Homodimeric PHD Domain-containing Rco1 Subunit Constitutes a Critical Interaction Hub within the Rpd3S Histone Deacetylase Complex*

Recognition of histone post-translational modifications is pivotal for directing chromatin-modifying enzymes to specific genomic regions and regulating their activities. Emerging evidence suggests that other structural features of nucleosomes also contribute to precise targeting of downstream chromatin complexes, such as linker DNA, the histone globular domain, and nucleosome spacing. However, how chromatin complexes coordinate individual interactions to achieve high affinity and specificity remains unclear. The Rpd3S histone deacetylase utilizes the chromodomain-containing Eaf3 subunit and the PHD domain-containing Rco1 subunit to recognize nucleosomes that are methylated at lysine 36 of histone H3 (H3K36me). We showed previously that the binding of Eaf3 to H3K36me can be allosterically activated by Rco1. To investigate how this chromatin recognition module is regulated in the context of the Rpd3S complex, we first determined the subunit interaction network of Rpd3S. Interestingly, we found that Rpd3S contains two copies of the essential subunit Rco1, and both copies of Rco1 are required for full functionality of Rpd3S. Our functional dissection of Rco1 revealed that besides its known chromatin-recognition interfaces, other regions of Rco1 are also critical for Rpd3S to recognize its nucleosomal substrates and functionin vivo. This unexpected result uncovered an important and understudied aspect of chromatin recognition. It suggests that precisely reading modified chromatin may not only need the combined actions of reader domains but also require an internal signaling circuit that coordinates the individual actions in a productive way.

It is now widely accepted that histone post-translational modifications (PTMs) 2 and other structural features of chromatin, such as variant histones and DNA modifications, mainly serve as signal platforms to direct downstream regulatory events (1,2). In general, four major components of chromatin structure contribute to the signals that can be interpreted by downstream chromatin regulators. (i) Histone peptides that carry certain PTMs contribute. Specifically modified histone peptides can be recognized by protein domains/motifs that are commonly referred as PTM "readers." A large repertoire of such domain/PTM interactions have been documented, and they are essential for targeting downstream factors to a given genomic locus (3). Reader domains not only distinguish different types of modifications and the locations of PTMs but also the specific states of modifications. However, due to relatively weak affinity, it is unclear whether these interactions are sufficient to mediate the recruitment of chromatin complexes. For example, although histone H3 methylation at residue Lys-36 (H3K36me) is an essential signal for the histone deacetylase Rpd3S, the contact between the chromodomain of the Eaf3 subunit (CHD Eaf3 ) of Rpd3S and H3K366me only contributes to the binding specificity but not to the overall affinity (4). (ii) Histone globular domains contribute. Because PTMs are mostly clustered at structurally flexible histone tails, it has been postulated that globular contact is needed as a stable anchor to properly orient chromatin binders on their substrates (5). All five crystal structures of protein/peptide-bound nucleosomes to date share a common "arginine anchor" motif that binds to the acidic patch of the histone globular domain surface (6 -10). (iii) Linker DNA contributes. Linker DNA can be recognized by the DNA binding regions mostly in a DNA-sequence-independent manner. In many cases, linker DNA interaction is essential for chromatin binding such as for the histone deacetylase Rpd3S and the Lsd1/CoREST demethylase complexes (11,12). (iv) The spacing between nucleosomes contributes. Rpd3S and the histone methyltransferase PRC2 complex prefer di-nucleosomes with shorter linker DNA (13,14), whereas the chromatin remodeler SWR1 complex prefers longer linkers (15).
Accumulating evidence suggests that multivalent interactions are a prevalent mode of chromatin recognition (16). One of the best-characterized examples of a chromatin complex interacting with modified nucleosomes is the PRC2 histone methyltransferase (17). Targeting the catalytic SET domain of the Ezh2 subunit of PRC2 to its cognate nucleosomal substrates involves several critical contacts between PRC2 and chromatin: the WD40 domain of the EED subunit binds to K27me3 peptides for allosteric activation (18); the VEFS domain of Suz12 binds to H3 of neighboring nucleosomes (13); the N termini of EED (Esc) and RbAp48 (Nurf55) interacts with histone H3 (19,20); the RbAp48 subunit alone can potentially bind to histone H4 (21), although this interaction might be blocked by Suz12 in PRC (19). However, how these individual interactions are utilized by chromatin-modifying enzymes in a coordinated fashion to achieve precise genome targeting remains largely elusive.
