Hypoxia-Inducible Factor-1α Is Associated With Sprouting Angiogenesis in the Murine Laser-Induced Choroidal Neovascularization Model

Abstract

Purpose: To investigate the expression and distribution of neoangiogenic molecules and the role of hypoxia during the development of experimental choroidal neovascularization (CNV).

Methods: Lesions were induced on C57Bl6 mice using laser photocoagulation. Animals were euthanized in a timely manner and eyecups were dissected from enucleated eyes. Choroids were immunostained for pericytes, sprouting endothelial cells (EC), or vascular EC. Choroidal neovascularization lesions where analyzed for tissue hypoxia, hypoxia-inducible factors (HIF), and heat-shock proteins (HSP).

Results: Choroidal neovascularization lesions showed a trend of increased cellular recruitment throughout the time-course and the lesions displayed positive staining for angiogenic markers. Both pericytes and sprouting EC displayed a radial progression, while vascular EC displayed a more uniform distribution across the CNV lesions. Furthermore, positive tissue hypoxia staining was observed and associated with expression of HIF-1α and vascular endothelial growth factor (VEGF).

Conclusions: Our data delimitate specific temporal windows during CNV initiation, propagation, maturation, and even recovery in experimental CNV. We show that murine CNV undergoes hypoxia-associated sprouting angiogenesis, and demonstrate involvement of pericytes. Moreover, we have shown expression of HIF-1α to the retinal pigment epithelium surrounding the CNV lesions, together with VEGF upregulation, independently of the HSP response induced by the laser thermal insult.

Age-related macular degeneration (AMD) is the leading cause of vision loss in developed countries,¹ with particular incidence in the elderly. An estimated two-thirds of the advanced cases of AMD will present signs of wet or exudative form of AMD.² The wet form of AMD is characterized by the presence of drusen deposits between the retinal pigment epithelium (RPE) and Bruch's membrane, concomitantly with choroidal neovascularization (CNV), and is designated neovascular AMD (nAMD). The process of CNV is vastly complex and to date the most acceptable model for this pathology is laser photocoagulation in mice.³ This model consists of rupturing Bruch's membrane by a laser insult, resulting in an ingrowth of choroidal vessels into the subretinal space. The murine laser-induced CNV model is commonly used to mimic nAMD,⁴ yet its robustness, rapid onset, and versatility of uses have recently expanded this model to be used as a neoangiogenic in vivo model.⁵ 

Angiogenesis, or the formation of new blood vessels from preexisting ones, is generally understood as the degradation of the capillary basement membrane, sprouting of endothelial cells (EC) growing toward an angiogenic stimulus (sprouting angiogenesis), followed by pericyte stabilization.⁶ Nevertheless, other types of angiogenesis, namely looping angiogenesis⁷ and intussusceptive angiogenesis,^8,9 have been described in adults. Vascular endothelial growth factor (VEGF) is a known key proangiogenic factor, involved in a multitude of physiological and pathological events, ranging from vessel formation in the embryo^10,11 to tumorigenesis.¹² The role of VEGF in inducing experimental and clinical CNV is evident,^2,13–18 and treatments targeting VEGF in nAMD have become standard practice in the clinic, halting vision loss in over 90% of patients.^2,19–21 Molecularly, VEGF is responsible, among others, for permeabilization of the endothelial basement membrane and tip-cell formation. An endothelial tip-cell probes VEGF concentrations through VEGFR2 (VEGF receptor)-rich filopodia, leading the endothelial sprouting in the direction of the angiogenic stimulus.^22,23 During early stages of neoangiogenesis, the newly formed vessels are highly permeable and leaky, promoting the formation of exudates and even hemorrhages, a clinical characteristic of nAMD,¹⁵ which suggests the involvement of the VEGF/VEGFR2 angiogenesis axis in CNV formation and progression. Subsequently, vessel maturation and stabilization is achieved by pericyte recruitment and deposition of extracellular matrix to reform the endothelial basement membrane and guarantee proper perfusion of the affected tissue.^6,23 

Oxygen plays a central role in vascular growth, and VEGF release from hypoxic tissues is a key event in regulation of angiogenesis.^6,24 Cells deficient in oxygen produce hypoxia-inducible factor (HIF)-1, a transcription factor responsible for upregulation of a multitude of angiogenic factors in response to hypoxia, among which is VEGF.^25,26 Hypoxia-inducible factors are heterodimeric transcription factors composed by a constitutive β-subunit and an oxygen-sensing α-subunit. Hypoxia-inducible factor-1α, together with its ortholog -2α, are considered master regulators of the response to cellular hypoxia, being involved in all adaptive responses to low oxygen,^27–29 from embryonic development to physiologic and pathologic conditions in adulthood. The presence of HIFs in CNV membranes from patients,³⁰ together with work conducted in RPE-conditional knock-out mice,³¹ implicates a role for these transcription factors in nAMD. 

