Hypoxia Inducible Factor 3α Plays a Critical Role in Alveolarization and Distal Epithelial Cell Differentiation during Mouse Lung Development

Lung development occurs under relative hypoxia and the most important oxygen-sensitive response pathway is driven by Hypoxia Inducible Factors (HIF). HIFs are heterodimeric transcription factors of an oxygen-sensitive subunit, HIFα, and a constitutively expressed subunit, HIF1β. HIF1α and HIF2α, encoded by two separate genes, contribute to the activation of hypoxia inducible genes. A third HIFα gene, HIF3α, is subject to alternative promoter usage and splicing, leading to three major isoforms, HIF3α, NEPAS and IPAS. HIF3α gene products add to the complexity of the hypoxia response as they function as dominant negative inhibitors (IPAS) or weak transcriptional activators (HIF3α/NEPAS). Previously, we and others have shown the importance of the Hif1α and Hif2α factors in lung development, and here we investigated the role of Hif3α during pulmonary development. Therefore, HIF3α was conditionally expressed in airway epithelial cells during gestation and although HIF3α transgenic mice were born alive and appeared normal, their lungs showed clear abnormalities, including a post-pseudoglandular branching defect and a decreased number of alveoli. The HIF3α expressing lungs displayed reduced numbers of Clara cells, alveolar epithelial type I and type II cells. As a result of HIF3α expression, the level of Hif2α was reduced, but that of Hif1α was not affected. Two regulatory genes, Rarβ, involved in alveologenesis, and Foxp2, a transcriptional repressor of the Clara cell specific Ccsp gene, were significantly upregulated in the HIF3α expressing lungs. In addition, aberrant basal cells were observed distally as determined by the expression of Sox2 and p63. We show that Hif3α binds a conserved HRE site in the Sox2 promoter and weakly transactivated a reporter construct containing the Sox2 promoter region. Moreover, Hif3α affected the expression of genes not typically involved in the hypoxia response, providing evidence for a novel function of Hif3α beyond the hypoxia response.


