Scientific Papers

Deep learning-assisted high-content screening identifies isoliquiritigenin as an inhibitor of DNA double-strand breaks for preventing doxorubicin-induced cardiotoxicity | Biology Direct


Cell culture and transfection

Rat cardiomyocyte cell line H9c2 (2–1) (H9c2 for short) was purchased from the Shanghai Cell Bank. Cervical cancer cell line HeLa and breast cancer cell line MDA-MB-436 were from ATCC (USA). DMEM complete medium was prepared by mixing DMEM medium (CORNING, New York, USA), fetal bovine serum (10% v/v, CORNING), and 100 × penicillin–streptomycin solution (1% v/v, CORNING). H9c2, HeLa and MDA-MB-436 were cultured in DMEM complete medium in a CO2 incubator at 5% CO2 and 37 °C. All cell lines were used within 10 passages after thawing from stocks.

The gene expression lentiviral vector of EGFP-53BP1 (1220 ~ 1711 aa) was constructed by directly linking EGFP to the gene sequence of human TP53BP1, and the vector map is shown in Fig. 2A. HEK 293 T cells were transfected to package the virus, and lentiviral particles were generated and collected. The construction of the gene expression lentiviral vector and the collection of lentiviral particles were performed by Yunzhou Biological Company (Guangzhou, China). Then, H9c2 was transfected to obtain H9c2 stably expressing EGFP-53BP1 (1220 ~ 1711 aa), which was called H9c2-EGFP-53BP1 and was cultured in a CO2 incubator at 5% CO2 and 37 °C using DMEM complete medium supplemented with 0.5 μg/mL puromycin (YEASEN, Shanghai, China). All cells were maintained mycoplasma-free.

Comet assay

Neutral comet assays were performed to determine the level of DSB. First, a 0.8% normal melting-point agarose (YEASON) layer was prepared on frosted slides. Second, 30 μL of cell suspension (1.5 × 106 cells/mL) was mixed with 80 μL of 1.5% low melting-point agarose (AMRESCO) and the mix was layered onto the precoated slides to form a cell-embedded agarose layer. The lysis was performed at 4 °C for 2 h. The lysed slides were subsequently soaked in TBE buffer for 30 min, and then electrophoresed in TBE for 20 min at 25 V, 15 mA. Finally, the slides were stained with Gel-Green (Beyotime, Shanghai, China), and images of more than 50 cells were acquired with a Leica DMi3000 B fluorescence microscope and analyzed by the OpenComet [45] plugin in ImageJ (National Institutes of Health).

Immunofluorescence assay

H9c2 cells were seeded in 96-well plates with 2000 cells or 3000 cells per well. After certain treatments, cells were fixed with 4% paraformaldehyde solution (Sangon, Shanghai, China) for 20 min, permeabilized with 0.2% Triton X-100 for 10 min, blocked with PBST (PBS + 0.1% Tween-20) containing 1% BSA and 22.52 mg/mL glycine for 30 min, then incubated with primary antibody at 4 °C overnight. Subsequently, cells were incubated with fluorophore-coupled secondary antibody at room temperature for 2 h, and finally with 1 μg/mL of Hoechst 33,342 (Invitrogen, Carlsbad, USA) for 10 min to stain the nuclei. Images were acquired by the ImageXpress Micro Confocal (Molecular Devices, San Jose, USA) high-content imaging system using a 40× objective. Images were analyzed by MetaXpress (Molecular Devices) software using the Cell scoring module to obtain the average γ-H2AX fluorescent intensity and using the Granularity module to obtain the γ-H2AX or 53BP1 foci per cell. The primary antibodies involved were anti-γ-H2AX (Ser139) (1:500, CST (Danvers, USA), #9718) and anti-53BP1 (1:250, Abcam (Waltham, USA), #ab175933).

Western blotting

Cellular proteins were extracted using RIPA lysis buffer (Beyotime) pre-cooled at 4 °C containing 1% PMSF (Beyotime), 1% protease inhibitor cocktail (MCE, New York, USA), and 1% phosphatase inhibitor cocktail (MCE). This was followed by adding 4 × Laemmli Sample Buffer (Bio-Rad, Hercules, USA) and β-mercaptoethanol. After that, incubate for 10 min at 100 °C to denature the proteins.

