Human IgG1 antibodies suppress angiogenesis in a target-independent manner

Aberrant angiogenesis is implicated in diseases affecting nearly 10% of the world’s population. The most widely used anti-angiogenic drug is bevacizumab, a humanized IgG1 monoclonal antibody that targets human VEGFA. Although bevacizumab does not recognize mouse Vegfa, it inhibits angiogenesis in mice. Here we show bevacizumab suppressed angiogenesis in three mouse models not via Vegfa blockade but rather Fc-mediated signaling through FcγRI (CD64) and c-Cbl, impairing macrophage migration. Other approved humanized or human IgG1 antibodies without mouse targets (adalimumab, alemtuzumab, ofatumumab, omalizumab, palivizumab and tocilizumab), mouse IgG2a, and overexpression of human IgG1-Fc or mouse IgG2a-Fc, also inhibited angiogenesis in wild-type and FcγR humanized mice. This anti-angiogenic effect was abolished by Fcgr1 ablation or knockdown, Fc cleavage, IgG-Fc inhibition, disruption of Fc-FcγR interaction, or elimination of FcRγ-initated signaling. Furthermore, bevacizumab’s Fc region potentiated its anti-angiogenic activity in humanized VEGFA mice. Finally, mice deficient in FcγRI exhibited increased developmental and pathological angiogenesis. These findings reveal an unexpected anti-angiogenic function for FcγRI and a potentially concerning off-target effect of hIgG1 therapies.

shows that the levels of endogenous mouse IgG2c in wild-type mouse corneas, retinae, or hind limb muscles (following whole body perfusion to remove blood), one day after injury, are a very small fraction of serum IgG2c levels, approaching the levels observed in the serum of Rag2 -/mice, which lack B and T lymphocytes. . Levels normalized to mean serum IgG2c levels in C57BL/6J mice. (b) ELISA shows that in wild-type mouse corneas, retinae, or hind limb muscles treated with bevacizumab, levels of exogenous human IgG1 (following whole body perfusion to remove blood), at one day after injection, are much higher than levels of endogenous mouse IgG2c. n = 6-7 (a, b). (c) Representative images of wild-type mouse retina shows immunolocalization of endogenous mouse IgG (red) that is abundant and spatially confined to the retinal blood vessels (CD31, green), and nearly undetectable in the extravascular tissue. Representative image of wild-type mouse retina probed with isotype control antibody shows specificity of IgG staining.

In vivo transfection and site directed mutagenesis of plasmids. Subretinal injections in mice
were performed using a 37-gauge Exmire microsyringe (Ito Corporation) 13 . In vivo transfection of plasmids was achieved using 10% Neuroporter (Genlantis, San Diego, CA) for subretinal injection. The pFUSE-hIgG1-Fc2, pFUSE-hIgG2-Fc2, pFUSE-mIgG2a-Fc2, and pFUSE-mIgG2b-Fc2 vectors (Invivogen, San Diego, CA), which encode Fc-fusion proteins expressing the Fc region (CH2 and CH3 domains) of the human IgG1, human IgG2, mouse IgG2a, and mouse IgG2b heavy chains, respectively, and the hinge region, and contain the IL2 signal sequence to enable secretion, as well as pFUSE-hIgG1e3-Fc2 and pFUSE-mIgG2ae1-Fc2 (Invivogen), which contain point mutations that reduce binding affinity for FcγRI, were injected into the subretinal space or into the corneal stroma. Using a site directed mutagenesis kit (Agilent Technologies, Santa Clara, CA), we engineered, and sequence-verified, the K322A point mutation into pFUSE-hIgG1-Fc2 to eliminate binding to C1q. A null control plasmid was created by digesting pFUSE-hIgG1-Fc2 with ApoI and religating the plasmid backbone to create an empty vector without the hIgG or mouse IgG coding sequences.

