repeat corneal neovascularization is characterized by more...

Cornea

Repeat Corneal Neovascularization is Characterized byMore Aggressive Inflammation and Vessel Invasion Thanin the Initial Phase

Anthony Mukwaya,1 Pierfrancesco Mirabelli,1 Anton Lennikov,1,2 Muthukumar Thangavelu,1,3

Lasse Jensen,4 Beatrice Peebo,1 and Neil Lagali1,5

1Department of Ophthalmology, Institute for Clinical and Experimental Medicine, Faculty of Health Sciences, Linkoping University,Linkoping, Sweden2University of Missouri-Columbia, Columbia, Missouri, United States3Deptartment of BIN Convergence Technology, Chonbuk National University, Jeonju, Republic of Korea4Department of Medical and Health Sciences, Division of Cardiovascular Medicine, Linkoping University, Linkoping, Sweden5Department of Ophthalmology, Sørlandet Hospital Arendal, Arendal, Norway

Correspondence: Neil Lagali, De-partment of Ophthalmology, Insti-tute for Clinical and ExperimentalMedicine, Faculty of Health Sciences,Linkoping University, Linkoping58183, Sweden;[email protected].

Submitted: May 23, 2019Accepted: June 14, 2019

Citation: Mukwaya A, Mirabelli P,Lennikov A, et al. Repeat cornealneovascularization is characterized bymore aggressive inflammation andvessel invasion than in the initialphase. Invest Ophthalmol Vis Sci.2019;60:2990–3001. https://doi.org/10.1167/iovs.19-27591

PURPOSE. Treatment of corneal neovascularization can lead to vessel regression and recoveryof corneal transparency. Here, we examined the response of the cornea to a repeatedstimulus after initial vessel regression comparing the second wave of neovascularizationwith the first.

METHODS. Corneal neovascularization was induced by surgical suture placement in the ratcornea for 7 days, followed by suture removal and a 30-day regression period. Corneas werethen re-sutured and examined for an additional 4 days. Longitudinal slit-lamp imaging, in vivoconfocal microscopy, and microarray analysis of global gene expression was conducted toassess the inflammatory and neovascularization response. Inhibitory effect of topicaldexamethasone for repeat neovascularization was assessed.

RESULTS. After initial robust neovascularization, 30 days of regression resulted in the recoveryof corneal transparency; however, a population of barely functional persistent vesselsremained at the microscopic level. Upon re-stimulation, inflammatory cell invasion, persistentvessel dilation, vascular invasion, and gene expression of Vegfa, Il1b, Il6, Ccl2, Ccl3, andCxcl2 all doubled relative to initial neovascularization. Repeat neovascularization occurredtwice as rapidly as initially, with activation of nitric oxide and reactive oxygen species, matrixmetalloproteinase, and leukocyte extravasation signaling pathways, and suppression of anti-inflammatory LXR/RXR signaling. While inhibiting initial neovascularization, a similartreatment course of dexamethasone did not suppress repeat neovascularization.

CONCLUSIONS. Persistent vessels remaining after the initial resolution of neovascularization canrapidly reactivate to facilitate more aggressive inflammation and repeat neovascularization,highlighting the importance of achieving and confirming complete vessel regression after aninitial episode of corneal neovascularization.

Keywords: corneal neovascularization, neovascularization, cornea, angiogenesis, inflamma-tion

Corneal neovascularization is a sight-threatening conditionthat can cause blindness, and may result from inflamma-

tory, infectious, and other causes, such as trauma, contact lenswear, and acquired or congenital limbal stem cell deficiency.1,2

Although to date there is no approved treatment specific forcorneal neovascularization,1 typically broad immunosuppres-sion (corticosteroids)3 and off-label antiangiogenic agents4–7 areused in the cornea, although newer therapies are undergoingclinical trials.8

These treatments, however, typically administered prophy-lactically after corneal transplantation or upon presentationwith a vascularized cornea, diminish corneal neovessels, but donot achieve complete vascular regression.3–9 Because regres-sion is not complete, some corneal neovessels inevitablyremain, and these could precipitate repeat neovascularization

if the underlying stimulus remains or where treatment isdiscontinued or loses efficacy. This repeated neovascular effectmay mimic the situation in the treatment of neovascular age-related macular degeneration (nAMD), where a single treatmentwith intravitreally injected anti-VEGF agents has only limitedeffect and is followed by robust repeat leakage and continuedchoroidal neovascularization necessitating repeated injec-tions.10 nAMD is a major disease, with global prevalence of0.37% of all persons over the age of 40.11 In addition, somestudy estimates show an approximate 30% recurrence ofchoroidal neovascularization in nAMD patients treated withanti-VEGF.12,13

