The interplay between H2A.Z and H3K9 methylation in regulating HP1α binding to linker histone-containing chromatin

Expression and purification of recombinant histone H1.4

The construction of the human H1.4 expression plasmid and the subsequent expression and purification of the protein have been described in detail in (38).

Expression and purification of recombinant wild-type HP1α and HP1α3KA

The pET15b expression plasmids containing HP1α and HP1α^3KA were transformed into Escherichia coli Rosetta2 (DE3) (Novagen) cells and spread onto LB-agar containing 50 μg/ml ampicillin (amp) and 34 μg/ml chloramphenicol (cml). After overnight incubation at 37°C, a single colony was picked, and used to inoculate 10 ml LB liquid media containing amp (50 μg/ml) and cml (34 μg/ml) and incubated at 37°C for ∼8 h. The starter culture was used to inoculate 1 l of ZYP-5052 auto-induction media (39) containing amp (50 μg/ml) and cml (34 μg/ml) in a 2 l flask. The culture was grown at 37°C overnight. The cells were collected by centrifugation at 5000 g for 12 min at 4°C, washed once with 25 ml cold PBS, and then the cell pellet was resuspended in ∼25 ml lysis buffer (20 mM Tris pH 7.5, 350 mM NaCl, 0.05% (v/v) β-mercaptoethanol, 20 mM imidazole, 0.1 mM 4-(2-aminoethyl)benzenesulfonyl fluoride hydrochloride, 1 μM pepstatin, 1 μM bestatin, 1 μM E-64) and either frozen and stored at −20°C or processed as follows. Cells were lysed by thawing cell suspensions then adding lysozyme (50 μg/ml) and DNaseI (10 μg/ml), followed by three cycles of freeze/thawing in liquid N₂ or the addition of 1× BugBuster (10× stock; MerckMillipore) reagent. The suspension was then incubated on ice for 20 min, followed by sonication on ice for 3–5 min, cycling on for 10 s and off for 20 s, at 30% power on a Branson sonifier. The lysate was centrifuged at 35 000 g for 25 min at 4°C, and the supernatant was carefully decanted and applied to a 2.5 ml Ni-NTA resin column pre-equilibrated in lysis buffer and allowed to flow under gravity. The flow-through was reapplied to the column, followed by washing with ∼10 column volumes of lysis buffer. The protein was eluted from the column in lysis buffer containing 250 mM imidazole. The eluate was exchanged to 20 mM Tris pH 7.5, 350 mM NaCl, 0.05% (v/v) β-mercaptoethanol and the sample treated with 30 units thrombin (Sigma) overnight at room temperature to remove the N-terminal hexahistidine-tag. The cleaved protein was then purified by anion exchange chromatography using 1 ml HiTrap Q cartridges (GE Healthcare) in 20 mM Tris pH 8.5 and 1 mM DTT over a 0–1 M NaCl gradient. Fractions containing full-length HP1α proteins were collected and concentrated in 10 kDa MWCO centrifugal concentrators and the protein purified further via gel filtration on a Superdex200 10/300 column (GE Healthcare) in 20 mM Tris pH 7.5, 250 mM NaCl, 1 mM DTT. HP1α-containing fractions were pooled and concentrated in 10 kDa MWCO centrifugal concentrators to a final concentration of 500 μM. Glycerol was added to a final concentration of 10% (v/v) and the protein aliquoted, snap frozen, and stored at −80°C. Protein concentration was determined by absorbance at 280 nm and using an extinction coefficient of 23.16 mM⁻¹ cm⁻¹.

