New insights into the epitranscriptomic control of pluripotent stem cell fate

Each cell in the human body has a distinguishable fate. Pluripotent stem cells are challenged with a myriad of lineage differentiation options. Defects are more likely to be fatal to stem cells than to somatic cells due to the broad impact of the former on early development. Hence, a detailed understanding of the mechanisms that determine the fate of stem cells is needed. The mechanisms by which human pluripotent stem cells, although not fully equipped with complex chromatin structures or epigenetic regulatory mechanisms, accurately control gene expression and are important to the stem cell field. In this review, we examine the events driving pluripotent stem cell fate and the underlying changes in gene expression during early development. In addition, we highlight the role played by the epitranscriptome in the regulation of gene expression that is necessary for each fate-related event.


INTRODUCTION
Stem cells exhibit the potential to differentiate into various cell types 1 . Therefore, stem cells have been useful in biological and medical research 2 . Stem cells have the advantage of inducibility to obtain and utilize the cell types suitable for a particular research purpose. In particular, stem cell types affect different developmental stages and differentiation lineages 3,4 . The usefulness of stem cells is based primarily on their ability to undergo various cell development stages, allowing researchers to obtain specific cell types with phenotypes at the appropriate developmental time based on the stage of acquisition 5 . Using lineage differentiated stem cells has enabled research model establishment for the direct investigation of molecular mechanisms, phenotypes, and therapeutic strategies in phenotype-relevant human cells [6][7][8] . Stem cells in the early stages of development exhibit a wide range of differentiation properties 2 . For example, zygote cells proliferate to form a morula, a group of stem cells with the same multicellular differentiation potential, including embryonic cells that form an organism and supporting cells needed to maintain proper embryonic development 9 . This broad differentiation potential and totipotency of stem cells gradually decrease as stem cells proliferate and acquire the necessary phenotypes during development, eventually progressing to a state called pluripotency 10 . Embryonic stem cells (ESCs) extracted from the inner cell mass (ICM) of a blastocyst are stem cells in the pluripotent state 11 . Although these cells have lost the ability to differentiate into cells that form extraembryonic tissues, such as the placenta and yolk sac, they retain the ability to differentiate into cells that can form fetal tissues and organs 12 . Thus, human ESCs (hESCs) in a pluripotent state are actively employed in developmental and pathological research that integrates aspects of human body pathophysiology 13,14 . However, both the totipotent and pluripotent states of cells are dynamic, and the range of differentiation potential changes gradually with the progressive maturation of stem cells 15 . Furthermore, whereas totipotent stem cells are characterized by their homogenous proliferation, human pluripotent stem cells (hPSCs) gradually show cellular diversity in the ICM of a blastocyst 16,17 . The pluripotency of stem cells is thus limited, leading to the generation of three heterogeneous germ populations through pattern and structural formation events and gastrulation 18 . Thus, understanding the biological events and molecular mechanisms that control the properties of hPSCs is critical. However, only a subset of the molecular pathways governing stem cell pluripotency is understood.
The state of stem cells is determined by the influence of spatiotemporal location and ambient environmental conditions that change on the basis of developmental processes 19 . In a proliferating mass of stem cells, the topological information of each cell leads to diverse cell polarity and plasticity 20,21 . Both the matrix surrounding stem cells and various environmentally induced signaling pathways trigger alterations in stem cell activity and behavior 22 . Through the control of gene expression, these factors eventually lead to biological changes, functioning as intrinsic fate regulators operated by the stem cells themselves 23,24 . Gene expression patterns regulate the biological properties of cells and define cell types 25 . Similar to those of somatic cells, the functions of stem cells are controlled by the molecular actions of proteins encoded by cell-type-specific genes 26 . Gene expression control, therefore, is a key factor in determining the state and function of stem cells, regardless of intrinsic or extrinsic triggers 24,27 . To ensure the timely and accurate regulation of required gene expression, stem cells engage well-established transcription machineries and epigenetic regulatory systems 28 . For instance, recent studies have suggested that an epitranscriptomic gene regulatory system changes the protein translation efficiency and persistence of messenger RNA (mRNA) through chemical modifications of transcribed mRNAs 29,30 . Compared with somatic cells, hPSCs have immature genetic and epigenetic regulatory mechanisms 31,32 , which can affect precise gene expression control, resulting in alterations in protein expression patterns that can be fatal 33 . Nonetheless, PSCs in the early stages of development preserve an accurate fate decision process based on tight gene expression control 34 . In addition to conventional gene regulation, other molecular mechanisms may play a role in the control of gene expression in PSCs.
