Polycomb Cbx2 Condensates Assemble through Phase Separation

Polycomb group (PcG) proteins are master regulators of development and differentiation. Mutation and dysregulation of PcG genes cause developmental defects and cancer. PcG proteins form condensates in the nucleus of cells and these condensates are the physical sites of PcG-targeted gene silencing. However, the physiochemical principles underlying the PcG condensate formation remain unknown. Here we show that Polycomb repressive complex 1 (PRC1) protein Cbx2, one member of the Cbx family proteins, contains a long stretch of intrinsically disordered region (IDR). Cbx2 undergoes phase separation to form condensates. Cbx2 condensates exhibit liquid-like properties and can concentrate DNA and nucleosomes. We demonstrate that the conserved residues within the IDR promote the condensate formation in vitro and in vivo. We further indicate that H3K27me3 has minimal effects on the Cbx2 condensate formation while depletion of core PRC1 subunits facilitates the condensate formation. Thus, our results reveal that PcG condensates assemble through liquid-liquid phase separation (LLPS) and suggest that PcG-bound chromatin is in part organized through phase-separated condensates.


Introduction 32
The genome in eukaryotic cells can be broadly classified as euchromatin (active transcription) 33 and heterochromatin (repression and silencing) (1,2). Heterochromatin can be further described 34 broadly as constitutive and facultative heterochromatin (3)(4)(5). Constitutive heterochromatin is 35 observed at and near centromeres and telomeres (5). Facultative heterochromatin is found at a 36 (22)(23)(24)(25)(26). PcG condensates function as specific nuclear compartments for target gene silencing 66 (22)(23)(24)(25)(26). Overall, these biochemical and genetic studies suggest that PRC1 and PRC2 67 coordinate to establish and maintain facultative heterochromatin. Despite these exciting 68 advances, the fundamental physicochemical principles that underpin how PcG proteins 69 establish, maintain, and regulate facultative heterochromatin remain incompletely understood. 70 Spatial and temporal compartmentalization of intracellular components into organelles in 71 eukaryotic cells is a generic theme for organizing biochemical reactions (27)(28)(29)(30)(31)(32). These 72 organelles can be membrane-bound or membraneless. A large number of membraneless 73 compartments, including the nucleolus, stress granules, Cajal bodies, PML nuclear bodies, and 74 others, are condensates formed by condensation of cellular components through liquid-liquid 75 phase separation (LLPS) (27)(28)(29)(30)(31). The forces that drive LLPS are multivalent interactions among 76 proteins and other macromolecular polymers such as . Phase-separated 77 condensates have been shown to be involved in multiple cellular processes and functions (27-78 29). Over the past year, phase-separated membraneless condensates have been suggested to 79 be implicated in transcriptional activation and repression (33)(34)(35)(36)(37)(38)(39). A phase-separated model has 80 been emerging to explain transcriptional activation: transcription factors and coactivators phase 81 separate to form condensates that interact with condensates of RNA polymerase II (RNA Pol II) 82 to efficiently activate transcription (33-37). Phase-separated condensates also function in 83 transcriptional repression. Heterochromatin protein 1 (HP1) phase separates to form 84 condensates that compartmentalize constitutive heterochromatin (38,39). Facultative 85 heterochromatin represents one major class of chromatin structures. Whether the PcG proteins 86 that are responsible for the formation of facultative heterochromatin phase separate to form 87 condensates remains unknown. 88 Here we provide the first experimental evidence that the PRC1 protein Cbx2 phase 89 separates to form condensates that can concentrate DNA and nucleosome. The conserved 90 residues within the Cbx2 IDR are required for the Cbx2 condensate formation in vitro and in 91 vivo. We show that H3K27me3 contributes little to the Cbx2 condensate formation, while 92 depletion of Cbx2-PRC1 subunits facilitates the condensate formation. Thus, our results provide 93 a general experimental framework that can explain how PcG condensates assemble and a 94 starting point for further exploring how phase separation facilitates efficient and specific control 95 of transcription. 96 Results 97

