International evaluation study of a highly efficient culture assay for detection of residual human pluripotent stem cells in cell therapies

Human pluripotent stem cells (hPSCs) have the potential to revolutionize regenerative medicine. However, the potential risk of tumorigenicity due to residual undifferentiated hPSCs persisting in the final product remains a particular concern, because hPSCs are intrinsically tumorigenic and can form teratomas [1,2]. The establishment of a robust and internationally harmonized methodology to evaluate the contamination of a cell therapy product with residual undifferentiated hPSCs is needed not only for product developers but also for regulatory authorities and patients [1].

A minimum number of hPSCs sufficient to induce teratoma in immunodeficient animals varies and depends on the administration procedure. For instance, the 50% tumor producing dose – that is, a minimal number of transplanted cells that generated teratomas in 50% of animals – is reported to be about 100 human induced pluripotent stem cells (hiPSCs) when cells were subcutaneously injected with Matrigel into severely immunodeficient mice [3]. On the other hand, 1 × 10⁵ human embryonic stem cells or 2 × 10⁵ hiPSCs were reported to be needed to induce teratomas when cells were injected intramyocardially [4] or intravenously [5], respectively. In clinical applications, it is presumed that approximately 10⁹ cells need to be transplanted into a patient to treat heart failure [6], and for chimeric antigen receptor-based cell therapies, 10¹⁰ cells may be administered [7]. The recommended sensitivity for the detection of residual undifferentiated hPSCs would be different for each product and depends on the administration route/environment of the injection site and the intended clinical dose level. However, for instance, if a higher clinical dose level is intended, leveraging sensitive in vitro assays will be required, considering the limitation of maximum feasible dose in in vivo studies.

In response to this emerging safety issue, the Health and Environmental Sciences Institute (HESI) Cell Therapy-TRAcking, Circulation & Safety (CT-TRACS) Technical Committee launched an international project to evaluate in vitro testing methods focusing on the detection of residual undifferentiated hPSCs [1,8] in collaboration with a Japanese experimental public–private partnership initiative (Multisite Evaluation Study on Analytical Methods for Non-clinical Safety Assessment of Human-Derived Regenerative Medical Products [MEASURE]).

Several in vitro testing methods have been reported for the detection of residual hPSCs: flow cytometry [9], the Glycostem test [10], quantitative reverse transcription PCR [9], droplet digital PCR [7,11,12] and highly efficient culture (HEC) assay [13,14]. In regard to the sensitivity of testing methods, the limit of detection (LOD) of assays is relatively high for flow cytometry and the Glycostem test (0.05–0.1%) [9,10]. On the other hand, PCR-based assays are highly sensitive (LOD: 0.0003–0.001%) [7,9,11,12], and recently an even more sensitive methodology based on reverse transcription loop-mediated isothermal amplification was reported (LOD: 0.00002%) [15]. However, some types of hiPSC-derived cells are known to retain the expression of pluripotency marker genes during the differentiation process [12,13]; therefore, the marker genes for these molecular biology-based assays need to be carefully validated prior to the assay for each cell therapy product (CTP) to avoid false positive results. The HEC assay is a simple, culture-based assay that can directly detect hPSCs by identifying hPSC-derived colonies under culture condition that favor the growth of hPSCs [13]. Its high sensitivity and robustness were confirmed and compared with PCR-based assays in a multisite study previously organized by MEASURE (LOD: 0.001%) [14]. That study further demonstrated that a detection sensitivity of 0.00002% could be achieved by adding a step consisting of enriching the targeted undifferentiated hPSCs by utilizing a magnetic-activated cell sorting system [14].