Corneal transplantation is the most common transplantationprocedure performed worldwide, with over 180,000 transplan-tations performed annually.14 Of these, typically 10% experi-

Copyright 2019 The Authors

iovs.arvojournals.org j ISSN: 1552-5783 2990

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Downloaded from iovs.arvojournals.org on 11/15/2019

https://creativecommons.org/licenses/by-nc-nd/4.0/

ence an immune reaction leading to graft failure.15 In cases ofgraft failure with corneal inflammation and neovascularization,the survival rate of subsequent transplants (termed ‘high risk’)is below 35% due to recurrent neovascularization.15 In a priorstudy describing a single case of corneal graft neovasculariza-tion, it was observed that a single subconjunctival injection ofanti-VEGF antibody led to initial reduction of vessel caliberafter 1 week, but vessel caliber increased again by 3 weeks,indicating a potential rebound effect.16

In addition to the cessation of antiangiogenic treatment,repeat corneal injury, inflammation, or persisting pathologycould present a risk of repeat episodes of corneal neovascu-larization. It is well known, for example, that rejected, failedcorneal grafts carry a high risk of subsequent failure after re-transplantation due to repeat neovascularization15,17 despitethe original invading vessels being physically removed duringtransplantation. It is also well known that repeat neovascular-ization episodes can develop rapidly and are more difficult totreat. Given the continued development of improved antian-giogenic treatments and alternative treatment strategies, suchas photodynamic therapy18 and UV light corneal crosslink-ing,19 that aim to regress initial neovascularization andsubstantially regain corneal transparency, it is unclear howthe cornea would respond to repeated injury/stimulus orremoval of such antiangiogenic treatment. Specifically, it is notknown whether repeat neovascularization and the accompa-nying inflammation would occur in the same manner as theinitial neovascularization episode, or if it would differ inphenotype and aggressiveness. Knowledge of the nature andaggressiveness of repeat neovascularization could potentiallyguide decisions regarding type, dosage, and timing oftreatment.

To investigate these questions, and given the importanceof inflammation as a driving force and modulator of cornealneovascularization,20–22 we used a model of suture-inducedinflammatory neovascularization in the rat cornea to investi-gate the process of repeat neovascularization after aspontaneous and incomplete regression. The analysis wasconducted at the phenotypic, and gene expression level,comparing the second episode of neovascularization with thefirst, by longitudinal examinations conducted in the samecorneas.

METHODS

Human Data

Anonymized human photographic data were extracted from aninstitutional database of clinical slit-lamp photographs and wasused for illustrative, nonquantitative purposes only. Patientconsent was obtained as per routine clinical practice for theacquisition and storage of slit-lamp photographs, following thetenets of the Declaration of Helsinki.

Animal Experiments

Animal experiments were conducted after receiving approvalfrom the Linkoping Regional Animal Ethics Committee underethical permit number #585 and in line with the ARVOStatement for the Use of Animals in Ophthalmic and VisionResearch. Male Wistar rats 5 to 6 weeks of age (Janvier Labs,France) were used. Two 10-0 nylon sutures placed in thetemporal cornea 1.5 mm from the limbus were used to induceneovascularization, as previously described.23 Corneal neovas-cularization was allowed to proceed for seven days followingsuture placement (initial neovascularization). Both sutureswere then removed from the cornea on day 7 to inducespontaneous vessel regression, which lasted for 30 days. Onday 30, the same cornea was re-sutured as described above toinduce repeat neovascularization. Following re-suture, theanimals were examined at 24, 72, and 96 hours (Fig. 1).16,20–26

For experiments involving treatment, dexamethasone (1mg/mL Opnol; Clean Chemical Sweden AB, Borlange, Sweden)was given topically to the re-sutured eye immediately aftersuture placement and after that three times daily for 3 days.The control group was not treated.

