Fruquintinib inhibits VEGF/VEGFR2 axis of choroidal endothelial cells and M1-type macrophages to protect against mouse laser-induced choroidal neovascularization

Wet age-related macular degeneration, which is characterized by choroidal neovascularization (CNV) and induces obvious vision loss. Vascular endothelial growth factor (VEGF) family member VEGF-A (also named as VEGF) and its receptor VEGFR2 contribute to the pathogenesis of CNV. Choroidal endothelial cells (CECs) secret C–C motif chemokine ligand 2 (CCL2), which attracts macrophages to CNV lesion and promotes macrophage M1 polarization. Accordingly, infiltrating macrophages secret inflammatory cytokines to promote CNV. In vivo, intravitreal injection of fruquintinib (HMPL-013), an antitumor neovascularization drug, alleviated mouse CNV formation without obvious ocular toxicity. Meanwhile, HMPL-013 inhibited VEGF/VEGFR2 binding in CECs and macrophages, as well as macrophage M1 polarization. In vitro, noncontact coculture of human choroidal vascular endothelial cells (HCVECs) and macrophages under hypoxia conditions was established. HMPL-013 downregulated VEGF/VEGFR2/phosphoinositide-3-kinase/protein kinase B (AKT)/nuclear factor kappa B pathway and CCL2 secretion in HCVECs, as well as VEGF/VEGFR2-induced macrophage M1 polarization under hypoxia condition. In addition, HMPL-013 inhibited HCEVC derived CCL2-induced macrophage migration and M1 polarization, along with macrophage M1 polarization-induced HCVECs proliferation, migration, and tube formation. Altogether, HMPL-013 alleviated CNV formation might via breaking detrimental cross talk between CECs and macrophages.


Introduction
Age-related macular degeneration (AMD), a leading cause of incurable vision loss in the elder people and accounting for 8.7% of all cases of blindness in the developed nations 1 , is categorized into dry and wet types. Dry AMD is featured by multiple drusen deposits and rarely impacts vision, developing not only to geographic atrophy but also to wet AMD, which is characterized by choroidal neovascularization (CNV) and induces obvious vision loss. Vascular endothelial growth factor (VEGF) family members, containing VEGF-A (also named as VEGF), VEGF-B, VEGF, VEGF-D, VEGF-E, and placental growth factor (PGF), promote CNV via binding to their respective receptors vascular endothelial growth factor receptor 1 (VEGFR1), VEGFR2, and VEGFR3. Intravitreal injection of anti-VEGF reagents is deemed to be the optimal treatment for CNV. However, any improvement is accompanied by long-term monthly intravitreal injections and ocular complications, such as endoophthalmitis 2 .
Therefore, the searching of cellular and molecular mechanisms of CNV is warranted. Clinical studies have shown that age-related changes in Bruch's membrane lead to choriocapillaris atrophy, as well as to decreased diffusion of oxygen toward the neuroretina. The resulting outer retina hypoxia may be an important driving force of CNV formation, by stimulating VEGF overexpression via the transcription factor hypoxia-inducible factor 1α (HIF-1α) in the retinal pigment epithelium (RPE) cells, and retinal Muller cells 3 . VEGF binds to its receptor VEGFR2, consequently phosphorylates Tyr1175 inside VEGFR2, finally promotes laser-induced CNV formation in mice 4,5 . Anti-VEGF-A/VEGFR2 or nonspecific small interfering RNA inhibits CNV and attenuates VEGF mRNA expression in a mouse laser-induced CNV model 6 . In addition, resveratrol inhibits HIF-1α accumulation and VEGF secretion induced by cobalt chloride through sirtuin 1 in human RPE cells 7 . In addition, in AMD-relevant models, VEGF/ VEGFR2 blockade does not cause retinal atrophy, which is a side effect caused by intraocular injections of VEGFneutralizing proteins 8 . These studies attract us to investigate the suppression of VEGF/VEGRR2 axis in CNV.
Accumulating study reveals infiltrating macrophages contribute to the progress of CNV. In different environments, macrophages polarize into M1 pro-inflammatory and M2 anti-inflammatory types. M1-type macrophages with pro-inflammatory functions can produce VEGF and promote neovascularization 9 , while it has higher transcript ratio of M1 chemokine C-X-C motif chemokine ligand 11 (CXCL11) to M2 chemokine C-X-C motif chemokine ligand 22 (CXCL22) in advanced AMD maculae compared to the control 10 . Besides, M1-type macrophages secret pro-inflammatory cytokines, such as interleukin-6 (IL-6), tumor necrosis factor-α (TNF-α), and C-C motif chemokine ligand 5 (RANTES) to promote ocular neovascularization 11 . C-C motif chemokine ligand 2 (CCL2) produced and released by choroidal endothelial cells (CECs) draws macrophages with CCL2 receptor CCR2 on the surfaces of macrophages to the site of CNV injury 12 . In addition, CCL2 facilitates macrophage M1 polarization 13 .
Fruquintinib (HMPL-013) is an antitumor neovascularization drug, which belongs to tyrosine kinase inhibitors, acting as a powerful and highly selective inhibitor for all types of VEGFRs, including VEGFR1, VEGFR2, and VEGFR3. The researchers have completed phase III clinical trials of colorectal cancer (CRC), and the results show that in patients with metastatic CRC who have received at least two chemotherapy regimens, oral HMPL-013 significantly improves the overall survival rate of the patients compared to the placebo group 14 . Following China's priority review of HMPL-013 in September 2017, on September 4, 2018, the National Medical Products Administration granted HMPL-013 for the first global approval for the treatment of progressive CRC 15 . Therefore, the question whether HMPL-013 can alleviate CNV attracts our attention.
Herein, mouse laser-induced CNV and in vitro endothelial cell and macrophage hypoxia models were applied to identify the functions and mechanisms of HMPL-013 on CNV. Our study could supply a potential therapeutic strategy for the treatment of wet AMD.

