Modulation of Wnt/BMP pathways during corneal differentiation of hPSC maintains ABCG2-positive LSC population that demonstrates increased regenerative potential

RESEARCH Open Access Modulation of Wnt/BMP pathways during corneal differentiation of hPSC maintains ABCG2-positive LSC population that demonstrates increased regenerative potential Meri Vattulainen 1† , Tanja Ilmarinen 1† , Laura Koivusalo 1 , Keijo Viiri 2 , Heidi Hongisto 1,3 and Heli Skottman 1* Abstract Background: The differentiation of corneal limbal stem cells (LSCs) from human pluripotent stem cells (hPSCs) has great power as a novel treatment for ocular surface reconstruction and for modeling corneal epithelial renewal. However, the lack of profound understanding of the true LSC population identity and the regulation of LSC homeostasis is hindering the full therapeutic potential of hPSC-derived LSCs as well as primary LSCs. Methods: The differentiation trajectory of two distinct hPSC lines towards LSCs was characterized extensively using immunofluorescence labeling against pluripotency, putative LSC, and mature corneal epithelium markers. Cell counting, flow cytometry, and qRT-PCR were used to quantify the differences between distinct populations observed at day 11 and day 24 time points. Initial differentiation conditions were thereafter modified to support the maintenance and expansion of the earlier population expressing ABCG2. Immunofluorescence, qRT-PCR, population doubling analyses, and transplantation into an ex vivo porcine cornea model were used to analyze the phenotype and functionality of the cell populations cultured in different conditions. Results: The detailed characterization of the hPSC differentiation towards LSCs revealed only transient expression of a cell population marked by the universal stemness marker and proposed LSC marker ABCG2. Within the ABCG2-positive population, we further identified two distinct subpopulations of quiescent ΔNp63α-negative and proliferative ΔNp63αpositive cells, the latter of which also expressed the acknowledged intestinal stem cell marker and suggested LSC marker LGR5. These populations that appeared early during the differentiation process had stem cell phenotypes distinct from the later arising ABCG2-negative, ΔNp63α-positive third cell population. Importantly, novel culture conditions modulating the Wnt and BMP signaling pathways allowed efficient maintenance and expansion of the ABCG2-positive populations. In comparison to ΔNp63α-positive hPSC-LSCs cultured in the initial culture conditions, ABCG2-positive hPSC-LSCs in the novel maintenance condition contained quiescent stem cells marked by p27, demonstrated notably higher population doubling capabilities and clonal growth in an in vitro colony-forming assay, and increased regenerative potential in the ex vivo transplantation model. (Continued on next page) © The Author(s). 2019 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. * Correspondence: [email protected] † Meri Vattulainen and Tanja Ilmarinen contributed equally to this work. 1 Faculty of Medicine and Health Technology, Tampere University, Arvo Ylpön katu 34, 33520 Tampere, Finland Full list of author information is available at the end of the article Vattulainen et al. Stem Cell Research & Therapy (2019) 10:236 https://doi.org/10.1186/s13287-019-1354-2 (Continued from previous page) Conclusions: The distinct cell populations identified during the hPSC-LSC differentiation and ABCG2-positive LSC maintenance may represent functionally different limbal stem/progenitor cells with implications for regenerative efficacy. Keywords: Human pluripotent stem cells, Stem cell differentiation, Limbal stem cells, Stem cell hierarchy, Stem cell maintenance, Wnt signaling, ABCG2, Limbal stem cell deficiency Background The constant homeostatic regeneration of the human corneal epithelium (CE) is maintained by limbal stem cells (LSCs) that reside in their specific niche structures in the palisades of Vogt of the limbus [1]. Disturbances in the renewal process due to LSC dysfunction or loss manifest as a clinical condition called limbal stem cell deficiency (LSCD) that may in unilateral cases be treated with autologous cultured limbal epithelial transplantation (CLET) [2]. On the other hand, differentiation of LSCs from human pluripotent stem cells (hPSCs), including both human embryonic stem cells (hESCs) and induced pluripotent stem cells (hiPSCs), represents a promising therapy option for patients suffering from