Type 5 phosphodiesterase expression is a critical determinant of the endothelial cell angiogenic phenotype

Abstract

Type 5 phosphodiesterase (PDE5) inhibitors increase endothelial cell cGMP and promote angiogenesis. However, not all endothelial cell phenotypes express PDE5. Indeed, whereas conduit endothelial cells express PDE5, microvascular endothelial cells do not express this enzyme, and they are rapidly angiogenic. These findings bring into question whether PDE5 activity is a critical determinant of the endothelial cell angiogenic potential. To address this question, human full-length PDE5A1 was stably expressed in pulmonary microvascular endothelial cells. hPDE5A1 expression reduced the basal and atrial natriuretic peptide (ANP)-stimulated cGMP concentrations in these cells. hPDE5A1-expressing cells displayed attenuated network formation on Matrigel in vitro and also produced fewer blood vessels in Matrigel plug assays in vivo; the inhibitory actions of hPDE5A1 were reversed using sildenafil. To examine whether endogenous PDE5 activity suppresses endothelial cell angiogenic potential, small interfering RNA (siRNA) constructs were stably expressed in pulmonary artery endothelial cells. siRNA selectively decreased PDE5 expression and increased basal and ANP-stimulated cGMP concentrations in these conduit cells. PDE5 downregulation increased network formation on Matrigel in vitro and increased blood vessel formation in Matrigel plug assays in vivo. Collectively, our results indicate that PDE5 activity is an essential determinant of angiogenesis and suggest that PDE5 downregulation in microvascular endothelium imparts a stable, enhanced angiogenic potential to this cell type.

de novo blood vessel formation is essential for embryonic vascular development and for postnatal vascular homeostasis and wound healing (5, 9, 12). Numerous factors have been identified that promote the angiogenic potential of endothelium. Whereas vascular endothelial cell growth factor (VEGF) is one such factor that is widely accepted to promote angiogenesis (9), vasoactive autocoids, such as nitric oxide (1, 4, 27, 32, 36), also promote angiogenesis. Nitric oxide activates soluble guanylyl cyclase and increases cGMP, and it is this increase in cGMP that ultimately promotes new blood vessel formation (1, 27). Increased cGMP activates protein kinase G, which phosphorylates vasodilator-stimulated phosphoprotein and reorganizes peripheral F-actin (6). Protein kinase G also activates extracellular signal-regulated kinase (ERK) 1/2 and p38 mitogen-activated protein kinase (MAPK) and promotes VEGF secretion (27, 36). Collectively, each of these cGMP effector pathways contribute to an increased angiogenic potential.

Both duration and magnitude of cGMP signals are controlled by phosphodiesterases (PDEs), which hydrolyze cGMP to 5′GMP. Endothelial cells variably express PDE isozymes 1–7 (10, 25, 30, 38, 39). Isoform-selective inhibitors have been used to test the importance of PDE enzymes in angiogenesis (25). The PDE5 inhibitors, sildenafil and tadalafil, promote vascular repair in multiple models of tissue damage and wound repair (20, 27, 35, 36), highlighting the enzyme's important role(s) in angiogenesis.

However, the angiogenic potential of endothelium varies among vascular beds and within vascular segments of an organ's circulation. In particular, microvascular endothelium appears to possess a greater angiogenic potential than does conduit endothelium; pulmonary microvascular endothelial cells (PMVECs) possess higher angiogenic capacity than do pulmonary artery endothelial cells (PAECs) (2, 8). These two cell types also differ in the PDE isoforms they express, as PAECs express PDE5, whereas PMVECs do not express PDE5 and possess only trace levels of cGMP hydrolyzing activity (39). The role that PDE5 expression plays in control of the angiogenic potential of PAECs and PMVECs is not known. The present studies were therefore designed to examine whether PDE5 activity is an essential determinant of the endothelial cell angiogenic potential.

METHODS

Isolation and culture of rat pulmonary endothelial cells.

PAECs and PMVECs were isolated and cultured in DMEM medium supplied with 10% FBS, as described previously (2). A panel of specific cell surface markers was used routinely to characterize the phenotype of cultured cells. Cells used in all experiments were below passage 12.

Retroviral constructs.

