Exploring the impact of human pluripotent stem cell heterogeneity on corneal limbal stem cell differentiation outcomes
Sonja Harjuntausta, Meri Vattulainen, Soile Nymark, Heli Skottman
TL;DR
This study shows that differences between human pluripotent stem cell lines affect how well they can be turned into corneal limbal stem cells, which are important for treating eye diseases.
Contribution
The study reveals that intrinsic hPSC line variability impacts LSC differentiation efficiency and consistency.
Findings
Variations in BMP4, LEF1, PAX6, and TGFB1 gene expression were observed even in well-differentiating hPSC lines.
Elevated TGFB1 expression correlated with fewer LSC markers and a mesenchymal morphology in some batches.
Protein expression variation between hPSC lines and batches was confirmed using flow cytometry.
Abstract
Differentiation of limbal stem cells (LSCs) from human pluripotent stem cells (hPSCs) shows promise for treating bilateral limbal stem cell deficiency. However, variation in hPSCs can compromise cell production protocols and increase unwanted heterogeneity. We hypothesize that intrinsic differences among hPSC lines affect their efficiency and consistency in LSC differentiation. Several hPSC lines were differentiated towards LSCs, and the process was analyzed in different timepoints using real-time qPCR (RT-qPCR) and immunofluorescence (IF) to assess differentiation efficiency. Flow cytometry (FC) was also used to quantify protein expression in the differentiated LSCs. Interestingly, variations in BMP4, LEF1, PAX6, and TGFB1 gene expressions, key factors in corneal epithelium induction, were high even in well-differentiating hPSC lines. Some batches failed to upregulate PAX6 and…
Genes, proteins, chemicals, diseases, species, mutations and cell lines named across the full text — each resolved to its canonical identifier and authoritative record.
Click any figure to enlarge with its caption.
Figure 1
Figure 2
Figure 3
Figure 4- —https://doi.org/10.13039/501100002341Research Council of Finland
- —Sigrid Juselius Foundation
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsCorneal Surgery and Treatments · Corneal surgery and disorders · Ocular Surface and Contact Lens
Introduction
Human pluripotent stem cell (hPSC)-based therapies hold great promise, yet several widespread challenges continue to hinder their successful application. Significant variation between individual hPSC lines^1,2^and also between induced pluripotent stem cell (iPSC) clones^3,4^remains a well-recognized issue in hPSC differentiation efficacy, making it difficult to develop standardized protocols for widespread clinical applications. Interestingly, a recent overview of current clinical trials using hPSC lines^5^summarizes that, to date, more than 1,200 patients have been treated with hPSC products. However, a limited number of different hPSC lines have been used for cell production, highlighting issues such as the lack of commercially available Good Manufacturing Practices (GMP)-grade hPSCs^5^. This variation in the differentiation rates of hPSC lines, combined with the limited availability of GMP-grade cell lines, further exacerbates the challenges in producing cells for clinical use. These factors contribute to significant problems in ensuring consistent and reliable cell production, which is crucial for the success of clinical applications.
The cell line-dependent variation in hPSC differentiation outcomes arises from intrinsic factors, including epigenetic memory of the cell of origin, variability in endogenous signaling pathway activation, and genetic diversity among hPSC lines^6–9^. These factors can affect lineage-specific propensity, leading to differences in the efficiency and fidelity of hPSC differentiation. Additionally, spatial effects, such as heterogeneity in cell-cell interactions and signaling gradients during differentiation, can further contribute to variability^10^. Importantly, this variation in differentiation potential is well acknowledged in the field and not specific for any certain differentiation pathway^11–13^.
The ability to generate functional limbal stem cells (LSCs) in vitro from hPSCs represents a promising approach for cell-based therapies to treat ocular surface disorders. Notably, hPSC-based therapies for these conditions have already advanced to clinical trials, as demonstrated by Soma and co-workers^14^. In addition to the recent clinical protocol used in the trial, several preclinical protocols for differentiating LSCs from hPSCs have been published, demonstrating diverse approaches to achieve LSCs and the corneal epithelium (CE)-like cells^14,15^. In our previous studies, our focus has been in the differentiation of LSCs from hPSCs^16–19^and their potential in homeostatic renewal of the CE which is critical for maintaining the health and integrity of the human ocular surface^20^. However, previous studies often base their findings on a limited number of hPSC lines, usually between one and three. This indicates a need for broader evaluation of LSC differentiation potential using a wider variety of hPSC lines.
