New insights into the regulation of placental growth factor gene expression by the transcription factors GCM1 and DLX3 in human placenta

Expression of placental growth factor (PGF) is closely associated with placental perfusion in early pregnancy. PGF is primarily expressed in placental trophoblasts, and its expression decreases in preeclampsia, associated with placental hypoxia. The transcription factors glial cells missing 1 (GCM1) and metal-regulatory transcription factor 1 (MTF1) have been implicated in the regulation of PGF gene expression through regulatory elements upstream and downstream of the PGF transcription start site, respectively. Here, we clarified the mechanism underlying placenta-specific PGF expression. We demonstrate that GCM1 up-regulates PGF expression through three downstream GCM1-binding sites (GBSs) but not a previously reported upstream GBS. Interestingly, we also found that these downstream GBSs also harbor metal-response elements for MTF1. Surprisingly, however, we observed that MTF1 is unlikely to regulate PGF expression in the placenta because knockdown or overexpression of GCM1, but not MTF1, dramatically decreased PGF expression or reversed the suppression of PGF expression under hypoxia, respectively. We also demonstrate that another transcription factor, Distal-less homeobox 3 (DLX3), interacts with the DNA-binding domain and the first transactivation domain of GCM1 and that this interaction inhibits GCM1-mediated PGF expression. Moreover, the GCM1–DLX3 interaction interfered with CREB-binding protein–mediated GCM1 acetylation and activation. In summary, we have identified several GBSs in the PGF promoter that are highly responsive to GCM1, have demonstrated that MTF1 does not significantly regulate PGF expression in placental cells, and provide evidence that DLX3 inhibits GCM1-mediated PGF expression. Our findings revise the mechanism for GCM1- and DLX3-mediated regulation of PGF gene expression.

