Abstract Introduction: Small cell lung cancer (SCLC) is a common and rapidly fatal malignancy for which no biomarker-targeted therapies have been developed. Despite a critical need, progress suffers from (1) scarcity of cutting-edge laboratory models and (2) absence of promising targets. Patient-derived xenografts (PDX) may faithfully model the clinical disease, but because SCLC is rarely biopsied or resected, specimens for PDX generation are scarce. PARP inhibition has recently emerged as a compelling strategy for SCLC, and an ongoing phase 1/2 trial of combination olaparib tablets and temozolomide (O/T) has shown promising activity in patients. However, biomarkers for patient selection remain elusive. Methods: We generated SCLC PDX models from circulating tumor cells (CTCs), biopsies and malignant effusions. CTCs were enriched with an automated microfluidic device, the CTC-iChip. To assess the genomic fidelity of the models, we performed comparative whole exome sequencing (WES) and RNA-seq on 6 sets of corresponding patient biopsies, founder (P0) PDX tumors, and early-passage PDXs. We then assessed the activity of combination O/T in a panel of PDX models, and compared PDX responses with molecular profiles to identify candidate biomarkers. Results: 44 PDXs were generated from 32 patients, including 6 sets of serial models and 4 synchronous CTC- and biopsy-derived models. PDXs were derived with high efficiency from both CTCs (35% per blood draw) and biopsies/effusions (82% per implant). WES demonstrated that somatic alterations in tumor biopsies were stably maintained in both CTC and biopsy-derived models, without significant accumulation of new mutations, and transcriptional profiles remained consistent through early passages. Six models were derived from O/T trial patients, including two sets of serial models before and after durable responses. The serial models faithfully recapitulated patient responses to O/T: pre-trial models were highly sensitive and post-relapse were highly resistant. The co-clinical trial was expanded to 30 models, using the models derived from trial patients to delineate the margins of clinical sensitivity (6 models), intermediate sensitivity (6 models) and resistance (18 models). Among the molecular features evaluated, basal protein PARylation best distinguished the O/T-sensitive category from both intermediate (p < 0.001) and resistant models (p < 0.0001). In addition, PARylation decreased after relapse in serial models from O/T trial patients. Conclusions: Both biopsy- and CTC-derived SCLC PDX models faithfully recapitulate the genomic and functional features of the donor patient tumor. O/T sensitivity in this panel correlated with basal PARylation. The value of the co-clinical trial is the potential to refine the clinical application of O/T in real time, to optimize follow-on clinical trials and to develop biomarker-directed therapy for SCLC. Citation Format: Benjamin J. Drapkin, Julie George, Marcello Stanzione, Beow Y. Yeap, Mari Mino-Kenudson, Camilla L. Christensen, Ruben Dries, Sarah Phat, Jun Zhong, David T. Myers, Joseph A. Licausi, Tilak Sundaresan, Marina Kem, Nima Abedpour, Leica V. Sequist, Alice T. Shaw, Aaron N. Hata, Mehmet Toner, Shyamala Maheswaran, Daniel A. Haber, Martin Peifer, Roman K. Thomas, Anna F. Farago, Nicholas J. Dyson. Co-clinical trial of olaparib and temozolomide in SCLC PDX models uncovers new biomarkers of sensitivity [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2018; 2018 Apr 14-18; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2018;78(13 Suppl):Abstract nr 2972.