In this review, we summarize changes in the state of PSCs during the early stages of development and the regulatory mechanisms underlying these changes. Additionally, we discuss recent research on gene regulation at the transcriptome level in relation to the regulation of PSC fate. Although a number of studies have reported distinct properties of hPSCs compared with mouse PSCs due to interspecies differences, the use of mouse embryos can provide sufficient knowledge for understanding human embryonic development. Here, we present studies on both mice and human cells and point out whether the results have been derived from human or animal models.

STATE OF PLURIPOTENT STEM CELLS (PSCS) DURING EARLY DEVELOPMENT
Stem-cell development generates diverse groups of cells with different properties via the generation of identical multiple-twin cells from a starting single cell (Fig. 1a) [35][36][37] . Immediately after fertilization, zygote cells continue to divide, forming a group of 16homogenous cells that comprise the early morula 38 . As this cell group continues to divide, differences in environmental conditions, including topographical forces applied to each cell in the mass, are created. These differences result in the generation of the first heterogeneous population: the ICM and trophectoderm (TE, Fig. 1b) 35,39,40 . Each cell group constituting the preimplantation blastocyst exhibits distinct differentiation potential for the production of nonoverlapping progeny. Trophoblasts organize surrounding structures such as the chorion, which supports embryogenesis, whereas the ICM is critical for the formation of the embryo 41 . After implantation of a blastocyst in the maternal endometrial epithelium, the ICM undergoes subsequent morphogenetic changes. The ICM of a postimplantation blastocyst contains epiblasts and hypoblasts (Fig. 1a) 42 . The morphogenetic events include the polarization of the epiblast, which forms the central lumen that develops into the amniotic cavity; creation of the amniotic epithelium, which forms the amniotic sac membrane; and differentiation of primordial germ cells, which are precursors of eggs or sperm [43][44][45] . Moreover, extraembryonic mesenchyme cells derived from the hypoblast surround the generated structure to isolate it from the outer cell membrane (OCM) formed by the trophoblast 37,43,46 . Thereafter, epiblasts in the ICM form a primitive streak, gastrulate and differentiate into three germ layers: The ectoderm, mesoderm, and endoderm 47 .
HESC cultures have been established by the in vitro cultivation of cells isolated from a morula, an entire blastocyst, or an ICM 11,48,49 . Morula-derived hESCs are thought to remain in a naive state, showing unbiased differentiation potential, including the generation of trophoblasts 50 , whereas blastocyst-or ICM-derived cells are considered to exhibit a primed-like state, retaining comparably late-stage stem cell characteristics (Fig. 1c) 15 . Despite differences in cell properties, both types of hESCs exhibit the developmental spectrum that enables the generation of all somatic and germline cells 51,52 . In contrast, murine naive and primed ESCs have shown to exhibit clear differences in the molecular and cellular properties 53,54 . Previously, murine ESCs were believed to show better differentiation plasticity than human ESCs, but the limited pluripotent state epiblast stem cells (EpiSCs) was revealed. In contrast to mouse ESCs, known to be in the naive state, mouse EpiSCs in the primed state clearly show not only differences in response to activation by external signals and a decrease in self-renewal capacity but also a massive reduction in whole-animal generation efficiency because of a tetraploid composition 55,56 . Although the definition of the naive and primed state of hPSCs remains ambiguous, evidence of s naive state of mammalian stem cells, including that of primates, has been reported, suggesting the need to understand the conversion of hPSCs to the pluripotent state 57,58 .

MECHANISMS THAT CONTROL PSC FATE
The state of PSCs, including their development and differentiation, is controlled by both intrinsic and extrinsic conditions. Deciphering the complicated signaling pathways governing the pluripotent state of stem cells is challenging, as multiple biological events, such as maturation, fate determination, and differentiation, concurrently affect stem cell development. In this context, the biological events regulating the pluripotent state of stem cells are categorized into two processes: "acquisition of heterogeneity from a homogeneous population" and "conversion from a naive to a primed state". Each process is accompanied by an explanation of the mandating molecular pathways.
