Metabolic reprogramming of donor T cells enhances graft-versus-leukemia effects in mice and humans

Abstract

Acute myeloid leukemia (AML) relapse after allogeneic hematopoietic cell transplantation (allo-HCT) has a dismal prognosis. We found that T cells of patients relapsing with AML after allo-HCT exhibited reduced glycolysis and interferon-γ production. Functional studies in multiple mouse models of leukemia showed that leukemia-derived lactic acid (LA) interfered with T cell glycolysis and proliferation. Mechanistically, LA reduced intracellular pH in T cells, led to lower transcription of glycolysis-related enzymes, and decreased activity of essential metabolic pathways. Metabolic reprogramming by sodium bicarbonate (NaBi) reversed the LA-induced low intracellular pH, restored metabolite concentrations, led to incorporation of LA into the tricarboxylic acid cycle as an additional energy source, and enhanced graft-versus-leukemia activity of murine and human T cells. NaBi treatment of post–allo-HCT patients with relapsed AML improved metabolic fitness and interferon-γ production in T cells. Overall, we show that metabolic reprogramming of donor T cells is a pharmacological strategy for patients with relapsed AML after allo-HCT.

INTRODUCTION

The major cause of death after allogeneic hematopoietic cell transplantation (allo-HCT) is relapse of the underlying hematological malignancies (1). Whereas some hematological malignancies such as chronic myeloid leukemia respond well to donor lymphocyte infusion (DLI), others are refractory to this type of cellular immunotherapy. Retrospective, multicenter studies using DLI to treat acute myeloid leukemia (AML) relapsing after allo-HCT reported complete response rates of only 17% at day 100 after treatment (2). A prospective, nonrandomized, monocenter study using cytarabine-based chemotherapy followed by DLI described a 2-year overall survival (OS) rate of 19% (3). A large retrospective, multicenter study published a 2-year OS upon DLI treatment of 20% versus 9% upon chemotherapy-only treatment (4). In aggregate, these reports not only show that donor T cells can exert antileukemic effects in a small fraction of patients but also show that strategies to improve these graft-versus-leukemia (GVL) effects are urgently needed.

The failure of the T cells in the allo-HCT recipient to prevent the relapse and the inability of the T cells contained in DLI to eliminate leukemia cells may be related to leukemia cells that impair activation and metabolic activity of T cells. A recent study has shown that the mitochondrial fitness of T cells dictates their ability to eliminate malignant cells (5). Metabolic competition for nutrients and the production of metabolites that paralyze T cell function in close proximity to the malignant cells were shown to promote cancer progression (6). To better understand how AML cells affect T cell function and metabolism, we studied T cells at primary diagnosis of AML and in patients who underwent allo-HCT. T cells isolated from patients relapsing with AML after allo-HCT displayed reduced glycolytic activity and oxidative phosphorylation (OXPHOS) compared to T cells from the same patients harvested at a time point before relapse, when the AML was still in remission. These changes in metabolic fitness of T cells correlated with diminished cytokine production and increased abundance of lactic acid (LA) in the serum of the patients. In mouse models, LA reduced the activity of glycolysis-related enzymes in T cells and impaired their proliferative capacity and antitumor immunity. Antagonizing acidosis with sodium bicarbonate (NaBi) improved metabolic fitness and T cell function and led to leukemia control in mice. On the basis of the preclinical data, patients suffering from relapse were treated with oral NaBi. The treatment enhanced the respiratory capacity and effector cytokine production in patients’ T cells. On the basis of the low toxicity profile of NaBi, our findings allow for clinical intervention to enhance GVL effects mediated by DLI in patients relapsing with AML after allo-HCT.

RESULTS

CD8⁺ T cells isolated at relapse exhibit reduced mitochondrial spare respiratory capacity and interferon-γ production compared to the status at remission

To characterize the metabolic activity of T cells in the presence of AML cells, we isolated CD8⁺ T cells from the peripheral blood of healthy volunteers and of patients at primary diagnosis of AML, before any treatment was given, at a purity of over 95% (fig. S1, table S1, and data file S1). We observed no difference with respect to extracellular acidification rate (ECAR), as a parameter for glycolytic activity, between the two groups (Fig. 1A). Oxygen consumption rate (OCR), an indicator of OXPHOS, was also not different between the two groups (Fig. 1B). Besides the intact metabolic activity, T cells isolated from patients at primary diagnosis exhibited increased interferon-γ (IFN-γ), tumor necrosis factor–α (TNFα), and perforin expression compared to healthy individuals (Fig. 1, C to E). These findings suggest that AML cells do not affect T cell metabolism and cytotoxic function in the absence of allogeneic immune pressure.

Fig. 1. Characterization of CD8⁺ T cells of patients with AML during disease progression from primary diagnosis to relapse after allo-HCT.

(A) ECAR of CD8⁺ T cells isolated from the peripheral blood of patients with AML at the time of primary diagnosis compared with samples from healthy individuals [healthy control (HC)] (n = 19). n.s., not significant. (B) OCR of CD8⁺ T cells (n = 19). (C) Flow cytometry–based analysis of CD8⁺ T cells (Tc). Representative fluorescence-activated cell sorting (FACS) plots (left) and statistical analysis (right) of the percentage of cells expressing IFN-γ (HC, n = 18; AML, n = 17). (D) Representative FACS plots (left) and statistical analysis (right) of the percentage of cells expressing TNFα (HC, n = 16; AML, n = 17). (E) Representative FACS plots (left) and statistical analysis (right) of the percentage of cells expressing perforin (HC, n = 18; AML, n = 17). (F) OCR measured at baseline and in response to oligomycin, carbonyl cyanide p-trifluoromethoxy-phenylhydrazone (FCCP), etomoxir (ETO), and rotenone + antimycin A (R/A) in CD8⁺ T cells isolated from one patient during remission and at the time of relapse after allo-HCT. (G) OCR of CD8⁺ T cells isolated from the peripheral blood of patients with AML during remission and at the time of relapse after allo-HCT (n = 21). (H) OCR of CD8⁺ T cells isolated from the peripheral blood of patients with AML at two time points (tp1 and tp2) in remission after allo-HCT (n = 17). (I) ECAR of CD8⁺ T cells isolated as described in (G) (n = 21). (J) ECAR of CD8⁺ T cells isolated as described in (H) (n = 17). (K) Representative FACS plots (left) and statistical analysis (right) of the percentage of CD8⁺ T cells expressing IFN-γ, isolated as described in (G) (n = 10). (L) Representative FACS plots (left) and statistical analysis (right) of the percentage of CD8⁺ T cells expressing IFN-γ, isolated as described in (H) (n = 18). Each data point represents the measurement of an individual patient at the indicated time point; n values represent individual patients. P values were determined using the two-sided Student’s unpaired t/Mann-Whitney (HC versus AML) or paired t/Wilcoxon (two time points of one patient) test.

