G9a controls placental vascular maturation by activating the Notch Pathway

The placenta mediates transport between the mother and the fetus through a complex vascular network. Defects in placental vascular development can cause embryonic death and abnormal organogenesis, can negatively affect fetal growth and can confer a higher risk of disease in the postnatal life (Barker et al., 1989). For example, defective patterning of the fetoplacental vasculature, also known as the labyrinth, results in abnormal heart development in mouse (Shaut et al., 2008) and human (Demicheva and Crispi, 2014), and causes intrauterine growth restriction (IUGR) (Barut et al., 2010). IUGR affects up to 32% of pregnancies in some developing countries (Ananth and Vintzileos, 2009), and can cause cardiovascular disease in utero and in adulthood (Demicheva and Crispi, 2014). Understanding the mechanisms controlling development of the placental vasculature is essential to uncover the developmental origins of disease.

A key time point during placental development is the transition from a developmental phase to a maturation phase, which occurs at mid gestation (Knox and Baker, 2008). The developmental phase spans from day 32 to month 5 of pregnancy in humans (Benirschke et al., 2012) and from embryonic day (E) 8.5 to E12 in the mouse (Knox and Baker, 2008). During the developmental phase, placental blood vessels branch to create a complex network. After the fifth month of pregnancy in humans, and E12 in mice, the placental vasculature transitions to the maturation phase during which placental blood vessels elongate (Benirschke et al., 2012). In humans, placental vascular elongation requires proliferation and migration of endothelial cells and is accompanied by a concomitant decrease in the proliferation of trophoblasts (Benirschke et al., 2012; Kaufmann et al., 1985), which provide angiogenic signals to the endothelial cells (Adamson et al., 2002; Kaufmann et al., 1985).

The mechanisms controlling blood vessel elongation are not well understood. Evidence in vitro and in the mouse retina suggests that modulating the Notch pathway in endothelial cells favors vessel enlargement (Pedrosa et al., 2015; Ubezio et al., 2016). Notch signaling is required for vascular development and remodeling (Roca and Adams, 2007), and the main Notch effector, the transcription factor recombination signal binding protein for immunoglobulin kappa J (Rbpj), is required for extra-embryonic blood vessel maturation (Copeland et al., 2011). The Notch pathway might also be involved in human placental maturation, as the Notch receptor NOTCH1 and its ligand JAG2 are downregulated in placentae affected by IUGR (Sahin et al., 2011). However, the mediators controlling the balance between endothelial cell and trophoblast proliferation in the maturing placenta, and the function of the Notch pathway in endothelial cells during vascular maturation are unknown.

The transition from the developmental to maturation phase is accompanied by a transcriptional switch in the expression of ∼700 genes that occurs from E12 to E13.5 (Knox and Baker, 2008). These dramatic gene expression changes imply a role for global transcriptional regulators during the transition. Epigenetic regulators coordinate global gene expression patterns by modifying chromatin structure and controlling the access of transcription factors to their target genes, and also control vascular development (Delgado-Olguin et al., 2014; Griffin et al., 2011). The euchromatic histone-lysine N-methyltransferase (G9a; also known as Ehmt2) can both repress and activate gene expression. G9a mediates gene repression by di-methylating lysine 9 of the histone H3 (H3K9me2), and by scaffolding the interaction of repressor protein complexes at target genes (Shinkai and Tachibana, 2011; Tachibana et al., 2002). In addition, G9a also co-activates gene expression through interaction with histone acetyl transferase complexes (Oh et al., 2014) and estrogen receptors (Purcell et al., 2011). G9a is required for embryonic development. Constitutive G9a mutant mice die at E12.5 (Tachibana et al., 2002), and fail to inactivate some imprinted genes in placental trophoblasts at E9.5, but appear to develop a normal placenta at this stage (Nagano et al., 2008; Wagschal et al., 2008). Thus, G9a is not required for early placental development up to E9.5. However, the function of G9a in specific placental cell types in later aspects of placental vascular development, such as vascular maturation, or in placental disease is unknown.

To uncover the function of G9a in vascular development, we inactivated G9a in endothelial progenitors and their derivatives by Tie2-Cre-mediated homologous recombination of a conditional allele (Proctor et al., 2005). G9a was efficiently inactivated as shown by decreased G9a and H3K9me2 in embryo body and placental endothelial cells in E10.5 G9a mutant embryos (Fig. S1). Thus, G9a is the major H3K9me2 methyltransferase in embryonic endothelial cells.

