Single-cell technologies: a new lens into epigenetic regulation in development

The totipotent zygote gives rise to diverse cell types through a series of well-orchestrated regulatory mechanisms. Epigenetic modifiers play an essential, though still poorly understood, role in the transition from pluripotency towards organogenesis. However, recent advances in single-cell technologies have enabled an unprecedented, high-resolution dissection of this crucial developmental window, highlighting more cell-type-specific functions of these ubiquitous regulators. In this review


Introduction
Mammalian embryogenesis is a highly orchestrated process characterized by precisely timed specification events that guide the formation of cellular hierarchies in the developing embryo.These events are in part coordinated by the modulated action of signaling molecules, which can induce genetic and epigenetic responses instrumental for cell state priming and acquisition [1][2][3][4].This highly synergistic interplay of several regulatory layers, including transcription factors, chromatin remodelers, histone modifiers, DNA methyltransferases (DNMTs), and noncoding RNAs (ncRNAs), shapes the dynamic epigenome of the developing embryo.It is well established that epigenetic regulators are central to these processes and many studies over the past decades have demonstrated their essential role in ensuring proper developmental progression [5,6].For instance, loss of their function during embryogenesis has been associated with cellular miss-specification events, developmental delay or arrest through mouse transgenesis, light microscopy, and next-generation sequencing approaches (Figure 1) [7][8][9][10].
Despite the consensus that these ubiquitously expressed regulators are essential, assigning specific roles and understanding the exact cause of lethality in mutant embryos remains largely unanswered.Recent advances in single-cell technologies have now provided an alternative route to classic perturbation and morphological assessment studies and enabled a deeper exploration of early embryogenesis, including improved interpretation of phenotypes associated with loss of function [11].In this review, we summarize recent insights and technical developments with a focus on epigenetic regulators, their expression profiles and cell-type-specific functions, as well as the associated phenotypic outcomes in embryonic knockouts (KOs).Finally, we close by describing the promise of advanced in vitro models of embryogenesis to further explore transcriptional and epigenetic regulator functions with higher throughput.

Epigenetic regulators are essential for early development
Advances in genetic engineering and sequencing technologies from the 1990s onwards have allowed the generation and investigation of phenotypes that emerge in embryos lacking core epigenetic regulators such as Setdb1, Ezh2, Suz12, G9a/Glp (histone modifiers), Brg1, Caf1 (chromatin remodelers), and DNA methyltransferases (DNMTs) [6,8].While their precise function remained obscure, these studies clearly highlighted the essential role the enzymes and their epigenetic modifications play in early embryonic development (Figure 1) [6][7][8][9][10]12].Nonetheless, linking the observed phenotypes, including embryonic lethality, to specific molecular and cellular functions remains challenging for these global and widely expressed actors.Next-generation sequencing-based technologies such as chromatin immunoprecipitation followed by sequencing, wholegenome bisulfite sequencing, and transcriptome sequencing permit researchers to characterize the molecular effects triggered by epigenetic regulators' disruption, but often only validate the loss of modification status without a clear connection to disrupted cellular or developmental functions.For example, zygotic DNMT1 or DNMT3A and B KO result in a global loss of DNA methylation (DNAme), but it is not yet clear which part of this deregulation is responsible for the observed lethality [10,13].Two recent studies have investigated the polycomb repressive complexes (PRC)mediated epigenomic reprogramming during pre-implantation development and showed that, in this context, PRC1 acts upstream of PRC2, guiding the zygotic deposition of H3K27me3 on maternally marked H2AK119ub1 promoters [14,15].As a result, loss of maternal H2AK119ub1 during zygotic gene activation (ZGA) causes premature developmental gene expression (otherwise targeted by PRC2) and developmental arrest at the blastocyst stage [14].
Using a combination of zygotic disruption and molecular characterization, we recently highlighted that DNAme maintenance via DNMT1 and PRC (both PRC1 and PRC2) play a central role in autosomal imprinting in both the embryo proper and the extraembryonic ectoderm (ExE).Notably, the latter relies on an expanded set of imprinting mechanisms, including H3K9 methylation by G9a/GLP for noncanonical imprinting of endogenous retrovirus containing promoters [16].Below we will further expand on the additional power provided by such an approach when combined with single-cell technologies [11].
Taken together, a large collection of studies highlighted the essential and early developmental function of epigenetic regulators.Molecular characterization moved this further over the past decade and now we are well equipped with new technologies to revisit central questions about their precise function and target regulation.

