Tpbpa-Cre-mediated deletion of TFAP2C leads to deregulation of Cdkn1a, Akt1 and the ERK pathway, causing placental growth arrest

In mice, the trophectoderm (TE) cells next to the inner cell mass form diploid extra-embryonic ectoderm (ExE) and ectoplacental cone (EPC), which further forms all trophoblast cells within the developing placenta (Rossant and Cross, 2001). Around E8.5, a subset of cells within EPC starts to express the gene Tpbpa (trophoblast-specific protein α) and differentiate into the junctional zone (JZ) of the placenta (Carney et al., 1993; Lescisin et al., 1988; Simmons et al., 2007). The JZ comprises spongiotrophoblast cells (SpTs) and glycogen trophoblast cells (GCs) interspersed by several blood canals lined by trophoblast giant cells (TGCs) called canal TGCs (C-TGCs) and the channel TGCs (Ch-TGCs) (Adamson et al., 2002; Gasperowicz et al., 2013; Guillemot et al., 1994). Aside from forming SpT and GCs, TPBPA⁺ cells give rise to all spiral artery-associated TGCs (SpA-TGCs) and half of C-TGCs (Simmons et al., 2007). The labyrinth, which comprises a dense network of fetal and maternal blood vessels lined by sinusoidal TGCs (S-TGCs) and syncytiotrophoblasts, is formed by the precursor cells within the chorion of the placenta (Cross et al., 2003). Labyrinth development is initiated by embryonic day (E)7.5 when a small cluster of cells within the chorion starts to express Gcm1 (glial cells missing-1), which initiates branching morphogenesis (Anson-Cartwright et al., 2000).

Controlled proliferation and differentiation of trophoblast progenitor cells is essential for placental development. In humans, aberrant trophoblast differentiation leading to placental insufficiency is implicated in several disorders such as miscarriage, pre-eclampsia and intrauterine growth restriction (IUGR) (Kingdom et al., 2000). The transcription factor AP-2γ (TFAP2C) belongs to the activator protein-2 (AP-2) family (Eckert et al., 2005). In mice, Tfap2c expression is first detected in oocytes and all through preimplantation development (Auman et al., 2002; Cao et al., 2015; Choi et al., 2012; Kuckenberg et al., 2010). In mouse preimplantation embryo, Tfap2c has been recently demonstrated to be one of the first expressed trophoblast transcriptional regulators involved in trophoblast specification (Cao et al., 2015). At the blastocyst stage, TFAP2C is mainly detected in TE (Kuckenberg et al., 2010). Upon TE differentiation, TFAP2C protein is found in all trophoblast derivatives. Furthermore, during placental development, TFAP2C is present in SpT, GCs and TGCs including mononuclear S-TGCs that line the maternal sinusoids, but absent in syncytiotrophoblasts (Kuckenberg et al., 2010). In humans, TFAP2C is present in all trophoblast derivatives including extravillous trophoblasts, cytotrophoblasts and syncytiotrophoblasts (Kuckenberg et al., 2012). Its major site of expression is detected in the cytotrophoblast precursor cells, which give rise to the syncytiotrophoblasts (Biadasiewicz et al., 2011). Of note, increased levels of TFAP2C have been reported in pre-eclamptic placentae (Kotani et al., 2009).

Tfap2c-deficient mice die around E7.5 as a result of a defect in the developing ExE (Auman et al., 2002; Werling and Schorle, 2002). Tfap2c has been implicated in the specification and maintenance of trophoblast stem cell (TSC) fate (Kuckenberg et al., 2010; Werling and Schorle, 2002). Genome-wide chromatin immunoprecipitation in TSCs revealed that TFAP2C governs a set of genes involved in TSC self-renewal (Kidder and Palmer, 2010). Although these data emphasize the requirement of TFAP2C in the maintenance of the progenitor state of TE cells, the role of TFAP2C in the differentiating trophoblast is not understood. To examine the effect of loss of TFAP2C in the committed precursor lineages, we used a conditional mouse model to ablate TFAP2C in trophoblast precursor cells at E6.5 and E8.5, respectively. Loss of TFAP2C in ExE at E6.5 leads to a reduced number of trophoblasts resulting in embryonic retardation at E8.5. Loss of TFAP2C in TPBPA⁺ precursor cells arrests the growth of the developing JZ. Downregulation of Akt1 and reduced activation of AKT in the placenta upon loss of TFAP2C inhibits glycogen synthesis. Microarray analysis revealed upregulation of Cdkn1a (previously p21) and several dual-specificity phosphatases such as Dusp1, Dusp4 and Dusp6, which are repressors of mitogen activated protein kinase (MAPK) signaling. TFAP2C binds to the promoter of Cdkn1a and Dusp6, thereby directly regulating the expression of these target genes. The reduced activation of MAPK (pERK1/2, pP38) and AKT pathways upon loss of TFAP2C was further validated in the human choriocarcinoma cell line (JAr). JAr cells represent human trophoblasts and cytotrophoblast cells in vitro in terms of morphology and hormone production (Aladjem and Lueck, 1981; Pattillo et al., 1971). Furthermore, the cells of cytotrophoblast columns in human placenta are analogous to the EPC in mouse, which differentiate into giant cells (Rossant and Cross, 2001). These results suggest that, in trophoblasts, TFAP2C represses Cdkn1a and negative regulators of the MAPK pathway (Dusp1, Dusp4 and Dusp6). TFAP2C is also required for activating the AKT pathway, which is involved in glycogen synthesis.

