Homocysteine inhibits neoangiogenesis in mice through blockade of annexin A2–dependent fibrinolysis

When plasma levels of homocysteine (HC), a thiol amino acid formed upon methionine demethylation, exceed 12 μM, individuals are at increased risk of developing large vessel atherothrombosis and small vessel dysfunction. The annexin A2 complex (termed “A2”) is the cell surface coreceptor for plasminogen and TPA and accelerates the catalytic activation of plasmin, the major fibrinolytic agent in mammals. We previously showed that HC prevents A2-mediated, TPA-dependent activation of plasminogen in vitro by disulfide derivatization of the “tail” domain of A2. We also demonstrated that fibrinolysis and angiogenesis are severely impaired in A2-deficient mice. We now report here that, although hyperhomocysteinemic mice had a normal coagulation profile and normal platelet function, fibrin accumulated in their tissues due to reduced perivascular fibrinolytic activity and angiogenesis was impaired. A2 isolated from hyperhomocysteinemic mice failed to fully support TPA-dependent plasmin activation. However, infusion of hyperhomocysteinemic mice with fresh recombinant A2, which localized to neoangiogenic endothelial cells, resulted in normalization of angiogenesis and disappearance of peri- and intravascular fibrin. We therefore conclude that hyperhomocysteinemia impairs postnatal angiogenesis by derivatizing A2, preventing perivascular fibrinolysis, and inhibiting directed endothelial cell migration. These findings provide a mechanistic explanation for microvascular dysfunction and macrovascular occlusion in individuals with hyperhomocysteinemia.

Endothelial cells express plasminogen activators, their receptors, and their inhibitors, all of which are key elements in the fibrinolytic surveillance mechanism of blood vessels (1). Together with an array of endothelial cell surface natural anticoagulants, this system is central to the maintenance of blood fluidity. While TPA is secreted by resting microvascular endothelial cells in many vascular beds, urokinase is produced by perturbed endothelial cells in settings such as wound repair and inflammation (2, 3). Cultured endothelial cells specifically bind both TPA and its substrate, plasminogen (4–6), via surface expression of the calcium-dependent, phospholipid-binding protein annexin A2 (herein referred to as “A2”) (7). A2 is a member of a large, multi-phylum superfamily of calcium-dependent, phospholipid-binding proteins (8, 9). Upon interaction with plasminogen and TPA, purified A2 confers a 60-fold increase in the efficiency of TPA-dependent plasmin generation (10). Complete deficiency of A2 in mice is associated with microvascular fibrin deposition and impaired clearance of injury-induced arterial thrombi (11). In addition, Anxa2^–/– mice show striking defects in postnatal angiogenesis, suggesting a mechanistic association between A2-based endothelial cell fibrinolytic activity and angiogenic potential.

Homocysteine (HC) is a thiol-containing amino acid intermediate that is generated during the conversion of methionine to cysteine (12). Due to the increased length of its R-group carbon chain compared to that of cysteine, the HC thiolate displays increased nucleophilic reactivity compared with that of the cysteine thiolate (13). HC circulates in normal human plasma at concentrations of 5 to 12 μM (14–16). Hyperhomocysteinemia is associated with mutations in genes encoding cystathionine-β synthase, methionine synthase, and 5,10-methylenetetrahydrofolate reductase. It also occurs in association with environmental factors such as gender, age, and diet, in which deficiency of folate, vitamin B₆, and vitamin B₁₂ appear to be primary determinants (15, 17, 18). Previous studies have linked mild to moderate hyperhomocysteinemia with macrovascular occlusion related to venous thromboembolism as well as coronary, cerebrovascular, and peripheral atherosclerosis (14). More recent studies have revealed an association between hyperhomocysteinemia and microvasculopathy, particularly cerebral small vessel disease (19). Despite these associations, the molecular mechanisms underlying HC-associated vascular disorders have not been clearly defined (20).

We previously demonstrated that the binding of TPA to A2 requires residues 7–12 (LCKLSL) of the A2 aminoterminal “tail” domain (21). Furthermore, we showed that the TPA-A2 interaction on the surface of endothelial cells is inhibited by pre-incubation with HC (22). In this reaction, it appears that uptake of HC and its condensation with the free sulfhydryl of cysteine 8 of A2 may require several hours to reach levels that are functionally significant. This modification of A2 decreases its ability to support plasminogen activation (21–23). From these studies, we concluded that HC might perturb the intrinsic fibrinolytic potential of the endothelial cell by selectively blocking TPA binding to A2.

Angiogenesis is the process by which new blood vessels arise from pre-existing vasculature. It entails a complex series of steps whereby activated endothelial cells dissociate from their underlying matrix, migrate toward a chemotactic stimulus, and proliferate (24, 25). Endothelial cells utilize MMPs to migrate through matrix barriers (26). In the mouse, MMP-3 and MMP-13 are directly activated by the serine protease, plasmin, and MMP-3 can subsequently activate MMP-9 (27). Thus, activation of the fibrinolytic system may initiate an extended, matrix-remodeling proteolytic cascade (28).

In the present study, we examine the effects of HC on postnatal angiogenic responses in vivo. In the setting of diet-induced hyperhomocysteinemia, mice developed microvascular fibrin accumulation, reduced perivascular fibrinolysis, and increased susceptibility to arterial thrombosis. A2 isolated from hyperhomocysteinemic mice under non-reducing conditions, furthermore, failed to bind TPA and did not support plasminogen activation. At the same time, the angiogenic response was reduced, and VEGF-directed migration of endothelial cells pretreated with HC was blunted. Importantly, angiogenic failure in the corneal pocket assay was accompanied by intra- and perivascular fibrin accumulation. Infusion of recombinant A2 associated specifically with neovascular endothelial cells restored the angiogenic response and inhibited fibrin accumulation. Interestingly, diet-induced hyperhomocysteinemia did not depress corneal angiogenesis beyond that observed in Anxa2^–/– mice. This suggested that the antiangiogenic effect of HC is exerted via a pathway that is dependent upon A2. Impairment of fibrinolysis and angiogenesis in the hyperhomocysteinemic mouse may be an etiologic factor underlying the role of HC as a risk factor for thrombosis, atherosclerosis, and small vessel disease.

