Neuronal identity is maintained in the adult brain through KAT3-dependent enhancer acetylation

Very little is known about the mechanisms responsible for maintaining cell identity in mature tissues. The paralogous type 3 lysine acetyltransferases (KAT3) CBP and p300 are both essential during development, but their specific functions in nondividing differentiated cells remains unclear. Here, we show that when both proteins are simultaneously knocked-out in excitatory neurons of the adult brain, the mice express a rapidly progressing neurological phenotype associated with reduced dendritic complexity and electrical activity, the transcriptional shutdown of neuronal genes, and a dramatic loss of H3K27 acetylation and pro-neural transcription factor binding at neuronal enhancers. The neurons lacking both KAT3 rapidly acquire a molecularly undefined fate with no sign of dedifferentiation, transdifferentiation or death. Restoring CBP expression or lysine acetylation reestablished neuronal-specific transcription. Our experiments demonstrate that KAT3 proteins act as fate-keepers in excitatory neurons and other cell types by jointly safeguarding chromatin acetylation levels at cell type-specific enhancers throughout life.

To elucidate the neuronal roles of CBP and p300 in the adult brain, we 95 selectively eliminated CBP, p300, or both KAT3 proteins in forebrain 96 7 Lipinski et al. 11/11/2019 simultaneous loss of CBP and p300 alters neuronal morphology and impairs 146 electrical responses leading to dramatic neurological defects. 147 Strikingly, these severe changes occur in the absence of cell death. 148 TUNEL staining (Fig. S3a) and immunostaining against active caspase 3 149 ( Fig. 1n) and the H2A.X variant (Fig. S3b), which respectively label cells 150 undergoing apoptosis and suffering DNA damage, were all negative. EM 151 images also show a largely normal stratum pyramidale in which neuronal 152 nuclei did not present apoptotic bodies, although the nucleoplasm appeared 153 clearer in dKAT3-ifKOs than in control littermates ( Fig. S3c-d). To monitor the 154 evolution of double knockout neurons beyond the limit imposed by the survival 155 of dKAT3-ifKO's, we infected the CA1 of the Crebbp f/f ::Ep300 f/f (dKAT3-floxed) 156 mice with adeno-associated virus (AAV) expressing Cre recombinase under 157 the human synapsin promoter (Fig. S3e). Immunostainings confirmed the 158 efficient and complete elimination of CBP and p300 in granular neurons (Fig. 159 S3f) in the absence of detectable cell death even two months after gene 160 ablation (Fig. S3g). 161

