1 Title: Functional Correlation of H3k9me2 and Nuclear Compartment Formation 1 Kei Fukuda , Chikako Shimura , Hisashi Miura , Ak

Home , H3K9me2

bioRxiv preprint doi: https://doi.org/10.1101/2020.08.28.271221; this version posted August 28, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder.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 Title: Functional correlation of H3K9me2 and nuclear compartment formation

2 Kei Fukuda1, Chikako Shimura1, Hisashi Miura2, Akie Tanigawa2, Takehiro Suzuki3, Naoshi Dohmae3,

3 Ichiro Hiratani2 and Yoichi Shinkai1

5 1 Cellular Memory Laboratory, RIKEN Cluster for Pioneering Research, Wako 351-0198, Japan

6 2 Laboratory for Developmental Epigenetics, RIKEN Center for Biosystems Dynamics Research,

7 Kobe 650-0047, Japan

8 3 Biomolecular Characterization Unit, Technology Platform Division, RIKEN Center for Sustainable

9 Resource Science, Wako, 351-0198, Japan

10 *Correspondence: [email protected] (Y.S.), [email protected] (K.F.)

12 Abstract

13 Background: Histone H3 lysine 9 dimethylation (H3K9me2) is a highly conserved silencing

14 epigenetic mark. Chromatin marked with H3K9me2 forms large domains in mammalian cells and

15 correlates well with lamina-associated domains and the B compartment. However, the role of

16 H3K9me2 in 3-dimensional (3D) genome organization remains unclear.

17 Results: We investigated the genome-wide H3K9me2 distribution, the transcriptome and 3D

18 genome organization in mouse embryonic stem cells (mESCs) upon the inhibition or depletion of

19 H3K9 methyltransferases (MTases) G9a/GLP, SETDB1, and SUV39H1/2. We found that H3K9me2 is

20 regulated by these five MTases; however, H3K9me2 and transcription in the A and B

21 compartments were largely regulated by different sets of the MTases: H3K9me2 in the A

22 compartments were mainly regulated by G9a/GLP and SETDB1, while H3K9me2 in the B

23 compartments were regulated by all five H3K9 MTases. Furthermore, decreased H3K9me2

24 correlated with the changes to the more active compartmental state that accompanied

25 transcriptional activation.

26 Conclusion: Our data showed that H3K9me2 domain formation is functionally linked to 3D

27 genome organization.

29 Keywords

30 H3K9 methylation, A/B compartment, Histone H3K9 methyltransferases

32 Background

33 Post-translational modifications of histone proteins regulate chromatin compaction and mediate

34 epigenetic transcriptional regulation. The methylation of histone tails is one of the fundamental

35 events of epigenetic signaling. In organisms ranging from the fission yeast Schizosaccharomyces

36 pombe to humans, repeat-rich constitutive heterochromatin is marked by H3K9me2 or H3K9me3

37 [1-3]. These modifications are catalyzed by a family of SET domain-containing lysine

38 methyltransferases, of which five are present in mammals. SETDB1 and the related enzymes,

39 SUV39H1 and SUV39H2, contribute to the formation of H3K9me3 [1, 4], while GLP and G9a (also

40 called EHMT1 and EHMT2, respectively) regulate H3K9me1 and H3K9me2 formation, respectively

41 [5-7]. In mESCs, H3K9me3 is enriched in retroelements and pericentromeric satellite repeats that

42 are mediated by SETDB1 and SUV39H1/2, respectively [8-10]. SUV39Hs are also involved in

43 H3K9me3 modification in intact retroelements and LINE1 [11]. Unlike H3K9me3, H3K9me2 forms

44 mega base-scale domains that comprise approximately half of the genomes of pluripotent and

45 differentiated cells [9, 12, 13]. Endogenous G9a and GLP mostly exist as a G9a/GLP heterodimer,

46 and the G9a/GLP heteromeric complex is the functional form for global H3K9 methylation in vivo

47 [5, 7].

48 Although H3K9me2 is primarily catalyzed by G9a/GLP in mESCs, not all of H3K9me2 is diminished

49 in either G9a or Glp knockout (KO) or G9a/Glp double knockout (DKO) mESCs [6, 7], which

50 suggests the involvement of other histone MTases in H3K9me2. However, the function and

51 mechanism of G9a/GLP-independent H3K9me2 are currently unknown. H3K9me2 is enriched in

52 pericentromeric satellite repeats [6, 14] and genomic regions with H3K9me3 [15], and SETDB1 and

53 SUV39H1/2 may be involved in G9a/GLP-independent H3K9me2.

54 Recent Hi-C technology has revealed that chromosomes are hierarchically folded at different levels

55 in the nucleus, such as chromatin loops, topologically associating domains (TADs), and genomic

56 compartments [16-18]. At the mega-base scale, genomic regions are spatially segregated into two

57 subnuclear compartments that predominantly consist of either euchromatin (the A compartment)

58 or heterochromatin (the B compartment). The B compartment is correlated with lamina-

59 associated domains (LADs) [19, 20], which interact with the nuclear lamina (NL) and highly overlap

60 with the H3K9me2 domains [13]. G9a-mediated H3K9me2 is required for LAD-nuclear lamina

61 interaction [21] .

62 Based on the links among H3K9me2, LADs, and B compartments, both G9a/GLP-dependent and

63 independent H3K9me2 could be involved in the A/B compartment formation.

64 In this study, we analyzed H3K9me2 genome-wide profiles and 3D genome organization in Setdb1

65 KO and/or Suv39h1/2 DKO cells treated with a G9a/GLP specific inhibitor, UNC0642 [22], to

66 identify the role of each H3K9 MTase in H3K9me2 domain formation and whether H3K9me2 is

67 involved in nuclear compartment formation.

69 Results

70 Compartment-dependent regulation of H3K9me2 in mESCs

71 It has been reported that H3K9me2 is enriched in LADs, which significantly overlap with the B

72 compartments in mESCs. Although H3K9me2 chromatin immunoprecipitation sequencing (ChIP-

73 seq) in mESCs showed higher H3K9me2 enrichment in the B compartments (negative score

74 regions) than in the A compartments (positive score regions), as expected, H3K9me2 was also

75 present in the A compartments, with a relatively low compartment score (Fig. 1A). As SETDB1 and

76 SUV39H1/2 have the potential to mediate H3K9me2, H3K9me2 ChIP-seq was performed in

77 previously established Setdb1 conditional KO mESCs[8] or Suv39h1/2 DKO mESCs[10] (Additional

78 file 1: Fig. S1A) to investigate the role of SETDB1 and SUV39H1/2 in H3K9me2 formation. Although

79 the global H3K9me2 profile did not change largely in Setdb1 KO or Suv39h1/2 DKO mESCs

80 (Additional file 1: Fig. S1B), H3K9me2 decreased in the A and B compartments in Setdb1 KO mESCs

81 and Suv39h1/2 DKO mESCs, respectively (Fig. 1B).

82 To obtain a more detailed view of the H3K9me2 profile, we identified H3K9me2 domains in mESCs

83 using Hiddendomains, which is a program that uses a hidden Markov model to identify broad

84 domains [23]. The H3K9me2 domains occupied 46.2% and 85.7% of the A and B compartments,

85 respectively (Fig. 1C). The total length of H3K9me2 domains in the A compartments slightly

86 decreased in Setdb1 KO mESCs, which indicates that SETDB1 may have a role in H3K9me2 domain

87 formation in the A compartments (Fig. 1D). It is difficult to further dissect SETDB1- and

88 SUV39H1/2-dependent H3K9me2 in each KO mESC owing to the dominant impact of G9a/GLP for

89 H3K9me2 formation. A G9a/GLP catalytic inhibitor, such as UNC0642, is useful for analyzing

90 G9/GLP-independent H3K9me2 [22]. Treatment with 0.5-2 μM UNC0642 for three days decreased

91 H3K9me2 to a level comparable with that of G9a/GLP DKO mESCs (Additional file 1: Fig. S1C).

92 Furthermore, H3K9me2 in G9a/GLP DKO mESCs and UNC0642-treated mESCs were enriched in

93 pericentromeric satellite repeats, characterized as DAPI-dense regions (Additional file 1: Fig. S1D,

94 E), which is consistent with a previous finding [14]. Compared to WT mESCs, G9a/GLP DKO mESCs

95 and UNC0642-treated mESCs formed relatively narrow H3K9me2 domains, which highly

96 overlapped between G9a/GLP DKO mESCs and UNC0642-treated mESCs (Additional file 1: Fig. S1F,

97 G). Therefore, UNC0642 treatment can mimic the H3K9me2 profile associated with the G9a/Glp

98 DKO phenotype. To investigate the roles of SETDB1 and SUV39H1/2 in G9a/GLP-independent

99 H3K9me2, we performed H3K9me2 ChIP-seq from Setdb1 KO or Suv39h1/2 DKO mESCs treated

100 with 2 μM UNC0642. Hierarchical clustering analysis of the H3K9me2 ChIP-seq data showed that

101 the H3K9me2 profile of UNC0642-treated Setdb1 KO mESCs was largely distinct from that of the

102 G9a/GLP DKO mESCs (Fig. 1E). Consistent with UNC0642-untreated mESCs, reductions of

103 H3K9me2 in Setdb1 KO and Suv39h1/2 DKO mESCs treated with UNC0642 were mainly observed in

104 the A and B compartments, respectively (Fig. 1F), and a significant reduction of the H3K9me2

105 domains in the A compartments was observed in UNC0642-treated Setdb1 KO mESCs (Fig. 1G, H,

106 Additional file 1: S1H). Approximately 80% of the G9a/GLP-independent H3K9me2 domains in the

107 A compartments, and approximately 20% of those in the B compartments were lost in the

108 UNC0642-treated Setdb1 KO mESCs, while almost all the G9a/GLP-independent H3K9me2 domains

109 were maintained in UNC0642-treated Suv39h1/2 DKO mESCs (Fig. 1I). Although SUV39H1/2

110 depletion did not have the same impact as SETDB1 depletion from the view of H3K9me2 ChIP-seq

111 data, H3K9me2 immunofluorescent analysis showed a noticeable loss of H3K9me2 signals in DAPI-

112 dense regions in UNC0642-treated Suv39h1/2 DKO mESCs (Additional file 1: Fig. S1I). This finding

113 suggests a crucial role of SUV39H1/2 in H3K9me2 in pericentromeric satellite repeats.

