A novel screening test to predict the developmental toxicity of drugs using human induced pluripotent stem cells

Abstract

In vitro human induced pluripotent stem (iPS) cells testing (iPST) to assess developmental toxicity, e.g., the induction of malformation or dysfunction, was developed by modifying a mouse embryonic stem cell test (EST), a promising animal-free approach. The iPST evaluates the potential risks and types of drugs-induced developmental toxicity in humans by assessing three endpoints: the inhibitory effects of the drug on the cardiac differentiation of iPS cells and on the proliferation/survival of iPS cells and human fibroblasts. In the present study, the potential developmental toxicity of drugs was divided into three classes (1: non-developmentally toxic, 2: weakly developmentally toxic and 3: strongly developmentally toxic) according to the EST criteria. In addition, the type of developmental toxicity of drugs was grouped into three types (1: non-effective, 2: embryotoxic [inducing growth retardation/dysfunction]/deadly or 3: teratogenic [inducing malformation]/deadly) by comparing the three endpoints. The present study was intended to validate the clinical predictability of the iPST. The traditionally developmentally toxic drugs of aminopterin, methotrexate, all-trans-retinoic acid, thalidomide, tetracycline, lithium, phenytoin, 5-fluorouracil, warfarin and valproate were designated as class 2 or 3 according to the EST criteria, and their developmental toxicity was type 3. The non-developmentally toxic drugs of ascorbic acid, saccharin, isoniazid and penicillin G were designated as class 1, and ascorbic acid, saccharin and isoniazid were grouped as type 1 while penicillin G was type 2 but not teratogenic. These results suggest that the iPST is useful for predicting the human developmental toxicity of drug candidates in a preclinical setting.

INTRODUCTION

Developmental toxicity caused by drugs, chemicals and pesticides is a serious issue that can affect subsequent generations. Developmental toxicity is roughly divided into embryotoxic (inducing growth retardation or dysfunction), teratogenic (inducing malformation) and death (inducing infertility or stillbirth). However, these toxicities cannot be evaluated in general toxicity studies using animals and are instead assessed in reproductive and developmental toxicity studies, which require an extensive number of animals and a long experiment term. Therefore, simple assays to achieve similar findings are desired, and many in vitro developmental toxicity assays have been developed, such as the mouse embryonic stem cell test (EST) (Annett et al., 2016; Seiler and Spielmann, 2011; Scholz et al., 1999; Genschow et al., 2002, 2004), zebrafish test (Truong et al., 2011) and devTOX^TM test (Palmer et al., 2013) using human embryonic stem cells or human induced pluripotent stem (iPS) cells. We also reported the preliminary method of in vitro iPS cells testing (iPST) (Aikawa et al., 2014) that is a modified version of the EST and a promising animal-free approach.

The EST has been thoroughly studied and is well-understood, and it has been validated by the European Center for the Validation of Alternative Methods (ECVAM). The EST uses two cell lines and three endpoints to predict the developmental toxicity of drugs, chemicals and pesticides. The cell lines are mouse embryonic stem (mES) cells and mouse BALB/c 3T3 clone A31 fibroblast (3T3) cells. The three endpoints of the assay are the 50% inhibition of cardiac differentiation as the beating generation of the embryonic bodies (EBs), mES cell aggregates and the 50% inhibition of the proliferation or survival of mES cells and 3T3 cells for 10 days. The drugs and chemicals are classified into three classes: 1 (non-developmentally toxic), 2 (weakly developmentally toxic) and 3 (strongly developmentally toxic). These classes describe the developmental toxicity potential based on the in vivo effects of drugs and chemicals in animals and/or humans, according to the findings for the linear discriminant functions integrating three endpoints.

Thalidomide is a well-known teratogen, which was launched in over 40 countries in 1957 but was withdrawn from the market in most countries by 1962 due to reports it induced human fetal teratogenicity. The teratogenic effect of thalidomide differs among species (Vargesson, 2015), and was caused in humans and non-human primates at low doses and in rabbits at high doses only but not in hamsters or rodents (Fratta et al., 1965; Schumacher et al., 1968; Teo et al., 2001, 2004). As thalidomide has not been tested using the EST, it was subjected to the commercial EST kit POCA^® Hand1-EST (Suzuki et al., 2011), a modified high-throughput EST. The developmental toxicity was negative, which coincided with the findings in the developmental toxicity study in mice. If human cells were used instead of mouse cells in the EST, the species-specific difference in the drug response would be negated, and the adverse effects of thalidomide would have been detected.

The iPST assesses three endpoints, similar to the EST, to predict the developmental toxicity of the drugs using iPS cells instead of mES cells and adult human dermal fibroblasts (fibroblasts) instead of 3T3 cells. One endpoint is the inhibition concentration of cardiac differentiation for iPS cells, and the others are the cytotoxic concentrations for iPS cells and fibroblasts. Since a cytotoxicity assay is the general method used for culturing iPS cells and fibroblasts in culture medium containing drugs, there are no technical problems. However, a cardiac differentiation assay is considerably difficult to perform because iPS cells, unlike mES cells, do not differentiate spontaneously into cardiomyocytes. The mES cells are differentiated spontaneously into cardiomyocytes after the withdrawal of leukemia inhibitory factor (LIF), which helps maintain the pluripotency of the undifferentiated mES cells, from culture medium (Evans and Kaufman, 1981; Martin, 1981; Wobus et al., 1984). We previously developed a simple protocol to differentiate iPS cells into cardiomyocytes within a week by the application of biological substances (Aikawa et al, 2015), which made it possible to carry out the cardiac differentiation assay. In a previous study, the teratogenicity effect of thalidomide had already been predicted (Aikawa et al., 2014).

