Chordin-like 1, a bone morphogenetic protein-4 antagonist, is upregulated by hypoxia in human retinal pericytes and plays a role in regulating angiogenesis

Cell culture and hypoxia

Primary cultures of human retinal pericytes (hRPC) were obtained from Cambrex (Nottingham, UK) and were cultured in MCDB131 medium (Invitrogen, Paisley, Scotland) supplemented with 2 mM L-glutamine 50 U/ml penicillin, 50 μg/ml streptomycin, and 10% (v/v) fetal bovine serum. For experiments, cells were used at passage four or less and maintained in medium with 10% serum. HeLa (LGC promochem, Teddington, UK) cells were cultured in minimum essential medium (Sigma, Dublin, Ireland) supplemented with 10% (v/v) fetal bovine serum, 2 mM L-glutamine, 50 U/ml penicillin, 50 μg/ml streptomycin, and non-essential amino acids (Sigma). Cos7 (LGC Promochem) cells were cultured in DMEM (Cambrex) supplemented with 10% (v/v) fetal bovine serum, 2 mM L-glutamine, 50 U/ml penicillin, and 50 μg/ml streptomycin.

For hypoxia experiments, cells were placed in an hypoxia chamber (Coy Laboratories, Grass Lake, MI) allowing the establishment of humidified, ambient, atmospheric hypoxia of 1% O₂ with 5% CO₂ and a balance of N₂. Temperature was maintained at 37 °C. Extracellular pO₂ measurements were made by using fluorescence quenching oxymetry (Oxylite-2000; Oxford Optronix, Oxford, UK).

RNA extraction and cDNA synthesis

RNA was extracted from hRPC using RNeasy kit (Qiagen, Crawley, UK) according to the manufacturer’s instructions. RT–PCR was performed as follows: 2 μg of total RNA was treated with DNaseI (Invitrogen), according to the manufacturer’s instructions, to remove chromosomal DNA. Reverse transcription was performed using random primer (Invitrogen) and Superscript II (Invitrogen) using the manufacturer’s protocol.

Limited cycle PCR was performed using the following primers, designed using the Primer3 software described in (Table 1) [11].

Real-time (Taqman) polymerase chain reaction

Real-time (Taqman) polymerase chain reaction transcript levels were determined by real-time PCR using a PerkinElmer 7700 analyzer (Perkin Elmer Life Sciences, Shelton, CT). Probe and primer sequences were described in Table 1. Probes were labeled with 5′-FAM and 3′-TAMRA as quencher with the exception of the ribosomal probe, which was labeled with 5′-VIC to facilitate multiplexing. All results were normalized to 18S rRNA (pre-developed assay reagent, PerkinElmer Life Sciences).

Gene microarray analysis

RNA isolation was performed as described in the previous section. In vitro transcription and microarray analysis were performed as follows: 5 μg of total RNA were annealed to 100 pmol of the T7-(dT)24 primer. Then first-strand cDNA synthesis was completed by incubating the mixture with 1 μl of Superscript II RT (200 units) at 42 °C for 1 h. Second-strand synthesis was performed in a total volume of 150 μl. Following purification, the cDNA was extracted with phenol chloroform, precipitated with ammonium acetate, and resuspended in RNase-free water. Synthesis of cRNA was performed by in vitro transcription. The amplified cRNA was purified with an affinity resin column. Biotin-labeled cRNA prepared from template cDNAs was fragmented and hybridized to the HG U133A array (Affymetrix, High Wycombe, UK). Arrays were then washed and fluorescently labeled before being scanned using a confocal scanner. Microarray Suite 5.1 software (Affymetrix) was used to analyze the relative abundance of each gene from the average difference of fluorescent intensities. Calculated signal/log ratios for each gene at all time points of hypoxia were compared to normoxia, and those genes with a value greater than 0.6 or less than –0.6 (1.5 fold difference) in all three biologic replicate experiments were examined further.

Cloning of the chordin-like 1 promoter and open reading frame

A 1537 bp fragment of the CHL-1 promoter was amplified using the Advantage2 PCR enzyme system (BD Biosciences, Oxford, UK) from a human genomic library EMBL3 SP6/T7 (HL 1067 J, BD Biosciences) with the following primers, sense 5′-GGA GAT AAC AAC CAG AGA GTA GTG G-3′ and, antisense 5′-GAA AAG GAG GTG AGG GAA GC-3′. The PCR product was immediately cloned into the TA cloning vector pCRII-TOPO (Invitrogen). The plasmid was restricted using the enzymes Kpn1 (NEB) and Xho1 (NEB), the resulting promoter fragment was subcloned into the reporter vector pGL3 Basic (Promega Southampton, UK).