G9a is dispensable for endothelial cell specification because expression of the endothelial marker Pecam1 and of GFP driven by a Tie2-Cre-activated Rosa^mT/mG reporter transgene (Muzumdar et al., 2007) revealed a normal vascular pattern in E9.5 G9a mutant embryos (Fig. S2). G9a heterozygotes were recovered at the expected ratios and grew to adulthood. In contrast, G9a homozygous mutants died by E16.5 (Table S1), indicating that G9a is essential for vascular development.

G9a mutant embryos appeared normal at E10.5, but became pale starting at E13.5. Endothelial cells are essential for liver development (Lammert et al., 2003); accordingly, G9a mutants had an underdeveloped liver (Fig. S3).

Knockout of G9a-like protein (GLP; also known as Ehmt1) concomitant with knockdown of G9a in cardiovascular progenitor cells has been shown to cause cardiac septation defects (Inagawa et al., 2013). In our experiments, G9a was inactivated in endocardial cushions (ECs) (Fig. S4A-D), which fuse to form the membranous cardiac septum. Apoptotic EC cells positive for activated caspase 3 were increased in G9a mutants (Fig. S4E-G). Accordingly, interventricular septation was incomplete in mutant hearts (Fig. S4H-K), indicating that G9a is required in ECs for interventricular septation. Thus, G9a is required for development of endothelial derivatives in specific organs, including the liver and heart.

Analysis of the umbilical cord revealed stenosis of the umbilical artery in G9a mutant E13.5 embryos (Fig. S5), suggesting a possible function of G9a in placental vascular development. Measurements of control whole placentae showed that the thickness of the capillary-irrigated portion of the normal labyrinth progressively increases from E12.5 to E15.5. The thickness of this area of the labyrinth was comparable between wild-type controls and G9a mutant placentae at E12.5; however, G9a mutant placentae failed to expand the capillary-irrigated area from this point onwards (Fig. 1A,B). Similarly, quantification on histological sections revealed that the entire labyrinth surface area was comparable between controls and G9a mutants at E12.5, but significantly decreased in G9a mutants at E13.5 and E15.5 (Fig. 1C,D).

To visualize endothelial cells, we incorporated the Rosa^mT/mG transgene (Muzumdar et al., 2007) in mice carrying the conditional allele of G9a and expressing Tie2-Cre. GFP fluorescence revealed a smaller labyrinth (Fig. 1E), and fluorescence activated cell sorting (FACS) recovered a smaller percentage of GFP⁺ cells from G9a mutant placentae (Fig. 1F,G). Thus, deficiency of G9a leads to decreased numbers of endothelial cells and decreased labyrinth expansion at E13.5.

To analyze the labyrinth structure in detail, we perfused the placental arterial vasculature in E15.5 control and G9a viable mutants with an X-ray contrast agent and scanned them using microcomputed tomography (microCT) (Rennie et al., 2011). We analyzed three-dimensional renderings to quantify the span and depth of the labyrinth, vessel curvature, the number of branching generations, and the diameter of the Strahler order vessels (Rennie et al., 2011, 2015). There was a trend towards a decreased labyrinth span in mutants (mean 6.3±0.4 mm) compared with controls (6.6±0.2) (P<0.08). In agreement with measurements on tissue sections (Fig. 1A), G9a mutant placentae had a significantly decreased labyrinth depth (mean 1.40±0.12 mm) compared with controls (mean 1.71±0.20 mm) (P<0.02) (Fig. 2A,B).

Large vessels of the labyrinth in the mutant group exhibited marked curvature on surface renderings (Fig. 2C, red line). This feature was not observed in control placentae. Accordingly, the median tortuosity ratio was slightly increased in mutants (mean 1.79±0.56) compared with controls (mean 1.62±0.47) (P<2×10⁻¹⁶) (Fig. 2D).

Although there was no significant difference in the total number of vessel segments in the arterial tree between groups (mean 3927±1574 and 2858±1435 for mutants and controls, respectively), there was a trend towards decreased numbers of larger diameter (>200 μm) vessels, and increased numbers of smaller diameter vessels in mutants compared with controls (P<0.08) (Fig. 2E). Analysis of branching structure of the vascular tree with a Strahler ordering scheme revealed that most of the trees had seven orders (two mutants and one control had eight orders). There was no difference in the number of the order 1-7 vessels between the groups. However, there was a significant decrease in the diameters of the Strahler order vessels in mutants compared with controls (P=0.006, two-way ANOVA). Furthermore, there was a significant interaction between Strahler order and group (P=0.02, two-way ANOVA), with the largest difference in diameters on Strahler order 3 vessels (P<0.001) (Fig. 2E,F). Thus, G9a is essential for vessel elongation and maturation, and labyrinth expansion.