The pAcAPDE5 plasmid (Pharmingen) containing full-length human PDE5A1 was kindly provided by Drs. Hengming Ke (Univ. of North Carolina) and Kenji Omori (Tanabe Seiyaku, Japan). This cDNA was fused to the COOH terminus of EGFP, and resulting fusion was inserted into a puromycin resistance-encoding retroviral vector thus generating pMA2508. The hPDE5A1-EGFP pMA2508 construct or EGFP vector control was used to generate amphotropic envelope-pseudotyped retroviral particles using Phoenix-Ampho packaging cell line (kindly provided by Gary Nolan, Stanford Univ.). PMVECs were infected by incubating retroviral supernatants with cells for 24 h in the presence of 5 μl/ml polybrene. Infected cells were selected with 20 μg/ml puromycin for 72 h. Cells with high expressions of EGFP were selected by cell sorting (BD FACSvantage).

Generation of siRNA for PDE5.

Retroviral small interfering RNA (siRNA) expressing pSilencer 5.1 vector (Ambion, Austin, TX) with an H1 promoter and encoding a puromycin resistance cassette was used to establish the stable suppression of PDE5 A1/A2 gene expression in cultured endothelial cells. Briefly, six hairpin structures of 21-bp target sequences, named siRNA-1 through siRNA-6, at the NH₂ terminus of the rat/human PDE5A1/A2 gene, were designed using online template siRNA software and checked by NCBI BLAST. Among them, siRNA-1 was designed as a native control with a specific sequence to the human PDE5A1 gene, but contained two nucleotides mutated from the rat sequence. siRNA-2 was designed to be more accessible to the PDE5A1 gene than to the PDE5A2 gene. The scramble siRNA was designed according to the same method. The retroviral supernatants of siRNA from Phoenix-Ampho packaging cells were incubated with PAECs with 5 μl of polybrene for 24 h. The stably transfected PDE5 siRNA cells were selected to homogeneity in 10 μg/ml puromycin for 72 h before PDE5 activity, mRNA, and protein levels were tested in cell lysates.

Phosphodiesterase activity assay.

The cGMP and cAMP hydrolyzing activities in whole cell lysate were performed as previously described (39). The substrate concentrations for PDE activities in the supernatant were measured with 0.25 μM ³H-cAMP or ³H-cGMP as substrate. The PDE5-selective inhibitor sildenafil was dissolved in DMSO (final concentration at 0.5%) and added to the reaction mixture 5 min before the assay.

cGMP measurements.

The whole cell acid extracts in 0.2 N HCl/50% methanol were prepared as described (39). The amount of cGMP in acetylated samples was measured using enzyme immunoassay kits (Cayman Chemical, Ann Arbor, MI) and expressed as picomoles of cGMP generated per total protein.

Immunoblots and immunoprecipitation.

Whole cell extraction and immunoblotting were performed as described previously (39). Endogenous PDE5 in PAECs was immunoblotted by the PDE5A1-specific antibody raised by the NH₂ terminal peptide (CMERAGPSFGQQRQQQQPQQQ). Antibody to GST-fusion protein of the high affinity cGMP-binding domain (cGB-I) of human PDE5 was prepared to detect the human PDE5A1 expression in PMVECs with the GFP-specific antibody (Bioversion). Antibody to PDE4 was prepared as previously described (10, 39). In Western blot analysis, protein samples and adsorbed proteins on an affinity gel were dissolved in 1× loading buffer before being subjected to 7% precast Novex gels (Invitrogen, Carlsbad, CA) for electrophoresis.

RT-PCR.

RT-PCR was performed with total RNA extracted from PAECs and PMVECs using an RNA isolation kit (Roche, Indianapolis, IN) and cDNA prepared by the First Strand RT-PCR kit (Stratagene, La Jolla, CA) with random primers. Primers for rat PDE5A1 and PDE5A2 were generated, with produced 352- and 328-bp products, respectively (39). Transcripts for PDE4 as a native control for PAECs and PMVECs produced a 287-bp product. Relative quantitative RT-PCR, using 18S as the internal standard, was performed to confirm that knockdown of PDE5 was specific.

In vitro matrigel experiments.

