bioRxiv preprint doi: https://doi.org/10.1101/2021.02.08.430127; this version posted February 8, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

1 Nucleosomal Asymmetry Shapes Histone Mark Binding 2 and Promotes Poising at Bivalent Domains 3 4 5

6

1 1¶ 1¶ 1¶ 7 Elana Bryan , Marie Warburton , Kimberly M. Webb , Katy A. McLaughlin , 1 3 1 3 8 Christos Spanos , Christina Ambrosi , Viktoria Major , Tuncay Baubec , Juri 1,2 1* 9 Rappsilber , Philipp Voigt

10

11

12

1 13 Wellcome Centre for Cell Biology, School of Biological Sciences, University of

14 Edinburgh, Edinburgh EH9 3BF, United Kingdom.

2 15 Chair of Bioanalytics, Institute of Biotechnology, Technische Universität Berlin,

16 13355 Berlin, Germany.

3 17 Department of Molecular Mechanism of Disease, University of Zurich, CH-

18 8057 Zurich, Switzerland

19

20

¶ 21 These authors contributed equally.

22

23

24

25 * Correspondence should be addressed to P.V.: [email protected]

26 bioRxiv preprint doi: https://doi.org/10.1101/2021.02.08.430127; this version posted February 8, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

27 Summary

28

29 Promoters of developmental genes in embryonic stem cells (ESCs) are marked by

30 histone H3 lysine 4 trimethylation (H3K4me3) and H3K27me3 in an asymmetric

31 nucleosomal conformation, with each sister histone H3 carrying only one mark. These

32 bivalent domains are thought to poise genes for timely activation upon differentiation.

33 Here we show that asymmetric bivalent nucleosomes recruit repressive H3K27me3

34 binders but fail to enrich activating H3K4me3 binders, despite presence of H3K4me3,

35 thereby promoting a poised state. Strikingly, the bivalent mark combination further

36 attracts chromatin proteins that are not recruited by each mark individually, including

37 the histone acetyltransferase complex KAT6B (MORF). Knockout of KAT6B blocks

38 neuronal differentiation, demonstrating that bivalency-specific readers are critical for

39 proper ESC differentiation. These findings reveal how histone mark bivalency directly

40 promotes establishment of a poised state at developmental genes, while highlighting

41 how nucleosomal asymmetry is critical for histone mark readout and function.

42

43

44

45

46 Keywords: Chromatin, transcription, histone methylation, histone acetylation, bivalent

47 domains, embryonic stem cells, differentiation, Polycomb.

48

2 bioRxiv preprint doi: https://doi.org/10.1101/2021.02.08.430127; this version posted February 8, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

49 Introduction

50

51 Histone modifications have emerged as key regulators of transcription and other

52 chromatin-templated processes. Often in combinatorial fashion, histone marks set up

53 chromatin environments that reflect, reinforce, and potentially instruct transcriptional

54 states by directly regulating access to the DNA or by recruiting factors that activate or

55 repress transcription. Promoters of developmental genes in embryonic stem cells

56 (ESCs) are marked by a distinctive histone modification signature comprised of the

57 active histone mark histone H3 lysine 4 trimethylation (H3K4me3) and the repressive

58 mark H3K27me3 (Azuara et al., 2006; Bernstein et al., 2006; Harikumar and Meshorer,

59 2015; Mikkelsen et al., 2007; Voigt et al., 2013). These so-called ‘bivalent domains’

60 are presumed to poise expression of developmental genes primarily in ESCs by

61 maintaining a repressed but plastic state that allows for prompt activation or stable

62 repression of these genes upon differentiation cues. Bivalent domains are established

63 by the histone methyltransferase complexes MLL2 and Polycomb repressive complex

64 (PRC) 2, catalyzing H3K4me3 and H3K27me3, respectively (Christophersen and

65 Helin, 2010; Piunti and Shilatifard, 2016; Schuettengruber et al., 2017; Yu et al., 2019).

66 Supporting a pivotal role of bivalent domains in regulating developmental genes,

67 knockouts or loss-of-function mutations in these complexes abolish bivalency and lead

68 to developmental defects in mice and compromised differentiation potential in ESCs

69 (Laugesen and Helin, 2014; Piunti and Shilatifard, 2016; Vastenhouw and Schier,

70 2012). However, the mechanisms by which bivalent domains poise genes for

71 expression are largely unclear and it remains elusive whether the bivalent histone

72 marks themselves are key drivers of poising.

73 Given the central role of binding proteins or ‘readers’ in executing histone mark

74 function, proteins that bind H3K4me3 or H3K27me3 are candidates for translating

75 bivalency into a poised state via a histone mark-based mechanism. Complexes

76 featuring binding proteins for the active H3K4me3 mark comprise the general

77 transcription factor TFIID, transcriptional co-activators, histone modifiers, and

78 chromatin remodelers, whereas the repressive H3K27me3 mark is primarily

79 recognized by PRC2 and canonical PRC1 complexes (Bartke et al., 2010; Eberl et al.,

80 2013; Musselman et al., 2012; Taverna et al., 2007; Vermeulen et al., 2010). However,

81 thus far recognition of H3K4me3 and H3K27me3 has only been studied individually,

3 bioRxiv preprint doi: https://doi.org/10.1101/2021.02.08.430127; this version posted February 8, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

82 rendering potential antagonistic or multivalent synergistic effects of their bivalent

83 coexistence elusive. Moreover, we and others have shown that bivalent nucleosomes

84 are asymmetrically modified, carrying H3K4me3 and H3K27me3 on two separate

85 histone H3 copies, rather than featuring two histone H3 copies modified with both

86 H3K4me3 and H3K27me3 (Shema et al., 2016; Voigt et al., 2012). However, it remains

87 unclear how nucleosomal asymmetry affects recruitment of histone mark binding

88 proteins—and thus histone mark function—at bivalent nucleosomes and beyond.

89 Here, we set out to determine how asymmetric bivalent nucleosomes are decoded

90 in order to clarify whether the bivalent marks directly act to poise developmental genes

91 in ESCs. Using an approach based on nucleosome pulldown assays with

92 recombinant, site-specifically modified nucleosomes followed by quantitative mass

93 spectrometry (MS) analysis, we reveal that bivalent nucleosomes recruit repressive

94 complexes as well as bivalency-specific binders, but fail to recruit H3K4me3 binders.

95 These findings provide evidence for a histone mark-based poising mechanism and

96 illustrate how nucleosomal asymmetry regulates histone mark function.

97

4 bioRxiv preprint doi: https://doi.org/10.1101/2021.02.08.430127; this version posted February 8, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

98 Results

99

100 Monovalent asymmetric nucleosomes fail to recruit H3K4me3 and H3K27me3

101 binding proteins

102 In order to resolve how the asymmetric bivalent mark combination might directly

103 establish or facilitate poising at bivalent domains, we generated site-specifically

104 modified histones via native chemical ligation and assembled them into recombinant

105 nucleosomes carrying symmetric and asymmetric H3K4me3 and H3K27me3

106 (Figure S1). We then carried out nucleosome pulldown assays with mouse E14 ESC

107 nuclear extract (NE) followed by label-free quantification (LFQ) liquid chromatography

108 (LC)-coupled mass spectrometry (MS) analysis (Figure 1A). We first sought to

109 determine how asymmetric monovalent presence of H3K4me3—as observed at

110 bivalent domains in vivo—regulates its readout and affects recruitment of H3K4me3

111 binding proteins. As expected from previous studies (Bartke et al., 2010; Eberl et al.,

112 2013), when performing pulldowns with symmetric H3K4me3 (denoted as

113 H3K4me3/3) nucleosomes, we observed specific enrichment of a host of known

114 H3K4me3 binding proteins, including members of the TFIID, SETD1, and SIN3A/B

115 complexes, chromatin remodelers such as NURF, CHD1, and CHD8, and other PHD

116 finger proteins such as PHF2 (Figures 1B and 1C, Table S1). We further observed

117 depletion of HMG20B, PHF21A, G9a/GLP, and PRC2 on H3K4me3/3 nucleosomes

118 (Figures 1B and 1C, Table S1). Strikingly, when we performed analogous pulldown

119 experiments with asymmetric (‘H3K4me0/3’) nucleosomes, enrichment of H3K4me3

120 binders and depletion of repressive factors was lost (Figures 1B and 1C, Table S1),

121 indicating that asymmetric nucleosomes fail to recruit H3K4me3 binding proteins,

122 despite presence of the mark on one of the two histone H3 copies in the nucleosome.

123 In line with these findings, asymmetric H3K27me3 nucleosomes likewise failed to

124 enrich known H3K27me3 binding proteins, while symmetric H3K27me3 nucleosomes

125 robustly recruited known H3K27me3 binders (Figures 1B and 1D, Table S2). Taken

126 together, symmetric modification appears to be required to achieve enrichment of

127 binding proteins for both H3K4me3 and H3K27me3 in nucleosome pulldown assays,

128 while recruitment was diminished for asymmetric nucleosomes. These findings

129 suggest that loci featuring asymmetric nucleosomes—such as bivalent domains—are

5 bioRxiv preprint doi: https://doi.org/10.1101/2021.02.08.430127; this version posted February 8, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

130 functionally distinct from loci containing nucleosomes symmetrically modified with

131 H3K4me3 or H3K27me3.

132

133 Symmetric modification enhances the apparent affinity of H3K4me3 and

134 H3K27me3 binders towards nucleosomes

135 We next sought to clarify why recruitment of mark binding proteins was markedly

136 less efficient for asymmetric compared to symmetric nucleosomes, despite carrying

137 the same modifications. To this end, we performed a series of nucleosome pulldown

138 assays with symmetric and asymmetric nucleosomes and analyzed recruitment of

139 select H3K4me3 binders in a semi-quantitative fashion using titration experiments

140 (Figures 2A and S2A). Robust, concentration-dependent binding of TAF3, PHF2, and

141 ING2 was detected for symmetric H3K4me3 nucleosomes, whereas binding to

142 asymmetric H3K4me3 nucleosomes was substantially weaker (Figure 2A), in line with

143 our LFQ-MS analysis (see Figure 1B and 1C). We reasoned that the marked decrease

144 in apparent affinity towards asymmetric nucleosomes could be due to factors binding

145 to the unmodified tail, precluding recruitment of H3K4me3 binders due to steric

146 exclusion. However, when performing pulldowns with asymmetric nucleosomes

147 containing a tail-less rather than full-length histone H3 as the unmodified counterpart,

148 enrichment of H3K4me3 binders was not recovered (Figure 2A). We next

149 hypothesized that the two-fold higher abundance of binding sites on symmetric

150 nucleosomes enhances interaction with binding proteins based on mass action.

151 However, while partially enhancing ING2 binding, matching the abundance of

152 H3K4me3 sites by doubling the amount of asymmetric nucleosomes did not recover

153 binding of TAF3 and PHF2 to levels seen for symmetric nucleosomes (Figure 2A).

154 These findings indicate that the symmetric arrangement itself—rather than mere

155 increased overall mark abundance—favors recruitment of H3K4me3 binders to

156 symmetric nucleosomes.

157 On symmetric nucleosomes, two equivalent binding sites are available for reader

158 proteins that contain a single cognate binding domain. This situation is akin to clusters

159 of near-equivalent phosphorylation sites that recruit signaling factors containing a

160 single phosphorylation binding domain. Conceptually distinct from multivalency, the

161 term ‘allovalency’ has been coined to describe the behavior of such systems that

6 bioRxiv preprint doi: https://doi.org/10.1101/2021.02.08.430127; this version posted February 8, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

162 consist of a single binding domain interacting with a polyvalent ligand containing

163 multiple identical binding sites in tandem (Ivarsson and Jemth, 2019; Klein et al., 2003;

164 Locasale, 2008; Mittag et al., 2008; Olsen et al., 2017). Upon dissociation from one

165 binding site, the binder can rapidly rebind another epitope nearby before leaving the

166 ‘capture sphere’ of the polyvalent ligand, resulting in rapid intracomplex exchange

167 rather than complete dissociation, thus increasing overall apparent affinity (illustrated

168 in Figure 2B for TAF3). As observed in our nucleosome pulldown assays, these effects

169 increase the apparent affinity of binders towards symmetric H3K4me3 modification,

170 resulting in markedly enhanced binding (Figure 2A, illustrated in Figure 2C).

171 Increased apparent affinity through allovalent effects on symmetric nucleosomes

172 and the associated shift in apparent KD would be expected to lead to larger gains in

173 binding for lowly compared to highly abundant binders that are already approaching

174 saturated nucleosome binding (illustrated in Figure 2C). Interestingly, most H3K4me3

175 binders are expressed at low levels in ESCs, whereas PRC1 and PRC2 are

176 considerably more abundant (Zhang et al., 2017). Indeed, while the lowly abundant

177 H3K4me3 binders exhibited markedly increased binding to symmetric nucleosomes,

178 this increase was much less pronounced for H3K27me3 binders (Figures 2D and

179 S2B). These findings indicate that the high abundance of PRC1 and PRC2 complexes

180 in ESCs overcomes the requirement of symmetric modification for efficient

181 recruitment, suggesting that nucleosomal asymmetry at bivalent nucleosomes in

182 ESCs would impact recruitment of H3K4me3 binders more strongly than H3K27me3

183 binders.

184

185 Bivalent nucleosomes recruit repressive H3K27me3 binders and bivalency-

186 specific binders, but not H3K4me3 binders

187 Having established that nucleosomal asymmetry affects the readout of monovalent

188 H3K4me3 and, to a lesser extent, H3K27me3, we next sought to determine whether

189 these effects are recapitulated for asymmetric bivalent nucleosomes by performing

190 pulldown experiments with bivalent nucleosomes. In line with the different sensitivity

191 of H3K4me3 and H3K27me3 reader binding towards asymmetric modification

192 (Figures 1 and 2), asymmetric bivalent nucleosomes failed to recruit H3K4me3 binding

193 proteins such as TAF3, PHF2, and ING2, but significantly enriched H3K27me3 binders

7 bioRxiv preprint doi: https://doi.org/10.1101/2021.02.08.430127; this version posted February 8, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

194 (Figures 3A–3C and S3A, Table S3). In the case of PRC1, enrichment on bivalent

195 nucleosomes was lower compared to symmetric H3K27me3 nucleosomes (Figures

196 3B and S3A), as expected from the reduced affinity of H3K27me3 binders towards

197 asymmetric H3K27me3 nucleosomes (Figures 1D and 2D). Interestingly, along with

198 the core subunit EZH2, the auxiliary PRC2 subunits EPOP and EZHIP were

199 specifically enriched on bivalent nucleosomes (Figures 3B and S3A), suggesting

200 differential interaction of specific PRC2 subcomplexes with bivalent nucleosomes.

201 Strikingly, we further identified several proteins that were specifically enriched on

202 asymmetric bivalent but not monovalent symmetric H3K4me3 or H3K27me3

203 nucleosomes (Figures 3B, 3D, S3A, Table S3), suggesting that asymmetric bivalent

204 nucleosomes recruit specific factors through multivalent interactions involving both

205 marks. These factors included two chromatin modifying complexes, the histone

206 acetyltransferase KAT6B (also known as MORF) and the histone chaperone SRCAP.

207 In addition to these complexes, we also found RNA binding proteins including

208 ZRANB2, SRSF2, and PUM1, metabolic enzymes such as acetyl-CoA carboxylase,

209 propionyl-CoA carboxylase, and 3-methylcrotonyl-CoA carboxylase, and transcription

210 factors including MITF, TFE3, and TFEB to be specifically enriched on bivalent but not

211 monovalent nucleosomes (Figures 3A–3C and S3A, Table S3). We also observed

212 depletion of factors from bivalent nucleosomes, most prominently members of the

213 INO80 complex, several zinc finger proteins, and general transcription factor subunits

214 (Figure 3D and S3A, Table S3). However, the majority of these depleted proteins

215 displayed favorable interactions with the asymmetric unmodified control nucleosomes

216 (Figure S3B), potentially confounding accurate assessment of the role of the bivalent

217 marks in recruitment of these factors.

218 Taken together, these findings suggest that the asymmetric bivalent histone mark

219 combination directly promotes poising by recruiting repressive and bivalency-specific

220 factors but excluding active factors, thereby maintaining a repressed but plastic state

221 (illustrated in Figure 3E). To examine whether the differential recruitment of H3K4me3

222 and H3K27me3 binders was recapitulated in vivo, we analyzed the genome-wide

223 distribution of TAF3 and CBX7 (Liu et al., 2011; Morey et al., 2013). CBX7 was found

224 enriched at bivalent promoters within the Hoxd cluster and on a genome-wide scale in

225 ESCs, whereas TAF3 was largely absent from bivalent promoters, despite presence

226 of H3K4me3 (Figures 4A–C). Conversely, TAF3 was enriched at promoters

8 bioRxiv preprint doi: https://doi.org/10.1101/2021.02.08.430127; this version posted February 8, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

227 monovalently marked with H3K4me3 (Figures 4A–C). These data are in agreement

228 with the in vitro nucleosome pulldown experiments and support a histone mark-based

229 poising mechanism via recruitment of specific repressive complexes and exclusion of

230 active binders at bivalent domains.

231 To understand how the bivalency-specific factors identified in our nucleosome

232 pulldown assays (Figure 3D) contribute to the regulation of developmental genes at

233 bivalent promoters in vivo, we focused on the chromatin modifying complexes SRCAP,

234 a histone chaperone complex responsible for incorporating H2A.Z–H2B dimers into

235 nucleosomes (Clapier et al., 2017; Giaimo et al., 2019; Mizuguchi et al., 2004), and

236 KAT6B, a multi-subunit histone acetyltransferase that, along with its paralog KAT6A/

237 MOZ, has been implicated in acetylation of H3K9, H3K14, and H3K23 (Huang et al.,

238 2016; Klein et al., 2019; Yang, 2015). Co-enrichment of several core subunits

239 suggested recruitment of intact, functional complexes for both SRCAP (subunits

240 DMAP1, YEATS4/GAS41, and VPS72) and KAT6B (subunits BRPF1/3, MEAF6, and

241 ING5; Figures 3B, 3D, S3A). Whereas SRCAP has been previously linked to bivalency

242 through presence of H2A.Z at bivalent domains in ESCs (Creyghton et al., 2008; Ku

243 et al., 2012) and the requirement of H2A.Z for ESC differentiation (Creyghton et al.,

244 2008; Hu et al., 2013b), the role of KAT6B and its acetylation marks at bivalent

245 domains is unclear. We therefore asked whether KAT6B is recruited to bivalent

246 domains in ESCs and whether KAT6B contributes to regulation of developmental

247 genes during differentiation.