The Rpd3S complex is evolutionarily conserved, and the orthologs of its key subunits can be found in many species including flies and mammals (36,37). The yeast Rpd3S is approximately a 410-kDa complex (38) composed of five subunits. Rpd3 is responsible for catalysis. Two key components of Rpd3S that were shown to involve chromatin recognition are Eaf3, which contains a chromodomain (CHD) that binds to H3K36me, and Rco1, which contains two PHD domains with PHD1 being essential for Rpd3S function (11). All Rpd3S complexes that have been reported appear to function at coding regions where H3K36me is enriched, although the functional consequences of the resulting hypoacetylation at coding regions vary among species (22,36,37). We recently showed that Rpd3S prefers dinucleosome substrates and bridges between two linked nucleosomes (38). We also showed that H3K36me can direct Rpd3S to deacetylate neighboring nucleosomes (14). Importantly, we discovered that the SID domain of Rco1 (SID) can bind to the MRG domain of Eaf3 (MRG) and allosterically activate the binding of CHD to H3K36-methylated peptides (4), suggesting that the ability to read PTMs can be a crucial regulatory step for a chromatin-modifying enzyme. We subsequently identified a minimal nucleosome recognition module of Rpd3S, which constitutes the PHD-SID region of Rco1 (PHD-SID) in complex with Eaf3, and demonstrated that it is tightly controlled by two autoinhibitory mechanisms (4). To gain deeper insights into how Rpd3S controls its precise chromatin recognition beyond the direct nucleosome contact interface and how the minimal nucleosome binding module exerts its function in the context of Rpd3S, we decided to first determine the subunit-interacting network of Rpd3S. We showed that Rco1 and Eaf3 are integrated to Rpd3S through the scaffold subunit Sin3, which is directly connected to the catalytic Rpd3. Interestingly, we found that Rpd3S contains two copies of the essential subunit Rco1, and both copies of Rco1 are required for full functionality of Rpd3S. Deletion of the SID domain from one copy of Rco1 is sufficient to abolish nucleosome binding ability of Rpd3S without disrupting the integrity of the complex. Moreover, our functional dissection of Rco1 revealed that besides the known chromatin recognition interfaces of Rco1, other regions of Rco1 are also critical for Rpd3S to recognize its nucleosomal substrates and function in vivo. This unexpected result uncovered an important and understudied aspect of chromatin recognition and suggests that precisely reading modified chromatin may not only need the combined actions of reader domains but also require an internal signaling circuit that coordinates the individual actions in a productive way.

Experimental Procedures
Construction of Plasmids and Yeast Strains-All yeast strains were constructed using standard procedures and are listed in Table 1. To test the cryptic transcription phenotype caused by mutations in Rpd3S, we employed a yeast reporter strain (YCR239) that carries an integrated HIS3 reporter gene at the STE11 locus. In this strain HIS3 is activated when the Set2-Rpd3S pathway is defective (4).
Protein Purification-Recombinant Rpd3S or subcomplexes were purified from an Sf21 insect cell-based baculovirus expression system as described previously (4,39). Briefly, freshly passed Sf21 cells were co-infected with an individual virus that encodes each subunit of Rpd3S for 48 h. Cells were collected and lysed in BV lysis buffer (50 mM HEPES, pH 7.9, 300 mM NaCl, 2 mM MgCl 2 , 0.2% Triton X-100, 10% glycerol, 0.5 mM EDTA, and freshly added protease inhibitors) on ice for 30 min. Cell lysates were clarified by ultracentrifugation and incubated with anti-FLAG M2 resin (Sigma) at 4°C for 2 h. After extensive washing, proteins were eluted using 500 g/ml 3ϫ FLAG peptides in BV elution buffer (50 mM HEPES, pH 7.9, 100 mM NaCl, 2 mM MgCl 2 , 0.02% Nonidet P-40, and 10% glycerol). If tandem purification was needed, 5 volumes of BV Lysis buffer were added back to the FLAG eluents and then mixed with anti-HA-agarose (Sigma) at 4°C for 2 h. HA elution was conducted using 500 g/ml 2ϫ HA peptides in BV elution buffer. All final purified proteins were concentrated using Amicon concentrators. All native yeast Rpd3S complexes were purified using the standard TAP method as described previously (11).
