A Genetically Encoded Approach for Breaking Chromatin Symmetry.

Nucleosomes frequently exist as asymmetric species in native chromatin contexts. Current methods for the traceless generation of these heterotypic chromatin substrates are inefficient and/or difficult to implement. Here, we report an application of the SpyCatcher/SpyTag system as a convenient route to assemble desymmetrized nucleoprotein complexes. This genetically encoded covalent tethering system serves as an internal chaperone, maintained through the assembly process, affording traceless asymmetric nucleosomes following proteolytic removal of the tethers. The strategy allows for generation of nucleosomes containing asymmetric modifications on single or multiple histones, thereby providing facile access to a range of substrates. Herein, we use such constructs to interrogate how nucleosome desymmetrization caused by the incorporation of cancer-associated histone mutations alters chromatin remodeling processes. We also establish that our system provides access to asymmetric dinucleosomes, which allowed us to query the geometric/symmetry constraints of the unmodified histone H3 tail in stimulating the activity of the histone lysine demethylase, KDM5B. By providing a streamlined approach to generate these sophisticated substrates, our method expands the chemical biology toolbox available for interrogating the consequences of asymmetry on chromatin structure and function.

Introduction

Nucleosomes are nucleoprotein complexes that spool the eukaryotic genome into a dynamic polymer known as chromatin.[1] These repeating structural units have been extensively characterized as pseudosymmetrical bioparticles composed of about 147 base pairs (bp) of double-stranded DNA wrapped around a hetero-octameric protein complex consisting of two copies each of histones H2A, H2B, H3, and H4.[2] Although nucleosomes block access to the associated genomic DNA, they also can guide DNA-templated processes by acting as docking platforms for chromatin-associated proteins through a combination of DNA– and protein–protein interactions, often mediated by post-translational modifications (PTMs) of the histone proteins themselves.[3−5] Dysregulation of such fundamental processes, via either mutation of histones or chromatin-associated proteins, has been implicated in many diseases.[6−12]

The nucleosome core particle possesses pseudo-2-fold symmetry, with each half containing single copies of the four canonical histones (Figure A). However, this apparent symmetry is broken in numerous circumstances, leading to important biological outcomes (Figure B). Each of a pair of degenerate histones can be differentially post-translationally modified.[13] Such heterotypic PTM patterns occur in unique “bivalent” genomic regions that contain both trimethylation of lysine 4 on histone H3 (H3K4me3) (an activating mark) and H3K27me3 (a repressive mark) on different H3 histones within the same nucleosome.[13−15] Histone variants have been shown to be deposited asymmetrically into chromatin. At active gene promoters, H2A.Z is deposited primarily on the distal half of the +1 nucleosome, whereas canonical H2A resides in the proximal half.[16] Furthermore, cancer-associated histone mutations, so-called “oncohistones”, have been shown to be stochastically incorporated into chromatin,[10,17,18] giving rise to heterotypic nucleosomes with distinct biochemical and biophysical properties.[6,8,9]

Figure 1

Structure and symmetry of nucleosomes. (A) Crystal structure (PDB 1kx5) of a mononucleosome, showing both copies of each canonical histone: H2A (yellow), H2B (pink), H3 (cyan), and H4 (green). (B) Cartoon depiction of a nucleosome illustrating the pseudo-2-fold symmetry. Top: symmetrical (homotypic) nucleosomes contain identical H3 monomers (blue). Bottom: asymmetric (heterotypic) nucleosomes contain different copies of H3 (blue and red).

