CELL FATE DETERMINATION IN 3D: REGULATION OF GENE EXPRESSION VIA CHROMATIN INTERACTIONS WITH THE NUCLEAR MEMBRANE

Abstract

In the nucleus of all cells, DNA is packaged in association with histone proteins to form chromatin. It is becoming increasingly clear that the organization of chromatin in three dimensions within the nucleus is highly regulated and can contribute to gene expression and cell function. The regions of the genome that are near the nuclear periphery are termed “lamin associated domains” or LADs (1–3). Here, I present a theory, based on recent results, for “chromatin competence” in which the organization of LADs in a progenitor cell accounts for the ability of that cell to respond to external factors to promote differentiation into one lineage or another. I propose that a cell can only respond to an inductive cue if the downstream genes are available for activation, and that genes are not available for activation if they are sequestered in LADs.

INTRODUCTION

Conrad Waddington is commonly considered the father of epigenetics, and he provided a frequently reprinted model of the “epigenetic landscape” to depict the interactions of the environment with a multipotent progenitor cell during development as it traverses the landscape of progressive lineage restriction. In his famous treatise Organisers and Genes published in 1940 (4), Waddington focused attention on the concept of cellular “competence,” i.e., the ability of a cell to respond to inductive signals. In modern-day terms, competence refers to the ability of a cell to respond to morphogens or growth factors. Despite the fact that all cells in an organism contain the same genome and DNA sequence, some cells will respond to a growth factor in one way, whereas other cells will respond to the same factor in a different way. For example, bone morphogenetic protein factor will cause mesenchymal stem cells to become bone, whereas bone morphogenetic protein will cause cardiac progenitor cells to become cardiac muscle. A mechanistic explanation for this phenomenon remains elusive.

Potential explanations for different responses to a common inductive signal might include differences in the expression of cell surface receptors, or differences in the expression of components of an intracellular signal transduction pathway that is downstream of a growth factor receptor, or differences in expression of transcription factors necessary for activation of downstream genes. However, recent data from my laboratory and others suggest another possibility. Differences in cellular competence may be due to differences in the three-dimensional (3D) packaging of chromatin in the responding cells. Specifically, the pattern of lamin-associated domains (LADs) may define whether downstream genes are accessible or available for activation (5).

For many years, it has been appreciated from electron microscopic studies that heterochromatic, tightly packed chromatin is clustered at the nuclear periphery in close association with the inner nuclear lamina. LADs are cell-type specific and change with cell differentiation (6–9). LADs are thought to be restrained at the nuclear periphery by protein complexes (LAD tethers) that interact with nuclear lamins. The specific components of LAD tethering complexes have not been well defined, although components including Hdac3, PRR14, HP1, cKrox, emerin, and Lap2ß have been implicated (10). LAD tethers presumably recognize specific regions of the genome via subunits that bind directly to DNA (such as the zinc finger protein cKrox) or by “reading” and binding to a specific histone mark. The nuclear lamina itself contains many proteins, several of which have been implicated in human diseases that are collectively known as “laminopathies.” These include premature aging syndromes such as Hutchinson-Gilford progeria and several forms of cardiomyopathy and muscular dystrophy. Why abnormalities in nuclear lamina proteins cause these specific phenotypes remains unclear, although changes in lamina-chromatin interactions have been postulated to contribute to underlying changes in gene expression in these disorders (11,12).

DELETION OF HDAC3 CAUSES PRECOCIOUS CARDIOMYOGENESIS

Several members of the family of histone deacetylase (Hdac) enzymes, which remove acetyl moieties from histone tails and repress transcription, play important roles in cardiac development and adult cardiac function (13). As part of our analysis of the specific roles of Hdac family members, we deleted Hdac3 from the hearts of mice using Cre-lox technology. Deletion in the adult heart produced no obvious phenotype when mice were fed a normal diet. However, all Hdac3-deficient animals developed cardiomyopathy and early lethality when fed a high fat diet (14). This fascinating example of genetic susceptibility to heart disease will not be discussed further here. Rather, the focus will be on the effects of deleting Hdac3 in cardiac precursor cells early in embryonic development. This genetic manipulation in mice results in abnormal cardiac development associated with precocious myocyte maturation and sarcomere assembly (5).