To understand the impact of AML relapse on metabolic activity of T cells, we isolated T cells from patients with AML after allo-HCT during remission and when relapse occurred (tables S2 and S3). T cells were collected within 11 to 60 days before relapse or at the time point of hematological relapse after allo-HCT in the same patient. In agreement with reduced mitochondrial fitness, OXPHOS was reduced in T cells isolated from patients at the time point of hematological relapse after allo-HCT as compared to the time point when the same patient was still in remission (Fig. 1, F and G). This reduction of OXPHOS was not seen when CD8⁺ T cells were isolated at two time points (tp1 and tp2, comparable time span as in remission/relapse samples) in patients with AML who remained in remission after allo-HCT (Fig. 1H and tables S4 and S5). In addition, glycolytic activity (ECAR) was reduced in T cells isolated from patients at the time point of hematological relapse after allo-HCT as compared to the time point when the same patient was still in remission (Fig. 1I). Reduction of ECAR was not observed when CD8⁺ T cells were isolated at two time points in patients with AML who remained in remission after allo-HCT (Fig. 1J). We calculated OCR and ECAR changes of individual patients and found a strong correlation between these values, suggesting a general reduction in metabolic activity in relapsed patients (fig. S2). IFN-γ, a surrogate parameter for cytotoxicity, was reduced in the CD8⁺ T cells at the time point of relapse (Fig. 1K). IFN-γ was not reduced in the CD8⁺ T cells of patients after allo-HCT without a relapse (Fig. 1L). These findings indicate that AML relapse in patients induces a profound change of glycolytic activity and OXPHOS in CD8⁺ T cells, which is associated with reduced IFN-γ production in these cells.

Murine AML cells inhibit glycolytic activity and transcription of glycolysis/OXPHOS-related genes in T cells in vitro and in vivo

To understand whether the AML-induced effects on T cells were also seen in established in vivo murine GVL models, we isolated CD8⁺ T cells from AML-bearing allo-HCT mice or mice undergoing allo-HCT alone (Fig. 2A). Glycolysis and OXPHOS of the donor T cells were reduced in AML-bearing allo-HCT mice compared to allo-HCT control mice (Fig. 2, B and C). For further studies on the mechanism by which leukemia cells influence T cell metabolism, we cocultured murine CD8⁺ T cells and murine AML (WEHI-3B) cells during a 48-hour activation phase with exposure to CD3/CD28 beads. We observed reduced glycolytic activity of T cells cocultured with AML cells compared to T cells cocultured with the same number of T cells (Fig. 2, D and E). To understand whether this effect was mediated by soluble factors or cell contact dependent, we exposed T cells to AML supernatant. We observed that AML supernatant was sufficient to cause the metabolic changes in T cells (Fig. 2, D and E). In addition, OCR as an indicator for OXPHOS was reduced in CD8⁺ T cells in the presence of AML cells or supernatant (Fig. 2F). In contrast to AML cells, myeloid dendritic cells (DCs) did not affect glycolytic activity or OXPHOS (Fig. 2, G and H). Transcriptomic analysis of T cells isolated after exposure to AML-conditioned medium identified OXPHOS and glycolysis under the top 10 down-regulated gene sets in T cells cultured with AML medium compared to the controls (Fig. 2I and fig. S3). Both pathways exhibited a lower expression of most of the genes in these pathways (Fig. 2, J and K). These findings indicate that AML cells exert a profound effect on glycolysis and OXPHOS of T cells via a soluble factor.

Fig. 2. Murine T cells are metabolically impaired by AML cells in vitro and in vivo.

(A) Recipient mice (BALB/c background) were transplanted with 5 × 10⁶ C57BL/6 bone marrow (BM) cells (allo BM) with or without additional 1 × 10⁵ syngeneic leukemia cells (WEHI-3B luc) and 2 × 10⁵ C57BL/6 T cells (allo Tc). Tumor growth was monitored by bioluminescence imaging (BLI). d0, day 0. (B) Analysis of ECAR of reisolated CD8⁺ T cells (n = 6, means ± SEM). (C) Analysis of OCR (n = 6, means ± SEM). The experiment was performed six times (with 18 mice in total); n values represent biologically independent experiments, pooling together three mice for each experiment. P values were determined using the one-sample Student’s unpaired t test. (D) ECAR measured at baseline and in response to glucose, oligomycin, and 2-deoxyglucose (2-DG) for CD8⁺ T cells isolated from 48-hour cell culture under CD3/CD28 stimulation either alone or in coculture with AML cells or incubated in medium originating from 48-hour AML cell culture. Graph is representative of four biologically independent experiments. (E) ECAR of CD8⁺ T cells (n = 4 to 6). (F) OCR of CD8⁺ T cells (n = 4 to 6). (G) ECAR of CD8⁺ T cells isolated from 48-hour cell culture and stimulated with CD3/CD28, from T cells incubated alone, in coculture with BM-derived dendritic cells (BMDCs), or T cells incubated in medium originating from 48-hour BMDC cell culture (n = 6). (H) OCR of CD8⁺ T cells (n = 6). The n values represent biologically independent experiments (means ± SEM). P values were determined using one-way ANOVA. (I) Microarray-based analysis of gene expression of CD8⁺ T cells isolated after 48-hour CD3/CD28 stimulation and incubated in fresh medium or medium originating from 48-hour AML or BMDC cell culture. Bar plot depicts the top 10 down-regulated gene sets in T cells cultured in AML medium compared to T cells cultured in DC medium. MTORC1, mammalian target of rapamycin complex 1. (J) Tile display shows differentially regulated OXPHOS genes as determined from linear model analysis. Only regulated genes with an adjusted P < 0.05 were selected. The color scale represents the row-wise Z score of gene expression with red being the highest and blue being the lowest. (K) Tile display shows differentially regulated glycolysis genes.

Reduced metabolic fitness and proliferation of AML-exposed T cells led to reduced GVL effects in multiple mouse models

To clarify the mechanisms responsible for the loss of GVL effect, we studied mitosis-related gene expression of T cells exposed to medium alone or AML (WEHI-3B) supernatant. Gene expression analysis of T cells isolated after exposure to AML-conditioned medium for 2 days exhibited a disturbed expression of genes related to mitosis (Fig. 3A). Functional studies on cell cycle activity (gating strategy in fig. S4) showed that T cells isolated after exposure to AML-conditioned medium exhibited a lower cell cycle activity with more cells being in G₀ phase (Fig. 3B). T cell proliferation measured by cell trace dilution was reduced (Fig. 3C). In vivo expansion of luc-transgenic T cells exposed to AML medium was reduced compared to T cells exposed to T cells only (Fig. 3D). These findings indicate that AML-derived soluble factors affect T cell cycle progression and proliferation.

Fig. 3. Proliferative and metabolic unfitness of T cells is accompanied by reduced antitumor activity.

