Therapeutic Potential of Anti-Angiogenic Multitarget N,O-Sulfated E. Coli K5 Polysaccharide in Diabetic Retinopathy

Proliferative diabetic retinopathy (PDR) is the main cause of blindness in the working-age population (1). Numerous angiogenic factors and cytokines related to retinal neovascularization have been identified (2–5) whose levels are elevated in the ocular fluid of patients with PDR (5–7). Among them, upregulation of vascular endothelial growth factor (VEGF) has been associated with the breakdown of the blood-retina barrier, increased vessel permeability, and retinal neovascularization (8–10).

Three different VEGF blockers have evolved over time for PDR therapy (pegaptanib, ranibizumab, and aflibercept); a fourth agent, bevacizumab, is used off label (11). Even though anti-VEGF interventions have shown better outcomes than alternative treatments, limitations to anti-VEGF therapies may exist, including short duration of action, local and systemic adverse effects, and poor response in a significant percentage of patients (12–14). Furthermore, the production of other angiogenic factors may cause resistance to anti-VEGF therapies. Thus, the identification of novel therapeutic approaches targeting additional pathways is warranted (15,16).

Heparan sulfates (HS) are N,O-sulfated glycosaminoglycans (GAGs) present on the cell surface and extracellular matrix as HS proteoglycans (HSPGs). HSPGs modulate the activity of heparin-binding angiogenic growth factors, including VEGF and fibroblast growth factor (FGF) family members, by mediating their interaction with signaling receptors (17,18). Thus, synthetic or chemically modified heparin/HS-like molecules may act as angiogenic factor inhibitors, hampering their binding to endothelial receptors (19). Relevant to PDR, intravitreal injection of high concentrations of heparin/HS or heparin analogs reduces retinal neovascularization in a model of oxygen-induced retinopathy (OIR) (20,21), originally developed in mice as a model of retinopathy of prematurity (22) and widely adopted also to study angiogenesis in PDR (23).

The Escherichia coli K5 polysaccharide, whose name coincides with the abbreviation of the structurally and functionally unrelated antiangiogenic plasminogen proteolytic fragment kringle 5 (24), has the same structure as the heparin precursor N-acetyl heparosan (25). K5 polysaccharide derivatives, produced by chemical and enzymatic modifications, represent a new class of heparin/HS-like compounds (26). In particular, the highly N,O-sulfated derivative K5-N,OS(H) is endowed with a negligible anticoagulant activity and acts as an antiangiogenic, anti-inflammatory, and antiviral molecule due to its capacity to interact with different heparin-binding proteins (26,27). Thus, K5 derivatives may represent a novel class of multitarget molecules able to interact with and block the activity of a broad range of heparin-binding angiogenic factors involved in the pathogenesis of PDR.

Here, we investigated the capacity of K5-N,OS(H) to affect the angiogenic activity of the major heparin-binding VEGF-A₁₆₅ isoform (hereafter named VEGF) expressed in PDR (3). The results demonstrate that K5-N,OS(H) inhibits VEGF-driven neovascular responses in vitro, ex vivo, and in vivo. Furthermore, K5-N,OS(H) binds a variety of heparin-binding angiogenic factors present in the vitreous humor of patients with PDR and inhibits the angiogenic activity of PDR vitreous fluid samples. Thus, K5-N,OS(H) represents an antiangiogenic multitarget molecule with potential therapeutic implications in PDR.

K5 derivatives (28) and unfractionated heparin were from Glycores 2000; M199 medium and FCS from Gibco Life Technologies (Grand Island, NY); FGF2 from Tecnogen (Caserta, Italy); connective tissue growth factor (CTGF) and stromal cell–derived factor-α (SDF1α) from PeproTech (Rock Hill, NJ); hepatocyte growth factor (HGF), IGF-1, interleukin-8 (IL8), and platelet-derived growth factor-B (PDGF-B) from ReliaTech (Wolfenbüttel, Germany); high mobility group box-1 (HMGB1) from ProSpec (Rehovot, Israel); bevacizumab from Roche (Basel, Switzerland); and cyclopeptide vascular endothelial growth inhibitor (cyclo-VEGI) and the soluble form of VEGFR2 (sVEGFR2) from Calbiochem (Darmstadt, Germany). VEGF (VEGF-A₁₆₅ isoform) was provided by K. Ballmer-Hofer (PSI, Villigen, Switzerland). Anti-VEGFR2, anti-pVEGFR2, and anti-rabbit horseradish peroxidase (HRP)-labeled antibodies were from Santa Cruz Biotechnology (Santa Cruz, CA); rat monoclonal anti-CD31 antibody from BD Biosciences (San Diego, CA); Alexa Fluor 488 from Molecular Probes (Eugene, OR); and anti–β-actin antibody from Sigma-Aldrich (St. Louis, MO).

