Nucleosome context regulates chromatin reader preference

Abstract

Chromatin is more than a simple genome packaging system but rather locally distinguished by histone post-translational modifications (PTMs) that can directly change nucleosome structure and/or be “read” by chromatin-associated proteins to mediate downstream events. An accurate understanding of histone PTM binding preference is vital to explain normal function and pathogenesis and has revealed multiple therapeutic opportunities. Such studies most often use histone peptides, though these cannot represent the full regulatory potential of nucleosome context. Here we apply a range of complementary and easily adoptable biochemical and genomic approaches to interrogate fully defined peptide and nucleosome targets with a diversity of mono- or multivalent chromatin readers. In the resulting data, nucleosome context consistently refined reader binding, and multivalent engagement was more often regulatory than simply additive. This included abrogating binding of the Polycomb group malignant brain tumor (MBT) protein L3MBTL1 to lysine methylated histone tails and confirmation that the CBX7 chromodomain and AT-hook-like motif (CD-ATL) tandem act as a functional unit to confer specificity for H3K27me3. These in vitro nucleosome preferences were confirmed by in vivo reader-CUT&RUN genomic mapping. Such data confirms that more representative chromatin substrates provide greater insight into biological mechanism and human disease.

Graphical Abstract

Graphical Abstract.

Introduction

Chromatin is an essential regulator of multiple biological processes, including transcription [1–4], DNA damage repair [5, 6], cellular differentiation [7–9], and pathogenesis [10–12]. Its basic repeating subunit is the nucleosome: a core histone octamer (two each of H2A, H2B, H3, and H4) wrapped with ∼147 bp DNA [13]. These structures are highly dynamic and distinguished by/associated with a compendium of histone variants (e.g. H2A.Z, H3.3, CenH3), post-translational modifications (PTMs; e.g. methylation, acetylation, phosphorylation, ubiquitylation), DNA modifications (e.g. methylation; primarily 5-methylcytosine), and chromatin-associated proteins (CAPs; including histone modifiers, chromatin remodelers, and transcription factors) [14–16].

The “histone code” hypothesis posits that combinatorial histone PTMs act as a molecular language “read” by CAPs to facilitate DNA transactions [17, 18]. This idea was highly controversial at inception [19–21], but has proven hugely insightful and greatly contributed to epigenetics research [22–24]. A multiplicity of “reader” domains have now been identified, and understanding how they interact with and interpret histone PTMs can reveal novel disease mechanisms, biomarkers, and therapeutic targets [25]. Such studies have been challenged by the dynamic nature of chromatin and (until recently) the limited availability of defined nucleosome substrates [24, 26–28]. Instead, the field has largely relied on modified histone peptides, where low manufacturing costs and compatibility with high-throughput screening (HTS) platforms have supported reader profiling to hundreds of PTM combinations [29–31]. However, peptide array screens are often compromised by high material consumption, low signal-to-background (S/B), and highly variable results [32–34]. They also rarely include post hoc optimizations, which can be essential to deliver maximal insight [28].

Perhaps the greatest underminer of peptide-based approaches is their inability to inform on CAPs that require nucleosome substrates, including NSD2 [35–37], LSD1 [38–40], DOT1L [41, 42], Set2 [43], EZH2 [44–46], or PR-DUB [47]. Such necessity is often due to obligate multivalent engagement, where the CAP contacts multiple histone PTMs, DNA modifications, and/or nucleosome surfaces [27, 48–57]. As an example, NSD2 lysine methyltransferase contains two PWWP domains that interact with nucleosomal DNA to effectively engage H3K36me2 [37]. Histone peptides are thus unable to recapitulate the context to fully characterize NSD2 PWWP domain binding [37] and develop maximally effective domain inhibitors [58, 59]. CAPs can also contain multiple reader domains that interact with individual histone PTMs in trans (i.e. on separate histones), as in the PHIP/BRWD2 Tudor-Bromodomain-Bromodomain (Tudor-BD-BD2) reader triplet with ([H3K4me3K14ac]₂•[H4K12ac]₂) nucleosomes [60]. Interactions with DNA, the histone tails, and/or the H2A/H2B acidic patch can play essential roles in CAP function, such as to allosterically activate DOT1 lysine methyltransferase after engaging H2BK120ub [51–54]. Finally, the positive charge on histone tails can be neutralized, such as by lysine acetylation, to relieve the association with negatively charged nucleosomal DNA and increase target accessibility for reader binding and further modification [27, 48, 56]. These examples underscore the need for reliable nucleosome-based approaches for effective CAP interrogation.

To address this, we developed the Captify^™ platform (Captify for brevity; formerly known as dCypher) using no-wash high-sensitivity Alpha [27, 28, 47, 61–63] or washable multiplexable Luminex [64] technologies for the rapid, sensitive, and robust screening of CAPs (the Queries) to fully defined histone peptides and nucleosomes (the targets). The ability to explore diverse conditions, including reactant concentrations and additives, often revealed multivalent engagement mechanisms and refined specificities on nucleosomes. This was confirmed by orthogonal approaches, including reader-CUT&RUN genomic mapping, indicating that in vitro CAP:nucleosome binding data more accurately reflects in vivo chromatin interactions.

Materials and methods

Histone peptides

PTM-defined peptides for Captify^™ assays were synthesized with a terminal PEG-Biotin (locations as indicated: Supplementary File 1), and identity/purity confirmed by mass spectrometry (MS).

Semi-synthetic nucleosomes

This study uses a nucleosome nomenclature recently devised for accurate scientific communication in the chromatin and epigenetic fields [57]. Here any distinguishing histones in a fully defined semi-synthetic homotypic nucleosome are in alphabetical and numerical order separated by “•,” as in ([H3K4me3K14ac]₂•[H4K12ac]₂). Other positions not denoted are understood as unmodified major histones.

Fully PTM-defined histones, octamers, and nucleosomes (dNucs^™ or versaNucs^™: Supplementary File 1) for Captify were synthesized, purified, and assembled on 147 bp 5′ biotinylated 601 DNA (EpiCypher 18-0005) as previously [32], but without DNA barcoding.

For dNucs, PTM-defined histones were mixed to defined stoichiometry, dialyzed, purified to octamers, and assembled onto 147 bp 5′ biotinylated 601 DNA [135]. The resulting products (e.g. ([H3K27me3]₂); EpiCypher 16-0317) contained full-length “scarless” histones and minimal (<5%) free DNA. PTMs were confirmed by MS and immunoblotting (if an antibody was available).

For versaNucs [27, 56], histone H3 tail peptides (aa1-31; A29L) with a designation of interest [PTM, mutation, or methyllysine analog (MLA)] were individually ligated to an H3 tailless nucleosome precursor (e.g. H3.1NΔ32; EpiCypher 16-0016). The resulting nucleosomes (assembled at 20–100 μg scale) contained undetectable levels of free peptide and ≥90% full-length H3.1 with the designation of interest. When interrogating readers and CAPs in this study, we observed no difference between dNuc and versaNuc behavior and thus used these substrates interchangeably. Of note, versaNucs contain A29L and are not recommended for studies near this position (e.g. modifiers/binders of H3R26, K27, or S28).

