Human PSCs determine the competency of cerebral organoid differentiation via FGF signaling and epigenetic mechanisms

Summary Various culture methods have been developed for maintaining human pluripotent stem cells (PSCs). These PSC maintenance methods exhibit biased differentiation; for example, feeder-dependent PSCs efficiently yield cerebral organoids, but it is difficult to generate organoids from feeder-free PSCs. It remains unknown how PSC maintenance conditions affect differentiation. In this study, we identified fibroblast growth factor (FGF) signaling in feeder-free PSC maintenance as a key factor that determines the differentiation toward cerebral organoids. The inhibition of FGF signaling in feeder-free PSCs rescued organoid generation to the same level in feeder-dependent cultures. FGF inhibition induced DNA methylation at the WNT5A locus, and this epigenetic change suppressed the future activation of non-canonical Wnt signaling after differentiation, leading to reliable cerebral organoid generation. This study underscores the importance of PSC culture conditions for directed differentiation into cerebral organoids, and the epigenetic status regulated by FGF signaling is involved in the underlying mechanisms.


INTRODUCTION
Human pluripotent stem cells (PSCs), including embryonic stem cells (ESCs) and induced PSCs (iPSCs), can generate virtually all cell lineages of the body, offering in vitro models of human developmental processes, platforms for studying disease pathogenesis, and cell sources for regenerative medicine (Imaizumi and Okano, 2021;Okano and Yamanaka, 2014). In particular, PSC-derived neural cells are of great interest given that it is difficult to obtain human neural cells or tissues because of the limited accessibility to the human brain. Conventional adherent two-dimensional (2D) culture systems provide highly pure populations of neural cells from PSCs; however, 2D cultures cannot reproduce the characteristic three-dimensional (3D) structures of the brain. Recently, some studies have addressed this disadvantage by developing 3D cultures that resemble developing brains, named organoids Lancaster et al., 2013;Pasca et al., 2015;Qian et al., 2016). Brain organoid technology has provided researchers with a unique opportunity to study human neurodevelopmental processes and neurological disease pathogenesis that were previously inaccessible.
Substantial innovations have also been made in the field of technology for maintaining human PSCs. Traditional human PSC cultures require mouse-derived feeder cells, often resulting in perturbations of PSC differentiation. The contamination of mouse-derived cells is also unfavorable for the clinical application of PSCs. Recent advances have made it possible to maintain PSCs without feeder cells under defined culture conditions (Chen et al., 2011;Ludwig et al., 2006;Nakagawa et al., 2015). Although it is widely accepted that PSCs have a similar cellular identity regardless of maintenance methods, some studies raised the possibility that PSC maintenance protocols with or without feeder cells affect the differentiation capacity of PSCs . Especially in the case of organoid technology, methods for organoid generation were initially established in feeder-dependent PSC cultures, and previous reports indicated that it is difficult to directly apply these methods to feeder-free PSC cultures (Kuwahara et al., 2019;Watanabe et al., 2019). These studies suggest that the culture conditions of PSCs affect organoid generation; however, such differences among PSC culture methods have yet to be extensively explored.