Although access to recombinant symmetrical nucleosome substrates is relatively straightforward and, indeed, is well-established at this point,[19,20] the preparation of chemically defined heterotypic chromatin substrates for biochemical studies remains nontrivial.[21] Asymmetric nucleosomes bearing two different copies of H2A or H2B have been obtained through stepwise incorporation of unique H2A/H2B dimers into nucleosomes, capitalizing on the ability to control the formation of subnucleosomal structures on certain DNA sequences.[9,22−24] Control of H3/H4 asymmetry is even more challenging, requiring the deposition of asymmetric H3/H4 tetramers during nucleosome reconstitution. Previous strategies have relied on stochastic mixtures of histones followed by consecutive affinity purifications, the use of compensatory “bump–hole” mutations, or chemical tethering of the histone into a heterodimer.[13,25−29] These methods are (bio)chemically onerous and often low-yielding and, in the case of the “bump–hole” approach, furnish non-native products.[28,29] Moreover, chemical tethering approaches rely on sophisticated synthetic chemistry manipulations that are not accessible to every laboratory.[25−27]

In this study, we describe a traceless approach to generate asymmetric nucleosomes through use of the SpyCatcher/SpyTag system as a genetically encoded bivalent linkage (Figure A). SpyCatcher and SpyTag are an engineered split protein pair that associate at low concentrations and form an isopeptide bond between the two domains.[30−32] We utilized this covalent linkage to tether histones bearing distinct modifications or mutations. From these, we assembled heterotypic nucleosomes to assess the consequences of asymmetry on nucleosome stability, chromatin remodeling, and demethylation of H3K4me3 by KDM5B. This system holds several advantages over prior methods, including ease of access, fewer steps, and high yields.

Figure 2

“Spytethering” method for the generation of asymmetric nucleosomes. (A) Schematic illustrating the general workflow for obtaining asymmetric nucleosomes using the SpyTag system. In phase 1, His6-SpyC-H3 and SpyT-H3 undergo isopeptide bond formation followed by refolding into tethered tetramers in the presence of H4. In phase 2, the tethered tetramers are combined with H2A/H2B dimers and 601 DNA to reconstitute mononucleosomes, followed by proteolytic removal of SpyTags. (B) ESI-TOF mass spectrometry analysis of His6-SpyC-H3. (C) ESI-TOF mass spectrometry analysis of SpyT-H3. (D) ESI-TOF mass spectrometry analysis of the WT spytethered H3 homodimer. (E) SDS-PAGE analysis of WT histone dimers, tetramers, and octamers compared to spytethered histone tetramers and octamers visualized by Coomassie staining. (F) Native PAGE analysis of WT mononucleosomes prepared by standard methods, WT/WT spytethered nucleosomes, and WT/WT TEV-cleaved nucleosomes visualized by ethidium bromide staining.

Results and Discussion

We began by exploring options for the chemoselective and traceless tethering of defined H3 monomers such that their association would be maintained throughout the nucleosome reconstitution process.[25−27] In particular, we wanted these tethers to be genetically encoded, biorthogonal, and irreversible. The so-called SpyCatcher/SpyTag method presented itself as an optimal method for efficiently generating obligate protein heterodimers, as it results in an autocatalyzed isopeptide bond between two fragments of a split protein domain, covalently fusing the two polypeptide chains.[30−33] A wide range of unnatural protein conjugates have been engineered using the SpyCatcher/SpyTag system, affording exotic protein topologies in high yield.[34−39] Recently, evolved versions of the SpyCatcher/SpyTag system have been developed that react with high efficiency, even under the mild to moderate denaturing conditions that are used routinely during the manipulation of histone monomers.[31,32] By combining distinct histone monomers together, each fused to either a SpyCatcher or SpyTag domain through the N-termini, we posited that the subsequent heterodimers would be stably tethered during histone octamer/tetramer preparation and nucleosome reconstitution, thus defining the nucleosome composition, and that these tethers could later be released by a protease-catalyzed cleavage event (Figure A).

To achieve this, we designed genetic fusions of histone H3 with either the His6-SpyCatcher002 (SpyC) or His5-SUMO-SpyTag002 (SpyT) sequences, separated by flexible linkers. A TEV protease cleavage site was inserted directly before the H3 protein sequence for traceless release from the SpyC and SpyT domains after nucleosome assembly. These fusion proteins were isolated from an Escherichia coli expression system (Figure B,C, Figure S1A,B) and then combined in neutral reaction buffer containing 1 M guanidinium chloride (Gdn.HCl), a chaotrope concentration previously shown to disrupt histone aggregation.[40] Using 1.2 equiv of SpyT-H3, an essentially complete reaction was observed after 3 h to form the “spytethered” SpyC/T-(H3)2 homodimer (Figure D, Figure S1C). The reaction also proceeded in the presence of 2 M Gdn.HCl, albeit more slowly (Figure S1C). Refolding of the purified H3 dimer in the presence of the complementary core histones H2A, H2B, and H4 under high-salt conditions afforded tethered histone octamers (Figure E, Figure S2A–C).[41]