Maintenance of neuronal identity requires at least one KAT3 gene 162
To determine the molecular basis of the abovementioned phenotypes, we 163 conducted a RNA-seq screen in the hippocampus of dKAT3-ifKOs and control 164 littermates. Differential gene expression profiling revealed 1,952 differentially 165 expressed genes (DEGs) in dKAT3-ifKOs, with a clear preponderance both in 166 number and magnitude of gene downregulations ( Fig. 2a-b, S4a and Table  167 S2). Gene Ontology (GO) enrichment analysis indicated that these 168 downregulations affect a large number of neuronal functions (Fig. 2c,  neuronal firing and lack of electrical responses. Gene upregulation was much 173 more restricted, including a modest inflammatory signature (Fig. 2c, red bars) 174 but no activation of cell death pathways (Fig. S4b). In fact, several positive 175 regulators of neuronal death were strongly downregulated in dKAT3-ifKOs 176 (e.g., Hrk; Fig. S4c). Consistent with the survival of these cells, housekeeping 177 genes remained largely unchanged ( Fig. 2b and S4d). Immunodetection 178 experiments for neuronal proteins like CaMKIV, NeuN and hippocalcin 179 confirmed the dramatic loss of expression of neuronal proteins ( Fig. 2e and  180 S4e). Notably, the loss of neuronal markers expression was not detected in 181 mice bearing a single functional KAT3 allele (Fig. S4f), indicating that this 182 minimal gene dose is sufficient to preserve the active status of neuronal loci. 183 To determine if these deficits were cell-autonomous, we analyzed the 184 hippocampus of dKAT3-floxed mice monolaterally infected with Cre-185 recombinase-expressing AAVs. We observed a dramatic loss of neuronal 186 marker expression only in the transduced hippocampus (Fig. 2f). Moreover, 187 the cells depleted of KAT3 proteins completely failed to respond to kainic acid, 188 a strong agonist of glutamate receptors, further confirming the loss of 189 excitatory neuron properties (Fig. S5). Similarly,experiments in primary 190 neuronal cultures (PNCs) produced from the hippocampi of Crebbp f/f ::Ep300 f/f 191 embryos infected with a Cre-recombinase-expressing lentivirus (Fig. 2g) 192 demonstrate that the cultured neurons do not die after simultaneous 193 elimination of CBP and p300 ( Fig. S6a-b (Fig. 2h), downregulated neuron-specific transcripts (Fig. 2i) and 195 proteins (Fig. S6c) and severe hypoacetylation of KAT3 targets (Fig. S6d). 196 Cells lacking KAT3 proteins acquire a novel, molecularly undefined fate 197 The altered morphology, electrophysiological properties and gene expression 198 all suggest that the excitatory neurons rapidly lose their identity after the 199 elimination of both KAT3 genes. To tackle this hypothesis, we compared the 200 set of DEGs in dKAT3-ifKOs with transcriptome information for the different 201 cell types in the adult mouse brain using single-cell RNA-seq (scRNA-seq) 202 data from the mouse cortex 21 (Fig. 3a). Our analysis revealed that the genes 203 typically expressed in CA1 and S1 pyramidal neurons were significantly 204 downregulated in the hippocampus of dKAT3-ifKOs, whereas other cell-type 205 specific transcriptional programs were unaffected except for a modest 206 increase of microglia genes related to inflammation. Importantly, although 207 identity loss is often associated with dedifferentiation (i.e., the regression to an 208 earlier stage of differentiation), we did not detect an upregulation of stemness 209 genes 22, 23 , nor neuronal stem cell (NSC)-or neuroprogenitor (NPC)-specific 210 gene expression 24 ( Fig. 3a and S7a-b). Trans-differentiation can be also 211 discarded because we did not detect the upregulation of the transcriptional 212 signatures of other brain cell-types. 213 To determine more precisely the fate of excitatory neurons after losing 214 their identity, we conducted single-nucleus RNA-seq analyses 2 and 5 weeks 215 after TMX treatment ( Fig. 3b-c, S7c and Table S3). We observed a 216 progressive confluence of CA1/CA3 pyramidal neurons and dentate gyrus 217 granule neurons in a common cell cluster depleted of neuronal type-specific 218 markers ( Fig. S7d-g) and in which no other distinctive marker appears ( Fig.  219 10 Lipinski et al.

11/11/2019
3d-e and S7h-i). Altogether, these results indicate that the principal neurons 220 of dKAT3-KOs lose their neuronal identity, but do not die, dedifferentiate or 221 transdifferentiate to other cell types. Instead, these cells seem to be trapped 222 in a non-functional interstate deadlock. 223 CBP and p300 bind to the same regulatory regions 224 The evidence above indicates that KAT3 proteins play a redundant role 225 preserving neuronal identity. To explore the basis of such redundancy, we 226 next mapped the occupancy of hippocampal chromatin by CBP and p300. 227 Chromatin immunoprecipitation followed by whole-genome sequencing (ChIP-228 seq) using antibodies that differentiate between the two paralogous proteins 229 ( Fig. S1b and S8a) retrieved 37,359 peaks in the chromatin of wild type mice 230 ( Fig. 4a, S8b and Table S4). After correcting for the different efficiencies of 231 the two antibodies, we detected an almost complete overlap between the CBP 232 and p300 peaks throughout the whole genome ( Fig. 4b and S8c). This 233 finding, supported for the combined analysis of genome occupancy by CBP 234 and p300 in wild type animals and single and double ifKOs, reveals for the 235 first time that these two essential epigenetic enzymes occupy basically the 236 same sites in neuronal chromatin, and underscores the large functional 237 redundancy of KAT3 proteins. 238 As both proteins are ubiquitously expressed yet gene ablation only 239 takes place in excitatory neurons, the signal detected in the hippocampal 240 chromatin of dKAT3-ifKOs must correspond to CBP/p300 binding in other cell 241 types. We combined this information with ATAC-seq (a technique that 242 investigates chromatin occupancy and requires much lower input than ChIP-243 seq) in sorted NeuN + hippocampal nuclei 25 to discriminate between neuronal-244 11 Lipinski et al. 11/11/2019 specific, non-neuronal-cell-specific and "pancellular" KAT3 binding ( Fig. 4c-

