Nucleosomal asymmetry: a novel mechanism to regulate nucleosome function

Nucleosomes are the basic building blocks of chromatin. They are comprised of 147 bp of DNA wrapped around a histone octamer that is formed of two copies each of the four core histones H2A, H2B, H3, and H4 [1,2]. By regulating access to genomic DNA, nucleosomes are at the center of virtually all chromatin-templated processes, controlling transcription, replication, and DNA repair. Histones within nucleosomes are subject to a wide range of site-specific posttranslational modifications, including methylation, acetylation, phosphorylation, and ubiquitination [3,4]. Moreover, the canonical core histones can be replaced with histone variants, providing additional means to regulate nucleosome structure and function [5,6]. Together, histone marks and histone variants set up chromatin states that reflect and direct events such as transcriptional activation and repression by directly altering chromatin structure or by recruiting specific binding or ‘reader’ proteins [3,7,8].

Since David Allis and colleagues coined the ‘histone code’ hypothesis in the early 2000s [9,10], it has become increasingly clear that histone marks commonly function in a combinatorial fashion. This is supported by genome-wide studies that have revealed co-occurrence of different histone marks and variants at specific genomic regions such as active promoters and poised enhancers [8,11,12]. At the same time, biochemical and structural studies have provided insight into the mechanisms that underlie synergistic and antagonistic relationships between different histone marks [13–15]. Many proteins that interact with chromatin comprise multiple binding domains that allow them to interact not only with combinations of different histone marks, but also with other chromatin features such as DNA or the nucleosome acidic patch. Such multivalent engagement governs the affinity, dynamics, and specificity of interactions in chromatin, rendering multivalency a key regulatory principle in chromatin biology [16–19].

As nucleosomes contain two copies each of the four core histones, each sister histone could theoretically carry different marks, allowing them to act independently rather than redundantly in mediating multivalent interactions with combinatorial histone codes. However, given that sister histones within nucleosomes are related through 2-fold rotational pseudo symmetry, they have for a long time been considered to be identical and to therefore carry the same modifications. However, when this assumption was eventually put to a test, mass spectrometry and single molecule imaging-based approaches revealed that nucleosomes in mouse and human cells can in fact be asymmetrically modified, carrying different marks on equivalent sister histones [20,21]. Rather than being a rare event, nucleosomal asymmetry was found to be pervasive, with about half of all nucleosomes modified with H3K27me2/3 or H4K20me1 carrying the mark on only one of the two H3 or H4 sister histones, respectively [21]. Moreover, studies based on high-resolution chromatin immunoprecipitation coupled to exonuclease treatment (ChIP-exo) showed that nucleosomes at the transcription start sites (TSS) of about half of all genes in budding yeast similarly display asymmetry [22]. In addition to uncovering the existence of asymmetric nucleosomes, these pioneering studies also provided first insights into the mechanisms that give rise to asymmetry and identified transcription and the establishment and function of bivalent chromatin as key cellular processes that are regulated by nucleosomal asymmetry. Since then, a growing number of studies have corroborated the notion that nucleosomal asymmetry represents a new layer of regulation in the establishment, readout, and potentially inheritance of combinatorial histone codes and chromatin states, opening an exciting new area of chromatin biology.

Here, we will review recent progress and current challenges in elucidating the origins and functional consequences of nucleosomal asymmetry. We will summarize our current understanding of how asymmetry arises, how it affects the function of chromatin remodelers and histone mark readers, and how it regulates cellular processes such as transcription. We will also highlight some of the methods developed to study nucleosomal asymmetry in vitro and in vivo.

Nucleosomal asymmetry manifests in several ways (Figure 1). Besides sister histones exhibiting distinct modifications, histone isoforms can be incorporated into nucleosomes asymmetrically, e.g. by pairing canonical histone H2A with the histone variant H2A.Z in a heterotypic nucleosome [22–24]. Nucleosomes may also lack an H2A-H2B dimer altogether, forming so-called hexasomes [22]. Mutations in histone genes have been found in various cancers [25–29]. As these mutations only affect one of multiple copies encoding the same core histone, the abundance of these ‘oncohistones’ is low compared with their wild-type counterparts, resulting in predominantly asymmetric incorporation into nucleosomes (see e.g. [30,31]). The nature of these types of asymmetry suggest that histone modifying enzymes, chromatin remodelers, and histone chaperones are responsible for the formation of asymmetric nucleosomes. These enzymes may generate asymmetry de novo as a consequence of their intrinsic catalytic properties, or they may recognize and propagate asymmetry that already exists within a nucleosome or its environment. Such pre-existing asymmetry may involve other histone marks and variants as well as the underlying DNA sequence. As we will discuss below, both modes of regulation appear to be at play in vivo.