To study requirement of TFAP2C in early placental development, we took advantage of a mouse model expressing Cre recombinase under the bovine K5 promoter active in early embryogenesis at E6.5 in the ExE (Crish et al., 2013; Ramirez et al., 2004) and crossed it with a mouse harboring a floxed Tfap2c allele (Tfap2c^fl/fl) (Werling and Schorle, 2002) to generate K5Cre:Tfap2c^−/− embryos. At E7.5, IHC showed loss of TFAP2C protein in the cells of ExE and EPC in the K5Cre:Tfap2c^−/− embryos but not in wild-type or heterozygous embryos (herein referred to as control) indicative of efficient loss of TFAP2C (Fig. S1A). K5Cre:Tfap2c^−/− implantations were considerably smaller than controls suggesting a developmental retardation of the embryo (Fig. S1B,C). Hematoxylin and Eosin (H/E)-stained sections revealed no formation of JZ and labyrinth in mutant placentae (Fig. 1A, H/E). Furthermore, in K5Cre:Tfap2c^−/− placentae, the region where the SpT are localized comprised loosely arranged TGCs interspersed by maternal blood (Fig. 1A, H/E, arrowheads). The number of Ki-67⁺ proliferating cells was severely reduced in the mutant placentae (Fig. 1A, Ki-67). The labyrinth of control placentae contained CD31⁺ fetal endothelial blood vessels comprising primitive nucleated fetal blood cells. Interestingly, the mutant placentae showed very few and small fetal endothelial vessels with no fetal blood (Fig. 1A, CD31) suggesting that labyrinth formation was initiated but severely impaired. In situ hybridization (ISH) revealed reduced expression of Prl3d1 (also known as PL1), which is a marker of parietal-TGCs (P-TGCs) and Prl2c2, a marker of SpA-TGCs in K5Cre:Tfap2c^−/− placentae (Fig. 1B). Also, Prl2c2-expressing SpA-TGCs within the mutant placentae showed poor invasion of maternal blood vessels compared with the controls (Fig. 1B, Prl2c2). The number of Tpbpa⁺ SpT cells was markedly reduced and displayed reduced invasion into the decidua (Fig. 1B, Tpbpa). Next, gene expression analysis of E8.5 whole placentae (n=4 each) showed a significant downregulation of the progenitor markers Tpbpa (P=0.009) and Gcm1 (P=0.032), the differentiation markers Prl3d1 (P=0.006), Prl2c2 (P=0.012) and Hand1 (P=0.007), which are important for TGC differentiation (Hemberger et al., 2004) and Ascl2 (P=0.029), which is required for the maintenance of JZ (Tanaka et al., 1997) (Fig. 1C). Thus, loss of TFAP2C in early placental morphogenesis most likely affects the trophoblast stem cell/progenitor compartment, leading to a general reduction of all trophoblast subtypes.

