Method for managing quality of stem cells by image analysis using ai

Abstract

Provided is a reference standard for use in the evaluation or determination of the quality of cells of interest, the reference standard including a rapidly expanding mesenchymal stem cell population characterized by employing, as an indicator, the positivity or negativity for at least one marker selected from LNGFR (CD271) and Thy-1 (CD90).

Description

Stem cell quality control method using AI-based image analysis

[0001] The present invention relates to a method for quality control of stem cells by image analysis, particularly by image analysis using AI.

[0002] Regenerative medicine is a medical treatment that regenerates dysfunctional organs, and examples of practical applications include heart transplants, liver transplants, kidney transplants, and hematopoietic stem cell transplants. Three types of stem cells that are expected to be applied to regenerative medicine include embryonic stem cells (ES cells) obtained from fertilized eggs, iPS cells obtained by introducing the four Yamanaka factors into skin fibroblasts, and somatic stem cells present in tissues and organs within the body. Currently, research and development into the use of these in drug discovery and clinical applications is being actively conducted around the world.

[0003] Stem cells exist in every tissue in the body, and mesenchymal stem cells (MSCs) are one type. MSCs are primarily found in bone marrow and are involved in the regeneration of bone and cartilage. In particular, in recent years, cell products using MSCs have been approved as regenerative medicine products, and the industrialization of regenerative medicine is rapidly accelerating as the mass production of cells using allogeneic cells has expanded the possibility of reducing costs and catering to a large market.

[0004] The present inventors developed a technique for directly isolating human MSCs from bone marrow using two antibodies, LNGFR (CD271) and Thy1 (CD90), and a cell sorter. They then successfully selected and isolated ultra-high-purity human MSCs, a cell population with extremely high proliferation and differentiation potential. These ultra-high-purity human MSCs were named REC (Rapidly Expanding Cells). They also demonstrated that RECs do not cause side effects such as pulmonary embolism even when transplanted in numbers 100 times higher than conventionally cultured MSCs (Patent Documents 1-4, Non-Patent Document 1).

[0005] However, in the current state of regenerative medicine, the price of a single regenerative medicine product undergoing clinical research is estimated to be in the hundreds to tens of millions of yen, making it difficult to deliver to many patients and placing a strain on medical finances. The main reason regenerative medicine is so expensive is the inefficiency of the cell manufacturing process. Currently, all cells for regenerative medicine are produced individually by hand, requiring skilled artisans to produce them. A particularly significant challenge is the lack of cell quality control technology, meaning that even in cutting-edge research facilities, hospitals, and cell banks around the world, cell quality assessment must rely on the discerning eye of skilled experts.

[0006] On the other hand, the present inventors have developed a technology that can non-destructively, in real time, at low cost, and quantitatively inspect and predict the quality of cells in culture for a wide variety of cells used in human cell therapy by using a technology (cell morphology information analysis) that combines image processing technology with AI (artificial intelligence) and machine learning (Patent Documents 5 to 8, Non-Patent Document 2).

[0007] Cell morphology analysis is a technology that combines image processing, data mining, and AI analysis technologies to predict the biological evaluation results (quality) of cells from images alone. Using cell morphology analysis, quality information that could previously only be obtained by staining after long-term culture can now be detected within three hours of seeding.

[0008] However, the development of the above-mentioned cell morphology information analysis technology requires a large amount of high-quality data (images and quality data of functional stem cells), and even using regular commercially available cells is extremely expensive, so until now, analysis of data from only 10 to 20 cells has been limited. Furthermore, the quality of commercially available cells is not consistent between lots, and they are a mixture of cells with various properties, so there are still issues with the suitability of the cells.

[0009] Japanese Patent No. 6463029, Japanese Patent No. 6850944, Japanese Patent No. 6932390, WO2021 / 145002, JP 2016-189701, JP 2016-189702, JP 2016-191565, JP 6721895

[0010] Mabuchi Y et al., "LNGFR+ Thy-1+ Vcam-1hi+ cells reveal functionally distinct subpopulations in mesenchymal stem cells". Stem Cell Reports 1(2): 152-165, 2013. NEDO FY2018-2019 Results Report: Development of Core Technologies for Next-Generation Artificial Intelligence and Robotics / Next-Generation Artificial Intelligence Technology Field / Research and Development on Advanced Quality Inspection of High-Purity Mesenchymal Stem Cells Using AI.

[0011] Given the above circumstances, there has been a demand for the development of cells suitable for cell morphology information analysis techniques.

[0012] As a result of intensive research to solve the above problems, the present inventors have found that REC is useful in cell morphology information analysis techniques, and have thus completed the present invention.

[0013] [1] A standard for evaluating or determining the quality of test cells, comprising a rapidly proliferating mesenchymal stem cell population characterized by the positivity or negativity of at least one marker selected from LNGFR (CD271) and Thy-1 (CD90). [2] The standard according to [1], wherein the cell population is characterized by any one of the following information (a) to (d): (a) at least one piece of cell information selected from the group consisting of cell size, degree of extension, complexity of contour, pattern, and brightness of individual cells within the cell population, and image information thereof; (b) information obtained by statistically processing the cell information of (a) across multiple cells within the cell population; (c) combined information obtained by combining the information of (b) across multiple cell populations; or (d) the information of (a) to (c) obtained over time. [3] The standard according to [2], wherein the statistically processed information is at least one piece of information selected from the group consisting of mean value, standard deviation, and distribution shape. [4] The reference standard according to [1], wherein the quality is at least one selected from the group consisting of cell proliferation rate, number of recovered cells, degree of undifferentiation, degree of differentiation, amount of production factor, and heterogeneity of cells. [5] The reference standard according to [1], wherein the test cells are stem cells. [6] The reference standard according to [5], wherein the test stem cells are mesenchymal stem cells. [7] A method for providing information for quality assessment of a test stem cell population, comprising the following steps: (a) characterizing individual cells of a fast-proliferating mesenchymal stem cell population using as an indicator whether at least one marker selected from LNGFR (CD271) and Thy-1 (CD90) is positive or negative, and acquiring image data (reference image information) of the individual cells, or acquiring parameter information characteristic of the cell population (reference parameter information) from the reference image information; (b) acquiring image data (test image information) of individual cells of a test stem cell population provided by a user, or acquiring parameter information characteristic of the test stem cell population (test parameter information) from the test image information; (c) comparing the test image information or test parameter information with the reference image information or reference parameter information; and (d) providing all or a portion of the information obtained by the comparison to the user for quality assessment of the test stem cell population.[8] A method for evaluating the quality of a test stem cell population, comprising the following steps: (a) characterizing individual cells of a rapidly proliferating mesenchymal stem cell population using as an index the positivity or negativity of at least one marker selected from LNGFR (CD271) and Thy-1 (CD90), and acquiring image data of the individual cells (reference image information), or acquiring parameter information characteristic of the cell population (reference parameter information) from the reference image information; (b) acquiring image data of individual cells of the test stem cell population (test image information) provided by a user, or acquiring parameter information characteristic of the test stem cell population (test parameter information) from the test image information; (c) comparing the test image information or test parameter information with the reference image information or reference parameter information; and (d) evaluating the quality of the test stem cells using all or a part of the information obtained by the comparison as an index. [9] An information provision system for evaluating the quality of a test stem cell population, comprising the following means: The system comprises: (a) a means for characterizing individual cells of a rapidly proliferating mesenchymal stem cell population using as an indicator whether at least one marker selected from LNGFR (CD271) and Thy-1 (CD90) is positive or negative, and for acquiring image data (reference image information) of the individual cells, or acquiring parameter information (reference parameter information) characteristic of the cell population from the reference image information; (b) a means for acquiring image data (test image information) of individual cells of a test stem cell population provided by a user, or acquiring parameter information (test parameter information) characteristic of the test stem cell population from the test image information; (c) a means for comparing the test image information or test parameter information with the reference image information or reference parameter information; and (d) a means for providing all or a portion of the information obtained by the comparison to a user for quality assessment of the test stem cell population.

[10] A system for evaluating the quality of a test stem cell population, comprising: (a) a means for characterizing individual cells of a rapidly proliferating mesenchymal stem cell population using as an index the positivity or negativity of at least one marker selected from LNGFR (CD271) and Thy-1 (CD90), and for acquiring image data of the individual cells (reference image information), or acquiring parameter information characteristic of the cell population (reference parameter information) from the reference image information; (b) a means for acquiring image data of individual cells of the test stem cell population (test image information) provided by a user, or acquiring parameter information characteristic of the test stem cell population (test parameter information) from the test image information; (c) a means for comparing the test image information or test parameter information with the reference image information or reference parameter information; and (d) a means for evaluating the quality of the test stem cells using all or a part of the information obtained by the comparison as an index.

[11] The system described in [9] or

[10] , wherein the characteristic of the cell population is selected from any of the following information (a) to (d): (a) at least one piece of cell information selected from the group consisting of cell size, degree of extension, complexity of contour, pattern, and brightness of individual cells in a cell population, and image information thereof; (b) information obtained by statistically processing the cell information of (a) by integrating it across multiple cells in a cell population; (c) information obtained by combining the information of (b) with multiple cell populations; and (d) information obtained over time from (a) to (c).

[12] The system described in

[11] , wherein the statistically processed information is at least one piece of information selected from the group consisting of mean value, standard deviation, and distribution shape.

[13] The system described in [9] or

[10] , wherein the quality is at least one piece of information selected from the group consisting of cell proliferation rate, number of recovered cells, degree of undifferentiation, degree of differentiation, amount of production factor, and cell heterogeneity.

[14] The system described in [9] or

[10] , wherein the test stem cells are test mesenchymal stem cells.

[15] An information provision program for quality assessment of a test stem cell population, the program causing a computer to function as the following means: (a) a means for characterizing individual cells of a fast-proliferating mesenchymal stem cell population using as an indicator whether at least one marker selected from LNGFR (CD271) and Thy-1 (CD90) is positive or negative, and for acquiring image data of the individual cells (reference image information) or acquiring parameter information characteristic of the cell population (reference parameter information) from the reference image information; (b) a means for acquiring image data of individual cells of a test stem cell population (test image information) provided by a user or acquiring parameter information characteristic of the test stem cell population (test parameter information) from the test image information; (c) a means for comparing the test image information or test parameter information with the reference image information or reference parameter information; and (d) a means for providing all or a portion of the information obtained by the comparison to a user for quality assessment of the test stem cell population.

[16] A program for evaluating the quality of a test stem cell population, the program causing a computer to function as the following means: (a) a means for characterizing individual cells of a rapidly proliferating mesenchymal stem cell population using as an index the positivity or negativity of at least one marker selected from LNGFR (CD271) and Thy-1 (CD90), and for acquiring image data of the individual cells (reference image information), or acquiring parameter information characteristic of the cell population (reference parameter information) from the reference image information; (b) a means for acquiring image data of individual cells of a test stem cell population (test image information) provided by a user, or acquiring parameter information characteristic of the test stem cell population (test parameter information) from the test image information; (c) a means for comparing the test image information or test parameter information with the reference image information or reference parameter information; and (d) a means for evaluating the quality of the test stem cells using all or a part of the information obtained by the comparison as an index.

[17] The program according to

[15] or

[16] , wherein the characteristic of the cell population is selected from any of the following information (a) to (d):(a) at least one piece of cell information selected from the group consisting of cell size, degree of extension, contour complexity, pattern, and brightness of individual cells within a cell population, and image information thereof; (b) information obtained by statistically processing the cell information of (a) by integrating it across multiple cells within a cell population; (c) combined information obtained by combining the information of (b) with multiple cell populations; and (d) the information of (a) to (c) obtained over time.