In the absence of any pre-existing asymmetry, nucleosomal asymmetry of histone marks may arise stochastically due to incomplete substrate turnover by histone modifying enzymes. In the case of histone marks, modifiers may be prevented from catalyzing symmetric modifications by limited enzyme availability, poor catalytic activity, or weak interaction with nucleosome substrates, resulting in residence times that are too short to ensure full substrate conversion. In support of such a scenario, mass spectrometry-based analysis of nucleosomes that were affinity-purified using antibodies against specific histone marks revealed that substantial amounts of nucleosomes are asymmetrically modified with respect to the marks tested [21]. Similarly, calibrated ChIP-seq studies with internal standards that allow to convert ChIP enrichments to biologically meaningful, normalized modification densities showed that histone marks rarely reach maximal modification density across the genome outside of specific loci [32]. Poor substrate turnover may therefore represent a simple, general mechanism that gives rise to nucleosomal asymmetry genome-wide. However, it remains to be determined exactly how these biochemical properties interface with other dynamic processes that take place in the complex environment of native chromatin, including nucleosome turnover, competition between different modifiers, and the action of eraser enzymes such as demethylases.

Several histone modifying enzymes exhibit loose substrate specificity and could thus establish asymmetry by stochastically modifying different residues on sister histones within a nucleosome. This has been demonstrated biochemically for checkpoint kinase 1 (CHK1) and protein kinase C (PKC) family enzymes [33]. CHK1 phosphorylates both H3S10 and H3T11 with similar efficiency, albeit in a mutually exclusive fashion, leading to generation of asymmetric nucleosomes. Similarly, PKCα and PKCβ phosphorylate either H3T6 or H3S10 in a mutually exclusive fashion and therefore also give rise to asymmetry in vitro [33]. Beyond kinases, many histone acetyltransferases (HATs) are also known to have broad substrate specificity, targeting several lysine residues within the same histone. For instance, p300 and its homolog CBP both target H3K18 and H3K27, among other substrates [34]. Mutually exclusive acetylation of residues on histone H3 would give rise to asymmetric acetylation. Even though histone methyltransferases usually only target a single specific histone residue, incomplete turnover may also lead to asymmetry through combination of different mono-, di- or trimethylation states across sister histones. Taken together, the de novo establishment of nucleosomal asymmetry may be a consequence of the catalytic properties of histone modifying enzymes. However, it remains to be determined how strongly these processes contribute to setting up asymmetry in vivo in the context of the complex chromatin environment within the cell nucleus, where site-specific recruitment of factors will lead to varying local abundance and residence times of specific enzymes. Moreover, such an environment will involve competition between different histone modifiers, leading to asymmetry via cross-talk between different marks, enzymes, and readers as discussed below.

In the case of heterotypic nucleosomes, similar mechanisms based on incomplete turnover may be responsible for the asymmetric incorporation of histone variants such as H2A.Z. The yeast SWR1C chromatin remodeling complex and its mammalian homolog SRCAP are responsible for replacing H2A-H2B dimers with H2A.Z-H2B dimers in nucleosomes [35]. Interestingly, the SWR1C complex catalyzes dimer exchange in a sequential, stepwise fashion, first exchanging one and then the other dimer. Kinetic analyses revealed that the second exchange reaction is markedly slower [36], thus promoting incomplete, asymmetric dimer exchange. The exchange reaction is further modulated by the DNA sequence surrounding the nucleosome substrate. At the TSS, SWR tends to bind on the side of the +1 nucleosome that borders the nucleosome free region (NFR), due to its preferential binding around long DNA linkers [37]. Moreover, runs of GC base pairs stimulate SWR activity and impart directionality onto dimer exchange, as H2A.Z-H2B incorporation is more efficient when the outgoing H2A-H2B dimer contacts DNA with higher GC content [38]. These findings suggest that nucleosomal DNA sequence plays an important role in directing H2A.Z to the distal side of the +1 nucleosome in an asymmetric fashion, as is observed in the yeast genome [22] (see also below). Indeed, the proximal and distal halves of nucleosomal DNA exhibit differences in base composition that render those nucleosomes asymmetric [39]. In addition to SWR, other chromatin remodelers such as Chd1 [40] and INO80 [41] exhibit sequences preferences in their binding and activity, giving directionality to nucleosome repositioning and downstream processes such as transcription. Moreover, transcription is an inherently asymmetric process, proceeding through a gene from its TSS to the transcription termination site in a directional fashion. While DNA sequence properties promote directionality of transcription, transcription itself likely further reinforces asymmetry and thus plays a key role in the generation of heterotypic nucleosomes and asymmetrically modified nucleosomes around the TSS and over the gene body.