To investigate the loss of TFAP2C in the TPBPA⁺ progenitor cells, Tpbpa-Cre mice (Simmons et al., 2007) were mated with Tfap2c^fl/fl mice (Werling and Schorle, 2002). IHC confirmed loss of TFAP2C in JZ of TpbpaCre:Tfap2c^−/− placentae at E14.5 (Fig. 2A). The JZ was extracted using laser microdissection and RNA was isolated (n=4 each). Here, the level of Tfap2c expression was significantly reduced in the JZ of TpbpaCre:Tfap2c^−/− placentae (P<0.0001, Fig. 2B) further validating the loss of Tfap2c in TPBPA⁺ cells. The remaining level of Tfap2c expression might emanate from TPBPA-negative cells within the JZ or might be due to incomplete excision of the Tfap2c allele in TPBPA-positive cells (Fig. 2B). Next, we observed that the JZ did not further increase in area from E12.5 onwards in the mutant placentae (Fig. 2C). Morphometric analyses demonstrated that the JZ area was comparable in size at E10.5 but significantly retarded in development at E12.5 (n=3 each, P=0.02), resulting in only 50% of the area of wild-type JZ at E14.5 (n=6 each; P=0.007), 75% at E16.5 (n=4 each; P<0.0001) and 80% at E18.5, respectively (n=6 each; P<0.0001; Fig. 2D). However, the overall area of mutant placentae displayed only a non-significant reduction from E14.5 onwards (Fig. S2A). Because we detected a slight increase in the labyrinth area in the mutant placentae at E16.5 and E18.5, we speculate that the growth impairment of the JZ might be compensated by the increase in labyrinth (Fig. S2B). Despite this, the mutant placentae at E16.5 (P=0.01, n=11 control, n=3 mutant) and E18.5 (P=0.0006, n=13 control, n=7 mutant), were significantly lighter compared with the controls (Fig. S2C). The labyrinth of mutant placentae displayed no gross morphological defect and the number of CD31⁺ fetal endothelial vessels within the labyrinth did not change (Fig. S2D). We next found a reduced number of Ki-67⁺ proliferating cells in mutant placentae from E10.5 onwards (Fig. 2E). Quantitative estimation revealed that in mutant placentae at E10.5, 51.87% (P=0.017) and at E12.5 (P=0.0009) only 25.47% of cells were Ki-67⁺ compared with the control (set as 100%) (Fig. 2F). TUNEL assay at E12.5 did not reveal any apoptotic cells in the control and mutant placentae (Fig. S2E). This result strongly suggests that the growth arrest seen in TFAP2C negative JZ is due to a lack of proliferation and not enhanced apoptosis.

To further characterize trophoblast subtypes affected, we performed ISH and found reduced numbers of Tpbpa-expressing cells in the mutant placentae from E12.5 onwards. Tpbpa-expressing cells were scarcely scattered in the JZ at E14.5 and completely absent from E16.5 onwards (Fig. 3A). This suggested that TFAP2C is required to maintain Tpbpa expression in these cells. Furthermore, Prl2c2⁺ SpA-TGCs invading the maternal arteries and Prl3b1⁺ C-TGCs were both reduced in the mutant placentae (Fig. 3B,C). The number of P-TGCs at E12.5 placental sections (n=3 each) in the control and mutant placenta was not altered (Fig. S3A). From E11.5, the GCs start to accumulate glycogen in their cytoplasm and increase in size and are characterized as vacuolated terminally differentiated GCs (Coan et al., 2006). Morphologically, GCs in mutant placentae were much smaller, with only weak accumulation of glycogen and were not as strongly vacuolated, suggesting that they are not terminally differentiated (Fig. 3D). Next, we detected 25% reduction of glycogen in the mutant placentae (P=0.0418) at E14.5 (n=5 mutant; n=9 control; Fig. 3E). Thus, loss of TFAP2C affects all trophoblasts derived from the TPBPA⁺ population.

We next performed an expression microarray analysis (Illumina Mouse WG-6 v2.0). Because microdissection of JZ yields little RNA, samples from TFAP2C mutant (n=3) and control placentae (n=3) at E13.5 were pooled to perform the analysis. A scatter plot revealed that in mutant placentae 146 genes were upregulated and 51 genes were downregulated significantly with fold change ≥1.5 in log₂ scale compared with the control (Fig. 4A). Deregulated genes are summarized in Tables S1 and S2 and a heat map depicting differential expression of deregulated genes is shown in Fig. S4A. Deregulated genes were categorized according to Gene Ontology into several groups (Table S3). Selected genes were further analyzed and are represented in the form of a heat map (Fig. 4B). The pregnancy-specific glycoprotein (Psg) and carcinoembryonic antigen-related adhesion molecules (CEACAMs) are synthesized by the TGCs and SpT cells in the mouse placenta (Kromer et al., 1996; Wynne et al., 2006; Zhou et al., 1997). Interestingly, many of the downregulated genes in the mutant placentae belong to the CEA family (Psg18, Psg22, Psg23, Ceacam13, Ceacam14, Ceacam3) and prolactin-associated protein family (Prl4a1, Prl3c1, Prl7c1, Prlpc3) (Fig. 4B). Additionally, we found upregulation of genes involved in vasculogenesis and angiogenesis (Vegfa, Pdgfa) (Arroyo and Winn, 2008; Mayhew et al., 2004). Genes representative for the different categories were further validated by qRT-PCR. Psg18 was significantly reduced (P=0.0022; Fig. S4B), whereas Vegfa was upregulated more than 3-fold in the mutant placentae (P=0.046; Fig. S4C). We confirmed downregulation of several endocrine hormones synthesized by different trophoblast population at E14.5 by qRT-PCR (n=4 each; Fig. 4C). Gjb3, a marker of differentiated GCs (Coan et al., 2006) was also reduced in the mutant placentae validating our finding that the GCs in the mutant placentae do not accumulate sufficient glycogen (Fig. 4C; fold changes and P values are summarized in Table S4). In line with this, we observed a significant reduction of Akt1 expression in mutant placentae (P=0.0183; Fig. S4D). AKT1 is involved in PI3K-AKT-mediated synthesis of glycogen stores within GCs (Cross et al., 1995; Yang et al., 2003). To confirm loss of activated AKT in the placenta, we performed IHC on E14.5 placental sections and found a 3-fold reduction (P=0.0038) in the number of phospho-AKT (pAKT)-positive cells in the JZ of mutant placenta (Fig. S4E,F).