A modified in vitro Matrigel assay was performed. Briefly, 50 μl of regular Matrigel (BD Biosciences, Bedford, MA) was added to prechilled 96-well plates. Matrigel was allowed to polymerize at room temperature for 2 h and maintained in a 37°C incubator overnight. PMVECs expressing PDE5A1 (2 × 10³ cells/well) and PAECs expressing siRNA-3 to inhibit PDE5 (5 × 10⁴ cells/well) were suspended in DMEM and seeded on Matrigel. For quantitative measurement of network formation, whole areas of each well were imaged, and the number of networks, composed of closed and continuous cords, was measured (2, 8). The nuclear positions of angiogenic cells in the networks were imaged by UV fluorescent microscopy after methanol fixation and DAPI staining.

Cell growth curve.

Cells were seeded in 60 mm dishes, in media containing either 10% or 0.5% serum. The cell numbers were counted every 24 h to generate a complete growth curve.

Live cell imaging and time-lapse video microscopy.

Cells were seeded on 96-well plates with Matrigel and incubated at 37°C in 5% CO₂ using a Neuve live cell chamber fitted into an Eclipse Nikon TE 2000-U microscope (Nikon Instruments, Melville, NY). Image acquisition was achieved using a CoolSnap ES monochrome camera and processed with MetaMorph Premier software (Universal Imaging, Downingtown, PA).

In vivo vessel formation assay.

Cells (3.75 × 10⁵) suspended in 0.25 ml of standard culture medium were mixed with 0.5 ml of unpolymerized Matrigel and kept at 4°C. The cold fluid containing a cell/Matrigel mixture was injected subcutaneously into the abdomen (2 plugs each side) of sedated CD40 rats (250–300 g, 75 mg/kg ketamine and 10 mg/kg xylazine ip, Charles River Laboratories). The injected mixture polymerizes at body temperature and becomes a solid plug within ∼2 min. In control experiments, the Matrigel/medium mixture was injected without cells. After 10 days of injection, Matrigel plugs were excised from the abdominal wall of rats (50 mg/kg ip pentobarbital sodium) and immersion fixed in 4% paraformaldehyde. Fixed specimens were dehydrated in ethanol, embedded in paraffin, and cut in 5-μm sections for hematoxylin and eosin staining. Stained sections were examined under light microscopy (Eclipse Nikon E600) to count the total number of tube-like vessels containing red blood cells (de novo formed by angiogenesis) inside the Matrigel plugs. For quantification purposes, the total number of vessels within each section cut in three different portions of the plugs were counted and normalized by the area of each section using QCapture image system (QIMAGING Micropublisher 5.0) and Spot Advance software. All animal studies were approved by the University of South Alabama Institutional Animal Care and Use Committee.

Statistical analysis.

Results were analyzed using two-way ANOVA with post hoc test, where appropriate, using GraphPad Prism 3.1 software. All data represent means ± SE. P < 0.05 is considered statistically significant for the comparisons.

RESULTS

PDE5 regulates endothelial cell cGMP concentrations.

PDE5 is expressed in PAECs, but is not typically expressed in PMVECs, either in vitro or in vivo (39). Therefore, whole cell PDE5 activity was first measured in confluent PAECs and PMVECs. Only trace PDE5 activity was detected in PMVECs, equal to 3% of the activity resolved in PAECs. In contrast, whole cell cAMP-PDE activity was similar in both cell types (refer to Figs. 1 and 2).

Fig. 1.

Overexpression of human type 5 phosphodiesterase (PDE5) A1 gene in pulmonary microvascular endothelial cells (PMVECs) reduces cellular cGMP. A: retroviral construct of human PDE5A1 with EGFP fusion was used to transfect PMVECs. B: Western blot analysis resolved a 125-kDa protein in hPDE5A1/EGFP expressing cells (5A1) using both GFP and human PDE5-specific antibodies (38, 40), which was confirmed by coimmunoprecipitation (IP:GFP and IB: hPDE5). C: the hPDE5A1/EGFP expressing cells possessed a greater than 50-fold increase in cGMP-hydrolytic activity, whereas no change in cAMP-PDE activity was detected (inset). D: the hPDE5A1 enzyme was inhibited by the PDE5 selective inhibitor, sildenafil, IC₅₀ = 20 nM. E: a decrease in basal cGMP was detected in hPDE5A1-expressing cells compared with PMVECs. F: expression of hPDE5A1 in PMVECs attenuated the increase in cGMP after atrial natriuretic peptide (ANP) (10 nM) stimulation compared with previously reported responses in PDE5-deficient PMVECs (39). Pretreatment of hPDE5A1-expressing cells with sildenafil (100 nM for 10 min) increased basal cGMP concentrations and increased the ANP-induced rise in cGMP. *P < 0.05, n = 3 compared with PMVECs or hPDE5A1 controls.