Redundancy in the regulation of DNA methylation
The observation that embryos lacking even single epigenetic regulators, independent of their specific functions, die soon after implantation indicates their crucial and indispensable roles (Figure 1).As may be expected for pathways such as DNAme where multiple and partially redundant enzymes are available, the phenotype can vary.For instance, single mutant embryos for either of the DNMT3s or their antagonistic enzymes, the teneleven translocation (TET) methylcytosine dioxygenases, progress much further in development than mutants of the nonredundant DNMT1 maintenance enzyme.The observation that embryos lacking both DNMT3s or all three TET enzymes also die around mid-gestation confirms their crucial but partially overlapping functions [10,17].Development in the absence of TET1/2 function, despite generating viable animals, results in genome-wide aberrant cytosine methylation and 5-hydroxymethylation (5hmC) distribution; in particular, germ cells are completely depleted of 5hmC, and embryonic as well as post-natal tissues display a general increase in genome-wide methylation, including at some classes of repetitive elements [18].Despite this redundancy, some of the regulators appear to also have distinct functions within specific lineages, such as the unique role of DNMT3A in the germline and DNMT3B in shaping the ExE methylome [19,20].Embryos lacking DNMT3B, but not DNMT3A, lose genome-wide and CpG island-specific methylation in the ExE, without drastic changes in the epigenome of the embryo proper [20].
Of the repressive pathways, DNAme enzymes display a redundancy that can at least buffer some early phenotypes, while both PRCs are required and therefore embryos are susceptible to loss of any individual core subunit (Figure 1).For instance, disruption of either Ezh2 or Rnf2 leads to mid-gestation embryonic lethality, indicating a necessary and nonredundant function of these two regulators during mouse development [9,[21][22][23][24].Genomic targets of EZH1-containing PRC2 are generally also bound by EZH2-containing PRC2 in F9 cells, and loss of Ezh1 does not induce changes in chromatin compaction or gene expression [25].Both in vitro and in vivo, EZH2 and RNF2 are crucial for PRC2 and PRC1 functions respectively, and can compensate for the loss of their paralogues (EZH1 and RNF1), but not vice versa.

Single-cell atlases provide a new lens on mammalian development
The recent curation of single-cell transcriptomic atlases of mouse embryogenesis has created an opportunity to investigate transcriptional and epigenetic regulator expression dynamics at high resolution across different lineages during pre- [26], peri- [27][28][29], post-implantation development [11,28,[30][31][32][33][34][35], and organogenesis [36] (Figure 2).As expected, DNMTs show some stage, tissue and cell cycle-dependent expression dynamics.In line with the genome-wide resetting of the methylome, DNMTs are generally lowly expressed during preimplantation development and Dnmt3b is activated during the late blastocyst stage concomitant with embryo implantation.Although both Dnmt3a and Dnmt3b are expressed at high levels in the embryo, Dnmt3a is only lowly expressed in the E6.5 ExE and, as compared to the epiblast, cannot compensate for the loss of Dnmt3b in this lineage, as described above (Figure 2, E6.5 stage).Other regulators such as Hdac1, Mbd3 and Nasp, in contrast, show a more stable expression profile across tissues and developmental stages.Beyond some of the unique patterns, the data highlight the surprisingly ubiquitous expression of most core regulators during gastrulation where cell state diversity increases rapidly (Figure 2, E6.5-8.5 window).The more precise expression dynamics that can be acquired from the analysis of single-cell data are a major advance that represents a driving focus of the community.However, single-cell transcriptomic states alone fail to inform on precise functions of the individual regulators and cannot yet assign specific roles in different lineages or across developmental stages.In the next sections, we describe how genetic perturbations and single-cell readouts provide a path forward.