(A) Microarray-based analysis of gene expression in CD8⁺ T cells cultured as in Fig. 2I. Tile display shows differentially regulated mitosis related genes as determined from linear model analysis. The color scale represents the row-wise Z score of gene expression, with red being the highest and blue being the lowest. (B) Cell cycle analysis of CD8⁺ T cells incubated in fresh medium or medium originating from 48-hour AML cell culture. Representative FACS plots (left) and statistical analysis (right) of four biologically independent experiments (means ± SEM) are shown. P value was determined using two-way ANOVA showing P value comparing G₀ phases. (C) Proliferation assay of T cells after 72 hours of culture in conditions described in (B). Representative histograms (left) and statistical analysis (right) of percentages of proliferating cells in five biologically independent experiments (means ± SEM) are shown. P value was determined using two-sided Student’s unpaired t test. (D) Recipient mice (BALB/c background) were transplanted with 5 × 10⁶ C57BL/6 BM cells (allo BM). C57BL/6 luc CD8⁺ T cells (allo Tc) were isolated from 48-hour cell culture described in (B), and 3 × 10⁵ CD8⁺ T cells were transplanted into each mouse. T cell expansion was monitored by BLI. Representative BLI images (left) and statistical analysis (right) of photon signals on day 6 after BMT pooled from two independent experiments with n = 10 per group, P value was determined using Mann-Whitney test. (E) Recipient mice (BALB/c background) were transplanted with 5 × 10⁶ C57BL/6 BM cells (allo BM) with or without additional 1 × 10⁴ syngeneic leukemia cells (WEHI-3B luc). C57BL/6 CD8⁺ T cells (allo Tc) were isolated from 48-hour cell culture described in (B), and 2 × 10⁵ CD8⁺ T cells were transplanted into each mouse. Survival was monitored. Data were pooled from two independent experiments with n = 5 mice per group in the BM group and n = 10 in the BM + AML groups. P values were determined using the log-rank (Mantel-Cox) test. (F) Recipient mice (C57BL/6 background) were transplanted with 5 × 10⁶ BALB/c BM cells (allo BM) with additional 5 × 10³ splenocytes derived from a syngeneic mouse carrying MLL^PTD/+;Flt3^ITD/+ mutated tumor. BALB/c CD8⁺ T cells (allo Tc) were isolated from 48-hour cell culture described in (B), and 2 × 10⁵ CD8⁺ T cells were transplanted into each mouse. Survival was monitored. Data were pooled from two independent experiments with n = 10 mice per group. P values were determined using the log-rank (Mantel-Cox) test. (G) Recipient mice (Rag2^−/−Il2rγ^−/− background) were transplanted with human leukemia cells (MOLM-13 luc). C57BL/6 CD8⁺ T cells (allo Tc) were isolated from 48-hour cell culture described in (B), and 2 × 10⁵CD8⁺ T cells were transplanted into each mouse. Survival was monitored. Data were pooled from two independent experiments with n = 12 mice per group. P values were determined using the log-rank (Mantel-Cox) test.

To understand whether the AML cell–mediated metabolic changes in T cells affect their ability to eliminate AML cells in vivo, we next transferred T cells exposed to medium alone or AML supernatant into WEHI-3B AML-bearing mice. Mice receiving the AML cells after allo-HCT died of the leukemia within 40 days (Fig. 3E). Conversely, transfer of T cells isolated from a culture without AML medium improved survival and led to complete leukemia elimination in a fraction of mice (Fig. 3E). Transfer of T cells exposed to AML medium had no impact on survival and on leukemia elimination (Fig. 3E). This was repeated in a leukemia model driven by internal tandem duplication in FMS-like tyrosine kinase 3 and partial tandem duplication of the mixed-lineage leukemia gene (FLT3-ITD/MLL-PTD), showing that the prior exposure of the donor T cells to the leukemia-conditioned medium led to loss of the GVL effect (Fig. 3F). In a humanized MOLM-13 xenograft leukemia model, T cells exposed to leukemia-conditioned medium were not able to protect from leukemia-related death (Fig. 3G). These findings indicate that the factors derived from the leukemic cells in vitro had an inhibitory impact on T cell–mediated control of leukemic cells.

AML-derived LA is responsible for the effects on T cells

To detect differences in metabolite concentrations in the AML supernatant, which may account for the loss of the GVL effect, we analyzed the AML (WEHI-3B) supernatant by magnetic resonance imaging and mass spectrometry for metabolites. The concentration of multiple metabolites increased or declined in the AML supernatant compared to the control (fig. S5, A to G). Mass spectrometric analysis of cell culture supernatants revealed LA to be highly abundant in samples with AML supernatant present compared to control medium or supernatant derived from a T cell culture (Fig. 4A). The results could be reproduced by nuclear magnetic resonance (NMR) as a second method to study metabolites in the medium. Again, we observed a similar pattern, with elevated lactate concentrations when AML cells but not normal myeloid DCs were part of the coculture (Fig. 4B). Increased lactate was also seen when AML cells were cultured alone (in the absence of T cells) in vitro (Fig. 4A and fig. S5H). The increased LA abundance found in the AML culture was associated with a lower pH, consistent with acidification of the medium, which was not observed when normal myeloid DCs were added to the culture (Fig. 4C). To understand whether LA could play a role in patients, we studied patients with AML at remission and at relapse after allo-HCT. We observed that LA increased in patients with AML relapse after allo-HCT but not in patients with AML who were in remission after allo-HCT (Fig. 4, D and E), supporting the concept that LA release is relevant for immune escape in the presence of allogeneic immune pressure. Accordingly, patients at primary diagnosis had lower LA in the plasma compared to patients with AML relapse (fig. S5I). The amount of LA correlated with the blast count in the peripheral blood of patients with AML relapse (fig. S5J). To test the functional role of LA, we added increasing concentrations to the T cell culture and observed reduced glycolytic activity (Fig. 4, F and G). Comparable to the AML supernatant, LA reduced cell cycle progression and T cell proliferation (Fig. 4, H and I). In contrast to LA, the addition of hydrochloric acid (HCl) (fig. S5K) did not reduce T cell proliferation, indicating that extracellular acidification alone was not responsible for the observed effects of LA (Fig. 4I). Consistent with the microarray data of T cells incubated in AML medium (Fig. 2I), mammalian target of rapamycin complex 1 signaling, for instance the phosphorylation of p70S6 kinase (S6K), was altered in LA-treated T cells, as were other key signaling pathways (Fig. 4J and fig. S6).

Fig. 4. LA is increased in the AML supernatant inhibiting T cell proliferation.

