Phase separation by the Sterile Alpha Motif of Polyhomeotic compartmentalizes Polycomb Group proteins and enhances their activity

Polycomb Group (PcG) proteins organize chromatin at multiple scales to regulate gene expression. A conserved Sterile Alpha Motif (SAM) in the Polycomb Repressive Complex 1 (PRC1) subunit Polyhomeotic (Ph) is important for chromatin compaction and large-scale chromatin organization. Like many SAMs, Ph SAM forms helical head to tail polymers, and SAM-SAM interactions between chromatin-bound Ph/PRC1 are believed to compact chromatin and mediate long-range interactions. To understand mechanistically how this occurs, we analyzed the effects of Ph SAM on chromatin in vitro. We find that incubation of chromatin or DNA with a truncated Ph protein containing the SAM results in formation of concentrated, phase-separated condensates. Condensate formation depends on Ph SAM, and is enhanced by but not strictly dependent on, its polymerization activity. Ph SAM-dependent condensates can recruit PRC1 from extracts and enhance PRC1 ubiquitin ligase activity towards histone H2A. Overexpression of Ph with an intact SAM increases ubiquitylated H2A in cells. Thus, phase separation is an activity of the SAM, which, in the context of Ph, can mediate large-scale compaction of chromatin into biochemical compartments that facilitate histone modification.

Polycomb Group (PcG) proteins organize chromatin at multiple scales to regulate gene expression. A 20 conserved Sterile Alpha Motif (SAM) in the Polycomb Repressive Complex 1 (PRC1) subunit 21 Polyhomeotic (Ph) is important for chromatin compaction and large-scale chromatin organization. Like 22 many SAMs, Ph SAM forms helical head to tail polymers, and SAM-SAM interactions between 23 chromatin-bound Ph/PRC1 are believed to compact chromatin and mediate long-range interactions. To 24 understand mechanistically how this occurs, we analyzed the effects of Ph SAM on chromatin in vitro. 25 We find that incubation of chromatin or DNA with a truncated Ph protein containing the SAM results in 26 formation of concentrated, phase-separated condensates. Condensate formation depends on Ph SAM, 27 and is enhanced by but not strictly dependent on, its polymerization activity. Ph SAM-dependent 28 condensates can recruit PRC1 from extracts and enhance PRC1 ubiquitin ligase activity towards histone 29 H2A. Overexpression of Ph with an intact SAM increases ubiquitylated H2A in cells. Thus, phase 30 separation is an activity of the SAM, which, in the context of Ph, can mediate large-scale compaction of 31 chromatin into biochemical compartments that facilitate histone modification. the Pcgf subunit 4 . At least in mouse embryonic stem cells, ncPRC1 is responsible for the bulk of 49 ubiquitylated H2A 7,8 . This suggests histone modification and chromatin organization may be partitioned 50 between nc and cPRC1s, although both types of complexes share many genomic targets 7-10 . All cPRC1 51 subunits can interact with DNA and/or chromatin, and both canonical and non-canonical PRC1s can 52 compact chromatin in vitro 9,11 , but Polyhomeotic (Ph), and thus canonical PRC1, is the most implicated 53 in large-scale chromatin organization 3,12-17 . 54 Ph is a core subunit of canonical PRC1, and its most notable feature is the presence of a 55 conserved Sterile Alpha Motif (SAM) in its C-terminus that can assemble into helical polymers 18 . SAMs 56 are present in many different types of proteins and in many cases can mediate protein polymerization 57 19 . The SAM of Ph is required for Ph function in Drosophila, and its full polymerization activity is 58 important for gene repression 20,21 . PRC1 forms visible foci both in Drosophila and in mammalian 59 cells 13,22 , and, in Drosophila cells, a much larger number of diffraction-limited clusters 14 . Disrupting the 60 Ph SAM impairs formation of PcG protein clusters and reduces long-range contacts among PcG bound 61 loci, suggesting the two processes are related 13,14 . Despite the wealth of in vivo data supporting the 62 phases" (Fig. 1I, J; Supplementary Fig. 2F, G). This produces a limited coarse-grained delineation of the 107 boundary between one-and two-phase regimes. Phase separation is sensitive to the concentration of 108 both components, and the ratio between the two. This is most notable for Mini-Ph-DNA titrations, 109 where we are able to add high concentrations of DNA, which prevent phase separation (Supplementary 110 Fig. 2F, G). From similar titrations of NaCl and Mini-Ph at a fixed DNA concentration, we find that phase 111 separation is observed in NaCl concentrations up to 125 mM ( Supplementary Fig. 3). We conclude that 112 Mini-Ph forms phase separated condensates with either DNA or chromatin. 