Interfacing noncoding RNAs and epigenetic regulation in early development
As shown in Figure 1, many ncRNA genes also exhibit early embryonic lethal KO phenotypes comparable to those of core epigenetic regulators.Two studies in this area addressed the regulatory interplay between ncRNAs and epigenetic regulators by combining singleblastomere injection with single-cell gene expression profiling (sc-qPCR), as well as through high-resolution microscopy.Among other interesting findings, these experiments demonstrated the concerted action of LincGET, a long ncRNA, and the histone methyltransferase (HMT) CARM1 in the differential induction of ICM and trophectoderm (TE) fates as early as the 2cell stage embryo (Figure 3a) [37,38].In LincGET expressing blastomeres, CARM1 catalyzes the deposition of histone H3 arginine 26 methylation (H3R26me2) on pluripotency gene promoters such as Nanog, Sox2, and Oct4.This results in their expression, which increases the accessibility of inner cell mass regulatory elements (ICM-RE), and supports the activation of an ICM-specific developmental program (Figure 3a) [38,39].
Many more such interactions occur throughout development, illustrating the relevance of the RNA-chromatin interface in cell state acquisition.

A multipronged assessment using single-cell multiomic technologies
In general, the opportunity to simultaneously profile multiple regulatory layers can provide additional insights beyond what can be recovered from transcriptomes alone, driving many groups to develop and apply scMultiomics technologies [40][41][42][43].Among other uses, these approaches have enabled the direct association between cis-regulatory element activity and gene expression to further define the activity of genetic programs that support cell state transitions during development [40,43].Wang and colleagues developed a technology to simultaneously profile gene expression, DNAme, and chromatin accessibility in the same cell (scNOMeRe-seq) and applied it to mouse early preimplantation embryos (Figure 3b, upper panel) [43].This study detailed the stepwise and asynchronous activation of fate-associated cis-regulatory elements, with ICM-RE already accessible at the 2-cell stage embryo, while TE-specific regulatory elements become active around the 8-cell stage [43].
As embryos proceed through implantation, cells in the pluripotent epiblast initiate the process of gastrulation, resulting in the generation of the three embryonic germ layers, ectoderm, mesoderm, and endoderm.In this context, another multiomics approach (scNMT-seq) was developed by Argelaguet and colleagues to investigate the dynamic epigenomic changes as cells exit from pluripotency and commit to new lineages [40,44].Similar to the differential activation of lineage-associated elements in the preimplantation embryo, this study found that ectoderm-specific regulatory elements (EC-RE) are accessible and free of methylation in the E6.5 EPI cells, while mesoderm and endoderm-specific REs (ME-RE and EN-RE, respectively) remained inactive.ME-REs and EN-REs are instead gradually demethylated and activated as a response to growth factors within their  [11,26,27,36]).Expression values were normalized to sequencing depth per cell and averaged across lineages and cell types for simplicity.TATA-binding protein (Tbp) expression was used as a reference to compare expression across studies and to allow for classification into low, intermediate or high expression with the transition between intermediate and high being defined at Tbp levels.E, embryonic day; ICM, inner cell mass; EPI, epiblast; Xecto, extraembryonic ectoderm; Xendo, extraembryonic endoderm; Ecto, ectoderm; Meso, mesoderm; Endo, endoderm; ExtraE, extraembryonic; Neuro, neural tissues; Hepa; hepatocytes.Embryo schematics have been drawn with BioRender.