[18] The program described in

[17] , wherein the statistically processed information is at least one piece selected from the group consisting of mean value, standard deviation, and distribution shape.

[19] The program described in

[15] or

[16] , wherein the quality is at least one piece selected from the group consisting of cell proliferation rate, number of recovered cells, degree of undifferentiation, degree of differentiation, amount of production factor, and cellular heterogeneity.

[20] The program described in

[15] or

[16] , wherein the test stem cells are test mesenchymal stem cells.

[21] A computer-readable recording medium having the program described in

[15] or

[16] recorded thereon.

[22] An image processing method for rapidly proliferating mesenchymal stem cells, comprising the following steps: (a) acquiring image data of the standard described in [1], and (b) removing noise from the acquired image data and / or normalizing the image data.

[23] A method for providing image data processed by the method described in

[22] for computer machine learning.

[24] A method for predicting the quality of a test cell by comparing image data processed by the method described in

[22] with image data of the test cell.

[0014] The present invention provides a standard for evaluating or determining the quality of test cells, including rapidly proliferating mesenchymal stem cell populations. The standard of the present invention makes it possible to achieve efficient quality control and cost reduction when producing stem cells as cell medicines, thereby realizing the cost of producing stem cell medicines.

[0015] Conceptual diagram of continuous passage potential prediction using morphological profiles to select a highly potent cell bank. Target RECs were selected from MSCs through clonal selection steps (P1 and P2). Cells were cryopreserved at P3 to form a primary cell bank for the preparation of early-passage cells for further experiments. Serial passage tests were conducted from P3 to the passage limit. For practical cell-based therapeutic product manufacturing, a candidate cell bank is formed during such serial passages. However, there is a risk that cells may exhibit unexpected growth arrest, resulting in failure to establish a cell bank. Furthermore, although the proliferation potential may be further reduced, the risk of obtaining bank cells with higher activity and higher potential is lower, i.e., a higher-quality candidate cell bank is expected to be formed. Such continuous passage potential was predicted from the morphological profile of cell images at P2. The morphological profile is composed of the time course, 12 morphological descriptors, and cell population information (mean and SD). Figure 1 shows the profiles of 15 clones of RECs. (a) Results of a serial passage test of RECs. b) Typical morphology of REC. c) Growth profile of REC. d) Correlation between proliferation rate and limit number of serial passages. ２ indicates the coefficient of determination. a) Comparison of size distribution and its time course between 15 clones of REC and conventionally treated bulk MSCs (cMSCs, 9 lots), as well as morphological characterization of 15 RECs. Only adherent and outgrowing cells were counted. The dashed vertical line represents the average cell size. b) Representative time-lapse images of RECs and bulk MSCs. Yellow and blue arrows represent proliferating and recently proliferating cells, respectively. c) Size distribution and time course changes among 15 clones. d) Correlation between "SD of area" and passage number, R ２is the coefficient of determination. (e) and (f) show PCA plots of 15 RECs profiled by 24 morphological descriptors. (e) is a PCA plot with clone color labels (e), and (f) is a PCA plot colored by a heat map of serial passage potential. Figures show a comprehensive evaluation of data utilization effects and the performance of serial passage prediction models. (a) is a diagram showing the evaluation of data utilization effects using LAS SO. Row values ​​represent the number of FOVs used, and column values ​​represent the size of the time window used for training data. The heat map shows the RMSE. An RMSE < 1.0 is considered a well-performing model. (b) is a scatter plot to visualize the performance of the serial passage prediction model, with each dot representing one clone. (c) is a comparison of model structures between two pairs of constructed models, with time windows of 6, 6-30, 6-60, and 6-90 hours selected. The heat map shows the correlation coefficients between all weights on all selected morphological descriptors within the model. The correlation coefficient is high when the combination of descriptors used is similar. d: Evaluation of data utilization effectiveness by RF. The rows indicate the number of FOVs used, and the columns indicate the time window size used for the training data. The heat map shows RMSE. RMSE < 1.0 is considered a well-performing model. This figure shows the robustness of the continuous passage prediction model to data variability included by bootstrap FOV selection (50 iterations) and an evaluation of its data utilization effectiveness. This figure is a block diagram of the system of the present invention. This figure is a flowchart showing the operation of the system of the present invention. This figure shows information extracted from stem cell image data. The red dots represent REC clone MSCs, indicating that the information is homogeneous and well-organized. The horizontal axis is a numerical index. The mountain shape in the figure is a histogram of thousands of cell populations. This figure shows information extracted from stem cell image data. The information extracted from REC image data is well-organized when viewed from multiple indexes. The blue dots represent bulk MSCs, scattered in various locations, while the red dots represent REC clones, forming a neat oval shape. This figure shows information extracted from stem cell image data. The left panel shows REC clones, and the right panel shows bulk stem cell populations (commercially available MSCs). The vertical axis represents time.REC clones change over time while maintaining a single distribution. In the bulk, two peaks are observed in the early stages of culture, and over time, these peaks fuse or mix. This means that cellular information is confusing. Figure showing the results of cell sorting using a flow cytometer. Through the selection process using a flow cytometer, populations are cloned, resulting in the production of various clones. This suggests that clones other than REC that would otherwise be discarded may potentially be utilized as cells with specific characteristics for machine learning. Figure showing the proliferation rate (top panel), degree of adipogenesis (middle panel), and degree of osteogenic differentiation (bottom panel) of each MSC. Figure showing the adipogenesis prediction model for each MSC. High accuracy is observed for REC clone data, but low accuracy is observed for bulk. Figure showing the results of bulk data prediction using a clonal model. Figure showing the integration of cell morphology information. Figure showing the integration of cell morphology information. Figure showing the morphological profile of a cell population. Figure showing the proliferation profile of bulk MSCs.

[0016] 1. Overview The present invention relates to a standard for evaluating or determining the quality of test cells, including a rapidly proliferating mesenchymal stem cell (REC) population. In the present invention, cell image information analysis of each REC clone was performed, and based on the results of comparison with actual culture data, a learning model was constructed using data from each REC clone, thereby completing a quality evaluation system.

[0017] More than 100 clones were used in the cell image information analysis, including commercial-grade clones, clones with confirmed deviations, clones from different donor lots, and clones grown under different culture environments. Detailed culture process data was also organized as metadata along with image data and compiled into a database. Furthermore, experimental results for each clone's osteogenic differentiation, adipogenic differentiation, and proliferation ability were added to the database.

[0018] Furthermore, in one aspect of the present invention, AI was utilized to select three nonlinear regression models in addition to the conventional linear regression model as a machine learning model, and the source code was developed and the environment was configured. Furthermore, in this invention, the learning model was validated using REC clone data, and by referencing and comparing REC data, a method for evaluating the quality of other mesenchymal stem cells was established. For example, it was found that the cell recovery rate after 5 days of REC can be predicted with high accuracy from data during as little as two days of expansion culture, and that deviations in differentiation potential after one month can be predicted from images during expansion culture. Furthermore, using a quality database comprehensively validated with REC cells, the heterogeneity of competitors' MSCs was quantitatively evaluated from REC data, and it was found that the quality of competitors' MSCs can be predicted with nearly identical performance.

[0019] 2. Standard for Evaluating or Determining the Quality of Test Cells The standard used in the present invention is a rapidly proliferating mesenchymal stem cell population characterized by the positive or negative expression of at least one marker selected from LNGFR (CD271) and Thy-1 (CD90). Rapidly proliferating mesenchymal stem cells that are positive or negative for at least one marker selected from LNGFR (CD271) and Thy-1 (CD90) can be obtained, for example, according to the method described in WO 2009 / 31678.

[0020] The method is outlined as follows: First, mesenchymal stem cells are highly enriched by selecting a cell fraction positive for at least one of CD271 and CD90 (CD271+ / CD90+) from a cell population containing human mesenchymal stem cells. If the cell population containing human mesenchymal stem cells also contains blood cells, a step of selecting cells that are co-negative for CD45 and CD235a (CD45-CD235a-) may be added to select non-blood cells.

[0021] A cell population containing mesenchymal stem cells can be prepared by flow cytometry or affinity chromatography. The material for obtaining this cell population is not particularly limited, but examples include bone marrow and peripheral blood (including peripheral blood after G-CSF administration). Bone marrow may be from the spine, sternum, ilium, or the like. During cell preparation, if the material is in the form of a cell mass containing mesenchymal stem cells, the material can be subjected to physical treatment such as pipetting or enzymatic treatment with trypsin, collagenase, or the like, as necessary. Furthermore, if the material contains red blood cells, it is preferable to hemolyze the red blood cells in advance.

[0022] CD271+ / CD90+ cells are selected using the cell population prepared as described above. Methods for selecting CD271+ / CD90+ cells include, for example, methods using antibodies. The antibodies are anti-CD271 and anti-CD90 antibodies, which can select CD271+ / CD90+ cells. When using flow cytometry for selection, live cells can be selected in a short period of time by appropriately combining anti-CD271 and anti-CD90 antibodies labeled with different fluorescent dyes, such as FITC, PE, or APC. In addition to flow cytometry, CD271+ / CD90+ cells can also be selected using methods using magnetic beads or affinity chromatography.

[0023] Next, the selected LNGFR and Thy1 co-positive cells are cultured as single cells (clones), and by selecting a batch with rapid proliferation, highly purified human mesenchymal stem cells (REC: Rapidly Expanding Clone) with excellent proliferation, differentiation, and migration capabilities are obtained. Mononuclear cells are prepared from human bone marrow or adipose tissue / placental chorion, and bone marrow mononuclear cells are stained with anti-LNGFR and anti-Thy1 antibodies. Next, LNGFR-positive and / or Thy1-positive cells are clonally sorted into a 96-well culture plate using flow cytometry (cell sorter). That is, one cell is seeded per well. After two weeks of single cell culture, the culture plate is photographed under a microscope, and wells that have become confluent or semi-confluent are selected, and the cells contained in those wells are designated as REC.

[0024] Here, "fast proliferation" and "rapid proliferation" refer to a proliferation rate (doubling time: 26±1 hours) sufficient to make the culture plate confluent or semi-confluent two weeks or earlier after the start of culture when one cell is seeded per well of a 96-well culture plate. Confluence refers to a state in which the cultured cells cover 90% or more of the surface of the culture vessel (culture surface). Semi-confluence refers to a state in which the cultured cells cover 70-90% of the surface of the culture vessel (culture surface). The size and type of culture equipment used can be adjusted appropriately depending on the cell proliferation rate. Moderately / slowly expanding cells, i.e., cells that do not reach semi-confluence or confluence after two weeks of single-cell culture, are discarded. RECs selected from each well are collected and transferred to a culture flask, and further cultured until confluence is reached (expansion culture). The expanded cells are then collected separately, and REC from one well is designated as one lot (one clone in the Examples below), which is used in the selection described below.

[0025] The cells of the present invention are obtained by clonal sorting, in which one cell is seeded per well, and therefore the genetic traits of the expanded cells are all identical. Therefore, in the present invention, the entire cell population may be referred to as a "clone," or each individual cell constituting the cell population may be referred to as a "clone."

