This research was approved by the ethics committee of East Hospital Affiliated with Tongji University in Shanghai. The human iPS cell lines (hiPSCs) used for in vitro differentiation was a gift from Professor Jin Ying (Chinese Academy of Sciences, China). The hiPSCs were cultured in basal medium [BM; DMEM/F12 medium (Thermofisher, United States)] supplemented with 1 × N2(Thermofisher, United States), 1 ×  B27(Thermofisher, United States), 0.1 mM minimum essential medium with nonessential amino acids (MEM NEAA; Gibco, United States), and 0.1 mM 2-mercaptoethanol (Gibco, United States) containing 10 ng/ml BFGF (R&D Systems, United States), as previously described (Jin et al., 2018). For the human corneal endothelial cells (HCEC-B4G12) (Zeye Culture Collection, China) cultured in DMEM/F12 medium (Thermofisher) supplemented with 10% fetal bovine serum (FBS, ExCell Bio, China). The cultures were incubated in a humidified atmosphere of 5% CO₂ at 37°C. The medium was changed every day.

There were three stages of CEC differentiation (Figure 1E). In the first stage, the hiPSCs were cultured with 2 µM SB431542 (Selleck, United States) and 2 µM DMH1 (Selleck, United States) for 6 days with daily medium changes to differentiate into NCCs. In the second stage, 30 small molecules related to stem cell differentiation and corneal development were screened on hiPSC-derived NCCs. On day 6, the cells were grown in chemically defined medium (CDM) supplemented with small-molecule compounds for 3 days during the second stage. In the last stage, 20 ng/ml epidermal growth factor (EGF, Thermofisher, United States) and 2 µM CHIR99021 (Selleck, United States) were added to the medium, and the cells were cultured for an additional 13 days with daily medium changes.

Firstly, the hiPSC-NCCs were passaged to the 24-well cell culture plate and cultured with the small molecules respectively as shown in Table 1. The total RNA was extracted to detect the expression of relative genes. The significant high expression of the relative genes was used to pick up the candidate molecule. Subsequently, the combination of two types of candidate molecules was applied to treat the hiPSC-NCCs. The expression level of the relative genes was applied to screen and evaluate the combination of the molecules.

The cultured neural crest cells (NCCs) were passaged to the 96-well plated for 24 h. The NCCs were treated with chemical molecules of different concentrations for another 24 h. The cell viability was detected with cell counting kit-8 reagent (TargetMol, United States) according to the manufacturer’s instructions.

RNA was extracted from cells with RNAiso Plus reagent (Cat. No. 9019, Takara, Japan) and chloroform. The concentration of RNA was measured with a NanoDrop spectrophotometer. Approximately 1.0 µg of total RNA was reverse-transcribed into complementary DNA with PrimeScript RT Master Mix (RR036A, Takara, Japan). qRT-PCR was run with SYBR Green PCR Master Mix (Tiangen Biotech, China) and the following cycling parameters: denaturation at 95°C for 5 min followed by 39 cycles of 95°C for 30 s and 60°C for 30 s. The relative expression level of each gene was analyzed using the 2^−ΔΔCT method. The primers used in this study are listed in Supplementary Table S1.

The cells were fixed with 4% paraformaldehyde (Sigma-Aldrich, Germany), permeabilized with 0.1% Triton X-100 (Sigma-Aldrich) in PBS for 10 min, washed 3 times for 5 min/wash with PBS, and then blocked with 3% BSA (Sangon Biotech, China) in PBS (Sangon Biotech, China) for 1 h at room temperature. The cells were incubated with the primary antibodies (Supplementary Table S2) and then incubated overnight at 4°C. Then, they were washed 3 times for 5 min/wash with PBS and incubated with the following fluorescent secondary antibodies for 1 h at room temperature: Alexa Fluor 555-conjugated donkey anti-mouse (1:500, Thermofisher, United States) and Alexa Fluor 488-conjugated donkey anti-mouse (1:500, ThermoFisher, United States). After three washes for at least 10 min each, the cells were exposed to DAPI (Sigma-Aldrich, Germany) for 5 min at room temperature to visualize nuclei. The samples were washed 3 times for 5 min each, and then pictures were taken with a Olympus microscope (Olympus, Japan).

Total proteins of cells were extracted using RIPA lysis buffer (Beyotime, China) supplemented with protease and phosphatase inhibitor cocktails (TargetMol, United States) on ice, and the protein concentrations were then determined with a BCA assay (Pierce, United States). Twenty micrograms of protein was run on a 10–15% polyacrylamide gel and transferred to a polyvinylidene fluoride (PVDF) membrane (Millipore, Germany). The blots were blocked with 5% BSA in TBS +0.1% tween 20 and incubated with primary antibodies (Supplementary Table S2) in 5% BSA overnight at 4°C. Then, the cells were incubated with the corresponding HRP-conjugated secondary antibodies (Proteintech, United States) for 1 h at room temperature. Images of the blots were obtained by using a Tanon system with enhanced chemiluminescence (ECL) reagent (Thermofisher, United States).