248

249 Bivalency promotes recruitment of KAT6B to developmental genes in ESCs

250 To determine whether KAT6B is recruited to bivalent promoters in ESCs, we

251 generated ESC lines expressing N-terminally FLAG- and HA-tagged KAT6B from the

252 endogenous Kat6b gene via CRISPR/Cas9-based genome editing (Figures S4A and

253 S4C) and performed chromatin immunoprecipitation (ChIP) using the HA epitope. In

254 line with our nucleosome pulldown data (Figure 3D), KAT6B was bound to bivalent

255 promoters, but not inactive promoters or gene bodies of bivalent genes (Figure 5A). In

256 addition, KAT6B was also found at promoters of active genes (Figure 5A). To test

257 whether recruitment of KAT6B to bivalent promoters in ESCs is dependent on the

258 bivalent state, we introduced a point mutation that renders MLL2, the

9 bioRxiv preprint doi: https://doi.org/10.1101/2021.02.08.430127; this version posted February 8, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

259 methyltransferase responsible for placing H3K4me3 at bivalent domains (Denissov et

260 al., 2014; Hu et al., 2013a), catalytically inactive (Figures 5A, S4B, S4C). Concomitant

261 with the reduction in H3K4me3, levels of KAT6B were markedly reduced at bivalent

262 promoters in MLL2 Y2688A-expressing ESCs (Figure 5A). These findings indicate that

263 recruitment of KAT6B to bivalent promoters is promoted by the bivalent modification

264 state, in line with our nucleosome pulldown data.

265 We next sought to clarify the mechanisms underlying recruitment of KAT6B to

266 bivalent promoters. The KAT6B complex comprises a host of well-characterized

267 histone mark binding domains (Figure 5B). The PHD finger of ING5 has been shown

268 to bind H3K4me3 (Champagne et al., 2008), whereas the double PHD fingers of both

269 KAT6B (Ali et al., 2012; Klein et al., 2017; Qiu et al., 2012) and BRPF1/2 (Lalonde et

270 al., 2013; Qin et al., 2011) bind histone H3 tails with unmodified but not trimethylated

271 H3K4. These interactions provide a mechanism for recognition of the asymmetric

272 H3K4me3 feature contained in bivalent nucleosomes (illustrated in Figure 5B).

273 Given failure of asymmetric H3K4me3 nucleosomes to enrich KAT6B in pulldown

274 assays in vitro (Figures 1C and 3B), recruitment of KAT6B requires H3K27me3 at least

275 in vitro (Figures 3B–3D). However, KAT6B does not contain any known H3K27me3

276 binding domains. Enok, the Drosophila ortholog of KAT6B, has been shown to interact

277 with PRC1 (Kang et al., 2017; Kang et al., 2015; Strübbe et al., 2011), which could

278 mediate interaction with H3K27me3 in a KAT6B–PRC1 complex. However, we did not

279 observe interaction of murine KAT6B with PRC1 (Figure 5C). Moreover, RING1B

280 binding at bivalent promoters in ESCs was diminished upon EZH1/2 double knockout

281 (Figure S4D), whereas KAT6B binding remained largely unaffected (Figure 5A),

282 further arguing against an essential role of PRC1 in recruiting KAT6B in mouse ESCs.

283 Given that KAT6B acetylates H3K23 (Huang et al., 2016; Klein et al., 2019; Simó-

284 Riudalbas et al., 2015), we reasoned that the catalytic MYST domain of KAT6B might

285 interact with a H3K27me3 mark present adjacent to H3K23. Indeed, when performing

286 histone acetyltransferase assays with KAT6B complex purified from ESCs, KAT6B

287 was markedly more active towards H3K27me3 nucleosomes than towards unmodified

288 or H3K4me3-modified nucleosomes (Figure 5D). These findings suggest that the

289 catalytic MYST domain interacts with H3K27me3 as part of substrate recognition,

290 providing a transient H3K27me3 binding site that promotes recruitment of KAT6B to

291 bivalent nucleosomes (illustrated in Figure 5B). After acetylation, H3K23ac could

10 bioRxiv preprint doi: https://doi.org/10.1101/2021.02.08.430127; this version posted February 8, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

292 interact with the double PHD finger of KAT6B or the bromodomain of BRPF1

293 (illustrated in Figure 5B), providing additional multivalent interaction sites for the

294 KAT6B complex on H3K23ac-marked bivalent nucleosomes. To test the role of

295 H3K27me3 in recruitment of KAT6B in vivo, we introduced an EZH1/2 double knockout

296 into the ESC lines harboring tagged KAT6B (Figures S4B and S4C). Despite absence

297 of H3K27me3, levels of KAT6B at bivalent promoters remained largely unchanged

298 (Figure 5A). These observations suggest that additional interactions at bivalent

299 domains in vivo could mask the dependence of recruitment on H3K27me3 that is

300 clearly discernible in nucleosome pulldown assays in vitro. These interactions could

301 involve other histone marks such as H3K14ac, which is present on both active and

302 bivalent promoters in ESCs (Figure S5C).

303

304 KAT6B is required for neuronal differentiation

305 Having established that KAT6B is recruited to bivalent domains in vivo, we next

306 asked whether KAT6B regulates expression of bivalent genes in ESCs and during

307 differentiation. To this end, we generated knockout ESC lines for KAT6B and its

308 paralog KAT6A using CRISPR/Cas9 genome editing (Figures S5A and S5B).

309 KAT6A/B double knockout ESC lines exhibited diminished H3K23ac, whereas single

310 knockouts displayed only partial reductions in H3K23ac (Figures 6A and S5B). In

311 contrast, acetylation of H3K14 remained largely unchanged (Figure S5C), indicating

312 that KAT6A/B are essential for H3K23 but not H3K14 acetylation in mouse ESCs.

313 H3K4me3 and H3K27me3 were unaffected by knockout of KAT6A/B (Figure S5C).

314 Knockout of KAT6A/B did not affect proliferation or maintenance of ESC identity

315 (Figure 6B). Moreover, despite diminished H3K23ac at their promoters, expression of

316 housekeeping genes was unaffected in KAT6A/B knockout ESCs (Figure S5D),

317 suggesting that KAT6A/B and the H3K23ac mark are dispensable for expression of

318 active genes in ESCs. Furthermore, we did not observe derepression of bivalent genes

319 in KAT6A/B knockout ESCs (Figure 6D), indicating that KAT6B is not required for

320 maintaining their repression in ESCs.

321 To test whether KAT6B is instead required to regulate proper induction or silencing

322 of bivalent genes upon differentiation, we carried out neuronal differentiation towards

323 the glutamatergic neuronal lineage via formation of embryoid bodies (EBs) and neural

11 bioRxiv preprint doi: https://doi.org/10.1101/2021.02.08.430127; this version posted February 8, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

324 progenitor cells (Bibel et al., 2007) (Figure 6B–6D). After 15 days, formation of mature

325 neurons was clearly evident for wild-type E14 ESCs by the presence of networks of

326 neuronal protrusions that were positive for the neuron-specific class III ß-tubulin Tuj1

327 (Figure 6B). Strikingly, KAT6A/B single and double knockouts failed to generate

328 mature neurons (Figure 6B), indicating diminished differentiation potential of KAT6A/B

329 knockout ESCs. To determine how loss of KAT6A/B compromises differentiation, we

330 analyzed expression of pluripotency and differentiation marker genes. Whereas

331 downregulation of pluripotency factors Oct4 and Nanog was largely unaffected in

332 KAT6A/B knockout cells, expression of early differentiation markers Otx2 and Fgf5

333 was misregulated after 2 days of differentiation (Figure 6D). These genes feature

334 bivalent promoters, indicating a role for KAT6B and KAT6A in regulation of bivalent

335 genes at the onset of differentiation. Moreover, KAT6A/B knockout lines failed to

336 upregulate early and late neuronal markers such as Pax6, Mash1, and Fabp7

337 (Figure 6D), further supporting a pivotal role of KAT6B and KAT6A in the regulation of

338 bivalent developmental genes during differentiation. Interestingly, Hoxb13 was

339 excessively upregulated in KAT6A/B knockout cell lines (Figure 6D), indicating a

340 failure to also properly curtail expression of some bivalent genes in addition to

341 promoting expression of others. Taken together, these findings indicate that KAT6B is

342 required for proper differentiation of ESCs through regulation of bivalent

343 developmental genes.

344

345

12 bioRxiv preprint doi: https://doi.org/10.1101/2021.02.08.430127; this version posted February 8, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

346 Discussion

347 Bivalent domains have been posited to poise developmental genes for timely

348 expression or terminal repression upon differentiation while keeping their expression

349 repressed in ESCs (Azuara et al., 2006; Bernstein et al., 2006; Mikkelsen et al., 2007;

350 Voigt et al., 2013). However, it has remained unclear whether and how the histone

351 modifications present at bivalent domains directly cause—or contribute to—

352 establishment of a poised state. Our findings reveal a histone mark-based poising

353 mechanism at bivalent domains and highlight how nucleosomal asymmetry regulates

354 histone mark readout and function through allovalency, uncovering a novel regulatory

355 mechanism in the recruitment of chromatin binders. We show that the bivalent histone

356 marks themselves are sufficient to set up a poised state, revealing a direct role for the

357 bivalent modification state in regulating expression of developmental genes. We

358 propose that bivalent nucleosomes promote poising by recruiting the repressive

359 H3K27me3 binders PRC1 and PRC2 along with bivalency-specific binders, while

360 excluding active mark binders such as TAF3, preventing gene activation despite

361 presence of H3K4me3 (Figure 3E).

362 The asymmetric modification state of bivalent nucleosomes appears to be crucial

363 for the differential recruitment of active and repressive binders at bivalent domains.

364 For symmetrically modified nucleosomes, allovalent effects associated with the

365 presence of two equivalent binding sites in close proximity boost interaction with mark

366 binding proteins (see Figure 2B). These effects support efficient recruitment of even

367 lowly abundant binding proteins at symmetric nucleosomes, whereas recruitment of

368 highly abundant binders would be favored at asymmetric nucleosomes, as their

369 abundance overcomes the reduced affinity associated with asymmetric modification

370 (see Figure 2C). Accordingly, in nucleosome pulldown assays recruitment of

371 H3K4me3 binders such as TAF3 was robust for symmetric but not asymmetric

372 nucleosomes, whereas the more highly abundant H3K27me3 binders were efficiently

373 recruited also at asymmetric nucleosomes (Figure 2A and 2D). Consistent with these

374 findings and our proposed model, bivalent promoters in ESCs are bound by PRC

375 complex members such as CBX7 but fail to recruit TFIID, as shown for TAF3 (Figure

376 4A–C; Liu et al., 2011) and TBP (Ku et al., 2012).

377 Beyond bivalent domains, nucleosomal asymmetry of H3K4me3 could control

378 global transcriptional output by regulating recruitment of TFIID and potentially of other

13 bioRxiv preprint doi: https://doi.org/10.1101/2021.02.08.430127; this version posted February 8, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

379 H3K4me3 binders that facilitate transcription. Strikingly, H3K4me3 abundance at

380 promoters in mouse ESCs, when calibrated and normalized to biologically meaningful

381 modification densities using internal standards in ChIP-seq, correlates with RNA

382 expression in a sigmoidal fashion (Grzybowski et al., 2015). Promoters with H3K4me3

383 densities corresponding to asymmetric modification exhibit low expression, whereas

384 expression is markedly increased above the asymmetric-symmetric transition

385 (Grzybowski et al., 2015), potentially reflecting more efficient recruitment of H3K4me3

386 binders to symmetrically modified nucleosomes due to allovalent effects. Interestingly,

387 based on recent cryo-EM structures of human TFIID (Patel et al., 2018), it has been

388 proposed that recognition of promoter nucleosomes by TFIID could precede

389 engagement of promoter DNA by TBP (Bhuiyan and Timmers, 2019; Patel et al.,

390 2018), supporting a crucial role for the TAF3–H3K4me3 interaction in TFIID

391 recruitment. Similar to multivalency, a central principle governing affinity and

392 specificity of chromatin-based interactions via combinatorial readout of histone marks,

393 DNA, and RNA (see e.g. Patel and Wang, 2013; Ruthenburg et al., 2007; Su and

394 Denu, 2016), allovalency could therefore be crucial to regulating the recruitment and

395 functional output of chromatin complexes, especially for lowly abundant, narrowly

396 distributed marks such as H3K4me3 and their cognate, lowly abundant binders.

397 In addition to curbing recruitment of H3K4me3 binders, asymmetric bivalent

398 nucleosomes enrich PRC1 and PRC2 complexes. Given the central role of binding

399 proteins in mediating histone mark function, the proteins recruited to bivalent

400 nucleosomes likely set up or promote a poised state. Several mechanisms have been

401 proposed for PRC-mediated repression, including compacting chromatin, curbing

402 productive elongation by RNAPII, promoting a poised, exclusively S5-phosphorylated

403 state of RNAPII, and blocking acetylation of H3K27 through PRC2-mediated H3K27

404 methylation (Aranda et al., 2015; Schuettengruber et al., 2017; Simon and Kingston,

405 2013; Voigt et al., 2013). Given that allovalent effects at symmetric H3K27me3

406 nucleosomes enhance recruitment especially of PRC1 (Figures 1D, 3A, 3B, S3A), it is

407 conceivable that these modes of repression are attenuated at asymmetric bivalent

408 nucleosomes, establishing a poised rather than fully repressed state. Indeed, removal

409 of asymmetric H3K4me3 at bivalent domains via knockout of MLL2 increases levels

410 of H3K27me3, likely converting asymmetric bivalent to symmetric H3K27me3

411 nucleosomes, leading to increased recruitment of PRC2 and PRC1 and, in turn,

14 bioRxiv preprint doi: https://doi.org/10.1101/2021.02.08.430127; this version posted February 8, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

412 reduced promoter accessibility and levels of Ser5-phosphorylated RNAPII at bivalent

413 domains (Mas et al., 2018). These findings provide additional support for a key role of

414 nucleosomal asymmetry in setting up a poised rather than fully repressed state at

415 bivalent domains.

416 Interestingly, asymmetric bivalent nucleosomes displayed preference for PRC2

417 complexes containing EPOP and EZHIP (Figures 3A and S3A). EPOP and EZHIP

418 have been shown to reduce H3K27me3 levels by competing with binding of PRC2

419 auxiliary subunits such as JARID2 that confer more potent repressive activity to PRC2

420 (Beringer et al., 2016; Liefke et al., 2016; Ragazzini et al., 2019) and, in the case of

421 EZHIP, also by directly inhibiting PRC2 activity (Hübner et al., 2019; Jain et al., 2019).

422 The bivalency-specific recruitment of EPOP and EZHIP may therefore further

423 contribute to poising by curbing PRC2 activity. In agreement with our nucleosome

424 pulldown data, EPOP has been shown to localize to bivalent domains in ESCs

425 (Beringer et al., 2016; Liefke et al., 2016; Liefke and Shi, 2015).

426 In addition, we show that bivalent nucleosomes recruit factors that specifically

427 interact with bivalent nucleosomes (Figures 3B–D, S3A), providing a set of candidates

428 for novel mediators of poising, functioning alongside PRC1 and PRC2 at bivalent

429 domains (Figure 3E). In contrast to the plant proteins EBS and SHL, which were

430 recently shown to recognize both H3K27me3 and H3K4me3 in a mutually exclusive

431 fashion (Qian et al., 2018; Yang et al., 2018), the bivalency-specific binders identified

432 here simultaneously and multivalently engage both marks to achieve recruitment to

433 bivalent domains. Interestingly, besides KAT6B and SRCAP, most identified factors

434 lack known histone binding domains, suggesting involvement of novel binding

435 domains or recognition of composite binding surfaces generated by binding of PRCs,

436 SRCAP, or KAT6B to bivalent nucleosomes.

437 To clarify how the novel bivalency-specific factors contribute to the regulation of

438 developmental genes at bivalent promoters we focused our analysis on KAT6B. We

439 show that the complex is recruited to bivalent domains in ESCs (Figure 5A) and crucial

440 to regulating expression of bivalent domains during differentiation (Figure 6). The

441 KAT6B complex features well-characterized histone mark binding domains,

442 suggesting a model for the recruitment of KAT6B to bivalent nucleosomes (Figure 5B).

443 Multivalent interactions involving additional binding domains present in the complex

444 could further contribute to its recruitment to bivalent domains as well as active

15 bioRxiv preprint doi: https://doi.org/10.1101/2021.02.08.430127; this version posted February 8, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

445 promoters in ESCs in vivo. Our findings reveal that KAT6B contributes to maintaining

446 a poised chromatin state while being dispensable for active transcription, potentially

447 by maintaining accessibility of bivalent promoters through placement of H3K23ac or

448 other means. In support of such a role in chromatin accessibility, KAT6B has recently

449 been shown to facilitate interaction of OCT4 and NANOG with chromatin in ESCs

450 (Cosentino et al., 2019). Moreover, KAT6B-mediated H3K23ac could further

451 contribute to poising by modulating interaction of PRC1 and PRC2 with H3K27me3,

452 affecting recruitment of repressive complexes.

453 This study provides insight into the role of nucleosomal asymmetry and bivalency-

454 specific multivalency in regulating recruitment of binding proteins to bivalent domains.

455 Although the exact roles of Polycomb complexes and other factors recruited to bivalent

456 domains remain to be uncovered, our data provides evidence that the bivalent histone

457 marks themselves are sufficient to set up a poised state at bivalent domains. The

458 recruitment of KAT6B and SRCAP to bivalent nucleosomes highlights the potential

459 significance of additional layers of histone-based regulation at bivalent nucleosomes.