Production of recombinant histones and histone octamers

Histone octamers were assembled using standard protocols from purified recombinant histones (40,41), either as unlabelled proteins (for mononucleosomes) or containing AlexaFluor488-labelled H2A (for arrays). AlexaFluor488 was attached to H2A via a single cysteine residue introduced at position 120, as described previously in (42). Labelling efficiency was ∼65–70%. The labelled octamers were mixed at a 1:4 ratio with unlabelled octamers such that only ∼2 nucleosomes per array were labelled with AlexaFluor488. H3K9_cme3 and H3K9_cme0 were produced as described in (43). Octamers were stored at −20°C in 10 mM Tris pH 7.5, 1 mM EDTA, 1.5 M NaCl, 40% glycerol, 1 mM DTT.

In vitro nucleosome and nucleosome array assembly

Nucleosome (12N₂₀₀) arrays were assembled on a 2400 bp DNA fragment consisting of twelve 200 bp tandem repeats that each contain the 147 bp 601 nucleosome positioning sequence (44) and 53 bp flanking DNA. The array fragment was propagated in a pUC18 vector, using standard methods. The array fragment was excised from the pUC18 vector backbone via digestion with EcoRV, and the backbone DNA cut into smaller fragments using DraI and HaeII enzymes. The array fragment was then separated on a 1× TAE 1.3% (w/v) agarose gel, the band excised and the DNA extracted using electroelution. The purified DNA fragment was then concentrated by ethanol precipitation and resuspended in 10 mM Tris pH 8.5 and stored at −20°C.

Mononucleosomes were assembled on AlexaFluor488-labelled DNA fragments that corresponded to the 147 bp 601 nucleosome positioning sequence (44) alone (N₁₄₇) or the 601 sequence flanked by 28 bp of additional DNA on either side (N₂₀₃). The fragments were produced by performing 96 parallel PCR reactions in 96-well plates using one AlexaFluor488-labelled and one unlabelled primer and MyTaq DNA polymerase (Bioline; as per the manufacturer's instructions). The PCR products were purified on native PAGE gels; 0.5× Tris–borate–EDTA (TBE; 44.5 mM Tris-Borate, 1 mM EDTA) 5% (w/v) acrylamide. The desired product bands were excised and the DNA extracted via electroelution in 0.5× TBE at room temperature for 30–60 min at 70–100 V. The DNA was concentrated by ethanol precipitation, resuspended in 10 mM Tris pH 8.5 and stored at −20°C.

Assembly of the 12N₂₀₀, N₂₀₃ and N₁₄₇ DNA fragments into the various nucleosome arrays/mononucleosomes (±H1.4) was performed by salt-gradient dialysis, as we have described in (38). Mononucleosome assembly reactions were analysed on 0.2 × TBE 5% (w/v) native PAGE gels run at 150 V at 4°C for 45–60 min. Nucleosome array assemblies were analysed on 0.2× TBE 0.8% (w/v) agarose gels run at 110 V for 90 min at 4°C. The gels were visualized by scanning for fluorescence AlexaFluor488 on a Typhoon™ FLA9000 laser scanner.

Electrophoretic mobility shift assays (EMSAs)

Dilutions of HP1α (0–20 μM) and HP1α^3KA (0–20 μM for array binding and 0–30 μM for mononucleosome binding) were incubated with AlexaFluor488-labelled 12N₂₀₀ (±H1.4) arrays (4.7 nM; effective nucleosome concentration is 56 nM) or mononucleosomes (50 nM; N₂₀₃, N₂₀₃+H1.4, N₁₄₇) in gel shift buffer (10 mM Tris pH 7.5, 25 mM NaCl, 25 mM KCl, 0.2 mM DTT, 4% (w/v) Ficoll 400) on ice for 20–30 min. Mononucleosome binding reactions were loaded onto 5% (w/v) acrylamide:bisacrylamide (49:1) gels and electrophoresed in 0.2× TBE at 150 V for 90 min at 4°C. Nucleosome array binding reactions were loaded onto 0.8% (w/v) agarose gels and electrophoresed in 0.2× TBE at 70 V for 180 min at 4°C. All gel images were acquired on an FLA9000 laser scanner (GE Healthcare) scanning for AlexaFluor488 signal.