Introduction
The lung originates from the primitive foregut early in the development of land dwelling organisms, and through a complex interplay of signaling molecules the future airway epithelium and surrounding mesenchyme develop into the highly structured arbor-like bronchial-vascular tree (reviewed in [1,2,3]). Normal development in mammals occurs in a relative hypoxic environment, which is beneficial for lung organogenesis [4,5]. Cellular responses to different levels of oxygen are important for development and homeostasis [6], and the most important oxygen-sensing mechanism to protect cells from oxygen toxicity is the transcriptional response mediated by Hypoxia Inducible Factors (HIF), which are also expressed in the lungs [7].
HIFs are critical mediators of the hypoxic cellular response and regulate cellular adaptation by transactivating genes involved in angiogenesis, metabolism and cellular homeostasis (for recent reviews see [6,8,9]). HIFs are heterodimeric transcription factors which have two structurally related subunits, an oxygen sensitive HIFa subunit and a constitutively expressed HIFß or ARNT subunit (Aryl hydrocarbon Receptor Nuclear Translocator). Both subunits belong to the transcription factor family containing a basic Helix-Loop-Helix (bHLH) and a Per/ARNT/Sim (PAS) domain at the N-terminus, which mediate heterodimerization and DNA binding [10,11]. HIFß is expressed ubiquitously and as such, the level and expression patterns of the HIFa proteins are mostly determining the activity of the heterodimers [12]. Currently, three genes have been identified in human and mouse that encode HIFa isoforms, HIF1a [10,11], HIF2a or EPAS1 [13,14,15], and HIF3a [16,17,18,19,20]. Aside from the N-terminal bHLH/PAS domain, the HIFa subunits contain an Oxygen-Dependent Degradation Domain (ODDD) in the center of the protein, an N-terminal transactivation domain (NTAD) and a C-terminal transactivation domain (CTAD) [21,22,23,24,25]. The CTAD is absent in the HIF3a subunit, which significantly reduces the transcriptional activity of the protein [16,26]. The three a subunits are posttranscriptionally regulated by prolyl hydroxylase domain-contain-ing enzymes (PHD1-3), which hydroxylate with different specificity the HIFa subunits at two critical prolyl residues in the ODDD under normoxic conditions [22,27]. The PHD proteins are dioxygenases which require oxygen for their function and as such are sensitive to oxygen concentrations, losing their activity under low oxygen concentration [22]. The hydroxylated HIFa proteins are poly-ubiquitinylated and targeted for 26S proteosomal degradation through the von Hippel-Lindau (pVHL)/Elongin BC/Cul2 ubiquitin-ligase complex [28,29,30,31,32,33,34]. Under low oxygen conditions, the PHD proteins are inactive, so the HIFa proteins are not hydroxylated and stable. They will translocate to the nucleus and dimerize with HIF1b, leading to the transcription of target genes, such as EPO and VEGF, through the binding to specific DNA seqences (Hypoxia Responsive Elements, HRE) [8,35,36]. Aside from the regulation of the stability of the HIFa isoforms by PHDs, additional regulatory activities are identified. The oxygen-dependent asparaginyl hydroxylase Factor Inhibiting HIF (FIH), a member of the Fe(II) and 2-oxoglutarate-dependent dioxygenase, hydroxylates a conserved asparaginyl residue in the CTAD, preventing the association of HIFa with the p300 coactivator [37,38,39]. In addition to these hydroxylation dependent regulation of HIFa isoforms, several other posttranslational modifications have been identified (for review, see [8,40,41]).
The regulation and functions of the HIF3a gene and isoforms is very complex, contrasting HIF1a and HIF2a. The HIF3a locus gives rise to different splice variants, resulting in three protein isoforms, HIF3a, NEPAS (neonatal and embryonic PAS) and IPAS (inhibitory PAS) [19,20,42]. HIF3a and NEPAS only differ in the first eight N-terminal amino acids due to alternative exon usage. IPAS and NEPAS are hypoxia inducible, whereas HIF3a is not because of alternative usage of promoters [43,44]. HIF3a expression is induced under hypoxia in several organs, including cortex, hippocampus, lung, heart, kidney, cerebral cortex [17,45,46]. NEPAS is almost exclusively expressed during late embryonic and neonatal stages of development, especially in the lung and heart, while HIF3a mRNA is rarely detectable during embryonic and neonatal stages [42]. HIF3a has a high homology to HIF1a and HIF2a at the N-terminus, but only a low degree of sequence similarity across the C-terminus [26]. The HIF3a/ HIF1ß (HIF3) and NEPAS/HIF1ß dimers suppress basal and hypoxia induced reporter gene activation, as well as HIF1 (HIF1a/HIF1ß) or HIF2 (HIF2a/HIF1ß) driven expression [16,42]. HIF3 binds to HRE sites in promoter regions, but the transcriptional activity is much weaker than that of HIF1 and HIF2, because it lacks the CTAD [16,26,42]. Therefore, both HIF3a and NEPAS serve as competitors of HIF1 and HIF2 dependent transcription, not only by occupying identical promoter regions, but also by associating with the same HIF1ß partner [16,42]. The splice variant IPAS lacks both the NTAD and CTAD domains producing a dominant negative regulator of the HIF1a and HIF2a dependent pathway [16,18,43]. It was shown that IPAS directly associates with HIFa isoforms, thereby displacing Hif1b, and the resulting IPAS/Hifa dimer is unable to bind to DNA [18]. Both short HIF3a isoforms related to IPAS in human and the IPAS in mouse have antagonistic effects on the expression of HIF1 and HIF2 dependent hypoxia regulated target genes [47]. Thus, the HIF3a locus encodes isoforms generally thought to act as negative regulators of the hypoxic response.
The importance of the hypoxia response was shown by the identification of mutations in the VHL-HIF pathway in different human diseases (reviewed in [9]). Specific gene ablation studies in mice also added to the knowledge on the pleiotropic effects of the members of the hypoxia response pathway. Complete ablation of this pathway through inactivation of Hif1ß resulted in a severe lethal phenotype with defective angiogenesis of the yolk sac and branchial arches, stunted development and embryo wasting [48,49]. Hif1a knockout mice also died early during development with cardiac malformations and vascular defects [50]. Hif2a null mice displayed a pleiotropic phenotype ranging from premature death until postnatal abnormalities, depending on the background of the mouse strain [51,52,53,54]. The neonates that survived suffered from breathing problems and did not produce sufficient surfactant phospholipids and surfactant associated proteins [51]. It is interesting to note that the inactivation and ectopic activation of Hif2a showed comparable phenotypes, suggesting that type II cells require different levels of Hif2a at distinct phases of type II cell maturation [51,55]. Homozygous mutant NEPAS/Hif3a -/mice were alive at birth, but displayed enlarged right ventricle and impaired lung remodelling, suggesting that NEPAS/Hif3a is important in lung and heart development during embryonic and neonatal stages [42]. Interestingly, the Hif3a gene contains hypoxia response elements in its promoter region and has been shown to be a transcriptional target of Hif1a [56].
In order to understand the precise role of Hif3a during pulmonary epithelium development, we generated transgenic mice with an inducible HIF3a gene. Mice expressing the HIF3a transgene in the developing airways showed a post-pseudoglandular branching defect with a reduced number of airspaces and a clear reduction in the number of alveolar type I and type II cells. Importantly, expression of the HIF3a transgene did not lead to changes in the levels of Hif1a, but affected Hif2a. The lungs of the HIF3a expressing mice showed an upregulation of genes normally expressed in the proximal parts of the lung, while genes only expressed in distal parts of the lung were downregulated. Specifically, Foxp2, a repressor of distal cell markers, and Rarß were induced in the lungs of Hif3a expressing mice, which may explain the reduction in the number of distal cell types. Furthermore, we showed that Hif3a binds a conserved HRE in the Sox2 promoter and induces the expression of a Sox2 promoter driven reporter gene, explaining the appearance of aberrant Sox2and p63 positive cells. Collectively, our results show that Hif3a is involved in modulating correct development of the lung epithelium.

