Introduction

Lack of oxygen occurs in a number of processes, e.g., embryonic development1, postnatal maturation2 and tumorigenesis3. Under hypoxia condition, hypoxia-inducible factors (HIFs) are activated and function as transcriptional regulators of genes involved in the hypoxic response4,5,6. HIFs are heterodimers consisting of an oxygen-labile α subunit (HIFα) and a stable β subunit (HIFβ). HIFα in mammals includes three isoforms, of which HIF1α and HIF2α are the best characterized7. In environments with sufficient oxygen, HIFα subunits are hydroxylated at conserved proline residues by prolyl hydroxylases (PHDs) and rapidly degraded by ubiquitin-proteosome system. While in the hypoxic environment, activity of PHDs is diminished and as a consequence, HIFα subunits become stabilized. The resulting stabilized HIFα proteins dimerize with HIFβ and activate the transcription of the target genes8. HIF1α is expressed ubiquitously, while the expression of HIF2α is more restricted, with highest expression levels in heart, lung and placenta9.

HIF2α is encoded by endothelial PAS domain protein 1 (EPAS1) gene. Studies on EPAS1 inactivation revealed that mice with loss of EPAS1 developed a syndrome of multiple-organ pathology10 and may die at mid-gestation11. Thus it is hypothesized that certain EPAS1 expression levels are important for its proper functions in the course of embryo development. Genetic variation in the regulatory region of EPAS1 gene may influence the transcription levels of EPAS1 and may further contribute to the changes of the biological functions. Recently, genome-wide studies on high-altitude adaptation have suggested several SNPs in the regulatory region (intron and 3′ downstream) of EPAS1 gene responsible for the genetic adaptation of high-altitude hypoxia in Tibetans12,13,14. However, whether there are variants in the promoter region of EPAS1 gene related to high-altitude adaptation awaits further studies. Functional studies focusing on the molecular mechanism need to investigate the direction and the magnitude of gene expression changes associated with the variants and figure out what tissues, developmental time points and what the downstream target genes involved.

Thus, we had three specific aims in this study. Firstly, we were interested to identify the variants in the promoter region with divergence between Tibetans and Han Chinese. Secondly, we wanted to explore the direction and the magnitude of EPAS1 gene expression changes associated with the variants. Thirdly, we intended to investigate the candidate targets that are regulated by EPAS1.

Results

Eleven variants were identified in the promoter region of EPAS1 gene

As compared with the 1000 Genome Data, we found that Tibetans have several specific variants in the promoter region of EPAS1 gene (Table 1), among which four variants have minor allele frequency (MAF) larger than 0.05. A 40-bp insertion was identified at −742 bp relative to the transcriptional start site of EPAS1 gene (NM_001430.4, Supplementary Fig. S1) in Tibetans but was absent in Chinese Han individuals (Supplementary Fig. S2). Another three variants newly identified in Tibetans lay at −706 bp, −138 bp and −129 bp, respectively and two of them (−138 bp and −129 bp) were in strong LD with the −742 indel polymorphism (R2 = 1, Supplementary Fig. S3). Moreover, rs56721780:G>C and rs13428739:C>T lying at −886 bp and −607 bp, respectively, had much higher MAF in Tibetans than that of Han Chinese (0.372 vs. 0.010 and 0.39 vs. 0.010, allele frequency in Han Chinese was from 1000 Genomes Project) and were also in strong LD with the −742 indel polymorphism (R2 = 0.8 and 0.95, respectively). The frequency of the haplotypes constructed by the three variants (rs56721780:G>C, −742 indel and rs13428739:C>T) in strong LD was listed in Table 2.

