Prmt5 promotes vascular morphogenesis independently of its methyltransferase activity

During development, the vertebrate vasculature undergoes major growth and remodeling. While the transcriptional cascade underlying blood vessel formation starts to be better characterized, little is known concerning the role and mode of action of epigenetic enzymes during this process. Here, we explored the role of the Protein Arginine Methyl Transferase Prmt5 in blood vessel formation as well as hematopoiesis using zebrafish as a model system. Through the combination of different prmt5 loss-of-function approaches we highlighted a key role of Prmt5 in both processes. Notably, we showed that Prmt5 promotes vascular morphogenesis through the transcriptional control of ETS transcription factors and adhesion proteins in endothelial cells. Interestingly, using a catalytic dead mutant of Prmt5 and a specific drug inhibitor, we found that while Prmt5 methyltransferase activity was required for blood cell formation, it was dispensable for vessel formation. Analyses of chromatin architecture impact on reporter genes expression and chromatin immunoprecipitation experiments led us to propose that Prmt5 regulates transcription by acting as a scaffold protein that facilitates chromatin looping to promote vascular morphogenesis.


INTRODUCTION 48
Blood vessel formation is an essential developmental process required for the survival of all 49 vertebrates and much effort has been devoted to understand the molecular pathways and to 50 identify key molecules that regulate different aspects of this process. Interestingly, the vascular 51 anatomy and the mechanisms involved in vessel formation are highly conserved among 52 vertebrates (for a review, (Isogai et al., 2001)). Hence, in the past two decades, zebrafish has 53 been proven to be a useful model to study vascular morphogenesis and blood cell formation 54 in vivo (Beis and Stainier, 2006;Lawson and Weinstein, 2002a;Thisse and Zon, 2002). 55 In vertebrates, blood cell formation is tightly associated with the development of the vascular 56 system. Hematopoietic Stem Cells (HSC), which give rise to the different blood cell lineages, 57 emerge directly from the ventral part of the dorsal aorta, an area referred to as the hemogenic 58 endothelium. Notably, the ETS transcription factor ETV2 functions as a master regulator for 59 the formation of endothelial and hematopoietic cell lineages through the induction of both blood 60 cells and vasculature transcriptional programs, in mouse and in zebrafish (Liu et al., 2015b;61 Wong et al., 2009). In endothelial cells, ETV2 regulates the expression of other ETS 62 transcription factors, VEGF (Vascular Endothelial Growth factor) signaling receptors and 63 effectors, Rho-GTPases and adhesion molecules (Liu et al., 2015b;Wong et al., 2009). 64 Besides, adhesion molecules have been shown to be crucial players in vascular 65 morphogenesis as Vascular Endothelial cadherin (VE-cad/ cdh5) and endothelial cell-selective 66 adhesion molecule (Esama) are essential for junction remodeling and blood vessel elongation 67 cardiovascular disease and development (Rosa-Garrido et al., 2018;Shailesh et al., 2018), 75 only few examples illustrate in detail the role of epigenetic enzymes during blood vessel 76 development. For instance, the chromatin-remodeling enzyme BRG1 affects early vascular 77 development as well as hematopoiesis in mice (Griffin et al., 2008), and the histone 78 acetyltransferase P300 has been proposed to be recruited at the promoter of specific 79 endothelial genes by the ETS transcription factor ERG (ETS Related Gene) to control their 80 expression both in vivo in zebrafish and in HUVEC (Human Umbilical Vein Endothelial Cell) 81 (Fish et al., 2017;Kalna et al., 2019). 