Generation of transgenic animal
The myc epitope encoding sequence (EQKLISEEDL) was cloned directly after the endogenous ATG start codon of the full length human HIF3a cDNA (GenBank: BC080551) and subcloned into a modified pTRE-Tight vector [55]. Transgenic lines were produced by pronuclear injection of FVB/N fertilized eggs, and tail tip DNA of transgenic lines was initially genotyped by Southern blot analysis, after which positive lines were routinely checked by PCR, using transgene-specific primers (sense: 59-GTCAAGCTTATGGCGCTGGGGCTGCAGCG; antisense 59-GCATCTAGATCAGTCAGCCTGGGCTGAGC). Three independent lines were initially analyzed, which all produced the same phenotype as described in this manuscript. Lung-specific expression of the HIF3a transgene, i-Tg-mycHIF3a, was obtained by crossing the mycHIF3a lines with the SPC-rtTA transgenic mice (A generous gift of Jeffrey Whitsett). Administration of doxycycline to pregnant mothers from gestational age 6.5 onward in the drinking water (2 mg/ml, 5% sucrose) resulted in lung epitheliumspecific expression. Each experiment was performed with at least three independent litters containing double transgenic, single transgenic and wild type pups. All double transgenic animals receiving doxycycline expressed mycHIF3a in the pulmonary epithelium and showed the described phenotype. Mice were housed under standard conditions at 40-50% relative humidity and 2161uC (12/12 hour dark/light cycle) with food and water ad libitum. All animal experiments were performed according to the Dutch and European guidelines and approved by the local ethics committee (DEC Nr 1657, 1833 and 2206).
Lungs were imaged using an Olympus BX41 microscope and DP71 camera (Olympus, Zoeterwoude, The Netherlands). Subsequent airspaces counting were performed with SIS Softward Cell D (Olympus). Three independent samples of control and doubletransgenic lungs of gestational age E18.5 were used to count the number of airspaces on a selected surface area (140000 mm 2 ) on those selected lung samples.

Microarray analysis
Lungs of three control and three double transgenic embryos were dissected at E18.5 and the middle and caudal lobes were used for total RNA isolation with Trizol reagent according to the manufacturer's instructions (Invitrogen life technologies, Carlsbad, CA, USA). RNA was purified using the RNeasy MinElute Cleanup kit. (Qiagen, Valencia, CA, USA) and cDNA was synthesized from 3 mg RNA using the GeneChip Expression 39-Amplification Reagents One-Cycle cDNA Synthesis kit (Affymetrix, Santa Clara, CA, USA). Biotin-labelled cRNA synthesis, purification and fragmentation were performed according to standard conditions. Fragmented biotinylated cRNA was subsequently hybridized onto Affymetrix Mouse Genome 430 2.0 microarray chips. After normalization, the data were analysed with OmniViz software, version 3.6.0 (Omniviz, Inc., Maynard, MA, USA).
Functional annotation of the statistical analysis of microarrays results was done using Ingenuity Pathway Analysis (Ingenuity, Mountain View, CA) and DAVID (http://david.abcc.ncifcrf.gov). The results are shown for biological processes, which are significantly (P ,0.05) enriched after multiple testing.

RT-PCR
RNA isolation and subsequent quantitative PCR analysis was essentially performed as previously described [7]. Gene-specific primer sets were Abca3 Luciferase reporter activity assays HEK293T cells were transfected in duplo with Lipofectamine 2000 (Invitrogen) with a total concentration of 500 ng DNA/well, using 9*HREluc (Gift from Manuel Landazuri), pGL3-mpSox2 and pGL3-mpSox2delta (Named Sox2-Luc and DSox2-Luc; Gift from Victoria Moreno), Hif2a-pcDNA3, (gift from Carole Peyssonnaux), Hif3a-pcDNA3 or pcDNA3. Cells were lysed with passive lysis buffer (Promega) 24-hours after transfection and processed for lucifease analysis by the addition of the LARII reagent (Promega), which was subsequently quantified with the VICTOR luminometer. A construct containing the renilla gene (10 ng/well) was co-transfected in each well to serve as an internal control for transfection efficiency. The renila luciferase activity was quantified by addition of Stop&Glio reagent and also detected with the VICTOR luminometer. The experiment was repeated three times, and all samples were measured at least in duplo. The average luciferase activity was calculated and divided by the average of renilla activity. Standard deviations were measured with the SPSS program (Independent-samples T test), which also generated the P values.

Chromatin immunoprecipitation (ChIP)
ChIP assay was performed essentially as previously described [58], with some modifications. Chromatin-protein complexes of confluent A549 cells were fixed by adding 1% formaldehyde to the cultures. Nuclear extracts were made and chromosomal DNA was fragmented by sonication. Equal amounts of DNA was diluted 1:10 with ChIP dilution Buffer (0.01% SDS, 1.1% Triton X-100, 1.2 mM EDTA, 16.7 mM Tris-HCl pH 8.1 and 167 mM NaCl) and the samples were pre-cleared with 80 ml prot A/G agarose beads for 1 hour, after which the sample was split in equal volumes and incubated O/N with 6 mg antibody specific for HIF3a (NBP1-03155) or control IgG (rabbit). Immune complexes were subsequently purified by adding 80 ml of prot A/G beads, which were washed several times before the immune-precipitated DNA was eluted with elution buffer (1% SDS and 0.1 m NaHCO 3 ). After de-crosslinking the DNA-protein complexes by incubation at 65uC O/N with 200 mM NaCl, the eluted DNA was phenolextracted, precipitated and qPCRs were performed to analyze the enrichment of HIF3a specific binding to the HRE in the SOX2 gene using the following primer set 59-CAAGTGCATTTTAGC-CACAAAG-39 and 59-CCCAAGAGGGTAATTTTAGCCG-39, while the primers for the ARRDC3 and EGLN3-D were described previously [36,59]. The data are the average of two independent ChIP assays, which were each analyzed by duplicate qPCRs, and are represented as the fold enrichment of the specific immune precipitation compared to the control IgG precipitation.