In Vivo Imaging

In vivo confocal microscopy (IVCM) with the laser-scanningHRT3-RCM system (Heidelberg Retinal Tomograph 3 withRostock Corneal Module; Heidelberg Engineering, Heidelberg,Germany) was used to image inflammatory cells and vessels inthe cornea, as previously described.16,20 To determine vesseldiameter, IVCM images showing distinct perfused blood vesselswere selected from four animals. From each animal, threeimage sequences were selected and used to measure thediameter of vessels using the measuring tool in ImageJ software

(http://imagej.nih.gov/ij/; provided in the public domain bythe National Institutes of Health, Bethesda, MD, USA), aspreviously described.20 For inflammatory cells, images fromfour animals per experimental group were analyzed. From eachanimal, three image sequences were selected, and the numberof cells was counted manually using the cell counter tool inImageJ. Analysis of cell count and vessel diameter wasperformed by two observers masked to the treatment andtime point, and averaged values of both observers werereported.

A rodent slit-lamp microscopy imaging system (Micron III;Phoenix Research Laboratories, Pleasanton, CA, USA) was usedto monitor the overall neovascularization response. Digital slit-lamp images were acquired from all experimental animals andanalyzed by two observers masked to the treatment and timepoint. Vessel length (from the limbus to the tip of the vessels)was measured using the measuring tool in ImageJ. The numberof vessels extending from the limbus toward the central corneawas counted manually. Using ImageJ, the vascularized area wasdefined by the area of the polygon with vertices defined by the

FIGURE 1. Experimental design followed to investigate repeat neovascularization in the cornea using the suture model of inflammatory-inducedcorneal neovascularization.16,20–26

Repeat Corneal Neovascularization is More Aggressive IOVS j July 2019 j Vol. 60 j No. 8 j 2991


furthest invading vessel tips in slit-lamp photographs of thecornea and by the limbal border.

Hematoxylin and Eosin (H&E) Staining

Harvested corneal tissue was fixed in 4% (wt/vol) paraformal-dehyde in PBS for 24 hours. Fixed samples were embedded inparaffin and sectioned into 5-lm thick sections. The sectionswere placed on a glass slide and deparaffinized, stained,rehydrated, rinsed in water, and imaged.

Microarray Whole Transcriptome Analysis

Whole transcriptome analysis was performed using GeneChip2.0 ST microarrays (Affymetrix, Santa Clara, CA, USA). Freshharvested corneal tissue was stored in RNA stabilizationsolution (RNAlater; ThermoFisher, Waltham, MA, USA) at 48Cuntil use. Total RNA was extracted from a single cornea peranimal as one biological sample, with groups consisting of fourindependent biological samples per time point withoutpooling of RNA samples within any group. The RNA integritywas determined using the bioanalyzer (Agilent bioanalyzer2000; Agilent, Palo Alto, CA, USA), with a RNA integritynumber (RIN) value of 7 or more used as a cut-off for sampleinclusion for downstream analysis. RNA samples were pre-pared for microarray hybridization according to the manufac-turer’s guidelines (GeneChip WT PLUS Reagent Kit, P/N703174 Rev. 2; Affymetrix Inc.). Following hybridization in aGeneChip Hybridization Oven 645 (Affymetrix, Inc.), themicroarray chips were washed and stained in a Fluidics station450 (Affymetrix, Inc). The microarray chips were scannedusing the scanner 30007G (Affymetrix, Inc.), and the raw CELfiles were normalized to the log2 using the Ranking Analysis ofMicroarray data (RAM) method. Hierarchal cluster analysis wasperformed using the Affymetrix expression console (Affyme-trix, Inc.). Fold change expression was obtained during initialangiogenesis and revascularization by normalizing to the naıvecornea, using transcriptome analysis console (TAC; Affymetrix,Inc). Differentially expressed genes were defined using filters:one-way ANOVA P value < 0.05, and false discovery rate (FDR)< 0.05.

Pathway Enrichment and Upstream RegulatoryAnalysis

Using the obtained differentially expressed genes, pathwayenrichment analysis was performed using QIAGEN’s IngenuityPathway Analysis (IPA; QIAGEN, Valencia, CA,USA). Coreanalysis was initially performed using default parameters, andthen pathway analysis was performed to identify activated/inhibited pathways, based on the z-score. Positive and negativez-score values indicated pathway activation and inhibition,respectively. The resultant list of canonical pathways wascompared between initial neovascularization and repeatneovascularization. The upstream regulatory analysis wasperformed following the core analysis. The upstream regula-tory analysis was performed to identify activated/inhibitedupstream regulators. The obtained upstream regulators werecompared between initial neovascularization and repeatneovascularization.