Quantitation of gels was performed by measuring the amount of unbound (mononucleosome or nucleosome array) signal in each gel lane using ImageJ (45). Normalization across replicate experiments was performed as follows: the signal in the mononucleosome/nucleosome array-only lane (i.e. the 0 μM HP1α binding reaction) was set as the 100% unbound signal in each individual experiment. The fraction unbound for each binding condition in an experiment was then normalized relative to this signal. These values were then converted to fraction bound by subtracting them from one. Normalized data sets from replicate experiments were then fit simultaneously using the concatenated fit option in MicroCal OriginPro 2017 with the in-built Hill equation, which is of the form:

where ‘max. bound’ is the maximum level of binding, and represents the upper asymptote of the binding isotherm (upper limit of one); K is the concentration of HP1α at half-maximal binding and is indicative of the relative affinity constant; n is the Hill coefficient. Supplementary Table S1 provides an overview of the results for each experiment.

Linker histone H1.4 interferes with binding of HP1α to mononucleosomes

Linker histones have been reported to interact with both the chromodomain and the hinge region of HP1α (20,21,37). However, these studies did not examine these interactions in the context of nucleosomes. Thus, we wanted to investigate how linker histone H1.4 affects the binding of HP1α to nucleosomes.

First, we assembled mononucleosomes, either in the absence or presence of a stoichiometric amount of linker histone H1.4 (Figure 1A). These nucleosomes contained a core 147 bp 601 positioning sequence (44) flanked by 28 bp of linker DNA on either side, yielding a 203 bp symmetric nucleosome (N₂₀₃); these nucleosomes were labelled with AlexaFluor488 on one end. We chose 28 bp of flanking DNA because shorter linker DNA lengths have been demonstrated to decrease the affinity of linker histones for nucleosomes (46). Figure 1A shows that H1.4-bound nucleosomes have an increased electrophoretic mobility (Figure 1A, lane 2) compared to the linker histone-free species (Figure 1A, lane 1). This is consistent with the linker histone interacting with and constraining the linker DNA, resulting in a more compact particle.

Next, we performed electrophoretic mobility shift assays (EMSAs) using these two different nucleosome species (N₂₀₃ and N₂₀₃+H1.4), incubating them with increasing concentrations of HP1α (noting that HP1α exists as a dimer, Figure 1B). Given the reports of a positive and direct interaction between HP1α and linker histones (20,21,37), we anticipated the presence of H1.4 would enhance the interaction between HP1α and nucleosomes. Surprisingly, this was not the case. HP1α shifts N₂₀₃ nucleosomes at much lower concentrations when H1.4 is absent (indicating a higher relative affinity); compare lanes 2–7 in the upper and lower panel of Figure 1B. Despite this apparent decrease in binding affinity in the presence of H1.4, HP1α forms a more discrete complex with the H1.4-associated nucleosomes; suggesting that either H1.4 alters the interaction to yield a more stable HP1α–nucleosome complex, or that the number of available binding sites for HP1α decreased in the presence of H1.4.

It is well-documented that HP1α is able to interact with DNA (18,19,26). One possible explanation for the reduced affinity of HP1α for H1.4-associated nucleosomes is that the linker histone inhibits HP1α from accessing the linker DNA arms of the nucleosome. To examine this, we assembled 147 bp nucleosomes that lack linker DNA (N₁₄₇), and assessed binding of HP1α by EMSA (Figure 1C). The binding of HP1α to N₁₄₇ compared to both N₂₀₃ and N₂₀₃+H1.4 nucleosomes is visibly different, and displays a lower apparent affinity (Figure 1C).

