Revascularization after angiogenesis inhibition favors new sprouting over abandoned vessel reuse
Inhibiting pathologic angiogenesis can halt disease progression, but such inhibition may offer only a temporary benefit, followed by tissue revascularization after treatment stoppage. This revascularization, however, occurs by largely unknown phenotypic changes in pathologic vessels. To investigate the dynamics of vessel reconfiguration during revascularization, we developed a model of reversible murine corneal angiogenesis permitting longitudinal examination of the same vasculature. Following 30 days of angiogenesis inhibition, two types of vascular structure were evident: partially regressed persistent vessels that were degenerate and barely functional, and fully regressed, non-functional empty basement membrane sleeves (ebms). While persistent vessels maintained a limited flow and retained collagen IV+ basement membrane, CD31+ endothelial cells (EC), and α-SMA+ pericytes, ebms were acellular and expressed only collagen IV. Upon terminating angiogenesis inhibition, transmission electron microscopy and live imaging revealed that revascularization ensued by a rapid reversal of EC degeneracy in persistent vessels, facilitating their phenotypic normalization, vasodilation, increased flow, and subsequent new angiogenic sprouting. Conversely, ebms were irreversibly sealed from the circulation by excess collagen IV deposition that inhibited EC migration and prevented their reuse. Fully and partially regressed vessels therefore have opposing roles during revascularization, where fully regressed vessels inhibit new sprouting while partially regressed persistent vessels rapidly reactivate and serve as the source of continued pathologic angiogenesis.
KeywordsNeovascularization Revascularization Cornea Regression Empty basement membrane sleeves Sprouting angiogenesis
Pathological angiogenesis can lead to serious progression of disease and is typically treated using molecular inhibitors of angiogenesis. Clinically, this strategy is widely used to limit retinal tissue damage in neovascular age-related macular degeneration (nAMD) and diabetic retinopathy [1, 2, 3], to avoid immune rejection in high-risk corneal transplantation [4, 5] or to prevent tumor growth and metastasis in various cancers [6, 7, 8, 9]. Although anti-angiogenic therapy (to date principally focused on VEGF inhibition) can partially regress vessels to maintain useful vision [1, 2, 3] or achieve a modest survival benefit in cancer patients [10, 11], discontinuation of anti-angiogenic therapy or a gradual decline in therapeutic effect eventually ensues [3, 12]. This carries the risk of rebound, whereby angiogenesis is re-activated by the underlying inflammatory, hypoxic, or angiogenic stimulus [3, 13, 14, 15, 16, 17, 18, 19, 20]. This angiogenic rebound or ‘revascularization’ of the tissue necessitates repeated or continued anti-angiogenic treatment; however, continued treatment comes at a cost, including adverse effects/toxicity of prolonged therapy [11, 21, 22], acquired drug resistance [23, 24], and an escalating economic and health care delivery burden of repeated treatments .
Mechanisms of tissue revascularization after initial angiogenic regression therefore deserve closer investigation, although to date relatively little is known about the manner by which vessels respond to a change in the tissue microenvironment from anti- to pro-angiogenic to mount a revascularization response. Earlier studies have examined revascularization of skeletal muscle [26, 27], the retina [28, 29], and the heart  following experimental injury; however, such vasculatures were not pathologically angiogenic in nature. In the eye, choroidal neovascularization in AMD and retinal neovascularization in proliferative diabetic retinopathy are caused by angiogenesis; however, even advanced clinical imaging methods have insufficient resolution to examine revascularization dynamics at the single-vessel level [3, 12]. Tumor vessels are similarly angiogenic, and although pre-clinical and clinical studies report tumor revascularization following stoppage of anti-VEGF treatment [9, 19, 20, 31], the detailed mechanisms by which revascularization may occur are less well studied. In a prior study by Mancuso and colleagues, revascularization of experimental tumors after a short duration (7 days) of anti-VEGF therapy was reported to occur by reuse of the empty basement membrane sleeve (ebms) remnants of regressed vessels . Anti-VEGF therapy, however, is often administered over a prolonged period, for example in tumors [19, 20] or to treat nAMD . It is unknown how a longer course of anti-angiogenic treatment affects the newly formed vessels, or how revascularization proceeds once the treatment effect subsides. In nAMD, a longer period between anti-angiogenic treatments  is becoming more common, driven primarily by economic considerations.
