to directly interact with Mit1, we wanted to understand how this in-teraction compares to the high-affinity Mit1-Chp2interaction [Fischer et al., 2009]. To this end, we purified the chromo shadow domain of Swi6 and subjected it to ITC measurements. In addition, we tested binding to a known Swi6interaction partner, a PxVxL-motif contain-ing Sgo1 peptide [Isaac et al.,2017]. In addition to reproducing the reported affinity for the Sgo1-Swi6interaction within error, we found that Mit1binds Swi6with the same affinity within error. Furthermore, we could show that Sgo1similarly interacted with Chp2(Figure16A).
This points to an interaction network between theS. pombe HP1 pro-teins and their partners that seems to be more general and is based on PxVxL-like motifs being accommodated by the binding groove with-out strict sequence specificity. The interaction between Chp2and Mit1 on the other hand is ~50times tighter and depends on more than just the motif binding.
Size-exclusion chromatography and gel electrophoresis allowed us to verify that also the Sgo1 peptide and Swi6 CSD are stable and not aggregated (Figure 16B, C). However, we noticed that while stoi-chiometry of the Swi6-Mit1complex seemed to be2as predicted from PxVxL-based interactions, for both Swi6-Sgo and Chp2-Sgo1we mea-sured a binding ratio of1:1. Since proteins elute as a single peak from size-exclusion chromatography and show no degradation on gel elec-trophoresis, we assume them to be functional and therefore cannot explain why the binding ratio appears to be different than expected.
Further experiments will have to be conducted to rule out artefacts and understand possible alternative binding modes.
Figure16: ITC measurements show interactions between HP1proteins and inter-action partners. (A) Injection of Chp2CSD or Swi6CSD into Mit1(1-81) or Sgo1, respectively. The heat rate trace for injection of Swi6CSD into Sgo1is shown as an example. (B) SEC and gel electrophoresis for samples used for ITC, Sgo2. (C) ITC sample Swi6CSD.
Disrupting the Chp2–Mit1interaction impairs silencing
To test the functional significance of an intact Chp2–Mit1interface we replaced the endogenous copy of Mit1with Mit1harboring mutations targeting the CkIvV motif, the linker region and the CDL domain.
Subjecting these strains to comparative growth assays on selective medium analogous to previous experiments (Figure8) revealed that disruptive mutations in any part of the Chp2–Mit1 interface lead to loss of gene silencing comparable to the deletion of Chp2 or to the deletion of the Mit1 N-terminus (Figure17). Deletion or mutation of the motif (rows 8 and 9) show a very similar silencing phenotype but seem to have slightly less impact than deleting or mutating the linker region (rows 6, 10, 11) or deleting the CDL (row 7). These
re-Figure17: Silencing assay shows the importance of an intact Chp2-Mit1interface for binding. Mit1-13myc denotes a strain harboring13myc-tagged Mit1 at its en-dogenous locus. All other Mit1constructs are also 13myc-tagged and replace the endogenous copy of Mit1.
sults are slightly different from our ITC data where deletion of the CDL impacted the affinity to Chp2 less than mutating the motif se-quence. However, both experiments point in the same direction and minor differences can be due to additional factors that arise from thein vitro nature of ITC measurements. We identified a basic patch towards the end of our crystallization construct (Mit1(1-81)) and hy-pothesized that this could be involved with DNA or RNA binding.
However, mutating these basic residues to alanines showed a very weak phenotype (lane12), arguing against a role of these residues in Mit1-mediated silencing. Taken together, this assay validates the im-portance of an intact interface between Chp2 and Mit1 for function and shows that further possible recruitment mechanisms, eg. Swi6 -mediated, are likely to play a minor role.
Mit1ATPase activity is stimulated by nucleosomes but not free DNA To understand whether the interaction partners of Mit1, namely Chp2 and Clr1, have any impact on the remodeling activity of Mit1 and maybe contribute to its regulation, we performed assays measuring the ATPase activity of Mit1coexpressed and -purified with Chp2and Clr1, and compared it to the well-studied chromatin remodeler Chd1. We found that compared to Chd1, Mit1exhibits a higher basal ATPase activity that is only marginally increased in the presence of DNA (Figure 18). Nucleosomes however significantly activate Mit1 activ-ity. To understand possible contributions of Chp2 and Clr1, we had
Figure18: Mit1ATPase activity increases upon addition of nucleosomes.The rate of ATP conversion is shown for no substrate, free DNA and nucleosomes in dupli-cates, striped and gray for Mit1and black and white for Chd1.
planned to repeat the same experiment with the N- or C-terminus of Mit1 deleted, thereby removing either one or both interaction do-mains. Unfortunately we could not perform this experiment due to time constraints and the instability of full-length Mit1.