The open reading frame of CHL-1 was amplified using the Advantage2 PCR enzyme system (BD Biosciences) using human retinal pericyte cDNA as a template. A sense primer with a BamH1 site, 5′-GCT CGG ATC CGC CTG TCA CCT TTT GCA GTG GTC-3′, and an antisense primer with an EcoR1 site, 5′-TGC AGA ATT CGC TTA CAG TGG CCC TTT TCA GAT C-3′, were used to generate CHL-1 expression from a parent vector, pcDNA6/V5-His C (Invitrogen), designated pcDNA6 CHL-1/V5-His.

Overexpression of hypoxia inducible factor-1α

Wild-type (WT) and double mutant (DM) constructs expressing HIF-1α were a gift from Thilo Hagen (University College London, London, UK). Prolines 564 and 402 were mutated to alanine, as these are the residues important for targeted degradation in normoxia [12]. Pericytes were transfected, according to manufacturer’s protocol, with 4 µg of each construct using 12 µl Fugene 6 (Roche, Dublin, Ireland). After 24 h following transfection, the medium was changed and the cells were incubated for a further 24 h. RNA was then isolated and analyzed by RT–PCR, or nuclear protein extracts were examined by western blot analysis to detect HIF-1α protein.

Luciferase assays

HeLa cells were transfected with plasmids pGL3-CHL-1 or pGL3-HRE [13] along with DM-HIF-1α, using 6 µl Fugene 6 (Roche). Cells were transfected for 24 h, medium replaced with fresh and incubated for a further 24 h. Cells transfected with DM-HIF-1α were incubated in normoxia, and cells transfected without DM-HIF-1α were incubated in hypoxia. Luciferase gene reporter assay was performed according to the manufacturer’s instructions (Promega).

Analysis of chordin-like 1 and bone morphogenetic protein −4 complex formation

Cos7 cells were transfected as described previously with either empty vector, pcDNA/V5-HisC, or with the CHL-1 expression vector, pcDNA6 CHL-1/V5-His. Cells were lysed, and whole cell extracts were made. Next, 50 µl of extract was incubated with 250 ng rhBMP-4 (R&D Systems, Abingdon, UK) for 1 h at 4 °C before incubation with 100 µl Ni-NTA magnetic beads, overnight at 4 °C with constant rocking. Ni-NTA magnetic beads had been previously washed with binding buffer, pH 8.0, that contained 50 mM NaH₂PO₄, 300 mM NaCl, 10 mM imidazole, and 0.05% Tween 20. Complexes were washed in wash buffer, pH 8.0, that contained 50 mM NaH₂PO₄, 300 mM NaCl, 20 mM imidazole, and 0.05% Tween 20, and eluted off the magnetic beads with elution buffer, pH 8.0, that contained 50 mM NaH₂PO₄, 300 mM NaCl, 250 mM imidazole, and 0.05% Tween 20. Samples were electrophoresed on SDS–PAGE gels, transferred to PVDF membranes, and probed for either V5-tagged CHL-1 or BMP-4.

Western blot analysis

Western blot analysis was performed by standard procedures [14]. HIF-1α protein levels were detected using a 1:500 dilution of either anti-HIF-1α (Biomol) or anti-HIF-1α antibody (AbCam, Canbridge, UK). V5-tagged CHL-1 was detected using a 1:5000 dilution of the anti-V5 antibody (Invitrogen). BMP-2 and BMP-4 was detected using 1:500 dilutions of the anti-BMP-2 and anti-BMP-4 antibodies (R&D Systems). CHL-1 was detected using a 1:1000 dilution of anti-CHL-1 antibody (R&D Systems).

Angiogenesis assay

Human umbilical vascular endothelial cells and human diploid fibroblasts were obtained (day 1) as cocultures in 24 well plates (AngioKit, TCS CellWorks Ltd., Buckingham, UK). Medium, with treatments or vehicle, was replenished on days 1, 4, 7, and 9. Treatments were recombinant human BMP-2, BMP-4, and CHL-1 (R&D Systems). In addition, 2 ng/ml VEGF and 20 μM Suramin (TCS Cellworks) were used as positive and negative controls respectively. Tubule formation was examined at day 11. Cells were fixed with ice-cold 70% ethanol and tubules were quantitated and visualized following fixing and staining for CD31 platelet-endothelial cell adhesion molecule 1 (PECAM-1) using a combined ELISA and histology kit (TCS CellWorks Ltd., UK) according to the manufacturer's instructions. Angiogenesis was quantitated by using anti-CD31 antibody coupled to a soluble substrate, ρ-nitrophenol phosphate, which permits quantitation by an optical density measurement. Plates were subsequently washed and stained using an insoluble substrate, 5-Bromo-4-chloro-3-indolyl phosphate (BCIP)/Nitroblue tetrazolium (NBT), for tubule visualization.