Deuterium Exchange Mass Spectrometry (DXMS)-Recombinant Rpd3S purified from co-infected Sf21 insect cells was used in DXMS as previously described (4). Briefly, 12 l of the Rpd3S (1.5 mg/ml) was equilibrated with 6 l of the nucleosome buffer (10 mM Tris-HCl, pH 7.5, and 5 mM ␤-mercaptoethanol) at 30°C for 60 min and then cooled to 0°C. 2 l of the above mixture was then mixed with 6 l of D 2 O buffer (8.3 mM Tris, 150 mM NaCl in D 2 O, pD read 7.2) at 0°C. Deuteration was stopped at different time points by adding 12 l of quenching buffer (1.6 M guanidinium HCl, 0.8% formic acid, and 16.6% glycerol) on ice followed by freezing at Ϫ80°C. Duplicate samples were collected at 3 time points: 10 s at 0°C and 100 s and 10,000 s at room temperature. Once all samples were collected, they were thawed on ice and passed over AL-20-pepsin columns (Sigma, 16-l bed volume). The digested peptides were loaded on a C18 trap (Michrom MAGIC C18AQ 0.2 ϫ 2) and separated by a C18 reverse phase column (Michrom MAGIC C18AQ 0.2 ϫ 50) running a linear gradient of 8 -48% solvent B (80% acetonitrile and 0.01% TFA) over 30 min. The column effluents were then directly injected into an OrbiTrap Elite mass spectrometer (Thermo Fisher) for analysis. Proteome Discoverer software (Thermo Fisher) was used to identify the sequence of the peptide ions. The centroids of the isotopic envelopes of non-deuterated, partially deuterated, and equilibrium-deuterated peptides were measured using DXMS Explorer (Sierra Analytics, Inc., Modesto, CA) and then converted to the corresponding deuteration levels. The data analysis was essentially carried out as described previously (40) with more streamlined computation tools for handling larger datasets.

Rpd3S Contains Two Copies of Rco1 as Revealed by Subunit-
interacting Network Analysis-We previously identified a minimal nucleosome recognition module of Rpd3S that can bind to mono-and di-nucleosomes in a K36me-dependent manner that resembles the intact Rpd3S (4). To investigate how this module functions and is regulated in the context of Rpd3S, we decided to first examine the subunit-interacting network within the Rpd3S. We took advantage of an established in vitro reconstitution system in which recombinant Rpd3S can be pro-duced using combinations of baculoviruses that express individual subunit of Rpd3S (4,39). To test the interdependence of each subunit for complex assembly, recombinant Rpd3S complexes were reconstituted in which a specific virus for an individual subunit was omitted. As shown in Fig. 1A, Rco1 is required for Eaf3 incorporation, whereas omission of Eaf3 only minimally impacts the association of Rco1 with the complex. This is consistent with previous results using native complexes (29). As expected, Sin3 is the major scaffold subunit of Rpd3S because (i) Sin3 links the Rco1/Eaf3 heterodimer to Rpd3 and Ume1 (Fig. 1B, lane 2); (ii) Sin3, Rpd3, and Ume1 form the Rpd3 core complex (Fig. 1C) (42), which is shared between Rpd3S and Rpd3L (29); and (iii) without Ume1 (Fig. 1C, lane 1) or Rpd3 (Fig. 1D, lane 2), Sin3 remains associated with the rest of the subunits of Rpd3S. Lastly, Ume1 seems to only contact Sin3, and its absence does not influence other components of Rpd3S, which may explain the fact that no obvious ortholog has been found in flies and mammals (36,37).
