(P 288) EVALUATION OF ATR AND WEE1 INHIBITORS COMBINED WITH NANO-LIPOSOMAL IRINOTECAN IN A LARGE COHORT OF RHABDOMYOSARCOMA XENOGRAFTS USING A TWIN MOUSE PRECLINICAL TRIAL DESIGN

Objective: Rhabdomyosarcoma (RMS) is an aggressive malignancy of childhood which has a poor prognosis when it has metastasized. Although there has been improvement in overall survival for low and intermediate risk patients, there has been no significant change in outcome in the past 25 years for patients with high-risk disease. Previous preclinical studies have shown that RMS orthotopic patient-derived xenografts (O-PDXs) are sensitive to cell cycle checkpoint inhibitors in combination with DNA damaging agents. Here, we compare a WEE1 inhibitor (adavosertib) to an ATR inhibitor (elimusertib) in combination with relevant standard of care agents across 36 O-PDX models (25 embryonal, 10 alveolar, 1 sclerosing).

Methods: In order to encompass the heterogeneity of patient tumors, we used a “twin-mouse” study design. We enrolled two mice for each model from our collection of fully characterized rhabdomyosarcoma O-PDXs (www.stjude.org/CSTN) for every treatment group. Each model was labeled with firefly luciferase for weekly bioluminescent monitoring of tumor burden. The following treatment groups were utilized for testing: placebo, elimusertib + nano-liposomal irinotecan (nal-IRI), adavosertib + nal-IRI + vincristine (VCR), and nal-IRI + VCR. Enrollment was randomized for each of the O-PDXs tested. Four courses (21 days/course) of therapy were given on a clinically relevant schedule using pharmacokinetic guided dosing.

Results: The majority of models (86%) treated with nal-IRI + VCR had some survival advantage compared with placebo. Of the 36 models tested, we found 1 tumor model generated from a patient with refractory disease that failed to respond to any combination (SJRHB013758_X2, ERMS). In contrast, a model generated from the same patient at diagnosis (SJRHB013758_X1) was sensitive to all combinations tested compared to placebo. Similarly, there was a model generated at first relapse (SJRHB013759_X1, ARMS) that responded to elimusertib + nal-IRI, adavosertib + nal-IRI + VCR and nal-IRI + VCR but subsequent models generated from the same patient at autopsy (SJRHB013759_X14, SJRHB013759_X15) only responded to elimusertib + nal-IRI. There were two models (SJRHB012405_X1, ERMS; SJRHB015720_X1, sclerosing) that responded to both combinations of elimusertib + nal-IRI and adavosertib + nal-IRI + VCR but not to nal-IRI + VCR alone.

Conclusion: The twin mouse experimental design is feasible and represents an attractive strategy for testing diverse models of disease to both identify and prioritize new exploratory combinations. By using two mice to represent each tumor sample, redundancy of each model is provided while still remaining cost effective. By testing a vast collection of RMS O-PDXs in this study, we obtained an understanding of the baseline response of each of the models to standard chemotherapy agents in comparison to the experimental combinations of ATR and Wee1 kinase inhibitors. Furthermore, the identification of resistant models represents an opportunity to better understand the underlying molecular mechanisms of disease resistance and additional studies of these models are warranted.