ECM-Dependent HIF Induction Directs Trophoblast Stem Cell Fate via LIMK1-Mediated Cytoskeletal Rearrangement

The Hypoxia-inducible Factor (HIF) family of transcriptional regulators coordinates the expression of dozens of genes in response to oxygen deprivation. Mammalian development occurs in a hypoxic environment and HIF-null mice therefore die in utero due to multiple embryonic and placental defects. Mouse embryonic stem cells do not differentiate into placental cells; therefore, trophoblast stem cells (TSCs) are used to study mouse placental development. Consistent with a requirement for HIF activity during placental development in utero, TSCs derived from HIF-null mice exhibit severe differentiation defects and fail to form trophoblast giant cells (TGCs) in vitro. Interestingly, differentiating TSCs induce HIF activity independent of oxygen tension via unclear mechanisms. Here, we show that altering the extracellular matrix (ECM) composition upon which TSCs are cultured changes their differentiation potential from TGCs to multinucleated syncytiotropholasts (SynTs) and blocks oxygen-independent HIF induction. We further find that modulation of Mitogen Activated Protein Kinase Kinase-1/2 (MAP2K1/2, MEK-1/2) signaling by ECM composition is responsible for this effect. In the absence of ECM-dependent cues, hypoxia-signaling pathways activate this MAPK cascade to drive HIF induction and redirect TSC fate along the TGC lineage. In addition, we show that integrity of the microtubule and actin cytoskeleton is critical for TGC fate determination. HIF-2α ensures TSC cytoskeletal integrity and promotes invasive TGC formation by interacting with c-MYC to induce non-canonical expression of Lim domain kinase 1–an enzyme that regulates microtubule and actin stability, as well as cell invasion. Thus, we find that HIF can integrate positional and metabolic cues from within the TSC niche to regulate placental development by modulating the cellular cytoskeleton via non-canonical gene expression.


Introduction
Mammalian development occurs in a physiologically hypoxic environment that drives the expression of dozens of genes via the Hypoxia-inducible Factor (HIF) family of transcriptional regulators [1]. A heterodimeric transcription factor composed of alpha and beta subunits, HIF can activate canonical target genes in response to oxygen deprivation by directly binding to hypoxia response elements (HRE) located within their regulatory regions [2,3]. Due to the short half-life of alpha subunits, HIF activity can be tightly regulated [4]. Mitochondrial oxygen sensing mechanisms produce highly labile reactive oxygen species to ensure that HIF-a subunit stabilization occurs only under hypoxic conditions [5]. Oxygen gradients that are generated as a function of tissue growth can thus activate HIF in a dynamic fashion to pattern the developing embryo [6]. Consistent with this, HIF activity is required for embryonic development [7,8,9].
During mammalian gestation, the placenta forms a vital transport interface between the maternal and fetal circulations and its development is also dependent on HIF activity [10,11,12,13]. In mice, two terminally differentiated cell types are primarily responsible for placental function: 1. trophoblast giant cells (TGC) and, 2. multinucleated syncytiotrophoblasts (SynT) [14]. TGCs anchor the placenta to the uterus and direct maternal blood flow to the conceptus while SynTs perform the transport functions of the placenta [15]. The derivation of trophoblast stem cells (TSCs) that differentiate into these placental cells in vitro has enhanced our understanding of placental development in vivo [16]. Interestingly, however, TSC differentiation using conventional techniques results primarily in the production of TGCs that is associated with the stabilization of HIF-a subunits independent of oxygen tension [10,17]. We previously demonstrated that HIF induction is critical for TGC differentiation as Hif-1/2a 2/2 (compound null) or Hif-1b (Arnt) 2/ 2 TSCs (collectively referred to as HIF-null) fail to produce TGCs in vitro, differentiating primarily into multinucleated SynTs, indicating that HIF activity can suppress cell fusion and SynT fate determination [10,17]. The mechanisms responsible for differentiation-dependent HIF induction and how HIF ultimately regulates cell fate in the placenta remain unknown, however.
Here, we show that TSC-extracellular matrix (ECM) interactions provide positional cues during normoxia that trigger differentiation-dependent HIF induction via signaling pathways that intersect with metabolic responses to oxygen deprivation. We find that altering the ECM substrate upon which TSCs are cultured impacts differentiation-dependent HIF stabilization and TSC fate. SynT formation is dependent on cell fusion -a process that is associated with significant cytoskeletal reorganization [18,19] [20,21]. Consistent with this, HIF-null TSCs exhibit dramatic morphological changes upon syncytialization [17]. We now show that non-canonical HIF-2 activity, induced in response to hypoxia or ECM composition, can prevent this process by promoting Limk1 expression and subsequent cytoskeletal stabilization.