In this study, we found that transient FGF inhibition at the undifferentiated stage rescued cerebral organoid generation from feeder-free PSCs. FGF inhibition in feeder-free PSCs could produce mature cerebral

OPEN ACCESS
As the culture conditions of PSCs affect organoid generation, we hypothesized that drug treatment can convert the initial PSC condition and enhance successive organoid generation. Indeed, it has been reported that the manipulation of some signaling pathways in PSC cultures before differentiation drastically enhanced the differentiation efficiency without changing derivation methods (Fujimori et al., 2017;Kuwahara et al., 2019;Watanabe et al., 2019). These manipulated signaling pathways in previous reports include TGFb, BMP, Wnt, Shh, and FGF, and we performed small-scale screening of drugs, including ActivinA, TGFb3, BMP4, SB431542 (a TGFb inhibitor), LDN193189 (a BMP inhibitor), IWR1e (a Wnt inhibitor), SAG (an Shh agonist), and PD173074 (an FGF inhibitor; referred to hereafter as PD17), for two days before differentiation ( Figure S1A). Only PD17 and SAG treatment enhanced the round shape formation with neuroepithelium-like structures (Figures S1B and 1C). FF-PSCs treated with TGFb3, SB431542, LDN193189, and IWR1e formed similar collapsed aggregates to the control condition, and BMP4-and ActivinA-treated aggregates were significantly smaller. The forebrain marker FOXG1 was highly upregulated in PD17-or SAG-treated cells; on the other hand, only PD17-treated cells had high expression of another forebrain marker, LHX2 (Figures S1C-S1E). The upregulation of the expression of these two markers was comparable to that of OnF-PSC-derived organoids ( Figures 1D and 1E). We also confirmed that PD17 downregulated the expression of DUSP6, a downstream gene of FGF signaling, which indicates that PD17 indeed suppressed FGF signaling ( Figure S1F). This PD17-induced organoid generation was also confirmed in other PSC lines [414C2 iPSCs  and KhES1 ESCs (Suemori et al., 2006)] and under the mTeSR1 method (Figures S1G-S1L).
With prolonged culture, OnF-PSC-derived cerebral organoids acquired neuroepithelial domes with a ventricle-like cavity inside, and these dome structures were also seen in FF + PD17-PSC cultures ( Figures 1F-1H and S2A). In OnF-and FF + PD17-PSC cultures, immunostaining analyses showed that there were SOX2-positive cell-dense ventricular zone (VZ)-like structures on the apical luminal side and TUBB3positive neurons outside of the VZ, reminiscent of the cortical plate (CP) ( Figures 1I, 1J, and S2B). On the other hand, FF-PSC-derived aggregates contained multiple cysts surrounded by a thin cell layer, and there were no VZ-or CP-like structures ( Figures 1F-1J), indicating that FF-PSC-derived aggregates have a different cellular identity from the cerebral cortex. These results suggest that FGF inhibition promoted the generation of cerebral organoids from FF-PSCs to a level similar to that of OnF-PSCs. iScience Article Cerebral organoids derived from feeder-free PSCs with FGF inhibition recapitulate corticogenesis FF + PD17-PSC-derived organoids were subjected to long-term culture. On Day 80, the deep-layer marker CTIP2 and the upper-layer marker SATB2 were mostly co-expressed in the same domain of cerebral organoids, and their distribution did not separate into distinguishable layers (Figure 2A), reminiscent of early stage fetal brains and organoids of a similar age (Qian et al., 2020). On the other hand, in the prolonged culture, by Day 120, upper-layer neurons had preferentially localized more superficially to deep-layer neurons ( Figure 2A). Another pair of markers, RORB and TBR1, also had biased expression patterns, indicative of the upper and deep layers, respectively, on Day 120 ( Figure 2B). These observations indicate that FF + PD17-PSC-derived organoids formed the laminar structures in a time-dependent manner, as previously reported for OnF-PSC-derived organoids Qian et al., 2016).
iScience 25, 105140, October 21, 2022 iScience Article enriched in dividing progenitor (DP) and radial glia (RG) clusters, and the expression of the mitotic marker MKI67 distinguished DP and RG clusters ( Figure 2D). EOMES, also known as TBR2, marked the intermediate progenitor cell (IPC) cluster, and NR4A2, also known as NURR1, was exclusively expressed in the subplate (SP) cluster ( Figure 2D). BCL11B (CTIP2) was enriched in the deep-layer, corticofugal projection neuron (CFuPN) clusters, and upper-layer, callosal projection neuron (CPN) clusters had high expression of SATB2 ( Figure 2D). A small portion of cells expressed DLX2, indicative of ganglionic eminence (GE) identity ( Figures 2C and 2D). RNA velocity analysis  showed the trajectories from DP and RG clusters to neuronal clusters ( Figure S3A). There were also distinct streams derived from IPC and SP clusters projecting to CFuPN and CPN clusters, respectively ( Figure S3A). These results imply that organoids derived from FF + PD17-PSCs had a cell-type diversity similar to that of the cerebral cortex, including radial glia, progenitors, and various neuronal subtypes.
