Triggering deletion of MCT1 in T_reg cells synergized with anti-PD-1 therapy, resulting in complete regressions in 37.5% of B16-bearing mice (typically insensitive to anti-PD-1) (Fig. 4l). Deletion of MCT1 from T_conv cells using Slc16a1^f/fCd4^Cre mice did not impact their capacity to produce IL-2 and IFN-γ, suggesting MCT1 was not generally required for T cell function (Extended Data Fig 4q). Thus, MCT1 expression and consequent lactic acid uptake is dispensable for most tissue derived T_reg cells but required to maintain high suppressor activity in the TME.

Our study highlights the antagonistic effects of glucose on T_reg cell function and identity. While previous data support this^9,10,23,30, some studies using transcriptomic and proteomic analysis suggest T_reg cells are highly glycolytic^31–33. We demonstrate T_reg glucose consumption is certainly heterogeneous between tissues (Fig. 1g) and within the T_reg population (Fig 2a); however, glucose uptake characterizes poorly suppressive T_reg cells, or at least those suppressing via mechanisms not potent in vitro coculture (Fig. 2a,c Extended Data Fig. 2a). Importantly, T_reg cells consuming low amounts of glucose still express glycolytic pathway genes (Fig. 2h). As most reactions in glycolysis are reversible, T_reg cells likely build higher-order intermediates from carbon sources like lactate (Fig. 3).

Lactate is often thought of as a waste product of glycolytic metabolism, but it represents an important metabolite for cellular function and a likely fuel source for tissues²⁰. Our data suggest T_reg cells do not require lactic acid for survival, but are endowed with metabolic flexibility to use this carbon source, both as fuel and as a means to protect their high suppressive capacity from negative effects of glucose (Fig 3i). Consequently, cells do not possess both high lactic acid and glucose uptake (Fig 1g, 2f, 3c,d, Extended Data Figure 3d). Thus, we propose a glucose:lactate axis exists to fine-tune T_reg cell functioning depending on the nutrient milieu (Fig 2g, 3i). Lactate not only fuels the TCA cycle but is also exported from the mitochondria, contributing to higher glycolytic pathways via PEP. This gluconeogenic component of lactate metabolism supports T_reg cell proliferation, critical in a cell type that does not readily take up glucose. However, the contribution of PEPCK-independent lactate metabolism appears also to be at the level of T_reg cell suppression (Fig. 4f), suggesting lactate plays a multi-pronged role in T_reg cell biology.

It is unlikely regulatory T cells evolved to thrive in tumors; rather cancers exploit this predilection for alternative substrates to maintain an immunosuppressive environment. Lactate is not only enriched in the tumor but is elevated in immunologically distinct tissue environments such as muscle, adipose, and nervous system^34,35. We also suspect lactate is one of a host of alternative metabolites T_reg cells can utilize. MCT1 itself can also transport acetate, succinate, propionate, and butyrate, which have been shown to play roles in T_reg cells^21,36,37. Indeed, our findings using transfer colitis suggest MCT1 deficiency may impact T_reg cell function in the gut. Notably, however, we did not observe overt gut inflammation at the steady state in Slc16a1^f/fFoxp3^Cre mice, suggesting both inflammation and MCT1 substrate availability play roles in maintaining T_reg cellsin that environment..

Our study suggests that, as T_reg cells must be active when immune activity is at its lowest, they exist metabolically ‘out-of-sync’ with their conventional counterparts. This requires utilizing fuels prevalent in tissue environments to maintain suppressor function when their conventional T cell targets may be competing for nutrients. As lactate can promote suppressive functions of other tolerogenic cell types, like tumor-associated macrophages³⁸, cell types sharing lactate metabolism may share suppressor activity. While this is exploited in the tumor microenvironment, causing suppressive cells to thrive, it is actionable: T_reg cell-specific deletion of the lactate transporter not only results in decreased tumor growth but synergy with checkpoint blockade immunotherapy. MCT1 inhibition to directly target lactate metabolism or inhibition of tumor acidity may break this metabolic symbiosis and lower the regulatory T cell barrier to cancer immunity.