ECM Composition Regulates TSC Fate and HIF Stability Independent of O 2 Tension
TSC proliferation depends on Fibroblast Growth Factor 4 (FGF4) as well as the presence of fibroblast ''feeder'' cells or fibroblast conditioned medium (Fib-CM) [16]. In the absence of either, TSCs default to a TGC differentiation program. While screening for culture conditions that could maintain FGF4dependent TSC growth independent of fibroblasts or Fib-CM, we identified the xeno-free defined ECM substrate, CELLstart TM (Invitrogen) [22,23]. This ECM substrate is composed primarily of Fibronectin, along with other ECM components [24], and thus represents a physiologically relevant substrate for TSC culture [25,26]. TSCs maintained on CELLstart TM in the presence of FGF4, but without fibroblasts or Fib-CM, proliferated indefinitely and expressed TSC-specific transcription factors such as CDX2 and EOMES [16] (Fig. 1A-D), the levels of which dramatically decreased following FGF4 withdrawal (not shown). Interestingly, however, differentiation in 21% O 2 (room air) following FGF4 withdrawal of TSCs maintained on CELLstart TM promoted cell fusion and resulted primarily in the formation of multinucleated SynTs (Fig. 1E, 1F), as opposed to the TGCs commonly observed with TSCs maintained on fibroblasts or on TC plastic in Fib-CM [16,17]. Importantly, differentiation under hypoxic (2% O 2 ) conditions could reverse this cell fate choice (Fig. 1G, 1H), blocking SynT formation and generating TGCs expressing the lineage specific transcription factor HOPX1 [27]. Lineage-specific gene expression analyses further confirmed that wild-type TSCs differentiated following culture on CELLstart TM expressed dramatically reduced levels of the TGC-specific markers Placental lactogens I and 22, Proliferin and Cathepsin Q, and exhibited increased levels of the SynT markers Tfeb and SynA, when compared with genetically identical TSCs differentiated following culture on TC plastic in Fib-CM (Fig. 1I). Importantly, this pattern of gene expression was similar to, though more pronounced than, that observed following differentiation of Arnt 2/2 or Hif-1/2a 2/2 TSCs that form SynTs following culture on TC plastic in Fib-CM (Fig. 1I) [10] [17]. We therefore asked whether the alteration of TSC fate following culture on CELLstart TM might be due to impaired HIF-a subunit stabilization that normally occurs during differentiation using standard techniques [17]. Indeed, TSCs differentiated following culture on CELLstart TM in 21% O 2 failed to stabilize HIF-2a and only slightly accumulated HIF-1a protein levels, whereas differentiation in 2% O 2 induced both proteins (Fig. 1J). Furthermore, TSCs derived from Vhlh 2/2 embryos, which exhibit constitutively elevated HIF-1a and -2a due to lack of VHL-dependent ubiquitination and proteasomal degradation [2,3], still formed TGCs in 21% O 2 following differentiation despite maintenance on CELLstart TM (Fig. 1K). These results suggest that TSC derivation on fibroblasts and maintenance on TC plastic in Fib-CM provides a set of extracellular cues that promote O 2 -independent HIF stabilization and subsequent TGC formation during differentiation that are lost when TSCs are maintained on the defined ECM substrate, CELLstart TM . To understand the mechanisms responsible, we concentrated on cell surface integrin expression, as these molecules play a central role in cell-ECM interactions [28]. TSCs maintained on CELLstart TM were compared with TSCs maintained on TC plastic in Fib-CM. Interestingly, culture on CELLstart TM completely blocked b3integrin expression in both undifferentiated and differentiated TSCs ( Fig. 2A-D), suggesting that ECM composition determines cell surface integrin expression and thereby modulates downstream signaling and TSC fate. Importantly, while undifferentiated TSCs maintained on TC plastic in Fib-CM expressed this b3integrin, its cell surface expression was largely restricted to differentiated TGCs, consistent with HIF induction being associated with TGC differentiation.