the bilateral condition [3,4]. One of the most critical current challenges in the field of LSC therapies, with both primary and hPSC-derived cells, is the identification and maintenance of the clinically relevant LSC population. No specific single marker has been discovered to fill this gap. Thus, the phenotypic identification of LSCs currently relies on the expression of several stemness-related markers, combined with the absence of cytokeratins (CKs) 3 and 12 that mark terminally differentiated CE [5]. The tumor protein p63 is currently acknowledged as the classical identifier of colony-forming LSCs with proven clinical relevance and thus serves as a hallmark of high-quality CLET transplants [6,7]. Moreover, among different p63 isoforms, ΔNp63αhas been found to be the most abundantly expressed by LSCs [8,9]. Other suggested LSC markers include ABCB5, BMI1, Frizzled-7, C/EBPδ,andCK15[10–13]. In addition, studies in mice have pointed to ABCG2 being a universal marker of stemness in several tissues [14] and describing a slow-cycling subpopulation of colony-forming primary LSCs in humans [15,16]. Thus, ABCG2 has been widely acknowledged as a marker for putative LSCs in vivo [5]. Especially in the development of hPSC-based therapies, the potential presence of undifferentiated (UD) cells with tumorigenic potential among the transplantable cells raises an obvious safety concern. Markers that allow purification of the clinically relevant cell material by sorting would significantly promote both the safety and efficacy of the future treatments [17]. However, this issue is complicated by the fact that a low level of pluripotency markers is also expressed in the primary human limbal epithelium [18], and some of these markers are critically involved in the regulation of stemness [19]. A thorough understanding of the LSC differentiation hierarchy and the functional roles of various LSC-related markers is required to be able to address these questions. Recently, Bojic et al. identified two novel corneal cell surface proteins in primary cultured LSCs, namely, CD200, which marks a small quiescent population, and CD109, which is a marker for a more abundant proliferative progenitor cell type [20]. Both of these markers were coexpressed with the ΔNp63 isoform and demonstrated great proliferative capacity in vitro; however, only CD200positive cells generated holoclones that are a hallmark of a self-renewing stem cell population in vitro. The more detailed knowledge of the mutual relations among various LSC markers and their exact positions in the functional LSC hierarchy have remained largely unknown. In recent years, it has become increasingly evident that several stem cell niches in various tissues possess heterogenic stem cell subpopulations with distinct roles. For example, in the well-studied gut epithelium, LGR5-positive intestinal stem cells (ISCs) are located at the bottom of intestinal crypts together with nurturing Paneth cells, from where their short-lived transiently amplifying cell (TAC) progeny migrate towards the villi and terminally differentiate into mature intestinal epithelial cell types. On the other hand, it has been suggested that in the intestinal niche, there is also quiescent subpopulations of LGR5negative ISCs, which upon injury activate to repopulate both the intestinal epithelium and the LGR5-positive cell pool, demonstrating the extreme flexibility of tissue repair mechanisms in the intestine [21]. It is currently not known whether corresponding stem cell compartments can also be found in the limbus. Intriguingly however, it has been shown that LSCs share some mutual markers with ISCs, such as LGR5 [22,23] and BMI1 [11], suggesting potential stem cell compartmentalization in the limbal niche as well. In this study, we utilized our established protocol [24, 25] to address the question of whether distinct stem cell populations can be identified during the in vitro differentiation process of hPSC-derived LSCs. Extensive characterization of the protein expression patterns during the differentiation process was carried using a set of putative stemness, LSC, and mature CE markers. With this approach, we revealed the subsequent emergence of three cell populations, each of which exhibited Vattulainen et al. Stem Cell Research & Therapy (2019) 10:236 Page 2 of 15 a different LSC-associated phenotype. The first two populations appeared early during the process and consisted mainly of cells with strong expression of ABCG2 but