Abstract Background Treatment outcomes have improved dramatically over the past several decades for many pediatric patients with acute lymphoblastic leukemia (ALL); however, progress is still lacking for those who experience relapse or refractory disease. To further improve outcomes for these patients, new strategies need to be devised to both identify patients most at risk of relapse a priori to treatment exposure and to optimally match them with emerging new therapeutic options. To that end, we report the development of a combination drug screening assay that correlates with clinical outcomes using standard pediatric ALL treatment protocols. This assay utilizes a microfluidic device to generate overlapping drug gradients and implements a novel analysis framework that characterizes drug synergy and efficacy in a ratiometric manner. Methods The developed device exposes cells embedded in a 3D culture system to three stable, overlapping concentration gradients of small molecule drugs. Concentration gradients overlap in a manner such that each differing location within the hydrogel region comprises a unique combination ratio of all three drugs and spans a cut plane in normalized, 3D concentration space. The system enables efficient analysis of response to all possible drug ratios within a three-drug combination, with reduced experimental burden and far fewer cells than would be required in standard systems. Single cell drug concentration exposure is determined from experimentally validated computational fluid dynamics (CFD) model predictions, and single cell viability is analyzed via calcein AM (live), propidium iodide (dead) and hoechst (nuclear) staining following 48 hours of treatment. We further implement a multi-device experimental schema in which we test multiple devices at serially decreasing input dose levels. This increases coverage of the drug combination concentration space and allows for custom implementation of established synergy models. Results Using this developed combination drug screening assay and analysis framework, we retrospectively evaluated response to a combination of chemotherapies commonly used in induction therapy across a set of aspirate samples collected from pediatric ALL patients at diagnosis, acquired frozen from an established biorepository (n = 10). When comparing patient specific responses, we observed that patients whose disease responded poorly to standard treatment had a unique functional response to standard chemotherapies in vitro in our system. Moreover, using supervised machine learning, we found that our combination response metrics were predictive of both end-of-induction minimal residual disease (MRD) as well as later clinical endpoints. Specifically, the developed model explained a large portion of the variance in clinical outcome metrics (cumulative R2Y = 0.99), was predictive when cross-validated with the leave-one-out method (Q2 = 0.84), and could accurately classify patients according to their clinical course. Lastly, we demonstrated that our system could evaluate response to experimental therapeutic regimens in patients who did not respond to standard treatment. Conclusion In summary, the work presented here establishes a new platform for ratiometric, combinatorial drug screening, and envisions a future in which this assay could be used, in the clinic and at the time of diagnosis, to actionably identify difficult-to-treat ALL patients, screen for alternative treatment strategies, and overall improve patient outcomes.
Williams et al. (Wed,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: