Dual engagement of the nucleosomal acidic patches is essential for deposition of histone H2A.Z by SWR1C

Summary The SWR1C chromatin remodeling enzyme catalyzes the ATP-dependent exchange of nucleosomal histone H2A for the histone variant H2A.Z, a key variant involved in a multitude of nuclear functions. How the 14-subunit SWR1C engages the nucleosomal substrate remains largely unknown. Numerous studies on the ISWI, CHD1, and SWI/SNF families of chromatin remodeling enzymes have demonstrated key roles for the nucleosomal acidic patch for remodeling activity, however a role for this nucleosomal epitope in nucleosome editing by SWR1C has not been tested. Here, we employ a variety of biochemical assays to demonstrate an essential role for the acidic patch in the H2A.Z exchange reaction. Utilizing asymmetrically assembled nucleosomes, we demonstrate that the acidic patches on each face of the nucleosome are required for SWR1C-mediated dimer exchange, suggesting SWR1C engages the nucleosome in a “pincer-like” conformation, engaging both patches simultaneously. Loss of a single acidic patch results in loss of high affinity nucleosome binding and nucleosomal stimulation of ATPase activity. We identify a conserved arginine-rich motif within the Swc5 subunit that binds the acidic patch and is key for dimer exchange activity. In addition, our cryoEM structure of a Swc5-nucleosome complex suggests that promoter proximal, histone H2B ubiquitinylation may regulate H2A.Z deposition. Together these findings provide new insights into how SWR1C engages its nucleosomal substrate to promote efficient H2A.Z deposition.


Introduction
Eukaryotic genomes regulate access to information stored in their genetic code through the dynamic nucleoprotein structure known as chromatin. The basic unit of chromatin, the nucleosome, sequesters ~147 base pairs (bp) of DNA around a histone octamer consisting of a H3-H4 tetramer flanked by two H2A-H2B (AB) hetero-dimers 1 . Rather than a static structure, chromatin is highly dynamic, due to the coordinated impact of histone post translational modifications, ATP-dependent chromatin remodeling enzymes, and the incorporation of histone variants 2 . Disruption of these processes can have deleterious consequences for all nuclear events, including transcription, DNA replication, and genome stability pathways.
H2A.Z, a variant of the core histone H2A, is enriched within nucleosomes adjacent to genes transcribed by RNA polymerase II, replication origins, centromeres, and at sites of DNA double strand break repair [3][4][5][6][7] . In yeast, H2A.Z enrichment correlates with higher rates of nucleosome turnover that facilitates transcription, as well as antagonizing the spread of heterochromatic regions from telomeres and the silent mating type loci [8][9][10] . Recently, we found that H2A.Z plays a global, positive transcriptional role in yeast strains lacking the nuclear RNA exosome 11 . While yeast H2A.Z is not essential for viability, H2A.Z is essential in metazoans where it is associated with both transcriptional activation and repression 12 . For instance, in embryonic stem cells, deposition and function of H2A.Z is interdependent with the polycomb repressive complex 2 (PRC2), and loss of either H2A.Z or PRC2 leads to transcriptional de-repression of a host genes and a subsequent failure in lineage commitment 13 .
The site-specific deposition of H2A.Z is catalyzed by ATP-dependent chromatin remodeling enzymes related to the 14-subunit, yeast SWR1C complex 14,15 . Mammals contain two enzymes related to SWR1C, the SRCAP and Tip60/p400 complexes, and subunits of the mammalian enzymes are highly conserved with those of SWR1C 16,17 . Each of these enzymes are members of the larger INO80 subfamily of chromatin remodeling enzymes, distinguished by a large 4 insertion domain between the two ATPase lobes of the catalytic ATPase (Swr1 in SWR1C) 2,18,19 .
This insertion domain serves as a docking site for several key subunits of SWR1C-like complexes, including the Rvb1/Rvb2 heterohexameric ring (RUVB1/RUVB2 in humans) which acts as a further scaffold for organizing additional subunits 20,21 . In addition to Rvb1/Rvb2, previous studies have demonstrated that the deposition of H2A.Z by SWR1C requires several key subunits, including Swc2, Swc4, Yaf9, and Swc5 20,22 . The Swc2 subunit and its mammalian ortholog, Yl-1, promotes nucleosome binding and assists SWR1C in the recognition of nucleosome free regions adjacent to gene promoter regions 20,23,24 . Swc4 (DMAP1 in mammals) is also part of the yeast NuA4 histone acetyltransferase complex, and it appears to promote DNA and histone binding 25,26 .
Swc5 (Cfdp1 in mammals) is essential for the ATPase activity of SWR1C, but how Swc5 promotes SWR1C activity is poorly understood 27 .
The subunit complexity of chromatin remodelers varies from the single subunit, yeast CHD1, to multi-subunit ~1MDa enzymes such as RSC, SWI/SNF, INO80C, and SWR1C. Despite differences in function and complexity, recent studies have shown that the activity of many remodelers require that they engage a solvent exposed, acidic surface on the nucleosome [28][29][30] . This "acidic patch" is composed of eight residues from histones H2B and H2A that are known to provide a docking interface for many proteins. For instance, PRC1, RCC1, and Orc1 bind to the nucleosome acidic patch through a common binding motif consisting of a loop region with an arginine residue that inserts directly into the acidic pocket [31][32][33] . Though these regions are primarily unstructured and basic, nucleosomal docking requires the specific side chain structure of an arginine, which cannot be substituted by lysine, resulting in these regions being referred to as "arginine anchors" 34,35 . In the context of remodelers, ISWI uses an arginine anchor within its catalytic ATPase subunit to relieve an autoinhibitory mechanism following nucleosome binding. Members of the SWI/SNF family of enzymes also use arginine-rich motifs within both the conserved Snf5-like and Snf2-like subunits to sandwich the complex to the octamer during remodeling activity [36][37][38] . Recent cryoEM structures of INO80 family members have predicted key interactions with the nucleosomal acidic patch 39 , and this domain has been found to play a crucial role in remodeling activity 29 .
Here we investigate the role of the nucleosome acidic patch in SWR1C dimer exchange.
Through Förster-resonance energy transfer (FRET), fluorescence intensity (FI), and fluorescence polarization (FP) assays, we interrogate how SWR1C is influenced by H2A/H2B acidic patch alterations, and we specifically probe how each acidic patch on the two faces of the nucleosome contribute to SWR1C activity. Surprisingly, we find that loss of even a single acidic patch results in the loss of dimer exchange activity, suggesting SWR1C must engage both acidic patches simultaneously in a "pincer-like" fashion to carry out dimer exchange. In addition, we found that both linker DNA and the contralateral incorporation of an H2A.Z/H2B dimer stimulated H2A.Z deposition. Efforts to identify potential SWR1C subunits responsible for acidic patch interactions led us to identify an arginine rich motif within the Swc5 subunit of SWR1C, and we show that this domain is essential for dimer exchange in vitro, the function of SWR1C in vivo, and Swc5 binding to the nucleosome. A cryoEM structure of the Swc5-nucleosome complex confirms an acidic patch interaction, as well as contacts with the histone H4 N-terminal tail and the H2B C-terminal helix.
These results indicate that SWR1C activity relies on communication between the Swc5 arginine motif and the nucleosome acidic patch to functionally engage nucleosomes for dimer exchange activity.
Spot assays were performed with the swc5∆ strain (Y03371) transformed with an empty CEN ARS URA3 vector PRS416, or ones containing WT and mutant SWC5. Cultures were diluted to an OD600 of 1 and then serially diluted 1:10. A dilution (7 μL) was spotted onto synthetic complete media lacking uracil with and without 2% formamide. Plates were incubated for 3 days before imaging.

Plasmid Construction
The vector (pQE80L) was used for recombinant Swc5 expression. The SWC5 gene and SWC5  were inserted in frame with the N-terminal 6x histidine tag through Gibson assembly 41 after digesting the vector with BamHI and SphI. Once WT and truncated Swc5 were in frame, the alanine mutants [swc5 (RRKR-4A) and swc5 (LDW-3A)] were generated using QuickChange site directed mutagenesis. The pRS416 CEN ARS URA3 vector was digested with XbaI and SacI to allow for insertion of the SWC5 gene with 500bp upstream of the start site and 500bp downstream of the stop. This region was initially PCR amplified from a W303 strain where the primers had 7 homology to the XbaI and SacI cut sites of the vector. Gibson assembly was used to ligate the vector and insert (CP1566) 41 . Once the WT SWC5 gene was inserted, site directed mutagenesis was used on CP1566 to make swc5 RRKR-4A (CP1579) and swc5 LDW-3A (CP1580). swc5 79-303 construct was generated using CP1566 to PCR amplify the 500bp upstream stretch of DNA and start site of SWC5 and PCR amplifying another DNA fragment starting from Swc5 K79 to 500bp downstream of the stop. Primers were designed so each PCR fragment had homology to each other and the pRS416 vector at XbaI and SphI so that the two fragments and the pRS416 vector could be ligated together by Gibson assembly. Site directed mutagenesis was used to generate cysteine mutants used for Oregon Green labeling, as well as for single amino acid substitution used for assembly of acidic patch mutant (AB-apm) dimers -H2A-apm (H2A-E63A, H2A-E65A, H2A-D74A, H2A-D92A, H2A-D93A, H2A-E94A) and H2B-apm (H2B-E109A, H2B-E117A).

Recombinant Swc5
Swc5 and variants were first cloned into a pQE80L expression vector in frame with a six histidine N-Terminal tag through Gibson assembly. Plasmids were transformed into Rosetta 2 DE3 competent cells (Novagen). 1L cultures of 2x YT (Yeast extract, tryptone) were grown at 37˚C to OD600 of 0.5-0.7 before adding 0.4 mM IPTG. After adding IPTG, cells continued to shake at room temperature for 3 hours. Cells were harvested by centrifugation at 4˚C and 1L of Swc5 bacterial pellet was resuspended in 20 mL lysis buffer (10 mM Imidazole, 50 mM Tris pH 8.0, 300 mM NaCl, 1 mM PMSF, 1 mM DTT) before being flash frozen in liquid nitrogen and stored at -80˚C.
Pellets were thawed in a 37˚C water bath then fresh 0.5 mM PMSF and 1 mM Benzamidine were added. Cells were sonicated 5x 30 seconds at 15% power, incubating on ice for 1 min between sonication intervals. Lysate was then centrifuged at 14000 rpm in a JA-17 rotor at 4˚C for 25 min.
Whole cell extract was incubated with 800 μL of HisPur Ni-NTA resin slurry, that had been prewashed with lysis buffer, for 2 hours at 4˚C on nutator. The resin and whole cell extract mixture was centrifuged at 750 rcf for 5 minutes. The supernatant was aspirated, and the resin was placed in a 25 mL gravity column and washed with 5x 10 mL wash buffer (500 mM NaCl, 10 mM Imidazole, 50 mM Tris pH 8.0). Swc5 was eluted by capping the gravity column and adding 400 μL elution buffer (500 mM NaCl, 300 mM Imidazole, 50 mM Tris pH 8.0) to resin and incubating at 4˚C for 10 min then collecting the flow through. The Swc5 eluate was dialyzed against 2x 250 mL of storage buffer (50 mM NaCl, 50 mM Tris pH 7.4, 10% glycerol) then flash frozen and stored at -80˚C. Concentration was determined by BSA standard on a SDS PAGE gel.

Nucleosome reconstitution
These experiments utilized a variety of nucleosomes, both in terms of histone variants, labels, and mutations. Broadly, all nucleosomes contained a tetramer of Xenopus laevis H3 and H4, along with Saccharomyces cerevisiae H2A/H2B heterodimers, with mutations or variant substitutions as appropriate, reconstituted on DNA containing the Widom 601 positioning sequence with or without linkers and/or fluorescent labels. We decided to use Xenopus laevis H3/H4 tetramers due to slightly increased stability of nucleosomal product in solution. These histones show high sequence conservation with Saccharomyces cerevisiae H3 and H4 and we have previously demonstrated there is no difference in SWR1C eviction activity on full yeast or hybrid yeast-xenopus nucleosomes 45 . Furthermore, the H4 tail region that interacts with Swc5 in our cryo-EM models is virtually identical in the orthologs.
For experiments using nucleosomes with symmetrical dimers, hybrid octamers were generated as previously described and nucleosomes were reconstituted via salt dialysis deposition on a Widom 601 positioning sequence containing DNA template 46,47 .
Asymmetrical nucleosomes were produced by first reconstituting Xenopus laevis H3/H4 tetramers as well as Saccharomyces cerevisiae dimers of interest separately instead of generating octamers.
Tetramers, a single dimer type, and template DNA were mixed in a 1.4:1.8:1.0 ratio followed by reconstitution via salt dialysis identically to regular nucleosome reconstitution. This ratio was found to reliably produce a mixture of hexasomes and nucleosomes, and in the case of hexasomes the dimer is known to be deposited on the "strong" side of the Widom position sequence 48 . For asymmetric nucleosomes containing a single dimer labeled with Cy5, the template DNA was varied such that the labeled dimer is deposited on the strong side of the 601 positioning sequence to eliminate weak versus strong side differences. Hexasomes are then purified using a Bio-Rad MiniPrep Cell apparatus, a method by which proteins are collected via fraction collector after passage through a native-PAGE tube gel, which is sufficient to separate hexasomes from nucleosomes [48][49][50] . After checking fractions on a native-PAGE gel stained with SYBR Gold, the hexasome containing fractions were pooled and concentrated, followed by buffer exchange into remodeling buffer (25 mM HEPES, pH 7.6, 0.2 mM EDTA, 5 mM MgCl2, 70 mM KCl, 1 mM DTT) with 20% glycerol, aliquoting, and flash-freezing. The introduction of a Saccharomyces cerevisiae heterodimer to hexasomes in solution leads to spontaneous incorporation and formation of a nucleosome, allowing for production of asymmetric nucleosomes with known orientation.
Six types of asymmetric nucleosomes for dimer exchange assays were produced in this manner containing linker-proximal or distal H2A-Cy5/H2B (AB-Cy5) heterodimers with placement of unlabeled AB, ZB, or AB-apm dimers on the contralateral side on a 77N0-Cy3 template DNA.
For fluorescence polarization and ATPase studies, unlabeled dimers were used to produce hexasomes containing linker-distal AB or AB-apm dimers on 77N0-Cy3. These hexasomes were then used to generate nucleosomes with dimers (listed here and throughout the manuscript in linker proximal/linker distal order) AB/AB, AB/AB-apm, AB-apm/AB, or AB-apm/AB-apm, all on the 77N0-Cy3 template.

