Quantification of the preclinical and clinical relationship between pRAD50 and efficacy after treatment with the ATR inhibitor ceralasertib (AZD6738)

Abstract

Ceralasertib is a potent and selective ATP competitive inhibitor of ataxia telangiectasia and Rad3 related (ATR) in clinical development as monotherapy and in combination with olaparib (Lynparza) and durvalumab (Imfinzi) in patients with advanced solid tumors. Paired pre- and on-treatment tumour samples from seven patients during ceralasertib monotherapy in Phase 1 studies (NCT02223923, NCT02264678), all ATM expressing, showed increases in pRAD50 on treatment. This study aimed to robustly understand the relationship between ceralasertib pharmacokinetics, pRAD50 (pSer635) induction by immunohistochemistry and anti-tumor efficacy, in mouse xenografted models; to enable interpretation of paired clinical tumour samples. First, in vivo data were generated in a range of xenografted models (HBCx9 [TNBC, Xentech], HCC1806 [BC], OCI-Ly19 [DLBCL] and OE21 [HNSCC]) and NSCLC and HNSCC PDX models at Champions). Mouse ceralasertib doses (6.25 - 25mg/kg BID and 50mg/kg QD) reflected the observed free drug exposure achieved at clinical doses. In pharmacodynamic studies, animals were dosed continuously for 5 days, with PKPD endpoints being assessed on the 5th day. In anti-tumor studies, animals were randomized and treated for at least 28 days continuously. Increases in pRAD50 on treatment, versus control, were both dose and time dependent. There was a wide range of baseline pRAD50 expression (H-Scores ranging from 2 to 48) but the fold increase was consistent; a 3-fold increase being seen at the highest dose. Similarly, the anti-tumor activity was dose dependent with a range of sensitivities being observed across the models; the OCI-Ly19 model was most sensitive with %TGI ranging 43% to 104%. Mathematical modelling confirmed a consistent PKPD relationship for the fold increase of pRAD50. The effect was not saturated at the tested drug exposures and the slope relating drug in plasma to fold pRAD50 increases was estimated to be 0.75 uM-1. Modelling also demonstrated a relationship between pRAD50 and the observed anti-tumor activity, across all models. Of note, the fold induction required for efficacy increased with the baseline pRAD50 The hypothesis for this observation is that the baseline pRAD50 H-score is a functional measure of DDR signaling and so low pRAD50 may be indicative of impaired signaling and thus sensitivity to inhibition of DDR signaling. The results demonstrate the utility of generating PKPD-Efficacy relationships across a range of patient relevant xenografts and PDXs. Here, a marker of ATR inhibitor pharmacology, pRAD50, has been related to anti-tumor activity increasing our understanding of predictive markers of drug sensitivity. Clinical translation of these findings is being tested in a Phase I study, where patients provide paired tumour biopsies pre-treatment and following ceralasertib monotherapy dosing