460 Adding to the roles exerted by the ‘core’ bivalent modification comprised of H3K4me3

461 and H3K27me3, CpG island promoters of developmental genes in ESCs appear to

462 adopt a state of ‘extended bivalency’ featuring H3K4me3, H3K27me3, H2A

463 ubiquitinylation, H2A.Z, H3.3, H3K14ac, H3K23ac, and possibly other marks.

464 Unraveling how these different modules interact at bivalent domains to recruit

465 chromatin complexes, promote poising, and regulate timely access to DNA will be

466 pivotal to further refining our understanding of the mechanisms that control expression

467 of developmental genes in ESCs.

468

16 bioRxiv preprint doi: https://doi.org/10.1101/2021.02.08.430127; this version posted February 8, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

469 Acknowledgments

470 We are grateful to Till Bartke for helpful discussions and advice on nucleosome

471 pulldown assays. We thank Tania Auchynnikava for help with initial MS analysis and

472 Giulia Bartolomucci for technical support in generating knockout and knockin ESC

473 lines. We are grateful to Dónal O’Carroll for critical reading of the manuscript. We

474 thank Adrian Bird, Atlanta Cook, and Ken Sawin for fruitful discussions and insightful

475 comments on this work. We also thank members of our labs for helpful discussions.

476 This work was supported by the Wellcome Trust ([104175/Z/14/Z], Sir Henry Dale

477 Fellowship to P.V., [091020] Equipment grant to J.R., and Multi-User Equipment grant

478 [108504]) and through funding from the European Research Council (ERC) under the

479 European Union’s Horizon 2020 research and innovation programme (ERC-STG grant

480 agreement No. 639253 to P.V.). The Wellcome Centre for Cell Biology is supported

481 by core funding from the Wellcome Trust [203149]. We are grateful to the Edinburgh

482 Protein Production Facility (EPPF) for their support. The EPPF was supported by the

483 Wellcome Trust through a Multi-User Equipment grant [101527/Z/13/Z]. Research in

484 the Baubec Lab is supported by the Swiss National Science Foundation (#157488 and

485 #180345) and the Swiss initiative in Systems Biology (SystemsX.ch).

486

487 Author contributions

488 E.B. conceived, designed, performed, and analyzed experiments. M.W., K.W.,

489 K.M., V.M. and C.A. designed and performed experiments. K.M. and E.B. further

490 analyzed and visualized data and generated figures. C.S. performed MS experiments

491 and analysis. T.B. provided conceptual input, analyzed and supervised experiments,

492 performed bioinformatic analysis, and secured funding. J.R. provided conceptual

493 advice, analysis tools, and supervision. J.R. further secured funding. P.V. conceived

494 the study, designed, performed, and analyzed experiments, wrote the manuscript with

495 input from E.B. and T.B., supervised the project, and secured funding. All authors

496 contributed to editing and revision of the manuscript.

497

498 Declaration of Interests

499 The authors declare no competing interests.

17 bioRxiv preprint doi: https://doi.org/10.1101/2021.02.08.430127; this version posted February 8, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

500 Figure Legends

501

502 Figure 1. Monovalent asymmetric nucleosomes fail to recruit H3K4me3 and

503 H3K27me3 binding proteins (A) Schematic overview of the experimental approach

504 used to determine the impact of asymmetric modification on histone mark binding.

505 (B) Enrichment and depletion of representative proteins on nucleosomes

506 symmetrically or asymmetrically modified by H3K4me3 or H3K27me3. Plotted are

507 label-free quantification (LFQ) intensities expressed as fold change relative to the

508 mean LFQ intensity of all pulldown conditions analyzed. Error bars represent 95%

509 confidence intervals of means, and diamonds denote individual values of three

510 independent experiments. Green bars, H3K4me3 binding proteins and associated

511 complex members; red bars, H3K27me3 binding proteins and associated PRC

512 members; dark red bars, non-PRC H3K27me3 binding proteins and associated

513 factors; H3K4me3/3, symmetric modification with H3K4me3; H3K4me0/3, asymmetric

514 modification with H3K4me3 on one copy of histone H3 per nucleosome. Analogous

515 notations used for H3K27me3-modified nucleosomes. (C,D) Volcano plots of LFQ MS

516 analysis of nucleosome pulldown experiments with symmetric (left panels) and

517 asymmetric (right panels) nucleosomes modified with H3K4me3 (C) and H3K27me3

518 (D), compared against unmodified symmetric and asymmetric (containing C-terminally

519 tagged histone H3) controls. The C-terminal tags present on histone H3 in asymmetric

520 nucleosomes did not significantly affect binding of the proteins studied here (see

521 Extended Data Fig. 1c). Means and p values are derived from three independent

522 pulldown experiments. Black dots, significantly enriched or depleted proteins (log2 fold

523 change ≤ 1.5 or ≥ 1.5, respectively, and p value ≤ 0.01), color scheme for known

524 H3K4me3 and H3K27me3 binders as in (B).

525 See also Figure S1 and Tables S1 and S2.

526

527 Figure 2. Symmetric modification enhances the apparent affinity of H3K4me3

528 and H3K27me3 binders towards nucleosomes (A) Western Blot analysis of titration

529 pulldown assays with nucleosomes carrying H3K4me3 in symmetric, asymmetric, and

530 asymmetric-tailless conformation. The pulldown labelled ‘mix’ was performed with an

531 equimolar mixture of symmetric and asymmetric nucleosomes. Blots shown are

18 bioRxiv preprint doi: https://doi.org/10.1101/2021.02.08.430127; this version posted February 8, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

532 representative of three independent experiments. (B) Scheme illustrating how

533 allovalent effects increase apparent affinity towards symmetric nucleosomes.

534 Availability of two binding sites enhances recruitment while concomitantly allowing for

535 rebinding to the neighboring mark upon dissociation, retaining the binder on the

536 nucleosome. (C) Allovalency shifts the binding equilibrium towards the bound state for

537 symmetric nucleosomes. The resulting gain in binding is expected to be stronger for

538 factors with concentrations below the apparent KD compared to highly abundant

539 binders, which would experience only minor gains in binding. Intensity-based absolute

540 quantification (iBAQ)-derived protein copy numbers per cell for representative

541 H3K4me3 and H3K27me3 binders are from Zhang et al. (2017). (D) Western Blot

542 analysis of titration pulldown assays with nucleosomes carrying H3K27me3 in

543 symmetric, asymmetric, and asymmetric-tailless conformation. Blots shown are

544 representative of three independent experiments.

545 See also Figure S2.

546

547 Figure 3. Bivalent nucleosomes recruit repressive H3K27me3 binders and

548 bivalency-specific binders, but not H3K4me3 binders (A) Volcano plot analysis of

549 LFQ MS data for nucleosome pulldown experiments with bivalent nucleosomes

550 compared against an unmodified asymmetric (containing C-terminally tagged histone

551 H3) nucleosome control. Means and p values are derived from three independent

552 pulldown experiments. Dark gray dots, significantly enriched or depleted proteins (log2

553 fold change ≤ 1.5 or ≥ 1.5, respectively, and p value ≤ 0.01). Known H3K4me3 and

554 H3K27me3 binders are highlighted by color as in Figure 1. (B) Enrichment and

555 depletion of representative proteins on symmetric as well as monovalent and bivalent

556 asymmetric nucleosomes. Plotted are fold changes of LFQ intensities relative to the

557 mean LFQ intensity of all symmetric (for first three bars) or asymmetric (for last four

558 bars) pulldown conditions analyzed. Error bars, 95% confidence interval of mean;

559 diamonds, individual values of three independent nucleosome pulldown experiments,

560 proteins color-coded as in panels (A) and (D). (C) Western Blot analysis of

561 nucleosome pulldowns for select H3K4me3 and H3K27me3 binding proteins as well

562 as HA-tagged KAT6B as bivalency-enriched protein. For the latter, NE of E14 ESCs

563 stably expressing FLAG-HA-KAT6B from its endogenous locus was used for pulldown

564 experiments. Data shown are representative of three independent pulldown

19 bioRxiv preprint doi: https://doi.org/10.1101/2021.02.08.430127; this version posted February 8, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

565 experiments. (D) Same Volcano plot analysis as in (A), this time highlighting proteins

566 specifically enriched on or depleted from bivalent nucleosomes. The C-terminal tags

567 present on histone H3 in asymmetric nucleosomes did not significantly affect binding

568 of proteins enriched on bivalent nucleosomes (see Figure S2B). Gold, KAT6B

569 subunits; orange, SRCAP subunits; purple, other highlighted bivalent binders; teal,

570 TAF subunits; blue, INO80 subunits; dark blue, zinc finger proteins; green and read,

571 as in panel (A). (E) Model illustrating how bivalent nucleosomes support poising of

572 genes by recruiting H3K27me3 binders as well as bivalency-specific binders, but not

573 binders of the active H3K4me3 mark.

574 See also Figure S3 and Table S3.

575

576 Figure 4. Bivalent nucleosomes in ESCs recruit CBX7 but not TAF3

577 (A) Genome browser snapshot showing distribution of H3K4me3, TAF3, H3K27me3,

578 and CBX7 over the Hoxd cluster and adjacent active genes. For all panels, TAF3

579 ChIPseq data is from Liu et al. (2011) and data for CBX7 from Morey et al. (2013).

580 Positions of CpG islands and genes are indicated below the tracks. (B) Heat maps of

581 H3K4me3, H3K27me3, TAF3, and CBX7 ChIP-seq read densities centered around

582 transcriptional start sites (TSS) and clustered as indicated, by high or low enrichment

583 of H3K4me3 and H3K27me3. Note presence of TAF3 on H3K4me3-positive but not

584 bivalent TSS, whereas CBX7 is bound to the latter. (C) Representation of data as

585 average density profiles around TSS clustered into H3K4me3-positive (blue), bivalent

586 (green), and H3K4me3-negative (red) TSS as in (B).

587

588 Figure 5. Bivalency promotes recruitment of KAT6B to developmental genes

589 in ESCs (A) ChIP-qPCR analysis with HA tag (upper panels), H3K4me3 (middle

590 panels), and H3K27me3 (lower panels) antibodies for select active, bivalent, and

591 inactive genes in E14 ESCs expressing FLAG-HA tagged KATB from its endogenous

592 locus (blue bars), and in ESC lines with additional mutation in MLL2 (catalytically

593 inactive MLL2 Y2688A, green bars) or knockout of EZH2 and EZH1 (red bars). Plotted

594 are means and SEM of four independent ChIP experiments. Statistically significant

595 differences (p ≤ 0.05, Student’s t test) in HA and H3K4me3 ChIP for control and

596 MLL2 Y2688A background are highlighted by asterisks. (B) Domain composition of

20 bioRxiv preprint doi: https://doi.org/10.1101/2021.02.08.430127; this version posted February 8, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

597 KAT6B complex and proposed model for its interaction with bivalent nucleosomes

598 based on previously described histone mark interaction and data obtained here.

599 (C) MS analysis of FLAG purification of FLAG-HA-KAT6B from NE of E14 ESC stably

600 expressing tagged KAT6B from its endogenous locus. Note co-purification of all known

601 KAT6B complex subunits but absence of PRC1 and PRC2 subunits. (D) Histone

602 acetyltransferase (HAT) assay with KAT6B complex purified from ESC NE as in (C).

603 KAT6B complex was incubated with the indicated nucleosome species and activity

3 604 was monitored via H incorporation from radiolabeled acetyl-CoA. Equal loading was

605 verified by Coomassie stain as shown below. Experiment shown representative of

606 three independent HAT assays.

607 See also Figure S4.

608

609 Figure 6. KAT6B is required for neuronal differentiation (A) ChIP-qPCR

610 analysis with H3K23ac antibody for select active, bivalent, and inactive genes in E14

611 ESCs (blue bars), KAT6B knockout (red bars), KAT6A knockout (yellow bars), and

612 KAT6A/B double knockout (purple bars) E14 ESCs. Plotted are means and SEM of

613 four independent ChIP experiments. (B) Oct4 and Tuj1 immunofluorescence staining

614 of ESCs before and after the neuronal differentiation protocol outlined in (C). Note that

615 neuronal protrusions positive for Tuj1 are formed in differentiated E14 control ESCs

616 but not in KAT6 knockout ESCs. Brightfield images further show presence or absence

617 of protrusions. Images shown are representative of three independent differentiation

618 experiments. Scale bars, 100 µm. (C) Outline of neuronal differentiation protocol.

619 (D) RT-qPCR analysis of mRNA expression changes in select pluripotency, early

620 differentiation, and neuronal marker genes in E14 ESCs and KAT6 knockout ESC

621 lines, expressed as fold changes relative to day 0 and normalized to expression in E14

622 ESC controls on day 0. Data are represented as mean and SEM of three independent

623 differentiation experiments.

624 See also Figure S5.

625

626

21 bioRxiv preprint doi: https://doi.org/10.1101/2021.02.08.430127; this version posted February 8, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

627 Supplemental Figure Legends

628

629 Figure S1, related to Figure 1 and Figure 3. Generation of modified

630 nucleosomes for pulldown experiments and control pulldowns. (A) Native

631 chemical ligation (NCL) was used to generate modified histone H3 with either

632 H3K4me3 or H3K27me3. Top left panel, schematic of NCL reaction involving

633 truncated histone H3 lacking the first 31 amino acids after the initiator methionine

634 (H3Δ1–31T32C) and a synthetic peptide spanning residues 1–31 of histone H3 and

635 containing trimethylated lysine at the desired residue as well as a C-terminal thioester

636 that is required for NCL. The reaction yields a full-length histone with the desired

637 modification and a threonine-to-cysteine point mutation at position 32. For some

638 experiments involving H3K27me3, T45 was used as the ligation site, resulting in a

639 T45C point mutation in the final modified histone. Bottom left panel, Representative

640 Coomassie gel analysis of NCL reaction for H3K4me3, showing unligated histone

641 (pre), crude reaction mixture (post), and final ligated histone purified by cation

642 exchange chromatography on a monoS column (purified). Right panel, representative

643 cation exchange purification performed after NCL reactions to separate ligated from

644 residual unligated histone. The elution profile of a linear NaCl gradient plotted as UV280

645 signal is shown along with a Coomassie stained SDS PAGE gel of fractions within the

646 indicated range of the gradient. Peak fractions containing ligated histones were pooled

647 and used for histone octamer assembly. (B) Analysis of symmetric and asymmetric

648 octamers used for assembly of nucleosomes. Presence of H3K4me3 and H3K27me3

649 was verified by Western blotting with H3K4me3 and H3K27me3 antibodies. H4 was

650 probed for as a loading control. Coomassie stain shows presence of histones H2B,

651 H2A, and H4 along with untagged (symmetric octamers) and C-terminally Strep- or

652 His-tagged histone H3 (asymmetric octamers). (C) Volcano plot analysis of LFQ MS

653 data for pulldown experiments with unmodified nucleosomes containing C-terminally

654 tagged His- and Strep-tagged histone H3 (used as controls for asymmetrically

655 modified nucleosomes) compared to unmodified and untagged nucleosomes (used as

656 controls for symmetrically modified nucleosomes) to assess the impact of the tags on

657 binding to asymmetric nucleosomes. Means and p values are derived from three

658 independent pulldown experiments. Black dots, significantly enriched or depleted

659 proteins (log2 fold change ≤ 1.5 or ≥ 1.5, respectively, and p value ≤ 0.01). Known

22 bioRxiv preprint doi: https://doi.org/10.1101/2021.02.08.430127; this version posted February 8, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

660 H3K4me3 and H3K27me3 binders are highlighted as in Figure 1. Note that presence

661 of the C-terminal tags on histone H3 does not adversely affect binding of those factors.

662 (D) Electrophoretic mobility shift assay (EMSA) performed to verify assembly of

663 recombinant nucleosomes used for pulldown experiments shown in Figures 1 and 3.

664 Mononucleosomes (nuc) were separated from free 601 DNA (601) on native 6%

665 polyacrylamide gels in TGE buffer and visualized with SYBR safe stain.

666

667 Figure S2, related to Figure 2. Nucleosome assembly for titration pulldowns

668 (A,B) EMSAs verifying assembly of recombinant nucleosomes used for pulldown

669 experiments shown in Figure 2. Nucleosome preparations with H3K4me3 (A) and

670 H3K27me3 (B) modifications were run on native 6% polyacrylamide gels in TGE buffer

671 and visualized with SYBR safe stain.

672

673 Figure S3, related to Figure 3. Nucleosome pulldown data for additional

674 bivalent binders and control pulldowns. (A) Bars plots showing representative

675 proteins specifically enriched or depleted on bivalent asymmetric nucleosomes.

676 Plotted are fold changes of LFQ intensities relative to the mean LFQ intensity of all

677 pulldown conditions analyzed. Error bars, 95% confidence interval of mean; diamonds,

678 individual values of three independent nucleosome pulldown experiments, proteins

679 color-coded as in Figure 3. (B) Volcano plot analysis of LFQ MS data for pulldown

680 experiments with unmodified nucleosomes containing C-terminally tagged His- and

681 Strep-tagged histone H3 (used as controls for asymmetrically modified nucleosomes)

682 compared to unmodified and untagged nucleosomes (used as controls for

683 symmetrically modified nucleosomes) to assess the impact of the tags on binding to

684 asymmetric nucleosomes (same data as in Figure S1C). Proteins enriched or depleted

685 on bivalent nucleosomes are highlighted as in Figure 3, showing that presence of the

686 tags does not adversely affect binding of factors enriched on asymmetric bivalent

687 nucleosomes, whereas several factors depleted from asymmetric bivalent

688 nucleosomes exhibit favorable interactions with the tags.