A major limitation of current models of revascularization is an inability to examine tissues in vivo. Analysis of different tissues post-harvest and processing precludes investigation of the dynamics of revascularization of the same vasculature. Typical dense vascular invasion within tumors also makes difficult an unambiguous discrimination between persistent and revascularizing vessels. Revascularization may also be influenced by factors such as the type and efficacy of anti-VEGF therapy, degree of inflammation present, and alternative pro-angiogenic factors that may be activated upon VEGF inhibition [34, 35, 36, 37, 38, 39, 40, 41]. To address these multiple issues, we opted to investigate revascularization in the cornea, one of the first tissues used to study angiogenesis [42, 43, 44, 45] and where anti-VEGF drugs originally developed for tumor treatment are used clinically [4, 5]. Advantages of the cornea as a model include its transparency and accessibility for in vivo examination, permitting longitudinal in vivo analysis of the same vessels over time . Cornea models further allow a controlled and reproducible pattern and timing of angiogenesis [47, 48, 49], while the relatively thin corneal tissue permits characterization of the entire neovascularized region in vivo using high-resolution imaging methods. Here, we induced inflammatory angiogenesis in the rat cornea in a reversible manner  by suture placement to stimulate inflammation and angiogenesis, followed by suture removal to abruptly inhibit angiogenesis and induce vessel regression. This method of inhibiting angiogenesis is not solely dependent on interfering with VEGF signaling, as a multiplicity of endogenous inflammatory and angiogenesis inhibitors present in the cornea are activated upon suture removal, in attempt to restore the cornea’s normal avascularity [47, 48, 50]. Following a 30-day regression phase, angiogenesis inhibition was reversed by re-suturing the same cornea to induce a revascularization response, simulating discontinuation of an anti-angiogenic treatment in the presence of an underlying pathology.
Using this model, opposing roles of different vascular structures in the regressed vascular bed were observed, which were additionally sensitive to the duration of angiogenesis inhibition. The study elucidates for the first time the relative importance of different vessel phenotypes present after angiogenesis inhibition, and the phenotypic changes by which revascularization occurs. The results suggest that revascularization is principally mediated by rapid recovery of persistent, partially regressed vessels, and not by reuse of the fully regressed structures (the empty basement membrane sleeves) as conduits by the EC.
A model for investigating revascularization dynamics in vivo
Two types of vascular structures are present following sustained regression
Ebms pervade the tissue, do not express CD31, and do not support blood flow during revascularization
Persistent vessels undergo de novo angiogenic sprouting during revascularization, independent of ebms
Acellularity of ebms is maintained during revascularization
Excess collagen IV deposition irreversibly isolates ebms from parent vessels
Persistent vessels are perfused, have EC and pericyte coverage, and EC rapidly normalize during revascularization
A shorter course of angiogenesis inhibition results in incomplete vessel regression
A 30-day course of inhibiting angiogenesis achieved effective regression to dramatically regress new angiogenic vessels, restoring murine corneal transparency; however, a sub-population of barely visible vessels with markedly narrowed diameter and restricted flow persisted in the tissue. These persistent vessels facilitated revascularization of the tissue through a process of hyper-dilation and subsequent de novo sprouting. The persistent vessels remained in the tissue despite a sustained anti-angiogenic milieu, retaining pericyte coverage and an intact, although abnormal, vascular endothelium. Persistent vessels remaining in the tissue are a well-known phenomenon , and consistent with the hypothesis of vascular normalization of initial angiogenic vessels following anti-angiogenic treatment [52, 53, 54]. Despite this ‘normalization,’ however, we observed abnormalities at the ultrastructural level. Pericytes appeared swollen and partly detached from the vascular wall, while EC bulged and extended into the vessel lumen and formed luminal and abluminal processes. Flow was present in the persistent vessels but consisted of plasma/plasma-like substance interrupted by occasional erythrocytes flowing in a serial manner. These findings are consistent with earlier studies of vessel regression in other tissues [55, 56, 57, 58], indicating a degeneracy of incompletely regressed vessels at the cellular level. Remarkably, however, this degeneracy reversed after discontinuation of the angiogenesis inhibition phase, with pericytes and EC of persistent vessels rapidly (within 24 h) reverting to a normal phenotype to permit increased perfusion and flow. This ‘reverse normalization’—in response to a sudden increase in vasodilating and angiogenic factors—re-activated the degenerate persistent vessels, to mount a strong revascularization response.