Cbx2 forms condensates in cells 98
We first investigated whether Cbx2 forms condensates in living cells. We integrated YFP-Cbx2 99 and HaloTag (HT)-Cbx2 into the genome of PGK12.1 mouse embryonic stem (wild-type mES) 100 cells, respectively. To observe the cellular distribution of HT-Cbx2, we labelled the fusion protein 101 by HaloTag TMR ligand in living cells. Both YFP-Cbx2 and HT-Cbx2 formed condensates in 102 living wild-type mES cells ( Fig. 1a-b), which is consistent with the previous observations 103 reporting that both exogenous and endogenous Cbx2 forms condensates (40,41). About half of 104 cells contained microscopically visible Cbx2 condensates. The average area of condensates 105 was 0.19 µm 2 (250 nm in radius), and their fluorescence intensity was 1.5-fold higher than the 106 average intensity. It is possible that the formation of Cbx2 fusion condensates in mES cells is 107 due to their overexpression. To resolve this possibility, we integrated YFP-Cbx2 and HT-Cbx2 108 into the genome of Cbx2 / mES cells. The distribution of YFP-Cbx2 and HT-Cbx2 in Cbx2 / 109 mES cells was similar to that in wild-type mES cells ( Fig. 1a-b). These data indicate that Cbx2 110 forms microscopically visible condensates in living cells. 111 Cbx2 forms a stable PRC1 complex (Cbx2-PRC1), including Phc1 and Ring1b (Fig. 1c)

112
(42), so we investigated whether YFP-Cbx2 condensates colocalize with Cbx2-PRC1 subunits. 113 We stained endogenous Ring1b and Phc1 as well as YFP-Cbx2. Ring1b and Phc1 formed 114 condensates in cells (Fig. 1d), consistent with the previous reports (23,41,43,44). YFP-Cbx2 115 condensates colocalized with condensates of Ring1b and Phc1 (Fig. 1d). Since H3K27me3 116 marks PcG-targeted genes, we investigated whether Cbx2 condensates colocalize with 117 H3K27me3. Immunofluorescence of H3K27me3 and YFP-Cbx2 showed that Cbx2 condensates 118 colocalize with chromatin with the dense H3K27me3 mark (Fig. 1d), suggesting that PcG-119 targeted genes are recruited to Cbx2 condensates, or vice versa. Thus, our results show that 120 Cbx2 condensates colocalize Cbx2-PRC1 subunits and H3K27m3-marked chromatin regions. 121 Next, we interrogated whether Cbx2 condensates exhibit liquid-like features that are 122 characterized with rapid exchange kinetics, which can be studied by measuring the recovery 123 rate using fluorescence recovery after photobleaching (FRAP). We performed FRAP 124 experiments on condensates of YFP-Cbx2 stably expressed in mES cells. FRAP showed that 125 80% of YFP-Cbx2 within condensates is recovered within 3 min ( Fig. 1e-f), consistent with our 126 previous reports (40). These results indicate that Cbx2 within condensates dynamically 127 exchange with surrounding environments and has liquid-like properties in cells. 128 If Cbx2 condensates are liquid-like, reducing concentration of YFP-Cbx2 would dissolve 129 these condensates. We lysed cells stably expressing YFP-Cbx2 in lysis buffer to cause a local 130 that treatment of Cbx2 condensates with increasing concentrations of NaCl and Triton X-100 165 causes a reduction in the number of Cbx2 condensates ( Fig. 2e-f). Hexanediol is known to 166 dissolve liquid-like condensates (33,35), possibly through disruption of hydrophobic interactions. 