Tumorigenicity Studies of Induced Pluripotent Stem Cell (iPSC)-Derived Retinal Pigment Epithelium (RPE) for the Treatment of Age-Related Macular Degeneration

Basic studies of human pluripotential stem cells have advanced rapidly and stem cell products are now seeing therapeutic applications. However, questions remain regarding the tumorigenic potential of such cells. Here, we report the tumorigenic potential of induced pluripotent stem cell (iPSC)-derived retinal pigment epithelium (RPE) for the treatment of wet-type, age-related macular degeneration (AMD). First, immunodeficient mouse strains (nude, SCID, NOD-SCID and NOG) were tested for HeLa cells’ tumor-forming capacity by transplanting various cell doses subcutaneously with or without Matrigel. The 50% Tumor Producing Dose (TPD50 value) is the minimal dose of transplanted cells that generated tumors in 50% of animals. For HeLa cells, the TPD50 was the lowest when cells were embedded in Matrigel and transplanted into NOG mice (TPD50 = 101.1, n = 75). The TPD50 for undifferentiated iPSCs transplanted subcutaneously to NOG mice in Matrigel was 102.12; (n = 30). Based on these experiments, 1×106 iPSC-derived RPE were transplanted subcutaneously with Matrigel, and no tumor was found during 15 months of monitoring (n = 65). Next, to model clinical application, we assessed the tumor-forming potential of HeLa cells and iPSC 201B7 cells following subretinal transplantation of nude rats. The TPD50 for iPSCs was 104.73 (n = 20) and for HeLa cells 101.32 (n = 37) respectively. Next, the tumorigenicity of iPSC-derived RPE was tested in the subretinal space of nude rats by transplanting 0.8–1.5×104 iPSC-derived RPE in a collagen-lined (1 mm×1 mm) sheet. No tumor was found with iPSC-derived RPE sheets during 6–12 months of monitoring (n = 26). Considering the number of rodents used, the monitoring period, the sensitivity of detecting tumors via subcutaneous and subretinal administration routes and the incidence of tumor formation from the iPSC-derived RPE, we conclude that the tumorigenic potential of the iPSC-derived RPE was negligible.


Introduction
Clinical cell therapy trials were recently initiated for treatment of Stargardt's disease and the dry type of age-related macular degeneration (dry AMD). The trials have used human embryonic stem cell (hESC)-derived retinal pigment epithelium (RPE) [1][2][3][4]. In addition, several groups are planning clinical trials with autologous human induced pluripotent stem cell (hiPSC)-derived RPE for the wet type of AMD. Thus, cell therapy using human pluripotent stem cells (hPSCs) has reached clinical application. However, in contrast to tissue stem cells that have a limited proliferation potential, tumor formation from residual undifferentiated or incompletely differentiated hPSCs in hPSC-derived cell products is an issue that must be carefully analyzed. This issue is particularly important when transplanting autologous hiPSCderived cells.
We recently reported a highly sensitive residual hiPSC detection method based on qRT-PCR using primers for the LIN28A transcript [5] in hiPSC-derived RPE. This method enables us to detect residual hiPSCs down to 0.002% of differentiated RPE cells. As we plan to transplant 4-8610 4 hiPSC-derived RPE cells into the subretinal space of patients, this method is sensitive enough to detect a few residual hiPSCs, if any, in a clinical setting.