For nucleosomes with MLAs, histone H3 peptides (aa1-31; A29L) containing K4C or K9C were site-specifically reacted with the corresponding haloalkylamine: (2-bromoethyl) trimethyl ammonium bromide for K_Cme3; 2-chloro-N,N-dimethyl-ethylamine hydrochloride for K_Cme2; or 2-chloroethyl(methyl)ammonium chloride for K_Cme1; under SN2 reaction conditions [56, 65, 66] and purified for individual ligation to H3.1N∆32 by the versaNuc approach.

Query proteins

Reader domains were recombinantly expressed as GST- or 6His- fusions (Supplementary File 1 and Supplementary Figs S1 and S2), and purity assessed after sodium dodecyl sulfate–polyacrylamide gel electrophoresis and Coomassie staining. Single domain VHH (variable heavy domain of a heavy chain antibody) chromatibody to the nucleosome acidic patch was recombinantly expressed with a C-terminal 6His-tag and epitope characterized as previously [64, 67].

EpiTriton^™ histone peptide arrays

Three identical arrays of biotinylated histone peptides [287× (Supplementary File 1) in two sets of triplicate plus a fluorescein transfer control] were printed on streptavidin-coated glass slides (EpiTriton^™; EpCypher 11-4001) and stored/handled under subdued lighting. Prior to use, each slide subarray was submerged in hybridization buffer [1× phosphate buffered saline (PBS), 5% bovine serum albumin (BSA), 0.1% (w/v) Tween-20], placed in a humidified chamber and incubated for 30 min at 4°C with gentle rotation. The liquid on each subarray was replaced with GST-tagged query (2 μM in hybridization buffer) and incubated for 3 h at 4°C with gentle rotation. Slides were washed three times with 1× PBS in a rotating slide chamber at 4°C. To detect query binding, slides were probed with primary anti-GST (Sigma G7781) for 2 h, washed, and secondary anti-rabbit IgG AlexaFluor 647 (Invitrogen A-21244) for 30 min (all under subdued lighting in a humidified chamber at 4°C with gentle rotation). After each step slides were washed with 1× PBS, and finally dried by centrifugation at 800 x g for 2 min. Fluorescent signal (526 and 670 nm) was measured on a Typhoon Trio+ (GE) using a + 3 mm focal plane and 25 μm resolution. Signal was analyzed using ImageQuant TL (Cytiva) and Excel (Microsoft). All related data in Supplementary File 2.

Captify-Alpha

The assay previously known as dCypher^™ is now named Captify^™, with no distinction in how the assay is performed or its capabilities. Assays on the no-wash Alpha (amplified luminescence proximity homogeneous assay) platform (PerkinElmer, Revvity) to examine the interaction of epitope-tagged reader domains (the Queries) with biotinylated PTM-defined peptides or nucleosomes (the targets) were as previously [28, 58, 60, 61]. Importantly, Captify-Alpha assay buffers and bead binding buffers vary by target type. When using peptides, assay buffer and bead binding buffer [50 mM Tris, pH 7.5, 50 mM NaCl, 0.01% (v/v) Tween-20, 0.01% (w/v) BSA, 0.0004% Poly-L Lysine, 1 mM Tris(2-carboxyethyl)phosphine (TCEP)] were identical unless otherwise noted. In nucleosome-based experiments, assay buffer was (20 mM Tris, pH 7.5, 100–250 mM NaCl, 0.01% BSA, 0.01% NP-40, 1 mM Dithiothreitol (DTT)); bead binding buffer was the same minus DTT.

Queries and targets were prepared in suitable assay buffer (as earlier). In a 384-well plate, 5 μl of GST- or 6His-tagged query was combined with 5 μl of biotinylated peptide (100 nM final) or nucleosome (10 nM final) and incubated for 30 min at room temperature. Anti-tag donor beads and streptavidin acceptor beads were prepared in appropriate bead binding buffers as earlier. For GST-tagged Queries, a 10 μl mixture of 2.5 μg/ml glutathione acceptor beads (PerkinElmer AL109) and 5 μg/ml streptavidin donor beads (PerkinElmer 6760002) was added to each well. For 6His-tagged proteins (including 6His-VHH to the nucleosome acidic patch [64, 67]), a 10 μl mix of 2.5 μg/ml Ni-NTA acceptor beads (PerkinElmer 6760619) and 10 μg/ml streptavidin donor beads was added. The plate was incubated at room temperature in subdued lighting for 60 min, and Alpha signal measured on a PerkinElmer 2104 EnVision (680 nm laser excitation, 570 nm emission filter ± 50 nm bandwidth). Each reaction was performed in triplicate. Binding curves [query:target] were generated using a non-linear 4PL curve fit in GraphPad Prism 10, with EC₅₀^rel values computed for specific comparisons. Where necessary, values beyond the hook point (indicating bead saturation/competition with unbound query) were excluded, and top signal constrained to average max signal for target. In cases where signal never plateaued, signal was constrained to the average max signal within the assay.

To optimize binding conditions for nucleosome assays, query was cross-titrated in two-dimensions (2D) with salt (NaCl) or salmon sperm DNA (salDNA: Thermo Fisher Scientific 15632011) against a predicted target and unmodified nucleosome [Positive and Negative (background) controls, respectively]. Optimal conditions produced robust signal-to-background and were used for discovery studies within the range EC₂₀^rel – EC₈₀^rel, as indicated, preferentially near physiological salt concentration in nuclei (∼150 mM NaCl) [136]. Concentrations tested are indicated in respective figures (see also Supplementary File 2).

Preparation of MagPlex nucleosome panels

Avidin-conjugated MagPlex^® beads (Luminex, Diasorin) with spectrally distinct regions were used to assemble multiplexed nucleosome panels (Supplementary File 1). The histone lysine methylation panel (K-MetStat) comprised 16 nucleosomes: me1, me2, and me3 on H3K4, H3K9, H3K27, H3K36, and H4K20 and unmodified control. The acidic patch assessment panel comprised six nucleosomes: H2AK119ub, H2BK120ub, H2AE61A, H2AE92K, H2AE105A/E113A, and unmodified control. All MagPlex bead handling and incubation was performed under subdued lighting. Briefly, beads stocks were vortexed and incubated for 30 s in a water bath sonicator to monodisperse. Desired bead volumes were transferred to 1.5 ml tubes, placed on a magnet, and washed twice with pre-conjugation buffer (50 mM Tris, pH 7.5, 0.01% Tween-20). Thoroughly mixed beads were added to the assigned nucleosome (e.g. MagPlex Region 13 & [H3K4me3]), adjusted to 500 μl (5 μg nucleosome per 1 × 10⁶ beads: i.e. loading to saturation) with pre-conjugation buffer, and incubated on a rotator for 30 min at room temperature. Nucleosome-bound beads were washed twice with post-conjugation buffer (50 mM Tris, pH 7.5, 0.01% Tween-20, 0.01% BSA), counted (LUNA cell counter; Logos Biosystems), and adjusted to 1 × 10⁶/ml. Beads were pooled to a master mix with an equal quantity (e.g. 500 000 beads) of each MagPlex bead region, resuspended at 1 × 10⁶/ml per region in storage buffer (10 mM cacodylate, pH 7.5, 0.01% BSA, 0.01% Tween-20, 1 mM ethylenediaminetetraacetic acid, 10 mM beta mercaptoethanol, 50% glycerol), and stored at −20°C.