MATERIALS AND METHODS

In Vitro Tissue Culture

B16-F10 were obtained from ATCC. MC38 cells were obtained from Dario Vignali (commercially available from Kerafast). Clone24 was originally produced by single cell sorting melanoma tumors arising from Pten^f/fBraf^LSL-V600ETyr^CreER mice⁵. MEER cells were obtained from Robert Ferris (originally from ³⁹). B16 cell line was authenticated in 2018 by independent sequencing. MC38 cells were authenticated in 2016 by independent sequencing. Clone24 was authenticated via metabolic profiling and tumor growth compared to Melan-A (normal melanocytes) in 2019⁵. MEER cells were authenticated in 2013 via immunoblot to show E6/E7 and Ras overexpression in MTEC cells. MC38 and MEER were confirmed mycoplasma free in 2016, B16 were confirmed in 2018, Clone24 in 2019. Both primary and immortalized cell lines were maintained in lab-made R10 media [RPMI1640, 10%FBS, 2mM L-glut, PenStrep, NEAA, 1mM Sodium pyruvate, 5mM HEPES, β-ME]. Cultures were incubated in temperature and partial pressure-stable conditions at 37°C and 5% CO₂. For in vitro conditioning experiments, isolated T_reg or T_conv cells were activated in complete R10 media with 0.1 ug/mL Phorbol 12-myristate 13-acetate (PMA) and 1ug/mL ionomycin (Sigma-Aldrich) with 1000U/mL IL-2 for ~20 hours at 37°C then placed in the conditions indicated in the figure with IL-2. For in vitro 3-mercaptopicolinic acid (3MP, Cayman Chemical) experiments a 250 μM concentration was used.

In Vivo Mice Studies

Sample sizes were not statistically predetermined but were chosen based on previous work in T_reg cell functional assays^8,29 and tumor immunology^4,5,8,14,29. These sample sizes are sufficient because they allowed for the determination of statistical significance between groups and minimized the number of animals or replicates needed for each experiment. Mice were placed into experimental group by nature of their genotype and/or if receiving treatment were randomized within a genotype. For experiments not involving mice, cells were randomized into experimental groups. Blinding was not possible as most of the data acquisition and analysis was done by a single person; two individuals performing every experiment was not feasible during the course of our study. C57BL/6J-Foxp3^YFPiCre, -Foxp3^GFP.Cre.ERT2, -Rag1^−/−, and -Thy1^a (Thy1.1 congenic) were purchased from The Jackson Laboratory. Slc16a1^f/f mice were a gift from Jeffrey Rothstein and Brett Morrison⁴⁰ used to generate Slc16a1^f/fFoxp3^GFP.Cre.ERT2 (termed Slc16a1^f/fFoxp3^Cre-ERT2) and Slc16a1^f/fFoxp3^YFP.Cre (termed Slc16a1^f/fFoxp3^Cre) mice. Foxp3^{FlpO Ametrine} mice were a gift from Dario AA Vignali. Spleens of OT-II Foxp3^RFP Thy1.1⁺ mice were a gift from Geoffrey Camirand. Animal work was done in accordance with the Institutional Animal Care and Use Committee of the University of Pittsburgh (protocol #17071235). All mice were housed in specific pathogen free conditions at an ambient temperature 20-26°C and humidity of 30-70% with a 12:12 hour light:dark cycle prior to use. The maximal tumor size of 15 mm in any direction was not exceed in any experiment. Both male and female mice were used in studies, from five (5) to ten (10) weeks of age.

