Synthetic Nucleosomes Reveal that GlcNAcylation Modulates Direct Interaction with the FACT Complex

Abstract Transcriptional regulation can be established by various post‐translational modifications (PTMs) on histone proteins in the nucleosome and by nucleobase modifications on chromosomal DNA. Functional consequences of histone O‐GlcNAcylation (O‐GlcNAc=O‐linked β‐N‐acetylglucosamine) are largely unexplored. Herein, we generate homogeneously GlcNAcylated histones and nucleosomes by chemical post‐translational modification. Mass‐spectrometry‐based quantitative interaction proteomics reveals a direct interaction between GlcNAcylated nucleosomes and the “facilitates chromatin transcription” (FACT) complex. Preferential binding of FACT to GlcNAcylated nucleosomes may point towards O‐GlcNAcylation as one of the triggers for FACT‐driven transcriptional control.

GlcNAc sites,H 2B-Ser112-O-GlcNAc has previously been associated with indirect transcriptional regulation through promotion of H2B-Lys120 ubiquitination that in turn facilitated transcriptional activation. [5c] We have previously characterized the effect of H2A-Thr101 GlcNAcylation on nucleosome structure; [6] H2A-Thr101 is located at the dimer-tetramer interface of the nucleosome and GlcNAcylation at this specific site leads to destabilization of the H2A/B dimer in the nucleosome.H owever,H 2B-Ser112 is remote from any such critical interface and so this mechanism cannot be readily invoked. In contrast to this structural modulation by PTMs,a nother mechanism by which other PTMs can establish afunctional output is through direct recruitment of chromatin "reader"/interactor proteins, [2b] yet such ad irect recruitment has not yet been observed for GlcNAcylation. Mass spectrometry (MS)-based proteomics provides an unbiased, powerful approach for both identification and quantification of such interactor proteins, [7] in contrast to traditional Western blotting methods. [8] Herein, we report the synthesis of modified nucleosomes containing aG lcNAcylation mimic at site H2B-S112. These synthetic nucleosomes were used for identification of "reader" proteins by MS-based interaction proteomics (Figure 1). These first direct interaction assessments in anucleosomal context suggest that, in fact, GlcNAcylation at H2B-S112 modulates the interaction between the nucleosome and the "facilitates chromatin transcript" (FACT) complex, ap ivotal histone chaperone complex.
Of the few described histone O-GlcNAcylation sites, H2B-Ser112 is particularly interesting being both ac ore domain modification and due to its proximity to the nucleosome "acidic patch" (Figure 2a), raising numerous potential mechanistic roles.The "acidic patch" is anegativelycharged binding interface on the nucleosome surface. [9] A number of proteins,s uch as LANA, [10] IL-33, [11] RCC1, [12] Sir3, [13] HMGN2, [14] RNF168, [15] and RING1B/BMI1, [15] as well as the H4 tail, [1] are known to interact competitively with the "acidic patch" leading to remodeling of chromatin structure;w es peculated that because of its proximity H2B-Ser112 O-GlcNAcylationc ould directly modulate binding. We tested this through the creation and characterization of aGlcNAcylated nucleosome.Although pulldown assays with certain modified nucleosomes have been used in conjugation with quantitative MS to study interacting nucleosomeprotein partners, [7,16] few [6] have investigated O-GlcNAcylation because of al ack of access to pure GlcNAcylated nucleosome.
Forsuch precise functional characterization of H2B-S112 O-GlcNAcylation, access to site-specifically GlcNAcylated protein is required. Access to homogeneous GlcNAcylated histones,and in general glycoproteins,remains abottleneck. [17] In vitro enzymatic O-GlcNAcylation of histone proteins by OGT typically leads to incomplete and heterogeneous product mixtures. [5c] We have previously reported as iteselective chemical protein modification strategy employing a" tag-and-modify" approach [18] for the generation of differently modified proteins,i ncluding GlcNAcylated histone H3. [19] Herein, by using this approach, we generated H2B-S112-GlcNAc protein for the reconstitution of site-specifically GlcNAcylatednucleosomes.