114 To investigate whether H3K9me2 in the B compartment was mediated by SETDB1 and SUV39H1/2

115 redundantly, we established Setdb1 and Suv39h1/2 triple KO (TKO) mESCs using the CRISPR-Cas9

116 system (Additional file 1: Fig. S1J). The complete removal of Suv39h1 exon 3, which contains SET

117 domain, and the removal of SUV39H2 H398, which is an essential for the methyltransferase

118 activity, were validated by RNA-seq in TKO mESCs[1] (Additional file 1: Fig. S1J). The depletion of

119 SUV39H1/H2 in the TKO mESCs was further validated by western blotting (Additional file 1: Fig.

120 S1K). We used 4-OHT inducible conditional Setdb1 KO mESCs as parental cells for the TKO mES line.

121 SETDB1 and H3K9me3 were undetectable seven days after 4-OHT treatment (Additional file 1: Fig.

122 S1L). Furthermore, UNC0642 treatment to 4-OHT-treated TKO mESCs decreased H3K9me2 to an

123 undetectable level, as analyzed using western blot (Additional file 1: Fig. S1M). We confirmed the

124 loss of H3K9me2 in the two regions where H3K9me2 was retained in both Setdb1 KO and

125 Suv39h1/2 DKO mESCs treated with UNC0642 (Fig. 1J and Additional file 1: Fig. S1N). In conclusion,

126 G9a/GLP-independent H3K9me2 in the A and B compartments in mESCs was mostly mediated by

127 SETDB1 and by both SETDB1 and SUV39H1/2, respectively.

128

129 Compartment-dependent regulation of H3K9me2 in immortalized mouse embryonic fibroblasts

130 (iMEFs)

131 G9a/GLP-independent H3K9me2 in the A and B compartments was mediated by SETDB1 and both

132 SETDB1 and SUV39H1/2 in mESCs, respectively. To investigate whether this trend is specific to

133 pluripotent stem cells, we analyzed the H3K9me2 profiles in iMEFs. Similar to mESCs, H3K9me2

134 was more enriched in the B compartments than in the A compartments in iMEFs (Additional file:

135 Fig. S2A). Compared to mESCs, the larger size of the H3K9me2 domains was retained by UNC0642

136 treatment: the average size of the H3K9me2 domains was 54.5 kb in iMEFs and 14.2 kb in mESCs

137 (Fig. 2A, B). Interestingly, a marked reduction of H3K9me2 in the A compartments was observed in

138 UNC0642-treated iMEFs but not in the B compartments. H3K9me2 decreased upon UNC0642

139 treatment in approximately 90.0% of the A compartments (36.5/4.1 + 36.5), but only decreased in

140 approximately 35.7% of the B compartments (21.2/21.2 + 36.5) (Fig. 2A, C). G9a/GLP-independent

141 H3K9me2 domains in iMEFs occupied 80.8% of the B compartments and 31.3% of A the

142 compartments (Fig. 2D). Therefore, it can be said that H3K9me2 in the B compartments was more

143 resistant to UNC0642 treatment than that in the A compartments in iMEFs. Despite the crucial role

144 of G9a/GLP in the A compartments, H3K9me2 was not completely diminished by UNC0642

145 treatment in iMEFs.

146 We also performed H3K9me2 ChIP-seq analysis on UNC0642-treated Setdb1 KO iMEFs, which is

147 previously established [24]. Further reduction of H3K9me2 in the A compartments was observed

148 in UNC0642-treated Setdb1 KO iMEFs (Additional file 1: Fig. S2B). It was observed that 80.8% of

149 the A compartments (32.8/7.8 + 32.8) and 48.7% of the B compartments (28.9/30.5 + 28.9)

150 showed lower H3K9me2 levels in the UNC0642-treated Setdb1 KO iMEFs than in the UNC0642-

151 treated WT iMEFs (Fig. 2E). The number and total length of the H3K9me2 domains also decreased

152 in the A compartments in the UNC0642-treated Setdb1 KO iMEFs, while those in the B

153 compartments largely remained unchanged (Fig. 2F). Therefore, SETDB1 is essential for G9a/GLP-

154 independent H3K9me2 in the A compartments in iMEFs, as observed in mESCs.

155 To investigate whether SETDB1 and SUV39H1/2 mediated all G9a/GLP-independent H3K9me2 in

156 iMEFs, we established Setdb1 and Suv39h1/2 TKO iMEFs using the CRISPR-Cas9 system (Additional

157 file 1: Fig. S2C). The established TKO iMEFs showed an undetectable level of H3K9me3, as

158 observed from the results of western blot analysis (Additional file 1: Fig. S2D). H3K9me2 still

159 remained, albeit at a low level, but again became undetectable after UNC0642 treatment in

160 Setdb1 and Suv39h1/2 TKO iMEFs (Additional file 1: Fig. S2E). Almost complete loss of H3K9me2 in

161 UNC0642-treated TKO iMEFs was also validated using mass spectrometry analysis (Additional file

162 1: Fig. S2F). ChIP-qPCR analysis of the selected regions in the B compartments, shown in Additional

163 file 1: Fig. S1N, also showed a loss of H3K9me2 in UNC0642-treated TKO iMEFs (Fig. 2G). Thus, we

164 concluded that H3K9me2 in B compartments is redundantly regulated by G9a/GLP, SETDB1, and

165 SUV39H1/2, both in mESCs and iMEFs.

166

167 G9a/GLP-independent H3K9me2 is correlated with efficient H3K9me2 recovery

168 Heterochromatin can spread along chromatin from the nucleation sites, such as recruiters of the

169 writer molecule binding regions [25, 26]. As G9a and GLP bind to nucleosomes that contain

170 H3K9me2 via ankyrin repeats of G9a and GLP [27] and methylate the adjacent nucleosome [28],

171 we hypothesized that H3K9me2 spreads from the G9a/GLP-independent H3K9me2 sites during

172 H3K9me2 domain formation. To investigate this, we analyzed H3K9me2 recovery in mESCs after

173 UNC0642 treatment (Fig. 3A). As H3K9me2 almost completely recovered three days after

174 UNC0642 removal (Additional file 1: Fig. S3A), we performed H3K9me2 ChIP-seq at each time

175 point after UNC0642 removal (0h, 24h, 32h, 40h, 48h, 56h, 64h, and 72h). From the H3K9me2

176 ChIP-seq time course data, we obtained H3K9me2 dynamics during the recovery period (Fig. 3B).

177 Principal component analysis of H3K9me2 Reads Per Kilobase Million (RPKM) in the 80-kb genomic

178 window showed gradual H3K9me2 recovery after UNC0642 removal (Fig. 3C). To investigate the

179 role of G9a/GLP-independent H3K9me2 in H3K9me2 recovery, we classified genomic regions as

180 “early,” “middle,” or “late,” based on the timing of H3K9me2 recovery after UNC0642 removal and

181 compared H3K9me2 levels among the different genomic region classes in the UNC0642-treated

182 mESCs (see Materials and Methods). In this analysis, regions with similar H3K9me2 in WT mESCs

183 were compared. The “early” class showed a higher H3K9me2 level after UNC0642 treatment (Fig.

184 3D, Additional file 1: Fig. S3B) and overlapped more frequently with G9a/GLP-independent

185 H3K9me2 and SETDB1-dependent H3K9me2 than the “middle” and “late“ classes (Fig. 3E, F). In

186 addition, interestingly, H3K9me2 recovery after UNC0642 removal was not induced well in Setdb1

187 KO mESCs and TKO iMEFs (Additional file 1: Fig. S3C, D). These data support the idea that

188 G9a/GLP-independent H3K9me2 helps de novo H3K9me2 domain formation mediated by G9a/GLP.

189

190 G9a/GLP-independent H3K9me2 is involved in transcriptional repression

191 To further address the function of G9a/GLP-independent H3K9me2, we investigated whether

192 G9a/GLP-independent H3K9me2, mainly catalyzed by SETDB1, regulates transcription, and

193 performed RNA-seq analysis in the WT and Setdb1 KO mESCs treated with or without UNC0642.

194 We identified 161, 93, and 75 upregulated genes in UNC0642-treated Setdb1 KO, Setdb1 KO, and

195 UNC0642-treated mESCs, respectively (Additional file 1: Fig. S4A). Fifty-four of the 161

196 upregulated genes in UNC0642-treated Setdb1 KO mESCs were identified in neither Setdb1 KO

197 without UNC0642 treatment nor UNC0642-treated WT mESCs, which suggests a redundant

198 function of SETDB1 and G9a/GLP in gene repression (Additional file 1: Fig. S4A). G9a/GLP mainly

199 repressed genes in the B compartments, while SETDB1 repressed genes in both the A and B

200 compartments (Additional file 1: Fig. S4B). Most upregulated genes both in A and B compartments

201 in UNC0642-treated WT mESCs were derepressed in Setdb1 KO mESCs (17 of 18 in the A

202 compartments and 56 of 57 in the B compartments, FC>2) (Additional file 1: Fig. S4C). In contrast,

203 upregulated genes in the B compartments in Setdb1 KO mESC was more derepressed in UNC0642

204 treated mESCs than genes in the A compartments (35 of 56 in the A compartments (62.5%) and 32

205 of 36 in the B compartments (88.9%), FC>2) (Additional file 1: Fig. S4D). Furthermore, the additive

206 upregulation effect that was observed upon UNC0642 treatment in Setdb1 KO mESCs was also

207 higher in genes in the B compartments than in the A compartments (Additional file 1: Fig. S4E).

208 Therefore, genes repressed by SETDB1 in the B compartments are co-regulated by G9a/GLP;

209 G9a/GLP-mediated H3K9me2 is more important for genes repressed by SETDB1 in the B

210 compartments than in the A compartments.

