Nuclear Factor Erythroid-derived 2 (Nfe2) Regulates JunD DNA-binding Activity via Acetylation

Background: Nfe2 restricts Gcm1 expression, placental vascularization, and embryonic growth. Results: Nfe2 induces hypoacetylation of JunD, thus limiting JunD binding to the Gcm1 promoter (at −1441). Conclusion: In trophoblast cells Nfe2 negatively controls Gcm1 expression and syncytiotrophoblast formation by repressing JunD-binding activity. Significance: This identifies a novel, acetylation dependent interaction of bZip transcription factors regulating placental and embryonic development. We recently demonstrated that the bZip transcription factor nuclear factor erythroid-derived 2 (Nfe2) represses protein acetylation and expression of the transcription factor glial cell missing 1 (Gcm1) in trophoblast cells, preventing excess syncytiotrophoblast formation and permitting normal placental vascularization and embryonic growth. However, the Gcm1 promoter lacks a Nfe2-binding site and hence the mechanisms linking Nfe2 and Gcm1 expression remained unknown. Here we show that Nfe2 represses JunD DNA-binding activity to the Gcm1 promoter during syncytiotrophoblast differentiation. Interventional studies using knockdown and knockin approaches show that enhanced JunD DNA-binding activity is required for increased expression of Gcm1 and syncytiotrophoblast formation as well as impaired placental vascularization and reduced growth of Nfe2−/− embryos. Induction of Gcm1 expression requires binding of JunD to the −1441 site within the Gcm1 promoter, which is distinct from the −1314 site previously shown to induce Gcm1 expression by other bZip transcription factors. Nfe2 modulates JunD binding to the Gcm1 promoter via acetylation, as reducing JunD acetylation using the histone acetyltransferase inhibitor curcumin reverses the increased JunD DNA-binding activity observed in the absence of Nfe2. This identifies a novel mechanism through which bZip transcription factors interact. Within the placenta this interaction regulates Gcm1 expression, syncytiotrophoblast formation, placental vascularization, and embryonic growth.

We recently identified a novel function of the bZip transcription factor Nfe2 in trophoblast cells (1). In trophoblast cells Nfe2 prevents excess syncytiotrophoblast formation during the later pregnancy stages in the mouse (after day 14.5 p.c., 2 corresponding to the 3rd trimester), thus ensuring sufficient vascularization of the placenta and normal embryonic growth (1). This new function of Nfe2 may provide novel insight into mechanisms underlying placental dysfunction and intrauterine growth restriction. These are frequent but incompletely understood pregnancy complications associated not only with increased peripartal morbidity and mortality of the newborn, but also with diseases such as diabetes mellitus or cardiovascular diseases later on in adult life (2,3). Hence, deciphering the mechanism through which Nfe2 modulates trophoblast differentiation and placental development and function may reveal new diagnostic or therapeutic targets for placental dysfunction and intrauterine growth restriction.
Nfe2 is the larger (45 kDa) subunit of the heterodimeric transcription factor NFE2. Until recently expression of Nfe2 was thought to be restricted to hematopoietic cells (cells of the megakaryocyte, erythrocyte, or mast cell lineage) (4). We identified a function of Nfe2 in nonhematopoietic cells (1). In trophoblast cells Nfe2 represses Gcm1 (also referred to as GCMa) activity (1). Gcm1 belongs to a unique family of transcription factors hallmarked by a highly conserved zinc-coordinated DNA-binding domain (5). Gcm1 is required for branching and syncytiotrophoblast fusion during placental development, and both loss and gain of Gcm1 function disturb syncytiotropho-* This work was supported by Deutsche Forschungsgemeinschaft Grants IS 67/4-1 and IS 67/4-2 (to B. I.) and a grant of the German Israeli Foundation (to B. I. and P. P. N.). □ S This article contains supplemental Figs. S1 and S2. 1 To whom correspondence should be addressed. Tel.: 49-0-391-13900; Fax: 49-0-391-13902; E-mail: berend.isermann@med.ovgu.de or ikcp@med. ovgu.de.
blast formation and placental development (1,6). Loss of Nfe2 function increases acetylation of histone H4 within the Gcm1 promoter and of Gcm1 itself, enhancing Gcm1 expression in murine placenta and trophoblast cells (1). Although these studies identified a novel function of Nfe2 as a repressor during syncytiotrophoblast differentiation, the positive gene regulators repressed by Nfe2 during normal trophoblast and placental differentiation remain unknown. Based on in silico analyses of the Gcm1 promoter the 6000-bp 5Ј of the ATG lack potential Nfe2-binding sites, indicating that Nfe2 must regulate Gcm1 expression through an indirect mechanism. Interaction of Nfe2 with CBP or other bZip transcription factors has been reported and may be mechanistically linked to the developmental function of Nfe2 (1,7,8). Indeed, bZip transcription factors have established functions for Gcm1 expression and placental development. Thus, the bZip transcription factors CREB and OASIS stimulate Gcm1 expression in vitro and enhance expression of syncytiotrophoblast markers in vivo (9). The relevant cAMP response element (CRE)-binding site was located at Ϫ1337 within the mouse Gcm1 promoter. In vivo, expression of CREB was replaced by expression of OASIS around late midgestation (E13.5), at which time the hemochorial placenta is fully established. This demonstrates a dynamic regulation of bZip transcription factors during placental development.