Beyond chromatin remodelers, inherent properties of DNA sequences might also regulate the activity of histone modifying enzymes. Differences in DNA sequence and base composition have been shown to differentially regulate DNA shape and rigidity as well as the strength of histone-DNA contacts across the two nucleosome halves and their adjacent linkers [41–44], providing an asymmetric environment for modifiers to act in. However, the mechanisms and physiological consequences of such DNA sequence and structure-based regulation of histone mark asymmetry remain largely unexplored so far.

Interestingly, many histone modifying enzymes exhibit properties that promote symmetric modification. Methyltransferase and acetyltransferase complexes often contain binding domains that recognize the mark placed by the enzyme, establishing autoregulatory positive feedback loops that promote recruitment and/or catalytic activity of the enzyme. These mechanisms may also operate in an intra-nucleosomal, trans-tail fashion. The SAGA HAT complex primarily acetylates histone H3 at lysine 14 but also targets lysines 9, 18, and 23 [45,46]. The complex comprises a bromodomain that interacts with acetylated lysines, including the catalytic products of the SAGA complex itself. Interaction of the bromodomain with one acetylated histone H3 tail within asymmetric nucleosomes stimulates activity of the yeast SAGA complex towards the unmodified sister histone tail [47,48], promoting symmetric modification. As another example, Polycomb repressive complex 2 (PRC2) catalyzes the trimethylation of histone H3 lysine 27 (H3K27me3), a mark associated with promoters and gene bodies of repressed genes [49–51]. The EED subunit contains a WD40 domain that binds H3K27me3, thereby stimulating PRC2 methylation activity [52]. This stimulation also occurs in trans within asymmetric nucleosomes carrying one H3K27me3 mark [21,53,54], thereby boosting symmetric modification (Figure 2).

In the case of the yeast Set1 complex, such positive feedback via complex-intrinsic reader domains appears to work in concert with asymmetry-specific demethylation to ensure symmetric modification in chromatin. Set1 catalyzes H3K4me3, a hallmark of actively transcribed promoters in yeast and higher eukaryotes [55,56]. Its Spp1 subunit (CFP1 in mammals) binds to H3K4me3 via a PHD finger [57,58], which may result in cooperative methylation of sister histones as observed for PRC2. As an additional mechanism promoting symmetric modification, the Set1 complex forms a dimer in yeast, mediated by its Sdc1 subunit (DPY30 in mammals) [59]. Mutations in Sdc1 that disrupt dimerization virtually abolish H3K4me3 and strongly reduce H3K4me2 levels, both products of Set1 [59]. At the same time, the yeast H3K4me2/3 demethylase Jhd2 appears to specifically remove asymmetric H3K4me3 [59]. Mechanistically, this asymmetry-specific demethylation is mediated by a PHD finger in the demethylase that binds unmodified H3K4 and promotes interaction with nucleosomes containing asymmetric H3K4me3 [60–63]. Together, these findings suggest that methylation and demethylation activities act in concert to focus symmetric H3K4me3 on promoter nucleosomes, while removing spurious asymmetric H3K4me3 present elsewhere. Given the high degree of conservation of the enzymes involved, similar mechanisms may be at play in mammalian systems. Interestingly, PRC2 has also been shown to dimerize [64–66], suggesting that enzyme dimerization-based mechanisms may also be at play in other modification systems.