Fig. 2.

PDE5 small interfering RNA (siRNA) decreases rat PDE5 mRNA in pulmonary artery endothelial cells (PAECs). A: 6 siRNA sequences (Si-1 to -6) targeting NH₂-terminal domains specific to rat (r) and/or human (h) PDE5A1 and PDE5A2 genes were designed using a gene-walking approach. The siRNAs were transfected into PAECs using retroviral constructs for establishing stable cell lines. B: 4 rat-specific siRNAs, siRNA-3, -4, -5, and -6, effectively inhibited both rPDE5A1 and rPDE5A2 mRNA in PAECs. C: inhibition of PDE5A1/5A2 mRNA was greatest in siRNA-3-expressing cells and was confirmed by RQ RT-PCR using 18S and PDE4D5 (4D5) as the endogenous controls (10). siRNAs consistently decreased rPDE5A1 protein levels (D) and enzyme activities (E). siRNA-3 inhibited more than 90% of PDE5 activity. Negative controls, scrambled RNA controls, and siRNA-1 specific to human PDE5A1 did not change rPDE5A1 and rPDE5A2 at mRNA levels (B), protein levels (D), or activity (E). None of the siRNA constructs changed expression of endogenous cAMP-PDE isoforms, including PDE4D5 (4D5, B and C) and PDE4A (4A, D), and did not influence cAMP-PDE activity in PMVECs (E). *P < 0.05, n = 3 compared with controls.

We sought to restore PDE5 activity to PMVECs to determine whether the presence of this enzyme was sufficient to decrease basal cGMP concentrations. Full-length human PDE5A1 (hPDE5A1) fused to EGFP (see methods) was inserted into a retroviral delivery system, and PMVECs were infected and selected to homogeneity using puromycin (Fig. 1A). hPDE5A1/EGFP expression was confirmed by coimmunoprecipitation and Western blot analysis (Fig. 1B); hPDE5A1/EGFP immunoprecipitation and GFP immunoblotting revealed the presence of hPDE5A1/EGFP. Similar results were obtained when GFP antibody was used for immunoprecipitation followed by hPDE5A1 immunoblotting. hPDE5A1 expression in PMVECs increased cGMP hydrolyzing activity 50-fold, without changing cAMP-PDE activity (Fig. 1C). The hPDE5A1 activity in PMVECs (40.2 ± 1.1 pmol·min⁻¹·mg⁻¹ protein) was nearly twofold higher than the endogenous PAEC PDE5 activity (23.3 ± 0.9 pmol·min⁻¹·mg⁻¹ protein). PDE5 inhibition using sildenafil dose dependently decreased PDE5A1 activity in PMVECs (Fig. 1D). The basal cGMP concentration in hPDE5A1-expressing cells was only 19% of the constitutive level measured in wild-type PMVECs (Fig. 1E). Both the peak (5 min) and sustained (>30 min) atrial natriuretic peptide (ANP)-induced rise in cGMP was attenuated in hPDE5A1-expressing cells (Fig. 1F). As with basal cGMP, sildenafil pretreatment potentiated the ANP-stimulated cGMP response in these cells. PDE activities and cGMP levels in GFP retroviral vector controls were not different from those observed in wild-type PMVECs (data not shown). Thus, heterologous hPDE5A1 expression increases cGMP-hydrolyzing activity in PMVECs and decreases the basal and ANP-stimulated rise in cGMP.