Nutritional effects of a high-methionine diet. In mice placed on a high-methionine (Met) diet, plasma HC levels were, on average, 4-fold higher than those observed in mice on the control high-glycine (Gly) diet and 6-fold higher than those in mice fed standard chow (69. 5 ± 5.9 vs 17.3 ± 1.4 vs 11.3 ± 1.0 μM, respectively, mean ± SEM, n = 11, P = 0.002 for Gly vs. Met vs. chow) (Table 1). The 1.5-fold increase in HC levels in mice on the Gly diet versus those on standard chow diet was most likely due to its lower folic acid content (1.98 vs. 3.0 mg/kg chow). Mice on the Met diet, in addition, exhibited an 18% reduction in body mass compared with those on the Gly diet, even though food consumption was equivalent.

Table 1

HC levels, body weight, and hematologic data

Hematologic parameters in hyperhomocysteinemic mice. In mice on the Gly diet, hematocrits were slightly higher than those in mice on either the chow or Met diets, possibly reflecting mild dehydration. At the same time, however, we found no significant difference in leukocyte and platelet counts or plasma fibrinogen levels among the 3 groups (Table 1). Platelet reactivity in platelet-rich plasma in response to either 2.5 or 10 μM ADP or 1 or 5 μg/ml collagen was also equivalent among the 3 groups. In addition, D-dimer and plasma clotting time, a reflection of the rate of thrombin generation, were not significantly altered. Finally, TPA-dependent plasma clot lysis time and circulating plasmin-antiplasmin complex concentrations were equivalent among the 3 experimental groups, indicating that plasma components associated with fibrinolytic function were unaltered.

Hyperhomocysteinemia and vascular fibrinolysis. Tissues from mice receiving chow, Gly, or Met diets were examined next. Immunohistologic analysis of sections of renal tissue revealed only minimal fibrin associated with glomeruli from mice fed normal chow or Gly diets but abundant fibrin associated with glomeruli from mice fed a Met diet (Figure 1, A–C). Sections of heart and lung tissue showed a similar pattern, with significant fibrin staining detected only in samples from Met diet–fed mice (Figure 1, D–I). In addition, casein overlay of frozen lung sections from mice fed the Met diet revealed perivascular zones of lysis that were significantly reduced in area (Figure 1L; 3,480 ± 360 μm²; mean ± SEM, n = 10) compared with those in overlays of tissue from mice fed either a chow diet (Figure 1J; 5,440 ± 600 μm²; mean ± SEM, n = 10, P < 0.001) or Gly diet (Figure 1K; 7160 ± 480 μm²; mean ± SEM, n = 10, P < 0.001). These data suggested a deficiency of vascular fibrinolysis in mice on the Met diet.

Figure 1

Fibrin accumulation and impaired perivascular fibrinolysis in hyperhomocysteinemic mice. (A–I) For fibrin immunofluorescence, representative sections from renal (A–C), heart (D–F), and lung (G–I) tissue from mice on chow (A, D, and G), Gly (B, E, and H), or Met (C, F, and I) diets were stained with either rabbit anti-human fibrin(ogen) or nonimmune rabbit IgG (not shown), followed by secondary CY3 (red) and DAPI (blue). Tissue autofluorescence is green. Original magnification, ×400. Scale bars: 100 μm. (J–L) For casein zymography, 15-μm sections of lung tissue from mice on chow (J), Gly (K), or Met (L) diets were overlaid with a plasminogen-containing casein gel and incubated (21°C, 90 minutes). Zones of lysis were photographed at 90 minutes under darkfield microscopy. Scale bars: 100 μm. (M) Perivascular caseinolysis surrounding bronchial microvessels was quantified in photographic images. Lysis zones were calculated for at least 10 vessels per photographic field within representative samples. Values shown are mean areas ± SEM, n = 10. *P < 0.001 versus both Gly and chow diets.

To further explore fibrinolytic function in the 3 groups of mice, we studied injury-induced thrombosis in the carotid artery. We used the FeCl₃ vascular injury model, which has been used extensively to define thrombotic tendencies in mice with fibrinolytic deficiencies (29–32). After establishing the initial rate of blood flow, recording was stopped while FeCl₃ was applied for 105 seconds to the adventitial surface of the carotid artery. Post-injury blood flow was then recorded over the next 30 minutes. In all 3 groups, initial blood flow was equivalent (1.08 ± 0.05, 1.11 ± 0.14, and 1.02 ± 0.11 ml/min; mean ± SEM, n = 7, for chow, Gly, and Met diets, respectively). Following vascular injury, however, complete arterial occlusion occurred in 6 of 7 mice on the Met diet but in only 1 of 7 mice on either the chow or Gly diet (Figure 2A). In addition, mean post-injury blood flow in mice on the chow and Gly diets (106% ± 8% vs. 88% ± 11% of baseline, respectively; mean ± SEM, n = 7) did not differ significantly from the initial rate, while post-injury blood flow in mice on the Met diet was markedly reduced (50% ± 11%; mean ± SEM, n = 7, P < 0.001; Figure 2C). These results suggest that the hyperhomocysteinemic mouse exhibits a prothrombotic defect.

Figure 2

Carotid artery thrombolysis. Carotid arteries were injured by application of FeCl₃ to the adventitial surface, and blood flow was monitored before and after injury using a Doppler flow probe. (A) Representative carotid blood flow tracings for mice on chow, Gly, and Met diets before and after arterial injury. Values for mean initial and post-injury flow are shown for representative mice. (B) Mean initial blood flow rates for mice on chow, Gly, and Met diets (mean ± SEM, n = 7). (C) Mean post-injury blood flow, calculated as the area under the flow curve, for mice on chow, Gly, and Met diets (mean ± SEM, n = 7). *P < 0.001 versus chow or Gly diet.

Hyperhomocysteinemia inhibits A2-dependent fibrinolysis. We next examined A2-associated fibrinolytic activity more directly. Immunoblot analysis of reduced gels containing membrane proteins extracted from lung tissue of mice on chow, Gly, or Met diets revealed that the total concentration of A2 did not differ among the 3 samples (Figure 3A). There was, however, a significant reduction in extractable TPA in tissues from the Met mice versus those from mice fed a chow or Gly diet. When TPA-binding capacity of A2 was tested by ligand blot analysis, we observed equivalent binding of TPA to a 36-kDa protein that comigrated with authentic A2 in samples from mice on chow, Gly, and Met diets. When the same experiment was conducted under non-reducing conditions, however, binding of TPA to the 36-kDa protein was evident in samples from mice fed chow and Gly diets, and in recombinant A2, but not in samples from Met diet–fed mice. These results revealed that, whereas A2 tissue levels are unaffected by hyperhomocysteinemia, the A2 extracted from hyperhomocysteinemic mice appeared to have lost its ability to bind TPA. Interestingly, this capacity was completely restored when the sample was treated with a reducing agent, suggesting that the loss of binding of TPA to A2 in hyperhomocysteinemic mice involved the formation of at least 1 disulfide bond. Similarly, while frozen sections of lung tissue from mice on chow and Gly diets bound biotinylated TPA in an overlay assay, sections from mice on the Met diet showed significantly less binding of TPA (Figure 3B).

