The epigenetic encoded memory of the innate immune system

Abstract

Stimulation or infection of innate immune cells induces profound epigenetic changes, including the induction of histone modifications and alterations in DNA methylation levels. While some of these changes are rapidly reversible, others appear to be long-lasting, even in mitotic populations, with important functional consequences for the stimulus-experienced cell. Here we discuss the individual contributions of each of the plethora of known epigenetic modifications to the initial transcriptional response to immune activation, their dynamics as cells return to homeostasis, and their contribution to memory of the initial stimulus.

Introduction

The mammalian immune system is comprised of a highly diverse repertoire of cell types that perform unique functions, ranging from the release of antigen-specific antibodies by B-cells, to the production of cytotoxic granules by CD8+ T-cells, and the phagocytosis of cellular debris and microbes by macrophages. Remarkably, all of these cells share the same DNA sequence, raising the fundamental question of how cell-type specific functions are encoded. Increasingly, we now understand that not all genes are equally activated in all cell types, due to the fact that DNA is differentially packaged into regions of open and closed chromatin [1,2]. Open chromatin is more accessible to the transcription factors (TFs) that promote gene expression, and often occurs at the promoters and enhancers (distal regulatory elements to which these TFs bind) of cell-type specific genes [1,2]. Conversely, inaccessible chromatin usually occurs at DNA regions regulating genes that are silenced or unimportant to the cell.

Epigenetic modifications, including post-translational modifications of histones and modulation of DNA methylation, are intimately tied to the degree of chromatin accessibility at any region of the genome. For example, promoters and enhancers are associated with distinct epigenetic modifications depending on their relative level of activity (see Table 1). For example, histone 3 lysine 27 acetylation (H3K27Ac) is found at active, but not inactive promoters and enhancers, and H3K4me1 and H3K4me3 are primarily found at enhancers and promoters, respectively [3,4]. While our current understanding of the relationships between the various epigenetic modifications and chromatin accessibility remains largely correlative, data suggest that, at least in immune cell development, the establishment of such lineage specific chromatin accessibility states is driven by the activity of pioneer TFs unique to each lineage [5–7]. These pioneer TFs bind to different sets of enhancers and promoters to promote a localized increase in chromatin accessibility via the displacement of nucleosomes [5,7,8] – the basic unit around which DNA is wrapped – followed by the recruitment of additional chromatin remodelers [7,8]. For some genes, pioneer TFs are sufficient to drive constitutive expression of that gene. However, in many cases, additional cooperative binding by stimulus-induced TFs is required to drive expression. In these cases, primed enhancer elements largely function as the framework within which stimulus-induced TFs, such as NF-κB, AP-1, STATs, or IRFs can bind to modulate the expression of associated genes, and enable different cell types to activate a unique set of genes in response to the same stimulus [9–12].

Table 1.

Specific combinations of histone modifications and DNA methylation levels occur at different regulatory elements and can also be used to identify these elements de novo.

Genomic element	Epigenetic marks	Notes
Silenced promoter [58]	High DNA methylation	Methylated promoters are usually irreversibly silent
Poised/inactive promoter [58]	H3K4me3, hypomethylated	May be more susceptible to DNA methylation in disease contexts
Active promoter [4,29,58]	H3K4me3hi, H3K27Ac, hypomethylated, high RNA PolII, moderate p300
Poised enhancer [4,15,29]	H3K4me1, moderate p300	Poised enhancers are established and bound by pioneering transcription factors
Active enhancers [4,15,29]	H3K4me1, H3K27Ac, hypomethylated under certain conditions, p300, H3K4me2
Latent enhancer [15]	unmarked	Defined by a lack of marks associated with poised or active enhancers, but gains these marks when a cell is stimulated
Active Gene body [29]	H3K36me3, DNA methylation	Unlike at regulatory regions, DNA methylation within gene bodies is associated with increased transcription
Heterochromatin, repressed regions [29]	H3K9me3, H3K27me3	Unlike at regulatory regions, DNA methylation within gene bodies is associated with increased transcription