We then asked if the tethered octamer assembly process could be further streamlined by conducting the reaction in situ during the histone refolding step, exploiting the tolerance of the spytethering reaction to denaturants (Figure S3A). Thus, we combined the SpyC-H3 and SpyT-H3 constructs along with histone H4 under denaturing conditions (6 M Gdn.HCl) and then dialyzed this mixture into 1 M Gdn.HCl followed by 2 M NaCl. This one-pot procedure afforded the desired tethered H3/H4 tetramers in excellent yield and high purity following size exclusion chromatography (SEC) (Figure E, Figure S3B,C). Notably, the SEC trace for tethered tetramers was nearly identical to that of tetramers prepared using traditional methods (Figure S3B). The tethered H3/H4 tetramers were then combined with H2A/H2B dimers and DNA and subjected to a standard nucleosome reconstitution protocol (Figure S4A).[41] As expected, the additional size of the SpyC/T domain resulted in a noticeable gel shift relative to wild-type (WT) untethered nucleosomes (Figure F).[42] Importantly, we did not detect any additional slower migrating species at this stage (Figure S4B), suggesting that the formation of undesired oligonucleosomes (containing domain swapped tetramers) is not an issue under the reconstitution conditions we employ. Treatment of the spytethered nucleosomes with TEV protease efficiently liberated the nucleosomes from the SpyC/T tethers (Figure S4B, Figure F). Capitalizing on the size difference between the nucleosome (120 kDa) and both SpyC/T (18 kDa) and TEV protease (30 kDa), purified nucleosomes could be isolated from these reaction mixtures via centrifugal filtration with high-molecular-weight cutoff membranes (Figure S4C,D). Importantly, these nucleosomes had near identical thermal stability, as indicated by a fluorescence-based assay, to those assembled using standard methods (Figure S4E,F).[43] Additionally, they also displayed indistinguishable chromatin remodeling activity, as indicated by a restriction enzyme accessibility assay (REA),[44] to those assembled via the standard assembly protocol (Figure S4G). Combined, these results argue that nucleosome integrity is not affected by the transient presence of the tether.

The basic protocol in place, we next generated a series of heterotypic nucleosomes designed to illustrate the different types of asymmetric species accessible through our strategy. We successfully prepared mononucleosomes in which the two copies of histone H3 were differentiated through (i) stable isotope incorporation, (ii) amino acid mutation, (iii) histone variant introduction, and (iv) post-translational modification (Figure A–D, Figures S5–S7). We also prepared a nucleosome in which two different histones were differentially modified with two different PTM chemotypes, specifically histone H2B carrying ubiquitin on lysine 120 (H2BK120ub) and H3K4me3 (Figure E–G). Notably, this necessitated the generation of a SpyT-H3K4me3 fusion protein through semisynthesis (Figure S8),[45] as well as the use of an assembly protocol that allowed the introduction of two different H2A/H2B dimers (one containing H2BK120ub) into the tethered asymmetric H3/H4 tetramers (Figure E).[9] To our knowledge, this is the first time that asymmetry has been introduced simultaneously into two different histone types (i.e., H2B and H3) within the same nucleosome.

