The measurement of drug dose-response is the cornerstone of pre-clinical assessments of novel therapeutics. High-throughput studies have attempted to identify intrinsic drivers of drug sensitivity by measuring the response of hundreds of cell lines to hundreds of drugs and it has been proposed to use the drug response of primary human tumor cells as a way to personalize therapy for individual patients.

The commonly used metrics to parameterize drug response (IC50, Emax, or AUC) are based on assessing the cell count of a treated condition relative to an untreated control. Yet all of these metrics suffer from a fundamental flaw: Cell lines undergoing more divisions over the course of an assay—be it due to the length of the experiment or speed of division—are scored as more sensitive than cell lines with less divisions, even if their inherent drug sensitivities are identical. We developed a new method to parameterize drug response, the growth rate inhibition (GR) metrics, which is based on the ratio of growth rates under treatment conditions in relation to an untreated control. The GR metrics are independent of cell growth over the course of the experiment and thus enable us to accurately compare cell lines with varying growth rates or experimental conditions, like genetic or micro-environment perturbations, that can alter growth rates.

We use the GR metric to reanalyze a high throughput dataset of approximately 50 breast cancer cell lines that were treated with therapeutic inhibitors. We find that the GR50 clearly identifies ErbB2amp cells as being highly sensitive to lapatinib, an ErbB2 inhibitor, a fact that is obscured by the classic IC50 metric due to the slow growth of ErbB2amp cell. For paclitaxel, the IC50 suggests that the sensitivity of the cell lines spans a range of more than 100-fold. Surprisingly, the GR50 shows that all breast cancer cell lines are about equally sensitive to paclitaxel, at a value consistent with the drug's binding affinity to stabilized microtubules in vitro. It is rather the maximal response (GRmax) that distinguishes the response of the different lines.

We also apply the GR to study how drug responses are affected by plating density, an often poorly controlled variable in drug response measurements. Previous reports suggest that at high density cells generally become more resistant to therapy, however, we find that this is often an artifact of the IC50 metric due to slowed growth at high confluence. The emerging picture is more complex, with trends that are rarely uniform across cell lines or drugs. It appears that drug- or cell line-specific biological mechanisms drive density-dependent drug responses, for example through media conditioning effects or metabolic changes.

Our work shows the importance of using a metric based on growth rates for any systematic comparison of drug sensitivity where the speed of growth is not uniform for all compared cells. This includes high-throughput profiling of cell lines, but also the comparisons of patient-derived tumor and normal cells, or any system where cells are manipulated in ways that can affect their growth rates, like changing the microenvironment or genetically altering the cells.

Citation Format: Mario Niepel, Marc Hafner, Peter K. Sorger. A novel analytical approach to accurately assess in vitro drug responses for breast cancer therapy. [abstract]. In: Proceedings of the AACR Special Conference on Advances in Breast Cancer Research; Oct 17-20, 2015; Bellevue, WA. Philadelphia (PA): AACR; Mol Cancer Res 2016;14(2_Suppl):Abstract nr B39.