Interestingly, the early vasodilation effect we report in the current model has also been observed in other tissues. For instance, following 7 days of VEGF blockade in healthy mice, discontinuation of treatment induced dilation and bulging of hepatic vessels . Likewise in tumor-bearing mice, 7-day VEGF blockade led to treatment-resistant vessels (persistent vessels) appearing dilated 2 days following treatment withdrawal . Finally, in a patient receiving aflibercept injections for nAMD, longitudinal optical coherence tomography angiography imaging suggested that dilated persistent vessels comprised the revascularization response and led to continued pathology necessitating re-treatment .
Regression of vessels in the present model was associated with formation of a network of extensive ebms in the tissue as a record of the initial angiogenesis. Despite the abundance of ebms in direct contact with persistent vessels, there was no evidence to support their reuse by EC after a 30-day course of angiogenesis inhibition. While injury models of normally vascularized tissue such as the retina and skeletal muscle provide some evidence of partial ebms reuse [26, 27, 28, 29], here we note that pathologic angiogenic vessels once fully regressed appear to remain irreversibly dormant. With reactivation of vasodilating/permeability factors such as VEGF-A, it may be more favorable for EC within the persistent vessels to reactivate to a functional state by normalizing and reconfiguring to accommodate an expanding lumen, rather than repopulate the acellular ebms by proliferation and migration. By contrast, if ebms were reused as conduits for EC, extremely fast EC proliferation and migration into ebms would have been required to account for the observed speed and extent of revascularization. In addition, at least a few partially colonized ebms would be present in the tissue during revascularization. None of these effects were observed during revascularization after a 30-day course of angiogenesis inhibition. In an earlier study of tumor revascularization after a short 7-day course of VEGF inhibition , α-SMA staining on ebms indicated that the structures may not have been acellular after a short duration of treatment. This effect was also observed in our model after only 7 days of angiogenesis inhibition; partially regressed vessels still had some EC and smooth muscle cell coverage and retained connection to the circulation. A very short course of anti-VEGF treatment may therefore lead to an intermediate stage of regression prior to full cellular abandonment of the vascular basement membrane, which could possibly support re-functionalization of partially regressed vessels upon abrupt discontinuation of the treatment.
By contrast, after a longer course of angiogenesis inhibition mimicking clinical time frames of anti-angiogenesis treatment, more vascular basement membrane had complete cellular abandonment leading to formation of ebms, with roughly equal distribution of ebms and persistent vessels in the tissue. Ebms, once formed, were unfavorable for repopulation by EC or pericytes during tissue revascularization. Based on our observations we suggest that during vascular regression, ebms are rendered irreversibly dormant by a pruning process of luminal closure at proximal sites by excess basement membrane deposition in the form of ‘plugs.’ These plugs exhibit strong type IV collagen expression, localize to the junction of ebms with persistent vessels, and could be deposited to prevent leakage from persistent vessels. Similar dense Coll IV plugs at junction points with perfused persistent vessels were apparent in a RIP-Tag2 tumor model , with the EC cytoskeleton closely following the form of the plugs but appearing not to penetrate them, as we also observed in the present model. In the tumor model, however, the intense Coll IV expression (plugs) was interpreted as “matrix heterogeneity accompanying new sprouts” . Conversely in the present study, basement membrane in new sprouts (1–2 days old) had very weak to absent collagen IV expression, consistent with endothelial tip cell formation requiring basement membrane degradation . Furthermore, earlier studies show Coll IV absence on new sprouts and deposition only in a later phase [60, 61, 62], consistent with new sprouts expressing Coll IV only at a later time point (7 days of sprouting) in our model. Observations of new Coll IV-negative angiogenic sprouts adjacent to plugged, dormant, and acellular Coll IV-positive ebms indicated that sprouting preferentially occurred away from the Coll IV plugs. EC cytoskeletal adaptation to the shape of the plugs moreover suggests Coll IV in ebms inhibiting EC migration, in accordance with known anti-angiogenic properties of Coll IV [63, 64].