Figure 3

Effect of HC on A2 binding of TPA and plasmin generation. (A) Lung membrane proteins from mice maintained on chow, Gly, or Met diets and rA2 (1 μg) were immunoblotted with rabbit anti-mouse TPA or mouse anti-A2 IgG. For ligand blotting (LB), proteins resolved under reducing (R) or non-reducing (NR) conditions were incubated with human TPA and then probed with rabbit anti-mouse TPA. GAPDH served as a loading control. (B) Left: Cryosections from mice on chow, Gly, or Met diets were incubated with biotinylated TPA and stained as described in Methods. Right: Paraffin-embedded lung sections were stained with rabbit anti–HC-A2 tail peptide IgG as described in Methods. Original magnification, ×200. (C) Purification of native A2. Reduced Coomassie blue–stained (CB) SDS-PAGE and immunoblot analysis with monoclonal anti-A2 of native A2 purified from lung tissue of mice on Gly and Met diets. Samples were run on noncontiguous lanes of the same gel. (D) TPA-dependent plasmin generation in the presence of A2 purified from lung tissue. Initial rates of hydrolysis of a fluorogenic plasmin substrate were calculated as described in Methods. RFU, relative fluorescence units. *P < 0.001 versus Gly.

We previously demonstrated that HC can derivatize A2 in vitro through formation of a disulfide linkage with cysteine 8, thereby impeding the A2-TPA interaction. To determine whether HC-A2 disulfide conjugates can be found in vivo, we employed rabbit IgG directed against the keyhole limpet hemocyanin–conjugated (KLH-conjugated), homocysteinylated A2 N-terminal peptide (Ac-STVHEILC[hCys]KLSLEGD[C]-[KLH]). This antibody displayed weak specific reactivity with lung tissue from mice on chow or Gly diets but strong reactivity when tissue from Met diet–fed mice was examined (Figure 3B). This polyclonal antibody failed to react with any proteins contained within total lysates from Anxa2^–/– endothelial cells (data not shown), indicating its specificity for A2. We concluded that Met tissues, in comparison with Gly or chow tissues, contain significantly higher concentrations of homocysteinylated A2 and bind significantly less TPA.

We then examined the effect of hyperhomocysteinemia on the ability of purified A2 to support TPA-dependent plasminogen activation. Native non-reduced A2 was purified to near homogeneity by calcium precipitation and detergent extraction (Figure 3C). In the absence of added A2 or plasminogen, plasmin generation was inefficient (data not shown). Furthermore, addition of A2 purified from lung tissue of Gly diet–fed mice accelerated plasmin generation by 5- to 6-fold over baseline, while addition of A2 from hyperhomocysteinemic Met mice failed to augment baseline catalytic activity above control levels (Figure 3D). These data establish that in hyperhomocysteinemia the fibrinolytic cofactor activity of A2 is impaired.

HC blocks directed endothelial cell migration. To determine whether HC-induced A2 loss of TPA binding capacity affects directed migration of endothelial cells in vivo, we employed the Matrigel implant assay using both VEGF and bFGF-containing implants. Both agents significantly stimulated vascular invasion in control mice (Figure 4). After 12 days, VEGF-supplemented implants harvested from Gly mice contained 50.2 ± 6.0 (mean ± SEM, n = 7) cells per 0.3 mm² and numerous erythrocyte-containing neovessels were noted. In contrast, implants from Met mice had 13.5 ± 1.6 cells per 0.3 mm² (mean ± SEM, n = 8, P < 0.001) and neovessels containing erythrocytes were not seen (Figure 4A). In all sections, about 90% of invading cells were found to express the endothelial cell marker von Willebrand factor (data not shown). These data suggest that moderate hyperhomocysteinemia impairs endothelial cell invasion of, or survival within, a Matrigel environment.

Figure 4

Effect of HC on endothelial cell invasion of Matrigel implants. (A) Left: Sections from Matrigel implants containing VEGF-A (100 ng/ml) harvested from mice on Gly and Met diets. Arrows indicate erythrocyte-containing neovessels. Right: Matrigel plug cellularity, assessed as described in Methods. Shown are mean ± SEM for 8 separate plugs, 10 fields per plug. Scale bars: 20 μm. (B) Left: Sections from Matrigel implants containing 75 ng/ml bFGF, 100 μg/ml heparin, and the indicated concentrations of DL-HC, harvested from mice maintained on a normal chow diet. Right: Matrigel plug cellularity. Shown are mean ± SEM for 3 separate plugs, 10 fields per plug. Scale bars: 50 μm. (C) Left: Sections from Matrigel implants containing 75 ng/ml bFGF and 100 μg/ml heparin with or without 100 μM DL-HC, L-cysteine, or L-methionine, harvested from mice on a normal chow diet. Right: Matrigel plug cellularity. Shown are mean ± SEM for 3 separate plugs, 10 fields per plug. Scale bars: 50 μm. All sections were stained with hematoxylin and eosin. *P < 0.01 versus control.

Similarly, when bFGF-containing Matrigel plugs were supplemented with HC and placed in mice on normal chow, we noted a dose-related diminution of implant cellularity and vascularity compared with plugs treated with bFGF alone (Figure 4B). Based upon manual cell counts and computer-assisted image analysis, the I₅₀ for this effect was approximately 90 μM (Figure 4B). In addition, only HC, and not cysteine and methionine, impaired Matrigel invasion (Figure 4C). While HC (100 μM) blocked 62.8% of cellular infiltration seen in the control (40.8 ± 10.1 versus 104.6 ± 21.2 cells/0.3 mm²; mean ± SEM, n = 3, P < 0.01), cysteine blocked only 2.3% (97.7 ± 18.5 cells/0.3 mm²; mean ± SEM, n = 3, P = NS), and methionine enhanced cellularity by 30.2% (142.7 ± 47.5 cells/0.3 mm²; mean ± SEM, n = 3, P = NS; Figure 4C). Similar values were obtained by computer-assisted image analysis (HC 64.7% ± 5.9% inhibition, mean ± SEM, n = 18; cysteine 21.6% ± 7.8% inhibition, mean ± SEM, n = 6; and methionine 35.3% ± 13.7% enhancement, mean ± SEM, n = 5). Therefore, we determined that among sulfur-containing amino acids, HC has a selective ability to block endothelial cell invasion of, or survival within, bFGF-containing implanted matrix.