Importantly, addressing the line dependent variability is essential for the development of clinically applicable protocols and to improve methods for the data-driven selection of the most suitable hPSC lines. Additionally, the absence of single, specific LSC marker to reliably identify therapeutically relevant LSCs^21,22^complicates the evaluation of differentiation efficiency and consistency across various cell lines used. Although previous studies have reported variability in differentiation efficiency^4,19^, the line-dependent variability in the effectiveness of hPSC differentiation into LSCs remains insufficiently investigated. In this study, this variation was further investigated to better understand and address the challenges in the differentiation of hPSC-derived LSCs (hPSC-LSCs).
Methods
hPSC culture and hPSC-LSC differentiation
Four in-house established hPSC lines, including one human embryonic stem cell (hESC) line (Regea08/017) and three human induced PSC (hiPSC) lines (WT001.TAU.bB2; WT003b.TAU.bC; WT004b.TAU.b2A) and a commercial hiPSC line (AICS-0016-184: WTC-mEGFP-ACTB-cl184 (mono-allelic tag)) from Allen Institute for Cell Science (Coriell Institute) were used for LSC differentiations. The establishment of hESC line Regea08/017 has been previously described by Skottman^23^. Human iPSC line WT001.TAU.bB2 has been previously established and characterized in-house as described by Grönroos et al.^24^ and hiPSC line WT003.TAU.bC as described in Vattulainen et al.^22^. WT004b.TAU.b2A and AICS-0016 were characterized as described in the supplemental information (Supplemental Table S3, Supplemental Figs. 1, 2). All methods were performed in accordance with the relevant guidelines and regulations. The faculty has the approval of the National authority Finnish Medicines Agency Fimea (Dnro FIMEA/2020/003758) to conduct research on human embryos, and supportive statements from Regional Ethics Committee of the Expert Responsibility Area of Tampere University Hospital have been obtained by the research group, granting the permissions to derive, culture, and differentiate hESC lines (R05116), and to establish and use hiPSC lines in ophthalmic research (R16116). No new cell lines were generated for this study.
The schematic outline of the hPSC-LCS differentiation and experimental design is presented in Fig. 1. Human PSC cultures were routinely maintained under serum- and feeder-free conditions and differentiated towards LSCs as previously described^16,18,19^, with the addition of a one-day Wnt inhibitor IWP-2 supplementation^17^. In brief, undifferentiated hPSCs were enzymatically dissociated into a single-cells and seeded into ultra-low attachment plates (Corning^®^ Costar^®^, United States) for induction. Embryoid body (EB) formation was promoted by using 5 µM blebbistatin (Sigma-Aldrich, United States) in the defined XFDM- medium (KnockOut™ Dulbecoo’s modified Eagle’s medium (DMEM) supplemented with 15% KnockOut™ SR XenoFree CTS™ (XFKOSR), 2 mM GlutaMAX™, 0.1 mM 2-mercaptoethanol, 1% MEM non-essential amino acids, and 50 U/ml penicillin-streptomycin (all from Gibco, Thermo Fisher Scientific, United States) for one day. To guide the EB differentiation toward the surface ectoderm lineage, cells were treated for one day with 10 µM transforming growth factor (TGF)-β inhibitor SB-505124 (PeproTech, Thermo Fisher Scientific), 50 ng/ml human basic fibroblast growth factor (bFGF; PeproTech) and 10 µM IWP-2 (Stemcell Technologies, Canada), followed by a two-day exposure to 25 ng/ml bone morphogenetic protein (BMP)-4 (PeproTech).
Following the induction, EBs were transferred and plated down onto cell culture plates (Corning^®^ CellBIND) coated with 0.5 µg/cm^2^ recombinant laminin-521 (LN-521, Biolamina, Sweden) and 5 µg/cm^2^ human placental collagen Type IV (Col IV, Sigma-Aldrich) and maintained in CnT-30 corneal differentiation medium (CELLnTEC Advanced Cell Systems AG, Bern, Switzerland) for adherent differentiation. Cells were thereafter cultured in CnT-30 medium, with regular medium changes three times a week until subjected to characterization analyses on day (d) 7, 10/11 and 24. Representative cell morphology during the differentiation was imaged with a Nikon Eclipse TE2000-S microscope equipped with a DS-Fi1 camera (Nikon Instruments, Amsterdam, Netherlands).