The iPST determined the developmental toxicity in two prediction models. One was the EST prediction mode established in the EST, which involved subdivision into three classes which was based on the in vivo effects in animals and/or humans. The other, as the developmental toxicity prediction model, defined the types of developmental toxicity (e.g., malformation, growth retardation and dysfunction) by comparing each endpoint in the three assays similar to the EST. Embryo and fetal development and growth are affected by the direct effects of drugs as well as by indirect effects on pregnant women, in whom a poor physical condition can adversely influence the supply of nutrition, oxygen, hormones and other molecules to the embryo or fetus. Developmental toxicity is a direct action of drugs on the embryo or fetus without affecting the physiological condition of the pregnant woman. In the iPST, which considers iPS cells as the embryo or fetus and fibroblasts as the pregnant woman, the concentration that results in toxicity to the iPS cells but not to the fibroblasts is considered the level of developmental toxicity (Fig. 1). Drugs that are more toxic to the iPS cells than to the fibroblasts have developmental toxicity, and strong toxicity of drugs to the iPS cells differentiation induces teratogenicity or death and strong toxicity to the iPS cells growth/survival induces growth retardation or dysfunction nor death (Fig. 1).

Fig. 1

Effects of developmental toxic signaling on the developmental stage. In the present study, human induced pluripotent stem cells were used instead of anaplastic embryos, and adult human dermal fibroblasts were used instead of human pregnant women. Developmental toxic signals from pregnant women to embryo/fetus include drugs, radiation, hypoxia, malnutrition and virus. Strong developmental toxic signaling induces death of the embryo in the preimplantation period, death of the embryo or malformation of newborns in the organogenesis period, and death of the fetus or dysfunction of newborns in the fetal growth period. Weak developmental toxic signaling induces death of the embryo or normal growth of the embryo in the preimplantation period, malformation of newborns or normal differentiation of the embryo in the organogenesis period, dysfunction of newborns or normal growth of the fetus in the fetal growth period. An abnormal condition of the pregnant women induces developmental toxic signals inhibits embryo implantation, inducing death of the embryo or growth retardation of the fetus. Thick dotted vertical arrows: strong developmental toxicity signal, thin dotted vertical arrows: weak developmental toxicity signal, hatched horizontal arrows: normal differentiation/growth, shaded arrows: abnormal differentiation/growth, solid arrows: death (infertility or stillbirth), open vertical arrows: normal/weak signal. iPS cells: human induced pluripotent stem cells. Fibroblasts: adult human dermal fibroblasts. (The present illustration is modified from Tuchmann-Duplessis, 1965).

The experimental protocol of the iPST in the present study was a partially modified of a previously reported protocol as follows: the drug treatment period was prolonged to 6 days (from 4 days) depending on an cardiac differentiation period, the ReLeSR™ as a cell dissociation reagent was used instead of dispase to equalize the number of cells seeding, and the author’s original prediction criteria, established by assay of the mouse cells, was improved to the classification of the type for developmental toxicity as the developmental toxicity prediction model (Aikawa et al., 2014, 2015). In the present study, the iPST was validated internally using various reference drugs. That is, the accuracies of the EST prediction model (Annett et al., 2016) and developmental toxicity prediction model, established using the mouse cells, were verified. First, as a pre-validation examination, the predictability of the developmental toxicity of drugs in humans with the modified protocol was confirmed using thalidomide, a well-known teratogen with high sensitivity for humans; valproate, a well-known teratogen in almost all animal species; and saccharin as a non-teratogen, as it does not affect the pH of the culture medium, unlike ascorbic acid. Next, several reference drugs for which the developmental toxicity has or has not been reported in experimental animals or in epidemiological surveys were assessed.

MATERIALS AND METHODS

The present study was conducted according to the policy of the Biological Material Handling Committee of our institute.

Test drugs

The drugs shown in Table 1 were obtained from the following sources: all-trans retinoic acid (retinoic acid), lithium carbonate (lithium), methotrexate, sodium valproate (valproate), tetracycline hydrochloride (tetracycline), warfarin and thalidomide from FUJIFILM Wako Pure Chemical (Osaka, Japan); aminopterin from Adooq Bioscience (Irvine, CA, USA); 5,5-diphenylhydantoin (phenytoin), 5-fluorouracil (5-FU) and isoniazid from Sigma-Aldrich Japan (Tokyo, Japan), and ascorbic acid, saccharin and penicillin G potassium salt (Penicillin G) from Nacalai Tesque (Kyoto, Japan). Ascorbic acid, isoniazid, saccharin and penicillin G have not been reported to show developmental toxicity in humans, whereas all other drugs have some degree of developmental toxicity in humans and experimental animals. All drugs were dissolved in appropriate solvent (dimethyl sulfoxide [DMSO] or medium) and then diluted with the same solvent. The final concentration of DMSO was set at ≤ 0.2 vol%, and equal concentrations of DMSO were used in the control and test groups. DMSO at ≤ 0.25 vol% had no effect on the proliferation or viability of iPS or fibroblast cells (unpublished observation).

Table 1. List of drugs used in the study.

Cell lines and cell culture

iPS cells

The iPS cells were generated from human bone marrow mononuclear cells in-house (Kunisato et al., 2011).

The iPS cells were suspended in modified tenneille serum replacer 1 medium (mTeSR™-1; STEMCELL Technologies, Vancouver, Canada), seeded in a cell culture dish (60 mm × 15 mm) coated with Matrigel^® (Corning^® International, Tokyo, Japan) and cultured in a 5% CO₂ incubator (37°C and humidity-controlled).

Fibroblasts

Human dermal fibroblasts-adult, isolated from adult human skin cryopreserved at primary culture, were obtained from ScienCell Research Laboratories (Carlsbad, CA, USA).

Fibroblasts were cultured using a typical procedure for cell lines, as follows: the cells were suspended in a dermal fibroblast growth medium (fibroblast medium; DS Pharma Biomedical, Osaka, Japan), seeded in a cell culture dish (100 mm × 20 mm) coated with collagen type I (Thermo Fisher Scientific, Waltham, MA, USA) and cultured in a 5% CO₂ incubator (37°C and humidity-controlled).

iPST

The iPST processes were shown in Fig. 2.