By electron microscopy and α-SMA expression in this study, and using the pericyte marker NG-2 in prior studies [48, 65], pericyte coverage of persistent vessels was apparent, whereas ebms remained acellular during revascularization. This selectivity of vessel regression associated with pericyte coverage has similarly been reported in tumors .
It deserves mention that the model used to induce angiogenesis and regression in this study differs from the clinical situation of angiogenesis mediated by disease-related processes, followed by treatment with pharmacological inhibitors of angiogenesis. It would therefore be useful to verify the mode of revascularization in models more closely paralleling disease, and additionally using long-term pharmacologic inhibition of angiogenesis. Nevertheless, much knowledge pertaining to angiogenesis has been obtained through use of models selected to simplify and separate the complex processes occurring during the course of disease and its treatment.
Animal ethics and study model
Experiments were conducted after receiving approval from the Linköping Regional Ethical Committee for Animal Experiments, under ethical permit number 585. Experiments were also conducted with adherence to the guidelines of the Association for Research in Vision and Ophthalmology (ARVO) for the use of animals in ophthalmic and vision research. Wild-type male Wistar rats aged 5–6 weeks (Janvier Labs, France) were used for all experiments. The suture model of inflammatory corneal angiogenesis was used as previously described . Briefly, two 10-0 nylon sutures were placed into the right eye cornea at 1.5 mm from the limbus. Sprouting of new blood vessels from the limbus toward central cornea occurred over a 7 days period (referred to as initial angiogenesis phase). On the 7th day, both sutures were removed from the cornea to induce vessel regression over a 30-day period (regression phase mimicking treatment). On the 30th day, the same cornea was again re-sutured with two nylon sutures placed at the same distance from the limbus as originally, to induce revascularization (revascularization phase), which was monitored longitudinally during a 4-day period, i.e., on days 0, 1, 3, and 4. In a separate experiment, the above procedure was repeated, however, terminating the experiment after only 7 days of regression.
In vivo confocal microscopy (IVCM)
In vivo confocal microscopy (IVCM) (Heidelbert Retinal Tomograph 3 with Rostock Corneal Module HRT3-RCM, Heidelberg Engineering, Germany) was used as previously described , to monitor vessel perfusion and new sprouting. Serial images showing the same vessel at different time points were selected.
Slit lamp imaging
To clinically monitor the overall neovascularization response, live images and video were captured using a rodent slit lamp  (Micron III, Phoenix Research Laboratories, USA). Vessels from still images were manually counted and compared between initial angiogenesis, regression, and revascularization phases.
Whole-mount immunofluorescence staining
Rats were anaesthetized with a combination of ketamine (Pfizer) and dexdomitor (Orion Pharma) and euthanized by intracardial injection of pentobarbital. The cornea was dissected for whole-mount immunofluorescent staining. Briefly, the harvested cornea was embedded in OCT media (Thermo scientific) and frozen at -80 °C until use. Frozen samples were thawed at room temperature, washed in PBS for 1 h, and then fixed in cold acetone (− 20 °C) for 30 min. The fixed samples were washed three times in PBS for 30 min each time and blocked for 2 h with 10% normal goat serum at room temperature. Primary antibodies CD31 (1:300, MAB13932Z-Merck Milipore) and α-SMA (1:100, ab7817-Abcam) were added and incubated overnight at 4 °C. Subsequently samples were washed three times in PBS for 30 min each, and incubated with secondary antibody (1:100, DLlight 549-Abcam) overnight at 4 °C. For double staining with collagen IV, samples were then washed three times in PBS for 30 min each time and blocked for 2 h with 10% normal goat serum at RT, and then incubated overnight at 4 °C with primary antibody against Coll IV (1:300, ab19808-Abcam). After overnight incubation, samples were washed three times in PBS for 30 min each and incubated with secondary antibody (1:100, Alexa Flour 488-Abcam) overnight at 4 °C. Samples were then washed in PBS for 1 h and mounted using quick hardening antifade mounting media (Sigma) and imaged using a laser scanning fluorescent confocal microscope (LSM 700, Zeiss).