Acquisition of heterogeneity in a homogeneous population Among the major questions in stem cell biology, when and how is the control of clonal cells within a homogeneous population changed to produce cells with distinct and orderly fates? To form a highly organized body from an embryo comprised of symmetrical cells, the embryonic symmetry must be disrupted 59 . The importance of acquiring asymmetric fate during gastrulation to enable balanced development of the human body has been investigated in previous studies 60,61 . Complex signaling at the cellular level, as well as geometrical and topological transformation at the embryonic level, are necessary for asymmetry 62,63 . Specifically, the process that ultimately disrupts embryonic radial symmetry and triggers anteroposterior axis specification, has been found to be driven by a morphogen signaling gradient 60 . Notably, the first disruption to the symmetric organization of human cells occurs much earlier than previously thought: during the division of stem cells in the totipotent state. The emergence of the TE in blastocyst-stage embryos is the first morphological variation during embryogenesis. TE and primitive endoderm lineages generate the placenta and yolk sac extraembryonic cells, whereas cells inside the blastocyst form the ICM, the cells of which eventually divide to form the primitive endoderm and pluripotent epiblast cells. According to the most recent understanding, each cell in the early embryonic state retains identical developmental potential. Therefore, in the absence of stimulants such as morphogens, cell fate has been assumed to be determined randomly, making the existence of initiating factors driving cellular heterogeneity enigmatic. Based on single-cell transcriptome analyses, recent studies with PSCs and mouse embryos have suggested another possibility [64][65][66][67] . Differences were detected in the totipotency between cells constituting murine or human blastomeres, with each cell in the 16-cell stage exhibiting a transcriptome that differs from that in the 2-cell stage immediately after the first cleavage 66 . The gradual increase in the differential tendency of genes to contribute to lineage specification suggests that the symmetry is disrupted earlier than traditionally thought; that is, it distinguishes trophoblasts from embryoblasts after the morula stage. This finding was consistent with the results observed in several studies suggesting that only one of the cells separated in the 2-cell stage can develop into a mouse 64,[68][69][70][71] .
The molecular mechanism causing this initial heterogeneity has not yet been clearly elucidated. A recent study reported that individual cells in the 4-cell-stage blastomere exhibit differential expression of epigenetic modifiers, such as that of the histone methyltransferase CARM1 (Fig. 1b) 72,73 . Differences in the expression of CARM1 regulate the DNA-binding ability of the pluripotency transcription factors SOX2 and OCT4, with activity that is affected by histone H3 arginine 26 dimethylation (H3R26me2) 74,75 . A differential increase in the expression of genes downstream of SOX2, such as Sox21, at the 4-cell stage has been reported to suppress the expression level of Cdx2 76 , a lineage specifier for TE differentiation in the 8-cell stage 77,78 . Subsequently, the reduced expression of Cdx2 results in the decreased expression of polarity markers that define the apical domain, in turn leading to TE differentiation. Interestingly, Cdx2 mRNA has been demonstrated to be asymmetrically localized to the apical pole in the 8-cell stage, when the embryo is undergoing apical-basal polarization 79,80 . Due to the apical localization of Cdx2 transcripts, differential inheritance during subsequent division to the 16-cell stage might result in cells on the outside of the apical domain possessing more Cdx2 transcripts than inner cells. Therefore, the determination of timing of these differences in gene expression between cells is important. The mechanisms underlying the determination of early mammalian stem cell fate remain unclear, largely because whether the first bifurcation in the cell fate sequence is a random event in mammalian embryo during the morula stage or is based on molecular features of differentiation that emerge prior to morphological changes remains unlcear 75,[81][82][83][84] . Depending on the true pathway, either the role played by an intrinsic gene regulatory mechanism or the cellular environment condition will be predominant. An analysis of the transcriptome in multiple blastomeres revealed that after the first cleavage each cell exhibited different gene expression patterns from the pattern evident in the 2-cell stage 66 . Although evidence supports an autonomous molecular cascade for the regulation of cell fate that directs differences between cells in blastomeres early in embryonic development, the mechanisms that control precise gene expression in the open chromatin structure of early stem cells remain unclear.