167 We found that treatment of Cbx2 condensates in vitro as well as in mES cells expressing YFP-168 Cbx2 with Hexanediol results in a reduction in the number of condensates (Fig. 2g-h) Cbx2 condensates (Fig. 3a). Previous studies have shown that Cbx2 can compact chromatin on 180 its own (20), so we tested whether Cbx2 can concentrate core nucleosome particles. We 181 prepared dye-labelled nucleosomes and mixed them with Cbx2. After dialysis, we observed that 182 Cbx2 condensates colocalize with the concentrated dye-labelled nucleosomes (Fig. 3a). Under 183 the same conditions, Cbx2 condensates could not enrich dye (Fig. 3a). Thus, these results 184 suggest that Cbx2 condensates can concentrate DNA and nucleosomes in vitro. 185 Given that PcG condensates are the repressive compartments for PcG-targeted genes, 186 we should be able to detect DNA with Cbx2 condensates isolated from cells. We cross-linked 187 cells stably expressing YFP-Cbx2 with formaldehyde. After sonication and centrifugation, we 188 resuspended the pellet and stained DNA with Hoechst. Fluorescence images showed that YFP-189 Cbx2 condensates contain concentrated DNA labelled by Hoechst (Fig. 3b). Our data indicate 190 that Cbx2 condensates can enrich chromatin/DNA within cells. interactions occur between aromatic residues with Lys or Arg residues (47-52). We found that 196 Cbx2 contains 48 Lys (9.0%) and 33 Arg (6.2%), whose frequency is higher than their 197 respective average frequency in vertebrate proteins (6.6% for Lys and 4.9% for Arg) (53). There 198 are 5 Phe (0.9%), 5 Trp (0.9%), and 3 Tyr (0.6%) in Cbx2, whose frequency is lower than their 199 respective average frequency in vertebrate proteins (3.6% for Phe, 1.3% for Trp, and 3.4% for 200 Tyr) (53). Given that proteins whose phase separations are promoted by the cation-pi 201 interactions contain a high content of aromatic residues (47), we focused on the distribution of 202 charged residues of Cbx2. Electrostatic interaction is one of the major driving forces that 203 promote the phase separation of IDR-containing proteins (39,51,(54)(55)(56)(57)(58). In Cbx2, many 204 positively and negatively charged residues are grouped into a series of clusters across the 205 sequence (Fig. 4b). It is interesting to note that the three conserved regions, AT-hook (ATH), 206 ATH-like 1 (ATHL1), and ATHL2 (59), are positively charged clusters ( Fig. 4a-b). We 207 substituted residues PRG (77-79) for AAA to generate Cbx2 ATH , PRG (134-136) for AAA to 208 generate Cbx2 ATHL1 , and RKKRGRK (161-167) for AAAAGAA to generate Cbx2 ATHL2 (Fig. 4b). 209 The net positive charge per residue of the mutated regions of Cbx2 ATH and Cbx2 ATHL1 was 210 slightly reduced compared to Cbx2, while the net positive charge per residue of the mutated 211 region of Cbx2 ATHL2 was completely eliminated (Fig. 4b). We generated these mutant proteins 212 and compared their ability to form condensates with Cbx2 in vitro. Our analysis indicated that 213 the phase-separation ability of the three mutants is greatly reduced compared to Cbx2 (Fig. 4c-214 d). Cbx2 ATH and Cbx2 ATHL1 had a better capacity to phase separate than Cbx2 ATHL2 (Fig. 4c-d), 215 consistent with the complete loss of positive charge in the ATHL2 of Cbx2 ATHL2 . Within the IDR 216 of Cbx2, there is a conserved serine-rich region (SRR) consisting of a stretch of consecutive 19 217 residues of serine and threonine (59). We substituted SKSKSSSSSSSSTSSSSSS (102-120) 218 for SKSKASASASASTASASAA to generate Cbx2 SRR (Fig. 4a). Cbx2 SRR greatly reduced its 219 ability to phase separate compared to Cbx2 ( Fig. 4c-d). Thus, these data indicate that these 220 conserved residues within the IDR of Cbx2 promote LLPS in vitro. 