To access homogeneous GlcNAcylated H2B,w e expressed and purified ar ecombinant Ser112Cys (S112C) mutant of the Xenopus laevis H2B. [20] Ad ehydroalanine (Dha) "tag" was site-selectively installed at Cys112 by treating H2B-S112C with 2,5-dibromohexanediamide (DBHDA) [19a] under denaturing conditions.T he resulting H2B-S112Dha intermediate protein was reacted with GlcNAc-thiol to yield H2B with aG lcNAc mimic installed at S112 (Figure 2b)t hat bears at hioether linkage instead of the natural ether linkage;t he conformation of the glycosidic bond remains similar for such linkages. [21] We anticipated that the use of athioglycosidic linkage,which is more resistant to corresponding glycoside hydrolases (e.g.h OGA), would allow us to ensure homogeneity even inside enzymatically active cell lysates,t hereby enabling better precision as an affinity probe.P rotein characterization by LC-MS and LC-MS/MS confirmed the full conversion and site-specific GlcNAc installation (see Figures S1-S3 in the Supporting Information). Circular dichroism (CD) spectra of refolded H2B-S112-GlcNAc protein resembled that for wt H2B protein, indicating no gross change in secondary structure of the protein upon GlcNAcylation ( Figure S4).
GlcNAcylated H2B-S112 was used to constitute GlcNAcmodified H2A/H2B heterodimers by refolding with histone protein H2A;s imilarly GlcNAc-modified octamers were constructed by assembling GlcNAcylated H2B-S112 with histone proteins H2A, H3, and H4 ( Figure 1; Figure 2). The resulting multimer species were purified and analyzed by sizeexclusion chromatography (SEC). Similar SEC traces for Figure 1. Workflow for identification of H2B-S112 GlcNAc interactor proteins. H2B-S112-GlcNAc was synthesized using a"tag-and-modify" approach ( Figure 2b). The modified histone protein (along with other canonical histone proteins and biotinylated DNA (shown in gray)) were used for nucleosomereconstitution. Wildtype (wt) and the GlcNAcylated nucleosome were immobilized on magnetic beads (brown) via streptavidin-biotin affinity to enrich nucleosome-binding proteins. Pooled proteins from each sample were separatelyd igested and identified by nLC-MS/MS(nLC = nanoflow liquid chromatography). Label-free quantification (LFQ) was applied for quantification. H2A/B dimers are shown in red, the H3/H4 tetramer is shown in green, GlcNAc shown as ab lue square. The green square, dark-red oval, yellow circle, light-blue star,a nd gray triangle represent nucleosome-bindingproteins. Figure 2. a) Electrostatic potential view of the nucleosome. Blue indicates the positivelyc harged surface, whereas red denotes the negatively charged surface;inset:t he acidic patch with H2B-S112. b) Synthesis of GlcNAcylated H2B-S112. H2B-S112C was reacted (i)with DBHDA to generate H2B-S112Dha,w hich upon reaction with GlcNActhiol (ii)generated H2B-S112-GlcNAc. This was then used to reconstitute synthetically GlcNAcylated nucleosome as per Figure 1( see the SupportingInformation for full details).c )Native PAGE analysis of wt (lane ii)a nd GlcNAc-modified (lane iii)nucleosomer econstitution: Sybr Gold staining (left) and Coomassie Brilliant Blue (CBB) staining (right). Lane ishows the 100 bp DNA ladder.Image of the full gel is shown in Figure S9.
both modified dimer and octamer when compared to the wt species suggested that the histone fold of the H2A/B dimer and the dimer-tetramer interface were not disrupted upon introduction of GlcNAcylationa tH 2B-S112 ( Figure S5). Comparison of CD spectra and melting temperature measured by variable-temperature CD analysis of GlcNAcylated and wt H2A/B dimers revealed no significant changes in both the spectral profiles and melting temperatures (T m S112-GlcNAc dimer = 50.73 + /À0.27 8 8C; T m wt = 52.9 + /À0. 1 8 8C) upon GlcNAcylation at H2B-S112, suggesting little or no influence of the modification on the structure and stability of the dimers (Figures S6, S7). Nucleosome reconstitution was accomplished by the salt-gradient dialysis method using 145 bp DNAa nd biotinylated DNAc ontaining the strong "601" nucleosome positioning sequence. [16,22] Ther econstitution of nucleosome was analyzed on native PAGE gels (Figure 2c). Similar reconstitution yield and mobility on the PAGE gel for the modified nucleosome as compared to the wt nucleosome also suggested no major structural changes upon introduction of the modification, as expected for as urface-exposed site.R econstituted nucleosomes were further analyzed by CD spectroscopy.T he CD spectra for both the reconstituted nucleosomes were essentially identical and the melting temperatures measured by variabletemperature CD analysis were similar (at 220 nm: T m wt = 73.25 + /À0.  (Figure S10-S14). These combined data suggest that the differential stability mechanism observed for H2A-T101 GlcNAcylation [6] does not operate in H2B-S112 GlcNAcylation.