Nucleosome dimer exchange assays
FRET based dimer exchanges assays were performed as previously described 43 using an ISS PC1 spectrofluorometer, a Tecan Infinite M1000 PRO microplate reader, or a Tecan Spark microplate reader. The nucleosome remodeling reactions (50-100 μL) were performed in remodeling buffer at room temperature with 30-60 nM SWR1C, 10 nM 77N0-Cy3 and at least one dimer-Cy5 containing nucleosome, 60-70 nM ZB dimers, and 1 mM ATP (to start reaction). Dimer exchange was observed by exciting the Cy3 fluorophore at 530nm and monitoring the Cy5 emission 670. Data for nucleosomes with both dimers labeled was fit to the two-phase decay equation (Equation 1; where Y denotes RFU, ! the initial RFU, " the final RFU, # the fast phase expressed as a fraction, $ the slow phase expressed as a fraction, # the rate constant of the fast phase, $ the rate constant of the slow phase, and t is time) while data for reactions with only one dimer labeled were fit to the one-phase decay equation (Equation 2; where Y denotes RFU, ! the intial RFU, " the final RFU, k is the rate constant, and t is time).

(Equation 1)
Gel based dimer exchange assays used a modified protocol from a previous study 15,23 .

Fluorescence polarization assay
Fluorescence polarization assays were performed as previously described 51  Data was processed with the cryoSPARC v4 suite (Supplemental Figure 1) 56 . Super-resolution movies were motion-corrected with "Patch Motion Correction" and binned using 2x Fourier cropping. CTF values were estimated with "Patch CTF Estimation". Initial particles were picked with "Blob Picker" from a subset of micrographs, then used for ab-initio 3D reconstruction, and the best obtained model was used to generate templates for particle picking with "Template Picker".
Template-picked particles were refined into 3 classes using "Heterogenous Refinement" with the previously generated ab initio reconstructions as initial volumes to remove junk particles. 1,846,495 nucleosomal particles were further filtered to remove low-quality particles using the "random-phase 3D classification" strategy 57 . Briefly, particles were sequentially subjected to multiple rounds of Heterogeneous Refinement against a low-pass filtered reconstruction (40, 30, 25, 20, 15, 10, 8 Å low-pass) of the output volume from the previous round. This strategy yielded two classes of highquality particles with the nucleosomal DNA either 1) completely wrapped around the core (585,573 particles, Non-uniform Refinement to 3.0-Å resolution), or 2) partially peeling off (392,348 particles, 3.1 Å). The Swc5 moiety is extremely noisy in these reconstructions due to apparent structural heterogeneity. We tried employing various methods including "3D Variability Analysis" 58 to either further classify the particles into more homogenous subsets or to describe the intrinsic conformational dynamics, but with limited success (data not shown). We obtained the most informative results with cryoSPARC's algorithm of "3D Classification" without alignment. The 585k particles of fully wrapped nucleosome were classified into 10 classes using a mask focused on the sides of the histone core in the regions with Swc5 moiety apparent. The obtained particle classes were refined to ~3.1 Å using "Non-uniform Refinement" 59 . Modeling was performed with UCSF ChimeraX 60 .