Other bZip transcription factors with known function during placentation belong to the AP-1 proteins. Knockout studies in mice revealed embryonic lethal placental defects in JunB and Fra1-deficient embryos (10,11), whereas placental development was normal in mice lacking other AP-1 proteins. Placental JunB or Fra1 deficiency impairs vascularization of the labyrinthine layer, causing early embryonic lethality between days 8.5 and 10.5 p.c. (10,11).
Whether Nfe2 regulates syncytiotrophoblast formation and vascularization of the placental labyrinthine layer through an interaction with TORC1, OASIS, JunB, Fra1, or yet to be identified transcription factors remained hitherto unknown. Within the current study we explored the interaction partners of Nfe2 during placental development, identifying a so far unknown interaction of Nfe2 with JunD. During placental development Nfe2 regulates through an acetylation dependent mechanism JunD-binding activity to the Ϫ1441 site of the Gcm1 promoter, regulating syncytiotrophoblast formation, placental vascularization, and embryonic growth.

EXPERIMENTAL PROCEDURES
Materials-Heparin sodium salt from porcine intestinal mucosa, human acidic fibroblast growth factor 4 (aFGF4), trichostatin A (TSA), Polybrene, and puromycin dihydrochloride from Streptomyces alboniger were obtained from Sigma. All cell culture media, FBS, glutamine, penicillin/streptomycin mixture, neuromycin, nonessential amino acids, and sodium pyruvate were from PAA, Cölbe, Germany. P24 tag antigen ELISA and the CHIP assay kit were from BioCat, Heidelberg, Germany. ECL reagent was from Amersham Biosciences. Murine Nfe2 and JunD shRNA constructs were from Open Biosystems, Huntsville, AL. GoTaq DNA polymerase, AP-1, and CREB consensus oligonucleotides for EMSA were obtained from Pro-mega. Rabbit polyclonal IgG antibodies against Nfe2, c-Jun, JunD, JunB, FosB, and Fra1 were purchased from Santa Cruz Biotechnology. Rabbit polyclonal IgG antibody against acetylated lysine was purchased from Cell Signaling Technology, Boston, MA. CellMask Deep Red stain was bought from Invitrogen. QuikChange XL Site-directed Mutagenesis kit was purchased from Stratagene. All primers were synthesized by Thermo Scientific, Langenselbold, Germany.
Mice-Nfe2-deficient mice were kindly provided by R. Shivdasani (Harvard Medical School, Cambridge, MA). All animals were housed and the experiments were performed at the Interdisciplinary Biomedical Research institution (IBF) of the University of Heidelberg, Germany. For in vivo intervention studies we used TSA (15 g/mouse (13)) or valproic acid (sodium salt, 155 mM, 25 ml/kg of body weight (14)) as HDAC (histone deacetylase) inhibitors and epigallocatechin-3-gallate (50 mg/kg in saline (15)) or curcumin (0.5 mg/kg (16)) as HAT inhibitors. Animal experiments were conducted following standards and procedures approved by the local Animal Care and Use Committee (Regierungspräsidium Karlsruhe, Germany).
Cell Culture-HEK293T and HEK293 cells were kindly provided by Dr. Peter Seeburg (Max-Planck-Institute for Medical Research, Heidelberg, Germany). They were used to generate lentiviral particles and perform the luciferase assay. These cells were maintained in DMEM supplemented with 4.5 g/ml of glucose, 10% FBS, 2 mM L-glutamine, 50 g/ml of penicillin/streptomycin, and 4 g/ml of neomycin at 37°C in a humidified incubator with 5% CO 2 . For transfection experiments medium without antibiotics was used.
In Vitro Syncytium Formation-Syncytiotrophoblast formation was monitored in nearly confluent TS cells differentiated for 6 days, washed with PBS, incubated with 2.5 M CellMask Deep Red for 5 min at 37°C, and then fixed for 3 min in 3.75% formaldehyde at RT. Cells were counterstained with 0.5 g/ml of DAPI in PBS for 10 min at RT in the dark. For quantification of syncytiotrophoblasts at least 30 random images per slide were captured (University of Texas Health Science Center San Antonio ImageTool software) and a blinded investigator determined the frequency of syncytial cells in these random images (1).