Ectopic expression of mycHIF3a causes late branching defects
Previously, it was shown that homozygous NEPAS/Hif3a knockout mice were viable, but displayed an enlarged right ventricle and impaired lung remodelling, suggesting that Hif3a plays an important role during pulmonary development. However, the precise role of Hif3a during the formation of the lung is not fully understood. We first analyzed the endogenous expression of Hif3a in normal fetal lungs isolated at the end of gestation (E18.5) and in lungs of adult mice (8 weeks). Hif3a positive cells were present in the epithelium of the developing lung, as well as in the type II pneumocytes of the adult lung (arrows in Figure 1A, B). In order to determine the precise role of Hif3a in the epithelium during lung development, and more specifically in type II pneumocytes, we generated transgenic mice carrying a mycepitope tagged HIF3a under the control of a doxycyclineinducible tet-on promoter (i-Tg-mycHif3a; Figure 1C). Expression of mycHIF3a in embryonic lung epithelium was established by crossing the i-Tg-mycHIF3a transgenic line with the established SPC-rtTA line, which drives the expression of the rtTA gene in epithelial cells of the embryonic lung [60]. Pregnant females from timed matings between SPC-rtTA and i-Tg-mycHIF3a mice received doxycycline to induce the expression of the HIF3a transgene in double-transgenic fetuses. Lungs isolated from doxycycline-induced or non-induced single i-Tg-mycHIF3a or SPC-rtTA transgenic mice, or lungs from non-induced double transgenic i-Tg-mycHIF3a/SPC-rtTA animals appeared indistinguishable from wild type lungs. Doxycycline-induced, doubletransgenic pups were born at Mendelian ratio and did not show obvious external abnormalities compared to their control litter mates.
In order to determine whether expression of mycHIF3a leads to pulmonary development defects, we analyzed lungs of doubletransgenic animals and control lungs at different gestational ages. Macroscopic analysis of isolated lungs did not show clear abnormalities in double-transgenic animals at gestational ages E16.5, E17.5, E18.5 days and postnatal day 1 (PN1) (Figures 1E and F, I and J; Figure S1). Histological examinations at E16.5 did not show clear differences between control and mycHIF3a transgenic lungs ( Figures S1C and D). However analysis of a series of developmental ages clearly showed aberrant alveolar airspaces in mycHIF3a expressing lungs starting at E17.5 compared to controls ( Figure S1G, H). mycHIF3a expressing lungs contained significant fewer alveolar spaces compared to control ones at E18.5 and PN1 (Figures 1D). Staining with a specific antibody against the myc-epitope confirmed the expression of transgenic mycHIF3a protein in the epithelium of doubletransgenic lungs (Figures 1H and L, Figure S1). The abnormal alveolar spaces remain present in the PN1 stages, but apparently, the mice do not suffer from respiratory distress, indicating that the initial requirements for life are present. So, we conclude that mycHIFa expression in epithelial cells leads to aberrant alveolar formation and affects late branching morphogenesis during pulmonary development.
This post-pseudoglandular branching defect prompted us to analyze the expression of the mycHIF3a at early embryonic stages of development. This showed that the transgene is expressed in a non-uniform manner in the epithelium of early E11.5 lungs (Figure 2A), but gradually all epithelial cells express the transgene ( Figure 2B-D). Next, we analyzed whether the primary airway branches appropriately expressed some of the major branchinducing genes [2]. Therefore, embryonic lungs of controls and double transgenic animals were isolated at gestational age 12.5. At this stage of development, the primary bronchi are already present, and these branches start to form secondary and tertiary branches. The expression of Fgf10, the growth factor with a very potent branch-inducing activity, was found in the mesenchymal compartment, alongside the epithelium that is in the process of branching ( Figure 2E and I, arrows). Moreover, its receptor, FgfR2, was detected at the tips of the epithelium, in close proximity of the Fgf10 signal ( Figure 2F and J). Next, we also analyzed the expression of two genes known to be induced as a result of the Fgf10-FgfR2 signalling, Shh and Bmp4. Both genes were also expressed in the epithelium at the same location as the FgfR2, indicating that the Fgf10-FgfR2 signalling cascade is intact ( Figure  2G and K; H and L). In addition, quantitative PCR analysis of embryonic lungs isolated at E12.5, E15.5 and E17.5 of controls and double transgenic mice using primers specific for FgfR2, FgfR2-IIIb, FgfR2-IIIc, Bmp4 and Spry did confirm the absence of differential expression of these important branch-inducing genes (data not shown). In conclusion, no differences in expression pattern were observed for the early branch-inducing genes between controls and double transgenics, suggesting that the initiation of the branching process occurred normally.