Biotinylated heparin was immobilized onto a Biacore SA sensor chip (GE Healthcare, Madison, WI) containing pre-immobilized streptavidin, whereas sVEGFR2 was immobilized on an activated CM5 sensor chip (29). For competition experiments, VEGF (150 nmol/L) plus increasing GAG concentrations were injected over heparin or sVEGFR2 surfaces, and bound analyte was measured (resonance units [RU]) at the end of injection. For evaluation of K5-N,OS(H)/growth factor interaction, proteins were injected over the heparin surface in the presence of increasing GAG concentrations.

Procedures involving animals were performed in agreement with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research and in compliance with Italian guidelines for animal care (DL 116/92) and European Communities Council Directive (86/609/EEC). Procedures were approved by the Ethical Committee in Animal Experiments of the universities involved in the study.

GAG-deficient A745 CHO cells were transfected with pcDNA3 harboring the human VEGFR2 extracellular domain (ECD) conjugated with yellow fluorescent protein (YFP) to generate stable ECD-VEGFR2-YFP transfectants. CHO-K1 cells were seeded in 24-well plates at 50,000 cells/cm². After 24 h, ECD-VEGFR2-YFP cells (50,000 cells/cm²) were added to CHO-K1 monolayers in the presence of 75 ng/mL VEGF and different competitors. After 2 h at 4°C, ECD-VEGFR2-YFP cells bound to the CHO-K1 monolayer were counted.

Human umbilical vein endothelial cells (HUVECs) treated with 5.0 ng/mL VEGF and increasing concentrations of K5-N,OS(H) were probed by Western blotting with anti-pVEGFR2 antibody. For the cell proliferation assay, HUVECs (15,000 cells/cm²) were treated with 10 ng/mL VEGF or vitreous fluid diluted in culture medium (1:4, volume:volume [vol:vol]) plus 2.5% FCS and increasing K5-N,OS(H) concentrations. After 24 h, cells were counted.

HUVEC aggregates were embedded in fibrin gel (30) and treated with 30 ng/mL VEGF or diluted vitreous fluid (1:4, vol:vol) plus increasing K5-N,OS(H) concentrations. Sprouts were counted after 24 h.

Cell monolayers made quiescent by overnight incubation in medium with 3.5% FCS were wounded with a 1.0-mm-wide rubber policeman. Then, cells were incubated in fresh medium with 3.5% FCS and 50 ng/mL of the angiogenic factor under test in the absence or presence of 100 ng/mL K5-N,OS(H). After 24 h, wound repair was quantified by computerized analysis.

Fibrin-embedded aortic rings were incubated in serum-free endothelial cell basal medium (Clonetics) with 30 ng/mL VEGF and increasing K5-N,OS(H) concentrations. After 5 days, vessel sprouts were counted under a stereo microscope (31).

Fibrin-embedded retina fragments (32,33) were treated with 75 ng/mL VEGF and increasing K5-N,OS(H) concentrations. At day 7, vessel sprouts were counted.

Alginate beads (3 µL) containing 2 µL of vitreous fluid or 100 ng/pellet of VEGF and increasing amounts of K5-N,OS(H) were placed on the chorioallantoic membrane (CAM) of fertilized white Leghorn chicken eggs at day 11 (30). After 3 days, microvessels converging toward the implant were counted.

Polymer Elvax-40 pellets containing 200 ng VEGF with or without 100 ng K5-N,OS(H) were implanted in the cornea of male New Zealand white rabbits (n = 9; Charles River) (34). Angiogenic score (vessel density × distance in millimeters) was calculated at day 11.