211 To further investigate the association between G9a/GLP-independent H3K9me2 and gene

212 expression, we performed an integrative RNA-seq, H3K9me2 ChIP-seq, and H3K9me3 ChIP-seq

213 analysis. G9a/GLP-independent H3K9me2 was frequently found within 5-kb from the

214 transcriptional start sites (TSS) of upregulated genes in each condition, while H3K9me3 was found

215 less frequently (Fig. 4A). Both H3K9me3 and G9a/GLP-independent H3K9me2 were enriched

216 around the upregulated genes in Setdb1 KO mESCs and these H3K9me3 and H3K9me2 were

217 dependent on SETDB1 (Fig. 4B, C). In contrast, H3K9me3 around the upregulated genes in the

218 UNC0642-treated mESCs was not dependent on SETDB1, and G9a/GLP-independent H3K9me2

219 decreased in the Setdb1 KO mESCs (Fig. 4B, C). Therefore, H3K9me2 correlated with gene

220 repression to a greater extent than H3K9me3. To investigate more clearly whether SETDB1-

221 dependent H3K9me2 functions as a transcriptional repressor, the expression of genes with

222 SETDB1-dependent H3K9me2 but without H3K9me3 around TSS were analyzed (280 genes).

223 Although these genes were not marked with H3K9me3, they were upregulated in the Setdb1 KO

224 mESCs (Fig. 4D, E). The upregulation of the three selected genes was validated using qRT-PCR (Fig.

225 4F). These data support the idea that SETDB1-dependent H3K9me2 has a role in transcriptional

226 repression and is not just an intermediate of H3K9me3.

227

228 Correlation of decreased H3K9me2 with reorganization of the active compartment setting

229 Lastly, to clarify whether H3K9me2 has a role in A/B compartment formation, we performed Hi-C

230 analysis in each H3K9 MTase deficient mESC treated with or without UNC0642 and calculated the

231 compartment scores in bins of 100-kb. The Hi-C analysis showed an overall maintenance of nuclear

232 compartment patterns in all the samples analyzed (Fig. 5A), and more than 94% of the

233 compartment profiles were conserved in all the samples (Fig. 5B). Moreover, the overall

234 compartment scores did not largely differ between samples (Additional file 1: Fig. S5A). To

235 elucidate the effect of the H3K9me2 changes on the nuclear compartments more specifically, we

236 compared the changes in H3K9me2 in the UNC0642-treated mESCs to the changes in the

237 compartment scores. This comparison clearly showed that the decreased H3K9me2 upon

238 UNC0642 treatment correlated with the increased compartment score (Fig. 5C). The further

239 decrease in H3K9me2 in the UNC0642-treated Setdb1 KO mESCs also correlated with the increased

240 compartment scores (Fig. 5D). In addition to the H3K9me2 changes, the increased gene expression

241 in the UNC0642-treated Setdb1 KO mESCs correlated with increased the compartment scores (Fig.

242 5E). Such increased compartment scores were observed in both the A and B compartments, and B-

243 to-A compartment changes were also observed in six genes (Fig. 5F). Therefore, decreased

244 H3K9me2 correlated with the re-location of the target genes to more active compartments,

245 accompanied by transcriptional activation.

246 Overall, this study demonstrated that H3K9me2 is regulated by different sets of H3K9 MTases

247 between the A and B compartments and that G9a/GLP-independent H3K9me2 has a role in

248 efficient H3K9me2 domain formation, transcriptional repression, and 3D genome organization (Fig.

249 6).

250

251 Discussion

252 This study investigated the role of five different H3K9 MTases in H3K9me2 domain formation and

253 the function of G9a/GLP-independent H3K9me2 in transcriptional repression and 3D genome

254 organization. Although many studies have implicated the presence of G9a/GLP-independent

255 H3K9me2 in mESCs and iMEFs [6, 14, 15], the mechanism and function of G9a/GLP-independent

256 H3K9me2 are unknown. G9a/GLP-independent H3K9me2 is enriched in pericentromeric satellite

257 repeats in mESCs [6, 14] and observed in H3K9me3-marked regions in mESCs and late-replicating

258 domains in MEFs [15]. Therefore, it has been predicted that G9a/GLP-independent H3K9me2 is

259 mediated by other H3K9 MTases, such as SETDB1 and SUV39H1/2. We formally demonstrated that

260 G9a/GLP-independent H3K9me2 is mediated by SETDB1 and SUV39H1/2. Surprisingly, the

261 contribution of SETDB1 and SUV39H1/2 to H3K9me2 differed in each compartment. SETDB1 is

262 essential for G9a/GLP-independent H3K9me2 in the A compartment, while both SETDB1 and

263 SUV39H1/2 mediate this process in the B compartment (Fig. 6 upper left), and it remains unknown

264 why SUV39H1/2 function is restricted only to the B compartment. A mixture of heterochromatin

265 protein 1 (HP1)-SUV39H1-TRIM28 complexes, H3K9me2, and 3-marked nucleosomal arrays

266 undergoes phase separation in vitro [29]. HP1 and SUV39H1 are enriched in DAPI-dense regions

267 and form droplet-like structures in cultured cells [30]. Thus, one possible explanation for SUV39H1

268 restriction to the B compartment is phase separation and the other possible reason could be the

269 genomic distribution of SUV39H1/2-target regions. SUV39H1/2 represses LINE1 retrotransposon,

270 and LINE1 is enriched in AT-rich isochores and gene-poor regions, which is consistent with the

271 features of the B compartment [31]. Thus, the enrichment of LINE1 in B compartments might

272 restrict SUV39H1/2 function in this location.

273 Heterochromatin can spread from the nucleation sites. G9a and GLP can bind to H3K9me1 and

274 H3K9me2 via their ankyrin repeat domain [27], and GLP might premethylate nucleosomes and

275 further methylate the premethylated neighboring nucleosomes [28]. We demonstrated that

276 H3K9me2 reduction upon UNC0642 treatment recovered more efficiently after inhibitor removal

277 in the regions with high G9a/GLP-independent H3K9me2 (Fig. 6 upper right). In addition, SETDB1

278 depletion resulted in no H3K9me2 recovery in mESCs (Additional file 1: Fig. S3C). These results

279 imply that G9a/GLP-independent H3K9me2 helps efficient H3K9me2 establishment or spreading.

280 The HP1-CAF1-SETDB1 complex monomethylates K9 on non-nucleosomal histone H3 [32], and

281 H3K9me1 is imposed during translation by SETDB1 [33]. Thus, H3K9me1 on non-nucleosomal

282 histone H3 mediated by SETDB1 may also provide a scaffold for G9a/GLP during H3K9me2 domain

283 establishment. As H3K9me2 domains are established in intergenic regions in mouse oocytes by

284 G9a, and the global H3K9me2 level is decreased in Setdb1 KO oocytes [34, 35], SETDB1-mediated

285 G9a/GLP-independent H3K9me2 and/or SETDB1-mediated non-nucleosomal H3K9me1 may be

286 important for H3K9me2 domain establishment during gametogenesis.

287 In addition to H3K9me2 spreading, our transcriptome analysis demonstrated that G9a/GLP-

288 independent H3K9me2 plays a role in transcriptional repression. Some SETDB1 target genes

289 harbor SETDB1-dependent H3K9me2 but not H3K9me3 (Fig. 4D-F). It is unknown why histone

290 methylation patterns differ among SETDB1 target genes. As SETDB1 catalytic activity is regulated

291 by ATF7IP and the monoubiquitination of lysine 867 of the hSETDB1, monoubiquitination of

292 SETDB1 or ATF7IP enrichment at the target genomic regions might affect the catalytic activity of

293 SETDB1 [4, 36].

294 Our Hi-C analysis showed that the overall nuclear compartment patterns were maintained in all

295 the analyzed samples. However, H3K9me2 decreased upon UNC0642 treatment in mESCs, which

296 correlated well with increased compartment scores, and the further decrease in H3K9me2 in

297 UNC0642-treated Setdb1 KO mESCs also correlated with the further increase in the compartment

298 scores (Fig. 5C, D). Therefore, we propose that H3K9me2 contributes to heterochromatin

299 compartment formation as one of the key players of determinant mechanisms (Figure 6 bottom).

300 Because Setdb1 KO mESCs treated with UNC0642 still possess H3K9me2 in the B compartment (Fig.

301 6 upper left) but H3K9me2 and H3K9me3 are almost depleted in Setdb1, Suv39h1/2 TKO mESCs,

302 and iMEFs treated with UNC0642, it will be of interest to analyze genome compartmentalization in

303 this cell line in order to fully dissect the role of H3K9 methylation in this process.

304

305 Methods

306 Cell culture. mESCs were maintained in Dulbecco's modified Eagle's medium (Sigma) containing

307 10% Knockout SR (Invitrogen), 1% fetal bovine serum (Biowest), MEM Non-Essential Amino Acids

308 Solution (Gibco), 0.1% LIF and 2-Mercaptoethanol (Nacalai tesque) described as ES medium

309 hereafter). Mouse embryonic fibroblasts were maintained in Dulbecco's modified Eagle's medium

310 (Nacalai tesque) containing 10% fetal bovine serum (Biosera), MEM Non-Essential medium and 2-

311 Mercaptoethanol (Nacalai tesque). To inhibit G9a/GLP catalytic activity, mESCs and iMEFs were

312 cultured for three days with 2 μM UNC0642. For conditional Setdb1 KO, Setdb1fl/fl mESCs were

313 cultured for five days with 800 nM 4-OHT, while Suv39h1/2 DKO/Setdb1fl/fl mESCs were cultured for

314 seven days with 800 nM 4-OHT.

315

316 Establishment of Setdb1, Suv39h1/2 TKO cells.

317 Setdb1 (conditional), Suv39h1/2 TKO mESCs: Setdb1 conditional KO mESC, #33-6 [8]was

318 transfected with the following four gRNAs: 1. Suv39h1 exon 3 upstream in PL-CRISPR.EFS.tRFP, 2.