The interface sequence is not well conserved
To answer the question whether this mode of interaction between chromatin remodeler and HP1 protein is conserved, we looked at the sequence of Chp2 CSD and the N-terminus of Mit1 (Figure 19).
The chromo shadow domain of Chp2shows some conservation both within the Schizosaccharomyces and the Pneumocystis clade. Interest-ingly, while a number of residues involved in Mit1 binding (Fig-ure 12A-D) are conserved between the different proteins, three in-terface residues are conserved everywhere but in Swi6, the other S. pombe HP1 protein (Figure 19A). The N-terminus of Mit1 is less conserved, with some degree of sequence homology only within the Schizosaccharomycesclade (Figure19B). Again, some residues involved in complex formation are conserved, among them the hydrophobic residues in position 3 and 5 of the motif, but the overall conserva-tion score is low. Taken together, this indicates that the Chp2-Mit1
interaction interface does not follow a general binding mode, but has evolved specifically between these two proteins.
Figure19: Sequence alignments reveal that the interaction interface is not well con-served.(A) Chp2has sequence homologs both within the Schizosaccharomyces and the Pneumocystis clade. Red arrows indicate conserved residues implicated in Mit1 binding. Red boxes indicate residues that are not conserved within Swi6. (B) The N-terminus of Mit1has sequence homologs only within the Schizosaccharomyces clade. Green arrows indicate conserved residues important for the interaction with Chp2, the green box marks the CkIvV motif.
The observed binding mode is not a universal mechanism for HP1binding Having identified the CDL as part of the interaction interface, we wondered if this involvement might be a more general mechanism for interaction with HP1 proteins. We decided to test this hypothe-sis on another HP1 interaction partner containing a chromodomain, the human methyltransferase Suv39h1. We reasoned that coexpres-sion of the N-terminal part of Suv39h1including the chromodomain together with the chromo shadow domain of its interaction partner HP1β would yield a complex that is resistant to protease treatment, analogous to Figure 9, if indeed the chromodomain of Suv39h1was involved with binding. While we could purify a stable HP1β-Suv39h1 complex (Figure 20A), it was not resistant to protease treatment and Suv39h1 was digested to smaller fragments (Figure 20B), indicating that the interaction interface was not equivalent to the one observed
in the Chp2-Mit1 complex. However, this finding does not rule out the general possibility that there are similar interactions within other HP1protein complexes.
Figure20: Suv39h1-HP1βcomplex is not resistant to protease treatment.(A) The Suv39h1-HP1βCSD complex elutes as a single peak from SEC and peak fractions were used for limited proteolysis. (B) Limited proteolysis experiment with increas-ing amounts of protease yields shows that Suv39h1is not stable and gets digested to smaller fragments, indicating that not the full construct is involved in complex formation.
The Chp2-Mit1structure reveals interaction beyond canonical PxVxL recog-nition
We set out to understand how Chp2interacts with the SHREC com-plex and to shed light on its role for SHREC recruitment to hete-rochromatin. We were able to unambiguously show that the chromo shadow domain of Chp2 binds to the N-terminus of the chromatin remodeler Mit1and could solve the crystal structure of this complex.
The data revealed that in contrast to previously reported structures of HP1-client interactions [Thiru et al.,2004, Kang et al.,2011], the in-terface between Chp2and Mit1extends beyond the PxVxL motif and also encompasses the elongated linker region and the CDL, a cryptic domain with a chromodomain fold.
Even though the special 2:1 stoichiometry of this complex was ex-pected from previous studies of PxVxL-motif recognition [Thiru et al., 2004], the binding mode we observed is even more unusual and can be divided into two parts. The first part is the CkIvV motif that is bound across the CSD dimer symmetry axis and follows the known peptide recognition of HP1 proteins, thereby defining the stoichiom-etry of the complex. The second part consists of interaction between two monomeric protein chains and would in theory allow for for-mation of a symmetric 2:2 complex. Binding of Mit1 in the bind-ing groove and around one copy of Chp2 breaks the symmetry of the CSD dimer and renders the two Chp2 monomers in this struc-ture nonequivalent, reminiscent of the asymmetric binding of TAF-subunits to the symmetric core-TFIID complex [Bieniossek et al.,2013].