SWR1C dimer exchange requires the nucleosomal acidic patch
To investigate the role of the nucleosomal acidic patch on SWR1C dimer exchange activity, is monitored by following the decrease in the 670 nm FRET signal due to eviction of the Cy5labeled AB-Cy5 dimer ( Figure 1A).
SWR1C dimer exchange reactions were performed under single turnover conditions (excess enzyme to nucleosomal substrate) and contained free H2A.Z/H2B (ZB) dimers which act as an essential co-substrate. Addition of SWR1C to a wildtype nucleosome led to a rapid drop in FRET, showing the biphasic kinetics consistent with the sequential exchange of the two nucleosomal H2A/H2B (AB) dimers 43,61 . Incubation of SWR1C with a nucleosome that lacks an intact acidic patch (AB-apm/AB-apm nucleosome) showed little decrease in the FRET signal, indicating that an intact nucleosomal acidic patch is essential for SWR1C to catalyze H2A/H2B eviction ( Figure   1B).
Since the FRET assay is limited to monitoring the eviction of nucleosomal H2A/H2B dimers, a gel-based assay that monitors deposition of H2A.Z was employed to confirm these findings 19 . In this assay, an unlabeled, 77N0 nucleosome was incubated with SWR1C, ATP, and free H2A.Z/H2B dimers in which H2A.Z contains a 3xFLAG tag at its C-terminus. Reaction products are separated on native PAGE, and formation of the heterotypic (AB/ZB) and homotypic (ZB/ZB) nucleosomal products are detected by their reduced gel migration due to the 3xFLAG tag on H2A.Z. Similar to the FRET-based assay, the SWR1C-catalyzed deposition of H2A.Z was largely complete by 30-60 minutes for the AB/AB nucleosomal substrate, but addition of SWR1C to the AB-apm/AB-apm nucleosome led to 10-fold reduction in H2A.Z deposition after 60 minutes (Supplemental Figure 2A, B, D). Together, these assays indicate that SWR1C requires an intact nucleosomal acidic patch for H2A.Z deposition.
To further interrogate the role for each of the two nucleosomal acidic patches in SWR1C dimer exchange activity, we performed dimer exchange reactions with asymmetrically assembled nucleosomal substrates. Such asymmetric nucleosomes allow the fluorescent labelling of only a single H2A/H2B dimer, such that we could monitor which dimer was targeted for replacement. We generated these substrates by leveraging the asymmetric features of the 601 nucleosome positioning sequence and the spontaneous incorporation of dimers into purified hexasomes 49 (Supplemental Figure 3A). Hexasomes were generated by varying the ratio of Xenopus laevis H3/H4 tetramers to yeast H2A/H2B heterodimers on 77N0, 0N77, or 0N0 DNA that was labeled with Cy3 (Supplemental Figure 3B). After hexasome purification, nucleosomes of different compositions were reconstituted by titration of the appropriate yeast heterodimer (Supplemental Figure 3C, D).
For nucleosomes with linker DNA, we list the asymmetric dimers in linker proximal, linker distal order throughout (e.g. AB/AB-apm for a nucleosome with a linker distal H2A/H2B dimer harboring the acidic patch alteration).
First, we monitored H2A.Z deposition with nucleosomes harboring an unlabeled H2A/H2B heterodimer with a disrupted acidic patch (AB-apm) assembled on the nucleosomal face opposite to the Cy5-labelled heterodimer that will be monitored for H2A.Z deposition (Supplemental Figure   4G,H). Strikingly, H2A.Z deposition was eliminated, irrespective of whether the linker distal or linker proximal H2A/H2B heterodimer was the target for exchange ( Figure 1C, AB-apm/AB-Cy5 and AB-Cy5/AB-apm substrates). We then tested if the Cy5-labelled heterodimer that is targeted for replacement also requires an intact acidic patch (Supplemental Figure 4I). In this case as well, H2A.Z deposition was eliminated ( Figure 1C, AB/AB-apm-Cy5 substrate  Figure 1C, compare AB/AB-Cy5 to AB-Cy5/AB; Supplemental Figure 4A,B). To rule out the possibility that this preferential exchange was influenced by the unlabeled H2A/H2B dimer on the opposing nucleosomal face, nucleosomes harboring an unlabeled ZB dimer were also reconstituted. In these cases as well, the linker distal H2A/H2B dimer was exchanged at a more rapid rate, compared to the linker proximal heterodimer ( Figure 1C,  Figure 4C,F). Interestingly, these assays also confirmed that linker DNA has a large stimulatory impact on the H2A.Z deposition reaction, with both linker distal and linker proximal eviction occurring at a faster rate than eviction from a nucleosome core particle 23 . To test whether the stimulation by linker DNA was due to differences in nucleosome binding affinity, rates of H2A.Z deposition were measured at varying SWR1C concentrations. At high levels of SWR1C, the stimulatory impact of linker DNA remained, confirming that reactions are performed under singleturnover conditions (Supplemental Figure 5A). Thus, both linker DNA and nucleosomal ZB dimers appear to stimulate H2A.Z deposition.

The nucleosomal acidic patch is a key driver of SWR1C binding affinity
To further investigate the role of the nucleosomal acidic patch for SWR1C-mediated dimer exchange, we explored how disruption of each acidic patch impacted the nucleosome binding affinity of SWR1C for nucleosomes, leveraging a fluorescence polarization assay. Using a Cy3labeled 77N0 nucleosome, we determined a Kd of 13.6 nM for an AB/AB nucleosome (Figure 2A), similar to prior studies 61 . However, disruption of even a single nucleosomal acidic patch severely weakened SWR1C binding, such that a Kd was unable to be determined ( Figure 2B-D). Given the range of SWR1C concentrations that could be used in this assay, we estimate that alteration of one or both nucleosomal acidic patches decreases binding affinity by at least 10-fold.