Gene Expression Analysis by qPCR

Rat corneas were harvested, and total RNA was extracted fromeach rat cornea as described above. Gene expression analysiswas performed using TaqMan fast advanced master mix(PN:4444554; ThermoFisher) using custom primers for Vefga

(PN4351370), Ccl3 (PN4448892), Cxcl2 (PN4453320), Il1b

(1209639 C7), Timp1 (1454605 H5), Il6 (PN4453320), andGapdh (PN4351370). Using Power SYBR Green PCR Mastermix (Applied Biosystems) gene expression for Ccl2 withprimer sequences (F-ATGCAGTTAATGCCCCACTC and R-TTCCTTATTGGGGTCAGCAC) and Gapdh (F-ATGGT-GAAGGTCGGTGTGAA and R-TGACTGTGCCGTTGAACTTG)was performed. Threshold cycle (Ct) values were normalizedto Gapdh, and fold change was calculated relative to the naıvecornea.

Statistical Analysis

The Student’s t-test was used when comparing two samplemeans from normally distributed data, while the nonparamet-ric Mann-Whitney U test was used where data were notnormally distributed. One-way ANOVA with Tukey’s post hocmultiple comparisons of normally distributed data was used forgroup analysis. A P value < 0.05 was considered significant.Statistical tests were performed using GraphPad Prism 8 forWindows, GraphPad Software (La Jolla, CA, USA) and dataerror bars are presented as the standard error of measurement(SEM).

RESULTS

Human Case of Corneal Neovascularization andGhost Vessels Following Regression

Clinically, corneal neovascularization (Fig. 2A) can result fromphysical or chemical injury, surgical procedures, foreign bodyreactions, localized hypoxia, limbal stem cell deficiency, orfrom infections.1,27,28 When successfully treated, the regres-sion of pathologic vessels leaves behind threadlike strands inthe regressed region after antiangiogenic treatment (Fig. 2B),but the structural and flow characteristics of these features,and their fate during re-stimulation of neovascularization areunknown.

Revascularization is Characterized by RapidHyperdilation of Persistent Vessels WithoutVascular Leakage

Sutures initially placed in the rat cornea induced neovascular-ization after 7 days (Figs. 3A, 3F), with live imaging by IVCMrevealing inflammatory cell infiltration and neovessel invasion(Figs. 3F–J). Subsequent removal of the suture led to clinicalregression with approximately 50% of original vessels lackingflow after a regression period of 30 days, and the remainingvessels being severely constricted with minimal flow, resultingin a dramatic restoration of corneal transparency as observedin the slit lamp (Fig. 3B). A population of persistent vesselsremained, however, and appeared as sparse, thin thread-likestructures evident by IVCM (Fig. 3G) and detailed slit-lampobservation, with markedly constricted lumen supportingmainly plasma flow, with only intermittent serial erythrocytepassage observed (Supplementary Video S1). By IVCM,extravascular inflammatory cells were almost completelyabsent in this regressed state (Fig. 3G). Following re-suture ofthe same cornea after 30 days of regression, revascularizationwas characterized by an influx of inflammatory cells andperfusion of the persistent vessels, starting as soon as 1 dayafter re-suture (Figs. 3C, 3H). By day 3, a dramatic hyperdilationof the same vessels was observed, with vessel dilationincreasing daily (Figs. 3D, 3E, 3I–K). The sequence of eventsobserved in vivo was similarly observed in fixed specimens byH&E staining (Fig. 3L). Notably, the vessels observed duringrepeat neovascularization did not appear leaky, with the



https://arvo.silverchair-cdn.com/arvo/content_public/journal/iovs/938091/iovs-60-08-06_s01.mp4?Expires=1562947788&Signature=QUXUgKsw37o5e3C4lZfxHeX03R0Px56XYiF8OwpjL9G9q3dQDuPO1wrDsW~YME4L1TDqAQ6s2YtJTnpt3VCH4logTaYKukV6rTulUiAQbMbgHwMkHCPdjgo9ir-TAji4RbAgj~1DFHPiEYtf3kJS1aPpbgZwNiwq3gEjtioT6Kn~RHZoqedct6bPKVlmuwBkjs83wy0e6Yql44yAkyyPtOytnW8nQbgYDbuH3fKp0sUBG7n6FfLC8GbuSP4vTa0OQ4ej94S-NnI5a8fCuezFlstAzyHcXR~vWveCTF2qgQgKrRLe4sFHiRr-lpPH7YQdh8fnS3dlC-s8eCTDtEwFuA__&Key-Pair-Id=APKAIE5G5CRDK6RD3PGA