Fig. 2

Schematic overview of the process of the iPST. The three endpoints of the iPST for assessing the developmental toxicity of drugs are shown here. Two human cell lines were used: iPS cells and adult human dermal fibroblasts. The method of myocardial differentiation of iPS cells was developed for the myocardial differentiation assay of the iPST. The cytotoxicity assay was a widely used in vitro method. iPST: human induced pluripotent stem cell test. iPS cells: human induced pluripotent stem cells. Fibroblasts: adult human dermal fibroblasts. EB: embryonic body (aggregated iPS cells).

Cytotoxicity assay with human iPS cells and human fibroblasts

iPS cells

Subconfluent iPS cell colonies were collected by ReLeSR™ (STEMCELL Technologies) treatment (few minutes, 37°C). After centrifugation (135 × g, room temperature), cell pellets were suspended carefully in fresh mTeSR™-1 at 1 × 10⁵ cells/mL. The cell suspensions were then seeded at 0.1 mL/well in 96-well clear-bottom black cell culture plate (Thermo Fisher Scientific) coated with Corning^® Matrigel^® matrix (Thermo Fisher Scientific) and cultured in a 5% CO₂ incubator (37°C and humidity-controlled). After 3 hr, the media of all wells were carefully discarded, and then mTeSR™-1 containing each concentration of drugs or the vehicle was immediately added at 0.2 mL/well in triplicate and cultured in a 5% CO₂ incubator. Six days later, the cytotoxic effects of the drugs were assessed. The assay medium was replaced with fresh medium of the same composition every day (0.2 mL/well for 5 days, 0.1 mL/well last day).

Fibroblasts

Subconfluent fibroblasts were collected by the 0.05% trypsin-EDTA (1%) (Thermo Fisher Scientific) treatment (few minutes, 37°C). After centrifugation (135 × g, room temperature), cell pellets were suspended carefully in fresh fibroblast medium at 1 × 10⁵ cells/mL. The cell suspensions were seeded 0.1 mL/well in a Corning^® 96-well clear-bottom black cell culture plate (Thermo Fisher Scientific) coated with collagen type 1 and cultured in a 5% CO₂ incubator (37°C and humidity-controlled). After 3 hr, the media of all wells was carefully discarded, and then fibroblast medium containing each concentration of drugs or the vehicle was added at 0.2 mL/well in triplicate and cultured in a 5% CO₂ incubator. Six days later, the cytotoxicity of drug was assessed. The assay medium was replaced with fresh assay medium of the same composition at Day 2, 4 and 5 (0.2 mL/well for 5 days, 0.1 mL/well last day).

Cytotoxicity assay

The viability of cells was assessed using the Cell meter™ colorimetric cell cytotoxicity assay kit (ATT Bioquest, Sunnyvale, CA, USA). Assay solution (0.02 mL/well) was added to all wells of the 96-well plate with cultured iPS cells or fibroblasts, which were then incubated in a 5% CO₂ incubator (37°C and humidity-controlled). Four hours later, the absorbance change at 570 nm and 605 nm was monitored using a microplate reader (SpectraMAX M3; Molecular Devices). The cell viability in each well was determined as the relative percentage to the vehicle-control wells. Because a high concentration of ascorbic acid disturbs the absorbance by discoloring the culture medium, in the ascorbic acid assay only, the test medium was replaced by no-drug medium (0.1 mL/well) on the day of the assay. After incubation for at least 1 hr in a 5% CO₂ incubator, the cytotoxicity of drug was assessed.

Cardiac differentiation assay with human iPS cells

Subconfluent iPS cell colonies were collected by dispase (1 mg/mL; STEMCELL Technologies) treatment (37°C, 20 min). After centrifugation (135 × g, 5 min, room temperature), supernatant was discarded, and the cell pellets were suspended by careful pipetting in fresh AggreWell™ EB formation medium (STEMCELL Technologies) supplemented with 10 µmol/L Y-26732 solution (5 mmol/L; FUJIFILM Wako Pure Chemicals). The cell suspension from one culture dish (60 mm × 15 mm) was seeded to 3 wells of an AggreWell™ 800 plate (STEMCELL Technologies) and then cultured in a 5% CO₂ incubator. One to 2 days later, the embryonic bodies (EBs) formed from the iPS cells were collected in a conical tube (15 mL) using a reversible cell strainer (37 µm, STEMCELL Technologies) and reseeded in an ultra-low-attachment 6-well culture plate (Sumitomo Bakelite) from one well to another. Culture medium was replaced with Dulbecco’s modified eagle medium: nutrient mixture F-12 + GlutaMax™ (DMEM/F12) supplemented with a serum-free Gibco^® B-27^® supplement (50X) and a Gibco^® MEM non-essential amino acids solution (100X) (NEAA) (B27-DMEM/F12; Thermo Fisher Scientific) (2 mL/well) containing each concentration of drugs, activin-A (100 ng/mL, which is more than 10-fold the ED₅₀; HumanZyme, Chicago, IL, USA), Wnt-3a (100 ng/mL, which is more than 10-fold the ED₅₀; R&D Systems, Minneapolis, MN, USA) and bone morphogenetic protein-4 (BMP-4) (100 ng/mL, which is more than 10-fold the ED₅₀; HumanZyme). After one day, medium was replaced with B27-DMEM/F12 containing the same concentrations of drug and noggin (300 ng/mL, which is more than 10-fold the ED₅₀; StemRD, Burlingame, CA, USA) and then cultured for 3 days in a 5% CO₂ incubator. Noggin-medium was replaced once with fresh medium of the same composition within 3 days. Thereafter, the medium was replaced with DMEM/F12 supplement with 5% Gibco^® qualified fetal bovine serum (Thermo Fisher Scientific) and NEAA (Thermo Fisher Scientific) (5% FBS-DMEM/F12) containing the same concentration of drug using a reversible cell strainer (STEMCELL Technologies), and the EBs with medium were transferred to a 96-well cell culture multi-plate (Sumitomo Bakelite) at 24 EBs (1 EB/well) per concentration or vehicle-control.