The gels were prepared according to the instructions of CFAS Any KD PAGE Protein Electrophoresis Gel Preparation Kit (ZHONGHUIHECAI, Xi’an, China). After being separated with a vertical polyacrylamide gel electrophoresis system (Bio-Rad), the proteins were transferred to a 0.45 μm PVDF membrane (Millipore, Bedford, USA). Subsequently, blocking was performed with 5% non-fat milk at room temperature for 1 h. The bands were incubated with the primary antibody at 4 °C overnight and then incubated with horseradish peroxidase-conjugated secondary antibody at room temperature for 1 h. Secondary antibodies were detected using an ECL chemiluminescence kit (Invitrogen) and visualized by an Imaging System (Bio-Rad). The relative protein content was analyzed with Image Lab software (Bio-Rad). Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) is used as the loading control.

The involved primary antibodies were anti-γ-H2AX (Ser139) (1:1000, CST, #9718), anti-BAX (1:1000, Proteintech (Wuhan, China), #50599-2-Ig), anti-Caspase 3 (1:1000, CST, #9662), anti-HO-1 (1:1000, Proteintech, #10701-1-AP), anti-topoisomerase II (1:1000, abcam, #ab109524), and anti-GAPDH (1:1000, Beyotime, #AF0006).

H9c2-EGFP-53BP1 cell sample preparation, image acquisition, and image analysis

Doxorubicin hydrochloride (MCE), reference compound dexrazoxane (Selleck, Houston, USA), reference compound resveratrol (SIGMA-ALDRICH, Saint Louis, China), and isoliquiritigenin (Yuanye, Shanghai, China) which all have a purity of ≥ 96% were purchased and prepared as a stock solution using DMSO (SIGMA-ALDRICH). Prior to use, the stock solution was diluted to the expected concentration with DMEM complete medium. H9c2-EGFP-53BP1 cells were seeded in black 96-well cell culture plates (Greiner (Monroe, USA), #655090), and treated with or without the indicated concentrations of DOX, and with or without the indicated concentrations of compounds for specific durations before fixation.

Cells were fixed with 4% paraformaldehyde solution for 10 min, then washed twice with PBS. Next, nuclei were stained with 1 μg/mL Hoechst 33,342 for 10 min and then washed three times with PBS. Finally, add PBS to each well to avoid dryness.

The images of DAPI channel and FITC channel were acquired through a 40× objective at room temperature. The imaging devices are as follows. In order to observe the fluorescence distribution of H9c2-EGFP-53BP1 cell lines (Fig. 2B) and to select appropriate treatment duration of DOX (Fig. 2C), Leica DMi3000 B manual microscope was used to carefully focus and obtain images manually one by one. For other H9c2-EGFP-53BP1 images, including images corresponding to Fig. 2E, Fig. 3A–B, Fig. 5A, Additional file 3: Fig. S3 and Additional file 4: Fig. S4, were all acquired using the ImageXpress PICO (Molecular Devices) or ImageXpress Micro Confocal (Molecular Devices) high-content imaging systems. It should be pointed out that different imaging devices were used in Fig. 2C and E, which may be the reason why there is a statistically significant difference of foci per cell between the control group and 0.5 µM DOX 1 h group in Fig. 2D, while there is no statistically significant difference between the control group and 0.5 µM DOX 1 h group in Additional file 5: Fig. S5A.

For the threshold-based traditional method, the images were analyzed using the Granularity module in MetaXpress software to obtain the number of foci in the nucleus. For the deep learning-based method, images were analyzed using the FociNet in Python to obtain the percentage of foci-positive cells. The percentage of foci-positive cells = the number of foci-positive cells/(the number of foci-positive cells + the number of foci-negative cells) × 100%.

High-content screening

H9c2-EGFP-53BP1 cells were seeded in black 96-well cell culture plates at 3000 cells/well. A normal control group and a DOX group were set up on each plate. The compounds to be tested were 315 compounds from the self-made compound library of the lab. Twenty-four hours after seeding, the cells were incubated with 0.5 μM DOX and 50 μM compounds for 1 h. Then, the cells were fixed, stained with Hoechst, and placed into the ImageXpress Micro Confocal imaging system to obtain images of the DAPI channel and the FITC channel through a 40× objective.