(A) Relative abundance of LA in fresh RPMI compared to T cell supernatant after 48-hour cell culture in fresh medium or medium originating from 48-hour AML cell culture or in 48-hour AML cell culture supernatant determined by liquid chromatography–mass spectrometry (LC-MS). Statistical analysis of six to nine biologically independent experiments (means ± SEM) is shown. P values were determined using one-way ANOVA, ****P < 0.0001. (B) Concentration of lactate in fresh RPMI compared to T cell culture supernatants after 48-hour culture alone, with AML cells or with BMDCs, or in medium originating from 48-hour AML/BMDC cell culture, determined by NMR analysis. Statistical analysis of seven to nine biologically independent experiments (means ± SEM) using Kruskal-Wallis test is shown. (C) pH measurements of T cell culture supernatants after 48-hour culture in fresh medium or medium originating from 48-hour AML/BMDC cell culture. Statistical analysis of three to six biologically independent experiments (means ± SEM) is shown. P values were determined using one-way ANOVA. (D) Relative abundance of LA in the serum of patients with AML during remission and at the time of relapse after allo-HCT obtained by LC-MS (n = 7). (E) Relative abundance of LA in the serum of patients with AML at two time points in remission after allo-HCT obtained by LC-MS (n = 16). Each data point in (D) and (E) represents the measurement of an individual patient at the indicated time point; n values represent individual patients. P values were determined using the two-sided Student’s paired t test. (F) ECAR measured at baseline and in response to glucose, oligomycin, and 2-DG for CD8⁺ T cells isolated from 48-hour cell culture treated with indicated concentrations of LA. Graph is representative of five biologically independent experiments. (G) ECAR of CD8⁺ T cells (n = 5). The n values represent biologically independent experiments (means ± SEM). P values were determined using one-way ANOVA. (H) Cell cycle analysis of CD8⁺ T cells. Statistical analysis of five biologically independent experiments (means ± SEM) is shown. P values were determined using two-way ANOVA showing P values of G₀ phases. (I) Proliferation assay of T cells treated with LA as described in (F) or at a similar pH achieved using HCl (pH 7.0) for 72 hours. Statistical analysis (left) of percentage of proliferating cells of four biologically independent experiments (means ± SEM) and representative histograms (right) is shown. P value was determined using one-way ANOVA. ****P < 0.0001. (J) Total lysates of CD8⁺ T cells after CD3/28 stimulation and treatment with 15 mM LA for the indicated time points were analyzed by Western blotting. Graph depicts quantification of pS6K. Statistical analysis of three biologically independent experiments is shown. For quantification, all proteins were normalized to the corresponding loading control 14-3-3. P values were calculated using two-way ANOVA.

Because glucose was decreased in the presence of AML cells in the T cell culture (fig. S7A) and because depletion of glucose by leukemia cells could influence T cell metabolism, we added glucose to the cultures including murine T cells alone or in combination with AML cells. We observed that glucose did not prevent the reduced ECAR of the T cells upon exposure to AML cells or AML cell medium (fig. S7B). The same pattern was seen when we added glutamine, which was reduced in the presence of AML cells (fig. S7, C and D). Neither the AML-affected ECAR nor OCR of T cells was rescued by glucose or glutamine substitution (fig. S7, E and F). Few metabolites increased in the AML supernatant compared to the T cell–only supernatant, one of them being alanine (Fig. 4, A and B, and figs. S5, B and C, and S7G). Addition of different amounts of alanine to the T cell culture did not reduce the ECAR of T cells (fig. S7H), which makes it unlikely that alanine was responsible for the AML-induced decline of the ECAR. Because the AML cells (WEHI-3B) produce interleukin-3 (IL-3), we studied whether IL-3 addition would affect ECAR or OCR of T cells and found no evidence for this (fig. S8). These findings indicate that LA contributes to the AML-induced dysfunctional T cell metabolism and reduced proliferation.

NaBi can counteract the LA-induced metabolic and functional T cell inhibition

NaBi is clinically used to antagonize metabolic acidosis in patients; hence, we investigated whether exogenous NaBi could help to overcome the metabolic inhibition of CD8⁺ T cells caused by LA. We analyzed the glycolytic activity of T cells treated with LA in combination with NaBi. Addition of NaBi completely rescued the T cell metabolism from the effects of LA (Fig. 5, A and B), leading to high glycolytic activity of T cells despite the presence of LA. In agreement with a functional rescue, T cells exposed to NaBi exhibited restored proliferation (Fig. 5, C and D) and intact T cell cycle progression (Fig. 5E) despite exposure to LA.

Fig. 5. NaBi reverses metabolic and functional T cell defects caused by LA.

(A) ECAR measured at baseline and in response to glucose, oligomycin, and 2-DG for CD8⁺ T cells isolated from 48-hour cell culture under CD3/CD28 stimulation and treated with LA (15 m M) and/or NaBi (15 mM) as indicated. Graph is representative of four biologically independent experiments. (B) ECAR of CD8⁺ T cells (n = 4). The n values represent biologically independent experiments (means ± SEM). P values were determined using one-way ANOVA. (C) Representative histograms of proliferation assay of T cells for 72 hours. (D) Statistical analysis of the percentage of proliferating cells in four biologically independent experiments (means ± SEM). P values were determined using one-way ANOVA. (E) Cell cycle analysis of CD8⁺ T cells. Statistical analysis of four biologically independent experiments (means ± SEM) is shown. P values were determined using two-way ANOVA showing P value for G₀ phases. (F) Polar metabolites were extracted from CD8⁺ T cells, acquired by LC-MS, and analyzed using an untargeted approach with XCMS software. Graph shows total number of changed metabolites between T cells without treatment and T cells treated with LA. (G) Total number of changed metabolites between T cells treated with LA with or without NaBi. (H) Top five metabolic pathways and two others of interest decreased in T cells treated with LA compared to control. The number of decreased metabolites flagged in each pathway is indicated. ABC, ATP (adenosine triphosphate)–binding cassette. (I) Top five metabolic pathways and two others of interest increased in T cells treated with LA and NaBi compared to LA alone. The number of increased metabolites flagged in each pathway is indicated. (J) Targeted analysis of metabolites of interest using assayR for all four treatment groups. The values have been normalized to control without treatment; the color scale depicts the fold change. PEP, phosphoenolpyruvate; PG, phosphoglycerol; AMP, adenosine monophosphate; GMP, guanosine monophosphate; CDP, cytidine diphosphate; UDP, uridine diphosphate; ADP, adenosine diphosphate; GDP, guanosine diphosphate; CTP, cytidine triphosphate; UTP, uridine triphosphate; GTP, guanosine triphosphate.

Untargeted metabolomic analysis revealed a marked decline of many metabolites in T cells treated with LA (Fig. 5F), whereas the addition of NaBi reversed the LA-induced metabolomic changes (Fig. 5G and fig. S9). LA induced a reduction of multiple metabolic pathways, including the biosynthesis of amino acids, pyrimidine metabolism, amino sugar and nucleotide sugar metabolism, and glycolysis/gluconeogenesis (Fig. 5H). The addition of NaBi reversed the LA-induced metabolic pathway changes (Fig. 5I and fig. S9). Targeted analysis identified a severe decline in nucleotide abundance in LA-treated T cells concurrent with the reduced proliferation and cell cycle activity (Fig. 5, D, E, and J). In addition, intermediates of the glycolytic pathway such as glucose, fructose-1,6-bisphosphate, or triose phosphate were markedly reduced, explaining the impaired glycolytic activity (Fig. 5, B and J).