Panel balance, nucleosome integrity, and [Nucleosome:bead region] identity were confirmed using anti-double-stranded DNA (EMD Millipore #MAB030; 1/5, 1/50, and 1/500), anti-histone (MilliporeSigma MAB3422), anti-H3.1/2 (Active Motif #61629), and an appropriate anti-PTM if available (e.g. chromatinantibodies.com/) (Supplementary File 1). Bead and antibody dilutions were prepared in QC assay buffer (50 mM Tris, pH 7.5, 250 mM NaCl, 0.01% Tween-20, 0.01% BSA). Briefly, 50 µl of the multiplexed bead panel (20 000 beads/ml/region; 1000 beads/region) was combined with 50 µl of antibody (1:125, 1:500, and 1:2000 unless otherwise specified) in a 96-well plate and incubated for 60 min with shaking (800 rpm) to maintain bead resuspension. Beads were washed for three cycles on a magnet using 100 µl QC assay buffer, shaking for 2 min between cycles. Anti-IgG PE (anti-rabbit, Biolegend 406421; or anti-mouse, Biolegend 405307) was diluted 1:100 in QC assay buffer, added to each well, and incubated for 30 min with shaking. Beads were then washed for two cycles and resuspended in 100 µl QC assay buffer. Median fluorescence intensity (MFI) was measured using the FLEXMAP-3D system (Diasorin) with a minimum of 50 events/region. Bar graphs were generated in GraphPad Prism 10.

Captify-Luminex

Assays to examine the interaction of epitope-tagged reader domains (the queries) with biotinylated PTM-defined nucleosomes (the targets) in the multiplex K-MetStat Luminex panel were performed in 96-well plates (GreinerBio 655900) with a modified CUT&RUN buffer (20 mM Hepes, pH 7.5, 150 mM NaCl, 0.01% BSA, 0.01% Tween-20) or optimal query buffer from Captify-Alpha (20 mM Tris, pH 7.5, 150–250 mM NaCl, 0.01% BSA, 0.01% Tween-20, 1 mM DTT). Briefly, 50 μl of multiplexed bead panel (20 000 beads/ml/region; 1000 beads/region) was combined with 50 μl of tagged query (fixed concentration or titration; noted in figures/legends). The reaction plate was incubated for 60 min with shaking (800 rpm) to maintain bead resuspension. Beads were washed for three cycles on a magnet using 100 μl assay buffer, shaking for 2 min between cycles. One hundred microliters of anti-GST (1:2000; Fortis A190-122A) was added to each well and incubated for 30 min with shaking. Beads were washed for three cycles and incubated with rabbit anti-IgG PE (1:100; Biolegend 406421) for 30 min with shaking. Beads were washed for two cycles, resuspended in 100 μl assay buffer, and MFI measured on a FLEXMAP-3D for a minimum of 50 events/region. Bar graphs or binding curves (non-linear 4PL curves) were generated in GraphPad Prism 10.

Assays to examine the interaction of 6His-VHH chromatibody [67] with biotinylated PTM-defined nucleosomes (the targets) in the multiplex acidic patch assessment Luminex panels were performed in VHH buffer (20 mM Tris, pH 7.5, 100 mM NaCl, 0.01% BSA, 0.01% NP-40, 1 mM DTT) [64]. Serially diluted VHH query (two-fold from 2 μM to 15.6 nM) was probed against the panel (1000 beads/region per well) following steps earlier. For detection, 100 μl PE-labeled anti-VHH (1:400; Jackson ImmunoResearch Laboratories 128-115-232) was added to each well and incubated for 30 min with shaking. Beads were washed, resuspended in 100 μl in VHH buffer, and MFI measured. Bar graphs or binding curves (non-linear 4PL curves) were generated in GraphPad Prism 10.

Native top-down mass spectrometry

GST-RAG2 PHD (5 μM) was incubated with nucleosomes (unmodified and/or ([H3K4me3]₂); 1 μM) in binding buffer (20 mM HEPES, pH 7.5, 250 mM NaCl, 0.01% (w/v) BSA, 0.01% (v/v) NP-40, 1 mM DTT) for 1 h at 4°C. Reactions were desalted and buffer exchanged to 150 mM ammonium acetate using a 30 kDa MWCO centrifugal filter (MilliporeSigma). Buffer exchange was performed at 10 000 x g for 10 min at 4°C and repeated up to 10 times. Protein complex mixtures were adjusted to 1 µM prior to native MS analysis. Samples were loaded to borosilicate glass emitter tips (Thermo Fisher Scientific), native MS analysis performed on an Ultra High Mass Range Orbitrap QExactive Mass Spectrometer (Thermo Fisher Scientific) and data acquired using Xcalibur QualBrowser (Thermo Fisher Scientific). First, RAG2-nucleosome complexes were analyzed in MS¹. Complexes of charge states z = 34⁺ and 35⁺ between 8000–8300 m/z range were then activated by collision with nitrogen gas, resulting in ejection and detection of obtained histones by MS². Additional procedures are described elsewhere [68–70]. MS¹ and MS² spectra were obtained from averaged scans within MS.raw files; then deconvoluted using UniDec.

CUT&RUN assays

CUT&RUN [71, 72] was performed with 500 000 native K562 cells immobilized on Concanavalin A magnetic beads with a SNAP-CUTANA K-MetStat Panel spike-in (EpiCypher 19-1002) to monitor antibody/reader binding. Cells were permeabilized with 0.01% digitonin and incubated overnight at 4°C with 0.5 μg antibody (Supplementary File 1) or 70 nM GST-tagged reader (Supplementary File 1). For Reader CUT&RUN reactions, a secondary antibody (0.5 μg anti-GST; Fortis A190-122A) was then added. After the addition and activation of pAG-MNase (EpiCypher 15-1016), CUT&RUN-enriched DNA was purified using a 2:1 ratio of SPRIselect beads (EpiCypher 21-1403). Five nanograms DNA was used to prepare sequencing libraries with the CUTANA CUT&RUN Library Prep kit (EpiCypher 14-1001). Libraries were analyzed by Agilent TapeStation, pooled to equivalence, and sequenced (Illumina NexSeq 2000), targeting ∼10 million paired-end reads per reaction.

Genomic data analysis

Paired-end FASTQ files were aligned to the T2T-CHM13v2.0 reference genome using Bowtie2 [73, 74]. DAC exclusion list regions and multi-aligned or duplicate reads were removed prior to subsequent analyses using SAMtools v1.6 and Picard v2.26.2 [75]. RPKM-normalized smoothed bigWig files were generated using deepTOOLS v3.5.1 [76] and visualized using Integrative Genome Viewer (IGV). Peaks were called using SICER2 v1.0.3 [77]. Peak overlap (≥50%) was determined using Intervene v0.6.5 [78] and visualized using matplotlib-venn v0.11.7. deepTools was used to generate TSS and gene body heatmaps and perform Pearson correlations. Antibody and reader domain preference in each reaction was determined by comparing the number of sequencing reads for each DNA barcode in the K-MetStat Panel. Anti-PTM and reader domain barcode counts were normalized to reads from the anticipated target; anti-IgG and anti-GST controls were normalized to total barcode read counts. All sequencing data are available at NCBI Gene Expression Omnibus (accession number GSE249239). CUT&RUN analyses were performed independently a minimum of three times with consistent results.

K562 RNA-seq data (two replicates of polyA-messenger RNA) was downloaded from ENCODE (SRR4235541, SRR4235542) [79] and aligned to T2T-CHM13v2.0 using the GTEx RNAseq pipeline (github.com/broadinstitute/gtex-pipeline). Gene expression values were FPKM normalized, and replicates were assessed for reproducibility (R² = 0.87). Non-protein coding and transcriptionally inactive genes (FPKM = 0) were removed from further analysis, and the remaining 15 568 genes were binned to deciles by average FPKM. deepTools [76] was used to visualize CUT&RUN signal by RNA-seq decile.