We further examined this effect by modeling cardiac development in the tissue culture dish. Embryonic stem cells (ESCs) were engineered such that addition of tamoxifen to the culture media would result in the genetic deletion of Hdac3. ESCs can be differentiated into embryoid bodies that contain cardiac progenitor cells characterized by expression of Pdgf receptor alpha and Flk1. These cells can subsequently differentiate into one of three alternative fates: cardiac muscle, smooth muscle, or endothelial cells. Deletion of Hdac3 at the cardiac precursor stage resulted in a greater percentage of cells adopting a cardiac muscle fate at the expense of the other lineages. When Hdac3 was deleted just 1 day after the cardiac progenitor stage (i.e., early after the time of lineage determination) little effect was seen (5). Thus, deletion of Hdac3 in cardiac precursor cells both in vivo and in vitro results in precocious cardiac myocyte differentiation.

To understand how Hdac3 regulates cardiac lineage determination, a rescue experiment was performed in the ESC model system. ESCs were differentiated into cardiac precursor cells, and Hdac3 was deleted by addition of tamoxifen, while simultaneously introducing virus expressing either wild-type or mutant forms of Hdac3. As expected, wild-type Hdac3 rescued the effects of tamoxifen-induced loss of Hdac3. Surprisingly, however, a mutant form of Hdac3 lacking catalytic deacetylase activity also rescued the loss of Hdac3. This result indicates that the function of Hdac3 that modulates cardiac lineage determination does not require histone deacetylase activity.

As noted above, Hdac3 has been implicated as a LAD tether (15,16), and this function does not necessitate enzymatic activity. Therefore, we postulated that Hdac3 affects cardiac lineage determination by acting as part of a LAD tethering complex. Consistent with this hypothesis, introduction of a virus encoding a fusion of catalytic-dead Hdac3 with the nuclear lamina protein Lap2ß (a fusion protein that is restricted to the nuclear periphery) successfully rescued genetic loss of Hdac3 in ESCs (5). Thus, LAD tethering appears to modulate the development of cardiac progenitor cells into mature cardiac myocytes.

CARDIAC DIFFERENTIATION INDUCES MOVEMENT OF CARDIAC GENES FROM THE NUCLEAR PERIPHERY

We have developed high-resolution techniques to image the location of specific genomic loci within the nucleus using fluorescence in situ hybridization coupled with immunohistochemistry and 3D reconstruction (Figure 1). Using a labeled probe to the Titan gene locus, e.g., we could show that this region of the genome is situated adjacent to the nuclear lamina in undifferentiated ESCs. On differentiation of ESCs into cardiac myocytes, which express Titan, this region of the genome moves away from the nuclear lamina. The same is true for several other cardiac-specific genes, whereas neuronal genes that are located at the periphery in ESCs do not change position on cardiac differentiation.

Fig. 1.

Cardiac genes are released from the nuclear lamina during cardiac differentiation. (A-C) Immunostaining (A) and 3D reconstruction (B,C) showing Titan (Ttn) locus in relation to nuclear lamina and H3K9me2-marked chromatin in ESC. (D) 3D reconstruction showing Ttn locus in relation to nuclear lamina in CM. Bottom panels in (C,D) are higher magnification, slightly rotated images, showing distance of each Ttn allele from lamina (Lmnb). Ttn is located further away from the nuclear periphery in CM (C versus D). Abbreviations: 3D, three-dimensional; ESC, embryonic stem cells; CM, Cardiomyocyte. Reprinted from Poleshko et al. (5) with permission.

Many cardiac-specific genes are not located at the periphery in ESCs; therefore, they do not change from peripheral to nucleoplasmic location on differentiation. But for those that do change position, we postulate that movement away from the highly repressed and heterochromatic environment of the nuclear periphery is associated with “availability” or competence for gene activation.

Consistent with this idea, genetic deletion of Hdac3 in ESCs is associated with movement of cardiac genomic loci, such as the Titan locus, away from the nuclear periphery although activation of Titan gene expression requires growth for several days in differentiation media.

HETEROCHROMATIN AT THE NUCLEAR PERIPHERY IS MARKED BY HISTONE 3 LYSINE 9 DIMETHYLATION

The choreographed movement of cardiac genomic loci away from the nuclear lamina during cardiac differentiation suggests that the tethering of LADs to the nuclear periphery is a regulated process. Therefore, we hypothesized that one component of an LAD tether might be a “reader” of a specific histone mark that designates specific genomic regions for peripheral localization. We screened numerous antibodies directed against specific histone modifications by immunohistochemistry in ESCs and various other cultured cell types (Figure 2). We also examined the specificity of these antibodies using peptide arrays containing a wide variety of modified peptides representative of histone tails. Intriguingly, one commercially available antibody against histone 3 lysine 9 dimethyl (H3K9me2) revealed localization of this mark exclusively to the nuclear periphery, whereas H3K9me3 gave an entirely different pattern of immunostaining with signal both at the periphery and within the nucleoplasm (5). The specificity of both antibodies was confirmed by specific interactions with the appropriately modified peptides on the arrays. Interestingly, the antibody specific for H3K9me2 was unable to recognize that mark on H3 peptides when the adjacent serine residue (S10) or the subsequent threonine (T11) was phosphorylated, suggesting that the phosphate on these residues “shielded” the dimethyl mark on lysine 9 from recognition by the antibody (Figure 3). None of the other histone marks examined showed a pattern of immunostaining restricted to the nuclear periphery (Figure 2). This finding, in addition to those mentioned above, convinced us that peripheral heterochromatin is uniquely decorated with at least one post-transcriptional modification of histones such that it could be uniquely recognized by an appropriate “reader” of this mark. We confirmed that H3K9me2 is restricted to the nuclear periphery in a wide range of cell types.