113 The disordered linker connecting Ph SAM to the FCS domain was previously demonstrated to 114 restrict Ph SAM polymerization, possibly due to its ability to contact Ph SAM in trans 5 . A scrambled 115 linker has the same effect, implicating amino acid composition rather than organization 5 . The sequence 116 properties of linkers that connect structured domain play a central role in phase separation 48 , by 117 restricting or promoting interactions between structured domains, and by contributing weak 118 interactions 49 . We therefore analyzed the sequence properties of the linker ( Supplementary Fig. 4), 119 both in Drosophila Ph, and in the three human homologues (PHC1-3). The Ph linker is acidic (pI 3.9), but 120 relatively uncharged (fraction charged residues (FCR) =0.15), and does not have strongly segregated 121 charge ( Supplementary Fig. 4B, E, Supplementary Table 1). Overall, the Ph linker is expected to be 122 collapsed ( Supplementary Fig. 4D). 123 The linker region is conserved between the two Drosophila Ph homologues (Supplementary  Table 1). Evolutionary tuning of the linker sequences is likely to affect phase 133 separation properties of PHCs, although this will need to be tested experimentally. 134 135 Chromatin is highly concentrated in Mini-Ph condensates. One potential function of phase separation is 136 to concentrate (compact) chromatin. To measure the concentration of chromatin in Mini-Ph-chromatin 137 condensates, we first prepared calibration curves using the same Cy3-labelled histone octamers 138 (labelled on H2A) that were used to assemble chromatin ( Supplementary Fig. 5A). The concentration of 139 nucleosomes in Mini-Ph condensates, starting from a mixture of 150 nM nucleosomes, and 5 µM Mini-140 Ph, was measured as 22.5 +/-4.4 µM ( Supplementary Fig. 5B). We note that this value is lower than the 141 reported concentration of chromatin in pure chromatin condensates induced by monovalent cations 142 (~340 µM 28 ). The reported measurements used free dye to prepare the calibration curve. When we 143 imaged calibration curves prepared from free Cy3, although the curves are linear, they predict at least a 144 60x higher concentration than curves prepared with labelled histone octamers using the same imaging 145 parameters. Because ladders prepared with free Cy3 do not accurately predict known concentrations of 146 Cy3-labelled histone octamers in our hands, we believe the chromatin concentrations measured using 147 the Cy3-labelled histone calibration curve ( Supplementary Fig. 5) are correct for Mini-Ph-chromatin 148 condensates. 149 150

Mini-Ph is dynamic in condensates, but chromatin intermixes slowly. A characteristic of liquid 151
condensates is that the components are dynamic. We carried out fluorescent recovery after 152 photobleaching (FRAP) experiments with Mini-Ph-chromatin condensates. A fraction of Mini-Ph is 153 mobile and exchanges in condensates, so that bleached Mini-Ph drops partially recover fluorescence 154 within several minutes ( Fig. 2A, B; Supplementary Fig. 6A-D). In contrast, when the histones (labelled 155 with H2A-Cy3) were bleached, less than 15% of the fluorescence is recovered after several minutes ( Fig.  156 2B, Supplementary Fig. 6E indicating that the chromatin in pre-formed condensates does not fully intermix when the condensates 180 fuse, at least over the 60 minutes that we monitored (Fig. 2H). This is in clear contrast to control 181 experiments in which the two chromatins are mixed prior to addition of Mini-Ph, where all structures 182 contain a uniform mix of both fluorophores (Fig. 2C, D). These experiments are consistent with the 183 coexistance of different dynamics in Mini-Ph-chromatin condensates. The persistence of unmixed 184 regions could also reflect dynamically arrested phase separation in the pre-formed condensates. We 185 note that in the mixtures shown in Fig. 2C-H, the Alexa-647 labelled chromatin (white in Fig. 2) 197 1A). The structure of Ph SAM, including its two polymerization interfaces, termed "End Helix" (EH) and 198 "Mid Loop" (ML) is well characterized 18 (Fig. 3B). Mutation of these interfaces blocks SAM 199 polymerization in vitro and impairs Ph function in vivo 5,18,20 . We therefore prepared Mini-Ph containing 200 a point mutation that disrupts the EH interface (L1565R) ("Mini-Ph EH"), or a single point mutation that 201 weakens but does not fully disrupt the ML interface (L1547R) ("Mini-Ph ML") ( Supplementary Fig. 1A). 202 Previous AUC experiments with these mutants indicate that Mini-Ph ML forms shorter polymers than 203 Mini-Ph, and Mini-Ph EH at most may form some dimers at high concentrations (see Fig. 3 Figure 8). While the concentration of nucleosomes in condensates is similar (Fig. 3H), condensates 223 formed with Mini-Ph EH are smaller (Fig. 3I, J). 224 To look more carefully at the effects of the Ph SAM mutations, we titrated Mini-Ph EH or Mini-225 Ph ML with DNA over a range of NaCl concentrations, and scored each reaction as one-phase or two 226 phases ( Supplementary Fig. 8). We find that both mutants are more sensitive to NaCl than Mini-Ph 227 ( Supplementary Fig. 3A, B, 8A-C). ATP has been shown to dissolve many protein-RNA condensates, and 228 is hypothesized to have a physiological role in regulating phase separation 56 . To test whether ATP might 229 also regulate Mini-Ph-chromatin condensates, we formed condensates with Mini-Ph, Mini-Ph ML, or 230 Mini-Ph EH, and challenged them with 2 mM ATP for 15 or 60 min. (Supplementary Fig. 9A). 