We next assessed the effect of homocysteine on endothelial cell survival, proliferation, or non-directed motility. Incubation of human dermal microvascular endothelial cells (HDMECs) with HC (0, 10, and 100 μM; 24 hours) had no significant effect on cell survival, as judged by the ability to exclude Trypan blue (10.9% ± 3.4%, 11.4% ± 5.3%, and 17.9% ± 2.7%, mean ± SD, n = 3, P = NS). In addition, numbers of proliferating endothelial cells recovered 72 hours after plating at low density in HC-containing medium (0, 10, or 100 μM) did not differ significantly (179,000 ± 62,200, 167,500 ± 23,300, and 162,500 ± 14,000, respectively, mean ± SD, n = 3, P = NS). Similarly, there was no difference in random cell motility through uncoated porous membranes between untreated and HC-treated (10 or 100 μM) cells (3,658 ± 1,033 cells; 2,773 ± 797 cells; and 2,626 ± 767 cells, respectively, mean ± SEM, n = 5, P = NS). These data suggest that basal endothelial cell survival, proliferation, and motility were not significantly altered by HC.

To test the potential effect of HC on microvascular endothelial cell matrix invasion, HDMECs were pre-incubated for 24 hours with HC (10–1,000 μM) and plated on porous membranes placed atop modified Boyden chambers containing VEGF-A–supplemented medium. In the untreated control samples, cells that migrated through the filter represented 28.6 ± 6.5, 24.0 ± 3.8, and 23.4 ± 4.6 cells per 0.3 mm² (mean ± SEM, n = 3) of those plated on filters coated with gelatin, Matrigel, and fibrin, respectively (~20 cells per 0.3 mm²). Upon pretreatment of cells with HC, however, we observed a dose-related diminution in the number of cells that penetrated filters coated with each of the 3 artificial matrices (Figure 5A). When cells were co-incubated, rather than pre-incubated, with HC, we observed no effect upon their invasive capability. Treatment with HC inhibited endothelial cell invasion through Matrigel, gelatin, and fibrin (I₅₀ values 5, 8, and 150 μM HC, respectively).

Figure 5

Effect of HC on directed migration of HDMECs. (A) Dose response. Confluent HDMECs were preincubated for 24 hours with the indicated concentrations of DL-HC, then plated (25,000 cells per filter) onto filters pre-coated with fibrin, Matrigel, or gelatin and positioned over wells containing complete medium supplemented with VEGF-A. After 12 hours, cells adherent to the upper surface of the filter were removed, and cells on the under surface were fixed and stained with hematoxylin. (B) Specificity. HDMECs were pretreated (24 hours) with the indicated sulfur-containing amino acids (100 μM), then plated at 50,000 cells per filter coated as described in A. In each experiment, cell migration was assessed in 10 fields from each of 3 duplicate filters. Data are mean ± SEM, n = 3 filters. ^†P < 0.05, ^#P < 0.01, *P < 0.001 versus control.

We next compared the effect of HC on directed endothelial cell migration with the effect of other sulfur-containing amino acids. While HC (100 μM) pretreatment of endothelial cells effectively blocked VEGF-induced migration through gelatin (21.8 ± 6.9 vs. 39.9 ± 4.4 cells per 0.3 mm², mean ± SEM, n = 3, P < 0.05), Matrigel (6.5 ± 0.4 vs. 21.9 ± 4.5 cells per 0.3 mm², mean ± SEM, n = 3, P < 0.01), and fibrin (8.5 ± 0.8 vs. 23.6 ± 1.4 cells per 0.3 mm², mean ± SEM, P < 0.001) by 44.5%, 70.3%, and 64.0%, respectively, cysteine, methionine (Figure 5B), and homocystine (data not shown) had no effect at the same concentration. Thus, HC specifically impaired directed migration of HDMECs in response to VEGF-A in vitro, the most statistically significant effect having been observed in the setting of a fibrin barrier.

Hyperhomocysteinemia inhibits corneal neoangiogenesis. To further evaluate the in vivo neoangiogenic response in the setting of hyperhomocysteinemia, we placed bFGF-containing pellets within the normally avascular cornea (33). In Met diet–fed mice, the area of neovascular response (angiogenic index) was 6-fold less (0.3 ± 0.2 vs. 1.8 ± 0.4 mm², mean ± SEM, n = 14) (data not shown) than in Gly diet–fed mice. Similarly, maximum neovessel length was reduced in Met mice (0.4 ± 0.1 vs. 1.1 ± 0.2 mm, mean ± SEM, n = 14), as was the number of vessels greater than 0.5 mm in length (0.7 ± 0.3 vs. 2.6 ± 0.6, mean ± SEM, n = 14) (data not shown). These data indicate that bFGF-driven neoangiogenesis was inhibited in diet-induced hyperhomocysteinemia.

Next we asked whether the defect in corneal neoangiogenesis could be repaired upon introduction of fresh recombinant A2. In mice on chow or Gly diets, we found no significant difference in neovascular area when mice were treated with supplemental A2 (1 mg/kg i.v. once per day for 5 days) versus BSA (2.1 ± 0.3 and 2.5 ± 0.6 mm² vs. 2.4 ± 0.01 and 2.2 ± 0.4 mm², mean ± SEM, n = 3) (Figure 6, A and B). On the other hand, angiogenesis in the Met diet–fed mice remained severely depressed following BSA treatment (0.6 ± 0.2 mm², mean ± SEM, n = 3, P < 0.01) but improved dramatically following treatment with recombinant His-tagged A2 (1.9 ± 0.1 mm², mean ± SEM, n = 3) (Figure 6, A and B). This effect was not due to contamination of the recombinant His-tagged A2 (rA2) preparation with bacterial products because material prepared in a parallel fashion from mock-transfected E. coli had no effect on corneal angiogenesis (data not shown). This result indicated that the homocysteinemia-associated defect in neoangiogenesis could be completely corrected with recombinant A2.