Experiments were performed on C57BL/6J mouse pups (22). Sixteen pups were used for experimental protocol setup, and five pups per group were randomly selected for K5-N,OS(H) treatment followed by Western blotting, immunohistochemistry, or quantitative RT-PCR analysis (71 pups total). Litters of mouse pups with their nursing mothers were room-air reared or exposed to 75 ± 2% oxygen between postnatal day (PD)7 and PD12 before return to room air. At PD12 and PD15, 1.0 µL K5-N,OS(H) in sterile PBS (from 0.06 to 60 ng/µL) was intravitreally injected into the left eye, whereas the right eye was injected with vehicle. At PD17, retinal extracts were probed by Western blotting with anti-pVEGFR2, anti-Bax, anti–Bcl-2, and anti–cytochrome c antibodies, normalizing their levels to those of VEGFR2 or β-actin (35). Also, quantitative RT-PCR analysis was performed on total RNA extracted from PD17 retinas, and data were analyzed by the formula 2^-ΔΔCT using Rpl13a gene expression as internal standard (35). Oligonucleotide primers were as follows: Vegf-A forward: 5′-GCACATAGGAGAGATGAGCTTCC-3′ and reverse: 5′-CTCCGCTCTGAACAAGGCT-3′; Igf-1 forward: 5′-TTCAGTTCGTGTGTGGACCGAG-3′ and reverse: 5′-TCCACAATGCCTGTCTGAGGTG-3′; Fgf-2 forward: 5′-GCGACCCACACGTCAAACTA-3′ and reverse: 5′-TCCCTTGATAGACACAACTCCTC-3′; angiopoietin-2 forward: 5′-GACTTCCAGAGGACGTGGAAAG-3′ and reverse: 5′-CTCATTGCCCAGCCAGTACTC-3′; erythropoietin forward: 5′-ACTCTCCTTGCTACTGATTCCT-3′ and reverse: 5′-ATCGTGACATTTTCTGCCTCC-3′; Tnfα forward: 5′-GCCTCTTCTCATTCCTGCTT-3′ and reverse: 5′-CTCCTCCACTTGGTGGTTTG-3′; Il-1β forward: 5′-TCCTTGTGCAAGTGTCTGAAGC-3′ and reverse: 5′-ATGAGTGATACTGCCTGCCTGA-3′; Il-6 forward: 5′-TCTGCAAGAGACTTCCATCCAGT-3′ and reverse: 5′-TCTGCAAGTGCATCATCGTTGT-3′; iNos forward: 5′-GGCAAACCCAAGGTCTACGTT-3′ and reverse: 5′-TCGCTCAAGTTCAGCTTGGT-3′; Rpl13a forward: 5′-CACTCTGGAGGAGAAACGGAAGG-3′ and reverse: 5′-GCAGGCATGAGGCAAACAGTC-3′.

Immunohistochemical analysis of CD31⁺ retinal vasculature and quantification of preretinal neovascular tufts were performed on retinal whole mounts (36).

Patients with PDR (Table 1) underwent pars plana vitrectomy at the Clinics of Ophthalmology of the University of Brescia. Collection and analysis of human samples was approved by the internal review board of the Spedali Civili of Brescia and followed the principles of the Declaration of Helsinki. In some patients, intravitreal injection of the VEGF blockers bevacizumab or ranibizumab was performed 10–15 days prior to vitrectomy (Table 1). Samples were stored at −80°C.

HSPGs mediate the engagement of VEGF with its signaling receptor VEGFR2, leading to the formation of HSPG/VEGF/VEGFR2 ternary complexes (18). To assess the impact of VEGF blockers on the formation of this bioactive complex, we developed a novel assay in which VEGF mediates cell-cell adhesion events due to its simultaneous interaction with VEGFR2 and HSPGs expressed in trans on neighboring cells (28). As shown in Fig. 1A, VEGF allows the dose-dependent adhesion of HSPG-deficient CHO cells overexpressing the ECD of VEGFR2 (ECD-VEGFR2-YFP cells) when seeded on a monolayer of HSPG-bearing CHO-K1 cells. Specificity of the interaction was demonstrated by the lack of activity of the unrelated heparin-binding FGF2 or of the VEGF-A₁₂₁ isoform devoid of the heparin-binding domain. Furthermore, HSPG/VEGF/VEGFR2 complex formation was prevented by the anti-VEGF antibody bevacizumab and by the VEGFR2 antagonist cyclo-VEGI peptide (37) (Fig. 1B). Thus, cell-cell adhesion in this model is VEGF dependent and due to the formation in trans of HSPG/VEGF/VEGFR2 ternary complexes on neighboring cells.