319 Suv39h2 exon 4 downstream in pKLV2-U6.gRNA (Bbs1)-PGK.puro-BFP, 3. Suv39h2 exon 3 in PL-

320 CRISPR.EFS.tRFP, 4. Suv39h2 exon 4 in pKLV2-U6.gRNA (Bbs1)-PGK.puro-BFP. 3 days after the

321 transfection, BFP and RFP double-positive cells were sorted by flow cytometry. From the sorted

322 cells, Suv39h1/2 DKO mES clone #4 was identified using PCR and western blotting. Partial deletion

323 of SUV39H1/2 SET domains were validated by RNA-seq analysis (Fig. S1J).

324 Setdb1, Suv39h1/2 TKO iMEFs: Setdb1 KO iMEFs [24]were transfected with Suv39h1 exon 4 gRNA

325 in pKLV2-U6.gRNA(Bbs1)-PGK.puro-BFP and Suv39h2 exon 4 upstream and downstream gRNAs in

326 pX330-BB and selected with puromycin. From the puromycin resistant cells, Setdb1, Suv39h1 DKO

327 iMEF clone #27 was identified by Western blot analysis. Then, Suv39h1 KO clone #27 was further

328 transfected with Suv39h2 exon 3 gRNA in pL-CRISPR.EFS.tRFP and pKLV2-U6.gRNA(Bbs1)-

329 PGK.puro-BFP and selected with Puromycin again. From the puromycin resistant cells, Setdb1,

330 Suv39h1/2 TKO KO cell clone #27-34 was identified by Western blot analysis. CRISPR-Cas9

331 mediated Suv39h1 and 2 gene mutations were confirmed by DNA sequencing analysis (Fig. S2C).

332

333 Native ChIP. At least 2x105 cells were lysed in 50 ul Buffer 1 ( 60mM KCl, 15mM NaCl, 5mM MgCl2,

334 0.1mM EGTA, 15mM Tris-HCl (pH7.5), 0.3M Sucrose, 0.5mM DTT, and protease inhibitors), then

335 were added 50 ul Buffer2 (Buffer 1 + 1% NP40) to the sample. After the incubation for 10 min on

336 ice, the sample was added 800 ul Buffer 3 (60mM KCl, 15mM NaCl, 5mM MgCl2, 0.1mM EGTA,

337 15mM Tris-HCl (pH7.5), 1.2M Sucrose, 0.5mM DTT, and protease inhibitors). After centrifugation

338 at 9,000 rpm for 10min at 4℃, the sample was added 100ul MNase buffer (0.32M Sucrose, 50mM

339 Tris-HCl (pH7.5), 4mM MgCl2, 1mM CaCl2, and PMSF) followed by incubation for 20min at 37 oC

340 with 0.3U MNase (TAKARA). After the incubation, the reaction was stopped by 10 ul 0.5M EDTA.

341 After centrifugation at 15,000 rpm for 10min at 4℃, the supernatant was added 900 ul Incubation

342 buffer (50mM NaCl, 20mM Tris-HCl (pH7.5), 5mM EDTA, 0.01% NP40, ad PMSF). 10% of the sample

343 was used for input, and the remains were proceeded the following procedure. The sample was

344 mixed with the antibody-beads complex which was formed by incubation of 20ul protein A/G

345 beads with antibody in Incubation buffer on ice for 1h. After rotating for overnight at 4 oC, the

346 complex was washed by 500ul Wash buffer A (75mM NaCl, 50mM Tris-HCl (pH7.5), 10mM EDTA,

347 and 0.01% NP40), 500ul Wash buffer B (100mM NaCl, 50mM Tris-HCl (pH7.5), 10mM EDTA, and

348 0.01% NP40), and 500ul Wash buffer C (175mM, 50mM Tris-HCl (pH7.5), 10mM EDTA, and 0.01%

349 NP40). The ChIP DNA was eluted by RNase and Protease treatment followed by DNA purification

350 using PCR purification kit (QIAGEN).

351

352 Western blotting. Briefly, cells were suspended in RIPA buffer and sonicated. The extract was

353 incubated for 30 minutes on ice, and then incubated at 95oC for 5 minutes. The extract was loaded

354 and run on SDS-PAGE gel as standard protocols. For histone proteins, intensity was analyzed by

355 Odyssey Clx (LI-COR).

356

357 Mass spectrometry analysis for H3K9 methylation

358 Histones were prepared by the acid-extraction method as previously described [6]. The samples

359 were subjected to SDS-PAGE and stained with Coomassie blue. The protein bands corresponding

360 to histone H3 were excised from the gel and were digested with Achromobacter protease I (Lys-C)

361 in gel. The digests were performed nano-liquid chromatography - tandem mass spectrometry

362 using EASY-nLC 1200 liquid chromatograph (Thermo Fisher Scientific) connected to Q-Exactive HFX

363 mass spectrometer equipped with nanospray ion source (Thermo Fisher Scientific). Peptides

364 containing the digests were separated with a linear gradient of 0-100%B buffer in A buffer (A:0.1%

365 formic acid, B:80%acetonitrile/0.1%formic acid) for 20min with a reversed-phase column (NTCC

366 analytical column, C18, φ75 μm ×150 mm, 3 μm; Nikkyo Technos, Japan). MS and MS/MS data

367 were acquired using a data dependent TOP 10 method. Protein quantification were performed

368 with Proteome Discoverer 2.4. (Thermo Fisher Scientific) using a sequence database search node

369 as a MASCOT program 2.7 (Matrix Science) with following parameters, Database : Histone 130611

370 (66 sequences; 15002 residues), Enzyme : Lys-C/P, Variable modifications : Acetyl (Protein N-

371 term),Oxidation (M),Gln->pyro-Glu (N-term Q),Acetyl (K),Methyl (K),Trimethyl (K),Dimethyl (K),

372 Mass values : Monoisotopic, Protein Mass : Unrestricted, Peptide Mass Tolerance : ± 15 ppm,

373 Fragment Mass Tolerance: ± 30 mmu, Max Missed Cleavages : 3, Instrument type : ESI-TRAP.

374 Protein methylation rates were obtained peak areas of selected ion chromatograms that are the

375 tetra charged protonated molecules at m/z 381.9706 (methyl), m/z 385.4745 (dimethyl) and m/z

376 388.9784 (trimethyl) of “QTARKSTGGKAPRK” related peptides ( 1xAcetyl [K10]; 1xGln->pyro-Glu

377 [N-Term]; 1xmethylations [K5]) using Qual Browser (Thermo xcalibur 4.1.50, Thermo Fisher

378 Scientific) with ±15ppm width.

379

380 RT-qPCR. RNA was isolated by RNeasy Plus Mini Kit (Qiagen) following manufacturer’s instructions.

381 cDNA synthesis was performed with Omniscript RT Kit (Qiagen). qPCR was carried out using Power

382 SYBR Green PCR Master Mix (ThermoFisher SCIENTIFIC) on StepOnePlusTM (ThermoFisher

383 SCIENTIFIC). The signals were normalized relative to Hprt. All qPCR data is represented as the mean

384 +/- standard deviation of three biological replicates.

385

386 Preparation of ChIP-seq library. The ChIP DNA was fragmented by Picoruptor (Diagenode) for 10

387 cycles of 30 seconds on, 30seconds off. Then, ChIP library was constructed by KAPA Hyper Prep Kit

388 (KAPA BIOSYSTEMS) and SeqCap Adapter Kit A (Roche) or SMARTER ThruPLEX DNA-seq kit

389 (TAKARA) and SMARTer DNA Unique Dual Index Kit (TAKARA) according to manufacturer

390 instructions. The concentration of the ChIP-seq library was quantified by KAPA Library

391 quantification kit (KAPA BIOSYSTEMS). ChIP sequencing was performed on a HiSeq 2000 platform

392 (Illumina).

393

394 Preparation of RNA-seq library. 500 ng of total RNA was used for RNA-seq library construction.

395 RNA-seq library was constructed by KAPA mRNA Hyper Prep Kit (KAPA BIOSYSTEMS) and SeqCap

396 Adapter Kit (Roche) according to manufacturer instructions. The concentration of the RNA-seq

397 library was quantified by KAPA Library quantification kit (KAPA BIOSYSTEMS). mRNA sequencing

398 was performed on a HiSeq 2000 platform (Illumina).

399

400 Preparation of HiC-seq library

401 Hi-C experiments were performed as previously described [37-39], based on DpnII enzyme (4-bps

402 cutter) using 2×106 fixed cells. Hi-C libraries were subject to paired-end sequencing (150 base pair

403 (bp) read length) using HiSeq X Ten.

404

405 Hi-C data processing and A/B compartment calculation

406 Hi-C data processing was done by using Docker for 4DN Hi-C pipeline (v43,

407 https://github.com/4dn-dcic/docker-4dn-hic). The pipeline includes alignment (using the mouse

408 genome, mm10) and filtering steps. After filtering valid Hi-C alignments, .hic format Hi-C matrix

409 files were generated by Juicer Tools [40] using the reads with MAPQ>10. The A/B compartment

410 (compartment score) profiles (in 100-kb bins) in each chromosome (without sex chromosome)

411 were calculated from .hic format Hi-C matrix files (intrachromosomal KR normalized Hi-C maps) by

412 Juicer Tools [40] as previously described[41].

413

414 ChIP-seq analysis.

415 ・Mapping and domain identification

416 Adaptor sequences in reads were trimmed using Trim Galore version 0.3.7

417 (http://www.bioinformatics.babraham.ac.uk/projects/trim_galore/). Then trimmed reads were

418 aligned to the mm10 genome build using bowtie version 0.12.7 [42] with default parameters.

419 Duplicated reads were removed using samtools version 0.1.18 [43]. H3K9me2 domains were

420 identified by Hiddendomains with some modifications: The option of Bin size and max.read.count

421 was 2,000 bp and 150, respectively. Read number in each bin was normalized by the following

422 formula to adjust difference in read number among samples. Normalized read number = read

423 number x 30,000,000 / total read number. For the hidden Markov parameter to determine

424 enriched and depleted states, average parameter among chromosomes rather than parameter

425 calculated by each chromosome was used, because very few domains were identified in some

426 chromosomes. Loss of H3K9me2 domains was identified using getDifferentialPeaks in Homer (fold

427 change >= 4 and P-value <= 0.0001). Conserved H3K9me2 domains ware the domains where was

428 identified also in a sample and were not identified as lost domains.