d, 245
S9a and Table S4). Intriguingly, pancellular KAT3 peaks are found in the 246 promoter of genes involved in basic cellular functions, such as RNA 247 processing and metabolism, while cell-type-specific peaks (neuronal and non-248 neuronal) primarily locate at introns and intergenic regions with enhancer 249 features ( Fig. S9b) and associate with cell-type-specific processes ( Fig. 4e-f  250 and Table S5). We used binding and expression target analysis (BETA) 26 to 251 integrate ChIP-seq and RNA-seq data (Fig. 4g), and found that the loss of 252 neuronal KAT3 binding (~14,000 peaks) is an excellent predictor of 253 transcriptional downregulation (p = 3.3x10 -72 , Fig. 4h). Up to 74% of the 254 downregulated genes in dKAT3-ifKOs are linked to the loss of KAT3s at 255 proximal regulatory regions. Comparison of the ATAC-seq accessibility 256 profiles of control and dKAT3-ifKO neuronal nuclei retrieved more than 6,000 257 differentially accessible regions (DARs) ( Table S6). Most of these DARs 258 displayed reduced accessibility in dKAT3-ifKO neurons (Fig. S9c), were 259 located at enhancers (Fig. S9d) and coincided with the downregulation of the 260 proximal gene (Fig. S9e). 261

KAT3 proteins control the acetylation levels of neuron-specific enhancers 262
The acetylation of histone H3 at lysine 27 (H3K27ac) is a likely mechanism for 263 mediating the role of KAT3 proteins in maintaining neuronal identity. This 264 histone modification is enriched in active enhancers and its levels correlate 265 with tissue specification 27 . In agreement with a recent acetylome analysis 11 , 266 immunostaining against specific lysine residues in the histone tails revealed 267 their variable dependence on CBP/p300 (Fig. S10a) as well as the particular 268 sensitivity of H3K27ac to the loss of CBP and p300 (Fig. 5a). The dramatic 12 Lipinski et al.

11/11/2019
reduction in H3K27ac is not accompanied by an increase in the signal or 270 changes in the distribution of H3K27me3 (Fig. S10b). H3K9me3, a histone 271 modification associated with heterochromatin, also appears unaffected (Fig.  272

S10c). 273
We next investigated the genomic distribution of H3K27ac in the 274 hippocampal chromatin of dKAT3-ifKOs and control littermates. We retrieved 275 37,732 H3K27ac-enriched regions (Fig. 5b) that largely overlap with KAT3 276 peaks (Fig. S10d). More than one third of these H3K27ac peaks were 277 strongly reduced in dKAT3-ifKOs (Table S7), particularly those that overlap 278 with neuronal KAT3 binding (~80%). In contrast, less than 5% of pancellular 279 and non-neuronal H3K27ac-enriched regions were affected (Fig. 5b). To 280 explore in greater detail these differences, we classified neuronal KAT3 peaks 281 into promoters and enhancers based on their location and H3K4me1/me3 282 content (Fig. 5c). In dKAT3-ifKOs, KAT3 binding, H3K27ac and ATAC-seq 283 signals were all strongly reduced in neuronal enhancers ( Fig. 5d and S10e), 284 while the promoters associated with pancellular-KAT3 peaks were spared, 285 indicating that other KATs maintain the acetylation of these loci. Consistent 286 with this view, the acetylation of H3K9,14 that decorates promoters only 287 showed a modest reduction in dKAT3-ifKOs (Fig. S10f), indicating that other 288 KATs are responsible for maintaining this form of acetylation at promoters. 289 We also observed a strong correlation between the loss of H3K27ac and 290 transcript downregulation. Of the 1,376 downregulated genes in dKAT3-ifKOs, 291 78% showed a strong reduction in H3K27ac. Reciprocally, 74% of the genes 292 with reduced acetylation were neuronal genes severely downregulated in 293 dKAT3-ifKOs. 294 13 Lipinski et al.

11/11/2019
Super-enhancers constitute a special type of regulatory region that 295 results from a cluster of several enhancers bound by master regulators and 296 the Mediator complex and plays a critical role in controlling cell identity 28 . 297 Their conspicuous features include the association with highly transcribed 298 genes, broad domains of H3K27 acetylation, and a high density of TF binding 299 sites. To identify putative neural super-enhancers, we fused together 300 enhancers that are closer than 5 kb from each other ( Table S8). The retrieved 301 super-enhancers were associated with highly expressed genes in 302 hippocampal neurons that are downregulated in dKAT3-ifKOs (Fig. 5e). 303 Furthermore, most of these super-enhancers-associated genes were strongly 304 hypoacetylated in dKAT3-ifKOs ( Fig. 5f) and encode bona fide neuronal 305 regulators related with synaptic transmission and neuronal plasticity (Fig. 5g). 306 Neuronal fate also relies on changes in chromatin architecture during neural 307 differentiation 29 . In agreement with this view, we found that the loss of 308 chromatin accessibility in dKAT3-ifKOs overlapped with the establishment of 309 cell type-specific enhancer-promoter contacts during neural differentiation, 310 which is concomitant with the activation of neuron-specific transcription ( Fig.  311 S10g-h). Together, these results show that KAT3 proteins control the status 312 of neuron-specific genes by regulating lysine acetylation levels and chromatin 313 interactions at enhancers and super-enhancers. 314 bHLH TFs drive KAT3 binding to neuron-specific genes and enhancers 315 Since KAT3 proteins do not directly bind to DNA, we next asked which 316 proteins are responsible for recruiting CBP and p300 to these neuron-specific 317 regulatory regions. Motif prediction analysis of KAT3 binding peaks revealed 318 remarkable differences between pancellular and neuronal-specific regions. 319 14 Lipinski et al.

