Dendrogenin A (DDA) is a newly discovered cholesterol metabolite with tumor suppressor properties. Here, we explored its efficacy and mechanism of cell death in melanoma and acute myeloid leukemia (AML). We found that DDA induced lethal autophagy in vitro and in vivo, including primary AML patient samples, independently of melanoma Braf status or AML molecular and cytogenetic classifications. DDA is a partial agonist on liver-X-receptor (LXR) increasing Nur77, Nor1, and LC3 expression leading to autolysosome formation. Moreover, DDA inhibited the cholesterol biosynthesizing enzyme 3β-hydroxysterol-Δ8,7-isomerase (D8D7I) leading to sterol accumulation and cooperating in autophagy induction. This mechanism of death was not observed with other LXR ligands or D8D7I inhibitors establishing DDA selectivity. The potent anti-tumor activity of DDA, its original mechanism of action and its low toxicity support its clinical evaluation. More generally, this study reveals that DDA can direct control a nuclear receptor to trigger lethal autophagy in cancers.

BackgroundDeficiency in dihydropyrimidine dehydrogenase (DPD) enzyme is the main cause of severe and lethal fluoropyrimidine-related toxicity. Various approaches have been developed for DPD-deficiency screening, including DPYD genotyping and phenotyping. The goal of this prospective observational study was to perform exhaustive exome DPYD sequencing and to examine relationships between DPYD variants and toxicity in advanced breast cancer patients receiving capecitabine.MethodsTwo-hundred forty-three patients were analysed (88.5% capecitabine monotherapy). Grade 3 and grade 4 capecitabine-related digestive and/or neurologic and/or hemato-toxicities were observed in 10.3% and 2.1% of patients, respectively. DPYD exome, along with flanking intronic regions 3’UTR and 5’UTR, were sequenced on MiSeq Illumina. DPD phenotype was assessed by pre-treatment plasma uracil (U) and dihydrouracil (UH2) measurement.ResultsAmong the 48 SNPs identified, 19 were located in coding regions, including 3 novel variations, each observed in a single patient (among which, F100L and A26T, both pathogenic in silico). Combined analysis of deleterious variants *2A, I560S (*13) and D949V showed significant association with grade 3–4 toxicity (sensitivity 16.7%, positive predictive value (PPV) 71.4%, relative risk (RR) 6.7, p<0.001) but not with grade 4 toxicity. Considering additional deleterious coding variants D342G, S492L, R592W and F100L increased the sensitivity to 26.7% for grade 3–4 toxicity (PPV 72.7%, RR 7.6, p<0.001), and was significantly associated with grade 4 toxicity (sensitivity 60%, PPV 27.3%, RR 31.4, p = 0.001), suggesting the clinical relevance of extended targeted DPYD genotyping. As compared to extended genotype, combining genotyping (7 variants) and phenotyping (U>16 ng/ml) did not substantially increase the sensitivity, while impairing PPV and RR.ConclusionsExploring an extended set of deleterious DPYD variants improves the performance of DPYD genotyping for predicting both grade 3–4 and grade 4 toxicities (digestive and/or neurologic and/or hematotoxicities) related to capecitabine, as compared to conventional genotyping restricted to consensual variants *2A, *13 and D949V.