Metabolic signatures mirror divergent modalities of leukemia immune evasion and relapse after transplantation

Abstract

Abstract Acute myeloid leukemia (AML) relapse after allogeneic hematopoietic cell transplantation (allo-HCT) is increasingly recognized as a consequence of immune escape mechanisms by which cancer cells evade the graft-versus-tumor effect. Recent studies have shown that, in most relapses, these immune changes are driven by epigenetic rewiring of AML cells, leading to two distinct and mutually exclusive immune evasion strategies: downregulation of HLA class II molecules or upregulation of T-cell inhibitory ligands (e.g., PD-L1, B7-H3) (Christopher et al., NEJM, 2018; Toffalori et al., Nat Med, 2019). To investigate whether these relapse modalities are associated with distinct metabolic patterns, we analyzed 14 longitudinally collected sample pairs from AML patients at diagnosis and post-transplantation relapse, evenly distributed between the two modalities. Following AML blast purification, we conducted bulk RNA sequencing and liquid chromatography–mass spectrometry (LC-MS) profiling of intracellular metabolites (endometabolites) and extracellular metabolites from supernatants after short-term culture in a metabolically defined medium (exometabolites). In parallel, we functionally characterized these samples using the Single Cell ENergetIc metabolism by profilIng Translation inhibition (SCENITH) assay (Argüello et al., Cell Metabolism, 2020). In line with our previous studies, transcriptional profiling clearly separated the two patterns of relapse: downregulation of HLA class II molecules (downHLA, n=7), and upregulation of T cell inhibitory ligands (upInib, n=7). Notably, we observed that most transcriptional and metabolic alterations were exclusive to each modality, with more evident metabolic rewiring in upInib relapses. In particular, biological processes enriched in these relapses included hypoxia (p=0.035) and nitric oxide (NO) biosynthesis (p=0.029), and endometabolome analysis highlighted increased intracellular levels of citrulline (FC=2.22, raw p=0.015) and arginine metabolites (FC=1.44, raw p=0.046), which are known substrates in nitric oxide synthesis. Notably, we also detected marked increase in intracellular pyroglutamic acid (FC= 1.71, raw p=0.046), a key intermediate in the γ-glutamyl cycle, suggesting a transcriptional and metabolic shift towards a pro-oxidative/nitrosative state at relapse. These findings indicate that upInib relapses adapt to a pro-inflammatory and pro-oxidative microenvironment and may rely more heavily on tumor microenvironment (TME)-driven metabolic reprogramming than their downHLA counterparts. Our analysis also revealed metabolic changes common to both relapse modalities. For instance, SCENITH demonstrated similar alterations in bioenergetic pathways in the two types of relapse, with increased mitochondrial dependence and reduced glycolytic capacity (p = 0.015). Furthermore, independently from the immune escape modality, in the exometabolome we detected a significant increase in lactic acid (LA) at relapse (p=0.007), which was confirmed by a validation colorimetric assay performed using both primary patient samples and AML patient-derived xenografts. Besides its known activity as immunosuppressive oncometabolite, LA has recently been shown to function also as a direct epigenetic regulator (Zhang et al, Nature, 2019). For this reason, we investigated by immunofluorescence analysis lactylation of specific hystone sites in our relapses, detecting a significant increase in H4K8la (p=0.04). CUT&Tag sequencing is currently underway to identify differentially lactylated genes and thus the role of pan- and residue-specific lactylation in driving relapse-associated gene expression programs.Taken together, our data provide a detailed landscape of the metabolic alterations associated with AML relapse, highlighting that most of them are modality-specific. These findings may serve as a roadmap to test therapeutic approaches aimed to untangle the epigenetic/metabolic TME rewiring driven by AML relapse, especially for upInib cases in which direct targeting by checkpoint blockade has yet largely failed at improving patient outcome.