221 To investigate whether these conserved residues of Cbx2 contribute LLPS in vivo, we 222 established wild-type mES cells stably expressing HT-Cbx2 ATH , HT-Cbx2 ATL1 , HT-Cbx2 ATL2 , or 223 HT-Cbx2 SRR . These Cbx2 mutants were labelled with HaloTag TMR ligand. We performed live-224 cell imaging of these mutants (Fig. 4e). Quantitative analysis showed that the size and the 225 number of condensates of these Cbx2 mutants are significantly reduced compared to wild-type 226 Cbx2 ( Fig. 4f-g). We also noted that the size and the number of condensates of Cbx2 ATHL2 and 227 Cbx2 SRR were slightly smaller than Cbx2 ATH and Cbx2 ATHL1 ( Fig. 4f-g), consistent with in vitro 228 analysis. Thus, our data demonstrate the conserved residues within the IDR that are critical for 229 the LLPS of Cbx2 in vitro are also critical for the formation of Cbx2 condensates in vivo. 230 231

H3K27me3 has minimal effects on the formation of Cbx2 condensates 232
The PRC2-catalyzed product H3K27me3 is the marker of PcG-targeted chromatin (7). 233 H3K27me3 has been hypothesized to be the mark for recruiting Cbx-PRC1 to chromatin ( Fig.  234 5a) (60). Thus, we asked whether H3K27me3 affects the formation of Cbx2 condensates in 235 vivo. To this end, we integrated HT-Cbx2 into the genome of Eed / mES cells. Eed is the core 236 component of PRC2 and Eed knockout results in a complete loss of H3K27me3 (15). Live-cell 237 imaging of HT-Cbx2 labelled with HaloTag TMR ligand showed that HT-Cbx2 forms 238 condensates in Eed / mES cells (Fig. 5c). Quantitative analysis indicated that the size and the 239 number of Cbx2 condensates in Eed / mES cells are not significantly different from that in wild-240 type mES cells ( Fig. 5d-e). Previous studies have shown that the aromatic cage, consisting of 241 three aromatic residues, of the chromodomain (CD) domain of Cbx2 is critical for the 242 H3K27me3 binding in vitro (61). We mutated the cage residue Phe-12 of Cbx2 to Ala (Cbx2 F12A ) 243 ( Fig. 5b). We stably expressed HT-Cbx2 F12A in wild-type mES cells. Live-cell imaging showed 244 that HT-Cbx2 F12A forms condensates (Fig. 5c). The size and the number of HT-Cbx2 F12A were 245 similar to HT-Cbx2 ( Fig. 5d-e). These data suggest that H3K27me3 contributes little to the 246 formation of Cbx2 condensates in living cells. 247 Since the CD of Cbx2 is the binding domain for H3K27me3 in vitro (61,62) ( Fig. 5a), we 248 investigated the effects of the CD on the formation of Cbx2 condensates in living cells. We 249 fused CD with HT, generating HT-CD Cbx2 (Fig. 5b). We also deleted CD to generate HT-Cbx2 ∆CD 250 ( Fig. 5b). We established mES cells that stably express HT-CD Cbx2 and HT-Cbx2 ∆CD , 251 respectively. HT-CD Cbx2 did not form condensates in living cells (Fig. 5c). HT-Cbx2 ∆CD phase 252 separated to form condensates (Fig. 5c); however, their size and number significantly reduced 253 in comparison with HT-Cbx2 ( Fig. 5d-e). Given that ATH is adjacent to CD and binds DNA (46) 254 ( Fig. 5b), we deleted both CD and ATH to generate HT-Cbx2 ∆CD-ATH (Fig. 5b), which was then 255 stably integrated into the genome of mES cells. HT-Cbx2 ∆CD-ATH did not phase separate to form 256 condensates within living cells (Fig. 5c). Thus, these results indicate that the interactions of 257 H3K27me3 and Cbx2 are not required for the formation of Cbx2 condensates; however, the 258 amino acid residues within CD are required for the condensate formation. 259 260

Depletion of Cbx2-PRC1 subunits facilitates the formation of Cbx2 condensates 261
Cbx2 phase separates on its own in vitro, so we speculate that removal of Cbx2-PRC1 subunits 262 would not prevent the formation of Cbx2 condensates in vivo. To address this, we integrated 263 HT-Cbx2 into the genome of Ring1a ─/─ /Ring1b fl/fl ; Rosa26::CreERT2 and Bmi1 ─/─ /Mel18 ─/─ mES 264 cells, respectively. Ring1b in Ring1a ─/─ /Ring1b fl/fl ; Rosa26::CreERT2 mES cells was depleted by 265 administrating hydroxytamoxifen as described previously (15,63). Live-cell imaging showed that 266 depletion of Ring1a and Ring1b or Mel18 and Bmil1 does not disperse Cbx2 condensates (Fig.  267   6a). Instead, we found