To uncover the cellular mechanism controlled by G9a required for labyrinth expansion, we analyzed the capacity of G9a mutant endothelial cells to establish tight junctions, and their cell shape and migratory capacity. Control and G9a mutant cells established tight junctions in confluent cultures, as revealed by membrane localization of zonula occludens-1 (ZO1; also known as Tjp1) (Fig. 3A). Transmission electron microscopy (TEM) showed smaller nuclei in G9a mutant than in control endothelial cells (Fig. 3B,C). Endothelial cell shape and nuclear shape are coordinated (Versaevel et al., 2012). To assess cell shape, GFP⁺ placental endothelial cells sorted from E12.5 control and G9a mutant embryos carrying the Rosa^mT/mG transgene were cultured at low confluence on gelatin. G9a mutant cells developed shorter filopodia, and were overall shorter than control cells after 72 h in culture (Fig. 3D,E) suggesting that G9a mutant cells are less sensitive to migratory stimuli. To assess migratory capacity, we seeded an equal number of control and G9a mutant placental endothelial cells and performed migration scratch assays. Fewer G9a mutant cells migrated into the scratched zone compared with controls (Fig. 3F,G), indicating decreased migratory capacity. Thus, decreased migratory capacity of G9a mutant endothelial cells might contribute to defective placental vascular maturation.

Histological sections showed wedges of spongiotrophoblast invading the labyrinth in G9a mutant placentae at E13.5 and E15.5 (Fig. 1C, arrows). To visualize the spongiotrophoblast, we performed in situ hybridization for trophoblast specific protein alpha (Tpbpa). Quantification on stained sections revealed that the spongiotrophoblast in normal placentae, unlike the labyrinth, which grows progressively from E12.5 to E15.5 (Fig. 1C,D), grows only marginally between these stages (Fig. 4A,B). In contrast, the spongiotrophoblast grows progressively from E12.5 to E15.5 in G9a mutant placentae (Fig. 4A,B). Spongiotrophoblast overgrowth as a result of inactivation of G9a in labyrinth endothelium suggests that growth of these placental cell types must be coordinated. We quantified proliferating endothelial cells and spongiotrophoblasts in E11.5 to E15.5 control placentae. Endothelial cells were identified by immunofluorescence for GFP in mice carrying the Rosa^mT/mG and Tie2-cre transgenes. Spongiotrophoblasts were identified by immunofluorescence for Tpbpa. Immunofluorescence for GFP and phosphorylated histone H3 showed that endothelial cells increase their proliferation from 31% at E11.5 to 39% at E12.5, and then gradually decrease to 9% at E15.5. In contrast, only 12% and 13.5% of endothelial cells proliferate at E11.5 and E12.5, respectively, and decrease down to 2% at E15.5 in G9a mutant placentae (Fig. 4C,D).

Few proliferating spongiotrophoblasts were present in control placentae (Fig. 4E); therefore, we obtained total numbers of proliferating trophoblasts in sections of normal placentae. This analysis revealed an increase of 20% (five to seven cells) from E11.5 to E12.5, and a decrease of 52% (seven to four cells) from E12.5 to E13.5. Proliferating trophoblasts then decreased by 82% (one cell from five at E11.5) by E15.5. G9a mutant placentae showed significant increases over control placentae of proliferating trophoblasts at E11.5, E12.5, E13.5, E14.5 and E15.5 (Fig. 4E,F). Thus, a significantly higher proportion of endothelial cells versus spongiotrophoblasts proliferate during the normal development-to-maturation transition, and G9a in placental endothelial cells is required to maintain this balance. Thus, decreased endothelial cell migration and proliferation, combined with increased trophoblast proliferation affect placental vascular maturation in the G9a mutant placenta.