To gain a more detailed picture of these differences, we generated binding curves from replicate EMSA experiments and fitted these with the Hill equation (Figure 1D) to determine the relative affinity (K) and cooperativity (Hill coefficient) of the HP1α interactions. These data are presented in Supplementary Table S1 (which summarizes data from the ∼140 individual EMSA experiments conducted in this study). This analysis reveals that HP1α binding to N₁₄₇ (K = 2.29 ± 0.36 μM) is almost 4-fold weaker than to N₂₀₃ (K = 0.58 ± 0.02 μM) and 2-fold weaker than to N₂₀₃+H1.4 (K = 1.26 ± 0.04 μM). Interestingly, HP1α exhibits cooperative binding to N₂₀₃ in both the absence and presence of H1.4 (Hill coefficients = 1.90 ± 0.10 and 1.88 ± 0.11, respectively), but not to N₁₄₇ (Hill coefficients = 1.21 ± 0.16). These data indicate that the interactions between HP1α and the linker DNA are important for nucleosome binding. However, the effect of H1.4 on HP1α nucleosome binding is not equivalent to the complete loss of linker DNA, suggesting that H1.4 only partially occludes linker DNA accessibility in a mononucleosome.

The hinge domain of HP1α drives association with nucleosomes

Next, we wanted to confirm that the interactions between HP1α and the linker DNA is central for nucleosome binding. The hinge region (Figure 2A) that links the N-terminal chromodomain and C-terminal chromoshadow domain of HP1α is known to bind DNA and is important for interactions with nucleosomes (19,47). It contains a series of conserved basic residues (Figure 2A) and mutation of these basic residues has previously been shown to affect HP1α−DNA/chromatin interactions (19,26,47). We expressed and purified a triple lysine to alanine hinge-region mutant of HP1α (HP1α^3KA; Figure 2A) and confirmed this mutation affects the DNA-binding activity of HP1α (Supplementary Figure S1B).

We performed quantitative EMSA experiments (Figure 2B) with HP1α^3KA and the same three mononucleosome species (N₂₀₃, N₂₀₃+H1.4, and N₁₄₇) used for the experiments in Figure 1. Consistent with previous reports (19,47), HP1α^3KA has a significantly diminished capacity for binding nucleosomes (Figure 2C). Further, the relative affinity of HP1α^3KA for N₂₀₃, N₂₀₃+H1.4 and N₁₄₇ is very similar (K = 9.83 ± 0.42; 12.10 ± 0.83; 10.62 ± 1.75 μM, respectively). This differs from wild-type HP1α, which has strong preference for binding N₂₀₃ mononucleosomes (Figure 1D). This means the HP1α^3KA mutation disproportionately affects binding to nucleosomes with more available linker DNA; the affinity for N₂₀₃ nucleosomes is reduced more than for N₂₀₃+H1.4, which in turn is affected more than N₁₄₇ nucleosomes (17-, 10- and 5-fold decrease relative to wild-type HP1α, respectively; Figure 2C). However, the interaction of HP1α^3KA with N₁₄₇ is still substantially weakened, suggesting there is a component to the hinge-mediated interaction that involves the nucleosome core. Clearly, the hinge region plays a critical role in HP1α binding to nucleosomes and this is mainly driven through interactions with the linker DNA.

Linker histones significantly inhibit binding of HP1α to nucleosome arrays

The action of linker histones extends beyond the single nucleosome, with a primary function to modulate higher-order conformations across multiple nucleosomes. In addition, HP1 proteins have been demonstrated to have enhanced interactions with multi-nucleosome substrates compared to mononucleosomes (25,27,47). Thus, we next examined the effect of linker histones on the binding of HP1α to nucleosome arrays. For this, we used arrays consisting of twelve 200 bp repeats, each containing a core 601 nucleosome positioning sequence (12N₂₀₀). To simplify in-gel visualization and quantitation in our EMSA experiments, we produced arrays containing H2A site-specifically labelled with AlexaFluor488, as we have described previously (38). To minimize any possible adverse effects due to the AlexaFluor488 moiety, we diluted our AlexaFluor488-H2A histone octamer preparations with unlabelled histone octamers thus limiting the number of labelled nucleosomes per 12N₂₀₀ array to approximately two.