Many histone modifying enzymes have been shown to be stimulated or inhibited by other histone marks, establishing cross-talk between different modification systems [15]. In contrast with autoregulatory feedback loops that promote symmetry, cross-talk between different histone marks may be key to establishing and reinforcing nucleosomal asymmetry. While H3K27me3 is associated with promoters and gene bodies of repressed genes, H3K4me3 and H3K36me3 mark promoters and gene bodies, respectively, of actively expressed genes [8,11]. PRC2 has been shown to be inhibited by H3K4me3 and H3K36me3 [67], explaining how repressive H3K27me3 is excluded from actively transcribed genes (Figure 2). However, together with the failure to detect histone H3 tails carrying both H3K4me3 and H3K27me3 in vivo, these findings challenged the existence of promoters marked with both H3K4me3 and H3K27me3. These so-called bivalent domains are thought to regulate expression of developmentally regulated genes in undifferentiated cells such as embryonic stem cells (ESCs), keeping them repressed in ESCs but poised for activation upon appropriate differentiation signals [68–72]. Resolving this conundrum, bivalent nucleosomes in ESC were found to be in an asymmetric configuration, carrying H3K4me3 and H3K27me3 on separate — rather than the same — histone H3 tails [20,21]. Moreover, using recombinant symmetrically and asymmetrically modified nucleosomes, inhibition of PRC2 by H3K4me3 and H3K36me3 was found to only occur within the same histone tail in cis [21,53,54]. When H3K4me3 or H3K36me3 were present asymmetrically on only one H3 sister histone within a nucleosome, PRC2 activity towards the unmodified histone H3 copy was unaffected, giving rise to an asymmetric bivalent nucleosome (Figure 2). These findings provided not only conclusive evidence for the existence of bivalent nucleosomes but also a mechanism for their generation.

Histone mark cross-talk has been found to generate asymmetry in other modification systems as well, such as H3K4 methylation-based inhibition of H3T3 phosphorylation by Haspin kinase [73]. Together, these findings highlight how histone mark cross-talk can establish asymmetric nucleosome states, thereby expanding the combinatorial space available for complex sets of histone marks at individual nucleosomes within chromatin. Cross-talk between mutually exclusive marks is a common feature of many histone modification systems and may therefore lead to asymmetry for many marks beyond the examples known so far.

The ability to generate recombinant nucleosomes that carry defined modifications in an asymmetric fashion was essential to uncovering the intra- but not inter-histone tail inhibition of PRC2 by active histone marks [21]. Initial protocols utilized affinity tags to generate asymmetric nucleosomes for biochemical studies. Each differentially modified sister histone was fused to an orthogonal affinity tag, allowing for sequential affinity purification to separate asymmetric histone octamers from symmetric byproducts of octamer reconstitution (Figure 3) [21]. These early approaches were laborious and characterized by a low yield of the desired asymmetric nucleosomes. Since these early studies, several groups have developed elegant novel approaches to generate asymmetrically modified nucleosomes (reviewed in-depth in [74]). As the first step, these approaches rely on native chemical ligation and other techniques to generate histones carrying specific modifications [75–77], which are then assembled into asymmetrically modified nucleosomes. These developments allow generating a plethora of asymmetric nucleosomes with ease, greatly facilitating determining the mechanisms that control the origins and functional consequences of nucleosomal asymmetry.

The Fierz group developed an approach based on joining the H3 sister histones in the desired asymmetric combination via a cleavable linker prior to octamer assembly, allowing for efficient generation of asymmetric octamers and subsequent traceless removal of the linker using TEV protease [53,54]. In that way, these approaches also eliminate any potential negative impact arising from the presence of affinity tags during downstream applications. This approach confirmed that PRC2 is inhibited by H3K4m3 and H3K36me3 only in cis, whereas H3K27me3 can stimulate PRC2 activity in trans in asymmetric nucleosomes [53,54]. The Muir group developed conceptually similar approaches for traceless generation of asymmetric histone octamers, most recently by employing the SpyCatcher/SpyTag protein heterodimerization system [63]. Using this spytethering-based approach, they provided direct biochemical evidence for the preferential demethylation of asymmetric H3K4me3 nucleosomes by the H3K3me3 demethylase KDM5B [63], as suggested for its Jhd2 homologue in yeast. Although other techniques are already available to generate nucleosomes with asymmetrically modified H2A or H2B (see next section), the SpyCatcher/SpyTag approach should be applicable to all four core histones.