[0026] In the present invention, REC to be used for selection and quality evaluation can be evaluated in advance using a REC marker (anti-Ror2). For example, after the expansion culture, adherent and proliferated cells are collected from all lots, and a portion (1 to 3 × 10) of each lot is collected. ３Approximately 100 cells are selected and single-stained with an anti-Ror2 monoclonal antibody. A technique for single-staining with an anti-Ror2 monoclonal antibody is known ( WO 2016 / 17795 ). In summary, the proportion of REC marker-positive cells in the collected cells is determined by flow cytometry analysis using the REC marker. The proportion may be determined by quantitative PCR to quantify Ror2 mRNA expression, or by manually determining the proportion using a microscope. A lot (cell population) with a positive proportion above a certain value (e.g., 65%) is considered acceptable.

[0027] 2. Characteristics of the cell population In the present invention, the characteristics of the cell population used as a standard for evaluating or determining the quality of the test cells can be obtained from any of the following information (a) to (d). The test cells are not particularly limited, but mesenchymal stem cells, embryonic stem cells (ES cells), iPS cells, etc. are preferred. (a) At least one piece of cell information selected from the group consisting of cell size, degree of extension, complexity of outline, pattern and brightness of individual cells within the cell population, and image information thereof.

[0028] Information on cell size, extension, and contour complexity can be obtained by measuring or calculating, for example, the area, cell diameter, width, length, circumference, arc length, or a combination thereof, as well as the proportion of these in the cell image (e.g., the number of pixels). Pattern and brightness can be obtained by using, for example, the number of pixels showing a particular hue in the image as an index.

[0029] (b) Information obtained by accumulating the cell information of (a) across multiple cells in the cell population and performing statistical processing. The cell information of (a) is accumulated across multiple cells in the cell population and then statistically processed. The method of statistical processing is not particularly limited. Examples of statistically processed information include average values, standard deviations, and distribution shapes.

[0030] (c) Combination information obtained from the information of (b) above for a plurality of cell populations The information of (b) above can be obtained from a plurality of cell populations and combined.

[0031] (d) The information (a) to (c) obtained over time. "Over time" means after a predetermined time has elapsed, starting from any time point after the preparation of the rapidly proliferating mesenchymal stem cell population, or continuously, for example, information obtained after 1 hour, 2 hours, 3 hours, 5 hours, 10 hours, etc.

[0032] In the present invention, the qualities that serve as indicators for evaluating or judging the test cells include, but are not limited to, cell proliferation rate, number of recovered cells, degree of undifferentiation, degree of differentiation, amount of production factors, and cell heterogeneity.

[0033] The cell proliferation rate refers to the rate of cell proliferation when a test cell or its progeny is cultured, and can be evaluated or determined using indicators such as the cell proliferation rate, the number of passages until cell proliferation stops, and the total number of cells obtained from a single cell through cell proliferation. The cell proliferation rate and these indicators are positively correlated. A high proliferation rate, i.e., a large number of these indicators, can be evaluated or determined as high quality (high proliferation potential). The number of recovered cells refers to the number of viable cells obtained after culturing a test cell or its progeny for a certain period of time. When high proliferation potential is important, cells with a high number of recovered cells can be evaluated or determined as "cells that meet the desired quality."

[0034] The degree of undifferentiation refers to the immaturity of the differentiation stage of the test cells in the process of functional differentiation of stem cells, or the diversity of cell types that can be differentiated from the test cells. When it is important to prepare and stock a large number of stem cells in an undifferentiated state, cells with a high degree of undifferentiation can be evaluated or determined as "cells that meet the desired quality." The degree of differentiation refers to the degree or speed at which the test cells normally differentiate into the desired type of cell, tissue, or organ. When it is important to differentiate stem cells to the desired state, cells with a high degree of differentiation can be evaluated or determined as "cells that meet the desired quality."

[0035] The amount of a production factor refers to the amount (e.g., concentration) of a specific cytokine produced by a test cell or the number of types thereof, and the amount of the production factor can be used to evaluate or determine whether or not the test cell has a target function (e.g., proliferation ability, migration ability, etc.), or the degree of this function, as well as the degree of aging of the test cell. For example, if the amount of cell growth factor produced by the test cell is high, it can be said that the proliferation ability is high, and if the amount of migration-promoting factor produced by the test cell is high, it can be said that the migration ability is high. In either case, it can be used to evaluate or determine the degree of conformity with the target cell quality.

[0036] Cell heterogeneity means that the phenotype of the test cells (cell state, including morphological characteristics, migration ability, responsiveness, etc.) is not homogeneous, indicating the possibility that a highly pure cell population meeting the desired quality cannot be produced (risk of production failure). Therefore, low heterogeneity indicates that process control parameters are being controlled to produce cells with the desired quality, while high heterogeneity indicates the risk that these culture controls are not being adequately performed. Sufficiently low heterogeneity can be evaluated or determined as indicating high cell production capability (sufficient process control) or a high cell production success rate (low culture risk, highly economical and efficient production).

[0037] 3. Information provision method, quality assessment method, and systems and programs used for these methods (1) Information provision method and quality assessment method The present invention provides an information provision method for quality assessment of a test stem cell population, and a quality assessment method for a test stem cell population, in which the cell standard of the present invention is used.

[0038] The information providing method of the present invention comprises the following steps: (a) characterizing individual cells of a rapidly proliferating mesenchymal stem cell population using as an indicator whether at least one marker selected from LNGFR (CD271) and Thy-1 (CD90) is positive or negative, and acquiring image data of the individual cells (reference image information) or acquiring parameter information characteristic of the cell population (reference parameter information) from the reference image information, (b) acquiring image data of individual cells of a test stem cell population provided by a user (test image information) or acquiring parameter information characteristic of the test stem cell population (test parameter information) from the test image information, (c) comparing the test image information or test parameter information with the reference image information or reference parameter information, and (d) providing all or a portion of the information obtained by the comparison to the user for quality assessment of the test stem cell population.

[0039] The present invention also provides a method for evaluating the quality of test cells, which includes a step (d) of evaluating the quality of the test stem cells using all or part of the information obtained by the comparison in step (c) as an index. These methods can use systems and programs including means for achieving steps (a) to (d). The system of the present invention is as follows.

[0040] (2) Information Processing Device and Program An information processing device and program for quality assessment according to one embodiment of the present invention will be described with reference to the drawings. Figure 6 is a schematic diagram of a test stem cell population quality assessment system 1 that includes an information processing device 10 according to this embodiment. In the following, "users" refer to institutions that handle test cells and institutions (e.g., testing companies, hospitals, universities) or individuals that wish to obtain stem cell information.

[0041] The information processing device 10 is a server-type computer equipped with a CPU, a ROM, a RAM, an input / output interface, etc. The information processing device 10 communicates with a user device 30 via a network 20 such as the Internet or a LAN. ａ１ , 30 ａ２ , ...30 ａｎ(collectively referred to as "user device 30") are communicably connected. Note that the number of user devices 30 may be one or more.

[0042] The information processing device 10 characterizes individual cells of a rapidly proliferating mesenchymal stem cell population, a cell standard, using the positive or negative markers of at least one of LNGFR (CD271) and Thy-1 (CD90), and acquires image data (reference image information) of the individual cells, or acquires parameter information (reference parameter information) characteristic of the cell population from the reference image information. The information processing device 10 then compares the stem cell information (including image information) of the user (referred to as the "target user") handling the test cells with each item of stem cell information of the REC standard stored in the database, and calculates and acquires parameter information based on the comparison results. The information processing device 10 then compares the calculated parameter information and evaluates and determines the quality of the test stem cell population of the target user. The information processing device 10 then transmits the user's stem cell information and the evaluation and determination results to the user device 30.

[0043] The user device 30 is a device such as a computer, smartphone, or tablet equipped with information input means and display means, and is communicably connected to the information processing device 10 via the network 20. The user device 30 displays various information received from the information processing device 10 on a display via a browser or an application, and provides an interface for inputting information to be sent to the information processing device 10.

[0044] The information processing device 10 includes functional units realized by the cooperation of a program stored in ROM or placed in RAM with the CPU, such as an information acquisition unit 101, a stem cell information extraction unit 102, an evaluation reference value setting unit 103, a comparison unit 104, an image information calculation unit 105, and a judgment unit 106. The information processing device 10 also includes a storage unit 110 provided in the ROM (or RAM), and the storage unit 110 includes a stem cell information database (DB) 111, a parameter information database (DB) 112, and an evaluation reference database (DB) 113.

[0045] The information acquisition unit 101 acquires various information, such as stem cell image information, from the user device 30 via the network 20, and stores it in the storage unit 110. The information received from the user device 30 includes stem cell information acquired by the user, and user identification information (such as the user's name, information on the stem cell source, information on the donor of the stem cell), or an ID set for the user).

[0046] When the information acquisition unit 101 receives a user's test stem cell data, it associates the data with user identification information and stores the data in the stem cell information DB 111. The stem cell information DB 111 stores stem cell data collected from multiple users in this manner. When the information acquisition unit 101 receives a user's stem cell information, it acquires parameter information of the characteristics of the test stem cell population from the stem cell information and stores the information in the parameter information DB 112. The evaluation criteria DB 113 stores information obtained by the statistical analysis process that extracts characteristics using the comparison unit 104. The information is stored in association with information that identifies the user population sample that was the subject of the statistical analysis process.

[0047] The statistically processed information includes information on the average value, standard deviation, distribution shape, or a combination of these. Details of this information are as described above.

[0048] 7 is a flowchart of processing by the information processing device 10 according to this embodiment. First, the information acquisition unit 101 of the information processing device 10 requests input of basic information and test stem cell information, such as cell images, from the user, acquires this information (S1), and stores it in the stem cell information DB 111 (S2). The stem cell information extraction unit 102 extracts quality indicators from the information stored in the stem cell information DB 111, such as at least one indicator selected from the group consisting of cell proliferation rate, number of recovered cells, degree of undifferentiation, both production factors, and cell heterogeneity, and stores this information in the parameter information DB 112 (S3). The evaluation reference value setting unit 103 also sets a judgment reference value for each quality evaluation item from the information stored in the parameter information DB 112, and stores this information in the evaluation reference DB 113.

[0049] The comparison unit 104 extracts standard stem cell population information (reference parameter information) and the user's stem cell information (subject parameter information) from the stem cell information DB 111 and the parameter information DB 112, and compares and collates the stem cell information between the reference parameter information and the subject parameter information (S4).

[0050] The image information calculation unit 105 calculates the numerical value of the quality evaluation of the test stem cells based on the comparison results between the reference parameter information and the test parameter information, and obtains information for quality evaluation and judgment (S5). The judgment unit 106 judges the quality of the test stem cells of the target user. The information processing device 10 references the data stored in the judgment criterion evaluation standard DB, judges whether the quality of the cells is good or bad using the reference value as a boundary, and displays the evaluation score. The display results are displayed on the user device by a display unit (not shown).

[0051] The program of the present invention can be stored on a computer-readable recording medium or a storage means connectable to a computer. Computer recording media or storage means containing the program of the present invention are also included in the present invention. Examples of recording media or storage means include, but are not limited to, magnetic media (flexible disks, hard disks, etc.), optical media (CDs, DVDs, etc.), magneto-optical media, flash memory, etc.

[0052] EXAMPLES The present invention will be explained in more detail below with reference to examples, although the scope of the present invention is not limited to these examples.