Comparable to the mononucleosome experiments (Figure 1), we observed that the binding of HP1α to 12N₂₀₀ arrays is significantly diminished when the arrays are assembled with a stoichiometric amount of H1.4 (Figure 3A). Quantitation of this effect (Figure 3B; Supplementary Table S1) reveals that the affinity of HP1α for the 12N₂₀₀ arrays (K = 0.71 ± 0.02 μM) is reduced 3-fold when H1.4 is present (12N₂₀₀+H14; K = 2.15 ± 0.23 μM). Yet, more striking is the fact that binding of HP1α to the H1.4-saturated arrays is non-cooperative (Hill coeff. = 1.32 ± 0.13), whereas strong positive cooperativity is exhibited on the linker histone-free arrays (Hill coeff. = 3.52 ± 0.29). This differs from the behaviour on N₂₀₃ mononucleosomes where some cooperativity is observed irrespective of whether H1.4 is present or not (Figure 1B; Supplementary Table S1). On the other hand, binding of HP1α to N₁₄₇ mononucleosomes is non-cooperative (Figure 1C, D). This suggests that H1.4 limits linker DNA accessibility in the 12N₂₀₀ arrays more so than in N₂₀₃ mononucleosomes and that the mode of HP1α binding to 12N₂₀₀+H1.4 arrays is most similar to the mode of binding to N₁₄₇ mononucleosomes, namely core particles that lack linker DNA.

We then assessed the binding of HP1^3KA to the 12N₂₀₀ arrays (± H1.4). HP1^3KA binds 12N₂₀₀ arrays more weakly than wild-type HP1α (Figure 3C; Supplementary Table S1), approximately 8-fold less for 12N₂₀₀ (K = 5.62 ± 0.26) and 7-fold less for 12N₂₀₀+H1.4 (K = 14.29 ± 5.58). These data also show that HP1α^3KA has a definite preference for the linker histone-free arrays, binding 12N₂₀₀ ∼2.6-fold more strongly than 12N₂₀₀+H1.4. This is different to the behaviour of HP1α^3KA on mononucleosomes, where it binds both the N₂₀₃ and N₂₀₃+H1.4 mononucleosomes with similar affinity (Figure 2B, C). This suggests that there is an additional component to the HP1α-array interaction that does not participate when interacting with mononucleosomes. This interaction(s) most likely involves connections that span multiple nucleosomes, and it is not directly dependent on the hinge region. This hinge independent interaction is disrupted by the presence of H1.4, since both wild-type HP1α and HP1α^3KA display a 2.5–3-fold reduction in binding affinity for 12N₂₀₀+H1.4 arrays compared to 12N₂₀₀ arrays (Supplementary Table S1).

Next, we investigated whether these effects were specific to H1.4 or were a more general property of linker histones. To this end, we assembled our 12N₂₀₀ arrays in the presence of an alternative linker histone, H1.0. We chose H1.0 as it is significantly sequence-divergent from H1.4 (∼37% sequence identity), as well as being biologically distinct; unlike H1.4, H1.0 is expressed independently of replication and is predominantly found in terminally differentiated cells (36). EMSAs (Figure 3D) using the 12N₂₀₀+H1.0 arrays showed that HP1α displays the same interaction pattern as with 12N₂₀₀+H1.4 arrays (Figure 3A; lower), indicating the effects on HP1α binding are not linker histone subtype-specific.

Linker histone stoichiometry modulates HP1α binding to nucleosome arrays

The experiments conducted thus far have all been in the context of a stoichiometric amount of linker histone, that is, each nucleosome is associated with one linker histone molecule. However, in cells this may not always be the case. Numerous studies have measured total linker histone stoichiometry in different cellular systems, with estimates ranging from ∼0.5–1.3 linker histone molecules per nucleosome (reviewed in (48)). We therefore explored whether varying linker histone stoichiometry can affect HP1α binding to nucleosome arrays.