Other approaches developed to generate asymmetric nucleosomes target the dimerization interface of the two H3 sister histones within the histone octamer. Using protein design followed by directed evolution in yeast, the Kaufman and Rando labs developed a pair of mutant H3 histones termed H3X and H3Y that efficiently heterodimerize but do not homodimerize, due to bump-hole mutations in the H3-H3′ interface that generate complementary binding surfaces [78]. An analogous approach was employed by the Zhou lab to generate a pair of H3 histones termed H3^D and H3^H with similar bump-hole mutations [79]. While these approaches are limited to histone H3, they eliminate the need for any additional processing steps after histone octamer assembly and yield almost exclusively asymmetric octamers. Crucially, these mutations can also be used to interrogate the consequences of nucleosomal asymmetry in vivo. In yeast, H3K36me3 is thought to silence transcription from cryptic internal promoters by recruiting the RPD3S histone deacetylase complex [80–82]. By blocking H3K36 methylation via introduction of a H3K36Q mutation in asymmetric fashion, a single H3K36 methylation site per nucleosome was shown to be sufficient to repress cryptic transcription in yeast [78,79]. Although application to mammalian systems is complicated by the presence of multiple genes encoding histone H3, these heterodimerization approaches should prove useful in studying the consequences of asymmetry in vivo.

Several methods have also been developed to generate hexasomes and heterotypic nucleosomes containing different H2A/H2B dimers [83,84]. These methods exploit asymmetry inherent in the 601 nucleosome positioning sequence, a synthetic DNA sequence with high affinity for histones that is routinely used for assembling recombinant nucleosomes in vitro [85]. First, a single H2A/H2B dimer can be assembled onto the 601 sequence in an oriented fashion to yield hexasomes. Subsequently, a different species of H2A/H2B dimers can be added to reconstitute asymmetric nucleosomes. The Bowman lab used this approach to study the mechanisms by which the chromatin remodeller Chd1 repositions nucleosomes, showing that it can utilize hexasomes as substrates and slide them in a unidirectional fashion [83]. The Narlikar lab used this approach to show that the yeast chromatin remodelling complex INO80 preferentially slides hexasomes in regulating nucleosome spacing [86]. The Muir lab developed these methods further, most recently by employing a truncated 601 sequence for the initial hexasome assembly, eliminating formation of nucleosome by-products. The missing part of the 601 sequence can then be ligated onto the truncated 601 DNA before a different type of dimer is added to form asymmetric nucleosomes [84]. This method can be combined with the spytethering approach described above [63] to assemble nucleosomes carrying asymmetric modifications on two different types of histones such as H3 and H2B.

Taken together, several approaches are now available for the generation of various types of asymmetric nucleosomes in vitro. These powerful tools will undoubtedly continue to facilitate the biochemical characterization of the mechanisms giving rise to asymmetry and their impact on the readout of histone marks, variants, and nucleosomal states. Despite these advances, however, it has remained challenging to study asymmetry in cells and organisms. While heterodimerization-based approaches such as the histone H3X/H3Y system allow enforcing asymmetry of specific marks, they rely on genetic mutations that prevent or facilitate a specific modification, thereby introducing non-native nucleosomes into chromatin. Moreover, these approaches currently lack the ability to modulate asymmetry in a locus-specific fashion. Controlling asymmetry by harnessing the potential of DNA sequence and base composition to differentially regulate histone modifier activity towards the two nucleosome halves represents an intriguing possibility. However, we currently lack sufficient mechanistic understanding to do so in a controlled and directed fashion, and changing DNA sequence would likely have additional unintended consequences beyond changing nucleosomal asymmetry. The development of methods to accurately control asymmetry in cells and to determine the genome-wide distribution of asymmetrically modified nucleosomes would greatly facilitate exploring the physiological roles of nucleosomal asymmetry.