Table 1 Variants in the promoter region of EPAS1 gene identified in Tibetans
Full size table
Table 2 Information of the haplotypes defined by rs56721780:G>C, −742 indel and rs13428739:C>T
Full size table

C allele of rs56721780:G>C decreased the binding affinity of IKZF1 to the cis-acting element in EPSA1 promoter

To further analyze the function of the three variants used to construct haplotypes, function prediction was performed by TFSEARCH software. The Prediction results demonstrated that transcription factor IKZF1 isoform 3 and 5 (also known as Ik-2 an Ik-3, respectively15) bind to the cis-acting element with allele G at rs56721780:G>C site, while the binding effect is missing for allele C (Supplementary Fig. S1). In this study, EMSA experiment was designed to confirm the binding effect of IKZF1 to the target site of EPAS1 promoter. Biotin-labeled G allele or C allele probe was incubated with nuclear extract from HEK293 cells transfected with pcDNA3.1-IKZF1-myc plasmid expressing IKZF1 isoform 3 and probe-protein complexes were observed for both of the two probes (Figure 1a and 1b). To verify the binding specificity, excess unlabeled G or C probe competitor was added to the EMSA reaction and the probe-protein complex binding revealed a different affinity for the two alleles. Higher concentration of unlabeled C competitor was required to compete for binding to both of the G probe and C probe, as compared to unlabeled G competitor. The addition of anti-myc antibody targeting IKZF1 resulted in a supershift of the upper complex (Figure 1a and 1b). These results suggest IKZF1 binds more avidly to the G allele, compared to the C allele.

Figure 1
figure 1

G allele of rs56721780:G>C increased the transcription binding affinity of IKZF1 and the 40-bp insertion provided the binding site for Sp1.

Electrophoretic mobility shift assay was performed with biotin-labeled (a) G probe or (b) C probe and (c) 40-bp insertion probe. These experiments were repeated more than three times independently with similar results.

Full size image

The 40-bp insertion fragment in EPSA1 promoter provided the binding site for Sp1

The 40-bp insertion fragment at −742 position was predicted to have the binding site for transcription factor Sp1 (Supplementary Fig. S1). EMSA experiment observed a shift band for the probe-protein complex (Figure 1c). The intensity of the shift band decreased upon pre-incubation with excess unlabeled 40-bp competitor or anti-Sp1 antibody, confirming the binding effect of Sp1 to the 40-bp insertion fragment. As to the −607 site of rs13428739:C>T, no transcription factor binding site was found in the EPAS1 promoter region by the prediction software.

C allele of rs56721780:G>C and the 40-bp insertion increased the promoter activity of EPAS1 gene

To examine the potential function of IKZF1 and Sp1 in regulating EPAS1 gene, pcDNA3.1-IKZF1-myc or pcDNA3.1-Sp1-myc and pGL3-EPAS1 promoter-luciferase plasmid (pGL3-G/deletion/C or pGL3-C/deletion/C or pGL3-C/insertion/T) were co-transfected into HEK293 cells. The results showed that IKZF1 could decrease the transcriptional activity of EPAS1-G/deletion/C promoter, EPAS1-C/deletion/C promoter and EPAS1-C/insertion/T promoter by 60%, 43%, 31%, respectively (Figure 2a), suggesting IKZF1 was a potential transcriptional repressor for EPAS1 gene and had stronger inhibition effects on EPAS1-G/deletion/C promoter than those on EPAS1-C/deletion/C and EPAS1-C/insertion/T promoters. Our results also demonstrated that Sp1 could increase the transcription activity of EPAS1-G/deletion/C, EPAS1-C/deletion/C and EPAS1-C/insertion/T promoters to 1.66 times, 1.69 times, 2.09 times, respectively (Figure 2b), indicating Sp1 was a potential transcriptional activator for EPAS1 gene and had stronger activation effects on EPAS1-C/insertion/T promoter, as compared to those on EPAS1-G/deletion/C and EPAS1-C/deletion/C promoters.

Figure 2
figure 2

IKZF1 and Sp1 were a potential repressor and activator for EPAS1 gene, respectively.

Luciferase activity was measured after HEK293 cells were co-transfected with pGL3-G/deletion/C or pGL3-C/deletion/C or pGL3-C/insertion/T luciferase reporter construct and (a) pcDNA3.1-IKZF1 plasmid or (b) pcDNA3.1-Sp1 plasmid. pcDNA3.1-empty plasmid was used as negative controls. The relative luciferase activity was calculated as the ratio of Firefly luciferase activity vs. Renilla luciferase activity. Fold changes were calculated as the ratio of relative luciferase activity with IKZF1 or Sp1 over-expression vs. relative luciferase activity without IKZF1 or Sp1 over-expression. Data represent as mean ± S.E. from 3 independent experiments with three replicates in each experiment. Statistical analysis was performed using t test. **P < 0.01.