Vasculogenesis and angiogenesis are essential for placental development. These events are modulated by vascular endothe-lial growth factor (VEGF), 2 placental growth factor (PGF), and their receptors kinase insert domain receptor (KDR) and Flt-1 (1). The human placenta is composed of villous tissues where fetal arterial and venous vessels are developed to convey nutrients and gases from the mother to the fetus (2,3). Indeed, VEGF and PGF are preferentially responsible for branching and nonbranching angiogenesis in terminal villi, which results in a complex and branched capillary web and elongated capillary loops, respectively (3). A balance of branching and nonbranching angiogenesis is maintained for proper geometry of the villous vascular bed for healthy pregnancy outcomes.
PGF, which is a member of the VEGF family, is primarily expressed in placental trophoblasts, and its serum level increases throughout pregnancy and peaks between gestational weeks 28 and 32 (4,5). In early pregnancy, a correlation of placental perfusion and PGF expression has been observed by measuring the systolic blood flow velocity at proximal and distal ends of the umbilical cord artery and maternal serum PGF levels (6). Clinically, a decreased circulating level of PGF has been detected as early as the first trimester and often leads to development of preeclampsia (PE) of which symptoms include hypertension and proteinuria (7). As PE is associated with poor uteroplacental circulation and placental hypoxia, PGF expression is also decreased under hypoxia (8,9). Importantly, administration of recombinant human PGF has recently been shown to relieve the clinical symptoms of PE in rat and baboon disease models induced by uteroplacental ischemia (10,11).
Previous studies have reported that glial cells missing 1 (GCM1), metal-regulatory transcription factor 1 (MTF1), and DLX3 are involved in the regulation of PGF gene expression in human placenta (12)(13)(14). GCM1 is a placental transcriptional factor responsible for trophoblast fusion, migration, and invasion and hormone production by transactivation of syncytin, HtrA4, and hCG␤ genes (15)(16)(17)(18). As cAMP signaling is crucial for trophoblast differentiation, GCM1 activity is up-regulated by the cAMP/PKA/CBP and cAMP/Epac1/Ca 2ϩ /calmodulindependent protein kinase I signaling cascades, respectively (19,20). A GCM1-response element (GRE) harbors a consensus GCM1-binding sequence of 5Ј-ATGCGGG(T/C)-3Ј (21). MTF1 was identified in the regulation of metallothionein gene expression in the presence of heavy metals such as copper, zinc, and cadmium (22). MTF1 binds to DNA sequence motifs known as metal-response elements (MREs) with a core consensus of 5Ј-TGC(A/G)C(G/A/T/C)C-3Ј (23). Chang et al. (12) have demonstrated that GCM1 up-regulates PGF promoter activity through a GRE at position Ϫ818 to Ϫ810 relative to the transcription start site (TSS). Conversely, Depoix et al. (24) performed a DNA microarray analysis to demonstrate that PGF is induced in BeWo cells treated with the cAMP stimulant forskolin (FSK). The same group further demonstrated that mutation of the GRE identified by Chang et al. (12) fails to inhibit FSK-induced PGF expression and questioned the role of GCM1 in regulation of PGF gene expression. Interestingly, Nishimoto et al. (13) reported that MTF1 up-regulates PGF gene expression through triple repeats of MRE downstream of TSS. Nevertheless, it is hard to reconcile the placenta-specific gene expression pattern of PGF with MTF1, which is a ubiquitous transcription factor. The Distal-less homeobox 3 (DLX3) transcription factor harbors a conserved homeodomain with DNAbinding activity and regulates trophoblast differentiation (25,26). Targeted gene disruption of Dlx3 in mice leads to embryonic lethality and placental vascularization defects in the labyrinth (27). DLX3 has recently been reported in the regulation of trophoblast differentiation and modulates PGF expression by inhibition of the transactivation activity of GCM1 in a homeodomain-dependent fashion (14,28).
In search of GCM1 functions in placenta, we have recently conducted chromatin immunoprecipitation (ChIP)-chip analysis to identify PGF as a candidate target gene of GCM1 (16). In the present study, we investigated the functional role of GCM1 in regulation of PGF gene expression and provide a revised mechanism underlying the placenta-specific expression of PGF gene. We identify three novel tandem repeats of GRE downstream of the TSS of PGF gene, which also constitute the above described triple repeats of MRE. Importantly, the identified GREs/MREs are recognized by GCM1, but not MTF1, in vivo and in vitro by ChIP analysis and EMSA, respectively. We further demonstrate that the transcript level of GCM1, but not MTF1, correlates with that of PGF in response to hypoxia and FSK. GCM1 overexpression significantly reverses the suppression of PGF gene expression by hypoxia, which was not observed by MTF1 overexpression. We also show that DNA-binding domain (DBD) and the first transactivation domain (TAD) of GCM1 are involved in interaction with DLX3, leading to suppression of GCM1-mediated PGF expression. DLX3 very likely blocks the interaction between GCM1 and CBP to suppress CBP-mediated GCM1 acetylation and activation. Collectively, our results provide new insights into regulation of PGF gene expression by GCM1 and DLX3 in the human placenta.