Fig. 1. Schematic outline of the hPSC-LCS differentiation workflow. Illustration created with Biorender.com.
Immunofluorescence based characterization of hPSC-LSC
Immunofluorescence staining (IF) was utilized in several stages during the study to analyze the marker protein expression of the cells. In the characterization of standard hPSC-LSC differentiation in CnT-30 culture medium, adherent cultures of hPSC-LSCs at d7, d10/11, and d24 were stained and expression of Paired box 6 (PAX6), tumor protein p63 alpha (p63α), ΔNp63 (p40) and cytokeratin (CK) 14 markers were investigated. Standard fixation and IF procedures with primary and secondary antibodies were performed essentially as previously described in Mikhailova et al.^17^. IF images were captured with Olympus IX51 fluorescence microscope (Olympus Corporation, Japan). All five hPSC lines were used for the full IF characterization of the differentiated hPSC-LSC, and the results were replicated with at least two independent cell differentiation batches for all lines. For additional characterization, at d24 time point protein expression of phosphorylated Smad2/3 (p-smad2/3) was investigated with two hPSC lines: Regea08/017 and WT003b.TAU.bC with two independent cell differentiation batches.
In addition, cytospin samples were prepared on d24 and stained with PAX6 and CK14 for quantification of LSC populations through cell counting analysis. These cytospin analysis were performed using two hPSC lines: Regea08/017 and WT003b.TAU.bC. Raw images of the stained cells were captured with an Olympus IX51 fluorescence microscope. ImageJ Image Processing and Analysis tools (v2.17.0, https://imagej.nih.gov/ij/) software were used for cell counting and image processing, respectively. Overview of the analysis strategy is presented within the Supplemental methods and Supplemental Fig. 3. Antibody specifics are provided in Supplemental Table 1.
Flow cytometry-based characterization of hPSC-LSC
The hPSC-LSCs after the 24-days of differentiation were characterized for their PAX6 and p63α antigen expression using flow cytometry (FC). FC analysis was performed with two cell lines Regea08/017 and WT003b.TAU.bC, both with at least three individual differentiation batches serving as biological replicates. Standard FC staining protocols were used in the sample preparation, following the recommendations provided by the antibody manufacturers. PE-conjugated rabbit monoclonal anti-human p63α (clone D2K8X) antibody (1:50, Cell Signaling Techology, #56687, United states), and PE-conjugated mouse monoclonal anti-human PAX6 (clone O18-1330) antibody (1:20, BD Pharmingen, #561552, United states) were used for indicated cell populations. Unstained cells were used as negative controls, respectively.
The FC analyses were performed using CytoFLEX S (Beckman Coulter, Brea, United states) flow cytometer. Dead cells, cell debris, and doublets were discarded during the gating and analyzing. At least 10 000 events were recorded from the initially gated populations and analyzed with FlowJo 10 software (v10.10.1, https://flowjo.com/flowjo10/download, BD Biosciences, San Jose, CA, USA). Negative control was used to gate the population of interest containing the cells. After excluding doublets from the analysis, the negative vs. positive gates were set with histograms (single stained samples) using 0.5% marginal. Finally, the established gates were copied to each sample of the experiment. Overview of the FC gating strategy is presented within the Supplemental Fig. 4.
Quantitative RT-qPCR characterization of hPSC and hPSC-LSC
The undifferentiated hPSCs (UD-hPSCs), as well as differentiated cells at d4, d7, d10/11 and d24 timepoints of the differentiation were analyzed for their mRNA expression with RT-qPCR. Genes PAX6,* BMP4*,* LEF1*,* AXIN2*, TGFB1,* Id1* and Id3 were analyzed from all time points, and the LSC marker genes ABCG2,* KRT15*,* LGR5 *from d24 and early ectodermal marker KRT18 from d7 and d24 time points using a sequence-specific TaqMan Gene Expression Assays. TaqMan assay details in Supplemental Table 2.
GAPDH was used as a housekeeping gene. RNeasy Mini Kit (Qiagen, Netherlands) were used for RNA isolation and cDNA was synthesized by using reverse transcription with a High-capacity cDNA kit (Applied Biosystems, United States) from the cell pellet samples collected in indicated time points. Biological replicates (at least n = 2) and controls were run as triplicate reactions with the ABI QuantStudio 12 K Flex Real-Time PCR system (Applied Biosystems). The results were analyzed using the − 2^∆∆Ct^ method^25^and are presented as the fold change in gene expression normalized to GAPDH and relative to the UD controls. Furthermore, the average Ct values from all Regea08/017 samples were utilized as generalized controls for each gene when assessing the initial differences amongst all UD-hPSC samples.