different levels of ΔNp63α. After this phase, the population phenotype gradually shifted to form the third population with strong expression of ΔNp63αand no ABCG2. By employing the Wnt/BMP signaling modifiers traditionally used for culturing ISCs, we were able to regulate the maintenance of ABCG2-positive hPSC-LSCs in vitro.Importantly, in the functional experiments, these ABCG2-positive hPSC-LSCs demonstrated increased regenerative potential in comparison to the cell population expressing the ΔNp63αpositive phenotype. Materials and methods Experimental design Initial experimental design and progression of the study is presented in Fig. 1. The study consisted of two main parts, the first being the detailed characterization of hPSC differentiation process towards LSCs (Fig. 1a), and the second being establishing novel culture conditions for the maintenance of an ABCG2-positive LSC phenotype and further characterization of the stemness and functionality of the distinct populations observed in indicated time points and culture conditions (Fig. 1b). Full descriptions of the cell culture and cell characterization methods are provided as Supplemental Materials and Methods (Additional file 1). hPSC differentiation and hPSC-LSC culture All three hPSC lines used in this study (hESC lines Regea08/017 and Regea11/013 and hiPSC line UTA.04607.WT) were derived and characterized inhouse, as described previously [26,27]. Human PSC cultures were routinely maintained in serumand feeder cell-free conditions and differentiated towards the corneal epithelial lineage as described by Hongisto et al. Fig. 1 Flow chart of the experimental design and progression. aStandard CnT-30-based hPSC-LSC differentiation protocol and characterization of the hPSC-LSC differentiation process. bNovel CnT-07+ENRC-based hPSC-LSC maintenance protocol, characterization, and comparison of distinct cell populations identified during the study. PSC pluripotent stem cell, UD-hPSC undifferentiated human PSC, LSC limbal stem cell, IF immunofluorescence, qRT-PCR quantitative real-time PCR, LN-521 laminin-521, Col IV collagen type IV, E8 Flex, E8 Flex pluripotent stem cell culture medium, CnT-30 CnT-30 corneal differentiation medium, CnT-07 CnT-07 epithelial proliferation medium, ENRC epidermal growth factor, Noggin, R-Spondin-1, CHIR99021 Vattulainen et al. Stem Cell Research & Therapy (2019) 10:236 Page 3 of 15 [24,25]. In brief, UD-hPSCs were enzymatically dissociated to a single-cell suspension and transferred onto low-attachment plates for induction. Formation of embryoid bodies (EBs) was supported by adding 5 μM blebbistatin (Sigma-Aldrich) to the defined XF-Ko-SR medium for 1 day. During the following 3 days, XF-KoSR was first supplemented with 10 μM SB-505124 and 50 ng/ml human basic fibroblast growth factor (bFGF; PeproTech Inc., Rocky Hill, NJ) for 1 day and with 25 ng/ml bone morphogenetic protein (BMP)-4 (PeproTech Inc.) for 2 days, to push the differentiation towards surface ectoderm. EBs were then transferred onto plates coated with 0.5 μg/cm 2 recombinant laminin-521 (LN521, Biolamina, Sweden) and 5 μg/cm 2 human placental collagen Type IV (Col IV, Sigma-Aldrich) for adherent differentiation in defined commercial CnT-30 corneal differentiation medium (CELLnTEC Advanced Cell Systems AG, Bern, Switzerland). Cells were thereafter cultured in CnT-30, with medium changes three times per week until subjected to characterization analyses or further experimental settings. Throughout the article, the term “CnT-30 differentiation/condition”refers to the standard CnT-30-based differentiation protocol described in this chapter. 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). Establishment of the ABCG2-positive hPSC-LSC culture For maintenance of the ABCG2-positive population, CnT-30 was replaced with CnT-07 (CELLnTEC Advanced Cell Systems AG, Bern, Switzerland) supplemented with ENRC (50 ng/ml mouse recombinant epidermal growth factor (EGF, Invitrogen), 100 ng/ml mouse recombinant Noggin, 1 μg/ml human recombinant R-spondin (both from PeproTech), and 3 μMCHIR99021 (Stemgent)) at d10–11 of the standard CnT-30 differentiation protocol. New medium was introduced directly to the adherent cultures; alternatively, the hPSCLSCs were concomitantly passaged onto fresh LN-521/ Col IV-coated wells at a density of 1 000 cells/cm 2 in the new medium. The cells were thereafter cultured following the standard feeding regimen. After the emergence of ABCG2-positive colonies in approximately 