Identification of PGF as a GCM1 target gene
We recently performed ChIP-chip experiments in BeWo31 cells, which stably express GCM1 with an N-terminal HA tag (HA-GCM1) and identified PGF as a candidate GCM1 target gene (16). As shown in Fig. 1A, strong HA-GCM1 binding signals were detected and mapped to a genomic region downstream of TSS of PGF gene. Interestingly, MTF1 has been suggested to regulate PGF promoter activity through a genomic region harboring triple repeats of MRE positioned between nucleotides Ϫ511 and Ϫ468 relative to the translation initiation site (i.e. nucleotides ϩ12 to ϩ55 relative to TSS) (13), which also overlaps with the aforementioned HA-GCM1-binding region in ChIP-chip experiments. Because PGF is primarily expressed in placenta and MTF1 is a ubiquitous transcription factor, we wished to re-examine whether MTF1 or GCM1 plays a key role in regulation of PGF gene expression in placenta. We compared the effects of GCM1 and MTF1 on the pGL4-PGF(Ϫ1000/ϩ100) luciferase reporter construct (note that nucleotides were numbered relative to TSS) in 293T cells in transient expression experiments. The luciferase activity directed by pGL4-PGF(Ϫ1000/ϩ100) was significantly stimulated by GCM1-FLAG in a dose-dependent manner (Fig. 1B,  upper). In contrast, transcriptional activation of the pGL4-PGF(Ϫ1000/ϩ100) construct by MTF1-FLAG was marginally detected after a closer examination (Fig. 1B, lower). To rule out the possibility that the C-terminal FLAG tag may impair MTF1-FLAG activity, we compared the MTF1-FLAG expression plasmid with another expression plasmid encoding MTF1 with an N-terminal HA tag (HA-MTF1) or a C-terminal HA tag (MTF1-HA) in transient expression experiments using an established MTF1 reporter construct, p4MREd-E1bLuc. Indeed, significant stimulation of luciferase activity directed by p4MREd-E1bLuc was detected in the presence of MTF1-FLAG, HA-MTF1, or MTF1-HA ( Fig.  1C), supporting that all three MTF1 expression constructs are functional in terms of transcriptional activity. Taken together, these results suggested that GCM1, but not MTF1, very likely plays an important role in regulation of PGF promoter activity.

Identification and characterization of GCM1 response elements in PGF gene
We characterized the genomic region Ϫ1000 to ϩ100 (relative to TSS) of PGF gene for potential GREs using a series of luciferase constructs harboring different parts of this genomic region ( Fig. 2A). 293T cells were transfected with pHA-GCM1 and the aforementioned PGF reporter plasmids in transient expression experiments. The luciferase activity directed by pGL4-PGF(Ϫ270/ϩ100) was also enhanced by HA-GCM1, although the effect was lower compared with pGL4-PGF(Ϫ1000/ϩ100) (Fig. 2B). Interestingly, deletion of nucleotides ϩ6 to ϩ100 dramatically diminished the enhancement effect of HA-GCM1 on the luciferase activity directed by pGL4-PGF(Ϫ1000/ϩ5) and pGL4-PGF(Ϫ270/ϩ5), suggesting that the deleted genomic region constitutes a major regulatory region responsive to GCM1. Chang et al. (12) have reported an GCM1, but not MTF1, regulates PGF expression in placenta upstream GRE (hereafter named as GBS-BOR) in the PGF promoter, i.e. nucleotides Ϫ818 to Ϫ810 relative to the translation initiation site or Ϫ295 to Ϫ288 relative to the TSS. Because the luciferase activity directed by pGL4-PGF(Ϫ345/ϩ100) or pGL4-PGF(Ϫ270/ϩ100) was not significantly different, it is very likely that GBS-BOR is not a functional GRE (Fig. 2B). Again, marginal effects of MTF-FLAG were observed on the tested PGF reporter constructs (Fig. S1). We then concentrated on the genomic region of Ϫ270 to ϩ100 of PGF gene for identification of functional GREs. There are five potential GCM1binding sites (i.e. GBS1-5) in the genomic region of Ϫ270 to ϩ100 (Fig. 2, A and C). The GBSs were changed by site-directed mutagenesis individually (M1-5) or in combination (M345) in pGL4-PGF(Ϫ270/ϩ100), and the resultant reporter constructs were cotransfected with pHA-GCM1 into 293T cells. As shown in Fig. 2D, mutagenesis of GBS3, -4, or -5 (the M3, -4, or -5 construct) alone significantly reduced the stimulatory effect of HA-GCM1 on the PGF promoter region of Ϫ270 to ϩ100. Moreover, this effect was further decreased on the M345 construct with combined mutagenesis of GBS3, -4, and -5 (Fig. 2D). These results suggested that GBS3, -4, and -5 are functional GREs required for up-regulation of PGF gene expression by GCM1.