Figure 3

“Spytethering” allows nucleosome symmetry to be broken in various ways. Panels show the asymmetric incorporation of uniformly 15N-labeled H3 (A), H3A96C and H3A110C mutations (B), and histone variant H3.3 (C), both before and after treatment with TEV protease. (D) H3K18ssUb Native-PAGE analysis including treatment with DTT to display a mobility shift following removal of ubiquitin. (E) Workflow for the generation of nucleosomes containing both H3 and H2B asymmetry. Spytethered WT/H3K4me3 tetramers were combined with 1.2 equiv of H2BK120Ub dimers to form a mixture of symmetric nucleosomes and oriented hexasomes. Hexasomes were formed into nucleosomes by an addition of H2B-CfaN-Strep dimers, which were purified away from symmetric ubiquitinated nucleosomes by Strep-Tactin affinity purification. Asymmetric nucleosomes were eluted from resin by thiolysis in the presence of a catalytically dead CfaC. (F) Native PAGE analysis of steps in the formation of H3/H3K4me3:H2B/H2BK120Ub asymmetric nucleosomes (DNA visualized by ethidium bromide staining). (G) Western blot analysis of WT/WT and dual asymmetric mononucleosomes demonstrating the presence of H3K4me3 (left) and two H2B bands (right) indicating the presence of both epitopes in a single asymmetric nucleosome.

The robustness of our strategy having been established, we then turned to study the impact of nucleosome desymmetrization on various aspects of chromatin biochemistry. We initially focused on cancer-associated histone mutations,[6,8,11,12] which remain poorly understood and are currently the focus of intense investigation.[7,9,10,17] Although oncohistones are present predominantly in a heterotypic context in cellular chromatin,[10] biochemical studies in this area have, with few exceptions,[9,22] been restricted to the use of homotypic nucleosomal substrates. A recent study from our group identified several oncohistone mutations that impact nucleosome remodeling processes when both copies of the histone carry the mutation.[9] Focusing on a subset of these, we asked whether this behavior extended to asymmetric contexts. Accordingly, we began by preparing heterotypic nucleosomes containing the mutations H3D81A or H3R83A that map to a region of the nucleosome, superhelical location (SHL) 2, where the ATPase subunit of the ATP-utilizing chromatin assembly and remodeling factor (ACF) binds (Figure S7H).[46,47] The ACF chromatin remodeler is a member of the imitation switch (ISWI) subfamily of chromatin remodelers and regulates nucleosome spacing. As controls, we also prepared the corresponding homotypic wild-type and mutant nucleosomes using our spytethering strategy (Figure S7I).

Consistent with our previous findings,[9] both homotypic mutants displayed reduced levels of ACF remodeling compared to the wild-type nucleosomes as shown by an established REA assay (Figure A–C, Figure S9).[44] Interestingly, the impact of desymmetrization differed for the two oncohistones. In the case of H3R83A, similar levels of inhibition were observed regardless of whether one or both copies of the histone were mutated. By contrast, for the H3D81A mutant, a clear dosage effect was observed in that a single copy of the mutant led to an intermediate level of inhibition. In considering the origins of this differential behavior, three features of the system must be considered: (i) The REA assay requires that the nucleosome is initially positioned toward one end of the DNA such that it can be moved to a centered position upon remodeling (Figure A). (ii) Our spytethering procedure yields a 50:50 mixture of nucleosomes in which the facial orientation of the histone mutant is directed toward or away from the direction of remodeler movement. (iii) ACF is believed to function as a dimer, binding both SHL (−2) and (+2) regions of the substrate, where this mode of engagement is required for normal remodeling activity.[36] Consequently, it is conceivable that, for the heterotypic H3D81A nucleosomes, where an intermediate effect was seen, one facial orientation of the mutant renders it able to be remodeled normally whereas the other leads to inhibition; thus, the observed rate represents the average of the two. Alternatively, each mutant orientation may be partially inhibitory toward remodeling such that the mutation of both residues has an additive effect in remodeling rate reduction, possibly due to an interaction between an unstructured loop in ISWI and H3D81 (Figure S7H). Although both H3D81 and H3R83 are found in the SHL 2 region of the nucleosome, unlike the former, the latter is positioned to form a histone/DNA contact that is lost upon mutation and results in destabilization of the nucleosome (see below). The binary effect on ACF remodeling observed in the case of R83A indicates that the disruption of a single SHL region through this mutation, irrespective of its orientation with regards to sliding direction, has the same impact as the disruption of both. Thus, in the context of asymmetry, remodeling is not always impacted in an intermediate manner by oncohistone mutations.