mycHIF3a expression inhibits Clara cells differentiation
Since we observed significant alveolar changes and aberrant branching morphogenesis, we analyzed the integrity and differentiation potential of fetal transgenic lungs by immunohistochemistry with cell-specific markers. The smooth muscle cell component of the mesenchyme (a-Sma) did not reveal striking differences between control and transgenic lungs ( Figures 3A, B). Thyroid transcription factor (Ttf1) was expressed in nearly all epithelial cells in both control and transgenic lungs ( Figures 3C, D). Ciliated cells (b-tubulin) and neuroendocrine cells (cGRP) were present in proximal conducting airways of control and transgenic lungs at gestational age E18.5 ( Figures 3E, F and 3G, H, arrows). Moreover, both type I (T1a; Figure 4A, B) and type II pneumocytes (Lpcat1; Figures 4C, D) were present in the alveolar regions. These results indicate that differentiation into the various epithelial cell types is not hampered by Hif3a, although the total number of each cell type may be different. In addition, no differences were observed in the proliferation of epithelial or mesenchymal cells between control and transgenic lungs as indicated by Ki67 staining (Figure 4E, F).
Next, three mycHIF3a-expressing lungs and three control lungs were processed at gestational age 18.5 days for microarray analysis, to elucidate the origin of the aberrant branching morphogenesis. Hierarchical clustering of differentially expressed genes revealed large differences between controls and double transgenic lungs ( Figure 5A) and the major biological processes ( Figure 5B) and molecular functions ( Figure 5C) are indicated. Although mycHIF3a does not prevent the differentiation of epithelial cells into Clara cells, we noticed that the number of Clara cells was significantly reduced. Both in the microarray analysis as well as the qPCR validation showed downregulation of the Ccsp gene in mycHIF3a transgenic mice. These gene expression results were confirmed by immunohistochemistry, showing that Ccsp positive cells were less prominent in the proximal airways of the Hif3a expressing lungs compared to control lungs (Figures 6A-D). Quantification of the total number of Clara cells revealed a significant reduction in the double transgenic mice ( Figures 6H). So, our data show that mycHIF3a expression inhibits Clara cells differentiation during pulmonary development.

mycHIF3a induces airway epithelial cells to differentiate into proximal cell types
Analysis of the microarray data revealed that genes associated with proximal cell types of the lung appeared to be upregulated, whereas genes specifically expressed in distal epithelial cells were downregulated (Table 1 and Table 2). The induction of proximal markers is reflected by the significant downregulation of genes specific for the distal lung epithelium. The type 1 pneumocyte cell Hif3a Inhibits Distal Pulmonary Epithelium PLOS ONE | www.plosone.org marker Aquaporin 5 (Aqp5) was dowregulated in the Hif3a expressing mice, as were three genes specifically expressed in type II pneumocytes, stearoyl-coenzyme A desaturase (Scd1), surfactant associated protein D (Sftpd) and ATP-binding cassette (ABC) subfamily A3 (Abca3) ( Figure 6E) [61,62,63]. Quantification of the number of type II pneumocytes present in the Hif3a expressing lungs using Sftpd in reference to Ttf1 confirmed a significant reduction in these cells ( Figure 6G). Since we are inducing the Hif3a family member of hypoxia inducible genes, we analyzed the expression of Hif1a and Hif2a in the transgenic lungs. Although no apparent difference could be detected for Hif1a ( Figure 6F), but we did notice a significant downregulation of Hif2a (Epas1) ( Figure  6E). Previously, we showed that Hif2a is involved in maturation of type II pneumocytes, so the reduction of Epas1 expression could be directly related to the loss of type II cells.
Among the upregulated genes are two transcription factors known to play important functions during lung development, Foxp2 and Sox2 [57,64]. Foxp2 is important during lung development and is expressed in the distal parts of the lung. It represses the transcription of several distal cell markers, such as T1a, Spc, and Ccsp [65]. In our microarray analysis, Foxp2 was significantly upregulated, which we validated by quantitative PCR (Table 1 and Figure 7G). Staining with a Foxp2 antibody show that the distribution of Foxp2 positive cells in Hif3a double transgenic lungs was expanded compared to control lungs ( Figures  7A, D), suggesting that Hif3a suppressed the transcription of genes specific for alveolar epithelial cells through the induction of Foxp2. In addition, Rarb, which is expressed at proximal sites in the lung from embryonic day 11 to 12 and not in the distal epithelium of the lung [66,67], was significantly induced in Hif3a transgenic mice ( Figure 7G), confirming the expansion of proximal cell makers in these lungs [64,65].
Sox2 is important for pulmonary branching morphogenesis, epithelial cell differentiation and is exclusively expressed in the proximal parts of the lung [57]. However, in mycHIF3a expressing lungs, Sox2 is present in epithelial cells of both proximal airways and certain alveoli at postnatal day 1, suggesting that Hif3a is able to induce proximal cell fate ( Figures 7B, E,  arrows). The basal cell marker p63 is expressed in the esophagal and tracheal epithelium, and previously we showed that ectopic Sox2 expression induced the appearance of p63 positive cells in the epithelium of the bronchioles and enlarged distal airspaces [57]. Therefore, we analysed the distribution of basal cells in the mycHIF3a expressing lungs and found that p63 is abnormally expressed in the alveolar epithelial cells of mycHIF3a expressing lungs, contrasting the unique expression in the trachea (Figures 7C  insert, arrows F). Collectively, our data indicate that mycHIF3a expression leads to the induction of crucial genes, such as Sox2, Foxp2 and Rarß, which cause airway epithelial cells to differentiate into proximal cell types.