To uncover the transcriptional pathways regulated by G9a that control placental vascular development, we analyzed the global gene expression profile in the labyrinth of control and G9a mutant placentae at E12 and E13.5, before and after the transition to maturation, respectively, by high-throughput sequencing of RNA. Similar to a previous report (Knox and Baker, 2008), our analysis revealed a transcriptional switch from E12.5 to E13.5. The expression of 3722 genes changed more than 1.5-fold (P<0.05) in normal placentae from E12.5 to E13.5; 1930 genes were down- and 1792 upregulated (Fig. 5A). Analysis of the genes that changed their expression using DAVID (Huang et al., 2009) identified pathways that regulate vascular development and cell proliferation. The top ten pathways associated with upregulated genes include cell adhesion, cell cycle, insulin resistance, and the Notch pathway. The top ten pathways associated with downregulated genes include cytokine-cytokine receptor interaction, TNF signaling, chemokine signaling, and p53 signaling (Fig. 5B). Finely tuned regulation of the Notch pathway is required for vascular maturation (Pedrosa et al., 2015; Ubezio et al., 2016). In addition, expression of the main effector of the Notch pathway, Rbpj, in developing endothelial cells is required for vascular maturation in extra-embryonic tissues (Copeland et al., 2011). Most genes in the Notch pathway identified in our analysis were downregulated after the development-to-maturation transition at E13.5. However, the Notch pathway effectors Rbpj and mastermind like transcriptional co-activator 2 (Maml2), were upregulated (Fig. 5C). Hence, despite downregulation of many Notch ligands and receptors, the activity of Notch effectors might increase during maturation of the normal placenta.

Only 2565 genes changed their expression during the transition to maturation in the G9a mutant labyrinth, from which only 1768 overlapped with genes for which expression changed in the control placentae (Fig. 5D). Consistent with the G9a mutant labyrinth phenotype, positive regulators of endothelial cell migration and proliferation, including TAL BHLH transcription factor 1 (Tal1) (Lazrak et al., 2004), mouse double minute 2 homolog (Mdm2) (Secchiero et al., 2007) and glutathione peroxidase 1 (Gpx1) (Galasso et al., 2006), were decreased in the G9a mutant labyrinth. Thus, G9a controls the transcriptional switch from the development phase to the maturation phase. We used gene set enrichment analysis (GSEA) (Subramanian et al., 2005) to identify pathways associated with genes that were misregulated in the G9a mutant labyrinth at E12.5 and E13.5. Interestingly, the analysis identified Notch amongst the most enriched pathway in upregulated genes at E12.5 and E13.5 (Fig. 5E). Accordingly, several Notch receptors and ligands were upregulated in G9a mutant labyrinth. However, the Notch effectors Rbpj and Maml2, which were upregulated during the developmental-to-maturation transition in the normal labyrinth (Fig. 5C), were decreased in the G9a mutant labyrinth (Fig. 5F). Analysis of the 5′ regulatory region of genes downregulated in E12.5 G9a mutant labyrinth using Enrichr (Kuleshov et al., 2016) revealed enrichment of RBPJ binding motifs (data not shown), suggesting that the Notch pathway might activate genes important for placental vascular maturation. Western blot analysis confirmed decreased Rbpj levels, and revealed a slight decrease of the cleaved, but not full-length, Notch1, and of Notch2 proteins in G9a mutant labyrinth (Fig. 5G,H). This analysis was carried out on labyrinth tissue; therefore, it could have underestimated protein decrease in endothelial cells. In situ hybridization (Fig. 5I) confirmed downregulation of Rbpj mRNA in G9a mutant placental labyrinth. Upregulation of Rbpj during normal placental maturation and its downregulation in G9a mutant placental endothelial cells was confirmed by qPCR on GFP⁺ cells sorted from the labyrinth of G9a mutant embryos carrying the Tie2-Cre and Rosa^mT/mG transgenes (Fig. 5J; Fig. S6). Thus, G9a controls the transcriptional switch from development to maturation and is required for activation of the Notch pathway to promote maturation of the placental vasculature.