82 Given the common origin of blood and endothelial cells, and their partially shared 83 transcriptional programs, it is plausible that known chromatin-modifying enzymes affecting 84 hematopoiesis could also control blood vessel formation. Along this line, the epigenetic 85 enzyme Prmt5 (Protein Arginine Methyltransferase 5) has been identified as a key player in 86 blood cell formation (Liu et al., 2015a) but its impact on endothelial development has not been 87 investigated to date. Prmt5 catalyzes the symmetric di-methylation of arginine residues on a 88 variety of proteins including histones and therefore acts on many cellular processes such as 89 genome organization, transcription, differentiation, cell cycle regulation or spliceosome 90 assembly, among others (Blanc and Richard, 2017;Karkhanis et al., 2011;Stopa et al., 2015). 91 Prmt5 is mainly known to repress transcription through the methylation of arginine residues on 92 histones H3 and H4 and has been shown to regulate several differentiation processes such as 93 myogenesis, oligodendrocyte and germ cell differentiation or hematopoiesis (Batut et al., 2011;94 Liu et al., 2015a;Shailesh et al., 2018;Zhu et al., 2019). In mice, prmt5 knock out prevents 95 pluripotent cells to form from the inner cell mass and is embryonic lethal (Tee et al., 2010). 96 Conditional loss of prmt5 in mice leads to severe anemia and pancytopenia and Prmt5 97 maintains Hematopoietic Stem Cells (HSCs) and ensures proper blood cell progenitor 98 expansion (Liu et al., 2015a). Loss of prmt5 leads to oxidative DNA damages, increased cell the transgene while the cardinal vein endothelial cells do not (Ninov et al., 2012;Quillien et al., 154 2014). In this transgenic context the area occupied by the dorsal aorta in prmt5 morphant 155 embryos was significantly reduced as compared to control embryos ( Fig. 2G-I). Prmt5 mutant 156 embryos also showed a defect of sprouting ISV to reach the most dorsal part of the trunk and 157 to connect with other ISVs and form the Dorsal Longitudinal Anastomotic Vessel (DLAV) (Fig.  158 2D, E, F). This defect was associated with a significant reduction of ISV length (Fig. 2E, F, K) 159 but with no impact on the number of endothelial cells (Fig. 2J). The observed size reduction of 160 ISVs is thus most likely the result of an elongation issue rather than a proliferation defect. Of 161 note, prmt5 morphants reproduced the phenotype observed in prmt5 mutants i.e. a reduced 162 ISV length at 28 hpf ( Fig. S2 A-D). 163 To get a better insight into the impact of Prmt5 on the dynamics of vascular system formation, 164 we performed time-lapse analyses in control and prmt5 morphant embryos. Time-lapse 165 confocal movies were carried out from 28 hpf to 38 hpf to follow the elongation of ISVs to the 166 formation of an effective lumen. As compared to control morphants, prmt5 morphants showed 167 an impaired formation of ISV lumen and DLAV. Indeed, in prmt5 morphants tip cells failed to 168 stay connected to the stalk cells and to contact other tip cells to allow the formation of the 169 DLAV ( Fig. 3A-B). Moreover, supernumerary connections were detected in the context of 170 prmt5-loss of function (Fig. 3B). Altogether, these data suggest a central role for Prmt5 in 171 vascular morphogenesis. 172 The master gene regulator ETV2, ETS transcription factors and adhesion proteins have been 173 shown to be involved in blood vessel formation (Craig et al., 2015;Hultin et al., 2014;Pham et 174 al., 2007;Sauteur et al., 2017;Sauteur et al., 2014). Analyzing single cell RNA-sequencing 175 data from Wagner et al. (Wagner et al., 2018), allowed us to determine that prmt5 is expressed 176 in endothelial cells at 10 hpf (like etv2 and fli1a) and that its expression decreases in later 177 stages, when the expression of fli1b, cdh5, agtr2, esama, and amotl2a starts to increase (Fig. that etv2 expression was not affected, the expression of ETS transcription factors (fli1a, fli1b) 181 and adhesion proteins (cdh5, agtr2, esama and amotl2a), all putative ETV2 target genes (Liu 182 et al., 2015b;Wong et al., 2009), was significantly reduced in prmt5 mutant (Fig. 3C). Of note, 183 we also detected a reduction of fli1a and cdh5 expression in prmt5 mutant by in situ 184 hybridization (Fig. S4). As etv2 expression was unaffected by the loss of prmt5 but its targets 185 were down-regulated, it is tempting to speculate that Prmt5 could modulate ETV2 activity at 186 post-translational level. 187