429 ・Classification of regions by the timing of H3K9me2 recovered

430 RPKM of each 10-kb bin was calculated and the RPKM was converted to z-score. The 10-kb bin

431 where Z-score increased by 0.3 or more at 32h, 48h, and 72h from 0h was annotated as early,

432 middle, and late, respectively.

433

434 RNA-seq analysis

435 Raw FastQ data were trimmed with Trim Galore (v0.3.7, default parameters)

436 (http://www.bioinformatics.babraham.ac.uk/projects/trim_galore/) and mapped to the mouse

437 GRCm38 genome assembly using TopHat (v2.1.1) [44]. After read mapping, mapped reads were

438 analyzed by TEtranscripts (v1.4.11, default parameters) [45] to calculate gene expression levels and

439 identify DE genes (adj. P-value < 0.05, FC > 10).

440

441 Visualization of NGS data

442 The Integrative Genomics Viewer (IGV) was used to visualize NGS data. Enrichment of

443 H3K9me2/H3K9me3 enrichment in specified regions was visualized by ngsplot. Scatter plot

444 analysis, principal component analysis boxplot and violin plot analysis were performed by R script.

445

446 Immunofluorescence analysis. 2x104 cells were seeded on laminin coated 12 well Chamber (Ibidi)

447 the day before fixation. The cells were fixed with 4% paraformaldehyde for 10 minutes at room

448 temperature, permeabilized with 0.1% Triton X-100 for 10 minutes, blocked with 3% BSA 0.1%

449 Tween20 in PBS and incubated overnight with primary antibodies at 4℃. Anti-mouse IgG

450 conjugated with Alexa Fluor 568 (ThermoFisher SCIENTIFIC) or anti-rabbit IgG conjugated with

451 Alexa Fluor 488 (ThermoFisher SCIENTIFIC) were used as secondary antibodies. The nuclei were

452 counterstained with DAPI, observed under fluorescence microscopy and analyzed Olympus

453 FluoViewTM FV3000 (Olympus).

454

455 Antibodies. For western blotting, antibodies specific for histone H3 (07-690, EMD Millipore),

456 H3K9me3 (2F3), H3K9me2 (6D11), SETDB1 (Cp10377, Cell Applications), SUV39H1 (#8729, CST),

457 SUV39H2 (LS-116360, LSBio), G9a (A8620A) and GLP (B0422B) were used as primary antibody. For

458 histone proteins, IRDye 800CW Goat anti-Mouse IgG (926-32210, LI-COR) and IRDye 680RD Goat

459 anti-Rabbit IgG (926-68071, LI-COR) were used as secondary antibody. For non-histone proteins,

460 HRP-linked anti-Rabbit IgG (NA934, GE Healthcare) and HRP-linked anti-Mouse IgG (NA931, GE

461 Healthcare) were used as secondary antibody. For immunofluorescence analysis, H3K9me2 (6D11),

462 H3K9me3 (39161, Active Motif) were used for primary antibody, and Goat anti-Mouse IgG Alexa

463 Fluor 568 (A-11031, Invitrogen) and Goat anti-Rabbit IgG Alexa Fluor 488 (A-11034) were used for

464 secondary antibody . For ChIP analysis, antibodies specific for H3K9me3 (2F3) and H3K9me2

465 (6D11) were used.

466

467 Oligonucleotides. Oligonucleotide sequences for PCR primers and production of gRNA targeting

468 vectors are listed in Supplementary Table S1 and S2, respectively.

469

470 DATA ACCESS

471 All reads from the RNA-seq, ChIP-seq and Hi-C experiments in this study have been submitted to

472 Sequence Read Archive (SRA; https://www.ncbi.nlm.nih.gov/sra) under accession number

473 SRPXXXX. Read number of NGS data are listed in Supplementary Table S3.

474

475 Figure legend

476 Fig. 1 Characterization of SETDB1 and SUV39H1/2-dependent H3K9me2 region in mESCs

477 A. Comparison between compartment score and H3K9me2 enrichment in 80-kb bin. RPKM of

478 H3K9me2 from WT mESCs is negatively correlated with compartment score. Darker blue

479 represents higher dot density.

480 B. Comparison between compartment score in WT mESCs and changes in H3K9me2 in Setdb1 KO

481 (left) or Suv39h1/2 DKO (right) mESCs in 80-kb bins. Decreased H3K9me2 is observed in the A

482 and B compartments in Setdb1 KO and Suv39h1/2 DKO mESCs, respectively.

483 C. A representative view of H3K9me2 domains in WT mESCs identified by Hiddendomains (left).

484 Top panel, H3K9me2 ChIP-seq data from WT mESCs; middle panel, H3K9me2 domains; bottom

485 panel, compartment score in 80-kb bins. Right, fractions of H3K9me2 domains in the A or B

486 compartments. Although H3K9me2 domains are more enriched in the B compartments than in

487 the A compartments, H3K9me2 occupies about half of the A compartments.

488 D. Length of H3K9me2 domains in each KO mESCs to WT mESCs. The length of H3K9me2 domains

489 in the A compartments is slightly decreased in Setdb1 KO mESCs, while that is increased in

490 Suv39h1/2 DKO mESCs.

491 E. A correlation matrix of H3K9me2 ChIP-seq data in mESCs. UNC0642-treated mESCs form a

492 cluster distinct from untreated mESCs. Among UNC0642-treated mESCs, Setdb1 KO mESCs

493 show a different H3K9me2 profile.

494 F. A comparison between compartment score in WT mESCs and changes in H3K9me2 in Setdb1

495 KO (left) or Suv39h1/2 DKO (right) mESCs treated with UNC0642. Consistent with UNC0642-

496 untreated mESCs, H3K9me2 is decreased in the A and B compartments in Setdb1 KO and

497 Suv39h1/2 DKO mESCs treated with UNC0642, respectively.

498 G. A representative view of H3K9me2 ChIP-seq from G9a/GLP DKO and UNC0642-treated mESCs.

499 G9a/GLP DKO mESCs show an H3K9me2 profile similar to UNC0642-treated mESCs. A drastic

500 loss of H3K9me2 domains in UNC0642-treated Setdb1 KO mESCs in the A compartments

501 (circled in red).

502 H. Heatmaps of H3K9me2 enrichment around G9a/GLP-independent H3K9me2 regions. The A

503 compartment-specific H3K9me2 loss is observed in UNC0642-treated Setdb1 KO mESCs.

504 I. Fractions of conserved or lost G9a/GLP-independent H3K9me2 in Setdb1 KO or Suv39h1/2 DKO

505 mESCs treated with UNC0642. More than 80% of G9a/GLP-independent H3K9me2 domains in

506 the A compartments are lost in UNC0642-treated Setdb1 KO mESCs, while about 80% of that in

507 B compartment are conserved. In UNC0642-treated Suv39h1/2 DKO mESCs, almost all

508 G9a/GLP-independent H3K9me2 domains are maintained.

509 J. H3K9me2 ChIP-qPCR analysis in regions shown in Fig. S1N. ChIP was performed five days after

510 4-OHT treatment in Setdb1 KO mESCs and 7 days after 4-OHT treatment in Setdb1/Suv39h1/2

511 TKO mESCs. H3K9me2 in the region 1 and 2 is completely lost in UNC0642-treated TKO mESCs.

512

513

514 Fig. 2 Characterization of SETDB1- and SUV39H1/2-dependent H3K9me2 regions in iMEFs

515 A. A representative view of H3K9me2 ChIP-seq data along chromosome 6 in iMEFs. Although

516 H3K9me2 in A compartment is decreased by UNC0642 treatment, low H3K9me2 signal is

517 retained. Those retained H3K9me2 is further diminished in UNC0642-treated Setdb1 KO iMEFs.

518 B. Difference in G9a/GLP-independent H3K9me2 between mESCs and iMEFs. H3K9me2 domain

519 size is larger in iMEFs than in mESCs.

520 C. A comparison between compartment score and H3K9me2 changes by UNC0642 treatment in

521 iMEFs. Compartment score in mESCs was used for the analysis. Overall H3K9me2 reduction in

522 the A compartments is induced by the UNC0642 treatment. Each plot represents data from

523 each 80-kb bin. Darker blue represents higher dot density.

524 D. Fractions of H3K9me2 domains in the A and B compartments in UNC0642-treated iMEFs.

525 H3K9me2 domains in the B compartments is largely retained after UNC0642 treatment.

526 E. A comparison between compartment score and H3K9me2 changes in UNC0642-treated Setdb1

527 KO iMEFs. Further reduction of H3K9me2 in the A compartments is observed in UNC0642-

528 treated Setdb1 KO iMEFs.

529 F. The number and the length of H3K9me2 domains in WT and Setdb1 KO iMEFs treated with

530 UNC0642. Both the number and the length of H3K9me2 domains in the A compartments are

531 reduced in UNC0642-treated Setdb1 KO iMEFs.

532 G. H3K9me2 ChIP-qPCR analysis in regions showing in Fig. S1N. H3K9me2 in the region 1 and 2 is

533 completely lost in UNC0642-treated TKO iMEFs (N = 3).

534

535 Fig. 3 Recovery of H3K9me2 after the UNC0642 treatment

536 A. An experimental design to investigate H3K9me2 recovery. mESCs were treated with UNC0642

537 for 3 days, and then H3K9me2 ChIP-seq was performed at each time point after the UNC0642

538 removal.

539 B. A representative view of H3K9me2 ChIP-seq data during H3K9me2 recovery.

540 C. A principal component analysis of H3K9me2 ChIP-seq data during H3K9me2 recovery.

541 D. Distinct kinetics of H3K9me2 recovery from UNC0642 treatment between three classes of

542 sequences. Each 10-kb region throughout the genome was classified to three types named

543 “early”, “middle”, and “late” based on the timing of H3K9me2 recovery. Z-scaling was

544 performed on H3K9me2 RPKM of 10-kb bins in each time point. Y-axis represents the average

545 Z-scores of each class. Only regions with Z-scores between 0.5 and 0.7 in WT mESCs were used

546 for this analysis to match the final H3K9me2 levels among three classes. The regions that show

547 faster H3K9me2 recovery tend to harbor higher H3K9me2 after UNC0642 treatment.