To further characterize the multivalent nature of the nucleosome interactions of Rpd3S, we sought to determine how many copies of each chromatin reading subunit (Rco1 and Eaf3) are in Rpd3S. We created two versions of each protein through differential epitope tagging and performed co-immunoprecipitation experiments in the reconstituted system. The results showed that Eaf3 appeared to be in a monomeric form, because the HA-tagged Eaf3 failed to pull down untagged Eaf3 (Fig. 1E). However, when we reconstituted Rpd3S with two differently tagged Rco1, tandem purification resulted in an intact Rpd3S that contains both versions of Rco1 (Fig. 1F), suggesting that Rpd3S possesses two copies of Rco1. This result is consistent with the observation that the fission yeast Rpd3S complex includes two PHD domain-containing subunits, Cph1 and Cph2, both of which are Rco1 orthologs (43). To test whether dimerization of Rco1 could be a common feature among other species, we performed phylogenetic analysis. The results showed that all species within the budding yeast branch carry only one Rco1 ortholog gene, whereas the fission yeast branch seems to have two genes encoding the Rco1 counterpart, with one of them containing a shorter N-terminal region and a lessconserved PHD2 (Fig. 1G). Therefore, it is likely that Rpd3S in different species might contain either two copies of the Rco1 protein or two closely related homologous proteins.
To confirm that Rpd3S contains two Rco1 in vivo, we constructed a diploid yeast strain in which the two copies of Rco1 are tagged differently. Indeed, we found that these two tagged Rco1 can physically interact with each other ( Fig. 2A), suggesting that dimeric Rco1 exists in Rpd3S in vivo. We then used the reconstituted system to demonstrate that incorporation of two copies of Rco1 into the Rpd3S complex does not rely on CHD and PHD (Fig. 2B). Rco1 forms a homodimer in the presence and absence of Eaf3 (Fig. 2, C and D). Combining these results with previous findings, we have updated the internal subunitinteracting network of Rpd3S as shown in Fig. 2E. This map is in a good agreement with a proteomics-based connectivity analysis (44). In addition, we found that Eaf3 weakly interacts with Rpd3 ( Fig. 1D and data not shown), which is consistent with a previous report that Drosophila MRG15, the Eaf3 homolog, inhibits the Rpd3 HDAC activity toward core histone substrates (36).
Both Copies of Rco1 Are Essential for Rpd3S Function-To address the functional significance of having two copies of the essential Rco1 subunit in Rpd3S, we wanted to determine the impact of mutating only one copy of Rco1 on Rpd3S function. We showed previously that deletion of PHD1 abolished the Rpd3S activity without disrupting complex integrity (11). Deletion of SID dissociated Eaf3 from the Rpd3S, which inactivated Rpd3S (4). Therefore, we reconstituted the mutant Rpd3S complexes in which these two essential domains were removed individually from only one copy of Rco1. Baculoviruses that express differently tagged WT and mutant Rco1 were combined, and the mutant complexes were prepared through a tandem-purification approach (Fig. 3A). The results showed that Rpd3S remains intact when one PHD1 or one SID was deleted (Fig. 3A, lanes 2 and 3). Interestingly, a mutant Rpd3S can be assembled with one copy of Rco1 that lacks PHD1 and the other copy of Rco1 that is missing SID (Fig. 3A, lane 4). Given that SID is required for Eaf3 association with Rpd3S, the fact that we did not observe any detectable loss of Eaf3 in the ⌬1SID and ⌬1SID/ ⌬1PHD1 mutants suggested that the remaining SID in the mutant Rpd3S should be preferentially occupied by Eaf3 (Fig.  3B).
