Allogeneic hematopoietic cell transplantation (allo-HCT) is historically one of the earliest routinely used forms of cancer immunotherapy (1, 2). Its application has been increasing worldwide since the 1980s (3). To date, allo-HCT is the only curative treatment option for patients with highly aggressive hematological malignancies, such as adverse-risk acute myeloid leukemia (AML). However, more than 30% of the patients with AML would eventually relapse after allo-HCT (4). Patients with relapse can be treated either with chemotherapy or with immune cell-directed therapies aiming at reinvigorating the GVL response (5). One such immunotherapy is the administration of donor lymphocyte infusions (DLI), but the response rates in AML are modest at approximately 20% (6, 7). Other strategies to boost the anti-leukemic immune response include immune checkpoint inhibitors and flotetuzumab, a CD3 x CD123 dual affinity retargeting antibody, which facilitates the recruitment of T cells to CD123⁺ AML cells. Adoptive cell transfer therapies with chimeric antigen receptor (CAR) T and natural killer (NK) cells are currently being developed (5). Nevertheless, re-establishing efficient anti-tumor immunity in a relapsed patient remains a challenge, resulting in dismal clinical outcomes. Why does the transfer of a foreign donor graft achieve a lifelong remission through immunosurveillance in some patients, while in others, the anti-leukemia immune response is inefficient or fails after time? How can donor immune cell activity be reinvigorated to boost the anti-leukemia activity? Is there a way to make tumor cells more susceptible to immune-mediated killing?

Cellular metabolism is now an established hallmark of cancer (8) and an essential determinant of T cell survival, differentiation, and function (9, 10). In this review, we focus on alterations of the metabolic environment as a mechanism of immune escape after allo-HCT. We discuss current knowledge about how immune cell and leukemia cell metabolism regulate the anti-tumor immune response, review recent insights about the role of systemic metabolism, and point out future research directions.

Leukemic cells, characterized by uncontrolled proliferation and impaired differentiation capacity, show extensive metabolic alterations compared to healthy hematopoietic cells. In fact, the deregulation of cellular metabolism is a hallmark of cancer (48). To expand rapidly, leukemic cells have to generate biomass. Thereby, they display a highly diverse and flexible metabolism (49) ( Figure 1A ).

Glucose metabolism plays a central role in generating the biomass required for the rapid expansion of leukemic cells. In an AML mouse model with BCR/ABL and NUP98/HOXA9 fusion proteins, the leukemic cells dramatically increased glucose uptake up to 20-fold compared to normal hematopoietic cells (64). The cellular uptake of glucose is mediated by glucose transporters (GLUT), which are upregulated in many types of cancer (65). Analysis of 108 AML patient samples via qPCR showed that GLUT1 was increased in bone marrow mononuclear cells of AML patients without remission compared to healthy individuals and patients with complete or partial remission (50). Moreover, a study using leukemia cell lines as well as patient samples demonstrated that AML cells additionally upregulated GLUT5, enabling fructose uptake to compensate for glucose deficiency under low extracellular glucose levels (52). The ingested glucose was preferentially fermented to lactate after passing through glycolysis instead of entering the tricarboxylic acid (TCA) cycle, regardless of the presence of oxygen. This shift towards aerobic glycolysis, named Warburg effect (51), enables the cells to produce ATP as well as accumulate precursor molecules and reducing equivalents needed for a rapid cell division (66). In AML patients’ serum, increased amounts of pyruvate and lactate were detected compared to the healthy controls (67). The deletion of pyruvate kinase M2 (PKM2), which catalyzes the rate-limiting conversion from phosphoenolpyruvate to pyruvate in glycolysis, in mouse BM cells carrying BCR-ABL or MLL-AF9 rearrangements reduced leukemia cell growth in vivo without perturbing HSCs (68). Overall, aerobic glycolysis is one of the essential pathways for leukemic cells.