Full size image

LOX was identified as a direct target of EPAS1

To investigate changes in gene expression associated with EPAS1, global gene expression profiling analysis of Flp-In-293 cells with/without stably expressed EPAS1 was carried out on microarrays. In total, 25 significantly up- or down-regulated (P < 9 × 10−7) genes were identified (Supplementary Table S1). The increased expression of lysyl oxidase (LOX) gene was particularly intriguing as (i) Both of the two probes of LOX gene were significantly up regulated with the most significant P values, 2.78 × 10−32 and 1.35 × 10−31, respectively; (ii) LOX gene encodes an extracellular copper enzyme that catalyzes formation of aldehydes from lysine residues in collagen and elastin precursors, resulting in cross-linking of collagen and elastin16; (iii) LOX gene was highly expressed in amnion tissues17, which contained a large percentage of collagen and were easier to collect compared with other tissues of human. Subsequent qRT-PCR analysis confirmed the results found in global gene expression profiling analysis (Figure 3a). ChIP experiment further verified that two hypoxia inducing elements (HREs), which contain the core recognition sequence 5′-RCGTG-3′, located upstream 183 to 230 bp relative to the transcriptional start site of LOX might be the binding sites for EPAS1 (Figure 3b). Luciferase expression derived by LOX promoter could be induced about 10-fold of increase by EPAS1 over-expression (Figure 3c) in HEK293 cells. Due to LOX involving in extracellular matrix protein cross-linking of amnion18, we further analyzed LOX and EPAS1 mRNA expression levels in amnion from Tibetan newborns. Our results showed that LOX mRNA levels were highly positively correlated with EPAS1 mRNA levels in amnion (Pearson correlation coefficient R2 = 0.45, P value < 0.0001, Figure 3d).

Figure 3
figure 3

LOX was a direct target of EPAS1.

(a) LOX was up-regulated when EPAS1 was over-expressed in Flp-In 293 cell line by qRT-PCR analysis. (b) LOX was enriched with anti-EPAS1 antibody as compared to normal IgG by ChIP-qPCR analysis. (c) EPAS1 over-expression could induce human LOX promoter luciferase expression up to 10-fold. (d) mRNA expression levels of EPAS1 and LOX in amnion were highly correlated by Pearson correlation analysis. Both of x axis and y axis were plotted in ΔCt manner, with GAPDH as the endogenous control. Data were represented as mean ± S.E. from three independent experiments. *P < 0.05, **P < 0.01, **P < 0.001.

Full size image

Three variants were associated with EPAS1 and LOX expression in the amnion from Tibetan newborns

Although we have showed that rs56721780:G>C and −742 indel regulated EPAS1 expression in human cells, whether the regulation effect works in human tissues was not clear. To further understand the role of rs56721780:G>C and −742 indel in transcriptional regulation, we compared the mRNA levels of EPAS1 and LOX between different genotype groups in amnions from Tibetan newborns. The results showed that the individuals with GG or GC genotypes had much lower EPAS1 and LOX mRNA levels compared with the individuals with CC genotypes at rs56721780:G>C (Figure 4a and Supplementary Fig. S4). Similarly, the individuals with deletion allele at −742 indel (Figure 4b and Supplementary Fig. S4) or C allele at rs13428739:C>T (Figure 4c and Supplementary Fig. S4) had much lower EPAS1 and LOX mRNA levels compared with the individuals without the corresponding alleles. As to the effects of different haplotypes on the expression of EPAS1 and LOX genes, it was found G/deletion/C haplotype (Figure 4e and Supplementary Fig. S4) was related to decreased EPAS1 and LOX mRNA levels and individuals with two copies of C/insertion/T haplotype had much higher EPAS1 and LOX mRNA levels (Figure 4d and Supplementary Fig. S4). Western blot analysis demonstrated that IKZF1 and Sp1 were present in amnion (Supplementary Fig. S5), suggesting that rs56721780:G>C and −742 indel may also regulate EPAS1 expression by means of IKZF1 and Sp1 binding, respectively, in human amnion.

Figure 4
figure 4

Three variants in the promoter region of EPAS1 gene were correlated with EPAS1 gene expression levels in amnion.