Results

Captify-Alpha has dramatically improved performance over peptide arrays and supports nucleosome studies

The dissection of CAP: chromatin engagement mechanisms has historically relied on isolated reader domains and PTM-defined histone peptides. This reductionist approach has advanced our understanding, but physiological relevance is potentially undermined by both components being removed from higher-order context (i.e. protein complexes and nucleosomes). We previously developed histone peptide arrays (e.g. EpiTriton^™) for such research but observed multiple drawbacks, including high query protein requirements (up to 800 pmol) and low dynamic range. Most notably, the arrays were nucleosome-incompatible, and thus unsuitable for exploring multivalent engagement—a feature of many CAPs. Consequently, ∼80% of tested Queries failed to generate reliable binding data (not shown).

To address these shortcomings, we developed Captify on the high-sensitivity no-wash Alpha platform [80] (hereafter Captify-Alpha). Here, biotinylated substrates (the targets) are mixed with epitope-tagged CAPs (the Queries) and relative engagement (expressed as EC^rel) quantified through proximity-dependent signal generation using streptavidin “Donor” and anti-tag “Acceptor” beads [27, 61–63] (Supplementary Fig. S3A andSupplementary File 1). Compared to histone peptide arrays, Captify-Alpha is highly adaptable, HTS-compatible, and provides substantial gains in sensitivity and S/B, while dramatically reducing material consumption. To demonstrate, we examined BRD4 bromodomain 1 (GST-BRD4 BD1; query) binding to histone peptides by EpiTriton array (2 µM query; hits defined as > two-fold S/B) and Captify-Alpha (30 and 1 nM query; hits defined as >100-fold S/B). In both formats, BRD4 BD1 bound H4 acetyl-lysine peptides, preferring those with multiple acetylations (i.e. H4_[1–23]K5acK8acK12acK16ac; aka. H4_[1–23]tetra^ac) over singles [31, 81] (Fig. 1A and B). However, the Captify-Alpha approach captured more hits, including all detected by peptide arrays, while using much less query (up to 2000-fold) and with a much higher dynamic range (>1000-fold) (Fig. 1A and B). Further, these binding partners and their rank-ordering were confirmed by orthogonal TR-FRET [81] (Supplementary Fig. S3B and C).

Figure 1.

Lysine-acetyl readers show refined PTM specificity in nucleosome versus histone peptide assays. (A) GST-BRD4 BD1 (query, concentrations as noted) interactions in histone-peptide based assays: EpiTriton array and Captify-Alpha. “Hits” are defined by signal-to-background (S/B) > 2 for peptide array and >100 for Captify-Alpha. (B) GST-BRD4 BD1 S/B in EpiTriton (2 μM) and Captify-Alpha (1 or 30 nM) for select H3 and H4 acetyl peptide targets. Each assay identifies the bromodomain preferentially interacting with various forms of H4 tail acetyl. (C) [Inset] GST-BRD4 BD1 query titration to potential nucleosome targets; dashed line represents EC₈₀^rel (5 nM) against ([H4K5acK8acK12acK16ac]₂) (aka. [H4tetra^ac]). [Main] GST-BRD4 BD1 (5 nM) discovery screen to indicated panel reveals strong preference for [H4tetra^ac] nucleosomes. (D) [Inset] GST-BRM BD titrations to potential nucleosome targets; dashed line represents EC₂₀^rel (30 nM) against ([H3K4acK9acK14acK18ac]₂) (aka. [H3tetra^ac]). The binding at near µM concentrations to unmodified nucleosomes is likely due to interactions with nucleosomal DNA [60] (Supplementary Fig. S4C). [Main] GST-BRM BD (30 nM) discovery screen to indicated panel identifies strong preference for [H3tetra^ac] nucleosomes. See Supplementary Figs S4–S6 for Captify-Alpha two-dimensional (2D) titrations with a range of lysine-acyl readers, and Supplementary File 2 for complete datasets from peptide (n = 287) and nucleosome (n = 77) target discovery screens.

Of particular utility, Captify-Alpha is also compatible with defined nucleosome targets, where continued improvements in synthesis and scaling have dramatically increased the diversity now available [24, 26, 82]. In this version of the workflow, a regular first step is query titration to potential negative and positive controls [usually unmodified and predicted engaged PTM(s)] [28], which can include an exploration of buffer components (e.g. salt) and supplements (e.g. exogenous DNA, cofactors, or divalent cations) to optimize S/B. Pinpointing a query concentration between EC₂₀^rel and EC₈₀^rel with robust S/B is considered ideal for further use, per HTS guidelines [83]. As an example, we tested GST-tagged forms of two disease relevant bromodomains that respectively engage acetylated H4 and H3 histone tails: BRD4 BD1 [31, 81] and BRM BD [84–86]. In each case, titration identified optimal reader concentrations (Fig. 1C and D, Inset: S/B ≥ 10) for discovery screening to nucleosomes (77-member panel; Supplementary File 2), where each bromodomain preferred multi-acetylated targets (Fig. 1C and D, Main). Indeed, BRD4 BD1 failed to engage single H4 tail acetylations in the nucleosome context despite robust binding to comparable peptides (compare Fig. 1B and C), while BRM BD only weakly bound ([H3K14ac]₂), contrasting with prior peptide studies [85].

The restriction of reader: PTM binding preference on nucleosomes versus peptides is commonly reported [27, 28]. To investigate further, we used Captify-Alpha to examine six additional acyl-readers, including two more bromodomains (BRD3 BD1 and BRG1 BD) and all four human YEATS domains: a family with strong links to transcriptional regulation and disease [87–89]. For each query, we established reader capability followed by a discovery screen with PTM-defined peptides (Supplementary Fig. S4); then optimized binding conditions to nucleosomes followed by a discovery screen to this target class (Supplementary Figs S5 and S6). In every case, we observed refined binding on nucleosomes versus peptides. Further, each reader preferred distinct conditions to yield an acceptable S/B (Supplementary Figs S5 and S6), which may inform on their individual means of target engagement, but also stresses the importance of extensive optimization in such studies.

Nucleosome context refines the binding preference of methyl-lysine readers

Having established assays for readers of lysine acylation, we extended our investigation to those of histone lysine methylation. This PTM class potentially exists at target residues in three major forms (me1, me2, and me3), and is regulated by families of writer and eraser enzymes (lysine methyltransferases and demethylases), with the distinct product states distinguished by genomic distribution and function [90–94].