We next sought to reduce cGMP-hydrolyzing activity in PAECs. Six separate siRNA constructs were generated, targeting discrete regions along the rat PDE5A1 and PDE5A2 mRNAs (Fig. 2A). Of the six constructs tested, siRNA-3 displayed the most prominent inhibitory effect (Fig. 2B). Relative quantitative RT-PCR confirmed that only ∼7% of rPDE5A1 and 5% of rPDE5A2 mRNA was expressed in siRNA-3-treated cells (Fig. 2C). Consistent with the RNA data, Western analysis indicated that siRNA-3 caused the most significant decrease of PDE5A1 (Fig. 2D) and PDE5A2 (data not shown) protein. PDE4 is responsible for the majority of cAMP-hydrolyzing activity in these endothelial cells (10, 30, 38, 39,); none of the siRNAs targeting PDE5A1/A2 altered PDE4 expression.

We confirmed that the siRNA-3 construct specifically inhibited cGMP-PDE activity in PAECs (Fig. 2E). siRNA-3, -4, -5, and -6 decreased cGMP-hydrolyzing activity. siRNA-3-expressing cells displayed only 8% of the PDE5 activity normally observed in PAECs. Basal cGMP was increased approximately twofold in PAECs expressing siRNA-3 (Fig. 3A), and the peak and sustained ANP-induced rise in cGMP was potentiated in these cells (Fig. 3B). Negative controls, including scrambled RNA and human PDE5A1-specific siRNA-1, showed no effect on rPDE5A1 and rPDE5A2 expression in PAECs (refer to Fig. 2, A, B, and D). None of the PDE5 siRNAs inhibited cAMP-hydrolyzing activity (refer to Fig. 2E). Thus, siRNA-3 selectively reduced PDE5 activity and increased cGMP in PAECs compared with other siRNA constructs.

Fig. 3.

Inhibition of PDE5 expression in PAECs increases cGMP. Inhibition of PDE5 expression in PAECs (open bars) by siRNA-3 (closed bars) potentiates cGMP levels under basal (A) and ANP-stimulated (10 nM) conditions (B). *P < 0.05, n = 3 compared with PAECs control.

PDE5 is a determinant of endothelial cell proliferation.

Previous studies have demonstrated that PMVECs have higher rates of DNA synthesis, cell proliferation, and migration than do PAECs (2, 7, 8, 19, 29, 39). Therefore, we examined the impact of PDE5 expression and activity on cell proliferation. PDE5A1 expression in PMVECs did not impact on serum-stimulated cell proliferation (Fig. 4A) but did reduce cell proliferation in low-serum (0.5%) media (Fig. 4B). In contrast, PAECs expressing siRNA-3 exhibited higher proliferation rates than did wild-type PAECs in the presence (Fig. 4C) or relative absence (Fig. 4D) of serum.

Fig. 4.

PDE5 regulates endothelial cell growth. A and C: hPDE5A1-expressing PMVECs and wild-type PMVECs showed similar growth under 10% serum conditions, whereas siRNA-3-expressing cells displayed a greater proliferative response than did PAECs. B and D: under 0.5% serum culture conditions, hPDE5A1-expressing cell showed a slower growth than PMVECs. Under the same low-serum culture, siRNA-3 cells showed a greater increase in cell number than did PAECs. Cells were cultured in 60-mm dishes for cell counting, n = 4. Note the different y-axes scales (A and C, 10% serum; B and D, 0.5% serum). *P < 0.05 compared with controls.

PDE5 regulates network formation on Matrigel in vitro.

PMVECs and PAECs were seeded on Matrigel matrix in vitro, and network formation was initially quantitated over a 72-h period. PMVECs rapidly generated complex networks 6–12 h after seeding (Fig. 5A). Network formation peaked at 24 h and was maintained from 24–72 h in culture. More than 120 networks/well formed from 2 × 10³ PMVECs seeded into 96-well plates (Fig. 5B). PAECs similarly developed networks, although the magnitude of their response was modest compared with PMVECs (Fig. 6A). PAECs generated nearly 15 networks/well when 5 × 10⁴ cells were seeded into 96-well plates (Fig. 6B). The networks formed by PAECs were not stable and disassembled 48 h after seeding. Thus, as in our earlier studies (2, 8), PMVECs possess a stable angiogenic imprint that can be resolved by measuring their enhanced capacity to form networks on Matrigel.

