Methods for mitochondrial health assessment by High Content Imaging System

Mitochondria are important organelles responsible for energy production. Mitochondrial dysfunction relates to various pathological diseases. The investigation of mitochondrial heath is critical to evaluate the cellular status. Herein, we demonstrated an approach for determining the status of mitochondrial health by observing mitochondrial H2O2 (one type of ROS), membrane potential, and morphology (fragmentation and length) in live primary fibroblast cells. The cells were co-stained with fluorescent dyes (Hoechst 33342 and MITO-ID® Red/MitoPY1/JC-10) and continuously processed by the High Content Imaging System. We employed the Operetta CLSTM to take fluorescent images with its given quickness and high resolution. The CellProfiler image analysis software was further used to identify cell and mitochondrial phenotypes in the thousand fluorescent images.• We could quantitatively analyze fluorescent images with high-throughput and high-speed detection to track the alteration of mitochondrial status.• The MMP assay is sensitive to FCCP even at the concentration of 0.01 µM.• The fibroblast cells treated with stress inducers (H2O2, FCCP, and phenanthroline) revealed a significant change in mitochondrial health parameters, with more ROS accumulation, depolarized MMP, increased fragmentation, and reduced length of mitochondria.


Introduction
Mitochondria are powerful organelles involved in various vital cellular functions. The key role is to maintain cellular homeostasis and energy status by producing ATP. ATP is generated via oxidative phosphorylation through the electron transport chain (ETC) in the inner membrane of mitochondria. The transferring of electrons through respiratory Complexes I, III, and IV creates an electrochemical gradient, a combination of mitochondrial membrane potential (MMP) and pH gradient. This electrochemical gradient provides free energy to drive ATP synthesis. During electron transports, reactive oxygen species (ROS) can occur from the leakage of electrons. The accumulation of ROS results in oxidative stress and mitochondrial damage. Therefore, the disruption of MMP and ROS accumulation further impairs the mitochondrial functions and often prevails cellular damage and apoptotic cell death pathway, eventually contributing to several diseases and aging [ 1 , 2 ].
However, mitochondria have a vital defensive mechanism against mitochondrial damage by destroying impaired mitochondria and keeping a balance of healthy mitochondria. This dynamically balanced check is manifested by changes in mitochondrial morphology. The elongation of mitochondria, tubular networks, is associated with mitochondrial health due to the fusion of mitochondria, which allows the exchange of healthy to damaged mitochondrial components to maintain the overall state in balance. On the other hand, mitochondrial fragmentation, the small pieces from the fission process, is related to an increase in ROS production and ATP depletion [3] . Excessive mitochondrial fission has been found to co-exist in cardiomyopathy [4] , cancer, and neurodegenerative diseases such as Alzheimer's disease.
Therefore, mitochondrial health is a key for maintaining cellular stability and pertained to health or disease. In this study, we explored the throughput method for monitoring mitochondrial status, including mitochondrial ROS, membrane potential status, and morphology by using the High-Content Imaging System. We employed the Operetta CLS TM (PerkinElmer), a high-content fluorescent imaging technology with a high-throughput detection system that automates the capture of cell images from 96-well microplate. We selected a water immersion objective lens, which gives substantially brighter images compared to air objectives. CellProfiler image analysis, a free public software, was used to identify cell components, enabling us to generate and automatically analyze thousands of images through a pipeline for statistical analysis [5][6][7] .

Cell culture
Primary fibroblast cells were established from skin biopsies of three volunteers (F1, F2, and F3). The Human Research Protection Unit approved this study with the certificate of approval No. Si 161/2019 (the Faculty of Medicine Siriraj Hospital, Mahidol University). The cells were grown in a T25 cell culture flask (Corning®) containing Dulbecco's modified Eagle's medium (DMEM, 5 mM glucose, Gibco®) supplemented with 10% fetal bovine serum (FBS), 1% penicillin (100 U/mL)/streptomycin (100 μg/mL), and 0.1% amphotericin B (1 mg/mL). Cells were incubated at 37 °C with a humidified 5% CO 2 atmosphere, and the culture medium was changed every other day.