Besides glucose, leukemia cells also increase the uptake and consumption of amino acids, most importantly glutamine (69). In the process of glutaminolysis, glutamine is converted into α-ketoglutarate, which can be used to replenish the TCA cycle (70). While the glutamine plasma concentration in healthy persons amounts to 0.6 nM to 0.8 nM, the glutamine concentrations measured in the serum of AML patients were below 0.3 nM (71, 72). Glutamine is mainly imported via the high-affinity glutamine transporter SLC1A5. A shRNA-mediated knockdown of SLC1A5 induced apoptosis in the AML cell lines MOLM-14 and HL-60 and reduced the growth of MOLM-14 tumors in a xenograft mouse model underlining the importance of glutamine for leukemia cell metabolism (54). Importantly, a genome-wide CRISPR-Cas9 knockout screen identified the enzyme glutaminase (GLS) as a major metabolic vulnerability in FLT3-ITD-mutated AML cells treated with FLT3 tyrosine kinase inhibitors (73). Genetic or pharmacological silencing of GLS function synergized with the FLT3 inhibitor AC220 to induce apoptosis in the FLT3-ITD-mutated human cell lines MV-4-11 and MOLM13 (73). Moreover, glutamine is an important metabolite for the synthesis of the reactive oxygen species (ROS) scavenger glutathione. Blocking glutamine metabolism with the glutaminase inhibitor CB-839 significantly impaired glutathione production in multiple AML cell lines, leading to the accumulation of mitochondrial ROS and apoptosis (73, 74).

Apart from glutamine, arginine has been proven to be important for AML cells (55, 75). Arginine is a semi-essential amino acid and is crucial to providing amine groups and aspartate for the production of nucleotides (76). The majority of leukemic cells do not express argininosuccinate synthetase-1 (ASS1), which is needed for the intracellular synthesis of arginine and thus depend on arginine uptake (55). In line with these findings, the arginine transporters CAT-1 and CAT-2B were found to be constitutively expressed on AML blasts, and the plasma arginine levels of AML patients were significantly reduced compared to healthy individuals (75). Moreover, the treatment with a human recombinant arginase (BCT-100), which catabolizes the cleavage of arginine and thus reduces its availability for uptake, inhibited the proliferation of primary AML blasts in vitro (75). Nevertheless, the clinical implementation of treatments targeting arginine metabolism has so far been challenging. A randomized clinical trial testing low-dose cytarabine in combination with the human recombinant arginase BCT-100 versus low-dose cytarabine alone in 86 AML patients over 60 years of age found no difference in the overall and median survival between the two groups (77). Another approach is to inhibit arginase together with inducible nitric oxide synthase (iNOS). In in vitro experiments, the combination of the small molecule inhibitor of arginase NOHA and the iNOS inhibitor L-NMMA enhanced the proliferation and cytotoxicity of T cells and CAR T cells (78, 79). Moreover, treatment with NOHA and L-NMMA restored the proliferation of HSCs and reduced the expression of CD206 on monocytes (78).

Furthermore, AML cells depend on the uptake of extracellular cysteine due to their inability to generate cysteine from methionine. Cysteine depletion in the cell culture medium decreased the viability and proliferation of primary AML cells and most AML cell lines. Since cysteine is an important precursor for the synthesis of glutathione, depletion resulted in increased ROS levels, which led to cell death via ferroptosis. While therapeutic targeting of cysteine uptake alone using the cysteine/glutamate antiporter inhibitor sulfasalazine showed limited efficacy in AML cell lines, combination with the γ-glutamylcysteine synthase (GCL) inhibitor L-Buthionine sulfoximine, which inhibits the production of glutathione, reduced proliferation in multiple AML cell lines and primary AML samples (56). Taken together, amino acids, specifically glutamine, arginine, and cysteine, play an essential role in the metabolism of leukemic cells.