Mutant alemtuzumab creation and binding. Mutant forms of alemtuzumab (G1Δab and
G1Δa) that have identical, humanized, CD52-specific variable domains to alemtuzumab 14 were created as previously described 15  were produced by solid-phase peptide synthesis by using amino acids in the D configuration 16 .
Purified peptides were dissolved at the working concentration in PBS.
Corneal angiogenesis. Two interrupted 11-0 nylon sutures (Mani, Utsunomiya, Japan) were placed into the corneal stroma, midway between the central corneal apex and the limbus (approximately 1.25 mm from the limbus), of both eyes of mice as previously described 17,18 . On day 10 after injury, we calculated the mean percentage CD31 + Lyve-1blood vessel areas for corneal flat mounts with ImageJ (US National Institutes of Health, Bethesda, MD) as previously reported 17,18 . Eyes were excluded in masked fashion from analyses (2% incidence) if anterior chamber hemorrhage suture loosening, or cataract formation occurred at any time during the study period.
Choroidal angiogenesis. Laser photocoagulation (OcuLight GL, IRIDEX, Mountain View, CA) was performed on both eyes of mice to induce CNV as previously described 19,20 . Choroidal angiogenesis volumes were measured by scanning laser confocal microscopy (TCS SP5, Leica, Solms, Germany) 7 days after injury as previously reported with 0.5% FITC-conjugated Isolectin B4 (Vector Laboratories, Burlingame, CA) 19,20 . Laser lesions were excluded in masked fashion from analyses (5% incidence) if laser photocoagulation did not induce a bubble, if it induced hemorrhage, or if lesions became confluent with one another.
Hind limb ischemia angiogenesis. Mice were anesthetized before undergoing unilateral proximal femoral artery ligation. The right femoral artery was gently isolated, ligated and excised distal to the deep femoral artery and 0.5 cm proximal to the bifurcation in saphenous and popliteal arteries, as previously described 21 . The non-ischemic left limb underwent sham surgery without arterial ligation. On day 7 after surgery, both anterior and posterior muscles from ischemic and non-ischemic hind limbs were harvested and processed for immunohistochemical analysis for vessel quantification. Animals were excluded in masked fashion from analyses (0% incidence) if hemorrhagic death from improper ligation occurred.
Immunostaining of mouse tissue. For staining of eye tissues, eyes were dissected 3 days after injury, cryoprotected in 30% sucrose, embedded in optimal cutting temperature compound (Tissue-Tek OCT; Sakura Finetek, Torrance, CA), and cryosectioned into 10 µm sections. For mouse IgG, immunofluorescent staining was performed with anti-mouse IgG (1:5000; Sigma). For the negative control staining, the primary antibody was omitted. For macrophage staining in the cornea or choroid, immunofluorescent staining was performed with rat antibody against F4/80 (1:50; AbD Serotec, Raleigh, NC) and, for CNV, rabbit antibody against CD45 (1:100; Abcam, Cambridge, MA). Isotype IgG was substituted for the primary antibody to assess the specificity of the staining. Bound antibody was detected with Alexa fluor-conjugated secondary antibodies. Sections were mounted with ProLong Gold Antifade reagent with DAPI (Life Technologies). In the cornea, a region of 600 µm starting at the limbus and including the whole width of the cornea was delineated, and positive cells within that area were counted. In the choroid, the area within and around the injury site that stained with F4/80 and CD45 antibodies was quantified (Adobe Photoshop, CS6.0; Adobe Systems, San Jose, CA) to obtain a quantitative index of macrophages, as described previously [22][23][24]  Cell stimulation and western blotting. Serum starved Py4 mouse blood endothelial cells were stimulated with 50 ng/ml of mouse Vegfa or human VEGFA for 10 min to assess Vegfr2 phosphorylation. 0.1 mg/ml bevacizumab (Genentech) or 0.75 µg/ml neutralizing anti-mouse Vegfa mAb (R&D Systems) were added to medium at the same time. Serum starved J774 mouse macrophages were stimulated with 0.1 mg/ml of bevacizumab or human IgG1 to assess c-Cbl phosphorylation. Serum starved RAW 264.7 mouse macrophages (ATCC, Manassus, VA) were stimulated with 0.1 mg/ml bevacizumab to assess Vegfr1 abundance. Bone marrow derived mouse macrophages (BMDMs) were prepared as previously described 26 . Bone marrow cells were cultured for 5 days in Iscove's modified Dulbecco's medium supplemented with 30% L929 cell supernatant containing sodium pyruvate, 10% heat-inactivated FBS (Life Technologies), non-essential Amino Acids (Sigma) and antibiotics. The differentiated BMDMs were serum starved and incubated with bevacizumab (0.1 mg/ml) for various time points. Human primary peripheral blood monocytes (Stemcell Technologies, Vancouver, Canada) and human monocytic were performed using Synergy 4 Microplate Reader and Gene5 software (BioTec, Suffolk, United Kingdom).

FcγRI interaction and phosphorylation in vivo.
Bevacizumab and denosumab were biotinylated using the EZ-Link Sulfo-NHS-LC-Biotinylation Kit (Thermo Scientific) by following the manufacturer's instructions. After the reaction, by using the HABA biotin assay (Pierce Biotin Quantitation Kit; Thermo Scientific), it was found that each protein molecule had been modified by ~ 3 biotin molecules. Biotinylated bevacizumab or denosumab (25 mg/ml; 4 µl) was injected into wild-type mouse corneas immediately following suture injury. Two days later, corneas were excised and cell lysates were prepared by pooling 4 corneas in each group.
Equal amounts of total corneal protein (400 ug) from each group were subjected to "pull-down" with Dynabeads® M-280 Streptavidin (Life Technologies) for 2 h at RT and eluted it by 1X protein sample buffer for western blotting. The blot was probed with anti-mouse Fcgr1 antibody (clone X54-5/7.1, BioLegend, San Diego, CA) overnight at 4 °C. The same blot was reprobed with HRP-conjugated Avidin (BioLegend). Bevacizumab (25 mg/ml; 4 µl) or PBS was injected into FcγR humanized mouse corneas immediately following suture injury. Two days later, corneas were excised and cell lysates were prepared by pooling 8 corneas in each group. Equal amounts of total corneal protein (300 µg) from each group were pre-cleared with Protein A/G Magnetic Beads (Pierce Classic Magnetic IP/Co-IP Kit, Thermo Scientific) for 1 h at RT to remove non-specific binding proteins. The pre-cleared samples were incubated with 10 µg of anti-human FcγRI antibody (clone 32.2, Acris Antibodies, San Diego, CA) overnight at 4 °C with gentle shaking. The immune complex was then precipitated with Protein A/G Magnetic Beads for 2 h at 4 °C, collected by magnets and washed three times with ice-cold IP Lysis/Wash Buffer and once with purified water, and eluted by 1× protein sample buffer for western blotting.
The blot was reprobed with anti-human FcγRI antibody to assess equal loading. The blot was probed with anti-phosphotyrosine antibody (Cell Signaling). Statistical analyses. Choroidal angiogenesis volumes per laser lesion were compared by hierarchical logistic regression using repeated measures analysis as previously described 11,12 . We used the Mann-Whitney U test with Bonferroni correction for statistical comparison of multiple