7–10 days, further expansion of the ABCG2-positive hPSC-LSCs in the ENRC medium was carried out by passaging subconfluent cultures onto fresh LN-521/Col IV-coated matrices at a density of 1 000 cells/cm 2 . Batches of the passaged cells were also cryopreserved following our routine cryopreservation protocol [24,25]. Throughout the article, the term “ENRC maintenance condition” refers to the novel CnT-07-based, ENRC-supplemented culture protocol described in this chapter. Immunofluorescence Immunofluorescence staining (IF) was utilized in several stages during the study to analyze the protein expression of the cells, as presented in Fig. 1. In the characterization of standard hPSC-LSC differentiation in CnT-30, adherent cultures of UD-hPSCs as well as hPSC-LSCs at d7, d9, d11, d14, d17, d21, and d24 were stained with antibodies against OCT3/4, PAX6, ABCG2, ΔNp63, p63α,CK14, CK15, and CK12. Cytospin samples were prepared at d10 and d24 and stained with OCT3/4, ABCG2, p63α,ΔNp63, CK14, and CK15 for quantification of the different populations by cell counting analysis. Two hPSC lines (Regea08/017 and UTA.04607.WT) were used for the full IF characterization, and the results were replicated with at least two individual cell differentiation batches for both lines. Expression of indicated markers during standard differentiation of the third hPSC line, Regea11/013, was analyzed less extensively at d10–11 and d24–25. Adherent d10–11 and d21–24 hPSC-LSCs during Cnt30 differentiation as well as d21–24 hPSC-LSCs in ENRC maintenance were stained against ABCG2, p63α, LGR5, Ki67, and p27. Day 24 hPSC-LSCs in ENRC maintenance were characterized also for their OCT3/4, PAX6, CK14, and CK15 expression. Standard fixation and IF procedures with primary and secondary antibodies were performed essentially as described previously in Mikhailova et al. [28]. Raw images of the stained cells were captured with an Olympus IX51 fluorescence microscope. ImageJ Image Processing and Analysis tools [29] and Adobe Photoshop CC 2019 software were used for cell counting and image processing, respectively. Antibody specifics are provided as Supplemental Information (Additional file 2: Table S1). Flow cytometry and fluorescence-activated cell sorting UD-hPSCs and d10–11 and d24-d29 hPSC-LSCs during Cnt-30 differentiation were characterized for their ABCG2 and CD200 surface antigen expression using flow cytometry. For practical reasons, cryopreserved d26–d29 hPSCLSCs were used in some repeats in place of freshly differentiated cells, as we have previously demonstrated preservation of the phenotype throughout this process [24]. Standard flow cytometry staining protocols were used in the sample preparation, following the recommendations provided by the antibody manufacturers. APC-conjugated monoclonal mouse anti-human CD338 (ABCG2) antibody, clone 5D3 (BD Pharmingen, #561451), and PE-conjugated mouse monoclonal CD200 (clone OX-104) antibodies from two manufacturers (BioLegend, #329205 and BD Pharmingen, #561762) were used for indicated cell populations. An APC-conjugated mouse IgG2b κantibody (BD Pharmingen, #555745) or unstained cells were used as isotype and/or negative controls, respectively. CD200-stained samples were also double-stained with ABCG2. Vattulainen et al. Stem Cell Research & Therapy (2019) 10:236 Page 4 of 15 For continued culture of pure ABCG2-positive cell population, 1 000 Regea08/017 hPSC-LSCs staining positive for ABCG2 at d11 were sorted directly onto LN521/Col IV-coated wells in CnT-30 supplemented with a 10 μM concentration of the Rho kinase inhibitor Y-27632 (Tocris Bioscience) and were then cultured following the standard feeding regimen for 17 days. The sorted cells were stained in IF against p63αand ABCG2, and expression of p63αwas quantified by cell counting. Both flow cytometry analyses and cell sorting were performed using a BD FACSAria™Fusion cell sorter operating with the FACSDiva™software (BD Biosciences, San Jose, California, USA). At least 10 000 events were recorded from the initially gated populations and analyzed with FlowJo 10 software (BD Biosciences, San Jose, CA, USA). Quantitative RT-PCR UD-hPSCs and d10–11 and d21–24 hPSC-LSCs during Cnt-30 differentiation as well as d21 hPSC-LSCs in the ENRC maintenance were analyzed for their ABCG2 mRNA expression with qPCR, using a sequencespecific TaqMan Gene Expression Assay for ABCG2 (#HS01053790_m1, Applied Biosystems). GAPDH (Hs99999905_m1) was used as a housekeeping gene. Standard methods