All data are expressed as the mean ± SEM. All analyses were performed with GraphPad Prism 9.0 software. One-way ANOVA was employed for the statistical comparisons. A value of p < 0.05 was considered to indicate statistical significance.

hiPSCs were cultured in CDM on Matrigel-coated cell culture plates and showed the typical cell morphology of pluripotent stem cells (Figure 1F). Moreover, the cultured hiPSCs were confirmed to express pluripotency markers such as OCT4, SOX2, NANOG, SSEA4 and TRA-1-60 by immunofluorescence (Figures 1A–D).

In this study, we induced hiPSCs to differentiate into neural crest cells (NCCs) and CECs in an orderly manner with CDM, as shown in Figure 1E. Briefly, hiPSCs were induced via dual-SMAD inhibition with SB431542 (to inhibit TGF-β-Smad-2/3 signaling) and DMH1 (to inhibit BMP-Smad1/5/8 signaling) to differentiate into NCCs (Figure 1G) for 6 days. Then, the NCCs were screened with small molecules for another 3 days to induce them to differentiate into cornea-destined cells and continuously cultured to differentiate them into CECs for another 13 days (Figures 1H,I).

NCCs, also named neural progenitor cells, are the original source of cornea-destined cells. Thus, we initially adopted a procedure to induce hiPSCs to differentiate into NCCs with dual-SMAD inhibition via SB431542 and DMH1. Initially, we detected the transcription levels at four time points to confirm the best time course for neural conversion of hiPSCs in the CDM. As shown in Figure 2A, the pluripotency marker OCT4 was significantly downregulated on day 2. Simultaneously, NCC-related markers (SOX9, SOX10, NTRK3, and NGFR) were upregulated on day 6 (Figures 2B–E).

The induced cells on day 6 showed a loss of hiPSC colonization and neuronal epithelial morphology (Figure 1G). We further analyzed the expression of the NCC-related genes PAX6, NGFR and B3GAT1 (Cheung et al., 2014). As shown in Figure 2F, almost all of the cells differentiated from the hiPSCs expressed NGFR, B3GAT1 and PAX6 on day 6. Thus, 6 days was selected for the induction of the NCC differentiation.

To identify small molecules facilitating the induction of NCCs to differentiate into CEC destined cells, we built a chemical library of 30 small molecules that targeted almost all the key pathways of stem cell differentiation and corneal development (Table 1). The final concentration of each compound added to the medium was based on previously reported data and testing for the ED50s in the NCCs via Cell Counting Kit-8 (CCK-8) assay (Miner et al., 2003; Bouzakri et al., 2004; Moreno et al., 2008; Johnson et al., 2012; Yap et al., 2012; Cho et al., 2013; Han et al., 2014; Schelleman et al., 2014; Kampa-Schittenhelm et al., 2017; Pinkosky et al., 2020; Zhang et al., 2020; Gampala et al., 2021; Ma et al., 2021). For initial screening, NCCs derived from dual-SMAD inhibition of hiPSCs for 6 days were cultured in a 24-well plate and treated with small molecules for another 3 days. 7 compounds, numbered 2 [ganetespib (STA-9090)], 4 (tenovin-6), 5 (SRT1720), 11 (purmorphamine), 25 [panobinostat (LBH589)], 26 (LY-2874455), and 30 (flavopiridol hydrochloride), were excluded, as almost of the cells died by the final time point. The effective candidate compounds were selected according to the transcriptome levels of CEC markers (AQP1, ZO-1, and COL8A1) at the final time point, as determined by qRT-PCR. Then, compounds 6 (A769662), 16 (AT13148), 20 (hydroxyprogesterone) and 23 (fisetin) were selected as compounds inducing high expression of AQP1, ZO1 and COL8A1 (Figures 3A–C). We further confirmed whether the combination of two of the four candidates improved the expression of CEC markers. The combination of 6 (A769662) and 16 (AT13148) clearly promoted the expression of AQP1, ZO1 and COL8A1 (Figures 3D–F). Additionally, the cell morphology after treatment with the combination of A769662 and AT13148 was more homogeneous than that after treatment with A769662 or AT13148 separately (Figure 3G). Taken together, these data demonstrate that the combination of AT13148 and A769662 promotes corneal endothelial differentiation from NCCs.