Statistical analysis
All data are presented as individual values or the mean ± standard deviation (SD). Kruskal–Wallis test (when 3 or more groups) or the Mann-Whitney U test (2 groups) was performed to analyze the differences between the groups using the GraphPad Prism 10 software (v10.3.1, https://www.graphpad.com/, GraphPad Software Inc.). Differences were considered statistically significant when P ≤ 0.05. Number of samples (n) is constituted of hPSC-LSC differentiation batches, each of individual cell batches being considered as biological replicates.
Results
Initial variability in gene expression levels observed in undifferentiated hPSCs
To investigate the cell line-dependent variability, the baseline expression of several genes critical for eye and corneal development was initially analyzed from undifferentiated hPSCs (UD-hPSCs). These genes included PAX6,* BMP4*,* LEF1*,* AXIN2*,* TGFB1*,* Id1* and Id3. The UD-hPSCs demonstrated some level of initial variation on gene expression levels (Fig. 2A). The relative expression level of both PAX6 and LEF1, when compared to a generalized control, varied not only between the different hPSC lines but also within the specific cell line as batch-to-batch variation (Fig. 2B). However, statistically significant differences were observed (P ≤ 0.05) only between Regea08/017 and WT003.TAU.bC lines with TGFB1 (Fig. 2C) and between Regea08/017 and AICS-0016 with LEF1 (Fig. 2D) expressions. Additionally, the investigation of TGF-β signaling pathway target genes Id1 and Id3 showed similarly initial variation in the gene expression levels, but no significant differences were observed between the cell lines (Supplemental Fig. 6A). These results suggest that some variations between cell lines and batches already exist in UD-hPSCs, which could potentially impact subsequent differentiation processes. To further illustrate the practical magnitude of batch-to-batch variation, data of individual UD-hPSC batches from Fig. 2B are presented in Supplemental Fig. 7A.
Fig. 2. Initial variation in the UD-hPSCs maintained in the E8 Flex + L521 system with weekend-free feeding regimen. Figure (A) displays the normalized Ct values, whereas Figure (B) presents the transformed data as relative gene expression levels, with the Regea08/017 UD-hPSCs sample average serving as an artificial generalized control for comparison. Figure (C) Statistically significant differences were observed between Regea08/017 and WT003.TAU.bC cell lines (P ≤ 0.05) in TGFB1 relative expression levels. Figure (D) Statistically significant differences were observed between Regea08/017 and AICS-0016 cell lines (P ≤ 0.05) in LEF1 relative expression levels.
The differentiation efficiency towards LSCs varies among the hPSC lines
To investigate the variability in gene and protein expression patterns during the differentiation of hPSCs towards LSCs, UD-hPSCs were differentiated towards the corneal lineage and characterized using RT-qPCR and IF. As a result, high variability was observed throughout the differentiation process. Importantly already at starting point, the used hPSC lines yielded different EB sizes at d4 of the differentiation and produced also morphologically variable cell cultures at the end of the differentiation protocol at d24 (Fig. 3A). IF analysis also revealed varying expression of typical LSC markers, including PAX6, p63α (or ΔNp63 (p40)), and CK14 at d24 (Fig. 3B). The low LSC marker expression levels coupled with mixed morphology indicated very inefficient differentiation in some of the used cell lines, such as iPSC line WT003.TAU.bC (Fig. 3A, B). In contrast, the hESC line Regea08/017 exhibited the most efficient differentiation, as indicated by both morphology and LSC marker expression (Fig. 3A, B). PAX6 staining were in line with RT-qPCR results, showing higher average PAX6 upregulation in Regea08/017 than in WT003.TAU.bC (Fig. 3C). Again, notable variation was detected between the cell lines and differentiation batches in the expression of PAX6,* LEF1*, and BMP4 genes (Fig. 3C). Notably, wide variability was observed in the expression of PAX6 at d7, LEF1 at d4 and d24, and BMP4 at d7 and d24. Although statistical analyses were conducted, no significant differences were detected between the lines, likely due to the limited number of biological replicates (n = 2–3) and high variability. In addition, these gene expression levels were also normalized to the generalized control, similarly than with the UD-hPSCs. These results are presented in Supplemental Fig. 5.