To further validate the cerebral identity of organoids from FF + PD17-PSCs, we compared our scRNA-seq dataset with that of a previous study . The integration between these two datasets revealed that each dataset has a similar heterogeneity ( Figures S3B-S3E). In addition, we examined the similarity between our organoids and mouse/human embryonic brains using VoxHunt  by mapping scRNA-seq data to reference atlases, including the Allen Developing Mouse Brain Atlas and the BrainSpan Developing Human Brain Atlas. Our organoids were specifically mapped onto the cerebral cortex of the embryonic day (E) 13.5 mouse brain (Figures 2E and S3F). When compared with human fetal primary tissues, every cluster had a high correlation with the cortical pallial structures, including the neocortex, hippocampus, and amygdala, whereas the GE cluster exhibited a relatively high similarity to the GE-derived striatum ( Figure S3G). Overall, these analyses demonstrate that FF + PD17-PSC-derived organoids have transcriptomic profiles close to those of the cerebral cortex.
Undifferentiated on-feeder and feeder-free PSCs have similar transcriptomic profiles and become apparently distinct upon differentiation Next, we investigated how FGF inhibition enhanced organoid generation in FF-PSC cultures. We performed bulk RNA-seq of undifferentiated OnF-, FF-, and FF + PD17-PSCs and their derivatives on Day 6 ( Figure 3A). Principal component analysis (PCA) indicated that among OnF-, FF-, and FF + PD17-PSCs, there were relatively small transcriptomic differences in the undifferentiated state, and the differences became more apparent on organoid induction ( Figure 3B). We confirmed that PD17 downregulated the expression of FGF signaling target genes, including DUSP6, ETV4/5, SPRY4, and IL17RD, in FF-PSCs, indicating that PD17 indeed suppressed FGF signaling ( Figure 3C and Table S1). We also validated that the cerebral markers FOXG1 and LHX2 were upregulated in FF-PSC-derived 6-day differentiating cells by PD17 treatment (Figure 3C and Table S2). As a previous report suggested that naive/primed state transitions are induced by OnF and FF cultures, thereby affecting organoid generation , we examined the transcriptomic differences on Day 0 in more depth with respect to naive pluripotency. To this end, we performed a comparative analysis of our RNA-seq dataset with a range of transcriptomic datasets of OnF-and FF-PSCs, as well as chemically reset and embryo-derived naive PSCs Guo et al., 2017;Matsuda et al., 2020;Takashima et al., 2014;Watanabe et al., 2019). Neither OnF-nor FF-PSCs expressed naive pluripotency markers, such as DPPA3 and TBX3 ( Figure S4A). In addition, PD17 treatment did not induce naive PSC marker expression ( Figure S4A). These observations suggest that the culture conditions of PSCs (OnF or FF) and FGF inhibition do not contribute to the naive/primed state transition.

Feeder-free PSCs predominantly differentiate into the neural crest lineage
We performed gene set enrichment analyses (GSEA) using these Day 0 and Day 6 RNA-seq data. Although no gene sets were enriched on Day 0, there were some statistically significant gene sets on Day 6 (Figure S4B and Table S3), supporting that the gene expression change by PD17 became more distinct on differentiation in contrast to the undifferentiated stage. In particular, the gene expression change on Day 6 was highly associated with the neural crest lineage ( Figures S4B-S4D). When compared with the RNAseq data from PSC-derived ectodermal lineages, including the neural crest and the neuroectoderm , the gene expression profile of FF-PSCs on Day 6 exhibited a high correlation with that of the neural crest lineage, whereas the gene expression profiles of OnF-and FF + PD17-PSCs on Day 6 exhibited a high similarity to that of the neuroectoderm ( Figure 3D). Indeed, neural crest markers, including SOX10 and FOXD3, were highly upregulated in FF-PSC cultures after 6 days of differentiation, and FF-PSC-derived cells had low expression of the neuroectodermal marker SOX21 and the forebrain markers FOXG1 and LHX2 ( Figures 3E-3G). Neural crest differentiation from FF-PSCs was also supported by the observation iScience Article that FF-PSC-derived cells expressed the neural-crest-derived peripheral neuronal marker BRN3A (Figure 3H). These data suggest that FF induces neural crest differentiation and that FGF inhibition reverses this step, thus enhancing cerebral organoid generation. Notably, at the undifferentiated stage, such apparent differences were not observed in terms of marker expression for pluripotency and neuroectodermal/neural crest lineage commitment ( Figure S4E).