Individuals possessing wild-type alleles at the positions of (a) rs56721780:G>C, (b) −742 indel and (c) rs13428739 had lower EPAS1 expression levels in amnion. Individuals with two copies of (d) C/insertion/T haplotype had much higher EPAS1 expression levels in amnion, whereas the (e) G/deletion/C haplotype decreased EPAS1 expression levels. Relative expression levels were log2 transformed for plotting, with GAPDH as the endogenous control. Data were presented as mean ± S.E. Statistical analysis was performed using the t test or t test with Welch's correction implemented in the software of Graphpad Prism. *P < 0.05, **P < 0.01. The numbers in parentheses indicated the counts of individuals with the corresponding genotypes. NA, not available or without the haplotype indicated.

Full size image

Three variants were associated with birth weight of Tibetan newborns

LOX gene encodes lysyl oxidase, which is responsible for normal extracellular matrix protein cross-linking. The reduced activity of lysyl oxidase is thought to be associated with abnormal development of the fetus, such as bone abnormalities, decreased integrity of vessels19 and preterm births17 with low birth weight fetus. Thus in this study, we were interested to test whether there was relationship between the expression levels of LOX and birth weight and we found that individuals with the genotypes or haplotype associated with lower mean LOX expression levels also had lower mean birth weight (Figure 5). The results demonstrated that the individuals with G allele at rs56721780:G>C (Figure 5a), deletion allele at −742 indel (Figure 5b) or C allele at rs13428739:C>T (Figure 5c) had much lower birth weight compared with the individuals without the corresponding alleles. The haplotype G/deletion/C (Figure 5e) was found to be related to decreased birth weight and individuals with two copies of C/insertion/T haplotype correlated with higher birth weight (Figure 5d).

Figure 5
figure 5

Three variants in the promoter region of EPAS1 gene were associated with birth weight of Tibetan newborns.

Individuals possessing wild-type alleles at the positions of (a) rs56721780:G>C, (b) −742 indel and (c) rs13428739 had lower birth weight. Individuals with (d) C/insertion/T haplotype had higher birth weight, whereas the (e) G/deletion/C haplotype decreased birth weight. Data were presented as mean ± S.E. Statistical analysis was performed using the t test or t test with Welch's correction implemented in the software of Graphpad Prism. *P < 0.05, **P < 0.01. The numbers in parentheses indicated the counts of individuals with the corresponding genotypes. NA, not available or without the haplotype indicated.