Prmt5 methyltransferase activity is not required for vascular morphogenesis 188
That Prmt5 modulates gene expression by methylating a variety of proteins including histones 189 but also transcription (co)factors led us to test whether Prmt5 methyltransferase activity was 190 required for vascular morphogenesis and lymphoid progenitor formation. To this end, prmt5 191 mutant or morphant embryos were injected with wild type human prmt5 mRNA (hprmt5WT) or 192 with a catalytic mutant form of this mRNA (hprmt5MUT) (Pal et al., 2003). In mice, the 193 expansion of lymphoid progenitor relies on Prmt5 methyltransferase activity (Liu et al., 2015a). 194 Consistent with this, hprmt5WT but not prmt5MUT mRNA, was able to restore normal lymphoid 195 progenitor expansion in prmt5 morphant embryos (Fig. S1 C-G). This underscores the 196 conserved requirement of PRMT5 methyltransferase activity for lymphoid progenitor formation 197 in human and zebrafish. We then tested whether the same was true for ISV elongation and the 198 expression of etv2 target genes. Surprisingly, we found that both mRNAs were able to restore 199 ISV elongation, albeit to a slightly different extend, as indicated by the average ISV length in 200 injected mutant embryos as compared to non-injected mutants ( Fig. 4A-E). Indeed, we 201 observed that the average length of ISVs in hprmt5WT-injected mutants was even longer than 202 intersegmental vessels of wild type embryos, while the average length in hprmt5MUT injected 203 mutants was significantly superior to non-injected mutants but shorter than control embryos 204 ( Fig. 4 E). Interestingly, no difference could be seen in the cell number per ISV in the different 205 contexts (Fig. 4F) thus ruling out the possibility that Prmt5 regulates cell proliferation at the were able to restore the expression of etv2 target genes, except for fli1a whose expression 208 was only rescued by hprmt5WT (Fig. 4G). In sum, these results indicate that Prmt5 209 methyltranferase activity is largely dispensable for its function in blood vessel formation. 210

Prmt5 might help to shape correct chromatin conformation in endothelial cells 211
As Prmt5 methyltransferase activity seems to be not required for gene expression regulation 212 in vascular morphogenesis, we speculated that Prmt5 could act as a scaffold protein in 213 complexes mediating transcription and chromatin looping. Indeed, Prmt5 has been proposed 214 to promote enhancer-promoter looping at the PPARg2 locus and more broadly to facilitate 215 chromatin connection in adipocytes, via the recruitments of Mediator subunit MED1 and 216 SWI/SNF chromatin remodeling complex subunit Brg1 ATPase (LeBlanc et al., 2016). Thus, 217 we decided to inspect the chromatin architecture of the flanking region of identified Prmt5-218 regulated genes using ATAC-seq data from zebrafish endothelial cells that we previously 219 generated (Quillien et al., 2017). Doing so, we found that putative enhancers are on average 220 distant of 16 kb from the transcriptional start site (TSS) (Table S1, Figure S5), indicating that 221 their expression could rely on proper chromatin looping. To further characterize these specific 222 cis regulatory regions, we turned into the mouse model and analyzed the ChIP-seq data of 223 Etv2 and Prmt5-dependent H4R3 di-methylation to determine whether Prmt5 target genes 224 identified in our study were conserved in mouse (Girardot et al., 2014;Liu et al., 2015a). We 225 found that Etv2 is recruited to the cis regulatory element of amotl2, cdh5 and fli1 (Table 1)  However, in this artificial context, prmt5 morpholino injection had no effect on the level of 249 KAEDE fluorescence intensity as compared to control morphants ( Fig. 5F-H). This result 250 suggests that in this particular context i.e. when chromatin looping between enhancer and 251 promoter was not needed, Prmt5 was not required either for gene expression. This finding 252 supports the idea that Prmt5 plays a role in the formation of the correct 3D environment for 253 endothelial genes expression. Finally, rescue experiments were performed by injecting either 254 wild type or a catalytic mutant of human prmt5 mRNA to determine whether Prmt5 255 methyltransferase activity was required for the transcriptional control of cdh5 expression in the 256 endogenous context. We found that both wild type and mutant hprmt5 mRNAs restored GFP 257 fluorescence intensity in prmt5 morphants as compared to control embryos ( Fig. 5B-D, I-J). 