231 Condensates are smaller after ATP treatment, and Mini-Ph EH is more sensitive than either Mini-Ph or 232 Mini-Ph ML (Supplementary Fig. 9  increasing the DNA amount to 16X prevents their formation (Fig. 4B). To display the data, we generated 257 a heat map of the average accessibility at each lysine under each condition ( We then compared accessibility of lysines in Mini-Ph to that in Mini-Ph EH, which does not form 280 polymers. The pattern of lysine accessibility in Mini-Ph EH is distinct from that of Mini-Ph, and 281 differences are not restricted to the SAM (Fig. 4D). Three lysines in the HD1, one in the FCS, and one in 282 the SAM are significantly altered in Mini-Ph EH versus Mini-Ph. When differences are considered over 283 each domain, they are more striking (Fig. 4E). While the overall accessibility of the HD1 is the same 284 between the two, probably because both increases and decreases in accessibility are observed, the FCS 285 is less accessible in Mini-Ph EH than in Mini-Ph. while the linker and the SAM are more accessible (Fig.  286 4D-F). The accessibility of the SAM is consistent with the expected monomeric state of Mini-Ph EH and 287 the positions of the lysines in the SAM polymer structure (Fig. 4F). However, thedifferences in the other 288 domains of Mini-Ph EH versus Mini-Ph indicate that SAM polymerization likely affects the whole 289 conformation of Mini-Ph and the interactions available for phase separation. The changes in the HD1 290 both on condensate formation and between Mini-Ph and Mini-Ph EH also raise the possibility that this 291 domain contributes interactions to phase separation, which will need to be directly tested. Whether Ph 292 SAM would also affect the conformation of Ph in the context of the full length protein, or when it is in 293 PRC1 (an interaction mediated by the HD1) remains to be determined. Finally, we attempted to analyze 294 lysine accessibility in Mini-Ph EH condensates ( Supplementary Fig. 13), but the condensates dissolved  biochemical compartments, we asked whether condensates can recruit proteins from nuclear extracts 325 (Fig. 5A). We prepared nuclear extracts from Drosophila S2R+ cells, and used an anion exchange resin to 326 deplete nucleic acids from the extracts. Even after depletion, the nuclear extracts contain substantial 327 amounts of RNA ( Supplementary Fig. 15A). Treatment of extracts with RNAseA resulted in precipitation 328 of most of the protein from the extracts, so that we used extracts containing RNA for our experiments. 329 Chromatin alone forms a few tiny structures in extracts ( Fig. 5B, C, reaction 1). Mini-Ph does not form 330 condensates in buffer (e.g. Fig. 1C), but does form small condensates in extracts, likely by binding to 331 RNA, since the condensates stain with YOYO-1 (Fig. 5B, C, reaction 2). When Mini-Ph is incubated with 332 chromatin to form condensates, and then nuclear extracts are added, the condensates are preserved, 333 although they are smaller than condensates in equivalent reactions incubated in buffer ( Fig. 5B, C, 334 compare reactions 3 and 4). Although the condensates are smaller, the concentration of chromatin in 335 them is similar to that in condensates incubated in buffer (Fig. 5D). We do not know why the 336 condensates are smaller after incubation in nuclear extracts. Post-translational modifications can 337 influence phase separation 57 , but the small molecule substrates needed for enzymes that mediate them 338 should be depleted in our desalted extracts. The presence of nucleic acids, in the extracts could disrupt 339 condensates, analogous to what is observed at high concentrations of DNA (Supplementary Fig. 2F, G). 340 Alternatively, proteins in the extracts that bind to Mini-Ph and/or chromatin may disrupt interactions 341 required for condensates. 342 We used low speed centrifugation (2 min. @ 2500*g) to isolate condensates and analyzed their 343 nucleic acid content on agarose gels. When Mini-Ph is incubated with extract in the absence of 344 chromatin, the pelleted condensates contain RNA (Fig. 5E). When Mini-Ph is incubated with chromatin, 345 and extract added subsequently, the isolated condensates contain both chromatin and RNA (Fig. 5E). 346 Since the amount of RNA that is pelleted with Mini-Ph is similar with and without chromatin, we infer 347 that Ub by about two-fold. This suggests the PRC1 recruited to condensates is functional, and that Mini-Ph-365 chromatin condensates enhance the ubiquitylation reaction ( Fig. 6B, C). 366 To determine if the Ph SAM polymerization state can influence condensate formation in the 367 more physiological environment of nuclear extracts, we prepared condensates with Mini-Ph ML, or 368 Mini-Ph EH, and added nuclear extracts to them. Mini-Ph ML condensates behave similar to those 369 formed with Mini-Ph in extracts ( Supplementary Fig. 16). In contrast, incubation of Mini-Ph EH 370 condensates in extracts transforms them into diffuse structures that occupy a larger area but have a 371 reduced chromatin concentration relative to condensates incubated in buffer ( Supplementary Fig. 16). 