Figure 6

Effect of hyperhomocysteinemia on corneal pocket neoangiogenesis. (A) Corneal angiogenesis in mice maintained on chow, Gly, or Met diets. Following implantation of bFGF-containing pellets, mice received 5 daily tail vein injections of either BSA or rA2. Shown are representative photographs of corneal neovessels. (B) Neovascular area shown as mean values ± SEM, n = 3. *P < 0.01 versus all other groups. (C) Neovascular area from corneal pocket assays in Anxa2^+/+ and Anxa2^–/– mice on chow, Gly, or Met diets. Shown are mean values ± SEM, n = 9. *P < 0.001. (D) Confocal immunofluorescence analysis of corneal neovessels in mice maintained on the Met diet and treated with i.v. BSA or rA2 as described in A. On day 5, corneal tissue was harvested and stained with anti–MECA-32 (red), Penta-His IgG (green), and DAPI (blue). Original magnification, ×75. (E) Sections prepared as described in D were stained with rabbit anti-human fibrin(ogen) IgG (brown) and rat MECA-32 IgG (blue). Original magnification, ×800.

Although Anxa2^–/– mice showed reduced neovascular area compared with Anxa2^+/+ mice, neoangiogenesis was not further depressed when Anxa2^–/– mice were placed on the Met diet (Figure 6C), suggesting that HC may impair angiogenesis through interference with an A2-related pathway. Confocal microscopy, moreover, revealed colocalization of anti-PentaHis immunoreactivity with immunostaining for the endothelial cell marker MECA-32 in corneal tissue from mice treated with rA2 (Figure 6D). Thus, the angiogenic defect associated with hyperhomocysteinemia can be corrected by supplementation with exogenous rA2.

Finally, we asked whether fibrin might constitute a barrier to neovessel growth in Met diet–fed mice (Figure 6E). While neovessels within the corneas of rA2-treated mice were essentially devoid of immunohistochemically detectable fibrin, those within corneas of BSA-treated mice showed significant deposition, with fibrin strands partially obliterating the vascular lumen. Results of these experiments strongly suggest that excessive fibrin deposition represents a significant barrier to neoangiogenesis in the hyperhomocysteinemic mouse cornea.

In 1969, McCully reported widespread atherosclerosis-like arterial disease at postmortem in an infant with hyperhomocysteinemia due to an inborn error of cobalamin metabolism (34). Since then, numerous studies, including recent meta-analyses, have confirmed an association between elevated circulating HC levels and macrovascular atherothrombotic disease, including coronary artery occlusion, stroke, and venous thromboembolism (35). Although there is now strong evidence that HC-lowering therapies reduce the incidence of major vascular events in patients with severe hyperhomocysteinemia, it remains unresolved as to whether treatment of mildly elevated HC lowers the risk of vascular disease (36–38).

Against this backdrop, there is renewed interest in HC as an etiologic agent in microvasculopathies. Hyperhomocysteinemia has emerged as an independent risk factor for cerebral small vessel disease, which may lead to lacunar infarction and accounts for about 25% of ischemic strokes (19, 39–42). In older individuals, several large cross-sectional studies have linked hyperhomocysteinemia with depression and with small vessel disease–related dementias, including Alzheimer disease (43–45). Hyperhomocysteinemia may also be a new risk factor for proliferative retinopathy and nephropathy in type 2 diabetes (46) as well as chronic lower extremity ulceration in patients with or without diabetes (47). While these associations could reflect impairment of microvascular thromboresistance or inhibition of microvascular remodeling, clear mechanistic insights have been lacking (48, 49).

In the current study, we report for the first time that moderate diet-induced hyperhomocysteinemia (an increase from 11 to 70 μM) in the mouse is associated with impairment of fibrin homeostasis and reduced angiogenesis. We observed fibrin accumulation and reduced perivascular fibrinolytic activity within microvessels of several tissues such as kidney, heart, and lung in hyperhomocysteinemic mice (Figure 1). Similarly, clot lysis following arterial injury was dramatically inhibited in hyperhomocysteinemic mice, further suggesting a significant defect in vascular fibrinolysis (Figure 2). Subsequent studies revealed dysfunction of A2, the major profibrinolytic receptor on resting endothelial cells, impaired endothelial cell migration, and blocked angiogenesis (Figures 3–6). Our data suggest, for what we believe is the first time, that vascular fibrinolysis and angiogenesis are closely aligned through regulation of fibrin homeostasis.

While elevated plasma HC concentrations have been associated with mild impairment of plasma fibrinolysis in humans, the clinical significance of these findings remains unknown. Mild elevations in plasma HC (15.75–16.80 vs. 12.35 μM) appear to correlate with somewhat prolonged plasma clot lysis times (8.7–8.8 vs. 7.9 minutes) (50). In patients with venous thromboembolism, moderate hyperhomocysteinemia (24.2 μM) was associated with elevations of thrombin-activatable fibrinolysis inhibitor (13%), and prolonged plasma clot lysis times (12%) (51). In rabbits, injected HC may induce a dysfibrinogenemia, leading to abnormal plasma clot structure ex vivo and resistance to fibrinolysis (52). Although these observations are of interest, the molecular mechanisms that underlie them remain obscure.

Vascular fibrinolysis was first observed by Todd in 1964 (53), who determined that when fibrin gels were applied directly to fresh tissue sections, fibrinolysis was specifically localized to blood vessels and required the integrity of the endothelium. Subsequently, TPA was identified as a secretory product of endothelial cells within specific vessel types in defined anatomic locations (3, 54, 55). Urokinase and its receptor, on the other hand, appear to be localized to either renal or cytokine-activated endothelial cells (56, 57). In contrast, the A2 complex has been identified as a major receptor for both TPA and its substrate, plasminogen, on resting and activated endothelial cells (7, 10). The A2-null mouse, moreover, displays normal plasma clot lysis but severe defects in endothelial cell–associated fibrinolytic activity and postnatal angiogenesis (11).