Full size image

Discussion

Mutations in coding sequence of EPAS1 gene have been found to be associated with several kinds of human disorders, such as polycythemia20,21, paragangliomas or somatostatinomas22. Such mutations affect the stability of HIF2α proteins and thus lead to dis-regulation of HIF2α. In addition to mutations, a common SNP rs17039192:C>T in the exon 1 of EPAS1 gene (+18 bp relative to the transcription start site) was reported to be associated with knee osteoarthritis in Japanese and the susceptibility allele C showed higher EPAS1 promoter activity in chondrogenic cells23. However, little is known about the underlying mechanisms of common SNPs in promoter region regulating EPAS1 transcription. In this study, we found a SNP variation rs56721780:G>C and an indel variation −742 indel in EPAS1 promoter region regulated EPAS1 expression by means of binding with two transcription factors, IKZF1 and Sp1, respectively. Both of the two minor alleles enriched in Tibetans, C allele at rs56721780:G>C and insertion allele at −742 indel, conferred the elevated transcription of EPAS1 gene. IKZF1 encodes a transcription factor that belongs to the family of zinc-finger DNA binding proteins and has been shown to regulate expression of its target genes in activation or repression manner via chromatin remodeling24, playing a key role in a variety of cellular functions, such as cell proliferation25, cell apoptosis26 and Notch cell signal transduction pathway27,28. IKZF1 is initially characterized as a hemo-lymphopoietic system-restricted transcriptional factor to regulate lymphocyte differentiation and subsequent studies also demonstrated that IKZF1 expresses in placenta and transactivates the expression of leucine aminopepetidse/insulin-regulated aminopeptidase gene29. IKZF1 gene possesses at least 13 spliced transcript variants encoding different isoforms (isoform 1–13). All isoforms share C-terminal domain containing two zinc finger motifs that required for hetero- or homo-dimerization and for interactions with other proteins, but differ in the number of N-terminal zinc finger motifs containing nuclear localization signal and requiring for DNA binding30. Both of the two isoforms (isoform 3 and 5) predicted as transcription regulators for EPAS1 gene contain requisite three N-terminal zinc fingers, but only isoform 3 has normal nuclear localization30 and thus confers high affinity binding to a specific core DNA sequence element in the promoters of target genes. Thus for in vitro experiment, we only studied IKZF1 isoform 3. The prediction results (Supplementary Fig. S1) showed G allele of rs56721780:G>C lies in the sequence of GGTTTCCCAGGA (the letter in bold), which is specific for IKZF1 binding to EPAS1 promoter. When G is changed to C, the binding of IKZF1 to EPAS1 promoter might be eliminated. In fact, our EMSA results revealed that the C allele only decreased the binding of IKZF1 to EPAS1 promoter. The inconsistency of the EMSA results with the prediction on the G>C change might due to the location of the G>C allele in the binding site of IKZF1 to EPAS1 promoter, i.e. it was not included in the core sequence (TTCCCA) of the high affinity binding element of IKZF131. IKZF1 was identified as a new transcriptional repressor for EPAS1 gene in this study. The results demonstrated that G allele of rs56721780:G>C increased the binding of IKZF1 to the cis-acting element, leading to the enhanced transcription repression of EPAS1 gene. The decreased inhibition of IKZF1 on EPAS1-C/insertion/T promoter as compared with that on EPAS1-C/deletion/C promoter (Figure 2A) might be caused by the steric hindrance of other transcriptional factors binding to the 40-bp insertion fragment. Sp1 is a zinc finger transcription factor that binds to GC-rich motifs of many promoters acting as an activator32 or a repressor33, which is involved in many cellular processes, including cell differentiation, cell growth, apoptosis34, immune responses35 and so on. Sp1 contains three spliced transcript variants encoding different isoforms (isoform 1–3). We focused on the longest isoform (isoform 1) in this study. We found Sp1 could bind to the 40-bp insertion fragment at −742 indel site and activate the transcription of EPAS1 gene. The results were not unexpected given a previous study demonstrating that Sp1 and the sequence −478/−445 in the promoter region of EPAS1 gene were responsible for the activation of EPAS1 expression during adipogenesis in 3T3-L1 cell36.

Another highlight in this study is that LOX gene was found to be a direct target of EPAS1. Numerous gene array studies37,38,39 including ours searching for novel HIF-regulated genes have drawn LOX as one of the highest regulated genes. Although functional studies have been successfully identified the binding site in the LOX gene for HIF140, whether there is direct binding effect of HIF2 to LOX gene remains unknown. In this study, ChIP experiment successfully determined the binding effect of HIF2α/EPAS1 to the promoter of LOX gene and the binding effect was further confirmed to be functional by luciferase report assay.

LOX gene encodes an extracellular copper enzyme, which has been well studied in human fetal membranes because of its importance in cross-linking both of collagens and elastin, thus increasing the tensile strength of these proteins17. Such strength is particularly required in the third trimester of pregnancy when the growth of the fetal membranes fails to keep pace with the growth of the fetus41. If the membranes fail to distend in the third trimester of pregnancy, they may rupture prematurely almost always resulting in preterm delivery17,18 with low-birth-weight newborns, who are likely to experience physical health problems or die during the early age of life as compared to the infants with normal weight18. The human fetal membranes constitute of amnion and chorion and resistance to tearing and rupture is provided primarily by the amnion18. This is the main reason that amnion tissues were used in this study. All the amnion tissues in our study were collected from laboring vaginal deliveries of healthy pregnant women from Shigatse (~3,800 meters in altitude), Tibet. Although a recent report demonstrated that placentas subjected to labor are oxidatively stressed compared with non-labored placentas delivered by cesarean section42, laboring placentas at high altitude (3,100 m) have little or no oxidative stress at the time of delivery irrespective of whether they are labored or delivered by cesarean section43. Thus, it is inferred that labor would be not an inducer of oxidative stress of the amnion tissues collected at ~3,800 m in our study. Our results showed that LOX mRNA levels were highly correlated with EPAS1 mRNA levels and Tibetans with the genotypes or haplotype associated with lower mean LOX expression levels also had lower mean birth weight. The relationship between LOX and birth weight has also been observed in a previous study investigating the effect of copper depletion throughout gestation on neonates44. The results showed the offspring born to female rabbits fed marginal copper diets (1.5 parts per million copper) had lower birth weight than that of the group fed control diets (10 parts per million copper). And the lungs of pups from the marginal copper diet group also had lower LOX activities. The findings indicating that lower LOX activities in lung or other organs might be related to the lower birth weight of the neonates. However, clear reports about the regulation of birth weight by LOX were limited. More efforts should be made to verify the regulation of birth weight by LOX in future studies.