372 We tested histone ubiquitylation in extracts in the presence of Mini-Ph ML or Mini-Ph EH, and find that 373 neither mutant stimulates histone ubiquitylation (Fig. 6B, C). We do not know if this is because the 374 condensates formed by the polymerization mutants have different properties (e.g. Supplementary Fig.  375 14), or because they recruit less PRC1, as would be expected if SAM-SAM interactions (between Mini-Ph 376 and Ph in PRC1) are directly involved in recruiting PRC1 to chromatin. 377 The observation that Mini-Ph condensates increase histone ubiquitylation might reflect the 378 increased concentration of PRC1 in condensates (Fig. 5H, I). It is not necessarily predicted, however, that 379 the environment of condensates, in which chromatin is compacted, would enhance enzyme activity. 380 Thus to determine whether Mini-Ph-chromatin condensates enhance PRC1 activity under optimal 381 conditions, we reconstituted the ubiquitylation reaction in vitro, using chromatin alone or Mini-Ph-382 chromatin condensates as the substrate ( Supplementary Fig. 17). We used PRC1ΔPh for these 383 experiments ( Supplementary Fig. 1B was not significant, even though Ph-ML is expressed at higher levels than Ph (Fig. 7A, C). 400 Because we find that Ph SAM polymerization activity is not strictly required for phase separation 401 in vitro, we wondered if Ph-ML might be able to phase separate in vivo, particularly when present at 402 high concentrations. Formation of highly concentrated foci in cells is consistent with phase separation, 403 although it can arise through other mechanisms, as has been pointed out 35  round, bright foci. These foci are mainly (although not exclusively) nuclear, and little Venus signal is 407 observed in the nucleoplasm outside the foci (Fig. 7E). In contrast, Venus-PhΔSAM is uniformly 408 distributed in the nucleus, and does not form foci (Fig. 7F). Venus-Ph-ML forms foci but is also 409 distributed throughout the nucleus (Fig. 7G) Fig. 4). The linker of PHC3, unlike the Drosophila Ph linker 5 , does not bind the PHC3 450 SAM in trans 50 , and allows much more extensive SAM polymerization than that of Ph 50 . It is therefore 451 possible that the linker has been tuned across evolution to control polymerization and its interplay with 452 phase separation. This is consistent with modeling based analysis indicating that the properties of 453 linkers connecting interacting domains tune phase separation properties 48 . Two other PcG proteins, 454 SCM and Sfmbt, also have SAMs, and the three SAMS have been shown to co-assemble 58 ; joining of 455 SAM-mediated polymers of these three proteins could allow formation of large and diverse polymers. 456 Evaluating the phase separation activity of these other PcG SAMs, alone or in combination, and of Ph 457 homologues, will be an important future goal. 458 The phase separation activity of Ph SAM is also likely subject to negative regulation. A 459 disordered, serine/threonine rich sequence adjacent to the HD1 undergoes O-linked glycosylation 460 mediated by the PcG protein Sxc 20,59 . This region, and Sxc, are both important for Ph function in 461 regulation of some genes 20,59 . In the absence of glycosylation, Ph undergoes SAM-dependent "non-462 productive aggregation", which is not alleviated by mutating the Ph SAM polymerization interfaces 20 . It 463 is possible that "non-productive aggregation" in fact reflects SAM-dependent phase separation (or 464 maturation of phase-separated protein into stable, insoluble aggregates) 23  and chromatin (or Ph and RNA), as captured by our in vitro assays, and possibly, in the foci observed 505 when Venus-Ph is overexpressed in cells (Fig. 7). Which mechanism dominates in any situation could be 506 modulated by the local concentration of PcG proteins (i.e. how strong a PcG recruitment site is, or the 507 density of recruitment sites). This could be analogous to the distinction between enhancers and super-508 enhancers, which recruit higher levels of transcription factors and co-factors and where LLPS is believed 509 to occur 34,66 . There is also no reason at this time to exclude hybrid models 54 . For example, LLPS could be 510 a mechanism to create biochemical compartments, and within these domains, strict SAM-SAM 511 interactions could establish precise chromatin contacts required for gene repression. LLPS may also 512 represent an extreme and transient state, used to silence large chromatin domains rapidly during 513 development 12,67 , or as a step in re-establishing gene expression patterns during the cell cycle. All of 514 these possibilities remain to be tested, but the separation of phase separation and polymerization 515 activity revealed by our simple in vitro assays may provide a means to do so. 516 517 Many proteins