To test whether these mutant complexes maintain normal functions, we first examined their binding ability to mononucleosomes and di-nucleosomes using gel mobility shift assays. Consistent with our previous results (38), wild type Rpd3S displayed higher affinity toward di-nucleosomes than mono-nucleosomes (Fig. 3, C and D). We found that deletion of one PHD1 only modestly reduced the binding of Rpd3S to mono-nucleosomes and had little effect on dinucleosome binding (Fig. 3, C and D). Strikingly, deletion of one copy of the SID region of Rco1 abolished the binding of Rpd3S to both monoand di-nucleosomes (Fig. 3, C and D) without disrupting complex integrity (Fig. 2E), suggesting that the presence of both functional copies of Rco1 is essential for Rpd3S activity. This result was reminiscent of the observation of the Rpd3S counterpart in fission yeast, which contains two closely related subunits, Cph1 and Cph2. It was shown that deletion of Cph1 alone is sufficient to cause the Rpd3S-defective phenotype (43). We next measured the histone deacetylase activity of mutant Rpd3S using an established histone deacetylation (HDAC) assay (41). Because during elongation Rpd3S should only FIGURE 1. Subunit-interacting network analysis suggests two copies of the essential Rco1 subunit within Rpd3S. A-D, Coomassie staining of the indicated recombinant Rpd3S complexes that were produced in a baculovirus expression system. The bands corresponding to each subunit were indicated either as green dots (the tagged subunits for purification) or red dots (untagged subunits). Indicated Rpd3S complexes were prepared via FLAG purification. Ϫ indicates that the particular virus was omitted from the reconstitution. Asterisks indicate degradation or contaminated proteins. E, Eaf3 is in a monomeric form in Rpd3S. The indicated viruses were co-infected to insect cells. Western blots were performed after FLAG purification. F, Rpd3S contains two copies of Rco1. Recombinant Rpd3S was prepared through FLAG and HA tandem purification and stained with Coomassie Blue. G, a multiple sequence alignment of Rco1 orthologs. The following sequences were used: SpCph2 (Q09698, Schizosaccharomyces pombe), SpCph1(Q09819, S. pombe), SjCph2(SJAG_01110, Schizosaccharomyces japonicas), SjCph1(SJAG_03884, S. japonicas), SkuRco1 (protSku1303, Saccharomyces kudriavzevii), SbaRco1 (Sbay_66.15, Saccharomyces bayanus), ScRco1 (Q04779, Saccharomyces cerevisiae), and SpaRco1 (protSpa836, Saccharomyces paradoxus). Conserved residues are colored with the ClustalX scheme in Jalview (56). The resulting graph was further condensed and is displayed. Conservation value and phylogenetic tree were also calculated in Jalview. The domain structure of Rco1 is illustrated at the bottom, and the critical boundary residues were labeled, although the sequence gaps within Rco1 were not removed.
deacetylate nucleosomes behind the elongating polymerase II but not acetylated free histones, which are important for chromatin assembly (45), it is critical to use physiologically relevant nucleosomal substrates for these assays. Recombinant nucleosomes were acetylated by cocktails of native histone acetyltransferases in the presence of [ 3 H]acetyl-CoA and further purified as substrates (Fig. 4B). Using this system, we showed previously that Rpd3S preferentially deacetylates H3K36methylated nucleosomes, and it also shows stronger activity toward di-nucleosomes than mono-nucleosomes (38). Here, we wanted to first test Rpd3S mutants without the essential CHD and PHD (Fig. 4A). Wild type Rpd3S displayed robust HDAC activity toward unmodified nucleosomes, which was further stimulated by H3K36me (Fig. 4C). Deletion of CHD  reduced Rpd3S activity on WT nucleosomes, and more importantly, H3K36me failed to elevate this basal activity (Fig. 4C). Deletion of both copies of PHD1 severely compromised Rpd3S activity, which is consistent with the defective nucleosome binding of this mutant that was reported previously (11). We next tested the Rpd3S that contains a single mutated Rco1. In general, all mutants showed reduced HDAC activities to the extent that mirrored their reduced nucleosome binding ability (Fig. 4, D and E). Of note, H3K36me can still stimulate mutant complex activity, which is likely due to the basal functions of CHD. Collectively, our data suggest that having two functional copies of Rco1 is required for the full Rpd3S activity.