Mouse laser-induced CNV model and treatment
Nine to 10-week-old male C57BL/6 mice were purchased from the Laboratory Animal Center of Nantong University (Nantong, China). The mouse laser-induced CNV model was constructed, as previous description 16 . At the time of laser photocoagulation, the production of a bubble was regarded as a rupture of the Bruch's membrane, indicating that the model was successfully established. Photocoagulation spots containing hemorrhage or failing to develop a bubble at the laser site were excluded. The mice were randomly assigned into five groups: normal, CNV 7 d, CNV 7 d + 1 μl of 0.1% dimethyl sulfoxide (DMSO), CNV 7 d + 1 μl of HMPL-013 (Elunate ® ; Chi-Med, China; 5 μg/μl in 0.1% DMSO), and CNV 7 d + 1 μl ranibizumab (RBZ; Lucentis; Genentech Inc.; 10 μg/μl; used as the positive control), which is a recombinant humanized monoclonal antibody fragment binding VEGF-A. In the animal experiments, the investigator responsible for all other experiments except for CNV model construction were blind to the group allocation.

Fundus angiography
Fundus fluorescein angiography (FFA) and indocyanine green angiography (ICGA) in mice were done following previous study 18 . For the grading of CNV leakage, two masked researchers not involved in laser photocoagulation or angiography evaluated the fluorescein angiograms at a single sitting. Grade 0 lesions had no hyperfluorescence. Grade 1 lesions exhibited hyperfluorescence without leakage. Grade 2A lesions exhibited hyperfluorescence in the early or midtransit images and late leakage. Grade 2B lesions showed bright hyperfluorescence in the transit images and late leakage beyond treated areas (grade 2B lesions were defined as clinically significant), as previous description 19 . For each ICGA examination, the entire lesion area was quantitatively measured using the Heidelberg software (Spectralis Acquisition and Viewing Modules; version 3.2, Heidelberg Engineering) by three independent observers. Mean observed values were calculated.

Choroidal flat mounts and immunofluorescence
Choroidal flat mounts and immunofluorescence were performed according to previous methods 20

Hematoxylin-eosin stain
On day 7, following euthanasia, the mouse eyes were enucleated and immersion-fixed 10 in 4% PFA for 2 h. After fixation, the eyes were embedded in Tissue-Tek ® optimum cutting temperature compound (#4583, Sakura Finetek, Japan), and cross-sectioned on a cryostat vertically through the center of the cornea and optic nerve. Slide of 5 µm thickness was stained with hematoxylin-eosin (HE).

Terminal deoxynucleotidyl transferase dUTP nick-end labeling
After fixation and permeabilization, the mouse cryosections were incubated with a terminal deoxynucleotidyl transferase dUTP nick-end labeling (TUNEL) reaction mixture (#11684795910, Roche, Switzerland) at 37°C for 60 min, and DAPI for 5 min at room temperature (RT), and then washed for 30 min in PBS.

Electroretinography
The mice were dark-adapted over 16 h. Then the mice were anesthetized and their pupils were dilated. Contact lens electrodes were placed on both eyes with a drop of methylcellulose. Full-field electroretinographies (ERGs) were recorded by using the universal testing and electrophysiologic system 2000 (UTAS E-2000, LKC Technologies). The responses were recorded at a gain of 2 k using a notch filter at 60 Hz, and were bandpass filtered between 0.1 and 1500 Hz. In the light-adapted photopic state, with a −1.02 log cds/m 2 background light (flash intensity) to desensitize the rods and isolate cones, photopic cone responses were recorded in response to a single flash of 0 dB. The amplitude of the a-wave was measured from the baseline to the lowest negative-going voltage, whereas peak b-wave amplitudes were measured from the trough of the a-wave to the highest peak of the positive b-wave.

