Chromatin “pre-pattern” and epigenetic modulation in the cell fate choice of liver over pancreas in the endoderm

Abstract

Understanding the basis for multipotency, whereby stem cells and other progenitors can differentiate into certain tissues and not others, provides insights into the mechanism of cell programming in development, homeostasis, and disease. We recently reported a screen of diverse chromatin marks to obtain clues about chromatin states in the multipotent embryonic endoderm. Genetic and pharmacologic tests of certain marks’ function demonstrated that the relevant chromatin modifying factors modulate the fate choice for liver or pancreas induction in the endoderm. The information about chromatin states from embryonic studies can be used to predict lineage-specific developmental potential and chromatin modifiers to enhance particular cell fate transitions from stem cells.

Keywords: :

• Chromatin
• competence
• endoderm
• histone marks
• liver
• pancreas

This article refers to:

Chromatin States in Pluripotent Cells

Epigenetic regulation, particularly through chromatin modification and DNA methylation, plays a critical role in controlling gene regulation and cell differentiation in the development of multicellular organism. Various studies revealed that chromatin alterations accompany cell lineage specification during mammalian development.^1,2 In the mammalian embryo, early pluripotent cells initially develop into the endoderm, ectoderm, and mesoderm germ layers. Germ layer cells are multipotent and each layer further differentiates into certain tissues and organs, but not others. Fully understanding the changes of chromatin states in cell lineage specification, which is fundamental to the studies of organogenesis and regeneration medicine, will allow the efficient induction of stem cells into desired cell lineages.³

Much has been learned about the chromatin states in pluripotent cells, specifically embryonic stem (ES) cells and how such states endow the competence to self-renew or initiate particular cell programs.^4,5 Master transcription factors control self-renewal and pluripotency by extensive chromatin binding and autoregulatory loops,^6-8 trimethylation of histone H3 on lysine residues K4 and K27 often marks silent genes with the potential to be activated in early development,⁹ pluripotent chromatin is more loosely structured and accessible than differentiated cell chromatin¹⁰ and P300, H3K27me3 and H3K4me1 are associated with poised enhancers at genes involved in early development.¹¹

While the combined pattern of transcription factors and histone modifications helps explain the competence for all tissue programs, it is not clear how such chromatin states will be relevant to multipotent cells, in which a restricted set of cell fates can be activated and other cell fates cannot. Furthermore, it is not known whether a common “open” chromatin state exists for silent genes in multipotent cells or if the silent genes for different fates are packaged into markedly different states. Superimposed upon these issues is the question of how chromatin states allow or deny the effect of inductive signals that promote cell fates. Lineage-restricted, multipotent progenitors exist at numerous stages of development, they are crucial for homeostatic cell replacement in adult tissues, and they are often the targets of human disease, such as cancer. Thus, it is important to understand how multipotent progenitors are made competent for their particular cell fate choices.

Chromatin States in Embryonic Endoderm Development

To address the above issues, we identified and tested the function of chromatin states in the embryonic endoderm. The ventral foregut endoderm of the mammalian embryo gives rise to liver and ventral pancreas progenitors and is an experimental model for investigating chromatin and inductive signaling.^12,13 The multipotency of the ventral foregut endoderm has been shown by the ability of mouse embryo tissue to activate either the earliest liver or pancreas genes in explants,¹⁴ by the ability of improperly positioned mouse endoderm to acquire a hepatic instead of a pancreatic fate in vivo¹⁵ and by the ability of individually labeled zebrafish endoderm cells to give rise to hepatic and pancreatic descendants.¹⁶ Endoderm cells contain occupied DNA binding sites for FoxA and GATA factors at an enhancer of the silent Alb1 gene;^17,18 the latter being a marker of hepatic specification.¹⁷ FoxA and GATA factors can bind and expose the local chromatin in compacted nucleosome arrays in vitro¹⁹ and elicit chromatin opening in vivo.^20,21 FoxA1 and FoxA2 are both required in the endoderm to activate the hepatic program²² and GATA4 and GATA6 are necessary for liver bud development.^23-25 The early developmental binding of the factors and their ability to open chromatin has led to the proposal that such factors function as competence or pioneer factors in the endoderm.^12,26

Recently, we found that FoxA1 and FoxA2 can recruit the corepressor Grg3, causing compaction of the local chromatin domain.²⁷ Grg3 is co-expressed in the endoderm with FoxA factors, but then is sharply downregulated during hepatic induction.²⁸ Thus liver genes in endoderm could remain competent but silent by the apparent combined action of positive (FoxA and GATA) and negative (Grg3) factors. However, the role of chromatin modifications has been unknown.