Rco1 Plays Multiple Essential Roles in Regulating Rpd3S
Functions-Because both copies of Rco1 play pivotal roles in regulating Rpd3S function, we wanted to determine whether other structural domains of Rco1 might also contribute to Rpd3S activity. Our previous studies have revealed several important domains of Rco1 (Fig. 5A): SID is responsible for allosteric activation of the binding of CHD to H3K36me, the autoinhibition domain (AID) suppresses the SID-mediated activation, and upon contacting nucleosomes, PHD1 somehow releases the AID repression, which leads to CHD activation and nucleosome engagement (4). The nucleosomal target of PHD1 remains unknown. Our unpublished results suggest that PHD2 is also essential for Rpd3S function (in collaboration with Dr. Strahl's laboratory at UNC Chapel Hill). However, the functions of the N-terminal disorder region and highly conserved C-terminal region have not been explored.
We have performed DXMS analysis to monitor the dynamic structural changes of Rpd3S upon nucleosome contact (4). This technique measures the hydrogen/deuterium exchange rates at each residue, which correlate with the solvent accessibility at the region (46). This assay can provide dynamic conformational information. DXMS analysis of the free form of Rpd3S may also reveal insights into the structure of the complex. Therefore, we inspected the deuterium exchange rates of Eaf3, which are displayed in a ribbon format in Fig. 5B. In general, the exchange rates nicely correlated with the predicted secondary structure (Fig. 5). In particular, both evolutionarily conserved regions of Eaf3, the methyl-lysine binding CHD and the C-terminal MRG domains, adopt well folded conformations within the Rpd3S complex because slow deuterium incorporation was detected at these regions. MRG appeared to be more compact than CHD as indicated by the slower kinetic of deuterium incorporation (Fig.  5B, comparing all three time points). The DNA binding region of Eaf3 (DBR, 122-203 amino acids) is highly solvent-accessible (Fig. 5B), likely due to structural flexibility, which is very consistent with the lack of discernable predicted secondary structure at this area. We then mapped the deuterium exchange profiles onto the available three-dimensional structural models of Eaf3. The overall exchange patterns of CHD and MRG are consistent with the solvent accessibility of a folded structure (Fig. 5, C and D). Interestingly, mosaic-like color patterns at some defined secondary structures suggest that the rates of exchange at these regions may also be influenced by contact with other subunits within the complex (Fig. 5, C and D). The N-terminal region of Rco1 is highly accessible, and only the middle conserved region (common among 24 Rco1 orthologs) showed moderate solvent protection (Fig. 5A). Consistent with predicted secondary structure, the highly conserved C-terminal regions displayed slower exchange rates, indicating that they are well structured within the complex. To directly test the functional importance of these two regions, we created a series of truncating mutants of Rco1 (Fig. 6A). Expression plasmids carrying these mutants were transformed into a yeast strain that contains an STE11-His3 reporter that can be activated when the Set2-Rpd3S pathway is defective (4) (Fig.  6B). Surprisingly, we found that all of the truncated Rco1 mutants caused a marked cryptic transcription phenotype (Fig.  6B). The control experiments indicated that all mutants were expressed at a comparable level and did not disrupt Eaf3 interaction (Fig. 6C). We then purified the reconstituted mutant complexes and found that removal of these two regions did not disrupt complex integrity (Fig. 6D). However, these mutant complexes showed severely compromised the ability to bind to nucleosomes (Fig. 6E). Because both copies of Rco1 are important for Rpd3S function (Figs. 3 and 4), we wanted to test whether these truncations altered the dimerization status of Rco1. Indeed, we observed that Rco1 still dimerizes when the entire C-terminal region (C⌬480) was removed (Fig. 6F), suggesting that the conserved C-terminal region plays a critical role in Rpd3S function. We attempted to purify the PHD1-PHD2 regions (Rco1 260 -470 amino acids) and test whether this fragment is involved in Rco1 dimerization. Unfortunately, this fragment appeared to be toxic both in bacteria and insect cells. Thus, we cannot rule out that PHD1-PHD2 is sufficient to mediate Rco1 dimerization. However, Rco1 dimerizes without PHD1 (Fig. 2B). PHD2 alone cannot form a homodimer as Histagged PHD2 failed to interact with GST-PHD2 under the same conditions where we detected Rco1 dimerization (data not shown). In addition, all Rco1 truncations tested can dimerize with the full-length Rco1 (Fig. 6G). These data collectively indicated that Rco1 dimerization is likely mediated through multiple regions. More importantly, our results strongly suggest that the N-terminal and C-terminal regions of Rco1 are also essential for Rpd3S functions.

