Paupar LncRNA Promotes KAP1 Dependent Chromatin Changes And Regulates Subventricular Zone Neurogenesis

Many long non-coding RNAs (lncRNAs) are expressed during central nervous system (CNS) development, yet their in vivo roles and molecular mechanisms of action remain poorly understood. Paupar, a CNS expressed lncRNA, controls neuroblastoma cell growth by binding and modulating the activity of genome-wide transcriptional regulatory elements. We show here that Paupar transcript directly binds KAP1, an essential epigenetic regulatory protein, and thereby regulates the expression of shared target genes important for proliferation and neuronal differentiation. Paupar promotes KAP1 chromatin occupancy and H3K9me3 deposition at a subset of distal targets, through formation of a DNA binding ribonucleoprotein complex containing Paupar, KAP1 and the PAX6 transcription factor. Paupar-KAP1 genome-wide co-occupancy reveals a 4-fold enrichment of overlap between Paupar and KAP1 bound sequences. Furthermore, both Paupar and Kap1 loss of function in vivo accelerates lineage progression in the mouse postnatal subventricular zone (SVZ) stem cell niche and disrupts olfactory bulb neurogenesis. These observations provide important conceptual insights into the trans-acting modes of lncRNA-mediated epigenetic regulation, the mechanisms of KAP1 genomic recruitment and identify Paupar and Kap1 as regulators of SVZ neurogenesis.


INTRODUCTION 38
A subset of nuclear long noncoding RNAs (lncRNAs) have been shown to act as transcription and 39 chromatin regulators using multiple different regulatory mechanisms. These include local functions 40 close to the sites of lncRNA synthesis (Engreitz,  LncRNAs show a high propensity to be expressed in brain nuclei and cell types relative to other tissues 57 (Mercer, Dinger et al., 2008, Mercer, Qureshi et al., 2010, Ponjavic, Oliver et al., 2009). The adult 58 neurogenic stem cell-containing mouse subventricular zone (SVZ) contributes to brain repair and can 59 be stimulated to limit damage, but is also a source of tumours ( Paupar and KAP1 functionally interact to control gene expression we first tested whether they 155 regulate a common set of target genes. We depleted Kap1 expression in N2A cells using shRNA 156 transfection and achieved approximately 90% reduction in both protein (Fig 2a) and transcript (Fig  157   2b) levels. Paupar levels do not change upon KAP1 knockdown indicating that KAP1 dependent 158 changes in gene expression are not due to regulation of Paupar expression (Fig 2b). Transcriptome 159 profiling using microarrays then identified 1,913 differentially expressed genes whose expression 160 significantly changed (at a 5% false discovery rate [FDR]) greater than 1.4-fold (log2 fold change ≈ 161 0.5) upon KAP1 depletion (Fig 2c and Supplemental Table S2). 282 of these genes were previously 162 identified to be regulated by human KAP1 in Ntera2 undifferentiated human neural progenitor cells 163 . Transient reduction of Kap1 expression by approximately 55% using a second 164 shRNA expression vector (Kap1 shB) also induced expression changes for 7 out of 8 Kap1 target 165 genes with known functions in neuronal cells that were identified in the microarray (Supplemental 166 Fig 1b). These data further validate the specificity of the KAP1 regulated gene set. 167 We previously showed that Paupar knockdown induces changes in the expression of 942 genes in 168 N2A cells . Examination of the intersection of KAP1 and Paupar transcriptional 169 targets identified 244 genes whose levels are affected by both Paupar and KAP1 knockdown in this 170 cell type (Fig 2d and Supplemental Table S3). This represents a significant 3.6-fold enrichment over 171 the number expected by random sampling and is not due to co-regulation because Kap1 is not a 172 