Intact nucleosomal acidic patches are required for SWR1C ATPase activity
The ATPase activity of SWR1C is stimulated by both its nucleosomal substrate and free ZB dimers 19 . Notably, the ATPase activity of SWR1C is not stimulated by a ZB/ZB nucleosome, implying that SWR1C can distinguish product from substrate 19 . To investigate the role of the nucleosomal acidic patch in the stimulation of SWR1C ATPase activity, SWR1C ATPase activity was measured in the presence or absence of different 77N0-Cy3 nucleosomes. Consistent with previous results 19 , SWR1C ATPase activity was stimulated by AB/AB nucleosomes as compared to the basal level. However, ATPase activity was not stimulated by either of the asymmetric acidic patch mutant nucleosomes (AB-apm/AB, AB/AB-apm) or by a nucleosome lacking both acidic patches (AB-apm/AB-apm), demonstrating the essential role of dual acidic patch engagement by SWR1C for stimulation of ATPase activity (Supplemental Figure 5B).

Swc5 interacts with the nucleosomal acidic patch
Recently, a potential arginine anchor domain, the archetypal motif for proteins interacting with the nucleosomal acidic patch, was identified within the N-terminal, acidic domain of the SWR1C subunit, Swc2. However, removal of this domain (Swc2-ZN) has only a modest impact on the dimer exchange activity of SWR1C, and this data, combined with the essential nature of the nucleosomal acidic patches for SWR1C activity, suggest that additional subunits must contribute to acidic patch recognition 62 . The Swc5 subunit is one candidate, as loss of Swc5 eliminates H2A.Z deposition and nucleosome-stimulated ATPase activity 20 (See Figure 3B, C). Furthermore, removal of the Swc5 subunit from SWR1C crippled nucleosome binding, as measured by fluorescence polarization ( Figure 3D). Previous studies of Swc5 identified an acidic N-terminal domain that interacts preferentially with AB dimers, and a conserved C-terminal domain, termed BCNT, that is essential for SWR1C dimer exchange activity 27,63 . An alignment of Swc5 homologs revealed a region adjacent to the BCNT domain, containing multiple arginine residues in a conserved basic region ( Figure 3A).
To investigate the role of this Swc5 arginine-rich domain, alanine substitutions were created in four basic residues (RRKR), and this derivative was recombinantly expressed and reconstituted into a SWR1C complex purified from a swc5 deletion strain (SWR1C swc5∆ ). In addition, SWR1C was reconstituted with wildtype Swc5, a Swc5 derivative lacking the acidic N-terminal domain (Swc5 79-303 ), and a Swc5 derivative harboring alanine substitutions in the essential BCNT domain (Swc5 LDW-3A ). In all cases, derivatives were incorporated into SWR1C complexes with equal efficiencies (Supplemental Figure 6). Consistent with previous studies, reconstitution with wildtype Swc5 fully restored dimer exchange activity to SWR1C swc5∆ ( Figure 3B, red curve). In contrast, SWR1C harboring Swc5 LDW-3A had no detectable exchange activity, and the complex that lacks the N-terminal, acidic domain of Swc5 (Swc5 79-303 ) exhibited a modest defect, as expected from previous studies ( Figure 3B) 27 . Strikingly, SWR1C that contained the Swc5 RRKR-4A derivative showed minimal activity in the FRET-based exchange assay ( Figure 3B), and this complex exhibited only ~10% the activity of wildtype SWR1C in the gel-based H2A.Z deposition assay (Supplemental Figure 2A, C, D). Dimer exchange activity for each SWR1C complex was also mirrored by their ATPase activity ( Figure 3D). SWR1C swc5∆ lost nucleosome-stimulated ATPase activity, with or without free ZB dimers, as compared to wild-type. As expected, reconstitution of SWR1C swc5∆ with recombinant Swc5 rescued ATPase activity. Underscoring the importance of the arginine-rich domain, reconstitution of SWR1C swc5∆ with Swc5 RRKR-4A showed a large defect in nucleosome-stimulated ATPase activity, revealing its importance in both dimer exchange and ATPase activity. Additionally, we reinforced prior findings that the ATPase activity of the Swc5 LDW-3A complex was not stimulated by either nucleosomes or free ZB dimers, while the complex harboring Swc5 79-303 had wildtype levels of ATPase activity (Supplemental Figure 5C).
To further investigate whether the arginine-rich domain within Swc5 binds to the nucleosomal acidic patch, in vitro binding assays were performed with nucleosome core particles (0N0) and recombinant Swc5 ( Figure 4A-F). Increasing amounts of Swc5 were incubated with nucleosomes, and binding was visualized by native PAGE. Addition of wildtype Swc5 to nucleosomes led to formation of discrete complexes with an apparent Kd of 125.7 nM ( Figure 4A).
Likewise, binding of Swc5 LDW-3A , which has a disrupted BCNT domain, bound to nucleosomes with nearly identical affinity, compared to wildtype Swc5 ( Figure 4B). Importantly, high affinity binding of Swc5 to nucleosomes required an intact acidic patch, with binding reduced ~3-fold when an APM nucleosome was used as the substrate ( Figure 4C). Furthermore, the Swc5 RRKR-4A derivative also showed an ~3-fold loss in affinity, with an apparent Kd of 379.1 nM ( Figure 4D).
Together these data demonstrate that Swc5 has nucleosome binding activity, and the data are consistent with direct interactions between the Swc5 arginine-rich domain and the nucleosomal acidic patch.
While the gel shift assays establish that Swc5 can bind to nucleosomes, data from these assays can only indirectly link the nucleosomal acidic patch to the Swc5 arginine-rich domain. To probe for more direct interactions, a fluorescence quenching assay was employed. These assays exploit nucleosomes harboring a site-specific, Oregon Green fluorophore whose emission is sensitive to the chemical environment whereby fluorescence is quenched by protein binding 32,64 .
Nucleosome core particles were reconstituted that contained an Oregon Green fluorophore covalently attached to either the histone H4 N-terminal domain (H4-tail) ( Figure 5A, top) or to a residue directly adjacent to the nucleosomal acidic patch ( Figure 5A, bottom). Titration of wildtype Swc5 led to the concentration-dependent quenching of the Oregon Green fluorophore positioned at the acidic patch, but little quenching was observed for the fluorophore located on the H4-tail ( Figure 5B) demonstrating the sensitivity of this assay. Importantly, Swc5 did not quench the acidic patch probe when this substrate also harbored alanine substitutions within the acidic patch ( Figure   5B). The Swc5 LDW-3A derivative also showed specific quenching of the acidic patch probe ( Figure   5C), but the Swc5 RRKR-4A derivative was clearly defective for specific binding in this quenching assay ( Figure 5D). Together these results are consistent with a direct interaction between the arginine-rich domain of Swc5 and the nucleosomal acidic patch, an interaction that is essential for SWR1C dimer exchange activity.