While a cell-type specific chromatin landscape maintains a cell’s identity and specifies its basic functional repertoire, the innate immune cell epigenetic landscape remains, paradoxically, highly dynamic post-differentiation. The microenvironment [13,14], as well as various Toll-like receptor agonists [15–17], bacteria [18], or fungal components [19–21] have the ability to dramatically induce epigenetic reprogramming of innate immune cells. Understanding the dynamics of these changes both during and following immune challenge is critical to understanding how cells combine the need for epigenetic plasticity with the maintenance of a cell-type specific baseline, or homeostatic state. The primary focus of this review, therefore, is on the dynamics of the post-differentiation innate immune chromatin landscape in response to such stimuli with a particular focus on key histone modifications and DNA methylation at enhancers, which have been shown to be the most dynamic between cell types [7] and activation states [13,22,23]. We discuss how post-mitotic innate immune cells, or self-renewing immune progenitors, undergo substantial chromatin remodeling in response to inflammatory stimuli and infection, and discuss the long-term functional impact that chromatin remodeling has on these cell types. Finally, we discuss critical remaining questions. Within a single cell type, what is the full repertoire of epigenetic changes induced by any given stimulus, and what are the relative kinetics with which these changes revert back to baseline upon removal of the stimulus? Do cells take active steps to revert stimulus-induced epigenetic changes back to baseline, or is it a passive process? And, are there certain types of induced epigenetic marks that remain post-stimulation, and how does this differ between dividing and post-mitotic cell types? In answering these questions, we not only gain a better understanding of the dynamics of innate immune cell epigenetics, but come closer to understanding how mutations in epigenetic regulators influence susceptibility to a number of modern-age diseases including cancer [24,25] and atherosclerosis [26,27].

Chromatin accessibility landscapes are responsive to immune stimulation

The most critical environmental cues take the form of a pathogen or danger signal to which innate immune cells are programmed to recognize and respond to by secreting cytokines, presenting antigens, and alerting the adaptive immune system. Although pre-existing enhancer elements marked by H3K4me1 provide a framework within which signal-induced TFs can bind to further regulate gene expression, we now know that this process itself induces epigenetic changes. For example, mouse bone marrow derived macrophages (BMDMs) stimulated with LPS for 24 hours gain H3K27Ac at more than 5000 pre-defined enhancers [15]. Similarly, human monocytes stimulated with certain β-glucans, BCG, or LPS, undergo genome wide changes in H3K27Ac levels, as well as more modest changes in H3K4me1 and H3K4me3 levels at enhancers and promoters, respectively [16,17,19–21].

In addition to the deposition of modified histones at preexisting promoters and enhancers, some pro-inflammatory stimuli also induce the de novo formation of enhancer elements absent prior to immune activation (also referred to as “latent enhancers”; Figure 1a,b). For example, stimulation of mouse BMDMs with LPS led to the induction of more than 1000 de novo enhancers, constituting more than 15% of all activated enhancers [15]. Likewise, the infection of human dendritic cells (DC) with Mycobacterium tuberculosis (Mtb) leads to the emergence of hundreds of de novo enhancers [22]. In a recent paper investigating the effects of IL-1β and IFNγ on human pancreatic islet cells, Ramos-Rodríguez et al. identified 3,800 regulatory elements responsive to cytokine stimulation and classified these elements either as primed-or neo-response elements according to whether they had some preexisting accessibility at baseline and further gained accessibility and H3K27Ac in response to stimulation, or whether they were previously inaccessible, but gained accessibility and H3K27Ac de novo, with neo-response elements reportedly making up about 45% of induced regulatory elements in this system [28].

Figure 1.

Infection or stimulation of innate immune cells induces multifaceted epigenetic changes at primed and de novo regulatory elements. (a) De novo regions are defined as being previously unmarked in naïve cells, while primed enhancers have a higher basal level of accessibility, are marked with H3K4me1, and are enriched for 5hmC. (b) Stimulation induces the binding of stimulus-induced TFs to primed enhancers, leading to gains in H3K27Ac, H3K4me1, increased chromatin accessibility, and gene expression. Pioneer TFs may cooperate with stimulus-induced TFs to bind to low affinity sites present at de novo enhancers. (c) TET2 interacts with the enhancer-bound protein complexes and catalyzes conversion of 5mC to 5hmC, which is later replaced with unmethylated cytosine. (d) Changes in DNA methylation have been shown to be longer lasting over the course of an infection compared to H3K4me1 and H3K27Ac. While removal of stimulus leads to the loss of stimulus-induced TF binding, as well as H3K27Ac, evidence suggests that the state of DNA hypomethylation, as well as H3K4me1 may remain long-term at both primed and de novo enhancers.