[0053] Mesenchymal stem cells (MSCs) are the most widely studied stem cells for cell therapy applications [1-3]. While various cell types are known to exist among MSCs [4-7], this is due to the fact that conventional MSC production involves simply isolating the adherent cell fraction from a cell suspension mixture [8, 9]. LNGFR (CD271) and THY-1 (CD90) co-positive cells (LT cells) are a specific subpopulation of MSCs in bone marrow. These cells exhibit unique characteristics that are useful for advancing MSC-based cell therapy: (1) high proliferation capacity, (2) diverse differentiation potential (adipogenic, osteogenic, and chondrogenic), (3) low expression of the senescence marker SA-β-ga 1, and (4) a uniform and small size that allows them to avoid entrapment in pulmonary capillaries after intravenous administration in a mouse model

[10] .

[0054] Relatively rapidly growing clones were observed in single-cell sorted LT cells and were named "rapidly expanding clones (REC)." Because the proliferation ability of REC is highly useful for establishing cell banks with less heterogeneity, REC is currently undergoing clinical trials for the treatment of hypophosphatasia and spinal stenosis

[11] .

[0055] Because MSCs have been used in numerous research projects for clinical applications [3, 12, 13, 14], expectations for the advancement of MSC-based cellular medicines are rising. Therefore, there is currently a strong demand for the development of technologies that can support the production of MSCs [15, 16]. The characteristics of RECs reduce two major critical risks in the current MSC manufacturing process and can greatly benefit from the increased feasibility of establishing an efficient and robust cell manufacturing process.

[0056] (1) Difficulty in controlling donor-to-donor variability: MSCs vary widely among donors [17, 18, 19], and it is virtually impossible to fully investigate donor cell variability before developing a robust manufacturing process [20-22]. Therefore, handling cells from various unknown donors is a highly risky task in MSC manufacturing and is a major source of unexpected and difficult-to-solve errors.

[0057] Establishing an allogeneic cell bank is a practical approach to controlling the quality of starting cells, which enables efficient MSC production. RECs can be highly beneficial in the process of establishing an allogeneic cell bank. Because RECs can be mass-produced by cell sorting, selecting RECs is more feasible and efficient than the difficulty of large-scale donor selection until an ideal cell bank is obtained. Furthermore, the proliferation potential of RECs significantly improves the success rate of establishing a cell bank while maintaining banked cells during the initial culture period. Therefore, manufacturing using RECs as starting cells allows for more feasible process development and stable quality control.

[0058] (2) Quality Deterioration During Cell Expansion Culture: MSCs lose their proliferation capacity after several passages, and other important quality characteristics also deteriorate during expansion culture [23-31]. However, achieving a certain cell number is an essential quality requirement in the manufacturing of MSC-based therapeutic products. From a therapeutic perspective, typical protocols for MSC-based therapies require more than one billion cells per treatment to ensure efficacy

[32] . From a manufacturing perspective, because most quality tests are invasive and cells are consumed for each test, the final cell bank must reach a high cell number to thoroughly test the final product against multiple quality criteria

[33] .

[0059] However, such cell manufacturing processes are an endeavor with no guarantee of success, as the possibility of quality deterioration cannot be denied. In fact, failure to establish a cell bank can only be found as a "fact" after all the work has been done, and it is difficult to prevent such failures in advance. In this context, the high and sustained proliferation capacity of REC, which allows for more than 10 consecutive passages, significantly minimizes the risk of failure to establish a cell bank and can achieve maximum culture efficiency for establishing a rich stock at early passages.

[0060] Although RECs have the advantage of low risk and efficient production, establishing a cell bank for RECs poses a dilemma. RECs can be effectively produced by cell sorting even from small cell sources. Therefore, the variation of the starting clones can be greater than that of allogeneic cell banks. However, expanding a large number of RECs to the final cell bank stage is costly, and furthermore, final quality testing to ensure the establishment of a cell bank can require additional costs and labor.

[0061] The variability of the starting clones of RECs raises the expectation that "better RECs" can be selected during the establishment of cell banks, but such expectation increases the number of cell banks established, which requires cost and labor. In practice, RECs are screened from LT cells based on their initial growth rate from single clones in 96-well plates, and all candidates are further expanded to form multiple "candidate cell banks," and "better REC cell banks" with higher quality are selected.

[0062] Although REC allows aggressive selection for better quality products (i.e., final products with lower risk and higher potency), it can increase the cost of the entire process. Therefore, it is necessary to identify such candidate RECs as early as possible to minimize excessive work. However, using conventional cell evaluation methods, it is extremely difficult to test cells at the early stage of cell bank establishment, especially RECs derived from single clones, because most of them are invasive and waste valuable cell resources.

[0063] In this study, we developed a non-invasive morphology-based method for predicting the potential of MSCs to select better RECs during the early stages of cell bank establishment, when cell numbers are extremely limited. Based on our previous findings based on early morphology-based prediction of MSC quality decline [34-37], we attempted to predict their potential for further serial passage using only their early morphological information to enhance the selection of better RECs for forming better REC cell banks (Fig. 1).

[0064] Furthermore, we anticipate that such capacity predictions can aid in the cell bank establishment process, which balances the "bank size" and "capacity of banked cells," a key dilemma in managing cell banks. Significant expansion of a cell bank can improve production efficiency, but this increases the risk of losing the proliferation capacity of banked cells. Therefore, we hypothesized that if future serial passage capacity could be predicted in advance, it would be possible to set a low-risk timing for achieving a maximum-sized cell bank with high capacity.

[0065] To develop a predictive model, we established 15 clones of RECs from bone marrow and experimentally confirmed their successive passage numbers up to the passage limit (defined as "continuous passage potential"). In this example, we attempted to develop a machine learning model to quantitatively predict such potential using morphological descriptors from the passage 1 (P1) stage. During the development of the predictive model, we performed a detailed analysis of the data utilization effects to robustly obtain a high-performance predictive model. Thus, our data demonstrated the most effective use of time-course data and the minimum number of images required to realize a predictive model in a practical manner. Our concept of predicting the future potential of RECs based on cell morphology suggests that an image-based, data-driven cell bank construction process for MSC production may achieve both efficiency and robustness.

[0066] Materials and Methods: Cells and Culture. To obtain RECs, bone marrow mononuclear cells were prepared from bone marrow aspirates (AllCells, Alameda, USA) collected from healthy donors using density gradient centrifugation with Ficoll (GE Healthcare, Chicago, USA). Bone marrow mononuclear cells were stained with anti-human rabbit anti-CD90 IgG (BD Biosciences, Cat# 559,869, Franklin Lakes, USA) and anti-human mouse anti-CD271 IgG (Thermo Fisher) for 1 hour at 37°C.

[0067] Single-cell sorting was performed for CD90 and CD271 co-positive cells using a cell sorter (JSAN, BayBioscience, Kobe, Japan). LT cells were further cultured as single clones (passage number = P0) in 96-well plates (Thermo Fisher Scientific, Waltham, USA) and cultured in a medium supplemented with 20% fetal bovine serum [FBS] (Cytiva HyClone, Marlborough, USA), 20 ng / ml basic fibroblast growth factor [bFGF] (Kaken Pharmaceutical, Tokyo, Japan), 0.01 M 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid [HEPES] (Thermo Fisher Scientific, Waltham, USA), and 1% penicillin / streptomycin (Meiji Seika). After 2 weeks, clones were grown to subconfluence in 96-well plates using maintenance medium (low-glucose Dulbecco's modified Eagle's medium [DMEM] (Wako, Osaka) containing 100% ethanol (Pharma, Tokyo, Japan), harvested, subcultured (passage number = P1) in a single well of a 6-well plate (Thermo Fisher Scientific) and set up for image acquisition.

[0068] Image acquisition for further morphological analysis was performed at this stage. Next, when the cells reached subconfluence in a 6-well plate (Thermo Fisher Scientific), they were harvested and subcultured in a T75 flask (Thermo Fisher Scientific) (passage number: P2). After incubation at 37°C for 3–5 minutes, the cells were harvested using TrypLE Select (Thermo Fisher Scientific).

[0069] P2 cells in a subconfluent T75 flask were cultured at 1.0 × 10 with CP-1 (KYOKUTO PHARMACEUTICAL, Tokyo, Japan) and 25% human serum albumin (CSL Behring, Tokyo, Japan). ６The cells were cryopreserved at 1000 cells / ml and designated as the "Primary Cell Bank (P3)." In order to preserve cells for serial passage studies, the Primary Cell Bank excluded in vitro and in vivo studies for multiple potencies.

[0070] To compare morphology, conventionally processed bulk MSCs (BMCs: Lot 0,000,394,413, 0,000,411,107, 0,000,413,042, 0,000,422,610, 0,000,423,370, 0,000,429,365, 0,000,446,319, 0,000,451,491, 0,000,458,207) purchased from Lonza Japan, Ltd. (Tokyo, Japan) were cultured at P3 in MSCGM (Lonza Japan, Ltd.) supplemented with BulletKit (Lonza Japan, Ltd.). All cultures were maintained at 37°C under CO ２ The concentration was maintained at 5% and changed every three days.

[0071] <Serial passage ability test> A primary cell bank (P3) vial was thawed and 2.0 x 10 ５ The cells were seeded at 100 mm dishes (Thermo Fisher Scientific) at 100 cells / dish and designated as P3 data. Cells were harvested from the 100 mm dishes using TrypLE Select (Thermo Fisher Scientific). When the cells reached subconfluence, they were subcultured at the same seeding density into new 100 mm dishes. This subculture process was repeated until cell growth stopped. The number of passages was counted, and the final passage number was designated the "limit number of continuous passages," which was defined as the "capacity for continuous passages" (Figure 1). At each passage, the total number of recovered cells was counted using a Cellometer Auto T4 (Nexcelom Bioscience, Lawrence, USA), and the result was recorded as the total number of recovered cells.

[0072] <Image Acquisition and Processing> In a 6-well plate (Thermo Fisher Scientific), P1 REC was photographed every 6 hours using an automatic image acquisition device Biostation CT (Nikon Corporation, Japan) at ×4 magnification (64 tilings / well, 16 mm ２Phase contrast images were acquired using a microscope (1000 pixels per image). Each clone lot was imaged in one well. At this stage, cell numbers were very limited.

[0073] For bulk MSCs, MSCs were plated in 6-well plates (Corning Incorporated, NY, USA) at 2000 cells / cm. ２ Cells were seeded at 100°C (n = 3 wells per lot). Time points were designated as time points 1 (6 hours after seeding) to 20 (120 hours after seeding). Time point 15 was selected as the final time point for morphology and proliferation rate measurements because some lots reached subconfluence or higher, making it difficult to accurately identify individual cells in images.

[0074] Raw images were processed to measure individual cell morphology to obtain a summary of the morphological profile of the cell population (a minimum of 300 cells and a maximum of 100,000 cells per well, covered by 64 images). Image processing was performed with original code using Python version 3.7.3, the package NumPy version 1.20.0, and OpenCV version 4.4.0. Eight processes were designed: 1) background adjustment, 2) texture enhancement, 3) binarization, 4) small object removal, 5) erosion, 6) small object removal, 7) fill holes, and 8) frame-touching object removal (Figure 17). After image processing, 12 basic morphological descriptors (Table 1) were measured for each cell site in each image.

[0075]

[0076] A summary of the cell population was described by calculating the "mean" and "standard deviation (SD)" using all single-cell-based morphological descriptor data. This summary was based on approximately 1 × 10 morphological data from a single well covered by 64 images. ４ ~1×10 ５ represented the cells.