The high-content screening images were analyzed using both the threshold-based traditional method and the deep learning-based method, FociNet. In the threshold-based method, MetaXpress was used to segment the cells and recognize the foci. The number of foci per cell in each well was obtained, and the normalized number of foci was calculated according to the following formula [46]: Robust Percent of Samples (RPS) = Si ÷ median(Sall) × 100. For this calculation, RPS = value of the compound ÷ median value of all compounds in the same plate × 100. In FociNet analysis, the percentage of foci-positive cells in each well was obtained and then normalized by the same RPS formula. For both image analysis methods, the mean value of the three RPSs obtained from the three experiments was used to rank the compounds, and the 12 compounds with the lowest mean RPS were selected for subsequent experiments.

Measurement of cellular ATP levels

H9c2 cells were seeded in white opaque 96-well cell culture plates. Cellular ATP levels were measured using the CellTiter-Glo kit (Promega, Madison, USA). The CellTiter-Glo solution was prepared according to the instructions and was added to the wells of the cell plate, shaken, and incubated at 37 °C for 10 min. Finally, the chemiluminescence intensity values were measured by a microplate reader (TECAN, Männedorf, Switzerland). The relative ATP level was calculated by dividing the value of each group by the value of the normal control group.

Apoptosis/necrosis assay

H9c2 cells were seeded in black 96-well cell culture plates. Twenty-four hours after seeding, cells were incubated with DOX and ISL for 24 h. Then, apoptosis/necrosis assays were performed using the Annexin V-FITC kit (Beyotime). A mixture was prepared by mixing Annexin V-FITC conjugate and Annexin V-FITC solution at a ratio of 39:1, and the mixture was then supplemented with 10 μg/mL Hoechst 33,342 solution to get the working solution. For staining, the plates were centrifuged at 1000 RCF for 5 min, and then the cells were washed once with PBS. After aspirating the PBS, staining working solution was added to each well in a light-protected environment and incubated at room temperature for 15 min, followed by immediate image acquisition on the ImageXpress Micro Confocal imaging system using a 20× objective, with DAPI channel and FITC channel selected. The percentage of Annexin V-FITC-positive (apoptotic/necrotic) cells was analyzed using the Cell Scoring module in MetaXpress software.

Detection of cellular peroxynitrite

The B545b probe used for the detection of peroxynitrite (ONOO) was a gift from Professor Xin Li [47]. The B545b was prepared as a 5 mM stock solution in DMSO, stored at − 20 °C. Before use, prepare the working solution by diluting the B545b to 5 μM with DMEM medium and adding Hoechst to a final concentration of 10 μg/mL. H9c2 cells were seeded in black 96-well cell culture plates with 3000 cells per well. Twenty-four hours after seeding, cells were incubated with 0.5 μM DOX and specific concentration of ISL for 24 h. For staining, cells were washed three times with PBS and then stained with B545b working solution at 37 °C for 20 min. The ImageXpress Micro Confocal Imaging System chamber was pre-heated to 37 °C, and DAPI-channel and TRITC-channel images were acquired through the 60× objective. The TRITC-channel B545b fluorescent signal was analyzed using the Cell Scoring module in the MetaXpress software.

Image adjustment and display

For EGFP-53BP1-H9c2 images, in order to show the distribution of EGFP-53BP1 protein more clearly, we performed different brightness adjustment for different images. Nevertheless, for the experiments where the fluorescent intensity is correlated with the concentrations of interested substances (i.e., γ-H2AX immunofluorescence, 53BP1 immunofluorescence, apoptotic/necrotic cells labeling with Annexin V-FITC, peroxynitrite labeling with B545b), quantitative analysis was performed on unadjusted 16-bit raw images, and the corresponding representative images have undergone the same brightness adjustment.

Molecular docking

The docking software was the Glide module of Schrödinger 2018 (Schrödinger, LLC, New York, USA). The crystal structure 1ZXM was processed by the Protein Preparation Wizard module of the Schrödinger software. The processing includes completing the missing side chain and loop region structure of the protein, removing all unbonded heteroatoms and water molecules, completing the missing hydrogen atoms, assigning protonated states and partial charges, and finally optimizing the protein structure using the OPLS3e force field until RMSD is optimized to a maximum value of 0.3 Å to reduce atomic space collisions. Next, a box of 20 Å × 20 Å × 20 Å centered on the center of mass of the ligand was identified as the docking area using the Receptor Grid Generation module, and the grid file for docking was generated.