ECM-and Oxygen-dependent HIF Stabilization and TGC Formation Occur via MAP2K1/2-dependent Signaling
We suspected that candidate pathways capable of integrating ECM-dependent signals with oxygen-dependent inputs might include members of the mitogen activated protein kinase (MAPK) cascade. This is due to the fact each set of stimuli can independently activate this signaling cascade [29] [30,31,32,33]. Consistent with this, pharmacological inhibition of MAP2K1/2 (MEK-1/2) activity prevented HIF-a subunit stabilization during hypoxic (2% O 2 ) differentiation of TSCs cultured on CELLstart TM (Fig. 3A) and during normoxic (21% O 2 ) differentiation following culture on TC plastic in Fib-CM (Fig. 3B). Furthermore, hypoxic TGC formation of TSCs following culture on CELLstart TM could be suppressed with the same MAP2K1/2 inhibitor  (Fig. 3G). And finally, pharmacological MAP2K1/2-inhibition could prevent TGC formation and promoted SynT differentiation in TSCs that had been cultured on TC plastic in Fib-CM (Fig. 3H, 3I). Northern blot analysis confirmed that MAP2K1/2 inhibition in wild-type TSCs suppressed expression of the TGC as well as spongiotrophoblast (SpT) marker genes, Placental lactogen I and 4311, respectively, nearly to levels observed in differentiated Arnt 2/2 TSCs (Fig. 3J). Collectively, these results confirm that ECM-and oxygen-dependent HIF-a subunit stabilization and subsequent TGC formation occurs through a MAP2K1/2-dependent pathway.

Cytoskeletal Rearrangement is Central to MAP2K1/2mediated TGC Formation
Similar to mitochondrial responses to changing O 2 levels, dynamic integrin ligation in response to changes in ECM composition allow a cell to sense its environment by converting positional information into downstream signals [34]. These frequently result in cytoskeletal reorganization [35] that can promote cell migration or other alterations in cell behavior [36]. Additionally, trophoblast differentiation has been associated with significant cytoskeletal changes [19,20,37]. We therefore examined the cytoskeletal organization of differentiated control and Hif-1/ 2a 2/2 TSCs that had been maintained on TC plastic in Fib-CM and analyzed its association with MAP2K1/2 activation. TGCs derived from control TSCs contained robust MTs extending the length of the cell (Fig. 4A, 4D), while multinucleated SynTs derived from Hif-1/2a 2/2 TSCs exhibited a disrupted microtubule (MT) network consisting of ''broken'' appearing MT fragments (Fig. 4B arrows, 4F). MT integrity associated strongly with MAP2K1/2 activity, as control TGCs with robust MTs stained strongly for the phosphorylated versions of the MAP2K1/ 2 target MAPK3/1 (ERK-1/2) (Fig. 4A), while multinucleated SynTs derived from Hif-1/2a 2/2 TSCs did not (Fig. 4B). Only unfused SynT progenitors that did not contain ''broken'' MTs continued to exhibit the active form of this kinase in Hif-1/2a 2/2 TSCs (Fig. 4B, arrowheads). Additionally, we observed dramatic differences in the actin cytoskeleton, with robust stress fibers noted in TGCs (Fig. 4C) while SynTs exhibited a disorganized actin cytoskeleton containing high amounts of diffusely distributed Factin (Fig. 4E). To formally test whether cytoskeletal integrity could regulate TSC fate, we investigated whether pharmacological MT or actin disrupting agents could promote SynT formation of TSCs cultured using conventional techniques. Indeed, the MT disrupting agent Taxol (Paclitaxel) (Fig. 4G) and the actin disrupting agent cytochalasin B (Fig. 4H) inhibited TGC