[0077] We accumulated summaries of these morphological descriptors over the time course (6–90 h) and designated them the "morphological profile" of the sample. The complete morphological profile for each condition (= 1 well) contained 360 parameters (= 12 descriptors × (mean and SD) × 15 time points). All data processing was performed using R version 4.0.2.

[0078] Visualization of Morphological Profiles Gaussian kernel density estimation was performed using single-cell data from each time point, and cell populations were visualized using a single morphological descriptor. Distributions were estimated using kernel functions in R. For distribution estimation, raw data were log10 transformed to describe the "area." Measurement data for single cells with an area >200 pixels were collected, allowing for detailed discussion of the "area" between REC and conventionally treated bulk MSCs.

[0079] Many of the objects smaller than 200 pixels were found in proliferating "round cells," and it was therefore considered difficult to discuss size differences, as the characteristics of larger cells are diminished in such data. Morphological profiles of all lots were analyzed using principal component analysis (PCA), and the relative similarity of clones was visualized using multiple descriptors.

[0080] In PCA comparing the morphological profiles of clones, dots were colored by clone label (15 colors) and serial passage limit number (7-level gradation: 7-13). All data with the same time window size, including the use of different FOV numbers, were integrated and used to set all principal components to cover the entire data diversity, allowing visualization of the effect of varying the data usage, time window, and FOV number.

[0081] Next, data using different FOVs were plotted individually on fixed PCA axes. In the PCA to compare the effects of data utilization, each plot represents one clone. To visualize the explanatory power of different data utilization, all single-cell measurement data for each data size (varying the combination of different time window sizes and FOVs) were resampled by bootstrap (50 repetitions with overlapping allowed). New means and SDs were re-obtained from each resampled data. A total of 50 plots per clone were plotted using PCA and the resampled data. Student's t-tests were used to test for differences between morphological descriptors and their population distributions. All data processing was performed using R (version 4.0.2).

[0082] <Construction of a continuous passage capacity prediction model> The time course morphological profile was used as an explanatory parameter, and the experimentally determined continuous passage limit number was used as the objective parameter of the machine learning dataset. Two types of machine learning models were tested: a linear regression model called lasso regression (LASSO) and a nonlinear machine learning model called random forest (RF). Parameter selection for LASSO was performed using the mean-reduced Gini index. Model performance was validated by leave-one-out cross-validation and compared using root mean square error (RMSE).

[0083] The effect of data usage on morphological information was examined by exhaustive combinations of two parameters: time window length (ranging from 6 to 90 hours) and number of FOVs (ranging from 1 to 64). To vary the time window length, the window was shortened by a total of 6 to 90 hours from the final time point (90 hours) at each time point (6 hours). Therefore, the maximum total morphological descriptor consisted of 360 parameters, and the minimum consisted of 24 parameters. Images from 64 images were randomly selected to vary the number of FOVs. Bootstrapping was implemented, in which the FOVs from 64 images were repeatedly resampled 50 times, increasing the dataset size from 15 to 750 samples. The effect of data variation on the selected data size conditions (6 hours, 6 to 30 hours, and 6 to 60 hours vs. 15, 40, and 60 FOVs) was evaluated. All data processing and machine learning were performed using R version 4.0.2.

[0084] Results <Obtaining and Characterizing REC for Training Data> To develop a machine learning model for predicting "serial passage potential" from initial cell morphology, we began work to obtain 15 clones of REC from bone marrow mononuclear cells (BMMCs) derived from a single donor (scheme in Figure 1). Fifteen clones were obtained from the same donor sample and selected according to the basic criteria for REC: (1) they were LNGFR and THY-1 co-positive at the time of cell sorting, and (2) they reached subconfluence within two weeks after single-clone sorting into 96-well plates (P0 stage).

[0085] Clones expanded in 6-well plates and then in T-75 flasks were cryopreserved as a temporary cell bank (P3), a bank for preserving early-passage cells. To further select "better RECs," the serial passage ability of candidate clones in P2 was evaluated. The ability of cells to be passaged repeatedly several times can be considered an ideal criterion for selecting "better-quality cell banks" for further use. The serial passage ability test showed that RECs retained high proliferation potential, recording an average of 9.8 serial passages (Figure 2a). This performance can be considered superior to that of commercially available MSCs processed using conventional methods.

[0086] Next, we evaluated the morphological characteristics of the clones (Fig. 2b). At the P1 stage (very early stage) in 6-well plates, morphology contains important cellular signatures for their evaluation. However, manual morphological observation (without quantification) makes it difficult to distinguish between RECs.

[0087] The proliferation profile of P1 stage RECs was analyzed using time-lapse images (Fig. 2c). Of the 15 clones, 12 showed a proliferation rate of more than fourfold. This proliferation rate was significantly higher than that of multiple conventionally treated bulk MSCs in our study (Fig. 18). The clone with the lowest serial passaging capacity (clone 13) showed a low proliferation rate, while the clone with the highest serial passaging capacity (clone 4) showed a high proliferation rate. However, the coefficient of determination between the "proliferation rate" and the "passage limit number" was low (R ２ = 0.09) (Fig. 2d). Thus, this data indicates that proliferation rate measurements at the P1 stage cannot predict future serial passage potential.

[0088] Morphological characteristics of REC We attempted to quantitatively characterize REC (at P1 time point) by image-based morphological analysis, following known analytical methods [33-36]. Compared with conventionally treated bulk MSCs (cMSCs, 9 lots), RECs were more uniform and remained smaller after adhesion (Fig. 3a, b). Both RECs and cMSCs were similarly sized cell populations (median area = 358 μm) at the very early stage of adhesion (6 h, T-test p < 0.22). ２ , 335 μm, respectively). The median size of RECs remained small (363 μm ２ ), but bulk cMSCs expanded in size after 30 hours of culture (median area = 473 μm ２ , T-test p<0.000001). Furthermore, in this series of time-lapse images, more proliferating cells (appearing white and round under a phase-contrast microscope) were observed in the REC (Fig. 3b).

[0089] Visualization of population changes in RECs at P1 revealed that the adherent cell population of clones with reduced serial passage capacity remained large in cell area (Fig. 3c). Thus, although high-performance cells formed a population with uniform cell size, analysis of such size data was suggestive, and the coefficient of determination between the "SD of area" and the "passage limit" was low (Fig. 3d). Analysis using a single morphological descriptor was not efficient enough for quantitative prediction.

[0090] Clones were profiled using multiple morphological information described by 24 descriptors derived from the basic descriptors (Table 1) [33-36]. Morphological similarities of RECs were visualized using principal component analysis (PCA) (Fig. 3e shows the clone number, and Fig. 3f shows the serial passage potential of each clone).

[0091] These results indicate that there are specific clusters of clones with similar morphological profiles that slightly separate low- and high-potency clones. In fact, clones with low potency cluster at the lower end of the PC2 axis, while clones with high potency cluster at the higher end of the PC2 axis and in the center of the PC1 axis. The descriptors contributing to each axis of the PCA map, particularly the most descriptive PC2 axis, can be interpreted as follows (Table 2).

[0092] If cells become more uniform and spindle-shaped during 24–30 h of growth, the clone has a higher serial passaging potential (capable of long-term serial passage), whereas disruption of cellular morphological uniformity reduces serial passaging potential. Such unsupervised model analysis suggests that a combination of multiple descriptors may provide a better explanation for morphologically characterizing performance.

[0093] Morphology-based machine learning to predict serial passage potential: Next, we investigated the construction of a machine learning model using morphological information to quantitatively predict the "serial passage potential" of RECs based solely on morphological information (Figure 1). Morphological profiles (24 descriptors x 15 time points) were used to predict the limit number of serial passages. In constructing this prediction model, we attempted to understand the extent to which morphological information effectively contributes to the development of the prediction model.

[0094] Therefore, we examined the effect of using morphological data by varying two parameters: the time window effect and the number of fields of view (FOV). The reason for examining these parameters is that in previous studies by the authors attempting to predict MSC proliferation and osteogenic differentiation rates from morphological descriptors, detailed examination of these parameters revealed that image data collection efforts could be effectively reduced [33-36]. Shortening the time window and minimizing the number of FOVs not only saves time and effort in acquiring image data, but also improves the efficiency of prediction.

[0095] After thoroughly examining both the "time window effect" and the "number of FOVs effect" using lasso regression (LASSO), we found that even using morphological information from P1 cells, several parameter combinations yielded high-performance predictive models (RMSE < 1.0, shown as a green heatmap in Fig. 4a). This data suggests that predictive performance can be maintained even if the time window and number of FOVs are reduced. Furthermore, we inferred that the effect of the "number of FOVs" is more important than the effect of the "time window." This is because the time window can be shortened without degrading performance when collecting more than 15 FOVs.

[0096] A scatter plot was plotted to further understand the performance of the prediction model for the continuous passage potential of each clone (Fig. 4b). These data clearly demonstrate that our prediction model predicts the "quantitative value of the continuous passage limit number" (Fig. 4a, b). Many high-performance prediction models have been developed using training data of various sizes. However, because LASSO is an algorithm that generates the optimal combination of descriptors for each dataset, all model structures are randomly different due to the lack of a common structure, and it is suspected that similar-performance models with completely different model structures may be generated.

[0097] If the model structure differs across data conditions, the modeling results will not be robust and therefore not practical. Therefore, we compared the correlations of all model structures (Fig. 4c). We confirmed that high-performing models robustly shared similar model structures. This result suggested that the relationship between morphological descriptor combinations and serial passage ability could be modeled by specific universal combinations of morphological descriptors.

[0098] Comparison of highly contributing descriptors shared among different models, such as "Correlation_SD(18h)", "Correlation_SD(6h)", and "Energy_SD(18h)", positively contributed to the prediction of clones with high serial passage potential. In addition, "Correlation_mean(6h)", "Length_SD(18h)", and "Compactness_SD(18h)" negatively contributed to the prediction of clones with low serial passage potential (Table 3).

[0099] Both "correlation" and "energy" are texture descriptors, and therefore, they change the intensity profile under a phase-contrast microscope, generally reflecting the three-dimensional pattern and complexity of cells. In fact, the intensity profile changes dramatically when cells change their three-dimensional circularity during cell division. Therefore, an increase in "texture SD" indicates the presence of a larger proliferating cell population. Therefore, when "texture mean" has a greater influence, the cell population becomes more uniform in texture, which means that the number of cellular events that change the texture decreases.

[0100] Length and compactness are shape-related descriptors, and therefore generally reflect two-dimensional responses such as elongation and expansion. Such shape changes, but with little or no texture change, indicate that cellular activity is being used more for elongation than proliferation. Therefore, information derived from the structure of such models suggests that the resulting serial passage predictive models are useful not only for early detection of future cell performance but also for quantitative extraction of descriptively recordable morphological rules. Therefore, this descriptive understanding of morphological profiles helps us move away from the old habit of perceiving morphological changes by sensation.

[0101] When the same condition matrix was tested using the nonlinear machine learning model RF, this model did not outperform LASSO under any of the data conditions (Fig. 4d). These data reflect the linear relationship between serial passage ability and morphological information.

[0102] While we examined the combination of two parameters (time window and number of FOVs), we ultimately attempted to confirm the predictive performance of the LASSO model in more detail because the model training was based on raw sample data that was relatively small in size compared to other fields of machine learning applications.