The 3D structure of the small molecule, ISL, was preprocessed with the Ligprep module, and the protonated state was generated at pH = 7.0 ± 2.0. The prepared small molecule was then docked to the pocket using the Glide program in standard scoring mode (standard precision, SP).

MTT assay for cell viability

Thiazolyl Blue Tetrazolium Bromide (MTT, Sigma-Aldrich) solution was prepared as 5 mg/mL in PBS, and the solution was filtered with a 0.22 μm filter for sterilization and diluted to 0.5 mg/mL with DMEM medium right before use. After three times washing with DMEM medium, cells were incubated with 0.5 mg/mL MTT in the CO2 incubator for 4 h. Then, replace the MTT solution in the wells with 100 μL DMSO, shake the plate for 10 min at 37 °C, and measure the absorbance at 570 nm in a microplate reader (TECAN). The absorbance of the treated group was divided by the absorbance of the control group to calculate the relative cell viability.

Animal experiments

Male C57BL/6 mice aged 6 ~ 8 weeks and weighing 22 ~ 26 g were purchased from Shanghai SILAIKE Laboratory Animal Company. The animal experiments were carried out following the guidance of the Care and Use of Laboratory Animals published by the US National Institutes of Health (NIH Publication No.85–23, revised 1996) and the protocols were approved by the Laboratory Animal Ethics Committee of Zhejiang University (Permit number: ZJU20220096).

The mice were randomly grouped into the normal control group, 15 mg/kg DOX group, DOX + DXZ group, DOX + 10 mg/kg/d ISL group, and DOX + 30 mg/kg/d ISL group. Except the normal control group, all mice were injected with 15 mg/kg DOX intraperitoneally and the day of injection was recorded as Day 1. DXZ was administered intraperitoneally at 60 mg/kg on Day 1 and Day 5. ISL was given via an intragastric administration once daily for 8 days (10 mg/kg/d or 30 mg/kg/d). The doses of ISL, DOX and DXZ were selected according to previously reported studies, respectively [48,49,50].

Transmission electron microscopy of myocardial tissue

Myocardial tissue samples of about 1 mm × 1 mm × 4 mm were cut out and immersed in Gluta fix solution (Solabio, Beijing, China), fixed overnight at room temperature, and transferred to 4 °C. The reagents used in the sample preparation were provided by the Electron Microscopy Center of Zhejiang University.

The tissues were rinsed three times with 0.1 M PBS for 10 min each. They were fixed with 100 μL of 1% osmium acid for 1 h, then rinsed 3 times with water for 10 min each. After that, fix/stain them with 100 μL 2% uranyl acetate for 30 min. Then, the samples were dehydrated by soaking in a gradient of 50% ethanol, 70% ethanol, 90% ethanol, and 100% ethanol each for 15 min, and then soaked in 100% acetone twice for 15 min each time. The samples were then permeabilized overnight at room temperature with a mixture of 3:1 embedding agent:100% acetone. Finally, embedding, sectioning, and image acquisition were performed by the Electron Microscopy Center of Zhejiang University, where images of myofilaments and mitochondria of myocardial tissue were acquired using 120 kV cryo-transmission electron microscopy.

Echocardiographic evaluation of cardiac function

M-mode echocardiograms were recorded on Day 9. The measurements were performed by experienced operators who were blinded to the study group assignment. Echocardiographic images of the left ventricle of the mice were collected in the anterior thoracic region after the removal of hair. The data were analyzed and calculated in Vevo2100 (Visualsonics, Toronto, Canada) to obtain the ejection fractions (EF) and fractional shortening (FS) of the left ventricle.

Statistical analysis

GraphPad Prism (San Diego, USA) was used for statistical analysis of the experimental data. Kruskal–Wallis test was performed for Fig. 1B and Fig. 5C, and Mann–Whitney test was performed for Fig. 4B (c) and Additional file 5: Fig. S5C (middle). For other data, if not specifically stated, two-tailed unpaired t test was employed to make comparisons between two groups, and One-Way ANOVA with Dunnett’s multiple comparisons test was performed for comparisons between multiple groups.



Source link