Considering the wide variety of hPSC-derived CTPs (e.g., direction and degree of differentiation), it is important that several sensitive methodologies for detecting residual hPSCs be validated at the international level and available to be leveraged by product developers based on which will best serve their specific use, context and needs. Therefore, CT-TRACS launched an international collaborative project for assay validation focusing on two methods – the HEC assay and ddPCR-based assay. In the present study, the authors evaluated the sensitivity and interlaboratory variability of the HEC assay at multiple facilities according to International Organization for Standardization (ISO) standards. This international, multisite study serves to confirm previous results published by the MEASURE program in Japan. The performance of the HEC assay was evaluated in the previous MEASURE study by spiking hiPSCs into somatic cells such as human mesenchymal stromal cells and human peripheral blood mononuclear cell-derived T cells, but not into hPSC-derived cells. Therefore, hiPSC-derived cardiomyocytes were selected in the present study, since they more closely mimic hPSC-derived CTPs.

Materials & methods

Cells

For hiPSCs, Cellartis human iPS cell line 18 (ChiPSC18) cells were purchased from Takara Bio, Inc., and were maintained with the Cellartis DEF-CS culture system (Takara Bio, Inc.) according to the manufacturer's instructions. The cultured ChiPSC18 (passages 19–22) were frozen in Stem-Cellbanker (Zenoaq Resource Co. Ltd) and cryopreserved in liquid nitrogen. The frozen hiPSCs were thawed and used for experiments after two to four passages. For hiPSC-derived cardiomyocytes, iCell cardiomyocytes (#C1106, Lot#103700) were purchased from FUJIFILM Cellular Dynamics, Inc., and were used for experiments immediately after thawing using iCell cardiomyocyte plating medium (#M1001; FUJIFILM Cellular Dynamics, Inc.) according to the manufacturer's instructions.

HEC assay

After thawing, the iCell cardiomyocytes were suspended with Essential 8 Flex medium (Thermo Fisher Scientific) containing 50 nM Chroman 1 (#HY-15392; MedChem Express), 5 μM emricasan (#S7775; Selleckchem), polyamine supplement (#P8483, 1:1000 dilution; Sigma-Aldrich) and 0.7 μM trans-ISRIB (#5284, Tocris; E8F+CEPT [16] medium). These cells were seeded in six-well plates coated with laminin-521 (BioLamina AB) [13] at a cell density of 1 × 10⁶ per well after checking the viability by cell counter, Countess automated cell counter (Thermo Fisher Scientific) or LUNA-FL dual fluorescence cell counter (Logos Biosystems) or manually using a counting chamber. The ChiPSC18 cells were dissociated into single cells with ×1 TrypLE Express and were suspended with E8F+CEPT medium. After confirming the viability (>∼80%) in the same manner, 5 × 10¹ cells/ml of the ChiPSC18 cell suspension was carefully prepared by serial dilution from 5 × 10⁵ cells/ml of the ChiPSC18 cell suspension at a ratio of 10. The 0.1 or 0.3 ml of 5 × 10¹ cells/ml of the ChiPSC18 cell suspension was added to the wells previously seeded with iCell cardiomyocytes to spike 5 (0.0005%) or 15 (0.0015%) hiPSCs into 1 × 10⁶ iCell cardiomyocytes. Four wells were prepared for each concentration, and wells in which only 1 × 10⁶ iCell cardiomyocytes were seeded (nonspiked sample) were prepared for each experiment (n = 1 or 2). Three days after seeding, the media were replaced with Essential 8 Flex medium, and after a total of 7 days of incubation at 37°C in 5% CO₂, the wells were fixed and stained using a Vector Blue alkaline phosphatase (ALP) substrate kit (Vector Laboratories, Inc.) according to a previously reported method [14]. 7 days in culture was sufficient to detect the hiPSC colonies, considering that the doubling time of hiPSC when cultured with Essential 8 medium on laminin-521-coated wells was approximately 21.5 h [13]. The ALP-positive hiPSC colonies were counted manually by two operators under a microscope, and the colony formation rate (the ratio of the total number of colonies to the number of spiked hiPSCs) was calculated. The experiment was conducted at four facilities (facilities A, B, C and D) and was repeated three-times at each facility except for facility A, where the experiment was repeated two times.