Treatment of stress inducers
Fibroblast cells at 80-90% confluency of a 25 mL flask were seeded on the 96-well plate (CellCarrier-96, PerkinElmer) with a density of 4,0 0 0-5,0 0 0 cells/well for 24 h. The next day, cells were treated with stress inducers including hydrogen peroxide (H 2 O 2 ), carbonyl cyanide-ptrifluoromethoxyphenylhydrazone (FCCP), or phenanthroline with the final concentrations as indicated in Table 1 in the culture media for the measurement of mitochondrial ROS, MMP, and fragmentation and length, respectively.

Cell staining
The stock solutions of fluorescent dyes were generated as shown in Table 2 . Whereas MitoPY1 stock was prepared by dissolving in methanol, aliquoted and vacuum-evaporating the solvent [8] . In each experiment, MitoPY1 was freshly dissolved in DMSO. After treatment with stress inducers, the cells were washed twice with PBS. The final concentrations of mixture dyes prepared in the phenolfree DMEM medium supplemented with 10% FBS were added into the 96-well plate at 50 μL/well to stain the cells for 30-45 min in the dark. The cells were washed with PBS twice and replaced with phenol-free DMEM medium supplemented with 2% FBS to reduce the background noise before fluorescent imaging. High-content analysis system setting The images of stained cells were visualized using the Operetta CLS TM High-Content Analysis System (PerkinElmer) in the confocal or non-confocal with two peaks of autofocus mode and a water objective lens, 40x high numerical aperture (NA), which had a high refractive index. With its high resolution and consistent illumination from the 8x LED light source, the Operetta CLS could rapidly track changes in mitochondrial status to reduce photodamage at a rate of around one field per 2-4 seconds, depending on the percentage of power and the exposure time. We used a non-confocal mode for MMP measurement due to the JC-10 dye being particularly sensitive to the light source resulting in fluorescent signals being easily dropped. As a result, this mode is better suited for signal conservation. The other experiments, on the other hand, employ confocal mode. We captured images of thousands of stained cells on a 96-well plate (15-40 cells/field, 60-80 fields/well, and 2 wells/conditions) under 37 °C and 5% CO 2 control. The binning mode was set to 2. The images' final resolution was 0.299 μm pixel size, 16 bit per pixel, and 1080 × 1080. The Operetta CLS setting via the Harmony® 4.9 software was provided in Table 3 .

Image analysis procedure by the CellProfiler program
CellProfiler is a high-throughput cell image analysis tool that user can be used to build a pipeline to analyze the organelles of interest in cells [5][6][7] . We created the pipeline to identify and measure the number of nuclei, H 2 O 2 in mitochondria, J-aggregate and J-monomers, and granularity and length of mitochondria from Hoechst 33342, MitoPY1, JC-10, and MITO-ID® Red fluorescent channels. The fluorescent images captured by the Operetta CLS were run through the pipeline in CellProfiler versions 3.1.9 or 4.0.7.

Metadata extraction (channel matching)
The fluorescent images from the Operetta CLS uploaded through the Columbus program were exported and extracted the metadata from the file name for specifying the image information and matching with the particular fluorescent channels using the regular expression; ).tif. The original images from each channel were continuously selected as grayscale type i.e., each pixel representing a single intensity value.

Illumination correction
One of the problems affecting precise fluorescence intensity measurements was that background illumination from the microscopy had an irregular pattern with bright or dark areas across the image. For reliable image analysis outcomes, we utilized the Correction Illumination Calculate and Apply modules: Fit Polynomial strategy to calculate the typical variance of illumination intensity and the Subtract strategy to apply the smoothed lightening pattern in each image ( Fig. 1 A) [ 9 , 10 ].

Organelle segmentations Identify the primary object (nuclei)
The primary object used to identify a cell's location was the stained nucleus with Hoechst 33342 dye. The threshold of fluorescent intensity of the Hoechst 33342 channel was determined by the Identified Primary Object modules: Global and Otsu thresholding methods defining the threshold into the foreground and background pixels by reducing the variation within each group. We further used the Three-classes thresholding strategy to classify fluorescent intensity above the threshold as foreground and below as background ( Fig. 1 B) [11] .