[0103] To evaluate the robustness of the prediction model against the influence of data variability, we introduced a bootstrap method to increase the variability of the image-derived morphological profiles. We verified the performance of the prediction model by introducing 50 bootstraps and collecting various FOV combinations from 64 tiling images per sample (Fig. 5).

[0104] The results indicate that the bias in image sampling caused by bootstrapping led to a decrease in the performance of some prediction models. However, the range of the "time window effect" and "number of FOVs effect" for achieving high-performance models (RMSE < 1.0) was narrowed by the more robust models compared to Figure 4a-c. However, the data collection scale could be minimized while maintaining prediction accuracy (RMSE < 1.0), with the earliest prediction requiring 40 FOVs (6-30 hours) and the least-effort prediction requiring 15 FOVs (6-90 hours).

[0105] When evaluating the effect of such bootstrapping in PCA, the number of FOVs clearly improved the robustness of the morphological profiles. Accumulating morphological profiles over a longer time window improved the discriminatory power of clonal differences. Such data size effect studies will contribute to the design of more effective processes for incorporating image-based quality checks in cell bank construction.

[0106] Discussion: RECs are clonal MSCs selected from human BMMCs. They not only retain the excellent quality of conventionally processed MSCs, but also possess advantageous characteristics for practical cell manufacturing processes for therapeutic products. In particular, the high proliferation capacity of RECs offers a significant advantage in developing efficient manufacturing processes for cell therapy products. Therefore, in this study, we investigated the possibility of predicting serial passage performance from initial morphological information alone, and the most practical method for their construction, so that REC performance can be evaluated from the early stages of cell bank construction.

[0107] A long-standing unresolved issue in any type of cell culture is determining the optimal time to generate a cryostock during expansion. Because most normal cells lose their proliferation potential when cultured in vitro [38, 39], creating a cryostock requires gambling on the passage number at which the cell will cease to grow. This gambling factor poses significant risks when building a production cell bank. Failure to grow cells as expected during a low-success expansion culture can result in significant costs, and premature cryopreservation of cells can result in a stock that is not productive.

[0108] In reality, the bottleneck in establishing a cell bank is the effort to recruit valuable donors, not the effort to greedily select candidate cell banks. However, REC is expected to implement a more rigorous and selective process to finalize candidate cell banks as "master cell banks." Our study presents a novel concept of using non-invasive morphological analysis as an "in-process analytical tool" to enhance and optimize the cell bank establishment process. This concept balances "cell yield" and "maintenance of performance of banked cells" and helps determine the optimal time for cryostock creation by predicting future serial passage potential. Such an approach will help abandon the current cell bank design concept, which uses dataless logic to limit the number of passages.

[0109] Although we have investigated methods for predicting the "serialization potential" of RECs, the serialization potential of cMSCs may require some discussion. Regarding induced pluripotent stem cells (iPSCs), it is understood that uncontrolled proliferation of iPSCs can have adverse effects on therapy (e.g., risk of teratoma formation) [40, 41]. However, although MSCs are known to have limited proliferation potential, their serialization potential is considered in some aspects.

[0110] If the "candidate cell bank" from this study is created as a master cell bank for the generation of further working cell banks, its serial passage ability will be beneficial to the entire process. However, if it is created as a final cell bank for transplantation, the effect of serial passage ability should be carefully verified. It can be a benefit if it increases the efficacy of the final product, but it is a risk if it negatively affects efficacy.

[0111] In any case, making such future performance predictions from the earliest stages of the process will help optimize the quality of the final cell bank, since this prediction can only be assessed by excessive serial evaluation of futile REC candidates. As RECs are now on the path to clinical trials, our next challenge is to validate such performance predictions and efficiently proceed with product manufacturing.

[0112] Finally, future performance predictions based on REC morphology will trigger discussions about whether this developed model can be applied to other studies of establishing MSC cell banks. Our detailed image-based morphological measurements showed that the major population of REC is composed of cells that are nearly two times smaller than bulk MSCs. This large size difference allows us to fit a predictive model structure, which is a combination of morphological descriptors, to REC. This is because we used "mean and SD" to reflect the cell population distribution in the morphological profile.

[0113] In Example 2, the population distribution of the entire cell population was compared. The results are shown in Figures 8 to 16.

[0114] Figure 8 shows the results of a comparison of the population distribution of entire cell populations. Clonal MSCs (fast-proliferating mesenchymal stem cells characterized by the positivity or negativity of at least one of the markers LNGFR (CD271) and Thy-1 (CD90)) were cultured with bulk MSCs (mesenchymal stem cells not characterized by these markers). Phase-contrast microscopic images obtained from the culture vessel were processed to describe the population distribution of several thousand cells in the field of view. Six figures show six types of cell morphological features, and the population distribution was plotted using these one-dimensional features. The population distributions of clonal MSCs and bulk MSCs differed significantly for each morphological feature. This figure differs from Figure 3 in terms of image quality, time series, and image processing method, and was measured using more stable cell image data.

[0115] Figure 9 shows the results of a comparison of population distribution across cell populations. Clonal MSCs (fast-proliferating mesenchymal stem cells characterized by the positive or negative status of at least one of the markers LNGFR (CD271) and Thy-1 (CD90)) were cultured (three lots each) with bulk MSCs (mesenchymal stem cells not characterized by these markers). Phase-contrast microscopic images obtained from the culture vessels were processed to depict the population distribution of tens of thousands of cells in the field of view as a scatter plot combining two indices rather than a single indicator. Figure 9 shows the differences between lots (differences in rows) and the differences in time-series data (differences in columns). By comparing clonal MSCs with bulk MSCs in two dimensions (two indices), it was possible to confirm the existence of multiple subpopulations (multiple distributions) in bulk MSCs. This image differs from Figure 3 in terms of image quality, time series time, and image processing method, and measurements were taken from more stable cell image data.

[0116] Figure 10 shows the results of a comparison of the population distribution of the entire cell population. Clonal MSCs (fast-proliferating mesenchymal stem cells characterized by the positive or negative expression of at least one of the markers LNGFR (CD271) and Thy-1 (CD90)) were cultured with bulk MSCs (mesenchymal stem cells not characterized by these markers). Phase-contrast microscopic images obtained from the culture vessel were processed to show the time-dependent changes in the population distribution of tens of thousands of cells in the field of view. The horizontal axis represents cell size. A comparison over time reveals differences in the omnidirectional distribution and the time periods at which this occurs between the left (clonal MSCs) and right (bulk MSCs). The y axis represents the time series (6-hour intervals). Clonal MSCs (RECs) maintained a single distribution over time (the higher the distribution, the later the culture period). Meanwhile, bulk MSCs (commercially available MSCs) initially showed two nodules that appeared to fuse or spread. In other words, it is clear that "cell populations are complex and contain multiple subpopulations."

[0117] Figure 11 shows the results of evaluating the proliferation rate, degree of adipocyte differentiation, and degree of osteogenic differentiation of each MSC. Ninety-one lots of clonal MSCs (fast-growing mesenchymal stem cells characterized by the positive or negative expression of at least one of the markers LNGFR (CD271) and Thy-1 (CD90)) and 21 lots of bulk MSCs (mesenchymal stem cells not characterized by these markers) were cultured, and (i) the proliferation rate after 7 days of culture, (ii) the degree of adipocyte differentiation after 2 weeks, and (iii) the degree of osteogenic differentiation after 2 weeks were evaluated. The proliferation rate was determined by counting the number of cells through image processing of time-lapse images. For adipocyte differentiation, cells were stained with Oil Red O after 2 weeks of culture in differentiation-inducing medium, and the differentiation potential score was calculated as the sum of the red-stained areas in the images. For osteogenic differentiation, cells were stained with Alizarin Red S after 2 weeks of culture in differentiation-inducing medium, and the differentiation potential score was calculated as the sum of the red-stained areas in the images. The results for all samples were then tabulated and displayed in ascending order of the degree of (i), (ii), and (iii). As a result, it was possible to obtain a large amount of data showing diverse qualities (proliferation, differentiation ability) from clonal MSCs that were not subjected to strong selection, but even if one tried to assemble a variety of variations by purchasing bulk MSCs, the comprehensiveness was very low (the blue bars show a bias in the possible quality values).

[0118] Figure 12 shows the quality scores for the degree of fat differentiation and the degree of osteogenic differentiation as bar graphs, sorted by proliferation rate from the data in Figure 11 . Figure 12 reveals that there is no simple correlation between a high proliferation rate and a high differentiation rate, or a low proliferation rate and a low differentiation rate, nor is there a simple correlation between the data, such as a high proliferation rate having a strong influence on the degree of bone or fat differentiation. This suggests that accumulating a variety of clones is effective for preparing large quantities of cell quality, and that simply collecting fast-growing clones is not enough to cover a wide range of qualities. Furthermore, because the quality ((i) (ii) (iii)) of cells generally cannot be determined without purchasing them, it is impossible to obtain data variation based solely on prior data. This demonstrates the effectiveness of clonal MSCs for obtaining data variation. Furthermore, it can be seen that clonal MSCs can cover a much wider variety of qualities than bulk MSCs. In particular, their proliferation capacity can achieve qualities far exceeding those attainable with bulk MSCs.

[0119] Figure 13 shows the results of measuring the proliferation rate of each MSC by time-course image analysis. Twenty lots of clonal MSCs (fast-growing mesenchymal stem cells characterized by the positivity or negativity of at least one of the markers LNGFR (CD271) and Thy-1 (CD90)) and 20 lots of bulk MSCs (mesenchymal stem cells not characterized by these markers) were cultured, and the proliferation rates (days 1 to 5) were measured by time-course image analysis (the horizontal axis of the scatter plot shows the actual measured values). Thereafter, each cell lot was treated as one sample, and a machine learning model was constructed using a dataset in which the cell morphology information of each sample (360 parameters (= population statistics of 12 morphological indicators for several thousand cells (mean and SD = 2 types) × 15 time periods (= 90 hours)) was used as the explanatory variable and the proliferation rate was used as the objective variable, and the proliferation rate after 5 days was predicted. Figure 13 shows that the machine learning model for the proliferation rate prediction model performed using clonal MSCs had a smaller error (RMSE value), demonstrating the effectiveness of the efficiency of building a machine learning model using clonal MSC data.

[0120] Figure 14 shows the results of measuring the proliferation rate of each MSC by time-course image analysis. Seventy-five samples of clonal MSCs (fast-growing mesenchymal stem cells characterized by the positivity or negativity of at least one of the markers LNGFR (CD271) and Thy-1 (CD90)) and 18 lots of bulk MSCs (mesenchymal stem cells not characterized by these markers) were cultured, and the proliferation rates (days 1 to 5) were measured by time-course image analysis (the horizontal axis of the scatter plot shows the actual measured values). Subsequently, using data from only the cloned MSCs, a machine learning model was constructed using a dataset in which the cell morphology information of each sample (360 parameters (= population statistics of 12 morphological indicators for several thousand cells (mean and SD = 2 types) × 15 time periods (= 90 hours)) was used as the explanatory variable and the proliferation rate was used as the objective variable, resulting in a model that could predict the proliferation rate after 5 days (gray dots in the scatter plot). Subsequently, the explanatory variables of the 360 ​​indicators obtained from bulk MSCs were used as input values ​​to construct a proliferation rate prediction model using cloned MSCs (black dots in the scatter plot). It was found that the proliferation rate of bulk MSC samples, which were not used for learning at all and were not cloned MSCs, could be predicted with high accuracy from morphological information alone.