FIGURE 2. Human case with active corneal neovascularization (arrow) before vessel regression (A) and a human case after regression of cornealneovessels with thin remnant vascular structures indicated by the arrows (B). The latter was a case of Herpes Simplex keratitis withneovascularization resolving after antiviral treatment combined with corticosteroids.

FIGURE 3. Longitudinal imaging of the same cornea indicating more rapid and aggressive neovascular phenotype during repeat cornealneovascularization. (A–E) Time series slit-lamp images are taken from the same cornea during initial, regression, and revascularization phases,indicating more rapid and aggressive neovascular response after re-suturing. (F–J) Corresponding IVCM images at each time point in the same eye,indicating the absence of inflammatory cells in the tissue after regression, but the presence of thin persistent vessels. Inflammation andhyperdilation of persistent vessels were observed during the revascularization phase. (K) The diameter of perfused, persistent vessels measuredfrom IVCM images indicated increased vessel diameter with time (ANOVA P < 0.0001) (black asterisks indicate pairwise comparisons relative toregressed state on day 30). During repeat neovascularization, the vessel diameter increased with time (asterisks in green indicate pairwisecomparisons relative to the prior time point during repeat neovascularization). (L) H&E sections indicated small sprouting vessels visible on day 4 ofinitial neovascularization (arrows). By day 7 vessels were larger with increased perfusion indicated by increased erythrocyte presence (arrows).After 30 days of regression, narrow vessels (arrows) with few erythrocytes (mainly plasma flow) were evident. By day 4 of repeatneovascularization, vessels became enlarged and hyperperfused with many erythrocytes (arrow). Note also the dense infiltration of inflammatorycells under the epithelium (asterisk). Scale bars in (L) are 20 lm. The error bars in (K) represent SEM, n¼5, and asterisks indicate P value < 0.05.Scale bars in IVCM images are 50 lm.



corneal tissue in nonvascularized regions remaining transpar-ent and relatively free of edema.

Inflammation is Enhanced During RepeatNeovascularization

Further morphologic changes were noted during repeatneovascularization. Relative to 4 days after the originalneovascular response, the hyperdilated persistent vessels 4days after induction of repeat neovascularization extendedtwice the distance from the limbus into the cornea (P ¼ 0.02,Figs.4A, 4B). Vessel caliber additionally doubled (from a meanof 24.1 to 52.2 lm, P¼ 0.001, Figs. 4C, 4 D), and the density ofinflammatory cells invading the cornea also doubled (P¼ 0.03,Figs. 4E, 4F).

Repeat Neovascularization has a UniqueTranscriptomic Signature and is Characterized byEnhanced Inflammatory Gene Expression

To investigate gene regulation of the observed inflammatoryresponse, comparative whole-transcriptome expression analy-sis was performed. Using gene chip microarrays in initial andrepeat neovascularization phases normalized to naıve corneas,differentially expressed genes (DEG) were compared. Principlecomponent analysis and hierarchical cluster analysis revealedclustering within groups and separation between groups,indicating a unique overall transcriptomic signature of repeatneovascularization, distinct from the initial phase (Figs. 5A,

5B). Of the DEG in the initial phase, inflammatory genes Ccl2,S100a9, and Cxcl6 were among the most upregulated with anaverage fold change of 26.6 among them (Fig. 5C). In therepeat phase, Ccl2, S100a9, and Cxcl6 were still among themost upregulated genes but with an average fold change of 55among them (Fig. 5E). Expression of genes with fold change 50or more, 20 or more, and 10 or more yielded higher numbersof genes in each category during the repeat relative to initialphase (Fig. 5D). Gene expression was stronger during repeatneovascularization compared with the initial phase fornumerous inflammatory and angiogenesis-related genes, in-cluding Timp1, Il6, Mmp9, Socs3, Icam, Vegfa, and Cxcr4 (P <0.05, Fig. 5F).