Figure 3
Current Opinion in Genetics and Development Single-cell based approaches to study epigenetic regulation during development.(a) Single blastomere injection combined with single-cell qPCR and imaging to explore the regulation of early lineage choice by ncRNAs and epigenetic regulators [37,38].TE, trophectoderm; ICM, inner cell mass; qPCR, quantitative polymerase chain reaction.(b) scNOMeRe-seq and scNMT-seq as representative multiomics approaches to explore regulatory elements in early development [40,43].RE, regulatory elements; TE-RE, trophectoderm-specific regulatory elements; EC, ectoderm; EN, endoderm; ME, mesoderm; E, embryonic day; CpG, unmethylated CpG; CpGme, methylated CpG.(c) Single-cell zygotic disruption combined with scRNA-seq profiling in mid-gestational WT and mutant embryos as a novel method to explore phenotypes at a high molecular resolution [11].Pie charts are generated from data reported in [11].WT, wild type; PRC, polycomb repressive complex.(d) scCUT&TAG and spatialCUT&TAG as examples of singlecell histone modification mapping during embryonic and adult neurogenesis [48,50].OPCs, oligodendrocyte progenitor cells; COPs, committed oligodendrocyte progenitor cells; MOLs, mature oligodendrocytes; C, spatial cluster.Mouse and embryo schematics have been drawn with BioRender.
respective cell populations, which cooccurs with EC-RE compaction and methylation (Figure 3b, lower panel) [40].Further characterization of these dynamic processes in embryoid bodies revealed that the demethylation of ME-REs in mesodermal cells is dependent on TET enzyme function, in line with reported TET activity at somatic enhancers in mouse and human pluripotent stem cells [40,45].
In both the pre and post-implantation cases, the application of these scMultiomics approaches revealed a temporal hierarchy for timed, lineage-specific enhancer activation and subsequent coordination of the respective gene expression programs.

Assigning cell-type-specific functions for global regulators
The ability to combine these technologies with perturbation approaches provides additional opportunities for insights, particularly for epigenetic regulators where their context-specific nature remains ambiguous.For example, we recently demonstrated the utility of combining a zygotic perturbation strategy with an in vivo single-cell gene expression readout to explore the developmental outcome and molecular phenotype of embryos lacking core epigenetic regulators [11].For instance, all the analyzed PRC subunit KO embryos (Eed, L3mbtl2, Rnf2, and Kdm2b) show delayed development and unbalanced cellular composition at E8.5, with an over-representation of extraembryonic tissues and under-representation or complete absence of mature embryonic cellular subtypes (Figure 3c) [11].Interestingly, although transcriptional phenotypes were not seen at E6.5, one could already observe clear epigenetic alterations compared to wildtype controls [11].In terms of the altered cell type composition, it is possible to envision different models for how these regulators might be acting, including failed repression of alternative lineages or selective advantage of tolerant lineages.In the former case, PRC subunits suppress one or more lineages that in their absence become active, while in the latter only certain lineages are able to successfully differentiate, proliferate and expand in the absence of PRC-based regulation (Figure 3c).Distinguishing these two possibilities is not trivial, as illustrated by embryos lacking maternal PRC1 (Pcgf1/6 matKO embryos).These embryos can develop to term, but display placental enlargement (almost doubled the volume as compared to the wild-type placenta) without affecting embryo size [15].Similarly, Eed disrupted embryos show expansion of the allantois as well as primordial germ cell (PGC)-like cells [11], which could arise through either path and would require molecular barcoding or similar genetic lineage tracing strategies to resolve if this phenotype represents a 'default' differentiation state within the epiblast or if instead PGCs and allantois are preferentially tolerant to the loss of Polycomb group regulation [46,47].
These recent advances helped to begin disentenglingthe exact function of PRCs and other regulators to successfully apportion the differentiating embryo into its appropriate lineages.