Observing these major changes in metabolic pathways of T cells treated with LA and NaBi (Fig. 5, H to J), we used ¹³C-labeled LA to investigate whether it is taken up by T cells and is used as an energy source in the tricarboxylic acid (TCA) cycle. Heavy-labeled LA was detected intracellularly upon treatment (Fig. 6). LA administration did not change the total abundance of intracellular LA, suggesting that the intra- and extracellular LA flux was unchanged (Fig. 6A). However, far less unlabeled LA was generated by glycolysis, emphasizing the severe metabolic impairment of the T cells (Fig. 6A). Addition of NaBi rescued the glycolytic activity, as evidenced by unlabeled LA being increased again. The treatment increased the total pools of LA because the cells also took up substantial amounts of extracellular ¹³C-LA (Fig. 6A). TCA cycle intermediates were decreased in LA-challenged T cells, with only a small portion of ¹³C-LA being further metabolized (Fig. 6, B to E). NaBi treatment rescued OXPHOS such that the unlabeled metabolites were increased, and additionally, a portion of extracellular ¹³C-LA was oxidized in the TCA cycle (Fig. 6, B to E). NaBi not only maintained glycolysis but also allowed for utilization of extracellular lactate in the TCA cycle. Inhibiting the uptake of extracellular LA using a monocarboxylate transporter 1 (MCT1) inhibitor led to a normalization of glycolytic activity comparable to T cells without treatment but also reduced their proliferative capacity (Fig. 6, F and G, and fig. S10). Increasing inhibitor concentrations did not further enhance the phenotype (fig. S11). NaBi treatment not only counteracted LA-mediated metabolic shutdown but also enabled the usage of extracellular LA as an additional fuel source for energy production.

Fig. 6. NaBi treatment of LA-challenged T cells leads to increased LA uptake and incorporation into the TCA cycle.

(A) Metabolites were extracted from CD8⁺ T cells isolated after 48 hours of CD3/CD28 stimulation. Cells were treated with ¹³C–heavy-labeled LA (15 mM) with or without NaBi (15 mM) as indicated (n = 4), and ¹³C-labeled metabolites were identified by mass spectrometry. Average background from ¹³C labeling, detected in the unlabeled samples, was subtracted from all samples. Graph shows relative abundance of unlabeled and labeled LA. (B) Relative abundance of citrate. (C) Relative abundance of succinate. (D) Relative abundance of fumarate. (E) Relative abundance of malate. The n values represent biologically independent experiments (means ± SEM). * in black represents the significance of differences in total abundance of each metabolite, calculated by one-way ANOVA or Kruskal-Wallis test. # in purple indicates the significance of differences in ¹³C metabolite isotopolog, calculated by two-way ANOVA. */#P < 0.05, **/##P < 0.01, ***/###P < 0.001, and ****/####P < 0.0001. (F) ECAR measured at baseline and in response to glucose, oligomycin, and 2-DG for CD8⁺ T cells isolated from 48-hour cell culture under CD3/CD28 stimulation treated with LA (15 mM), NaBi (15 mM), and/or MCT1 inhibitor AZD3965 (100 nM) as indicated. Graph is representative of three biologically independent experiments. (G) ECAR of CD8⁺ T cells (n = 3 to 5). The n values represent biologically independent experiments (means ± SEM). P values were determined using one-way ANOVA.

NaBi restores the low intracellular pH induced by the leukemia cells

Because we could rescue LA-challenged T cells by treating with NaBi, we analyzed the metabolic profile of CD8⁺ T cells when AML supernatant alone or in combination with NaBi was included in the culture. Addition of NaBi completely reversed the AML-induced inhibition of glycolytic activity in the T cells (Fig. 7, A and B). NaBi also blocked the reduction in T cell proliferation caused by the AML supernatant (Fig. 7, C and D). In agreement with this result, cell cycle activity increased when NaBi was included in the T cell culture containing AML-conditioned medium (Fig. 7E). Addition of AML supernatant or LA to a CD4⁺ T cell culture resulted in similar metabolic impairment, which could be reversed by NaBi (fig. S12, A and B). Accordingly, cell cycle activity and proliferative capacity of CD4⁺ T cells were diminished with AML supernatant or LA and restored through NaBi treatment (fig. S12, C to F). Inhibition of lactate dehydrogenase A (LDHA) in AML cells reduced the concentration of lactate in the AML supernatant (fig. S13, A and B). T cells incubated with medium originating from LDHA-treated AML cells showed better metabolic activity compared to controls (fig. S13, C and D). Mechanistically, LA decreased the intracellular pH in T cells, which could be reversed by NaBi (Fig. 7, F and G). In contrast to LA, HCl did not decrease the intracellular pH compared to the control (fig. S14, A and B). This supports the concept that a purely extracellular acidification does not affect the intracellular pH. In contrast to NaBi, other molecules that can antagonize acidosis (Hepes and NaOH) did not reverse LA effects and did not change the intracellular pH, indicating that the protective effect was specific to NaBi (fig. S14, C and D). Consistent with this result, Hepes and NaOH did not restore the glycolytic activity of T cells (Fig. 7H). In addition, the effect was not caused by sodium itself, because the addition of NaCl did not reverse LA-induced T cell dysfunction (Fig. 7H, and fig. S14, C and D). Analysis of microarray data revealed that CD8⁺ T cells show a strong expression of several transporters of the solute carrier family 4a (SLC4a) and SLC26a (fig. S15). These transporters enable bicarbonate transport and can regulate intracellular pH (7). Untargeted metabolomic analysis of T cells treated with LA for 30 min showed a strong increase in carbon metabolism compared to control independent of NaBi treatment (fig. S16). Targeted analysis revealed an increase in intracellular LA, which was strongly reduced in cells treated with NaBi (Fig. 7I). NaBi was able to inhibit the initial massive and potentially damaging influx of LA, enabling the T cells to adapt and use the incorporated LA as an energy metabolite. These results show that NaBi specifically rescues T cell metabolism and proliferation.

Fig. 7. Leukemia-derived LA induces intracellular pH changes and causes the impaired T cell phenotype.