Production of Lentiviral Particles-Highly purified plasmid DNA prepared by alkaline lysis preparation was used for lentiviral particle production. Vesicular stomatitis virus-G pseudotyped lentiviral particles were generated as described (1,24). Briefly, HEK 293T cells were transfected with pLKO1-JunD, pLV-JunD, or pLV-JunD-shRes plasmids together with the packaging plasmid (pCMV-dR8.91) and vesicular stomatitis virus-G-expressing plasmid using the standard calcium phosphate method. Lentiviral particles were harvested 2 and 3 days after transfection and concentrated by ultracentrifugation (50,000 ϫ g, 2 h, 2 times). After resuspension in HBSS buffer the viral particle concentration was determined by measuring the p24 gag antigen by ELISA according to the manufacturer's instruction (BioCat, Heidelberg, Germany). Viral particles were stored at Ϫ80°C.
Luciferase Assay-The HEK293 cells were transfected in 60-mm dishes with 1 g of luciferase reporter construct and 0.1 g of CMV driven expression vectors using a standard calcium phosphate method. At 48 h post-transfection, cells were harvested for luciferase assays using the luciferase assay system according to the manufacturer's instruction.
Transduction of Preimplantation Embryos-Female Nfe2 Ϫ/Ϫ or wild type (WT) mice were superovulated by intraperitoneal injection of pregnant mare's serum gonadotropin (5 units) followed by human chorionic gonadotropin (5 units) 48 h later and then mated with Nfe2 Ϫ/Ϫ or WT males, respectively. Twocell stage embryos were collected from the females at day 1.5 p.c. and incubated for 2 days to obtain blastocysts. The zona pellucida was removed with acidic Tyrode solution. Zona pellucida-free embryos were incubated individually in 50 l of KSOM medium containing JunD with either shRNA or JunD expressing lentiviral particles (1 ϫ 10 3 ng of p24/ml) for 4 h (1, 12). Transduced blastocysts were transferred into pseudopregnant females. The transfer day was counted as day 3.5 p.c. and analyses were performed at day 14.5 or 16.5 p.c. Embryos generated from blastocysts transduced with lentiviral particles generated from nonspecific shRNA or empty vector served as controls.
Morphological Analyses of Placenta-Placental tissues were obtained from Nfe2 Ϫ/Ϫ or WT females after lentiviral transduction at day 14.5 p.c. Half of the placental tissue was fixed in 4% buffered formalin, whereas the other half was further divided and stored either in RNAlater or quick frozen in liquid nitrogen for protein isolation. The fixed tissue was embedded in paraffin, sectioned, and stained with hematoxylin and eosin.
For vascular space quantification randomly selected microscopic fields (magnification ϫ10) of the labyrinthine region from 3 nonconsecutive placental sections (each section Ն10 images) were acquired at 1280 ϫ 960 resolution using a digital camera (Nikon digital sight DS-U1) connected to a light microscope (Nikon Eclipse TE2000). We implemented an image analysis routine using NIS Elements AR2.30 imaging software. Briefly, after acquisition, the images underwent an automated analysis procedure and were given a color threshold to cover the area corresponding to the lumen of blood spaces. The coverage percentage was calculated as the ratio between the number of pixels covered by the area defined by the threshold and the overall number of pixels in the image (1). A blinded investigator performed analyses.
Transmission Electron Microscopic Analyses of Placenta-Placenta tissues obtained at day 16.5 p.c. were fixed with 2.5% glutaraldehyde, 2.5% polyvinylidone 25, 0.1 M sodium cacodylate, pH 7.4. After washing with 0.1 M sodium cacodylate buffer, pH 7.4, samples were post-fixed with 2% osmium tetroxide, 1.5% potassium ferrocyanide in 0.1 M sodium cacodylate buffer, pH 7.4, for 1 h, washed, contrasted en bloc with uranyl acetate, dehydrated with an ascending series of ethanol, and embedded in glycid ether 100-based resin. Ultrathin sections were cut with a Reichert ULTRACUT S ultramicrotome (Leica Microsystems, Wetzlar, Germany), contrasted with uranyl acetate and lead citrate, and viewed with an EM 10 CR electron microscope (Carl Zeiss NTS, Oberkochen, Germany).
Gene Knockdown and Knockin in Mouse Trophoblast Cells-Generation of stably transfected Nfe2 knockdown mouse trophoblast cells (Nfe2 kd ) has been previously described (1). Transient knockdown or knockin of JunD in control, Nfe2 kd , or Nfe2 kd JunD kd TS cells was performed by lentiviral transduction. Briefly, pLKO1-JunD, pLV-JunD-shRes, or pLV-JunD viral particles (5 ϫ 10 5 TU/ml) were generated as described above and used to infect Nfe2 kd (yielding Nfe2 kd JunD kd cells), Nfe2 kd JunD kd (yielding Nfe2 kd JunD kdRes cells), or control (yielding Nfe2 c JunD ϩ cells) TS cells, respectively, in the presence of 8 g/ml of Polybrene in TS cell medium. The efficiency of the JunD knockdown or knockin was determined after 3 days of differentiation by RT-PCR and EMSA. In all experiments cells transduced with viral particles generated from nonspecific shRNA or empty pLV-CAG1.1 plasmid served as controls.