Polycomb Group protein L3MBTL1 contains three malignant brain tumor (MBT) repeats with a reported preference for lower lysine methylations (Kme1 or Kme2) on multiple linker and core histone residues, including H1.5K27, H3K4, H3K9, H3K27, and H4K20 [95–100]. Query titration of GST-L3MBTL1 MBT to H4_[11–27]K20 peptides (me0-1-2-3) confirmed preferential reader binding to the mono and dimethyl states (Fig. 2A), with discovery testing (at EC₈₀^rel; 3 nM) to the lysine-methyl status panel (K-MetStat; me0-1-2-3 at H3K4, H3K9, H3K27, H3K36, and H4K20) confirming interaction to each Kme1/Kme2 peptide (Fig. 2B). Such PTM state preference but residue promiscuity is due to a loose binding mode, where the MBT domain interacts with the mono- or dimethyl-lysine without contacting surrounding residues [98, 99]. However, it would seem of questionable biological utility since each of these PTMs is distinctly regulated in vivo [90–94]. We thus explored L3MBTL1 MBT binding to the K-MetStat nucleosome panel, and observed no engagement (Fig. 2C). This may suggest either that additional elements are required for binding in the chromatin context, or the peptide data are of limited in vivo significance. Given the cancer relevance of L3MBTL1, including its potential roles in tumor suppression and drug resistance [101, 102], further investigation of mechanisms governing its nucleosome interactions is warranted.

Figure 2.

Lysine-methyl readers show refined PTM specificity in nucleosome versus histone peptide assays. (A) GST-L3MBTL1 MBT query titration to H4_[11–27]K20 methyl peptide targets (maximal S/B at EC₈₀^rel = 3 nM) identifies the reported preference for Kme1 and Kme2. (B) GST-L3MBTL1 MBT (3 nM) discovery screen against 287-member peptide panel identifies a preference for Kme1 and Kme2 independent of histone residue (data subset shown). (C) [Inset] GST-L3MBTL1 MBT titration to nucleosome targets bearing distinct methyl states at H4K20 identifies no binding preference. [Main] GST-L3MBTL1 MBT (3 nM) discovery screen shows background binding to nucleosomes independent of lysine-methyl status. (D) [Inset] GST-RAG2 PHD titration to H3_[1–20]K4 methyl peptides (EC₇₀^rel = 1.8 nM). [Main] GST-RAG2 PHD (1.8 nM) peptide discovery screen shows selective binding to methylations at H3K4 (me1-2-3) over all other histone methyl states (data subset shown). (E) [Inset] GST-RAG2 PHD titration to nucleosomes with H3K4 methyl states (EC₈₀^rel = 11.9 nM). [Main] GST-RAG2 PHD (11.9 nM) nucleosome discovery screen reveals a refined preference for ([H3K4me3]₂). For (B–E), see Supplementary File 2 for complete discovery screen datasets. (F) Native top-down mass spectrometry (nTDMS) distinguishes a 1:1 mix of unmodified and [H3K4me3] nucleosomes (both 1 μM). (G) GST-RAG2 PHD (5 μM) selectively associates with ([H3K4me3]₂) over unmodified nucleosomes. Reader was incubated with a 1:1 nucleosome mix (both 1 μM) and analyzed by nTDMS. Mass-to-charge (m/z) peaks correspond to dimerized GST-RAG2 PHD bound to ([H3K4me3]₂) nucleosomes (further characterized in Supplementary Fig. S7).

RAG complex (Recombination Activating Gene: RAG1-RAG2) is critical for adaptive immunity [103–105] and mediates DNA cleavage during V(D)J recombination at immunoglobulin and T-cell receptor genes. Within RAG, the plant homeodomain (PHD) of RAG2 binds H3K4me2/me3 for effective genomic targeting [106–108]. To explore any impact of target context, we titrated GST-RAG2 PHD query to H3K4 methyl-focused peptides and nucleosomes (Fig. 2D and E, insets), followed by discovery screens to the respective K-MetStat panel. In peptide testing, the reader bound all three methyl states (me1-2-3) over the unmodified residue at H3_[1–20]K4 (Fig. 2D); while nucleosomes refined binding to a preference for the higher methylations (H3K4me3 > me2: Fig. 2E). Orthogonal validation of this engagement was provided by mixing unmodified and ([H3K4me3]₂) nucleosomes (1:1), confirming their distinct identities by nTDMS [69, 70] (Fig. 2F), then adding GST-RAG2 PHD and confirming the specific formation of a (GST-RAG2 PHD : ([H3K4me3]₂)) complex (Fig. 2G and Supplementary Fig. S7).

We continued our analysis of methyl-lysine readers, selecting Queries to represent a diversity of domain types, targets and functions: chromodomains (CDs) from heterochromatin protein HP1β (CBX1) [109–111] and Polycomb protein CBX7 [112–116]; PWWP domains from histone methyltransferases NSD2 and NSD3 [36, 37], DNA methyltransferase DNMT3A [61, 117], and non-enzymatic nuclear factors GLYR1 [118] and BRD1 [119]; and the ADD domain of ATRX, a SWI2/SNF2 chromatin remodeler [120–122]. In each case query engagement was established/optimized on PTM-defined peptides and nucleosomes, followed by discovery screens where each reader showed refined binding on nucleosome targets. For example, GST-HP1β CD and GST-ATRX ADD engaged H3_[1–20]K9 peptides containing all three methyl states (me3/2 > 1), but only [H3K9me3] nucleosomes [28] (Fig. 3A and B and Supplementary Figs S8–S10). As expected, the PWWP domains largely failed to bind histone peptides [28, 37], with the notable exception of DNMT3A, which preferred H3_[21–44]K36me2/3 but also engaged H3_[15–34]K27me2/3 (Supplementary Fig. S8A). However, on nucleosomes all PWWPs showed varying selectivity for the higher methyl states of H3K36, which free DNA competitor was often required to discern (e.g. GLYR1: see below) [28] (Fig. 3C–E and Supplementary Fig. S8B).

Figure 3.

Nucleosome context is required for an accurate determination of multivalent reader engagement. Titration of GST-HP1β (CBX1) CD (A) and GST-ATRX ADD (B) to indicated nucleosomes identifies differential impact of the ([H3K9me3S10ph]₂) combinatorial relative to ([H3K9me3]₂). (C) salDNA optimization for 6His-GLYR1 PWWP. Nonspecific nucleosome binding is reduced by the free DNA, revealing a preference for ([H3K36me3]₂). Dashed line represents optimal salDNA concentration (120 ng/ml) used for (D and E). (D) Salt optimization for 6His-GLYR1 PWWP. Query was titrated at noted NaCl concentrations against ([H3K36me3]₂) and unmodified nucleosomes, where increasing salt reduced binding to latter (EC₈₀^rel = 170 nM). (E) GLYR1 PWWP (170 nM) binding in a nucleosome discovery screen with optimized buffer conditions (120 ng/ml salDNA, 175 mM NaCl) confirms a preference for ([H3K36me3]₂). (F) GST-CBX7 CD-ATL titration to histone peptides identifies equivalent binding to H3_[1–20]K9me3 and H3_[15–34]K27me3. (G, H) CBX7 CD-ATL titration to nucleosomes reveals a preference for ([H3K27me3]₂), but only in presence of salDNA competitor (123 ng/ml: compare G and H) (EC₅₀^rel = 7.1 nM). (I) CBX7 CD-ATL (7.1 nM) binding in a nucleosome discovery screen with optimized buffer conditions (120 ng/ml salDNA), confirms a preference for ([H3K27me3]₂). See Supplementary Figs S9 and S10 for Captify-Alpha 2D optimizations of a range of lysine-methyl readers; Supplementary Figs S12 and S13 for further dissection of the CBX7 CD-ATL tandem; and Supplementary File 2 for complete datasets from all discovery screens.