To date, a detailed analysis of molecular and cellular events has not been described in the murine laser-induced CNV model, despite its widespread use. This work has focused on characterizing the neoangiogenic events, particularly the different types of angiogenesis, occurring during the murine laser-induced CNV model of nAMD. Here, CNV-induced eyes from mice were analyzed at specific time intervals and characterized for cellular events relevant during angiogenesis. As such, choroidal membranes of these eyes were immunostained for platelet-endothelial cell adhesion molecule-1 (PECAM-1/CD31) as an endothelial cellular marker and VEGF, VEGFR2, and NG2 chondroitin sulfate proteoglycan (a marker for pericytes) were examined for their role in the referred mouse model of angiogenesis. The use of vascular-specific markers allowed us to evaluate which type of angiogenesis could be present at each of the studied time intervals. Our data indicate that mouse laser-induced CNV undergoes exclusively sprouting angiogenesis, after the initial laser-induction, and culminates in wound healing and tissue remodeling events. As involvement of ischemia is directly associated with sprouting angiogenesis,^6,24 we further analyzed the role of tissue hypoxia and HIFs in CNV lesions. These data clearly indicate a role for hypoxia and HIF-1α in CNV progression and VEGF upregulation. In addition, the heat-shock protein (HSP) response to the laser insult, was analyzed in parallel with the angiogenic events observed in the mouse laser-induced CNV model. 

Materials and Methods

Animals and CNV Induction

Six- to 8-week-old C57Bl6J mice (Charles River, Cologne, Germany) were anesthetized by subcutaneous injection of ketamine (30 mg/kg; Pfizer, Cambridge, UK) and xylazine (5 mg/kg; Bayer, Leverkusen, Germany), and both pupils were dilated by topical administration of tropicamide (0.5%; Alcon, Puurs, Belgium). Choroidal neovascularization lesions were induced in one eye by diode laser (532 nm; IRIS Medical, Mountain View, CA, USA) set for 75 μm spot, 200 mW intensity, 100 ms duration. Induction of CNV was estimated by formation of a characteristic vapor bubble^3,32 and all visual hemorrhagic lesions were excluded from the study. After laser-induction, animals were treated twice with 1 mL of saline (9 mg/mL NaCl; B. Braun, Melsungen, Germany) subcutaneously and the eyes were kept lubricated by topical administration of a mix of paraffin and vaseline (APL, Gothenburg, Sweden). Animal handling was performed in accordance with the ARVO statement for the Use of Animals in Ophthalmologic and Vision Research; the study protocols were approved by Stockholm's Committee for Ethical Animal Research. 

Immunofluorescence

Mice (six animals per time-point over two independent experiments) were euthanized by cervical dislocation at specific times postlaser-induction (day 0, 3, 6, 9, 15, 21). Two distinct methods of dissection and fixing were used, depending if the experimental focus was cellular events (choroidal membranes representing exclusively the choroid vascular aspects of the CNV lesions) or tissue-markers (full CNV lesions). For choroidal membranes, laser-induced eyes (four spots per eye) were enucleated and the posterior eyecups comprehending sclera and choroid were microdissected from the surrounding tissues. Such dissections most probably comprehend the RPE layer through the sclera, though all neuroretina and visible subretinal vessels were excluded. Eyecups were fixed 10 minutes at room temperature (RT) in a 4% formaldehyde solution (FA; Solveco, Rosersberg, Sweden) in PBS (Gibco, Paisley, UK). After extensive washing with PBS, antigen retrieval was achieved by microwave heating for 3 minutes in Diva Decloaker (Biocare Medical, Concord, CA, USA), and proceeded to 1 hour incubation at RT in blocking solution: 10% normal goat serum (Invitrogen, Camarillo, MD, USA) in PBS containing 0.1% Triton X-100 (Sigma-Aldrich Corp., St. Louis, MO, USA). Primary antibodies were incubated overnight (ON) at 4°C, while secondary antibodies were incubated for 1 hour at RT; Hoechst 33258 (5 g/L in PBS; Sigma-Aldrich Corp.) was added to the secondary antibodies for counter-staining. Antibodies were prepared in blocking solution (Table). Each antibody step was extensively washed with PBS, and eyecups were postfixed as mentioned before, prior to flat-mounting with fluorescent mounting medium (Dako, Carpinteria, CA, USA). For full CNV lesions, laser-induced eyes (four spots per eye) were enucleated, cleared from the surrounding tissues, and immediately fixed in FA ON at 4°C. Animals included in tissue hypoxia experiments were injected intraperitoneally with 60 mg/kg pimonidazole hydrochloride (100 g/L in PBS; Chemicon, Burlington, MA, USA) 1 hour prior to being euthanized. After fixing, the eyeballs were extensively washed with PBS and dissected into eyecups (most plausibly comprehending RPE layer, Bruch's membrane, choroid, and sclera, including the subretinal vasculature yet not the neuroretina). Immunostainings proceeded as described for CNV membranes. Primary and secondary antibodies and stainings are described in the Table. Images of the choroidal side were acquired using an Axioskop 2 plus fluorescence microscope with the AxioVision software (Zeiss, Gottingen, Germany). 