Once established, nucleosomal asymmetry may regulate the interaction between chromatin binding proteins and nucleosomes, shaping chromatin structure and controlling chromatin-templated processes such as transcription. The symmetry state of a histone mark or variant changes its abundance and spatial arrangement at a given stretch of chromatin. By increasing the local concentration of available binding sites symmetric modification could enhance interaction with cognate reader domains. In addition, asymmetry controls which combinations of histone marks can be incorporated into a single nucleosome, affecting availability and relative orientation of different binding sites. In this way, asymmetry regulates multivalent interactions between nucleosomes and chromatin complexes that contain multiple reader domains for different histone marks, controlling affinity and specificity of their recruitment. In the case of dimeric reader proteins, symmetric modification may allow for simultaneous, multivalent engagement of the same mark on a single nucleosome, if such an interaction is sterically feasible. Crucially, symmetric modification is also required for dimeric readers to efficiently bridge neighboring nucleosomes. For instance, such a binding mode is thought to be key to the function of HP1 in mediating chromatin compaction and heterochromatin spreading [87,88].

Several studies have begun to explore how these consequences of asymmetry affect interactions in chromatin and regulate chromatin function. Thus far, these studies have primarily focused on transcription, bivalent chromatin, and oncohistone mutations in cancer.

By employing a high-resolution ChIP approach, the Pugh lab observed the presence of asymmetric nucleosomes around the TSS of around half of all genes in budding yeast [22]. This asymmetry manifests as increased levels of H3K9ac and H2BK123ub on the side of the +1 nucleosome that faces the NFR [22]. In contrast, the histone variant H2A.Z is enriched on the distal side [22], in line with the role of the NFR and adjacent nucleosomal DNA sequence in regulating SWR-mediated dimer exchange (see above). Moreover, histones were asymmetrically lost from the +1 nucleosome, leading to the formation of hexasomes and even half-nucleosomes [22]. This asymmetric nucleosome structure may promote transcription by facilitating passage of RNA polymerase II through the +1 nucleosome and its progression into gene bodies. In addition, the presence of hexasomes likely affects the activity of specific chromatin remodelers as highlighted above [83,86]. In mouse trophoblast stem cells, TSS of actively transcribed genes were also found to be marked by heterotypic H2A–H2A.Z nucleosomes [23], suggesting that the presence of such asymmetric nucleosomes may similarly facilitate transcription in mammals. RNA polymerase II travelling through the gene body in a directional fashion during elongation may impose further asymmetry beyond the TSS, resulting in asymmetric nucleosomes interacting with the transcription, chaperone, and chromatin remodeling machinery during transcription.

Promoters of active genes in mammals are marked by H3K4me3, which is thought to facilitate transcription by recruiting the basal transcription factor TFIID and other effectors such as HAT complexes and chromatin remodelers [56,58]. Using a calibrated ChIP-seq approach based on spike-in of internal standards that allows to obtain normalized histone modification densities that can be directly compared between different loci, the Ruthenburg lab observed that transcriptional output correlated with H3K4me3 levels at promoters in mouse ESCs [32]. Strikingly, in the resulting sigmoidal curve, mRNA abundance responded most strongly to increases in promoter H3K4me3 levels where they transition from asymmetric to those corresponding to symmetric nucleosomes. These findings suggest that the symmetry state of H3K4me3 may regulate recruitment of TFIID and other H3K4me3 readers by altering the availability and valence of binding sites for these readers and thus the avidity of their interaction with promoter nucleosomes. Supporting this notion, recent work from our lab showed that symmetric H3K4me3 modification allows efficient recruitment of TFIID and other H3K4me3 readers in nucleosome pulldown experiments, whereas asymmetric H3K4me3 failed to do so [89]. Together, these findings suggest that nucleosomal asymmetry of H3K4me3 and potentially other marks may regulate transcription by controlling the recruitment of histone mark readers.