Paupar target . A large majority (87%; 212/244) of these common targets are 173 positively regulated by Paupar and for two-thirds of these genes (161/244) their expression changes 174 in the same direction upon Paupar or KAP1 knockdown (Fig 2e). Furthermore, Gene Ontology 175 enrichment analysis of these 244 genes showed that Paupar and KAP1 both regulate a shared set of 176 target genes enriched for regulators of interphase, components of receptor tyrosine kinase signalling pathways as well as genes involved in nervous system development and essential neuronal cell 178 functions such as synaptic transmission (Fig 2f). Genes targeted by both Paupar and KAP1 are thus 179 expected to contribute to the control of neural stem-cell self-renewal and neural differentiation. 180 Paupar, KAP1 and PAX6 associate on chromatin within the regulatory region of shared target genes 181 In order to investigate Paupar mediated mechanisms of distal gene regulation we next sought to 182 determine whether Paupar, KAP1 and PAX6 can form a ternary complex on chromatin within the 183 regulatory regions of their shared target genes. To do this, we first integrated our analysis of PAX6 184 regulated gene expression programmes in N2A cells  and identified 87 of the 244 185 Paupar and KAP1 common targets, which is 35.8-fold greater than expected by random sampling, 186 whose expression is also controlled by PAX6 (Fig 3a and Supplemental Table S3). We found that 34 of 187 these genes contain a CHART-Seq mapped Paupar binding site within their GREAT defined putative 188 regulatory regions (Vance, 2016 and predicted that these represent functional 189 Paupar binding events within close genomic proximity to direct transcriptional target genes (Fig 3a  190 and Supplemental Table S3). 191 ChIP-qPCR analysis previously identified four of these Paupar bound locations within the regulatory 192 regions of the Mab21L2, Mst1, E2f2 and Igfbp5 genes that are also bound by PAX6 in N2A cells 193 . We therefore measured KAP1 chromatin occupancy at these regions as well as 194 at a negative control sequence within the first intron of E2f2 using ChIP and identified a specific 195 enrichment of KAP1 chromatin association at the Mab21L2, Mst1, E2f2 and Igfbp5 genes compared 196 to an IgG isotype control (Fig 3b). KAP1 binding to these regions is only 2-to 4-fold reduced 197 compared to the Zfp382 3' UTR positive control (Fig. 3b)  Paupar-KAP1-PAX6 bound regions upon Paupar depletion and that the extent of KAP1 chromatin 217 association appears to be dependent on Paupar transcript levels (Fig. 3d). KAP1 chromatin association 218 is also not reduced at the Ezh2 gene Paupar-KAP1 binding site or at the control sequence that is not 219 bound by Paupar (Fig 3d), whilst total KAP1 protein levels do not detectably change upon Paupar 220 knockdown (Fig 3e), further confirming specificity. 221 These results imply that Paupar functions to promote KAP1 chromatin association at a subset of its 222 genomic binding sites in trans and that this requires the formation of a DNA bound ternary complex 223 containing Paupar, KAP1 and PAX6. Consistent with this, co-expression of the Paupar lncRNA 224 promotes KAP1-PAX6 association in a dose dependent manner in an immunoprecipitation 225 experiment (Fig 3f). This effect is specific for the Paupar transcript because expression of a size- using two different shRNAs resulted in a significant decrease in histone H3K9me3 modification at 3 of 238 4 of these shared binding sites tested using ChIP (Fig. 3g, h). No change in histone H3K9me3 was 239 detected at Ezh2 gene whose expression does not change upon PAX6 depletion. Together, these data 240 show that Paupar functions to modulate KAP1 chromatin association and histone H3K9me3 241 deposition at a subset of its shared binding sites in trans. 242