Flow Sorting and Cytometric Analysis

Single-cell suspensions from murine tumors (day 14 post tumor inoculation), lymph node, liver, thymus, or spleen were derived through mechanical separation and 70 μM filter (Fisherbrand) passage. For hypoxia detection, mice were injected intravenously with pimonidazole (80mg/kg, Hypoxyprobe) in PBS 1.0 hour before sacrifice. Tumors dissolution was aided through lab-made tumor lysis buffer [ 2mg/mL Collagenase Type IV (Gibco), 2U/mL Dispase in Hanks’ Balanced Salt Solution (Stemcell Technologies), 10U/mL Deoxyribonuclease I (Sigma), serum free RPMI medium] injected with a 20G needle and incubated for 30 minutes at 37°C. Prior to sort tumor suspensions were purified by negative selection with MojoSort magnetic bead separation (BioLegend) and biotin-conjugated anti-mouse antibodies to CD105 (MJ7/18, Cat# 120403, Lot# B266720, BioLegend).The skin (ears of mice were used), were finely minced using scissors and resuspended in RPMI1640 (Gibco, Grand Island, NY) with 2.5 mg/mL collagenase XI (Sigma-Aldrich), 0.25 mg/mL hyaluronidase (Sigma-Aldrich), 0.1 mg/mL DNase (Sigma-Aldrich), 0.01 M HEPES (Sigma-Aldrich), and 10% FBS followed by incubation in a shaking incubator for 1 h at 37C at 250 rpm. Following incubation skin was subjected to mechanical separation and passed through a 70 μM filter. Similarly, muscles (quadriceps were used) were finely minced using scissors then resuspended in 0.05% type II collagenase (ThermoFisher) in HBSS for 30 min at 37 °C under constant agitation. Islets were isolated as previously described⁴¹. Pimonidazole was visualized using anti-pimonidazole antibodies (Hypoxyprobe, Cat# HP7-100, Lot# 5914) after 5 minutes of 4% PFA fixation followed by the FoxP3 Fix/perm kit (Biolegend). Staining for expression markers was performed with anti-mouse specific antibodies obtaining from the following companies; Biolegend: anti-CD4 (GK1.5, Cat# 100412, Lot# B184560, dil 1:1000), CD8 (53-6.7, Cat# 100707, Lot# B171971, dil 1:1000), CD39 (Duha59, Cat# 143806, Lot# B186014, dil 1:500), CD44 (IM7, Cat# 103032, Lot# B267976, dil 1:500), CD73 (TY/11.8, Cat# 127215, Lot# B260585, dil 1:500), CD279 (PD-1, 29F.1A12, Cat# 135221, Lot# B194160, dil 1:250), HAVcr-2 (TIM-3, RMT3-23, Cat# 119705, Lot# B224472, dil 1:250), Helios (22F6, Cat# 137220, Lot# B155089, dil 1:250), IFNγ (XMG1.2, Cat# 505842, Lot# B270630, dil 1:250), Ki67 (16A8, Cat# 652403, Lot# B208445, dil 1:250), CD304 (Nrp-1, 3E12, Cat# 145212, Lot# B175814 dil 1:200), TNFα (MP6-XT22, Cat# 506322, Lot# B218553, dil 1:500). BD Horizon: CD62L (MEL-14, Cat# 564109, Lot# 7341887, dil 1:500). eBioscience: FoxP3 (FJK-16s, Cat# 53-5773-82, Lot#2011698, dil 1:250). Intracellular staining was performed using FoxP3 Fix/Perm buffer set (BioLegend) as per manufacturer’s protocol and staining was performed on overnight at 4°C. Stained cells were analyzed on a LSRFortessa (BD). Cell doublets were excluded by comparison of side-scatter and forward-scatter width to area. Flow cytometry data and proliferation modeling were analyzed with FlowJo v10 software (Tree Star) and figures were produced in Prism v8 (GraphPad).

Glucose Uptake Assay

Single cell suspensions of tumor and lymph node and other tissues of ~1 million cells/mL were placed in serum free RPMI containing 40 μM 2NBDG or 0.4 μM Glucose-Cy5 for 25 minutes at 37°C. Glucose-Cy5, a Cy5-linked 1-amino-glucose tracer was synthesized in collaboration with Dr. Marcel Bruchez¹⁷. For OT-II experiments, CD4⁺ T cells were isolated via negative selection using magnetic bead separation as previously described⁴ then injected retroorbital into C57BL/6J mice. Simultaneously, recipient mice were injected I.P. with 1 x 10⁶ vaccinia-OVA viral particles and harvested 5 days later.

Microsuppression & Proliferation Assays

In vitro microsuppression assays were performed as previously described²⁹. Briefly T_reg cells (CD4⁺FoxP3⁺) populations were isolated via flow assisted sorting from tumor or lymph node from C57BL/6J-Foxp3^GFP.Cre.ERT2, -Foxp3^YFP.Cre mice or mice crossed to these strains. Responder cells (CD4⁺Foxp3⁻ T_conv) and antigen presenting cells (CD4⁻CD8⁻ APCs) were isolated via flow assisted sorting from the spleen of a C57BL/6J-Thy1^a mouse. T_reg cells were co-cultured for 72 hours at 37°C with APCs and CellTrace Violet (CTV, ThermoFisher Scientific) labeled responder cells at ratios from 1:2 (T_reg:responder) to 1:32 in complete RPMI media with 1ug/mL anti-CD3. Proliferation assays were performed by labeling isolated T_reg, T_conv, and CD8 T cells with CTV then activating at 1 million cells/mL using anti-CD3/CD28 Dynabeads according to the manufacture’s protocol (ThermoFisher Scientific) and IL-2 in complete RPMI media containing 0 to 10mM lactic acid (Fisher Bioreagents). Cell proliferation of responding cell populations was modeled using FlowJo v10 software (Tree Star).