Statistical analysis

Each result was representative of at least three independent experiments. All values are represented as the mean ±standard error of measurement. The Student t-test was used to determine statistical differences. Significance of results was indicated when p<0.05.

Upregulation of chordin-like 1 in human retinal pericytes in response to hypoxia

The transcriptomic response of human retinal pericytes to hypoxia was examined. The mean signal–log ratio of three independent experiments is shown (Table 2). Using oxygen quenching oxymetry, we determined extracellular oxygen tensions at the level of the monolayer. Cells cultured in normoxia had a mean pO₂ measurement of 140±2.79 mmHg, cells cultured in hypoxia had a mean pO₂ measurement of 2.12±0.94 mmHg. The ambient pO₂ in the hypoxia chamber was 6.21±0.08 mmHg (data not shown).

CHL-1 (upregulated gene number 20) is a novel gene not previously associated with diabetic retinopathy [9,15,16]. CHL-1 was upregulated in pericytes by hypoxia with a mean SLR of 0.9, corresponding to a 1.9 fold increase in expression (p<0.001). Two genes previously known to be associated with diabetic retinopathy were also upregulated: VEGF (upregulated gene number 7) is implicated in diabetic retinopathy, contributing to endothelial proliferation, permeability, and angiogenesis [2], and is known to be upregulated in pericytes in response to hypoxia [17]. VEGF was upregulated in pericytes by hypoxia with a mean SLR of 1.26, corresponding to a 2.4 fold increase in expression (p<0.001), and Cox2 (upregulated gene number 2) is associated with neovascularisation in models of retinopathy [18,19]. Cox2 was upregulated in pericytes by hypoxia with a mean SLR of 1.43, corresponding to a 2.7 fold increase in expression (p<0.05). To examine the upregulation of CHL-1 along with genes known to be upregulated in diabetic retinopathy, we performed real-time quantitative PCR. As determined by quantitative PCR over a time course (0, 6, 24 and 48 h) of exposure to hypoxia (1% O₂), CHL-1 mRNA was upregulated 1.7 fold by 48 h (p<0.05; Figure 1A), and VEGF and Cox 2 mRNAs were upregulated twofold by 48 h (p<0.05; Figure 1B,C).

Chordin-like 1 expression is directly transactivated by hypoxia inducible factor-1α

The upstream region of CHL-1 was identified using the Human Genome Browser Gateway at the UCSC interface. A 1537 bp fragment of the CHL-1 promoter was cloned, sequenced and inserted into the reporter vector pGL3 Basic (Promega), and named pGL3 CHL-1. Bioinformatic analysis of transcription factors binding to this region of the promoter was performed using MatInspector^TM software, and a putative HIF-1α binding site was located in the proximal promoter (data not shown).

Human retinal pericytes were exposed to 1% O₂ for 0, 6, 24, and 48 h. Nuclear and cytosolic extracts were prepared and used for western blotting. HIF-1α protein expression was dramatically increased by 6 h of hypoxia and remained elevated by 48 h (Figure 2A).

To determine if the CHL-1 promoter with the putative HIF-1α binding sequences was truly HIF-1α responsive, we transfected WT and DM constructs expressing HIF-1α into retinal pericytes. Semiquantitative RT–PCR was performed on RNA isolated from these cells for genes known to be HIF-1α responsive, VEGF and Cox2, and for the novel hypoxia regulated gene CHL-1. In addition, 18S rRNA PCR was performed as a loading control (Figure 2B). Expression of the transiently transfected HIF-1α is also shown in Figure 2. The level of CHL-1 mRNA expression in these transiently transfected cells was quantitated using real-time PCR (Figure 2C). Overexpression of HIF-1α was confirmed by western blotting (Figure 2B).