Paupar co-occupies an enriched subset of KAP1 binding sites genome-wide 243
We next examined the intersection between Paupar and KAP1 bound locations genome-wide in 244 order to generate a more comprehensive view of the potential of Paupar for regulating KAP1 245 function. ChIP-seq profiling of KAP1 chromatin occupancy showed that KAP1 associates with 5510 246 genomic locations compared to input DNA in N2A cells (1% FDR) (Supplemental Table S4). only one of which is located within the 3'UTR of a ZNF gene (zfp68) (Supplemental Table S4). Notably, 253 this represents a significant (p < 0.001) 4-fold enrichment of Paupar and KAP1 co-occupied locations 254 as estimated using Genome Association Tester (GAT) (Fig. 4b). In addition, plotting the distribution of 255 peak intensities across these co-occupied regions revealed a precise coincidence of Paupar and KAP1 256 binding (Fig. 4c). These data therefore show that Paupar co-occupies an enriched subset of KAP1 257 bound sequences genome-wide and suggest that Paupar mediated genomic recruitment of KAP1 258 may involve interactions with other transcription factors in addition to KRAB-ZNF association. 259

Paupar and Kap1 regulate the SVZ neurogenic niche and olfactory bulb neurogenesis 260
Our results indicate that Paupar and KAP1 regulate the expression of shared target genes important 261 for proliferation and neuronal differentiation in N2A cells. We next expanded this observation and 262 tested whether Paupar and Kap1 can regulate the same neurodevelopmental process in vivo. To do 263 this, we used the mouse SVZ system as it is experimentally convenient for discovering many different We first showed using RT-qPCR that Paupar is expressed in the SVZ, as well as in neurospheres 269 cultured from P4 SVZ (Supplemental Fig S3a), and then confirmed the efficiency of the Paupar 270 targeting shRNA expression vectors to deplete Paupar transcript in neurospheres cultured from P4 SVZ 271 (Supplemental Fig S3b). sh165 caused robust Paupar knockdown (KD) whereas sh408 moderately 272 reduced Paupar expression enabling us to identify dose dependent regulatory effects. Nucleofection 273 of Paupar KD constructs and a scrambled (scr) control plasmid targeted ~60% of cells, as measured 274 using GFP, but we determined Paupar levels in all cells. Thus on a cell-by-cell basis the relative level of 275 knockdown is predicted to be greater than shown (Supplemental Fig S3b). To study Paupar function in 13 neurogenesis, we electroporated P1 pups with Paupar KD constructs or scr controls and examined the 277 SVZ 24 hours post electroporation (24hpe) and 3 days post electroporation (3dpe). To control for 278 differences in the number of cells electroporated in the different groups we measured the percentage 279 of GFP+ cells expressing lineage markers (Fig. 5a, c). Immunostaining showed that at 24hpe, the 280 percent of GFP+ cells expressing the TAP marker MASH1 was increased by more than 50% with sh165 281 knockdown (Fig. 5b). This was confirmed by immunostaining with the TAP and neuroblast marker DLX2 282 which showed a greater than 30% increase with both knockdown constructs (Fig. 5b). Additionally we 283 showed that the percentage of GFP+ cells positive for the proliferation marker Ki67 was significantly 284 increased in the sh165 group (Fig. 5b). At 24hpe the majority of cells in scramble controls are radial 285 glia-like neural stem cells (Boutin et al., 2008a, Chesler et al., 2008. These results thus suggest that 286 after Paupar KD a larger percentage of cells are progressing into the next phase of the SVZ lineage and 287 are actively proliferating. We next carried out immunohistochemistry for the same markers at 3dpe 288 and quantification showed that fewer GFP+ cells expressed the radial glial/neural stem cell marker 289 GFAP upon KD with the sh165 construct ( Fig. 5c-e). This further suggests that Paupar loss increases 290 lineage progression and/or diminishes SVZ stem cell maintenance. 291 The Allen Brain Atlas shows Kap1 expression in the SVZ. To study the functional effect of Kap1 on SVZ 292 neurogenesis, P1 pups were electroporated with either a scr control or the Kap1 shRNA expression 293 vectors that we used to deplete Paupar in N2A cells ( Fig. 3 and Supplemental Fig S1b) and sections 294 were immunostained for GFP and SVZ markers (Fig. 5f). At 3dpe of Kap1 shA and shB, the percentage 295 of GFP+ cells that expressed the radial glial/neural stem cell marker GFAP significantly decreased (Fig.  296   5g). This is similar to Paupar KD and is consistent with accelerated lineage progression. Also similar to 297 Paupar KD at 3dpe, Kap1 KD did not alter the percent of GFP+ cells which expressed DLX2 or Ki67. 298 However, the percentage of MASH1+ cells decreased slightly but significantly at 3dpe post Kap1 shA 299 KD, which was not found upon Paupar KD. Since we showed that Paupar and Kap1 regulate similar as 300 well as different genes this result may be due to differential gene regulation. Furthermore, these because we did not detect changes in the number of CASPASE3+ cells (Supplemental Fig S4a), or in the 303 percentage of GFP+ cells that are Tunel+ between scr control and any of the Paupar or Kap1 shRNA 304 expression vectors (Supplemental Fig S4b, c). 305 We next studied how Paupar or Kap1 affects the number of electroporated cells that reach the OB 306 7dpe. There were significantly fewer GFP+ cells in the OB after Paupar KD using sh165 KD compared to 307 the scr control whilst KD with sh408 caused a slight but statistically non-significant decrease in OB 308 GFP+ cell numbers (Fig. 6a, b). Co-staining with the immature neuroblast marker DCX (Yang, 309 Sundholm-Peters et al., 2004) showed that all GFP+ cells in the OB were DCX+ and this was not altered 310 by Paupar KD (Supplemental Fig S3d). Similar to Paupar, at 7dpe of either Kap1 KD construct, there 311 was a significant reduction in the number of GFP+ cells that had migrated from the SVZ to the OB (Fig.  312 6c, d). These results suggest that both Paupar and Kap1 are required for the production of newborn 313 OB neurons. 314 Interestingly, Paupar as well as Kap1 knockdown altered the morphology of newborn neurons that 315 migrated to the OB (Fig. 6e-h). In scr controls many GFP+ neurons in the OB granule layer had 316 processes extending radially towards the pial surface and some of the processes were branched and 317 these were classified as class I cells (Fig 6e, f). By contrast, after Paupar KD, a variety of abnormal 318 morphologies were observed, which we classified as class II or class III (Fig 6e). Class II cells were rare 319 but were distinguished by many short branched processes. Class III cells were stunted with only short 320 or no processes (Fig 6e). Quantification revealed that after Paupar KD the percentage of cells with 321 Class I morphology was 34±2% in scr controls but only 8±3% in sh165 and 6±3% in sh408 (P=0.0005 322 and P=0.0009, respectively) (Fig. 6g). Conversely, after Paupar KD there were more class III neurons in 323 the sh165 group 87±4% as well as in the sh408 group 85±6% compared to 58±5% controls (P=0.003 324 and P=0.02, respectively). Kap1 knockdown showed similar effects (Fig 6f, h) Gene Ontology analysis was performed as previously . 514