that Cbx2 condensates in these double knockout mES cells are typically 268 more and larger compared to wild-type mES cells (Fig. 6a). Some of Cbx2 condensates in the 269 double knockout mES cells were irregular, instead droplet-like shape. Quantitative analysis 270 demonstrated that both the size and the number of Cbx2 condensates in the double knockout 271 mES cells are significantly larger than that in wild-type mES cells (Fig. 6b-c) The phase behavior of proteins containing IDRs can be described by the theory of 291 associative polymers (47,64). Associative polymers phase separate through interactions 292 between associative motifs called stickers, which are separated from one another by spacers 293 (47,64). Spacers can impart the material properties of polymers and modulate the phase-294 separation ability of polymers (47,64). The stickers can be residues that involve cation-pi, 295 electrostatic, hydrophobic, or dipolar interactions (27,28,47). In the case of Cbx2, the stickers 296 appear to be appositively charged clusters. Perturbation of these charged clusters reduces the 297 phase separation of Cbx2 both in vitro and in vivo, which is consistent with the notion that the 298 phase separation of IDR-containing proteins can be promoted by interactions between blocks of 299 appositively charged residues (39,51,54-58). The SRR also appears to be a sticker since 300 substitution of the Ser residues of SRR with Ala prevents the phase separation of Cbx2 both in 301 vitro and in vivo. Because the content of aromatic residues is low and there is no apparent 302 pattern for aromatic residues across the Cbx2 sequence, we hypothesize that these aromatic 303 residues are unlikely to be stickers. However, further experiments are required to test this 304

hypothesis. 305
It has been long thought that the PRC2-catalyzed product H3K27me3 acts as a binding 306 site for Cbx-PRC1 through its interactions with the Cbx proteins in vivo (61,62). This may not be 307 the case for how Cbx2 is recruited to chromatin. We show that Cbx2 condensates can 308 concentrate DNA and core unmodified nucleosomes and demonstrate that H3K27me3 has 309 minimal effects on the formation of Cbx2 condensates, which is consistent with the previous 310 observations that Cbx2 can bind DNA and compact unmodified oligonucleosomes (20,46). 311 These results also reconcile with our live-cell single-molecule tracking results in which depletion 312 of Eed or Ezh2 has negligible effects on the chromatin bound level of Cbx2, but greatly reduces 313 the bound level of Cbx7 and Cbx8 (15). Given that Cbx2 condensates colocalize with dense 314 H3K27me3-marked chromatin regions, H3K27me3 could play other functions, such as 315 organizing PcG-targeted chromatin within condensates. 316 Depletion of Ring1a and Ring1b or Bmi1 and Mel18 does not impair the formation of 317 Cbx2 condensates, instead increases the size and the number of Cbx2 condensates. Our data 318 also indicate that Cbx2 can form condensates in the absence of Cbx2-PRC1 subunits. 319 Together, these results suggest that the subunits of Cbx2-PRC1 regulate the assembly and 320 structure of Cbx2 condensates. These data can be explained by a client-scaffold model of LLPS 321 (28,65). We propose that Cbx2 is the scaffold and the other subunits of Cbx2-PRC1 are clients. 