HIF-dependent Limk1 Expression Promotes TGC Differentiation
We next focused our efforts on identifying cytoskeleton regulatory molecules that may be misregulated in the absence of HIF activity. Gene array studies indicated that expression of the gene encoding Lim domain kinase 1 (LIMK1), an enzyme responsible for regulating MT and Actin integrity [38], was significantly downregulated in the absence of HIF activity (not shown). Immunofluorescence microscopy confirmed that HIF-null TS cells differentiated into SynTs expressed no detectable LIMK1 expression, while control TGCs expressed robust LIMK1 protein levels in a perinuclear distribution (compare Fig. 5A and 5B). Immunoblot analyses confirmed this (see below) and indicated that the MAP2K1/2 inhibitor U0126 significantly decreased LIMK1 levels in wild-type TSCs differentiated following maintenance on TC plastic in Fib-CM (Fig. 5H), consistent with its ability to block HIF stabilization during differentiation. Furthermore, we found that the LIMK1 target Cofilin was highly phosphorylated, also in a perinuclear distribution, and therefore inactive in TGCs (Fig. 5D), but not in SynTs (Fig. 5E), which frequently contained prominent Cofilin rods that form when Cofilin is hyperactive [39] (Fig. 5E, arrowheads). The residual phosphorylation of Cofilin rods in SynTs may be due to their residual low-level LIMK2 levels (see below) [38]. Furthermore, transient LIMK1 expression in differentiating Hif-1/2a 2/2 TSCs promoted the appearance of large TGCs in the majority of transfected cells, as opposed to the SynTs commonly observed following their differentiation (Fig. 5F, 5G). These results suggest that HIF-dependent LIMK1 expression can regulate TSC fate downstream of ECM-or oxygen-dependent MAP2K1/2 activation by modulating the cytoskeleton.

Non-Canonical HIF-2a Activity Drives TSC Fate via LIMK1mediated Cytoskeletal Stabilization
We next investigated how HIF regulates Limk1 gene expression during TSC differentiation. We first determined whether we could detect canonical HIF-DNA interactions in differentiated TSCs via electrophoretic mobility shift assays (EMSA). As seen, control TSCs maintained on TC plastic in Fib-CM and differentiated in room air (21% O 2 ) contained abundant HRE-bound HIF complexes (Fig. 6A), consistent with our prior observations of HIF activation during normoxic differentiation of TSCs maintained using conventional methods (13). Interestingly, however, while two different HIF-1a-specific antibodies produced a ''supershift, SS'' in differentiated TSC nuclear extracts, a HIF-2a-specific antibody did not (Fig. 6A), suggesting that HIF-1a was predominantly responsible for canonical HRE-mediated gene expression in differentiating TSCs. To interrogate the requirement for direct HIF-DNA binding during TSC fate determination, we stably reconstituted Hif-1/2a 2/2 TSCs [10] utilizing a PiggyBac transposon system [40] either with HA-tagged wild-type HIF-1a or HIF-2a individually, or with mutant forms of each lacking their DNA binding basic domains (HIF-1aDb, HIF-2aDb) (Fig. 6B). Deletion of the basic domain prevents HRE binding and canonical target gene expression by HIF complexes without affecting their stability [41]. As expected, full-length versions of each were capable of activating their respective canonical downstream target genes in HIF-null TSCs, while versions lacking their basic domains could not (Fig. 6C, 6D). Using this system, we investigated the HIF-a subunit dependence of LIMK1 expression. Interestingly, while neither HIF-1a nor HIF-1aDb restored LIMK1 protein levels in Hif-1/ 2a 2/2 TSCs to that observed in wild-type TSCs, both HIF-2a and HIF-2aDb did (Fig. 7A, 7B), suggesting it to be a noncanonical HIF-2-specific target gene. We suspected that the HIF-2-specificity may be due to its known interaction with and activation of c-MYC-dependent transcription [42,43]. To test this, we immunoprecipitated HA-tagged HIF-2aDb in reconstituted Hif-1/2a 2/2 TSCs and detected strong c-MYC interaction (Fig. 7C). We next identified a 100% conserved c-MYC binding E box (CACGTG) in the Limk1 promoter, but no canonical HIF binding sites, and detected both c-MYC as well as HIF-2a binding to this site in situ in differentiated TGCs by chromatin immunopreciptation (Fig. 7D). We then tested whether pharmacological c-MYC inhibition could block non-canonical HIF-2a-dependent Limk1 gene expression, and found that it could decrease LIMK1 protein levels in HIF-2aDb expressing cells (Fig. 7E). This indicates that HIF-2a interacts with c-MYC containing transcriptional complexes during TSC differentiation and that this interaction contributes to Limk1 expression during TGC differentiation (Fig. 7F). Stable HIF-2aDb expressing HIF-null TSCs, whether cultured on CELLstart or on TC plastic in Fib-CM, failed to fuse into multinucleated SynTs and differentiated along the TGC lineage (Fig. 8A, 8B, 8C). Importantly, pharmacological cytoskeleton disruption (Fig. 8D, 8F, 8G) could reverse this effect, promoting SynT formation in these cells. Additionally, pharmacological LIMK1 inhibition [44,45] promoted the formation of multinucleated cells (Fig. 8E, 8H) in HIF-null TSCs reconstituted with HIF-2aDb.

