Clinical safety of cell-based therapies requires the use of highly defined and preferably xenogeneic-free techniques with economically reasonable cost. In this study, we developed simple but efficient methods for high-quality hPSC culture and differentiation towards RPE and LESC lineages. A minor modification of the inductive cues during the first week of differentiation steered hPSCs toward either the RPE or LESC fate, enabling production of two clinically relevant ocular cell types with the same basal method.

The culture of hPSCs has advanced from mouse or human feeder cell-based systems to feeder-independent culture on ECM proteins or their fragments in chemically defined, xeno-free media such as NutriStem® hESC XF medium (Biological Industries) and TeSR™2 medium [37]. The E8 medium was a major breakthrough as an albumin-free, current Good Manufacturing Practice (cGMP)-compliant medium containing only eight components essential for hPSC culture [38]. Consistent with earlier studies, we were able to achieve efficient monolayer culture of eight genetically distinct hESC and hiPSC lines in E8 medium combined with LN-521 matrix [3941]. We found that prolonged feeder-free culture using single cell passaging led to karyotype changes that generated growth and differentiation advantages. In our analysis, two hPSC lines acquired karyotypic changes in the 20q arm—a known hot spot for chromosomal changes in hPSCs [42]. The gain of 20q11.21 allows hPSCs to escape apoptosis, allowing abnormal cells to take over the culture [43]. Feeder-free cultures thus require frequent karyotyping and use of low-passage hPSCs for differentiation. The safe window for usage is short; a maximum of 15 feeder-free passages equals 7.5 weeks of culture, while we have routinely cultured hPSCs on hFF feeders using manual passaging for up to 100 weeks with stable karyotypes. However, the feeder-free monolayer culture on LN-521 is more efficient, scalable, and less laborious compared to feeder-based systems. In practice, an average of approximately threefold cell multiplication pace was achieved within each passage (3–4 days) during successful culture. This means that a single 24-well of hPSCs (~100,000 cells) can expand into ~2.7 × 106 cells within 3 passages (1.5 weeks). This efficiency enables the use of low-passage feeder-free hPSCs to produce clinically relevant numbers of differentiated cell types. Minimized variability due to culture medium and matrix simplicity, and flexibility as a weekend-free feeding regimen, also add value and decrease costs.

Besides the well-defined and high-quality hPSCs, successful stem cell-based therapy requires differentiation methods with sufficient yield and purity. Recently, several hPSC differentiation strategies mimicking whole eye development have been reported to generate multiple ocular cell types in zones of two-dimensional colonies [44], or complex three-dimensional corneal [13, 45] or retinal organoids [29]. Ocular organoids recapitulate the early developmental events in vitro and provide powerful three-dimensional model systems for studying developmental or disease processes. However, these protocols are complex and time-consuming, difficult to standardize for production of high cell yields for cell therapy applications, and would require rigorous sorting of the desired cell populations [46].

Within the last 2 years, several research groups introduced xeno-free protocols for spontaneous or directed clinical grade RPE differentiation [2427, 29]. In our hands, attempts to replicate some of these methods were unsuccessful. Plaza Reyes et al. used spontaneous RPE differentiation in suspension in tailored NutriStem hESC XF medium lacking bFGF, and further plating of selected, pigmented RPE cells to LN-521 [25]. Our attempts to achieve RPE differentiation in EB suspensions led to disaggregation of EBs and insufficient levels of pigmentation. On the other hand, differentiation in X-VIVO™ 10 medium (Lonza), previously reported for xeno-free RPE differentiation [24], led to extreme pigmentation at the expense of RPE and epithelial properties (Additional file 6: Supplementary Dataset 1). RPE differentiation with XF-Ko-SR basal medium was recently reported by Choudhary and co-workers as an adherent monolayer-based method relying on temporal and sequential dual SMAD inhibition, BMP4, and Activin A activation [26]. Their approach was efficient and fast, with RPE-specific genes upregulated in 40 days, allowing whole plate passaging without the need for manual selection of pigmented cells. However, they found lower expression levels of key RPE genes in these cells compared with spontaneously differentiated hPSC-RPE. This again emphasizes the need for a prolonged culture of several weeks followed by purification and expansion to allow for proper RPE maturation. Our approach of using the XF-Ko-SR, together with LN-521 and col IV combination matrix, was suitable for either spontaneous or simple, directed, xeno-free RPE differentiation producing high-quality RPE. We also demonstrate that a similar method, with minor modifications, yields corneal LESCs; the induction stage was performed in XF-Ko-SR, and adherent culture was continued in CnT-30, a commercially available, fully defined culture medium developed for primary corneal epithelial cells. In its current formulation, CnT-30 contains animal-derived (but clinical grade) heparin, which could be replaced for a fully xeno-free medium. Most previously published studies to generate corneal epithelial cells from hPSCs have relied on the use of undefined, xenogeneic factors such as conditioned medium, PA6 feeder cells, Bowman’s, or amniotic membranes [16, 19, 47, 48]. A recent, defined protocol to differentiate a single hESC line to corneal epithelial progenitor cells utilized simple keratinocyte serum-free medium (KSFM) mixed with DMEM/F12 for increased calcium concentration, together with a simple col IV matrix [30]. ABCG2- and p63-positive populations were produced in just 6 days but differentiation required elevated CO2 levels. The mechanism behind these differentiation conditions was not discussed [30].