Table

 

List of Antibodies and Stains

Immunoblotting

Upon designated times postlaser-induction, three mice were cervically dislocated and eyes enucleated (nine spots on laser-induced eyes), and the eyeball was dissected from the surrounding tissues and microdissected to eyecups as described for CNV membranes. Control samples were collected from two nonlaser-induced animals. Three eyecups (four for controls) were suspended in 200 μL CelLytic-MT (Sigma-Aldrich Corp.) containing a protease inhibitor cocktail (Roche, Mannheim, Germany). Tissues were kept on ice and immediately homogenized by 3 × 10 second pulses at maximum speed with a VDI12 rotorstator (VWR, Dresden, Germany) and 1 minute vortex in the presence of 0.5 mm diameter glass beads (Scientific Industries, Bohemia, NY, USA), followed by freezing in liquid nitrogen and stored at −80°C. Whole-tissue protein extracts were clarified by centrifugation (14000g, 20 minutes, 4°C) and quantified by a modified Bradford protein assay (Bio-Rad Laboratories, Hercules, CA, USA). Thirty micrograms of total protein extracts were separated by SDS-PAGE and transferred onto nitrocellulose filters (Bio-Rad Laboratories). Blots were blocked using 5% non-fat milk in Tris-buffered saline (nfm/TBS-T; Bio-Rad Laboratories, containing 0.05% Tween-20; Sigma-Aldrich Corp.). For primary antibodies, membranes were incubated ON at 4°C. Secondary antibodies were used for 1 hour at RT. For a full description of antibodies used consult the Table. Both antibody steps were performed in 1% nfm/TBS-T, followed by extensive TBS washes. Visualization of proteins of interest was achieved by exposure to Pierce enhanced chemiluminescence plus (Thermo Scientific, Rockford, IL, USA). 

Immunohistology

Two animals were euthanized 6 days postlaser-induction, and laser-induced eyes (four spots per eye) were enucleated and cleared from the surrounding tissues. Eyeballs were fixed in FA for 24 hours and processed for paraffin embedding. Four micrometer sections were subsequently deparaffinized and immunohistochemistry was performed in a Bond III robotic system (Leica Biosystems, Newcastle, UK). User-based modification consisted of antigen retrieval with EDTA buffer pH8 (Leica Biosystems). Primary antibodies are described in the Table. Secondary antibody, alkaline phosphatase staining, and hematoxylin counter-staining were performed using the Bond polymer refine red detection (Leica Biosystems), according to the manufacturer's instructions. Images were acquired on an Axioskop 40 microscope (Zeiss) coupled to a VisiCam TC10 (VWR, Lutterworth, UK). 

Data Analysis

Due to the dissection method used by us for CNV membranes, the area of laser impact and the area of vapor bubble could discernibly be observed from day 0 through days 15 to 21 transition, therefore rendering the classic analysis of this model by lesion area very biased. Consequently, six CNV membranes of at least three animals (selected by a blinded user based on CD31 staining conformity) were quantified using AxioVision software (Zeiss). Channels intensities (densitometric mean) and areas were calculated and corrected to the total lesion area identified by the laser-induced halo and increased CD31 staining. Quantitative data are displayed as box plots and statistical analysis was performed by one-way ANOVA with Bonferroni corrected post-hoc tests. 

Results

CNV Membranes Undergo Cellular Recruitment

One initial observation for all immunostainings was the positive immunoreactivity displayed for the site of laser impact, even in nonprimary antibody controls (data not shown). The area of laser impact was generally reduced past day 15, but was equally observed in all filter settings of the fluorescence microscope. As such, cellular events in the choroidal membranes were analyzed in the area between the laser-impact site (centrally located in the CNV lesion) and the thermal vapor bubble halo (limiting the hedges of the CNV lesion). 

Total nuclei were stained with Hoechst and intensity of the blue signal was analyzed in the CNV area (Fig. 1A). Densitometric analysis of the blue channel intensity over the CNV lesion area through time (Fig. 1B) depicted a near void of signal (i.e., cellular denudity) of the choroidal membranes immediately after laser-induction (day 0, P < 0.001), most probably attributed to the thermal events created by the laser impact. Due to day 0 cellular denudity, this time-point was excluded from all subsequent analyses. Following this initial denudity, nuclear recruitment to the CNV area followed a trend similar to that observed for vascular leakage (see Supplementary Fig. S1), suggesting a parallelism between cellular recruitment and cellular remodeling, and vessel permeability in the CNV lesions. 

Figure 1

 

Cellular recruitment in laser-induced CNV. (A) Laser-induced CNV membranes were stained with the nuclear marker Hoechst to analyze cellular recruitment. Total lesion area is delimited (white circle) according to the vapor bubble and increase in vascular density (Figs. 21552–4). Scale bar: 100 μm. (B) Blue fluorescence intensity was measured within CNV membranes through time. After an initial cellular denudity induced by the laser at day 0, a rapid cellular recruitment to the CNV membranes could be observed, indicating that cellular events in the CNV membranes are a result of cellular repopulation. *P < 0.05; **P < 0.001.