Ch-TGCs were not altered in the mutant placentae most probably because they are derived from a TPBPA-negative population (Fig. 4C). We did not find any significant changes in the expression levels of Hand1 and Cdkn1c (P57), which are involved in TGC differentiation (Hemberger et al., 2004; Takahashi et al., 2000) (Fig. S3B) or the P-TGC marker Prl3d1 (Fig. 4C). Expression levels of Tfap2c and Gcm1 in the labyrinth were unaffected; however, the S-TGC markers Prl2c2, Ctsq and Prl3b1 were slightly elevated in mutant placentae (Fig. S3C). Also, an increase in the number of Prl3b1-expressing S-TGCs was observed within the labyrinth of the mutant placenta (Fig. S3D).

Because imprinted genes are known to be crucial for placental development, we investigated expression of several imprinted genes and found significant upregulation of H19 (P=0.03), downregulation of Ascl2 (P=0.00054) and Slc38a4 (P<0.0001) (Fig. 4C), whereas levels of Phlda2 and Igf2 were unaffected (Fig. S3E). Deregulation of imprinted genes suggested loss of imprinting. To check this further, we sought to analyze methylation status of the imprinted gene H19, which is known to act as a trans-regulator of imprinted gene network thereby controlling mouse embryonic growth. The H19/Igf2 imprinted control region (ICR), which is located upstream of the H19 promoter contains several differentially methylated regions (DMRs) including CCCTC binding factor (CTCF) binding sites which perform diverse regulatory functions such as transcriptional activation/repression and imprinting (Gabory et al., 2009; Phillips and Corces, 2009). We looked for the methylation pattern of 6 CpGs within a CTCF3 binding site using cells from the JZ of Tfap2c mutant and control placentae. We found no differential methylation (n=14 control, n=15 mutant; Fig. S4G), indicating that imprinting within this region is not affected upon loss of TFAP2C. Loss of TFAP2C also deregulated Tex19.1 (P=0.03; Fig. 4C).

We further utilized GENEMANIA, a web-based tool prioritizing gene interaction based on the available genomics and proteomics data to find interactive gene networks between Tfap2c and deregulated genes (Fig. S5). Tfap2c was linked to several endocrine genes, suggesting coordinated expression of Tfap2c and these genes (Fig. S5A). Interestingly, we found co-expression of Tfap2c and Tpbpa (Fig. S5A), supporting our speculation that upon loss of Tfap2c expression, cells also downregulate expression of Tpbpa in the placenta. The deregulated genes were further subjected to analysis using STRING web-based tool to find known and predicted protein interactions between Tfap2c and deregulated genes (Fig. S5B).

