Christa Ringers

Cell Painting is a popular high-content imaging assay that captures the shape and internal organization of 2D cell cultures at scale. While 3D cell cultures are more physiologically relevant models – especially for studying drug perturbations – they remain challenging to image because light scatters before it reaches the cultures’ interior. Here, we introduce a scalable method to adapt Cell Painting for 3D spheroids. We show that tissue-clearing, a technique that makes spheroids transparent, combined with a focusing routine that locates spheroids on the fly, enables Cell Painting at single-cell level inside intact spheroids. By profiling 52 clinically relevant drugs, we show that single-cell fingerprints clustered by biological pathway and resolved drug effects more robustly than those of flattened spheroid-wide views. Notably, the effect of two drugs, 5-fluorouracil and olaparib, nearly indistinguishable in 2D monolayers, diverged in spheroids. This finding was then corroborated by DNA-damage, cell-cycle, and gene-expression assays. The workflow runs on a spinning-disk confocal microscope in a standard 384-well plate format, making single-cell 3D profiling accessible to laboratories already performing Cell Painting.

Detailed description of methods.

Sensitivity of all samples as sensitive or resistant to oxaliplatin and SN-38 using median GR50 and median AOC.

Classification of the eight fastest growing samples as either sensitive or resistant to oxaliplatin and SN-38 at day 7 of the experiment according to either median ED20 or median GR50.

Growth rate (relative total area) of negative controls in all growing samples.

Classification of Samples 1, 3, 7, 8, 10, 12, and 13 as either sensitive or resistant to oxaliplatin and SN-38 for the total duration of the experiment, according to the median GR50 of all 16 samples on day 7 of the experiment.

A and B, Point estimates of sample library AOC. The dotted red line represents median AOC, and the blue area represents the IQR. Dot color indicates sample sensitivity according to the median, with red indicating resistant and blue indicating sensitive. C and D, Correlation plots of GR50 and AOC. The y-axis is log-transformed, and the x-axis is linear. The dotted blue line indicates the median GR50, the dotted red line indicates the median AOC, and the whole blue line indicates linear regression of log2(GR50) and AOC with CIs. GR50 is in μmol/L for oxaliplatin (left) and nmol/L for SN-38 (right).

Dose–response curves for SN-38 generated using the log-logistic 3-parametric model fitted to either relative total area (A and B) or GR metrics (C and D). The x-axis displays drug concentration in nanomolar (nmol/L), and the y-axis displays inferred response (normalized relative total area/GR metric). Solid black line indicates the library mean inferred response across both fast- and slow-growing samples per readout, and dashed red line indicates the mean inferred response for only the fast- or slow-growing samples, respectively, per readout. Vertical red lines indicate median ED values for increasing levels of inhibition (20%, 30%, 40%, and 50%) across both fast- and slow-growing samples. A, Dose–response curves of the eight slowest-growing samples fitted to relative total area. B, Dose–response curves of the eight fastest-growing samples fitted to relative total area. C, Dose–response curves of the eight slowest-growing samples fitted to GR metrics. D, Dose–response curves of the eight fastest-growing samples fitted to GR metrics.

Barplot of estimated ED20 of simulated samples.