On this basis, different K5 derivatives chemically sulfated in the N position or N,O positions (Table 2) were investigated for the capacity to affect HSPG/VEGF/VEGFR2 complex formation in this assay. Similar to unmodified heparin, the highly N,O-sulfated K5-N,OS(H) prevents the adhesion of ECD-VEGFR2-YFP transfectants to HSPG-bearing cells; the low N,O-sulfated derivative K5-N,OS(L) was ∼20 times less potent than K5-N,OS(H), whereas K5-NS and K5 were ineffective (Fig. 1C and D).

To assess whether this activity was due to the ability of K5 derivatives to affect HSPG/VEGF and/or VEGF/VEGFR2 interactions, GAGs were investigated by surface plasmon resonance (SPR) analysis for their capacity to prevent the binding of VEGF to heparin or sVEGFR2 immobilized to Biacore sensor chips, a VEGF interaction that occurs with K_d values equal to 100 (38) and 3.0 nmol/L (39) for the two molecules, respectively. As shown in Fig. 1E and F, K5-N,OS(H) inhibits the binding of VEGF to immobilized heparin and sVEGFR2, being more effective than free heparin. At variance, K5-N,OS(L) and K5-NS exerted an inhibitory effect only at high doses and K5 was ineffective. The ID₅₀ values calculated for K5 derivatives in cell-cell adhesion and SPR assays are summarized in Table 2.

In conclusion, K5-N,OS(H) acts as a potent VEGF antagonist able to prevent its interaction with heparin/HSPGs and VEGFR2, hampering the formation of bioactive HSPG/VEGF/VEGFR2 complexes. On this basis, the capacity of K5-N,OS(H) to inhibit the biological activity of VEGF was assessed in a series of angiogenesis bioassays.

The antiangiogenic potential of K5-N,OS(H) was assessed in vitro, ex vivo, and in vivo on the endothelium of different species and tissue origin when stimulated by an optimal dose of VEGF able to elicit the maximal response in each assay.

K5-N,OS(H) inhibits VEGF-induced VEGFR2 phosphorylation in HUVECs, maximum effect being observed at 50 ng/mL of the GAG (Fig. 2A). Accordingly, a significant decrease in the cell proliferation rate was observed when HUVECs were treated with VEGF in the presence of K5-N,OS(H) (Fig. 2B). Also, K5-N,OS(H) prevents VEGF-mediated endothelial cell sprouting of fibrin-embedded HUVEC spheroids, murine aortic rings, and murine retinal fragments (Fig. 2C–E).

Next, K5-N,OS(H) was tested in the chick embryo CAM and rabbit cornea assays. In the first assay, alginate beads adsorbed with 100 ng VEGF and increasing concentrations of K5-N,OS(H) were implanted on the CAM of 11-day-old chick embryos (30). As shown in Fig. 2F, K5-N,OS(H) caused a significant inhibition of VEGF-induced angiogenesis in the absence of any effect on embryonic development and survival. In the rabbit cornea assay, pellets of polymer Elvax-40 containing 200 ng VEGF with or without 100 ng K5-N,OS(H) were implanted in the cornea and the angiogenic score was evaluated 11 days thereafter (40). K5-N,OS(H) significantly inhibited VEGF-induced neovascularization also in this assay (Fig. 2G).

The effect of K5-N,OS(H) on retinal angiogenesis was evaluated in vivo in a murine OIR model in which VEGFR2 is the main mediator of the proangiogenic activity of hypoxia-upregulated VEGF (41). As shown in Fig. 3, intravitreal injection of increasing concentrations of K5-N,OS(H) (ranging from 0.06 to 60 ng/µL) caused a significant decrease of VEGFR2 phosphorylation in the retinas of PD17 OIR mice with no effect on the total levels of VEGFR2.