Inflammation-Related Pathways and Genes andTheir Putative Upstream Regulators are MoreStrongly Activated During RepeatNeovascularization

To further investigate regulatory control of repeat neovascu-larization, pathway enrichment and upstream regulatoryanalysis were performed using the microarray data. Pathwayenrichment analysis revealed a differing pattern of pathwayactivation and inhibition during the repeat relative to the initialphase (Figs. 6A, 6B). During the repeat phase, a higher numberof inflammatory pathways were enriched and to a higherdegree. Notably, the ‘leukocyte extravasation’ pathway activa-tion increased, while the dual pathways ‘LPS/IL-1–mediatedinhibition of RXR function’ and ‘LXR/RXR activation’ had a

FIGURE 4. Repeat neovascularization is characterized by increased inflammation at the tissue level, relative to the initial neovascular response. (A)Slit-lamp photographs indicate the neovascularization response on day 4 of initial angiogenesis and day 4 of repeat neovascularization, in the samecornea. (B) Revascularization was of a stronger magnitude, expressed as the distance of neovessel invasion into the cornea from the limbal arcade,relative to the suture position. (C) Typical IVCM images with vessels on day 4 of initial angiogenesis and day 4 of revascularization indicating thehyperdilated phenotype. (D) During repeat neovascularization, vessel diameter was doubled (asterisk, P¼0.001). (E) IVCM images of inflammatorycell infiltration into the corneal stroma just under the epithelium on day 4 of the initial and repeat neovascular phases. (F) Infiltration ofinflammatory cells into the cornea increased during revascularization as indicated by the average number of cells per IVCM field in the subepithelialregion (asterisk, P¼ 0.03). n ¼ 4 per group. Scale bars in the IVCM images ¼ 50 lm.



stronger proinflammatory profile.25 Specific to the repeatphase, both innate and adaptive immune pathways wereactivated, as well as the production of nitric oxide (NO) andreactive oxygen species (ROS) pathway. The upstreamregulatory analysis revealed that in initial neovascularization,Il1b was the most active upstream regulator with activationscore 3.8, and with a P value of overlap (between the DEG andQIAGEN knowledge base) of 7.5E�04 (Fig. 6C). During repeatneovascularization, Il1b remained the most active regulator,but with increased activation score 7.4 and P value 1.1E�16(Fig. 6D). Other upstream regulators activated strongly duringrepeat neovascularization were Vegfa and Il6 with activationscores 4.5 and 6.25, respectively, and P values 1.5E�10 and4.1E�09, respectively. To corroborate these results, geneexpression of several inflammatory regulators in the cornealtissue was examined by qPCR. Genes Vegfa, Il6, Il1b, Ccl2,Ccl3, and Cxcl2 were more strongly upregulated during repeatneovascularization compared with the initial phase, with thegeneral trend of gene expression by qPCR confirming thatobtained by microarray analysis (Fig. 6E).

Dexamethasone Inhibits Initial NeovascularizationBut is Ineffective in Suppressing Vasodilation andInflammation During Repeat Neovascularization

Previously, we showed that dexamethasone, a corticosteroidbroadly targeting inflammation, vasodilation, and angiogenesispotently suppresses initial corneal neovascularization.20,23 Todetermine whether corticosteroid treatment could also sup-press the repeat neovascular response, dexamethasone wasadministered topically to the re-sutured eye immediatelyfollowing re-suture (referred to here as day 0) and for 3 days.Analysis of slit-lamp images at days 1, 3, and 4 revealed thatrepeat neovascularization still occurred in both nontreatedcontrol and dexamethasone-treated groups (Fig. 7), and wastherefore not suppressed by the treatment.

In vivo imaging in the treated groups (Figs. 8A–C) revealed asimilar inflammatory cell infiltration occurred in both groups(Fig. 8 D) and persistent vessels dilated similarly in both groupson day 1 but dexamethasone treatment suppressed dilation onday 3 relative to controls (Fig. 8E). Despite this, dexamethasone

FIGURE 5. Gene regulation of repeat versus initial corneal neovascularization as revealed by whole transcriptome microarray expression analysis.(A) Principle component analysis indicates separation and grouping of individual biological samples according to phase. (B) Hierarchical clusteranalysis of whole transcriptome data showing sample partitioning with neovascular phase. (C, E) Volcano plots indicating the distribution of foldchange of DEG relative to the naıve cornea, with a P value cut-off < 0.05. Note the increased number and expression level of inflammatory genes inthe repeat phase (red dots). (D) Comparison of fold change expression of selected genes between day 4 of initial and day 4 of repeatneovascularization. More genes were represented in each category during the repeat phase. (F) Comparison of fold change expression of selectedinflammatory and angiogenesis genes at day 4, indicating higher expression of many genes in the repeat phase. *P < 0.05, n ¼ 4. Error bars

represent SEM.