Forcing Notch activation in endothelial cells should improve placental vascular maturation in G9a mutants. To test this hypothesis, we integrated the Rosa^Notch transgene into G9a mutants. Rosa^Notch contains a Lox-STOP-Lox cassette between the coding sequence of the intracellular portion of the Notch1 receptor and the GT(ROSA)26Sor promoter. Cre-mediated recombination constitutively activates the Notch1 receptor (Murtaugh et al., 2003). We crossed G9a^fl/+;Tie2-Cre males with G9a^fl/fl;Rosa^Notch/+ females, to activate the Notch pathway in G9a mutant endothelial progenitor cells. Activation of Notch in control G9a heterozygotes did not affect placental morphology (Fig. S7). Only two out of 41 G9a live (with beating hearts) mutant embryos were recovered at E16.5 (4% rather than the expected 12.5%), whereas seven out of 41 live G9a mutants carrying Rosa^Notch were recovered at E16.5 (17% rather than the expected 12.5%) (Table S2). Survival curves for G9a mutants positive and negative for the Rosa^Notch transgene analyzed using the Kaplan–Meier method coupled with Gehan–Breslow–Wilcox test revealed that the increase was significant (P<0.01) (Fig. S8). Furthermore, the placental phenotype was rescued in G9a mutants carrying the Rosa^Notch transgene. The labyrinth surface area was normalized in G9a mutants expressing the Rosa^Notch transgene (Fig. 6A-D). In addition, activation of the Notch pathway in G9a mutants blunted abnormal trophoblast expansion (Fig. 6E,F). Notch activation did not rescue liver underdevelopment and anemia (Fig. S2B,C), suggesting that defective placental vascular development in G9a mutants is not secondary to anemia. Viable G9a mutants carrying the Rosa^Notch transgene were not recovered after E16.5, suggesting that G9a is required in non-placental endothelial precursors and their descendants for embryonic survival. Alternatively, G9a might regulate additional placental endothelial functions not assessed in our study. Thus, G9a controls placental vascular maturation by activating the Notch pathway.

Our results open the possibility that G9A-mediated modulation of the Notch pathway might be required in human placental maturation and involved in placental diseases such as IUGR. We determined the presence of G9A mRNA and protein in the human placenta. In situ hybridization and immunohistochemistry revealed G9A in endothelial cells in placental blood vessels of the chorionic plate (Fig. 7A,B). To assess the potential involvement of G9A and the Notch pathway in human placental vascular maturation, we quantified the expression of G9A and RBPJ in chorionic villi of human placentae affected by IUGR using qPCR. G9A and RBPJ mRNA and protein, as well as the cleaved but not the full-length NOTCH1 and NOTCH2, were slightly decreased in placentae from pregnancies affected by IUGR (Fig. 7C-E). Our analysis was carried out on placental tissue and thus could have underestimated cell-specific decreases in Notch regulators. Altogether, our results suggest that G9A is a key regulator of placental vascular maturation potentially implicated in human placental disease.

Vascular elongation during placental maturation requires proliferation of endothelial cells (Benirschke et al., 2012) and concomitant rearrangement of the surrounding trophoblast. Such rearrangement is accompanied by reduced proliferation of the trophoblast along the entire length of the elongating vessel (Benirschke et al., 2012). The cellular and molecular events controlling maturation of the placental vasculature are poorly understood in large part because this process has not been systematically investigated in mouse models. We found that mouse endothelial cells actively proliferate, whereas trophoblasts markedly decrease their proliferation during the transition from development to maturation in the normal placenta (Fig. S9). We identified G9a as a key regulator required for coordinated proliferation of endothelial and trophoblast cells, and labyrinth expansion. Inactivation of G9a in endothelial progenitors and their derivatives caused a decrease in endothelial proliferation with a concomitant increase in trophoblast proliferation and decreased labyrinth expansion. Furthermore, deficiency of G9a results in decreased expression of Notch pathway effectors (Fig. S9), and activation of the Notch pathway rescues labyrinth expansion. This study describes the first mouse model of deficient placental vascular maturation and provides insight into the regulatory pathways.

It has been proposed that a reduction in trophoblast proliferation during placental maturation is required for vascular elongation driven by endothelial cell proliferation, and that the balance between VEGFA (a member of the vascular endothelial growth factor family) and placental growth factor (PlGF; also known as PGF), which are secreted by both endothelial cells and the trophoblast (Barker et al., 2010), is involved. In contrast to VEGFA, PlGF, a regulator of trophoblast differentiation (Maglione et al., 1993), suppresses angiogenesis. Therefore, paracrine signaling between the endothelium and the trophoblast might regulate placental vasculature maturation. We found that deletion of G9a in endothelial progenitors causes a decrease in proliferation of endothelial cells and increased proliferation of the trophoblast. This indicates that G9a is required in the labyrinth to limit trophoblast expansion. However, neither Plgf nor Vegfa was misregulated in G9a mutant placentae, suggesting that alternative molecules might mediate communication from the endothelium to limit trophoblast proliferation. Our RNAseq data revealed that the G9a mutant labyrinth overexpresses secreted protein-encoding genes, including secreted modular calcium-binding protein 1 (Smoc1) and arylsulfatase family member 1 (Arsi). Smoc1 is expressed in endothelial cells and positively regulates cell proliferation (Awwad et al., 2015), and Arsi is secreted from proliferating long-term self-renewing neuroepithelial-like stem cells (Doerr et al., 2015). Thus, placental endothelial cells might limit trophoblast proliferation via G9a-mediated repression of secreted activators of cell proliferation.

