Page 173 - IJB-8-2

Tröndle Kevin, et al.
by providing a high degree of automation and spatial 2.2. DoD bioprinting
resolution. Various bioprinting techniques were developed For bioprinting, we used a piezo-actuated DoD
and applied to fabricate 3D cell models with defined size bioprinting technology (PipeJet , Biofluidix GmbH).
®
and shape by spatially controlling the cell distribution in an A detailed description of the printing process can be
artificial ECM . In this study, we used a drop-on-demand found in our previous study . In brief, single droplets
[6]
[6]
(DoD) bioprinting technology which was previously (10 nl) each containing about 150 cells were deposited
applied in the production, handling, and treatment of cell onto a Matrigel substrate layer. The printed cell clusters
spheroids , attributed to its capability to precisely deposit were encapsulated with a second layer of Matrigel and
[7]
low volumes of low viscous bioinks, such as spheroids incubated for subsequent incubation supporting the
or cells in suspensions. Compared to classical tissue cellular self-assembly of one spheroid per cluster.
engineering approaches, bioprinting provides increased
sample reproducibility, which is one key requirement for 2.3. LDH toxicity assay
systematic screening applications. In previous studies, we
have described a scalable concept of DoD bioprinting and We used a colorimetric LDH Assay Kit (ab65393, Abcam)
controlled cellular self-assembly to fabricate size-defined to quantify the cellular release of LDH enzyme caused
renal spheroids and tubules in a hydrogel scaffold . These by a treatment with different concentrations of cisplatin
[8]
structures comprise of a hollow lumen surrounded by an (ab141398, Abcam). The cell death rate Ψ was determined
organized epithelial cell layer, thereby closely mimicking the relative to a lysed control (Lysed Ctrl) obtained by
nephron tubule structure. Here, we applied this concept to sample treatments with Triton X (30 min, 37°C). The
fabricate 3D renal spheroids from iRECs for a head-to-head treatment was conducted by incubating the cells for 24 h
comparison with 2D monolayers of the same cell type. The and changing medium subsequently. To determine Ψ,
sensitivity to the nephron-toxicant cisplatin was investigated the supernatant of treated samples was collected (10 µl)
with different readout methods. First, we determined the and compared to an un-lysed control (Neg. Ctrl). The
cell death rate Ψ using a lactate dehydrogenase (LDH) measured absorbance was determined by measurements
quantification assay. Next, we fluorescently labeled a of the optical density at 450 nm wavelength (OD ).
450
nephrotoxicity biomarker for microscopic imaging. In the A normalized solution of LDH enzyme (0.25 µg/µl, LDH
context of bioprinting, machine learning image processing Ctrl) was used to assess the assay performance.
could prospectively contribute to improve 3D cell model 2.4. Fluorescent image acquisition
generation, fabrication, and readouts [9,10] . In the latter,
microscope images of cell morphology could be used to The treated samples were imaged using a fluorescence
assign biochemical values (e.g., viability) in an automated microscope (Axio Observer.Z1/7, Carl Zeiss) with a
manner without requiring additional assays to be conducted 20-fold magnification objective (EC Plan-Neofluar
for each experimental setting. This again addresses 20x/0.5 M27), LED excitation, and fluorescently labeled
important aspects of prospective screening applications, biomarkers for nephrotoxicity. The obtained images
such as automation and high throughput. Here, we present were correlated with the treatment dose and Ψ, which
a feasibility study for an automated toxicity readout using was chemically determined as the relative cytotoxicity
deep learning image classification based on bioprinted renal with the released LDH amount. As primary biomarker
spheroids [11,12] . We trained a convolutional neural network of cytotoxicity, the integrity of the cell membranes was
(CNN) through supervised learning to predict the Ψ of a observed, which were labeled in the cell line (iRECs)
spheroid from its microscopic image. with a stable expression of membrane-localized green
fluorescent protein (GFP). The kidney injury molecule
2. Materials and methods 1 (KIM-1) was labeled as a specifically expressed
biomarker of nephrotoxic effects . For this, the protein
[13]
2.1. Cell culture and hydrogels was fluorescently labeled post-treatment with a primary
For all sample preparations, we used iRECs . A detailed antibody (Invitrogen, #PA5-79345), and a secondary
[4]
description of the cell type and culture conditions can antibody (Abcam, #ab6939) for the microscopic detection
be found in previous studies . The cells were cultured within spheroids. For cultivation and microscopy,
[4]
in Dulbecco’s Modified Eagle Medium (DMEM, the spheroid arrays were fabricated in an 8-chamber
#41966029, Gibco), with additives of fetal bovine microscopy slide (µ-Slide 8 well; ibidi GmbH, #80826).
serum (10%) and penicillin/streptomycin (1%), for cell 2.5. Deep learning
expansion. Matrigel (100%, #356231, Corning) was used
as artificial ECM material. The 3D spheroid models were The code was written in Python 3.8.10 using Pytorch
cultured in renal epithelial growth medium (REGM, 1.8.1. Detailed information is listed in the Supplementary
#CC-3190, Lonza), without addition of additives. File. The dataset was made of 4974 spheroid images taken

International Journal of Bioprinting (2022)–Volume 8, Issue 2 165