Hif3a binds the promoter region of Sox2 and induces transcription of Sox2
The promoter region of the Sox2 gene contains two functional HREs, which are bound by Hif2a [68]. Since Sox2 is upregulated in Hif3a transgenic lungs, we analyzed whether Hif3a can directly induce the transcription of Sox2. Therefore, we first performed transcription reporter assays using a luciferase reporter construct under the influence of the Sox2 promoter containing two HREs, or two mutated HREs (Sox2-Luc and DSox2-Luc [68]). Hif3a induced the expression of the Sox2-Luc promoter about 2 fold, whereas the DSox2-Luc promoter was hardly induced compared to controls ( Figure 7H). The positive control, HRE, was considerably induced by Hif2a, but only mildly by Hif3a, corresponding with the weak transcriptional activity of Hif3a [16,42]. Under hypoxia-mimicking conditions, induced by adding CoCl 2 to the medium, which inhibits prolyl hydroxylases by displacement of Fe(II) from their catalytic center [22], Hif3a could induce the 9*HRE-Luc considerably, and the difference with the Hif2a induced expression was much reduced (10 times versus 2 times). Moreover, the induction of the Sox2-Luc construct by Hif3a was 4 times higher than under normoxic conditions, and was comparable between Hif2a and Hif3a ( Figure 7H). Subsequent analysis of the 1 kilobase region immediately upstream of the Sox2 transcriptional start site revealed that the most upstream of the two putative HRE sites was highly conserved between mice and human [68]. In order to investigate whether Hif3a could directly bind this conserved HRE site, we performed a chromatin immunoprecipitation of chromatin-protein complexes isolated Hif3a Inhibits Distal Pulmonary Epithelium PLOS ONE | www.plosone.org from human A549 cells. Analysis of the HIF3a precipitated chromatin showed that the region containing the conserved HRE site in the SOX2 promoter region was indeed preferentially enriched compared to the IgG fraction ( Figure 7I). ARRDC3 was used as a potential positive control, as it is bound by both HIF1a and HIF2a, and the enhancer region D of the EGLN3 gene served as negative control [36,59]. Indeed, HIF3a did not bind to the EGLN3-D region, but did bind to the ARDDC3-HRE. This indicated that HIF3a could bind the HRE site present in the Sox2 promoter, suggesting a potential direct regulatory role of Hif3a in the transcription of Sox2.
So, Hif3a binds to the conserved HRE in the Sox2 promoter and weakly induces Sox2 expression, resulting in an abnormal Sox2 expression in airway epithelial cells of HIF3a transgenic lungs.

Discussion
Hypoxia inducible factors are an important family of proteins involved in the regulation of the cellular response to hypoxia. Its functions are required from the earliest steps of mammalian life to the correct development of multiple organs and tissues, like the placenta, trophoblast formation, bone development, heart and vascular development (reviewed in [6,8]). The importance of the hypoxia response was shown by the identification of human mutations in the VHL-HIF pathway in different diseases [9]. Gene ablation studies in mice have revealed in more detail the specific and important roles of the different subunits of the Hifa/Hifß heterodimers. Inactivation of the stable subunit, Hif1ß, resulted in severe embryonic defects and premature death [48,49]. The disruption of the different Hifa genes identified specific roles for the individual Hifa isoforms. Hif1a knockout mice die early at gestation, have multiple developmental defects in neural tubeforrmation, vascularization, heart development, neural crest migration [69,70,71], whereas depending on the genetic background of the mouse strain, Hif2a knockout out mice ranging from early embryonic lethality to adulthood [51,52,53,54].

Hif genes and lung development
The lung is under continuous exposure of external oxygen and several (patho)-physiologic conditions trigger global or local hypoxia in the lung, resulting in pulmonary abnormalities to which HIFs contribute, such as lung cancer, acute lung injury and pulmonary hypertension (reviewed in [72]). Long term changes in oxygen levels, as experienced at high altitude gives rise to lung damage as a result of chronic mountain sickness. Recently, the EPAS1 gene, encoding for HIF2a, was shown to be associated with adaptation of living at high altitude [73,74,75,76].
Inactivation of Hif2a in mice resulted in respiratory distress and surfactant deficiency in newborns on a mixed genetic background [51]. Remarkably, heterozygous Hif1a +/or Hif2a +/mice showed a reduced increase in pulmonary arterial pressure and right ventricular hypertrophy upon exposure to chronic hypoxia in comparison with wild type mice [77,78]. Ectopic expression of an oxygen-insensitive Hif1a transgene in lung epithelial cells during development resulted in defective branching, impaired epithelial maturation and respiratory distress. Moreover, increased expres-  Hif3a Inhibits Distal Pulmonary Epithelium PLOS ONE | www.plosone.org sion of VegfA and VegfC was observed, leading to sub-pleural hemorrhaging [79]. We recently showed that the transgenic expression of an oxygen-insensitive mutant of Hif2a also lead to a late branching defects with enlarged alveoli and altered epithelial differentiation [55]. Contrasting the Hif1a transgenic study, we did not find increased levels of VegfA or endothelial abnormalities, even though the transgenes were expressed in the same manner. This indicates that Hif1a and Hif2a have different effects. In addition, the expression of Hif1a had not changed, whereas Hif3a expression was reduced in our Hif2a transgenic mice [55]. It seems that the effects of Hif1a are more widespread, whereas the number of affected genes by Hif2a is restricted, which is in line with previous reports describing target genes of Hif1a and Hif2a [35,36,80,81,82,83,84,85].
The occurrence of the Hif3a isoforms is well described transcriptionally, but the functional analysis is complicated by the appearance of different splice variants [19,26,42,43,86]. Hif3a isoforms act as negative regulators of the traditional Hif1 (Hif1a/ Hif1ß) and/or Hif2 (Hif2a/Hif1ß) driven hypoxia response by functioning as dominant negative modulators, effectively resulting in the transcriptional competition with Hif1 and Hif2 [16,18,26,42,43]. Gene ablation of Hif3a/NEPAS/IPAS, resulted in mice that were born alive with enlarged right ventricles and impaired lung remodelling [42]. Furthermore, they showed that expression of endothelin-1 is negatively influenced by Hif3a/ NEPAS, by regulating the binding of Hif1a and Hif2a to the HRE sites if the ET-1 promoter, which may contribute to the observed phenotype. Remarkably, the expression of Vegf, a direct target of Hif1 and Hif2, had not changed, even though the expression of Hif1a and Hif2a was not affected. This hinted at a selective regulation of target genes by NEPAS/Hif3a during pulmonary development. Therefore, we conditionally expressed mycHIF3a in airway epithelial cells during embryonic development in order to further elucidate the role of Hif3a in pulmonary development.

