Label-Free Longitudinal Imaging of Single Cell Drug Response with a 3D-Printed Cell Culture Platform

Abstract

Image-based phenotypic screening has emerged as a powerful tool for revealing single-cell heterogeneity and dynamic phenotypic responses in preclinical drug discovery. Compared to traditional static end-point assays, live-cell longitudinal imaging captures the temporal trajectories of individual cells, including transient morphological adaptations, motility shifts, and divergent subpopulation behaviors, enabling high content features and more robust early prediction of treatment outcomes. Fluorescence-based screening, while highly specific, is constrained in live-cell contexts by broad spectral overlaps (limiting multiplexing to fewer than six channels), bulky fluorophores that may perturb small-molecule interactions, and photobleaching or phototoxicity under repeated excitation. Stimulated Raman scattering (SRS) microscopy overcomes these barriers by delivering label-free, quantitative chemical contrasts alongside morphological information. Here, we present a low-cost, 3D printed cell culture platform compatible with the stringent optical requirements of SRS microscopy. This set up enables real-time drug delivery and continuous monitoring of biochemical and morphological changes in living cells during 24-hour time-lapse imaging with minimal photodamage. We outline a processing pipeline for longitudinal SRS images to extract chemical and morphological features of single live cells. Using this system, we showcase time-lapse SRS microscopy as a tool to map heterogenous drug-induced single-cell response over time, enabling the identification of varying trajectories within complex cell populations. By parallelizing multi-well perfusion with label-free chemical imaging, our approach offers a pathway toward high-throughput pharmacodynamic assays for the acceleration of phenotypic screening and personalized medicine.

Figure 1.. SRS microscope with live cell imaging compatibility using a 3D-printed perfusion chamber.

A. 3D-rendered representation of the top-view and side-view of device with labelled device height. B. Image of perfusion chamber with addition of tubing and heating components. A ruler is provided for size reference. C. Trace of internal chamber temperature with perfusion against room temperature over 24 hours. A second temperature sensor was placed alongside the thermistor in the same experimental conditions as on-device culture for temperature recording. Measurements were recorded every 10 s immediately after the onset of perfusion. D. Schematic diagram of SRS microscope setup highlighting the imaging and cell culture components. PD: photodiode; BPF: bandpass filter; Cond: microscope condenser; Obj: microscope objective; DCM: dichroic mirror.

Figure 2.. Evaluation of live cell viability in perfusion chamber.

A. Fold-change in cell count over 24 h in a six-well plate (N = 10 FOV), 3D-printed chamber inside a CO₂ incubator (N = 10 FOV), and 3D-printed chamber with perfusion (N = 10 FOV). B. Fold-change in cell count when cells were imaged continuously (N = 8 FOV) versus only at the start and end of the session (N = 10 FOV). C. Representative SRS intensity projections of combined protein and lipid bands in DMSO-treated A549 cells imaged every 20 min. Scale bar: 50 μm. ns: not significant.

Figure 3.. Single-cell feature extraction from 3-band longitudinal SRS images.

A. Averaged intensity protein/lipid SRS images of Osi-treated cells over time. Scale bar: 50 μm. B. A mask is created from segmentation model, where each color is a different cell. C. Single-cell lineages are tracked from the segmentation mask. D. Cellular features are extracted from each frame and connected throughout temporal images using the tracked lineage tree. E. Features such as circularity and mean protein signal intensity of one cell can be observed over time.

Figure 4.. Dynamic cellular responses in drug-treated populations.

A. Representative images of drug-treated cells over 24 hours of imaging. Each image is an intensity projection of lipid and protein SRS bands of one FOV of A549 cells treated with (left to right) 0.04% DMSO, 6 μM Osi, 10 μM Lap, 4 μM Dox, and 50 nM Pac. Scale bar: 50 μm. B. Zoomed in images at 24 hours. Scale bar: 20 μm. C. The number of tracked cells in each drug treatment condition over time. D. The percentage of mitotic events out of all tracked cells in each drug treatment condition represented in six-hour increments.

Figure 5.. Selected features of phenotypic changes in drug-treated populations.

Violin plots of single-cell feature ratios compared against DMSO-treated cells at six-hour time points. A. Aspect ratio changes in Pac-treated cell population. B. Texture angular second moment (ASM) ratio changes in Pac-treated cells. C. Protein kurtosis ratio changes in Dox-treated cells. D. Area ratio changes in Osi-treated cells.

Figure 6.. Vacuolization occurrence in drug treatment conditions.

A. Vacuole segmentation (colored dots) from whole-cell image. B. Number of vacuoles (mean ± IQR) formed per cell in each drug-treated condition. C. Area fraction (mean ± IQR) of vacuoles per cell in each drug-treated condition. Mean vacuole area (mean ± IQR) per cell in each drug-treated condition.

Figure 7.. Drug-induced differences in cell motility.

A-E. Single-cell trajectory plots (μm) where color bar indicates time and each line is one cell track. F. Instantaneous speed (μm/min) (mean ± interquartile range) calculated at different time intervals.