RT-qPCR

Total RNA was extracted using Trizol reagent (Invitrogen) and cDNA was transcribed using a High Capacity cDNA Reverse Transcription Kit (Applied Biosystems), according to manufacturer protocol. Transcript levels were measured with SybrGreen (ThermoFisher Scientific) and using primers specific to genes of interest, including; Slc16a1: FWD 5’-GCAGTGTTAGTCGGAGCC-3’ REV 5’-GCGATCATTACTGGACGGC-3’, Ldha: FWD 5’-AGCTTCCATTTAAGGCCCCG-3’ REV 5’-TCTTTTGAGACCGCTAGTGC-3’ cDNA concentration was normalized per samples relative to β-actin. All experiments were performed in technical triplicates.

Cytokine Production Assay

Cell suspensions were stimulated in complete R10 media with 0.1 ug/mL Phorbol 12-myristate 13-acetate (PMA) and 1ug/mL ionomycin (Sigma-Aldrich) for ~20 hours at 37°C. Five hours prior to antibody staining, Golgi-Plug (BD Biosciences) was added to the samples. Cells were surfaced stained and then intracellular stained using FoxP3 Fix/Perm buffer set (Invitrogen) as per manufacturer’s protocol. Unstimulated controls were used to determine gating strategies.

Interstitial Lactate Quantification

Tumor and spleen were harvested and placed in empty 15mL conical tubes. Tissues were cut up with scissors then wrapped with a 5-micron Nylon filter paper (Sterlitech) and stuffed filter down into a 1.5 mL conical tube making sure the tissue did not touch the bottom. Tissues were centrifuged at 4000 rpm for 2 hours. Interstitial fluid was assayed for L-lactate concentration using a colorimetric detection kit according to manufacturer’s protocol (abcam).

Isotopic Flux Analysis

Cell suspensions at 1 million cells/mL in complete RPMI media were stimulated for 24 hours at 37°C with plate-bound anti-mouse CD3ε (10 μg for Treg, 3 μg for Tconv, clone 145-2C11, Cat#100239, Lot# B306297, Biolegend) and soluble anti-mouse CD28 (2 μg/mL, clone 37.51, Cat#102115, Lot#B283231, Biolegend) and IL-2 (1000U/mL T_reg 50U/mL T_conv). After activation media was exchanged for complete RPMI with 5mM uniformly labeled ¹³C-L-Lactate (Sigma Aldrich) and IL-2. For some experiments PEPCK inhibitor, 3-mercaptopicolinic acid (3MP, Cayman Chemical) was added at 250 μM. Media was then titrated to pH 6.88 with 1 N HCl. Suspensions were pulsed for 25 hours then washed 2x with room temperature PBS then resuspended in ice cold 80% methanol.

Metabolic quenching and polar metabolite pool extraction was performed using ice cold 80% methanol in water with 0.1% formic acid at a ratio of 400μL per 100μL. Deuterated (D4)-taurine and (D3)-lactate (Sigma-Aldrich) was added to the sample lysates as an internal standard for a final concentration of 100μM. The supernatant was cleared of protein by centrifugation at 16,000xg. 5μL of cleared supernatant was subjected to online LC-MS analysis.

Analyses were performed by untargeted LC-HRMS. Briefly, Samples were injected via a Thermo Vanquish UHPLC and separated over a reversed phase Phenomenex Kinetex C18+ column (2.1×100mm, 1.7μm particle size) maintained at 40°C. For the 20 minute LC gradient, the mobile phase consisted of the following: solvent A (1.5mM ammonium fluoride) and solvent B (100% acetonitrile). The gradient was the following: 0-12.0 min 5% B, to 1000% B, 12.0-15.0 min hold at 100% B, 15.0-15.1100% to 5% B, 15.1-20.0 min 5%B. The Q Exactive mass spectrometer was operated in polarity switching mode, using both positive and negative ion mode, scanning in Full MS mode (2 μscans) from 66.7 to 1000 m/z at 70,000 resolution with an AGC target of 3e6. Source ionization settings were 4.5/3.0 kV spray voltage respectively for positive and negative mode. Source gas parameters were 20 sheath gas, 10 auxiliary gas at 250°C, and 4 sweep gas. Calibration was performed prior to analysis using the PierceTM Positive and Negative Ion Calibration Solutions (Thermo Fisher Scientific). Integrated peak areas were then extracted manually using Quan Browser (Thermo Fisher Xcalibur ver. 2.7). Fully untargeted analysis was completed by using Thermo Compound Discoverer 3.0 software suite to extract and align molecular features. Graphs and statistical analyses (either t-test or ANOVA) were prepared with GraphPad Prism 7.0 (GraphPad Software, Inc., La Jolla, CA, USA).