Figure 4

Biochemical interrogation of oncohistone asymmetry. (A) Schematic of the restriction enzyme accessibility (REA) assay. Nucleosomes bear 45 and 15 bp DNA overhangs (5′ and 3′, respectively, to the 601 DNA sequence) and a protected PstI site. Upon ACF remodeling, the PstI site is exposed and cut by the restriction enzyme. (B) Representative native gel analyses of DNA products from the ACF REA time course for WT, D81A/WT, D81A/D81A, R83A/WT, and R83A/R83A mononucleosomes as visualized by SybrGold staining. The bands represent uncut (top, unremodeled) and cut (bottom, remodeled) DNA, respectively. (C) Calculated rate constants for remodeling assays determined using densitometry values for cut and uncut DNA. Data were modeled as single-phase exponential decays from which observed rate constants [kobs parameter ± s.d. (n = 6)] were calculated. Statistical significance was determined by a t test (*p < 0.05, ****p < 0.0001). (D–F) Melt curves for WT (blue), asymmetric mutant (pink), and symmetric mutant (gold) mononucleosomes generated from a sigmoidal fit for the dimer melt (n = 3). (G) Summary of Thalf values calculated using fitted melt curves for WT and oncohistone mutant mononucleosomes presented as ± s.d. (n = 3). Statistical significance was determined by a t test (ns, not significant; ***p < 0.001, ****p < 0.0001).

We also examined the impact of asymmetric histone mutants on nucleosome stability, which can be assessed directly using the aforementioned thermal stability assay.[43] We found that two mutants, H3R83A and H3E97K, resulted in reduced nucleosome stability (as reflected by a reduced temperature for dimer melting) in both heterotypic and homotypic contexts, with the heterotypic mutant nucleosomes showing destabilization intermediate that of the wild-type and the homotypic mutant (Figure D,E,G, Figure S10). By contrast, the H3D81A mutation had little impact on stability even when present in both histone copies in the nucleosome (Figure F,G, Figure S10). Together with the sliding data presented above, our studies show that cancer-associated mutations can impact nucleosome stability and mobility when incorporated asymmetrically, as is the case in cells. This data also builds upon our previous findings that mutations can impact thermodynamic stability without affecting remodeling,[9] since here we uncovered that the inverse can also be true (e.g., in the case of H3D81). It is also notable that the H3R83A mutant inhibits ACF chromatin remodeling and reduces stability, demonstrating that nucleosome destabilization does not necessarily translate to easier remodeling; in fact, quite the opposite is observed in this case.

Next, we asked whether our spytethering system could be used to introduce asymmetry in a multinucleosome context. Access to such substrates is especially important given that many chromatin effectors have the potential to engage more than one nucleosome as part of their function.[48,49] A case in point is the H3K4me3 demethylase, KDM5B, which is overexpressed in numerous cancers.[50] Like all members of the KDM5 demethylase family, KDM5B contains a plant homeodomain (PHD) reader domain flanked by a split Jumonji histone demethylase domain (Figure S11A). Binding of this nested PHD domain (PHD1) to unmodified H3K4 histone tails has been shown to stimulate the demethylation activity of both KDM5A and KDM5B.[51,52] This positive feedback provides an attractive mechanism for efficient, and potentially processive, erasure of the H3K4me3 mark; however, previous biochemical studies have left it unclear whether stimulation occurs exclusively in an internucleosomal fashion or whether intranucleosomal crosstalk can occur also (Figure A). With respect to the latter, recent studies of the yeast homologue, Jhd2, have suggested a role for intranucleosomal stimulation to help maintain symmetrical nucleosome modification.[53]

Figure 5

Role of symmetry in KDM5B-mediated H3K4me3 demethylation. (A) Schematic illustrating the two potential modes of KDM5B stimulation involving either intra- or internucleosomal engagement of the PHD1 domain with the unmodified H3 tail. (B) Dinucleosome substrates used to investigate the geometry of KDM5B-mediated demethylation of H3K4me3. Types of stimulation possible by the KDM5B PHD1 domain for each substrate are represented by check marks. (C) Native gel showing mononucleosome starting materials and resultant dinucleosomes following ligation with T4 DNA ligase. (D) Immunoblot for H3K4me2 to assess KDM5B demethylase activity on dinucleosome substrates (top) and immunoblot for H4 as a loading control (bottom). (E) Histogram of H3K4me2 immunoblot, corrected for H4 loading. Densitometry was conducted using ImageJ and is shown ± s.d. (n = 3). Statistical significance was determined by a t test (*p < 0.05, **p < 0.01, ***p < 0.001).

