Multi-Omics Profiling Reveals Tumor Cellular Landscape and Metabolism Remodeling during Extranodal Natural Killer/T-Cell Lymphoma (ENKTL) Metastasis and Progression

Background :ENKTL is an aggressive lymphoma associated with EBV infection. ENKTL mostly occurs in the upper aerodigestive tract (UAT), especially the nasal cavity, but it also manifest in the non-UAT (NUAT) including the skin, lung, etc. Compared to UAT-NKTL, NUAT-NKTL (Model-III) is associated with more inferior survivals. Detailed in UAT-NKTL, patients (pts) with advanced stage (Model-II) responded poorly to first-line treatment leading to worse prognosis compared with that of early stage (Model-I). Since significantly different clinical outcomes were observed among these 3 models which were also validated in the result of 341 newly-diagnosed (ND) pts in our center, we try to use multi-omics profiling to reveal tumor cellular landscape especially combined with EBV tag and metabolism remodeling during ENKTL metastasis and progression in order to better understand the origins of the heterogeneity and pathogenesis. Methods:The study enrolled 65 ENKTL pts from our center. We profiled using DNA-target gene sequencing (N=42), RNAseq (N=35), metabolomic assay (N=42) and scRNAseq with EBV tag (molecular tags including EBNA1, EBNA2, EBERs, ZEBRA) (N=23). Results: Firstly, at the mutational level, there was no significant difference in the proportion and profiling of mutated genes among the 3 models. Subsequently, as anticipated, immense differences were found among the three models at transcriptional level. KEGG enrichment analysis of signature genes revealed that Model-I was characterized by focal adhesion and proteoglycans in cancer. Model-II featured activation of multiple carcinogenic pathways (PI3K-Akt and Hippoa) while Model-III exhibited upregulation of lipid and nucleotide metabolism pathways reprogramming. Further untargeted metabolomics analysis revealed that Model-III displayed significant enrichment of multiple lipid metabolites and relative deficiency in amino acid metabolites at diagnosis. In contrast, Model-I was characterized by a lack of lipid metabolites. We further analyzed the dynamic change of the metabolic features of the 3 models according to the response. To further investigate cellular and EBV heterogeneity underlying different models and tumor cell from UAT to NUAT, we performed scRNAseq on 23 samples: tumor cell from nasal cavity for ND pts [GroupA, N=4] and relapse pts (GroupB, N=4) with Model-I; GroupC: tumor cell from nasal cavity for ND pts with Model-II [N=5]; tumor cell outside UAT for ND pts (GroupD, N=5) and relapse pts (GroupE, N=5) with Model-II). After quality control, 280,939 cells were categorized into 12 main cell types, with T/NK cells further subdivided into 19 subclusters (T: NaiveT, Treg, Tex, Teff, NKT; NK: C1-C14). For the first time, we characterized the spectrum of EBV infection in the cellular landscape using EBV tags introduced at the single-cell level. EBV primarily infected NK cells (with infection rates ranging from 10% to 67%, except NK_C13) and Teff cells, and inferCNV predicted that all of these cells were tumor cells. Interestingly, these EBV+ cells exerted completely different functions compared with EBV- cells underlying the pathogenic role in ENKTL metastasis. Further analysis revealed heterogeneity in cellular composition and function. In GroupA, EBV+ cells are extremely rare, NK_C1 exerts cytotoxic effects to eliminate EBV and promote apoptosis of tumor cells, and adhesion and tight junctions confine lesions. In GroupB, the proportions of NK_C5 and NK_C6 are similar, and the immune and cytotoxic functions are comparable. However, unlike GroupA, EBV is difficult to clear and the adhesion function still confines the lesions to the nasal cavity. GroupC exhibits a downregulation of adhesion, making it more likely to metastasize to non-nasal areas, and the cytotoxic ability related to NK cells is also reduced. Instead, it has stronger biosynthetic capabilities and antigen presentation abilities. GroupD and E, despite being distinctly different in cell type (GroupD is the only Teff-dominated group), both showed significant metabolic reprogramming features: GroupD showing an upregulation of stronger folate biosynthesis while Group E focuses more on the activation of the oxidative phosphorylation pathway. Conclusions: Multi-omics profiling combined EBV target sequencing of different models in our study reveals tumor cellular landscape and metabolism remodeling during ENKTL metastasis and progression.