In Vivo Tumor Growth and Therapy

Slc16a1^f/fFoxp3^Cre or Foxp3^Cre mice were inoculated with 1.5 x 10⁵ B16-F10 cells in complete RPMI media. To monitor tumor growth, tumors area was determined 3x weekly using digital calipers and stopped once tumors reached 15mm in any direction. For 3MP treated mice, Foxp3^Cre or Foxp3^Cre.ERT2 mice were inoculated with 1.5 x 10⁵ B16-F10 cells. Starting on day 10 or 12 post tumor inoculation mice were treated daily either with water or 25 mg/kg 3MP (Caymen Chemical) via intraperitoneal (I.P.) injection. On day 14 one hour prior to sacrifice, mice were given a final dose of 25 mg/kg 3MP or water. For 3MP tumor growth, treatment started day 7 post tumor injection and continued daily with 25 mg/kg 3MP or H₂O I.P. until tumors reached 15mm in any direction.

For the inducible deletion of MCT1 (Slc16a1) Slc16a1^f/fFoxp3^GFP.Cre.ERT2 or Foxp3^GFP.Cre.ERT2 mice were treated I.P. or P.O. with 1 mg of tamoxifen (T5648, Sigma) in corn oil (C8267, Sigma) from day −4 to 0. On day 0, mice were inoculated intradermally with 2.5x10⁵ B16-F10, MC38, or MEER cells in complete RPMI media and given a dose of 1mg Tamoxifen. Tamoxifen treatment continued 3x weekly until the conclusion of the experiment. On day 7, when tumors reached 1-10 mm² mice were treated with 0.2mg anti–PD-1 or hamster IgG isotype control (Bio X Cell) 3x weekly until the conclusion of the experiment.

pHrodo Red Lactic Acid Uptake Assay

Cell suspensions from lymph nodes and day 14 B16-F10 tumors were loaded with pHrodo Red AM (ThermoFisher Scientific) according to manufacturer’s protocol in a 20mM HEPES in PBS solution. Cells were surfaced stained for multicolor flow cytometry following the normal protocol in 20mM HEPES/PBS buffer. At the flow cytometer, lactic acid was spiked into each sample at a final concentration of 5mM pH ~6.7. Samples were read at 0, 10, and 30 mins after addition of lactic acid.

Extracellular flux analysis

Isolated T_reg and T_conv cells were plated on Cell-Tak coated Seahorse culture plates (250,000/well) in media consisting of minimal, unbuffered DMEM supplemented with 2mM glutamine. Basal rates were taken for 30 min, and then streptavidin-complexed anti-CD3^biotin at 3 μg/mL was injected and readings continued for 3 hrs. 2μM oligomycin and 10mM 2-DG were injected to obtain maximal and minimal ECAR values, respectively. Similarly, isolated T_reg and T_conv cells were activated using plate bound anti-CD3 (5 μg/mL) and soluble anti-CD28 (2 μg/mL) with 250 U/mL IL-2 for 48hrs then plated on Cell-Tak coated Seahorse culture plates (100,000/well) in media consisting of minimal, unbuffered DMEM supplemented with 2mM glutamine. Basal extracellular acidification rates were taken for 30 mins. Cells were stimulated with 10mM Glucose, 2μM oligomycin, 10mM 2-DG.

Transfer Colitis

The transfer colitis model was performed as previously described⁴². Briefly, 2NBDG^hi and 2NBDG^lo T_reg cells from Foxp3-Ametrine reporter mice or T_reg cells from Slc16a1^f/fFoxp3^Cre or Foxp3^Cre were isolated via flow assisted sorting (CD4-APC⁺ Foxp3-reporter⁺). T_conv cells (CD4⁺ CD25⁻) from C57BL/6J-thy1^a were isolated via flow assisted sorting. Rag1^−/− mice were simultaneously given 2 x 10⁴ sorted T_reg cells and 1 x 10⁵ thy1.1⁺ T_conv cells via retroorbital injection. Mice were weighed once a week for 7 weeks (experiment end). Sections of colons were fixed in 10% formalin and paraffin embedded. Slides were stained with hematoxylin and eosin by the University of Pittsburgh Pathology Department, and histology scores were blindly determined using criteria previously described⁴².