On this basis, retinal vascularization was evaluated by CD31 immunohistochemistry after intravitreal administration of vehicle or K5-N,OS(H) in OIR mice. When compared with normoxic retinas (Fig. 4A), PD17 retinas of untreated and vehicle-treated OIR mice show the formation of a large avascular area in their central region and a prominent vessel regrowth leading to preretinal neovascular tufts in the midperipheral retina (Fig. 4B and C). In K5-N,OS(H)–treated eyes (Fig. 4D–G), the extent of avascular area was similar to those observed in vehicle-treated retinas (data not shown). However, the vessel tuft area appeared to be reduced with no preferential regional distribution. Quantitative analysis of CD31 immunoreactivity confirmed a significant dose-dependent decrease of the neovascular tuft area in K5-N,OS(H)–treated OIR retinas (Fig. 4H). Also, in keeping with its anti-inflammatory activity (42–44), K5-N,OS(H) caused downregulation of the expression of various markers of inflammation in OIR retinas (Supplementary Fig. 1), with no significant changes in the expression of angiogenic growth factors known to play a role in OIR, including VEGF (Supplementary Fig. 2), and in the amount of apoptosis-related proteins, including Bax, Bcl-2, and cytochrome c (Supplementary Fig. 3).

Various heparin-binding proinflammatory/angiogenic factors other than VEGF play an important role in PDR, including FGF2, IGF-1, IL8, PDGF-B, SDF1α, CTGF, HGF, HMGB1, and placental growth factor-2 (PlGF2) (2–5). When tested by SPR analysis, these proteins bind immobilized heparin with K_d values ranging between 10 and 600 nmol/L. Notably, K5-N,OS(H) caused a dose-dependent inhibition of the binding of all the factors tested to the heparin sensor chip, with ID₅₀ values ranging between 0.01 and 0.2 µg/mL (Fig. 5A). Of note, no correlation exists between the capacity of K5N,OS(H) to hamper the interaction of the various proteins with immobilized heparin (expressed as ID₅₀ value) and their affinity for the immobilized GAG (expressed as K_d value) (Fig. 5B).

In keeping with its binding capacity, K5N,OS(H) inhibits the proliferative/motogenic response elicited by the various heparin-binding angiogenic factors on mechanically wounded HUVEC monolayers (Fig. 5C). Thus, K5-N,OS(H) acts as a multitarget antagonist for a variety of heparin-binding proinflammatory/angiogenic factors involved in PDR.

The evaluation of the angiogenic effects exerted by the vitreous humor obtained after pars plana vitrectomy may provide a useful tool for a preclinical screening of angiostatic molecules with potential implications for the therapy of PDR (6,45–47). To this aim, preliminary experiments were performed to define the amount of PDR vitreous fluid that exerts an optimal angiogenic response in the HUVEC sprouting and chick embryo CAM assays. Given the limited amount of fluid available from each patient, samples obtained from five to eight patients were pooled together (Table 1). The results showed that a 1:4 (vol:vol) dilution of PDR vitreous fluid in cell culture medium provided an angiogenic stimulus in the HUVEC sprouting assay similar to that exerted by 30 ng/mL VEGF and that 2.0 µL/pellet of the same sample induce neovessel formation in the CAM in a manner similar to 100 ng/pellet of VEGF. Comparable results were obtained with three distinct pools from different patients with PDR (data not shown), whereas no angiogenic response was elicited by a pool of vitreous fluid samples from macular hole patients (Supplementary Fig. 4). As shown in Fig. 6A and B, K5-N,OS(H) hampered the proangiogenic activity exerted by pools of PDR vitreous fluid both in the HUVEC sprouting assay and in the CAM assay with ID₅₀ values equal to 1.0 ng/mL and 0.1 μg/pellet, respectively.

On this basis, vitreous fluids from 10 patients with PDR (Table 1) were individually tested in the HUVEC sprouting assay with or without 10 ng/mL K5-N,OS(H) (corresponding to 0.7 nmol/L final concentration) or 150 μg/mL bevacizumab (corresponding to 1.0 µmol/L final concentration). As shown in Fig. 6C, PDR vitreous fluid samples exert a different angiogenic response when their activity was normalized to that exerted by 30 ng/mL VEGF in the same assay. In 8 out of the 10 patients tested, K5-N,OS(H) caused a significant inhibition of the angiogenic activity of PDR vitreous fluid whereas bevacizumab was effective on three samples only, with an average percent of inhibition equal to 69 ± 8 and 33 ± 7% (P < 0.01) for the two compounds, respectively (Fig. 6C and D). Accordingly, the inhibitory action of K5-N,OS(H) was significantly more potent than bevacizumab in eight vitreous fluid samples (Fig. 6D). For both inhibitors, their efficacy was not related to the angiogenic potency of the PDR vitreous fluid tested nor to the clinical features of the enrolled patients with PDR (data not shown), including intravitreal pretreatment with VEGF blockers (Fig. 6E).