To further investigate the activity of TGF-β signaling pathway, the expression of TGFB1 together with the target genes Id1 and Id3 were analyzed during the differentiation. These results did not show any significant differences between the cell lines. However, slight TGFB1 upregulation trend was observed (Fig. 3C) together with downregulation of Id1 and Id3 (Supplemental Fig. 6B, C) at the d24 time point in some cell lines although differences were not statistically significant (P ≥ 0.05). All datapoints of investigated genes from Fig. 3C were presented by differentiation batches and cell lines in Supplemental Fig. 7B-G to further illustrate the practical magnitude of batch effects.
In addition, the expression levels of the LSC markers ABCG2, KRT15, and LGR5 at d24 revealed differences in the differentiation efficacy among the cell lines (Supplemental Fig. 8). Notably, ABCG2 expression was significantly lower in the Regea08/017 line compared to the WT003.TAU.bC line. Conversely, KRT15 expression was significantly higher with Regea08/017 than WT001.TAU.b2B and AICS-0016 lines. Furthermore, the expression of the early ectodermal marker KRT18 was assessed at d7 and d24 time points. While expression was generally higher at d7, no statistically significant differences (P ≥ 0.05) were observed between lines or time points (Supplemental Fig. 9). Nevertheless, the results suggest apparent differences in the differentiation patterns among the hPSC lines used.
Fig. 3hPSC-LSC differentiation using five different hPSC lines. (A) The hPSC lines formed embryoid bodies (EBs) of varying size at d4 and displayed morphological heterogeneity at the end of the differentiation at d24. Scale bar 250 μm. (B) Immunofluorescence of d24 cell cultures showed variable expression of PAX6 and LSC/corneal basal layer-related proteins p63 and CK14. Scale bar 100 μm. (C) RT-qPCR quantification of the expression levels of PAX6,* LEF1*,* TGFB1* and BMP4 during the 24-day differentiation, showing variation between the cell lines and differentiation batches.
Comparison of two hPSC lines at a late timepoint of the differentiation
To further investigate TGF-β signaling pathway activity, we analyzed the protein expression of p-smad2/3 with two selected hPSC lines, Regea08/017 and WT003.TAU.bC which demonstrated a marked difference in differentiation efficacy. The expression of p-smad2/3 was not detected at d7 and d10 (Supplemental Fig. 10A, B). However, at d24, dim cytoplasmic and nuclear staining was observed, indicating activated TGF-β signaling (Supplemental Fig. 10C). Interestingly, no differences in p-Smad2/3 expression were detected between the two cell lines at these time points.
Next, the differences at the d24 time point between the two selected hPSC lines, Regea08/017 and WT003.TAU.bC, were further investigated. The cells expressing corneal epithelial progenitor markers CK14, PAX6 and p63α at the late timepoint (d24) were quantified using FC and cytospin.
At d24 time point, cytospin analyses (Fig. 4A) revealed a slightly higher proportion of PAX6-positive cells in the Regea08/017 line. Notably, the proportion of PAX6+/CK14 + double-positive cells was observed to be substantially higher in Regea08/017, exceeding 45%, compared to only 6% in the WT003.TAU.bC line based on cytospin quantifications (Fig. 4A). The notable difference in differentiation efficacy between these compared cell lines was further confirmed, as two of the four differentiations batches with WT003.TAU.bC did not even reach the d24 timepoint due to major cell death, indicating inefficient differentiation towards LSCs. However, statistical analyses were not feasible for the cytospin quantification results due to the limited replicative samples (n = 2) and no statistically significant differences were observed (P ≥ 0.05) with FC analyses due to high variability between replicative samples in both used cell lines. To further highlight the practical impact of batch effects, the data shown in Fig. 4A were re-visualized by differentiation batches in Supplemental Fig. 11A.
Complementary FC analyses (Fig. 4B) showed that over 70% of Regea08/017 cells expressed the LSC marker p63α, and nearly 40% expressed PAX6. In contrast, in the WT003.TAU.bC line, only slightly over 35% of cells expressed p63α, and merely 14% expressed PAX6. However, no statistically significant differences were observed (P ≥ 0.05) with FC analyses due to high variability between replicative samples in both used cell lines. Taken together, the characterization analyses consistently demonstrated very distinct expression levels of the LSC markers and confirmed the earlier results indicating inefficient differentiation with WT003.TAU.bC cell line as compared to Regea08/017 cell line. The data of Fig. 4B were presented by differentiation batches in Supplemental Fig. 11B to further illustrate the practical magnitude of batch variation.