The tumorigenic potential of hiPSC-derived RPE cells is attributable to contamination by undifferentiated hiPSCs, intermediate products having proliferation potentials and/or tumorigenic transformed cells. Contamination by these cells should be assessed by nonclinical testing using suitable animal models [6,7]. However, there is no internationally recognized guideline for tumorigenicity testing in cell therapy products. The most relevant guideline is the WHO TRS 878, ''Recommendation for the evaluation of animal cell cultures as substrates for the manufacture of cell banks'' [8,9]. The guideline recommends transplanting 1610 7 test cells subcutaneously to 10 nude mice and monitoring tumor formation for more than 16 weeks. Transplantation of the same dose of a well-known tumorigenic cell line such as HeLa in parallel is suggested as a tumor-forming positive control. The WHO guideline covers animal cell substrates for the production of biological medicinal products and specifically excludes viable animal cells that are intended for therapeutic transplantation into patients. To examine the tumorigenicity of hiPSC-derived cells intended for administration to patients, several teratoma-forming tests exploring dose and administration route were studied using immuno-deficient mice [6], [10]. However, discussions how we can interpret and extrapolate the results of tumorigenicity testing with immuno-deficient or immuno-suppressant animals to human patients continue [6,7]. Recently a commentary report from FDA/CBER pointed out the issues to be considered for cell-based products and associated challenges for preclinical animal study [11]. The report stated that although the nature of cells used for cellular therapy is diverse, tumorigenic test results from the administration of cells through nonclinical routes would not be considered relevant as it would not track the behavior of transplanted cells in a micro-environment. When tumorigenicity testing of ESC-derived cellular products is undertaken, the study design should include groups of animals that have received undifferentiated ESCs, serial dilutions of undifferentiated ESCs combined with ESC-derived final products and the final intended clinical products. This approach would thereby address the tumorforming potential of these cell groups in animal models.
Tumorigenicity testing via the clinical route of administration could recapitulate the fate of transplanted cells in a microenvironment of host tissue and could be fairly extrapolated to human application. However, elaborate surgical intervention requires skills that greatly influence the outcome of transplantation. For example, it is difficult to determine whether the cells were transplanted into the right location or organ in small rodents. These concerns can be overcome by conducting a subcutaneous tumorigenicity test in addition to testing via the clinical route.
In this report, we conducted 2 types of in vivo tumorigenicity tests by transplanting hiPSC-derived RPE cells into subcutaneous and sub-retinal spaces in immuno-deficient animals. The results and limits of these tests are discussed.

Tumorigenicity Tests with Several Types of Immunodeficient Mice
The tumor-forming potential of human iPSC-derived cell products should be examined using a suitable animal transplantation model. One should take into account the number of cells to be transplanted, the method of transplantation, the microenvironment of the transplantation site, the monitoring period and the status of the immune-deficient animals.
First, we checked the tumor-forming potential of several immune-deficient animals by subcutaneously transplanting HeLa cells over a wide range of doses (1 to 1610 6 in 10-fold increments) with or without Matrigel (BD) and observed tumor formation every day for up to 36 weeks on a daily basis. Matrigel is known to enhance the tumor-forming potential of transplanted cells [16]. Recipient animals included immune-deficient nude, SCID [10], NOD-SCID [17], and NOG [18] mice. The minimal dose of transplanted cells that generated tumors in 50% of the transplanted animal (TPD 50 ) was calculated statistically to evaluate the sensitivities of tumor formation in each animal model [19]. We found the NOG mouse was most susceptible to tumors. That is, when transplanted subcutaneously with Matrigel, tumors were generated by the lowest number of HeLa cells. The TPD 50 for HeLa was 10 1.1 (n = 75), in agreement with a previous report [20], (Figure 1, Table 1). It is interesting to note that among the conditions tested, the highest number of HeLa cells was required to form tumors in nude mice without Matrigel. TPD 50 for nude mice without Matrigel