Here, we identified K5-N,OS(H) as a biotechnological HS-like molecule highly effective in inhibiting the activity of VEGF in different angiogenesis assays, including a murine model of OIR. Of note, K5-N,OS(H) also inhibits the angiogenic responses triggered by human vitreous fluid samples collected from patients with PDR, being more effective than the specific VEGF blocker bevacizumab. The remarkable efficacy of K5-N,OS(H) may be related to its capacity to act as a multitarget antagonist for a variety of heparin-binding proangiogenic/proinflammatory factors upregulated in PDR vitreous humor.

K5-N,OS(H) is a highly N,O-sulfated K5 derivative (28) with antiangiogenic, antineoplastic, and antiviral activities (26,27) but endowed with negligible anticoagulant activity when compared with heparin (26). Also, K5-N,OS(H) inhibits proinflammatory cytokine production by mononuclear cells (44). Accordingly, systemic administration of K5-N,OS(H) at doses as high as 1.0 mg/kg body weight exerts an anti-inflammatory activity in different models of ischemia/reperfusion injury in the absence of toxic effects (42,43).

Previous observations had shown the capacity of K5-N,OS(H) to bind FGF2 and HIV-Tat, thus inhibiting their angiogenic activity (26–28). Here, K5-N,OS(H) also binds with high affinity the major heparin-binding VEGF-A₁₆₅ isoform expressed in PDR (3). Indeed, given that VEGF-A₁₆₅ interaction with immobilized heparin occurs with a K_d value equal to 100 nmol/L (38), the competition experiments shown in Fig. 1C demonstrate that VEGF/K5-N,OS(H) interaction occurs with an 80-fold higher affinity, leading to a K_d value equal to ∼1.3 nmol/L. Accordingly, K5-N,OS(H) is more potent than unfractionated heparin in hampering the binding of VEGF to immobilized sVEGFR2 and in preventing the formation of HSPG/VEGF/VEGFR2 ternary complexes. Furthermore, in all these assays, K5-N,OS(H) was more active than the low sulfated derivatives K5-N,OS(L) and K5-NS whereas K5 was inactive (Table 1). Thus, the capacity of K5 derivatives to bind VEGF appears to depend on the backbone structure, charge distribution, and degree of sulfation of the GAG chain.

In keeping with these observations, K5-N,OS(H) inhibits different steps of the angiogenic process mediated by VEGF in vitro, ex vivo, and in vivo. Remarkably, intravitreal injection of K5-N,OS(H) also reduced the formation of preretinal neovascular tufts in a murine OIR model in which VEGFR2 mediates the proangiogenic activity of hypoxia-upregulated VEGF (41). In keeping with the aforementioned anti-inflammatory potential of K5-N,OS(H), this was paralleled by a significant downregulation of proinflammatory mediators in OIR retinas, with no effect on the expression of various angiogenic factors known to play a role in OIR, including VEGF. Of note, K5-N,OS(H) did not affect retinal health, as shown by the analysis of the levels of the apoptotic markers Bax, Bcl-2, and cytochrome c.

The OIR model recapitulates various aspects of proliferative retinopathies, including PDR (23). Indeed, although the metabolic changes that characterize PDR are not present in OIR mice, many features of human PDR are observed in the OIR model in which early vessel loss leads to neovascularization with consequent blood-retinal barrier breakdown and neuronal damage (23). K5-N,OS(H) inhibits the formation of preretinal neovascular tufts without affecting the extent of the avascular area in OIR retinas. The mechanisms that control the balance between pathological and healthy angiogenesis in the OIR model are not fully elucidated (47). Further experiments will be required to assess the impact of K5-N,OS(H) on healthy retinal regeneration.

The inhibitory effect of injected K5-N,OS(H) in the OIR model occurred at concentrations as low as 0.6 ng/μL (corresponding to an intravitreal concentration equal to 7.0 nmol/L when considering a vitreous volume of about 5.0 µL [48]) and was paralleled by a significant inhibition of VEGFR2 phosphorylation in OIR retinas, thus confirming the capacity of K5-N,OS(H) to act in vivo as a VEGF antagonist in the retinal microenvironment. Relevant to this point, previous observations had shown that intravitreal injection of heparin/HS or of a synthetic heparin analog reduces the angioproliferative changes in the OIR model (20,21). However, the effect occurred at intravitreal concentrations (equal to 1.0 and 10 mmol/L, respectively) >100,000-fold higher than the minimal K5-N,OS(H) effective dose.