689

690 Figure S4, related to Figure 5. Generation and characterization of knockin

691 and knockout cell lines and additional ChIP data (A) Generation of E14 ESC lines

23 bioRxiv preprint doi: https://doi.org/10.1101/2021.02.08.430127; this version posted February 8, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

692 expressing FLAG-HA-tagged KAT6B via CRISPR genome editing and homologous

693 repair. (B) After successful generation of the ESC line expressing tagged KAT6B

694 described in (A), additional CRISPR genome editing was performed to introduce a

695 point mutation in the Kmt2b gene encoding the catalytically inactive Y2688A point

696 mutation in MLL2 (top panel) as well as to generate a double knockout of EZH2 and

697 EZH1 by introducing a series of premature stop codons in Ezh2 and Ezh1 genes

698 (middle and bottom panels). (C) Western blot analysis on whole cell extracts of ESC

699 lines generated, verifying presence of the HA tag on KAT6B, reduced levels of

700 H3K4me3 in MLL2 Y2688A ESCs, and absence of EZH2 and H3K27me3 in

701 EZH2/EZH1 double knockout ESCs. H4 was used as a loading control. Data shown

702 are representative of two Western blot experiments. (D) ChIP-qPCR analysis with

703 RING1B antibody for select active, bivalent, and inactive genes in E14 ESCs

704 expressing FLAG-HA tagged KATB (blue bars), and cell lines with additional mutation

705 in MLL2 (catalytically inactive MLL2 Y2688A, green bars) or knockout of EZH2 and

706 EZH1 (red bars). Plotted are means and SEM of four independent ChIP experiments.

707 Note that RING1B enrichment is diminished in absence of H3K27me3.

708

709 Figure S5, related to Figure 6. Generation and characterization of KAT6A/B

710 knockout cell lines and additional ChIP and RT-qPCR data (A) Generation of

711 KAT6B and KAT6A knockout E14 ESCs via CRISPR genome editing and homologous

712 repair to introduce a series of premature stop codons in Kat6b and Kat6a genes.

713 (B) Western blot analysis confirming knockout of KAT6B and concomitant loss of

714 H3K23ac in KAT6B and KAT6A knockouts, highlighting partial redundancy between

715 both acetyltransferases. H4 and EZH2 were used as loading controls. Blots shown are

716 representative of two independent experiments. (C) Additional ChIP-qPCR analysis of

717 KAT6A/B knockout ESC lines with H3K14ac, H3K4me3, and H3K27me3 antibodies

718 showing that these marks remain largely unchanged in the knockout ESC lines. Select

719 active, bivalent, and inactive genes were analyzed in E14 ESCs (blue bars), KAT6B

720 (red bars), KAT6A (yellow bars), and KAT6A/B double (purple bars) knockout ESCs.

721 Plotted are means and SEM of four independent ChIP experiments. (D) RT-qPCR

722 analysis of housekeeping gene expression in E14 ESCs and KAT6 knockout ESC

723 lines, showing that housekeeping gene expression is not affected in knockout cell lines

724 despite diminished H3K23ac levels in those cell lines. Expression was normalized to

24 bioRxiv preprint doi: https://doi.org/10.1101/2021.02.08.430127; this version posted February 8, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

725 expression of Gapdh. Data are represented as mean and SEM of three independent

726 experiments.

727

25 bioRxiv preprint doi: https://doi.org/10.1101/2021.02.08.430127; this version posted February 8, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

728 Supplemental Tables

729

730 Table S1. Proteins significantly enriched or depleted for symmetric and

731 asymmetric H3K4me3 nucleosomes.

732

Complex Direct binder Significantly Enriched or Depleted Subunits Enriched for symmetric H3K4me3 nucleosomes: SPIN1 SPIN1 SPIN1, SPINDOC NURF BPTF BPTF, HMGXB4, BAP18, BAP18.1 SIN3A ING2 ING2, BRMS1L, ARID4B, SUDS3, BRMS1, SAP30, SAP130, ARID4A, SAP30L, SIN3A, FAM60A SET1 CXXC1 CXXC1, RBBP5, DPY30, SETD1A TFIID TAF3 TAF3 EMSY JARID1A SIN3B, JARID1A, GATAD1, EMSY, PHF12 not assigned to complexes ZGLP1, PHF2, MORC3, PHF13 (SPOC1), CHD8, PHF23, CHD6, DIDO1, JARID1B, CHD1

Depleted for symmetric H3K4me3 nucleosomes: PRC2 EPOP, C10orf12/PALI1, EZH1, AEBP2, EED.1, SUZ12, EZH2, MTF2 PRC1 PCGF6 G9A/GLP G9A, GLP, WIZ not assigned to complexes PHF21A, HMG20B, BEND3, UHRF1, TAF1C, THAP1, BAZ2A, PRDM10, ZFP219, PAX6, BAZ2B, TOP3A, HIF1A

Enriched for asymmetric H3K4me3 nucleosomes*: not assigned to complexes ZRANB2, DNAJC9, MBD1

733 *: not significant by adjusted p value.

734

735

736

737

738

26 bioRxiv preprint doi: https://doi.org/10.1101/2021.02.08.430127; this version posted February 8, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

739 Table S2. Proteins significantly enriched or depleted for symmetric and

740 asymmetric H3K27me3 nucleosomes.

741

Complex Direct binder Significantly Enriched or Depleted Subunits Enriched for symmetric H3K27me3 nucleosomes: PRC1 CBX7 CBX7, PHC1 ORC ORC3, LRWD1, ORC2, ORC1 not assigned to complexes ARID4B, PHF2

Depleted for symmetric H3K27me3 nucleosomes: PRC2 AEBP2 not assigned to complexes G2E3,ZBTB2.

Enriched for asymmetric H3K27me3 nucleosomes*: not assigned to complexes NFYB, NFYC, SIX4

Depleted for asymmetric H3K27me3 nucleosomes: not assigned to complexes SURF6*, GNL3

742 *: not significant by adjusted p value.

743

744

745

746

747

748

749

750

751

752

753

754

27 bioRxiv preprint doi: https://doi.org/10.1101/2021.02.08.430127; this version posted February 8, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

755 Table S3. Proteins significantly enriched for asymmetric bivalent

756 nucleosomes.

757

Complex Significantly Enriched Subunits PRC1 CBX7, PHC1 PRC2 EPOP, EED, EZHIP, PHF13 (SPOC1), EZH2* ORC ORC2, LRWD1 KAT6B BRPF3, KAT6B, MEAF6*, BRPF1*, ING5* SRCAP SRCAP, YEATS4, DMAP1, VPS72 Splicing factors ZRANB2, SRSF2 Chromatin modifiers SETD8, JARID1B, EP300, HMG20A Transcription Factors MITF, TFEB, TFE3, SREBF2, MYCN, MYC, MLX RNA binding proteins PUM1, PUM2 Metabolic enzymes ACACA, MCCC1&2, PCCB, PCX, PCCA, TECR DNA repair DNAJC9, XRCC5&6, PARP2, DDB1&2, ANKRD32, PARP1.1 Other functions FHL2&3, RAI1

758 *: log2 enrichment between 1.5–1.25.

759

760

761

762

763

28 bioRxiv preprint doi: https://doi.org/10.1101/2021.02.08.430127; this version posted February 8, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

764 Methods

765 Cell culture

766 E14 ESCs were grown on plates coated with gelatin (0.1%, Sigma) in DMEM

767 containing 4.5 g/l glucose (Gibco) supplemented with 15% fetal bovine serum (Gibco,

768 South American, A3160802), 2 mM l-glutamine, 1 mM pyruvate, 1x MEM nonessential

769 amino acids, 50 units/ml penicillin and 50 µg/ml streptomycin (all Gibco), 0.2 mM β-

770 mercaptoethanol (Sigma), and heterologously expressed recombinant LIF (made in-

771 house and batch-tested for maintenance of self-renewal) at 37°C and 5% CO2.

772

773 CRISPR-mediated genome editing of ESCs

774 To perform genome editing in E14 ES cells, sgRNAs specific to the gene of interest

775 were designed and cloned into pX458 ((addgene #48138)Ran et al., 2013). For

776 homology repair, 200-base single stranded oligodeoxynucleotide (ssODN, IDT)

777 donors were designed to encode the desired mutations or insertions flanked by

778 homology regions of at least 30 bp. Low-passage E14 ESCs were transfected with

779 pX458 plasmid and the corresponding ssODN donor using Lipofectamine 2000 or

780 3000 (Invitrogen). After GFP fluorescence-based FACS sorting of transfected single

781 cells, single cell colonies were expanded for 7–10 days on 15-cm plates before picking

782 of single colonies, trypsinisation, and further expansion in 96-well plates in duplicate.

783 DNA was extracted from one plate using QuickExtract DNA extraction solution

784 (Epicentre) to perform genotyping by PCR using primer pairs specifically recognizing

785 the mutated sequences introduced by the ssODN and a genomic sequence adjacent

786 to the region of integration. To confirm correct genotypes in cell clones positive for the

787 desired mutations from the genotyping PCR, genomic sequences surrounding and

788 including the integration sites were amplified by PCR and analyzed by Sanger

789 sequencing on PCR-amplified genomic material to confirm presence of the desired

790 mutations only.

791

792 Preparation of ES cell nuclear extract

793 E14 ESC nuclear extract (NE) was prepared based on modifications of existing

794 protocols (Conaway et al., 1996; LeRoy et al., 1998). E14 ESCs were washed in PBS

29 bioRxiv preprint doi: https://doi.org/10.1101/2021.02.08.430127; this version posted February 8, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

795 (Gibco) and harvested by scraping into PBS supplemented with 1 mM EDTA. All

796 subsequent steps were performed at 4°C. Cells were collected by centrifugation at

797 500 g for 8 min, washed with TMSD (20 mM HEPES pH 7.9, 10 mM KCl, 1.5 mM

798 MgCl2, 250 mM sucrose, 0.5 mM DTT, 0.2 mM PMSF), and collected by centrifugation

799 as before. The cells were resuspended in TMSD containing 0.05% NP40 and

800 incubated on ice for 5 min to allow for cell lysis to occur. After successful cell lysis was

801 confirmed by staining an aliquot of the cell suspension with Trypan Blue (Gibco), nuclei

802 were pelleted by centrifugation for 10 min at 1000 g. The supernatant (cytosolic

803 fraction) was removed and the pellet containing nuclei resuspended in BC420 buffer

804 (20 mM HEPES pH 7.9, 420 mM KCl, 20% glycerol, 1.5 mM MgCl2, 0.2 mM EDTA,

805 0.5 mM DTT, 0.2 mM PMSF) and incubated at 4°C with gentle rotation for 1 h. After a

806 15-s vortex, the suspension was centrifuged for 30 min at 25,000 g. The supernatant

807 was dialysed three times against 50 volumes of BC150 buffer (20 mM HEPES pH 7.9,

808 150 mM KCl, 20% glycerol, 1.5 mM MgCl2, 0.2 mM EDTA, 0.5 mM DTT, 0.2 mM

809 PMSF) for 1 h each. The dialysate was centrifuged for 30 min at 25,000 g. The

810 resultant supernatant is the NE. For all nucleosome pulldowns, 3 independent

811 extractions were performed and the resulting NEs combined to compensate for batch

812 variations between individual preparations.

813

814 Expression and purification of histones

815 Xenopus H3 and H4 and human H2A and H2B were expressed from pET-3a or

816 pET-3d vectors in BL21 (DE3) pLysS for H3, H2A, and H2B or BL21 (DE3) for H4

817 through induction with 0.2 mM IPTG for 4 h at 37°C. Histones were purified from

818 inclusion bodies and solubilised in unfolding buffer (20 mM Tris-HCl pH 7.5, 7 M

819 guanidine HCl, 10 mM DTT). Extracted histones were dialysed against three changes

820 of urea dialysis buffer (10 mM Tris HCl pH 8, 7 M urea, 100 mM NaCl, 1 mM EDTA,

821 5 mM β-mercaptoethanol). This and all subsequent histone dialysis steps were carried

822 out at 4°C. Histones were then purified further by passing over a HiTrap Q column

823 (GE Healthcare) before binding to and NaCl gradient elution from HiTrap SP cation

824 exchange chromatography columns (GE Healthcare). Fractions containing histones

825 were pooled and dialyzed three times against water containing 5 mM β-

826 mercaptoethanol and lyophilised for long-term storage at -80°C.

30 bioRxiv preprint doi: https://doi.org/10.1101/2021.02.08.430127; this version posted February 8, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

827 To express histones for native chemical ligation (NCL), constructs encoding

828 truncated Xenopus histone H3 were generated in pET-3a. For generation of

829 H3K4me3- and H3K27me3-modified histones, truncated H3 lacking residues 1–31

830 after the initiator methionine, with a threonine-to-cysteine substitution at position 32 of

831 Xenopus H3 and a cysteine-to-alanine substitution at position 110 (H3Δ1–31T32C

832 C110A) was expressed in BL21 (DE3) pLysS and purified as above, except for the

833 final dialysis, which was carried out as two rounds of dialysis against 1 mM DTT in

834 H2O and one round against 0.5 mM TCEP (Sigma) before lyophilisation and storage.

835 For generation of H3K27me3-modified histones for some experiments, a similarly

836 truncated Xenopus H3 construct was used, lacking the first 44 residues and carrying

837 a threonine-to-cysteine mutation at residue 45 (H3Δ1–44T45C C110A).

838

839 Generation of modified histones by native chemical ligation

840 NCL reactions were carried out in 6 M Guanidine HCl, 250 mM sodium phosphate

841 buffer pH 7.2, 150 mM 4-mercaptophenylacetic acid (MPAA, Sigma), 50 mM TCEP

842 for 72 h at room temperature with constant agitation. Reactions were then dialyzed

843 three times against urea dialysis buffer (see above, but with 1 mM DTT instead of

844 5 mM β-mercaptoethanol). Ligated full-length modified histone was separated from

845 unligated histone through cation exchange chromatography on a HiTrap SP column

846 (GE Healthcare) and then dialysed against three changes of water containing 5 mM

847 β-mercaptoethanol before lyophilization and storage at -a80°C until use. For

848 H3K4me3- or H3K27me3-modified histones, H3Δ1–31T32C C110A was reacted with

849 a synthetic peptide spanning residues 1–31 of histone H3.1 containing trimethylated

850 lysine at position 4 or 27, respectively, and a C-terminal benzyl thioester (Peptide

851 Protein Research). For some H3K27me3-modified histones, H3Δ1–44T45C C110A

852 was reacted with a synthetic peptide spanning residues 1–44 of histone H3.1 including

853 trimethylated lysine at position 27 and a C-terminal benzyl thioester.

854

855 Reconstitution of recombinant nucleosomes

856 To reconstitute histone octamers, the four core histones were resuspended in

857 unfolding buffer (see above), mixed in a mass ratio of 1:1:1.2:1.2 (H3:H4:H2A:H2B),

858 and dialysed against three changes of refolding buffer (10 mM Tris HCl pH 8, 2 M

31 bioRxiv preprint doi: https://doi.org/10.1101/2021.02.08.430127; this version posted February 8, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

859 NaCl, 1 mM EDTA, 5 mM β-mercaptoethanol) at 4°C. After centrifugation to remove

860 precipitate formed during dialysis, correctly assembled histone octamers were purified

861 by size exclusion chromatography in refolding buffer on a S200 column (GE

862 Healthcare), using an Akta PURE system (GE Healthcare).

863 Asymmetrically modified octamers were further purified as described previously

864 (Voigt et al., 2012), with slight modifications; the two copies of histone H3 incorporated

865 into asymmetric octamers were differentially tagged with either His or Strep tag.

866 Sequential affinity chromatography was performed on the combined fractions

867 containing the desired asymmetric protein complexes from the size exclusion

868 chromatography step by first using a Ni-sepharose HisTrap column (GE Healthcare)

869 and then a StrepTrap HP column (GE Healthcare). Symmetric His-His-tagged

870 complexes and the desired asymmetric complexes containing one His- and one Strep-

871 tagged histone H3 were eluted from the HisTrap column with a gradient (0–250 mM)

872 of imidazole in elution buffer (2 M NaCl, 25 mM Tris pH 8). Fractions containing His-

873 tagged protein as verified by Western blot with an anti-His antibody (Sigma) were

874 pooled and loaded onto the StrepTrap column. Symmetric His-His-tagged complexes

875 eluted in the flowthrough, whereas asymmetric His-Strep-tagged complexes were

876 eluted from the StrepTrap column in elution buffer containing 2.5 mM desthiobiotin

877 (IBA). After each chromatography step, elution profiles and sample purity were

878 confirmed by Western blot with antibodies against His tag (Sigma, H1029, lot

879 033m4785) and Strep tag (IBA, 2-150-7001).

880 The DNA template for mononucleosome assembly was generated by PCR with a

881 biotinylated forward primer, amplifying a 209-bp fragment centred around the 147-bp

882 601 nucleosome positioning sequence (Lowary and Widom, 1998), followed by PCR

883 purification and elution into TE buffer (10 mM Tris pH 8, 1 mM EDTA). To reconstitute

884 recombinant mononucleosomes, DNA and histone octamers were combined in

885 refolding buffer supplemented with 5 M NaCl to compensate for reduction in NaCl

886 concentration due to introduction of TE buffer with the DNA, followed by gradient

887 dialysis against TE buffer down to 400 mM NaCl and then a step dialysis against TE

888 buffer at 4°C. Optimal ratios of DNA and histone octamer were determined so that at

889 least 95% of DNA was complexed while avoiding over-assembly and unspecific DNA

890 binding of histones. Assemblies were routinely checked by native gel electrophoresis

891 on 6% acrylamide gels in TGE buffer.

32 bioRxiv preprint doi: https://doi.org/10.1101/2021.02.08.430127; this version posted February 8, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

892

893 Nucleosome pulldown assays

894 For pulldown assays with recombinant modified nucleosomes and E14 ESC NE,

895 streptavidin sepharose high performance beads (GE Healthcare, 10.5 µl of slurry per

896 pulldown) were briefly washed three times with pulldown buffer (20 mM HEPES

897 pH 7.9, 150 mM NaCl, 10% glycerol, 1 mM EDTA, 1 mM DTT, 0.2 mM PMSF,

898 0.1% v/v NP-40). All centrifugation steps were carried out at 1,500 g for 2 min at 4°C.

899 All incubation steps were carried out with constant rotation at 30 rpm on a rotator mixer

900 (Starlab) at 4°C in the cold room. After washes, beads were incubated overnight with

901 10.5 µg of assembled recombinant nucleosomes (in TE buffer diluted with pulldown

902 buffer and adjusted to a final concentration of 0.1% NP-40 and 150 mM NaCl by

903 addition of 10% NP-40 and 5 M NaCl, respectively). For pulldown titrations with

904 increasing amounts of recombinant nucleosomes, nucleosomes were assembled in 4

905 batches of 10 µg each, combined, and incubated with beads in the amounts indicated

906 in Fig. 1e, g. The amount of beads used was scaled with the amount of nucleosomes

907 (6.4, 6.4, 12, 24 µl slurry for 3, 6, 12, 24 µg of nucleosomes). Beads were then

908 collected by centrifugation and washed briefly with three changes of pulldown buffer.

