Procurement of oral mucosa samples and culturing primary oral keratinocytes
The procurement of oral mucosa samples and procedure for oral keratinocyte cultures were described previously.29 All methods were performed in accordance with relevant guidelines and regulations. The details were described in the Supplementary Methods.
Time-lapse microscopic imaging
Under appropriate conditions, randomly-chosen five locations within the dish of the p1 cell culture were subjected to time-lapse observation. Phase-contrast images were taken at 8-min intervals for 4 h, until a total of 31 images were produced using a ×4 PlanFluor NA0.13 PhL objective lens. The images were converted to video files using a BZ-X analyser (Keyence) (Fig. 5A; Supplemental movie 2A). More detailed information was provided in the Supplementary Methods.
Determination of the cell area using image segmentation
Previously, we utilised a conventional segmentation method of applying the threshold to the intensity value by extracting areas containing targeted cells/colonies within the microscopic image. Briefly, using the MS as an index of cell locomotive ability, we selected areas with an MS of p0 cells/colonies that was equal to or greater than 1.00 pixel/frame, which corresponded to a p0 oral keratinocyte cell/colony area.29 However, it is time consuming and laborious to apply to p1 cell cultures instead of p0 cells/colonies in the current experimental design. Hence, we implemented a different approach to achieve high-precision segmentation.
For image segmentation before determining the MMS of the p1 oral keratinocyte cells/colonies by using the OF algorithm, we used the variational segmentation method of Chan and Vese43 to identify the surface area covered by all cells/colonies. The algorithm developed by Chan and Vese for active contours is a flexible method that enables the segmentation of many types of images. It has been applied to segment biological images such as cells, and its details have been reported.44 There are four parameters (λ1, λ2, ν and μ) in this algorithm, and the parameters should be determined by the user. Details of each parameter are explained in a previous study.44 Tuning of parameters was conducted by comparing manual and automatic segmentations and was selected to minimise the deviation between manual and automatic segmentations as much as possible. The validation data used was the image data shown in Fig. 5A. In this study, the preferred settings are λ1 = 1.2, λ2 = 1.0, ν = 0.02 and μ = 0.8.
Cell segmentation is conducted using the following three steps:
Step 1. Smoothing: All frames are smoothed using a Gaussian kernel (size is 3 pixel × 3 pixel window).
Step 2. Detect edges: 3 × 3 convolution kernels (Sobel filter) are applied to the image and vertical and horizontal derivatives are generated. Edges are detected by combining the two derivatives using the square root of the sum of the squares (Fig. 5B).
Step 3. Segmentation of p1 cells/colonies: By applying the Chan–Vese algorithm to the obtained image, the image is binarised into the background and cell area (Fig. 5C).
In the Chan–Vese algorithm, the use of small-sized images is recommended, because computation is time consuming. However, since the image size obtained by current time-lapse microscopic imaging is high resolution (1920 pixel × 1440 pixel), the algorithm is programmed in Python language with CuPy45 to calculate based on the GPU (graphics processing unit). The runtime of Step 3 is approximately 10 min on a 1920 × 1440 resolution for cell images (31 frames) using a single GPU (GeForce GTX 1080 Ti, 11 GB) core on a common desktop PC (memory, 16 GB).
Measurement of the overall cell/colony growth
Because of the implementation of a different image segmentation, it is necessary to determine what cell growth conditions of p1 oral keratinocytes are appropriate to start time-lapse imaging. The simplest way to evaluate cell growth kinetics is to measure the overall cell/colony growth,46 which was obtained via micro-photographing cells grown in the same culture dish at four different cell confluencies of 30%, 50%, 70% and 90% using manual observation. From the binary image after segmentation, the overall cells/colonies growth (%) was calculated, which was obtained by dividing the surface area covered by all cells/colonies shown in the full-screen mode by the entire image size (in pixel) for each frame over 4 h.
Determination of MMS of the p1 cells
The MS was determined in this study by using the identical OF algorithm of our previsou study.29 A total of 31 sequence frames were created, in which vectors are drawn for each full-screen image (Fig. 5D; Supplemental movie 2B). The following procedure was described in the Supplementary Methods.