Fig. 4. Comparison of hPSC-LSC differentiation outcomes at d24 timepoint for two hPSC lines selected based on previous results. (A) Protein expression levels of CK14, PAX6 and double positive cells analyzed by cytospin quantification and (B) protein expressions of p63α and PAX6 assessed using FC analysis (P ≥ 0.05). Plots are presented with mean and standard deviation.
Discussion
Variability in hPSC differentiation rate is a well-recognized challenge^1,26,27^, often compromising the differentiation protocol robustness and increasing the unwanted heterogeneity of the differentiated cells^8^. Gaining a deeper understanding of this variability would be essential for improving control over the differentiation process and advancing the development of safe and effective hPSC-based therapies. In this study, we systematically examined and quantified hPSC line–specific differences in LSC differentiation efficacy which has been only sparsely addressed in the current hPSC-LCS literature. The LN521/E8 Flex culture system utilized in this study is widely accepted to support consistent and scalable hPSC expansion. The combination of single-cell passaging, xeno-free and chemically defined components, and a weekend-free maintenance regimen offers a robust platform for standardized hPSC production^16^. As a starting point, we decided to determine the baseline status of UD-hPSCs, by comparing the expression levels of the selected genes playing a key role in the ocular and corneal development in UD-hPSCs.
The gene selection was based on processes in the ocular development, where BMP4 and Wnt signaling is precisely balanced and regulated^28,29^. Early downregulation of Wnt signaling is critical for surface ectoderm specification, establishing the basis for ocular surface development^29^. Subsequently, the direction of BMP4 signaling determines the fate of these common progenitors, guiding their differentiation toward either lens or corneal epithelial lineages. PAX6, a master regulator of eye development, modulates Wnt signaling together with TGF-β^29^. In addition, LEF1 serves as a marker of canonical Wnt/β-catenin pathway activation, while AXIN2, a well-known target gene of this pathway, functions as a negative feedback regulator^28^.
As expected, we observed gene expression variation in UD-hPSCs, particularly in the PAX6 and LEF1 genes. This indicates heterogeneity in the starting material, which likely impacts the outcome of subsequent differentiation towards LSCs. Interestingly, the variation between different batches of individual hPSC lines was also high, making it impossible to propose that any of the selected genes could be used as reliable indicators of successful further differentiation. However, these results clearly indicate that, despite efforts to use standardized reagents, cell counting, and single-cell passaging of UD-hPSCs at similar densities, the UD-hPSC cultures exhibit a certain level of initial variation and further efforts would be needed to standardize the quality of the starting hPSCs materials for differentiation. Moreover, the individual values, particularly in comparison with generalized control, appear to be quite scattered with hESC line Regea08/017, suggesting considerable variability even in the most efficiently differentiating hPSC line.
Next, we differentiated all five hPSC lines towards LSCs with our previously established differentiation method^16–18^. Interestingly, already in the four days induction phase of the differentiation, additional variation was observed with the size of the formed EBs, their distinctive behavior during the four days induction phase, and challenges to ensure equal uniform plating densities of the EBs at d4 for following adherent differentiation. It is possible that the initial variability in gene expression at the UD-hPSC state may be associated with differences in the formation of EBs, but further investigation is needed to understand the possible correlations. Additionally, further analysis is required to determine if the size and form of EBs impact differentiation results. However, EBs formed through spontaneous aggregation can vary significantly in size and shape, which impacts the cell lineages that emerge during differentiation^30^. Moreover, the EB size can influence Wnt signaling pathways affecting to the lineage specification^31^. Taken together, these observations highlight intriguing early-stage differences between the hPSC lines suggesting potential associations between UD-hPSC variability and EB formation. While these findings provide valuable descriptive insight into the differentiation process, they do not yet uncover the mechanistic drivers underlying these differences. Future targeted studies will be required to identify the regulatory factors responsible for this variability. In addition, since part of the heterogeneity may arise specifically from the EB-based induction phase, future research could explore whether such variability can be mitigated through methodological refinements—for example, by implementing adherent differentiation approaches that offer more controlled and uniform early-stage conditions^32^.