As represented in Figure 1, H2B-S112-GlcNAc-modified nucleosome (bait) and wt nucleosome (control) were immobilized on magnetic streptavidin beads.T hese were then incubated with HeLa cell nuclear extract for affinity enrichment of nucleosome-binding protein partners.N on-specifically bound proteins were removed by washing. Theenriched nucleosome-binding proteins were digested in-solution;t he resulting peptide mixtures were separated and analyzed by ultra-high performance LC (UHPLC) coupled to ah ybrid quadrupole-orbitrap (Q-Exactive) mass spectrometer.T o obtain ar obust data set, all pulldown experiments were performed as three independent biological replicates. MaxLFQ [23] (a MS-based label free quantitation (LFQ) algorithm) analyses allowed us to identify and quantify 584 protein groups.E xperimental correlation among LFQ intensities within group and across replicates were monitored using the Pearson correlation coefficient ( Figure S15). False discovery rate (FDR) based t-test statistical analyses revealed FACT subunits,s uppressor of Ty (SPT16), and structure specific recognition protein 1( SSRP1) as being both the statistically most significant and the most enriched interacting protein partners for H2B-S112-GlcNAc-modified nucleosome ( Figure 3; see also Table S1 and Figure S16).
TheF ACTcomplex is areplication [24] and transcription [25] factor functioning in various chromatin processes.F ACTi s ah eterodimer protein complex consisting of two subunits (SSRP1 and SPT16) and displays histone chaperone activ-ity. [26] FACT plays an important role in chromatin remodeling by modulating nucleosome stability and thereby DNA accessibility.F ACTc an interact with multiple sites on the nucleosome [27] and can decrease the nucleosomal barrier allowing productive transcriptional elongation, primarily by chaperoning the H2A/B dimer. [28] Interestingly,w eh ave previously shown that H2A-T101 GlcNAcylation can destabilize H2A/B dimers in nucleosomes. [6] We proposed this destabilization as as eparate structural mechanism to facilitate transcriptional elongation in am anner that is complementary to FACT recruitment. In addition, FACT is known to stimulate and function cooperatively with H2B-K120 monoubiquitination, [29] which has previously been suggested to be S112GlcNAc-dependent in the regulation of transcriptional elongation. We did not observe the ligase responsible for H2B-K120 ubiquitination BRE1A/B as suggested by Fujiki et al. [5c] in our pulldowns both by LC-MS and Western blot ( Figure S17), suggesting that this interaction is not strong enough to be detected under these conditions (see the Supporting Information). Since FACT has been reported to associate with both nucleosomes and separate histone proteins,w ew anted to investigate how the presence of different components influences the FACTwith-GlcNAc interaction. Fort his,w eperformed pulldowns with recombinant FLAG-tagged Xenopus laevis H2B containing the chemically installed S112-GlcNAcylation, both using isolated GlcNAcylated-H2B protein and also ar econstituted GlcNAcylated-H2A/B dimer in asimilar way to that tested with the nucleosome ( Figure S20). This experiment required the construction of anew H2B protein substrate bearing asuitable affinity motif for retrieval. Thus, N-terminally FLAG-tagged wt-H2B (FLAG-H2B) and the corresponding S112C mutant (FLAG-H2B-S112C) were designed, expressed, and purified in as imilar manner to before.Essentially identical two-step GlcNAcylation chemistry (Figure 2b)t othat used for H2B-S112C (Figure 1) also proved equally robust and successful for GlcNAcylation of FLAG-H2B-S112C (See the Supporting Information and Figures S21, S22). FLAG-tagged wt H2B and H2B-S112-GlcNAc monomers were refolded and, again, CD analyses revealed no significant structural changes upon modification ( Figure S23).