were used for RNA isolation and cDNA synthesis from the cell pellet samples collected in indicated time points. All samples and controls were run as triplicate reactions with the 7300 Real-Time PCR system (Applied Biosystems). The results were analyzed using the −2 ΔΔCt method [30]andarepresented as the fold change in gene expression normalized to GAPDH and relative to the UD controls. Cell surface antigen screening Cell surface marker screening was performed for the Regea08/017 hPSC-LSCs at d10 during Cnt-30 differentiation, using the LEGENDScreen™Lyophilized Antibody Array, Human PE Kit (BioLegend, #700007) and essentially following the instructions of the manufacturer. The experiment was carried out in four parts using 120 000– 150 000 cells per sample that were double-stained with 3μl of APC-conjugated ABCG2 antibody (BD Pharmingen, #561451). At least 10 000 initially gated events per sample were recorded with FACSCanto II flow cytometer (BD Biosciences, San Jose, CA, USA) and analyzed with FlowJo software. The results were compared to the previously published screening data of cultured primary human LSCs, produced by Bojic et al. [20], with special emphasis on the two novel LSC markers, CD109 and CD200. Population doubling analyses Population doubling calculations were carried out for the Regea11/013 hPSC-LSCs cultured in ENRC, both prior and after the standard cryopreservation protocols described in Hongisto et al. [24]. Population doublings (PDs) at the end of each subculture were calculated using the following formula: log (N/N0)/log2, where N0 is the number of plated cells and Nis the number of cells at the end of the culture period. Similarly, the population doubling time (PDT) for each passage was calculated with the following formula: T× log2/log(N −N0), where Tis the duration of the culture in hours. Ex vivo transplantation into a porcine cornea model Porcine corneas were obtained and processed for corneal ex vivo culture as previously described in [31,32]. The excised corneas were maintained in CnT-CC medium (CELLnTEC Advanced Cell Systems AG, Bern, Switzerland) for up to 3 weeks prior to transplantation experiments. LSCD mimicking state was induced to the ex vivo corneas by placing a filter paper disc soaked with 1 M sodium hydroxide (NaOH) onto the corneal surface for 40 s, followed by thorough removal of the epithelium by scraping. Human PSC-LSCs cultured in CnT-30 and ENRC conditions were seeded onto both sides of a fibrin carrier membranes at a density of 30 000 cells/cm 2 and thereafter cultured for 2 additional weeks in their initial medium conditions. The carrier membranes were transplanted into the ex vivo corneas using four interrupted 9-0 Vicryl sutures. Soft contact lenses were placed on top of the corneas to prevent drying. After ex vivo transplantation, the hPSC-LSCs from CnT-30 conditions were further cultured in CnT-30 for 1 week (n= 2). The hPSC-LSCs from ENRC conditions were cultured in CnT-07 for 1 week (n= 2) or CnT-30 for 2 weeks (n= 3). The medium was gradually changed from CnT-07 + ENRC to CnT-30 through one intermediate step with 1:1 ratio of the two media. During ex vivo culturing, all media were supplemented with 5% fetal bovine serum (FBS, Sigma-Aldrich), 1% penicillin/streptomycin and 0.1% amphotericin B (Sigma-Aldrich). Media were replaced three times a week. After 1 to 2 weeks, the ex vivo corneas were fixed, processed into paraffin-embedded tissue sections, and stained with hematoxylin-eosin (HE), following the standard methods. Images of the HE-stained tissue sections were captured with the Nikon Eclipse TE2000-S microscope and DS-Fi1 camera. Statistical methods All data are presented as the mean ± standard deviation (SD). Whenever n≥3, the Mann-Whitney Utest was performed to analyze the differences between the groups using the GraphPad Prism 5 software (GraphPad Software Inc.). Differences were considered statistically significant when P≤0.05. Vattulainen et al. Stem Cell Research & Therapy (2019) 10:236 Page 5 of 15 Results Dissection of the hPSC-LSC differentiation hierarchy reveals the subsequent emergence of three separate cell populations, marked by distinct expression patterns for ABCG2 and ΔNp63α To investigate protein expression patterns during the differentiation of hPSCs towards LSCs, UD-hPSCs were differentiated towards the corneal lineage and characterized using IF. During the time frame ranging from d7 to d24, expression of OCT-3/4 was markedly downregulated, whereas