Tumor Histology

After pimonidazole pulsing, tumors were dissected and frozen at -80 deg C in Optimal Cutting Temperature Compound (OCT) (Tissue-Tek) and sectioned (Cryostat microtome). Tissue was fixed in histology-grade acetone (Fisher) at −20 deg C, then rehydrated in staining buffer, stained with hypoxyprobe (Hypoxyprobe, catalog# HP7-100Kit), and DAPI (Life Technologies), and mounted with ProLong Diamond AntifadeMountant (Life Technologies). Sections were imaged with an Olympus IX83 microscope and analyzed with ImageJ and NIS-Elements Imaging Software.

RNA Sequencing

1000 low glucose and high glucose consuming T_reg cells were sorted directly into lysis buffer in a 96-well plate. Immediately after sorting cDNA was generated using SMART-seq HT kit (Cat # 634456 Clontech) using 1000 cells. cDNA product was checked by Tape Station D5000 from Agilent technologies 2200 to make sure cDNA was successfully generated. Library construction was done using Nextrera XT kit # 15031942 from Illumina. 1ng cDNA was used in total volume 5ul. Sequencing was done using NextSeq 500 System. High Output 75 Cycles kit with run Parameter Paired Read 150 cycles (2X75).

Sequencing reads were trimmed for adapters using Cutadapt⁴³ prior to being aligned to Mus musculus reference genome (mm10) using the RNA-seq aligner HISAT2. Subread’s featureCounts function was used for gene level quantification and results were normalized to Transcripts Per Kilobase Million (TPM). Using the raw quantification, differential genes were found with the R package DESeq2 and statistical cutoffs of q-value <0.05 and |log2foldchange|>1.5.

Statistical Analysis

The data presented in the figures are mean ± standard error of the mean (S.E.M.). Multiple group comparison in in vivo and ex vivo assays was accomplished with one- or two-way analysis of variance (ANOVA). For single comparison, unpaired two-tailed Student’s t test was used, unless otherwise denoted. In cases of non-Gaussian distribution, nonparametric t-test was used. Survival and tumor growth in in vivo implantable tumor models are presented as Kaplan–Meier survival curves and plotted tumor area with respect to time and were statistically analyzed using log rank test. All analysis was completed with on Prism v5 software (GraphPad). A value of P < 0.05 is statistically significant. In the figures, standard designations of significance were given; *p < 0.05; **p < 0.01; ***p < 0.001 and ****p < 0.0001. The specific analysis used per figure in the manuscript can be found within the legends.

Data availability

RNA sequencing data that support the findings of this study (Fig. 2b, h) have been deposited in GEO with the GSE158801 accession code. The authors declare that the data supporting the findings of this study are available within the paper and its supplementary information files.

Extended Data

Extended Data Figure 1. 1-amino-Cy5-glucose (GlucoseCy5) can act as a surrogate for 2NBDG in GFP/YFP T_reg reporter mice.

(a) Gating strategy for T_reg and T_conv cells (b) Representative plot of percent T_reg (CD4⁺Foxp3⁺) in the lymph node (LN) and tumor (TIL) of B16 bearing C57BL/6 mice day 14 post tumor inoculation. Representative plots gated on total CD4⁺ cells. (c) Glycolytic extracellular acidification rates (ECAR) of T_reg and T_conv cells sorted from LN and B16 TIL preparations as in (b) 48hrs after activation with αCD3/CD28 and IL-2. Oligo = oligomycin, 2-DG = 2-deoxy-D-glucose, ΔECAR = max reading after glucose minus basal ECAR. Max ECAR = max reading at after oligo minus basal ECAR. (d) Ex vivo 2NBDG uptake mean fluorescence intensity (MFI) by T_reg, T_conv, and CD8⁺ T cells from the LN and B16 TIL (**p=0.0045). (e) Ex vivo 2NBDG uptake by CD44⁺CD62L⁻ and CD44⁻CD62L⁺ T_reg and T_conv cells isolated from the LN of Foxp3-Ametrine reporter mice. (f) Lymphocytes from Foxp3-Ametrine reporter mice were simultaneously pulsed with 2NBDG and GlucoseCy5. Representative plot is gated CD4⁺ Foxp3⁺ with tabulation of percent GlucoseCy5⁺ of 2NBDG⁺ T_reg. (g) Ex vivo GlucoseCy5 uptake by CD44⁺ T_reg and T_conv cells from the LN and B16 TIL. Representative plots gated on CD44⁺ CD4⁺ cells (**p=0.0025). (h) GlucoseCy5 positivity in Nrp1 negative and positive T_reg cells. (i) Ex vivo GlucoseCy5 uptake by T_conv cells isolated from various tissues. Results are representative of three (a-d,f-i) or two (e) independent experiments. Significance (*p< 0.05, **p <0.01, ***p < 0.001, ****p<0.0001) was determined by paired two-tailed t test (e), unpaired two-tailed t test (d,f,g,h) or two-way ANOVA with Sidak's multiple comparisons test (c). Data are mean values of biological replicates ±SEM.