The unpredictability of methyllysine analogs for binding studies

Structural analogs of the Kme1-3 states (K_Cme1-3; methyl-amino-alkylated cysteines [65, 66, 123] (e.g. Supplementary Fig. S11A and B)) continue to be employed for binding and structural studies, most particularly in NMR spectroscopy, which regularly uses ¹⁵N-labeled recombinant histones for K_Cme conversion [48, 124, 125]. Despite this technical convenience, the value of such studies is undermined if the MLA does not functionally behave as the structurally similar native Kme. To investigate, we created nucleosomes with native H3K4me3 and H3K9me3 and their MLAs (respectively H3K4_Cme3 and H3K9_Cme3), and compared antibody and reader binding to each class. Here, an anti-H3K4me3 and anti-H3K9me3 bound the native methyl and MLA with similar efficiencies, while the readers showed contrasting tolerances: RAG2 PHD bound equivalently to ([H3K4me3]₂) and ([H3K4_Cme3]₂), while ATRX ADD effectively engaged ([H3K9me3]₂) but was blind to ([H3K9_Cme3]₂) (Supplementary Fig. 11C–F). Such observations are a salutary reminder of the unpredictable nature of MLAs, which often support weaker reader domain engagement relative to native methylations [126–129], and where the safest path appears to directly compare reader binding with each entity before adopting MLAs for technical convenience.

Interrogating the impact of histone PTM crosstalk and multivalent engagement

Chromatin functionality through PTMs requires coordinated engagements, where various elements can have a positive or negative impact [14, 22, 24, 49, 130]. As an example, the HP1β CD and ATRX ADD reader domains both interact with H3K9me3 to regulate gene silencing and genome stability [121, 122, 131–134]. However, their binding is differentially impacted by H3S10 phosphorylation in cis, which displaces HP1β CD [131, 132], while ATRX ADD is unaffected [133, 134]. This mechanism was recapitulated by Captify-Alpha with nucleosome targets, where combinatorial S10ph abolished HP1β CD binding to H3K9me3, but three-fold enhanced that of ATRX-ADD (Fig. 3A and B).

CAP complex: nucleosome engagement is largely (if not invariably) multivalent, with multiple contact points on the histone tails (e.g. at PTMs in cis or trans) and other nucleosome surfaces, such as the H2A/H2B acidic patch and/or DNA (by charge, sequence motif, or modification). Such modes of engagement are exhibited by individual reader domains with multimodal binding ability (e.g. the PTM and DNA binding PWWPs [58]), or grouped reader domains that independently bind co-occurring PTMs in cis and/or trans, such as the BPTF PHD-BD [27, 48] or PHIP/BRWD2 Tudor-BD1-BD2 [60].

PWWP domains have a general preference for H3K36me2/me3 (Supplementary Fig. S8), but many also bind DNA [135]. As an example, we examined the PWWP domain from GLYR1, a cofactor of the H3K4me3 demethylase LSD2 [136, 137] and H3K36me3 reader [118] with putative roles in transcriptional elongation [138, 139]. In initial testing 6His-GLYR1 PWWP showed equivalent binding to unmodified and ([H3K36me3]₂) nucleosomes, but a profound preference for the latter in the presence of salDNA competitor (Fig. 3C), that was further improved by salt optimization (Fig. 3D and Supplementary Fig. S10G). Screening the K-MetStat nucleosome panel with these optimized conditions confirmed H3K36me3 as the primary target (Fig. 3E and Supplementary Fig. S8B).

Polycomb repressive complex 1 (PRC1) co-operates with PRC2 to establish repressive H3K27me3 heterochromatin domains that regulate gene expression, development, and higher-order chromatin architecture [140]. The PRC1 CBX7 subunit contains an H3K27me3-binding chromodomain (CD) [115, 141–143] and adjacent DNA-binding AT-hook-like (ATL) motif [112–116]. Both are required for stable chromatin interaction in vivo and proposed to act as a functional unit, with initial H3K27me3 engagement of the chromodomain triggering an allosteric change that enables the ATL to engage DNA [115]. However, CBX7 CD is also reported to bind equivalently to H3K9me3 and H3K27me3 [112–114], a capability of unknown relevance.

Captify-Alpha was used to examine target binding by diverse GST-CBX7 queries, including the individual CD or ATL, the tandem CD-ATL, and a tandem containing a loss-of-function point-mutated CD [115] (CBX7 CD(F11A)-ATL). Studies with histone peptides (H3_[1–20] or H3_[15–34]), revealed equivalent binding of CBX7 CD to K27me3 and K9me3 > H3K27me2 ≈ H3K9me2 (Supplementary Fig. S12A), while CBX7 CD-ATL bound H3K27me3 ≈ H3K9me3 > H3K27me2 >> H3K9me2 (Fig. 3F), and the chromodomain mutation ablated all binding (Supplementary Fig. S12B). So, the CBX7 CD mediates effective and indistinguishable binding to trimethylated H3K9 and H3K27 in the peptide context, possibly driven by the sequence similarity of these regions (…AR[K9]ST… versus …AR[K27]SA…): a distinction difficulty shared by many antibodies to the methyl states of K9 or K27 (chromatinantibodies.com/). On peptides, the CBX7 ATL is of minor consequence, somehow contributing to selectivity between the dimethyl states.

We next investigated the binding of CBX7 CD to PTM-defined nucleosomes. Here, the higher-order context refined chromodomain binding preference to ([H3K27me3]₂) over other methyl states at H3K27 and H3K9 (Supplementary Fig. S12C), consistent with the CBX7 role in Polycomb-repressed chromatin. The native context for CBX7 CD is with the adjacent DNA-binding ATL, but the tandem constructs [CD-ATL or CD(F11A)-ATL] were dominated by DNA binding, with no selectivity between unmodified and methylated nucleosomes (Fig. 3G and Supplementary Fig. S12D). However, salDNA competitor revealed the preference of CD-ATL for ([H3K27me3]₂) (Fig. 3H and Supplementary Figs S12 and S13), which was further demonstrated on testing the tandem reader to K-MetStat nucleosome panel under optimized conditions (EC₅₀^Rel: Fig. 3I). In this manner, the CBX7 CD-ATL tandem uses a multivalent mechanism to engage nucleosomes, with the higher-order context also conferring a chromodomain specificity not displayed on histone peptides. This is in general agreement with a model where the CBX7 CD-ATL comprise a functional unit to engage chromatin [115], and reconciles conflicting published data, where all suggesting CBX7 CD bound to H3K9me3 used histone peptides.

Development of a wash-based multiplex platform that predicts genomic mapping capability for reader binding studies

No-wash Captify-Alpha can effectively profile CAP:nucleosome binding preferences. However, the approach also requires adaptations to interrogate multivalent binding, such as inclusion of salDNA to isolate a DNA-mediated element (e.g. Fig. 3C and H). Stringency can also be increased by washing, spurring our development of Captify^TM-Luminex (Fig. 4A). This offers numerous advantages: high sensitivity, reproducibility, and throughput; multiplex capabilities (i.e. testing multiple targets in a single query reaction); and compatibility with CUT&RUN wash conditions, offering a cost-effective means to optimize reader binding before genomic analyses.

Figure 4.

Captify-Luminex enables interrogation of chromatin reader binding to multiplexed nucleosome targets. (A) Biotinylated nucleosome targets (the K-MetStat panel: me0-1-2-3 at H3K4, H3K9, H3K27, H3K36, and H4K20) are individually conjugated to distinct Luminex avidin-coated MagPlex bead regions, pooled, and probed in multiplex with GST-tagged reader domains. Interactions are detected using anti-GST and anti-IgG*PE secondary, with bead/PE signals measured on a FLEXMAP-3D system. (B) GST-HP1β CD, GST-TAF3 PHD, and GST-CBX7 CD-ATL binding across the K-MetStat panel. For each query concentration (nM; labeled columns), heatmap depicts signal as percentage of max MFI. (C) Signal range for each query in panel (B). Figure 4A created in BioRender https://biorender.com/178t3sy.