Quantitative Real Time PCR (qRT-PCR)-RNA was isolated from mouse placental tissue or TS cells. Placental tissues were stored initially in RNAlater and placed into TRIzol for RNA isolation, whereas TRIzol was added directly to TS cells. Following incubating with TRIzol (5 min at RT), 0.2 ml of chloroform/1 ml of TRIzol was added, mixed, and centrifuged (12,000 ϫ g, 15 min, 4°C). The aqueous phase containing the RNA was transferred to a fresh tube and RNA was precipitated by adding 0.5 ml of 2-propanol/1 ml of TRIzol reagent initially used. Samples were incubated at RT (10 min) and centrifuged (12,000 ϫ g, 10 min, 4°C). Supernatant was removed and RNA was washed by adding 1 ml of 75% ethanol/1 ml of TRIzol. Samples were vortexed, centrifuged (7,500 ϫ g, 5 min, 4°C), the RNA pellet was air dried (5 min), dissolved in 20 l of diethyl pyrocarbonate/water, and incubated for 10 min at 55°C. The RNA concentration was measured in a photometer, and RNA purity and integrity were verified on a 1.8% agarose gel. cDNA was synthesized according to the manufacturer's protocol (SuperScript First-Strand Synthesis System). All primers were custom synthesized by Thermo Fisher Scientific and quantitative RT-PCR was performed as described previously (1,26).
Chromatin Immunoprecipitation-Chromatin immunoprecipitation (ChIP) was performed on placental tissue extracts using the Epiquik ChIP assay kit for tissue following the manufacturer's instructions. Briefly, tissue samples were minced into small fragments using a scalpel and exposed to 1% formaldehyde in TS cell medium for cross-linking (20 min at RT on a rocking platform). Tissue extracts were washed in 125 mM glycine in TS cell medium, centrifuged, and the pellet was disaggregated using a Dounce homogenizer in 1 ml of homogenization buffer/200 mg of tissue. Following centrifugation the pellet was resuspended in lysis buffer CP3 containing protease inhibitors and incubated on ice for 10 min. DNA was sheared by sonication (5 intervals, each 7 s with 5-s breaks). Following centrifugation (8,000 ϫ g, 10 min, 4°C) the supernatant was diluted with buffer CP4, an aliquot was removed as "input" control, and from the remaining solution proteins were precipitated using rabbit polyclonal anti-JunD IgG antibody. Following incubation (60 min at RT) supernatants were removed and the precipitates were washed 6 times (buffer CP1). DNA was released using proteinase K in buffer CP5 (15 min at 65°C) and then captured, washed, and finally eluted using spin columns. Purified DNA was stored at Ϫ20°C until final analyses. Recovered DNA was analyzed by PCR using the following primers for the various mouse Gcm1 promoter regions: Ϫ3150 to Ϫ2950 (5Ј-GAA GAA ACC ACC TAA AGC TA-3Ј and 5Ј-GGG AAG ACT CAC CCA ACT CA-3Ј), Ϫ2930 to Ϫ2720 (5Ј-AGG ATG CTG TAT AAG CCC TTC-3Ј and 5Ј-CCT CAG GCA TGC TAT GGG TGA-3Ј), Ϫ2140 to Ϫ1930 (5Ј-TGG AAC ACT AGT TGA AAT AG-3Ј and 5Ј-GCA TTC ATA ACC CAG GCA TC-3Ј), Ϫ1910 to Ϫ1720 (5Ј-CAG AAT GTA ATA TCT AGG GA-3Ј and 5Ј-GGA GTA GCT GAC CTT ACC TG-3Ј), Ϫ1630 to Ϫ1460 (5Ј-ACG GAG GCG ACA AAG AC-3Ј and 5Ј-CTA AAT TAC GCT TGA TGC GC-3Ј), Ϫ1450 to Ϫ1300 (5Ј-CAG CGT GGC CTC TGA TTC AT-3Ј and 5Ј-GAG ATG GGA GTG ACC TCC TA-3Ј), and Ϫ1100 to Ϫ900 (5Ј-AAA AAC CCT TCG ACT CAG AG-3Ј and 5Ј-TTA TCA TTG TGG CTA ATC CA-3Ј). Cycling conditions were 35 cycles with an annealing temperature of 58°C.
Statistical Analysis-All in vitro experiments were performed at least three times in duplicates. All data were expressed as mean Ϯ S.E. Statistical analyses were performed using Student's t test or analysis of variance as appropriate. Post FEBRUARY 17, 2012 • VOLUME 287 • NUMBER 8 hoc comparisons of analysis of variance were corrected using the method of Tukey. Excel and StatistiXL software were used for all statistical analyses. Statistical significance was accepted at p Ͻ 0.05.