11/11/2019
Pancellular KAT3 binding is associated with general TFs such as Sp, Fox and 320 Ets, all of which have been reported to bind to CBP or p300 10 (Fig. 5h). In 321 contrast, neuronal KAT3 peaks presented a very prominent enrichment (E-322 value = 1.0 x 10 -1223 ) for the DNA binding motif of basic helix-loop-helix (bHLH) 323 proteins (Fig. 5h). The regions losing occupancy in dKAT3-ifKOs were also 324 highly enriched for bHLH-recognized motifs ( Fig. 5i and S11a). 325 Among the bHLH proteins there are pro-neural TFs critically involved in 326 neuronal development 30, 31 (e.g., Ascl1, Neurod1-6, Neurogenin1-3 and 327 Atoh1). Our differential expression analysis retrieved 20 bHLH encoding 328 genes that are downregulated in the hippocampus of dKAT3-ifKOs ( Fig. S11b  329 and Table S2). In fact, the NeuroD2 and NeuroD6 genes, which encode TFs 330 with putative terminal selector activity 5, 32 , were among the most 331 downregulated genes in the hippocampi of dKAT3-ifKOs (Fig. 5j). These 332 results suggest that the change in accessibility is a likely consequence of the 333 loss of both the KAT3 proteins and the recruiting TFs. To directly assess the 334 occupancy of these sites before and after the ablation of KAT3s, we analyzed 335 their digital footprints in DARs and detected robust differences for pro-neural 336 bHLH TFs ( Fig. 5k and S11c), confirming that some changes in accessibility 337 reflect the reduced binding of one of several member of this family. This result 338 is consistent with the large overlap between the bHLH footprints detected in 339 the ATAC-seq profiles and the Neurod2 ChIP-seq data 33 (Fig. S11d). Overall,340 these experiments indicate that CBP and p300 interact with bHLH proneural 341 TFs in neuronal-specific genomic locations to maintain neuron-specific 342 transcription ( Fig. 5l and S11e show some representative loci). 343 Other cell types also require KAT3 protein for maintaining their identity 344 11/11/2019 Previous evidence indicates that KAT3 proteins play a critical role in the 345 differentiation of other cell types and the establishment of cell type-specific 346 transcription 34,35,36,37 . To investigate if cell fate maintenance in other cell 347 types also requires the KAT3 proteins, we next examined astrocytes derived 348 from dKAT3-floxed mice (Fig. 6a). We prepared astrocyte cultures from the 349 hippocampi of Crebbp f/f ::Ep300 f/f embryos and infected them with a Cre-350 recombinase-expressing lentivirus. Similar to our results in neurons, primary 351 cultures of cortical astrocytes missing both KAT3 proteins lose both the 352 expression of glial genes such as Gfap (Fig. 6b-c) and their characteristic 353 morphology (Fig. 6d), indicating that KAT3 proteins are also responsible for 354 identity maintenance in other cell types. 355 To examine this possibility in cells from other lineages, we investigated 356 the profiles of p300 binding in different tissues available at ENCODE (there 357 are no tissue-specific profiles for CBP). We found that, similarly to our 358 observations in brain tissue, p300 peaks were associated with genesets 359 highly enriched in tissue-specific functions (Fig. 6e). Furthermore, in each 360 tissue, p300-bound regulatory regions were enriched for distinct binding 361 motifs (Fig. 6f). Profuse and tissue-specific p300 binding was observed into 362 and upstream of loci presenting tissue-specific transcription (Fig. 6g) To explore the specific contribution of the scaffolding and KAT activities of 371 KAT3 proteins to the loss of cell type-specific transcription, we turned to 372 PNCs. Importantly, transfection of a heterologous full-length CBP in neurons 373 that have loss the endogenous expression of both CBP and p300 (Fig 7a-b) 374 fully prevented the rapid loss of expression and neuronal markers (Fig 7c-d). 375 We next examined whether the expression of the N-terminus or the C-376 terminus (bearing the KAT domain) halves of CBP, as well as the full-length 377 reconstituted protein (Fig. 7e-f) could also rescue the transcriptional 378 impairment. The specifically assess the contribution of KAT activity, we also 379 assessed a variant of the C-terminus half of CBP (referred to as KATmut) 380 bearing the R1378P mutation linked to RSTS 39 . We found that only the 381 reconstruction of full-length CBP with an intact KAT domain prevented the 382 downregulation of neuronal markers (Fig. 7g-h). These results indicate that 383 both activities of CBP, as KAT and molecular scaffold, are necessary to 384 preserve locus activity. We also investigated whether heterologous NeuroD2 385 expression exerted the same protection. This bHLH TF is highly expressed in 386 mature excitatory neurons, regulates its own expression and holds the 387 features of a neuronal terminal selector 5 . Furthermore, it is strongly 388 downregulated in dKAT3-ifKOs neurons ( Fig. 5j and S11b), and its 389 occupancy profile in cortical chromatin 33 shows a large overlap with DARs in 390 dKAT3-ifKOs ( Fig. S12a-b). However, conversely to CBP, NeuroD2 alone 391 was not sufficient to re-establish neuronal-specific transcription in dKAT3-KO 392 PNCs (Fig. S12c). 393 17 Lipinski et al.