Identify the secondary object (whole cell)
The secondary object, which was related to the primary object (the nuclei), was employed to determine a cell's boundary. We adopted the MITO-ID® Red channel to recognize the secondary object ( Fig. 1 C) for the mitochondrial H 2 O 2 , fragmentation, and length measurements. For the MMP measurement, the Alexa 488 channel was employed to determine J-monomer in the cytoplasm ( Fig. 3 A). The fluorescent intensity thresholds of the MITO-ID® Red and Alexa 488 channels were estimated by Global, Otsu, and classified intensity using the Three-classes Thresholding strategy. Similarly, the Propagation method was appointed as the determination of the dividing lines between the cells in contact.

Identify the tertiary object (cytoplasm)
The cytoplasmic area was defined by the calculated areas of the secondary objects (whole-cell area) minus the primary objects (nuclear area) using the Identified Tertiary Object modules, which could subtract the smaller object from the larger one ( Figs. 1 C, 2 A, and 3 A).

Filter objects
The Filter Object module was a method for excluding objects that had unsatisfied characteristics, for instance, clumping cells, unhealthy or degenerating cells, and other artifacts. These objects have a high intensity, a small nucleus, and an irregular shape, all of which can affect the actual results, such as representing a higher intensity. As a result, the cutoff value was adjusted to restrict the object's size, shape, eccentricity, and intensity by the Filtering measurement mode. Furthermore, we used the Image or Mask Border mode to exclude objects that are touching the image's edge because these cells do not complete the entire cell.

Quantification of mitochondrial H 2 O 2
Mitochondrial hydrogen peroxide (H 2 O 2 ), one of the ROS, is the parameter used to evaluate mitochondrial functions. ROS are generated when an oxygen-receiving electron escapes and becomes a superoxide anion ( •O 2 − ), an initial form of ROS, during the OXPHOS process. The •O 2 − can be converted to H 2 O 2 , and H 2 O 2 can be formed into a hydroxyl radical ( •OH), which is very reactive and can cause substantial damage.
Although H 2 O 2 produced by mitochondria plays an important role in human health and disease, it is difficult to monitor selectively inside living cells. Mitochondrial peroxy yellow 1 (MitoPY1) is a boronate-based compound designed to selectively and efficiently probe H 2 O 2 coupled with a phosphonium moiety to target mitochondria of living cells [12] . MitoPY1 structure can alter to become a bright fluorescence when exposed to H 2 O 2 [ 8 , 13 ]. It is more sensitive to H 2 O 2 than the commonly used MitoSOX Red. MitoSOX Red is a dihydroethidium derivative with a net positive charge that is used to detect mitochondrial •O 2 − . The high concentration of MitoSOX Red being used could alter mitochondrial morphology and enables it to pass through the nucleus rather than mitochondria [14] . MitoPY1 has been used to detect H 2 O 2 in cell culture and tissues in a few recent articles [15][16][17] . The methods for measuring H 2 O 2 in fibroblast culture using the MitoPY1 are described here.
To evaluate mitochondrial H 2 O 2 , the Hoechst 33342 and MITO-ID® Red channels were designated into the pipeline of organelle segmentation to reflect the nuclei and cell boundaries ( Fig. 2 A). The green-fluorescent intensity from the MitoPY1 channel was reinforced by using the Enhance or Suppress Feature modules; Speckle Feature Enhancement method, which heightens a region of greater intensity in comparison to its surroundings (supplementary data; Fig. S4). The green-speckled Each plate was treated as an individual experiment, n = 1 per condition. The graphs were presented as mean values with 95% CI. Statistical significance is analyzed using ANOVA with Tukey's multiple comparison post-test and the student's independent sample t -test. Statistical significance was considered when * P < 0.05, * * P < 0.01 and * * * P < 0.001.

Table 4
The process for determining mitochondrial H 2 O 2 levels by CellProfiler program.  ( Fig. 2 C). Table 4 and supplementary data contain a table and schematics that simplify and highlight each phase of the pipeline. (Fig. S1).