Based on the further characterization of the hPSC-LSCs during the differentiation process until d24, clear difference between hPSC lines in the differentiation efficacy was observed. This was evident already at the morphological level, where some cell lines, WT003.TAU.bC in particular, displayed highly heterogenous cell populations and absence of typical epithelial cell shapes in comparison to e.g. Regea08/017 and AICS-0016. The more LSC-like phenotype of Regea08/017 and AICS-0016 lines was further supported by robust expression of PAX6, p63α, and CK14 marker proteins while WT003.TAU.bC exhibited low levels of these markers, which indicated variable differentiation efficacy between the lines. PAX6 is a widely acknowledged marker of the ectodermal differentiation which is strongly expressed in tissues derived from the surface ectoderm of the eye, including the CE^33,34^, as well as in neurons^35^. Together with basal epithelial stem cell/progenitor markers p63α and CK14^36,37^, it indicates the progression of differentiation in the correct direction, towards ocular lineage and the corneal LSCs. However, the WT003.TAU.bC line exhibited absence or dim IF staining of these markers compared to other cell lines, particularly the Regea08/017 hESC line. Considering the observations of morphology alongside protein expression patterns, the low expression of LSC markers coupled with mixed morphology indicated very inefficient differentiation, especially with the iPSC line WT003.TAU.bC. Notably, with this cell line, the differentiating cell cultures were successfully maintained up to d24 only in a few batches, highlighting the inefficiency of this line for the used protocol. In contrast, the hESC line Regea08/017 exhibited the most efficient differentiation, as indicated by both morphology and marker expression.
To further understand the origin of these differences, we investigated the expression patterns of genes critical for eye and corneal development. Similarly to morphology and protein expression results, the gene expression results highlighted the variation between the cell lines but also between individual differentiation batches. Notably high variation was observed with PAX6 and BMP4 at d7 timepoint but also in LEF1 expression in all time points. Importantly, all the observed genes play an important role in the specification of ocular surface ectoderm and this variability may impact the overall effectiveness of the differentiation protocol by d24. Although the investigated genes play a crucial role in eye development, they do not elucidate the factors contributing to this variability and further studies are required to fully understand the possible factors inducing variation and how these could be controlled. Moreover, the detected significant differences in LSC markers ABCG2 and KRT15 gene expressions at d24 between the cell lines highlights the cell line dependent variation in the differentiation efficacy. Even though, WT003.TAU.bC had higher expression of widely acknowledged markers of stemness the ABCG2 and LGR5^38,39^ the significantly higher expression of LSC marker KRT15^40^ with Regea08/017 line highlighted more efficient differentiation towards KRT15, CK14 and p63a positive LSC population with Regea08/017 line as compared to WT003.TAU.bC line. These findings, particularly the elevated expression of the ABCG2 marker in the WT003.TAU.bC line, were consistent with our previous observations^41^. However, these associations remain descriptive in nature and do not uncover the mechanistic basis of the observed differences in lineage outcomes.
To further understand, if the TGF-β signaling pathway activity could have an effect to the differentiation efficacy between lines we investigated expression of p-smad2/3 with two of the used hPSC lines—one with poor differentiation efficiency (WT003.TAU.bC) and one with high efficiency (Regea08/017). The phosphorylation of Smad2 and 3 indicates active TGF-β signaling pathway^42^. As no clear difference in the expression of p-smad2/3 was detected we conclude that difference in the differentiation efficacy cannot be explained by TGF-β signaling pathway activity at these timepoints. Moreover, the assessed TGFB1, Id1/Id3 and p-Smad2/3 data indicate that TGF-β signaling does not distinguish the efficient and inefficient lines, suggesting that other pathways or cell-intrinsic factors are more likely to underlie the divergent differentiation outcomes. Further studies are required to fully understand the potential factors influencing the distinct differentiation rates across lines.
Finally, to systematically conclude our investigation, we examined the differences between these two hPSC lines (Regea08/017 and WT003.TAU.bC) on day 24 of the differentiation using protein quantification. To emphasize the low differentiation efficiency, it is crucial to mention that differentiation results for WT003.TAU.bC were not obtainable from all batches due to inadequate cell yield at later stages. Here our focus was in the PAX6 and p63α which are considered as important markers of corneal epithelial fate specification during the differentiation of hPSCs^29,43^as well as in CK14 which is expressed in the basal layers of the limbal and CE and considered as CE marker^36^.
