Clinical Responses to Oral Vorinostat (Suberoylanilide Hydroxamic Acid, SAHA) Are Associated with Specific Gene Expression Signatures in Patients with Advanced Leukemias: Results of a Phase I Trial.

Vorinostat is a histone deacetylase inhibitor with potent in vitro and in vivo anti-leukemia activity. Several gene expression signatures were developed from preclinical models as biomarkers for clinical response to vorinostat. The signatures were evaluated in peripheral blood mononuclear cell (PBMC) samples from a Phase I study of oral vorinostat in patients with advanced leukemias or myelodysplastic syndromes (MDS). Of 41 registered patients, 31 had acute myeloid leukemia (AML), 4 had chronic lymphocytic leukemia, 3 had MDS, 2 had acute lymphoblastic leukemia, and 1 had chronic myeloid leukemia. Seven of 41 patients showed improvement with treatment, 2 complete responses, 2 complete responses that were not maintained for 28 days, and 3 with >50% decreases in blast counts without count recovery. All responses occurred in patients with AML at or below the MTD. The following three clinical gene expression signatures were shown to be predictive of clinical response in PBMC samples of AML patients. First, a positive prediction trend was observed when a preclinical 40-gene resistance prediction signature was evaluated in baseline (pre-dose and first day of dosing) AML PBMC samples (8 responder and 21 non-responder samples) or all PBMC samples (14 responder and 56 non-responder samples, including non-AML patients). If this trend is maintained in pretreatment samples of a larger patient population, it may allow for exclusion of up to 50% of non-responders from future clinical trials. Second, a similar prediction of resistance was obtained based on expression levels of known reactive oxygen species scavengers that were upregulated in non-responders relative to responders, indicating that cellular ability to withstand oxidative stress may be protective with respect to vorinostat exposure. This supports the hypotheses that oxidative stress pathways play an important role in the response to vorinostat. Finally, overexpression of p21 and several p53-responsive genes was also associated with resistance, consistent with preclinical observations. These results are being validated at the protein level in a tissue array platform. Post-dose downregulation of genes associated with cellular proliferation was also observed in responders but not in non-responder PBMC samples. This suggests that proliferation may be used as a post-dose efficacy biomarker. Further evaluation of vorinostat in patients with leukemias and MDS, including analysis of a potentially predictive gene expression signature for response to vorinostat, is warranted.