We assembled a series of 12N₂₀₀ arrays with differing H1.4-to-nucleosome ratios (1.26, 1.02, 0.78, 0.54, 0.27 and 0; Figure 4A) and then performed EMSAs with HP1α (Figure 4B). It should be noted that the H1.4-to-nucleosome ratio of 1.26 represents the fully saturated nucleosome array, as we have described previously (38), whereas the 1.02, 0.78, 0.54, and 0.27 ratios are approximately equivalent to 80, 60, 40 and 20% of the saturating amount of H1.4, respectively.

These experiments show that there is an apparent gradual weakening of HP1α binding to the nucleosome arrays as the H1.4-level increases. Quantitative analysis of these EMSAs (Figure 4C, D) reveals that both the affinity of the interaction and binding cooperativity undergo non-linear transitions as H1.4 is varied. At low levels of H1.4 saturation (less than ∼50%) only modest weakening of HP1α binding (relative to the affinity for the H1.4-free array) is observed; ∼1.6-fold increase in K at 0.54 H1.4 per nucleosome (Figure 4D). However, at the very next increment (0.78 H1.4 per nucleosome), a much greater increase in K (∼2.8-fold) occurs, with further increases as the array becomes saturated with H1.4 (Figure 4D). Similarly, the positive cooperativity of binding is unaffected at the lowest ratio of H1.4 (0.27 H1.4 per nucleosome), but then a transition to a non-cooperative interaction occurs as the level of H1.4 is increased across the range of 0.54–1.02 H1.4 per nucleosome (Figure 4D). These data show that variations in linker histone stoichiometry are able to modulate the binding of HP1α to nucleosome arrays.

H3K9me3 and the histone variant H2A.Z both enhance binding of HP1α to nucleosome arrays

Given the established association between HP1α and the constitutive heterochromatin-associated histone modification H3K9me3, we next investigated whether this modification has any effect on the linker histone-mediated modulation of HP1α binding. We employed site-specific cysteine alkylation to install a trimethylated lysine analogue at the H3K9 position (H3K9_cme3; Supplementary Figure S2A and B) (43,49), as has been used previously to study Swi6- and HP1β-chromatin interactions (12,25). We also produced the non-methylated equivalent, H3K9_cme0, and confirmed that the methylation procedure had no non-specific effect on HP1α array binding (Supplementary Figure S2C and D). We then assembled 12N₂₀₀ arrays containing H3K9_cme3 (± H1.4) (Supplementary Figure S3A and B) and performed EMSA analysis with HP1α (Figure 5A).

When H3K9_cme3 is incorporated into the 12N₂₀₀ arrays HP1α binding is enhanced relative to the unmodified counterpart (compare Figure 5A with Figure 3A). This is both in terms of the HP1α concentration at which ‘shifting’ of the 12N₂₀₀ band occurs (compare lanes 2–4 in the upper panels of Figures 3A and 5A) and also in the ‘sharpness’ of the band of the shifted complex—the final H3K9_cme3 12N₂₀₀-HP1α complex appears as a much sharper homogeneous band in the gel (compare lanes 15–17 in the upper panels of Figures 3A and 5A). This indicates that H3K9_cme3 may also ‘lock-in’ HP1α at its preferred binding location.

Quantitative analysis (Figure 5B) shows H3K9_cme3 enhances HP1α binding to the 12N₂₀₀ array ∼2.5-fold relative to the unmodified array (H3K9_cme3 K = 0.31 ± 0.02 versus unmodified K = 0.71 ± 0.02; Supplementary Table S1). However, this enhancement only occurs in the absence of H1.4. When H1.4 is incorporated into the H3K9_cme3 12N₂₀₀ arrays (Figure 5A, lower panel), HP1α binding is inhibited with the relative affinity of HP1α for the H3K9_cme3 H1.4-saturated arrays (K = 2.30 ± 0.29) being comparable to the unmodified counterpart (K = 2.15 ± 0.23; Figure 5B). Therefore, H3K9_cme3 cannot overcome the H1.4-mediated inhibition of HP1α binding to 12N₂₀₀ arrays.