Statistical analysis

The true positive rate (TPR) for detection of hiPSC colonies with the HEC assay was calculated based on the assumption that the number of hiPSC colonies formed on wells follows a Poisson distribution, as previously reported [14]. The probability of no hiPSC colony formation (k = 0; i.e., the false negative rate [FNR]) in a single experiment was expressed using the mean number of actual hiPSC colonies obtained by repeated experiments (λ = n):

The FNR of a set of two experiments was calculated as the probability of no hiPSC colony formation in a single experiment (e^-n) raised to the second power:

The probability of hiPSC colony formation (k >0; i.e., TPR) in a single experiment and of a set of three experiments led to the calculation 1 - FNR:

In this study, use of the term “TPR” was identical to “clinical/diagnostic sensitivity” as defined in ISO/TS 17822-1:2014 [17], whereas the term ‘sensitivity’ was used in a broad sense.

The repeatability and reproducibility of the number of hiPSC colonies were statistically analyzed, according to ISO 5725 [18,19]. The repeatability is defined as precision under conditions where independent test results are obtained with the same method on identical test items in the same laboratory by the same operator using the same equipment within short intervals of time, and reproducibility is defined as precision under conditions where test results are obtained with the same method on identical test items in different laboratories with different operators using different equipment [20]. Variability from a collaborative validation study can be modeled as:

where s_R² is the reproducibility variance, s_L² is the between-laboratory variance and s_r² is the repeatability variance.

The value of s_r² is calculated by

where p is the number of laboratories, n_i is the number of test results in the i-th laboratory and s_i is the standard deviation of the test results in the i-th laboratory.

The value of s_L² is calculated by:

where:

To identify outliers, Mandel's k and h statistics, Cochran's statistics and Grubbs' statistics were used as consistency tests. The critical values of less than 5% were used as criteria for outliers.

Results

For all facilities, well-defined ALP-positive hiPSC colonies could be detected in both 5 (0.0005%) and 15 (0.0015%) hiPSC-spiked conditions (Figure 1). The mean number of hiPSC colonies showed a slight variation among the facilities. Colonies for 5 and 15 hiPSC-spiked conditions were 2.7 (2.1–3.3) and 7.9 (5.1–12.1), respectively (Table 1), and revealed a good dependency on the number of spiked hiPSCs. The colony formation rate calculated as the ratio of the total number of colonies to the number of spiked hiPSCs was similar between the 5 and 15 hiPSC-spiked conditions, and the overall colony formation rate was confirmed to be 54% (42–67%) for the 5 hiPSC-spiked condition and 53% (34–81%) for the 15 hiPSC-spiked condition (Table 1). For nonspiked samples, no colonies were observed in any wells at any of the facilities.

To confirm the TPR needed to detect spiked hiPSCs under the experimental conditions, the authors used the Poisson distribution formula and calculated the probability of detecting hiPSC colonies, with one or more colony present. The TPR of a single experiment varied slightly between facilities when spiking with five (0.0005%) hiPSCs (88–96%), but the mean value at all the facilities was sufficiently high (92%), and spiking with 15 (0.0015%) hiPSCs gave a TPR of 99% or higher at all facilities (Table 1). Two repeated experiments increased the TPR for detection of hiPSC colonies, and a TPR of 100% was also calculated in most facilities when spiking with five cells (Table 1).

The statistical analysis on repeatability and reproducibility of the number of hiPSC colonies revealed that the standard deviation of repeatability (s_r) and reproducibility (s_R) was 1.0 and 1.0 for the 5 hiPSC-spiked condition and 1.9 and 3.5 for the 15 hiPSC-spiked condition (Table 2). The coefficients of variation (CV) of repeatability and reproducibility were both 36% for the 5 hiPSC-spiked condition and 24% and 44% for the 15 hiPSC-spiked condition, and the ratio of s_R/s_r was 1.00 and 1.80 for each condition, respectively (Table 2). There were no excluded outliers from the consistency test for both the 5 and 15 hiPSC-spiked conditions.

