Deep Learning-Assisted Nephrotoxicity Testing with Bioprinted Renal Spheroids.

Area of Science:

• Biotechnology
• Toxicology
• Renal cell biology

Background:

• Bioprinting enables the creation of 3D cell culture models that better mimic in vivo tissue architecture.
• Renal epithelial cells are crucial for kidney function and are susceptible to drug-induced toxicity.
• Assessing drug toxicity in vitro often relies on 2D cell cultures, which may not accurately reflect in vivo responses.

Purpose of the Study:

• To evaluate the utility of bioprinted renal epithelial cell spheroids for cisplatin toxicity testing.
• To compare the sensitivity of spheroids versus monolayers to cisplatin.
• To develop a deep learning-based method for automated toxicity assessment using spheroid images.

Main Methods:

• Arrays of bioprinted renal epithelial cell spheroids were used for toxicity testing.
• Lactate dehydrogenase (LDH) assay was employed to determine concentration-dependent cell death rates.
• Fluorescent labeling was used to visualize kidney injury molecule 1 (KIM-1) accumulation.
• A deep learning algorithm was trained on microscopic images of spheroids with known cell death rates.

Main Results:

• Bioprinted spheroids exhibited enhanced sensitivity to cisplatin compared to monolayers, with a lower inhibitory concentration (IC50 = 9 ± 3 μM vs. 17 ± 2 μM).
• Fluorescent labeling revealed accumulation of the nephrotoxicity biomarker KIM-1 within the central lumen of the spheroids.
• The deep learning algorithm achieved a balanced accuracy of 78.7% in distinguishing between no, mild, and severe treatment effects.

Conclusions:

• Bioprinted renal spheroids represent a more sensitive in vitro model for evaluating nephrotoxicity.
• The accumulation of KIM-1 in spheroids suggests their potential for mechanistic toxicity studies.
• Deep learning offers a promising approach for automated, high-throughput toxicity screening using 3D cell models.