We found that all three parts of Mit1 contribute to complex forma-tion with hydrogen bonds or hydrophobic interacforma-tions, providing a large interaction surface and yielding specific recognition. To our knowledge, a similar interaction mode has not been proposed be-fore, therefore expanding our view of how HP1proteins interact with their client proteins. Furthermore, it raises the possibility that previ-ously reported structures missed this extended interaction surface by performing crystallization experiments only with the PxVxL peptide.
Hence, it would be interesting to revisit those complexes and have a closer look at the extent of interaction in order to understand if fur-ther parts of the client protein are involved in complex formation in vivo.
In this line of thought, we probed the interaction partners Suv39h1 and HP1βfor an extended interface, knowing that Suv39h1possesses a chromodomain close to its HP1-interaction motif that might be in-volved with HP1binding like we see it for the Mit1CDL. While our experiments suggested that in this case complex formation does not include more domains of the client protein, it does not rule out the general possibility. For further studies, candidates worthwhile test-ing are human chromatin remodelers CHD3 and CHD4. Like Mit1, CHD3 and CHD4 belong to the Mi-2 subfamily of CHD-type chro-matin remodelers and are subunits of the NuRD complex, the SHREC homolog in higher organisms [Tong et al., 1998]. Furthermore, both have been reported to interact with HP1proteins in a DNA-indepen-dent manner and with isoform specificity [Hoffmeister et al.,2017 ,Ver-meulen et al., 2010, Rosnoblet et al., 2011]. It is tempting to propose a similar recognition mechanism as observed for the Mit1-Chp2 com-plex that might involve either the tandem chromodomain within the N-terminus of CHD3/CHD4or a cryptic domain that is yet to be de-fined. This could be tested in future experiments with co-expressed and purified CHD3/CHD4 and HP1 proteins by performing limited proteolysis experiments to pinpoint interacting parts of the protein, followed by structural studies thereof.
High-affinity binding is mediated by the extended interaction interface Using ITC, we could measure the affinity between Chp2and Mit1to be in the nanomolar range, which indicates significantly tighter in-teraction than binding energies reported for Swi6and its interaction partners [Isaac et al., 2017]. We performed all experiments measur-ing affinities with the Chp2 crystallization construct, which consists of the chromo shadow domain and proved to be sufficient for com-plex formation. However, we cannot rule out the possibility that the linker region, the chromodomain or the N-terminus of Chp2 are to some extent involved with binding as well. This was first suggested by our yeast two-hybrid screens, where under the most stringent se-lection conditions full-length Chp2 was required for the assay to in-dicate interaction (Figure 8C). While the observed result could also be due to reduced stability of our CSD construct used in the yeast two-hybrid screen compared to full-length protein, we will perform binding experiments with full-length Chp2to test whether affinity for Mit1changes compared to the chromo shadow domain alone.
Our ITC experiments showed that mutation of the motif leads to a ~40 fold reduction, while loss of the Mit1 CDL reduces affinity
~20fold. This clearly supported our interpretation of the crystal
struc-ture to the effect that motif recognition alone is not sufficient to achieve full specificity. Experiments to measure the contribution of just the Mit1linker region are still ongoing, but we expect affinity to be in the same range as observed for other parts of Mit1. Furthermore, our initial yeast two-hybrid experiments suggested that a monomeric form of Chp2 (I370E, [Sadaie et al., 2008]) was still capable of bind-ing Mit1, as were Chp2constructs lacking important residues lining the binding groove (data not shown). We wanted to test this in ITC experiments but due to the instability of purified monomeric Chp2 we did not obtain any results. However, knowing that loss of motif recognition does not completely abolish binding and that Mit1 only contacts both copies of Chp2 within the motif binding groove, we can imagine binding of Mit1to monomeric Chp2 to happen with an affinity similar to the one observed for Mit1I11R. This assumption is based on the finding that the Mit1I11R-Chp2CSD complex showed a stoichiometry of ~1in our ITC experiments, indicating that Mit1did indeed contact only one copy of Chp2.