909 Bead-bound nucleosomes were incubated with 500 µg of NE for 2 h. Beads were then

910 washed with pulldown buffer by 5-min incubations under rotation, first with two washes

911 of pulldown buffer containing NP-40, followed by three washes of pulldown buffer

912 without NP-40.

913 After washes, bound proteins were eluted from beads. For subsequent LC-MS/MS

914 analysis, beads were resuspended in elution buffer (2 M Urea, 100 mM Tris pH 7.5,

915 10 mM DTT) and incubated for 20 min in a thermomixer (Eppendorf) at 1000 rpm at

916 25°C. After incubation, iodoacetamide (Sigma) was added to a final concentration of

917 55 mM and incubated for 10 min in the thermomixer. The eluted proteins were

918 digested with 0.3 µg of trypsin (Thermo Fisher Scientific) for 2 h whilst shaking and

919 then collected by centrifugation at 1500 g for 2 min. The supernatant containing eluted

920 peptides was removed and 50 µl elution buffer was added to the beads followed by

921 5 min incubation in a thermomixer. The beads were centrifuged at 1,500 g for 2 min

922 and the two elutions combined. 0.15 µg trypsin added and the solution was digested

923 overnight at room temperature before digestion was stopped by the addition of 10%

33 bioRxiv preprint doi: https://doi.org/10.1101/2021.02.08.430127; this version posted February 8, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

924 TFA. For LC-MS/MS LFQ analysis, pulldowns were performed in triplicate to ensure

925 robustness of LFQ-MS analysis.

926 For Western Blot analysis, elution was performed by boiling for 5 min at 95°C with

927 1.5x SDS sample buffer (95 mM Tris HCl pH 6.8, 15% glycerol, 3% SDS, 75 mM DTT,

928 0.15% bromophenol blue). 30% of bound sample was loaded for Western blot

929 analysis. Binding was analyzed by probing with antibodies against TAF3 (abcam,

930 AB188332, lot GR220332-3), ING2 (abcam, AB109594, lot YH102514), PHF2 (clone

931 D45A2, Cell Signaling, 3497s, lot 2), CBX7 (abcam, AB21873, lots GR3210651-1 and

932 -2), EZH2 (BD Biosciences, 612667, lot 415817), RING1B (clone D22F2, Cell

933 Signaling, 5694s, lot 3), HA tag (abcam, AB9110, lot GR3199553-1, or Cell Signaling,

934 clone C29F4, 3724s, lot 9). Antibodies against histone H4 (Cell Signaling, 13919, lot

935 3), H3K4me3 (abcam, ab8580, lot GR273043-2), and H3K27me3 (Cell Signaling,

936 9733, lot 8) were used as loading controls. Blots were developed with Clarity ECL

937 reagent (Bio-Rad) and imaged with a Chemidoc Touch imaging system (Bio-Rad). For

938 Western analysis, pulldowns were performed in duplicate or triplicate.

939

940 LC-MS/MS analysis

941 Following digestion, half of each sample was diluted with an equal volume of 0.1%

942 TFA and loaded onto home-made StageTips as described (Rappsilber et al., 2003).

943 Peptides were eluted in 40 μl 80% acetonitrile in 0.1% TFA and concentrated down to

944 1 μl by vacuum centrifugation (Concentrator 5301, Eppendorf). Samples were then

945 prepared for LC-MS/MS analysis by diluting them to 6 μl with 0.1% TFA.

946 LC-MS/MS analyses for samples that were digested on beads were performed on

947 a Q Exactive mass spectrometer and for samples processed by in-gel digestion on an

948 Orbitrap Fusion Lumos Tribrid mass spectrometer (both Thermo Fisher Scientific),

949 both coupled on-line to Ultimate 3000 RSLCnano Systems (Dionex, Thermo Fisher

950 Scientific). Peptides were separated on a 50-cm EASY-Spray column (Thermo Fisher

951 Scientific) assembled in an EASY-Spray source (Thermo Fisher Scientific) and

952 operated at 50°C. In both cases, mobile phase A consisted of 0.1% formic acid in

953 water, while mobile phase B consisted of 80% acetonitrile and 0.1% formic acid in

954 water. Peptides were loaded onto the column at a flow rate of 0.3 µl/min and eluted at

955 a flow rate of 0.25 µl/min with the following gradient: 2–40% buffer B in 150 min, then

34 bioRxiv preprint doi: https://doi.org/10.1101/2021.02.08.430127; this version posted February 8, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

956 to 95% in 11 min. For Q Exactive runs, FTMS spectra were recorded at 70,000

957 resolution (scan range 350-1400 m/z) and the ten most intense peaks of the MS scan

958 with charge ≥ 2 were selected with an isolation window of 2.0 Thomson for MS2 (filling

959 1.0E6 ions for MS scan, 5.0E4 ions for MS2, maximum fill time 60 ms, dynamic

960 exclusion for 50 s). For Orbitrap Fusion Lumos runs, survey scans were performed at

961 120,000 resolution (scan range 350-1400 m/z) with an ion target of 4.0e5. MS2 was

962 performed in the ion trap at a rapid scan rate with ion target of 2.0E4 and HCD

963 fragmentation (Olsen et al., 2007) with a normalized collision energy of 27. The

964 isolation window in the quadrupole was 1.4. Only ions with a charge between 2 and 7

965 were selected for MS2.

966

967 Processing and visualisation of LC-MS/MS data

968 The MaxQuant software platform (Cox and Mann, 2008) version 1.6.1.0 was used

969 to process raw files and searches were conducted against the Mus musculus

970 complete/reference proteome (Uniprot, released in July, 2017), using the Andromeda

971 search engine (Cox et al., 2011). The first search peptide tolerance was set to 20 ppm

972 while the main search peptide tolerance was set to 4.5 ppm. Isotope mass tolerance

973 was 2 ppm and maximum charge was set to 7. A maximum of two missed cleavages

974 was allowed. Carbamidomethylation of cysteine was set as fixed modification.

975 Oxidation of methionine and acetylation of the N terminus were set as variable

976 modifications. Label-free quantification (LFQ) analysis was performed by employing

977 the MaxLFQ algorithm as described (Cox et al., 2014). Peptide and protein

978 identifications were filtered to 1% FDR.

979 Data was analysed and visualised in R version 3.5.3. The DEP package (Zhang et

980 al., 2018) version 1.4.1 was used to determine differential enrichment of proteins

981 between nucleosome pulldowns. Data was filtered on reverse hits and potential

982 contaminants, before missing values were filtered to retain only proteins quantified in

983 at least two replicates across all nucleosome pulldown conditions and replicates

984 analysed. The data was normalized using vsn, and imputation was carried out using

985 the MinProb function in DEP. Proteins were considered differentially enriched when

986 the log2 fold change was greater than 1.5, and the p value was lower than 0.01. Bar

987 plots of fold changes of LFQ intensities relative to the mean LFQ intensity for select

35 bioRxiv preprint doi: https://doi.org/10.1101/2021.02.08.430127; this version posted February 8, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

988 proteins were generated in DEP using the accompanying Shiny package. Volcano

989 plots were generated in ggplot2.

990

991 Chromatin immunoprecipitation

992 To prepare chromatin for ChIP, E14 ESCs were grown as described above.

993 Crosslinking was carried out directly on 15-cm culture plates with 1% formaldehyde in

994 fixation buffer (DMEM supplemented with 10 mM HEPES pH 7.6, 15 mM NaCl,

995 0.15 mM EDTA, 0.075 mM EGTA for 10 min at room temperature with gentle rocking.

996 Crosslinking was stopped by addition of glycine to 0.125 M final concentration and

997 incubation for 3 min at room temperature with gentle rocking. Fixed cells were washed

998 with cold PBS, scraped off the plate in cold PBS, and collected by centrifugation at

999 2500 g for 5 min at 4°C. Cell pellets snap frozen in liquid nitrogen and stored at -80°C.

1000 To prepare fragmented chromatin, cell pellets were resuspended in lysis buffer 1

1001 (50 mM HEPES pH 7.6, 140 mM NaCl, 1 mM EDTA, 10% glycerol, 0.5% NP-40,

1002 0.25% Triton-X100) and incubated with constant rotation for 10 min at 4°C. Cells were

1003 then centrifuged at 3000 g for 5 min and cell pellets resuspended in lysis buffer 2

1004 (10 mM Tris-HCl pH 8.0, 200 mM NaCl, 1 mM EDTA, 0.5 mM EGTA) and incubated

1005 and centrifuged as for lysis buffer 1. Pellets were then resuspended in lysis buffer 3

1006 (10 mM Tris-HCl pH 8.0, 1 mM EDTA, 0.5 mM EGTA, 0.5% N-Lauryl Sarcosine) at a

1007 density of 75 mg cell pellet weight/ml. The suspension was transferred to polystyrene

1008 tubes (BD Falcon) and sonicated using a Bioruptor (Diagenode) at maximum output

1009 setting using 30 s ON, 30 s OFF cycles. Number of cycles (15–20) was experimentally

1010 determined to yield decrosslinked fragments of 250–350 bp size. Size of chromatin

1011 fragments was verified on a 1% agarose gel and on a high sensitivity DNA chip in a

1012 2100 bioanalyzer (Agilent) after decrosslinking overnight at 65°C with shaking in

1013 elution buffer (100 mM NaHCO3, 1% SDS, 200 mM NaCl, 0.4 mg/ml each proteinase

1014 K and RNAse A). Sonicated suspensions were clarified by centrifugation at 20,000 g

1015 for 10 min at 4°C. DNA concentration as a proxy for chromatin concentration was

1016 determined by Nanodrop ND-1000 (Thermo Fisher Scientific).

1017 For ChIP, chromatin was clarified by centrifugation at 20,000 g for 10 min and

1018 incubated overnight with antibodies in 1x IP buffer (10 mM Tris-HCl pH 8.0, 150 mM

1019 NaCl, 1 mM EDTA, 0.5 mM EGTA, 0.25% N-Lauryl Sarcosine, 1% Triton X100) with

36 bioRxiv preprint doi: https://doi.org/10.1101/2021.02.08.430127; this version posted February 8, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

1020 constant rotation at 4°C. Per IP, 50 μl protein A or G magnetic dynabeads (Invitrogen)

1021 were prepared by washing twice, then incubating overnight with PBS containing 0.5%

1022 (w/v) BSA overnight at 4°C with rotation, then washing 5 times in TE buffer (10 mM

1023 Tris pH 8, 1 mM EDTA), and finally resuspending in TE. After overnight incubation,

1024 IPs were centrifuged at 20,000 g for 10 min at 4°C and the supernatants were

1025 combined with blocked dynabeads and incubated at 4°C for 2–3 h under constant

1026 rotation. Beads were then washed 5 times in cold RIPA buffer (10 mM HEPES pH 7.6,

1027 1 mM EDTA, 0.5 M LiCl, 1% NP-40, 0.1% N-Lauryl Sarcosine) with 2 min rotating

1028 incubation at room temperature for each wash. Beads were then washed briefly with

1029 TEN buffer (TE with 50 mM NaCl). Immunoprecipitated DNA was eluted from washed

1030 beads by addition of elution buffer (100 mM NaHCO3, 1% SDS, 200 mM NaCl,

1031 0.4 mg/ml each proteinase K and RNAse A) and incubation overnight at 65°C with

1032 constant agitation to reverse crosslinking. The next day samples were purified using

1033 PCR cleanup kit columns (Monarch, NEB) and eluted using 50 μl kit elution buffer.

1034 Suitable dilutions of eluted DNA were used for qPCR analysis on a Lightcycler 480

1035 (Roche) using qPCRBIO SyGreen Blue Mix (PCRBiosystems). Enrichments were

1036 calculated as percentage of input. The following primers were used for qPCR analysis:

1037 Polm, tgacgggcacaattacacca, aaaggcttccgcgtcctaga; Gapdh, ggg ttc cta taa ata cgg

1038 act gc, ctg gca ctg cac aag aag atg; Pou5f1 (Oct4), ggc tct cca gag gat ggc tga g, tcg

1039 gat gcc cca tcg ca; Nanog, TGGCCTTCAGATAGGCTGAT,

1040 caagaagtcagaaggaagtgagc; HoxC5, gtactgctacggcggattgg, taccccgtggagagagttgg;

1041 Olig1, gggttacaggcagccaccta, atgcggtggaagaggatgag; Fgf5,

1042 GCGACGTTTTCTTCGTCTTC, ACGAAACCCTACCGGACTCT; HoxA7,

1043 GAGAGGTGGGCAAAGAGTGG, CCGACAACCTCATACCTATTCCTG; Fabp7,

1044 TGAGCAAATCACAAGGAGGA, TGGAGGAACTCGGGTCTTAC; Bcor promoter,

1045 gtaaaaccgaaagcgagcaa, GAGGGTTTCTCCTCCGACTT; Bcor gene body,

1046 GGGGGTAACTGTGGGAATCT, TCTCCACACCTTCAGCCTCT; Fzd1,

1047 ACATGAGCCCGTAAACCTTG, GGTGCCCTCCTACCTCAACT.

1048 The following antibodies were used for ChIP: H3K4me3 (Cell Signaling, 9751, lot

1049 10, 4 µg antibody for 40 µg chromatin), H3K27me3 (Cell Signaling, 9733, lot 14, 4 µg

1050 antibody for 40 µg chromatin), H3K23ac (abcam, AB177275, lot GR3213108-2, 2 µl

1051 antibody for 40 µg chromatin), H3K14ac (abcam, AB52946, lot GR3252548-2, 4 µl

1052 antibody for 40 µg chromatin), RING1B (Cell Signaling, 5694, lot 3, 4.5 µl antibody for

37 bioRxiv preprint doi: https://doi.org/10.1101/2021.02.08.430127; this version posted February 8, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

1053 50 µg chromatin), and HA tag (Cell Signaling, 3724, lot 9, 4 µl antibody for 200 µg

1054 chromatin).

1055

1056 Bioinformatic analysis of ChIP-seq data

1057 Obtained sequences were filtered for low-quality reads, and adapter sequences

1058 were removed using Trim Galore. Filtered reads were mapped to the mouse genome

1059 (version mm9) using the BOWTIE in QuasR (Gaidatzis et al., 2015) using default

1060 settings. Identical reads from PCR duplicates were filtered out from the obtained .bam

1061 files using Samtools rmdup. Wiggle tracks, for visualisation in the UCSC genome

1062 browser, were obtained using QuasR qExportWig() on the aligned reads. Heatmaps

1063 and average density-profiles 4 kb around RefSeq TSS were generated from .bam files

1064 using genomation in R using a 500 bin approach and strand-aware settings (Akalin et

1065 al., 2015). TSS were sorted into three categories based on the obtained binding

1066 profiles. Towards this, the k-means clustering option was applied using H3K4me3,

1067 H3K27me3, TAF3, and CBX7 profiles to calculate the clusters. The three clusters

1068 representing TSS in bivalent state, H3K4me3-only TSS, and TSS with low signal were

1069 then used to calculate average density around TSS.

1070

1071 Western Blot Analysis of ESC whole-cell extracts

1072 To generate ESC whole cell extract for Western Blot analysis, ESCs were washed

1073 with PBS and detached from culture plates by trypsinisation (Gibco). After addition of

1074 ESC culture media to stop trypsin digest, cells were collected, resuspended, and

1075 counted using a haemocytometer. Suitable amounts of ESCs were collected by

1076 centrifugation at 400 g for 5 min at room temperature and lysed by resuspending in a

1077 volume of 1X SDS sample buffer (63 mM Tris HCl pH 6.8, 10% glycerol, 2% SDS,

1078 50 mM DTT, 0.1% bromophenol blue) suitable for loading onto an SDS-PAGE gel. For

1079 histone modification analysis, the equivalent of 3x10^4 cells were loaded per lane. For

1080 chromatin factor analysis, 5–10x10^5 cells were used per lane for Western Blot

1081 analysis.

1082 Expression or modification status was analyzed by probing with antibodies against

1083 EZH2 (BD Biosciences, 612667, lot 415817), HA tag (Cell Signaling, clone C29F4,

1084 3724s, lot 9), KAT6B/MORF (abcam, ab191994, lot GR3180184), H3K4me3 (abcam,

38 bioRxiv preprint doi: https://doi.org/10.1101/2021.02.08.430127; this version posted February 8, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

1085 ab8580, lot GR273043-2), H3K23ac (abcam ab177275, lot GR3213108-2), and

1086 H3K27me3 (Cell Signaling, 9733, lot 8). Antibody against histone H4 (Cell Signaling,

1087 13919, lot 3) was used as a loading control. Blots were developed with Clarity ECL

1088 reagent (Bio-Rad) and imaged with a Chemidoc Touch imaging system (Bio-Rad).

1089 Western analysis was performed in duplicate.

1090

1091 Purification of endogenously FLAG-HA-tagged KAT6B from E14 ESCs

1092 About 1.625 x 10^9 ESCs expressing endogenously FLAG-HA-tagged KAT6B

1093 were grown as described above for E14 ESCs. NE was generated as described above,

1094 albeit using 0.1% NP40 in the TMSD lysis step. After solubilization in BC420,

1095 vortexing, and centrifugation as above, the resulting NE was adjusted to 300 mM KCl

1096 using BC buffer without KCl (BC0) by monitoring conductivity with a conductivity meter

1097 (Mettler Toledo). To prepare Anti-FLAG M2 affinity agarose beads (Sigma) for affinity

1098 purification, beads were incubated with 100 mM glycine, washed with 1M HEPES

1099 pH 7.9, and washed to equilibrate with BC300 buffer (see BC420 above but with

1100 300 mM KCl). The pretreated and washed anti-FLAG beads were combined with the

1101 NE and incubated for 2 h with constant rotation on a rotator mixer (Starlab) at 4°C in

1102 the cold room. After incubation, beads were collected by centrifugation at 800 g for

1103 2 min at 4°C and washed twice with BC300 buffer and then once with BC100 buffer

1104 (see BC420 above but with 100 mM KCl). Bound FLAG-HA-tagged KAT6B and

1105 associated proteins were eluted with 0.3 µg/µl 3x FLAG tag peptide (Sigma) in BC100

1106 buffer for 1 h with gentle rotation at 4°C.