EBs were observed the occurrence of spontaneous beating under a microscope at least once a day. Two days later, the medium was replaced with the 5% FBS-DMEM/F12 containing no drug. Thereafter, medium was replaced with the same fresh medium every few days. The occurrence of spontaneous beating of EBs was observed until the beat ratio exceeded 90% in the vehicle-control group.

Data analyses

The three endpoints were calculated, from which the potential risks of the developmental toxicity induced by drug was predicted using the EST prediction model shown in Fig. 3 and also the type of developmental toxicity was predicted using the developmental toxicity prediction model shown in Fig. 3.

Fig. 3

Prediction of the developmental toxicity. (a) Three endpoints resulting from the cytotoxicity and myocardial differentiation assays. (b) EST prediction model, which classified the predicted risks of developmental toxicity of drugs by comparing the variables from the linear discriminant functions integrating three endpoints. (c) Developmental toxicity prediction model, which classified the types of developmental toxicity of drugs by a relative comparison of three endpoints. Fibroblasts: adult human dermal fibroblasts. iPS cells: human induced pluripotent stem cells. EST: mouse embryonic stem cell test.

The calculation of three endpoints

The endpoint of cardiac differentiation assay was calculated using the probit method in the SAS software program ver. 9.4 (SAS Institute Japan, Tokyo, Japan) when the cardiac differentiation ratio of vehicle-control exceeded 90% and expressed as the 50% differentiation inhibitory concentration (ID₅₀) of test drugs. Each endpoint of the cytotoxicity assay on iPS cells and fibroblasts was determined using the logit method in the SAS software program and expressed as the 50% inhibitory concentration (IC₅₀PS for iPS cells and IC₅₀F for fibroblasts) of test drugs.

Classification by the EST prediction model

The EST prediction model was published as EURL-ECVAM protocol No. 113: EST (Annett et al., 2016). The linear discriminant functions I, II and III shown in Fig. 3 were calculated by the incorporation of the three endpoints (ID₅₀, IC₅₀PS and IC₅₀F), and classified according to the classification criteria shown in Fig. 3 as follows: class 1 (non-developmentally toxic), class 2 (weakly developmentally toxic) and class 3 (strongly developmentally toxic), which indicate the potential risk for the drug-induced developmental toxicity.

Classification by the developmental toxicity prediction model

The developmental toxicity prediction model was developed by our lab divided the developmental toxicity induced by the test drugs into three types (1: non-effective, 2: embryotoxic [inducing growth retardation or dysfunction] or deadly [inducing infertility or stillbirth] or 3: teratogenic [inducing malformation] or deadly [inducing infertility or stillbirth]) (Fig. 3). As already mentioned in introduction section, the developmental toxicity prediction model assumed the fibroblasts to be pregnant women and the iPS cells to be the embryo or fetus. In this model, the adverse effects of the drugs on the iPS cells at concentrations not affecting the fibroblast viability reflect the developmental toxicity (Figs. 1 and 3).

In the pre-validation examination, the period of cardiogenesis was calculated from the application of cardiac inducers (activin-A, BMP-4 and wnt-3a) until the completion of cardiac differentiation (observed for up to two weeks) to exam whether it can use as parameter for developmental toxicity.

RESULTS

The results of the pre-validation examination with the partially modified protocol and the subsequent internal validation examination were shown.

Pre-validation of iPST protocol modified partially

The results of the thalidomide experiment are shown in Fig. 4 and Table 2. The three endpoints of IC₅₀F, IC₅₀PS and ID₅₀ were > 100, 9.07 and 1.41 µg/mL, respectively. The results of the linear discriminant functions were as follows: function I was -5.77, function II was 0.644, and function III was -3.28 (II > I and II > III). According to the EST prediction model, thalidomide was designated as class 2 drug. In contrast, in the developmental toxicity prediction model shown in Fig. 3, both the IC₅₀PS and ID₅₀, which indicate surrogate toxicity to an embryo or fetus, were markedly lower than the IC₅₀F, which indicates surrogate toxicity to the pregnant woman. Thalidomide was considered to affect the development of the embryo or fetus at concentrations that were non-toxic in pregnant women, and therefore thalidomide was designated as type 3 drug. Thalidomide was classified as a developmentally toxic/teratogenic drug.

Fig. 4

Pre-verification study of the iPST modified partially from the preliminary protocol. a-(1), b-(1) and c-(1): Results of the iPST. The open triangles (fibroblasts) and open circles (iPS cells) represent the mean ± S.D. of triplicate samples in the cytotoxicity assay. The closed circles represent the ratio of beating generation in the 24 embryonic bodies per sample concentration. The solid lines indicate linear approximation curves. a-(2), b-(2) and c-(2): Days of myocardiogenesis. Each column with a bar represents the mean ± S.D. of days of the myocardiogenesis of embryonic bodies, except for undifferentiated embryonic bodies. iPST: induced pluripotent stem cell test.

Table 2. Three endpoint values, functions values and classification of the developmental toxic potential of test drugs.

The period of cardiogenesis was 6.7 ± 0.9 days in the control group, while that in the 0.3 µg/mL group was 7.8 ± 1.7 days, that in the 0.6 µg/mL group was 8.6 ± 1.3 days, that in the 1.3 µg/mL group was 8.8 ± 2.0 days, and that in the 2.5 µg/mL group was 9.6 ± 1.8 days. The cardiogenesis period was prolonged with increasing thalidomide concentrations.

The results of the valproate experiment are shown in Fig. 4 and Table 2. The IC₅₀F, IC₅₀PS and ID₅₀ of valproate were > 100, 51.1 and 9.63 µg/mL, respectively. The results of the functions were as follows: I was -2.69, II was 2.54, and III was -4.84 (II > I and II > III). According to the EST prediction model, valproate was designated as class 2 drug. In the developmental toxicity prediction model, both the IC₅₀PS and ID₅₀, which indicate surrogate toxicity to an embryo or fetus, were markedly lower than the IC₅₀F, which indicates surrogate toxicity to pregnant women. Valproate was considered to affect the development of the embryo or fetus at concentrations that were non-toxic in pregnant women; thus, valproate was designated as type 3 drug. Valproate was classified as a developmentally toxic/teratogenic drug.