Fig. 2.

H3K9me2 marks peripheral heterochromatin. Immunofluorescence images of C2C12 skeletal myoblasts with indicated repressive chromatin marks and LaminB. Reprinted from Poleshko et al. (5) with permission.

Fig. 3.

A model kinase regulation of LAD tethering. The schematic depicts an epigenetic “reader” that binds to H3K9me2 and tethers heterochromatin to the nuclear periphery. Phosphorylation of S10 or T11 “shields” the H3K9me2 mark such that the “reader” can no longer bind and the LAD is released from the nuclear periphery allowing genes located within the genomic region to be accessible to transcriptional activators. Abbreviation: LAD, lamin-associated domains.

Further confirmation of this finding was provided by experiments involving chromatin immunoprecipitation followed by next-generation sequencing (ChIP-seq). LADs are most commonly defined as the regions of chromatin that associate with lamin B in ChIP-seq experiments using an anti-lamin B antibody for chromatin immunoprecipitation. We compared lamin B and H3K9me2 ChIP-seq datasets from ESC and ESC-derived cardiomyocytes and found a high degree of overlap, further confirming the peripheral restriction of H3K9me2 (Figure 4). These datasets also delineate the genome-wide changes in LADs that take place during cardiac myocyte differentiation revealing the spectrum of cardiac-specific genes released from repression at the nuclear periphery during cardiogenesis.

Fig. 4.

H3K9me2-marked chromatin mirrors lamina bound chromatin. Correlation of LaminB and H3K9me2 occupancy derived from ChIP-seq data sets across the genome in 10 kb bins; Pearson’s correlation r = 0.84. Reprinted from Poleshko et al. (5) with permission.

DISCUSSION

Our findings add to a rapidly accumulating literature suggesting that the 3D organization of chromatin within the nucleus is highly regulated and contributes to cellular identity and function. Although the existence of peripheral heterochromatin has been recognized for many years, the dynamic nature of genome-nuclear lamina interactions has only been appreciated more recently. Our studies confirm that dysregulation of LAD dynamics associated with differentiation alters the differentiation process and can change the cell fates of progenitor cells exposed to differentiation cues. We propose that release of genomic regions from tethering to the nuclear periphery makes those regions more accessible to transcriptional activation complexes, and thus makes them “competent” for gene activation. Thus, a cell is competent to respond to a developmental morphogen only when critical downstream genes are not sequestered and silenced at the nuclear periphery where associated histones are decorated with H3K9me2. We believe that this is one of perhaps several molecular mechanisms that correlate to Waddington’s notion of cellular competence, and we refer to this mechanism as “chromatin competence.”

When a population of cells is exposed to conditions that can change the fate of the cells, such as reprogramming factors used to make induced pluripotent cells or transdifferentiation factors such as MyoD that can transform one cell type to another, the efficiency of cellular conversion is low. Some cells respond by changing cellular identity while others do not. Perhaps this is because LAD tethering is a stochastic process with “on” and “off” rates like any other chemical reaction, and cells can only respond when necessary downstream genes are “available” or untethered. This model would predict that forced untethering or release of LADs from the periphery would enhance the efficiency of cellular conversion. Likewise, it is possible that pathological conditions in which LADs are inappropriately released might not immediately result in altered gene expression, but might make those cells more susceptible to altered gene expression under the right conditions. Perhaps this would make those cells more susceptible to transformation under the influence of environmental carcinogens. These predictions will need to be tested experimentally.