FIGURE 6. Whole transcriptome pathway and upstream regulatory gene expression analysis of repeat corneal neovascularization. (A, B) Analysis ofenriched pathways based on whole transcriptome DEG during initial and repeat phases of corneal neovascularization. Blue bars indicate inhibitedpathways while orange bars indicate activated pathways, with corresponding inhibition and activation z-scores, respectively. Pathway activationand z-scores increased during repeat neovascularization and consisted primarily of innate and adaptive immune pathways. (C, D) Volcano plots ofupstream regulators of the identified pathways indicated stronger activation of regulators during repeat neovascularization. (E) Fold changeexpression of several key inflammatory genes and upstream regulators as determined by microarray and separate qPCR analysis, compared acrossnaıve, day 4 initial and day 4 repeat neovascularization groups. ANOVA with multiple comparisons was used to compare fold change expression. N¼4 samples per group. Error bars represent SEM. ****P < 0.0001, ***P < 0.001, **P < 0.01, *P < 0.05.



treatment did not suppress repeat neovascularization of thetissue as indicated by the presence of neovessels, resulting in alarger vascularized area relative to suppressed initial neovascu-larization after treatment (Figs. 8F, 8G). Also, dexamethasonetreatment did not reduce vascular area relative to untreatedcontrols. While dexamethasone was 91.2% effective in sup-pressing initial neovascularization, with the same treatmentregimen, it was only 23.7% effective in reducing the vascularizedarea during repeat neovascularization (Fig. 8H).

In summary, inflammation and the neovascular response areenhanced during repeat neovascularization of the corneaowing to multiple pathways and factor expression (Fig. 9).

DISCUSSION

A stronger inflammatory and neovascular response in thetissue was evident, and numerous genes were more stronglyexpressed during repeat neovascularization. Vegfa, a potentvasodilator,29 was upregulated to a significantly greater degreeduring repeat neovascularization, along with other proinflam-matory and proangiogenic genes, such as IL1b, IL-6, Ccl2, Ccl3,

Cxcr2, Mmp9, Icam-1, and Cxcr4. Previous studies showedthat inhibition of Ccl2 and IL1b suppresses corneal neovascu-larization by inhibiting inflammatory cell recruitment30 in themouse cornea. Also, Ccl3 acting through Ccr5 has been shownto result in macrophage infiltration in the cornea to promoteneovascularization.31

Cxcr2, a key receptor for chemokinesignaling, promotes angiogenesis by facilitating chemotaxis ofmicrovascular endothelial cells in a mouse corneal neovascu-larization model.32

Pathway enrichment analysis revealed NO and ROSproduction pathways activated during repeat neovasculariza-tion but not during the initial phase. NO is a strongvasodilator,33 and ROS is reported to regulate vasodilationand vascular permeability34,35 promoting vessel remodeling inresponse to high-flow conditions.16 ROS additionally functionsas an endothelial cell (EC) activator leading to EC phenotypechange and increase in permeability36 and plays a vital role intissue reperfusion.37 Moreover, ROS produced by inflammatorycells at the site of inflammation has been shown to promoteleukocyte migration across the endothelial barrier.38 Thefunction of NO, ROS, and VEGFA signaling provides strongevidence corroborating the rapid reconfiguration of persistentvessels and strong vasodilation responses observed in thisstudy. The observed inflammatory response is also likely to beregulated by enhanced activation of leukocyte extravasationsignaling pathway. It is important here to note that the repeatneovascularization occurred in an incompletely regressedvascular bed containing persistent, mature vessels. Althoughvery narrow and barely visible upon slit-lamp examination,these mature vessels responded rapidly and dramatically tohyperdilate, but without the leakage normally associated withnew sprouting vessels. Repeat neovascularization, acting onestablished and mature vessels, may differ fundamentally frominitial neovascularization, where new, leaky sprouts invade thecornea and result in edema. At the pathway level, NO and ROSsignaling (among other pathways and genes) were notactivated in the initial but only in the repeat phase, suggestingdifferential regulatory control of initial versus repeat episodesthat could be driven by the vessel phenotype.