Thec orresponding interactomes were analyzed using quantitative MS,a se arlier.T he H2B monomers were immobilized on anti-FLAGm agnetic beads and used to affinity enrich the interacting partners from nuclear extract (prepared under non-reducing conditions,see the Supporting Information);experiments were performed in duplicate.MSbased proteomics allowed us to identify and quantify 948 proteins,amongst which FANCI, INF2, and COAX6A1 were the most significant interacting protein partners for H2B-S112-GlcNAc protein as compared to wt-H2B.(see Table S2, Figure S24). Notably,wedid not see asignificant enrichment of FACT upon H2B-S112 GlcNAcylation. We were also not able to detect BRE1A in these experiments,a lthough we do detect BRE1A by MS in the nuclear extract. Next, FLAGtagged H2B-S112-GlcNAc and FLAG-tagged wt-H2B were each combined with wt-H2A protein to reconstitute FLAGtagged, wt, and GlcNAcylated H2A/B heterodimers,e ssentially as before.A gain using quantitative MS,w ei dentified 886 interacting proteins in both samples (see Table S3 and Figure S25). As with the interactome data with the monomer, we did not see any significant enrichment of FACT upon GlcNAcylation and no BRE1A was observed. Together these data suggest ac ontext-dependent interaction;t hus,t he difference in our observations here to those published previously [5c] appears to lie in our use of an intact nucleosomal structure as opposed to the prior use of isolated, partially GlcNAcylated protein (this might also be due to preferential binding of isolated H2B by different histone chaperones, occluding the GlcNAcylation site).
Based on our findings,w ep ropose ap ossible updated mechanism facilitating ubiquitination of H2B-Lys120 upon GlcNAcylation at H2B-Ser112 (Figure 4). This speculative mechanism is the simplest that is consistent with the data gathered here,a lthough, of course,o thers cannot be discounted. GlcNAcylation of histone H2B at Ser112 by OGT leads to FACT association. FACT recruitment results in nucleosome remodeling making the H2A/B dimer accessible for BRE1A ubiquitination. FACT can in turn directly or indirectly recruit BRE1A complex (RNF20/40) facilitating ubiquitination of H2B-Lys120. [29,30] In agreement with this, FACT is required in vivo [30] for BRE1A/B localization to chromatin in DNAd amage responses.I na ddition, many residues in the acidic patch have been shown to be essential for H2B ubiquitination, [31] suggesting that this surface is important for the anchoring of BRE1A/B.Strong enrichment of FACT complex may also point towards GlcNAcylation as at rigger for FACT-driven transcription processes as well as a" relaxed" chromatin state facilitating transcription elongation. While we expected to find BRE1A/B in our pulldown experiments,alack thereof might be explained by the previous observation that both active transcription (dependent on the presence of dNTPs) and FACT is necessary for H2B-K120 ubiquitination. It might be that even aF ACTbound nucleosome is not sufficient for BRE1A/B association and that further structural changes are required.
In summary,b yu sing a" tag-and-modify" approach we have synthesized GlcNAcylated H2B histones and corresponding nucleosomes.C oupled with interaction proteomic analyses,w ew ere able to explore the mechanistic details of atranscriptionally relevant segment involving O-GlcNAcylation. In contrast to H2A-T101 GlcNAcylation, H2B-S112 GlcNAcylation does not affect nucleosome assembly,b ut directly influences the nucleosome interactome,h ighlighting different possible signaling mechanisms for histone GlcNAcylation. It is also important to note that we use here ad esigned nonhydrolyzable mimetic that despite the anticipated similarity [21] might give rise to unexpected artefacts.The chemical synthetic approach used here can be employed in principle for the generation of differently (and multiply) modified nucleosomes to complement other biochemical approaches and/or other powerful, multiplexed methods, such as those achievable by,f or example,e xpressed protein ligation [32] or native chemical ligation. [33] These synthetic nucleosomes in combination with MS-based proteomics can elucidate the role of various PTMs as well as revealing "crosstalk" between PTMs.I nt urn, we anticipate that they will allow elucidation of the key players to create ap recise mechanistic picture of this biology at the molecular level. The field of chromatin biology has relied heavily on short peptides and isolated proteins that are mere fragments of true nucleosomal contexts;t he work presented herein, as well as other reports, [34] suggests that the use of suitably complex probe molecules that provide the correct context may prove vital for relevant interrogation.