(A) ECAR measured at baseline and in response to glucose, oligomycin, and 2-DG for CD8⁺ T cells isolated from 48-hour cell culture under CD3/CD28 stimulation incubating in fresh medium or medium originating from 48-hour AML cell culture. NaBi (15 mM) was added to T cell culture. Graph is representative of five biologically independent experiments. (B) ECAR of CD8⁺ T cells (n = 5). The n values represent biologically independent experiments (means ± SEM). P values were determined using one-way ANOVA. (C) Representative histogram of proliferation assay of T cells for 72 hours. (D) Statistical analysis of percentage of proliferating cells in four biologically independent experiments (means ± SEM). P values were determined using one-way ANOVA. (E) Cell cycle analysis of CD8⁺ T cells. Statistical analysis of five biologically independent experiments (means ± SEM) is shown. P values were determined using two-way ANOVA showing P value for G₀ phases. (F) Intracellular pH of CD8⁺ T cells after 30-min incubation with LA (15 mM), NaBi (15 mM), and/or medium originating from 48-hour AML cell culture, calculated using pH-controlled buffers (pH 5.5, pH 6.5, and pH 7.5) for calibration [shown in (G)]. Statistical analysis of four to six biologically independent experiments (means ± SEM), P values were determined using one-way ANOVA. (G) One representative of four biologically independent experiments described in (F) is shown. (H) ECAR of CD8⁺ T cells isolated from 48-hour cell culture under CD3/CD28 stimulation incubating in fresh medium or medium originated from 48-hour AML cell culture with administration of NaCl (15 mM), Hepes (15 mM), or increased pH achieved using NaOH (pH adjustment to pH 8) (n = 4). The n values represent biologically independent experiments (means ± SEM). P values were determined using one-way ANOVA. (I) Polar metabolites were extracted from CD8⁺ T cells after 30-min CD3/CD28 stimulation and treatment with LA (15 mM) and/or NaBi (15 mM), acquired by LC-MS, and analyzed using a targeted approach with assayR. Graph shows relative abundance of lactate for all four treatment groups (n = 4). The n values represent biologically independent experiments (means ± SEM). P values were determined using one-way ANOVA.

NaBi improves the GVL effect and counteracts the AML-mediated metabolic and functional inhibition in human T cells

To assess the efficacy of NaBi in clinically relevant settings, we administered NaBi to AML (WEHI-3B)–bearing mice that underwent allo-HCT. NaBi treatment did not improve the survival of mice in the absence of donor T cells (Fig. 8A), indicating that NaBi did not exert any direct toxic effects on AML cells in vivo. When T cells were given to AML-bearing mice, the animals survived longer when treated with NaBi (Fig. 8A). The results could be reproduced in the genetic FLT3-ITD/MLL-PTD–based GVL model (Fig. 8B). Because robust T cell activation in the allogeneic setting could exacerbate graft-versus-host disease (GVHD), we administered NaBi to mice with developing GVHD and observed no aggravation of GVHD pathology (fig. S17). On the basis of the potential for clinical translation, we next analyzed whether the observed effects of LA and NaBi could be reproduced in human T cells. We challenged human CD8⁺ T cells derived from healthy donors with LA or AML supernatant in the presence or absence of NaBi (Fig. 8, C and D). We observed that LA also affected glycolysis in human T cells and that this could again be restored by NaBi administration (Fig. 8, C and D). The results indicate that NaBi can antagonize the LA-induced effects on human T cells, comparable to our findings in mouse T cells. Encouraged by the preclinical results, we prescribed bicaNorm (NaBi) for 1 week to a cohort of 10 patients with AML who had a relapse after allo-HCT and received DLIs (table S6). We compared the metabolic and immune phenotypes of CD8⁺ T cells isolated from the peripheral blood before and after NaBi treatment. Blood gas analysis revealed an increase in pH and bicarbonate concentrations in the blood of these patients (Fig. 8, E and F). The T cells received a metabolic boost with increased respiration and spare respiratory capacity (Fig. 8, G and H). The rescue of metabolic activity was accompanied by increased IFN-γ and TNFα production (Fig. 8, I and J), whereas other parameters such as glycolytic activity and activation markers did not change (fig. S18). Together, these findings demonstrate the mechanism of metabolic evasion of the immune cells by LA production in aggressive blood cancers and pave the way for a clinical trial to combine T cell transfer with bicarbonate for AML relapse treatment. Our findings are summarized in fig. S19.

Fig. 8. NaBi promotes the GVL effect and rescues human T cell function.

(A) Recipient mice (BALB/c background) were transplanted with 5 × 10⁶ C57BL/6 BM cells (allo BM) with additional 1 × 10⁴ syngeneic leukemia cells (WEHI-3B luc). C57BL/6 T cells (allo Tc) were transplanted 2 days later (3 × 10⁵). NaBi (200 mM) was administered for 14 days (days 2 to 16 after BMT) in the drinking water. Survival was monitored. Data were pooled from two independent experiments with n = 10 per group, and P values were determined using the log-rank (Mantel-Cox) test. (B) Recipient mice (C57BL/6 background) were transplanted with 5 × 10⁶ BALB/c BM cells (allo BM) with additional 5 × 10³ splenocytes derived from a syngeneic mouse carrying MLL^PTD/+;Flt3^ITD/+ mutated tumor. A total of 3 × 10⁵ BALB/c CD8⁺ T cells (allo Tc) were transplanted 2 days later. NaBi (200 mM) was administered as described in (A). Survival was monitored. Data were pooled from two independent experiments with n = 9 to 10 mice per group, and P values were determined using the log-rank (Mantel-Cox) test. (C) ECAR measured at baseline and in response to glucose, oligomycin, and 2-DG for human primary CD8⁺ T cells from healthy donors after 48 hours of cell culture under CD3/CD28 stimulation incubating in fresh medium or medium originating from 48-hour AML cell culture. LA (15 mM) and/or NaBi (15 mM) were added to T cell culture. Graph is representative of three biologically independent experiments. (D) ECAR after oligomycin injection of CD8⁺ T cells (n = 3). The n values represent individual healthy donors (means ± SEM). P values were determined using one-way ANOVA. (E) Metabolic analysis of patients with relapsed AML after allo-HCT receiving DLIs before and after 7-day oral treatment with bicaNorm (sodium bicarbonate). Graph shows serum pH. (F) HCO₃⁻ concentration in the peripheral blood of the patients in (E). (G) OCR measured at baseline and in response to oligomycin, FCCP, and rotenone + antimycin A in isolated CD8⁺ T cells. (H) OCR after FCCP injection. (I) Representative FACS plots (left) and statistical analysis (right) of the percentage of CD8⁺ T cells expressing IFN-γ versus side scatter (SSC) (n = 10). (J) Representative FACS plots (left) and statistical analysis (right) of the percentage of CD8⁺ T cells expressing TNFα (n = 10). Each data point represents the measurement of an individual patient at the indicated time point; n values represent individual patients. P values were determined using the two-sided Student’s paired t test or Wilcoxon test.

DISCUSSION

Relapse of leukemia is the major cause of death after allo-HCT (1), indicating that strategies extending the mode of action of existing approaches are needed. Previous studies have used tyrosine kinase inhibitors (8), hypomethylating agents (9), cytokines (10), or immune checkpoint inhibitors (11) to enhance GVL effects [reviewed in (12)]. We investigated whether allogeneic T cell–mediated cytotoxicity against AML is inhibited by metabolic alterations induced by the leukemia cells. This hypothesis was based on our observation that both glycolysis and OXPHOS decreased in T cells when patients experienced an AML relapse but not at primary AML diagnosis, supporting the concept that the metabolic alterations play a role in the context of allogeneic immune pressure. Concurrent with the metabolic impairment, we observed strongly reduced effector molecule production by T cells isolated from patients with relapsed AML. Metabolic activity has been directly linked to T cell effector function because impaired glycolysis enables the enzyme glyceraldehyde-3-phosphate dehydrogenase to bind the mRNA of IFN-γ, thereby interfering with T cell cytokine production (13).