Quantification of mitochondrial membrane potential
During electron transferring activity in mitochondria, proton pumps provide a proton gradient across the inner membrane [18] . With a few exceptions [19] , the majority of the energy of the proton gradient must be described as membrane potential. Therefore, the mitochondrial membrane potential (MMP) is a key parameter for monitoring mitochondrial functional status.   We selected JC-10, a JC-1 derivative with enhanced solubility to detect the MMP. It is a cationic, lipophilic dye that is concentrated and exists in two forms: J-monomeric and J-aggregated, depending on the potential of the mitochondrial membrane. In healthy mitochondria, the protons are efficiently pumped into intermembrane space resulting in JC-10 favorably diffusing into the mitochondria due to the negative potential of inner membranes and matrix and accumulating in the reversible form of orange-fluorescent signal (J-aggregates), which indicate polarized MMP [20] . Whereas the unhealthy mitochondria (a cell is injured) with depolarized membrane potential, JC-10 exists in the cytosol and shifts to a green-fluorescent monomer [ 21 , 22 ]. Therefore, changes in MMP can be detected quantitatively using the fluorescence intensity ratio of orange-fluorescent aggregates to green-fluorescent monomers. To assess MMP status, J-monomers (green fluorescence) in the cytoplasm from JC-10 Alexa Fluor 488 channels were related to the nuclei and segmented by using the Identify Primary, Secondary, and Tertiary Object modules ( Fig. 3 A). To sort out the polarized mitochondria, we specified the J-aggregate by determining the threshold of orange fluorescence intensity in the JC-10 Rhodamine channel with the adjustment of object diameter between 3 to 15 pixel units using the Identified Primary Object Each treatment condition was represented by 60 -80 fields per well (15 -40 cells per field), leading to 3,600 -12,800 cells (for four wells) and 1,80 0 -6,40 0 cells (for duplicate wells) overall for each condition. Each plate was treated as an individual experiment, n = 1 per condition. The graphs were presented as mean values with 95% CI. Statistical significance is analyzed using ANOVA with Tukey's multiple comparison post-test. Statistical significance was considered when * P < 0.05, * * P < 0.01 and * * * P < 0.001.

Table 5
The process for determining MMP levels by the CellProfiler program. Identify Tertiary Object: Identifying the boundary of J-monomer 7 Identify Primary Object: J-aggregates identification (Rhodamine channel) 8 Mask Objects: Identifying the J-aggregate area masked by the J-monomer area 9 Mask Objects: Identifying the J-monomer area masked by the J-aggregate area 10 Measure Object Intensity: Measuring the orange intensity in the J-aggregate area 11 Measure Object Intensity: Measuring the green intensity in the J-monomer area 12 Calculate Math: Calculating the ratio of orange intensity to green intensity 13 Export to Spread Sheet: Exporting as an Excel file modules: the Adaptive and Otsu methods and classifying by the Three-classes thresholding strategy ( Fig. 3 B). The sizes of these pixels matched the J-aggregate in the mitochondria in the raw images. The J-aggregate looked to be a little point of 3 to 15 pixels in size. The J-aggregate appeared to be a small point in the range of 3 to 15 pixels in size; otherwise, the J-aggregate might not have been detected. Subsequently, we applied the Mask object module: Keep overlapping region and Retain strategies to improve the specificity of the J-aggregate and J-monomer regions. The intensity of the J-aggregate (orange-fluorescent region masked by green region, Fig. 3 C middle) and J-monomer (green-fluorescent region masked by orange region, Fig. 3 C right) was further gauged by the Measure Object Intensity module. Lastly, the Calculate Math module was applied to quantify the fluorescence intensity ratio of J-aggregate to J-monomer. The table and schematic to simplify and showcase each step of the pipeline was shown in the Table 5 and supplementary data (Fig. S2).

Quantification of mitochondrial fragmentation and length
Mitochondria are dynamic organelles with the ability to change the structure by fission or fusion to maintain their function and health [23] . In this study, we used MITO-ID® Red fluorescence labeling to observe the alteration in mitochondrial morphology in any energetic state. It is a mitochondria-specific staining dye that is harmless to live cells and produces consistent fluorescence signals. According   , leading to 1,800 -6,400 cells (for duplicate wells) overall for each condition. Each plate was treated as an individual experiment, n = 1 per condition. The graphs were presented as mean values with 95% CI. Statistical significance is analyzed using the student's independent sample t -test. Statistical significance was considered when * P < 0.05, * * P < 0.01 and * * * P < 0.001.