Activating the Notch pathway in G9a mutant placental endothelial progenitors rescued placental endothelial cell proliferation and labyrinth expansion (Fig. 6). This suggests that the Notch pathway stimulates endothelial cell proliferation in the placenta. Notch signaling has been shown to both repress and promote endothelial cell proliferation in different systems. Dll4/Notch signaling restricts angiogenesis to tip cells, which guide new sprouts, and inhibits angiogenic responses in other endothelial cells (Siekmann and Lawson, 2007). Accordingly, Notch deficiency leads to excessive proliferation of endothelial cells in segmental arteries in zebrafish (Leslie et al., 2007). In contrast, the Notch ligand jagged 1 promotes angiogenesis by antagonizing Dll4/Notch signaling in mouse retinal endothelium (Ehling et al., 2013). In addition, inhibition of Dll4/Notch signaling increases sprouting and density of nonfunctional blood vessels (Noguera-Troise et al., 2006; Ridgway et al., 2006). Our finding that Notch effectors are downregulated in G9a-mutant labyrinth endothelium and in the chorionic villi of human placentae affected by IUGR opens the possibility that pharmacological activation of the Notch pathway in the placenta might be a potential strategy to prevent or treat abnormal maturation of the placental vasculature. Valproic acid, which is used to control epileptic seizures and psychiatric disorders, was found to activate the Notch pathway resulting in increased activity of Rbpj (Pinchot et al., 2011). The use of valproic acid during pregnancy has limitations as it can promote a higher risk of neural tube closure defects in babies when administered during the first trimester of pregnancy (Jentink et al., 2010). However, the transition from development to maturation occurs in the second trimester (fifth month of pregnancy) (Benirschke et al., 2012). Therefore, valproic acid treatment after the first trimester could be tested as a way to prevent or correct abnormal maturation of the placental vasculature. Our mouse model will be useful for testing the capacity of valproic acid to activate the Notch pathway in placental endothelium and its effect on placental endothelial cell and trophoblast proliferation and vascular maturation.

IUGR is diagnosed by ultrasound only when the placenta is malfunctioning (Albaiges et al., 2000) or the fetus is smaller than expected for the gestational age (Yoshida et al., 2000). Uncovering the cellular and molecular events dysregulated in the developing placenta that affect vascular maturation and precede clinical manifestation is required to develop methods to identify fetuses at risk of IUGR or other abnormalities associated with placental insufficiency. We found that deficiency of G9a in placental endothelial cells alters placental morphology starting at E13.5. Importantly, we found that the decreased ratio of endothelial cell and spongiotrophoblast proliferation at E11.5 precedes morphological defects in the labyrinth and spongiotrophoblast, which were first detected at E12.5 (Figs 1 and 4). Therefore, it is possible that an altered ratio of endothelial cells versus trophoblast proliferation is an early manifestation of placental maturation defects. Our mouse model will be useful to identify molecules associated with an abnormal endothelial cell/spongiotrophoblast proliferation ratio that could be investigated as potential biomarkers of placental maturation defects.

Mouse lines: G9a^fl/fl (Sampath et al., 2007), Tie2-Cre (Proctor et al., 2005), Rosa^mT/mG (Muzumdar et al., 2007) and Rosa^Notch (Murtaugh et al., 2003). Procedures followed the Canadian Council for Animal Care guidelines, and were approved by the Animal Care Committee at The Centre for Phenogenomics. Embryos and placentae were dissected in cold PBS and fixed in 4% paraformaldehyde for 2 h at 4°C, dehydrated in an ethanol series and stored at −20°C until processing. Embryos and placentae were rehydrated by reversing the ethanol series before processing. Littermates from at least two litters were analyzed.

Samples were fixed in 2% glutaraldehyde in 0.1 M sodium cacodylate buffer, rinsed in buffer, post-fixed in 1% osmium tetroxide in buffer, dehydrated in a graded ethanol series followed by propylene oxide, and embedded in Quetol-Spurr resin. Sections (90 nm thick) were cut on a Leica Ultracut ultramicrotome, stained with uranyl acetate and lead citrate and viewed in an FEI Tecnai 20 transmission electron microscope.