548 E. Overlap between G9a/GLP-independent H3K9me2 domains and each class. “Early” class shows

549 the most frequent overlap with G9a/GLP-independent H3K9me2 regions among three classes.

550 Only regions with Z-scores between 0.5 and 0.7 in WT mESCs were used for this figure.

551 F. Overlap between SETDB1-dependent H3K9me2 domains and each class. Only regions with Z-

552 scores between 0.5 and 0.7 in WT mESCs were used for this figure.

553

554 Fig. 4 Function of G9a/GLP-independent H3K9me2 in transcriptional regulation

555 A. Overlap of upregulated genes with G9a/GLP-independent H3K9me2 or H3K9me3 domains. The

556 graph shows the fractions of upregulated genes in each condition that overlap with G9a/GLP-

557 independent H3K9me2 domains or H3K9me3 domains within 5kb of TSS.

558 B. Enrichment of H3K9me2 in WT or Setdb1 mESCs treated with UNC0642 around upregulated

559 genes in each condition. G9a/GLP-independent H3K9me2 is enriched from upstream of TSS to

560 gene body of upregulated genes identified in Setdb1 KO mESCs treated with or without

561 UNC0642. These H3K9me2 is largely dependent on SETDB1.

562 C. Enrichment of H3K9me3 in WT or Setdb1 mESCs treated with UNC0642 around upregulated

563 genes in each condition. Different from G9a/GLP-independent H3K9me2, H3K9me3 is mainly

564 enriched upstream of upregulated genes identified in Setdb1 KO mESCs treated with or without

565 UNC0642.

566 D. Expression and H3K9 methylation profiles of genes with SETDB1-dependent H3K9me2 and

567 without H3K9me3. Genes that satisfy the following criteria are extracted: 1. H3K9me2 domains

568 are located within 0.5 kb of TSS in UNC0642-treated mESCs, but not in UNC0642-treated

569 Setdb1 KO mESCs. 2. H3K9me3 domains are not located within 0.5 kb of TSS in any condition. 3.

570 H3K9me2 RPKM in UNC0642-treated mESCs around 0.5 kb of TSS is more than 1, and that in

571 UNC0642-treated Setdb1 KO mESCs is less than 1. 4. H3K9me3 RPKM around 0.5 kb of TSS is

572 less than 1 in any condition. The left panel is a boxplot of log2 fold change of expression of

573 genes satisfying the above criteria. Middle figure shows the enrichment of H3K9me2 around

574 the TSS in UNC0642-treated WT or Setdb1 KO mESCs. The right figure shows the enrichment of

575 H3K9me3 around the TSS in WT or Setdb1 KO mESCs.

576 E. Heatmap of fold change of expression, H3K9me2 RPKM, and H3K9me3 RPKM in genes

577 extracted in Fig. 4D. H3K9me2 and H3K9me3 RPKM were calculated around 0.5 kb of the TSS.

578 F. qRT-PCR of representative genes extracted in Fig. 4D. Top panel shows H3K9me2, H3K9me3,

579 and RNA-seq profiles of Nlrp4c, Gm4971, and Trpd52I3. SETDB1-dependent H3K9me2 is

580 present around TSS of these genes, but H3K9me3 is absent in all conditions. Star mark

581 represents SETDB1-dependent H3K9me2 around TSS. Bottom graph shows relative expression

582 of these three genes in each condition (N = 3).

583

584 Fig. 5 Correlation of decreased H3K9me2 and movement toward more active compartment

585 A. A representative view of compartment score in each 100-kb bin in mESCs. The regions colored

586 by red and blue represent the A and B compartments, respectively.

587 B. Compartment change in each condition. More than 94% of compartments are maintained in all

588 conditions.

589 C. A comparison between change in H3K9me2 and compartment scores in UNC0642-treated

590 mESCs. Each 100-kb region was classified by compartment score changes in UNC0642-treated

591 mESCs. Decreased H3K9me2 in UNC0642-treated mESCs is correlated with an increased

592 compartment score.

593 D. A comparison between changes in H3K9me2 and compartment scores in UNC0642-treated

594 Setdb1 KO mESCs. Each 100-kb region was classified by compartment score change between

595 UNC0642-treated Setdb1 KO mESCs and UNC0642-treated mESCs. Decreased H3K9me2 in

596 UNC0642-treated Setdb1 KO mESCs is correlated with an increased compartment score.

597 E. Change in compartment score around upregulated genes UNC0642-treated Setdb1 KO mESCs

598 and randomly selected genes. Increased compartment score is observed in upregulated genes

599 in UNC0642-treated Setdb1 KO mESCs.

600 F. Pattern of compartment changes in genes derepressed in UNC0642-treated mESCs. Only genes

601 of which compartment scores are increased were used in this analysis. Six genes of them

602 showed compartment change from the B compartments to the A compartments.

603

604 Fig. 6 Summary of this study

605 Although H3K9me2 (shown in green) is largely mediated by G9a/GLP in mESCs, some residual

606 H3K9me2 is observed in G9a/GLP DKO mESCs. SETDB1 and SUV39H1/2 mediate G9a/GLP-

607 independent H3K9me2 in a compartment-dependent manner. SETDB1 and SUV39H1/2 are

608 essential for G9a/GLP-independent H3K9me2 in the A compartments (red) and in pericentromeric

609 satellite repeats (dark blue), respectively, while H3K9me2 in the B compartments (light blue) is

610 mediated by both SETDB1 and SUV39H1/2 redundantly (Fig. 6 upper left). During H3K9me2

611 domain establishment after recovery from an UNC0642 treatment, H3K9me2 is recovered

612 efficiently in regions with high G9a/GLP-independent H3K9me2. Thus, G9a/GLP-independent

613 H3K9me2 might facilitate the efficient establishment of H3K9me2 domains (Fig. 6 upper right).

614 Finally, we compared H3K9me2, transcription and 3D genomes, and found that the reduction of

615 H3K9me2 resulted in a movement toward more active compartment, accompanied by

616 transcriptional activation (Fig. 6 bottom). Thus, G9a/GLP-independent H3K9me2 functions in

617 H3K9me2 domain establishment, transcriptional repression and 3D genome organization.

618

619 Refferences

620 1. Rea S, Eisenhaber F, O'Carroll D, Strahl BD, Sun ZW, Schmid M, Opravil S, Mechtler K, 621 Ponting CP, Allis CD, Jenuwein T: Regulation of chromatin structure by site-specific 622 histone H3 methyltransferases. Nature 2000, 406:593-599.

623 2. Nakayama J, Rice JC, Strahl BD, Allis CD, Grewal SI: Role of histone H3 lysine 9 624 methylation in epigenetic control of heterochromatin assembly. Science 2001, 292:110- 625 113. 626 3. Martens JH, O'Sullivan RJ, Braunschweig U, Opravil S, Radolf M, Steinlein P, Jenuwein T: 627 The profile of repeat-associated histone lysine methylation states in the mouse 628 epigenome. EMBO J 2005, 24:800-812. 629 4. Wang H, An W, Cao R, Xia L, Erdjument-Bromage H, Chatton B, Tempst P, Roeder RG, 630 Zhang Y: mAM facilitates conversion by ESET of dimethyl to trimethyl lysine 9 of histone 631 H3 to cause transcriptional repression. Mol Cell 2003, 12:475-487. 632 5. Tachibana M, Sugimoto K, Fukushima T, Shinkai Y: Set domain-containing protein, G9a, is 633 a novel lysine-preferring mammalian histone methyltransferase with hyperactivity and 634 specific selectivity to lysines 9 and 27 of histone H3. J Biol Chem 2001, 276:25309-25317. 635 6. Tachibana M, Sugimoto K, Nozaki M, Ueda J, Ohta T, Ohki M, Fukuda M, Takeda N, Niida H, 636 Kato H, Shinkai Y: G9a histone methyltransferase plays a dominant role in euchromatic 637 histone H3 lysine 9 methylation and is essential for early embryogenesis. Genes Dev 638 2002, 16:1779-1791. 639 7. Tachibana M, Ueda J, Fukuda M, Takeda N, Ohta T, Iwanari H, Sakihama T, Kodama T, 640 Hamakubo T, Shinkai Y: Histone methyltransferases G9a and GLP form heteromeric 641 complexes and are both crucial for methylation of euchromatin at H3-K9. Genes Dev 642 2005, 19:815-826. 643 8. Matsui T, Leung D, Miyashita H, Maksakova IA, Miyachi H, Kimura H, Tachibana M, Lorincz 644 MC, Shinkai Y: Proviral silencing in embryonic stem cells requires the histone 645 methyltransferase ESET. Nature 2010, 464:927-931. 646 9. Karimi MM, Goyal P, Maksakova IA, Bilenky M, Leung D, Tang JX, Shinkai Y, Mager DL, 647 Jones S, Hirst M, Lorincz MC: DNA methylation and SETDB1/H3K9me3 regulate 648 predominantly distinct sets of genes, retroelements, and chimeric transcripts in mESCs. 649 Cell Stem Cell 2011, 8:676-687. 650 10. Lehnertz B, Ueda Y, Derijck AA, Braunschweig U, Perez-Burgos L, Kubicek S, Chen T, Li E, 651 Jenuwein T, Peters AH: Suv39h-mediated histone H3 lysine 9 methylation directs DNA 652 methylation to major satellite repeats at pericentric heterochromatin. Curr Biol 2003, 653 13:1192-1200. 654 11. Bulut-Karslioglu A, De La Rosa-Velazquez IA, Ramirez F, Barenboim M, Onishi-Seebacher M, 655 Arand J, Galan C, Winter GE, Engist B, Gerle B, et al: Suv39h-dependent H3K9me3 marks 656 intact retrotransposons and silences LINE elements in mouse embryonic stem cells. Mol 657 Cell 2014, 55:277-290. 658 12. Filion GJ, van Steensel B: Reassessing the abundance of H3K9me2 chromatin domains in 659 embryonic stem cells. Nat Genet 2010, 42:4; author reply 5-6. 660 13. Wen B, Wu H, Shinkai Y, Irizarry RA, Feinberg AP: Large histone H3 lysine 9 dimethylated 661 chromatin blocks distinguish differentiated from embryonic stem cells. Nat Genet 2009, 662 41:246-250. 663 14. Peters AH, Kubicek S, Mechtler K, O'Sullivan RJ, Derijck AA, Perez-Burgos L, Kohlmaier A, 664 Opravil S, Tachibana M, Shinkai Y, et al: Partitioning and plasticity of repressive histone 665 methylation states in mammalian chromatin. Mol Cell 2003, 12:1577-1589.