The period of cardiogenesis was 7.4 ± 1.0 days in the control group, while that in the 1 µg/mL group was 7.3 ± 1.2 days, that in the 3 µg/mL group was 7.1 ± 0.8 days, that in the 10 µg/mL group was 9.0 ± 1.7 days, and that in the 30 µg/mL group was nonexistent. The cardiogenesis period was prolonged by 1 day and more at 10 µg/mL compared with the control group.

The results of the saccharin experiment are shown in Fig. 4 and Table 2. The IC₅₀F, IC₅₀PS and ID₅₀ of saccharin were 66.9, 283 and > 1000 µg/mL, respectively. The results of the functions were as follows: I was 77.7, II was 33.8, and III was -28.5. According to the EST prediction model, saccharin was designated as class 1 drug. In contrast, in the developmental toxicity prediction model, both the IC₅₀PS and ID₅₀, which indicate surrogate toxicity to an embryo or fetus, were higher than the IC₅₀F, which indicates surrogate toxicity to pregnant women. Saccharin did not affect the development of the embryo or fetus at concentrations causing adverse reactions in adult or pregnant women; thus, saccharin was designated as type 1 drug. Saccharin was classified as a non-developmentally toxic/non-effective drug.

The period of cardiogenesis was 7.3 ± 0.1 days in the control group, while that in the 10 µg/mL group was 7.5 ± 0.2 days, that in the 30 µg/mL group was 7.2 ± 0.1 days, that in 100 µg/mL group was 7.5 ± 0.2 days, that in the 300 µg/mL group was 7.6 ± 0.1 days, and that in the 1000 µg/mL group was 8.6 ± 0.2 days. The cardiogenesis period was prolonged by approximately 1 day at 1000 µg/mL compared with the control group.

These findings show that the iPST with partially modified protocol accurately classified the potential risk and type of the developmental toxicity induced by drugs. The period of cardiogenesis was closely related to the ratio of cardiogenesis and was not necessary/additional parameter in the prediction of developmental toxicity.

Validation of the iPST

Table 2 summarizes the endpoints (IC₅₀F, IC₅₀PS and ID₅₀) by the test drugs in the cytotoxicity assay and the cardiac differentiation assay.

The classification of the potential risk of developmental toxicity of the drugs in the EST prediction model

Table 2 lists the results of the linear discriminant functions. The drugs with the highest values for function III were designated as class 3 according to the EST prediction model, including aminopterin, methotrexate and 5-FU (Table 2). The drugs with the highest values for function II were designated as class 2, including retinoic acid, tetracycline, lithium, phenytoin and warfarin (Table 2). The drugs with the highest values for function I were designated as class 1, including ascorbic acid, isoniazid and penicillin G (Table 2). The test drugs were accurately classified into three classes depending on the potential risks of developmental toxicities in human. It is concluded that the EST prediction model established using mouse cells would be useful for predicting human developmental toxicities of drugs in the iPST using human cells.

The classification of the type of developmental toxicity of the drugs in the developmental toxicity prediction model

The drugs for which both the IC₅₀PS and ID₅₀ were higher than the IC₅₀F included ascorbic acid and isoniazid, which were classified as type 1 drugs (Table 2). These drugs may not adversely affect iPS cells (taken to be embryos or fetuses) at concentrations affecting fibroblasts (taken to be pregnant women). The drugs for which the ID₅₀ (and IC₅₀PS) were lower than the IC₅₀F included aminopterin, methotrexate, retinoic acid, tetracycline, lithium, 5-FU and warfarin, which were classified as type 3 drugs according to the criteria dependent on the comparison of the three endpoints shown in Fig. 3 (Table 2). These drugs may adversely affect iPS cells (taken to be embryos or fetuses) development at concentrations not affecting fibroblasts (taken to be pregnant women). Only penicillin G among the non-teratogenic drugs had a lower IC₅₀PS than the IC₅₀F, and it was designated as a type 2 drug (Table 2). This drug may have adversely affected iPS cells (taken to be embryos or fetuses) growth at the concentrations not affecting fibroblasts (taken to be adult or pregnant women).

Phenytoin was not classified into a group because all 3 endpoints were the same values (> 30 µg/mL) due to the solubility limit, and therefore comparison was impossible. The type of developmental toxicity induced by drugs except for penicillin G and phenytoin was predicted by the developmental toxicity prediction model. It is suggested that the developmental toxicity prediction model may help to predict the types of developmental toxicity of drugs in humans.

DISCUSSION

The classification of the EST prediction model classified the positive control test drugs known to have developmental toxicity (teratogenicity) in humans as developmentally toxic drugs (class 2 or 3) and the negative control test drugs as non-developmentally toxic (class 1). Even though this model was established using mouse cells, the risk of developmental toxicity of test drugs could be accurately predicted using human cells. Therefore, the EST prediction model was verified to be useful in the iPST.