[0121] Figure 15 shows the results of measuring the proliferation rate of each MSC by image analysis over time. Clonal MSCs (fast-proliferating mesenchymal stem cells characterized by the positivity or negativity of at least one of the markers LNGFR (CD271) and Thy-1 (CD90)) were cultured (separate lots were cultured for each experiment, for a total of 75 culture lots), and the proliferation rate (days 1 to 5) was measured by image analysis over time (the horizontal axis of the scatter plot shows the actual measured values). When a machine learning model (explanatory variable: morphological information (360 parameters (= population statistics of 12 morphological indicators for several thousand cells (mean and SD = 2 types) × 15 time periods (= 90 hours)), objective variable: proliferation rate) was constructed for each lot (different experimental days), it was found that a low prediction error (RMSE value) could be obtained for each experimental day (top row). However, when all of the data was simply mixed and a proliferation prediction model was constructed for all 75 culture lots, the prediction error deteriorated significantly. This indicates that even though each image data was taken with the exact same automated culture observation device, it contained data bias for each experimental day, and the performance of the machine learning model created from simply integrated data was poor.

[0122] Figure 16 shows the results of data integration. In the data integration shown in Figure 15, we performed preprocessing on each data set, combining data curation, noise reduction, and data normalization. As a result, we found that the prediction error (RMSE value) of the machine learning model was lower and more stable (right figure) than when this preprocessing was performed (left figure). This indicates that even if each image data set was captured using the exact same automated culture observation device, it contains data bias for each experimental day, resulting in poor performance for the machine learning model created using simply integrated data.

[0123] References 1. Marofi F, Vahedi G, Biglari A, Esmaeilzadeh A, Athari SS. Mesenchymal stromal / stem cells: a new era in the cell-based targeted gene therapy of cancer. Front Immunol. 2017;8:1770. https: / / doi.org / 10.3389 / fimmu.2017.01770. 2. Sherman LS, Shaker M, Marriotti V, Rameshwar P. Mesenchymal stromal / stem cells in drug therapy: new perspective. Cytotherapy. 2017;19(1):19 - 27. https: / / doi.org / 10.1016 / j.jcyt.2016.09.007. 3. Squillaro T, Peluso G, Galderisi U. Clinical trials with mesenchymal stem cells: an update. Cell Transplant. 2016;25(5):829 - 48. https: / / doi.org / 10.3727 / 096368915X689622. 4. Wakao S, Kitada M, Kuroda Y, Shigemoto T, Matsuse D, Akashi H, Tanimura Y, Tsuchiyama K, Kikuchi T, Goda M, Nakahata T, Fujiyoshi Y, Dezawa M. Multilineage-differentiating stress-enduring (Muse) cells are a primary source of induced pluripotent stem cells in human fibroblasts. Proc Natl Acad Sci USA. 2011;108(24):9875−80.https: / / doi.org / 10.1073 / pnas.11008 16108. 5.Vogel W,Grunebach F,Messam CA,Kanz L,Brugger W,Buhring HJ.Heterogeneity among human bone marrow−derived mesenchymal stem cells and neural progenitor cells.Haematologica.2003;88(2):126−33. 6.Trivanovic D,Jaukovic A,Popovic B,Krstic J,Mojsilovic S,Okic−Djordjevic I,Kukolj T,Obradovic H,Santibanez JF,Bugarski D.Mesenchymal stem cells of different origin:comparative evaluation of proliferative capacity,telomere length and pluripotency marker expression.Life Sci.2015;141:61−73.https: / / doi.org / 10.1016 / j.lfs.2015.09.019. 7.Sacchetti B,Funari A,Remoli C,et al.No identical“mesenchymal stem cells”at different times and sites:human committed progenitors of distinct origin and differentiation potential are incorporated as adventitial cells in microvessels.Stem Cell Reports.2016;6(6):897−913.https: / / doi.org / 10.1016 / j.stemcr.2016.05.011. 8.Bakopoulou A,Apatzidou D,Aggelidou E,Gousopoulou E,Leyhausen G,Volk J,Kritis A,Koidis P,Geurtsen W.Isolation and prolonged expansion of oral mesenchymal stem cells under clinical−grade,GMP−compliant conditions differentially affects“stemness”properties.Stem Cell Res Ther.2017;8(1):247.https: / / doi.org / 10.1186 / s13287−017−0705−0. 9.Colter DC,Class R,DiGirolamo CM,Prockop DJ.Rapid expansion of recycling stem cells in cultures of plastic−adherent cells from human bone marrow.Proc Natl Acad Sci USA.2000;97(7):3213−8.https: / / doi.org / 10.1073 / pnas.97.7.3213. 10.Mabuchi Y,Morikawa S,Harada S,Niibe K,Suzuki S,Renault−Mihara F,Houlihan DD,Akazawa C,Okano H,Matsuzaki Y.LNGFR+THY−1+VCAM−1hi+cells reveal functionally distinct subpopulations in mesenchymal stem cells.Stem Cell Reports.2013;1(2):152−65.https: / / doi.org / 10.1016 / j.stemcr.2013.06.001. 11.Ukeba D,Yamada K,Tsujimoto T,Ura K,Nonoyama T,Iwasaki N,Sudo H.Bone marrow aspirate concentrate combined with in situ forming bioresorbable gel enhances intervertebral disc regeneration in rabbits.J Bone Joint Surg.2021;103(8):e31.https: / / doi.org / 10.2106 / JBJS.20.00606. 12.De Bari C,Roelofs AJ.Stem cell−based therapeutic strategies for cartilage defects and osteoarthritis.Curr Opin Pharmacol.2018;40:74−80.https: / / doi.org / 10.1016 / j.coph.2018.03.009. 13.Kabat M,Bobkov I,Kumar S,Grumet M.Trends in mesenchymal stem cell clinical trials 2004−2018:is efficacy optimal in a narrow dose range?Stem Cells Transl Med.2020;9(1):17−27.https: / / doi.org / 10.1002 / sctm.19−0202. 14.Viswanathan S,Keating A,Deans R,Hematti P,Prockop D,Stroncek DF,Stacey G,Weiss DJ,Mason C,Rao MS.Soliciting strategies for developing cell−based reference materials to advance mesenchymal stromal cell research and clinical translation.Stem Cells and Development.2014;23(11):1157−67.https: / / doi.org / 10.1089 / scd.2013.0591. 15.Dwarshuis NJ,Parratt K,Santiago−Miranda A,Roy K.Cells as advanced therapeutics:state−of−the−art,challenges,and opportunities in large scale biomanufacturing of high−quality cells for adoptive immunotherapies.Adv Drug Deliv Rev.2017;114:222−39.https: / / doi.org / 10.1016 / j.addr.2017.06.005. 16.Jossen V,van den Bos C,Eibl R,Eibl D.Manufacturing human mesenchymal stem cells at clinical scale:process and regulatory challenges.Appl Microbiol Biotechnol.2018;102:3981−94.https: / / doi.org / 10.1007 / s00253−018−8912−x. 17.Lo Surdo J,Bauer SR.Quantitative approaches to detect donor and passage differences in adipogenic potential and clonogenicity in human bone marrow−derived mesenchymal stem cells.Tissue Engineering−Part C:Methods.2012;18(11):877−89.https: / / doi.org / 10.1089 / ten.tec.2011.0736. 18.Mohamed−Ahmed S,Fristad I,Lie SA,Suliman S,Mustafa K,Vindenes H,Idris SB.Adipose−derived and bone marrow mesenchymal stem cells:a donor−matched comparison.Stem Cell Res Ther.2018;9(1):168.https: / / doi.org / 10.1186 / s13287−018−0914−1. 19.Fennema EM,Renard AJS,Leusink A,Van Blitterswijk CA,De Boer J.The effect of bone marrow aspiration strategy on the yield and quality of human mesenchymal stem cells.Acta Orthop.2009;80(5):618−21.https: / / doi.org / 10.3109 / 17453 67090 32782 41. 20.Collart−Dutilleul P−Y,Chaubron F,De Vos J,Cuisinier FJ.Allogenic banking of dental pulp stem cells for innovative therapeutics.World Journal of Stem Cells.2015;7(7):1010−21.https: / / doi.org / 10.4252 / wjsc.v7.i7.1010. 21.Nievaleve S.Concise review:umbilical cord derived mesenchymal stem cell bank.Progress in Stem Cell.2017;4:3−4.https: / / doi.org / 10.15419 / psc.v4i3−4.397. 22.Alzahrani FA,Saadeldin IM,Ahmad A,Kumar D,Azhar EI,Siddiqui AJ,Kurdi B,Sajini A,Alrefaei AF,Jahan S.The potential use of mesenchymal stem cells and their derived exosomes as immunomodulatory agents for COVID−19 patients.Stem Cells International.2020;2020:8835986.https: / / doi.org / 10.1155 / 2020 / 88359 86. 23.Rombouts WJC,Ploemacher RE.Primary murine MSC show highly efficient homing to the bone marrow but lose homing ability following culture.Leukemia.2003;17(1):160−70.https: / / doi.org / 10.1038 / sj.leu.24027 63. 24.Dhanasekaran M,Indumathi S,Poojitha R,Kanmani A,Rajkumar JS,Sudarsanam D.Plasticity and banking potential of cultured adipose tissue derived mesenchymal stem cells.Cell Tissue Banking.2013;14(2):303−15.https: / / doi.org / 10.1007 / s10561−012−9311−7. 25.Dave C,McRae A,Doxtator E,Mei SHJ,Sullivan K,Wolfe D,Champagne J,McIntyre L.Comparison of freshly cultured versus freshly thawed(cryopreserved)mesenchymal stem cells in preclinical in vivo models of inflammation:a protocol for a preclinical systematic review and meta−analysis.Syst Rev.2020;9(1):188.https: / / doi.org / 10.1186 / s13643−020−01437−z. 26.Hladik D,Hofig I,Oestreicher U,Beckers J,Matjanovski M,Bao X,Scherthan H,Atkinson MJ,Rosemann M.Long−term culture of mesenchymal stem cells impairs ATM−dependent recognition of DNA breaks and increases genetic instability.Stem Cell Res Ther.2019;10(1):218.https: / / doi.org / 10.1186 / s13287−019−1334−6. 27.Jin HJ,Bae YK,Kim M,Kwon SJ,Jeon HB,Choi SJ,Kim SW,Yang YS,Oh W,Chang JW.Comparative analysis of human mesenchymal stem cells from bone marrow,adipose tissue,and umbilical cord blood as sources of cell therapy.Int J Mol Sci.2013;14(9):218.https: / / doi.org / 10.3390 / ijms1 40917 986. 28.Yang YHK,Ogando CR,Wang See C,Chang TY,Barabino GA.Changes in phenotype and differentiation potential of human mesenchymal stem cells aging in vitro.Stem Cell Res Ther.2018;9(1):131.https: / / doi.org / 10.1186 / s13287−018−0876−3. 29.Wagner W,Horn P,Castoldi M,Diehlmann A,Bork S,Saffrich R,Benes V,Blake J,Pfister S,Eckstein V,Ho AD.Replicative senescence of mesenchymal stem cells:a continuous and organized process.PLoS ONE.2008;3(5):e2213.https: / / doi.org / 10.1371 / journ al.pone.0002213. 30.Turinetto V,Vitale E,Giachino C.Senescence in human mesenchymal stem cells:functional changes and implications in stem cell−based therapy.Int J Mol Sci.2016;17(7):1164.https: / / doi.org / 10.3390 / ijms1 70711 64. 31.Binato R,de Souza FT,Lazzarotto−Silva C,Du Rocher B,Mencalha A,Pizzatti L,Bouzas LF,Abdelhay E.Stability of human mesenchymal stem cells during in vitro culture:considerations for cell therapy.Cell Prolif.2013;46(1):10−22.https: / / doi.org / 10.1111 / cpr.12002. 32.Galipeau J,Sensebe L.Mesenchymal stromal cells:clinical challenges and therapeutic opportunities.Cell Stem Cell.2018;22(6):824−33.https: / / doi.org / 10.1016 / j.stem.2018.05.004. 33.Samsonraj RM,Raghunath M,Nurcombe V,Hui JH,van Wijnen AJ,Cool SM.Concise review:multifaceted characterization of human mesenchymal stem cells for use in regenerative medicine.Stem Cells Transl Med.2017;6(12):2173−85.https: / / doi.org / 10.1002 / sctm.17−0129. 34.Imai Y,Yoshida K,Matsumoto M,Okada M,Kanie K,Shimizu K,Honda H,Kato R.In−process evaluation of culture errors using morphology−based image analysis.Regenerative Therapy.2018;9(9):15−23.https: / / doi.org / 10.1016 / j.reth.2018.06.001. 35.Matsuoka F,Takeuchi I,Agata H,Kagami H,Shiono H,Kiyota Y,Honda H,Kato R.Morphology−based prediction of osteogenic differentiation potential of human mesenchymal stem cells.PLoS ONE.2013;8(2):e55082.https: / / doi.org / 10.1371 / journ al.pone.00550 82. 36.Sasaki H,Takeuchi I,Okada M,Sawada R,Kanie K,Kiyota Y,Honda H,Kato R.Label−free morphology−based prediction of multiple differentiation potentials of human mesenchymal stem cells for early evaluation of intact cells.PLoS ONE.2014;9(4):e93952.https: / / doi.org / 10.1371 / journ al.pone.00939 52. 37.Takemoto Y,Imai Y,Kanie K,Kato R.Predicting quality decay in continuously passaged mesenchymal stem cells by detecting morphological anomalies.J Biosci Bioeng.2020;131(2):198−206.https: / / doi.org / 10.1016 / j.jbiosc.2020.09.022. 38.Sawada R,Ito T,Tsuchiya T.Changes in expression of genes related to cell proliferation in human mesenchymal stem cells during in vitro culture in comparison with cancer cells.J Artif Organs.2006;9(3):179−84.https: / / doi.org / 10.1007 / s10047−006−0338−z. 39.Muraglia A,Cancedda R,Quarto R.Clonal mesenchymal progenitors from human bone marrow differentiate in vitro according to a hierarchical model.J Cell Sci.2000;113(7):1161−6.https: / / doi.org / 10.1242 / jcs.113.7.1161. 40.Okita K,Ichisaka T,Yamanaka S.Generation of germline−competent induced pluripotent stem cells.Nature.2007;448(7151):313−7.https: / / doi.org / 10.1038 / natur e05934. 41.Gutierrez−Aranda I,Ramos−Mejia V,Bueno C,Munoz−Lopez M,Real PJ,Macia A,Sanchez L,Ligero G,Garcia−Parez JL,Menendez P.Human induced pluripotent stem cells develop teratoma more efficiently and faster than human embryonic stem cells regardless the site of injection.Stem Cells.2010;28(9):1568−70.https: / / doi.org / 10.1002 / stem.471.;