Cell conversion from a naive to a primed state The differentiation capacity of stem cells refers to the ability to generate a range of differentiated cells, which is the basis of classifying cells as totipotent, pluripotent, and multipotent 10 . Although pluripotency is a transient property of stem cells in vivo, regulation of signaling pathway activity involved in fate maintenance enables stem cells to remain in a pluripotent state for an extended period in vitro 85 . The pluripotent stage involves multiple stem cells in dynamically changing states, and the differentiation potential of each cell is gradually limited; therefore, the stem cell potential is determined on the basis of the developmental stage of the blastocyst from which the ESCs are derived. Due to the limited accessibility to human embryos 86 , studies using murine cells first elucidated the correlation between the origin, status, and differentiation capacity of PSCs. Two types of PSCs have been established from mouse blastocysts (Fig. 1c): ESCs 87,88 and EpiSCs 53,54 . ESCs derived from the ICM of preimplanted mouse blastocysts were in a naive state and were capable of generating all three germ layers and primordial germ cells [87][88][89] ; in contrast, EpiSCs isolated from postimplanted epiblasts were in the primed state and capable only of generating the three germ layers 53 . Although both ESCs and EpiSCs show the ability to form the three germ layers, these two types of stem cells are clearly distinguished by their developmental potential, as only ESCs, not EpiSCs, can efficiently generate chimeric progeny from injected blastocysts 53,54,90 . Differences in the ability of PSCs to generate chimeric animals are representative of stem cell potential in the in vivo counterparts of pre-or postimplantation blastocysts 91 . Various efforts have been made to elucidate the mechanism underlying the differences in pluripotency attributed to differences in PSC origin. In contrast to ESCs, which express a variety of pluripotency genes, including Oct4, Nanog, Sox2, ESRRβ, Rex1, Klf2, and Klf4, EpiSCs exhibit elevated expression levels of differentiationinducing genes, such as Otx2 and Zic2, and low levels of Nanog, Rex1, Klf2, or Klf4 54, 89,92 . Studies designed to optimize in vitro culture conditions to maintain naive or primed PSCs have suggested that the endogenous activity of genes is regulated by signals transduced from outside a cell, such as fibroblast growth factor (FGF) 93 , bone morphogenic protein (BMP) 94 , transforming growth factor β (TGFβ) 95 , and WNT 96 . Therefore, early methods for culturing naive murine ESCs from mitotically inactive mouse embryonic fibroblasts and fetal bovine serum 97 led to modulation of JAK-STAT3 pathway and NODAL signaling via the addition of leukemia inhibitory factor (LIF) 98 and BMP 94 . Similarly, treatment with MEK/FGF/GSK3 inhibitors led to murine ESC maintenance in the naive state, even without LIF stimulation 96 . When FGF inhibition was released, cell proliferation was promoted while the naive state of the stem cells was maintained by a MEK-ERKindependent mechanism of endogenous FGF 93,99 . Alternatively, FGF was reported to induce the emergence of ESCs from the naive state. In addition, FGF together with activin A, abrogated the differentiation of murine ESCs into primordial germ cells 89 , whereas it promoted location-specific differentiation in an embryo; that is, it induced the acquisition of the anterior lategastrula primitive streak cell phenotype by EpiSCs 100,101 . WNT signaling has also been found to function ambivalently with respect to the naive state of PSCs. For instance, the stabilization of nuclear β-catenin has been shown to enhance naive pluripotency by neutralizing the repressive activity of transcription factor 3 (TCF3) on nuclear target genes such as Oct4, Nanog, and Sox2 102-106 . The transcriptional activity of β-catenin on differentiation-associated target genes was diminished, thereby comporting with the increased membrane stability of cytoplasmic β-catenin rendered through complex formation with OCT4 and E-cadherin in cells in the naive state 107 . In contrast, by promoting mesodermal gene expression under differentiation-optimized conditions, nuclear β-catenin was shown to facilitate differentiation-related priming over naive pluripotencypromoting functions 108 .
Although hESCs are derived from blastocysts similar to those of mice, they differ from murine ESCs. For example, in contrast to murine ESCs, which require LIF for survival, hESCs derived from the ICM of blastocysts require FGF2 and TGF-β1. In addition, hESCs share hallmarks of murine EpiSCs, including partial expression of naive pluripotency markers, deposition of H3K27me3 at genetic loci required for development, lack of global hypomethylation, and an inactive X-chromosome in female ESC lines 51,109,110 . However, recent evidence has indicated that hESCs in the primed state resemble murine ESCs more closely than murine EpiSCs. Furthermore, hESCs maintain the expression of E-cadherin, which is highly expressed in naive murine ESCs, but not FGF5 or N-cadherin expression, although they are highly expressed in murine EpiSCs 51 . The level of expression of naive pluripotency factors such as NANOG, PRDM14, and REX1 is relatively stable in hESCs, and ablation of NANOG and PRDM14 induced the differentiation of hESCs 111 . In addition, even when cultured with FGF2/activin A, the DNA methylation pattern of hESCs has been shown to be comparable to that of naive ESCs cultured in LIF/FBS not murine EpiSCs 112,113 .