with diverse localizations and functions have SAMs. Some SAMs have been 518 shown to polymerize in a concentration dependent manner, while others require additional recruitment 519 mechanisms to induce polymerization. The SAMS of a subset of proteins, including Ets1, Fli1, and p63 68 , 520 have not been observed to polymerize. It is therefore possible that phase separation is a property of the 521 SAM that is distinct from polymerization, a hypothesis that is testable by measuring the phase 522 separation activity of proteins with monomeric SAMs. 523 524 Ph SAM and histone ubiquitylation. We find that Ph SAM driven chromatin condensates can enhance 525 PRC1-mediated histone ubiquitylation. We do not know what the mechanism of stimulation of H2A-Ub 526 is. It is unlikely to be concentration of the reaction components in condensates because all of the 527 components (except PRC1ΔPh) are present at saturating concentrations in these reactions. PRC1ΔPh 528 binds chromatin tightly (Kd for 150 bp DNA is <=1 nM 69 ) so that Mini-Ph is also not needed to recruit 529 PRC1ΔPh to chromatin. Although further experiments will be needed to determine the mechanism, the 530 environment of condensates may stimulate steps in the reaction subsequent to substrate binding, which 531 could include the actual ubiquitin transfer or steps affecting processivity 70 . It has recently been shown 532 that H2A-Ub mediated by PRC1 is stimulated by chromatin compaction 71 , and that spreading of H2B-Ub 533 along chromatin is facilitated by formation of structured, phase-separated compartments by the 534 ubiquitylation machinery 72 , which may be relevant to our observations. Formation of protein-chromatin 535 condensates with the heterochromatin protein HP1 alters the conformation of the nucleosome, 536 rendering specific regions of the histone proteins more accessible 73 . It is possible that nucleosome 537 conformation is also changed in Mini-Ph condensates, and that these changes facilitate histone 538 ubiquitylation. Detailed characterization of chromatin in condensates will be an important future goal. 539 Stimulation of H2A-Ub is unlikely to be the essential function of the Ph SAM in Drosophila, since 540 the modification is not required for PRC1-dependent gene repression in vivo, including repression of 541 genes that depend on Ph SAM 74,75 . However, H2A-Ub is required for full development 74,75 . Drosophila 542 cPRC1 also does not seem to mediate most H2A-Ub in tissue culture cells, and it is likely that another 543 ncPRC1 containing L3(73)Ah, a homologue of mammalian Pcgf3, in place of PSC, is present in these cells 544 76 . This also means that in our experiments with nuclear extracts, although we observe PRC1 recruitment 545 to condensates, we cannot be certain that it is responsible for the ubiquitylation activity we observe 546 (Fig. 6). 547 Histone ubiquitylation by PRC1 has been most intensively studied in mouse embryonic stem 548 cells (mESCs), where systematic analysis of the effect of disrupting PRC1 subunits implicates ncPRC1 (i.e. 549 non PHC-containing) in creation of most H2A-Ub 7-9 . However, using an artificial tethering system that 550 allows PcG proteins to be reversibly targeted to a reporter gene so that persistent effects on chromatin 551 and gene expression (i.e. memory) can be measured, Moussa et al. 77 found that heritable gene 552 repression and propagation of H2A-Ub depend on cPRC1. Recent work indicates a central role for H2A-553 Ub in PcG-dependent gene regulation in mESCs 78,79 , in seeming contrast with observations in 554 Drosophila; it will be interesting to determine how Ph SAM contributes to H2A-Ub activity in mammals. 555 The ability of Ph SAM to condense chromatin and to promote H2A-Ub could be important for rapidly 556 building PcG chromatin domains, or restoring them at the end of mitosis. H2A-Ub is not detected on 557 mitotic chromosomes in mammalian cells 80,81 , suggesting it is re-acquired after cells exit mitosis. 558 Finally, Cbx2, a member of some mammalian canonical (PHC-containing) PRC1s, which has a 559 strong chromatin compacting activity 82 , has also been shown to form phase separated condensates with 560 chromatin in vitro, and to form 1,6-hexanediol-sensitive foci in ES cells 40,42 . This phase separation 561 activity is mediated by a charged IDR in Cbx2 that is important for the developmental function of Cbx2 562 83 . Further, as shown in Supplementary Fig. 17 Cultures were grown at 37°C to an OD of 0.8-1.0, and then shifted to 15°C for overnight induction with 579 1mM IPTG. Cells were pelleted, flash frozen, and stored at -80°C. Cells were resuspended in 2 ml/g lysis 580 buffer (50 mM Tris, pH 8.5, 200 mM NaCl, 10 mM β-ME, 100 µM ZnCl2, 0.2 mM PMSF, 0.5 mM 581 Benzamidine). Cells were incubated on ice for 10 min, flash frozen in liquid nitrogen, thawed at 37°C, 582 and sonicated 6*30 sec. at 30% intensity. Freeze-thaw and sonication were repeated, and the lysate 583 centrifuged for 1 hour at 100,000*g and 4°C. Cleared lysate was sonicated 6*30" at 