Nfe2
Modulates AP-1 Binding in Murine Placenta-We have recently shown that Nfe2 suppresses Gcm1 expression and syncytiotrophoblast formation in trophoblast cells. Considering the negative gene regulatory function of Nfe2 in regard to Gcm1 and the absence of a Nfe2-binding site within the first 6000 bp of the Gcm1 promoter (both in humans and mice) we hypothesized that bZip transcription factor Nfe2 regulates Gcm1 indirectly, potentially by interacting with other bZip transcription factors. To this end we analyzed binding activity to two bZip transcription factor-binding sites, which have been previously shown to regulate placental development and/or Gcm1 expression (CREB and AP-1) (9 -11, 27). We observed a marked induction of AP-1 DNA-binding activity (251 versus 100% in Nfe2 ϩ/Ϫ embryos, p Ͻ 0.001), but not of CREB DNAbinding activity (121 versus 100% in Nfe2 ϩ/Ϫ embryos, p ϭ 0.8, Fig. 1, A and B), in placental tissue nuclear extracts obtained from day 14.5 p.c. Nfe2 Ϫ/Ϫ murine placenta. Next, we performed EMSA using cellular nuclear extracts obtained from control (Nfe2 c ) and Nfe2 knockdown (Nfe2 kd ) trophoblast cells differentiated for 3 days. Again, AP-1 DNA-binding activity (206 versus 100% in Nfe2 c , p Ͻ 0.01), but not CREB DNAbinding activity (117 versus 100% in Nfe2 c , p ϭ 0.12, Fig. 1, C  and D), was markedly enhanced in Nfe2 kd samples. These results suggest that in the absence of Nfe2 expression of Gcm1 is induced through an AP-1-dependent mechanism.  Fig. 1F). No relevant contribution of c-Jun, JunB, Fra1, or FosB to the enhanced AP-1-binding activity could be detected. Thus, in the absence of Nfe2, binding activity of JunD is markedly increased. This indicates that Nfe2 may reduce Gcm1 expression and trophoblast syncytium formation by repressing JunD, an AP-1 protein not previously known to be relevant for placental development.
Additional experiments were conducted to confirm the role of JunD in regulating syncytiotrophoblast formation. First, transient overexpression of a shRNA-resistant JunD mutant in Nfe2 kd JunD kd trophoblast cells (Nfe2 kd JunD kdRes ) reversed the phenotype to that observed in single Nfe2 kd trophoblast cells. Thus, syncytiotrophoblast formation increased to 2.6 per random image (Fig. 2, A and C) and the expression levels of Gcm1 (173%), Cebpa (131%), and Synb (158%) increased to levels not different from those in single Nfe2 kd trophoblast cells (Fig. 2C). Second, isolated overexpression of JunD in trophoblast cells (Nfe2 c JunD ϩ ) increased syncytiotrophoblast formation (to 1.93 syncytial trophoblast cells per random image versus 0.56 in Nfe2 c , p Ͻ 0.01, Fig. 2, A and B) and induced expression of Gcm1 (149%, p Ͻ 0.001), Cebpa (120%, p Ͻ 0.001), and Synb (138%, p Ͻ 0.001, Fig. 2C). These experiments establish that gain of JunD function is sufficient for increased syncytiotrophoblast formation, confirming the mechanistic relevance of increased JunD activity for the enhanced syncytiotrophoblast formation in the absence of Nfe2. Furthermore, these in vitro studies demonstrate that syncytiotrophoblast formation depends on a cell autonomous cross-talk between Nfe2 and JunD.
Nfe2 and JunD Coordinately Control Placental Vascularization and Syncytiotrophoblast Formation in Vivo-We have recently shown that Nfe2 restricts syncytiotrophoblast formation, thus regulating vascularization of the labyrinthine layer of the placenta and embryonic growth (1). Given the before men-tioned results we next explored whether cross-coupling of Nfe2 with JunD in trophoblast cells represses Gcm1 in vivo. We reduced JunD expression in cells of the trophectodermal lineage by transducing Nfe2 Ϫ/Ϫ blastocysts with JunD shRNA expressing lentiviral particles. This resulted in a marked reduction of JunD expression and of JunD DNA-binding activity within the placenta of Nfe2 Ϫ/Ϫ JunD kd embryos as compared with Nfe2 Ϫ/Ϫ embryos (see supplemental Fig. S2). Reduction of JunD expression normalizes placental vascularization within the labyrinthine layer of day 14.5 p.c. placental tissues (93% in Nfe2 Ϫ/Ϫ JunD kd versus 75% in Nfe2 Ϫ/Ϫ embryos, p Ͻ 0.001, Fig.  3, A and B). Restoration of placental vascularization was associated with a normalization of growth in Nfe2 Ϫ/Ϫ JunD kd embryos (weight 97% and length 95% of that of control and Fig.  3C). Furthermore, electron microscopic analyses revealed a normalization of the total width of the feto-maternal placental barrier (108% in Nfe2 Ϫ/Ϫ JunD kd versus 122% in Nfe2 Ϫ/Ϫ embryos, p Ͻ 0.001, Fig. 3, D and E) and syncytiotrophoblast layer II (110 versus 138%, p Ͻ 0.001, Fig. 3, D and E). Using quantitative RT-PCR we observed a normalization of Gcm1 (112 versus 192%, p Ͻ 0.001), Cepba (107 versus 142%, p Ͻ 0.001), and Synb (110 versus 172%, p Ͻ 0.001, Fig. 3F) expression in Nfe2 Ϫ/Ϫ placenta following JunD knockdown in cells of the trophectodermal lineage. In agreement with our previous observations we did not observe any morphological change in syncytiotrophoblast layers I and III, in endothelial cells, or in the expression of Syna or Cstq (Fig. 3F and data not shown) (1).