258 Collectively, these data indicate that the transcriptional control of cdh5 is independent of Prmt5 259 methyltransferase activity and could rather rely on a role of Prmt5 as a scaffold protein to 260 provide a proper chromatin conformation context. 261 Here we have demonstrated a role for Prmt5 in both hematopoiesis and blood vessel formation 264 in zebrafish. Our results suggest that Prmt5 promotes vascular morphogenesis through the 265 transcriptional control of ETS transcription factor and adhesion proteins in endothelial cells. 266 Intriguingly, we have shown that the methyltransferase activity of Prmt5 was not absolutely 267 required to regulate gene expression, leading us to propose a role of scaffold protein for Prmt5 268 to facilitate chromatin looping formation in endothelial cells. 269 We found that, similarly as in mouse ( playing a critical role in HSC quiescence through the splicing of genes involved in DNA repair 286 (Tan et al., 2019). Of note, this study showed that Prmt5 methyltransferase activity was 287 required for controlling HSC quiescence, in agreement with our findings in the present work. endothelial cells, reinforcing the idea that Prmt5 regulates transcription by different 290 mechanisms in these two processes (Fig. 6). 291 Prmt5 has been shown to facilitate ATP-dependent chromatin remodeling to promote gene 292 expression in skeletal muscles and during adipocyte differentiation (Dacwag et al., 2009;293 LeBlanc et al., 2012;LeBlanc et al., 2016;Pal et al., 2003). Here, we propose that Prmt5 could 294 also be essential for proper chromatin looping in endothelial cells. Our data suggest that Prmt5 295 influences gene expression only in an endogenous context where chromatin looping is 296 required (e.g. chd5 and TgBAC(cdh5:GAL4FF)), while it is dispensable for gene expression 297 when enhancer and promotor regions are artificially associated (e.g. Tg(cdh5:Gal4VP16)) or 298 close by (e.g. fli1a). This implies that Prmt5 could interact with Brg1 ATPase of SWI/SNF 299 chromatin remodeling complex and with the Mediator complex in endothelial cells as it does in 300 muscle cells and adipocytes. Consistent with this hypothesis, brg1 mutant mouse embryos 301 display an anemia coupled to vascular defects in the yolk sac, characterized by thin vessels 302 and supernumerary sprouts (Griffin et al., 2008), which is reminiscent to our present findings 303 in zebrafish prmt5 mutant. Interestingly, it has been proposed that the mediator complex 304 regulates endothelial cell differentiation (Napoli et al., 2019). Moreover, our analyses of the 305 published single cell expression data (Wagner et al., 2018) indicate that, similarly to prmt5, the 306 expression of smarc4a/brg1 and med12 in zebrafish endothelial cells is detected as early as 307 10 hpf and decreases in subsequent stages. It is thus tempting to speculate that Prmt5, Brg1 308 and the Mediator could act together to regulate chromatin organization in endothelial cells (Fig.  309   6). 310 ChIP-seq data available in mouse revealed that some flanking regions of orthologues of 311 identified Prmt5 target genes are bound by ETV2 and present histone marks associated with 312 the recruitment of Prmt5. In zebrafish, both prmt5 and etv2 genes are expressed at early stage 313 in endothelial cells, and Etv2 binding motif is enriched in cis-regulatory regions identified by displaying abnormal vasculature at 48 hpf characterized by a lack of lumen formation, a lack 317 of vessel extension and aberrant connections (Craig et al., 2015;Pham et al., 2007). Here, we 318 proposed that Etv2 could be involved in the recruitments of Prmt5 to cis regulatory regions of 319 endothelial genes. Another crucial player of blood