It is possible that the decreased LOX expression regulated by EPAS1 in amnion might cause inadequate cross-linking of interstitial collagen and reduced tensile strength, leading to premature rupture of the membranes or intrauterine growth restriction of the Tibetan fetus and producing babies with lower birth weight. Reduced birth weight likely compromises survival during the early postnatal period. Thus, it is presumed that the genotypes and haplotype which were enriched in Tibetans and associated with higher LOX expression levels in amnion might be helpful for fetal growth and provide Tibetans with high infant survival rate that tightly correlates with heavier birth weight and may contribute to high-altitude adaptation of Tibetans45.

A previous genome-wide study13 on high-altitude adaptation of Tibetans showed that H1 haplotype of EPAS1 gene was overrepresented in Tibetans (with a frequency of 89.1%) but was significantly less prevalent in other populations (e.g., 3.4% in East Asian populations), indicating the roles of the H1 haplotype in high-altitude adaptation of Tibetans. In this study, we constructed a C/insertion/T haplotype with a frequency of 34.6% in Tibetans but absent in Chinese Han individuals. All the Tibetans with C/insertion/T haplotype also possessed the EPAS1-H1 haplotype, suggesting the C/insertion/T haplotype was full concordance with EPAS1-H1 haplotype.

In conclusion, we showed that EPAS1 gene was regulated by IKZF1 and Sp1 at the sites of rs56721780:G>C and −742 indel, respectively and might play important roles in the development of amnion, fetus growth and high-altitude adaptation of Tibetans by regulating LOX expression. Further studies will increase our understanding of the important roles of EPAS1 during development.

Methods

Amnion and umbilical cord blood

Fresh placental tissues and umbilical cord blood samples were obtained from 43 newborns delivered at the People's Hospital of Shigatse District, Tibet. All the deliveries were vaginal deliveries of healthy pregnant women from Shigatse (~3,800 meters in altitude). Amnion was manually separated from choriodecidua and immediately washed extensively with sterile physiological salt solution and cut into segments before storing in RNAlater (Ambion, Austin, TX, USA) at −20°C following the manufacture's instructions. Birth weight of the newborns was measured immediately after delivery using a regularly calibrated weighing scale. This study also randomly selected 25 Chinese Han DNA samples from our database to serve as controls13. This study was conducted with an approval from the Internal Review Board of the School of Life Sciences at Fudan University in Shanghai, China. The methods were carried out in accordance with the principles of the Declaration of Helsinki. Written informed consent was obtained from all the subjects included in the study or their parents.

Gene cloning and plasmid construction

Since IKZF1 is mainly expressed in the hemo-lymphopoietic system, Jurkat cells which are an immortalized line of T lymphocyte cells were used for IKZF1 isoform 3 (NCBI accession no. NM_001220766.1) cloning. Approximately 1 × 106 cultured Jurkat cells were subjected to total RNA extraction and cDNA synthesis. The coding region of human IKZF1 isoform 3 was amplified using primers pcDNA3.1-IKZF1-F and pcDNA3.1-IKZF1-R (Table S2) at 98°C, 5 min; 98°C, 30 s; 68°C, 2.5 min for 35 cycles; then 72°C, 10 min and was cloned into the EcoRI and HindIII restriction sites of pcDNA3.1/myc-His(-)A vector. The generated plasmid was named pcDNA3.1-IKZF1-myc.