Regarding the classification of the type of toxicity according to the developmental toxicity prediction model, all test drugs except for penicillin G and phenytoin were accurately classified as positive control test drugs for teratogenic (malformation) and deadly (type 3) and negative control drugs for non-effective (type 1). Phenytoin, a positive control drug, could not calculate for its three endpoints due to its solubility limit, and therefore was not used in the developmental toxicity prediction model. Penicillin G, a negative control drug, was classified embryotoxic (growth retardation and dysfunction) and deadly (type 2). In the original EST, penicillin G was deemed a borderline drug between weakly and non-developmentally toxic (Scholz et al., 1999) and is considered difficult to classify correctly. The clinical total C_max of penicillin G is 3.44 µg/mL (134.6 µmol/L) (Palmer et al., 2013). The endpoints (50% inhibition concentrations) of the three assays in the present study were between 1420 and 3370 µg/mL, which was > 400-fold the C_max. At a lower concentration (e.g., at the 20% inhibition concentration), the cytotoxicity to fibroblasts in the three-endpoint assay was 831 µg/mL, that to iPS cells was 1086 µg/mL and the cardiac differentiation inhibition was 3200 µg/mL (unpublished observation). Even these concentrations were >240-fold the C_max and the type of toxicity according to the developmental toxicity prediction model was type 1 (unpublished observation). Therefore, if administered at a therapeutic concentration, the toxicity of penicillin G will be type 1. In fact, penicillin G is frequently administered during pregnancy because it is thought to lack human toxicity and not be teratogenic (Briggs et al., 2011). Therefore, because the safety margin of penicillin G is sufficient, no action in the clinical setting will occur, and penicillin G is unlikely to induce any teratogenic developmental toxicity. Given the above findings, the developmental toxicity prediction model is expected to accurately predict the developmental toxicity of drugs in the clinical setting by considering the clinical therapeutic concentration.

Phenytoin was designated as a class 2 drug, indicating weak developmental toxicity, from 30 μg/mL of solubility limit in the EST prediction model. In the original EST, which used mouse cells, phenytoin was also a class 2 drug (Table 2, EURL-ECVAM protocol No. 113: EST). However, other developmental toxicity assays, such as devTOX (Palmer et al., 2013) and ZET (Truong et al., 2011), incorrectly determined phenytoin to have a negative developmental toxicity (Table 2). Whereas, the original EST, which are undetected the developmental toxicity of thalidomide having high sensitivity in humans. The iPST using human cells showed positive findings for phenytoin and thalidomide, making it possible to predict the most accurately the clinical developmental toxicity of drugs.

Among the test drugs classified into classes 2 and 3 according to the EST prediction model, the three endpoints of retinoic acid, lithium and warfarin were higher than the clinical therapeutic blood concentrations (Tables 1 and 2). Retinoic acid is an essential factor for embryo and fetal development but impairs development at high exposure (Vandersea et al., 1998). A metabolite of vitamin A (retinoid), retinoic acid is necessary for supporting the life of vertebrates and is usually absorbed from food, as it is not synthesized in the body. An excessive intake of retinoic acid by pregnant women increases the exposure of the fetus to retinoic acid above the normal level and increases the teratogenic risk. Many of the teratogenic effects of retinoic acid are said to be caused by dietary supplements rather than medicine. Retinoic acid administered as a medicine is not expected to cause embryotoxicity, provided the exposure in pregnant women is properly controlled.

Lithium shows equivalent blood concentrations in pregnant women and embryos/fetuses, and infants with high cord blood levels have a high frequency of complications (Newport et al., 2005). Intake of lithium by pregnant women increases the risk of congenital heart disease in infants (Diav-Citrin et al., 2014). In the iPST, lithium was shown to be a weakly developmentally toxic drug, whereas the three endpoints in the iPST were more than 10-fold the clinically therapeutic concentration (Tables 1 and 2). On extrapolating the findings of the present study to humans, a high exposure to lithium is expected to induce developmental toxicity in embryos and fetuses. A retrospective analysis showed that lithium does indeed induce developmentally toxic effects, although at clinically therapeutic blood concentrations in pregnant women, it was unlikely to induce such effects. Of note, the use of lithium as a medication during pregnancy is contraindicated in Japan, whereas in the West, it is administered with cautions as a medication and is not explicitly contraindicated.

Warfarin is a vitamin K antagonist and an anticoagulant drug. It reduces the vitamin K redox cycle by inhibiting NAD(P)H-dependent reductase, resulting in the inhibition of the activation of proteins, such as blood coagulation factor in the liver and osteocalcin in the bone tissue (Nelsestuen et al., 1974; Kim et al, 2013; Gallieni and Fusaro, 2014). Osteocalcin is a vitamin K-dependent Ca-binding protein involved in bone remodeling (Moser and van der Eerden, 2019). Decreased vitamin K increases the incidence of fractures (Hao et al., 2017). Since warfarin passes through the placenta, it was considered that warfarin inhibits the vitamin K redox cycle in the embryo/fetus and subsequently affects bone formation in fetuses. In the present study, the developmentally toxic effects of warfarin manifested at a concentration (three endpoints in the iPST) higher than the clinical therapeutic concentration, suggesting that warfarin might not directly induce developmental toxicity at its therapeutic concentration. However, warfarin may affect the embryo/fetus through indirect activity, such as its anticoagulant action.

Isoniazid, a negative control drug, was shown to induce fetal hypoplasia (embryotoxicity) when given orally during pregnancy in rats and rabbits, whereas no teratogenesis was noted in reproduction studies in humans (Ludford et al., 1973). In the clinical setting, isoniazid administered during pregnancy showed no significant difference in birth rates, birth weight, sex ratios or birth abnormalities compared with placebo (Palmer et al., 2013; Shaul and Hall, 1977). Therefore, isoniazid is considered non-embryotoxic and non-teratogenic. In the present study, isoniazid was classified as a class 1, type 1 drug. The iPST was able to accurately predict the human developmental toxicity of isoniazid. In addition, our assay also accurately predicted the teratogenicity of thalidomide, which has high sensitivity in humans despite its effects being difficult to predict in animals, especially rodents. The iPST can therefore overcome species-specific differences in the developmental toxicity of drugs.

In the iPST, the EST prediction model was considered to predict the general developmental toxicity of drugs in the clinical setting. However, the developmental toxicity prediction model was considered to roughly predict the type of developmental toxicity induced by the drugs. Using both prediction models was thought to allow for a more detailed and accurate risk assessment than using either alone. Incorporating the predicted plasma concentration of clinical therapeutics may support the accurate prediction of the developmental toxicity induced by drugs.