Fig. 5.

Overexpression of hPDE5A1 in PMVECs inhibits network formation on Matrigel and decreases angiogenesis. A: reduced network formation was seen in hPDE5A1-expressing cells compared with PMVEC controls, which was reversed when cells were pretreated with sildenafil (100 nM) for 24 h. Branch-like tubes, or microspikes, were seen in hPDE5A1 cells after 24–48 h in culture, which were also eliminated by sildenafil pretreatment. B: network formation of PMVECs (closed squares) on Matrigel was counted over the time of cell culture. Overexpression of hPDE5A1 delayed network formation and reduced the total number of networks. The decrease in network formation observed in hPDE5A1-expressing cells (open squares) was acutely reversed by sildenafil pretreatment (open circles). 2 × 10³ cells/well were seeded on Matrigel in 96-well plates, n = 4. *P < 0.05 compared with control PMVECs. C: dynamic changes of branch-like tubes, or microspikes, grown on Matrigel were seen in hPDE5A1-expressing cells. Solid arrows indicate those tubes undergoing a growth and elongation response, whereas dashed arrows indicate tubes undergoing regression. *Center of the network.

Fig. 6.

Inhibition of PDE5 expression in PAECs promotes network formation on Matrigel. A: inhibition of PDE5 activity using siRNA-3 increases network formation on Matrigel culture when compared with wild-type PAECs. B: numbers of networks counted on Matrigel indicated more networks in siRNA-3-expressing cells than in PAECs. 5 × 10⁴ cells/well were seeded on Matrigel in 96-well plates, n = 4. C: pretreatment of PAECs with PDE5-specific inhibitor sildenafil (100 nM, 24 h) increased network formation on Matrigel. Inset: PAECs network formation with sildenafil pretreatment after 72 h of culture on Matrigel. *P < 0.05 compared with controls.

We next sought to determine whether PDE5 activity contributes to the angiogenic PMVEC phenotype. To address this issue, hPDE5A1-expressing PMVECs were seeded on Matrigel matrix in vitro, and network formation was compared with wild-type cells (Fig. 5A). The ability to rapidly form networks was greatly impaired in hPDE5A1-expressing cells. Indeed, networks had not begun to form by 12 h, and only rudimentary branching was observed at 24 h. Since network formation was reduced and delayed in these hPDE5A1-expressing cells, we extended the time course through 192 h (Fig. 5, B and C). Not only was there a time delay required for networks to form, but the magnitude of the networks that formed was greatly reduced; hPDE5A1-expressing cells generated a maximum of 60 networks/well. To confirm that this observation was due to PDE5A1 activity, the study was repeated after pretreatment of hPDE5A1-expressing cells with sildenafil (Fig. 5, A and B). Sildenafil rescued network formation to levels reminiscent of wild-type PMVECs. Thus, PDE5 activity reduces cGMP concentrations and limits network formation in PMVECs.

We noted that while the number of networks formed in hPDE5A1-expressing PMVECs was low, the number of branches that formed in these cells was not inhibited. Rather, hPDE5A1 expression appeared to induce de novo branches emanating from established tubes (Fig. 5C, >72 h of culture); this de novo branching pattern was not evident in wild-type PMVECs. We therefore tracked network formation in PMVECs and hPDE5A1-expressing cells 0–96 h after cells were seeded on Matrigel, using time-lapse microscopy, to assess the fate of individual branches that formed in hPDE5A1-expressing cells. Wild-type PMVECs exhibited a highly dynamic connection and alignment within the first 24 h of Matrigel culture (Supplemental Movie S1, 0–24 h; supplemental data for this article is available at the AJP-Lung web site), followed by well-organized network formation (Supplemental Movie S2, 24–96 h). Stable networks formed within 24 h in rapidly proliferating PMVECs. In contrast, slower cell-cell recognition and network formation were observed in hPDE5A-expressing cells (Supplemental Movie S3, 0–24 h). Although complex networks did not form in hPDE5A1-expressing cells, these cells displayed a remarkable number of new branches that failed to recognize neighboring cells necessary to complete network formation (Supplemental Movie S4, 24–96 h). Enlarged images illustrated that the dynamic extended branches, or microspikes, fail to recognize neighboring cells (Supplemental Movies S5–S8). By 96–120 h, branches regressed, and, as a consequence, few mature networks were observed by 192 h (Fig. 5C, branches in hPDE5A1-expressing cells are marked 1 through 6, and Supplemental Movie S9–S10). Thus, PDE5 activity does not inhibit the initial branching required for network formation, but it prevents the migration or cell-cell recognition that is necessary to transition from a branch into a mature network.