To explore the geometric constraints on KDM5B stimulation, we designed a series of asymmetric dinucleosomes containing various combinations and arrangements of either wild-type H3 (WT), H3K4me3, or a truncated version of H3 (“H3tr”) that excludes the first 14 amino acids needed for binding the KM5B PHD1 domain (Figure B). The preparation of these dinucleosomes involved two steps. First, we employed our spytethering strategy to assemble three different heterotypic mononucleosomes, namely, a WT/H3K4me3 combination, a WT/H3tr combination, and a H3K4me3/H3tr combination (Figure S11B,C). In addition, we generated two homotypic nucleosomes in which both copies of H3 were either WT or the truncated version (Figure S11B,C). In the second step, these mononucleosome building blocks were then ligated together in specific pairings using a DNA ligation approach,[22] taking advantage of strategically incorporated complementary sticky end overhangs (Figure C). Importantly, each of the four dinucleosomes was constructed such that only one building block is asymmetrically methylated (serving as a substrate for KDM5B), allowing the geometric dependence of stimulation by the unmodified H3 tail to be examined (Figure B).

We utilized a minimal KDM5B construct used in previous studies of enzyme stimulation by PHD1 that contains the catalytic Jumonji domains, a DNA binding AT-rich interaction domain (ARID), and the PHD1 domain (Figure S11D,E).[51,54] Upon incubation with KDM5B, we saw clear and significant differences among the four asymmetric dinucleosome substrates (Figure D,E, Figure S12). Substrates 1 and 2 had nearly identical levels of demethylation, whereas substrates 3 and 4 displayed decreased levels of demethylation. The difference in activity between substrates 2 and 3 was especially informative since these substrates were designed to restrict stimulation to either an intranucleosomal (substrate 2) or internucleosomal (substrate 3) context. That the former is a significantly better substrate than the latter argues for the superiority of an intranucleosomal mode of stimulation (Figure A). However, the decrease in activity between substrates 3 and 4 also indicates that stimulation can occur between nucleosomes, although this does not appear to be as efficient a process. Whether this difference in stimulation as a function of geometry simply reflects preferential binding of the PHD1 domain to the unmodified H3 tail within the same nucleosome or some other more complex phenomenon related to how binding is coupled to catalysis (i.e., an allosteric process) remains unclear and will require additional investigation. The key finding here is that an unmodified H3 tail can stimulate H3K4me3 demethylation by KDM5B in an intranucleosomal fashion, consistent with a role for the enzyme in eliminating H3K4me3-based nucleosome asymmetry, as has been proposed in yeast.[53] More generally, our data adds to a growing body of work that indicates that nucleosome symmetry must be considered when exploring the role of histone PTMs in the regulation of epigenetic factors.[13,21]

In this study, we successfully developed a genetically encoded approach for the preparation of asymmetric nucleosomes. Our “spytethering” strategy provided streamlined access to a series of chromatin substrates whose inherent symmetry was broken through the introduction of histone variants, mutations, or one or more PTMs. This allowed us to show that cancer-associated oncohistones disrupt nucleosome stability and chromatin remodeling processes when present in a physiologically relevant asymmetric context. By employing a dinucleosome system and multiple facets of asymmetry, we were able to define the geometric constraints on the stimulation of KDM5B by its PHD1 domain. The operational simplicity of our approach combined with its compatibility with established approaches in the chromatin biochemistry area, such as histone semisynthesis and nucleosome ligation, make it a powerful addition to the chemical biology toolkit available for investigating the role of asymmetry on chromatin structure and function.

Safety Statement

No unexpected or unusually high safety hazards were encountered.