Association of GCM1, but not MTF1, with PGF promoter
The identified GREs (GBS3-5) in PGF gene also harbor the previously reported triple repeats of MRE (here defined as MRE6 -8) (13). Careful inspection of the PGF promoter region from nucleotides Ϫ1000 to ϩ100 revealed additional potential MREs (MRE1-5; Fig. 3A). We then studied whether GCM1 or MTF1 is associated with PGF gene through GBS3-5/MRE6 -8 and/or other MREs in placental BeWo cells by ChIP analysis. As shown in the left panel of Fig. 3B, GCM1, but not MTF1, was able to recognize the PGF promoter region harboring GBS3-5, which was enhanced in the presence of FSK. Association of GCM1 with GBS3-5 in the PGF promoter was also detected in primary human trophoblasts (Fig. 3B, right). We further performed ChIP analysis in 293T cells transfected with pGL4-PGF(Ϫ270/ϩ100) and the expression plasmids encoding MTF1-FLAG and GCM1-FLAG, respectively. Specific interaction between GBS3-5 in pGL4-PGF(Ϫ270/ϩ100) and GCM1-FLAG, but not MTF1-FLAG, was detected using a pGL4 vector-specific primer and a primer upstream of GBS3-5 ( Fig.  3C). To study direct interaction between GCM1 and GBS3-5, we performed EMSA using a radiolabeled GBS3 oligonucleotide probe and recombinant GCM1-FLAG or MTF1-FLAG

GCM1, but not MTF1, regulates PGF expression in placenta
protein. Indeed, GCM1-FLAG specifically bound to the radiolabeled GBS3 probe, which was competed out by unlabeled GBS3, but not mutant GBS3, at a 100ϫ molar ratio to the probe (Fig. 3D, left panel, lanes 2-4). Of note, MTF1-FLAG failed to bind to the GBS3 probe (Fig. 3D, left panel, lane 7), although it was able to bind to a radiolabeled MRE oligonucleotide probe of the consensus MTF1-binding sequence (Fig. 3D, right panel). Together with the observation that unlabeled GBS4 and GBS5 were capable of competing with the GBS3 probe for GCM1-FLAG (Fig. 3D, left panel, lanes 5and 6), these results strongly suggested that GCM1 is the bona fide transcription factor that recognizes GREs (GBS3-5) in the PGF promoter.

Regulation of PGF gene expression by GCM1
Trophoblast differentiation and GCM1 activity are stimulated by cAMP (20, 29 -31). We tested the effect of FSK on the transcript and protein levels of PGF, GCM1, and MTF1 in BeWo cells. As shown in Fig. 4A, FSK up-regulated the transcript and protein levels of PGF and GCM1 but not MTF1. As a complementary approach, knocking down GCM1 in BeWo cells stably expressing GCM1 short-hairpin RNA (shRNA) significantly decreased the transcript level of PGF, whereas MTF1 knockdown failed to significantly affect PGF gene expression (Fig. 4B, left). Moreover, GCM1 knockdown also decreased the PGF transcript level in primary human trophoblasts (Fig. 4B,  right). Because placental hypoxia and decreased PGF expression are associated with PE, we measured the GCM1, MTF1, and PGF transcript levels in BeWo cells under normoxic and hypoxic conditions. Compared with the MTF1 transcript level, the GCM1 and PGF transcript levels in BeWo cells were dramatically decreased by hypoxia (Fig. 4C, left panel). Correspondingly, the GCM1 protein level in BeWo cells was greatly reduced by hypoxia, whereas the MTF1 protein level was only modestly reduced (Fig. 4C, right panel). As hypoxia suppresses PGF expression, we introduced GCM1-HA or HA-MTF1 into hypoxic BeWo cells by a lentiviral expression system to test whether GCM1-HA or HA-MTF1 could reverse the PGF transcript level. As shown in Fig. 4D, GCM1-HA, but not HA-MTF1, significantly elevated the PGF transcript level in the BeWo cells under hypoxia. Taken together, these results suggested that GCM1 is a key factor controlling PGF gene expression in placenta in response to environmental cues such as FSK and hypoxia.