Detailed methods for microCT imaging have been described (Rennie et al., 2015, 2014; Whiteley et al., 2006). Briefly, uteri were dissected at E15.5 and immersed in ice-cold PBS. Each fetus was dissected from the uterus while maintaining the vascular connection to the placenta. Embryos were immersed in warm PBS to resume blood circulation. A catheter was inserted into the umbilical artery and the fetus was perfused with saline (with heparin, 100 units/ml) followed by radio-opaque silicone rubber contrast agent (Microfil; Flow Technology). Following perfusion, specimens were fixed (10% formalin) for 24-48 h at 4°C. Specimens were scanned using a Bruker SkyScan1172 high-resolution microCT scanner at a resolution of 7.1 μm and 996 views were acquired via 180-degree rotation with an X-ray source at 54 kVp and 185 μA. Three-dimensional data were reconstructed using SkyScan NRecon software.

The structure of the vasculature was identified automatically using a segmentation algorithm (Rennie et al., 2011). The leaves of the vascular tree were pruned to 35 μm for data consistency. Analysis was performed on seven controls and four mutants. Each group contains a minimum of three dams and two to three specimens per litter. The span and depth of the fetoplacental arterial tree were measured from the surface renderings via digital calipers using the Amira software package (Visage Imaging). To quantify vessel curvature, a tortuosity ratio was computed (Rennie et al., 2011) by determining the ratio of the vascular path length to the Euclidian (i.e. beeline) distance from the umbilical artery to each terminal vessel. The number of branching generations in each specimen was determined using a Strahler ordering scheme (Rennie et al., 2015), which numbers vessels in an upstream direction starting from the terminal vessel segments. The geometric and morphological features of the vascular trees were analyzed using t-tests. A two-way ANOVA determined whether there was an effect of group on the average diameter of the Strahler order vessels, followed by t-tests to compare mutants with controls at each Strahler order number. A value of P<0.05 was considered significant.

RNA was isolated from the dissected labyrinth of control G9a^fl/fl and mutant G9a^fl/fl;Tie2-cre placentae at E12.5 and E13.5 using the Direct-zol RNA MiniPrep Plus Kit (Zymo Research). mRNA was pulled down using the NEBNext Poly(A) mRNA magnetic isolation module (New England Biolabs). RNAseq libraries were prepared using the NEBNext Ultra Directional RNA Library Prep Kit (New England Biolabs), and sequenced on the Illumina HiSeq 2500 platform. Sequencing reads were trimmed using Trimmomatic (Bolger et al., 2014) and mapped to the mouse genome (mm10) using STAR (Dobin et al., 2013). Mapped read counts were obtained using HTseq (Anders et al., 2015). Differential expression analysis was performed using DESeq2 (Love et al., 2014). Clustering was performed using Cluster 3.0 and visualized with Treeview (Saldanha, 2004). Enrichment of gene ontology categories was determined using DAVID (Huang et al., 2009). GSEA pre-ranked enrichment analysis with GSEA v2.2.0 (GSEA) (Subramanian et al., 2005) identified pathways associated with genes misregulated in G9a mutant placental endothelial cells. Transcripts were ranked according to fold change. Analysis involved 1000 random permutations using the ‘h.all.v5.1.symbols.gmt’ Hallmarks gene set. Gene sets with more than 125 or fewer than ten genes were excluded from the analysis. RNAseq data were deposited in Gene Expression Omnibus with entry GSE97579.

Sections (7 µm thick) were obtained from placental samples embedded in paraffin and processed as described previously (Chi and Delgado-Olguin, 2013).

GFP⁺ endothelial cells were sorted from control (G9a^fl/+;Tie2-Cre;Rosa^mT/mG) and G9a mutant (G9a^fl/fl;Tie2-Cre;Rosa^mT/mG) labyrinth at E12.5 and E13.5. cDNA was synthesized from 10 ng of total RNA using the SuperScript VILO cDNA Synthesis Kit (Thermo Fisher Scientific). For qPCR, 10 pg of cDNA were used with SsoAdvanced Universal SYBR Green Supermix (Bio-Rad). qPCR was performed using a CFX384 Touch Real-Time PCR Detection System (Bio-Rad) using the primers listed in Table S3.