The coding sequence clone of human EPAS1 gene (NCBI accession no. NM_001430.4) was purchased from Genecopoeia (Guangzhou, China). The coding sequence was subcloned into the 3× flag-pEGFP-N1 vector which was provided by Dr. Jing Zheng, East China University of Science and Technology. The fragment containing coding sequence and 3× flag tag was then subcloned to pcDNA5/FRT/TO expression vector (Invitrogen, Carlsbad, CA, USA). The coding sequence of human Sp1 gene was subcloned into the pcDNA3.1 vector using the pcDNA3.0-Sp1 expression plasmid as a template which was kindly provided by Dr. Xingzhong Wu from Fudan University46. A 2,088 bp of DNA fragment (from −1988 bp to +100 bp relative to the transcription start site) of the human EPAS1 gene (Figure S1) and the promoter sequence from −268 bp to +300 bp relative to the transcriptional start site47 of lysyl oxidase (LOX) gene (NCBI accession no. NM_002317.5) were cloned by PCR utilizing the genomic DNA as the template. The amplified products were inserted into pGL3-basic vector (Promega, Madison, WI, USA) to generate pGL3-EPAS1 (pGL3-G/deletion/C) plasmid and pGL3-LOX plasmid, respectively. pGL3-G/deletion/C plasmid was further used as a template to generate pGL3-C/deletion/C plasmid by PCR-based site-directed mutagenesis at the position of SNP rs56721780:G>C (−886 bp relative to the transcription start site). pGL3-C/insertion/T plasmid was constructed using the genomic DNA of the individual with C/insertion/T haplotye as the template. The PCR primers used for DNA amplification were listed in Supplementary Table S2. All plasmids were sequence-verified before use.

Cell culture and stable cell line establishment

Human embryonic kidney (HEK) 293 cell and Jurkat cell lines were purchased from the Cell Bank of the Chinese Academy of Sciences and were cultured under recommended conditions. Flp-In-293 cells (Invitrogen) were maintained in DMEM supplemented with 10% (v/v) heat-inactivated fetal bovine serum, 2 mM L-glutamine, 100 μg/mL of Zeocin, 15 μg/mL of blasticidin at 37°C with 5% (v/v) CO2. Stably transfected Flp-In-293 cell lines were generated by integrating the pcDNA5/FRT/TO vector containing EPAS1 coding sequence into a single defined genomic locus via Flp recombinase.

RNA extraction, microarray assay and quantitative real-time PCR (qRT-PCR)

Total RNA from Flp-In-293 cells and amnion was extracted with TRIzol Reagent (Sigma-Aldrich, St. Louis, MO, USA). Microarray assays were performed to figure out the differentially expressed genes between Flp-In-293 cells stably expressed EPAS1 gene and the control Flp-In-293 cells according to the manufacturer's protocol (Affymetrix, Santa Clara, CA, USA). Briefly, total RNA was used to produce double-stranded cDNA, which was then used to synthesize biotin-labeled antisense RNA (aRNA) by in vitro transcription. aRNA was cleaned, fragmented and hybridized to Affymetrix HG U133 Plus 2.0 GeneChips followed by washing and scanning. Standardized Fold Change method designed in house (Wang Y, et al.) was used for microarray gene expression difference analysis. The gene that displayed the most significant change between the samples was selected for further investigation. qRT-PCR was performed on an Applied Biosystems 7900 HT analyzer using SYBR Green dyes (Takara, Otsu, Shiga, Japan) according to the manufacturer's protocol. First-strand cDNA for qRT-PCR was synthesized using a high-capacity cDNA reverse transcription kit (Applied Biosystems, Foster City, CA, USA) with total RNA as template. Relative quantification was performed by means of the 2−ΔΔCt method, with GAPHD as the endogenous control. qRT-PCR primers used were listed in Supplementary Table S2.