Figure 2

 

Vascular endothelial growth factor displays an even distribution in CNV lesions. Laser-induced CNV lesions were identified by increased endothelial CD31 staining (A, C; red) and VEGF distribution was determined (B, C; green) through time; choroidal flat-mounts nuclei were costained with Hoechst (A, B; blue). The relative distribution of VEGF to the endothelium was evaluated by their merged signals (C; orange). Vascular endothelial growth factor staining displayed an even distribution throughout the CNV lesions on the times studied, with a slight increase of the merged signal to endothelium in later time-points (days 15 and 21). Scale bar: 100 μm.

EC Distribute Radially in the CNV Membranes

CD31-positive cells were immunostained in CNV membranes and analyzed through time (Figs. 2A,3A,4A). Overall, a drastic increase in CD31-positive cells was present in all lesions on day 3. Albeit high CD31 immunostaining of CNV membranes on day 3, these cells did not seem to organize in any noteworthy structures. Past day 6, CD31-positive cells clearly gained organization, suggesting EC proliferation in a radial manner, from the laser impact to the halo, creating vessel-like structures past day 9. 

Figure 3

 

Vascular endothelial growth factor receptor-2 suggests sprouting angiogenesis in CNV lesions. Laser-induced CNV lesions are characterized by increased endothelial CD31 staining (A, C; red) and costained for nuclei with Hoechst (A, B; blue). Vascular endothelial growth factor receptor-2 staining (B, C; green) was determined within the lesions in the times studies, with a clear dissipation to the overall tissue in later days (B21). Vascular endothelial growth factor receptor-2 staining, within the CNV membranes, progressed in a radial distribution and with high merged signal with CD31 (C; orange), suggesting sprouting angiogenesis of the EC. Scale bar: 100 μm.

Figure 4

 

Choroidal neovascularization maturation through pericytes recruitment. Choroidal neovascularization lesions were immunostained for endothelium by CD31 (A, C; red) and pericytes by NG2 (B, C; green), with costained nuclei by Hoechst (A, B; blue). Merged signals from pericytes and endothelium were analyzed (C; orange). NG2-positive cells displayed an increased radial distribution that merged with CD31-positive cells on later time-points (C9, C15, C21; orange), indicative of vessel maturation through pericyte recruitment. Scale bar: 100 μm.

VEGF Expression Is Associated With EC Distribution in Mouse CNV

Vascular endothelial growth factor immunostaining was achieved on choroidal membranes of laser-induced CNV lesions for all studied time-points (Fig. 2). Moreover, CNV membranes were costained for EC (by CD31) and nuclei (Hoechst). The distribution of VEGF signal seemed to cover an area closely related to the total CNV membrane (Fig. 2B). Nevertheless, statistical differences could be observed between days 6 (P < 0.05) and 9 (P < 0.001) to day 15 when the relative area of VEGF versus CD31 coverage was determined (Fig. 5A). These statistical differences suggest, as expected, a mandatory involvement of VEGF for the endothelial growth occurring during CNV progression, an indication of angiogenic mechanisms by sprouting. In addition, the lack of merged signals during days 6 and 9 (displayed by orange signal, Fig. 2 panel C) between CD31- and VEGF-positive cells, suggests that other cellular compartments, rather than the EC, are producing VEGF themselves, as has been suggested previously.^33,34 Interestingly, by day 15 VEGF- and CD31-positive signals displayed merged localization, most possibly due to an overall distribution of the newly formed endothelia throughout the total area of VEGF-positive CNV membrane, as is clear by day 21 (Fig. 2, C15 and C21; orange). 

Figure 5

 

Analysis of the angiogenic markers in the CNV lesions. The area coverage of the green signal (A, VEGF; B, VEGFR2; C, NG2) was measured in relation to the total CNV membrane area. All VEGF, VEGFR2, and NG2 showed an increase in relative area coverage, indicating that these markers progress in an inner to outer manner, concomitant with radial sprouting and maturation of the CNV vessels. *P < 0.05; **P < 0.001.

Day 0 (as stated previously) related to the laser impact alone, and day 21 to an overall choroidal staining of both CD31 and VEGF, and therefore were not included in the quantifications (Fig. 5A). 