40% intensity, and 584 filtered through a 22 µm filter. Lysate (from 1 L) was applied to a 1 ml His-Trap column using an AKTA 585 FPLC, and eluted with a gradient of imidazole (from 10-300 mM) in lysis buffer. Fractions with Mini-Ph 586 were dialyzed overnight against 1 L of 20 mM Tris, pH 8.5, 50 mM NaCl, 100 µM ZnCl2, and 10 mM β-ME. 587 Dialyzed fractions were centrifuged for 10 min. at 20,800*g, and loaded on a 1 ml HiTrapQ-HP column 588 and eluted with a gradient from 50 mM to 1 M NaCl in binding buffer. Fractions were pooled and 589 dialyzed overnight into 20 mM Tris, pH8, 50 mM NaCl, 10 µM ZnCl2, 1 mM βME, aliquotted and stored at 590 -80. In some cases, Mini-Ph was further purified by size exclusion chromatography on a Superose 12 size 591 exclusion column. 592 Mini-Ph∆SAM and Mini-Ph∆FCS: Both proteins were expressed in BL21 (DE3) Gold cells pre-593 transformed with the pRARE plasmid. The transformed cells were grown at 37°C in LB media to an OD600 594 of ~0.7 -0.8 and induced overnight at 15°C. Cells harvested from 1 L of culture were resuspended with 595 10 ml of lysis buffer (50 mM Tris pH 8.0, 200 mM NaCl, 5 mM βME, 30 mM imidazole pH 7.5, 1 mM 596 PMSF) and lysed by sonication. The soluble lysates were introduced onto am Ni-NTA column, washed 597 with lysis buffer (without PMSF), and bound proteins eluted using 300 mM imidazole, 200 mM NaCl, 5 598 mM βME. The leader sequence was cleaved using TEV protease, and the cleaved sequence and 599 uncleaved proteins removed by passing through a Ni-NTA column. Further purification was performed 600 using a HiTrap Q-HP column. Fractions containing protein were pooled, buffer exchanged into 50 mM 601 Tris pH 8.0, 100 mM NaCl, 5 mM βME, and concentrated. Mini-PhΔSAM was further purified on a 602 Superdex 200 size exclusion column in 50 mM Tris pH 8.0, 100 mM NaCl, 5 mM βME. with 10% FBS. 20*15 cm dishes were used to prepare nuclear extracts as described 89 , except that nuclei 650 were purified through a sucrose cushion prior to extraction. Cells lysed in hypotonic buffer were layered 651 over two volumes of 30% sucrose in hypotonic buffer, and centrifuged 18' @ 1400*g. Nuclei were 652 washed once in hypotonic buffer, and extracted as described. The high salt extraction buffer was 1.2 M 653 KCl, and extracts were not dialyzed. To use the extracts to treat condensates, up to 100 µl of extract was 654 buffer exchanged into 20 mM Tris, pH 8, 50 mM NaCl using a Zeba column. Extracts were centrifuged 2' 655 @ 20,000*g and incubated for 15' on ice with 60% volume of Q-sepharose. Extracts were spun through 656 an empty column (2' @ 10,000*g), and then centrifuged 2' @20,000 *g. All procedures were carried out 657 on ice or at 4°C and contained protease inhibitors and 0.4X PhosStop phosphatase inhibitor. 658 659 Chromatin preparation: Most experiments were carried out with the plasmid p5S*8, which contains 5 660 blocks of 8-5S nucleosome positioning sequences (repeat length 208 base pairs). Plasmids were 661 assembled by salt gradient dialysis as described 90 . Chromatin was finally dialyzed into HEN (10 mM 662 Hepes, pH 7.9, 0.25 mM EDTA, 10 mM NaCl) buffer and stored at 4˚C. To measure chromatin assembly, 663 100 ng of each assembly was digested overnight with 10 U of EcoRI in NEB buffer 2.1, and loaded on a 664 0.5X TBE, 5% acrylamide native gel. Gels were stained with Ethidium bromide and imaged on a buffer, which contains 2M NaCl, and in which histone octamers remain assembled), and Mini-Ph 700 chromatin condensates. Histones are the same histones used to prepare the chromatin; 43% of the 701 histone octamers are labelled (measured both using the NanoDrop and by loading histones and free dye 702 on SDS-PAGE gels), corresponding to a 21.5% labeling efficiency on H2A (since there are two copies of 703 H2A in each octamer). Image J "measure" was used to measure the mean grey intensity for each of 9 704 images for each point. Images were manually checked and images with bright artifacts removed, 705 although these had little impact on the measured intensities. A linear regression was fit to the 706 calibration curve and used to convert measured intensities to nucleosome concentrations. To measure 707 intensities in condensates, Image J was used to threshold the images (AutoThreshold-->Li); Analyze 708 Particles was used to measure the mean grey intensity in each thresholded structure. Particle size was 709 set as 100-infinity pixels. The mean grey intensity from the buffer image was subtracted from all 710 measurements, which were converted to nucleosome concentrations using the calibration curve. 711 H2A, although we were only able to record FRAP images from one channel. Two pre-bleach images were 714 collected, followed by an image every 5 or 10 sec. All FRAP analysis of Mini-Ph was done by bleaching 715 single complete structures. Images were analyzed in Image J (Fiji). An ROI was selected for the bleach 716 area, background, and a non-bleached structure. Background subtracted, normalized data were fit with 717 a double exponential fit using GraphPad Prism 8. 