Sections were stained as previously described (Delgado-Olguin et al., 2014). Primary antibodies and dilutions were: G9a (Cell Signaling Technology, 3306, 1/200), H3K9me2 (Cell Signaling Technology, 9753, 1/500), GFP (GeneScript, A01694, 1/1000), Pecam1/CD34 (BD Pharmingen, 553370, 1/200), phosphorylated histone H3 (Abcam, AB5176, 1/200), activated caspase 3 (Sigma, C8487, 1/200), trophoblast specific protein alpha (Abcam, ab104401, 1/50), ZO1 (Thermo Fisher Scientific, 61-7300, 1/100), pHH3 (Santa Cruz, SC-8656-R, 1/200), Notch1 (bTAN 20, 1/200) and Notch2 (C651.6DbHN, 1/200) (Developmental Studies Hybridoma Bank). Secondary antibodies were goat anti-rabbit Alexa Fluor 568 (A-11011), goat anti-chicken Alexa Fluor 488 (A-11039), donkey anti-rat Alexa Fluor 488 (A-21208) and goat anti-mouse Alexa Fluor 568 (A-11004). (Thermo Fisher Scientific, 1/700). DAPI (Sigma, D9564, 1/1000) was used for nuclear staining.

GFP⁺ endothelial cells were sorted from control G9a^fl/+;Tie2-Cre;Rosa^mT/mG and mutant G9a^fl/fl;Tie2-Cre;Rosa^mT/mG placentae at E13.5. Placentae were disassociated in TrypLE Express (Thermo Fisher Scientific, 12604-013) at 37°C for 30 min. After centrifugation, pellets were incubated in 1× Red Blood Cell Lysis Solution (MACS Miltenyi Biotec, 130-094-183) at room temperature for 10 min. Cells were diluted with 250 µl of DMEM (Wisent, 319-005-CL) containing 1% fetal bovine serum, 1 mM EDTA and 2 µg/ml propidium iodide (Sigma, P4170) for detection of dead cells. Cells were sorted using MoFlo-Astrios BYRV equipment then 10,000 cells were seeded on 96-well plates coated with 0.1% gelatin and cultured for 3 days in endothelial cell medium containing Endothelial Cell Growth Supplement (ECGS; Sigma, E2759), 10% fetal bovine serum and 100 ng/ml of human VEGF (R&D Systems, 293-VE-010). Cultured endothelial cell length was measured on micrographs using ImageJ.

The Biobank at Mount Sinai Hospital, and the Sant'Anna Hospital, University of Turin, Italy collected tissues from elective pregnancy terminations after informed consent. Specimen collection followed ethics guidelines of the University of Toronto, the Research Ethics Board of Mount Sinai Hospital, and the World Medical Association Declaration of Helsinki. Ten placentae from singleton normal and ten from intrauterine growth restriction (IUGR) pregnancies were obtained. IUGR placentae were selected according to the American College of Obstetricians and Gynaecologists guidelines (American College of Obstetricians and Gynecologists, 2013). Participants were healthy women without signs of hypertension, pre-eclampsia, or known causes of IUGR including renal, endocrine and autoimmune disorders. IUGR cases included exhibited abnormal umbilical artery Doppler defined as absence or reverse of end diastolic velocity and a birth weight below the fifth percentile for gestational age. Normotensive age-matched preterm controls were selected based on the absence of placental disease with appropriate-for-gestational-age fetuses. Tissues were collected from pregnancies that did not exhibit any fetal and chromosomal abnormalities. Maternal characteristics are shown in Table S4.

We thank Alexander Tarakhovsky (The Rockefeller University) for the G9a floxed line; Tullia Trodros (University of Turin) for placental samples; Doug Holmyard (Advanced Bioimaging Centre, Mount Sinai Hospital and The Hospital for Sick Children) for TEM; Lisa X. Yu (SickKids Mouse Imaging Centre) for micro-CT; Andrea Tagliaferro (Mount Sinai Hospital) for obtaining human samples; Rong Mo (Chi-chung Hui's Lab, The Hospital for Sick Children) for sectioning and staining; Sheyun Zhao (SickKids-UHN Flow Cytometry Facility) for cell sorting; Paul Paroutis (Imaging Facility at The Hospital for Sick Children) for help with confocal microscopy; Sergio Pereira (The Centre for Applied Genomics) for next generation sequencing; TCP (The Centre for Phenogenomics) for mouse husbandry and care; Jason Fish (University Health Network) for advice and critical reading of the manuscript; Koroboshka Brand-Arzamendi for graphics in Fig. S9; and Natalia Kaniuk (Grant Development Office, The Hospital for Sick Children) for editorial assistance.