666 15. Chen CCL, Goyal P, Karimi MM, Abildgaard MH, Kimura H, Lorincz MC: H3S10ph broadly 667 marks early-replicating domains in interphase ESCs and shows reciprocal antagonism 668 with H3K9me2. Genome Res 2018, 28:37-51. 669 16. Dixon JR, Selvaraj S, Yue F, Kim A, Li Y, Shen Y, Hu M, Liu JS, Ren B: Topological domains in 670 mammalian genomes identified by analysis of chromatin interactions. Nature 2012, 671 485:376-380. 672 17. Rao SS, Huntley MH, Durand NC, Stamenova EK, Bochkov ID, Robinson JT, Sanborn AL, 673 Machol I, Omer AD, Lander ES, Aiden EL: A 3D map of the human genome at kilobase 674 resolution reveals principles of chromatin looping. Cell 2014, 159:1665-1680. 675 18. Lieberman-Aiden E, van Berkum NL, Williams L, Imakaev M, Ragoczy T, Telling A, Amit I, 676 Lajoie BR, Sabo PJ, Dorschner MO, et al: Comprehensive mapping of long-range 677 interactions reveals folding principles of the human genome. Science 2009, 326:289-293. 678 19. van Steensel B, Belmont AS: Lamina-Associated Domains: Links with Chromosome 679 Architecture, Heterochromatin, and Gene Repression. Cell 2017, 169:780-791. 680 20. Kind J, Pagie L, de Vries SS, Nahidiazar L, Dey SS, Bienko M, Zhan Y, Lajoie B, de Graaf CA, 681 Amendola M, et al: Genome-wide maps of nuclear lamina interactions in single human 682 cells. Cell 2015, 163:134-147. 683 21. Kind J, Pagie L, Ortabozkoyun H, Boyle S, de Vries SS, Janssen H, Amendola M, Nolen LD, 684 Bickmore WA, van Steensel B: Single-cell dynamics of genome-nuclear lamina 685 interactions. Cell 2013, 153:178-192. 686 22. Liu F, Barsyte-Lovejoy D, Li F, Xiong Y, Korboukh V, Huang XP, Allali-Hassani A, Janzen WP, 687 Roth BL, Frye SV, et al: Discovery of an in vivo chemical probe of the lysine 688 methyltransferases G9a and GLP. J Med Chem 2013, 56:8931-8942. 689 23. Starmer J, Magnuson T: Detecting broad domains and narrow peaks in ChIP-seq data 690 with hiddenDomains. BMC Bioinformatics 2016, 17:144. 691 24. Kato M, Takemoto K, Shinkai Y: A somatic role for the histone methyltransferase Setdb1 692 in endogenous retrovirus silencing. Nat Commun 2018, 9:1683. 693 25. Fukuda K, Shinkai Y: SETDB1-Mediated Silencing of Retroelements. Viruses 2020, 12. 694 26. Allshire RC, Madhani HD: Ten principles of heterochromatin formation and function. Nat 695 Rev Mol Cell Biol 2018, 19:229-244. 696 27. Collins RE, Northrop JP, Horton JR, Lee DY, Zhang X, Stallcup MR, Cheng X: The ankyrin 697 repeats of G9a and GLP histone methyltransferases are mono- and dimethyllysine 698 binding modules. Nat Struct Mol Biol 2008, 15:245-250. 699 28. Liu N, Zhang Z, Wu H, Jiang Y, Meng L, Xiong J, Zhao Z, Zhou X, Li J, Li H, et al: Recognition 700 of H3K9 methylation by GLP is required for efficient establishment of H3K9 methylation, 701 rapid target gene repression, and mouse viability. Genes Dev 2015, 29:379-393. 702 29. Wang L, Gao Y, Zheng X, Liu C, Dong S, Li R, Zhang G, Wei Y, Qu H, Li Y, et al: Histone 703 Modifications Regulate Chromatin Compartmentalization by Contributing to a Phase 704 Separation Mechanism. Mol Cell 2019, 76:646-659 e646. 705 30. Lachner M, O'Carroll D, Rea S, Mechtler K, Jenuwein T: Methylation of histone H3 lysine 9 706 creates a binding site for HP1 proteins. Nature 2001, 410:116-120. 707 31. Ichiyanagi K, Li Y, Watanabe T, Ichiyanagi T, Fukuda K, Kitayama J, Yamamoto Y, 708 Kuramochi-Miyagawa S, Nakano T, Yabuta Y, et al: Locus- and domain-dependent control 709 of DNA methylation at mouse B1 retrotransposons during male germ cell development. 710 Genome Res 2011, 21:2058-2066.

711 32. Loyola A, Tagami H, Bonaldi T, Roche D, Quivy JP, Imhof A, Nakatani Y, Dent SY, Almouzni 712 G: The HP1alpha-CAF1-SetDB1-containing complex provides H3K9me1 for Suv39- 713 mediated K9me3 in pericentric heterochromatin. EMBO Rep 2009, 10:769-775. 714 33. Rivera C, Saavedra F, Alvarez F, Diaz-Celis C, Ugalde V, Li J, Forne I, Gurard-Levin ZA, 715 Almouzni G, Imhof A, Loyola A: Methylation of histone H3 lysine 9 occurs during 716 translation. Nucleic Acids Res 2015, 43:9097-9106. 717 34. Eymery A, Liu Z, Ozonov EA, Stadler MB, Peters AH: The methyltransferase Setdb1 is 718 essential for meiosis and mitosis in mouse oocytes and early embryos. Development 719 2016, 143:2767-2779. 720 35. Au Yeung WK, Brind'Amour J, Hatano Y, Yamagata K, Feil R, Lorincz MC, Tachibana M, 721 Shinkai Y, Sasaki H: Histone H3K9 Methyltransferase G9a in Oocytes Is Essential for 722 Preimplantation Development but Dispensable for CG Methylation Protection. Cell Rep 723 2019, 27:282-293 e284. 724 36. Sun L, Fang J: E3-Independent Constitutive Monoubiquitination Complements Histone 725 Methyltransferase Activity of SETDB1. Mol Cell 2016, 62:958-966. 726 37. Kadota M, Nishimura O, Miura H, Tanaka K, Hiratani I, Kuraku S: Multifaceted Hi-C 727 benchmarking: what makes a difference in chromosome-scale genome scaffolding? 728 Gigascience 2020, 9. 729 38. Ikeda T, Hikichi T, Miura H, Shibata H, Mitsunaga K, Yamada Y, Woltjen K, Miyamoto K, 730 Hiratani I, Yamada Y, et al: Srf destabilizes cellular identity by suppressing cell-type- 731 specific gene expression programs. Nat Commun 2018, 9:1387. 732 39. !!! INVALID CITATION !!! {Ikeda, 2018 #696;Kadota, 2020 #702}. 733 40. Durand NC, Shamim MS, Machol I, Rao SS, Huntley MH, Lander ES, Aiden EL: Juicer 734 Provides a One-Click System for Analyzing Loop-Resolution Hi-C Experiments. Cell Syst 735 2016, 3:95-98. 736 41. Miura H, Poonperm R, Takahashi S, Hiratani I: Practical Analysis of Hi-C Data: Generating 737 A/B Compartment Profiles. Methods Mol Biol 2018, 1861:221-245. 738 42. Langmead B, Trapnell C, Pop M, Salzberg SL: Ultrafast and memory-efficient alignment of 739 short DNA sequences to the human genome. Genome Biol 2009, 10:R25. 740 43. Li H, Handsaker B, Wysoker A, Fennell T, Ruan J, Homer N, Marth G, Abecasis G, Durbin R, 741 Genome Project Data Processing S: The Sequence Alignment/Map format and SAMtools. 742 Bioinformatics 2009, 25:2078-2079. 743 44. Trapnell C, Pachter L, Salzberg SL: TopHat: discovering splice junctions with RNA-Seq. 744 Bioinformatics 2009, 25:1105-1111. 745 45. Jin Y, Tam OH, Paniagua E, Hammell M: TEtranscripts: a package for including 746 transposable elements in differential expression analysis of RNA-seq datasets. 747 Bioinformatics 2015, 31:3593-3599. 748 749

750 Acknowledgement

751 We thank Thomas Jenuwein (Max Planck Institute) for providing Suv39h1/2 DKO mES and iMEF

752 cell lines, the staff of the Support Unit for Bio-Material Analysis (BMA) at RIKEN Center for Brain

753 Science (CBS) Research Resources Division (RRD) for DNA sequencing, and NGS library construction

754 with special thanks to K. Ohtawa. We also thank all of the Cellular Memory laboratory members

755 for helpful discussions.