We next explored whether gain of JunD function is sufficient for the changes observed in Nfe2 Ϫ/Ϫ placenta. To this end we transduced wild type blastocysts with JunD expressing lentivi-  FEBRUARY 17, 2012 • VOLUME 287 • NUMBER 8
Nfe2 Regulates JunD Binding to Gcm1 Promoter Particularly at Ϫ1441-To define the promoter region required for JunDdependent Gcm1 expression we conducted further in vitro experiments. Using computational analyses we identified several potential AP-1/JunD-binding sites with the Gcm1 promoter (Fig. 4A). ChIP analyses revealed increased JunD binding in 3 days differentiated Nfe2 kd trophoblast cells to all in silico identified binding regions except the most proximal one (Ϫ1100 to Ϫ900, Fig. 4B).
To analyze the functional relevance of these promoter regions in the context of JunD binding and Gcm1 expression we next performed luciferase reporter assays in HEK293 cells. HEK293 cells were transfected with reporter constructs consisting of Gcm1 promoter fragments of various lengths cloned to the luciferase reporter gene (9). These cells were co-transfected either with empty control vector (control), a JunD D and E, the characteristic enhanced width of syncytiotrophoblast layer II (SynT-II, marked with a white star) in the absence of Nfe2 (Nfe2 Ϫ/Ϫ cshRNA), which is associated with an increased width of the total embryonic-maternal barrier (total), is reversed following reduction of JunD expression in Nfe2 Ϫ/Ϫ embryos (Nfe2 Ϫ/Ϫ JunD kd , gray bars). Isolated gain of JunD function is sufficient to increase the total width of the embryonic-maternal barrier and the width of syncytiotrophoblast layer II (Nfe2 c JunD ϩ , striped bars). For control Nfe2 Ϫ/Ϫ embryos were infected with lentiviral particles expressing control shRNA (Nfe2 Ϫ/Ϫ cshRNA). EM images of the maternal embryonic barrier (D) and bar graph summarizing results of EM analyses (E, 6 embryos per group, for each embryo at least 50 images were analyzed) are shown. F, reduction of JunD normalizes expression of syncytiotrophoblast markers in Nfe2 Ϫ/Ϫ placenta (gray bars), whereas isolated overexpression of JunD is sufficient to increase expression of these markers (striped bars). Bar graph summarizing results of qRT-PCR analyses of Gcm1, Cebpa, Synb, Syna, and Cstq expression in placenta tissue samples (n ϭ 6 embryos each) is shown. Mean value Ϯ S.E., size bars are 50 m (A) and 1 m (D); *, p Ͻ 0.05, **, p Ͻ 0.01 (compared with Nfe2 ϩ/Ϫ cshRNA, analysis of variance). expression construct (JunD), or both JunD and Nfe2 expression constructs (JunD ϩ Nfe2). The Ϫ4041, Ϫ2982, and Ϫ1637 reporter constructs, but not the Ϫ396 reporter construct, contained inducible JunD-binding site(s). Hence, the functional relevant JunD-binding site must be within the Ϫ1637 to Ϫ396 promoter fragment. Of note, co-transfection with Nfe2 suppressed JunD-binding activity, directly demonstrating the repressive effect of Nfe2 in regard to JunD driven Gcm1 promoter activation.
Considering that the Ϫ1637 to Ϫ396 region contained inducible JunD-binding site(s) and that, based on the above ChIP analyses, JunD binds to the Ϫ1630 to Ϫ1460 and the Ϫ1450 to Ϫ1300 promoter regions, but not to the Ϫ1100 to Ϫ900 promoter fragment (see above, Fig. 4B), we hypothesized that the functional relevant JunD-binding site(s) must reside within the Ϫ1637 to Ϫ1300 region of the Gcm1 promoter. This region contains three potential JunD-binding sites (at Ϫ1576, Ϫ1441 and Ϫ1314). To determine their functional relevance for Gcm1 expression these sites were inactivated by targeted point mutations. JunD-mediated induction of the reporter construct containing the mutated JunD-binding site at Ϫ1441 was reduced to 45% (p ϭ 0.02, Fig. 4D), whereas induction of the reporter construct containing the mutated JunD-binding site at Ϫ1576 or Ϫ1314 was not significantly reduced (94 and 88% compared with the control construct, p ϭ 0.48 and 0.26, respectively, Fig. 4D).