Image processing and analysis
Fluorescence images were analyzed using Huygens software (Scientific volume imaging). Image files were deconvoluted in express mode, and using the generated images, co-localization analysis was performed using Costes threshold and Pearson’s correlation coefficient (r). IMARIS software (Bitplane) was used for 3D surface rendering of z-stack fluorescence confocal image files (LSM 700, Zeiss).
Transmission electron microscopy
Harvested cornea samples were fixed in 2% glutaraldehyde in 0.1 M Na cacodylate, pH 7.4. Fixed samples were washed in the same buffer and post fixed in 2% Osmium tetroxide. Following en block staining with 2% uranyl acetate in 50% ethanol, samples were dehydrated in a series of ascending concentrations of ethanol and acetone. A three-step infiltration was performed prior to embedding in epoxy embedding medium kit (SIGMA-ALDRICH GmbH). Blocks were initially trimmed and sectioned using a Leica Ultracut UCT microtome (Leica UC7 ultra microtome (Leica Microsystems GmbH, Vienna, Austria). Ultrathin sections (70-nm thickness) were collected onto a formvar-coated copper slot grid, and counterstained with uranyl acetate and lead citrate. Images were taken using a 100 kV transmission electron microscope (EM JEM 1230, JEOL Ltd, Tokyo, Japan).
The Shapiro–Wilk normality test was used with alpha = 0.05 to test for normal distribution. When comparing two sample means, the student t test was used for normally distributed data, while the non-parametric Mann–Whitney U test was used where data were not normally distributed. When comparing multiple groups, one-way ANOVA with Turkey’s post hoc multiple comparisons test was performed. A P value < 0.05 was considered significant. The Pearson’s correlation coefficient and the Costes threshold were used in the co-localization analysis.
Open access funding provided by Linköping University. The authors kindly acknowledge the contribution Camilla Hildesjö from the Departments of Clinical Pathology and Clinical Genetics, Region Östergötland (Linköping, Sweden) for technical assistance with tissue embedding and sectioning. Thanks to Vesa Loitto of the core facility microscopy unit of Linkoping University for assistance with Huygens and IMARIS software and spinning disk microscopy. Thanks also to Simin Mohseni of the Department of Clinical and Experimental Medicine, Linkoping University for assistance with electron microscopy analysis. The authors also wish to thank Theresa Lagali-Hensen for assistance in the design and preparation of Fig. 9.
AM, BP, LJ, and NL designed the experiments. AM, BP, AL, and PM performed microsurgery. AM, PM, MT, AL, BP, and NL collected in vivo data. AM, BP, MT, LJ, and NL analyzed in vivo data. AM and MN performed wet lab experiments. AM, LJ, BP, MN, and NL analyzed TEM data. AM and NL wrote the manuscript. All authors reviewed, revised, and/or approved the final version of the manuscript.
This work was supported by a grant from the Swedish Research Council (Grant No. 2012-2472).
Compliance with ethical standards
Conflict of interest
The authors declare that they have no conflict of interest regarding this work.
- 7.Wang Z et al (2015) Broad targeting of angiogenesis for cancer prevention and therapy. In Seminars in cancer biology. ElsevierGoogle Scholar
- 64.Colorado PC et al (2000) Anti-angiogenic cues from vascular basement membrane collagen. Can Res 60(9):2520–2526Google Scholar
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.