Discussion
Collectively, our results indicate that oxygen-and canonical target gene-independent HIF activity can drive TSC fate in response to positional cues encoded by ECM components within the TSC microenvironment (Fig. 9). Initially thought to function as a mere scaffold, the ECM is now known to regulate many aspects of cell behavior, including proliferation and growth, survival, migration, and differentiation [28,46]. Primary components of the ECM are structural proteins (e.g., collagens, laminins, fibronectin, vitronectin and elastin) and specialized glycoproteins that can interact with molecules having important biological functions such as growth factors. The precise composition varies by location. In the stem cell niche, the ECM can provide instructive cues for cell fate decisions via the integrin family of heterodimeric cell surface receptors [26,47,48]. In erythropoiesis, for example, adhesion of primary erythroid progenitors to fibronectin mediated by a 4 b 1 integrin is necessary for proper proliferation in vitro [49]. In this system, signals from the ECM cooperate with signals from the soluble factor erythropoietin to activate pathways necessary for terminal differentiation and proliferation. In TSCs, altering their ECM in this way alters their cell surface b3-integrin expression, and subsequent HIF induction during differentiation. We have shown here that in TSCs, altering their ECM in this way alters their cell surface b3-integrin expression, and subsequent HIF induction during differentiation. Supporting a link between HIF activity and ECM-dependent integrin ligation, avb3 activation can trigger HIF accumulation in some cancer cells [50] and HIF-deficiency negatively affects TSC surface b3-integrin localization [51], suggesting that HIF-alpha subunit stability can both be activated by, as well as further promote, cell surface integrin activity. Importantly, trophoblast adhesion to ECM is governed by integrins [52][53] and hypoxic conditions promote trophoblast invasion in utero [54].
The mitogen activated protein kinase (MAPK) cascade is a frequent point of convergence from multiple environmental inputs [35,55]. We therefore reasoned that candidate signaling pathways responsible for oxygen-dependent HIF induction that could interface with ECM-dependent signaling would include members of this family. The classical extracellular regulated kinase (ERK)/ MAPK signaling pathway is an obvious candidate because it: 1. Regulates cell fate in a broad range of organisms [56,57,58,59]; 2. Responds to hypoxia downstream of mitochondrial ROS [60] and, 3. Is activated by integrin ligation [61]. Importantly, in mice, MEK1 (renamed MAP2K1), as well as its downstream target ERK2 (renamed MAPK1), are critical regulators of placentation [62,63,64,65]. Interestingly, MAP2K1 deficiency results in placental malformation characterized by an excess of multinucleated cells within affected placentas in vivo [66], similar to the formation of multinucleated SynTs with HIF-deficient TSCs in vitro, while genetic disruption of the MAP3K B-Raf results in placental malformation associated with diminished HIF-1a protein levels [67], consistent with our results.
Our results also shed novel insights into HIF-dependent cell fate determination. We found that HIF-2acan interact with c-MYC to enhance Limk1 gene expression and promote cytoskeletal integrity, thereby enhancing TGC differentiation, via non-canonical means. While HIF-2a has previously been shown to activate c-MYCdependent target genes [42,43], to our knowledge this is the first demonstration of a role for this mechanism during normal development. Importantly, c-MYC is required for normal placentation in mice [68]. Also, while HIF activity is generally assumed to result from the reduced oxygen tension frequently encountered during development, our results provide evidence that ECM composition can be added to a growing list of O2independent factors that can also drive HIF-dependent developmental programs such as Insulin-like Growth Factor 1 [69], or Runx2-mediated HIF stabilization [70]. Additionally, we show that the dramatic cytoskeletal changes observed in SynTs are not a simple by-product of cell fusion, but help regulate cell fate decisions in the placenta. LIMK1 is likely activated downstream of Rho kinase which is known to be induced during TGC formation  [37]. Interestingly, HIF stability can be influenced by MT integrity [71,72], suggesting the possibility of a feed-forward mechanism whereby MAPK-dependent HIF activity promotes MT integrity, which further enhances HIF stability during TGC differentiation.
In addition to playing important roles during normal placental development, the pathway outlined here is likely to be involved in pregnancy complications such as preeclampsia wherein fetal trophoblasts fail to properly invade maternal tissues. While aberrant oxygenation and HIF activity have previously been associated with preeclampsia [73,74], its ability to regulate LIMK1 has particular relevance, given that LIMK1 has recently been shown to be important for tumor cell invasion in humans [44] as well as collective cell migration in D. melanogaster [75]. Furthermore, in addition to altered hypoxia signaling, ECM remodeling is also frequently disrupted in PE [76], suggesting novel roles for HIF in linking these disparate processes. Further investigation of the intersection of these pathways by HIF activation should therefore yield novel insights into the etiology of this intractable syndrome.