In addition to neural crest-related genes, FGF-related genes were also enriched in GSEA ( Figure S4B and Table S3). Indeed, FGF-related genes were upregulated in FF + PD17-derived cells ( Figure 3C and Table S2), indicating that FGF signaling was activated 6 days after differentiation of PD17-treated iPSCs. Given that PD17 was treated only at the undifferentiated stage ( Figure 3A), it is natural that FGF signaling was not suppressed after the organoid generation. Rather, this FGF activation after the organoid generation mirrors the proper cerebral specification, because the activation of FGF signaling, especially by FGF8, is observed in embryonic cerebral development (Shimamura and Rubenstein, 1997).

FGF inhibition strategy in the hypothalamus organoid generation
We next examined whether this FGF inhibition at the undifferentiated stage can be generalized to other brain region-specific organoid technologies. We generated hypothalamus organoids using SFEBq methods as previously described (Kasai et al., 2020) from 201B7 iPSCs under feeder-free conditions with the StemFit method ( Figure S5A). Although FF-PSCs failed to establish neuroepithelium-like structures, FF + PD17-PSCs successfully organized pseudostratified epithelium ( Figure S5B). Immunostaining analysis revealed that the hypothalamic marker RAX was upregulated by PD17 treatment ( Figure S5C). Moreover, PD17 treatment downregulated the expression of the neural crest marker SOX10 ( Figure S5C). These results indicate that the FGF inhibition strategy enhanced not only cerebral but also other brain region-specific organoid generation by preventing neural crest differentiation.

WGBS identified epigenomic changes induced by FGF inhibition that reflect organoid competency
We sought to determine how FGF inhibition at the undifferentiated stage promoted neural specification on differentiation from FF-PSCs. Although the undifferentiated PSCs were treated with the FGF inhibitor, the transcriptomic change became apparent only after differentiation. This time lag effect of PD17 suggests two possibilities: 1. PD17 induced the expression change of a few key genes at the undifferentiated stage and these key genes drove the gene regulatory network to suppress the neural crest differentiation; 2. Epigenetic status was changed by PD17 at the undifferentiated stage and these epigenetic changes affected the future gene expression on differentiation. We did not find any significant gene set enrichment by PD17 treatment at the undifferentiated stage, and it is difficult to identify putative key genes. On the other hand, it has been widely accepted that the epigenetic status, such as DNA methylation, biases the differentiation propensity of various stem cells (Kim et al., 2010;Sanosaka et al., 2017). Thus, we performed whole-genome bisulfite sequencing (WGBS) with PSCs (KhES1 ESCs) cultured in FF conditions with or without PD17 treatment, and their global methylation statuses were generally similar ( Figures 4A, 4B and S6A). Differentially methylated region (DMR) analysis identified 20 hypomethylated and 39 hypermethylated regions ( Figure 4C and Table S4). About 70% of DMRs are overlapped with candidate cis-regulatory element (cCRE) annotations in ENCODE project (Abascal et al., 2020), indicating that these DMRs had regulatory activities ( Figure 4D). By using RNA-seq data, we examined the expression level of genes associated with these DMRs ( Figure S6B). We found that only one DMR-associated gene (FIGNL2) was differentially expressed between FF-and FF + PD17-PSCs on Day 0, whereas 8 genes (FIGNL2, CNTNAP3B, GPRIN1, OPRD1, PNCK, TMSB15A, WDR45, and WNT5A) were differentially regulated on Day 6 (p = 0.35 on Day 0; 1.9 x 10 À4 on Day 6, hypergeometric test) ( Figure 4E). These results indicate that FGF inhibition changes the DNA methylation status of the distinct gene set in undifferentiated PSCs but does not directly induce transcriptomic changes. Instead, these results suggest that priming for future expression of some genes after differentiation is regulated by the methylation status at the undifferentiated stage.