The iPST, as well as the original EST, assessed the cardiac differentiation inducing from the mesoderm formed from iPS cells like embryonic stem cells isolated from blastoderm. The mesoderm is the middle layer of the trilaminar germ layers formed from blastoderm at a very early stage of embryonic development. The other two layers are the ectoderm (outer side) and endoderm (inner side). The neural tissue is generally thought to be generated from ectoderm, although a previous report stated that the neural tissue (hindbrain and spinal cord) is generate from axial stem cells, which are the same progenitor cells as seen in the mesoderm, rather than the ectoderm (Takemoto et al., 2011; Kondoh and Takemoto, 2012). This finding suggests that the cardiac differentiation assay in the iPST simultaneously evaluates the neural tissue differentiation pathway.

In summary, the iPST using human cells can predict the developmental toxicity induced by drugs in humans. This test may be a better drug screening model than other in vitro assay models for assessing the developmental toxicity induced by drug candidates, as well as chemicals and pesticides, in a preclinical setting. One potential issue associated with the iPST is the potential lot-to-lot variation in iPS cells, which has been shown to have cell-specific differentiation to tissue/organ. The low rate of the cardiac differentiation regarding the induction of iPS cells may not calculate the endpoint ID₅₀ of the cardiac differentiation assay in the iPST. Three in-house generated lots were confirmed to have no difference in cardiac differentiation, but publicly available commercial lots of iPS cells were not tested. Another potential issue is the difference in embryo/fetus exposure due to the absence of a placenta compared to in vivo. Drugs cross the placenta into the fetus by passive diffusion according to the lipid solubility and molecular weight of the drugs and by the active diffusion via placental active transporters. They are then excreted from the placenta by ATP-binding cassette transporters (Lankas et al., 1998). In the iPST, because the test drugs are administered to the iPS cells, which are the same as the inner cell mass of a blastocyst at the early-stage of pre-implantation (before placenta development), the absence of placenta might have a negligible effect.

To conduct the cardiac differentiation assay efficiently, the EB formation process is currently being improved, moving from the AggreWell™800 plate to the PrimeSureface^® plate (96 wells, V-type, low cell binding; Sumitomo Bakelite) with the appropriate edge inside of the V-well. This method will enable testing without moving EBs to another well.

Conflict of interest

The authors declare that there is no conflict of interest.