To establish whether JunD binds to the endogenous Gcm1 promoter sequence we next performed EMSA using an oligonucleotide derived from the Gcm1 sequence (Ϫ1426 to Ϫ1448). Binding of nuclear extracts to this oligonucleotide was enhanced in the absence of Nfe2 (Nfe2 kd , Fig. 4E). Using a mutated oligonucleotide (CT 3 AG) no binding activity could be detected (Fig. 4E). Supershift analyses confirmed binding of JunD to the Gcm1-derived oligonucleotide (Fig. 4E). Taken  ). B, binding of JunD to the mGcm1 promoter is increased in the absence of Nfe2. Potential JunD-binding sites identified in silico were analyzed by ChIP assay. In the absence of Nfe2 enhanced binding of JunD can be detected in all regions except Ϫ1100 to Ϫ900. Representative images of semiquantitative RT-PCR using Nfe2 expressing control (c) or Nfe2-deficient (kd) trophoblast cells; input DNA serves as positive control. C, the Ϫ1637 to Ϫ396 Gcm1 promoter region contains a functional relevant JunD-binding site, which is negatively controlled by Nfe2. Analyses of promoter activity using mGcm1 promoter deletion constructs fused to a luciferase reporter gene and transfected into HEK293 cells. Cells were co-transfected with empty vector (control), JunD alone (JunD), or JunD and Nfe2 (JunD ϩ Nfe2) expression constructs. D, the functional JunD-binding site within the mGcm1 promoter is at position Ϫ1441. The three potential AP-1/JunD-binding sites within the Ϫ1637 to Ϫ1430 promoter fragment were inactivated by targeted mutagenesis. Only inactivation of the Ϫ1441 JunD-binding site results in impaired promoter activity. HEK293 cells were transfected with the Gcm1 Ϫ1637 promoter/luciferase reporter constructs (nonmutated control; Ϫ1576 AC/TG mutant; Ϫ1441 TC/AG mutant; and Ϫ1314 AC/TG mutant) and a JunD expression construct. E, JunD-binding activity determined using a probe corresponding to the mGcm1 promoter region Ϫ1426 to Ϫ1448 with the JunD-binding site at Ϫ1441 intact or mutated (TC 3 AG). Binding to the intact but not to the mutated Gcm1 probe is enhanced in together, JunD binds to the Gcm1 promoter at Ϫ1441, which is required for JunD-mediated Gcm1 promoter activation.
Regulation of JunD-binding Activity by Nfe2 Is Acetylation Dependent-We have previously shown that the regulation of Gcm1 by Nfe2 is acetylation dependent (1). Given the current finding that JunD is crucially involved in the regulation of Gcm1 we next evaluated the role of acetylation for the Nfe2-dependent regulation of JunD.

DISCUSSION
Previously we identified a novel function of the bZip transcription factor Nfe2 during trophoblast differentiation. We demonstrated that Nfe2 represses Gcm1 expression, a key regulator of syncytiotrophoblast differentiation and placental vascularization, in trophoblast cells through an unknown modus operandi. In the present study we establish that the AP-1 protein JunD is a positive gene regulator during syncytiotrophoblast differentiation. Nfe2 prevents excess syncytiotrophoblast formation, which would impair placental vascularization and embryonic growth (1) by limiting JunD DNA-binding activity. The Nfe2-dependent repression of JunD DNA-binding activity is mediated via acetylation.
The acetylation dependent cross-coupling of Nfe2 with JunD during trophoblast differentiation provides a novel mechanism controlling Gcm1 expression, syncytiotrophoblast formation, placental vascularization, and embryonic growth. A function of other members of the AP-1 protein family, namely JunB and Fra1, for placental development has been previously described (10,11). The function of JunD must be different from JunB or Fra1. Thus, loss of JunB or Fra1 function results in embryonically lethal defects, whereas in mice with a gain of JunB or Fra1 function no embryonic phenotype has been reported (28). Contrary, loss of JunD function has no impact on embryonic development (29), whereas gain of function impairs placental vascularization and embryonic growth (this study).
In contrast to the nonlethal consequences of increased JunDbinding activity in Nfe2 Ϫ/Ϫ embryos after day 14.5 p.c. the absence of JunB or Fra1 results in a lethal placental defect associated with defective yolk sac vascularization during establishment of the hemochorial placenta (day 8.5 p.c. in JunB Ϫ/Ϫ and 10.5 p.c. in Fra1 Ϫ/Ϫ embryos) (10,11). Thus, whereas loss of function of JunB and Fra1 results in early fetal loss, bZip transcription factors Nfe2 and JunD coordinately regulate placental development and embryonic growth at later developmental stages. Such diverse functions of distinct Jun family proteins likely reflect their ability to form various heterodimers with other bZiP family proteins resulting in either positive or negative regulation of target genes (28,30). The newly identified interaction of JunD and Nfe2 expands our knowledge about the role of bZip transcription factors during placental development and demonstrates that this expanding network conveys crucial roles for placentation at various developmental stages.