These findings in patients were the basis for controlled in vitro studies in which we identified LA concentrations to be increased in the supernatant of leukemia cells. This was consistent with our finding that LA increased when patients had an AML relapse after allo-HCT. Accordingly, the plasma of patients at primary diagnosis contained a lower abundance of LA, whereas the T cells exhibited a robust metabolic and cytokine profile.

Functional studies showed that AML culture–derived supernatant reduced glycolysis and OXPHOS in T cells. Consistent with our work, other studies had shown that LA can affect T cell function in different cancer entities including AML (14, 15). A potential therapeutic advantage of the AML/post–allo-HCT setting is that the T cells contained in the DLI are nonmanipulated and derived from a healthy donor, whereas in solid tumors, a therapeutic intervention toward metabolic T cell manipulation is hampered, because the T cells are already paralyzed by the tumor microenvironment–derived metabolites.

Mechanistically, we found that the metabolic changes in T cells were connected to reduced T cell proliferation in vitro and in vivo. Expansion of T cells is essential for the GVL effect to keep pace with the rapidly expanding leukemia cells. Our previous work had shown that increased T cell longevity was associated with improved GVL effects (8), and others had shown that T cell exhaustion is connected to relapse (16). We observed that exogenous LA impaired leukemia control in vivo, while antagonizing LA with NaBi increased GVL effects. Because recent data emphasized the importance of CD4⁺ T cells in immune surveillance for AML relapse (17), we investigated functional and metabolic characteristics of CD4⁺ T cells challenged with LA and found a similar pattern as for CD8⁺ T cells.

Our data indicate that there are several reasons why NaBi is able to antagonize the LA effects and is therefore special compared to other buffers. First, NaBi treatment could reverse the severe metabolic damage and functional defects induced by LA. Second, NaBi reduced the massive initial influx of exogenous LA. Third, NaBi enabled a controlled uptake of LA into T cells and its metabolization in the TCA cycle. Thereby, LA served as an additional energy source that supported the proliferative capacity of T cells. Fourth, NaBi itself can serve as a substrate for multiple carboxylase reactions such as pyrimidine metabolism (18, 19). Consistent with these reports, the untargeted pathway analysis indicated that pyrimidine metabolism increased in T cells rescued with NaBi.

We administered NaBi to 10 patients with AML with a relapse after allo-HCT who received DLIs and observed an enhanced metabolic and immune profile of isolated T cells. The therapeutic intervention with NaBi has little toxicity and is often used in patients with tumor lysis syndrome (20). Bicarbonate has been explored as an experimental treatment for metastases in solid tumors, showing that the acidic pH in the microenvironment could be reversed (21, 22). The acid-base balance would likely prevent massive long-term changes in the pH of the whole individual. We propose extensive clinical studies with larger patient cohorts and monitoring long-term outcome measures such as OS to assess the benefit of NaBi administration for patients with AML relapsed after allo-HCT.

A limitation of our study is that most functional findings were made in the mouse model, and only correlative studies were performed with human cells. Determining whether NaBi improves the GVL effect in humans will require a prospective clinical study. Our findings provide a scientific rationale for a combination of DLI and NaBi in patients relapsing with AML after allo-HCT, in particular when increased LA concentrations are detected as observed in the patients with AML relapse whom we had studied.

Overall, our findings show that AML cells reduce glycolytic activity of T cells in patients and mice. Mechanistically, we demonstrated that this was mediated by LA release reducing intracellular pH in T cells, which changed the transcriptional profile leading to metabolic and proliferative dysfunction. NaBi treatment not only counteracted LA-mediated metabolic shutdown in T cells but also enabled the usage of extracellular LA as an additional fuel source for energy production, what translated into improved GVL effects. Our findings provide a pharmacological strategy to overcome metabolic reprogramming of transferred T cells to enhance GVL effects in patients relapsing after allo-HCT.

MATERIALS AND METHODS

Study design

For the sample size in the murine survival experiments, a power analysis was performed. A sample size of at least n = 8 per group was determined to reach a statistical significance of 0.05 to detect an effect size of at least 1.06 with 80% power. Mice were assigned randomly to the experimental groups. Experiments were performed in a nonblinded fashion, except for the blinded GVHD severity scoring. To obtain unbiased data, a pathologist blinded to the treatment groups performed the histopathological scoring of GVHD severity. Only after finalization of the quantitative GVHD severity scores, the samples were allocated to their treatment group. All samples and mice were included in our analysis.

Human participants

For the prospective clinical study, peripheral blood of patients with AML at different stages of disease (primary diagnosis, remission, and relapse) was analyzed. The trial was registered at clinicaltrials.gov ID: NCT04259372. For the interventional early phase 1 clinical study, patients with AML, relapsed after allo-HCT and receiving DLIs, received bicaNorm (NaBi) for 1 week, and peripheral blood was analyzed. The trial was registered at clinicaltrials.gov ID: NCT04321161. Human sample collection and analysis were approved by the Institutional Ethics Review Board of the Medical Center, University of Freiburg, Germany (protocol numbers: 10024/13, 26/11, and 509/16). Written informed consent was obtained from each patient. All analyses of human data were carried out in compliance with relevant ethical regulations. The patients’ characteristics, including recipient age and gender, AML characteristics, donor type, conditioning regimen, immunosuppressive regimen, and remission status before transplantation are detailed in tables S1 to S6.

Mice

C57BL/6 (H-2Kb) and BALB/c (H-2Kd) mice were purchased from Janvier Labs or from the local stock at the animal facility of Medical Center–University of Freiburg. Rag2^−/−Il2rg^−/− mice were obtained from the local stock at the animal facility of Medical Center–University of Freiburg. Mice were used between 6 and 14 weeks of age, and only female donors were used for female recipients and vice versa. Animal protocols were approved by the animal ethics committee Regierungspräsidium Freiburg, Freiburg, Germany (protocol numbers: X15/10A, G17-093, G17-063, and G20-075).