DNA hypomethylation at activated enhancers

In the context of the immune response most epigenetic studies have focused exclusively on posttranslational modification of histone tails, leaving the role of other epigenetic marks in the transcriptional response to infection unclear. DNA methylation has been particularly understudied, due to the belief that methylation marks are highly stable, and unlikely to respond to environmental perturbations on a short time scale. Recent studies, however, strongly challenged this idea. For example, infection of post-mitotic human DCs or macrophages with live Mtb leads to an active loss of DNA methylation at thousands of enhancers throughout the genome [22,29], and these changes are strongly predictive of changes in expression levels of nearby genes. Other studies correlate these two processes [30–37], yet current data does not support a causal relationship between changes in DNA methylation and changes in gene expression in response to infection. Indeed, Pacis et al., have shown that virtually all changes in gene expression in response to infection occur before any detectable changes in DNA methylation, indicating that the observed losses in methylation are a downstream consequence of transcriptional activation (Figure 1a,b,c). Based on our current knowledge, it appears that site-specific regulation of DNA demethylation is primarily mediated by the binding of stimulus-induced TFs (e.g., NF-kB and IRFs), which instigate chromatin opening followed by the recruitment of histone acetyltransferase p300, and the subsequent deposition of activating H3K27ac marks to these regions [38]. Interestingly, p300 can enhance the enzymatic activity of ten-eleven translocation (TET) enzymes [39]— the enzymes that convert 5-methylcytosine (5mC) to 5-hydroxymethylcytosine (5hmC) [40] – which might account for the eventual loss of DNA methylation in response to infection. Several recent reports have shown that other epigenetic modifications, such as the H3K4me1 enhancer mark, have a similar passive regulatory function [41–43] raising questions about the role played by infection-induced epigenetic changes to the establishment of the core regulatory program engaged upon infection.

Functional consequences of stimulus-induced epigenetic remodeling

Stimulus-induced epigenetic reprogramming can have consequences for cellular behavior and function after removal of the stimulus. However, the nature in which epigenetically reprogrammed cells become functionally altered is highly variable and context dependent. In what is called “endotoxin tolerance”, LPS exposure induces significant alterations in the levels of H3K27Ac at promoters and enhancers [16,21]. Subsequently, when LPS exposed macrophages are re-stimulated with LPS [21,44] or other ligands [16], they mount an epigenetically and transcriptionally altered response characterized by decreased levels of inflammatory agents [16,21,44], and a decreased ability to upregulate H3K27Ac at the promoters of those genes [21,44]. On the other hand, exposure of innate immune cells to certain stimuli (e.g., Candida albicans, its cell wall component β-glucan, BCG, or IFN-β) can prime the cells to secrete increased levels of pro-inflammatory cytokines, as well as to respond faster to secondary stimulation. [19,45,46] This priming is generally thought to be due to the deposition of H3K27Ac, H3K4me1, and H3K4me3 at the enhancers and promoters of transcriptionally upregulated genes [19,20,46], which can further recruit additional chromatin remodelers to promote increased chromatin accessibility. The role that changes in DNA methylation plays in the process, needs much further investigation.

Interestingly, the tolerization of macrophages with LPS, either in vitro or in vivo, followed by an in vitro β-glucan stimulation is enough to restore the transcriptional response of 60% of tolerized genes to levels comparable to that of naïve macrophages [21]. These findings point to high levels of epigenetic plasticity, and suggest that the transcriptional output of innate immune cells depends on their previous history of exposures, not only the most recent.

Maintenance and functional consequences of stimulus-induced epigenetic changes in mitotic cells

In all of the above cases, studies were performed in post-mitotic cells, or in slowly dividing cells over a short time period. Arguably, the short lifespan of cells such as peripheral blood monocytes limits the biological relevance of stimulus-induced epigenetic memory in these systems, given that any experienced cells, would be rapidly replaced by new “naïve” monocytes from the bone marrow. A few recent publications suggest that some self-renewing cell populations, however, may also have the capacity to maintain stimulus-induced epigenetic changes, which, given their continuous self-renewal could potentially be maintained indefinitely [45,47–50]. For example, intravenous injection of mice with BCG followed by antibiotic treatment leads to epigenetic, transcriptional, and functional differences in the BMDMs derived from the bone marrow of exposed mice that last up to 5 months post BCG vaccination. These BMDMs have significantly altered levels of H3K27Ac at thousands of sites overlapping primarily with enhancers and are more effective at controlling a subsequent in vitro Mtb infection [48]. In another study, Naik et al. report that skin exposed to topical imiquimod (IMQ), and subsequently challenged with a punch wound 30 days later, is able to close the wound at a significantly faster rate compared to previously unexposed skin. ATAC-seq performed on isolated EpSCs 6 days after the start of IMQ application revealed more than 40,000 differentially accessible sites compared to unexposed EpSCs. Notably, more than 2000 differentially accessible peaks remained after 180 days, suggesting that in the absence of any inflammatory stimulus, at least some changes in chromatin accessibility could be maintained, even in mitotically active stem cells [47].