In addition, leukemia cells undergo changes in lipid metabolism (80). Profiling of the lipid composition in five AML cell lines revealed significant differences in the expression of lipid species between the cell lines at steady state. However, under global serum deprivation, the cell lines showed variations in the same lipid species, suggesting a shared adaptive mechanism to nutrient stress. More precisely, phosphatidyl-serine, phosphatidyl-ethanolamine, and phosphatidyl-choline were decreased. Lipid profiling of plasma samples from AML patients revealed an incresed concentration of phosphatidyl-inositol and free fatty acids in the plasma of patients with an unfavorable prognosis (n=5) compared to patients with a favorable prognosis (n=5) (81). Another study analyzing the serum lipid levels in 214 AML patient samples found, among others, apolipoprotein B and high-density lipoprotein cholesterol to be prognostic factors for the clinical outcome (82). Furthermore, AML blasts can induce the phosphorylation of hormone-sensitive lipase in bone marrow adipocytes and activate lipolysis, leading to the release of fatty acids from adipocytes (83). The absorbed fatty acids can be oxidized to acetyl-CoA, thus serving as an alternative fuel for the TCA cycle and OXPHOS (53). Analysis of public microarray datasets revealed an increased expression of the carnitine palmitoyltransferase 1A (CPT1A), which catalyzes the first step of fatty acid oxidation, in the bone marrow, peripheral blood, and CD34⁺ cells of AML patients compared to healthy individuals. Moreover, a high expression of CPT1A was associated with a shorter survival (84). Likewise, the carnitine transporter CT2 was found to be highly expressed in AML cell lines and primary AML cells. An shRNA-mediated knockdown of CT2 in the AML cell lines OCI-AML2 and HL-60 decreased the viability and growth (85). In summary, leukemia cells exhibit modulated lipid metabolism that favors their rapid growth.

AML is a heterogeneous disease group characterized by a variety of chromosomal and cytogenetic aberrations, and it is conceivable that different AML subtypes might be characterized by distinct metabolic programs. Many genetic alterations lead to constitutive PI3K-AKT-TOR pathway activation, enhancing glucose uptake, glycolytic flux, amino acid uptake, acetyl-CoA production, and synthesis of lipids and nucleotides (86). Additionally, several studies investigated the connection between specific mutations and distinct metabolic abnormalities. For example, internal tandem duplication (ITD) mutation in FLT3 has been shown to upregulate mitochondrial hexokinase and increase the dependency on aerobic glycolysis in murine and human AML cell lines (87). Additionally, FLT3-ITD⁺ primary human AML cells had a lower abundance of ceramides compared to samples from FLT3-ITD^- patients. Mechanistically, FLT3 inhibition with sorafenib, crenolanib, or AC220 restored the ceramide abundance in MV-4-11 and MOLM-14 cell lines, and the accumulated mitochondrial ceramides facilitated mitophagy leading to cell death (88). FLT3-mutant AML was also shown to be particularly susceptible to ferroptosis, an iron-dependent form of cell death. Inhibition of FLT3 or the transcription factor C/EBPα downregulated the expression of stearyl-CoA-desaturase leading to impaired incorporation of fatty acids into lipids and elevated susceptibility to lipid redox stress (89). Moreover, metabolomics of AML patient samples revealed increased serum levels of choline, trimethylamine N-oxide, and leucine in patients carrying an NPM1 mutation together with mutations in cohesion complex and DNA damage response (90). Furthermore, somatic mutations in isocitrate dehydrogenase 1 and 2 (IDH1 and IDH2) lead to a neomorphic enzyme activity. Instead of catalyzing the oxidative decarboxylation of isocitrate to α-ketoglutarate, mutant IDH catalyzes the reduction of α-ketoglutarate into R-2-Hydroxyglutarate (R-2-HG) (91, 92). Overall, the knowledge about the relationship between genomic mutations and AML metabolism is limited to date, and further research is needed to explore the metabolic dependencies associated with distinct genetic profiles.

Notably, the metabolism of leukemic stem cells (LSCs) is distinct from that of rapidly proliferating leukemia blasts (93, 94). LSCs are a subpopulation of AML cells that reside in the bone marrow niche and have a quiescent cell cycle status and self-renewal properties (95). Moreover, LSCs are often resistant to chemotherapy and are thought to drive relapse (96, 97). In contrast to leukemic blasts, LSCs predominantly use OXPHOS and not glycolysis for ATP production (57). Since almost no pyruvate is available due to low glycolytic activity, LSCs require amino acids and/or fatty acids to fuel the TCA cycle and OXPHOS (98, 99). Similar to bulk leukemic cells, the amino acid cysteine has been found to be particularly important for the survival of LSC. In human LSCs, the depletion of cysteine in vitro led to impaired glutathione synthesis and a reduced glutathionylation of succinate dehydrogenase A, the key component of electron transport chain complex II. The resulting inhibition of OXPHOS led to LSC death (60). The specific metabolism of LSCs represents an opportunity for therapeutic targeting (58, 59). However, while de novo LSCs are described as metabolically inflexible, relapsed/refractory LSCs display a higher metabolic plasticity (99). For instance, the withdrawal of amino acids led to the cell death of LSCs isolated from de novo AML patients, whereas LSCs from relapsed AML patients were not dependent on amino acids because of their ability to balance energy requirements through fatty acid metabolism (98).