Among these DMR-associated genes, we focused on WNT5A because this gene is involved in the neural crest specification (Ossipova and Sokol, 2011). The WNT5A DMR was hypermethylated by PD17 treatment ( Figure 4F). The expression level of WNT5A was low in undifferentiated PSCs and was upregulated after differentiation in FF-PSC cultures but not in FF + PD17-PSC cultures ( Figure 4G). As DNA methylation is known as a mode of transcriptional repression, these results suggest that PD17-induced WNT5A hypermethylation suppressed WNT5A expression during differentiation. Indeed, the DMR in the WNT5A locus was ll OPEN ACCESS iScience 25, 105140, October 21, 2022 7 iScience Article enriched with the enhancer mark H3K4me1 in PSCs ( Figure 4F). In addition, H3K27me3 (repressive mark), but not H3K27ac (active mark), was enriched in this region. These data indicate that the WNT5A DMR has a regulatory activity under repressive control in PSCs. By Sanger sequence-based targeted methylation analysis, we reconfirmed PD17-induced methylation change in this WNT5A region in 201B7 iPSCs and found that OnF-PSCs also had a high methylation rate ( Figure S6C).   To further validate the effect of WNT5A on organoid generation, we added recombinant WNT5A protein to FF + PD17-iPSC-derived differentiating cells ( Figure 5A). WNT5A treatment downregulated FOXG1 expression and upregulated SOX10 expression ( Figure 5B), indicating that WNT5A prevented cerebral organoid generation by enhancing the neural crest specification.
Given that WNT5A generally acts as a non-canonical (b-catenin-independent) Wnt ligand (van Amerongen et al., 2008), we speculated that non-canonical Wnt signaling activation by WNT5A resulted in poor organoid generation from FF-PSCs. Indeed, the downstream genes of non-canonical Wnt signaling (Schambony and Wedlich, 2007;Voloshanenko et al., 2018) were highly expressed in FF-iPSC-derived differentiating cells, and were downregulated by PD17 treatment ( Figure 5C). We therefore applied various Wnt inhibitors during organoid generation from FF-PSCs ( Figures 5D and S6D). Inhibitors of canonical (b-catenin-dependent) Wnt signaling, including IWR1e and XAV939, did not induce FOXG1 and LHX2 expression ( Figure 5E). On the other hand, IWP2, which blocks both canonical and non-canonical Wnt signaling pathways (Chen et al., 2009;Mazzotta et al., 2016), drastically enhanced forebrain specification by preventing neural crest differentiation ( Figure 5E). These IWP2-responsive and IWR1e/XAV939-irresponsive reactions suggest that the non-canonical Wnt pathway was activated during differentiation from FF-PSCs ( Figure S6D) (Chen et al., 2009;Mazzotta et al., 2016), thereby preventing FF-PSCs from acquiring cerebral identity. Together, the data indicate that WNT5A locus is hypomethylated in FF-PSCs, which contributes to priming for future activation of non-canonical Wnt signaling on differentiation. FGF inhibition enhances cerebral organoid generation by inducing DNA methylation at the WNT5A locus, through which non-canonical Wnt signaling is tempered. iScience Article DISCUSSION This study suggests that the culture conditions of PSCs are important for cerebral organoid generation and that feeder-free PSCs poorly differentiate into cerebral organoids. This poor organoid generation resulted from a specific DNA methylation status and can be rescued by FGF inhibition at the undifferentiated stage. FGF inhibition induced DNA methylation at the WNT5A locus, leading to the suppression of the future activation of non-canonical Wnt signaling after differentiation and contributing to the reliable generation of cerebral organoids.