was 10 4.9 (n = 120), which is also in agreement with a previous report [19] (Figure 1, Table 1). Therefore, we selected NOG mice and Matrigel for embedding the test cells for further assays as it provided sensitive tumor detection using small numbers of transplanted cells. The tumor-formation potential of iPSCs was assessed by subcutaneously transplanting several doses of the iPSC cell line 201B7 with Matrigel into NOG mice. The TPD 50 for iPSC was 10 2.12 (n = 30) over 12 months' monitoring ( Table 2). The TPD 50 value for iPSCs in subcutaneous transplantation provided a reference cell number for the contamination of iPSCs in iPSC-derived RPE cells. Retinal differentiation was subsequently initiated. The resulting RPE cell lines were established as follows. 59-G3 RPE was derived from hiPSC clone 59-G3; 101-EV3 RPE was from 101-EV3; K11-EV9 RPE was from K11-EV9; K21-EV15 RPE was from K21-EV15; K21-G18 RPE was from K21-G18; 101-G25 RPE and RNT9 RPE were from RNT9-2-8; and, RNT10 RPE was from RNT10-24. The protocol for RPE differentiation from hiPSC was shown in our recent report [21]. It requires 3 months for RPE differentiation and another 2 months to prepare the RPE sheet. The following quality control tests for the hiPSC-derived RPE cell lines were conducted. (1) The EBNA plasmid fragment was not detectable by PCR. (2) The cells showed the characteristic morphology and pigmentation of RPE with a single or double layer cell structure. (3) BEST1 and PAX6 molecules were detected by immunohistochemistry in over 95% of final hiPSC-derived RPE cells. (4) RPE-specific markers RPE65, CRALBP, MERTK and BEST1 were confirmed by RT-PCR. (5) LIN28A was not detected by qRT-PCR. (6) Migration of non-RPE cells into the collagen layer lining the hiPSC-derived RPE cell sheet shall be below 0.1% of the total RPE cells. (7) The RPE cell sheet shall consist of over 70% viable cells with a density of over 4500 cells/mm 2 . Items (6) and (7) were quality control tests for the RPE cell sheet. All the cell culture processes including establishment of hiPSCs from a patient's fibroblasts and differentiation to RPE were conducted in a GMP-grade cell processing facility. The morphology and immunostaining of hiPSC-derived RPE cell lines 59-G3 RPE, K21-G18 RPE and 101-G25 RPE are shown in Figure 2A, B. The other hiPSC-derived RPE cell lines showed the same phenotype. The gene expression patterns of these cell lines are shown in Figure 2C. Primary RPE was used as a reference. It is notable that neither LIN28A nor POU5F1 (OCT3/4) was detected above background levels in hiPSC-derived RPE cells 5 . This finding serves as a useful criterion to eliminate immature hiPSCs in hiPSC-derived RPE ( Figure 2D, 2E).

Tumorigenicity Testing of hiPSC-derived RPE (Subcutaneous Transplantation)
To assess the tumorigenic potential of hiPSC-derived RPE cells with high sensitivity, subcutaneous transplantation of a large number of RPE cells would be ideal. However, the maximal number of RPE cells available for transplantation was limited by the culture capacity of the cell processing facility. We hypothesized that transplanting 1610 6 hiPSC-derived RPE cells was an acceptable cell number to address the tumorigenic potential of the final cell product when embedded in Matrigel in NOG mice. This hypothesis was based on the facts that we expect to transplant 4-8610 4 hiPSC-derived RPE in a clinical setting and as few as 10 undifferentiated hiPSCs or HeLa cells embedded in Matrigel could generate tumors in NOG mice ( Table 1, Table 2).
Thus, we subcutaneously transplanted 1610 6 hiPSC-derived RPE cells embedded in Matrigel into NOG mice [total n = 42; 59-G3 RPE (n = 14), K21-G18 RPE (n = 13), 101-G25 RPE (n = 15)]. Tumor formation was monitored for more than 70 weeks. Teratoma derived from subcutaneously transplanted iPSCs was analyzed as a positive control for tumor formation event ( Figure 3A-3E). The proliferative status of living cells was assessed by HE, Hoechst 33258 and anti-Ki67 antibody staining ( Figure 3G-3I). iPSC-derived neural rosette-like human cells were stained by anti-Lamin A antibody to check the specificity of anti-Lamin A antibody for human cells ( Figure 3F). We assumed that  anti-human Lamin A antibody could stain a wide range of human cell types and was not limited to human RPE. Transplantation of 1610 6 primary RPE cells embedded in Matrigel (n = 3) was used as a transplantation control. No tumor formation was observed from transplanted 1610 6 hiPSC-derived RPE of several origins in various administration forms. All of the subcutaneous tumorigenicity tests conducted for hiPSC-derived RPE under various conditions using NOG mice are shown in Table 2.