We further examined the role of PDE5 in network formation using PAECs expressing siRNA-3. Network formation was examined over a 72-h time course in both wild-type PAECs and siRNA-3-expressing cells (Fig. 6A). Inhibiting PDE5 activity dramatically increased the number of tubes that formed and were sustained over the entire time course (Fig. 6B). Consistent with the findings in siRNA-3-expressing cells, sildenafil treatment of wild-type PAECs increased tube formation (Fig. 6C). Thus, inhibiting PDE5 expression or activity in PAECs increases their angiogenic potential in the Matrigel assay.

PDE5 determines angiogenic potential in vivo.

PMVECs possess a greater angiogenic capacity than do PAECs when mixed with Matrigel and subcutaneously implanted (2, 8). We utilized the in vivo Matrigel model to further test whether PDE5A1 impacts on the angiogenic potential of endothelial cells. Although both wild-type and hPDE5A1-expressing PMVECs developed normal blood vessels that were contiguous with the host circulation, hPDE5A1-expressing PMVECs generated 50% fewer de novo vessels (Fig. 7A). Angiogenesis was also tested in wild-type and siRNA-3-expressing PAECs. Decreased expression of PDE5 in the siRNA-3-expressing cells increased de novo vessel formation (Fig. 7B). These results indicate that PDE5 activity reduces the angiogenic potential of endothelium, suggesting that the absence of PDE5 in PMVECs, both in vitro and in vivo, contributes to the enhanced angiogenic potential of microvascular endothelium.

Fig. 7.

PDE5 regulates vessel formation in Matrigel plug assay in vivo. A: hPDE5A1-expressing cells possessed fewer vessels in Matrigel plugs. B: siRNA-3 cells showed more vessel formation in Matrigel plugs. Each animal was injected with Matrigel-cell mixtures, and plugs were harvested after 10 days for vessel staining assay. N = 4, *P < 0.05 compared with PMVEC or PAEC controls.

DISCUSSION

Both nitric oxide and ANP increase cGMP and promote the angiogenic potential of endothelium. Therefore, increased cGMP has been implicated in promoting angiogenesis (1, 6, 9, 20, 25, 27, 32, 36). Since PDE5 is the principal enzyme responsible for cGMP hydrolysis, controlling both the magnitude and duration of the cGMP signal, it has been incriminated in control of angiogenic potential (25, 27, 35). However, not all endothelial cells express PDE5. Indeed, pulmonary microvascular endothelial cells do not express PDE5 in vivo or in vitro, bringing into question whether this enzyme, and the cGMP signal that it controls, similarly controls angiogenesis in endothelial cells derived from both conduit and microvascular origins. Findings presented herein demonstrate that the absence of PDE5 expression in PMVECs increases basal cGMP and critically contributes to the enhanced angiogenic capacity of this cell type. The expression of recombinant hPDE5A1 in PMVECs reduces basal cGMP and suppresses angiogenesis. Similarly, inhibition of endogenous PDE5 expression in PAECs increases basal cGMP and promotes the angiogenic potential of this cell type. Therefore, PDE5 is an important determinant of angiogenic potential due to the control of basal cGMP in endothelium.

The mechanisms by which increased basal cGMP responses accelerate angiogenesis are less clear, although activation of protein kinase G is essential to neovascularization. Angiogenic responses are attenuated in protein kinase G I knockout mice (34) and promoted in protein kinase G I overexpressing mice (34); both PAECs and PMVECs express protein kinase G I (data not shown). Several downstream signaling networks are recruited by protein kinase G activation, including ERK1/2 and p38 MAPK (27). The angiogenesis response requires that endothelial cells break down associated basement membrane proteins, migrate, proliferate, form new lumens, and stabilize the emerging vascular structure. cGMP-PKG-ERK1/2 activation has been shown to promote migration, whereas cGMP-PKG-p38 MAPK activation promotes proliferation (27). It is therefore likely that protein kinase G activates parallel downstream signals to coordinate neovascularization.