Cellular effects of mycHIF3a transgene expression
Since the NEPAS/Hif3a knockout mice suggested a selective regulation of genes by Hif3a, and our Hif2a transgenic mice showed a selective reduction in Hif3a expression, we conditionally expressed mycHIF3a in airway epithelial cells during embryonic development in order to further elucidate the role of Hif3a in pulmonary development. Analysis of mice expression a transgenic mycHIF3a in lung epithelium revealed a late branching morpho- Surprisingly, no apparent defects are observed early during lung development, even though the transgene is expressed. This may be due to the fact that at these stages of development, putative associating factors of Hif3a, like Hif2a and Hif2a, are not expressed yet. After the pseudoglandular stage of lung development, endogenous Hif2a becomes expressed in the cells positive for mycHIF3a and the effect of the mycHIF3a transgene starts to be noticeable. Histological analysis and gene expression profiling revealed changes in the differentiation of the developing pulmonary epithelium. We found reduced numbers of Clara cells, alveolar type I and type II cells, and in addition, basal cells were observed in atypical spatial positions. The expression pattern of diverse sets of genes was affected, and revealed that mycHIF3a expression mainly affects Hif2-directed transcription, although not all Hif2 target genes are equally affected. We show that expression of mycHIF3a in epithelial cells results in a down regulation of Hif2a, but not of Hif1a. This suggests that Hif3a is not a global regulator of the hypoxic response, but that Hif3a may selectively function to modulate Hif2a controlled target genes, supporting previous work [42]. The reduction in the expression level of Hif2a late in gestation may be due directly to the presence of mycHIF3a, or due to the impaired differentiation of the type II cells. However, it is clear that mycHIF3a does affect the differentiation of epithelial cells, and this could partly be explained by the aberrant activation of specific genes that are not part of the hypoxic response. Gene expression analysis does not show significant changes in typical hypoxia responsive genes, which indicates that Hif3a may have specific functions beyond the hypoxia response. Therefore, we provide first evidence for novel Hif3a functions beyond the hypoxia response.
The apparent increase in the mesenchymal compartment after the pseudoglandular stage does not seem to be induced by proliferation, as we did not observe an increase in mitotic cells in the mycHIF3a lungs. It may be due to either a delayed development of the double transgenic lungs, or, alternatively, to a specific response in epithelial cells triggered by mycHIF3a. Lysyl oxidase may be activated, which subsequently activates a cascade of proteins, such as Snail, involved in the repression of E-cadherin, and ultimately leading to epithelial-mesenchymal transition, as described for metastatic tumors [87,88].