We validated the upregulation of Cdkn1a by more than 3-fold (P=0.00018) in the JZ of the mutant placentae at E14.5 by qRT-PCR (Fig. 5A). Furthermore, several genes involved in repression of the MAPK signaling pathway (Hspa2, Gadd45a, Dusp6, Rasgrf1, Dusp1, Pdgfa, Dusp4, Fos; Table S3; Keyse, 2008) were deregulated (Fig. 4B). Dusp1, Dusp4 and Dusp6 were further validated by qRT-PCR and were significantly upregulated (P=0.00058, P=0.00804, P=0.0182, respectively) upon loss of TFAP2C in the JZ (Fig. 5B). These data demonstrated that TFAP2C represses Cdkn1a and negative regulators of MAPK signaling. Thus, we hypothesized that the MAPK pathway is downregulated in the mutant placentae. To further verify our findings, we analyzed phospho-ERK1/2 (pERK1/2) signal in the JZ at E14.5 (Fig. 5C). On average, 3.2% of cells within the JZ were positive for pERK1/2 in the mutant placentae (n=3) compared with 7.5% positive cells in the control placentae (n=2) (Fig. 5D). Because loss of TFAP2C reduced the activation of ERK1/2 at E14.5, we asked whether the levels of ERK1/2 activation would also be compromised at E7.5 upon loss of TFAP2C in ExE and EPC. IHC at E7.5 showed absence of pERK1/2 signal in the mutant embryos (Fig. S1A). Hence, TFAP2C modulates MAPK signaling in the murine placenta. We next asked whether TFAP2C is able to directly regulate cell proliferation and MAPK signaling by binding to promoter regions of Cdkn1a and Dusp6. To this end, we performed ChIP analysis utilizing trophoblast stem cell TS-6.5 (Kubaczka et al., 2014), differentiated for 5 days as a corollary for placenta. Information about TFAP2C binding sites in the promoter region of Cdkn1a were taken from Schemmer et al. (2013) and the binding site in the Dusp6 promoter region was identified by the rVista algorithm. Both were enriched for TFAP2C (Cdkn1a, 2.3×, P=0.045; Dusp6, 1.7×, P=0.032), implicating both as direct targets of TFAP2C (Fig. 5E).

To further validate that MAPK and AKT signaling pathways are indeed compromised upon loss of TFAP2C, we utilized the human choriocarcinoma JAr cells as a human trophoblast model system (White et al., 1988). JAr cells were transduced with two independent shRNA knockdown constructs targeting TFAP2C and analyzed thereafter. As in murine placenta, expression levels of DUSP1 and DUSP4 were upregulated upon loss of TFAP2C (Fig. 5F,G). Western blot analysis demonstrated that transduction of both shRNA knockdown constructs resulted in a high reduction of TFAP2C protein (Fig. 5H). Downregulation of TFAP2C, led to a dramatic reduction in pERK1/2 and phosphorylated P38 (MAPK14), whereas the total levels of ERK and P38 remained constant (Fig. 5H). Furthermore, upon loss of TFAP2C, AKT signaling was also inhibited (Fig. 5H). Taken together, loss of TFAP2C leads to a downregulation of AKT and MAPK signaling in murine, as well as human trophoblast models.

Given that TFAP2C-deprived JZ in the mutant placentae displays morphological defects from E12.5 onwards with a severely compromised JZ by E14.5 and altered expression of several trophoblast subtypes, we weighed embryos post E14.5 and found that the fetal mass at E16.5 (P<0.0001, n=11 control and n=3 mutant) and E18.5 (P<0.0001, n=13 control and n=7 mutant) was significantly reduced (Fig. 6A). The mutant embryos were approximately 19% lighter from E16.5 onwards and maintained the lower mass until term. After birth, the mutant pups were 17% lighter in the first week and 10% lighter in the second and third week, respectively (n=39 control, n=11 mutant; Fig. 6B). Following weaning, the mass of mutant animals caught up and remained comparable to the mass of the controls. Fig. 6C shows mutant pups at E18.5, postnatal day 4 and 14, respectively. Thus, placental insufficiency caused by loss of TFAP2C within the JZ leads to a fetal growth restriction post E16.5, which is overcome after birth.

We demonstrate that the loss of TFAP2C in the stem/progenitor compartment of the ExE leads to a severe placental malformation with strikingly reduced numbers of trophoblasts. Loss of TFAP2C in TPBPA⁺ cells arrests the development of JZ, affecting TPBPA-derived trophoblasts. Upregulation of Cdkn1a and downregulation of Akt1 upon loss of TFAP2C is indicative of imbalance between proliferation of trophoblasts and differentiation of GCs leading to 75% smaller JZ. Furthermore, TFAP2C represses Dusp1, Dusp4 and Dusp6, which are involved in repressing MAPK signaling. We observed reduced ERK1/2 activation upon loss of TFAP2C in the JZ of placenta. This is achieved by direct regulation of Cdkn1a and Dusp6 by TFAP2C. Knockdown of TFAP2C in JAr cells confirms reduced activation of the MAPK (ERK1/2 and P38) and AKT signaling pathways.