Luminex xMAP uses fluorescently barcoded MagPlex^® microspheres (aka. bead regions) to bind distinct targets and enable multiplexing. For Captify-Luminex, biotinylated PTM-defined nucleosomes (targets) were individually conjugated to discrete MagPlex-Avidin bead regions (providing a fluorescent barcode of histone PTM identity) and then pooled to multiplex panels (e.g. K-MetStat: me0-1-2-3 at H3K4, H3K9, H3K27, H3K36, and H4K20). GST-tagged queries were then combined with the panel in 96-well plates, detected with a fluorescent label [using anti-GST and anti-IgG*phycoerythrin (PE)], and relative binding assessed by co-localization of query (fluorescent PE) and target (fluorescent bead region) (Methods and Fig. 4A). Notably, multiplexing also provides a competitive environment, reflective of studies with native chromatin.

As an initial capability test, we focused on an integral nucleosome surface: the acidic patch. This cluster of negatively charged residues in histones H2A and H2B is a central hub for nucleosome interactors, including diverse chromatin-modifying and remodeling enzymes to regulate their activity, and the H4 tail to promote chromatin fiber formation [144–147]. H2BK120ub partially occludes the acid patch, with this “gatekeeping” function of the labile and mobile PTM a potential means to regulate access for pathway-specific patch binders within actively transcribed genes [64, 148–151].

Captify-Alpha and -Luminex were used to examine acid-patch binding by the VHH chromatibody [67], and each approach confirmed effective engagement with unmodified nucleosomes but not an acid-patch mutant (H2AE92K) (Supplementary Fig. S14). However, the [H2BK120ub] nucleosome was indistinguishable from unmodified in Alpha but closer to the acid-patch mutant in Luminex [64] (Supplementary Fig. S14). In this manner, acid-patch binding by chromatibody was stable enough to tolerate H2BK120ub in the no-wash platform, but PTM impact was revealed by the multi-step washable approach.

We next applied Captify-Luminex to three GST-tagged lysine-methyl readers: HP1β CD with a preference for H3K9me3 (Fig. 3A), TAF3 PHD with a preference for H3K4me3 [152, 153], and CBX7 CD-ATL with a preference for H3K27me3 in the presence of salDNA (Fig. 3H and I). Each preference was recapitulated with the Luminex approach, noting that CBX7 CD-ATL no longer required salDNA competitor to selectively engage H3K27me3. Additionally, each S/B was regulatable by query concentration, and under optimal conditions provided a >55-fold ratio from target to unmodified nucleosome (Fig. 4B and C).

Previously, we established the reader-CUT&RUN method using the BPTF PHD-BD tandem, observing a high degree of overlap with its combinatorial PTM targets {H3K4me4K14acK18ac} and endogenous BPTF [27]. Supported by fully defined DNA-barcoded spike-in nucleosome controls, reader-CUT&RUN enables the in vivo dissection of binding preference for chromatin features, including combinatorial histone PTMs, DNA modifications, and higher-order structures. Building on this, we considered that Captify-Luminex optimized conditions might be further suitable for the genomics approach. In this manner, each GST-tagged lysine-methyl reader manifested spike-in nucleosome recoveries (Fig. 5A) and genomic enrichment patterns similar to an antibody to their ostensible PTM targets (Fig. 5B): TAF3 PHD/anti-H3K4me3 formed sharp peaks at active promoters (Fig. 5C, F, and G and Supplemental Fig. S15); HP1β CD/anti-H3K9me3 covered broad regions across constitutively repressed locations (Fig. 5D, F, and H); and CBX7 CD-ATL/anti-H3K27me3 spanned polycomb-repressed genes (Figs. 5E, F, and I).

Figure 5.

Readers can deliver antibody-like PTM profiling in CUT&RUN genomic mapping. (A) DNA-barcoded nucleosome spike-in (K-MetStat panel) recovery from Reader-CUT&RUN reactions. Columns group data by antibody/GST-tagged reader; rows depict recovery of each PTM in K-MetStat panel normalized to predicted target (100%; orange in heatmap), except for anti-IgG and anti-GST (% of total spike-in reads). (B) Pearson correlation plot compares CUT&RUN read overlap from antibodies versus GST-readers (1 = perfect correlation). (C–F) Venn overlap of called peaks between PTM-specific antibodies and GST-readers (≥50% peak overlap classed as positive). Table shows average peak width and % overlapping peaks in each pairwise comparison (target 1, GST-reader; target 2, PTM-specific antibody). Representative peak comparisons: H3K4me3 and GST-TAF3 PHD (G); H3K9me3, H3K27me3, and GST-HP1β CD (H); H3K9me3, H3K27me3, and GST-CBX7 CD-ATL (I). Each IGV browser window (genomic region noted) is group-scaled to * track. IgG and αGST are background controls.

Discussion

Accurate interrogations of chromatin reader interactions are crucial to advancing our understanding of normal biology and the disease state. In this work, we develop, validate, and integrate a range of biochemical and genomic mapping approaches to study chromatin readers in the nucleosome context. Direct comparison showed histone peptides often failed to accurately predict the nucleosome binding preferences of even single reader domains, and were uninformative or even misleading for multivalent engagement, particularly when a DNA-binding component was involved (e.g. Figs 1–3). In stark contrast, in vitro performance in Captify-Luminex could be applied to CUT&RUN genomic mapping, where various lysine-methyl readers displayed antibody-like capability to distinguish various histone lysine-methyl states (e.g. Figs 4 and 5). This offers a path to the more efficient study of novel chromatin regulators and their disease-associated mutations.

The limitations of surrogates

Histone PTM peptides have been truly enabling for decades of chromatin studies [24]. Their primary strength lay in convenience: fully PTM-defined histone regions at an accessible price, control over experimental variables, and the ability to consistently generate data. However, histones peptides lack the structural complexity required to fully explore CAP interactions [21, 28, 35–42, 44–47, 56], and here we show that many readers bind differently, or not at all, when tested on nucleosomes (e.g. Figs 1–3). Several key features of nucleosomes help explain this discrepancy. Nucleosomes, unlike peptides, are complex assemblies of histones and DNA. Though often depicted as protruding from the nucleosome [19, 21], the histone tails are not freely accessible, but instead use positively charged residues to dynamically interact with DNA and the acidic patch [24, 48, 154]. This creates competition that can occlude certain reader engagements and may underlie observations such as restricted RAG2 PHD binding preference (H3K4me3 > H3K4me2/1; Fig. 2D and E), and the complete abrogation of L3MBTL1 MBT binding (Fig. 2A–C).

Fully defined nucleosomes are now readily available and wholly compatible with biochemical, proteomic, structural, and genomic approaches [26, 28, 56, 60–64, 69, 82, 126, 144, 155–159]. Our systematic analyses of multiple CAP domains that engage PTMs, DNA, or the acidic patch reveal some general observations, including: that nucleosomes consistently refine binding preferences relative to histone peptides; that multivalent nucleosome engagement by CAPs is more often regulatory than additive; and that, while reductive assays with surrogates (i.e. isolated domains, histones peptides, or MLAs) can yield data, it too often conflicts with data from multi-domain CAPs, nucleosomes, and native PTMs. Given this, the most valid mechanistic insight most likely resides in the more complicated systems, and a move away from the surrogates is largely warranted.