Functional interaction between GCM1 and DLX3
Li and Roberson (14) have recently reported that DLX3 and GCM1 modulate PGF promoter activity through the genomic region Ϫ369 to Ϫ320 (relative to TATA box, i.e. Ϫ622 to Ϫ592 relative to TSS. To test whether DLX3 affects GCM1-up-regulated PGF promoter activity, we transfected 293T cells with different combinations of pGL4-PGF(Ϫ1000/ϩ100), pGL4-PGF(Ϫ270/ϩ100), pHA-GCM1, and pDLX3-FLAG. DLX3-FLAG alone did not affect luciferase activity directed by pGL4-PGF(Ϫ1000/ϩ100) or pGL4-PGF(Ϫ270/ϩ100) (Fig. 5A). Instead, the stimulatory effect of HA-GCM1 on pGL4-PGF(Ϫ1000/ϩ100) or pGL4-PGF(Ϫ270/ϩ100) was significantly decreased when a higher amount of pDLX3-FLAG was cotransfected (Fig. 5A). A similar inhibitory effect of DLX3 on    Ab (B, left). ChIP analyses of GCM1-GBS interaction were also performed in primary human trophoblasts using GCM1 Ab (B, right). The numbers underneath indicate the band intensities normalized against input. In a separate experiment, 293T cells were transfected with pGL4-PGF(Ϫ270/ϩ100), pGCM1-FLAG, and pMTF1-FLAG for 48 h. Cells were harvested for ChIP analysis using FLAG mAb. The sequence of nucleotide Ϫ270 to ϩ100 in pGL4-PGF(Ϫ270/ϩ100) is shaded in gray, primers for PCR are underlined, and GBS3-5 are boxed (C). The protein levels of GCM1-FLAG and MTF1-FLAG in the transfected 293T cells were analyzed by immunoblotting (IB). D, EMSA of GCM1 binding to the GBS3-5 (GREs) of PGF promoter. Recombinant GCM1-FLAG or MTF1-FLAG protein was incubated with radiolabeled PGF-GBS3 probe in the presence or absence of the indicated unlabeled oligonucleotide at a molar ratio of 100 over the probe followed by electrophoresis and autoradiography. In a separate experiment, recombinant MTF1-FLAG protein was incubated with radiolabeled MRE probe in the presence or absence of unlabeled MRE oligonucleotide in EMSA. The asterisk indicates the position of the protein-DNA complex. GCM1-mediated transcription was observed in 293T cells transfected with the GCM1 reporter construct p(GBS) 4 -E1bLuc, pHA-GCM1, and a higher amount of pDLX3-FLAG (Fig. 5B). These results suggest that DLX3 may suppress PGF expression through GCM1.

GCM1, but not MTF1, regulates PGF expression in placenta
Consequently, we examined the interaction between GCM1 and DLX3 by coimmunoprecipitation analysis of 293T cells transfected with pDLX3-FLAG and pHA-GCM1. Specific interaction between HA-GCM1 and DLX3-FLAG was detected and was further enhanced in the presence of the proteasome inhibitor MG132, which is known to stabilize GCM1 (Fig. 6A). To characterize endogenous GCM1-DLX3 interaction, we first performed immunofluorescence microscopy to detect colocalization of GCM1 and DLX3 in the nuclei of BeWo cells (Fig.  6B). We then performed coimmunoprecipitation analyses to demonstrate the interaction between endogenous DLX3 and GCM1 in BeWo cells (Fig. 6C). To map the interaction domain in GCM1 for DLX3, 293T cells were transfected with pHA-DLX3 and expression plasmids encoding Gal4-GCM1 fusion proteins harboring full-length or truncated GCM1 for coimmunoprecipitation analysis. As shown in Fig. 6C, HA-DLX3 interacts with the GCM1 polypeptide region, amino acids 1-167 or 167-349, corresponding to the N-terminal DBD or the first TAD of GCM1, respectively.

