Long-acting protein drugs for the treatment of ocular diseases

Abstract

Protein drugs that neutralize vascular endothelial growth factor (VEGF), such as aflibercept or ranibizumab, rescue vision in patients with retinal vascular diseases. Nonetheless, optimal visual outcomes require intraocular injections as frequently as every month. Here we report a method to extend the intravitreal half-life of protein drugs as an alternative to either encapsulation or chemical modifications with polymers. We combine a 97-amino-acid peptide of human origin that binds hyaluronan, a major macromolecular component of the eye's vitreous, with therapeutic antibodies and proteins. When administered to rabbit and monkey eyes, the half-life of the modified proteins is increased ∼3-4-fold relative to unmodified proteins. We further show that prototype long-acting anti-VEGF drugs (LAVAs) that include this peptide attenuate VEGF-induced retinal changes in animal models of neovascular retinal disease ∼3-4-fold longer than unmodified drugs. This approach has the potential to reduce the dosing frequency associated with retinal disease treatments.

Figure 1. Schematic of LAVAs.

The HA-binding peptide (HABP) was fused to the C-terminal end of scFvs, Fabs, Fc Traps and IgGs using a single GSGGG linker. The Fc traps and IgGs have two copies of the HABP, whereas the scFv and Fab have one copy.

Figure 2. Duration of activity of anti-VEGF molecules in the rabbit retinal leakage model.

Two days before fluorescein angiography, retinal vessel permeability was induced by an intravitreal injection of 10 pmol per eye hVEGF-A₁₆₅. hVEGF-induced fluorescein leakage was measured in six to eight rabbit eyes for each anti-VEGF molecule and compared with control eyes that received intravitreal hVEGF but no anti-VEGF molecule. Each data point represents the average inhibition of vascular leakage compared with controls. LAVA1 inhibited fluorescein leakage when administered up to 28 days prior the hVEGF challenge (30 days before fluorescein angiography), whereas equimolar doses of ranibizumab, bevacizumab, aflibercept, brolucizumab or DARPin-PEGOL did not substantially inhibit fluorescein leakage when administered 18–21 days before the hVEGF challenge (20–23 days before fluorescein angiography). Data shown are from several independent studies.

Figure 3. Efficacy of NVS0 and LAVA1 in the cynomolgus monkey laser-CNV model.

CNV lesions were induced 4–9 months before dosing. Representative images of cynomolgus monkey eyes injected with NVS0 or LAVA1. The fundus images are shown as negatives (black and white reversed) to highlight the vascular leakage of fluorescein during fluorescein angiography. Each eye received four laser burns producing neovascular tufts (CNV) that appear as dark spots in the images; the darkness corresponds to the degree of vascular leakage from the lesions. The asterisks ‘*' in the fundus images correspond to the lesion from which retinal thickness was measured by optical coherence tomography (OCT). Total retinal thickness (values shown in the bottom of each OCT scan) was measured at the center of the lesion using calipers provided in the Spectralis software as the distance between the inner limiting membrane and retinal pigment epithelium (RPE) as shown by the arrow heads in the OCT of 362081 left eye. Both NVS0 and LAVA1 markedly reduced CNV-induced retinal oedema within 4 weeks. The plot shows the number of grade IV lesions at the time of dosing, 1 week and 4 weeks post dosing with NVS0 or LAVA1. Both NVS0 and LAVA1 reduced the number of grade IV lesions and retinal oedema within 4 weeks of injection.

Figure 4. Efficacy of ranibizumab and LAVA1 in the cynomolgus monkey laser-CNV model.

Laser photocoagulation to induce CNV lesions was applied 1–5 weeks after drugs were dosed. Late-phase fundus images from cynomolgus monkey eyes injected with ranibizumab or LAVA1 taken approximately 5 min post-intravenous fluorescein injection. The fundus images are shown as negatives (black and white reversed) to highlight the CNV-induced vascular leakage of fluorescein during fluorescein angiography. Cynomolgus monkeys 1–9 were imaged 22 days after dosing, whereas cynomolgus monkeys 10–15 were imaged 49 days after dosing. Each eye received four laser burns producing neovascular tufts (CNV) that appear as dark spots in the images; the darkness corresponds to the degree of vascular leakage from the CNV lesions. When dosed 8 days before laser, a single dose of ranibizumab only partially inhibited laser-induced CNV, whereas a single dose of LAVA1 completely inhibited laser-induced CNV. The plot (bottom left) shows average terminal levels of free (unbound to VEGF) LAVA1 and free ranibizumab in the cynomolgus monkey eyes 22 and 49 days post-intravitreal injection. Drug concentrations less than the lower limit of quantification (0.1 ng per eye) were categorized as ‘not detected'. At both day 22 and 49, free LAVA1 levels in the vitreous, retina and RPE-choroid were markedly higher than free ranibizumab. Error bars represent 1 s.d. Measurements were done in triplicates.

Figure 5. Distribution of average CNV lesion grades.

In the cohort of monkeys that were imaged 22 days post dosing, eyes that received ranibizumab had six grade IV lesions, whereas none of the eyes that received LAVA1 had grade IV lesions (all lesions were grade I). In the cohort of monkeys that were imaged 49 days post dosing, eyes that received ranibizumab had 18 grade IV lesions, whereas eyes that received LAVA1 had 13 grade IV lesions.

Figure 6. Efficacy of engineered variants of LAVA1 in the rabbit retinal vascular leakage model.

Variants of LAVA1 that were engineered to be resistant to proteolysis (LAVA2, 3, 45 and 46) all potently inhibited fluorescein leakage and had terminal vitreal levels similar to LAVA1 when administered 20 days before fluorescein angiography. In contrast, non-glycosylated variants of LAVA1 (LAVA24 and LAVA25) minimally inhibited fluorescein leakage. The numbers in boxes above the datapoints are the mean terminal drug levels in the vitreous. LAVA24 and LAVA25 had terminal levels that were 39-fold lower than LAVA1. Terminal vitreal levels measured in ng ml⁻¹ were converted to pmoles per eye using a molecular weight of 59 kDa for all LAVAs. Vitreal volume in rabbit was assumed to be 1.25 ml.

Figure 7. Efficacy of ranibizumab and LAVA2 in the rabbit retinal vascular leakage model.

LAVA2 at a dose of 210 pmol demonstrated similar efficacy to ranibizumab doses of 6,300 or 10,500 pmol when administered intravitreally 30 days before fluorescein angiography. The terminal vitreal levels (values in brackets above the datapoints) of LAVA2 (8.9 pmol) are higher than the terminal vitreal levels of ranibizumab at all doses (2.4 to 6.2 pmol) even though ranibizumab doses were 22.5- to 50-fold higher than LAVA2 doses. Terminal vitreal levels measured in ng ml⁻¹ were converted to pmoles per eye using a molecular weight of 49 kDa for ranibizumab and 59 kDa for LAVA2. Vitreal volume in rabbit was assumed to be 1.25 ml.