When directing feeder-free hPSC differentiation towards ocular epithelial lineages, suspension culture as EBs greatly improved cell survival, while small molecule induction helped direct differentiation. For feeder-free LESC differentiation, ectodermal induction alone was insufficient, resulting in neural network formation or cell death. A mesodermal induction with BMP4 was beneficial in directing differentiation towards surface ectoderm, and consequently LESCs. BMP4 is important in anterior eye segment development, induces corneal and epidermal differentiation, and inhibits neural differentiation [49, 50]. In our study, this two-stage surface ectodermal induction followed by adherent culture on LN-521/col IV combination matrix in CnT-30 medium resulted in > 70% p63α-positive cell populations in 24 days. As for RPE, the TGF-β signaling inhibitor SB-431542 has been previously shown to enhance pigmentation rate during initial differentiation, but in a cell line-dependent manner [51]. Meanwhile, Wnt signaling induces RPE differentiation once the cells are committed to the retinal fate [52, 53]. Our short neuroectodermal induction with Wnt and TGF-β inhibitors repeatedly increased pigmentation of mature RPE layers in all seven experiments and with both cell lines. The role of pigmentation resulting from melanin content in the RPE remains somewhat speculative. The pigment serves to store Ca2+, absorb stray light, and minimize scatter within the eye, but melanosomes also act as antioxidants, eliciting cytoprotective functions [54]. We have previously shown the inverse relationship of TEER and pigmentation [36], and continuously see this effect with hPSC-derived RPE. The mechanism behind this phenomenon and its clinical significance requires further studies. It was recently shown that the level of pigmentation does not correlate with hPSC-RPE maturity, although it is commonly used as such a marker [55]. Interestingly, highly pigmented hPSC-RPE appear to have significantly higher expression of pathways related to calcium signaling and adherens junction remodeling, compared with lightly pigmented cells [55]. Pigmentation may also affect other aspects of physiology and functionality of the intact RPE monolayer, and its relevance for transplantation success calls for further studies.

Ocular cell replacement therapies require relatively small populations of RPE (1.3 × 3.0 mm patch or 0.2 × 106 cells in suspension) [11, 12] or a few thousand p63-bright LESCs [56] per eye, but quality assurance for GMP compliance requires additional cell populations and thus scale-up of culture and differentiation. RPE differentiation from hPSCs tends to be time consuming, so using fresh hPSC-RPE for cell replacement therapy requires a long waiting time for the patient. In contrast, the hPSC-LESC differentiation period is too rapid to allow for quality and safety testing prior to transplantation. Cryobanking offers a solution by allowing preparation of readily available cell stocks in advance. A recent study showed successful cryostoring and recovery of xeno-free hiPSC-derived neuroretinal organoids, and further isolated photoreceptor precursor cells, as well as RPE cells [29]. With our xeno-free methods, a starting culture of 1 × 106 feeder-free hPSCs can yield cryostocks of > 5 × 106 RPE cells in 9 weeks, and nearly 1 × 106 LESCs within 23 days. Controlled passaging of hPSC-RPE once or twice is essential to achieve purity and successful cryostoring at the exponential growth phase, while passaging more than three times changes the hPSC-RPE phenotype [57]. Meanwhile, LESC purity improved after cryostoring and passaging, as suggested by ubiquitous p40 and p63α protein expression.

GMP-compliant derivation, culture, and differentiation methods are a requisite for cell-based medicinal products, including hPSC-derived cells [58]. Xeno-free, GMP-quality, clinical-grade hPSC-RPE products such as OpRegen® cells are emerging [28], but simplified protocols and raw materials for cell culture and differentiation are an advantage. The simple xeno-free methods described here could be upgraded to GMP-quality for future preclinical testing.