Choi et al. demonstrated increased Cdkn1a expression in TFAP2C-deficient preimplantation embryos, resulting in growth arrest at morula stage, which could be rescued by RNAi-mediated depletion of Cdkn1a (Choi et al., 2012). In line with this, we show that loss of TFAP2C in the JZ leads to upregulation of Cdkn1a and consequently growth arrest. The MAPK signaling pathway has been demonstrated to be essential in trophoblast differentiation. Although expression of ERK1/2 (MAPK3 and MAPK1) in the human villous cytotrophoblast cells but not in the syncytiotrophoblasts has been observed throughout pregnancy, activated ERK1/2 in the villous cytotrophoblast cells is detected up to 12 weeks of gestation (Kita et al., 2003). The presence of activated ERK1/2 within the committed cytotrophoblast progenitors both in human placenta and primary trophoblast culture suggests that activation of the MAPK pathway is important for initiation of trophoblast differentiation (Daoud et al., 2005). In mice, disruption of the Erk2 gene leads to embryonic lethality due to failure in formation of EPC and ExE (Saba-El-Leil et al., 2003). Also, knockdown of TFAP2C reduces activation of the ERK1/2 pathway in MCF-7 cells (Spanheimer et al., 2014). Thus, TFAP2C stimulates the MAPK signaling pathway in different cell types. TFAP2C does seem to directly affect proliferation and MAPK signaling in differentiated trophoblast by binding to the promoter of its targets, Cdkn1a and Dusp6, which are involved in repressing MAPK signaling. In 2010, Kidder and Palmer showed that promoters of several Dusp genes are bound by TFAP2C in murine TSCs (Kidder and Palmer, 2010). Loss of TFAP2C in murine placenta, as well as in JAr cells, leads to upregulation of DUSPs, suggesting that TFAP2C seems to regulate MAPK activity by suppressing levels of DUSPs.

Akt1 is involved in placental development and fetal growth in mice because Akt1^−/− placenta shows a complete loss of GCs causing placental dysfunction and growth retardation of the fetus (Yang et al., 2003). Furthermore, Prl2^−/− (phosphatase of regenerating liver 2) placenta also displays reduced activation of AKT, leading to inhibition of glycogen synthesis (Dong et al., 2012). Mechanistically, pAKT inhibits GSK3B, which leads to glycogen synthesis (Cross et al., 1995; Diehl et al., 1998). Thus, the reduced expression of Akt1 and reduced activation of AKT observed upon loss of TFAP2C leads to a decrease in glycogen stores.

Does apoptosis contribute to the growth arrest of the JZ? Ablation of TPBPA⁺ cells with diptheria toxin led to an embryonic lethality at E11.5 (Hu and Cross, 2011). If, upon loss of TFAP2C, the cells within the JZ were to undergo apoptosis, it should have resembled a phenocopy of the TPBPA-DTA mouse model, with much earlier embryonic lethality. Therefore, loss of TFAP2C within the JZ affects proliferation and not apoptosis of trophoblasts derived from TPBPA⁺ progenitors. The fact that there is no apoptosis in the JZ and the remnant cells are no longer TPBPA⁺ suggests that loss of TFAP2C directly leads to a loss of Tpbpa expression.

Loss of TFAP2C results in a severe deficit of hormones. These are synthesized by the TGCs and SpT cells in the mouse placenta and syncytiotrophoblast cells in the human placenta (Kromer et al., 1996; Wynne et al., 2006; Zhou et al., 1997). In humans, reduced levels of placental lactogens and pregnancy-specific glycoproteins in the maternal blood are associated with impaired placental function, leading to conditions such as IUGR and pre-eclampsia (Bersinger and Odegard, 2004; Pihl et al., 2009). Upon loss of TFAP2C, we see reduced number of SpA-TGCs, suggesting inefficient vascular remodeling, leading to restricted blood flow into the placenta. The increased expression of Vegfa and Pdgfa transcripts upon loss of TFAP2C are most likely a compensation for the inefficient angiogenesis and vascular remodeling.

Expression of imprinted genes has been shown to be crucial in the development of placenta. Ascl2-null mutants die at E10.5 as a result of placental failure due to lack of an SpT layer, reduced labyrinth and an expansion of the P-TGC layer (Guillemot et al., 1994). Unlike the phenotype associated with a complete loss of Ascl2, we observed reduced levels of Ascl2 upon loss of TFAP2C, which reduces the SpT and GCs in a similar manner but does not completely lack the JZ. Oh-McGinnis, also reported an expansion in P-TGC layer in Ascl2^−/− placenta (Oh-McGinnis et al., 2011) which does not correlate with our finding where P-TGCs are unaffected. P-TGCs are also derived from TPBPA-negative population and hence could be compensated as the loss of TFAP2C is specifically in TPBPA⁺ cells in this model system. Interestingly, the placenta from H19^−/− mice display increased number of GCs and glycogen reserves, which is correlated with an increase in AKT protein level (Esquiliano et al., 2009). Upon loss of TFAP2C, we observed an increase in H19 expression and a smaller placenta with reduced glycogen stores. We also found reduced expression of Tex19.1 upon loss of TFAP2C. Tex19.1 is a genome defense gene involved in placental development and embryos exhibit growth retardation with small placentae upon loss of Tex19.1 because of a reduction in the number of SpT, GCs and S-TGCs (Reichmann et al., 2013). TFAP2C mutant placentae, however, do not show a reduction in S-TGCs. On the contrary, S-TGCs lining the maternal sinusoids seem to be increased in number in the TFAP2C mutant labyrinth. Deregulated levels of SLC38A4, a neutral amino acid transporter in the placenta, have been associated with abnormal fetal birth weight in humans (Desforges et al., 2006; Li et al., 2012).