New insights await, but any exploration of nucleosome engagement must accommodate the multivalent milieu

Reader domain:nucleosome binding studies often require extensive optimization (e.g. the 2D-query concentration and salt explorations of Supplementary Figs S6 and S10; or competitor DNA titrations of Fig. 3C, G, and H, and Supplementary Fig. S13). Here, a major strength of no-wash Captify-Alpha is its flexibility for modified conditions to support weak binders removed from their native context (e.g. Supplementary Fig. S6), or to isolate a particularly dominant element (e.g. the DNA-binding ATL motif in the CBX7 CD-ATL tandem: Fig. 3F–I and Supplementary Figs S12 and S13); though it proved unable to reveal H2BK120ub-attenuated binding of chromatibody VHH to the acidic patch (Supplementary Fig. S14). However, a major strength of washable and competitive Captify-Luminex is using its higher stringency to explore multivalent engagement. As an example, the approach had no difficulty identifying the H3K27me3 preference of CBX7 CD-ATL without competitor DNA (Fig. 4), and even suggested optimized conditions where the recombinant reader was capable of mapping H3K27me3 with similar metrics to a PTM-specific antibody in CUT&RUN (Fig. 5).

Our analyses of individual bromodomains often revealed a strong preference for multiply modified nucleosome substrates, despite each BD generally only being able to accommodate a single acetylated lysine in their PTM binding pocket [31, 160–163]. For example, BRD4 BD1 displayed strong binding to ([H4K5acK8acK12acK16ac]₂) but none of the single H4 acetylations (Fig. 1C), while BRM BD had > 10-fold stronger binding to ([H3K4acK9acK14acK18ac]₂) over ([H3K14ac]₂) (Fig. 1D andSupplementary File 2). This could represent increased target accessibility, where multiple acetylations weaken the histone tail interaction with nucleosomal DNA [28, 56, 147, 164–169], but only one acetylated residue is primarily bound by the BD. Alternatively, high local concentrations of acetylated lysines (as frequently observed on in vivo chromatin [162, 170]) could contribute transient engagements to the recruitment and/or stabilization of a site-specific Kac binding event. However, individual bromodomains are almost certainly not biologically expected to manifest an individual role. As an example, BRD4 contains two adjacent bromodomains (BD1 and BD2), with the reader tandem effectively enriching native nucleoforms containing di- and tri-acetylated histone H4 tails [171]. Further, while BPTF BD weakly binds ([H3K4acK9acK14acK18ac]₂), the PHD-BD tandem binds similarly to ([H3K4me3K9acK14acK18ac]₂) and ([H3K4me3K14ac]₂) [27]. In each case, combinatorial context synergizes at least two individually weak site-specific interactions, likely representing the biological mechanism of action. With this in mind, we used the GST-epitope tag through this study to take advantage of its dimerization [172], and thus provide an “avidity boost” for target engagement, just as prior work did by multimerizing the HP1β CD [173]. However, while useful for nucleosome studies, this is not a necessity, and non-dimerizing epitope tags (e.g. MBP, 6HIS, FLAG, or HA) have also been successfully employed [28, 60–62, 173].

So, can we study everything now?

Combinatorial interactions are inherent to many CAPs, enabling robust and specific binding to their chromatin targets. These interactions typically involve two or more CAP moieties engaging distinct PTMs, histone sequence motif, DNA (by general charge, position on the nucleosome surface, sequence motif, or modification), nucleosome surfaces (e.g. the H2A/H2B acidic patch), or indeed other CAPs. Evaluating such complex interactions offers a formidable challenge, as the study platform(s) must be sufficiently sensitive and quantitative to isolate and evaluate individual contributions. Here, we describe easily adoptable biochemical and genomic platforms with such capabilities. Further, the tools now exist to create highly defined nucleoforms, including those containing histone mutants, PTMs, or combinatorials, in cis or trans, alongside DNA sequence motifs or modifications, and is stepping toward synthetic chromatin arrays that even more closely represent physiological targets [57, 174–176]: all in the service of dissecting biological mechanism with ever greater precision.

The new approaches will eventually hit a limit when interrogating multi-subunit CAP complexes and chromatin arrays: how many points of contact (cis/trans) can they reveal? There appears a lot of space to explore: Captify-Alpha can discern all partners in a trivalent interaction between PHIP/BRWD2 Tudor-BD1-BD2 and ([H3K4me3K14ac]₂•[H4K12ac]₂) nucleosomes, and indeed showed the impact of disease-associated mutations on chromatin binding [54]. The use of reader domains for genomics is particularly exciting, since tandems can find combinatorial PTM signatures; either native (e.g. BPTF PHD-BD as earlier) or by design (e.g. the DNMT3A-MPP8 PWWP-CD chimera that enriches {H3K9me3K36me3}) [171, 177, 178] that overlaid signals from PTM-specific antibodies can only infer. As such, the current technologies offer a highly accommodating starting point for future discoveries.

Supplementary data

Supplementary Data is available at NAR online.

Conflict of interest

EpiCypher is a commercial developer and supplier of reagents (e.g. PTM-defined semi-synthetic nucleosomes) and platforms (e.g. Captify and CUTANA) used in this study. All EpiCypher authors own shares in the company with J.R.B., M.W.C., and M.-C.K. also directors of same. EpiCypher holds patents related to technologies used in this study (#WO2019173565A1, #WO2020132388A1, and #WO2023159045A1) with M.R.M., E.N.W., B.J.V., Z.W.S., M.W.C., M.-C.K., and J.M.B. as listed inventors. N.L.K. serves as a consultant to Thermo Fisher Scientific and engages in entrepreneurship in the area of Top-Down Proteomics.

Funding

The Structural Genomics Consortium (SGC) is a registered charity (No. 1097737) that receives funds from Bayer AG, Boehringer Ingelheim, Bristol Myers Squibb, EU/EFPIA/OICR/McGill/KTH/Diamond Innovative Medicines Initiative 2 Joint Undertaking [EUbOPEN grant 875510], Genentech, Genome Canada through the Ontario Genomics Institute [OGI-196], Janssen, Merck KGaA (aka EMD in Canada and USA), Pfizer, and Takeda. This work was supported by the National Institutes of Health (NIH) through grant P41GM108569 for the National Resource for Translational and Developmental Proteomics at Northwestern University. A.S.L. is a trainee fellow under the Chemistry of Life Processes Predoctoral Training Grant (5T32GM105538-10) at Northwestern University. Work in the Musselman laboratory is funded by NIH grant R35GM128705. EpiCypher is supported by NIH grants R43GM134834, R44GM117683, R44GM145007, R44HG010595, R44CA214076, R44GM116584, R44GM119893, and R44HG010640.

Data availability

All MS files were uploaded as a MassIVE dataset with assigned identifier MSV000098235. https://massive.ucsd.edu/ProteoSAFe/dataset.jsp?task=2e5c7ad72b57417589747615c7bed1de

All CUT&RUN datasets are on the Gene Expression Omnibus with assigned accession number GSE249239

ncbi.nlm.nih.gov/geo/query/acc.cgi?acc = GSE249239