expression of PAX6, ΔNp63α, CK15, and CK14 increased, indicating the emergence of an LSC-like population (Fig. 2a). Interestingly, ABCG2 was expressed only transiently, peaking between d9 and d11 and then gradually decreasing to very low levels by d24 (Fig. 2a). In accordance with our previous results [24], CK12 remained undetectable within this timeframe (data not shown). Quantification of cell populations by protein expression confirmed major differences between the d10 and d24 time points, as shown for the representative hESC line (Regea08/017) in Fig. 2b. To be precise, the expression of ABCG2 decreased from 62.3% (SD 6.7) to 1.8% (SD 0.9), while the expression of ΔNp63α(as demonstrated by double-staining with ΔNp63 and p63αantibodies, Fig. 2c) increased from 23.2% (SD 14.1) to 54.3% (SD 6.2). CK15 and CK14 were undetectable at d10 but increased to 37.0% (SD 12.4) and 56.2% (14.3) by d24, respectively. OCT3/4, on the other hand, was expressed in less than 1.5% of the cells at d10 and was further diminished to under 1% by d24. Due to the distinct expression profiles of ABCG2 and ΔNp63αduring the differentiation process, we further characterized the coexpression of ABCG2 with p63αin d10 and d24. Interestingly, the most intense expression of both markers was typically observed in separate cell populations, as demonstrated in Fig. 2d (see also Additional file 3: Figure S1), suggesting a transitional rather than a stable phase for the coexpression. At d10, ABCG2 and p63αwere coexpressed in 31.6% of the cells, whereas at d24 only 1% of the cells were doublepositive (Fig. 2e). Flow cytometry confirmed the IF results by showing that both UD-hPSCs and more differentiated d24–26 hPSC-LSCs had low expression of ABCG2 (0.8%, SD 1.3 and 1.5%, SD 2.0, respectively, for the representative hESC line Regea08/017), whereas d10–11 hPSC-LSCs expressed significantly higher levels of ABCG2 (21.6%, SD 8.2) than both UD-hPSCs (P= 0.0238) and d24–26 hPSC-LSCs (P= 0.0275) (Fig. 2f, see also Additional file 4: Figure S2A for representative flow cytometry plots). Lower percentage of ABCG2 expressing cells in flow cytometry in comparison to cytospin quantification is likely in large part a technical issue related to the chosen methods, as standard IF sample processing (including permeabilization) also allows staining of the intracellular ABCG2, resulting in more abundant positive signal, whereas in live cell flow cytometry only the cells expressing ABCG2 on their surface membrane label as positive. Notably, isolation of ABCG2-positive hPSC-LSCs at d11 leads to the formation of homogeneous hPSCLSC monolayers, with 99.9% (SD 0.2, n= 2800 cells) nuclear p63αexpression and very low to undetectable levels of ABCG2 (Additional file 4:FigureS2B). Additionally, changes in ABCG2 expression at the mRNA level further verified notably higher expression level of ABCG2 in the d10 hPSC-LSC population than in UD-hPSCs and the d24 hPSC-LSC population (Fig. 2g). Taken together, the characterization analyses consistently demonstrated very distinct expression profiles for the proposed LSC/progenitor markers ABCG2, ΔNp63α, CK15, and CK14 between the d10–11 and d24 time points during the hPSC-LSC differentiation. Importantly, the described expression patterns were reproduced with the hiPSC line UTA.04607.WT (Additional file 5: Figure S3). Screening of LSC-associated surface markers revealed a high level of CD200-positive cells among d10 hPSC-LSCs At d10 in Cnt-30 culture, Regea08/017 hPSC-LSCs expressed several limbus-associated markers at levels comparable to those of human primary cultured LSCs, as reported by Bojic et al. [20](Table1). Furthermore, we specifically investigated the two markers highlighted by Bojic et al. as novel candidates for quiescent LSCs and the proliferative progenitor phenotype, namely, CD200 and CD109. CD200 was expressed at a considerably higher level in our d10 hPSC-LSCs than in cultured primary human LSCs (42.6% vs. 2.3%, respectively). On the other hand, CD109 was expressed in only 25.7% of hPSC-LSCs, in comparison to 56.3% of primary cultured LSCs. Interestingly, ABCG2 was expressed in approximately half of both the CD200and CD109-positive hPSC-LSC subpopulations, thus showing no preferred coexpression with either of these markers. To be precise, there were 48.1% ABCG2-positive and 51.9% ABCG2-negative cells in the CD200positive population, and in the CD109-positive population, these numbers were 58.5% and 41.5%, respectively (Additional file 