Fig. 7 summarizes the role of TFAP2C in murine placental development. Although JZ has been shown to be essential for embryonic survival (Guillemot et al., 1994), its reduction by three quarters upon deletion of TFAP2C leads to growth retardation of the embryo. Here, we present a non-invasive genetic model of IUGR which allows the advantage of studying the growth-restricted embryos together with the control embryos within the same uterine setting, thus overcoming intraspecies heterogeneity and offering a more precise analysis. IUGR is often associated with an increased risk of metabolic disorders and cardiovascular diseases later in adulthood (Godfrey, 2002). Thus, our model representing IUGR is a useful tool in gaining further insights into fetal metabolic programming. Additionally, IUGR models such as this could allow us to answer key questions in the hypothesis of developmental origin of diseases.

All experiments were conducted according to the German law of animal protection and in agreement with the approval of the local institutional animal care committees (Landesamt für Natur, Umwelt und Verbraucherschutz, North Rhine-Westphalia; approval ID: #8.87-50.10.31.08.238). The experiments were conducted in accordance with the International Guiding Principles for Biomedical Research Involving Animals as announced by the Society for the Study of Reproduction.

129-SV Tpbpa-Cre transgenic mice (Simmons et al., 2007) and male B6 K5-Cre transgenic mouse (Ramirez et al., 2004) were crossed with female 129-SV Tfap2c^fl/fl mice (Werling and Schorle, 2002) to generate TpbpaCre:Tfap2c^−/− and K5Cre:Tfap2c^−/− placentae, respectively, and the pregnant females were dissected at embryonic days E7.5-E18.5 (noon on the day of the vaginal plug was designated as E0.5). All animals used were 3-6 months old. Genotype of the embryos was determined by isolating tail DNA and performing PCR, using previously described primers listed in Table S5 (Werling and Schorle, 2002). Tissue preparation and histology staining was done using standard protocols as described in supplementary Materials and Methods.

For the detection of glycogen stores in the cells of the placentae, routine PAS (periodic acid-Schiff) reaction was performed (Autostainer 480, Medac, Germany). Briefly, the sections were deparaffinized, rehydrated with several ethanol steps, incubated in periodic acid, washed in tap water, incubated in Schiff reagent and counterstained with hematoxylin. The glycogen was extracted from E14.5 whole placenta as previously described (Lo et al., 1970). Frozen placentae were dissolved into 30% KOH by boiling for 20-30 min. Samples were cooled and incubated on ice with 1.2 volumes of 95% ethanol for 30 min. The samples were centrifuged and the glycogen precipitate was dissolved in distilled water. The amount of glycogen was calculated using standard biochemical colorimetric assay with Phenol and sulphuric acid.

Cryo sections of 10 µm thickness were cut from the frozen placentae and RNA in situ hybridization was performed according to the established protocol (Simmons et al., 2007). Detailed protocol described in supplementary Materials and Methods. The following probes were used: Prl3d1 (Pl1), Prl3b1 (Pl2), Prl2c2 (Plf) and Tpbpa (Simmons et al., 2007).

The DeadEnd Colorimetric TUNEL system was used to detect apoptosis (Promega, G7130) according to the manufacturer's instructions.