The assessment of the angiogenic potential of the PDR vitreous humor may provide a useful tool for a preclinical evaluation of therapeutic angiostatic molecules (6,45–47). Here, PDR vitreous fluid elicits a potent angiogenic response in vitro and in vivo. Interestingly, significant differences in angiogenic potency were observed among the individual vitreous samples tested. These differences were not related to any of the clinical features of the enrolled patients with PDR (listed in Table 1), including intravitreal treatment with VEGF blockers performed 10–15 days prior to vitrectomy, presence of vitreous hemorrhage, or macular edema (Fig. 6E and data not shown). As anticipated, K5-N,OS(H) hampers the angiogenic activity exerted by PDR vitreous fluid pooled from different patients. Also, when tested on individual samples, the inhibitory effect of K5-N,OS(H) was more potent than that exerted by bevacizumab in 8 out of 10 patients with PDR that were examined. This occurred despite the fact that K5-N,OS(H) was tested at a final molar concentration 1,000-fold lower than the anti-VEGF antibody. Interestingly, we did not observe a significant relationship between the inhibitory effect exerted by bevacizumab or K5-N,OS(H) and the clinical features of patients with PDR, including intravitreal pretreatment with VEGF blockers (Fig. 6E). Further studies on a larger cohort of patients will be required to confirm these observations.

The remarkable efficacy shown by K5-N,OS(H) in reducing the angiogenic responses in the murine OIR model and in the endothelium treated with PDR vitreous fluid may be related to its capacity to bind and inhibit the biological activity of a variety of heparin-binding proinflammatory/angiogenic factors besides VEGF, including FGF2, IGF-1, IL8, PlGF2, PDGF-B, SDF1α, CTGF, HGF, and HMGB1, all present in the vitreous humor from patients with PDR (2–5). Of note, the potency of the K5-N,OS(H) interaction was unrelated to the affinity of these proteins for unfractionated heparin, indicating that distinct molecular determinants (e.g., differences in backbone structure and charge distribution) mediate their binding to the two GAGs. The possibility that the overall angiogenic potential of PDR vitreous humor may represent the result of the synergistic action of the various angiogenic factors is supported by the observation that the amount of recombinant VEGF required to induce in our assays an angiogenic response similar to that exerted by the pooled PDR vitreous fluid far exceeds the concentration of the growth factor detectable in PDR vitreous humor (median VEGF levels equal to 345 pg/mL [45]). The better performance of K5-N,OS(H) versus bevacizumab in inhibiting the angiogenic activity of PDR vitreous fluid substantiates this hypothesis.

In conclusion, our data provide compelling experimental evidence that K5-N,OS(H) represents a novel multitarget molecule with potential implications for more efficacious therapeutic interventions aimed at inhibiting the neovascular responses that occur in the retina of patients with PDR.

Acknowledgments. The authors thank I. Fornaciari (University of Pisa) for technical assistance and D. Rusciano (BIOOS Italia), M. Ziche (University of Siena), and G. Casini (University of Pisa) for helpful discussion and criticisms. CHO-K1 and A745 CHO cells were provided by J.D. Esko (University of California, La Jolla, CA). Human recombinant VEGF and pcDNA3 harboring the ECD-VEGFR2-YFP cDNA were provided by K. Ballmer-Hofer, and PlGF2 was provided by S. De Falco (IGB, Naples, Italy).

Funding. This work was supported in part by grants from Istituto Toscano Tumori to L.M., Regione Toscana to P.B., Ministero dell’Istruzione, Università e Ricerca (FIRB project RBAP11H2R9 2011), and Associazione Italiana per la Ricerca sul Cancro (AIRC grant 14395) to M.P., and BIOOS Italia to L.M., P.B., and M.P.

Duality of Interest. P.O. is an employee of Glycores 2000 and inventor of the patent EP1366082 and its international counterparts. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. S.R. designed and performed experiments, analyzed data, and wrote the manuscript. M.D.M., M.B., A.B., P.C., M.Co., M.Ca., A.C., L.M., and P.O. designed and performed experiments and analyzed data. P.B. and F.S. analyzed data, contributed to discussion, reviewed the manuscript, and secured funding. M.P. supervised the project, designed experiments, analyzed data, wrote the manuscript, and secured funding. M.P. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.