To date, the distinction between naive and primed states remains unclear; however, clearly, the stem cell-specific regulation of intrinsic FGF and β-catenin signaling is precisely regulated at the early pluripotent naive stage.

CONTROL OF GENE EXPRESSION IN THE PLURIPOTENT STATE
Cellular identity is driven by epigenomic, transcriptomic, and proteomic heterogeneity. Studies creating pluripotent cells by introducing the exogenous genes OCT4, SOX2, KLF4, and MYC have provided evidence for the key role played by the expression of specific genes in the regulation of stem cell fate 27,114 . More specifically, the aforementioned differences in the expression of symmetry disrupting and pluripotency genes are evident before the important division of PSCs into the three germ layers 66,87,115,116 . The regulation of autonomous molecular factors participating in the control mechanism of stem cell fate determines the homogeneous and naive states, with the function of these factors leading to cell-specific differences that lead to achieve heterogeneity and a primed cell state [117][118][119] .
In general, the regulation of gene expression required by cells is described by the central dogma 120 . Specifically, transcription, translation, posttranslational modifications or subsequent proteolytic processes precisely control the levels of molecular expression. In contrast to other processes that require the systematic organization and regulation of subsequent processes to control the level of a final molecular product, transcriptional regulation is an efficient process for regulating PSC fate because the molecular control mechanisms are immature [121][122][123] . Although the epigenome shapes stem cell hierarchies 124 , differential transcriptomes only partially explain protein abundance 121,122 . As hPSCs are characterized by completely open chromatin conditions and immature epigenetic systems, posttranscriptional control is thought to play a key role in the functional output of genetic programs (Fig. 1d) 125,126 . A zygote survives and functions through the support of molecular components passed down through gametes, which carried these traits before fertilization 127 . In particular, PSCs in the early stages of development undergo zygotic gene activation (ZGA), which allows precise regulation of selective expression of genes necessary for survival 128 . ZGA of hPSCs begins in the 2-cell stage and proceeds to the morula stage 129 , which is still characterized by immature chromatin reorganization, even in the ICM of a preimplantedblastocyte 130,131 . Although epigenetic control is thought to occur during this stage, epigenetic modifications are deposited during the heterogeneity acquisition process of the cell transition from a primed state to three germ layer development 31 . This timeline has been supported by studies showing de novo heterochromatin formation and X-chromosome inactivation in hPSCs after ICM formation [132][133][134] . In addition, as described above, WNT, FGF, or activin signaling regulates the transcription of molecular factors that affect global methylation through the activity of OCT4 and NANOG in the naive state 135,136 . This regulatory program requires additional mechanisms to compensate for coarse transcriptome level changes caused by immature transcriptional control prior to the completion of epigenome-driven transcriptional regulatory mechanisms 123 . Increasing data have suggested a role for posttranscriptional modifications in the early stages of stem cell development that enables the selective regulation of required functional outputs (i.e., proteome content) based on the global transcriptome.

The epitranscriptome in development and stem cells
In addition to reported posttranscriptional modifications during oocyte development 137 , more than 100 chemical modifications in addition to typical posttranscriptional modifications, such as 5′capping, polyadenylation, and splicing, have been identified 138 . Methylation is the most common enzyme-catalyzed modification, with N6-methyladenosine (m 6 A) 29 , N1-methyladenosine (m 1 A) 139 , pseudouridine (Ψ) 140 , and C5-methylcytosine (m 5 C) 141 of RNA being reported. However, the role played by RNA m 6 A in posttranscriptional regulation during stem cell development is unknown.