While chemotherapy eliminates most bulk leukemic cells, it is rarely effective in eliminating LSCs (58). A growing body of recent work has focused on identifying unique metabolic features in therapy-resistant LSCs (100) ( Figure 1B ). The combination treatment of the BCL2 inhibitor venetoclax and the hypomethylating agent azacytidine has shown significantly better clinical outcomes than conventional treatment in newly diagnosed elderly AML patients (101, 102). Further analysis demonstrated that in patients treated with venetoclax and azacytidine, the LSC population was decreased due to disruption of LSC energy metabolism (98, 101). More precisely, the combination of venetoclax and azacytidine reduced the uptake of amino acids, resulting in a decreased OXPHOS (98). Since LSCs depend on OXPHOS (57, 103), disruption of this process causes selective cell death of LSCs (98, 101). Relapse/refractory LSCs, however, exhibit a higher metabolic plasticity (99). As mentioned above, LSCs isolated from relapsed AML patients were able to compensate for the amino acid deprivation through upregulation of fatty acid metabolism (98). These LSCs could be re-sensitized by combining venetoclax/azacytidine with inhibitors of fatty acid transport, such as etomoxir and sorbitan sesquioleate (98, 104). The combination of etomoxir with venetoclax/azacytidine reduced OXPHOS in human LSCs ex vivo, leading to a lower viability as well as a decreased tumor burden in a xenograft mouse model with primary AML samples (104). Treatment of LSCs from relapsed AML patients with sorbitan sesquioleate and venetoclax/azacytidine also significantly decreased OXPHOS and reduced viability (98). Moreover, metabolomics profiling of six de novo AML patients and six patients with relapsed AML after induction therapy revealed enhanced nicotinamide levels and increased nicotinamide metabolism in relapsed LSCs. The enhanced nicotinamide metabolism resulted in increased total energy metabolism, thus bypassing the cytotoxic effect of venetoclax/azacytidine therapy in relapsed/refractory LSCs (105). Collectively, these examples demonstrate the importance of LSC metabolism in therapy resistance.

The metabolism of T cells is intricately linked to their cell function, and distinct metabolic programs are required at each stage of activation and differentiation. Naive CD8⁺ T cells, which remain in a quiescent state, fulfill their relatively low metabolic requirements primarily by pyruvate and fatty acid oxidation via the TCA cycle and subsequent OXPHOS (106). Upon antigen recognition, CD8⁺ T cells expand clonally, for which both a sufficient energy supply and a large number of biomolecules are needed. Consequently, the transition from resting naive T cells to activated T cells requires substantial metabolic reprogramming (107), which we will review briefly here.