Reestablishment of the pre-stimulation epigenetic landscape in quiescent and mitotically active cells

What are the kinetics with which reprogrammed cells return to a “baseline” epigenetic state? The data reported in the above studies suggests that to some extent, certain inflammatory agents may induce epigenetic modifications that are never fully erased. In post-mitotic systems we know that not all epigenetic modifications are maintained equally well. For instance, H3K27Ac is deposited very quickly onto activated enhancers and promoters [51], but also lost very rapidly (by 4 hours post stimulation in one study) even if the stimulus is still present [15]. On the other hand, H3K4me1 has been shown to be a much more stable mark that was reported to increase slowly in comparison to various histone acetylation marks (H3K27Ac was reported to increase within one hour, while H3K4me1 appear between 6 and 24 hours in this system) [51], but is also much longer lasting-remaining at high levels at 24 hours of stimulation (compared to 4 hours for H3K27Ac) [15]. Similarly, DNA methylation changes have a slow onset but a high level of stability. In human dendritic cells infected with Mtb, it takes as long as 18 hours before active enhancers first become DNA demethylated compared to both H3K27Ac and gene expression levels being upregulated by 2 hours of infection, although these enhancers remained demethylated over the course of the 72hour study period (Figure 1d) [29].

Surely, one would expect these dynamics to change in dividing cells. During mitosis, histones are displaced from the DNA wrapped around them during passage of the replication fork [52]. As of now, there is no known mechanism by which histone modifications can be faithfully copied from mother cell to daughter cells independently of the presence of TFs. Work in the fission yeast S. pombe demonstrated that H3K9me3 (critical for heterochromatin formation) can be inherited though more than 50 cell divisions only if the demethylase Epe1 was deleted [53,54]. However, in the native context, inheritance of H3K9me3 was shown to require the DNA binding of CREB family TFs indicating that this histone mark cannot be independently maintained [55]. Although studies investigating this question are lacking, the fact that NF-κB or STAT activity is lost upon stimulus removal suggests that their associated histone modifications would concomitantly be rapidly lost in dividing cells.

DNA methylation as a source of inheritable memory in dividing cells

Given that yeast model systems lack DNA methylation, and thus, do not enable investigations into the interplay between histone modifications and DNA methylation in mammalian cells [56], the possibility remains that DNA methylation serves as a mammalian-specific driver of inheritance of stimulus-induced epigenetic changes. We hypothesize that this is the case because (1) the data show that DNA methylation can be induced by the stimulation of extremely diverse mammalian cell types – both developing and mature-exposed to various live and inert compounds [22,28,29,57], (2) these modifications occur at regions of stimulus-induced TF binding, enabling specificity of memory [29], (3) there is a known mechanism of inheritance by Dnmt1 [58], and (4) various histone modifying complexes are known to interact with 5mC [56], leading to the possibility that DNA methylation may serve as a mitotically inheritable epigenetic mark that drives the secondary deposition of histone modifications following each mitotic cycle. In support of this hypothesis, hippocampal progenitor cells stimulated with DEX retained 24.4% of differential methylated sites after 20 days [57] and in a study comparing differences in CpG methylation across adult mouse tissues it was discovered that while the majority of inter-tissue differences occurred at active or poised enhancers, some sites of tissue-specific hypomethylation corresponded to sites that were once active in the respective embryonic tissue but was no longer active, as indicated by a lack of both H3K4me1 and H3K27Ac [59].

Conclusions

Epigenetics plays a central role in specifying cell identity. Post-differentiation, innate immune cells face the conflicting requirement to both rapidly modulate their gene expression profiles in response to stimuli, but also to remember the original state to which they must return to after removal of the stimulus. We now know that immune cell activation involves epigenetic reprogramming occurring via many different mechanisms that likely include changes in TF binding, histone acetylation and methylation, as well as changes in DNA methylation in a coordinated fashion; however, we lack a fundamental understanding of which of these alterations is required for engagement of the gene expression program, and which are simply by-products of the stimulation-event itself, and may rather serve a role in memory. Several studies in dividing cell populations have indicated that some stimulus-induced changes in the chromatin landscape may be long-lasting and mitotically inheritable. No mechanistic evidence exists to show that histone modifications can be inherited through cell division. We speculate, therefore, that inheritance of epigenetic states may be driven by DNA methylation, which can be induced by various immune stimuli, inherited through mitosis, and is known to interact with histone modifying enzymes. Individually targeted deletions and additions of histone modifications and DNA methylation, as well as in vivo knockout models are needed to fully elucidate the individual contributions of these epigenetic modifications to epigenetic memory of stimulation, and the functional consequences of this memory.

Highlights.

• Stimulation of innate immune cells induces multifaceted epigenetic changes.

• Epigenetic changes induced by primary stimuli alter the functional behavior of immune cells and their response to a secondary stimulus

• DNA methylation may serve as a mammalian-specific driver of inheritance of stimulus-induced epigenetic changes