A direct consequence of nucleosomal asymmetry is the ability to generate patterns of combinatorial histone marks that would otherwise be prohibited. As a prominent example, the so-called bivalent domains adopt an asymmetric configuration, carrying H3K4me3 and H3K27me3 on separate H3 sister histones, thereby accommodating both marks within the same nucleosome. Mass spectrometry- and single molecule imaging-based studies conclusively showed that asymmetrically modified bivalent nucleosomes exist while symmetrically modified bivalent nucleosomes are essentially absent from mouse ESCs [20,21]. Although they cannot distinguish between asymmetric and symmetric nucleosomes, several re-ChIP-seq studies have confirmed the simultaneous presence of H3K4me3 and H3K27me3 at nucleosomes of bivalent promoters genome-wide [90–93]. Based on internal standard-calibrated ChIP-seq, promoters of developmentally regulated genes in mouse ESCs have a combined H3K4me3 and H3K27me3 modification density of ∼100% [32], consistent with asymmetric bivalent nucleosomes being the prevalent modification state at these promoters. In addition, engineered bivalent readers consisting of the H3K4me3-binding TAF3 PHD domain fused to the CBX7 or Drosophila Polycomb chromodomain, which bind H3K27me3, showed strong enrichment at bivalent promoters [94]. In contrast, engineered monovalent PHD-PHD and chromodomain-chromodomain readers exhibited poor enrichment at bivalent loci but were efficiently recruited to their cognate monovalent loci [94], arguing against presence of monovalent symmetric nucleosomes at bivalent promoters. Taken together, these studies indicate that bivalent promoters predominantly feature asymmetrically modified nucleosomes.

Once established, the asymmetric state of bivalent nucleosomes may regulate their function in poising developmental genes for expression. As a first layer, the inhibition of PRC2 by H3K4me3 on the same histone tail and the attenuation of SET1 and MLL1 activity by H3K27me3 provides a mechanism that prevents conversion of these promoters to a fully active or repressed state [67,95]. Upon differentiation, recruitment of transcription factors could skew the balance between activation and repression maintained at bivalent promoters, overriding this interplay to trigger activation. In addition, the asymmetric state may regulate the recruitment of effector proteins. As bivalency effectively enforces an asymmetric state for H3K4me3, recruitment of TFIID and other transcriptional activators may be diminished as discussed above, hindering activation of bivalent promoters. Moreover, asymmetric bivalent nucleosomes may recruit multivalent readers capable of interacting with both marks. Indeed, in Arabidopsis, the proteins EBS and SHL each contain both a PHD finger and a bromo adjacent homology (BAH) domain, which interact with H3K4me3 and H3K27me3, respectively [96–98]. While each protein can only bind either H3K4me3 via the PHD finger or H3K27me3 via the BAH domain but not both marks simultaneously, EBS and SHL homo- and heterodimerize to form bivalent reader complexes [98]. These complexes bind the bivalent state at the cold memory element region of the flowering control locus, driving cold-induced switching to a Polycomb repressed chromatin state [98]. Our recent work shows that factors capable of multivalent readout of the asymmetric bivalent state also exist in mammals [89]. Nucleosome binding of the HAT complex KAT6B is enhanced by the asymmetric bivalent modification, promoting recruitment of KAT6B to bivalent promoters in ESCs, where the complex is required for proper expression of developmental genes during differentiation [89]. Together, these studies suggest that asymmetry at bivalent promoters not only allows the bivalent methylation marks to coexist but also actively regulates their readout and function.

The discovery that several types of cancers are associated with specific somatic mutations in genes encoding core histones has spurred intense investigation into the mechanisms by which these mutations give rise to cancer phenotypes. Since the initial identification of the H3K27M and H3G34R/V mutations in pediatric gliomas, whole genome sequencing has revealed a plethora of histone mutations in a diverse range of cancers [25–29]. Given that these mutations only affect a single copy of several genes encoding the same histone, the resulting mutant histones only represent a small fraction of the total cellular histone pool. Oncohistones are therefore predominantly incorporated into heterotypic nucleosomes in chromatin, as has been experimentally verified in some cases (see e.g. [30,31]). The resulting asymmetry may contribute to the mechanisms by which oncohistones disrupt chromatin regulation. So far, the role of oncohistone asymmetry has mainly been studied for mutations affecting the nucleosome acidic patch. The acidic patch is a binding pocket formed by acidic residues of histones H2A and H2B, which is engaged by many chromatin factors in their multivalent interaction with nucleosomes [99], including many chromatin remodelers. In line with the importance of acidic patch-mediated interactions, cancer mutations in histones H2A and H2B have been shown to affect the activity of the ISWI family remodeler Snf2h in vitro [100,101]. Interestingly, when these mutations were incorporated into nucleosomes asymmetrically, they caused the remodeler to slide nucleosomes in a unidirectional rather than bidirectional fashion, leading to aberrant nucleosome spacing [101]. Directionality was similarly imposed on the activity of the cBAF chromatin remodeling complex [84]. These findings suggest that this group of oncohistones may modulate nucleosome positioning and chromatin accessibility, dysregulating transcription and gene expression profiles in cancer. Moreover, these studies underscore how nucleosomal asymmetry is an important regulator of chromatin states and gene expression.