Neurosphere Assay 515
Neurospheres were cultured according to standard protocols as previously described (Dizon, Shin et 516 al., 2006). In brief, age P3-P6 CD1 pups were anesthetized by hypothermia and decapitated, and the 517 brains were immediately dissected out and sectioned in the coronal plane with a McIlwain tissue 518 chopper. The SVZ was then dissected out in ice-cold HBSS in a sterile laminar flow hood. Accutase 519 was used for 15 mins for dissociation. Cells were cultured in defined Neurobasal media 520 supplemented with 20ng/ml EGF (Sigma) and 20ng/ml bFGF (R&D). Cells were seeded at a density of 521 100 cells/l and passaged every 3-4 days. 522 Neural stem cell nucleofection 523 3-4 x 10 6 dissociated neurosphere cells were nucleofected according to the protocol of LONZA (VPG-524 1004). Cells were mixed with 100μl nucleofection solution (82μl of Nucleofector Solution + 18μl of 525 supplement) and 5-10μg DNA and transferred into cuvettes. 500μl of culture medium was added into 526 the cuvette and the sample was then transferred into 1ml medium and centrifuged at 1200rpm for 527 5min and resuspended with fresh medium and plated at 200 000 cells/ 2ml in a polyheme coated 6-528 well plate. 529

Postnatal electroporation 530
Electroporation was performed as published (Boutin, Diestel et al., 2008b, Chesler et al., 2008. DNA 531 plasmids were prepared with Endofree Maxi kit (Qiagen) and mixed with 0.1% fast green for tracing. DNA concentrations were matched in every individual experiment. P1 CD1 pups were anesthetized with 533 hypothermia and 1-2 μl of plasmids were injected with glass capillary. Electrical pulses (100V, 50ms ON 534 with 850ms intervals for 5 cycles) were given with tweezer electrodes (CUY650P5). Pups were 535 recovered, then returned to dam and analysed at the indicated time. 536 Immunohistochemistry and imaging 537 Immunohistochemistry was as previously described ( Scanning Microscopy. For co-localization in GFP+ cells, a 40X oil immersion objective was used and 2μm 544 intervals were used for generating Z-stacks. Confocal images were analysed with ImageJ. 545

Morphological evaluation 546
All GFP+ neuroblasts in the granule layer of the OB were binned into Class I, II, or III groups. Only cells 547 with obvious cell bodies and that were entirely found in the field were included. Cells in the rostral 548 migratory stream in the core of the OB, and in OB layers outside of the granule layer were not 549 included. N=3-5 mice per group. 550

Ethics 551
All mouse experiments were performed in accordance with institutional and national guidelines and 552 regulations under UK Home Office Project Licence PPL 3003311. 553

Data availability 554
Microarray and ChIP-Seq data will be deposited in the GEO database. terms were used to annotate Paupar associated proteins according to biological process. The 704 Bonferroni correction was used to adjust the P-values to correct for multiple testing. (c) Endogenous 705 Paupar transcript interacts with transcription and chromatin regulatory proteins in N2A cells. Paupar 706 association with the indicated proteins was measured using native RNA-IP. Whole cell lysates were 707 prepared and the indicated regulatory proteins immuno-precipitated using specific antibodies. 708 Bound RNAs were purified and the levels of Paupar and the U1snRNA control detected in each RIP 709 using qRT-PCR. Paupar transcript directly interacts with KAP1 and RCOR3 in N2A cells. Nuclear 710 extracts were prepared from UV cross-linked (d) and untreated (e) cells and immuno-precipitated 711 using either anti-KAP1, anti-RCOR3 or a rabbit IgG control antibody. Associated RNAs were 712 stringently washed and purified. The levels of Paupar and a U1snRNA control transcript were 713 detected in each UV-RIP using qRT-PCR. Results are presented as fold enrichment relative to control 714 antibody. Mean values +/-SEM., N=3. One-tailed t-test, unequal variance *p<0.05, **p<0.01, 715 ***p<0.001 (f) PAX6 associates with KAP1 in N2A cells. FLAG-PAX6 and KAP1 or RCOR3 expression 716 vectors were transfected into N2A cells. Lysates were prepared two days after transfection and 717 FLAG-PAX6 protein immuno-precipitated using anti-FLAG beads. Co-precipitated proteins were 718 detected by western blotting. 719