All subcutaneous transplants consisting of RPE cells embedded in Matrigel were excised and subjected to histological examination. The size of transplants in subcutaneous tissue ( Figure 4A, 4B) was similar to that of Matrigel without RPE ( Figure 4C). Histological and immunohistological study showed that Lamin A-, BEST1-and Hoechst-positively staining RPE cells were present in all the Matrigel transplants ( Figure 4F-4M). None of the cells transplanted in Matrigel stained with anti-Ki67 antibody, suggesting the absence of active proliferation in these transplanted cells ( Figure 4D, 4E). Human cells derived from transplanted iPSC-derived RPE could be detected by Alu PCR at a level of $0.1% in mouse cells. However, we could not detect human cells in subcutaneous mouse tissue just beneath the transplants, in liver, spleen, kidney or lung by this method ( Figure 5A, 5B).

Tumorigenicity Test of hiPSC-derived RPE (Subretinal Transplantation)
Next, we conducted tumorigenicity tests by transplanting test cells into the subretinal space, a procedure that is technically demanding. We chose large albino nude rats to facilitate transplantation and minimize variability of test results. This choice also permitted us to transplant larger doses of human hiPSC-derived RPE cells to the subretinal spaces. First, we assessed the tumor-formation potential of HeLa and iPSC 201B7 via subretinal transplantation. The TPD 50 for HeLa was 10 1.32 (n = 37) and 10 4.73 for iPSC (n = 20) (Table 3) Next, we conducted tumorigenicity tests of iPSC-derived RPE by transplanting 1 mm 2 -sized (1 mm61 mm) iPSC-derived RPE sheets consisting of 0.8-1.5610 4 RPE cells into the subretinal space of nude rats (n = 26). The RPE cell number was assessed by cell density and sheet size transplanted. Considering the relative sizes of humans and rats, we estimated that transplanting 0.8-1.5610 4 RPE cells into the subretinal space of nude rats would provide the information required to determine the incidence of tumor formation in humans, as we expect to transplant 4-8610 4 RPE cells in a clinical setting. Thus, we transplanted 5 different hiPSC-derived RPE cell sheets from 5 different patients to minimize individual variations. hiPSC-derived RPE sheets [59-G3 RPE (n = 9), K21-G18 RPE (n = 4), 101-G25 RPE (n = 3), RNT9 RPE (n = 5), RNT10 RPE (n = 5)] were prepared and transplanted under various conditions. Transplanted nude rats were monitored for tumor formation and physical condition daily for 8 to 82 weeks.
No tumor was found during the period of observation (Table 3). All transplanted eye balls were excised and subjected to histological examination. The location of transplanted RPE sheet was shown by the brown color of the RPE sheet in albino nude rats ( Figure 7A-7B). Histological and immunohistological studies showed that Lamin A-and Hoechst-positively staining or BEST1and Hoechst-positively staining transplanted RPE cells were present in the subretinal space ( Figure 7H-7O). Although we used serial sections for this staining, we believe more than half of Lamin A positive-human cells were stained with Hoechst, suggesting that these cells were live human transplanted cells at

Discussion
Here, we presented the results of nonclinical tests assessing the tumorigenic potential of hiPSC-derived RPE sheets. These studies represent a portion of the nonclinical testing of our scheduled clinical study for the use of autologous hiPSC-derived RPE sheets for the treatment of wet type AMD. The clinical study is scheduled to commence in 2014. The hiPSC-derived RPE cells used in this study were prepared in a GMP-grade cell processing facility using the same procedures that will be used for patient treatment.