6:TableS2). As the role of CD200 was recently investigated also in hPSC-derived corneal cells [33], we performed flow cytometry analysis to further analyze the expression pattern of CD200 during differentiation of hPSC-LSCs. The results were consistent with both used CD200 antibodies, unambiguously showing that CD200 was expressed in over 99% of UD-hPSCs (Fig. 3a). During the Cnt-30 differentiation, CD200 expression decreased Vattulainen et al. Stem Cell Research & Therapy (2019) 10:236 Page 6 of 15 from 87.6% (SD 0.7) at d11 to 38.5% by d29 (Fig. 3b, c), whereas ABCG2 expression followed a typical trend, with low expression in 3.3% (SD 1.8) of UD-hPSCs, 57% (SD 5.9) of d11 cells, and 7.4% (SD 0.9) of d29 cells (Fig. 3a–c). Again, based on our results, the expression pattern of CD200 did not correlate with that of ABCG2, and in line with the screening results, there were both ABCG2-positive and ABCG2-negative hPSCLSCs in the CD200-positive subpopulation. However, we were unable to confirm this specific finding with IF visualization using the unconjugated primary CD200antibody from BioLegend. Fig. 2 Characterization of putative LSC marker expression during hPSC-LSC differentiation. aRepresentative morphology and protein expression of the cultures at selected time points. Scale bars, 100 μm for all images in the same column. Cell nuclei counterstained with DAPI (blue). b Marker expression differences in the d10 and d24 populations. Five images per sample and a minimum of 1400 cells per time point were analyzed for each marker from cytospin samples. cRepresentative IF image of ΔNp63 and p63αdouble-staining in a d24 cytospin sample. Scale bar, 100 μm for both cand d.dRepresentative IF image of ABCG2 and p63αdouble-staining in a d10 cytospin sample. ep63αand ABCG2 expression in d10 and d24 hPSC-LSCs. Five images per sample and a minimum of 3 000 cells per time point were analyzed from cytospin samples. P> 0.05. fThe level of ABCG2 protein expression in UD-hPSCs and in d10 and d24–26 hPSC-LSCs, analyzed with flow cytometry. gThe ABCG2 mRNA expression levels in UD-hPSCs and in d10 and d24 hPSC-LSCs analyzed with qRT-PCR. All representative data are presented with the hESC line Regea08/017. All quantitative data are presented as the mean + SD, and nmarks the individual cell differentiation batches serving as biological replicates. Statistical analyses were carried out using the Mann-Whitney Utest. *P≤0.05 Vattulainen et al. Stem Cell Research & Therapy (2019) 10:236 Page 7 of 15 Maintenance of ABCG2 expression during hPSC-LSC culture is achieved with Wnt/BMP pathway signaling modifiers previously established for the long-term sustained growth of intestinal organoids in vitro To study the properties of the ABCG2-positive cell population in more detail, culture conditions aiming to preserve the ABCG2 expressing cell population for extended periods were established. Due to similarities between the limbal and intestinal crypt stem cell niches and the signaling pathways involved in the regulation of stemness, we hypothesized that Wnt agonists and a BMP antagonist previously used for intestinal crypt organoid culture could be utilized for regulating the differentiation process of hPSC-LSCs as well. Remarkably, replacing the corneal differentiation medium CnT30 with the epithelial maintenance medium CnT-07 and specific combination of EGF, Noggin, R-spondin-1, and CHIR-99201 (ENRC) at day 11 of differentiation, resulted in the preservation of the colonial morphology and strong ABCG2 protein expression that were observed only transiently during the CnT-30 differentiation. The main differences between the standard CnT-30-based differentiation condition and the novel CnT-07+ENRC-based maintenance condition are shown in Fig. 1.Notably, colonies positive for ABCG2 in the ENRC condition expressed only low levels of p63α, whereas time pointmatched cells in CnT-30 were ABCG2-negative and p63α-“bright”(Fig. 4a). These IF results were confirmed with all three studied hPSC lines. Importantly, the effect of the ENRC condition was consistent in all lines, despite cell line-specific variations in the efficacy. Prominent upregulation of ABCG2 at the mRNA level in the ENRC condition in comparison to the CnT-30 condition was confirmed also with qRTPCR using the hESC line Regea08/017 (Fig. 4b). Additional characterization of PAX6, CK14, CK15, and OCT3/4 expression in Regea11/013 hPSC-LSCs in the maintenance condition demonstrated positive expression for PAX6, weak expression