Co-immunoprecipitation
Proteins of mouse retina-RPE-choroid complex tissues from normal, CNV 7 d, CNV 7 d + DMSO, and CNV 7 d + HMPL-013 groups were obtained by the protein extraction kit (Merck) and precleared. The beads coated with p-VEGFR2 or VEGF antibody were incubated with the precleared whole proteins 4°C for overnight. The beads were washed with cell lysis buffer four times. Finally, the beads were boiled for 10 min. The eluents were analyzed by Western blot with VEGF and p-VEGFR2 antibodies.

Cell noncontact coculture and treatment
Human choroidal vascular endothelial cells (HCVECS; #36052-03, Celprogen, USA) were cultured in Dulbecco's Modified Eagle Medium (DMEM) containing 4.5 g/l glucose supplemented with 10% (v/v) FBS, 100 U/ml penicillin, and 100 mg/ml streptomycin. The human macrophages were derived from human peripheral blood mononuclear cells and cultured, as the previous description 22 . The cells were kept at 37°C in a humidified atmosphere containing 5% CO 2 . HCVECs were cultured in the lower well and macrophages were cultured in the upper well of the transwell plate (#CLS3397, Corning, USA) and verified by STR profiling.

Transwell assay
After the coculture with HCVECs or macrophages for 24 h, the macrophages or HCVECs were collected and seeded into the upper chamber (8 µm) at a density of 1 × 10 5 cells/well (Corning) with non-serum DMEM medium. The lower chamber was filled with 500 µl DMEM supplemented with 10% FBS. Twelve hours later, the human macrophages or HCVECs on the upper surface of the membrane were removed with a cotton swab. Then, the lower cells were fixed with formaldehyde and stained by crystal violet for 30 min. The number of migrated cells was counted under a microscope.

5-Ethynyl-2-deoxyuridine assay
The 5-ethynyl-2-deoxyuridine (EdU) incorporation assay was conducted with an EdU kit (#C10310-1, Ribo-Bio, China) according to the manufacturers' instructions. The cell nuclei were counterstained with DAPI. The EdUpositive ratio was calculated as the cells. The number of cells was counted using Image-Pro Plus software (Media Cybernetics, USA).

Tube formation assay
HCVECs tube formation was analyzed using extracellular matrigel vessel-like formation assay. Firstly, precooled growth factor reduced matrigel (#354230, Corning) was coated in the bottom of 48-well plate, 300 µl/well and incubated in a humidified atmosphere of 5% CO 2 at 37°C for 1 h. After coculture with macrophages for 24 h, HCVECs were harvest and 1 × 10 5 cells/ well was seeded to the coated plate. Then, 20 ng/ml human VEGF-A recombinant protein (#ab55566, Abcam) was added to each well after cells were seeded in triplicates. After culture for 24 h, tube formation was visualized by Olympus microscope (Japan), and total tube length were analyzed using Image-Pro Plus software.

Statistical analysis
The data were shown as mean ± SEM. Statistical analysis was performed by the one-way ANOVA followed by Tukey's test. P < 0.05 was considered statistically significant. Analysis was done by statistical software SPSS 15.0.

HMPL-013 mitigates mouse laser-induced CNV formation
To explore the effects of HMPL-013 on mouse CNV formation, HMPL-013 intravitreal injection was done at day 3 after laser coagulation, and analysis was done at day 7 (Fig. 1a). The concentration-time profiles of HMPL-013 in the mouse retina-RPE-choroid complex tissues after [ 14 C] HMPL-013 intravitreal injection showed that HMPL-013 reached its peak concentration at 4 h (13649.21 ng/ml ± 1024.80 ng/ml) and the high concentration sustained until 20 h (Fig. 1b). CNV leakage was alleviated in the HMPL-013 and RBZ groups compared to the CNV 7 d group (Fig. 1c). The leakage score analysis further showed that the percentage of 0 and 1 scores increased, while the percentages of 2a and 2b scores decreased, in the HMPL-013 and RBZ groups compared to the CNV 7 d group (Fig. 1d). CNV area also decreased in HMPL-013 and RBZ groups (Fig. 1e). Accordingly, the mean intensity values showed that CNV leakage decreased in the HMPL-013 and RBZ groups (Fig. 1f). In addition, staining of IB4 (a vascular endothelial cell marker) and phalloidin was performed on choroidal flat mounts, indicating that HMPL-013 alleviated CNV formation (Fig. 1g, h). The results suggested that HMPL-013 intravitreal injection reduced CNV leakage and area.