Cell transient transfection and dual-luciferase reporter gene assay

Prior to transfection, 4 × 104 HEK293 cells were split into 24-well plates and incubated in DMEM. Next day, pcDNA3.1-IKZF1 plasmid or pcDNA3.1-Sp1 plasmid or pcDNA3.1-empty plasmid was co-transfected with pGL3-G/deletion/C or pGL3-C/deletion/C plasmid or pGL3-C/insertion/T plasmid at a ratio of 300 ng:180 ng into the HEK293 cells using Lipofectamine 2000 (Invitrogen). pcDNA5-EPAS1 plasmid or pcDNA5-empty plasmid was co-transfected with pGL3-LOX plasmid at a ratio of 240 ng:240 ng into the HEK293 cells to investigate the regulation of EPAS1 on LOX promoter. In addition, 20 ng of pRL-SV40 plasmid was used as an internal control in each transfection. Luciferase activity was measured with Dual-luciferase reporter assay system (Promega). The relative luciferase activity was calculated as the ratio of Firefly luciferase activity vs. Renilla luciferase activity for each sample.

Electrophoretic mobility shift assay (EMSA)

Non-radioactive EMSAs were performed with IKZF1 isoform 3 proteins and biotin-labeled double stranded probes, which were selected on the basis of predicted IKZF1 binding sites in the promoter of the EPAS1 gene (Supplementary Fig. S1). Non-radioactive EMSAs were also performed with Sp1 proteins and biotin-labeled double stranded probes which were made by the 40-bp insertion fragment. EMSAs were performed using a LightShift chemiluminescent electrophoretic mobility shift assay kit (Thermo Scientific) according to the manufacturer's instructions.

Chromatin immunoprecipitation (ChIP) analysis

ChIP experiment was carried out using ChIP Assay Kit (Millipore) according to the manufacturer's instructions. Briefly, Flp-In-293 cells stably expressed EPAS1 were treated with 1 μg/mL tetracycline (Invitrogen) for 24 h and treated with chemical hypoxia-mimetic 2,2′-dipyridyl during the last 16 h before harvest. Cells were then crosslinked with 1% formaldehyde for 10 min at 37°C. Nuclear lysate was harvested and sonicated to obtain 250 bp to 1,000 bp DNA fragments. The chromatin was immunoprecipitated with rabbit polyclonal EPAS1 antibodies (Novus Biologicals, Littleton, CO, USA) or normal rabbit IgG (Santa cruz, Dallas, TX, USA). The precipitated genomic DNA crosslinks were reversed by heating at 65°C overnight with NaCl, recovered by phenol/chloroform extraction and analyzed by quantitative PCR (qPCR). Primers used for ChIP-qPCR were listed in Supplementary Table S2.

DNA extraction and sequencing

Genomic DNA was extracted from umbilical cord blood using AxyPrep Multisource Genomic DNA Minprep kit (Axygen Scientific, Union City, CA, USA) following the procedure detailed in the kit. A 2,088 bp DNA fragment (from −1988 bp to +100 bp relative to the transcription start site) of the human EPAS1 gene (NCBI accession no. NM_001430.4) was amplified by PCR and the variants included this region were determined by direct dye terminator sequencing of the PCR products with the Applied Biosystems Prism BigDye system according to the manufacturer's instructions. The products for sequencing were run on an Applied Biosystems 3730 automated sequencer and analyzed by SeqMan software. Insertion and deletion (indel) variants were further verified by PCR method. Primers used for PCR and sequencing were presented in Supplementary Table S2.

Statistical analysis

Linkage disequilibrium (LD) and haplotype analyses for the variants identified in the promoter of EPAS1 gene were performed in the 43 unrelated individuals. Pair-wise R2 for six variants with MAF larger than 0.25 were calculated and LD structures were plotted by the program haploview48. The Hardy–Weinberg equilibrium for each variant was also tested by haploview program. Population haplotypes for the variants in strong LD (0.8 ≤ R2 < 1) were inferred using PHASE v2.1.1 software49. Potential transcription factor binding sites were predicted using TFSEARCH (version 1.3, threshold score 85.0, classification: vertebrate)50.

The data were presented as mean ± S.E. unless otherwise indicated. The data were tested for normal distribution and homogeneity of variances by Ryan-Joiner test implemented in Minitab software and by Levene's test implemented in SPSS software, respectively. If the data were normal distributed and had similar variances, Student's t test was performed to compare means between measured groups. Otherwise, data transformation was conducted before t test or Mann-Whitney test or t test with Welch's correction was performed using GraphPad Prism software. The correlation between mRNA levels of the interested genes was presented by Pearson coefficient using GraphPad Prism software. P values less than 0.05 were considered significant.