We have previously reported that the histone variant H2A.Z, like HP1α, is an important component of pericentric heterochromatin (28) and can enhance the binding of HP1α to chromatin in vitro (27). Thus, we investigated whether the incorporation of H2A.Z into our nucleosome arrays could influence HP1α binding when H1.4 is present (Supplementary Figure S3A and B). In these experiments, H2A.Z only accounts for 80% of the H2A in the arrays. This is because the fluorescent label (AlexaFlour488) is incorporated into our arrays via maleimide linkage to a site-specific cysteine mutant of H2A (H2AT120C), which we then mix one-to-four with unlabelled H2A, as described above. In this instance, we replaced the unlabelled H2A with unlabelled H2A.Z. However, this scenario is perhaps a better reflection of the in vivo situation where interspersion of core histones and histone variants occurs (50,51).

Incorporation of H2A.Z into the 12N₂₀₀ arrays results in a significant enhancement of HP1α binding compared to the unmodified arrays (compare Figures 5C and 3A). Results also show that band shifting occurs at lower concentrations and a sharper homogeneous final complex band is observed (compare Figures 3A and 5C, lanes 15–17, upper panels). Interestingly, H2A.Z increases the relative affinity of HP1α for the linker histone-free 12N₂₀₀ arrays ∼2.5-fold (K = 0.28 ± 0.01; Figure 5D, Supplementary Table S1), which is the same as H3K9_cme3. However, when H1.4 is present H2A.Z has little effect, and HP1α binding to the array is still significantly inhibited (Figure 5C and D). Again, this is similar to the observations made on H3K9_cme3 12N₂₀₀ arrays. These data indicate that H2A.Z is able to functionally mimic the effect of H3K9me3 on the binding of HP1α to nucleosome arrays.

Notably, the HP1α^3KA hinge mutant is still unable to bind to H3K9_cme3 or H2A.Z 12N₂₀₀ arrays. The relative affinity of HP1α^3KA for the H3K9_cme3 (K = 7.56 ± 0.29) or H2A.Z (K = 5.73 ± 0.0.64) 12N₂₀₀ arrays is comparable to the unmodified arrays (K = 5.62 ± 0.26; Supplementary Table S1). The same trend is also observed when H1.4 is present in the arrays. Thus, the hinge region of HP1α is required for the proper binding of HP1α to H3K9_cme3 and H2A.Z nucleosome arrays.

Concerning the affinity of histone H1.4 to H2A.Z-containing arrays, two studies have produced conflicting observations. In one study, H2A.Z inhibited the binding of histone H1 (52) whereas in the second study, H2A.Z did not (53). Our results are consistent with the latter study where histone H1 binds equally well to canonical arrays, H2A.Z-containing arrays and arrays modified with H3K9me3 (Figure 4, Supplementary Figure S3A and B).

Combination of H3K9me3 and H2A.Z partially reverses H1.4-mediated inhibition of HP1α binding to nucleosome arrays