Discussion

During the manufacturing process of hPSC-derived CTPs, most undifferentiated hPSCs are likely eliminated because most differentiation protocols are not optimal for the survival of undifferentiated hPSCs [21]. However, efforts to purify the final product and to minimize its contamination with undifferentiated hPSCs are critical to mitigate the potential tumorigenicity risks of hPSC-derived CTPs [1]. The HEC assay is a simple and sensitive culture-based assay and has, in fact, been employed in the tumorigenicity assessment of hPSC-derived CTPs, under the requirements of the regulatory agencies [22].

In the present study, the authors evaluated the sensitivity of the HEC assay at international and multisite levels and confirmed that all four facilities in the study could detect hiPSC colonies in wells in which 5 hiPSCs were spiked into 1 × 10⁶ iCell cardiomyocytes (LOD: 0.0005%). The mean number of hiPSC colonies showed an increase depending on the number of spiked hiPSCs between the 5 and 15 hiPSC-spiked conditions. The overall colony formation rate was constant among the facilities and was higher than 50% regardless of the number of spiked hiPSCs. Residual undifferentiated hPSCs in the final product are considered ‘contaminants’ or ‘impurities’ to the final differentiated CTP [1]. Therefore, it is important to assess both the LOD and the specificity of the assay methods to detect such residual cells, as outlined in International Conference on Harmonisation (ICH) Q2 (R1) [23]. In terms of the specificity of the assay, the basis of the HEC assay is to directly detect the residual hPSCs by identifying the colony derived from each residual hPSC with ALP activities. Actually, no colonies stained with an ALP substrate were detected in nonspiked samples at the four facilities, which indicates no false positive signals in this study and results in high specificity of the HEC assay. Additionally, its specificity was previously secured by visually identifying the colonies positively stained with antibodies and a lectin against several hPSC markers [14] and will also be confirmed with spiked samples prepared at each assay for the determination of the LOD [22]. For confirmation of the validity of LOD in the present study, the authors calculated the TPR based on the FNR, which is the probability of no colony formation in hiPSC-spiked samples. As a result, the mean value of TPR for detecting 0.0005% hiPSC contamination was 92% even with a single experiment and approached 100% with two repeated experiments in all facilities. These results indicate the high sensitivity and the robustness of the HEC assay to detect residual hiPSCs and that a negative result obtained from a product with an unknown contamination rate means that the hiPSC content is less than 0.0005% at a probability of about 90% or higher even with a single experiment. If the experiment were repeated, the probability would be almost 100%. This suggests that the assay appears to be capable of an even lower detection limit, but that would have to be experimentally verified.

Interestingly, the sensitivity with which the authors could detect undifferentiated cells in a differentiated population is highly superior to the sensitivity reported from in vivo teratoma assays. The results show that at least 5 hiPSC cells in a 1 million dose can be detected. While there are conflicting reports on how many hPSCs are needed to form a teratoma in vivo, the lowest reported numbers from Kanemura et al. [3] and Yasuda et al. [24] showed that the 50% tumor producing dose was 132 and 631, respectively, when 201B7 hiPSC line was subcutaneously transplanted with Matrigel into NOD/Shi-scid IL2Rγnull mice. The authors of the present study speculate that with further validation and evaluation of the HEC assay, it will be possible to take full advantage of this increased sensitivity, providing more precise safety predictions of hPSC-derived CTPs and reduced use of animals in preclinical testing.