Transcriptional activation of AP-1 proteins is regulated by MAPKs, in particular by the c-Jun NH 2 -terminal kinases (JNKs) and the extracellular signal-regulated kinases (ERKs), which phosphorylate AP-1 proteins (31,32). The molecular mechanism of transcriptional activation of AP-1 proteins by MAPKs remained unknown until recently. Aguilera and colleagues (27) demonstrated that transcriptional activation of the AP-1 protein c-Jun is linked to increased histone acetylation within AP-1-dependent promoters. Nonphosphorylated c-Jun recruits the Mbd3/nucleosome remodeling and histone deacetylation (NuRD) repressor complex through an interaction with its transactivation domain, resulting in gene repression of AP-1 target genes. This repression is relieved by JNKmediated c-Jun NH 2 -terminal phosphorylation. In the current study we demonstrate that DNA-binding and transcriptional activity of JunD is likewise modulated by acetylation, indicating that the mechanism identified by Aguilera and colleagues (27) is not specific for c-Jun. We previously demonstrated that acetylation of histone H4 within the Gcm1 promoter is modulated by Nfe2 in trophoblast cells and placenta (1), which is entirely consistent with the observation made by Aguilera et al. (27). In addition, we demonstrate now that the AP-1 protein JunD itself is modulated by acetylation. Whether other AP-1 proteins are directly modulated by acetylation and the relative importance of either histone or AP-1 protein acetylation for transcriptional activation needs to be explored in future studies.
Whether Nfe2-dependent acetylation of JunD depends on MAPK activity remains currently unknown. Unlike c-Jun, JunD lacks a functional JNK docking site within the NH 2 -terminal region (28). However, following heterodimerization with c-Jun, the serine residues of JunD can be phosphorylated (33), indicating that MAPK-mediated JunD phosphorylation may be involved in the regulation of acetylation. However, whereas regulation of Nfe2 by JNK has been demonstrated (34), the reverse, e.g. regulation of MAPK activity by Nfe2, has to the best of our knowledge not been reported. Hence the mechanism underlying Nfe2-dependent regulation of acetylation remains presently unknown. Based on previous and the current findings we propose a MAPK independent mechanism through which Nfe2 may inhibit acetylation. Interaction of Nfe2 with CBP in hematopoietic and trophoblast cells is established (1,7,35). CBP/p300 are coactivators generally influencing gene expression by functioning as scaffolding proteins facilitating proteinprotein interactions and/or through their intrinsic acetyltransferase activity toward histone proteins and other nearby nuclear factors (36). Considering that Nfe2 acts as a repressor during trophoblast differentiation we speculate that Nfe2 may inhibit the intrinsic acetylation activity of CBP in trophoblast cells, thus inducing hypoacetylation of JunD and histones within the Gcm1 promoter (1) (and current results). The Nfe2dependent repression of acetylation may result from a competition of bZip transcription factors for a limited CBP pool (37). Indeed, a competitive interaction between CBP and bZip proteins, including Nfe2 and c-Jun, has been previously reported (38). In addition, we observed a markedly increased interaction between CBP and Gcm1 in the absence of Nfe2 (1), which is in agreement with a competitive interaction of CBP with various transcription factors during trophoblast differentiation. Such competitive interaction with CBP may be the mechanism underlying the functional antagonism between bZip transcription factors like Nfe2 and JunD during trophoblast differentiation and may constitute an alternative, MAPK independent pathway through which AP-1 proteins regulate acetylation and gene expression. The existence of such a competitive mechanism, as proposed, needs to be explored in future studies.
Given the current and previous findings Nfe2 acts as an epigenetic and post-translational repressor during placental development, providing a mechanism limiting syncytiotrophoblast formation. It appears that all mechanisms through which Nfe2 modulates placental vascularization converge at the regulatory level of Gcm1. Thus, the current finding underscores a crucial function of Gcm1 in regulating placental labyrinthine formation (6). This novel, Nfe2-dependent pathway controlling Gcm1 activity is important to ensure sufficient placental vascularization and embryonic growth during later pregnancy stages (1) (and current findings).
Future studies are required to determine whether the same pathway is of functional relevance in the human placenta. We observed expression of Nfe2 in human placenta, specifically in human syncytiotrophoblast cells 3 and expression of Nfe2 within human syncytiotrophoblast is depicted on the human protein atlas webpage. Interestingly, JunD, but not other AP-1 proteins, is specifically expressed in human syncytiotrophoblast (39). Hence it appears possible, but remains to be shown, that cross-coupling of Nfe2 and JunD via acetylation constitutes a novel mechanism regulating placental development in humans as well. Providing that the relevance of this pathway can be confirmed in humans this may identify new diagnostic and therapeutic targets for placental diseases associated with altered expression of Gcm1, such as pre-eclampsia.