1107

1108 Histone acetyltransferase assays

1109 Recombinant nucleosomes were reconstituted as described above using a plasmid

1110 containing 12 177-bp repeats containing the 147-bp 601 sequence. Nucleosomes and

1111 KAT6B enzyme preparations were combined in HAT buffer (50 mM Tris-Cl pH 7.5,

1112 10% glycerol, 0.1 mM EDTA) and 4 mM DTT. Reactions were started by the addition 3 1113 of H-acetyl coA (90 kBq; Hartmann) and incubated at 30°C for 2 h. Reactions were

1114 stopped addition of 3x SDS sample buffer (see above) to a final concentration of 1x

1115 by boiling for 5 min at 95°C with. Proteins were separated by SDS-PAGE, transferred

3 1116 to PVDF membrane, and stained with Coomassie to assess loading. Incorporated H-

39 bioRxiv preprint doi: https://doi.org/10.1101/2021.02.08.430127; this version posted February 8, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

1117 acetyl was detected using BioMax MS film (Kodak, Sigma) and BioMax Transcreen

1118 LE amplifying screens (Kodak, via Sigma).

1119

1120 LC-MS/MS analysis of HA-tagged KAT6B complex purified from E14 ESC

1121 4x NuPAGE LDS Sample buffer (Invitrogen) and 1 M DTT was added to purified

1122 endogenously FLAG-HA-tagged KAT6B complexes to a final concentration of 1x and

1123 50 mM, respectively. Samples were boiled at 95°C for 5 min before running on a

1124 NuPAGE 4–12% Bis-Tris protein gradient mini gel (Invitrogen) at 200 V for 5 min in

1125 MOPS running buffer (Invitrogen) to concentrate the sample. Gels were washed three

1126 times with deionized water for 5 min each before staining with InstantBlue (Expedeon)

1127 for 1 h. Stained gels were washed three times with deionized water for 5 min each to

1128 remove excess stain. For in-gel digestion, the stained band containing the protein

1129 sample was excised and de-stained by washing with 50 mM ammonium bicarbonate

1130 (Sigma) and 100% (v/v) acetonitrile (Sigma). Proteins were reduced in 50 mM

1131 ammonium bicarbonate with 10 mM DTT for 30 min at 37°C and alkylated in 55 mM

1132 iodoacetamide (Sigma) for 20 min at ambient temperature in the dark. Proteins were

1133 then digested overnight at 37°C with 12.5 ng/µl trypsin (Pierce). Samples were then

1134 analyzed by LC-MS/MS as described above.

1135

1136 Neuronal differentiation of ESC

1137 Neuronal differentiation was performed as previously described (Bibel et al., 2007).

1138 Embryoid body (EB) formation was induced by removal of LIF. On day 0, ESCs were

1139 plated onto non-adherent bacterial plates (Greiner) in 10 ml EB medium (ESC medium

1140 described above but without LIF and with FBS reduced to 10%) at a density of 4 x 10^6

1141 cells per 10-cm plate. On day 2, the media was changed by gently transferring EBs to

1142 a 15-ml conical tube (Sarstedt), allowing EBs to settle for 5 min, then carefully

1143 aspirating the supernatant. EBs were then gently resuspended in 13 ml of fresh EB

1144 media using a 5-ml serological pipette (Sarstedt) and transferred to a fresh plate. On

1145 day 4 and day 6, media was changed in the same manner to 15 ml EB medium with

1146 10 μM all-trans retinoic acid (Sigma). On day 8, EBs were washed 3 x in 20 ml PBS,

1147 trypsinised (Sigma, prepared at 0.5 mg/ml in 0.05% v/v EDTA/PBS) for approximately

1148 3 min at 37°C with agitation and then quenched in 10 ml EB medium. Dissociated

40 bioRxiv preprint doi: https://doi.org/10.1101/2021.02.08.430127; this version posted February 8, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

1149 neural progenitor cells were then centrifuged for 5 min at 300 g, resuspended in 10 ml

1150 EB medium and passed through a 40-μm cell strainer (Thermo Fisher Scientific). Cells

1151 were counted, centrifuged again and resuspended in Advanced DMEM/F12 (Gibco)

1152 with 1x N2 supplement (Gibco) before being plated onto poly-D-lysine (Sigma) and

1153 laminin (Sigma) coated 6-cm dishes at a density of 3 x 10^6 cells/plate. On day 9, half

1154 of the media was replaced with Neurobasal medium (Gibco) containing 1x B27

1155 supplement (Gibco). This was repeated on day 12.

1156

1157 RNA preparation and RT-qPCR

1158 Cells were resuspended in 1ml TriPure isolation reagent (Roche) per 10-cm plate

1159 of ESCs or EBs (or per 2x 6-cm plate for plated neural cells) and RNA extracted

1160 following the manufacturer’s instructions, resuspended in BTE (10 mM Bis-Tris

1161 pH 6.7, 1 mM EDTA), and treated with Turbo DNAse (Ambion) for 30 min at 37°C.

1162 1 ml TriPure was then added and RNA extraction repeated before resuspending the

1163 RNA pellet in 50 μl BTE. Integrity of RNA was assessed using RNA 6000 Nano chips

1164 on a Bioanalyzer (Agilent). cDNA was prepared from of 1 μg RNA per sample using

1165 SuperScript IV reverse transcriptase (Invitrogen). qRT-PCR experiments with the

1166 resulting cDNA samples were carried out on a Lightycler 480 (Roche) using qPCRBIO

1167 SyGreen Blue Mix (PCRBiosystems). Results were analysed using the 2-ΔΔCT

1168 method with Gapdh as the reference.

1169 Primers used for qPCR were: GAPDH, CATGGCCTTCCGTGTTCCT,

1170 GCGGCACGTCAGATCCA; Oct4, AGATCACTCACATCGCCAATCA,

1171 CGCCGGTTACAGAACCATACTC; Otx2, CATGATGTCTTATCTAAAGCAACCG,

1172 GTCGAGCTGTGCCCTAGTA; Nanog, AGGCTTTGGAGACAGTGAGGTG,

1173 TGGGTAAGGGTGTTCAAGCACT; Fgf5, AAACTCCATGCAAGTGCCAAAT,

1174 TCTCGGCCTGTCTTTTCAGTTC; Mash1, TAACTCCCAACCACTAACAGGC,

1175 TGAGGAAAGACATCAACCCAG; HoxB13, CATTCTGGAAAGCAGCGTTTG,

1176 TGTTGGCTGCATACTCCCG; Pax6, GATTGAGGCTCTGGAGAAAG,

1177 CCATTTGGCCCTTCGATTAG; Fabp7, TGTAAGTCTGTGGTTCGGTTG,

1178 AGCAACGATATCCCCAAAGG; Eed, CAGCCACCCTCTATTAGCAG,

1179 GCATTTCCATGGCCAACATAG; Polm, ATGGGCTGTTTGATCCTGAG,

1180 ACAGGCATTTCTCAGTTCAGG; Tbp, TGTATCTACCGTGAATCTTGGC,

41 bioRxiv preprint doi: https://doi.org/10.1101/2021.02.08.430127; this version posted February 8, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

1181 CCAGAACTGAAAATCAACGCAG; Polr2a, GGTCCTTCGAATCCGCATC,

1182 CAGGGTCATATCTGTCAGCATG; Sox2, GGGAGAAAGAAGAGGAGAGAGA,

1183 GCCGCGATTGTTGTGATTAG; Tbx3, AGGAGCGTGTCTGTCAGGTT,

1184 GCCATTACCTCCCCAATTTT; Ring1b, CGAACACCTCAGGAGGCAATA,

1185 ATACAATCCGCGCAAAACCG; Kmt2b, AAGTTCCGCTACCACCAGC,

1186 TGTCAAAGGTACACTTCCGGAGATA.

1187

1188 Immunofluorescence analysis

1189 Cells were grown on 16-mm glass coverslips in 12-well plates. Coverslips were

1190 coated with 0.1% gelatin (ESCs) or with poly-D-lysine and laminin (for neural cells).

1191 Cells were fixed by washing once in PBS and adding 4% paraformaldehyde (Sigma)

1192 for 20 min at room temperature. Coverslips were then washed 3 x 10 min in PBS then

1193 stored in PBS at 4°C. To carry out immunofluorescence analysis, fixed cells were

1194 washed once in PBS and blocked in 10% donkey serum (Sigma) in PBS with 0.1%

1195 Triton X-100 for 1 h and incubated overnight with primary antibodies at 4°C. The

1196 following day, cells were incubated with Alexa Fluor 488 or 555 secondary antibodies

1197 (Invitrogen) at a dilution of 1:1000 in 1% donkey serum in PBS with 0.1% Triton X-100

1198 for 1 h at room temperature in the dark. Coverslips were then washed 3x with PBS for

1199 10 min. Nuclei were counterstained with 50 nM DAPI (Roche) for 5 min, and washed

1200 2x 10 min with PBS before being mounted on slides using vectashield (Vector

1201 Laboratories). Imaging was carried out using a Zeiss Axio imager with a 40x objective.

1202 The microscope was equipped with a Hamamatsu Flash sCMOS camera. Micro-

1203 manager software (version 1.4) was used to capture images (Edelstein et al., 2014).

1204

1205

42 bioRxiv preprint doi: https://doi.org/10.1101/2021.02.08.430127; this version posted February 8, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

1206 References

1207

1208 Akalin, A., Franke, V., Vlahoviček, K., Mason, C.E., and Schübeler, D. (2015). 1209 Genomation: a toolkit to summarize, annotate and visualize genomic intervals. 1210 Bioinformatics 31, 1127-1129.

1211 Ali, M., Yan, K., Lalonde, M.-E., Degerny, C., Rothbart, S.B., Strahl, B.D., Côté, J., 1212 Yang, X.-J., and Kutateladze, T.G. (2012). Tandem PHD fingers of MORF/MOZ 1213 acetyltransferases display selectivity for acetylated histone H3 and are required for 1214 the association with chromatin. J Mol Biol 424, 328-338.

1215 Aranda, S., Mas, G., and Di Croce, L. (2015). Regulation of gene transcription by 1216 Polycomb proteins. Sci Adv 1, e1500737.

1217 Azuara, V., Perry, P., Sauer, S., Spivakov, M., Jørgensen, H.F., John, R.M., Gouti, 1218 M., Casanova, M., Warnes, G., Merkenschlager, M., et al. (2006). Chromatin 1219 signatures of pluripotent cell lines. Nat Cell Biol 8, 532-538.

1220 Bartke, T., Vermeulen, M., Xhemalce, B., Robson, S.C., Mann, M., and Kouzarides, 1221 T. (2010). Nucleosome-interacting proteins regulated by DNA and histone 1222 methylation. Cell 143, 470-484.

1223 Beringer, M., Pisano, P., Di Carlo, V., Blanco, E., Chammas, P., Vizán, P., Gutiérrez, 1224 A., Aranda, S., Payer, B., Wierer, M., et al. (2016). EPOP Functionally Links Elongin 1225 and Polycomb in Pluripotent Stem Cells. Mol Cell 64, 645-658.

1226 Bernstein, B.E., Mikkelsen, T.S., Xie, X., Kamal, M., Huebert, D.J., Cuff, J., Fry, B., 1227 Meissner, A., Wernig, M., Plath, K., et al. (2006). A Bivalent Chromatin Structure 1228 Marks Key Developmental Genes in Embryonic Stem Cells. Cell 125, 315-326.

1229 Bhuiyan, T., and Timmers, H.T.M. (2019). Promoter Recognition: Putting TFIID on 1230 the Spot. Trends Cell Biol 29, 752-763.

1231 Bibel, M., Richter, J., Lacroix, E., and Barde, Y.-A. (2007). Generation of a defined 1232 and uniform population of CNS progenitors and neurons from mouse embryonic 1233 stem cells. Nat Protoc 2, 1034-1043.

1234 Champagne, K.S., Saksouk, N., Peña, P.V., Johnson, K., Ullah, M., Yang, X.-J., 1235 Côté, J., and Kutateladze, T.G. (2008). The crystal structure of the ING5 PHD finger 1236 in complex with an H3K4me3 histone peptide. Proteins 72, 1371-1376.

1237 Christophersen, N.S., and Helin, K. (2010). Epigenetic control of embryonic stem cell 1238 fate. J Exp Med 207, 2287-2295.

1239 Clapier, C.R., Iwasa, J., Cairns, B.R., and Peterson, C.L. (2017). Mechanisms of 1240 action and regulation of ATP-dependent chromatin-remodelling complexes. Nat Rev 1241 Mol Cell Biol 18, 407-422.

43 bioRxiv preprint doi: https://doi.org/10.1101/2021.02.08.430127; this version posted February 8, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

1242 Conaway, R.C., Reines, D., Garrett, K.P., Powell, W., and Conaway, J.W. (1996). 1243 Purification of RNA polymerase II general transcription factors from rat liver. Meth 1244 Enzymol 273, 194-207.

1245 Cosentino, M.S., Oses, C., Vázquez Echegaray, C., Solari, C., Waisman, A., 1246 Álvarez, Y., Petrone, M.V., Francia, M., Schultz, M., Sevlever, G., et al. (2019). 1247 Kat6b Modulates Oct4 and Nanog Binding to Chromatin in Embryonic Stem Cells 1248 and Is Required for Efficient Neural Differentiation. J Mol Biol 431, 1148-1159.

1249 Cox, J., Hein, M.Y., Luber, C.A., Paron, I., Nagaraj, N., and Mann, M. (2014). 1250 Accurate proteome-wide label-free quantification by delayed normalization and 1251 maximal peptide ratio extraction, termed MaxLFQ. Mol Cell Proteomics 13, 2513- 1252 2526.

1253 Cox, J., and Mann, M. (2008). MaxQuant enables high peptide identification rates, 1254 individualized p.p.b.-range mass accuracies and proteome-wide protein 1255 quantification. Nat Biotechnol 26, 1367-1372.

1256 Cox, J., Neuhauser, N., Michalski, A., Scheltema, R.A., Olsen, J.V., and Mann, M. 1257 (2011). Andromeda: a peptide search engine integrated into the MaxQuant 1258 environment. J Proteome Res 10, 1794-1805.

1259 Creyghton, M.P., Markoulaki, S., Levine, S.S., Hanna, J., Lodato, M.A., Sha, K., 1260 Young, R.A., Jaenisch, R., and Boyer, L.A. (2008). H2AZ is enriched at polycomb 1261 complex target genes in ES cells and is necessary for lineage commitment. Cell 135, 1262 649-661.

1263 Denissov, S., Hofemeister, H., Marks, H., Kranz, A., Ciotta, G., Singh, S., 1264 Anastassiadis, K., Stunnenberg, H.G., and Stewart, A.F. (2014). Mll2 is required for 1265 H3K4 trimethylation on bivalent promoters in embryonic stem cells, whereas Mll1 is 1266 redundant. Development 141, 526-537.

1267 Eberl, H.C., Spruijt, C.G., Kelstrup, C.D., Vermeulen, M., and Mann, M. (2013). A 1268 Map of General and Specialized Chromatin Readers in Mouse Tissues Generated by 1269 Label-free Interaction Proteomics. Mol Cell 49, 368-378.

1270 Edelstein, A.D., Tsuchida, M.A., Amodaj, N., Pinkard, H., Vale, R.D., and Stuurman, 1271 N. (2014). Advanced methods of microscope control using μManager software. J 1272 Biol Methods 1, e10.

1273 Gaidatzis, D., Lerch, A., Hahne, F., and Stadler, M.B. (2015). QuasR: quantification 1274 and annotation of short reads in R. Bioinformatics 31, 1130-1132.

1275 Giaimo, B.D., Ferrante, F., Herchenröther, A., Hake, S.B., and Borggrefe, T. (2019). 1276 The histone variant H2A.Z in gene regulation. Epigenetics & Chromatin 12, 37-22.

1277 Grzybowski, A.T., Chen, Z., and Ruthenburg, A.J. (2015). Calibrating ChIP-Seq with 1278 Nucleosomal Internal Standards to Measure Histone Modification Density Genome 1279 Wide. Mol Cell 58, 886-899.

44 bioRxiv preprint doi: https://doi.org/10.1101/2021.02.08.430127; this version posted February 8, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

1280 Harikumar, A., and Meshorer, E. (2015). Chromatin remodeling and bivalent histone 1281 modifications in embryonic stem cells. EMBO Rep 16, 1609-1619.

1282 Hu, D., Garruss, A.S., Gao, X., Morgan, M.A., Cook, M., Smith, E.R., and Shilatifard, 1283 A. (2013a). The Mll2 branch of the COMPASS family regulates bivalent promoters in 1284 mouse embryonic stem cells. Nat Struct Mol Biol 20, 1093-1097.

1285 Hu, G., Cui, K., Northrup, D., Liu, C., Wang, C., Tang, Q., Ge, K., Levens, D., Crane- 1286 Robinson, C., and Zhao, K. (2013b). H2A.Z Facilitates Access of Active and 1287 Repressive Complexes to Chromatin in Embryonic Stem Cell Self-Renewal and 1288 Differentiation. Cell Stem Cell 12, 180-192.

1289 Huang, F., Abmayr, S.M., and Workman, J.L. (2016). Regulation of KAT6 1290 Acetyltransferases and Their Roles in Cell Cycle Progression, Stem Cell 1291 Maintenance, and Human Disease. Mol Cell Biol 36, 1900-1907.

1292 Hübner, J.-M., Müller, T., Papageorgiou, D.N., Mauermann, M., Krijgsveld, J., 1293 Russell, R.B., Ellison, D.W., Pfister, S.M., Pajtler, K.W., and Kool, M. (2019). EZHIP / 1294 CXorf67 mimics K27M mutated oncohistones and functions as an intrinsic inhibitor of 1295 PRC2 function in aggressive posterior fossa ependymoma. Neuro Oncol 21, 878- 1296 889.