TSC Culture
Mouse TSCs were derived on human placental fibroblasts as previously described [17]. Prior to differentiation experiments, TSCs were subjected to a differential plating to remove fibroblasts and subsequently cultured for two passages in 70% Fib-CM with FGF4 and heparin [16]. Differentiations were performed in standard TSC medium without FGF4 or heparin, on tissue culture

Adaptation of TSCs on CELLstart TM
For the adaptation of TSCs to the xeno-free substrate CELLstart TM (Invitrogen), cells were passaged from feeder culture using mild trypsinization and plated on CELLstart TM coated tissue culture dishes according to manufacturer's instructions. Cells were passaged in TSC medium with FGF4 and heparin, but without Fib-CM, upon reaching approximately 75% confluence. Following 5-6 passages, TSC lines on CELLstart TM exhibited a distinct morphology and maintained that morphology for greater than 20 generations in the absence of Fib-CM.

Northern Blot
Northern blot hybridization was performed with the probes for placental lactogen I and 4311 as described previously [17].

Immunoblotting and Immunofluorescence Staining
Whole cell lysates were prepared using a buffer consisting of 150 mM NaCl, 50 mM Tris-HCl (pH 7.4), containing 1 mM EGTA, 1 mM EDTA, 1% Triton X-100, 1% SDS and 10% glycerol. TSCs were incubated with PHEMT buffer (60 mM Pipes, 25 mM Hepes, ph 6.9, 10 mM EGTA, 4 mM MgCl 2 and 0.5% Triton X-100) for 30 min (4uC) and centrifuged to fractionate TSCs into soluble (supernatant) and insoluble fractions (pellet). Immunoblotting was performed with ECL (Amersham) or Odyssey Western blot methodologies and analyzed by Odyssey infrared imaging system (LI-COR Biosciences, Lincoln, NE) using the appropriate secondary antibodies. Briefly, 50-100 ug of whole cell lysates were run on 7.5-12% SDS-PAGE gel, transferred on PVDF membrane, applied with the primary antibody as indicated. For immunoprecipitation, 500 ug of whole cell lysates were immunoprecipitated with the indicated antibody and immuno-

Immunoprecipitation
Adherent cells were washed twice by addition of ice cold PBS to the monolayer and disposal of the supernatant. 1 ml of freshly made ice cold lysis/was buffer (50 mM Tris-HCl, 150 mM NaCl pH 7.5, 1% Nonidet P40 0.5% sodium deoxycholate supplemented with 1 complete tablet from Roche) was added to the washed cell monolayers to achieve a concentration of 10 6 -10 7 cells/ml. Cells were scraped into an eppendorf, and sonicated on ice with 5 pulses each for 8 seconds. Lysate was spun down at 13000 rpm for 5 minutes. Supernatant (except 200 ul) was put onto a new tube. The un-lysed pellet was resuspended into the 200 ul remaining lysate, and sonicated again, the tube centrifuged at 13000rpm for 5 minutes and the new lysate added to the original lysate. 50 ul of this lysate was kept aside as input. To reduce background a preclearing step was performed overnight. 50 ul of the homogeneous protein G-agarose (Roche) suspension, equilibrated in the lysis buffer, was added to the 1 ml lysate at 2-8uC on a rotating platform overnight. Beads were then pelleted by centrifugation at 20006g for 2 minutes at 4uC. Supernatant was transferred to a new tube. 50 ul of Agarose-coupled chicken anti-HA (Aves Labs, Inc. Oregon) was equilibrated in the wash/lysis buffer, centrifuged for 2 minutes at 20006g, and supernatant discarded. The cell lysate was added to these beads and rotated (gentle end-over-end mixing) overnight at 4uC. The lysate/bead complex was then centrifuged for 2 minutes at 20006g. Pellet was washed 46 by resuspending in lysis/wash buffer. A final wash was performed once for 30 minutes. Beads were then resuspended in 90 ul of 26 SDS sample buffer, boiled for 10 minutes at 95uC. Beads were collected by centrifugation at 27006g for 2 minutes at 4uC and SDS-PAGE performed with the supernatant.