VEGFR2 Is Associated With EC Sprouting in CNV

One of the events in VEGF-driven angiogenesis is EC sprouting and formation of a VEGFR2-rich tip-cell.^22,23 Accordingly, CNV membranes were immunostained for CD31 and VEGFR2 (Fig. 3). The relative area of VEGFR2 coverage was significantly increased between days 6 and 9, and day15 (P < 0.001; Fig. 5B). Interestingly, apart from day 3 when VEGFR2-positive staining displayed an overall disarranged distribution (P < 0.05 when compared with all other days in study), VEGFR2 displayed the highest overall merged signal in relation to CD31, suggesting a high probability of coexistence of the VEGFR2 signal to the CD31 signal in the studied times (Fig. 3C; orange). Other cellular compartments than EC have been identified as sources of VEGFR2-positive immunostaining.³⁵ Nevertheless, the coexistence of VEGFR2 and CD31 signals indicates that tip-cell formation is present in the mouse laser-induced CNV model, a mechanism related to sprouting rather than intussusceptive angiogenesis.⁶ Together, these data indicate that VEGFR2 expression by the EC of CNV membranes is associated with sprouting angiogenesis. 

Days 0 and 21 were excluded from the quantifications as mentioned above. 

Pericyte Recruitment Promotes Vessel Maturity During CNV Progression

To evaluate pericyte recruitment to the newly formed vessels within CNV membranes, NG2 immunostaining was conducted in parallel with CD31 (Fig. 4). The relative area covered by NG2-positive cells increased through time in relation to the total CNV membrane area (days 6–15, P < 0.001; days 9–15, P < 0.05; Fig. 5C). Concomitantly, a noticeable increase in the merged CD31 and NG2 immunostainings, as well as an increase in merged signal area (Fig. 4C; orange), was observed. The correlation between the increased area distribution of NG2-positive cells and their merged signal to the CD31-positive cells, in a radial fashion, from the laser impact to the periphery, indicates the formation of more mature vessels in a time-dependent manner (see Supplementary Fig. S2). 

As previously, days 0 and 21 were excluded from the quantification. 

Although NG2 positivity is the most common indicator of pericyte cells, other microvascular mural cells can express this marker.³⁶ In order to distinguish pericytes from vascular smooth muscle cells (vSMC), CNV lesions were stained for CD31, NG2, and alpha-smooth muscle actin (αSMA; Fig. 6). Our data displayed that αSMA staining was absent in the CNV lesions throughout the times studied. It was noteworthy that larger αSMA-positive vessels could be visualized in the choroidal tissue surrounding the laser-induced lesions. This result indicates that NG2-positive cells recruited during CNV progression are in fact pericyte lineage and not vSMC lineage. 

Figure 6

 

Vascular smooth muscle cells are absent in CNV lesions. Choroidal neovascularization lesions were identified by increased CD31 staining (A, B; red). Pericyte cells were stained with NG2 (B, C; green), displaying staining within the CNV lesions as well as the larger vessels surrounding the lesion. To distinguish pericytes, vSMC where labeled with αSMA (A, C; blue). A clear absence of vSMC was denoted throughout all studied times within CNV lesions, yet present in larger vessels outside the lesions. Scale bar: 100 μm.

Hypoxia Is Involved in CNV Lesion Progression

Tissue ischemia and associated hypoxia are a well-characterized source of VEGF in pathology.^6,24 The indication of a VEGF/VEGFR2-driven sprouting angiogenesis suggested by our data, lead us to evaluate the role of tissue hypoxia during CNV lesions' progression. To determine tissue hypoxia in CNV lesions, animals were exposed to pimonidazole, an imidazole activated by hypoxic cells that creates effective immunogens detectable by immunofluorescence. Staining of CNV lesions with anti-pimonidazole antibodies, together with isolectin (an EC marker), allows tracing of hypoxia within the CNV lesions (Fig. 7A). Tissue hypoxia could be observed as early as day 3 postlaser-induction, with a clear near-lesion distribution on days 6 and 9 (Fig. 7A6, 7A9), dissipating through days 15 to 21. 

Figure 7

 

Presence of tissue hypoxia and HIFs in CNV lesions. Isolectin-IB₄ staining identified laser-induced CNV lesions (A; blue), and tissue hypoxia was determined by the presence of pimonidazole (A; green). Tissue hypoxia could be observed in CNV lesions on early times studied (days 3–9), with a marked decreased on later time-points (days 15 and 21). The presence of HIF transcription factors was denoted by positive staining for CAIX (B; red) while laser-induced CNV lesions were identified by Isolectin-IB₄ staining (B; blue). A diffuse CAIX staining was observed on all studied times within the CNV lesions, accompanied by a clear intensification of the stainings during days 6 and 9, an indication of transcriptionally active HIFs. The relative distribution of pimonidazole to CAIX staining was evaluated by their merged signals (C; orange), displaying a coexistance of tissue hypoxia and HIFs tissue marker, particularly during days 6 and 9 (C6, C9; orange). Scale bar: 100 μm.