Sample size estimation
The minimum required sample size was a priori determined by power analysis using G*power software.47,48 As a result, a total sample size of 26 was required by setting two-tails with an effect size of 0.5, a significance level of 0.05 and a power of 0.8.49
Standard protocol for culturing p1 oral keratinocytes and evaluating the correlation between the motion index and proliferative capacity
To evaluate the correlation of the proliferative capacity with the MMS under the standard protocol according to the current experimental design, p1 cells grown at approximately 50% confluence in the 35-mm dish were subject to time-lapse microphotography. The mean MMS of the five locations was represented as the MMS of the sample (n = 32). Regardless of the confluency, 24 h after the completion of time-lapse microphotography, p1 cells were collected, and the number of cells was counted to determine the proliferative capacity. As a parameter of proliferative capacity, population doublings (PDs) and PDT of p1 cells were calculated as described in Supplementary Methods. Subsequently, a total 32 of MMS and PDT were plotted on the scatter plot to evaluate the correlation between MMS and PDT.
Metabolic challenge protocols for oral keratinocyte culture
We hypothesised that there should be a threshold for the MMS of p1 cells that can be differentiated as substandard cell populations. We compared the proliferative capacity of p1 cells under standard protocol and under two types of metabolic challenge protocols that include a lower nutrition challenge created by the dilution of completed EpiLife® with D-PBS (Wako Chemical, Osaka, Japan) by 1:5 (5xPBS) and 1:20 (20xPBS) and a no-feeding challenge without a fresh culture medium change for four days (4dNoF) and seven days (7dNoF). Different from the standard protocol, which was used as a control for the metabolic challenges, a density of 1.25 × 105 cells was plated into a 35-mm dish for those challenges except for 20x PBS, for which the cells were plated at a density of 1.5 × 105 cells in a 60-mm dish with complete EpiLife® medium. To secure random sampling, cells were alternately subject to culturing under metabolic challenge protocols using a total of 32 oral keratinocytes (n = 16).
To test a lower nutrition challenge protocol, the cells were fed every 2 days with complete EpiLife®, and at approximately 50% confluency, they were fed with each medium diluted with D-PBS by the indicated ratio, respectively. The cells of 5xPBS in the 35-mm dish were fed every 2 days with the diluted medium and time-lapse micro-photographed 3 days later. By contrast, the cells of 20xPBS in the 60-mm dish were photographed under the time-lapse microscope 4 h after switching the diluted medium.
For the no-feeding challenge protocol, no medium change was made until they were subjected to time-lapse microphotography, which took three days for 4dNoF and six days for 7dNoF after the medium change, respectively.
Regardless of the confluency rate, 24 h after time-lapse microphotography, the p1 challenged cells were collected using the previously described method, and the number of cells was counted. PD and PDT were calculated using the formula provided in Supplementary Methods in which N = the number of challenged p1 cells collected, N0 = the cell number inoculated, which is 1.25 × 105 for 5xPBS, 4dNoF and 7dNoF or 1.5 × 105 for 20xPBS) and I = days in the culture of challenged p1 cells. A step-by-step chart of the four metabolic challenge protocols is shown in Supplemental Figs. 3A and 3B.
Manufacturing of ex vivo produced oral mucosa equivalents (EVPOME)
Histologic and immunohistochemical examination of EVPOMEs
Evaluation of the proliferative activity of challenged oral keratinocytes in EVPOME
Manufacturing of EVPOMEs, their histologic and immunohistochemical examinations and evaluation of the proliferative activity of challenged oral keratinocytes in EVPOME were described in the Supplementary Methods.
To examine the strength of a linear association between MMS and PDT, Spearman’s correlation coefficient was calculated for non-normal distribution of the PDT, and the coefficient, r, and p values were determined using Prism 7.05 (GraphPad Software, San Diego, CA, USA). The results of PI are presented as the mean ± standard deviation (SD). The comparisons between the cells under the standard protocol and metabolic challenge protocols (5xPBS and 4dNoF) were examined using a paired t-test. A p-value <0.05 was considered statistically significant.
The software and datasets generated and analysed during this study can be provided by the corresponding author upon reasonable request due to pending patent application.