DLX3 suppresses GCM1 activity by blockade of CBP-mediated GCM1 acetylation
To explore the mechanism underlying the inhibitory effect of DLX3 on GCM1 activity, we studied whether DLX3 inhibits the DNA-binding or transactivation activity of GCM1 or both using a mammalian two-hybrid system. 293T cells were transfected with different combinations of p(GBS) 4 -E1bLuc, pVP16, pVP16-GCM1(1-167), and pDLX3-FLAG to measure the effect of DLX3-FLAG on the DNA-binding activity of GCM1. As shown in Fig. 7A, stimulation of the luciferase activity directed by p(GBS) 4 -E1bLuc was detected in the presence of VP16-GCM1(1-167), which is a fusion protein containing the TAD of VP16 and the DBD of GCM1. The transcriptional activity of VP16-GCM1(1-167) was only modestly inhibited by a higher amount of pDLX3-FLAG (Fig. 7A). To analyze the effect of DLX3 on the transcriptional activity of GCM1, 293T cells were transfected with different combinations of pG5-Luc, pVP16, pGal4-GCM1(167-436), and pDLX3-FLAG. Indeed, the luciferase activity directed by pG5-Luc was up-regulated by the Gal4-GCM1(167-436) fusion protein containing the DBD of GAL4 and the TAD of GCM1 (Fig. 7B). Interestingly, the transcriptional activity of Gal4-GCM1(167-436) was significantly decreased by DLX3-FLAG in a dose-dependent manner (Fig. 7B).

GCM1, but not MTF1, regulates PGF expression in placenta
The biological function of DLX3 in regulation of PGF gene expression was studied by RNAi in BeWo cells treated with or without FSK. The PGF transcript level was not affected by DLX3 knockdown in the absence of FSK (Fig. 7C). As expected, the PGF transcript level was stimulated by FSK in the scrambled control cells. Interestingly, this FSK-stimulated PGF transcript expression was further elevated when DLX3 was knocked down (Fig. 7C). These observations suggest that DLX3 may interfere with the cAMP signaling pathway up-regulating GCM1-medi-ated PGF gene expression. We tested whether DLX3 jeopardizes the interaction between GCM1 and CBP, a transcriptional coactivator known to stimulate GCM1 acetylation and activity under cAMP (20). To this end, 293T cells were transfected with pGCM1-Myc, pCBP-FLAG, and pHA-DLX3 for coimmunoprecipitation analysis. As shown in Fig. 7D, acetylation of GCM1-Myc was detected in the presence of CBP-FLAG, which was decreased by increased interaction between GCM1-Myc and HA-DLX3. Correspondingly, the interaction between Figure 6. Characterization of the interaction between GCM1 and DLX3. A, GCM1 specifically interacts with DLX3. 293T cells were transfected with the indicated combinations of pDLX3-FLAG, pHA-GCM1, and pGATA3-FLAG for 48 h and then harvested for coimmunoprecipitation analysis using FLAG and HA mAbs. Note that the interaction between DLX3-FLAG and HA-GCM1 was enhanced after 6 h of treatment with 10 M proteasome inhibitor MG132. Specific interaction between GATA3-FLAG and HA-GCM1 was detected and used as a positive control. B, panels a-f, colocalization of GCM1 and DLX3. BeWo cells were stained with mouse anti-DLX3 and rabbit anti-GCM1 Abs. Cells were then incubated with secondary Abs and examined under a confocal microscope. Arrows indicate GCM1 and DLX3 nuclear colocalization. Bar in d, 20 meter; bar in f, 100 meter. C, endogenous interaction of GCM1 and DLX3. BeWo cells were harvested for immunoprecipitation (IP) with rabbit IgG or rabbit anti-DLX3 Ab followed by immunoblotting (IB) with rabbit anti-GCM1 Ab. D, mapping of the DLX3-interacting domains in GCM1. 293T cells were transfected with pHA-DLX3 and pGal4-FLAG or pGal4-GCM1-FLAG encoding for full-length or deleted GCM1 polypeptide. At 48 h post-transfection, cells were harvested for coimmunoprecipitation analysis using FLAG and HA mAbs.