718 Y=Y0+ SpanFast*(1-exp(-KFast*X)) + SpanSlow*(1-exp(-KSlow*X)). We excluded data sets that could not 719 be fit, and obvious technical artifacts (e.g. if a drop fuses with the bleached condensate during the 720 experiment). 721 722 Image analysis of condensates: Images for display were prepared using Zen2 (blue edition). For 723 quantification, images were exported as TIFs from Zen (original data). ImageJ (Fiji) was used to threshold 724 the images (Li algorithm); thresholds were manually checked and images with too few structures to 725 threshold were removed. Areas of thresholded structures were measured using ImageJ ("Analyze 726 Particles", size=10-infinity pixel), and intensities using Analyze Particles. For colocalization analysis, the 727 GDSC-->Colocalization-->Particle Overlap was used. Masks were created in the Alexa 647 (Mini-Ph) and 728 Cy3 ( vacuum was turned off, and reactions loaded on the filters. Slots were immediately washed with 2*100 742 µl binding buffer. Filters were air dried, exposed to a phosphoimager screen, and scanned on a Typhoon. 743 ImageQuant was used to quantify top (bound) and bottom (unbound) filters, and fraction bound 744 calculated in Excel. Curve fitting was done in GraphPad Prism 8, using the following equation: 745 Y=ABmax*X/(X+Kd)+b 746 747 Protein footprinting assay: The acetylation footprinting assay is described in detail in Kang et al. 6 . Phase 748 separation reactions were directly scaled up to use 4 µg of protein for each sample. Condensates were 749 allowed to form at room temp. for 15 min.; an aliquot of each sample was removed to confirm phase 750 separation using microscopy. Sulfo-NHS-acetate was dissolved immediately before use, and added to a 751 final concentration of 0.5 mM. An aliquot of each sample was removed to monitor phase separation by 752 microscopy, and reactions were stopped after 15 min. by addition of Trifluoroacetic acid to a final 753 concentration of 1%. For Mini-Ph EH, acetylation of condensates was restricted to 5 min. because these 754 condensates dissolved rapidly on exposure to Sulfo-NHS-acetate. We therefore analyzed Mini-Ph EH 755 alone, and bound to DNA (16X DNA, Fig. 4)  Mass Spectrometry data were analyzed using Maxquant (v 1.6.10.43) with Acetyl(K) and 762 Propionylation(K) as variable modifications. 10 missed cleavages were allowed since lysine modification 763 will block trypsin digest. All data files were analyzed together, with the "match between runs" option. 764 Intensities for identified Acetyl and Propionyl sites were used for quantification. Accessibility was 765 calculated (in Excel) as (intensity acetylated)/(intensity acetylated+intensity prop +0.5) for each site. To 766 compare accessibility between samples, GraphPad Prism 8 was used to conduct student's t-test, 767 assuming equal variance across samples, and with the Holm-Sidak method of correction for multiple 768 comparisons, with alpha=0.05 (unpaired, 2-tailed test). Heat maps were prepared from averaged 769 accessibilities using Morpheus (https://software.broadinstitute.org/morpheus). 770 771 Analysis of condensates after incubation in nuclear extracts: Phase separation reactions were set up in 772 40 µl with 80 nM nucleosomes, 7.5 µM Mini-Ph, in 20 mM Tris pH 8.0 and 50 mM NaCl. After incubating 773 10 min. at room temp., 12 µl of nuclear extracts were added, and reactions mixed by gently pipetting up 774 and down. 7.5 µl were removed and diluted to 10 µl for imaging, and 7.5 µl mixed with the 775 uibiquitylation machinery to assay histone ubiquitylation. After 60 min. of total incubation, samples 776 were pelleted by centrifugation for 2 min. at 2500*g, 4°C. Supernatants were removed and SDS-sample 777 buffer added to 1X. Pellets were resuspended in 2X SDS sample buffer. 2 µl of each pellet and 778 supernatant were removed and digested with Proteinase K for at least 1 hour at 55°C before analysis on 779 1.2% agarose, 1X TAE gels, which were stained with SYBR Gold to visualize nucleic acids. The remainder 780 of the samples were boiled and loaded on 8% SDS-PAGE gels, transferred to nitrocellulose, and used for 781 Western blotting. Membranes were blocked with 5% nonfat dry milk in PBST (PBS + 0.3% Tween-20), 782 and incubated with primary antibodies diluted in 5% milk-PBST overnight at 4°C. Membranes were 783 washed 3*10 min. in PBST, incubated in secondary antibody diluted in 5% milk-PBST for 1-2 hours, 784 washed 3*10 min. in PBST, and visualized using a Li-Cor Odyssey imaging system. Image J (Fiji) was used 785 to quantify band intensities. 786 787 Histone Ubiquitylation assays: For ubiquitylation assays, 125 ng chromatin per 5 µl was pre-incubated 788 with 5 µM Mini-Ph (or buffer) for 15 min. at room temp. to induce phase separation, followed by 789 addition of the ubiquitylation machinery and PRC1ΔPh. Final reaction conditions are 40 nM 790 nucleosomes, 20 mM Hepes, pH 7.9, 0.25 mM MgCl 2, 0.25 mM ATP, 0.6 mM DTT, 60 mM KCl, 25mM 791 NaCl, 700 nM E1, 800 nM E2, and 500 ng Ub. Titrations of the E1, E2, and His-Ub indicate that none are 792 limiting under these conditions. Reactions were further incubated for 45 min. at room temp. Aliquots 793 were removed for imaging, and the remainder of the reaction stopped by addition of SDS-Sample buffer. 