HIF Transcription Factors Are Implicated in CNV

Cells deprived of oxygen induce protein expression of HIFs.²⁶ Unfortunately, immunostainings for HIF-α transcription factors subunits proved very challenging in the CNV lesions, possibly due to a partial tissue permeabilization of the thick choroidal flat-mounts. Nevertheless, carbonic anhydrase IX (CAIX), a membrane-bound protein strongly induced by HIFs,³⁷ could be used as a tissue marker for the presence and transcriptional activation of hypoxia transcription factors in CNV lesions (Fig. 7B). On day 3 after laser-induction, CAIX immunoreactivity was weakly detected throughout the CNV lesions, with a discreet incidence by the laser-impact site. This pattern was observed at all studied times, and might not be directly related to HIFs. On days 6 and 9, a clear distinct pattern could be observed for CAIX stainings, a signal that appeared associated with the EC marker, isolectin (Fig. 7B6, 7B9). A colocalization of pimonidazole and CAIX stainings could be observed, particularly on days 6 and 9 (Fig. 7C6, 7C9). These findings indicate that tissue ischemia and cellular hypoxia are intrinsically involved in progression, yet possibly not initiation of CNV lesions in the murine laser-induced model. 

HIF-1α Upregulation Is Related With Angiogenesis in Laser-Induced CNV

To assess the relevance of HIF-α subunits during progression of laser-induced CNV, whole-tissue protein samples were analyzed by immunoblotting, in a time-dependent manner (Fig. 8A). Analysis of HIF-1α depicted a peak in protein expression between days 6 and 9, a result concomitant with the CAIX stainings (Fig. 7). Interestingly, a similar protein expression could be observed for VEGF. In addition, HIF-2α could not be detected, albeit previously tested antibodies (data not shown), suggesting that HIF-1 is the transcription factor mediating VEGF expression upregulation and CNV lesions' progression in the mouse model. 

Figure 8

 

Hypoxia-inducible factors-1α mediates VEGF expression during CNV. (A) Whole-tissue protein extracts from laser-induced eyecups and nonlaser induced controls (C) were analyzed with anti–HIF-1α, anti-VEGF, anti-HSP90, and anti-actin as a loading control. Hypoxia-inducible factor-1α protein levels displayed a peak of expression during days 6 and 9, an expression that was paralleled by VEGF. Analysis of HSP90 protein levels indicates that HIF-1α expression is independent of the laser-induced HSP response (day 3). (B) Hypoxia-inducible factor-1α and VEGF immunohistology of mouse CNV lesions from day 6 display positive staining (red) through the vascular lesion, including the surrounding RPE cells (arrowheads), distinguished by the characteristic pigment of the epithelium. Sections were counterstained with hematoxylin. Lesion area was labeled with a tissue marking dye (green) to facilitate sectioning. Scale bar: 25 μm.

Being a laser-induced model, with a thermal component, we analyzed the presence of HSP response in relation to CNV lesions. An increase in HSP90 protein levels could be observed on day 3, a clear evidence of HSP response to the laser insult, with a follow-up increase of protein expression past day 9, possibly related to tissue remodeling and chaperoning activity associated with HSP90. Moreover, the differences observed in protein expression for HIF-1α and HSP90 suggest that intracellular VEGF expression in the laser-induced CNV model is occurring after induction of this model by the thermal insult of the laser and associated with HIF-1α expression. 

To address which cells could be expressing HIF-1α, immunohistology was performed on paraffin sections from mouse CNV lesions of day 6 (Fig. 8B). As previously described, basal levels of HIF-1α were detected in mouse adult retina³⁸ as well as a generalized staining through the vascular CNV lesions,³⁹ with several cells displaying positive staining. More interestingly, RPE cells surrounding the vascular lesion also displayed positive immunoreactivity for HIF-1α. Subsequently, sequential sections of the same CNV lesion were analyzed for VEGF expression (Fig. 8B). Vascular endothelial growth factor–positive pattern correlated with HIF immunoreactivity, displaying a basal staining in the retina, and a generalized staining through the CNV lesion with a clear intensity in RPE cells surrounding the lesion. These data suggest that the RPE layer of laser-induced mouse CNV lesions could be directly involved in the hypoxia response and VEGF expression determined by immunoblotting. 

Discussion

Age-related macular degeneration is the leading cause of blindness in the Western world.² Experimental animal models of CNV became essential to elucidate the complex cellular and molecular mechanisms underlying AMD pathogenesis, but also as screening tools for new drugs.⁴ The robust and easily accessible laser-induced CNV model in mice is the most widely accepted and commonly used model of nAMD,³ becoming a generally used model of angiogenesis in recent years.⁵ 

The present study characterizes, for the first time, the subsequent events regarding angiogenesis, with both cellular and molecular insights, during the process of laser-induced murine CNV. The results of cellular recruitment to the CNV-induced area denoted by nuclear staining indicate that two distinct events occur in laser-treated eyes, at specific times. These results were further characterized by the distribution of VEGF, VEGFR2, and NG2 in relation to the newly formed endothelium within CNV lesions. Furthermore, the role of tissue hypoxia and HIF transcription factors, as well as HSP response, was further characterized in this model through the times studied. 