vessel formation is the transcription factor 320 Npas4l, which is expressed during late gastrulation and regulates etv2 expression (Marass et  321 al., 2019). Npas4l ChIP-seq data and ATAC-seq data from npas4l mutant also revealed the 322 binding of this transcription factor to a certain number of cis-regulatory regions of Prmt5 target 323 genes identified in the present work. In light of these findings, we speculate that Npas4l could 324 contribute to the recruitment of Prmt5 to endothelial genes (AQ and LV, unpublished data). In vitro, the addition of Prmt5 to Brg1-immunopurified complexes enhanced histone 334 methylation, while the addition of a catalytic dead version of Prmt5 did not (Pal et al., 2003). 335 Altogether these data suggest that wild type Prmt5, when recruited to target gene promoter 336 regions, acts most likely by dimethylating histone proteins. However, these studies did not 337 assess the ability of Prmt5 to facilitate chromatin looping independently (or not) of its 338 methyltransferase activity. Our data suggest that chromatin looping favored by Prmt5 does not 339 necessarily require its methyltransferase activity. Indeed, rescue experiments demonstrated 340 that Prmt5 was able to restore gene expression independently of its enzymatic activity, with 341 the exception of fli1a expression. Since fli1a putative enhancer is located only at 700 pb from and SpeI. A multisite LR recombination reaction (Gateway LR Clonase II Enzyme mix, 398 Invitrogen) was then performed using p5E_cdh5E, pme_cdh5P:Gal4VP16, with pminTol-R4-399 R2pA to give pminTol-cdh5E-cdh5P: Gal4VP16. Oligonucleotide sequences are listed in 400 Table S2.  Table S2.

RNA extraction, Reverse transcription and real-time PCR 420
Embryos were dissected at the indicated stage after addition of Tricaine Methanesulfonate. corresponding dissected tails were conserved in TRIzol Reagent at -20°C. After identification 423 of wild type and mutant embryos, total RNAs from at least 6 identified tails were extracted 424 following manufacturer's instructions (Invitrogen). Total RNAs were converted into cDNA using 425 Prime Script cDNA Synthesis Kit (Takara) with Oligo(dT) and random hexamer primers for 426 15 min at 37 °C according to manufacturer's instructions. cDNAs were then diluted 20-fold and 427 quantified by qPCR using SsoFast Evagreen Supermix (Bio-rad) and specific primers. Data 428 were acquired on CFX96 Real-Time PCR detection System (Bio-rad). Samples were analyzed 429 in triplicates and the expression level was calculated relative to zebrafish housekeeping gene 430 EF1α. Oligonucleotide sequences are listed in Table S2. Probes) for at least 2 h at room temperature and washed three times. Nuclei were then stained 447 with TO-PRO3 (Molecular Probes) and washed twice with PBST. Embryos were dissected, flat-mounted in glycerol and images were recorded on a confocal microscope as above. 449 Fluorescent in situ hybridization was carried out as previously described (Quillien et al., 2014). 450

Image processing and measurements 451
Confocal images and stacks were either analyzed with ImageJ software or LAS X. Nuclei of 452 ISV cells and gata2b+ cells were counted using the Multipoint tool of ImageJ. ISV lengths were 453 measured by drawing a line between the base and the tip of ISV on ImageJ. Contours of the 454 Dorsal Aorta were drawn using the Freehand Selection Tool with a digital pen and the area 455 was then measured. Fluorescence intensity corresponded to the measurement of average 456 gray value for each entire image. 457

Statistical analysis 458
Statistical comparisons of datasets were performed using GraphPad Prism software. For each 459 dataset, we tested the assumption of normality with D'Agostino-Pearson tests and accordingly, 460 unpaired t-test, Mann-Whitney test, One-way ANOVA, two-way ANOVA or Kruskal-Wallis test 461 were used to compare dataset; means (± SEM) are indicated as horizontal bars on dot plots. 462 The test used as well as the number of independent experiments performed and the minimal 463 number of biological replicates are indicated in each figure legend. 464