217 Cytotoxicity of nicotinamide enhanced natural killer cells GDA201 is based on metabolic modulation as demonstrated by artificial intelligence assisted analysis of NK cell transcriptome and metabolome

BackgroundNicotinamide (NAM), an allosteric inhibitor of NAD-dependent enzymes, has been shown to preserve cell function and prevent differentiation in ex vivo cell culture. GDA-201 is an investigational natural killer (NK) cell immunotherapy derived from allogeneic donors and expanded using IL-15 and NAM. In previous preclinical studies, NAM led to increased homing and cytotoxicity, preserved proliferation, and enhanced tumor reduction of NK cells. In a phase I clinical trial, treatment with GDA-201 showed tolerability and clinical responses in patients with refractory non-Hodgkin lymphoma (NHL) (Bachanova, et. al., Blood 134:777, 2019). While NAM is known to affect cellular metabolism and participate in 510 enzymatic reactions −in 66 as an inhibitor or activator− its mechanism of action and role in GDA-201 cytotoxicity is unknown.MethodsIn order to define the network of intracellular interactions that leads to the GDA-201 phenotype, flow-cytometry, next generation sequencing (NGS), and liquid chromatography–mass spectrometry (LC-MS)-based metabolite quantification were performed on NK cells cultured for 14 days with IL-15 and human serum in the presence or absence of NAM (7 mM). Artificial Intelligence (AI) machine learning analysis was applied by Pomicell in order to analyze the data using the Pomicell databases supporting data extracted from multiple origins including scientific articles organized using natural language processing tools. AI training was done using a combined algorithm designed to blindly explain and predict the transcriptomic and metabolomic (omics) profile.ResultsOmics analyses defined 1,204 differentially expressed genes, and 100 significantly modified metabolites in the presence of NAM. An in silico model was created that successfully predicted the experimental data in 83% of the cases. Upregulation of TIM-3 expression in GDA-201 was predicted to be mediated by inhibition of IL-10 and SIRT3, via CREB1/HLA-G signaling and adrenoceptor beta 2 (ADRB2) upregulation. Adenosine metabolite reduction supports this and suggests dopaminergic activation of NK cytotoxicity. Upregulation of CD62L in the presence of NAM was predicted to be mediated by transcription factor Dp-1 (TFDP1) via dihydrofolate reductase (DHFR) activation and intracellular folic acid reduction. Interferon-gamma and CASP3 modulation (via JUN and MCL1, respectively), via PPARa inhibition, support that finding.ConclusionsIn conclusion, AI machine learning of transcriptome an