We further identified the hiPSC derived CECs (hiPSC-CECs) with the qRT-PCR, western blot and immunofluorescence. hiPSC-CECs showed a tightly packed hexagonal/polygonal appearance and similar to CEC morphology on day 22 (Figure 4D). Compared with hiPSCs and CECs at previous stages (day 6 and day 9), hiPSC-derived CECs highly expressed ZO1, AQP1 and COL8A1, as determined by qRT-PCR (Figures 4A–C). Additionally, the expression of Na⁺/K⁺-ATPase was significantly expressed in the hiPSC-CECs compared to the hiPSC-NCCs (Figure 4E). The hiPSC-CECs expressed Na⁺/K⁺-ATPase, AQP1, ZO1 and vimentin. The cell connection protein ZO1 was regularly distributed on the edges of the cells (Figure 4F). The protein expression and distribution is similar with the human CEC cell lines (B4G12) (Figure 4G). Thus, we considered the protocol shown the successful differentiation of hiPSC-CECs from hiPSC.

The combination of AT13148 and A769662 clearly promoted the differentiation of NCCs into CECs. To further elucidate the mechanism of AT13148 and A769662 in the differentiation of CECs from hiPSCs. We analyzed the protein level of the NCCs treated with AT13148 and A769662. A769662, similar to activators of adenosine monophosphate (AMP), is a selective and effective molecule that activates AMP kinase (AMPK), maintains energy balance, preserves endothelial cell vitality, and enhances endothelial cell differentiation and migration (Göransson et al., 2007). AT13148 is an ATP-competitive inhibitor of multiple AGC kinases, including AKT, phosphoinositide-dependent kinase 1 (PDK1), p70S6 kinase (p70S6K), p90 ribosomal S6 kinase (RSK), glycogen synthase kinase 3β (GSK-3β) and ROCK.

We detected the related key signaling proteins to explore the potential mechanism, as shown in Figure 5A. The phosphorylation level of Akt was significantly upregulated in the AT13148-treated group and the combination-treated group, while the control group and the A769662-treated group had almost no phosphorylation of AKT. The total Akt levels in the AT13148 and combination groups were downregulated (Figures 5B,C). It has previously been reported that AKT inhibitors can lead to hyperphosphorylation of AKT on regulatory sites (including Thr308), leading to AKT activation that can counteract the effects of small-molecule inhibitors (Okuzumi et al., 2009). A769662 can significantly increase the phosphorylation level of the Thr172 site of AMPK. However, AT13148 inhibited the phosphorylation of AMPK and upregulated total AMPK (Figures 5D,E). Additionally, the expression of GSK3β, the phosphorylation of p90RSK (Thr359 and Thr573) and the expression of total RSK were inhibited by AT13148. This inhibition by AT13148 was partially counteracted by A769662 (Figures 5F–I).

The above results suggest that AT13148 and A769662 can upregulate AMPK and PI3K/AKT signaling pathways during the differentiation of NCCs into CECs. To explore the regulation of the key transcription factors in this process, FOXO, PAX6, PITX2 and FOXC1 were detected by western blotting. The forkhead box class O (FOXO) family, which includes FOXO1, FOXO3a, FOXO4, and FOXO6, can be regulated by the phosphoinositol-3-kinase (PI3K)-Akt signaling pathway and specifically activate a coordinated transcriptional program to regulate developmental processes and energy metabolism in embryo development (Hosaka et al., 2004; Maiese, 2015; Martins et al., 2016).

As shown in Figure 6, AT13148 inhibited the expression of FoxO3a and FoxO4 and promoted the expression of FoxO1. The combination of AT13148 and A769662 downregulated the expression of FoxO3a and FoxO4 and upregulated the expression of FoxO1. With regard to the phosphorylation levels of FoxO1, FoxO3a and FoxO4, the phosphorylation at sites Ser253 and Thr32 of FoxO3a was significantly upregulated compared with that in the control group. Moreover, compared to AT13148, phosphorylation at the Ser256 and Thr24 sites of FoxO1 and the Ser318 and Ser321 sites of FoxO3a was inhibited by the combination of AT13148 and A769662. Phosphorylation at the Thr28 site of FoxO4 was inhibited by the combination treatment. Thus, the combination of AT13148 and A769662 can upregulate FoxO1 by inhibiting the phosphorylation of FoxO1 at Ser256 and Thr24. In addition, the combination of AT13148 and A769662 can inhibit FoxO3a and FoxO4 expression by upregulating the phosphorylation of FoxO3a at Ser253 and Thr32 and FoxO4 at Thr28.

The key transcription factors consist of PAX6, PITX2 and FOXC1, which are required for the development of the ocular anterior segment and corneal endothelium (Zavala et al., 2013). As shown in Figure 7A, PITX2 was upregulated by AT13148, A769662 or the combination treatment. With regard to the expression of FOXC1, AT13148 partially inhibited the expression of FOXC1, while the combination of AT13148 with A769662 recovered the expression of FOXC1(Figures 7B–D). The combination of AT13148 and A769662 did not change the level of PAX6, while AT13148 significantly upregulated the expression of PAX6.