To further determine the response of the CHL-1 promoter to HIF-1α, we transfected HeLa cells with pGL3 CHL-1 and either stimulated with 1% O₂ or transfected with pGL3 CHL-1 and DM-HIF-1α. As a control, the experiment was also performed using a promoter reporter construct containing four hypoxia responsive elements driving the luciferase gene, HRE-Luc, instead of pGL3 CHL-1. The CHL-1 promoter was driven greater than threefold by HIF-1α (p<0.05) and greater than twofold by hypoxia (p<0.05). The control promoter reporter construct, which contains four hypoxia responsive elements (HREs) was driven greater than fourfold by HIF-1α and greater than sixfold by hypoxia (both p<0.001; Figure 2D).

Chordin-like 1 expressed in human retinal pericytes is secreted and binds to bone morphogenetic protein-4

CHL-1 mRNA was expressed in response to hypoxia in retinal pericytes (Figure 1C). The protein sequence for CHL-1 contained a signal peptide. Conditioned media (20 μl) from retinal pericytes exposed to hypoxia was examined by western blotting for secretion of CHL-1 using an anti-CHL-1 antibody (R&D Systems). Secreted CHL-1 was detectable in the conditioned media and was upregulated by hypoxia, correlating with the mRNA data (Figure 3A). CHL-1 is a BMP antagonist and previously shown to bind BMP-2 [20] and BMP-4 [9]. We investigated the expression of both these BMPs in retinal pericytes in response to normoxia and hypoxia. Both were expressed at the mRNA level (Figure 3B) and were secreted into the culture media as detected by western blotting (Figure 3C). Exposure to hypoxia had no significant expression on either BMP-2 or BMP-4 expression.

The open reading frame of CHL-1 was cloned and sequenced, and inserted inframe with the COOH-terminal V5 6xHis tag in the expression vector pcDNA6/V5His, and named pcDNA6 CHL-1/V5His. Cos7 cells were transfected with empty pcDNA6V5His (E), or with the expression vector pcDNA6 CHL-1/V5His (CHL-1) expressing V5His tagged CHL-1. Whole cell extracts were examined by western blotting for V5 tagged CHL-1 expression using an anti-V5 antibody (Invitrogen; Figure 3D) Whole cell extracts from Cos7 cells transfected with empty pcDNA6/V5His (E), or with the expression vector pcDNA6 CHL-1/V5His (CHL-1) expressing V5His tagged CHL-1, were incubated with 250 ng rhBMP-4 (R&D Systems) and 100 ml NiNTA magnetic beads at 4 °C overnight with rotation. The interaction of the two proteins was demonstrated by pulling down any complexes formed with the tagged CHL-1 using NiNTA magnetic beads (Qiagen) which bound the 6xHis tag on the expressed CHL-1. Proteins bound in a complex to CHl-1 were washed with wash buffer and then eluted using elution buffer. Western blotting was used to identify eluted proteins. CHl-1 was detected using anti-V5 antibody (Invitrogen), and BMP-4, using an anti-BMP-4 antibody (R&D Systems). The complexes bound to the Ni-NTA magnetic beads contained both CHL-1 and BMP-4 (Figure 3E).

Bone morphogenetic protein-4 inhibits angiogenesis and is antagonized by chordin-like 1

BMPs have been previously reported to be negative growth regulators in the adult retinal pigmented epithelium [21], and, more specifically, BMP-4 has been demonstrated to mediate apoptosis of capillary endothelial cells [22]. We wanted to determine the angiogenic effects of BMPs and CHL-1 using an in vitro angiogenesis assay (AngioKit, TCS Cellworks). Using 2 ng/ml VEGF and 20 μM Suramin as positive and negative controls, respectively, we successfully stimulated (p<0.001) and inhibited (p<0.001) angiogenesis (Figure 4A,D). While 10 ng/ml BMP-2 appeared to negatively regulate angiogenesis, this effect was not significant; however, 10 ng/ml BMP-4 significantly inhibited angiogenesis (p=0.0015; Figure 5B,D). CHL-1 alone had no effect on angiogenesis; however, it reversed the inhibitory effect of BMP-4 (Figure 4C,D).

Vascular endothelial growth factor and bone morphogenetic protein-4 co-regulate angiogenesis

To examine whether the pro- and anti-angiogenic effects of VEGF and BMP-4 were mutually exclusive, we used these growth factors alone and in combination in an angiogenesis assay. We found 2 ng/ml VEGF stimulated angiogenesis and 10 ng/ml BMP-4 inhibited angiogenesis. In combination the two growth factors have no effect on angiogenesis (Figure 5).