Several cancers also exhibit overexpression of the histone H3 variant CENP-A [102]. CENP-A (also known as CenH3) defines the location of centromeres in most eukaryotes and serves as the platform for the assembly of kinetochores, which attach chromosomes to spindle microtubules during cell division to ensure accurate chromosome segregation [103]. While several lines of evidence suggest that CENP-A normally forms symmetric, homotypic nucleosomes at centromeres [104], overexpression of CENP-A in cancers can lead to its incorporation outside of centromeres in the form of heterotypic CENP-A–H3.3 nucleosomes [105]. In HeLa cells, such ectopic deposition has been shown to occur at active promoters and enhancers and CTCF binding sites [105]. Even though gene expression remained largely unchanged in this overexpression system, heterotypic CENP-A nucleosomes might deregulate gene expression in cancer by altering transcription factor binding and histone modification states. Moreover, heterotypic CENP-A nucleosomes at ectopic sites may also destabilise native centromeres by acting as a sink for centromere and kinetochore proteins, leading to chromosome segregation defects and aneuploidy [106].

Since the discovery that nucleosomes can be asymmetrically modified, this novel facet of epigenetics and chromatin biology has attracted increasing attention. It is now well established that nucleosomal asymmetry is a hallmark of bivalent chromatin, and several lines of evidence suggest that asymmetry is a key regulator of transcription. So far, the roles of asymmetry have only been explored for a small subset of histone modification systems. Future studies will likely uncover additional regulatory systems and processes that rely on nucleosomal asymmetry. The development of novel methods to generate asymmetric nucleosomes has been pivotal to our ability to determine the mechanisms by which asymmetry is established, sets up complex histone modification patterns, and regulates multivalent interactions with chromatin reader complexes. Going forward, tools to map and quantify asymmetry in chromatin with nucleosome resolution as well as approaches to modulate and perturb asymmetry in a locus-specific and dynamic fashion in cells will be invaluable to gain a complete understanding of the physiological roles of nucleosomal asymmetry in regulating chromatin structure and function.

• Despite their inherent symmetry, nucleosomes can carry different modifications on sister histones or contain different variants of the same core histone. The resulting nucleosomal asymmetry represents a novel layer of regulation in the establishment, readout, and function of chromatin states.

• Asymmetry expands the combinatorial space available to set up complex histone mark codes and regulates multivalent interactions between nucleosomes and chromatin complexes, thereby regulating processes such as transcription and the poising of developmental gene expression by bivalent chromatin.

• Furthering insight into the physiological role of nucleosomal asymmetry and the mechanisms by which it regulates chromatin structure and function requires the development of novel methods to generate asymmetric nucleosomes in vitro, map their localization and abundance in cells, and to alter asymmetry in a locus-specific fashion to examine how asymmetry regulates transcription and other processes.

The authors declare that there are no competing interests associated with this manuscript.

This work was supported by the UK Biotechnology and Biological Sciences Research Council [BBS/E/B/000C0421]. D.V. was supported by a PhD studentship from the Darwin Trust of Edinburgh.

P.V. conceived and wrote the manuscript. P.V. and D.V. prepared figures and edited the manuscript.

We are grateful to members of the Voigt lab for helpful discussions. We thank Peter Rugg-Gunn and Sarah Elderkin for critical reading of the manuscript.

BAH

bromo adjacent homology

ChIP

chromatin immunoprecipitation

CHK1

checkpoint kinase 1

ESC

embryonic stem cells

HAT

histone acetyltransferase

NFR

nucleosome free region

PKC

protein kinase C

PRC2

Polycomb repressive complex 2

TSS

transcriptional start site