H2A.Z and H3K9me3 are both components of constitutive heterochromatin and can co-exist in the same nucleosomes (27,28,54). Therefore, we investigated whether H2A.Z and H3K9me3 could work cooperatively to enhance HP1α binding. We produced 12N₂₀₀ arrays containing both H2A.Z and H3K9_cme3 (± H1.4) and performed quantitative EMSAs with HP1α. The relative affinity of HP1α for H2A.Z/H3K9_cme3 12N₂₀₀ arrays (K = 0.23 ± 0.01 μM) was not significantly different to arrays containing H2A.Z or H3K9_cme3 alone (Figure 5E; Supplementary Figure S4 and Supplementary Table S1). However, when the H2A.Z/H3K9_cme3 arrays are saturated with H1.4, there is a marked increase (approximately two-fold) in HP1α binding affinity over the other H1.4-saturated arrays (K = 1.15 ± 0.09 μM; Figure 5E; Supplementary Figure S3 and Supplementary Table S1). Furthermore, a similar trend is observed with the affinities of the HP1α^3KA mutant: enhanced binding only occurs on the H1.4-saturated H2A.Z/H3K9_cme3 12N₂₀₀ arrays, although, the overall binding affinity of HP1α^3KA (K = 6.04 ± 0.70 μM) is still much weaker than wild-type HP1α (K = 1.15 ± 0.09 μM; Supplementary Table S1). Thus, the combination of H2A.Z/H3K9_cme3 enhances the interaction of HP1α with nucleosome arrays only when saturated with H1.4, and this enhancement appears independent of the hinge region of HP1α.

Interestingly, while the H2A.Z/H3K9_cme3 combination partially restores HP1α binding affinity to the H1.4-saturated arrays, there is no re-establishment of the positive cooperative binding of HP1α (Supplementary Table S1). This suggests the combination of H2A.Z/H3K9_cme3 enhances binding affinity to individual nucleosomes in the array, but does not affect HP1α interactions that bridge multiple nucleosomes.

To investigate this further, we returned to examining HP1α binding to mononucleosomes. We assembled N₂₀₃, N₂₀₃+H1.4 and N₁₄₇ mononucleosomes with either H3K9_cme3, H2A.Z, or the H2A.Z/H3K9_cme3 combination and performed quantitative EMSAs with HP1α (Supplementary Figure S5A–C, Figure 5F, Supplementary Table S1). Unlike 12N₂₀₀ arrays, the incorporation of H3K9_cme3, H2A.Z, or the H2A.Z/H3K9_cme3 combination in N₂₀₃ mononucleosomes has no effect on the binding of HP1α either in the presence or absence of H1.4—the relative affinities are all approximately the same as the unmodified counterpart nucleosomes (Figure 5F). Further, the HP1α^3KA mutant binds each of the H2A.Z, H3K9_cme3 or H2A.Z/H3K9_cme3 N₂₀₃ mononucleosome species poorly, and displays no clear preferences (Supplementary Table S1). This is consistent with HP1α-linker DNA interactions being the dominant mechanism upon which HP1α binds to these mononucleosomes (see Figures 1 and 2). Therefore, the positive effect of H3K9_cme3 and H2A.Z on HP1α binding to 12N₂₀₀ arrays is not preserved in mononucleosomes with linker DNA.

The behaviour of HP1α changes on N₁₄₇ mononucleosomes. The incorporation of H2A.Z into N₁₄₇ nucleosomes results in a 1.5-fold increase in the relative binding affinity of HP1α over the unmodified N₁₄₇ nucleosomes, whereas H3K9_cme3 elicits no change (Figure 5F; Supplementary Table S1). Most significantly, when H3K9_cme3 is combined with H2A.Z, a ∼2.5-fold increase in HP1α binding affinity is observed (relative to unmodified or H3K9_cme3 N₁₄₇ nucleosomes). This result parallels that of HP1α binding to the H1.4-saturated 12N₂₀₀ H2A.Z/H3K9_cme3 arrays (Figure 5E) and is consistent with the H2A,Z/H3K9_cme3 combination enhancing binding to individual nucleosome core particles. These results also indicate that 12N₂₀₀+H1.4 arrays essentially behave like linker DNA-free N₁₄₇ mononucleosomes.

Taken together, these data suggest that there are two alternative modes of interaction of HP1α with nucleosomes. The first mode is through interactions with linker DNA. The second mode involves HP1α binding directly to the nucleosome core region when linker DNA is not available (either missing or occluded by linker histones). This second interaction mode can be promoted by H2A.Z and further enhanced by H3K9me3.