Structure of a Swc5-nucleosome complex
We analyzed the structure of Swc5 79-303 in complex with a 147 bp nucleosome core particle using cryo-EM ( Figure 6). We collected over 13,000 micrographs of the cross-linked sample and were able to easily resolve the nucleosome structure to 3-Å resolution (Supplemental Figure 1L-O).
On the other hand, the moiety corresponding to Swc5 displayed severe structural heterogeneity, hindering our ability to build an atomic model for Swc5. The protein appears to be flexible on the nucleosome, with continuous conformational dynamics apparent ( Figure 6A

The Swc5 arginine-rich domain is required for SWR1C function in vivo
Previous studies have shown that yeast lacking functional SWR1C are inviable on media containing formamide 14,27 . To investigate the functional role of the Swc5 arginine-rich domain in vivo, growth assays were performed with isogenic swc5∆ strains that harbor low copy vectors that express different Swc5 derivat ives ( Figure 7). As expected, the swc5∆ strain grew well on rich media, but was inviable on formamide media (vector), while cells expressing wildtype Swc5 grew well on both media (Swc5). Consistent with a previous study, the derivative with a disrupted BCNT domain was inviable on formamide media (Swc5 LDW-3A ), and the Swc5 79-303 derivative that lacks the acidic N-terminal domain showed a moderate growth defect 27 . Importantly, the Swc5 RRKR-4A derivative had a severe growth defect on formamide media, consistent with an important role of the arginine rich region for SWR1C function.