Two types of tumorigenicity tests are summarized in this report. The first was a subcutaneous tumorigenicity test in NOG mice using Matrigel and the second was a subretinal tumorigenicity test in nude rats. It is intriguing to compare the objectives and the validity of the 2 tests in regard to hiPSC-derived RPE cell transplantation. Rationales for conducting subcutaneous tumorigenicity test are as follows: 1. Large numbers of test cells can be transplanted without difficulty, and bias-inducing variations in technical skills can be neglected. Moreover, the tumors are easy to detect. Therefore, statistical and endpoint analysis (TPD 50 assessment) can be conducted in a timely and accurate manner.
2. It is possible to conduct comparison studies of the tumorforming potential of different cellular products under the same transplantation conditions. 3. Above all, this test could serve as a substitute for in vitro soft agar assays of PSCs. PSCs cannot survive in soft agar, so that mode of testing is not feasible [5]. In contrast, PSCs or PSCderived cells can survive long-term (more than 12 months) in Matrigel when subcutaneously transplanted in NOG mice. We can detect tumors derived from as few as 10 iPSCs or HeLa cells in this system ( Table 1, Table 2). Thus, it provides a highly sensitive tumorigenic test for detecting both residual hiPSCs and tumorigenic transformed cells in hiPSC-derived cell products.
In this context, subcutaneous transplantation testing can be considered a quality control test of final cell products to ensure the absence of tumorigenic cells rather than characterizing the tumorigenic potential of the final cell products at a clinical transplantation site.
Next, we conducted tumorigenicity tests of hiPSC-derived RPE via clinical administration route. Under physiological conditions, the RPE is a monolayer that secretes various cytokines to maintain its structure in the retina. Diniz et al [3] reported that RPE transplanted in sheets retain better survival than when transplanted in suspension. For clinical application, we plan to transplant hiPSC-derived RPE in a sheet form and our preclinical testing was designed accordingly. Although we do not have RPE cell survival data in suspension form, it would be logical to presume that the transplantation of RPE in a sheet form exerts physiological function more effectively than in suspension, which may also facilitate the adaption of transplanted cells to the subretinal tissues. With regard to the tests' ability to detect immature (undifferentiated) hiPSCs in the subretinal space as a growing tumor, we demonstrated the trans-effects of RPE on hiPSC in our recent studies. We reported that RPE secreted pigment epithelium derived-factor (PEDF) that markedly induced apoptosis in hiPSC and hESC [21]. hiPSCs or ESCs in culture inserts ceased to survive when co-cultured with RPEs. Further addition of hPEDF induced apoptosis in hiPSCs or ESCs drastically. In fact, when hiPSCs were transplanted into the subretinal space of nude rats, the log 10 TPD 50 value was 4.73 (n = 20), whereas the value was only 2.12 (n = 30) when hiPSCs alone were transplanted subcutaneously into NOG mice with Matrigel. The 400-fold difference in the TPD 50 values under these conditions are at least partly explained by an environmental effect related to the subretinal   space, besides the difference in the status of immunodeficiency in these species or use of Matrigel. We suggest that the environmental effects of the subretinal space are mediated by PEDF secreted by RPE. The close protein sequence identity between human PEDF and the rat counterpart support this idea.