GCM1, but not MTF1, regulates PGF expression in placenta
GCM1-Myc and CBP-FLAG was decreased (Fig. 7D). Therefore, DLX3 may suppress PGF gene expression by blocking the interaction between GCM1 and CBP and thereby inhibiting GCM1 acetylation and activation. The inhibitory effect of DLX3 on GCM1 activity was also effective on the expression of another GCM1 target gene, hCG␤, as DLX3 knockdown enhanced the stimulatory effect of FSK on hCG␤ gene expression (Fig. 7C).

Discussion
It is generally believed that PGF is a critical factor regulating vasculogenesis under physiological and pathological conditions (32,33). PGF expression is elevated during pregnancy and is crucial for normal pregnancy outcomes as decreased PGF expression is associated with preeclampsia featuring poor vasculogenesis and placental hypoxia (7). Indeed, administration of PGF was shown to reverse the clinical symptoms in animal models of preeclampsia, including hypertension, proteinuria, and endotheliosis (10,11). Although GCM1 and MTF1 have been reported to regulate PGF gene expression in placenta (12)(13)(14), there are issues that need to be resolved before the physio-logical underpinnings of PGF expression can be truly revealed. In this regard, the regulatory elements for GCM1 in stimulation of PGF gene expression have not yet been clearly defined. Moreover, MTF1 is a ubiquitous transcription factor, which is difficult to reconcile with the primary expression of PGF gene in placenta.
In the present study, we reveal a new mechanism to revise the current knowledge about PGF gene expression in placenta. We redefined the key regulatory elements (GREs) in PGF gene to be GBS3-5 that are highly responsive to GCM1 and demonstrated that MTF1 barely regulates or does not recognize GBS3-5 in placental cells even though GBS3-5 also harbor previously reported MREs. This statement is supported by the following lines of evidence. First, ChIP-chip or ChIP analysis demonstrated the association of GCM1, but not MTF1, with GBS3-5. Second, transcriptional activation of PGF promoter by GCM1 was primarily dependent on GBS3-5. Third, direct binding of GCM1 to GBS3-5 was demonstrated by EMSA. Fourth, suppression of PGF gene expression by hypoxia was significantly reversed by GCM1, but not MTF1, in placental cells.

GCM1, but not MTF1, regulates PGF expression in placenta
An upstream GRE (i.e. GBS-BOR) has been reported in PGF gene and localized to nucleotides Ϫ295 to Ϫ288 relative to TSS (12). Because pGL4-PGF(Ϫ345/ϩ100) and pGL4-PGF(Ϫ270/ ϩ100) exhibit a similar response to GCM1 (Fig. 2B), i.e. deletion of nucleotides Ϫ345 to Ϫ269 did not affect GCM1-upregulated PGF promoter activity, we think that the reported GBS-BOR is not a functional GRE. Conversely, deletion of nucleotides Ϫ1000 to Ϫ269 decreased the responsiveness of PGF promoter to GCM1 (Fig. 2B), suggesting potential upstream GREs within this region. This possibility was overshadowed by the fact that deletion of downstream GBS3-5 in pGL4-PGF(Ϫ1000/ϩ5) and pGL4-PGF(Ϫ270/ϩ5) further decreases the responsiveness of PGF promoter to a similar level (Fig. 2B). Therefore, we think that GBS3-5 constitute the essential GREs for GCM1 to regulate PGF gene expression in placenta. PGF expression has been reported in nonplacental cells, mostly tumor cells, in which PGF collaborates with VEGF family members to stimulate vessel formation for oxygen supply for tumor growth (34). Regulation of PGF gene expression in nonplacental cells has been attributed to MTF1, NF-B, or FoxD1 (35)(36)(37). Therefore, MTF1 may be involved in the regulation of PGF gene expression in pathological conditions to support tumor growth in the absence of GCM1.
Li and Roberson (14,28) have recently reported that DLX3 alone stimulates PGF gene expression in JEG-3 cells. Contradictorily, they also showed that DLX3 inhibits GCM1-up-regulated PGF gene expression. Our results in the present study did not support DLX3 positively stimulating PGF expression based upon two lines of evidence. First, DLX3 alone failed to activate the luciferase activity directed by pGL4-PGF(Ϫ1000/ ϩ100) that harbors the claimed DLX3 response element. Second, DLX3 knockdown did not affect the PGF transcript level in BeWo cells. Nevertheless, our results further indicated that DLX3 inhibits GCM1-transactivated PGF expression via interaction with the DBD and the first TAD of GCM1. We further ascribe this observation to the interference of CBP-mediated GCM1 acetylation and activation by DLX3, which was corroborated by the fact that DLX3 knockdown enhances GCM1 target gene expression (i.e. PGF and hCG␤) in response to the cAMP-stimulant FSK. Therefore, DLX3 may play a suppressor role to fine-tune GCM1-mediated trophoblast differentiation and PGF gene expression.