1297 Ivarsson, Y., and Jemth, P. (2019). Affinity and specificity of motif-based protein- 1298 protein interactions. Curr Opin Struct Biol 54, 26-33.

1299 Jain, S.U., Do, T.J., Lund, P.J., Rashoff, A.Q., Diehl, K.L., Cieslik, M., Bajic, A., 1300 Juretic, N., Deshmukh, S., Venneti, S., et al. (2019). PFA ependymoma-associated 1301 protein EZHIP inhibits PRC2 activity through a H3 K27M-like mechanism. Nat 1302 Commun 10, 2146.

1303 Kang, H., Jung, Y.L., McElroy, K.A., Zee, B.M., Wallace, H.A., Woolnough, J.L., 1304 Park, P.J., and Kuroda, M.I. (2017). Bivalent complexes of PRC1 with orthologs of 1305 BRD4 and MOZ/MORF target developmental genes in Drosophila. Genes Dev 31, 1306 1988-2002.

1307 Kang, H., McElroy, K.A., Jung, Y.L., Alekseyenko, A.A., Zee, B.M., Park, P.J., and 1308 Kuroda, M.I. (2015). Sex comb on midleg (Scm) is a functional link between PcG- 1309 repressive complexes in Drosophila. Genes Dev 29, 1136-1150.

1310 Klein, B.J., Jang, S.M., Lachance, C., Mi, W., Lyu, J., Sakuraba, S., Krajewski, K., 1311 Wang, W.W., Sidoli, S., Liu, J., et al. (2019). Histone H3K23-specific acetylation by 1312 MORF is coupled to H3K14 acylation. Nat Commun 10, 4724-4713.

1313 Klein, B.J., Simithy, J., Wang, X., Ahn, J., Andrews, F.H., Zhang, Y., Côté, J., Shi, X., 1314 Garcia, B.A., and Kutateladze, T.G. (2017). Recognition of Histone H3K14 Acylation 1315 by MORF. Structure 25, 650-654.

45 bioRxiv preprint doi: https://doi.org/10.1101/2021.02.08.430127; this version posted February 8, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

1316 Klein, P., Pawson, T., and Tyers, M. (2003). Mathematical modeling suggests 1317 cooperative interactions between a disordered polyvalent ligand and a single 1318 receptor site. Curr Biol 13, 1669-1678.

1319 Ku, M., Jaffe, J.D., Koche, R.P., Rheinbay, E., Endoh, M., Koseki, H., Carr, S.A., and 1320 Bernstein, B.E. (2012). H2A.Z landscapes and dual modifications in pluripotent and 1321 multipotent stem cells underlie complex genome regulatory functions. Genome Biol 1322 13, R85.

1323 Lalonde, M.-E., Avvakumov, N., Glass, K.C., Joncas, F.-H., Saksouk, N., Holliday, 1324 M., Paquet, E., Yan, K., Tong, Q., Klein, B.J., et al. (2013). Exchange of associated 1325 factors directs a switch in HBO1 acetyltransferase histone tail specificity. Genes Dev 1326 27, 2009-2024.

1327 Laugesen, A., and Helin, K. (2014). Chromatin repressive complexes in stem cells, 1328 development, and cancer. Cell Stem Cell 14, 735-751.

1329 LeRoy, G., Orphanides, G., Lane, W.S., and Reinberg, D. (1998). Requirement of 1330 RSF and FACT for transcription of chromatin templates in vitro. Science 282, 1900- 1331 1904.

1332 Liefke, R., Karwacki-Neisius, V., and Shi, Y. (2016). EPOP Interacts with Elongin BC 1333 and USP7 to Modulate the Chromatin Landscape. Mol Cell 64, 659-672.

1334 Liefke, R., and Shi, Y. (2015). The PRC2-associated factor C17orf96 is a novel CpG 1335 island regulator in mouse ES cells. Cell Discov 1, 15008.

1336 Liu, Z., Scannell, D.R., Eisen, M.B., and Tjian, R. (2011). Control of embryonic stem 1337 cell lineage commitment by core promoter factor, TAF3. Cell 146, 720-731.

1338 Locasale, J.W. (2008). Allovalency revisited: an analysis of multisite phosphorylation 1339 and substrate rebinding. J Chem Phys 128, 115106.

1340 Lowary, P.T., and Widom, J. (1998). New DNA sequence rules for high affinity 1341 binding to histone octamer and sequence-directed nucleosome positioning. J Mol 1342 Biol 276, 19-42.

1343 Mas, G., Blanco, E., Ballaré, C., Sansó, M., Spill, Y.G., Hu, D., Aoi, Y., Le Dily, F., 1344 Shilatifard, A., Marti-Renom, M.A., et al. (2018). Promoter bivalency favors an open 1345 chromatin architecture in embryonic stem cells. Nature Genetics 8, 532.

1346 Mikkelsen, T.S., Ku, M., Jaffe, D.B., Issac, B., Lieberman, E., Giannoukos, G., 1347 Alvarez, P., Brockman, W., Kim, T.-K., Koche, R.P., et al. (2007). Genome-wide 1348 maps of chromatin state in pluripotent and lineage-committed cells. Nature 448, 553- 1349 560.

1350 Mittag, T., Orlicky, S., Choy, W.-Y., Tang, X., Lin, H., Sicheri, F., Kay, L.E., Tyers, 1351 M., and Forman-Kay, J.D. (2008). Dynamic equilibrium engagement of a polyvalent 1352 ligand with a single-site receptor. Proc Nat Acad Sci USA 105, 17772-17777.

46 bioRxiv preprint doi: https://doi.org/10.1101/2021.02.08.430127; this version posted February 8, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

1353 Mizuguchi, G., Shen, X., Landry, J., Wu, W.-H., Sen, S., and Wu, C. (2004). ATP- 1354 driven exchange of histone H2AZ variant catalyzed by SWR1 chromatin remodeling 1355 complex. Science 303, 343-348.

1356 Morey, L., Aloia, L., Cozzuto, L., Benitah, S.A., and Di Croce, L. (2013). RYBP and 1357 Cbx7 define specific biological functions of polycomb complexes in mouse 1358 embryonic stem cells. Cell Rep 3, 60-69.

1359 Musselman, C.A., Lalonde, M.-E., Côté, J., and Kutateladze, T.G. (2012). Perceiving 1360 the epigenetic landscape through histone readers. Nature Structural & Molecular 1361 Biology 19, 1218-1227.

1362 Olsen, J.G., Teilum, K., and Kragelund, B.B. (2017). Behaviour of intrinsically 1363 disordered proteins in protein-protein complexes with an emphasis on fuzziness. Cell 1364 Mol Life Sci 74, 3175-3183.

1365 Olsen, J.V., Macek, B., Lange, O., Makarov, A., Horning, S., and Mann, M. (2007). 1366 Higher-energy C-trap dissociation for peptide modification analysis. Nat Methods 4, 1367 709-712.

1368 Patel, A.B., Louder, R.K., Greber, B.J., Grünberg, S., Luo, J., Fang, J., Liu, Y., 1369 Ranish, J., Hahn, S., and Nogales, E. (2018). Structure of human TFIID and 1370 mechanism of TBP loading onto promoter DNA. Science (New York, NY) 362, 1371 eaau8872.

1372 Patel, D.J., and Wang, Z. (2013). Readout of epigenetic modifications. Annual 1373 Review of Biochemistry 82, 81-118.

1374 Piunti, A., and Shilatifard, A. (2016). Epigenetic balance of gene expression by 1375 Polycomb and COMPASS families. Science 352, aad9780.

1376 Qian, S., Lv, X., Scheid, R.N., Lu, L., Yang, Z., Chen, W., Liu, R., Boersma, M.D., 1377 Denu, J.M., Zhong, X., et al. (2018). Dual recognition of H3K4me3 and H3K27me3 1378 by a plant histone reader SHL. Nat Commun 9, 2425-2411.

1379 Qin, S., Jin, L., Zhang, J., Liu, L., Ji, P., Wu, M., Wu, J., and Shi, Y. (2011). 1380 Recognition of unmodified histone H3 by the first PHD finger of bromodomain-PHD 1381 finger protein 2 provides insights into the regulation of histone acetyltransferases 1382 monocytic leukemic zinc-finger protein (MOZ) and MOZ-related factor (MORF). J 1383 Biol Chem 286, 36944-36955.

1384 Qiu, Y., Liu, L., Zhao, C., Han, C., Li, F., Zhang, J., Wang, Y., Li, G., Mei, Y., Wu, M., 1385 et al. (2012). Combinatorial readout of unmodified H3R2 and acetylated H3K14 by 1386 the tandem PHD finger of MOZ reveals a regulatory mechanism for HOXA9 1387 transcription. Genes Dev 26, 1376-1391.

1388 Ragazzini, R., Pérez-Palacios, R., Baymaz, I.H., Diop, S., Ancelin, K., Zielinski, D., 1389 Michaud, A., Givelet, M., Borsos, M., Aflaki, S., et al. (2019). EZHIP constrains 1390 Polycomb Repressive Complex 2 activity in germ cells. Nat Commun 10, 3858-3818.

47 bioRxiv preprint doi: https://doi.org/10.1101/2021.02.08.430127; this version posted February 8, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

1391 Ran, F.A., Hsu, P.D., Wright, J., Agarwala, V., Scott, D.A., and Zhang, F. (2013). 1392 Genome engineering using the CRISPR-Cas9 system. Nat Protoc 8, 2281-2308.

1393 Rappsilber, J., Ishihama, Y., and Mann, M. (2003). Stop and go extraction tips for 1394 matrix-assisted laser desorption/ionization, nanoelectrospray, and LC/MS sample 1395 pretreatment in proteomics. Anal Chem 75, 663-670.

1396 Ruthenburg, A.J., Li, H., Patel, D.J., and Allis, C.D. (2007). Multivalent engagement 1397 of chromatin modifications by linked binding modules. Nat Rev Mol Cell Biol 8, 983- 1398 994.

1399 Schuettengruber, B., Bourbon, H.-M., Di Croce, L., and Cavalli, G. (2017). Genome 1400 Regulation by Polycomb and Trithorax: 70 Years and Counting. Cell 171, 34-57.

1401 Shema, E., Jones, D., Shoresh, N., Donohue, L., Ram, O., and Bernstein, B.E. 1402 (2016). Single-molecule decoding of combinatorially modified nucleosomes. Science 1403 352, 717-721.

1404 Simó-Riudalbas, L., Pérez-Salvia, M., Setien, F., Villanueva, A., Moutinho, C., 1405 Martínez-Cardús, A., Moran, S., Berdasco, M., Gómez, A., Vidal, E., et al. (2015). 1406 KAT6B Is a Tumor Suppressor Histone H3 Lysine 23 Acetyltransferase Undergoing 1407 Genomic Loss in Small Cell Lung Cancer. Cancer Res 75, 3936-3945.

1408 Simon, Jeffrey A., and Kingston, Robert E. (2013). Occupying Chromatin: Polycomb 1409 Mechanisms for Getting to Genomic Targets, Stopping Transcriptional Traffic, and 1410 Staying Put. Mol Cell 49, 808-824.

1411 Strübbe, G., Popp, C., Schmidt, A., Pauli, A., Ringrose, L., Beisel, C., and Paro, R. 1412 (2011). Polycomb purification by in vivo biotinylation tagging reveals cohesin and 1413 Trithorax group proteins as interaction partners. Proc Nat Acad Sci USA 108, 5572- 1414 5577.

1415 Su, Z., and Denu, J.M. (2016). Reading the Combinatorial Histone Language. ACS 1416 Chem Biol 11, 564-574.

1417 Taverna, S.D., Li, H., Ruthenburg, A.J., Allis, C.D., and Patel, D.J. (2007). How 1418 chromatin-binding modules interpret histone modifications: lessons from professional 1419 pocket pickers. Nat Struct Mol Biol 14, 1025-1040.

1420 Vastenhouw, N.L., and Schier, A.F. (2012). Bivalent histone modifications in early 1421 embryogenesis. Curr Opin Cell Biol 24, 374-386.

1422 Vermeulen, M., Eberl, H.C., Matarese, F., Marks, H., Denissov, S., Butter, F., Lee, 1423 K.K., Olsen, J.V., Hyman, A.A., Stunnenberg, H.G., et al. (2010). Quantitative 1424 Interaction Proteomics and Genome-wide Profiling of Epigenetic Histone Marks and 1425 Their Readers. Cell 142, 967-980.

48 bioRxiv preprint doi: https://doi.org/10.1101/2021.02.08.430127; this version posted February 8, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

1426 Voigt, P., Leroy, G., Iii, William J.D., Zee, Barry M., Son, J., Beck, David B., Young, 1427 Nicolas L., Garcia, Benjamin A., and Reinberg, D. (2012). Asymmetrically Modified 1428 Nucleosomes. Cell 151, 181-193.

1429 Voigt, P., Tee, W.-W., and Reinberg, D. (2013). A double take on bivalent promoters. 1430 Genes Dev 27, 1318-1338.

1431 Yang, X.-J. (2015). MOZ and MORF acetyltransferases: Molecular interaction, 1432 animal development and human disease. Biochim Biophys Acta 1853, 1818-1826.

1433 Yang, Z., Qian, S., Scheid, R.N., Lu, L., Chen, X., Liu, R., Du, X., Lv, X., Boersma, 1434 M.D., Scalf, M., et al. (2018). EBS is a bivalent histone reader that regulates floral 1435 phase transition in Arabidopsis. Nature Genetics 50, 1247-1253.

1436 Yu, J.-R., Lee, C.-H., Oksuz, O., Stafford, J.M., and Reinberg, D. (2019). PRC2 is 1437 high maintenance. Genes Dev 33, 903-935.

1438 Zhang, X., Smits, A.H., van Tilburg, G.B.A., Jansen, P.W.T.C., Makowski, M.M., 1439 Ovaa, H., and Vermeulen, M. (2017). An Interaction Landscape of Ubiquitin 1440 Signaling. Mol Cell 65, 941-955.

1441 Zhang, X., Smits, A.H., van Tilburg, G.B.A., Ovaa, H., Huber, W., and Vermeulen, M. 1442 (2018). Proteome-wide identification of ubiquitin interactions using UbIA-MS. Nat 1443 Protoc 13, 530-550.

1444

49 bioRxiv preprint doi: https://doi.org/10.1101/2021.02.08.430127; this version posted February 8, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. A unmodified asymmetric symmetric B Cxxc1 Ezh2 Phc1

2

0 recombinant nucleosomes −2 + Phf2 G9a Orc2

2 pulldowns

nucleosome ESC NE bead-bound nucleosomes

(mean intensity, ±95% CI) 0 2 unmodified asymmetric symmetric log −2 LFQ intensity

LC-MS/MS m/z m/z m/z

unmodifiedH3K4me3/3H3K4me0/3 unmodifiedH3K4me3/3H3K4me0/3 unmodifiedH3K4me3/3H3K4me0/3 H3K27me3/3H3K27me0/3 H3K27me3/3H3K27me0/3 H3K27me3/3H3K27me0/3 data analysis MaxQuant/DEP/R asymmetric unmod. asymmetric unmod. asymmetric unmod. C 12 12

Spin1 10 10 Hmgxb4 C10orf12/Pali1 Ing2 Bap18 Hmg20b 8 Cxxc1 8 Cxxc1 Phf21a EPOP Ezh1 Chd8 Bptf Phf21a Chd1 6 Ezh2 Chd1 Phf2 6 Hmg20b Dpy30

( p value) Eed Bap18 G9a Dpy30 10 GLP Suz12 Bptf 4 Wiz Taf3 4 C10orf12/ Zranb2 Spin1 Ezh1 Pali1 Taf3 −log EPOP Dnajc9 Eed Ing2 Phf2 Chd8 2 Suz12 2 Wiz Ezh2 Rnf2 Hmgxb4 0 0 GLP Rnf2 G9a −8 −6 −4 −2 0 2 4 6 8 −8 −6 −4 −2 0 2 4 6 8

log2FC(H3K4me3/3 vs. unmodified) log2FC(H3K4me0/3 vs. unmodified) D 12 12 Phc1

10 Cbx7 10

8 8

6 Orc2 6 Rnf2 ( p value) Cbx7 10 LRWD1 Ezh1 4 Aebp2 Ezh1 4 Suz12 Rnf2 Orc3 Ezh2 Phc1 Orc2

−log C10orf12/Pali1 EPOP Eed C10orf12/Pali1 LRWD1 G2e3 Arid4b Suz12 G9a 2 2

G9a Eed 0 0 Ezh2 −8 −6 −4 −2 0 2 4 6 8 −8 −6 −4 −2 0 2 4 6 8

log2FC(H3K27me3/3 vs. unmodified) log2FC(H3K27me0/3 vs. unmodified) Bryan et al. Figure 1 bioRxiv preprint doi: https://doi.org/10.1101/2021.02.08.430127; this version posted February 8, 2021. The copyright holder for this A preprint (which was not certified by peer review) is the author/funder. All rightsB reserved. No reuse allowed without permission.

mix H3K4me3/3 H3K4me0/392 me3/tailless TAF3 + --- 23 1854692 23 185469292 23 1854692 pmol nucleosomes TFIID

TAF3

TAF3 PHF2 TFIID binders H3K4me3 ING2

C K app. K app. EZH2 D D sym. nuc. asym. nuc. depleted H3K4me3

H3K4me3

loading H4 nucleosome bound fraction D

concentration [binder]

mix TAF3 H3K27me3/3 H3K27me0/392 me3/tailless + TFIID EED --- 23 1854692 23 185469292 23 1854692 pmol nucleosomes SAGA PRC2 CBX EZH2 SET1A/B CHD1 PRC1

CBX7 H3K4me3 binders H3K27me3 binders binders

H3K27me3 RING1B molecules/cell: molecules/cell: TAF3 ~2,200 RING1B ~560,000 PHF2 ~200 CBX7 ~39,000 H3K27me3 CHD8 ~1,200 PHC1 ~14,000 CxxC1 ~6,400 EZH2 ~80,000 BPTF ~1,700 EZH1 ~10,000 loading H4 ING2 ~12,800 EED ~475,000 nucleosome