PDE5 expression influenced endothelial cell growth in our present studies. In PAECs, inhibiting PDE5 increased serum-stimulated growth, whereas in PMVECs, hPDE5A1 expression did not decrease serum-stimulated growth, but rather inhibited proliferation under low-serum conditions. Increased proliferation in siRNA-3-expressing PAECs likely contributed to the potentiated angiogenic response in these cells. Reduced proliferation may also have contributed to the decrease in angiogenic capacity of hPDE5A1-expressing PMVECs seeded in Matrigel and implanted in vivo, as serum is restricted under these experimental conditions, until developing vessels fuse with the host circulation.

hPDE5A1-expressing PMVECs displayed evidence for reduced migration, or improper cell-cell recognition, in in vitro Matrigel studies. Similar decrements in cell-cell recognition were not observed in wild-type PAECs, likely reflecting the distinct signaling properties of PAECs and PMVECs. When densely seeded in Matrigel in vitro, PMVECs formed networks within 6–12 h. These networks were stable for more than 100 h. However, hPDE5A1-expressing PMVECs formed only 30% of the networks generated by wild-type cells, and peak network formation in the overexpressing cells was not observed until 72 h after seeding. hPDE5A1-expressing cells possessed numerous small branches, or microspikes, that reflected cell projections incapable of successfully recognizing adjacent cells and completing network formation. Indeed, time-lapse videos revealed that whereas cells extended projections toward adjacent cells, the projections retracted back to the cell body. Therefore, the inability to efficiently recognize adjacent endothelium in nascent vessel formation is likely to have reduced the angiogenic potential of hPDE5A1-expressing cells in in vivo Matrigel studies, as it did in the in vitro Matrigel experiments.

The molecular basis by which cGMP regulates network formation is presently unknown. mRNA expression levels between wild-type PMVECs and hPDE5A1-expressing cells have recently been compared. Two gene groups, including growth factors such as fibroblast growth factor (FGF) 21 and chemokines/cytokines like CXC-L1, were downregulated in PDE5A1-expressing cells (B. Zhu, unpublished observations). Future studies will be necessary to resolve whether decreased expression of FGF21 and CXC-L1 contribute to the angiogenic phenotype of endothelial cells (14, 17, 18, 22, 24, 26, 28, 37).

The collection of our present studies indicates that PDE5 activity suppresses the angiogenic capacity of endothelium and demonstrates that rapidly angiogenic cells, such as PMVECs, lack PDE5 expression. Whereas postnatal angiogenesis is essential for maintaining normal vascular homeostasis, and for vascular repair following injury, “disordered” angiogenesis has been described in patients with pulmonary hypertension (21, 31). In this case, endothelial cells, or cells sharing phenotypic features of endothelium, lose the “law of the monolayer.” These cells invade the vessel lumen, and ultimately occlude it, forming a vasoocclusive plexiform lesion. Cells within the lumen express PDE5 (23, 33). It is interesting to note that sildenafil is an approved therapy for patients with pulmonary hypertension, at a dosing regimen that yields a drug concentration consistent with those used in our present study (3, 11, 13, 15, 16). While sildenafil's intended therapeutic target is PDE5 in pulmonary artery smooth muscle cells, with a goal of causing vasodilation and reducing pulmonary artery pressure, the impact of prolonged sildenafil therapy on endothelial cell function is uncertain. Therefore, our present findings demonstrate a central role for PDE5 in regulation of the angiogenic potential of normal endothelium. Future studies will be required to determine how PDE5 activity influences vascular function in disease, such as pulmonary hypertension.

GRANTS

This work was supported by National Institutes of Health Grants HL-60024 and HL-66299 (T. Stevens) and RR-023961 (M. Alexeyev), an American Heart Association Beginning Grant-in-Aid 0665169 (B. Zhu), and an American Heart Association Scientist Development Grant 0835134N (D. F. Alvarez).