The present study shows that, while levels of total lung A2 did not differ between normo- and hyperhomocysteinemic mice, the TPA-binding ability of A2 from hyperhomocysteinemic lungs was dramatically diminished (Figure 3). This loss of binding capacity was completely restored upon reduction of A2 thiols. Lung tissue from Met diet–fed mice, furthermore, showed impaired TPA binding, leading to the conclusion that modification of A2 through a HC-mediated oxidative mechanism blocks its ability to interact with TPA. In further support of this postulate, we found that IgG, directed at an HC-derivatized N-terminal A2 peptide, bound avidly to vessels and perivascular lung tissue from Met diet–fed mice, but weakly to vessels in lungs of chow diet–fed or Gly diet–fed mice. Finally, A2 purified from hyperhomocysteinemic mice failed to significantly stimulate TPA-dependent plasminogen activation in vitro. Together these data indicate that direct HC derivatization of A2 impairs TPA binding and prevents plasmin generation at the endothelial cell surface.

Our second major finding was that diet-induced hyperhomocysteinemia is associated with impaired angiogenesis in the mouse. Matrigel implants were far less extensively vascularized in hyperhomocysteinemic mice than in normohomocysteinemic mice (Figure 4). Matrigel supplemented directly with HC showed impaired neoangiogenesis, thus ruling out a non-HC–related anti-angiogenic effect of the Met diet. HC also blocked directed endothelial cell migration in a specific and dose-related fashion in vitro (Figure 5). These data indicate that HC inhibits endothelial cell migration by blocking the cell’s ability to generate cell surface plasmin and by diminishing its activation of downstream proteases such as MMP-3 and -13 (11).

To date, there are few studies that link fibrin to dysangiogenesis. In our study, hyperhomocysteinemic mice displayed a deficiency in angiogenesis, which could be completely and specifically reversed upon infusion of recombinant A2 (Figure 6, A and B). While the hyperhomocysteinemic diet specifically reduced corneal angiogenesis in Anxa2^+/+ mice, no further reduction, beyond that due to A2 deficiency itself, was observed when Anxa2^–/– mice were placed on the Met diet, suggesting that HC targets the A2 pathway (Figure 6C). We noted further that infused rA2 localized to corneal neovascular endothelial cells in hyperhomocysteinemic mice (Figure 6D) and prevented fibrin accumulation and microvascular occlusion (Figure 6E). Taken together, these data reveal that homocysteine derivatizes A2 in vivo, leading to blockade of endothelial cell surface plasmin generation, formation of a fibrin barrier, loss of endothelial cell migratory ability, and angiogenic failure.

Our findings have several significant implications. First, although fibrin has been suggested to play a potentiating role in angiogenesis, based largely upon in vitro observations (58), we demonstrate that fibrin, the provisional matrix that forms in response to vascular injury, may act as a physical barrier to directed endothelial cell migration. Second, while hyperhomocysteinemia is strongly associated with atherothrombotic vascular disease in humans, there is, to our knowledge, no prior report of its association with dysangiogenesis in non-ischemic tissues. Third, our findings suggest that mice with diet-induced elevations of HC closely phenocopy the A2-null mouse, suggesting that HC and A2 belong to a common pathway relevant to fibrin homeostasis and angiogenesis. Taken together, the data provide a mechanistic explanation for vascular dysfunction in the setting of hyperhomocysteinemia, particularly at the level of the microvasculature. Direct blockade of A2 profibrinolytic function by HC might underlie the pathogenesis of small vessel disease associated with dementia, depression, chronic wounds, and diabetes.

We thank Kaili He, Emil Lev, Guangzhi Sui, LunBiao Yan, and Jing Ling for expert technical assistance. This work was supported by NIH grants HL 42493, HL 67839, and HL 90895 to K.A. Hajjar; HL 46403 to K.A. Hajjar, A.J. Marcus, and M.J. Broekman; and HL 47073 to A.J. Marcus and M.J. Broekman and March of Dimes grant 6-FY05 to K.A. Hajjar. A.J. Marcus is the recipient of a Merit Review Grant from the Department of Veterans Affairs and a grant from the Cancer Research and Treatment Fund. A.T. Jacovina was supported by NIH grant T32 HL 07423.

Conflict of interest: The authors have declared that no conflict of interest exists.

Reference information: J. Clin. Invest.119:3384–3394 (2009). doi:10.1172/JCI39591.

See the related article at Homocysteine-mediated thrombosis and angiostasis in vascular pathobiology.