By 8.5 d gestation (~7–8 somite pairs; 7–8S), the liver genes Alb1, Afp, and Ttr are induced by BMP and FGF signals, while Pdx1, a pancreatic determination gene,^29-31 is suppressed in the liver domain and activated in the adjacent caudal domain.^16,32-35 The BMP and FGF signals also differentially regulate Prox1, Hnf1b, and Hnf6;³⁵ which are necessary for both liver and pancreatic development. All three of these genes are expressed in endoderm spanning both the hepatic and pancreatic domains.^36-40 By contrast, the induction of Alb1, Afp, Ttr and of Pdx1 are specific to the nascent hepatic and pancreatic domains, respectively, and thus the chromatin states of their promoters and enhancers in the endoderm should provide insight into the regulation of the liver or pancreas fate choice.

Due to the small cell numbers, 400–800 per embryo, the chromatin states in the endoderm, as well as in the mesoderm and ectoderm germ layers, have been elusive. To address this, we developed protocols for fluorescence activated cell sorting (FACS) of ventral foregut endoderm cells and for low cell number chromatin immunoprecipitation (ChIP).⁴¹ We performed a screen for the presence of 15 different chromatin marks at regulatory elements for silent liver and pancreas specific genes in the endoderm. We also investigated such marks in nascent hepatoblasts and tested the functionality of the marks by genetic studies in mice.

Surprisingly, the results reveal a markedly different “pre-pattern” of chromatin states at the liver vs. pancreas regulatory elements, where the genes are poised but not active, in undifferentiated endoderm.⁴¹ In contrast to the liver regulatory elements in undifferentiated endoderm, which were devoid of the positive (H3K9acK14ac) and negative (H3K27me3) marks, the pancreas regulatory elements, except area IV of the Pdx1 gene, contained both marks. Despite the diversity of chromatin marks tested, only H3K9acK14ac showed a consistent, significant increase at the liver elements when the undifferentiated endoderm cells differentiated into hepatoblasts. The co-existence of H3K9K14 hyperacetylation and H3K27me3 at the Pdx1 elements is retained in hepatoblasts. Currently, the low cell numbers preclude a sequential ChIP assay on the native embryonic cells. However, the persistence of both H3K9acK14ac and H3K27me3 marks on the silent Pdx1 gene in sorted Liv2+ hepatoblasts is consistent with their co-existence on individual genes. Such co-existence has been seen in embryonic stem cells.⁴² The co-existence of the positive/negative marks of H3K9acK14ac and H3K27me3 may constitute a new kind of “bivalent” mark.

We next investigated histone modifiers for H3K9acK14ac and H3K27me3, to assess their roles in the liver vs. pancreas fate choice. Of the genes for mammalian lysine acetyltransferases (KATs), mice null for Gcn5l2, CBP or P300 die near gastrulation.^43,44 By contrast, the KAT P/CAF is expressed at low levels in embryos and is dispensable, though it is expressed ubiquitously in adult cells.^44,45 Mice heterozygous for Gcn5l2 or P300 are viable, though the latter can be lethal in certain backgrounds.^43,44 Our genetics studies revealed that P300, but not Gcn5l2 or P/CAF, plays key roles in adding acetylation at the liver-specific regulatory elements, in liver-specific gene activation, and in modulating the liver over pancreas fate choice during embryogenesis.⁴¹

Since H3K27me3 plays an important role in maintaining the silence of developmental regulatory genes,⁴⁶ we investigated whether diminishing this modification in the endoderm would be sufficient to alter the initiation of the pancreatic program. EZH2, a member of the PRC2 Polycomb complex, is a key histone methyltransferase for H3K27me3.⁴⁷ In embryos that we made deficient in Ezh2 in the endoderm, we discovered enhanced ventral pancreas specification with multiple bud-like structures, and the liver bud was smaller.⁴¹ H3K9K14 acetylated chromatin and/or other locally acetylated substrates could play a role in allowing activation when the repressive H3K27me3 methyltransferase is lost. Thus, the chromatin of the pancreas gene in endoderm is set up so that simple loss of the repressive mark facilitates rapid induction. Indeed, the H3K27me3 mark is not seen at Pdx1 later in pancreas development.⁴⁸

SMAD proteins are downstream effectors of BMP signaling⁴⁹ and can interact with P300.^50,51 To directly test the role of SMAD4 in regulating P300 and histone acetylation, we conditionally disrupted Smad4 in the foregut endoderm.^35,52 The study showed that the liver-inducing BMP signal is mediated by SMAD4 and P300 and results in histone acetylation at liver target elements and liver gene activation (Fig. 1). Our studies reveal chromatin “pre-programming” for different lineages in multipotent progenitor cells, they provide approaches that can be applied to other systems, and they provide new landmarks and molecular targets to track and modulate liver and pancreas specification from stem cells.