Bryan et al. Figure 2 bioRxiv preprint doi: https://doi.org/10.1101/2021.02.08.430127; this version posted February 8, 2021. The copyright holder for this A preprint (which was not certified by peer review) is the author/funder.B All rights reserved. No reuse allowed without permission. 12 TAF3 CBX7 4 4 2 2 10 0 0 −2 −2 8 −4 −4 EPOP KAT6B SRCAP Phc1 2 2 6 Orc2 (p value) 1 1

10 Ezh2 Chd1 Cbx7 Phf13 0 0

−log Taf3 Bptf1 Ezhip

4 (mean intensity, ±95% CI)

Chd8 Rnf2 2 −1 −1 Dpy30 LRWD1 Ezh1 log −2 −2 Cxxc1 Eed Suz12 2 Phf2 C10orf12/Pali1 Spin1 Bap18 bivalent bivalent unmodified unmodified 0 Ing2 unmodifiedH3K4me3/3 unmodifiedH3K4me0/3 H3K4me3/3 H3K4me0/3 Phf21a H3K27me3/3 H3K27me0/3 H3K27me3/3 H3K27me0/3 −8 −6 −4 −2 0 2 4 6 8 symmetric asymmetric symmetric asymmetric

log2FC(bivalent vs. unmodified)

C nucleosomes D 12 Actr5 Zfp296 Nfrkb Uchl5 10 Tfpt --- Actr8 Srcap unmodifiedH3K4me3/3H3K27me3/3bivalent Znf652 Srsf2 Mcrs1 Zbtb2 Yeats4 Pcca Zranb2 TAF3 Yy1 8 Acaca Taf10 Pum1 Tfe3 Dmap1 Taf4a Vps72 binders 6 Kat6b H3K4me3 ING2 Ino80b

(p value) Zbtb10 Ino80c Hmg20a 10 Ino80e Mitf Zfp219 Zbtb14 Meaf6 −log 4 Zfp263 Zfp91 Brpf1 EZH2 Zfp655 Znf787 Zfp568 Zfp600 Ing5 Zfp64 Zbtb48 Brpf3 2

binders CBX7 Taf6 Taf1d Kat6a

H3K27me3 Taf1a Taf1b Taf7 Taf5 Taf1 Taf8 0 Taf1c Taf4b HA-KAT6B −8 −6 −4 −2 0 2 4 6 8

log2FC(bivalent vs. unmodified)

H3K4me3 E TAF3 TFIID EED H3K27me3 SAGA PRC2 CBX loading

nucleosome SET1A/B PRC1 H4 CHD1

H3K4me3 binders H3K27me3 binders

novel binders SRCAP KAT6B bivalent binders

Bryan et al. Figure 3 bioRxiv preprint doi: https://doi.org/10.1101/2021.02.08.430127; this version posted February 8, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

A 50 kb 38 _ H3K4me3

20 _ TAF3

35 _ H3K27me3

61 _ CBX7

CpG islands Evx2 Hoxd11 Hoxd8 Lnp AK144266 Hoxd13 Mtx2 Hoxd10 Hoxd1 genes Hoxd4 Hoxd12 Hoxd9 BC060302 Mir10b 1700109F18Rik Hoxd3

B H3K4me3 H3K27me3 TAF3 CBX7 C H3K4me3 H3K27me3 40 15

30 high 10 H3K4me3 20 5 coverage 10

bivalent 0 0 0 100 300 500 0 100 300 500 TAF3 CBX7 20 20

15 15 low

H3K4me3 10 10

coverage 5 5

0 0 0 100 300 500 0 100 300 500 −4 −2 0 +2+4 −4 −2 0 +2+4 −4 −2 0 +2+4 −4 −2 0 +2 +4 index index distance (kb)

0 10 30 50 0 5 10 15 20 0 2 4 6 10 14 0 2 4 6 8 12

Bryan et al. Figure 4 bioRxiv preprint doi: https://doi.org/10.1101/2021.02.08.430127; this version posted February 8, 2021. The copyright holder for this A preprint (which was not certified by peer review) is the author/funder.B All rights reserved. No reuse allowed without permission. 0.3 WT KAT6B MLL2 Y2688A Ezh1/2 dKO H15 PHD PHD MYST ED SM 0.2 * * * *

H3K14ac H3K4me3 0.1 Eaf6

H3K27me3 HA ChIP (% of input) 0.0 H3K4me0

Polm Oct4 Fgf5 Olig1 Fzd1 Gapdh Nanog Hoxc5 Hoxa7 BCoR Fabp7 BCoR GB ING5 H3K4me3 PHD active bivalent inactive 30 ZnF PHD PHD Bromo PWWP BRPF1 20 * * * * *

H3K14ac H3K36me3 H2AK5ac H4K12ac 10 unique C protein uniprot MW (kDa) peptides peptides 0 H3K4me3 ChIP (% of input) KAT6B Q8BRB7 208.52 54/53/45 54/4/44 Polm Oct4 Fgf5 Olig1 Fzd1 Gapdh Nanog Hoxc5 Hoxa7 BCoR Fabp7 A0A0N4SVG7 137.99 23/29/– 1/1/– BCoR GB BRPF1 B2RRD7 137.31 23/29/24 1/1/2 active bivalent inactive BRPF2 Q3UE68 119.71 16/25/25 15/25/24

15 BRPF3 B2KF05 135.29 6/7/5 5/6/4 ING5 Q9D8Y8 27.798 4/5/4 4/5/4 EAF6 V9GX42 20.518 2/3/4 2/3/4 10 EED Q921E6-3 48.765 –/3/3 –/3/3 EZH2, SUZ12, RNF2, CBX7 –/–/– –/–/– 5

D nucleosomes 0 H3K27me3 ChIP (% of input)

Polm Oct4 Fgf5 Olig1 Fzd1 Gapdh Nanog Hoxc5 Hoxa7 BCoR Fabp7 BCoR GB unmodifiedH3K4me3/3H3K27me3/3bivalent active bivalent inactive H3

3H fluorography

H3

coomassie

Bryan et al. Figure 5 bioRxiv preprint doi: https://doi.org/10.1101/2021.02.08.430127; this version posted February 8, 2021. The copyright holder for this A preprint20 (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. E14 WT KAT6B KO 15 KAT6A KO

10 KAT6A/B dKO

5

H3K23ac ChIP (% of input) 0

Polm Oct4 Fgf5 Fzd1 Gapdh Nanog Hoxc5 Hoxa7 Olig1 BCoR Fabp7 BCoR GB active bivalent inactive

B mESCs C E14 WT KAT6A KO KAT6B KO KAT6A/B dKO day 0 ESCs embryoid bodies day 2

DAPI / Oct4 day 4

day 15 differentiation

low serum medium day 6 neural progenitors retinoic acid

day 8 DAPI / Tuj1

day 9 mature neurons

day 12

brightfield day 15 neurobasal serum-free medium D E14 WT KAT6B KO KAT6A KO KAT6A/B dKO

Oct4 Otx2 Pax6 Hoxb13 1.25 6 20 160 1.0 5 4 15 120 0.75 3 10 80 0.5 2 5 40 fold change 0.25 1 0 0 0 0 d0 d2 d6 d15 d0 d2 d6 d15 d0 d2 d6 d15 d0 d2 d6 d15

Nanog Fgf5 Mash1 Fabp7 15 1000 1.00 6000 750 0.75 10 4000 0.5 500 5 2000 0.25 250 fold change

0 0 0 0 d0 d2 d6 d15 d0 d2 d6 d15 d0 d2 d6 d15 d0 d2 d6 d15 differentiation (d) differentiation (d) differentiation (d) differentiation (d) Bryan et al. Figure 6 bioRxiv preprint doi: https://doi.org/10.1101/2021.02.08.430127; this version posted February 8, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. A 150 40

Cys 30 1 M NaCl (%) + R-S 100 H3(T32C)K4me3 H3 1-31T32C H3 1-31)-SR Δ ( (mAU) 20 280 Cys + UV 50 R-S 10 H3(T32C)K27me3

0 0 NCL reaction 9 13 17 21 25 volume (ml) pre post purified

H3(T32C)K4me3 H3(T32C)K4me3 15 kDa 15 kDa H3Δ1-31T32C H3Δ1-31T32C

200 400 mM NaCl B C 12

10

unmodifiedasymmetricH3K4me3/3 unmod.H3K27me3/3 H3K4me0/3H3K27me0/3bivalent 8 H3K4me3 Bptf 6 H3K27me3 ( p value)

10 Hmg20b Taf3 Phf21a 4

H4 −log C10orf12/Pali1 Hmgxb4 Ing2 Aebp2 Cxxc1 20 kDa 2 G9a H3-His/Strep/untagged Ezhip Ezh1Wiz Phf2 H2B/H2A GLP Phf13 Chd1 15 kDa H4 Bap18 Eed Chd8 coomassie 0 Suz12 EPOP Ezh2 Spin1 Dpy30 −8 −6 −4 −2 0 2 4 6 8

log2FC(control vs. asymmetric control) D

601 DNAunmodifiedasymmetricH3K4me3/3 H3K27me3/3unmod.H3K4me0/3 H3K27me0/3bivalent

1 kb mononucleosome

500 bp

200 bp free 601 DNA

Bryan et al. Figure S1 bioRxiv preprint doi: https://doi.org/10.1101/2021.02.08.430127; this version posted February 8, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

A

601 DNAunmodifiedH3K4me3/3H3K4me0/3H3K4me3/tailless

mononucleosome 1 kb

500 bp

200 bp free 601 DNA

B

601 DNAunmodifiedH3K27me3/3H3K27me0/3H3K27me3/tailless

mononucleosome 1 kb

500 bp

200 bp free 601 DNA

Bryan et al. Figure S2 bioRxiv preprint doi: https://doi.org/10.1101/2021.02.08.430127; this version posted February 8, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. A EZH2 RNF2 BRPF1 INO80

1.0 1.0 2 1 0.5 0.5 1 0 0.0 0.0 0

−0.5 −0.5 −1 −1 −2 −1.0 −1.0 −3

(mean intensity, ±95% CI) bivalent bivalent bivalent bivalent 2 unmodified unmodified unmodified unmodified unmodified unmodified H3K4me3/3 H3K4me0/3 H3K4me3/3 H3K4me0/3 H3K4me3/3 H3K4me0/3 unmodifiedH3K4me3/3 unmodifiedH3K4me0/3 H3K27me3/3 H3K27me0/3 H3K27me3/3 H3K27me0/3 H3K27me3/3 H3K27me0/3 H3K27me3/3 H3K27me0/3 log symmetric asymmetric symmetric asymmetric symmetric asymmetric symmetric asymmetric

EPOP PHC1 YEATS4 NFRKB 1.5 3 1.0 2 2.5 1 1 0.5 0 0 0.0 0.0 −1 −1 −0.5 −2.5 −2 −2 −5.0

(mean intensity, ±95% CI) bivalent bivalent bivalent bivalent

2 unmodified unmodified unmodified unmodified H3K4me3/3 H3K4me0/3 unmodifiedH3K4me3/3 unmodifiedH3K4me0/3 unmodifiedH3K4me3/3 unmodifiedH3K4me0/3 H3K4me3/3 H3K4me0/3 H3K27me3/3 H3K27me0/3 H3K27me3/3 H3K27me0/3 H3K27me3/3 H3K27me0/3 H3K27me3/3 H3K27me0/3 log symmetric asymmetric symmetric asymmetric symmetric asymmetric symmetric asymmetric

JARID2 ORC2 ZRANB2 ZBTB10 5.0 2 1.0 2.5 0.5 1 2

0.0 0 0.0 0

−0.5 −1 −2.5 −2 −1.0

(mean intensity, ±95% CI) bivalent bivalent bivalent bivalent 2 unmodified unmodified unmodified unmodified H3K4me3/3 H3K4me0/3 unmodifiedH3K4me3/3 unmodifiedH3K4me0/3 unmodifiedH3K4me3/3 unmodifiedH3K4me0/3 H3K4me3/3 H3K4me0/3 H3K27me3/3 H3K27me0/3 H3K27me3/3 H3K27me0/3 H3K27me3/3 H3K27me0/3 H3K27me3/3 H3K27me0/3 log symmetric asymmetric symmetric asymmetric symmetric asymmetric symmetric asymmetric

B Zbtb2 12 Zbtb10 Zbtb14 Zfp64 Zbtb48 10 Zfp568 Actr8 Tfpt Zfp600 Zfp91 Znf787 Ino80c Zfp263 8 Actr5 Znf652 Zfp296 Taf4a Zfp655 Taf7 Zfp219 6 Srcap Taf1 ( p value) Yeats4 Uchl5 Taf6 10 Taf10 Brpf1 Taf8 Kat6b Nfrkb Taf1c −log 4 Taf1d Acaca Taf1b Hmg20a Tfe3 Mcrs1 Taf1a Pcca Yy1 Taf5 2 Ing5 Mitf Ino80b Taf4b Vps72 Ino80e Dmap1 Brpf3 Zranb2 0 Kat6a Srsf2 Meaf6 Pum1 -8 -6 -4 -20 2 4 6 8

log2FC(control vs. asymmetric control)

Bryan et al. Figure S3 bioRxiv preprint doi: https://doi.org/10.1101/2021.02.08.430127; this version posted February 8, 2021. The copyright holder for this A preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. CRISPR design FLAG-HA-KAT6B ssODN donor 5’ UTR * protein coding region Kat6b locus exon 2

3 x FLAG 1 x HA C -/-

; EZH1 GGATTCGACTACAAGGACCACGACGGCGATTATAAGGATCACGACATCGACTACAAAGACGACGATGACAAGGGCTCCGGATACCCATACGATGTGCCAGATTACGCTGGCTTC Y2688A/Y2688A-/- linker 3xFLAG linker 1xHA linker

5’ UTR 3 x FLAG 1 x HA ** protein coding region E14 HA-KAT6BHA-KAT6B;HA-KAT6B; MLL2 EZH2 exon 2 edited Kat6b locus BamHI BsrGI HA-KAT6B edited sequence AAT CCG TTG TAC ACA GAG TGG ATT C AAC CCA CTT TAC ACA GAG TGG ATT C original sequence Asn Pro Leu Tyr Thr Glu Trp Ile PAM guide EZH2

ssODN donor H3K4me3 B H3K27me3 CRISPR design MLL2 Y2688A protein coding region * 3’ UTR Kmt2b locus exon 37 H4

NdeI

edited sequence CTG ACA TAT GAC GCC AAG TTT CCC ATT GAA GAT Leu Thr Tyr Asp Ala Lys Phe Pro Ile Glu Asp original sequence Leu Thr Tyr Asp Tyr Lys Phe Pro Ile Glu Asp CTC ACC TAT GAC TAC AAG TTT CCT ATC GAG GAT guide PAM D 4 ssODN donor WT MLL2 Y2688A CRISPR design EZH2 KO 3 Ezh1/2 dKO * protein coding region Ezh2 locus exon 6 2

HindIII

AT TAA TAA GCT TGA TCA CCT CGG AAA TTT CCT GCT G 1 edited sequence * * Ala * Ser Pro Arg Lys Phe Pro Ala Ring1B ChIP (% of input) Lys Glu Thr Cys Pro Pro Arg Lys Phe Pro Ala original sequence AT AAA GAA ACT TGC CCA CCT CGG AAA TTT CCT GCT G 0

PAM guide Polm Oct4 Fgf5 Olig1 Fzd1 Gapdh Nanog Hoxc5 Hoxa7 BCoR Fabp7 ssODN donor BCoR GB active bivalent inactive CRISPR design EZH1 KO * protein coding region Ezh1 locus exon 8

AAC TAA TAG GGA TCC TAG TAA CAG TTT CCA AAC edited sequence Asn * * Gly Ser * * Gln Phe Pro Asn Asn Lys Lys Ser Ser Lys Lys Gln Phe Pro Asn original sequence AAC AAA AAG AGT TCC AAG AAA CAG TTT CCA AAT

PAM guide

ssODN donor

Bryan et al. Figure S4 bioRxiv preprint doi: https://doi.org/10.1101/2021.02.08.430127; this version posted February 8, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

A C E14 WT CRISPR design KAT6B KO 12 KAT6B KO 5’ UTR * protein coding region Kat6b locus exon 2 KAT6A KO 9 KAT6A/B dKO

BglII

TCC TAA TGA TAG CCA TGC TAA GAT CTA CGC AAT 6 edited sequence Ser * * * Pro Cys * Asp Leu Arg Asn Ser Lys Gly Pro Pro Cys Asn Asp Leu Arg Asn original sequence TCC AAA GGA CCG CCG TGT AAT GAT CTA CGC AAT 3

PAM guide H3K14ac ChIP (% of input) 0 ssODN donor Polm Oct4 Fgf5 Fzd1 Gapdh Nanog Hoxc5 Hoxa7 Olig1 BCoR Fabp7 CRISPR design KAT6A KO BCoR GB 5’ UTR * protein coding region active bivalent inactive Kat6a locus exon 3 30

edited sequence CAT TGA TAA TAG GAT ACT TAG TGA AGT GTC GAT TGG His * * * Asp Thr * * Ser Val Asp Trp 20 His Gly Lys Leu Asp Thr Lys Gln Ser Val Asp Trp original sequence CAT GGA AAA TTG GAT ACT AAG CAA AGT GTG GAT TGG 10 guide PAM

ssODN donor H3K4me3 ChIP (% of input) 0

-/- B Polm Oct4 Fgf5 Fzd1 Gapdh Nanog Hoxc5 Hoxa7 Olig1 BCoR Fabp7 BCoR GB KAT6B -/- -/- -/-; active bivalent inactive

E14 KAT6BKAT6AKAT6A 250 kDa 25 KAT6B 20

15 EZH2 10

5 IN PROGRESS H3K23ac H3K27me3 ChIP (% of input) 0

Polm Oct4 Fgf5 Fzd1 Gapdh Nanog Hoxc5 Hoxa7 Olig1 BCoR Fabp7 H4 BCoR GB active bivalent inactive D 0.100 E14 WT KAT6A KO 0.075 KAT6B KO KAT6A/B dKO

0.050

0.025

0.000 expression (norm. to GAPDH) Polm TBP Polr2a Sox2 Tbx3 Eed Ring1B Kmt2B

Bryan et al. Figure S5