N6-methyladenosine (m 6 A) m 6 A is the most prevalent modification of the 3′ untranslated regions (UTRs), long internal exons, intergenic regions, and 5′ UTRs in human mRNAs 30,142 . The functional m 6 A network is accurately orchestrated through a series of protein effector molecules, including "writers" for methylation, "erasers" for demethylation, and "readers" for interpretation of methylated mRNAs. A microRNA-guided protein complex consisting of Wilms tumor 1-associating protein (WTAP), methyltransferase-like protein 3 (METTL3), and methyltransferase-like protein 14 (METTL14) targets specific regions on mRNA, depositing a methyl group at the N6position of adenosine [143][144][145] . Subsequent demethylation by eraser proteins, such as fat mass and obesity-associated protein (FTO) or human AlkB homolog H5 (ALKDH5), results in m 6 A being highly enriched in specific transcriptome sites 146,147 . The presence of eraser proteins that correct a methylation pattern suggests that m 6 A-associated posttranscriptional control might be a dynamic process that can quickly reflect environmental changes such as developmental progression and cellular stress. An m 6 A-loaded mRNA is recognized by RNA-binding protein YTH domaincontaining families 1 (YTHDF1), 2 (YTHDF2), or 3 (YTHDF3) after being delivered to the cytoplasm through 5′ capping and polyadenylation 30,148,149 . Emerging evidence has suggested that YTHDF1, YTHDF2, and YTHDF3 target distinct mRNAs for capdependent translation, acceleration of mRNA decay, or promotion of translation and decay, respectively. In contrast to the prevailing model, recent studies have reported that all 3 YTHDFs act redundantly to mediate biological functions through specific m 6 A sites and shared target mRNAs 150 .
Recent findings have also demonstrated the role played by the m 6 A modification in various molecular processes, such as translation efficiency, stability, localization, and splicing, which are involved in stem cell development and fate control ( Fig. 2) [151][152][153] . Deficient METTL3 and METTL14 have been reported to inhibit the expression of pluripotent genes such as SOX2, NANOG, and DPPA3 but to promote the expression of developmental regulators such as FGF5, CDX2, and SOX17 151 . Furthermore, deletion of Mettl3 and Mettl14 increased mRNA stability in a HuRand miRNA-dependent manner 152 , a finding was further supported by a subsequent study showing that ZC3H13 forms biochemical complexes with WTAP, VIRMA, and CBLL1 154 . This complex formation led to a decrease in global levels of the m 6 A modification and self-renewal ability, triggering the differentiation of murine ESCs. These results differed from those of a study reporting that Mettl3-knockout mice presented with embryonic lethality between embryonic days 3.5 and 6.5 when PSCs lost lineage differentiation potential 153 . In Mettl3-knockout murine ESCs, stabilized FGF5 mRNA activated pErk and downregulated Nanog expression, FGF5-mediated concurrently coactivated pAkt to reestablish the expression of Nanog. The induction of differentiation and a concurrent delay in pluripotency loss might explain the discrepancy in these studies 155 . Ultimately, m 6 A loss causes confusion in determining whether cells maintain and advance in their pluripotent state and or undergo differentiation. Another study investigating the interactome of SMAD2/3 revealed that SMAD2/3 promoted m 6 A deposition on nuclear RNA by interacting with the METTL3-METTL14-WTAP complex in response to the activation of activin A/TGFβ signaling in primed hESCs 156 . Accumulation of the mRNA of NANOG, NODAL, and LEFTY1, which are activin A signaling targets, likely affects processes such as the transition from homogeneity to heterogeneity or from the naive to the primed state by dynamically reflecting extracellular signaling in hPSCs.

Pseudouridine (ψ)
Pseudouridine is the most abundant modification on transfer RNAs (tRNAs) and ribosomal RNAs (rRNAs); it has also been also found on mRNAs, long noncoding RNAs (lncRNAs), and small nuclear RNAs (snRNAs) 157,158 . RNA is modified by pseudouridine synthases (PUSs) in a guide-RNA-independent manner 159 , affecting RNA structures by promoting base stacking interactions via a hydrogen bond donor to increase RNA backbone rigidity, thereby regulating the interaction of other biomolecules with RNAs 160 . Ultimately, pseudouridylation plays a role in increasing tRNAs stability and simultaneously alters translation termination, thereby regulating protein synthesis 161,162 . Interestingly, guide-RNAdependent pseudouridylation is regulated by dyskerin (DKC1) 163 , which promotes the elongation of telomeres and increases the expression of OCT4 and SOX2 164 . However, mutations in DKC1 result in failed iPSC reprogramming 165 . In line with this outcome, DKC1 mutations have been found to be involved in the pathogenesis of human diseases characterized by short telomeres 163 . Moreover, deletion of Dkc1 led to early embryonic lethality in a mouse model 166 . Although its association with the state of PSCs has not yet been clarified, pseudouridine might be related to stem cell development, maturation, and aging (Fig. 2).