Figure 1. Chromatin “pre-pattern” and epigenetic modulation during liver and pancreas specification. In undifferentiated endoderm cells, the poised liver and pancreas regulatory elements have distinct chromatin “pre-patterns.” The pioneer factors of FoxA and GATA4/6 occupy liver regulatory elements, which have low or undetectable levels of the positive (H3K9acK14ac) and negative (H3K27me3) chromatin marks. By contrast, the pancreas regulatory elements of the Pdx1 gene, except for area IV, contain both marks. BMP signaling induces SMAD4 to recruit P300 to the liver elements, stimulating histone acetylation there and promoting liver specification. During this period, EZH2 restricts the specification of pancreas progenitors.

Distinct Types of Positive-Negative Components at Poised Genes

We found the co-existence of the positive and negative marks of H3K9acK14ac and H3K27me3, respectively, at the silent Pdx1 elements in the endoderm, where the Pdx1 gene is competent or poised, to be activated. As described above, bivalent domains of H3K4me3/H3K27me3 often mark poised early developmental genes.⁹ Also, the positive and negative effectors FoxA/GATA and Groucho/Grg3, respectively, appear to co-exist at the silent liver regulatory elements in the endoderm, where the genes are poised to be activated. Other examples of positive and negative components at poised genes include transcription factors bound to enhancers or promoters of silent genes in ES and iPS cells that are activated late in differentiation. In ES cells, the enhancer of the hypomethylated, silent Alb1 gene is occupied by FoxD3, which represses transcription.⁵³ During endoderm induction, FoxD3 is replaced by FoxA1, all prior to Alb1 activation.⁵⁴ Similarly, in nascent endoderm, the induced FoxA1 binds within the methylation-free Afp distal promoter, which is necessary for Afp activation during stem cell differentiation.⁵⁵ The positive factors are bound at unmethylated CpG residues amidst many methylated CpGs, which are typically considered repressive; thus together constituting a distinct kind of positive/negative mark. In ES cells, an intergenic enhancer of λ5-VpreB1 genes,⁵⁶ which are expressed in pro- and pre-B cells, is occupied with Sox2 and FoxD3. Sox2 acts as a positive factor, contributing to the establishment of the H3K4me2 mark, whereas FoxD3 is a negative factor, repressing intergenic transcription from the enhancer.⁵⁷ Another example is the pre-loading of RNA polymerase II, but its pausing, at silent genes with the potential to be activated in development.^58-61 Since the poising of genes by simultaneous positive and negative chromatin components may be general, but involving diverse mechanisms, we suggest broadening the definition of “bivalency” in stem and progenitor cell chromatin to refer to the simultaneous presence of positive and negative components at poised genes, and not solely to the original H3K4me3/H3K27me3.⁹ By having the poised state result from a dynamic equilibrium between positive and negative effectors, it may facilitate the synchronous activation of genes in the rapidly developing embryo.

Future Directions

While in principle it should be possible to characterize chromatin states in germ layer cells derived from embryonic stem cells in vitro, early in this study we found that the chromatin states in ES-derived endoderm can be different from those we have characterized in vivo (C.-R. Xu, P. Gadue, C. Nostro, G. Keller and K.Z. unpublished data). Knowing the chromatin states in native embryonic tissue can be used to define benchmarks of proper progenitor cell programming in stem cell studies and molecular targets for enzymatic modifiers that function in the cell fate transitions. It will be informative to perform genome wide analysis for H3K9/K14 acetylation and H3K27me3 in the undifferentiated endoderm, as the marks were found to be different between in vitro and in vivo populations. There may be key genes that demonstrate altered chromatin marks in ESC-derived endoderm compared with in vivo-derived endoderm that could limit in the in vitro differentiation process. The initial analysis of the genome wide ChIP data will focus on genes known to be involved in pancreatic and liver.

The mechanistic studies to date have focused on the chromatin regulation of hepatic specification. We don’t yet know how the repressive chromatin mark of H3K27me3 is lost from Pdx1 regulatory elements, which appears to be required for initiation of the pancreatic program. We hypothesize that cell signaling regulates the recruitment of a histone demethylase to Pdx1 elements upon pancreas specification. Such information could provide insight into ways to enhance pancreas specification from stem cells.

Our work has provided a paradigm for using antagonists of histone modifiers at specific developmental stages to alter cell lineage specification. This was enabled by a careful analysis of chromatin prepatterns of native progenitor cells in vivo. Such approaches can be taken with any stem or progenitor cell type to predict developmental potential and enhance differentiation of stem cells.

Acknowledgments

We thank the members of the Zaret group for advice and comments on the manuscript and ongoing projects, and Eileen Pytko for help in preparing the manuscript. This work was supported by NIH grants R37GM36477 and U01DK072503 to K.S.Z, and K01DK093886 to C.-R.X.