As part of the experimental design, laser-induced CNV lesions were followed at regular time intervals (days 0, 3, 6, 9, 15, and 21) in order to evaluate the status of all assayed vasculature markers during mouse CNV lesions' progression. Furthermore, unlike the classic analysis of hemorrhage and recovery of the laser impact,^4,5 the dissection methods performed in this study allow analysis of CNV lesions with focus on the choroidal membranes, permitting evaluation of the newly formed vasculature in a radial fashion. This method has the advantage of assessing multiple factors simultaneously using fluorescence microscopy, and allowing spatial correlations between the evaluated factors. Nonetheless, it also carries the disadvantage that it is focused on choroidal membranes' angiogenesis and therefore represents primarily events occurring in the sub-RPE space (i.e., omitting potential events that may occur in the overlying neurosensory retina). Also, because these dissection and staining protocols display choroidal membranes of CNV-induced mouse lesions, area studies have to be carefully assessed due to the fact that the initial area of laser-induced lesion (laser impact plus vapor halo) is observed through all studied times. Albeit this minor limitation, the cellular events emerging during CNV progression can be determined more clearly in a timely manner. 

Examination of choroidal membranes of murine laser-induced CNV lesions in relation to endothelial formation displayed a cellular denudity of the CNV lesions on day 0 followed by a dramatic increase of CD31-positive cells on day 3. The number of CD31-positive cells was reduced on day 6. These occurrences were paralleled by nuclear recruitment (and Evan's blue dye extravasation; see Supplementary Fig. S1), as well as VEGF, VEGFR2, and NG2 signal intensity. Although consistent in all methods used, the early events observed (before and to day 3 postlaser-induction) are unclear. One possible interpretation of these results could be attributed to acute events mediated by laser-induction and/or recruitment of bone marrow–derived cells,^39–41 a mechanism similar to the recruitment of angioblasts observed during vasculogenesis in the embryo.⁴² Nevertheless, the result that laser-induced CNV is initiated with a clear void of cells followed by a marked cellular recruitment is a bonafide indicator that intussusceptive angiogenesis is absent during these early time points. Furthermore, the cellular recruitment observed by day 3 failed to display any defined organization, particular in merged markers for EC and pericytes (NG2 versus CD31; Fig. 4C3), suggesting that looping angiogenesis is absent in this model, in contrast to previous observations in cornea neovascularization.⁷ 

Analysis of days 6, 9, and 15 lead to a general characterization of an angiogenesis phase of this animal model. The distribution of EC migrating to a VEGF-driven signaling, by a VEGFR2-mediated response, was established between the referred days. Simultaneously, recruitment of pericytes to the newly formed endothelial structures, with increased merged signals, associated with vessel stabilization and maturation, also depicted by the decrease in vessel permeability between days 9 and 15 (see Supplementary Fig. S1). Furthermore, the absence of vSMC suggests that laser-induced angiogenesis in mouse choroids occurs exclusively by capillarization. Concurrently, these data are a clear indication of VEGF/VEGFR2-mediated sprouting angiogenesis in the choroids of laser-induced mouse model, as opposed to intussusceptive angiogenesis.⁶ This angiogenesis phase thus mimics the presumed cellular events of clinical CNV, including initial vascular sprouting, leakage, and subsequent maturation through pericyte recruitment (see Supplementary Fig. S3). 

The recovery of the lesions with diffusion of VEGFR2 staining to the overall choroid and NG2 to larger vessels (together with Evan's blue permeability assay on day 21; see Supplementary Fig. S1), illustrate the final phase of laser-induced CNV mouse model, indicating a recovery of laser-mediated lesions in the animals' eyes. 

Together, these cellular events demonstrate that laser-induced mouse CNV lesions are initiated by cellular recruitment to the laser-impact area, progress exclusively by sprouting angiogenesis, and terminate by tissue remodeling and wound healing, with scar formation. 

The presence of ischemia is considered directly linked to sprouting angiogenesis in a multitude of models.⁶ In that manner, we addressed whether hypoxia was involved in mouse laser-induced CNV lesions. Our results demonstrated that tissue hypoxia could be observed as early as day 3 and was sustained to day 9 postlaser-induction, while the HIFs tissue-marker, CAIX, was upregulated particularly between days 6 and 9. The concomitant presence of tissue hypoxia and HIF tissue-marker in a time-dependent manner raised the question of the role of HIF-α subunits on VEGF/VEGFR2-driven angiogenesis observed by us during CNV lesions' progression. Curiously, HIF-1α displayed near undetectable protein levels on day 3, even though tissue hypoxia and a faint CAIX staining could be observed (Fig. 7). Such indicated that tissue hypoxia could be present in the laser-induced CNV lesions from day 3, subsequently leading to HIF-1α protein stabilization and transcriptional activation, with VEGF upregulation (Fig. 8). In addition, the faint CAIX staining of the CNV lesions with discreet incidence on the laser-impact area observed on all studies, could be a result of CAIX presence in areas associated with necrosis, as has been described in tumor models.³⁷ 

The mechanisms of regulation and degradation of HIF-1α are intricate and complex,^43,44 and a link between HSP response and HIF-1α degradation has been suggested.⁴⁵ Being a thermal-induced model, by creation of a laser impact and a vapor bubble, the mouse CNV model could be associated with HSP responses. Here, we show first evidence that HSP90 is upregulated in this model as a response to the thermal-insult (day 3). Furthermore, we show that during HSP response, albeit the presence of tissue hypoxia, no HIF-1α could be detected (Fig. 8). These data suggest that HIF-1α could be negatively regulated by the HSP response during the early molecular events of laser-induced CNV. 