Table 6
The process for determining mitochondrial fragmentation and length levels by CellProfiler program.
The pipeline of mitochondrial fragmentation and length quantification To quantify mitochondrial fragmentation and length, the nuclei and cell boundaries (whole-cell area) were identified from the Hoechst 33342 and MITO-ID® Red channels using the organelle segmentation pipeline. The cytoplasmic area was further determined by subtracting the whole-cell area from the nuclear area ( Fig. 1 ). To analyze the mitochondria, the fluorescent intensity of MITO-ID® Red channels was augmented using Enhance or Suppress Feature modules; Neurites type and Tubeness Enhancement method ( Fig. 4 A). The enhanced red fluorescence was further labeled using the Threshold module; Global and Minimum cross-entropy thresholding strategy to compute a single threshold value and classify pixels above the threshold as foreground (supplementary data; Fig. S5)    [11] . Thereafter, the mitochondria were spotted by the Identify Primary Object module; the Global and Manual thresholding methods with a selective object diameter between 3 to 100 pixel units ( Fig. 4 B). The diameter of mitochondria was discovered to be around 0.2 μm [24] or 0.5 μm [25] , and the length of mitochondria was determined to be around 26 μm [25] or 46 μm [26] . According to our observations, the mitochondrial diameter was less than or equivalent to 3 pixels. As a result, we assumed that 3 pixels were approximately 0.3 μm, which corresponded to the previous study's minimal mitochondrial diameter. Consequently, we set the object diameter range from 3 to 100 pixels for the identification of mitochondria.
We employed the Convert Objects to Image module with the Binary (white and black) color format to assign the mitochondria as a white color and the background as a black color. For the analysis of mitochondrial fragments, the mitochondrial granularity was accordingly assessed by the Measure Object Granularity module from white and black color format of mitochondrial images that were associated to the cytoplasm. This module could calculate the percentages of fragmented mitochondria by evaluating the divided portions of the mitochondria and designating the number of those portions as a granular spectrum with varying pixel sizes. The high levels of granularity indicated to the highly fragmented mitochondria from the fission process.
For the mitochondrial length analysis, the Morph module was conducted to examine mitochondrial morphology by shrinking the mitochondria to a single line (skeleton) ( Fig. 4 C left) using the Skelpe method and denoting the branch end of the mitochondrial skeleton by the Endpoint method ( Fig. 4 C  right). The mitochondrial skeleton endpoints were then dwindled by the Expand or Shrink Object module; Shrink objects to a point ( Fig. 4 D left). Eventually, the Measure Object Skeleton was utilized for measuring the distance between shrunken endpoints within the mitochondrial skeleton in the pixel unit ( Fig. 4 D right). The table and schematic to simplify and showcase each step of the pipeline were shown in the Table 6 and supplementary data (Fig. S3).

Statistical Analysis
We used one-way ANOVA to compare the differences of more than two groups. Next, Tukey's multiple comparisons post-hoc testing was used to analyze the differences between the mean of all possible pairs. Student's independent samples t -test was used to analyze the differences between the two groups by Jamovi version 1.2.22 software. Data were presented as the mean ± 95% CI (confidence interval). Statistical significances were considered when * P < 0.05, * * P < 0.01, and * * * P < 0.001.