1. Hajjar, K.A. 2006. Vascular biology. InClinical hematology . N.S. Young, S.L. Gerson, and K.A. High, editors. Mosby Elsevier. Philadelphia, Pennsylvania, USA. 123–133.
2. Cines, D.B., et al. 1998. Endothelial cells in physiology and in the pathophysiology of vascular disorders. Blood. 91:3527-3561.
   View this article via: PubMed Google Scholar
3. Levin, E.G., et al. 1997. The expression of endothelial tissue plasminogen activator in vivo: a function defined by vessel size and anatomic location. J. Cell Sci. 110:139-148.
   View this article via: PubMed Google Scholar
4. Hajjar, K.A., et al. 1987. Binding of tissue plasminogen activator to cultured human endothelial cells. J. Clin. Invest. 80:1712-1719.
   View this article via: JCI CrossRef PubMed Google Scholar
5. Hajjar, K.A., Hamel, N.M. 1990. Identification and characterization of human endothelial cell membrane binding sites for tissue plasminogen activator and urokinase. J. Biol. Chem. 265:2908-2916.
   View this article via: PubMed Google Scholar
6. Hajjar, K.A. 1991. The endothelial cell tissue plasminogen activator receptor: specific interaction with plasminogen. J. Biol. Chem. 266:21962-21970.
   View this article via: PubMed Google Scholar
7. Hajjar, K.A., Jacovina, A.T., Chacko, J. 1994. An endothelial cell receptor for plasminogen and tissue plasminogen activator: I. Identity with annexin II. J. Biol. Chem. 269:21191-21197.
   View this article via: PubMed Google Scholar
8. Gerke, V., Moss, S.E. 2002. Annexins: From structure to function. Physiol. Rev. 82:331-371.
   View this article via: PubMed Google Scholar
9. Gerke, V., Creutz, C.E., Moss, S.E. 2005. Annexins: Linking Ca±± signalling to membrane dynamics. Nat. Rev. Mol. Cell Biol. 6:449-461.
   View this article via: CrossRef PubMed Google Scholar
10. Cesarman, G.M., Guevara, C.A., Hajjar, K.A. 1994. An endothelial cell receptor for plasminogen/tissue plasminogen activator: II. Annexin II-mediated enhancement of t-PA-dependent plasminogen activation. J. Biol. Chem. 269:21198-21203.
    View this article via: PubMed Google Scholar
11. Ling, Q., et al. 2004. Annexin II is a key regulator of fibrin homeostasis and neoangiogenesis. J. Clin. Invest. 113:38-48.
    View this article via: JCI PubMed Google Scholar
12. Selhub, J. 1999. Homocysteine metabolism. Annu. Rev. Nutr. 19:217-246.
    View this article via: CrossRef PubMed Google Scholar
13. Jacobsen, D. 2001. Practical chemistry of homocysteine and other thiols. InHomocysteine in health and disease . R. Carmel and D. Jacobsen, editors. Cambridge University Press. New York, New York, USA. 9–20.
14. Nygard, O., et al. 1997. Plasma homocysteine levels and mortality in patients with coronary artery disease. N. Engl. J. Med. 337:230-236.
    View this article via: CrossRef PubMed Google Scholar
15. Malinow, M.R., Bostom, A.G., Krauss, R.M. 1999. Homocyst(e)ine, diet, and cardiovascular diseases: A statement for healthcare professionals from the nutrition committee, American Heart Association. Circulation. 99:178-182.
    View this article via: PubMed Google Scholar
16. Pietrzik, K., Bronstrup, A. 1997. Causes and consequences of hyperhomocyst(e)inemia. Int. J. Vitam. Nutr. Res. 67:389-395.
    View this article via: PubMed Google Scholar
17. Moat, S.J., et al. 2004. Folate, homocysteine, endothelial function and cardiovascular disease. J. Nutr. Biochem. 15:64-79.
    View this article via: CrossRef PubMed Google Scholar
18. Splaver, A., Lamas, G.A., Hennekens, C.H. 2004. Homocysteine and cardiovascular disease: Biological mechanisms, observational epidemiology, and the need for randomized trials. Am. Heart J. 148:34-40.
    View this article via: CrossRef PubMed Google Scholar
19. Markus, H.S. 2007. Genes, endothelial function and cerebral small vessel disease in man. Exp. Physiol. 93:121-127.
    View this article via: CrossRef PubMed Google Scholar
20. Ling, Q., Hajjar, K.A. 2000. Inhibition of endothelial cell thromboresistance by homocysteine. J. Nutr. 130:373S-376S.
    View this article via: PubMed Google Scholar
21. Hajjar, K.A., et al. 1998. Tissue plasminogen activator binding to the annexin II tail domain: direct modulation by homocysteine. J. Biol. Chem. 273:9987-9993.
    View this article via: CrossRef PubMed Google Scholar
22. Hajjar, K.A. 1993. Homocysteine-induced modulation of tissue plasminogen activator binding to its endothelial cell membrane receptor. J. Clin. Invest. 91:2873-2879.
    View this article via: JCI CrossRef PubMed Google Scholar
23. Hajjar, K.A., Jacovina, A.T. 1998. Modulation of annexin II by homocysteine: implications for atherothrombosis. J. Investig. Med. 46:364-369.
    View this article via: PubMed Google Scholar
24. Carmeliet, P. 2003. Angiogenesis in health and disease. Nat. Med. 9:653-660.
    View this article via: CrossRef PubMed Google Scholar
25. Risau, W. 1997. Mechanisms of angiogenesis. Nature. 386:671-674.
    View this article via: CrossRef PubMed Google Scholar
26. Werb, Z. 1997. ECM and cell surface proteolysis: Regulating cellular ecology. Cell. 91:439-442.
    View this article via: CrossRef PubMed Google Scholar
27. Loskutoff, D.J., Quigley, J.P. 2000. PAI-1, fibrosis, and the elusive provisional fibrin matrix. J. Clin. Invest. 106:1441-1443.
    View this article via: JCI CrossRef PubMed Google Scholar
28. Lijnen, H.R. 2001. Plasmin and matrix metalloproteinases in vascular remodeling. Thromb. Haemost. 86:324-333.
    View this article via: PubMed Google Scholar
29. Farrehi, P.M., Ozaki, K., Carmeliet, P., Fay, W.P. 1998. Regulation of arterial thrombolysis by plasminogen activator inhibitor-1 in mice. Circulation. 97:1002-1008.
    View this article via: PubMed Google Scholar
30. Fay, W.P., Parker, A.C., Ansari, M.N., Zheng, X., Ginsburg, D. 1999. Vitronectin inhibits the thrombotic response to arterial injury in mice. Blood. 93:1825-1830.
    View this article via: PubMed Google Scholar
31. Konstantinides, S., Schafer, K., Thinnes, T., Loskutoff, D.J. 2001. Plasminogen activator inhibitor-1 and its cofacor vitronectin stabilize arterial thrombi following vascular injury in mice. Circulation. 103:576-583.
    View this article via: PubMed Google Scholar
32. Kurz, K.D., Main, B.W., Sandusky, G. 1990. Rat model of arterial thrombosis induced by ferric chloride. Thromb. Res. 60:269-280.
    View this article via: CrossRef PubMed Google Scholar
33. Kenyon, B.M., et al. 1996. A model of angiogenesis in the mouse cornea. Invest. Ophthalmol. Vis. Sci. 37:1625-1632.
    View this article via: PubMed Google Scholar
34. McCully, K.S. 1969. Vascular pathology of homocysteinemia: Implications for the pathogenesis of atherosclerosis. Am. J. Pathol. 56:111-128.