CHIP Assay
Cells were washed with D-PBS and cross-linked by 1% formaldehyde (37 wt% from Sigma-Aldrich) for 10 min at 37uC. Glycine (2.5M) was added and incubated in room temperature for 10 minutes. Cells were then washed 36 in D-PBS for 5 minutes and harvested with D-PBS in the presence of protease inhibitor (EDTA-free Complete, Roche Applied Science). These cells were then centrifuged at 3000 rpm for 5 minutes and lysis buffer (1% SDS, 10 mM EDTA, 50 mM Tris-HCl, PH 8.1, with fresh protease inhibitor) was added, sonicated and incubated overnight at 65uC. Rnase A was added and incubated at 37uC, after which 2 ul 0.5 M EDTA, 4 ul 1 M Tris-HCL PH 8.1, 1 ul proteinase K was added and incubated for 2 hours at 45uC to generate 200-to 500-bp DNA fragments, which were subsequently confirmed by agarose gel electrophoresis. Pre-clearing was performed by using 50 ul protein G Sepharose (washed in dilution buffer, 0.01% SDS, 1.1% Triton X-100, 2 mM EDTA, 20 mM Tris-HCl, PH 8.1, 150 mM MgCl2), 30 ul normal IgG, 20 ug salmon sperm and rotated for 2 hours at 4uC. 20 ul was taken as input. Protein G sepharose and antibody complexes were prepared by re-suspending protein G with dilution buffer, 1 ug antibody and incubated on a rotator at 4uC overnight, and then washed twice with dilution buffer and centrifuged at 3000 rpm for 1 minute. Precleared samples were added to the antibody complex beads and rotated at 4uC overnight to collect antibody/antigen/DNA complex. Protein G complex was centrifuged at 3000 rpm for 1 minute, and supernatant removed. Protein G complex was washed sequentially for 5 minutes twice with: low salt buffer (0.1% SDS, 1% Triton X-100, 2 mM EDTA, 150 mM NaCl, 20 mM Tris-HCl, PH 8.1), high salt buffer (0.1% SDS, 1% Triton X-100, 2 mM EDTA, 500 mM NaCl, 20 mM Tris-HCl, PH 8.1), LiCl buffer (0.25 M LiCl, 1% NP-40, 1% deoxycholate, 1 mM EDTA, 10 mM Tris-HCl, PH 8.1), with TE buffer 1 mM EDTA, 10 mM Tris-HCl, PH 8.1). 200 ul of elution buffer was used containing 20 ul 10% SDS, 20 ul 1 M NaHCO3, 160 ul H2O. 100 ul of elution buffer was added to each tube containing the agarose/antibody complex or the input and incubated in room temperature for 15 minuites. Agarose complex was pelleted by centrifugation (50006g, 1 minute) and supernatant collected. This was repeated with another 100 ul elution buffer, and added to the first eluate (total volume 200 ul). Protein/DNA complexes were reversed to free DNA by adding 8 ul 5 M NaCl, 1 ul 10 mg/ml RNase A, and incubated at 65uC overnight. DNA was purified by using spin columns from Qiagen. Before purification, 4 ul 0.5 M EDTA, 8 ul 1 M Tris-HCl, and 1 ul protein kinase k was added to each tube and incubated at 45uC for 1 hour. Specifically bound purified DNA fragments were visualized by PCR using specific primers: (forward) 59-tgcatgcaccctaaataaaaata-39, (reverse) 59-ccttgaggagcacacataaccat-39.