REFERENCES

• Agwuh, K.N. and MacGowan, A. (2006): Pharmacokinetics and pharmacodynamics of the tetracyclines including glycylcyclines. J. Antimicrob. Chemother., 58, 256-265.
• Aikawa, N., Kunisato, A., Nagao, K., Kusaka, H., Takaba, K. and Ohgami, K. (2014): Detection of thalidomide embryotoxicity by in vitro embryotoxicity testing based on human iPS cells. J. Pharmacol. Sci., 124, 201-207.
• Aikawa, N., Suzuki, Y. and Takaba, K. (2015): A simple protocol for the myocardial differentiation of human iPS cells. Biol. Pharm. Bull., 38, 1070-1075.
• Annett, J.R., Agnieszka, S. and Manuela, F. (2016): The European Union Reference laboratories (EURL)-the European Center for the Validation of Alternative Methods (ECVAM) database service on alternative methods to animal experimentation, protocol No. 113: embryonic stem cell test (EST).
• Briggs, G.G., Freeman, R.K. and Yaffe, S.J. (2011): A reference guide to fetal and neonate risk. In: Drugs in pregnancy and Lactation 9^th edition, pp.772-773, Lippincott Williams & Wilkins, Philadelphia.
• Diav-Citrin, O., Shechtman, S., Tahover, E., Finkel-Pekarsky, V., Arnon, J., Kennedy, D., Erebara, A., Einarson, A. and Ornoy, A. (2014): Pregnancy outcome following in utero exposure to lithium: a prospective, comparative, observational study. Am. J. Psychiatry, 171, 785-794.
• Erić, M. and Sabo, A. (2008): Teratogenicity of antibacterial agents. Coll. Antropol., 32, 919-925.
• Evans, M.J. and Kaufman, M.H. (1981): Establishment in culture of pluripotential cells from mouse embryos. Nature, 292, 154-156.
• Fratta, I.D., Sigg, E.B. and Maiorana, K. (1965): Teratogenic effects of thalidomide in rabbits, rats, hamsters, and mice. Toxicol. Appl. Pharmacol., 7, 268-286.
• Gallieni, M. and Fusaro, M. (2014): Vitamin K and cardiovascular calcification in CKD: is patient supplementation on the horizon? Kidney Int., 86, 232-234.
• Genschow, E., Spielmann, H., Scholz, G., Seiler, A., Brown, N., Piersma, A., Brady, M., Clemann, N., Huuskonen, H., Paillard, F., Bremer, S. and Becker, K. (2002): The ECVAM international validation study on in vitro embryotoxicity tests: results of the definitive phase and evaluation of prediction models. Altern. Lab. Anim., 30, 151-176.
• Genschow, E., Spielmann, H., Scholz, G., Pohl, I., Seiler, A., Clemann, N., Bremer, S. and Becker, K. (2004): Validation of the embryonic stem cell test in the international ECVAM validation study on three in vitro embryotoxicity tests. Altern. Lab. Anim., 32, 209-244.
• Hao, G., Zhang, B., Gu, M., Chen, C., Zhang, Q., Zhang, G. and Cao, X. (2017): Vitamin K intake and the risk of fractures: A meta-analysis. Medicine (Baltimore), 96, e6725.
• Jaju, M., Jaju, M. and Ahuja, Y.R. (1981): Combined action of isoniazid and para-aminosalicylic acid in vivo on human chromosomes in lymphocyte cultures. Hum. Genet., 56, 375-377.
• Kim, M., Na, W. and Sohn, C. (2013): Vitamin K1 (phylloquinone) and K2 (menaquinone-4) supplementation improves bone formation in a high-fat diet-induced obese mice. J. Clin. Biochem. Nutr., 53, 108-113.
• Kondoh, H. and Takemoto, T. (2012): Axial stem cells deriving both posterior neural and mesodermal tissues during gastrulation. Curr. Opin. Genet. Dev., 22, 374-380.
• Koren, G., Pastuszak, A. and Ito, S. (1998): Drugs in pregnancy. N. Engl. J. Med., 338, 1128-1137.
• Kunisato, A., Wakatsuki, M., Shinba, H., Ota, T., Ishida, I. and Nagao, K. (2011): Direct generation of induced pluripotent stem cells from human nonmobilized blood. Stem Cells Dev., 20, 159-168.
• Lankas, G.R., Wise, L.D., Cartwright, M.E., Pippert, T. and Umbenhauer, D.R. (1998): Placental P-glycoprotein deficiency enhances susceptibility to chemically induced birth defects in mice. Reprod. Toxicol., 12, 457-463.
• Loto, O.M. and Awowole, I. (2012): Tuberculosis in pregnancy: a review. J. Pregnancy, 2012, 379271.
• Ludford, J., Doster, B. and Woolpert, S.F. (1973): Effect of isoniazid on reproduction. Am. Rev. Respir. Dis., 108, 1170-1174.
• Martin, G.R. (1981): Isolation of a pluripotent cell line from early mouse embryos cultured in medium conditioned by teratocarcinoma stem cells. Proc. Natl. Acad. Sci. USA, 78, 7634-7638.
• Moser, S.C. and van der Eerden, B.C. (2019): Osteocalcin-A versatile bone-derived hormone. Front. Endocrinol. (Lausanne), 9, 794.
• Nahum, G.G., Uhl, K. and Kennedy, D.L. (2006): Antibiotic use in pregnancy and lactation: what is and is not known about teratogenic and toxic risks. Obstet. Gynecol., 107, 1120-1138.
• Nelsestuen, G.L., Zytkovicz, T.H. and Howard, J.B. (1974): The mode of action of vitamin K. Identification of gamma-carboxyglutamic acid as a component of prothrombin. J. Biol. Chem., 249, 6347-6350.
• Newport, D.J., Viguera, A.C., Beach, A.J., Ritchie, J.C., Cohen, L.S. and Stowe, Z.N. (2005): Lithium placental passage and obstetrical outcome: implications for clinical management during late pregnancy. Am. J. Psychiatry, 162, 2162-2170.
• Palmer, J.A., Smith, A.M., Egnash, L.A., Conard, K.R., West, P.R., Burrier, R.E., Donley, E.L. and Kirchner, F.R. (2013): Establishment and assessment of a new human embryonic stem cell-based biomarker assay for developmental toxicity screening. Birth Defects Res. B Dev. Reprod. Toxicol., 98, 343-363.
• Scheinhorn, D.J. and Angelillo, V.A. (1977): Antituberculous therapy in pregnancy. Risks to the fetus. West. J. Med., 127, 195-198.
• Scholz, G., Genschow, E., Pohl, I., Bremer, S., Paparella, M., Raabe, H., Southee, J. and Spielmann, H. (1999): Prevalidation of the embryonic stem cell test (EST)—a new in vitro embryotoxicity test. Toxicol. In Vitro, 13, 675-681.
• Schumacher, H., Blake, D.A., Gurian, J.M. and Gillette, J.R. (1968): A comparison of the teratogenic activity of thalidomide in rabbits and rats. J. Pharmacol. Exp. Ther., 160, 189-200.
• Seiler, A.E. and Spielmann, H. (2011): The validated embryonic stem cell test to predict embryotoxicity in vitro. Nat. Protoc., 6, 961-978.
• Shaul, W.L. and Hall, J.G. (1977): Multiple congenital anomalies associated with oral anticoagulants. Am. J. Obstet. Gynecol., 127, 191-198.
• Suzuki, N., Ando, S., Yamashita, N., Horie, N. and Saito, K. (2011): Evaluation of novel high-throughput embryonic stem cell tests with new molecular markers for screening embryotoxic chemicals in vitro. Toxicol. Sci., 124, 460-471.
• Takemoto, T., Uchikawa, M., Yoshida, M., Bell, D.M., Lovell-Badge, R., Papaioannou, V.E. and Kondoh, H. (2011): Tbx6-dependent Sox2 regulation determines neural or mesodermal fate in axial stem cells. Nature, 470, 394-398.
• Teo, S.K., Evans, M.G., Brockman, M.J., Ehrhart, J., Morgan, J.M., Stirling, D.I. and Thomas, S.D. (2001): Safety profile of thalidomide after 53 weeks of oral administration in beagle dogs. Toxicol. Sci., 59, 160-168.
• Teo, S.K., Denny, K.H., Stirling, D.I., Thomas, S.D., Morseth, S. and Hoberman, A.M. (2004): Effects of thalidomide on developmental, peri- and postnatal function in female New Zealand white rabbits and offspring. Toxicol. Sci., 81, 379-389.
• Truong, L., Harper, S.L. and Tanguay, R.L. (2011): Evaluation of embryotoxicity using the zebrafish model. Methods Mol. Biol., 691, 271-279.
• Tuchmann-Duplessis, J.M. (1965): Design and interpretation of teratogenic tests. In: Embryopathic activity of drugs (Robson, J.M., Sullivan, F.M., Smith, R.L., ed), pp.56-87, Little Brown & Company, NY, USA.
• Türck, D., Heinzel, G. and Luik, G. (2000): Steady-state pharmacokinetics of lithium in healthy volunteers receiving concomitant meloxicam. Br. J. Clin. Pharmacol., 50, 197-204.
• Vandersea, M.W., McCarthy, R.A., Fleming, P. and Smith, D. (1998): Exogenous retinoic acid during gastrulation induces cartilaginous and other craniofacial defects in Fundulus heteroclitus. Biol. Bull., 194, 281-296.
• Vargesson, N. (2015): Thalidomide-induced teratogenesis: history and mechanisms. Birth Defects Res. C Embryo Today, 105, 140-156.
• Wobus, A.M., Holzhausen, H., Jäkel, P. and Schöneich, J. (1984): Characterization of a pluripotent stem cell line derived from a mouse embryo. Exp. Cell Res., 152, 212-219.