    View this article via: PubMed Google Scholar
35. Humphrey, L.L., Rongwei, F., Rogers, K., Freeman, M., Helfand, M. 2008. Homocysteine level and coronary heart disease incidence: A systematic review and meta-analysis. Mayo Clin. Proc. 83:1203-1212.
    View this article via: CrossRef PubMed Google Scholar
36. Maron, B.A., Loscalzo, J. 2006. Homocysteine. Clin. Lab. Med. 26:591-609.
    View this article via: CrossRef PubMed Google Scholar
37. Eldibany, M.M., Caprini, J.A. 2007. Hyperhomocysteinemia and thrombosis. Arch. Pathol. Lab. Med. 131:872-884.
    View this article via: PubMed Google Scholar
38. Antoniades, C., Antonopoulos, A.S., Tousoulis, D., Marinou, K., Stefanadis, C. 2009. Homocysteine and coronary atherosclerosis: From folate fortification to the recent clinical trials. Eur. Heart J. 30:6-15.
    View this article via: CrossRef PubMed Google Scholar
39. Hassan, A., et al. 2003. Homocysteine is a risk factor for cerebral small vessel disease, acting via endothelial dysfunction. Brain. 127:212-219.
    View this article via: CrossRef PubMed Google Scholar
40. Fassbender, K., et al. 1999. Homocysteine in cerebral macroangiography and microangiopathy. Lancet. 353:1586-1587.
    View this article via: CrossRef PubMed Google Scholar
41. Wong, A., et al. 2006. Hyperhomocysteinemia is associated with volumetric white matter change in patients with small vessel disease. J. Neurol. 253:441-447.
    View this article via: CrossRef PubMed Google Scholar
42. Khan, U., Hassan, A., Vallance, P., Markus, H.S. 2007. Asymmetric dimethylarginine in cerebral small vessel disease. Stroke. 38:411-413.
    View this article via: CrossRef PubMed Google Scholar
43. Folstein, M., et al. 2007. The homocysteine hypothesis of depression. Am. J. Psychiatry. 164:861-867.
    View this article via: CrossRef PubMed Google Scholar
44. Selhub, J. 2008. Public health significance of elevated homocysteine. Food Nutr. Bull. 29:S116-S125.
    View this article via: PubMed Google Scholar
45. Smith, D.A. 2008. The worldwide challenge of the dementias: A role for B vitamins and homocysteine? Food Nutr. Bull. 29:S143-S172.
    View this article via: PubMed Google Scholar
46. Looker, H.C., et al. 2003. Homocysteine as a risk factor for nephropathy and retinopathy in Type 2 diabetes. Diabetologia. 46:766-772.
    View this article via: CrossRef PubMed Google Scholar
47. Schwartzfarb, E.M., Romanelli, P. 2008. Hyperhomocysteinemia and lower extremity wounds. Int. J. Low. Extrem. Wounds. 7:126-136.
    View this article via: PubMed CrossRef Google Scholar
48. Castro, R., Rivera, I., Blom, H.J., Jakobs, C., Tavares de Almeida, I. 2006. Homocysteine metabolism, hyperhomocysteunaemia and vascular disease: an overview. J. Inherit. Metab. Dis. 29:3-10.
    View this article via: CrossRef PubMed Google Scholar
49. Abahji, T.N., et al. 2007. Acute hyperhomocysteinemia induces microvascular and macrovascular endothelial dysfunction. Arch. Med. Res. 38:411-416.
    View this article via: CrossRef PubMed Google Scholar
50. Undas, A., et al. 2006. Plasma homocysteine affects fibrin clot permeability and resistance to lysis in human subjects. Arterioscler. Thromb. Vasc. Biol. 26:1397-1404.
    View this article via: CrossRef PubMed Google Scholar
51. Colucci, M., et al. 2008. Mild hyperhomocysteinemia is associated with increased TAFI levels and reduced plasma fibrinolytic potential. J. Thromb. Haemost. 6:1571-1577.
    View this article via: PubMed Google Scholar
52. Sauls, D., Wolberg, A.S., Hoffman, M. 2003. Elevated plasma homocysteine leads to alterations in fibrin clot structure and stability: Implications for the mechanism of thrombosis in hyperhomocysteinemia. J. Thromb. Haemost. 1:300-306.
    View this article via: CrossRef PubMed Google Scholar
53. Todd, A.S. 1964. Localization of fibrinolytic activity in tissues. Br. Med. Bull. 20:210-212.
    View this article via: PubMed Google Scholar
54. Levin, E.G., del Zoppo, G.J. 1994. Localization of tissue plasminogen activator in the endothelium of a limited number of vessels. Am. J. Pathol. 144:855-861.
    View this article via: PubMed Google Scholar
55. Levin, E.G., Osborn, K.G., Schleuning, W.D. 1998. Vessel-specific gene expression in the lung: tissue plasmingen activator is limited to bronchial arteries and pulmonary vessels of discrete size. Chest. 114:68S.
    View this article via: PubMed Google Scholar
56. Wojta, J., Hoover, R.L., Daniel, T.O. 1989. Vascular origin determines plasminogen activator expression in human endothelial cells. Renal endothelial cells produce large amounts of single chain urokinase type plasminogen activator. J. Biol. Chem. 264:2846-2852.
    View this article via: PubMed Google Scholar
57. Pepper, M.S., et al. 1993. Upregulation of urokinase receptor expression on migrating endothelial cells. J. Cell Biol. 122:673-684.
    View this article via: CrossRef PubMed Google Scholar
58. Wojtukiewicz, M.Z., Sierko, E., Klement, P., Rak, J. 2001. The hemostatic system and angiogenesis in malignancy. Neoplasia. 3:371-384.
    View this article via: CrossRef PubMed Google Scholar
59. Velez-Carasco, W., Merkel, M., Twiss, C.O., Smith, J.D. 2008. Dietary methionine effects on plasma homocysteine and HDL metabolism in mice. J. Nutr. Biochem. 19:362-370.
    View this article via: CrossRef PubMed Google Scholar
60. Kopp, H.G., et al. 2006. Thrombospondins deployed by thrombopoietic cells determine angiogenic switch and extent of revascularization. J. Clin. Invest. 116:3277-3299.
    View this article via: JCI CrossRef PubMed Google Scholar
61. Nagashima, M., et al. 2002. Thrombin-activatable fibrinolysis inhibitor (TAFI) deficiency is compatible with murine life. J. Clin. Invest. 109:101-110.
    View this article via: JCI PubMed Google Scholar
62. Moore, G.M., Goulet, F., Sartorelli, A.C. 1992. Purification of annexin I and annexin II from human placental membranes by high performance liquid chromatography. Protein Expr. Purif. 3:1-7.
    View this article via: CrossRef PubMed Google Scholar
63. Khanna, N.C., Tokuda, M., Waisman, D.M. 1987. Purification of three forms of lipocortin from bovine lung. Cell. Calcium. 8:217-228.
    View this article via: CrossRef PubMed Google Scholar
64. Hajjar, K.A., Guevara, C.A., Lev, E., Dowling, K., Chacko, J. 1996. Interaction of the fibrinolytic receptor, annexin II, with the endothelial cell surface: essential role of endonexin repeat 2. J. Biol. Chem. 271:21652-21659.
    View this article via: CrossRef PubMed Google Scholar
65. Kenyon, B.M., Browne, F., D’Amato, R.J. 1997. Effects of thalidomide and related metabolites in a mouse corneal model of neovascularization. Exp. Eye Res. 64:971-978.
    View this article via: PubMed Google Scholar