Previous reports suggest marked differences in differentiation propensity among various PSC maintenance cultures Ojala et al., 2012;Shoji et al., 2018). In particular, neural organoid generation from PSCs has largely been optimized for on-feeder PSC cultures, and feeder-free PSC cultures generally do not match the same organoid generation protocol (Kuwahara et al., 2019;Watanabe et al., 2019). Kuwahara et al. (2019) showed that the organoid generation from feeder-free PSCs was rescued by the pretreatment with TGFb inhibitor (SB431542 and LDN193189) and Shh agonist (SAG). Watanabe et al. (2019) identified TGFb ligands as molecules that restored the capacity of feeder-free PSCs to generate organoids. In our culture, however, neither TGFb ligands (BMP4, Activin A, and TGFb3) nor TGFb inhibitors (SB431542 and LDN193189) improved cerebral organoid generation from feeder-free PSCs (Figures S1A-S1E). Although SAG treatment partially upregulated FOXG1 expression, SAG did not recover the expression of another cerebral marker LHX2. There are some possibilities for this discrepancy. One possibility is that each molecule was added alone in our current study, whereas in previous studies they were added in combination. Another possibility is different duration of drug treatment. PSCs were treated with molecules for 1 day in Kuwahara et al. (2019); 2 days in our study; and 3-4 days in Watanabe et al. (2019). In addition, Kuwahara et al. (2019) primarily focused on retinal organoids, and they did not deeply investigate cerebral organoids; for example, they only checked FOXG1 expression and did not evaluate the cerebral organoid morphology or the expression of other forebrain markers, such as LHX2. These considerations indicate that the treatment with TGFb ligands, inhibitors, and Shh agonist has an effect on the organoid generation competency under some optimized conditions, and that these molecules were not so effective in our experimental setting. Watanabe et al. (2019) also suggested that feeder-dependent PSCs highly expressed genes associated with the naive pluripotency, and that the TGFb ligand treatment converted feeder-free PSCs into the naive-like pluripotent state, whereby restoring the competency to produce cerebral organoids. However, we could not confirm this naive-like pluripotent state of feeder-dependent PSCs, nor the primed-to-naive transition in recovering the organoid generation competency of feeder-free PSCs ( Figure S4A). On the other hand, in this study, we found that FGF signaling inhibition led to endowing feeder-free PSCs with the competency to produce cerebral organoids. Although FGF signaling has an essential role in maintaining pluripotency, FGF also functions as an inhibitor of neural induction (Greber et al., 2011). In addition, feeder-free cultures require a higher dose of FGF than on-feeder cultures (Xu et al., 2005). These studies support our FGF inhibition strategy for generating cerebral organoids from feeder-free PSCs.
Our FGF inhibition strategy was confirmed in both widely used methods for feeder-free PSC maintenance, StemFit (Ajinomoto) and mTeSR1 (Stemcell Technologies). Moreover, FGF inhibition enhanced not only cerebral but also hypothalamic organoid generation. These indicate that the FGF inhibition strategy can be generalized to various feeder-free culture systems and various brain region-specific organoid technologies. On the other hand, it should be noted that a previous study reported a highly reproducible derivation of brain organoids from feeder-free PSC cultures with another maintenance medium, Essential 8 (Thermo Fisher Scientific), which also contains a high concentration of FGF (Yoon et al., 2019). This protocol does not require exogenous FGF inhibition and seems to be inconsistent with our FGF inhibition strategy. However, Essential 8 medium has been known to induce endogenous FGF inhibition because it does not contain any lipid components, and this lipid-free condition suppresses FGF signaling . Thus, the Essential 8-based method autonomously inhibits endogenous FGF signaling in PSCs (although FGF inhibitors are not used in this method) and does not conflict with our study.