The statistical analysis on the number of hiPSC colonies to estimate the precision of HEC methodology revealed that the CV values of repeatability (CV [s_r]) and reproducibility (CV [s_R]) were 36% and 36% for the five hiPSC-spiked condition and 24% and 44% for the 15 hiPSC-spiked condition, respectively. These results show the precision of the HEC assay within an international, multisite study for the first time. Repeatability is used to evaluate the ability of the method to generate similar results with the successive measurements. As expected, a higher number of hiPSCs spiked into iCell cardiomyocytes lowered the CV value of repeatability. Although the CV value of reproducibility with 15-cell spiking was higher than that of 5-cell spiking, the s_R/s_r ratio was quite improved with 15-cell spiking. It should be noted that s_R was calculated as the same value of s_r (i.e., s_R/s_r is 1) based on the formula just because repeatability variance was very high in the 5-cell spiking condition. These results would help clarify the value and limitations of the HEC assay for evaluating CTPs. The CV (s_r) and CV (s_R) values were the same for the 5 hiPSC-spiked conditions; therefore, the main part of the reproducibility was considered to be repeatability. In addition, an increase in the ratio of s_R to s_r showed a lower contribution of repeatability to reproducibility when more hiPSCs (15 cells) were spiked. This indicates that possible variation in the actual number of spiked hiPSCs may cause the variances of repeatability and reproducibility, especially when a smaller number of hiPSCs is spiked. Thus, accurate execution of both preparation of hiPSC cell suspension and spiking of these cells into the samples is essential for the sensitivity and precision of the HEC assay.

Since the sensitivity of the HEC assay is determined by the hPSC-spiked sample, the efficiency of the colony formation from each spiked hPSC is highly important to lower the detection limit. Additionally, insufficient efficiency of colony formation may cause false negative results when hPSC-derived CTPs are evaluated; therefore, securing a higher colony formation rate is essential for this assay. It is well known that enzymatic dissociation causes cell death by apoptosis and anoikis [25–27], and a landmark paper has reported that the ρ-associated coiled coil-forming protein serine/threonine kinase inhibitor Y-27632 improves the survival of hPSC [28]. Recently, Chen et al. reported that the combination of chroman 1, emricasan, polyamines and trans-ISRIB (CEPT) enhanced cell survival of hPSCs by simultaneously blocking several stress mechanisms, and its effects on cytoprotection of hPSCs were superior to that of Y-27632 [16]. Therefore, CEPT, instead of Y-27632, was used in the present study. The mean values of colony formation rate (54% ± 19%, mean ± standard deviation, under the five hiPSC-spiked condition) was higher than that in the MEASURE study using Y-27632 (31% ± 21%, mean ± standard deviation, under the 10 hiPSC [ChiPSC18]-spiked condition [14]), indicating that CEPT treatment has beneficial effects on the efficiency of colony formation in the HEC assay, though further experiments, especially experiments using a head-to-head comparison, are necessary.

In the present study, the experiments were performed using hiPSC-derived cardiomyocytes only. However, it has been reported that the HEC assay can be applied for various cell types, including human mesenchymal stromal cells [13,14], human neurons [13] and hPSC-derived neural stem cells [22]. Furthermore, the MEASURE study using human peripheral blood mononuclear cell-derived T cells indicates that the HEC assay can also be applied for suspension cell products [14] with some modifications (the assay uses adherent cultures and should be optimized for cell suspensions). Therefore, it is expected that the HEC assay can be broadly applied for hPSC-derived CTPs with, perhaps, the exception of the CTPs that inhibit survival of undifferentiated hPSCs such as hiPSC-derived differentiated retinal pigment epithelial cells that secret pigment epithelium-derived factor and induce apoptotic cell death of hPSCs [29].

Conclusion

The HESI CT-TRACS international, multisite study using hiPSC-derived cardiomyocytes was able to prove the high sensitivity of the HEC assay (LOD: 0.0005%) and confirm that the mean TPR for detecting 0.0005% hiPSC contamination was sufficiently high (92%) even when looking at partial data from a single experiment. The findings indicate the robustness of the HEC assay for the evaluation of product contamination with residual undifferentiated hPSCs and that this in vitro methodology can be generally applicable for the evaluation of the potential tumorigenicity risk of hPSC-derived CTPs. It is considered to be a valuable in vitro approach to be included in a ‘toolbox’ for tumorigenicity assessment, which will contribute to the future standardization of tumorigenicity risk assessment of hPSC-derived CTPs.