Evaluation of mitochondrial hydrogen peroxide
To evaluate the method of mitochondrial hydrogen peroxide (H 2 O 2 ) measurement, fibroblast cells were challenged with H 2 O 2 at 10 0, 20 0, and 40 0 μM in culture media for 1 h compared with the untreated cells (0 μM of H 2 O 2 ). Following the treatment, the cells were co-stained with MitoPY1, MITO-ID® Red, and Hoechst 33342 dyes. The live-cell images from the Operetta CLS were processed through the pipeline for estimating mitochondrial H 2 O 2 levels by the CellProfiler program. The fluorescence images exhibited nuclei from the Hoechst 33342 channel along with the mitochondrial network from the MITO-ID® Red channel ( Figs. 5 A , 6 A , and 7 A). The areas of greenspeckled fluorescence from MitoPY1 ( Figs. 6 B and 7 B) increased in the H 2 O 2 -treated cells compared with untreated cells ( Fig. 5 B). Similarly, the overlay channels revealed that H 2 O 2 levels relating to mitochondria were elevated in the H 2 O 2 -treated cells ( Figs. 6 C and 7 C) compared to untreated cells ( Fig. 5 C).
The colocalized areas disclosed the increasing levels of mitochondrial H 2 O 2 in a dose-dependent manner of H 2 O 2 treatment. The mitochondrial H 2 O 2 levels significantly increased in the cells treated with H 2 O 2 at 10 0, 20 0, and 40 0 μM compared with untreated cells ( Fig. 8 A). Moreover, in distinct fibroblast samples (F1, F2, and F3), the mitochondrial H 2 O 2 levels were considerably elevated in cells treated with H 2 O 2 at 200 μM compared to untreated cells ( Fig. 8 B). However, the number of greenspeckled fluorescent areas in images ( Fig. 5 , 6 , and 7 ) appeared to be greater than the quantitative results of mitochondrial H 2 O 2 levels ( Fig. 8 ). The reason for this was that the Relate Object module might select just the high intensity of green-speckled fluorescence for relating to the nucleus, where mitochondrial H 2 O 2 largely accumulated around the nucleus (perinuclear) while peripheral areas were mostly low-intensity.
To assess the quality control, we displayed the distribution of mitochondrial H 2 O 2 levels in the triplicate wells for the titration of H 2 O 2 concentrations ( Fig. 9 A) and the duplicate wells for the comparison of 200 μM of H 2 O 2 treated cells and untreated cells in individual fibroblast samples ( Fig. 9 B). In all conditions, the well-related distribution graph of mitochondrial H 2 O 2 levels displayed a left-skewed distribution. There was a minor difference between the wells (1, 2, and 3) ( Fig. 9 A and B). For the distribution of H 2 O 2 levels related to the field of observation, for example, in the conditions of F2_H 2 O 2 0 μM_Well 1 ( Fig. 9 C), F2_H 2 O 2 0 μM_Well 2 ( Fig. 9 D), F2_H 2 O 2 200 μM_Well 1 ( Fig. 9 E), and F2_H 2 O 2 200 μM_Well 2 ( Fig. 9 F), the data revealed a minor difference between the fields. We next utilized the R tool to find the outlier of the data points in each well, and we noticed that each condition in Fig. 9 C -F had only one outlier.

Evaluation of mitochondrial membrane potential levels
FCCP (Carbonyl cyanide-p -trifluoromethoxyphenylhydrazone) is widely used in mitochondrial research as an uncoupling reagent. FCCP disrupts the proton gradient between the matrix and intermembrane space of mitochondria, resulting in a decrease in MMP [27] . In this study, we treated the cells with 0.01, 0.1 μM of FCCP, or ethanol (vehicle control) in the culture media for 1 h. The cells were then stained with Hoechst 33342 and JC-10 dyes after treatment. The CellProfiler program was used to run live-cell images from the Operetta CLS through a pipeline for determining MMP levels. The nuclei from the Hoechst 33342 channel coexisted with the green-fluorescence (J-monomers) from the JC-10 Alexa Fluor 488 channel in the fluorescence images ( Figs. 10 A, 11 A, and 12 A). The fluorescent images showed that vehicle control cells (0 μM of FCCP) ( Figs. 10 B) exhibited more orangefluorescence (J-aggregates) representing polarized MMP than the cells treated with FCCP at 0.01 and 0.1 μM, as observed from the JC-10 Rhodamine channels ( Figs. 11 B and 12 B). As well, the merged channels revealed that areas of J-aggregates masked by J-monomers ( Fig. 10 C middle image) increased more than in the FCCP-treated cells ( Figs. 11 C and 12 C middle image). The intensity ratios of orangefluorescent aggregates to green-fluorescent monomers, which indicated the MMP levels, significantly decreased in cells treated with FCCP with a dose-dependence compared to vehicle control cells in distinct fibroblast samples (F1 and F3) ( Fig. 13 ). Our findings indicated a more sensitive detection of MMP changes in response to 0.01 and 0.1 μM of FCCP treatment, compared to the previous study that used TMRM fluorescence in rat ventricular myocytes [28] .
We displayed the well-related distribution of MMP levels in the duplicate wells of the 0.01 μM and 0.1 μM of FCCP treated cells and the untreated cells in individual fibroblast samples ( Fig. 14 A) for validation of MMP measurement. We discovered the anomalous data in well 2 of F2 samples in the conditions of F2_FCCP 0 μM_Well 2, F2_FCCP 0.01 μM_Well 2, and F2_FCCP 0.1 μM_Well 2 (the red cross mark in Fig. 14 A). When compared to well 1 and the other samples (F1 and F3) in the identical settings, the distribution of MMP levels in well 2 of the F2 samples was in the lower range. Therefore, the results from F2 samples in measuring of MMP levels were omitted. For the observation of the distribution of MMP levels associated with the field, the data showed minor variation between the fields, such as in the conditions of F1_FCCP 0 μM Well 1 ( Fig. 14 B), F1_FCCP 0.01 μM Well 1 ( Fig. 14 C), and F1_FCCP 0.1 μM Well 1 ( Fig. 14 D). We further utilized the R program to find the outlier in each well's data, and we noticed each condition ( Fig. 14 B -D) had only one outlier.