A characteristic of activated CD8⁺ T cells is a switch towards aerobic glycolysis as the main energy source, commonly referred to as the Warburg effect (61). In this process, the pyruvate produced from glucose during glycolysis is fermented to lactate despite the availability of oxygen. Although aerobic glycolysis yields only two ATP molecules per glucose molecule, in contrast to up to 36 ATP molecules generated by OXPHOS, it offers the advantage of providing metabolic intermediates crucial for cell growth and proliferation (107). Moreover, it induces the pentose phosphate pathway, which supplies the cells with ribose 5-phosphate, a building block for nucleic acid synthesis, and NADPH (108). NADPH is required for catabolic pathways, such as the synthesis of fatty acids and amino acids, and the maintenance of the antioxidant glutathione (109). The neutralization of ROS is particularly important because activated T cells experience high levels of oxidative stress due to the increased non-mitochondrial oxygen consumption rate (110). Furthermore, CD8⁺ T cells increase glutamine uptake. In the process of glutaminolysis, glutamine is converted into α-ketoglutarate, which can subsequently enter the TCA cycle. Additionally, glutaminolysis provides NADPH to support lipid and nucleotide biosynthesis and the maintenance of glutathione (63). Apart from glucose and glutamine, lipids are an effective energy source as well as biosynthetic intermediates. Activated CD8⁺ T cells upregulate fatty acid synthase while decreasing fatty acid oxidation, leading to the accumulation of fatty acid metabolites needed for membrane synthesis (62). Furthermore, complex lipids support effector T cell signaling and function. Activated CD8⁺ T cells accumulate saturated phosphatidylinositides (PIPn) as opposed to naive CD8⁺ T cells, which are abundant in polyunsaturated PIPn. Saturated PIPn act as superior substrates for the enzymes PIP5K and PLC for the generation of second messengers and thus support downstream signaling essential for survival and cytokine production (111). In summary, CD8⁺ T cells exhibit metabolic changes upon activation, favoring aerobic glycolysis, glutaminolysis, and lipid accumulation to support proliferation and T cell functions ( Figure 1C ).

After antigen elimination, a small subset of cells differentiates into long-lasting memory T cells, which provide a fast and robust immune response upon re-exposure to the antigen. As a quiescent cell population, memory T cells have a catabolic metabolism and rely on OXPHOS fueled mainly by fatty acid oxidation (112–114).

Moreover, the metabolism of T cells also varies depending on their lineage commitment. The proinflammatory CD4⁺ T cell lineages Th1, Th2, and Th17 mainly use glycolysis to cover their energy needs (115). In contrast, CD4⁺ regulatory T cells (T_regs), which dampen the immune response and promote tolerance, are less reliant on glycolysis and instead use mitochondrial metabolism and OXPHOS for energy production (116). Due to the limited reliance on glycolysis, T_regs may have a metabolic advantage to survive and function in glucose-depleted conditions (117, 118). Notably, fatty acid metabolism is essential for T_reg generation. Blockade of fatty acid oxidation using the carnitine palmitoyl-transferase-1 inhibitor etomoxir suppressed the differentiation of mouse CD4⁺ T cells into T_regs in vitro (115). These distinct metabolic requirements are critical for the optimal function of different T cell subsets and result in distinct fates in the metabolically altered tumor microenvironment.

An increasing body of evidence suggests that the strength of T cell anti-leukemia responses is regulated by metabolic cues, both within localized interactions between AML cells and immune cells, and on the organismal level. In the following sections, we will first discuss how altered AML metabolism affects immune function by shaping the microenvironment. Next, we will review the literature on how metabolism regulates the expression of inhibitory checkpoint ligands on the AML cell surface. We will then address the potential impact of myeloid cell metabolism on the anti-tumor response. Finally, we will focus on the metabolic regulation of immune responses on the organismal level. In each section, we will additionally highlight ways in which the metabolism could be therapeutically modulated to enhance the GVL effect.

In recent years, considerable progress has been made in deciphering the mechanisms by which metabolic processes instruct T cell functionality. In the same time, our understanding of how AML cells shape the metabolic BM environment, and consequently allogeneic T cell immunity, is still developing. It has been uncovered that AML-derived metabolic products, such as lactate and R-2-HG, have an immunosuppressive effect on T cells and thus might promote immune escape after allo-HCT. Furthermore, there is increasing evidence that the metabolism of AML cells contributes to the upregulation of immune checkpoint ligands, which inhibit T cell activation. Additionally, systemic metabolic events, such as fever or oxidative damage caused by alloreactivity itself, were shown to influence the T cell activity. The identified mechanisms of how the GVL activity is modulated by metabolism open new possibilities for therapeutic targeting. Improving the redox balance of alloreactive T cells using antioxidants (204) and reversing the lactic acid-induced inhibition of glycolysis in T cells (130) are two of the most promising options. New treatment options are urgently needed since the rate of AML patients relapsing after allo-HCT is still high (4). Nevertheless, most insights about the potential metabolic regulation of the GVL immune response stem from preclinical or in vitro studies and are partly extrapolated from other diseases. Defining which of these metabolic alterations recurrently happen in allo-HCT recipients and how they contribute to immune evasion is one of the essential tasks in the field.