Our FGF inhibition approach does not require the modification of derivation methods but changes only the initial culture condition of PSCs before differentiation. This indicates that the competency of organoid generation is determined at the undifferentiated stage. Such variations in differentiation propensity are reminiscent of epigenetic memory in iPSCs (Kim et al., 2010;Miura et al., 2009 iScience Article DNA methylation signatures characteristic of their somatic tissue of origin. We found that the DNA methylation status, regulated by FGF signaling, determines the differentiation capability to produce cerebral organoids by analogy to epigenetic memory. Notably, in this study, the DNA methylation status of a specific gene target, WNT5A, was identified as a potential regulator of differentiation propensity. The hypomethylation status of WNT5A led to the activation of non-canonical Wnt signaling on differentiation and prevented PSCs from differentiating into the neural lineage while inducing neural crest lineage specification. WNT5A has pivotal roles in various aspects of embryogenesis, especially in the patterned spatial arrangement (''cell-polarity'') and the lineage specification (''cell-fate'') (Loh et al., 2016). Cell-polarity signaling by WNT5A controls planar cell polarity, convergent extension, and directed cell migration, and is involved in the body axis elongation, the limb development, and the face formation (Yamaguchi et al., 1999). WNT5A also exerts cell-fate activities; for example, WNT5A is required in the neural crest induction (Ossipova and Sokol, 2011) and the mesoderm/cardiomyocyte specification (Mazzotta et al., 2016). Given that the epigenetic priming affected WNT5A expression in a relatively early phase of differentiation, this epigenetic priming is expected to have impacts on the cell-fate activities. Thus, our FGF inhibition strategy would be useful not only for the brain organoid generation but also for the directed differentiation of PSCs into various cell types whose specification is regulated by WNT5A signaling. In addition, such gene priming before activation has been observed in a variety of biological systems (Wang et al., 2015;Zalc et al., 2021), and it would be interesting to investigate whether this epigenetic priming of WNT5A affects the in vivo embryogenesis.
Our FGF inhibition strategy offers a reliable method for generating cerebral organoids from PSCs under various maintenance culture conditions. This method would help in biological and medical applications of brain organoids, such as in vitro brain organogenesis models (Benito-Kwiecinski et al., 2021;Trevino et al., 2020), organoid-based phenotyping for neurological diseases (Bowles et al., 2021;Nakamura et al., 2019;Samarasinghe et al., 2021), and organoid transplantation for regenerative medicine (Dong et al., 2021;Kitahara et al., 2020). In addition, this study contributes to the elucidation of molecular mechanisms for biased differentiation among PSC maintenance methods, which have not been extensively examined.

Limitations of the study
Future experiments are needed to elucidate the molecular mechanisms by which FGF inhibition alters the methylation status of specific DNA targets. In mouse ESC culture, FGF inhibition leads to global DNA demethylation by some mechanisms, including impaired maintenance of methylation, the induction of hydroxylases Tet1 and Tet2, and the repression of de novo methyltransferases (Ficz et al., 2013). Additional mechanisms may be supposed for targeted methylation control. Furthermore, it is not clear whether organoid generation competency is regulated by WNT5A alone or together with other targets. In addition to WNT5A, our WGBS analysis identified additional DNA methylation signatures regulated by FGF signaling. Investigations of these targets would be beneficial for further understanding the mechanisms of the culture-condition-dependent differentiation propensity of PSCs.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:  We assessed the morphological quality of organoids on Day 36-37 based on the existence or the distribution of the neuroepithelial dome structures with the following criteria: Score A, domes all around the aggregate; Score B, domes more than halfway around; Score C, at least one dome; Score D, no dome structures.

Hypothalamus organoid generation
To generate hypothalamus organoids, we used an SFEBq-based method (Kasai et al., 2020) with some modifications. Briefly, on Day 0, PSCs were dissociated into single cells with TrypLE Select, and 10,000 cells per well were reaggregated in low-cell-adhesion 96-well plates with V-bottomed conical wells in growth ll OPEN ACCESS