Single-cell chromatin profiling in the developing brain
The development of technologies to quantitatively measure histone modifications in individual cells has revealed the presence of "cell-specific epigenetic signatures" that can distinguish transcriptionally similar subpopulations based on their histone modifications at precise genomic locations [48,49].Bartosovic and colleagues developed scCUT&TAG to simultaneously profile four histone modifications in individual cells of a post-natal mouse brain.Their single-cell resolution data could then be used to identify distinct chromatin sub-states during oligodendrocyte maturation in vivo, revealing an increasing amount of promoter-associated H3K4me3 along the maturation axis from oligodendrocyte progenitor cells to mature oligodendrocytes (Figure 2d, upper panel) [48].Advances in spatially resolved epigenomic technologies have also facilitated the geographical mapping of diverse cellular epigenetic states in developing tissues [50,51].For example, a recent study leveraged the scCUT&TAG method to jointly profile three histone modifications in the developing brain of a post-implantation embryo with spatial and near singlecell resolution (spatialCUT&TAG) [50].The integration of this dataset with conventional scRNA-seq enabled researchers to visualize enhancer activity both genomewide and within the spatial contexts of the tissue, alongside quantitative measurements of their target gene expression to develop models of gene regulation in situ.Their analysis demonstrated the presence of distinct chromatin state zones in the developing brain and identified the acetylation of lysine 27 of histone H3 (H3K27ac) at regulatory elements as a primary epigenetic determinant for radial glia diversification during brain development (Figure 3d, bottom panel) [50].
These examples representrecent technological advances that have provided new means to investigate cell-typespecific epigenetic regulation in both time and space, highlighting complex networks of epigenetic interactions that safeguard cell state identity and transitions in a developing embryo.
The ability to combine these approaches is providing conceptual evidence for the opportunity to explore the functions of different regulator classes at higher throughput as well as with greater replicate numbers and across more diverse cellular systems (Figure 4b).The use of different preimplantation and postimplantation models offers the unique opportunity to dissect early and late dependencies on epigenetic regulator function for cell identity acquisition.Moreover, the ability to modulate inhibitor treatment combination, timing, and dosage allows for the quantitative investigation of dependencies and synergistic functions.Finally, the accessible nature of these models also allows for easy modification of the extracellular environment at the metabolic (e.g. via oxygen deprivation) or mechanical levels (e.g.modulating extracellular matrix stiffness and modular components) [66,72].These investigations are furthering our appreciation of the direct crosstalk between metabolism and epigenetic regulation [73].For example, hypoxic conditions can inhibit the action of several epigenetic regulators, including TETs and several histone demethylases, resulting in DNA and histone hypermethylation that appears to influence cellular identities and potential [74][75][76].Interestingly, gastruloids developing in hypoxia display an enhanced and more embryo-like lineage representation, indicating a direct effect of oxygen deprivation on early differentiation events [72].
Taken together, the modularity, accessibility, tractability, and scalability of these new in vitro models of embryogenesis offer a toolkit to complement in vivo studies, and could provide a powerful new lens to explore long-standing questions regarding the specific functions of epigenetic and environmental cues in regulating cell identities.

Conclusion
The recent development and increased accessibility of high-resolution single-cell technologies offer unprecedented avenues to study early embryogenesis.This new lens will greatly facilitate the interpretation of the complex and dynamic epigenetic regulation during early cell state transitions.
As outlined above, many epigenetic regulators contribute to cell identity and fate choice.However, they often appear to act in a tissue-specific fashion despite being broadly expressed in most of the cell types present in a developing embryo.How is cell-type-specific epigenetic regulation then achieved, and what commonalities are shared across the responses seen in diverse progenitor states?How can broadly expressed regulators identify and act on cell-specific targets in the genome?Recent studies have shown that the expression of sequence-specific co-factors can drive tissue-specific epigenetic regulation by tethering them to defined genomic locations [77][78][79].The highly tissue-specific nature of long long ncRNAs may also support this fine-tuned regulation [7].The resulting network implies the presence of ubiquitously expressed epigenetic effectors (such as DNMTs, PRCs, TETs, SWI/SNF) combined with often lowly expressed, tissue-specific co-factors (such as transcription factors, long ncRNAs, and posttranslational modifiers) that act as a scaffold to provide the cell-type-specific positional information necessary to acquire diverse cell identities during development.
Given their accessibility and scalability, in vitro models of embryogenesis offer the potential for high-throughput readouts of broad chemical screens, offering an exciting toolkit to accelerate our understanding of epigenetic regulator function in the establishment of cell identities.Finally, the recent implementation of human in vitro models of embryogenesis faithfully recapitulating critical stages of pre- [80,81] and postimplantation [82][83][84] development may help overcome ethical limitations associated with human embryo research and offer a unique path to understand how conserved epigenetic principles are across mammals as well as to more directly address aspects of these findings as they relate to reproductive and fetal health.

Figure 1 Current
Figure 1

Figure 2 Current
Figure 2

Figure 4 Current
Figure 4