322 Previous studies have mapped the physical interactions between the subunits within the Cbx-323 PRC1 complexes. It has been suggested that one of Ring1a/Ring1b, Mel18/Bmi1, and 324  To generate recombinant Cbx2 in E.coli, we amplified the Cbx2 sequence by PCR and 371 inserted it to the downstream GST sequence within the pGEX-6P-1-GST vector (GE Healthcare, 372 Pittsburg, PA) to generate pGEX-6P-1-GST-Cbx2. To facilitate double-affinity purification, we 373 added a FLAG tag downstream Cbx2 sequence to generate pGEX-6P-1-GST-Cbx2-FLAG. We 374 amplified the YFP sequence to insert the upstream Cbx2 sequence to generate pGEX-6P-1-375 GST-YFP-Cbx2-FLAG. To generate plasmids for expressing Cbx2 variants in E.coli, we 376 amplified the sequence encoding the Cbx2 variants by PCR and used them to replace the Cbx2 377 sequence in the plasmid pGEX-6P-1-GST-Cbx2-FLAG. We generated the following Cbx2 To determine the critical/saturation concentration of phase separation of Cbx2, a serial 438 of concentrations of Cbx2 (1.2, 2.4, 4.8, and 12 µM) was dialyzed under the same conditions 439 and the number of condensates per frame was counted as described above. To investigate 440 driving forces that contribute the formation of Cbx2 condensates, to 10 µL of the dialyzed 441 sample, NaCl, Triton X-100, and 1, 6-Hexandiol (Sigma-Aldrich, 240117) was added, 442 respectively. The mixture was incubated at 4 °C for 30 min. Condensates were imaged and 443 analyzed as described above. 444 To prepare Cbx2 condensates that concentrate DNA or nucleosome, 4.8 µM of Cbx2 445 was mixed with Alexa 488-labelled DNA (0.5 µM), Cy5-labelled mononucleosome (40 nM), or 446 Alexa 488 (1.0 µM), respectively. The mixture was dialyzed as described above. DIC and 447 fluorescence images were taken by using an Axio Observer D1 Microscope as described above. Images of cells were acquired by using an Axio Observer D1 Microscope as described above. 463 For the excitation and emission of TMR, a Brightline® single-band laser filter set (Semrock; 464 excitation filter: FF01-561/14, emission filter: FF01-609/54, and dichroic mirror: Di02-R561-25) 465 was used. Visible condensates were counted by using ImageJ. Images were presented by using 466 Photoshop. 467 468

Live-cell imaging of YFP-Cbx2 treated with 1, 6-Hexandiol 469
We seeded mES cells stably expressing YFP-Cbx2 to gelatin-coated cover-glass bottom dish 470 24 hours before the imaging. Cell culture medium was replaced with the live-cell imaging 471 medium and maintained at 37 °C using a heater controller. 1, 6-Hexandiol was added to the 472 medium to reach a final concentration of 10%. Image stack was taken at every 2-min interval for 473 20 min by using an Axio Observer D1 Microscope as described above. For the excitation and 474 emission of YFP, an YFP-2427B filter set (Semrock; excitation filter: FF01-500/24, emission 475 filter: FF01-542/25, and dichroic mirror: FF520-Di02) was used. 476

Immunofluorescence 478
Cells stably expressing YFP-Cbx2 were seeded to coverslip and cultured for 24 hours. Cells 479 were fixed with 1.0% paraformaldehyde for 10 min at room temperature. After treatment with 480 0.2% Triton X-100 for 10 min, cells were washed with basic buffer (10 mM PBS pH 7.2, 0.05% 481 Tween 20, and 0.1% Triton X-100, and 0.05% Tween 20) and incubated with blocking buffer 482 (basic buffer supplemented with 3% goat serum and 3% bovine serum albumin) overnight. 177.32 (µs/pixel). Before photobleaching, two images were taken. Immediately after 500 photobleaching, 20 images were taken with 15-s intervals. The images were analyzed using 501 ImageJ. We used TurboReg to correct images for movement in the XY plane. After correcting 502 fluctuations in background and total signal, the fluorescence intensities were normalized to the 503 signal before photobleaching to obtain the fluorescence recovery as described previously 504 (40,63).    for HT-Cbx2 in Eed / mES cells, and for HT-Cbx2 F12A , HT-CD Cbx2 , HT-Cbx2 ∆CD , and HT-833 Cbx2 ∆CD-ATH in wild-type mES cells. Data were obtained from at least 10 cells. P-value was 834 calculated based on student's t-test. 835 (e) Box plot of the number of condensates for HT-Cbx2 in wild-type mES cells replicated from 836 Fig. 4g, for HT-Cbx2 in Eed / mES cells, and for HT-Cbx2 F12A , HT-CD Cbx2 , HT-Cbx2 ∆CD , and 837 HT-Cbx2 ∆CD-ATH in wild-type mES cells. Data were obtained from at least 10 cells. P-value was 838 calculated based on student's t-test. 839