11/11/2019
Locus-specific epi-editing rescue transcriptional impairments 394 Next, we examined whether increasing lysine acetylation is sufficient to 395 rescue the transcriptional deficit. To this end, we took advantage of recently 396 developed tools for epi-editing based on the expression of an inactive Cas9 397 enzyme (dCas9) fused to the KAT domain of p300 (dCas9-KAT) 40 . This 398 system enables targeting specific genomic regions for p300-dependent 399 acetylation using adequate guide RNAs (gRNA). We selected Neurod2 as the 400 target gene because it is both severely downregulated and H3K27-401 deacetylated in dKAT3-ifKOs (Fig. 8a). The infection of PNCs with lentiviruses 402 that express dCas9-KAT and a gRNA that recruits this chimeric protein to the 403 most proximal KAT3 peak of Neurod2 ( Fig. 8a-b) prevented the 404 downregulation of NeuroD2 that was observed in dKAT3-KO neurons (Fig.  405   8c-d). Moreover, co-transfection of plasmids carrying dCas9-KAT and the 406 Neurod2 gRNA in dKAT3-KO cells that had already lost Neurod2 expression 407 also caused a significant recovery of the expression of this gene ( Fig. 8e-g). 408 These experiments demonstrate that the increase of KAT activity at the locus 409 preserves and can even restitute its functionality, suggesting that 410 histone/lysine deacetylation at regulatory regions is the main cause for the 411 downregulation of neuronal genes after KAT3 loss. 412 413

Discussion 414
From gate-to fate-keepers 415 The differentiation of neuronal lineages during brain development requires the 416 participation of TFs and chromatin-modifying enzymes such as CBP and p300 417 6,12,41,42,43,44,45,46  genes present a dramatic downregulation that is accompanied by the loss of 438 CBP/p300 binding and the H3K27 deacetylation of the locus (Fig. S11e, left). 439 Together these results suggest that KAT3 proteins are essential to safeguard 440 cell identity but also to activate any alternative cell fate, including stemness 441 and programmed cell death.

11/11/2019
A conserved role in cell fate maintenance 494 Importantly, the role of KAT3 proteins as fate-keepers is not restricted to 495 neurons. Our experiments in astrocytes, analysis of KAT3 peaks in different 496 tissues and previous observations in other cell types demonstrating that the 497 combined loss of CBP and p300 causes impaired cell type-specific 498 transcription 34,35,36,37 , all support this view. What may differ between cell 499 types is the specific set of TFs that recruit the KAT3 proteins to cell type-500 specific loci (the bHLH and GATA family of TFs in the tissues explored here). 501 Further studies should be aimed at identifying the cell type-specific partners 502 and targets governing tissue specification and signaling. 503 The role of KAT3 proteins in preserving cell type-specific gene statuses 504 can be particularly relevant for neurons given their tremendous diversity, 505 elaborate connectivity patterns and long lifespan. The critical importance of 506 KAT3s for brain function may explain the duplication of the ancestral KAT3 507 gene in the first vertebrates coinciding with the emergence of a neural crest 508 and cephalization 12, 70 . Although the two KAT3 proteins have evolved some 509 individual functions in postmitotic cells (e.g., forebrain restricted CBP 510 knockouts show phenotypes related to cognitive dysfunction 67 and p300 511 seems to play a prominent role in muscle biology 71 ), our study demonstrates 512 that they still share a joint and more essential role preserving epigenetic 513 identity.