Plasmid constructs
HA-and FLAG-tagged GCM1 expression plasmids, pHA-GCM1 and pGCM1-FLAG, have been described previously (38). A DNA fragment encoding human MTF1 with an N-terminal HA tag or a C-terminal FLAG or HA tag was subcloned into the pcDNA3.1 expression vector (Invitrogen) to generate expression plasmids for HA-MTF1, MTF1-FLAG, or MTF1-HA. Likewise, DLX3 expression plasmids, DLX3-FLAG and HA-DLX3, were generated. The PGF promoter region from nucleotides Ϫ1000 to ϩ100 (relative to TSS) was subcloned into pGL4-basic (Promega, Madison, WI) to generate the pGL4-PGF(Ϫ1000/ϩ100) reporter plasmid. Derivatives of pGL4-PGF(Ϫ1000/ϩ100) harboring different regions in the PGF promoter from nucleotides Ϫ1000 to ϩ100 were also generated. The p4MREd-E1bLuc reporter plasmid was generated by subcloning four copies of the MREd in the mouse metallothionein-I gene into pGL3E1bLuc (20,39). The p(GBS) 4 -E1bLuc reporter plasmid for GCM1 has been described previously (20). The pVP16, pGal4-VP16, and pG5-Luc plasmids were derived from the CheckMate mammalian two-hybrid system (Promega). The pVP16 plasmid harbors the herpes simplex virus VP16 transactivation domain, and the pGal4-VP16 plasmid encodes a fusion protein containing the yeast Gal4 DNAbinding domain and the VP16 transactivation domain.

Cell culture, transfection, and lentivirus transduction
293T and BeWo cells were obtained from the American Type Culture Collection (Manassas, VA) and maintained at 37°C in minimal essential medium ␣ medium (for 293T) or F-12K medium (for BeWo) supplemented with fetal bovine serum, streptomycin, and penicillin. Preparation of primary human trophoblasts from term placentas has been described previously (16,17). For transient expression, cells were transfected with expression plasmids using Lipofectamine 2000 reagent (Invitrogen). For luciferase reporter assays, cells were harvested and analyzed with a commercial kit (Promega). Specific luciferase activities were normalized by protein concentration, which was measured using the BCA protein assay kit (Pierce). For generating stable knockdown cell lines, BeWo cells were infected by lentiviral pLKO.1-Puro shRNA expression plasmids harboring a scrambled sequence (5Ј-CCTAAGGTTAAGTCG-CCCTCG-3Ј) or target sequences for GCM1 (5Ј-CCTCAGCA-GAACTCACTAAAT-3Ј), MTF1 (5Ј-CAGAACTTACAATG-GATATTA-3Ј), and DLX3 (5Ј-CCTTCTCTACTCCTCCAG-TAA-3Ј) provided by the National RNAi Core Facility of Taiwan. The infected cells were subjected to antibiotic selection using 5 g/ml puromycin, and the puromycin-resistant clones were pooled for studies.