The observation that the H3K9me2 epitope is shielded by S10 or T11 phosphorylation (Figure 4) has exciting implications. If the identity of a cell is, at least in part, defined by the regions of the genome stored in LADs (i.e., the “LAD-map” of the cell), then it would seem likely that the cellular LAD-map would be remembered through mitosis such that daughter cells have the same LAD-maps as the parent cell. However, the nuclear lamina disappears during mitosis, and is re-established after separation of the daughter chromatids. This process is known to require Aurora B kinase as well as a phosphatase activity late in mitosis. Aurora B kinase phosphorylates H3S10 early in mitosis, and could therefore function to release LADs from the nuclear lamina at mitotic onset by shielding the H3K9me2 mark (without removing it) and allowing the nuclear lamina to be degraded. Near the end of mitosis, after sister chromatid separation, a phosphatase could remove the phosphate from S10 thus revealing the H3K9me2 mark and allowing re-establishment of designated LADs. In fact, the unmasked H3K9me2 mark could serve as a nidus around which the new nuclear lamina is assembled. This would provide an elegant mechanism of epigenetic memory to allow daughter cells to “remember” the cellular identity of its parent. In the setting of an asymmetric cellular division, perhaps an asymmetrically distributed S10 or T11 kinase results in the re-establishment of slightly different LADs in the two daughter cells, producing two different cell types. Furthermore, it is attractive to speculate that some signal transduction cascades might converge upon S10 or T11 kinases during interphase and thus untether specific regions of the genome from peripheral repression, making those regions competent for gene activation. One could imagine, for example, signals that promote changes in cellular metabolism (e.g., from gluconeogenesis to fatty acid metabolism) resulting in the expression of an entire spectrum of metabolic genes by inducing S10 or T11 phosphorylation at specific regions across the genome and the simultaneous release of many genomic loci which would then be available for activation. We are eager to explore the implications of signal transduction at the nuclear envelope.

DISCUSSION

Schreiner, Los Altos: Jonathan, absolutely fascinating presentation. I was interested at some of the mutations that you described are associated with any type of aging. Have you actually looked at normal chronological aging and do you see a confusion of these lineage patterns that you have described so elegantly in the developing heart? Is this a potential contribution to the eventual failure of the aging heart in terms of errors in this kind of packaging sequence that you are describing?

Epstein, Philadelphia: Yes. So, in progeria mutations, there are clearly abnormalities of the structure of the nuclear membrane. It looks abnormal when you just look at the cells. And people have hypothesized a number of different reasons why that could lead to premature aging including altered gene expression in the cells. So, we’re looking at senescence right now in the laboratory. We’re looking to see if the lamin-associated domains (LADs) or the pattern of LADs is altered. There are changes that have been reported in those but whether they cause the phenotype is hard to prove. I can’t conclude that yet.

Dennery, Providence: That was really great. I was about to ask the senescence question but also wanted to know whether these same patterns of expression are noticed in other stem cell components in other tissue. For example, the lung seems to have some of those same sort of multi-potent stem cell progenitors which then leads to various stem cell types. Is there any knowledge around other tissues?

Epstein, Philadelphia: It’s early days. It was initially thought that these lamin-associated domains were constant among different cell types and hence not so interesting. We now know by looking more closely that there are constant LADs that are common amongst most cell types but also variable LADs that change between different cell types, and we’re just now creating what we call the “LAD maps” of different lineages and cells. So, as you know, I have some interest in lung development and lung stem cells and we are looking there as well.

Hempstead, New York: Jon, very nice. Do you want to speculate at all about how this may — what the implications may be in tumorigenesis?

Epstein, Philadelphia: Tumorigenesis has many parallels to multi-potency and to stem cell biology. My guess is that when these LADs are inappropriately released, the cell becomes more competent to respond to signals including carcinogenic signals, and that stimuli that turn on pathways inappropriately, that are inappropriate for the cell, can promote carcinogenesis. And so, I would guess that the inappropriate release of LADs would make cells more susceptible to those signals. But it’s all theory.

Epstein, Philadelphia: Can I just add one more answer to your question if I may. One of the kinases that’s been reported to phosphorylate one of these residues that would release LADs is actually the nuclear form pyruvate kinase. It has been implicated in the Warburg effect and some of the metabolic shifts that takes place as cells undergo carcinogenic changes. So, that might be a direct mechanistic answer to your question as to how one could understand the connection between carcinogenesis and lamin-associated domains.

Prabhu, Birmingham: Beautiful presentation. My question is, you know you’ve shown this related to development — does a similar patterning or a reversal of patterning occur after injury and is this a basis for differentiation in the microenvironment?

Epstein, Philadelphia: That’s a great question. So, we’ve begun to ask whether if we release the LADs by removing some of these tethers — are the cells more competent to transdifferentiate from one cell type to another? Actually Hdac inhibitors have been used to promote making induced pluripotent stem cells. So far those experiments haven’t worked for us. We have not seen an enhancement of transdifferentiation.