We believe that future studies could focus on several biological aspects. First, the importance of the local metabolic niche (in the BM) versus systemic metabolism (in the peripheral blood) is not understood. The BM niche is a hypoxic environment (214), and it is likely that oxygen tension is even lower in the presence of an overt leukemic cell population. Additionally, little is known about how the BM concentrations of polar metabolites and lipids differ from their plasma levels. Recent studies in the metabolism field have uncovered that the metabolite concentrations in frequently used cell culture media differ sometimes substantially from their plasma concentrations and that utilizing media with “physiological” metabolite levels uncovers novel aspects of cancer and immune cell metabolism (215–218). It is likely that local metabolic conditions in the BM affect immune cell function, and further studies are required to define the BM niche metabolic composition in health and disease. Furthermore, it is possible that microbiota-derived metabolic products shape the phenotype and function of alloreactive T cells. In the context of GVHD, it is now established that microbiota-derived products, such as butyrate, affect intestinal inflammation (219, 220). Recent evidence suggests that distinct bacterial enterotypes are associated with alterations in the fecal and plasma metabolome, and consequently can influence T cell function (209). Further studies are warranted to dissect the contribution of individual (microbiota-derived) metabolites on the systemic level to anti-leukemia immunity.

Several studies discussed above focus on how leukemia cells might suppress the immune response by altering the concentration of metabolites in the microenvironment and thus impairing T cell activity by nutrient depletion or toxic metabolite accumulation. However, how metabolic processes affect the recognition of AML cells by the immune system has not been reported. In other tumors, a correlation between MHC expression and metabolism was found (221, 222). Inhibition of dihydroorotate dehydrogenase (DHODH), an enzyme essential for de novo pyrimidine synthesis, was shown to upregulate the expression of MHC class I in pancreatic tumor cell lines and the melanoma cell line B16F10 (221). DHODH inhibitors have demonstrated anticancer activity in different preclinical tumor models (223–225), including AML (226), and are currently tested in a phase I trial for patients with acute myeloid leukemia and myelodysplastic syndrome (NCT04609826). Since the downregulation of MHC represents a mechanism of immune escape in AML (27), future work could investigate how MHC expression is regulated by AML cell metabolism. Moreover, emerging evidence indicates that metabolic processes can drive the expression of inhibitory checkpoint ligands on AML cells, but further studies are required to understand whether this plays a role for post-allo-HCT relapses.

Broadly speaking, while most studies focus generally on how leukemia metabolism might affect T cell function (78, 79, 135, 166), very little published research has focused on mechanisms specific to the post-allo-HCT relapse setting in humans (130). It is likely that metabolic adaptations occur as the disease progresses and that relapsing AML after allo-HCT imposes unique metabolic constraints on allogeneic T cells. The specific nature of this metabolic reprogramming remains still to be understood.

Overall, more research is needed to examine the complex interaction between AML cells, T cells, and other cells present in the tumor microenvironment. Although they were not the primary focus of this review, other immune cells, such as macrophages and NK cells, are also influenced by AML metabolism. For instance, the secretion of arginase II by leukemic cells plays a role in the polarization of surrounding monocytes into tumor-supporting “M2-like” macrophages (78). These tumor-supporting macrophages can likewise contribute to the immunosuppressive tumor microenvironment, for example by expressing CD39 and CD73, which convert ATP into immunosuppressive adenosine (175, 176). In addition, the metabolism of AML cells additionally affects non-immune cells. BM adipocytes, for example, were demonstrated to release fatty acids after induction of phosphorylation of hormone-sensitive lipase by AML blasts (83) and BM stroma cells were found to transfer mitochondria to AML cells (193, 194). The immune and non-immune cells modulated by AML metabolism can, in turn, interact with the T cells and thus affect their activity. Understanding the metabolic interplay of all different cell types holds the potential to inform novel therapeutic developments.

A-CB: Conceptualization, Investigation, Writing – original draft, Writing – review & editing. PA: Conceptualization, Writing – original draft, Writing – review & editing.