Evaluation of mitochondrial fragmentation and length
To evaluate the method for assessing mitochondrial morphology, we selected 1,10-phenanthroline, which has been shown to induce mitochondrial fragmentation in HeLa cells. Reportedly, phenanthroline stimulates DRP1, a protein that operates mitochondrial fission [29] . In this study, the fibroblast cells were treated with 50 μM of phenanthroline or ethanol (vehicle control) in culture media for 4 h. After drug treatment, cells were stained with MITO-ID® Red and Hoechst-33342. The CellProfiler program was used to determine the levels of mitochondrial fragmentation and length. The fluorescent images from MITO-ID® Red channel were converted to grayscale and the red fluorescent intensity was increased ( Figs. 15 A and 16 A). The mitochondria were then identified from the threshold labeling ( Figs. 15 B and 16 B). The mitochondrial networks were reduced to a single line (skeleton) and then shrank into the endpoints ( Figs. 15 C and 16 C). The distance between shrinking endpoints inside the mitochondrial skeleton was related to the cytoplasm ( Figs. 15 D and 16 D). The length between endpoints inside the mitochondrial skeleton in vehicle control ( Fig. 15 ) was obviously longer than in cells treated with 50 μM of phenanthroline ( Fig. 16 ). Furthermore, the number of fragmented mitochondria in phenanthroline-treated cells ( Fig. 16 ) seemed to higher than in vehicle control ( Fig. 15 ).
We applied the granularity parameter from the CellProfiler program to determine the levels of mitochondrial fragmentation. From the texture analysis of the previous study, the granularity (granular spectrum) has been described as the size distribution normalized with integrated intensity [30] . Thus, the granularity parameter was used as a representative for the quantitative measurement of mitochondrial fragmentation. From the results, the percentage of mitochondrial fragmentation (granularity) from three different fibroblast cells (F1, F2, and F3) were significantly increased after 50 μM phenanthroline treatment compared with vehicle control ( Fig. 17 A). Conforming to the mitochondrial granularity, the average mitochondrial length substantially decreased in cells treated with phenanthroline at 50 μM when compared with vehicle control ( Fig. 17 B).We showed the distribution of mitochondrial granularity ( Fig. 18 A) and length ( Fig. 18 B) levels in duplicate wells in the condition of 50 μM of phenanthroline treated cells and untreated cells in individual fibroblast samples to account for any variations that might have an impact on the quality of the outcomes. The well-related distribution graph of mitochondrial granularity and length demonstrated a normal distribution with negligible variation between wells under all conditions (1 and 2). From the fieldrelated distribution of mitochondrial granularity of observation, the graph in the conditions of F1_phe 0 μM_Well 1 ( Fig. 18 C), F1_phe 0 μM_Well 2 ( Fig. 18 D), F1_phe 50 μM_Well 1 ( Fig. 18 E), and F1_phe 50 μM_Well 2 ( Fig. 18 F) revealed a minor variation between the fields, as well as field-related the distribution of mitochondrial length, F2_phe 0 μM_Well 1 ( Fig. 18 G), F2_phe 0 μM_Well 2 ( Fig. 18 H), F2_phe 50 μM_Well 1 ( Fig. 18 I), and F2_phe 50 μM_Well 2 ( Fig. 18 J). We additionally utilized the R program to identify the outlier in each well's data. We discovered that there was just one outlier in each condition in Fig.18 C -J.

Conclusions
Our study suggested a method for assessing the mitochondrial status, including ROS accumulation, MMP, and their morphology in fibroblast cells stained by specific-fluorescent dyes and observed with the High Content Imaging System. It was a powerful technology for high throughput detection providing faster and more reliable results than the conventional fluorescent microscope. We hope that our work will be useful for research studies involving mitochondrial-related pathological diseases in humans. Additionally, we believe that these methods can be applied to other types of cells.