N1-methyladenosine (m 1 A)
In human cells, m 1 A is widely distributed on tRNAs 167 . The m 1 A modification is mediated by TRMT10C, TRMT61B, and TRMT6/61A methyltransferases and is associated with the maintenance of the structure, stability, and function of tRNAs in mitochondria or cytoplasm 162,168,169 . Less than 0.1% of mRNAs contain an m 1 A modification, which is generally located near the 5′ UTR close to a translation initiation site 139 . Although the role played by TRMT6/ 61A in depositing m 1 A modifications in cytoplasmic mRNAs carrying the GUUCRA tRNA-like motif is known, its contribution is minor 167 . Interestingly, similar to the m 6 A modification, the m 1 A modification involves a dynamic and reversible process. m 1 A demethylation is mediated by ALKBH1 and ALKBH3, and m 6 A and m 1 A share a reader protein 170,171 . YTHDF2 and YTHDF3 are known to contribute to the interpretation of m 1 A, causing the rapid decay of modified mRNAs 172,173 . Recent studies have suggested that YTHDF3 promotes the degradation of IGF1R mRNA, leading to the inhibition of trophoblast invasion in trophoblast-associated pregnancy disorders 173 . Although neither the biological role nor the molecular mechanism of m 1 A has been established to date, it is thought to be involved in the pathophysiology of stem cells because of the similarity in the dynamics and shared molecular machineries of the m 1 A and m 6 A modifications (Fig. 2).
C5-methylcytosine (m 5 C) Similar to m 6 A, the m 5 C modification is reversible and controlled by methyltransferases (writers: NSUNs, TRDMT1, and DNMT2) 174 and demethylases (erasers: TETs and ALKBH1) [175][176][177] . The molecular function of m 5 C is mediated by reader proteins (ALYREF and YBX1) 178 . Although the m 5 C modification is highly enriched on tRNAs and rRNAs, it reportedly plays a role in mRNA export 178 , RNA stability maintenance 178 , and translation 179 . In zebrafish, deletion of maternal Ybx1 enhanced translation and triggered the unfolded protein response, resulting in oogenesis and embryogenesis defects 180 . In addition, Ybx1 has been shown to play a role in maintaining maternal m 5 C-containing mRNAs during the maternal-to-zygotic transition (MZT) during initial embryogenesis 181 . These mRNAs have been reported to facilitate the MZT in early zebrafish development. A pathological association between m 5 C and the activation of JAK-STAT signaling and dysregulation of tRNAs has been reported in various cancers and neurodegenerative disorders in humans 178,182 . Although the role played by the m 5 C modification in the regulation of early human stem cell properties has not been elucidated, further research into the function of m 5 C in health and disease is warranted because of its effect on early zebrafish embryogenesis and importance of TET proteins in early stem cell development (Fig. 2).

CONCLUSIONS AND PERSPECTIVES
Although somatic cells control chromatin accessibility to transcription machineries through their well-organized heterochromatin structure and epigenetic control, hPSCs accommodate the uncontrolled accessibility of transcriptional machineries to euchromatin throughout the nucleoplasm. However, PSCs strictly control the levels of expression of certain proteins, which are the functional end products of strictly controlled gene expression. More specifically, the molecules that determine the differentiation potential and developmental stage of PSCs are controlled at the epitranscriptomic level. Epitranscriptomic regulation, which controls the stability and translational efficiency of transcribed mRNAs, overcomes unfavorable conditions faced by PSCs. In particular, m 6 A is the most prevalent epitranscriptomic modification of mRNAs and is involved in the regulation of major signaling pathways in hPSCs. However, a number of studies have shown that mRNA modifications in addition to m 6 A were correlated with early developmental changes in animal models or modulated by molecular mechanisms during early developmental changes in PSCs. Therefore, future studies should be directed to investigating the complex epitranscriptomic gene regulatory system, which involves the mutual interaction of multiple types of RNA modifications, as well as the molecular mechanism and biological implications of certain types of RNA modifications.