Extended Data Figure 2. Glucose avid T_reg cells harbor a weaker T_reg cell signature but retain viability and some suppressor activity.

(a) T_reg cells were sorted based on GlucoseCy5 uptake and assayed for their ability to suppress the proliferation of CellTrace Violet (CTV) labeled T_conv cells at 1:4 (Treg:Tconv) (LN*p=0.02, TIL *p=0.041). (b) Experimental diagram for (c, d). (c) Representative histogram and quantification of viability of sorted 2NBDG^hi and 2NBDG^lo T_reg cells after 72hrs in a suppression assay. (d) 2NBDG uptake by 2NBDG^hi or 2NBDG^lo T_reg as in (b). (e) Representative histograms and tabulation of T_reg signature gene expression between 2NBDG low and high T_reg cell subsets (Nrp1 **p=0.0016, CD73 *p=0.029 **p=0.0076, TIGIT LN*p=0.04 TIL*p=0.016, CD25 *p=0.037). (f) Weights of Rag1^−/− mice that received an adoptive transfer of either 2NBDG^lo or 2NBDG^hi T_reg cells plus Thy1.1⁺ T_conv cells I.V. (g) Representative histogram of T_conv responder cell proliferation after 72hrs of co-culture with T_reg cells (1:4) conditioned in 0, 5, or 25mM glucose for 3 days. The suppression assay occurred in 25 mM glucose conditions. (h) Geneset enrichment plot of cellular lipid catabolic process from TIL 2NBDG^hi vs 2NBDG^lo T_reg cells. (i) mRNA expression of Slc16a1 and Ldha in LN- and B16-infiltrating T_reg and T_conv cells by qPCR (Ldha **p=0.004, Slc16a1 **p=0.009). (j) Slc16a1 and Ldha mRNA expression in T_reg or T_conv cells activated overnight and conditioned in the glucose concentration indicated for 3 days (normalized to 25mM glucose, significance between T_reg and T_conv) (Ldha **p=0.002 Slc16a1 5 *p=0.030, 0 *p=0.046). Results are representative of three (a,e,h,i,j), or two (c,d,f) independent experiments. Significance (*p< 0.05, **p <0.01, ***p < 0.001) determined by unpaired two-tailed t test (a,f,i,j) or paired two-tailed t test (c,d) or two-way ANOVA with Sidak’s multiple comparisons test (e). Data are mean values of biological replicates ±SEM.

Extended Data Figure 3. T_reg cells are resistant to lactic acid and use PEPCK-mediated metabolic pathways to support their proliferation.

(a) Proliferation of CD8⁺, T_conv, and T_reg cells labeled with CellTrace Violet (CTV) activated in media with lactic acid for 3 days (**p=0.0058). (b) IFN-γ production of CD8 and T_conv cells conditioned as in (a) then restimulated overnight with PMA/ionomycin (CD8 **p=0.0042, Tconv *p=0.02). (c) Suppression assay using T_reg cells conditioned as in (a) performed in absence of additional lactic acid. (d) Representative histogram and quantification of pHrodo⁺ T_reg cells taking up 2NBDG. (e) Representative flow plot of Ki67 expression by B16-infiltrating T_reg cells from mice treated ±3MP for 3 days. (f) Suppressive capacity of T_reg cells, isolated from mice as in Fig.3k. (g) Percent Foxp3⁺ cells, T_conv:T_reg ratio, and CD8:T_reg ratio within the TIL of mice treated as in (f). (h) Tumor growth curve of B16 in C57BL/6 mice treated with H₂O or 3MP. (i) Proliferation of T_reg and T_conv cells activated and cultured in 10mM lactic acid media ±3MP for 3 days (*p=0.016). (j) Capacity of activated T_reg cells conditioned in 10mM lactic acid media ± 250μM 3MP for 3 days to suppress the proliferation of CTV labelled T_conv cells. (k) IFN-γ expression by CD8 and T_conv cells in Foxp3^Cre or Slc16a1^f/fFoxp3^Cre mice treated as in (f). (l) Percent Foxp3⁺ cells, T_conv to T_reg ratio, and CD8 to T_reg ratio within LN and tumor of Slc16a1^f/fFoxp3^Cre mice treated as in (f). Results are representative of four (d), three (a,b,c,f,g,i), or two (h,j,k,l) independent experiments. Significance (*p< 0.05, **p <0.01, ***p < 0.001, ****p<0.0001) determined by unpaired two-tailed t test (d,e,g,i,k,l) or one-way ANOVA with Dunnett’s multiple comparisons test (a,b) or two-way ANOVA with Sidak's multiple comparisons test (c,f,h,j). Data are mean values of biological replicates ±SEM.

