Genomic profiling identifies distinct genetic subtypes in extra-nodal natural killer/T-cell lymphoma

Abstract

Extra-nodal NK/T-cell lymphoma, nasal type (ENKTCL) is a highly aggressive Epstein-Barr virus associated lymphoma, typically presenting in the nasal and paranasal areas. We assembled a large series of ENKTCL (n = 209) for comprehensive genomic analysis and correlative clinical study. The International Lymphoma Prognostic Index (IPI), site of disease, stage, lymphadenopathy, and hepatomegaly were associated with overall survival. Genetic analysis revealed frequent oncogenic activation of the JAK/STAT3 pathway and alterations in tumor suppressor genes (TSGs) and genes associated with epigenomic regulation. Integrated genomic analysis including recurrent mutations and genomic copy number alterations using consensus clustering identified seven distinct genetic clusters that were associated with different clinical outcomes, thus constituting previously unrecognized risk groups. The genetic profiles of ENTKCLs from Asian and Hispanic ethnic groups showed striking similarity, indicating shared pathogenetic mechanism and tumor evolution. Interestingly, we discovered a novel functional cooperation between activating STAT3 mutations and loss of the TSG, PRDM1, in promoting NK-cell growth and survival. This study provides a genetic roadmap for further analysis and facilitates investigation of actionable therapeutic opportunities in this aggressive lymphoma.

Fig. 1:. Patient cohort and clinical information.

(a) Schematic flowchart of the study. (b) Kaplan-Meier curve of OS by treatment type for all cases (left panel) and the 171 Asian cases with more complete clinical information selected for further studies (right panel). Two Asian cases without treatment information were excluded from the analysis. (c-d) OS (c) and OS according to IPI (d) of the 171 Asian patients. Four patients without IPI information were excluded from the analysis. RT, radiotherapy; SCT, stem cell transplantation.

Fig. 2:. Mutations identified in 209 ENKTCL cases.

(a) Waterfall plot depicting the frequency of non-silent mutations identified in at least 4% of the cases. (b) Frequency of mutations affecting the noted pathways, integrating whether a case has a mutation in any gene in the listed pathway. (c) Schematic domains affected in the top mutated genes. (d) Pathway enrichment of mutated genes predicted by IPA with mutated gene number and their percentage in the whole gene list of the pathway represented by dot size and color respectively. (e) Pairwise Fisher’s Exact Test to identify mutations that co-occur or are mutually exclusive.

Fig. 3:. Frequency of gCNAs in 209 ENKTCLs and comparison between two ethnic groups.

(a) Regions identified by GISTIC 2.0 as significantly gained or lost are noted in pink bars at the top or turquoise bars at bottom of the graph, respectively. Genes potentially significant in tumor development located at the peaks are noted. Genes in red were found to be mutated in at least 4% of cases. (b) Frequency of gCNAs in Asian and Hispanic cohorts. (c) Comparison of the most frequent mutations and gCNAs identified by GISTIC2.0 between the Asian and Hispanic cases. *P < 0.05.

Fig. 4:. Genetic clusters identified using NMF Consensus Clustering.

(a) NMF consensus clustering was performed using mutations and CNAs in 209 ENKTCL cases. Clusters C1-C7 represent the genetic clusters identified by unique genomic signatures. CN gains and losses are represented in red and blue respectively, and mutations are represented in black. The header shows cluster annotation with clinical parameters/patient demographics such as race, site, stage, and IPI score. Alterations significantly different across more than one cluster are shown. The lower two bar plots represent the number of mutated genes and proportion of aberrant genome in each case. (b) Kaplan-Meier curve showing OS across all clusters. (c) Kaplan-Meier curve showing OS across patients in combined clusters. C1, C2, C3, C4 were combined as a single group due to similar survival probability. Similarly, patients in clusters C5 and C7 showed good prognosis, and thus were combined. Patients in cluster C6 showed the worst prognosis.

Fig. 5:. Cooperation between PRDM1 loss and STAT3 mutation in promoting NK-cells growth.

(a) Schematic flow chart of the in vitro experiments. Freshly isolated human NK-cells were cultured with irradiated feeder cells before introduction of PRDM1 deletion (RFP⁺). Unmodified or PRDM1-null cells were further transduced with indicated lentiviral construct that expressed GFP. The growth of the transduced cells was monitored by the percentage of GFP⁺ cells. (b) Western blot analysis of PRDM1 loss and STAT3 phosphorylation in control (Cas9) and modified NK-cells (sorted RFP⁺ PRDM1 KO, sorted RFP⁺ GFP⁺ PRDM1 KO + STAT3 WT/STAT3^D661Y/STAT3^Y640F labeled as “KO + STAT3 WT/D661Y/Y640F”). The intensity of pSTAT3 (Y705) and total STAT3 bands was normalized to that of GAPDH bands. (c) Transduced NK-cells were expanded with irradiated feeder cells added every week. The percentage of GFP⁺ cells was monitored by flow cytometry. (d, e) At day 3 after adding feeder cells, apoptosis (d) or proliferation (e) of the indicated groups of cells was analyzed by staining with Annexin V and DAPI, or by the EdU assay, respectively. n = 3. (f) Irradiated feeder cells were added at day 0 and most feeder cells were dead after three to five days. At day 7, 0.5 million NK-cells were seeded in a 12-well plate in triplicate with the indicated single cytokines (left panel) or the combinations of cytokines (right panel). Cells were counted periodically with trypan blue. The Y-axis is shown in a log-2 scale. ***, P < 0.001.