Claims

1. A standard for evaluating or determining the quality of test cells, comprising a rapidly proliferating mesenchymal stem cell population characterized by the positive or negative expression of at least one of the markers LNGFR (CD271) and Thy-1 (CD90).

2. The reference material according to claim 1, wherein the characteristic of the cell population is selected from any of the following information (a) to (d): (a) at least one piece of cell information selected from the group consisting of cell size, degree of extension, complexity of contour, pattern and brightness of each cell in the cell population, and image information thereof; (b) information obtained by accumulating the cell information of (a) across multiple cells in the cell population and performing statistical processing; (c) information obtained by combining the information of (b) with respect to multiple cell populations; and (d) information obtained over time from the information of (a) to (c).

3. The standard according to claim 2, wherein the statistically processed information is at least one selected from the group consisting of average value, standard deviation, and distribution shape.

4. The reference product according to claim 1, wherein the quality is at least one selected from the group consisting of cell proliferation rate, number of cells recovered, degree of undifferentiation, degree of differentiation, amount of produced factors, and cell heterogeneity.

5. The standard described in claim 1, wherein the test cells are stem cells.

6. The standard preparation according to claim 5, wherein the test stem cells are test mesenchymal stem cells.

7. A method for providing information for quality assessment of a test stem cell population, comprising the steps of: (a) characterizing individual cells of a fast-proliferating mesenchymal stem cell population using as an indicator the positivity or negativity of at least one marker selected from the group consisting of LNGFR (CD271) and Thy-1 (CD90), and acquiring image data (reference image information) of the individual cells, or acquiring parameter information characteristic of the cell population (reference parameter information) from the reference image information; (b) acquiring image data (test image information) of individual cells of a test stem cell population provided by a user, or acquiring parameter information characteristic of the test stem cell population (test parameter information) from the test image information; (c) comparing the test image information or test parameter information with the reference image information or reference parameter information; and (d) providing all or a part of the information obtained by the comparison to a user for quality assessment of the test stem cell population.

8. A method for evaluating the quality of a test stem cell population, comprising the steps of: (a) characterizing individual cells of a fast-proliferating mesenchymal stem cell population using as an index the positivity or negativity of at least one marker selected from the group consisting of LNGFR (CD271) and Thy-1 (CD90), and acquiring image data of the individual cells (reference image information), or acquiring parameter information characteristic of the cell population (reference parameter information) from the reference image information; (b) acquiring image data of individual cells of a test stem cell population provided by a user (test image information), or acquiring parameter information characteristic of the test stem cell population (test parameter information) from the test image information; (c) comparing the test image information or test parameter information with the reference image information or reference parameter information; and (d) evaluating the quality of the test stem cells using all or a part of the information obtained by the comparison as an index.

9. An information provision system for quality assessment of a test stem cell population, comprising the following means: (a) means for characterizing individual cells of a fast-proliferating mesenchymal stem cell population using as an indicator the positive or negative result of at least one of the markers LNGFR (CD271) and Thy-1 (CD90), and acquiring image data (reference image information) of the individual cells, or acquiring parameter information characteristic of the cell population (reference parameter information) from the reference image information; (b) means for acquiring image data (test image information) of individual cells of a test stem cell population provided by a user, or acquiring parameter information characteristic of the test stem cell population (test parameter information) from the test image information; (c) means for comparing the test image information or test parameter information with the reference image information or reference parameter information; and (d) means for providing all or a part of the information obtained by the comparison to a user for quality assessment of the test stem cell population.

10. A system for evaluating the quality of a test stem cell population, comprising: (a) a means for characterizing individual cells of a fast-proliferating mesenchymal stem cell population using as an index the positivity or negativity of at least one marker selected from the group consisting of LNGFR (CD271) and Thy-1 (CD90), and for acquiring image data of the individual cells (reference image information), or acquiring parameter information characteristic of the cell population (reference parameter information) from the reference image information; (b) a means for acquiring image data of individual cells of a test stem cell population provided by a user (test image information), or acquiring parameter information characteristic of the test stem cell population (test parameter information) from the test image information; (c) a means for comparing the test image information or test parameter information with the reference image information or reference parameter information; and (d) a means for evaluating the quality of the test stem cells using all or a part of the information obtained by the comparison as an index.

11. The system according to claim 9 or 10, wherein the characteristic of the cell population is selected from any of the following information (a) to (d): (a) at least one piece of cell information selected from the group consisting of cell size, degree of extension, complexity of contour, pattern and brightness of each cell in the cell population, and image information thereof, (b) information obtained by accumulating the cell information of (a) across multiple cells in the cell population and performing statistical processing, (c) combined information obtained by acquiring the information of (b) for multiple cell populations, and (d) the information of (a) to (c) acquired over time.

12. The system according to claim 11, wherein the statistically processed information is at least one selected from the group consisting of an average value, a standard deviation, and a distribution shape.

13. The system according to claim 9 or 10, wherein the quality is at least one selected from the group consisting of cell proliferation rate, cell recovery number, degree of undifferentiation, degree of differentiation, amount of production factor, and cell heterogeneity.

14. The system according to claim 9 or 10, wherein the test stem cell is a test mesenchymal stem cell.

15. An information provision program for quality assessment of a test stem cell population, the program causing a computer to function as: (a) a means for characterizing individual cells of a fast-proliferating mesenchymal stem cell population using the positive or negative result of at least one of the markers LNGFR (CD271) and Thy-1 (CD90) as an indicator, and for acquiring image data (reference image information) of the individual cells, or acquiring parameter information characteristic of the cell population (reference parameter information) from the reference image information; (b) a means for acquiring image data (test image information) of individual cells of a test stem cell population provided by a user, or acquiring parameter information characteristic of the test stem cell population (test parameter information) from the test image information; (c) a means for comparing the test image information or test parameter information with the reference image information or reference parameter information; and (d) a means for providing all or a part of the information obtained by the comparison to a user for quality assessment of the test stem cell population.

16. A quality assessment program for a test stem cell population, the program causing a computer to function as: (a) a means for characterizing individual cells of a fast-proliferating mesenchymal stem cell population using as an index the positivity or negativity of at least one marker selected from LNGFR (CD271) and Thy-1 (CD90), and for acquiring image data of the individual cells (reference image information), or acquiring parameter information characteristic of the cell population (reference parameter information) from the reference image information; (b) a means for acquiring image data of individual cells of a test stem cell population provided by a user (test image information), or acquiring parameter information characteristic of the test stem cell population (test parameter information) from the test image information; (c) a means for comparing the test image information or test parameter information with the reference image information or reference parameter information; and (d) a means for evaluating the quality of the test stem cells using all or a part of the information obtained by the comparison as an index.

17. The program according to claim 15 or 16, wherein the feature of the cell population is selected from any of the following information (a) to (d): (a) at least one piece of cell information selected from the group consisting of cell size, degree of extension, complexity of contour, pattern and brightness of each cell in the cell population, and image information thereof, (b) information obtained by accumulating the cell information of (a) across multiple cells in the cell population and performing statistical processing, (c) combined information obtained by acquiring the information of (b) for multiple cell populations, and (d) the information of (a) to (c) acquired over time.

18. The program according to claim 17, wherein the statistically processed information is at least one selected from the group consisting of an average value, a standard deviation, and a distribution shape.

19. The program according to claim 15 or 16, wherein the quality is at least one selected from the group consisting of cell proliferation rate, cell recovery number, degree of undifferentiation, degree of differentiation, amount of production factor, and cell heterogeneity.

20. The program according to claim 15 or 16, wherein the test stem cell is a test mesenchymal stem cell.

21. A computer-readable recording medium having the program according to claim 15 or 16 recorded thereon.

22. An image processing method for rapidly proliferating mesenchymal stem cells, comprising the steps of: (a) acquiring image data of the standard specimen described in claim 1; and (b) removing noise from the acquired image data and / or normalizing the image data.

23. A method for providing image data processed by the method of claim 22 for machine learning in a computer.

24. A method for predicting the quality of a test cell by comparing image data processed by the method according to claim 22 with image data of the test cell.

Images