At d24 time point, cytospin analyses together with complementary FC analyses indicate that the Regea08/017 line exhibits a markedly higher efficiency in differentiating toward the LSC lineage compared to WT003.TAU.bC. This is supported by both cytospin and FC data, which consistently show elevated proportions of key LSC markers, including PAX6, CK14, and p63α, in Regea08/017 cells at d24. The substantially greater proportion of PAX6⁺/CK14⁺ double-positive cells further highlight the enhanced lineage commitment of this cell line. However, batch-to-batch variation was once again evident in these later time point analyses, underscoring the necessity for further studies to elucidate the underlying causes of this variability in efficiency and to develop strategies for controlling the heterogeneity.
The detected line- and batch-dependent variability aligns with previous studies that describe intrinsic heterogeneity among hPSC lines across both ocular and non-ocular differentiation systems^2,8,10,44^. These studies have demonstrated that differences in early transcriptional states, residual epigenetic memory from the donor cell type, and line-specific responsiveness to developmental signaling cues can influence downstream lineage commitment and maturation efficiency. In ocular differentiation, similar line-dependent effects have been reported during retinal organoid formation^44,45^together with our previous results in LSC specification^22^, indicating that the variability is a general property of hPSCs rather than a protocol-specific artifact. Early multi-omics profiling at initial differentiation stages (e.g. days 3–7) may identify molecular predictors of lineage competence before fully committing^1,46^. Our previously published data from single-cell RNA sequencing analyses of three hPSC lines across three LSC differentiation time points support our detected patterns and highlights hPSC–LSC heterogeneity^22^. However, these omics-level analyses did not identify predictive markers or additional mechanistic insights beyond phenotypic assessments presented. The prior work and our findings underscore the need for future studies and identify several directions for further investigation. In addition to early time-point profiling, incorporation of functional readouts such as clonogenicity, epithelial barrier formation, or wound-healing capacity could enable a more quantitative assessment of LSC potency beyond marker expression alone^47,43^. Expanding the number of hPSC lines studied may also help disentangle genetic from culture-induced sources of variability^46^. Together, these approaches could support more predictive qualification strategies for selecting hPSC lines suitable for scalable LSC manufacturing.
In conclusion, this study underlines the variability between hPSC-lines and among differentiation batches within individual cell lines. The data observed in this study highlights variability in the key signaling pathways already in the starting material. This variability is associated with heterogeneity during the differentiation phase. However, as our analyses were primarily descriptive, the dataset does not suggest any gene or protein with adequate predictive value for differentiation efficacy. Further work will be required to establish molecular markers that reliably indicate high quality starting material for efficient LSC differentiation. In addition, further standardization both to the UD-hPSCs and differentiation phases are required in order to improve differentiation consistency and to meet the clinical requirements. Implementing a clinically relevant differentiation protocol may require selecting an hPSC line with proven high differentiation efficiency. However, if several GMP-grade lines exhibit poor or inefficient performance, additional screening becomes necessary delaying the therapy development and increasing costs. In this context, assessing a minimal marker or phenotype panel at early time points across multiple batches and hPSC lines could provide an early quality control step in translational pipelines, enabling the efficient identification of promising lines for cell therapy development. Furthermore, such early-stage screening is directly compatible with GMP-compliant workflows and would support scalable LSC manufacturing for clinical applications. However, the substantial variation in differentiation rates between hPSC lines, combined with the limited availability of GMP-grade lines, further magnifies the challenges of producing clinically suitable cells. Together, these factors complicate efforts to achieve consistent and reliable cell production, which remains essential for successful clinical translation.
Finally, given the observational nature of this study, several limitations should be taken into account. This study was designed as exploratory analysis and is therefore constrained by the relatively small number of biological replicates, which limits statistical power and precludes mechanistic inference. Although multiple lineage‑ and pathway‑associated markers were examined, the dataset did not suggest any baseline marker capable of reliably predicting subsequent LSC differentiation performance. In addition, our analyses focused on early and intermediate stages of LSC differentiation and did not include functional stratification assays or the assessment of terminal corneal epithelial markers such as Keratin 3 or Keratin 12. Accordingly, the findings should be viewed as documenting reproducible patterns of variability and association rather than establishing causal determinants of LSC fate or functional maturity.
Supplementary Information
Below is the link to the electronic supplementary material.
Supplementary Material 1
The reference list from the paper itself. Each links out to its DOI / PubMed record.