Collectively, these data are the first to schematically and timely characterize the murine laser-induced CNV model, regarding cellular and molecular angiogenesis, identifying three independent windows for this model: an acute phase (days 0 and 3); a neoangiogenenic phase (days 6, 9, and 15); followed by a wound healing phase (day 21 and onward). Furthermore, it also reveals that the angiogenic phase of this mouse model undergoes exclusively sprouting angiogenesis by the VEGF/VEGFR2 axis, through a HIF-1α–mediated mechanism, with vessel maturation and stabilization by pericyte recruitment (see Supplementary Fig. S3). Interestingly, clinical observations demonstrate that anti-VEGF treatment effectively reduces CNV leakage but has limited efficacy on reducing the CNV area.⁴⁶ A rapid recruitment of pericytes, as demonstrated herein, is a likely explanation for the relative anti-VEGF resistance. This is further supported by the observation that recombinant anti-VEGF and anti–platelet-derived growth factor (PDGF) treatment, presumably targeting vascular endothelial and mural cells, respectively, synergistically inhibits neovascularization in ocular models, including laser-induced murine CNV.⁴⁷ 

Moreover, an overall parallelism can be established between the experimental mouse model of CNV and clinical observations of nAMD (see Supplementary Fig. S3), with preclinical events relating to the acute experimental phase, early clinical symptoms of macular exudates and edema, together with metamorphopsia and vision loss relating to the hemorrhagic sprouting phase of neoangiogenesis, followed by the vascularization phase, culminating in disciform scar formation relating to the wound healing phase of the animal model.^2,14,48 To date, the role of hypoxia, and subsequently HIFs, in clinical nAMD is still considered somewhat controversial due to the high levels of vascularization and oxygenation of the choroid. Evidence suggest that cellular hypoxia and HIF expression are associated with RPE cells,^30,49,50 rather than with the choroidal endothelium. Our data in the mouse-induced CNV model are in agreement with these findings, with VEGF stainings displaying a poor merged signal with the endothelium (Fig. 2), and tissue hypoxia and HIF tissue-marker stainings localizing peripherally to the CNV lesions (Fig. 7). These, together with intracellular VEGF protein upregulation being concomitant with HIF-1α expression (Fig. 8), could suggest that hypoxia is mediating secretion of VEGF though HIF-1α by the RPE cells. In fact, both HIF-1α and VEGF were detected in RPE cells surrounding CNV lesions in the mouse laser-induced model (Fig. 8), easily identified by the characteristic cytosolic pigment in this epithelium. As such, we indicate a role for HIF-1α in CNV formation and propagation, as the choroidal vasculature responds to the neighboring tissue hypoxia. Nevertheless, the cellular and molecular mechanisms that initiate nAMD in patients, leading to hypoxia in the RPE layer remain unraveled, and might be biased in the mouse model of CNV as a consequence of the laser-induction. 

The presented findings have implications on how to best take advantage of the widely used laser-induced mouse model of CNV, particularly in windows for screening of new antiangiogenic treatments, as well as time frames for evaluation of specific molecular and cellular related results. The results also give insight into how clinical CNV may progress through the various stages of maturation and how drug intervention may affect this process. Furthermore, the evidence of HIFs in late-stage human CNV membranes³⁰ can contribute to limiting effects in reducing the CNV area observed in certain patients treated with anti-VEGF therapies,⁴⁶ due to the maintenance of the CNV lesion in a VEGF-producing proliferative state. This could be the result of a negative feedback loop in nAMD patients between tissue hypoxia, HIF stabilization, VEGF production, and vessel leakage, suggesting the possibility of novel treatments for nAMD patients by anti-HIF therapies. 

Acknowledgments

The authors thank Parviz Mammadzada for helpful discussions, Linnea Tankred for animal husbandry, and Emma Lardner for help with immunohistology. 

Supported by grants from The Crown Princess Margareta Association for the Visually Impaired (Valdemarsvik, Sweden), Edwin Jordan Foundation (Stockholm, Sweden), Tore Nilsson Foundation (Stockholm, Sweden), The Swedish Research Council (Stockholm, Sweden), The Swedish Eye Foundation (Umeå, Sweden), and the Scientific and Technological Research Council of Turkey (ST; Ankara, Turkey). 

Disclosure: H. André, None; S. Tunik, None; M. Aronsson, None; A. Kvanta, None