Bone marrow transplantation model

Bone marrow (BM) transplantation (BMT) experiments were performed as described previously (23). The major mismatch strain combinations used were C57BL/6 into BALB/c or vice versa, as indicated in the respective experiments. Briefly, recipients were injected intravenously (iv) with 5 × 10⁶ BM cells after lethal irradiation with 1000 centigrays (cGy) (BALB/c recipient) or 1200 cGy (C57BL/6 recipient), using a ¹³⁷Cs source, split into two equal doses and performed 4 hours apart. CD4⁺ and CD8⁺ T cells were isolated from donor spleens and enriched by negative selection with the MACS cell separation system (Miltenyi) according to the manufacturer’s instructions. Pan T cell isolation kit II (Miltenyi) was used. CD4⁺/CD8⁺ T cell purity was at least 90% as assessed by flow cytometry. CD4⁺/CD8⁺ T cells were given at a dosage of 2 × 10⁵ to 3 × 10⁵ cells iv on day 2 after BMT as indicated in the respective experiments. For GVHD induction, 4 × 10⁵ CD4⁺/CD8⁺ T cells were injected on the same day as the BM.

GVL models

AML^{MLL-PTD FLT3-ITD} leukemia model

For the AML^{MLL-PTD FLT3-ITD} leukemia model (24), C57BL/6 recipients were transplanted with 5 × 10³ AML^{MLL-PTD FLT3-ITD} cells and 5 × 10⁶ BALB/c BM cells iv after lethal irradiation with 12 Gy split into two equal doses. A total of 2 × 10⁵ to 3 × 10⁵ BALB/c (allogeneic model) splenic CD4⁺ and CD8⁺ T cells were introduced intravenously on day 2 after initial transplantation.

WEHI-3B leukemia model

For the WEHI-3B leukemia model, BALB/c recipients were transplanted with 1 × 10⁴ AML (WEHI-3B) cells for survival studies or 1 × 10⁵ AML (WEHI-3B) cells for ex vivo T cell metabolic analysis, respectively, and 5 × 10⁶ C57/BL6 BM cells iv after lethal irradiation with 10 Gy split into two equal doses. A total of 2 × 10⁵ to 3 × 10⁵ C57/BL6 (allogeneic model) splenic CD4⁺ and CD8⁺ T cells were introduced intravenously on day 2 after initial transplantation.

MOLM-13 AML^FLT3-ITD xenograft model

For the MOLM-13 AML^FLT3-ITD xenograft model (25), Rag2^−/−Il2rγ^−/− recipients were transplanted with 1 × 10⁴ MOLM-13 cells iv after sublethal irradiation with 5 Gy. A total of 2 × 10⁵ C57/BL6 (allogeneic model) splenic CD4⁺ and CD8⁺ T cells were introduced intravenously on day 2 after initial transplantation.

In vivo bioluminescence imaging

In vivo bioluminescence imaging was performed as described previously (26).

Drug treatment

Mice in GVL experiments received 200 mM NaBi (Sigma-Aldrich) (27) in the drinking water for 2 weeks starting from the day of T cell transplantation on day 2 after BMT. Control mice received regular drinking water as provided by the animal facility.

Statistical analysis

All data were tested for normality applying the D’Agostino and Pearson or Shapiro-Wilk test. For statistical analysis of two groups, an unpaired or paired two-tailed Student’s t test was applied. If the data did not meet the criteria of normality, then the Mann-Whitney U or Wilcoxon signed-rank test was applied. If more than two groups were analyzed, then we used the Kruskal-Wallis test if nonparametric testing was suggested, and we performed one-way analysis of variance (ANOVA) in case of normally distributed data. For grouped analyses, two-way ANOVA was performed. Log-rank test (Mantel-Cox) was used to calculate significance of differences between survival curves. Corrections for multiple comparisons were performed by Tukey’s or Sidak’s test for parametric and Dunn’s test for nonparametric tests.

Statistical analysis was performed using GraphPad Prism (Graph-Pad Software). Data are presented as means ± SEM. In all cases, P < 0.05 was considered significant. The numbers of experimental repeats and numbers of mice used are indicated in the figure legends. All data values and appropriate statistical tests of each graph are listed in the data files S2 and S3.

Data and materials availability:

The complete gene expression data are available in the GEO repository under the access ID GSE136132 (www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE136132) via the token “chcfugekvvwzlmp.” MATLAB scripts are provided as data files S4 and S5. All data associated with this study are present in the paper or the Supplementary Materials.

Supplementary Material

Supp Materials

stm.sciencemag.org/cgi/content/full/12/567/eabb8969/DC1

Materials and Methods

Fig. S1. Flow cytometric gating strategy for human CD8⁺ T cell analysis.

Fig. S2. Correlation of glycolytic and respiratory changes in CD8⁺ T cells of patients with AML relapse after allo-HCT.

Fig. S3. Transcriptomic analysis of CD8⁺ T cells incubated with AML supernatant.

Fig. S4. Flow cytometric gating strategy for cell cycle analysis.

Fig. S5. Mass spectrometric analysis of metabolites of cell culture supernatants.

Fig. S6. Altered signaling profile in LA-challenged T cells.

Fig. S7. Influence of soluble factors on T cell metabolism.

Fig. S8. Influence of IL-3 addition on T cell metabolism.

Fig. S9. NaBi treatment of LA-challenged T cells normalizes metabolite pools.

Fig. S10. Inhibition of MCT1 leads to reduced proliferative activity.

Fig. S11. Higher concentrations of MCT1 inhibitor do not further decrease glycolytic and proliferative activity.

Fig. S12. LA impairs CD4⁺ T cell metabolism, proliferation, and cell cycle, which can be rescued by NaBi administration.

Fig. S13. LDHA inhibition in AML cells reduces lactate secretion and improves metabolic activity of T cells challenged with AML medium.

Fig. S14. Intracellular pH under extracellular acid pH, sodium, or buffer treatment.

Fig. S15. Expression of bicarbonate transporters on CD8⁺ T cells.

Fig. S16. Short incubation with LA leads to initial increase in carbon metabolism independent of NaBi administration.

Fig. S17. NaBi treatment does not increase GVHD severity.

Fig. S18. Metabolic and immune phenotypic analysis of patients with relapsed AML under bicaNorm treatment.

Fig. S19. Proposed mechanism for metabolic boost of NaBi treatment on T cells challenged with leukemia-derived LA.

Table S1. Clinical, molecular, and cytogenetic characteristics of patients with AML at primary diagnosis.

Table S2. Clinical, molecular, and cytogenetic characteristics of patients with AML at relapse.

Table S3. Chemotherapy conditioning regimens and transplant characteristics of patients with relapse.

Table S4. Clinical, molecular, and cytogenetic characteristics of patients with AML in remission.

Table S5. Chemotherapy conditioning regimens and transplant characteristics of patients in remission.

Table S6. Clinical, molecular, and cytogenetic characteristics of patients with AML treated with NaBi.

Table S7. Antibodies for flow cytometry.

Table S8. Additional cell staining dyes.

Table S9. Antibodies for Western blots.

Data File 1

Data file S1. Karyotype and molecular abnormalities of each patient.

Data File 2

Data file S2. Source data for main figures.

Data File 3

Data file S3. Source data for supplementary figures.

Data File 4

Data file S4. MATLAB for NMR analysis.

Data File 5

Data file S5. MATLAB for NMR analysis options.