756

757 Funding

758 This research was supported by KAKENHI (18H03991 and 18H05530) and a RIKEN internal research

759 fund to YS, and KAKENHI (19K16049) and the Special Postdoctoral Researcher (SPDR) Program of

760 RIKEN to KF.

761

762 Author contributions

763 Conceptualization: YS, KF. Cell culture and western blotting analysis: CS, KF. NGS analysis: KF, AT,

764 HM, HI. Mass spectrometry analysis: TS, ND. Paper writing: YS, KF, IH, TD.

765

766 Ethics approval and consent to participate

767 Not applicable.

768

769 Consent for publication

770 Not applicable.

771

772 Competing interests

773 Not applicable.

A B Fukuda et al. Fig. 1 RPKM in 80-kb bin 0.2 M M K K 1.2 P 0.2 P R R T 2 T 2 W e e - M m W m O K s 9 - 9 K P C 0.8 K O K R S 3 K 3 D0.0 2 E H 0.0 H /2 e m n 1 n 1 m i b i h 9 T e td e 9 K W c Se c 3 3 0.4 en en v H r r Su fe -0.2 fe if if -0.2 D D 0.0 -0.04 0.00 0.04 -0.04 0.00 0.04 -0.04 0.00 0.04 Compartment score Compartment score Compartment score C D Chr6: 118,773,618- 131,554,770 100.0 1.5 14.3 80.0 T 53.8 W o 1 WT mESCs 60.0 t % ve 40.0 85.7 ti a 0.5 el H3K9me2 domain 20.0 46.2 R 0.0 0 AB Setdb1 KO / WT Suv39h1/2 DKO / Compartment score WT Compartment A B Domain Non domain Compartment F E UNC0642-treated mESCs UNC0642-treated mESCs

1 M M C K K C N P P N R 2 R T U U 2 2 C + C e T e W N + m W m - O N O 9 - 9 O U T K U T K K K K + O 3 O 3 0 K W D + W D K /2 T D /2 /2 O O T /2 /2 H 1 H 1 W P 1 1 K K W 1 1 in b 0 in h L h h h h e td e 9 1 G 9 9 1 1 1 9 9 c e c v3 b / 3 3 b b b 3 3 n S n u td a v v td td td v v 2 re re S e 9 u u e e e u u T fe fe S G S S S S S S S T if if D D -1 -2 -0.04 0.00 0.04 -0.04 0.00 0.04 TT2 G Compartment score Compartment score Suv39h1/2 DKO Chr6: 77,857,971-105,719,664 Suv39h1/2 WT Setdb1 WT WT

Setdb1 KO G9a/GLP DKO Setdb1 KO + UNC Setdb1 WT+ Suv39h1/2 DKO + UNC UNC Suv39h1/2 WT + UNC Setdb1 KO+ UNC G9a/GLP DKO Suv39h1/2 WT+ Setdb1 WT + UNC UNC Suv39h1/2 DKO+ UNC Compartment score

H I J

G9a/GLP Setdb1 Setdb1 KO Suv39h1/2 Suv39h1/2 A compartment B compartment WT DKO t DKO WT +UNC u +UNC +UNC +UNC 100% 100% p t In en 80% 80% % tm 60% 60% 20 ar p 40% 40% m 0 o 20% 20% c Gapdh B comp B comp A 0% 0% region1 region2 t en WT WT + UNC tm ar p Setdb1 KO Setdb1 KO + UNC m o Suv39h1/2 DKO Suv39h1/2 DKO + UNC c Loss Conserve B TKO TKO + UNC bioRxiv preprint doi: https://doi.org/10.1101/2020.08.28.271221; this version posted August 28, 2020. 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.

Fukuda et al. Fig. 2

B A UNC0642-treated iMEFs and mESCs

Total length (Gb) Number (x1000) Average length (Kb) WT iMEF 60 UNC0642-treated 2 60 iMEF 1.5 40 UNC0642 treated 40 Setdb1 KO iMEF 1 20 20 Compartment 0.5 score (mESCs) 0 0 0 iMEFs mESCs iMEFs mESCs iMEFs mESCs

C D E UNC0642-treated iMEFs s 100% M M F K K E 38.2 % 4.1 % 19.2 P 30.5 % 7.8 % P iM R R 80% 2 0.4 2 T 0.4 e T e W m W 68.7 m - 9 – 60% 9 K d 3K O 3 e H K0.0 H at 0.0 40% 80.8 1 n e in b i tr e td ce 2- 20% c e n 4 31.3 n S e 6-0.4 re -0.4 r 0 21.2 % 36.5% e fe C 0% f 28.9 % 32.8% if N i A B D D U -0.04 0.00 0.04 Compartment -0.04 0.00 0.04 Compartment score Domain Non domain Compartment score

F G UNC0642-treated iMEFs 60 Number Total length 50 (x1000) (Gb) 30 1.5 40 t u 25 p n 30 20 I 1 % 15 20

10 0.5 10 5 0 0 0 Gapdh B comp region1 B comp region2 AB A B Compartment Compartment WT WT + UNC Setdb1 KO

WT Setdb1 KO iMEFs Setdb1 KO + UNC Suv39h1/2 DKO Suv39h1/2 DKO + UNC

TKO TKO + UNC bioRxiv preprint doi: https://doi.org/10.1101/2020.08.28.271221; this version posted August 28, 2020. 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.

Fukuda et al. Fig. 3 A B

UNC0642 Culture without treatment UNC0642

0h -3 days 0h 24h 32h 40h 48h 56h 64h 72h 24h H3K9me2 ChIP-seq 32h 40h

C PCA analysis of H3K9me2 RPKM in 80-kb bin 48h 56h 0.5 56h 64h 72h

64h TT2 48h 72h ) 40h % DEF 5 .3 90.0 ( 0.5 <=H3K9me2 in WT <= 0.7 25 2 C 70 h P 32h 1.2 ) it ) 24h 0h th % % 1 i ( w ( w 60 g 2 20 0.8 g e2 in e in m p m 0.6 p 9 50 p 9 e p K la K r 0.4 la 3 er 3 15 co r H v H s 0.2 ve t 40 o t Z- o n s n s e n e -0.5 0 n d 30 io d 10 o n g en -0.2 gi e e p e ep r e -0.4 r d 20 f d f n o - -0.6 o i n 1 5 WT n P io b o L 10 t td 0h 32h 48h 72h TT2 ti G ac e -0.5 0.0 0.5 c a/ r S ra 9 0 F 0 PC1 (82.43%) F G early middle early middle late early middle late

late bioRxiv preprint doi: https://doi.org/10.1101/2020.08.28.271221; this version posted August 28, 2020. 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.

Fukuda et al. Fig. 4 A B G9a/GLP-independent H3K9me2 Up in Setdb1 KO Up in UNC Up in Setdb1 KO+UNC within 5kb from TSS ) UNC UNC UNC 2 Setdb1 KO + UNC Setdb1 KO + UNC Setdb1 KO + UNC 100 e 9m 3K 80 (H M P R 60 % -10kb -5kb TSS 5kb 10kb -10kb -5kb TSS 5kb 10kb -10kb -5kb TSS 5kb 10kb 40 C Up in Setdb1 KO Up in UNC 20 Up in Setdb1 KO+UNC 3) WT WT WT e Setdb1 KO Setdb1 KO Setdb1 KO 0 9m Up in Up in Up in 3K Setdb1 UNC Setdb1 (H KO KO+UNC M P R H3K9me2 H3K9me3 -10kb -5kb TSS 5kb 10kb -10kb -5kb TSS 5kb 10kb -10kb -5kb TSS 5kb 10kb

D Genes with Setdb1-dependent E H3K9me2 and without H3K9me3 RNA H3K9me2 H3K9me3 (fold change) (RPKM) (RPKM) C C C RNA H3K9me2 H3K9me3 N N N U U U 0.8 0.8 + + + O O O O O O WT+UNC K K K K K K WT 1 1 1 1 1 1 8 Setdb1 KO b b b b b b Setdb1 KO C d d C d d C d d ) +UNC tr N t t tr N t t tr N t t C C U Se Se C U Se Se C U Se Se (F 2 0.4 0.4 g 4 Lo

0.0 0 0.0 -5kb-10kb 5’ 3’ 5kb 10kb -5kb-10kb 5’ 3’ 5kb 10kb Ctr UNC Setdb1 Setdb1 KO KO+UNC

F Nlrp4c Gm4971 Trpd52l3 1200 2 Fold change RPKM

UNC H3K9me2 Setdb1KO+UNC Ctr UNC H3K9me3 Setdb1 KO Setdb1 KO+UNC Ctr UNC RNA Setdb1 KO Setdb1 KO+UNC

Nlrp4c Gm4971 Trpd52I3

0.8 0.3 0.03 rt p 0.6 H 0.2 0.02 to 0.4 0.1 0.01 ve 0.2 ti a 0 0 0 el R bioRxiv preprint doi: https://doi.org/10.1101/2020.08.28.271221; this version posted August 28, 2020. 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.

Fukuda et al. Fig. 5

A B

100% 80% Setdb1 WT 60% Setdb1 WT +UNC 40% 20% Setdb1 KO 0% Setdb1 KO +UNC Suv39h1/2 WT Suv39h1/2 WT +UNC Suv39h1/2 DKO Suv39h1/2 DKO +UNC

B to B A to A A to B B to A

CD UNC0642-treated mESCs – WT mESCs UNC0642-treated Setdb1 KO mESCs – UNC0642-treated WT mESCs

M M K K P P 2 R 2 R 2 e e2 m m 9 9 K K 3 3 H H n i in ce e n c e en r 0 r 0 fe fe if if D D

~-0.02 -0.02~-0.01 -0.01~0.01 0.01~0.02 0.02~ ~-0.02 -0.02~-0.01 -0.01~0.01 0.01~0.02 0.02~ Difference in compartment score Difference in compartment score

E P = 4.1x10-8 F 100%

re 80% o c ) 0.01 s T t W en – ” 60% tm C ar N p U m +0.00 40% co O K in 1 e b c d 20% n et re S fe (“-0.01 if D 0% Random Upregulated in Setdb1 KO A to A B to A B to B + UNC bioRxiv preprint doi: https://doi.org/10.1101/2020.08.28.271221; this version posted August 28, 2020. 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.

Fukuda et al. Fig. 6

H3K9me2 profile in each condition H3K9me2 profile during recovery H3K9me2 A compartment WT B compartment Pericentromere G9a/GLP DKO Time Setdb1 KO + UNC

Suv39h1/2 DKO + UNC

TKO + UNC

3D genome organization, H3K9me2 and transcription

WT Setdb1 KO + UNC Nuclear membrane

B B

Silenced gene

Activated gene A A H3K9me2