mRNA Expression Analysis by Real Time PCR
RNA was extracted using TRIzolH Reagent (Invitrogen), and isolation was carried out according to manufacturer's instructions. 2 mcg of RNA per sample was made into cDNA using MMLV reverse transcriptase (Applied Biosystems). Prepared cDNA was amplified using SYBRH Green PCR Master Mix (Life Technologies) and the Bio-Rad iCycler iQ multicolor real time PCR detection system. Cycle threshold (Ct) values were normalized for amplification using Hypoxanthine guanine phosphoribosyl transferase (Hprt). Data analysis for real time quantitative PCR was done using the deltaCt method. Primer sequences are as follows: Prolilferin (Plf) Sense -tgaggaatggtcgttgcttt, antisensetctcatggggcttttgtctc. Placental Lactogen 1 (Pl-1) Sense -tggtgtcaagcctactccttt, Antisense -caggggaagtgttctgtctgt. Placental Lactogen 2 (Pl-2) Sense -ccaacgtgtgattgtggtgt, Antisense -tcttccgatgttgtctggtg. Hypoxanthine guanine phosphoribosyl transferase (Hprt) Sense -aaacaatgcaaactttgctttcc, Antisense -ggtccttttcaccagcaagct. Syncitin A (SynA) Sense -tactcctgcccgatagatga, Antisense -ccgtttttcttaacagtgggt. Syncytin B (SynB) Sense -ccaccacccatacgttcaaa, Antisenseggttatagcaggtgccgaag. Transcription factor EB (Tfeb) Senseaacaaaggcaccatcctcaa, Antisense -cagctcggccatattcacac. Trophoblast specific protein alpha (Tpbpa), Sense -cggaaggctccaacatagaa, Antisense -tcaaattcagggtcatcaacaa. Mammalian achaete-scute homolog 2 (Mash2) Sense -TTTTCGAGGACGCAATAAGC, Antisensecactgctgcaggactcccta. Statistical analysis of real time PCR results were performed as follows: All data points were performed in triplicate. One-way analysis of variance of the results was performed in Microsoft Excel 2007 to determine the presence of significant differences within the data sets. When analysis of variance indicated that a significant difference may be present, a two-sample Student's t-test was performed to compare experimental data with appropriate controls [77]. Statistical significance was determined at a value of P,0.05 and is represented with an asterisk.

Plasmid Constructs and Ectopic Expression
The full-length open reading frame of Hif-1a and Hif-2a were PCR amplified from cDNA constructs. Deletional mutants were generated removing the basic domain, Hif-1aDb (deletion of amino acids 4-27) and Hif-2aDb (deletion of amino acids 6-24) by high fidelity PCR. All expression constructs were modified to include the 9 amino acid hemagglutinin epitope YPYDVPDYA fused directly to the C-terminus and cloned into the ENTRD-TOPO vector (Invitrogen). The integrity of the constructs was confirmed by DNA sequencing. For expression in cell culture, a derivative of the Piggybac transposon system was employed allowing high efficiency expression. The parental plasmid EBXN Figure 9. Model of HIF-dependent integration of positional and metabolic cues in the TSC niche. ECM composition regulates HIF stabilization likely downstream of cell surface integrin ligation via MAP2K1/2 activation. Inside-out integrin signaling mechanisms may also be operative. Oxygen sensing and signaling pathways intersect with this signaling cascade to stabilize HIF, when ECM-dependent cues are absent. Stabilized HIF can act via canonical and non-canonical target genes. Non-canonical HIF-2, by interacting with MYC:MAX heterodimers, bind the Limk1 promoter to activate its expression. LIMK1 promotes microtubule and actin stability, critical for TGC formation, and thereby prevents SynT formation. HIF, therefore, can integrate divergent environmental inputs from within the placenta to regulate cell fate via non-canonical gene expression. doi:10.1371/journal.pone.0056949.g009 containing the minimal Piggybac 59 and 39 inverted terminal repeats as well as a CMV enhancer chicken Beta-actin promoter expression cassette was modified to include the SV40 promoter Blasticidin cassette allowing for eukaryotic selection in cell culture. The plasmid was further modified to include the Invitrogen Gateway Rfa cassette allowing for phiC31 mediated recombination. For monitoring transfection efficiency, EMCV IRES upstream of palmitoylated EGFP was inserted, PBX2.2. A control construct containing monomeric EGFP (Karel Svoboda, Addgene Plasmid 18696) was inserted into the parental plasmid PBX2.1.
Transfection of TSC lines was performed with Lipofectamine LTX and PLUS reagent (Invitrogen) in placental fibroblast-free culture. A 2:1 molar ratio of Piggybac transposase helper plasmid, PB, was combined with the transposon expression construct to mediate integration and high level expression. Selection with Blasticidin 5 mcg/ml was performed to identify stable integrants, which were subsequently passaged on placental fibroblasts.