794 Boiled samples were loaded on 16% SDS-PAGE gels, which were scanned for Cy3 to detect H2A, and 795 then stained with SYPRO Ruby. Histone ubiquitylation assays in nuclear extracts were carried out under 796 the same conditions except that the pre-incubation of chromatin with Mini-Ph was 10 min., nuclear 797 extracts were added just before the ubiquitylation components, and reactions were incubated for 80 798 min. at room temp. 799 800 Cell culture: Wild type S2 cells (from Expression Systems) and S2 cell lines harbouring stable Ph or Ph-ML 801 14 transgenes were grown in suspension in ESF-921 media with 5% FBS. Protein expression was induced 802 with 0.5 µM CuSO4 for 4 days. For whole cell extracts, cells were resuspended in 2X-SDS sample buffer 803 and boiled. For histone extraction, we followed the protocol of Abcam 804 (https://www.abcam.com/protocols/histone-extraction-protocol-for-western-blot); HDAC inhibitors 805 were not included. Western blots were carried out as described above, and ImageQuant was used to 806 quantify bands. 807 Live cell imaging: For live cell imaging, S2 (Fig. 7), or S2R+ ( Supplementary Fig. 17) cells were plated at 808 10 6 cells per well in 6-well plates the night before transfection. Transfection was carried out using Trans-809 IT lipid (Mirus), according to the manufacturer's protocol. 2 µg of each Venus-Ph construct was used 810 along with 0.5 µg of pAct5C-H2A-RFP 93 . One to two days after transfection, cells were replated on ConA-811 coated imaging dishes (Ibidi). Heat shock was for 8 min. (S2R+) or 12 min. (S2) at 37°C, and cells were 812 analyzed within 24 hours of protein induction. Confocal stacks of thick slices (3 µm) were collected on 813 the spinning disc microscope described above using the 63X objective to capture foci throughout the 814 cell. 815

Image analysis of live cells:
The .czi files of image stacks were opened in Image J (Fiji), converted to 816 maximum intensity projections, and the channels split. The red channel (H2A-RFP) was used to segment 817 nuclei as follows. Images were thresholded with the Li algorithm, followed by removing outliers less 818 than 5 pixels, and 3 rounds of erosion. Thresholded images were converted to masks, processed with a 819 watershed algorithm, and "Analyze Particles" used with a size threshold of 200-inifinity pixels to select 820 nuclei. The green channel (YFP fusion proteins) was then processed with "Find maxima" with the 821 following parameters: Prominence: 20000; strict; exclude edge maxima; output type: single points. The 822 nuclei selected from the red channel were used as ROIs, and the # maxima per ROI (i.e. # foci/nucleus) 823 obtained using Measure in the ROI tool, followed by dividing the raw integrated density by 255. This 824 entire pipeline is explained here: https://microscopy.duke.edu/guides/count-nuclear-foci-ImageJ. To 825 compare the # foci per cell, cells with zero foci were excluded; since Venus-PhΔSAM does not form foci, 826 the majority of cells were excluded. 827 828 829 830 Data Availability: Mass spectrometry raw files will be uploaded to MassIVE. The Source Data file 831 includes data for FRAP traces (Fig. 2, Supplementary Fig. 14) and MaxQuant output (intensities) for 832 acetylation footprinting experiments (Fig. 4), filter binding data (Fig. 3C), nucleosome and condensate 833 measurements (Fig. 3H, I, J), western blot quantification (Fig. 5I, 7C, D), ubiquitylation activity (Fig. 6E), 834 foci measurements (Fig. 7H). All other raw data are available on reasonable request.

GV GS GE T NGL GT GGI V GV DAMAL V DR L DE AMAE E K MQT E AT P K L S E S F P I L GAS T E V P P M GV GS GE T NGL GT GGI V GV DAMAL V DR L DE AMAE E K MQT E S Y QT V S DAL P I QAAT P E V P P I S L P V QAAI S AP S P L AMP L GS P L S V AL P T L AP L S V V T S GA-----AP K S S E V NGT DR S MP V L AAMS T S S P L S L P L T L P L P I AI AP T V S L P V V S AGV V AP V L AI P S S NI NGS DR
Ph-p Ph-d Ph-p Ph-d          Fig. 13F). The ML mutation, which weakens but does not eliminate SAM-SAM interactions may behave similar to the schematic in A (i.e. the SAM-SAM interaction is dynamic under phase separation conditions, unlike wild-type Mini-Ph which forms short, limited polymers before phase separation). It is important to point out that this is the simplest scheme; it is possible that there are other interactions in the system that have not been characterized. These could include interactions involving the HD1, hinted at by the difference in accessibility measured for Mini-Ph and Mini-Ph EH (Fig. 4D, E). We also do not know how the structure of Mini-Ph polymers influences binding to large chromatin or DNA templates, and whether Mini-Ph binding influences nucleosomenucleosome interactions (as suggested in the diagrams).

Mini-Ph
Mini-Ph ML