Discussion
The nucleosomal acidic patch has emerged as a key binding pocket for nearly all ATPdependent chromatin remodeling complexes. Here, leveraging asymmetrically constructed nucleosomes, we have found that both acidic patches are essential for the ATP-dependent deposition of H2A.Z by SWR1C. Each acidic patch is key for both binding and ATPase activity of SWR1C, and the requirement of both acidic patches suggests that SWR1C adopts a pincer conformation, contacting both acidic patches. One possibility is that adoption of this conformation may be the mechanism by which nucleosomes stimulate the ATPase activity of SWR1C from basal levels. Like others have noted 65 , the thermodynamic landscape of the nucleosome provides a much higher barrier to dimer exchange than sliding, and we suspect this is likely responsible for the greater effect of an asymmetric acidic patch mutant nucleosome on SWR1C as compared to other remodelers. Additionally, while the acidic patch plays a key role in the activity of SWR1C and other chromatin remodeling enzymes, it also serves as a docking point for other non-chromatin remodeling nuclear proteins 35,66,67 . From our data, we can conclude that other acidic patch interacting proteins may consequently inhibit the activity of SWR1C even if only one acidic patch is sterically occluded.
Recently, the nucleosomal acidic patch has been shown to differentially impact the behavior of chromatin remodeling enzymes in an asymmetric manner 48,49  The cryoEM structure of a SWR1C-nucleosome complex suggests that the enzyme has two major nucleosomal contacts -the Swr1 ATPase lobes interact with nucleosomal DNA ~2 helical turns from the nucleosomal dyad (SHL2.0), and the Swc6/Arp6 module interacts with DNA at the nucleosomal edge 21 . These contacts encompass the DNA gyre that wraps the AB dimer that is destined to be exchanged in the first round of catalysis (the linker distal dimer 43 ). The structure also indicated that each of the two nucleosomal acidic patches may make contacts with SWR1C subunits, although in both cases the unambiguous identification of the amino acid sequence was restricted by the resolution of cryoEM maps. Unfortunately, the Swc5 subunit was not visualized in a previous cryoEM structure 21 , but here we find that Swc5 can makes an arginine anchor-like interaction with the nucleosomal acidic patch. SWR1C that lacks Swc5 shows the same spectrum of defects as disruption of the nucleosome acidic patch, consistent with Swc5 playing a key role in anchoring SWR1C to the nucleosome. Based on the existing cryoEM map of the SWR1Cnucleosome complex, we propose that Swc5 contacts the H2A/H2B dimer that is targeted for eviction, positioned between the Arp6/Swc6 complex and the Swr1 ATPase. Notably, this places Swc5 close to its interaction surfaces on Swr1 53,71 , as well as positioning its acidic N-terminal domain in an ideal location to capture the evicted AB dimer, as previously suggested 63 .
Previous studies have investigated the impact of Swc5 on the nucleosome binding activity of SWR1C, concluding that this subunit does not play a role in nucleosome recognition. For instance, initial work from Wu and colleagues employed a nucleosomal array pulldown assay 23 , while subsequent studies from Wigley and colleagues used a native gel, electrophoresis shift assay 53 . In both cases, loss of Swc5 from SWR1C had no detectable impact on nucleosome binding.
Furthermore, recent live cell imaging from the Wu group showed that Swc5 is not required for general chromatin association by SWR1C 72 . Together with our results, these data suggest that SWR1C may make strong nonspecific interactions with chromatin, most likely through the Swc2 subunit which is known to bind DNA 23 . Indeed, deletion of Swc2 results in a substantial reduction in the localization of SWR1C to chromatin in vivo 72 . Furthermore, stabilization of a SWR1Cnucleosome complex in a gel matrix may mask the defects in binding that we observe here in a solution-based assay. Notably, our fluorescence polarization assay is unable to detect nucleosome binding by SWR1C if the affinity is reduced by more than 10-fold, and consequently our work indicates that Swc5 is essential for high affinity interactions that are required for subsequent H2A.Z deposition. Whether Swc5 plays roles subsequent to nucleosome recognition is currently unclear.
In addition to contacting the nucleosomal acidic patch, our Swc5-nucleosome structure indicates that Swc5 makes contacts with the H4 N-terminal tail and the H2B C-terminal helix.
Previous work has shown that acetylation of the H4 tail can influence both SWR1C recruitment and activity, raising the possibility that the Swc5-H4 interaction may be functionally important 73 .
Interestingly, we find that Swc5 interacts with the H2B C-terminal helix via contact with H2B-K123 which is subject to transcription-associated modification with ubiquitin by the Bre1-Rad6 ubiquitin ligase complex. Furthermore, H2B-K123ub is enriched at the +1 nucleosome, the same nucleosome targeted by SWR1C for H2A.Z deposition. Strikingly, a ChIP-exo study found that H2B-K123ub is enriched at the NFR proximal face of the +1, while H2A.Z is enriched at the NFR distal surface 74 . The anti-correlation between H2B-K123ub and H2A.Z is consistent with the idea that H2B ubiquitinylation disrupts the Swc5-nucleosome interaction and regulates H2A.Z deposition. Interestingly, H2B-K123ub is also known to regulate the Chd1 remodeler, but in this case H2B-K123ub does not appear to contact Chd1, but rather it interacts with unwrapped nucleosomal DNA, stimulating sliding activity 75 .         complex.

Supplemental Movie 1.
Conformational heterogeneity of Swc5. The 3D class depicted in Figure 6C and Supplemental