Discussion
Recognition of modified histone peptides by reader domains is critical to localize chromatin-modifying enzymes to specific genomic regions and direct subsequence enzymatic activities toward cognate substrates (3). However, these macromolecular complexes typically contain many ancillary subunits that are not involved in contacting chromatin, and the functional importance of these components has been largely ignored in the literature. Our recent work suggests that PTM reading can be allosterically regulated (4), implying that internal signaling pathways within chromatin-modifying enzymes may be of great importance for its regulation. Here, we showed that Rpd3S contains one copy of the H3K36me binding Eaf3 and two copies of Rco1, which carries two essential PHD domains that may contact nucleosomes. Importantly, both copies of Rco1 are required for full functionality of Rpd3S. To our surprise, we found that besides the known chromatin binding interface, other regions of Rco1 also have important functions beyond simple architectural roles.
Rpd3S Is Controlled by an Interaction Circuit-We have shown previously that the combined actions of CHD and PHD drive Rpd3S to H3K36-methylated nucleosomes (11). Subsequent mechanistic studies revealed that PHD-SID/Eaf3 constitutes the minimal nucleosome binding module of Rpd3S in which SID-mediated Eaf3 activation is controlled by an autoin-hibitory domain, and MRG intrinsically represses the binding of the CHD and DBR (DNA binding region) to H3K36me and linker DNA, respectively (4). In fact, Rpd3S needs to have the ability to deacetylate all of the histone tails of a single nucleosome even though they project in different directions (38). Rpd3S must also work on di-nucleosomes with varying length linkers and the nucleosomes adjacent to H3K36-methylated targets (14), which demands tremendous structural flexibility of the complex. More importantly, because Rpd3S is believed to travel with elongating polymerase II (39), it also needs to adapt to rapid catalysis on a moving machine. Therefore, it appears that a sophisticated internal regulatory system is required for Rpd3S to accommodate these complex functionalities.
Here, we showed that Rpd3S contains two copies of Rco1, and both of them seem to be essential for its full functionality. In a defined biochemical system, removal of the essential  (4). The yeast were serially diluted and spotted on the indicated plates. C, deletion of the N-terminal or C-terminal region of Rco1 does not compromise the binding of Eaf3 to Rpd3S. Whole cell extracts from the yeast strains in B were immunoprecipitated (IP) using anti-FLAG antibody and then subjected to Western blots. D and E, deletion of the N-terminal or C-terminal region of Rco1 did not disrupt complex integrity but abolished Rpd3S nucleosome binding. D, Coomassie staining of Rpd3S complexes purified from the baculovirus system. E, EMSA using mono-nucleosomes. F, Rco1 can still dimerize when the C-terminal region is deleted. The experimental scheme is indicated on the left. Two differently tagged Rco1 were used in a coimmunoprecipitation assay in a baculovirus system. WCE, whole cell extracts. G, the N-terminal and C-terminal truncated Rco1s can interact with full-length Rco1.
domain (SID) from only one copy is sufficient to abolish Rpd3S nucleosome binding. Due to technical limitations, this observation cannot be confirmed in vivo as dimerization of the wild type copy of Rco1 should lead to a functional Rpd3S, thereby disguising the potential phenotypes caused by mutations in the other copy of Rco1. However, in fission yeast, deletion of CPH1 causes total Rpd3S defects (43), suggesting that both functional copies of the Rco1 counterpart are essential under physiological conditions. We showed that removing one PHD1 can be mostly tolerated by Rpd3S (Fig. 3C), which might be due to functional redundancy of the two PHD1. Nevertheless, it is possible that both PHD1 are required under more stringent competition in vivo or when Rpd3S moves with a fast-traveling machine. Remarkably, deletion of SID from one copy of Rco1 totally abolished Rpd3S nucleosome binding and severely compromised its HDAC activity (Figs. 3 and 4). As discussed above (Fig. 3B), the remaining SID (first SID) should bind to Eaf3. What is the essential function of the second SID in WT Rpd3S? In mammalian systems, multiple regions of Pf1, the Rco1 homolog, contact Sin3 through the PAH1, PAH3, and HID domains of Sin3 (37). However, unlike in mammals, the yeast SID does not bind to the PAH2 domain of Sin3 (data not shown). Moreover, in the yeast Rpd3S, deletion of PHD1, SID, or PHD2 does not dissociate Rco1 from Sin3. Therefore, we speculate that the essential function of the second SID might be to contact other parts of Sin3 and drive conformational changes of the complex when Rpd3S senses a part of the nucleosomes.