Control and Tpbpa:Tfap2c^−/− placentae from day E13.5 and E14.5 were collected, snap frozen in liquid nitrogen and stored at −80°C. Laser microdissection was performed as described previously (Kaiser et al., 2015). Detailed protocol is given in supplementary Materials and Methods. The samples of SpT and labyrinth area were collected separately and RNA extraction was performed immediately with the RNeasy Micro Kit (Qiagen). Additionally, whole placental RNA from E8.5 K5Cre:Tfap2c^−/− mice was isolated using Trizol reagent method. 1 µl total RNA from laser microdissection Tpbpa:Tfap2c^−/− sections and 1 µg RNA from E8.5 K5Cre:Tfap2c^−/− placentae were used for cDNA synthesis by Revert Aid Premium (Fermentas, ThermoScientific). Maxima SYBR Green Master Mix (Fermentas, ThermoScientific) was used to perform quantitative real-time PCR (qRT-PCR) on ViiA 7 (Applied Biosystems, Life Technologies). For primer sequences, see Table S6. Expression of target genes was normalized to glyceraldehyde-3-phosphate dehydrogenase (GAPDH). Each reaction was performed in triplicate. Microarray analysis using Illumina Mouse WG-6 v2.0 expression bead chip was performed with 200 ng total RNA pooled from three different laser microdissected samples each of control and mutant placentae. Data pre-processing was performed using the lumi pipeline (Du et al., 2008) using quantile normalization and a variance stabilizing transformation (Lin et al., 2008). Scatter plots were generated using R v3.1.3 software (http://www.R-project.org). The package pheatmap available from CRAN was used to generate the heatmap. Log-transformed expression values were used for the preselected genes in the heatmap. Genes with fold change ≥1.5 log₂ scale are represented in the heatmap. Microarray data generated in the course of this work were deposited in National Center for Biotechnology Information's Gene Expression Omnibus (GEO) database (Edgar et al., 2002) and are accessible through GEO Series accession number GSE70962.

JZ from 8 serial sections of 20 µm each of control and Tpbpa:Tfap2c^−/− placentae were scraped off and DNA was isolated using a QIAamp FFPE DNA isolation kit according to the manufacturer's protocol (Qiagen). 500 ng gDNA was used for bisulfite conversion using EZ DNA Methylation-Direct Kit (ZYMO Research) following the manufacturer's recommendations. Following conversion, bisulfite-converted DNA was PCR amplified (primer sequences in Table S6), cloned into pCR2.1 vector and sequenced. DNA methylation profile over 6 CpG islands in the H19/Igf2 ICR region containing CTCF3 binding site was analyzed. Data analysis was performed using BISMA online tool (Rohde et al., 2008).

Trophoblast stem cells L5 were differentiated in standard medium for 5 days (Kubaczka et al., 2014). ChIP was performed by enzymatic shearing using ChIP kit (Cell Signaling, 9003S) according to the manufacturer's instructions. For immunoprecipitation (n=2), 5 µg antibody against TFAP2C (clone H77/sc-8977 X; Santa Cruz Biotechnology, Santa Cruz, USA) was used. As a negative control, a species-matched rabbit-IgG antibody provided with the kit was used (Cell Signaling, 9003S). Primers used for q-PCR following ChIP are listed in Table S6.

Human choriocarcinoma JAr cells were provided by Dr Daniel Nettersheim and tested for contamination (Nettersheim et al., 2013) and were grown and maintained in DMEM medium (10% FCS, 1% penicillin/streptomycin, 200 mM L-glutamine) at 37°C and 5% CO₂. For knockdown of TFAP2C, two different shRNA constructs were used (Table S7). Production of VSV-G pseudotyped retroviral particles was as following: HEK293T cells (maintained in DMEM supplemented with 10% FCS, 2 mM L-glutamine, 1% penicillin/streptomycin) were seeded at 80-90% confluency and transfected after 6-7 h by calcium phosphate transfection with the following plasmids at a 9:9:1 ratio: (1) pRetroSuper, (2) pCMV gag-pol, (3) pCMV VSV-G. The medium was replenished the following day in the morning and afternoon. Viral supernatants were collected at day 2 after transfection and were filtered through a 0.45-µm syringe filter. Briefly, viral supernatants were added to the cells for 24 h followed by growth in normal medium. Selection was done using 1 µg/ml puromycin for 5 days after transduction and the proteins were harvested after 1 week of transduction. Western blotting was performed as previously described (Kuckenberg et al., 2010). Detailed protocol and antibodies are listed in supplementary Materials and Methods.

Statistical significance (P-value) was determined using the Student's t-test (unpaired two-tailed distribution for Tpbpa:Tfap2c^−/− placentae and paired two-tailed distribution for K5Cre:Tfap2c^−/− placentae). All data are represented as means±s.d. Values of P≤0.05 were considered to be statistically significant. All graphs were made using GraphPad Prism software.

We thank Gaby Beine and Susanne Steiner for excellent technical assistance and Angela Egert, Andrea Jaeger and Renate Gammel for helping with mouse work. Additionally, we thank Andrea Hoffmann, Department of Genomics, Life and Brain Center, Bonn for performing Illumina expression arrays.