HMPL-013 causes no obvious intraocular toxicity
Next, on the basis of the efficacy of HMPL-013 on the mouse CNV, we wondered whether HMPL-013 could exert ocular side effects. HE stain on retinal cryosections (Fig. 2a) and quantification of retinal thickness (Fig. 2b) revealed no differences in histologic morphology or retinal thickness between the normal and HMPL-013-injected eyes. In addition, ocular cell apoptosis was unchanged in the HMPL-013 group compared to the normal and CNV 7 d groups (Fig. 2c, d). In CNV 7 d group, a-wave and b-wave amplitudes decreased compared to normal group, while HMPL-013 elevated a-wave and b-wave amplitudes ( Fig. 2e-g), indicating that HMPL-013 improved scotopic response in mice with CNV. These data suggested that HMPL-013 intravitreal injection caused no obvious intraocular toxicity.

HMPL-013 inhibits macrophage M1 polarization-induced HCVECs proliferation, migration, and tube formation
Finally, we investigated the effects of pro-inflammatory cytokines derived from macrophages on the proangiogenic behaviors of HCVECs. HCVECs proliferation was induced by hypoxia, reduced by macrophage polarization modulator geraniin or HMPL-013. The downregulatory role of HMPL-013 on the proliferation of HCVECs was weakened by macrophage M1-type polarization agonist LPS (Fig. 8a, b). Similarly, HMPL-013 played the inhibitory role on hypoxia-induced HCVECs migration (Fig. 8c, d) and tube formation (Fig. 8e, f). The results suggested that HMPL-013 inhibited macrophage M1 polarization, consequently suppressed HCVECs proliferation, migration, and tube formation. To sum up, HMPL-013 mitigated mouse CNV formation via inhibiting VEGF/VEGFR2 binding in CECs and macrophages, consequently blocking the detrimental cross talk between these two kinds of cells (Fig. 8g).
type VEGF family members to their receptors and downstream signaling pathways. We also found HMPL-013 at the dose of 5 μg alleviated mouse CNV leakage and area, even better than RBZ at the dose of 10 μg. To date, HMPL-013 has been progressed in clinical trial for advanced non-small cell lung cancer 28 and advanced gastric cancer 29 . Whether HMPL-013 can be used for the treatment of CNV needs further investigation.
VEGF and VEGFR2 expression, and their binding in CECs and macrophages.
Previous study reveals the M1-associated cytokines increased to a greater extent in the RPE/choroid complexes, whereas the M2-associated cytokines were highly expressed in the retinas 33 . While human advanced AMD macula has a higher M1 to M2 chemokine transcript ratio compared to normal autopsied eyes 10 . Study using transgenic mice advance inflammatory M1 phenotype monocyte (CCR2 + ) infiltrate as the drive for experimental CNV 34 . However, another study illuminates that VEGF + Arg1 + macrophages drive the onset of CNV in mice 35,36 . These contradictory results about the roles of M1 and M2 macrophages in the pathogenesis of CNV potentially due partially to the complex and kinetic microenvironment that governs macrophage polarization and function 37,38 . Hereon, we found HMPL-013-mitigated CNV formation via inhibiting macrophage M1 polarization.
The mRNA expressions of M1-related markers are dramatically upregulated in the early stage, while the M2related markers are slightly upregulated in the middle stage and sustained until the late stage in the aqueous humors of wet AMD patients 39 . In the study, the mouse choroidal flat mounts on day 7 following laser treatment were used for the detection of macrophage polarization, showing that both of M1-and M2-type markers increased. The difference might attribute to the different species and tissues. VEGFR1 knockdown inhibits MCP-1 (CCL2) expression of clear cell renal cell carcinoma cells 40 . It has been reported that MCP-1 (CCL2) via nuclear factor kappa β in bovine retinal endothelial cells 41 . HMPL-013 is a potent inhibitor for all kinds of VEGFRs. Therefore, we speculate that HMPL-013 inhibits the binding of VEGFs to VEGFRs in HCECs to downregulate the expression of CCL2. Thereafter, downregulated CCL2 restrain the infiltration of macrophages, resulting in the decreased expression of CCR2.
In summary, HMPL-013 ameliorated mouse CNV formation via inhibiting VEGF/VEGFR2 binding in CECs and macrophages, thereby blocking the detrimental cross talk between these two kinds of cells. However, the study still left certain questions needed for further exploration, such as the inhibitory roles of HMPL-013 on other kinds of VEGF family members, and dynamic observation for HMPL-013 treatment on the progress of mouse CNV.