Although we detected two copies of Rco1 within Rpd3S, and Rco1 can form a homodimer on its own, it is still formally possible that two Rco1 proteins independently bind to Sin3 and do not form homodimers in Rpd3S. Because Rpd3S can bind to dinucleosomes with varying linker distances and our preliminary EM study of Rpd3S (in collaboration with Dr. Asturias, Scripps) suggests that Rpd3S appears to be very flexible, we postulate that the Rco1 homodimer might contribute to this structural flexibility, particularly considering that the disordered and highly solvent-exposed N-terminal region may be involved in the formation of homodimers.
Lastly, we showed that PHD-SID/Eaf3 can efficiently bind to a nucleosome in a manner that mimics the intact Rpd3S (4). For almost all of the Rco1 mutants tested here, this minimal module remains undisrupted. However, the mutant Rpd3S showed severe nucleosome binding defects, suggesting that this module is tightly regulated in the context of Rpd3S. Similarly, we showed previously that deletion of a part of the signal transduction region within SID compromises Rpd3S function without altering complex integrity (4). In both scenarios, all key chromatin contacts are still available, but the complex loses its functions, which strongly suggests that an internal interaction network of chromatin-modifying complexes is critical for regulating their targeting and subsequent activities. In summary, dimeric Rco1 appears to be the key signal hub that directs Rpd3S function in multiple distinct steps.
General Principles for Recognition of Modified Chromatin-Rpd3S is a highly conserved histone deacetylase complex and it possesses many physical features that are shared among other chromatin complexes. The regulatory principles of this complex should have general implications on how histone modifi-cation signals can be precisely interpreted by downstream chromatin complexes. First, besides the apparent chromatinreading domains, other parts of the complex can also play important regulatory roles. We showed that deletion of almost any part of Rco1 imposes detrimental effects on Rpd3 functionality. These regions could be involved in contacting DNA, binding to the histone globular domain, providing structural flexibility for the complex to adopt different conformations, serving as an internal ruler for linkers, or mediating signal transduction. Second, the histone-reading domains are highly regulated in the context of the intact complex. We showed previously that SID can activate the binding of CHD to H3K36, which is autoinhibited in full-length Eaf3 (4). Here, we further demonstrated that this minimal binding module, which can bind to nucleosomes on its own, failed to do so within Rpd3S when Rco1 is mutated. This result suggested another level of self-control mechanism that is implemented in the context of the complex. In fact, auto-inhibitory mechanisms are prevalent among chromatin-modifying enzymes (47)(48)(49)(50)(51)(52)(53), which presumably is helpful to establish spatial and temporal specificity of the modification reactions. For instance, H3K36me can be recognized by opposing enzymes Rpd3S and NuA4/Tintin (34, 54) through a common subunit Eaf3. Thus, the ultimate decision and timing of recruitment could be determined by the corresponding regulatory partners of Eaf3 within each complex. An elegant recent study showed that the subunit of the yeast PRC2 complex Ccc1 can bind to H3K27me3 and suppress a latent affinity of another subunit, EED, toward K9me2 (55). Interestingly, in a Ccc1 mutant, the yeast PRC2 redistributes to the regions where K9me2 is enriched via EED interaction (55), suggesting that histone-reading domains can be dynamically influenced by another reading domain. Third, stoichiometry of chromatinreading subunits might play an important role in regulating complex activity. Considering the symmetric nature of nucleosome structure, which can easily provide binding surfaces for the same reading module, having more than one copy could present tremendous advantages for the binding strength and conformation flexibility of the complex. Alternatively, multiple copies of the same reader may enable the complex to distinguish two histone tails within one nucleosome that are differentially modified.