As many as 1610 4 hiPSC cells were required to form tumors in the subretinal space of nude rats. Similar numbers (0.8-1.5610 4 ) of hiPSC-derived RPE were transplanted into the subretinal space in the tumorigenicity test. Note that tumorigenicity testing of iPSC-derived RPE via clinical administration route will always give ''negative'' results, if we aim to detect a tumor from remaining small number of undifferentiated hiPSCs in final product. In this context, tumorigenicity tests conducted by transplanting serial dilutions of hiPSCs combined with hiPSC-derived RPEs into the subretinal space might not be informative. We suggest that tumorigenicity testing via clinical administration route might be useful to detect tumors from intermediate or incompletely differentiated RPE cells, but not for those relatively rare remaining  undifferentiated iPSC. For these reasons, we conducted high dose (1610 6 ) subcutaneous RPE transplantation in parallel to examine tumor-forming events from rare hiPSC in hiPSC-derived RPE and full dose hiPSC-derived RPE subretinal transplantation without diluting them with hiPSCs. This was the basis for our rationale in designing multiple tumorigenicity tests for iPSC-derived RPE. As the FDA commentary report 11 stated, the design of tumorigenicity tests should be tailored for each specific product. We hope our approach will facilitate a further discussion related to tumorigenicity testing of iPSC-derived cell products. Considering the number of rodents used, the duration of the monitoring period, the sensitivity to detect tumors in immunodeficient rodents via both subcutaneous and subretinal administration routes and the overall incidence of tumor formation from iPSC-derived RPE final cell products in these rodents, we conclude that the tumorigenic potential of the hiPSC-derived RPE cells produced by our methods is negligible. Of course, in considering the overall safety of the procedure in humans, discussion should include the site of transplantation as well as the source of the cells (autologous or allogeneic) and the immunesuppression status of the patients.

Materials and Methods
All the experiments using human samples and animal studies were reviewed and approved by the IRB of the Foundation for Biomedical Research and Innovation (FBRI) and Riken Center for Developmental Biology (Riken CDB), and the committee for animal experiments of the FBRI.

RT-PCR and qRT-PCR
Total RNA was isolated with the RNeasy plus Mini Kit (Qiagen) in accordance with the manufacturer's instructions. Contaminating genomic DNA was removed using a gDNA Eliminator spin column. cDNA was generated from 50 ng of total RNA using PrimeScript RT Master Mix (Takara Bio) and PrimeSTAR MAX DNA Polymerase (TaKaRa Bio). Real-time PCR was then performed with an ABI 7000 Sequence Detection System (Applied-Biosystems) and SYBR-green in accordance with the manufacturer's instruction. Gene expression levels were normalized to that of GAPDH. qRT-PCR was performed using the QuantiTect Probe one-step RT-PCR Kit (Qiagen). The expression levels of target genes were normalized to those of the RNase P transcript, which were quantified using TaqMan human RNase P control reagents (Applied Biosystems). All qRT-PCR reactions were run for 45 cycles. The sequences of primers and probes used in the present study are listed in Table S1.

Alu PCR
Alu sequences specific to human cells were used to design the primers. The Alu primer 59-AAGTCGCGGCCGCTTGCAGT-GAGCCGAGAT-39 and 50 ng of DNA template, PrimeSTAR Max DNA Polymerase (Takara) were used for PCR reactions (28 cycles). The DNA templates in various ratios (human HeLa DNA: mouse NIH3T3 DNA) were used to determine the human cell detection sensitivity by Alu PCR. PCR products were separated by electrophoresis (MyRun, Cosmobio) with 1% agarose gel (Nacalai), and the image was digitally captured (Bio-Pyramid, Mecan).

Conclusion
We tested the tumorigenic potential of hiPSC-derived RPE using immuno-deficient rodents. These preclinical tests laid the foundation for upcoming clinical studies using autologous hiPSCderived RPE sheets for treatment of wet type age-related macular degeneration (AMD). One million hiPSC-derived RPE cells were transplanted subcutaneously into 65 NOG mice and 0.8-1.5610 4 hiPSC-derived RPE cells were transplanted into the subretinal space of 26 nude rats. No tumors were found after 6-15 months of monitoring. Considering the number of rodents used, the duration of the monitoring period, the sensitivity to detect tumors in immuno-deficient rodents, we conclude that the tumorigenic potential of the hiPSC-derived RPE cells prepared by our method is negligible.