Genes affected by mycHIF3a
The appearance of proximal cells at the expense of distal cells in the mycHIF3a lungs is paralleled by transcriptional changes in several genes, such as Sox2, Rarß and Foxp2. At this point, it remains to be seen whether all effects observed are directly related to mycHIF3a, or that the expression of mycHIF3a affects Hif1a and Hif2a specific complexes, thereby interfering with transcription of specific genes. The increased expression of mycHIF3a could lead to the formation of complexes that normally are not present in the cell, which would shift the balance between Hif2a and Hif3a.
We observed Sox2 positive cells at unusual sites in the lung, which was supported by the aberrant presence of p63 positive basal cells. Previously, we showed that Sox2 directly induces the appearance of basal cells [57]. Since a link was found between Hif2a and Sox2 transcription [68], we analyzed the putative regulation of the Sox2 gene by Hif3a. We show that Hif3a is capable of inducing basal expression of a reporter construct under the control of the Sox2 promoter containing two HRE sites. In addition, we show that HIF3a binds to the conserved HRE sequence in the Sox2 promoter, which suggests that Hif3a may contribute directly to the regulation of Sox2 expression. However, the minimal transcriptional activity of Hif3a, as also shown previously, may explain the appearance of only scattered Sox2 positive cells in the lungs of mycHIF3a mice [16,26,42]. In addition, depletion of individual HIFa genes by siRNA in human ES cells suggested that HIF3a upregulates HIF2a, which subsequently induced the expression of stem cell marker genes, like SOX2 [89]. Although this hypothesis is intriguing, no direct relationship was established, yet. It was also shown that ectopic expression of HIFs in cancer cell lines can induce embryonic stem cell markers, like SOX2 and NANOG [90]. The combination of weak transcriptional activity and the ability to act as a dominant negative modulator of Hif2a may be responsible for the transcriptional regulation of Sox2. These results directly show that through the expression of HIF3a, Sox2 + and p63 + basal cells appear and suggest that the balance between Hif2a and Hif3a may function as a modulator of basal cell emergence [68].
Besides the aberrant induction of Sox2 and p63, the expression domain of Rarb was expanded distally in the mycHIF3a transgenic lungs. Rarb knockout mice exhibited premature septation, and formed alveoli twice as fast as wild-type mice [66,67,91]. So, upregulation of Rarb in mycHIF3a transgenic mice may in part explain the observed inhibition of pulmonary alveoli formation. We also detected an increase of Foxp2, which is a transcriptional repressor able to inhibit the expression of Ccsp and markers specific for distal epithelial cells, such as Spc and T1a [64,65,92]. Therefore, the reduced numbers of Clara cells (Ccsp + ), alveolar type I (Aqp5 + ) and alveolar type II (Sftpd + ) cells could be directly related to the upregulation of Foxp2. Recent findings showed that depletion of cells with CCSP promoter activity was associated with  alveolar hypoplasia and respiratory failure, adding to the idea that Ccsp downregulation as a result of Hif3a-mediated Foxp2 upregulation, directly leads to reduced numbers of Clara cells [93]. The increase d expression of key genes in lung development, which lead to major changes in epithelial differentiation, was confirmed by the loss of expression of other cell type specific markers,, such as Sftpd, Scd1 and Abca3 for type II cells. At this point it is not clear if the reduced expression of the type II cell markers is the cause, or the result of the loss of type II cells. Previously, we showed a significant downregulation of Scd1 and . The induction of the Sox2 promoter is higher with Hif2a than with Hif3a under normoxic conditions (4,8 versus 2,5), but equally strong under hypoxia mimicking conditions (8,8 versus 7,3). Data are presented as the induction (n-fold) relative to cells transfected with the corresponding reporter plasmid and control vector (pcDNA3). The values are the average of two duplicates, and standard deviations are: 0,04 (HRE-Hif2a), 0,02 (Sox2-Hif2a), 0,03 (DSox2-Hif2a), 0,08 (HRE-Hif3a), 0,24 (Sox2-Hif3a), 0,06 (DSox2-Hif3a), 0,53 (HRE-Hif2a+CoCl2), 0,007 (Sox2-Hif2a+CoCl2), 0,03 (DSox2-Hif2a+CoCl2), 0,88 (HRE-Hif3a+CoCl2), 0,02 (Sox2-Hif3a+CoCl2), 0,1 (DSox2-Hif3a+CoCl2). (I) Chromatin immunoprecipitation (ChIP) using anti-HIF3a antibody and chromatin isolated from A549 cells. Graph represents the fold enrichment of the HIF3aspecific binding to the conserved HRE of the SOX2 promoter compared to the IgG control ChIP. HIF3a also bound the ARRDC3 HRE region, and the enhancer region D of the EGLN3 gene served as negative control (EGLN3-D). doi:10.1371/journal.pone.0057695.g007 Hif3a Inhibits Distal Pulmonary Epithelium Abca3 in Hif2a expressing transgenic mice, which suffered from respiratory distress and surfactant deficiency [55]. However, the mycHIF3a transgenic mice appeared to produce sufficient levels of Scd1 and Abca3 to support respiration, even though the expression of Hif2a is decreased.
Thus, the increased expression of Sox2, Rarß and Foxp2 in the developing mycHIF3a lungs may directly contribute to the cellular changes observed and explain the phenotypic abnormalities observed in these lungs. The effects may also be cell type specific, as increased HIF3a expression in vascular cells resulted in an antagonistic effect on hypoxia induced HIF1/HIF2 target genes [47].

Concluding remarks
Although we cannot conclude that the dominant negative role of Hif3a as part of the hypoxic response is absent, our previous and current data do suggest that Hif2a and Hif3a have different target genes, during pulmonary development [55]. This is in line with previous findings describing common targets, as well as specific genes induced by Hif1a and Hif2a [80,81,82]. However, these studies used overexpression of Hif1a and Hif2a, which may cause aberrant complexes and loss of target gene specificity, as was reported for certain tumor cells [94]. Using siRNA and chromatin immuno-precipitation approaches, HIF1 and HIF2 target genes were identified [35,36,83,84,85]. Interestingly, it was shown that ETS transcription factors were involved in the regulation of HIF1 and HIF2 driven gene activation in MCF7 cells [83]. Knock down of ELK1 resulted in a reduction of hypoxia induced HIF2 dependent transcription. These data suggested a cooperation between ETS family members and HIF1 and HIF2 in the selection of target genes. An interesting idea is that target selection by HIFs may be cell specifically regulated by additional factors, adding to the complexity of the hypoxic response [8,95]. This is also observed in the analysis of the different transgenic mouse models expressing Hif1a [79], and our studies with HIf2a or Hif3a, showing similarities and differences [55].
Thus, in spite of the limited functional significance of Hif3a/ NEPAS in development as a global regulator of the hypoxia response, we demonstrate that Hif3a does contribute by balancing the function of the Hif regulated genes. Furthermore, Hif3a contributes to late branching morphogenesis, alveolar formation and epithelial differentiation. Moreover, the level of Hif3a, as well as Hif1a and Hif2a, is tightly regulated to ensure balance between the total number of proximal cells and distal cells. Figure S1 Expression of mycHIF3a leads to late branching defect. External appearances of control (A and E) and mycHIF3a transgenic lungs (B and F) at E16.5 and E17.5 showed no apparent differences. Histological analysis of control (C and G) and mycHIF3a transgenic (D and H) lungs showed a gradual decrease in the number of air spaces and aberrant, late branching morphogenesis in mycHIF3a transgenic lungs. Anti-Myc epitope staining confirmed the expression of the mycHIF3a transgene in double transgenic lungs (D and H), which is absent in control lungs (C and G).