Extended Data Figure 4. MCT1 is efficiently deleted in T_reg cells of Slc16a1^f/fFoxp3^YFP-Cre mice and is not required for peripheral T_reg cell function.

(a) Slc16a1 expression in T_conv and T_reg cells from the lymph nodes (LN) of Foxp3^Cre or Slc16a1^f/fFoxp3^Cre mice. (b) T_reg cell mRNA expression of Slc16a3 (MCT4) from mice as in (a). CD44, Ki67, and IFN-γ production by LN CD8⁺ T cells (c, *p=0.039) or T_conv cells (d, **p=0.007) from mice as in (a). (e) Capacity of LN-derived T_reg cells as in (a) to suppress proliferation of CellTrace Violet (CTV) labelled T_conv cells. (f) Percent Foxp3⁺ and Nrp1⁺Helios⁺ LN T_reg cells from mice as in (a) (*p=0.013). (g) CD44, Ki67, and CD62L expression by LN T_reg cells as in (a). (h) CD39 and CD73 expression by LN T_reg cells as in (a). Percent of Foxp3⁺ cells (i) or GlucoseCy5⁺ T_reg cells (j) from various tissues of mice as in (a) (thymus *p=0.014, B16 *p=0.019). (k) Normalized ear thickness of imiquimod treated mice as in (a). (l) Percent marker positive T_reg cells from mice treated as in (k). (m) Foxp3 expression in transferred wild type or Slc16a1-deficient T_reg infused with Thy1.1⁺ T_conv cells into Rag1^−/− mice iv. (*p=0.02) (n) Weight of Rag1^−/− mice over time from (m). (o) Representative sections of the colon and quantified histology scores 7 weeks post transfer from mice as in (m) and Fig.2d (**p=0.001, **p=0.009). (p) Anti-pimonidazole area over tumor area (B16) calculated from immunofluorescence from Foxp3^Cre and Slc16a1^f/fFoxp3^Cre mice. (q) IL-2 and IFN-γ expression by Slc16a1-deficient T_conv cells stimulated overnight with PMA/ionomycin. Results are representative of four (c,d,f,g), three (a,b,e,h,j,p,q), or two (k,l,m,i,o) independent experiments. Significance (*p< 0.05, **p <0.01, ****p < 0.0001) determined by unpaired two-tailed t test (a-d,f-q) or two-way ANOVA (e). Data presented as mean values of biological replicates ±SEM.

Extended Data Figure 5. Acute deletion of MCT1 results in similar immunologic phenotypes in B16 melanoma and predilection towards a fragile T_reg cell phenotype.

(a) Foxp3^Cre-ERT2 and Slc16a1^f/fFoxp3^Cre-ERT2 mice were treated 5 consecutive days with tamoxifen I.P. prior to inoculation with 1.5 x 10⁵ B16 tumor cells. Following inoculation, tamoxifen was administered I.P. 3 times a week until sacrifice at day 14. (b) Percent Foxp3⁺ CD4⁺ cells in the lymph node (LN) and tumor (TIL) as in (a) (**p=0.005). (c) Tabulation of IFN-γ production by T_reg cells from the LN and TIL of mice as in (a) (**p=0.002). (d) Glucose consumption by T_reg cells from the TIL of mice as in (a). (e) Percent proliferating T_conv cells from the TIL of mice as in (a) (*p=0.037). (f) Representative flow plot and tabulation of IFN-γ and TNFα production by CD8 T cells from the TIL of mice as in (a) (**p=0.003, *p=0.02). Tabulation of IFN-γ production by T_conv cells from the TIL of mice as in (a) (**p=0.003). (g) Nrp1 mean fluorescence intensity on MEER intratumoral Treg cells derived from Foxp3^Cre-ERT2 and Slc16a1^f/fFoxp3^Cre-ERT2 mice treated as in (a) (*p=0.034). (h) Tim-3 expression by MEER derived CD8 T cells from mice as in (g) (**p=0.003). Results are representative of four (b,c,d,e), or three (f), or two (g,h) independent experiments. Significance (*p< 0.05, **p <0.01) determined by unpaired two-tailed t test (b-h). Data presented as mean values of biological replicates ±SEM.