for CK14 and CK15, and negative expression for OCT3/4 (Additional file 7: Figure S4A). ABCG2-positive limbal colonies coexpress the stem cell marker LGR5 Due to the finding that ENRC supplementation, which preserves LGR5-positive ISCs, also supports the maintenance of the ABCG2-positive LSC phenotype, we decided to analyze the expression of LGR5 in our cells and compare its expression pattern to those of both ABCG2 and ΔNp63. Indeed, prevalent expression of LGR5 was observed during CnT-30 culture at d11, followed by decreased expression upon further differentiation up to Table 1 Expression of selected cell surface markers in d10 Regea08/017 hESC-LSCs Surface marker hPSC-LSCs (%) Primary human LSCs [20] (%) EGFR 97.7 88.8 CD71 81.9 88.8 Integrin β5 93.5 91.5 Integrin α6 (CD49f) 98.6 92.5 E-cadherin (CD324) 85.9 88.5 CD40 24.7 26.0 CD146 84.0 67.0 CD166 99.5 95.1 CD200 42.6 2.3 CD109 25.7 56.3 Fig. 3 CD200 and ABCG2 expression patterns during differentiation of hPSC towards LSCs, analyzed with flow cytometry. Scatter plots and adjunct histograms as well as tables showing the distribution of cells in the CD200 + /ABCG2 − (Q1), CD200 + /ABCG2 + (Q2), CD200 − /ABCG2 + (Q3), and CD200 − /ABCG2 − (Q4) subpopulations of UD-hPSCs (a) as well as at d11 (b) and d29 (c) during hPSC-LSC differentiation. For each sample, 10,000 initially gated events were analyzed, and the experiment was carried out once for the representative line Regea08/017 Vattulainen et al. Stem Cell Research & Therapy (2019) 10:236 Page 8 of 15 d24. On the other hand, in time point-matched colonies in ENRC, the expression of LGR5 was highly conserved (Fig. 4c). Interestingly, at d11 during the CnT-30 culture, LGR5 was coexpressed with ΔNp63, as marked by distinct membrane-localized staining at cellular junctions of ΔNp63-positive cells (Fig. 4c). Further differentiation under CnT-30 conditions, resulting in the formation of a ΔNp63-positive epithelial monolayer, was accompanied by concomitant loss of strong LGR5 expression on the cell surface, a phenomenon similar to the expression pattern of ABCG2. Under ENRC conditions, LGR5 expression was localized to cellular junctions, again similar to ABCG2. Fascinatingly, as in the ENRC, the colonies expressed only low levels of ΔNp63; LGR5 thus appeared to switch its preferred coexpression with ABCG2 and ΔNp63, depending on the culture conditions. ABCG2-positive hPSC-LSCs retain clonal growth and proliferative capacity during passaging in the novel maintenance condition Population doubling analyses were carried out to determine the proliferative capacity of hPSC-LSCs. Under ENRC maintenance, colony morphology and the ABCG2/p63αexpression pattern were preserved at least up to passage 10 (Fig. 5a). As comparison, passaging and subsequent culturing in CnT-30 medium resulted in loss of the colony morphology and ABCG2 expression and promoted further differentiation towards ΔNp63α-positive epithelial monolayers, as described for the hPSCLSC differentiation process (shown in, e.g., Fig. 4a). In addition, the proliferation of hPSC-LSCs is rapidly diminished upon passaging in CnT-30 and generally ceases after the third passage, as repeatedly demonstrated with various cell lines during our standard hPSCLSC cell culture routine. To quantify the proliferative capacity of the cells, Regea11/013 hPSC-derived LSCs were frequently passaged under ENRC conditions, and their PDs and PDTs were calculated. Freshly differentiated Regea11/013 hPSC-LSCs were cultured for five passages, during which the cells went through over 20 population doublings, with an average PDT of 50.9 h (SD 16.4) (Fig. 5b, c, black line). Notably, cryopreservation between passages 2 and 3 (p2-p3) did not have a marked effect on Fig. 4 Effect of culture conditions on hPSC-LSC morphology and p63α, ABCG2, and LGR5 expression.aRepresentative cell morphology and p63α/ABCG2 protein expression under different culture conditions, as demonstrated by IF. Scale bars, 100 μm. bABCG2 mRNA expression under different conditions, analyzed with qRT-PCR in d21 cells. Data are presented as the mean + SD, n= 3 technical replicates from one sample, P> 0.05. cCharacterization of LGR5 protein expression in relation to ABCG2 and ΔNp63 at d11, as well as after continued culture in CnT-07+ENRC or CnT-30 at d24. Cell nuclei counterstained with DAPI (blue). Scale bars, 50 μm. Data shown for the representative hESC line Regea08/017 Vattulainen et al. Stem Cell Research & Therapy (2019) 10:236 Page 9 of 15