Anti-inflammatory pathways were also modulated duringthe repeat phase. LXR/RXR signaling was doubly inhibited,directly through inhibition of anti-inflammatory pathway LXR/RXR activation39,40 and indirectly by activation of the LPS/IL-1–mediated inhibition of RXR function pathway. This is theopposite of the LXR/RXR activation profile observed duringvessel regression in this model,25 indicating an upregulation ofinflammation to reverse the regression phenotype during therepeat phase of neovascularization. Activation of these andother inflammatory pathways may partially be attributed toupstream regulators, such as Il1b, Vegfa, and Il6 that weremore strongly activated during repeat neovascularization.Interestingly, Th1, Th2, and interferon signaling pathwayswere additionally activated, suggesting a possible role of theadaptive immune response in priming the tissue for faster andmore aggressive repeat neovascularization. In line with this,CD8þ T cells were shown to be important for cornealneovascularization in HSV-1 infected mice.41 Notably, stimulat-ed T helper cells are a source of VEGF,42,43 while adaptiveimmune activation has been linked to the pathogenesis ofnAMD.44 The involvement of other vessel types, such aslymphatics during revascularization, is open for discussion,given that it has been shown that lymphangiogenesis occursalongside hemangiogenesis, for example during inflammationin penetrating keratoplasty.45 Using our rat model of suture-induced corneal angiogenesis however, we observed previous-ly that lymphangiogenesis occurs typically at 14 days afterinduction of angiogenesis,46 and lymphatics were therefore notobserved during initial angiogenic stimulation in the present

FIGURE 7. Serial slit-lamp images of the same nontreated anddexamethasone-treated corneas during the repeat phase of neovascu-larization. The left column indicates nontreated cornea, and the right

column indicates dexamethasone-treated cornea. In both groups, re-suturing was followed by rapid reperfusion of vascular loops and denovo sprouting occurred from these persistent vessels on day 4(arrows in the zoomed region corresponding to dashed circles indicatenew angiogenic sprouts).



study. Interestingly, however, recurrent inflammation isthought to accelerate the development of functional lymphaticvessels,47 in conjunction with enhanced inflammation asobserved here during revascularization. In addition, macro-phages, cells we have previously shown to dominate aregressing corneal vascular bed,21 are known to supportlymphangiogenesis in many ways, including by transdifferen-tiation into lymphatic endothelial cells.48 A recent studyshowed that targeting lymphatics promotes graft survival viaa reduced activation of the immune system,49 suggestinglymphatics as mediators of immunological memory, that mayimpact rapid corneal revascularization.

Although highly effective in suppressing initial neovascu-larization, the broad-acting corticosteroid dexamethasone wasineffective in inhibiting repeat neovascularization. This resultcorroborates the gene regulatory and tissue-level evidence ofincreased inflammation and angiogenesis during revasculariza-

tion, spanning multiple pathways and numerous genes. Theentire transcriptome is altered during repeated neovasculariza-tion, and this alteration rendered dexamethasone only mini-mally effective in suppressing vasodilation and ineffective inpreventing new angiogenic sprouting. This suggests that atissue primed for revascularization may require strongersuppression of inflammation and angiogenesis; however, ahigher dosage of corticosteroids may increase the risk forglaucoma50 and cataract51 development in the eye. Alterna-tively, other nonsteroidal approaches should be considered forsuppressing the aggressive repeat neovascularization response.

In the present study, using longitudinal in vivo imagingserially in the same corneas, it was observed that resolution ofthe initial neovascular response—in this model triggered bythe removal of the initial pathologic stimulus—led to therestoration of corneal transparency. Although this transparencywas clinically evident paralleling the ghost vessels observed in

FIGURE 8. Dexamethasone treatment efficacy in initial versus repeat neovascularization. (A–C) IVCM images indicate inflammatory cell infiltrationinto the cornea and hyperdilation of persistent vessels 1 and 3 days after re-stimulation in both nontreated and dexamethasone-treated groups, withtreatment initiated at the time of re-stimulation (after 30 days of regression). Quantification of inflammatory cell invasion (D) and vasodilation (E) byIVCM did not reveal an effect of treatment during the repeat neovascularization phase, except reduced vasodilation at day 3 in dexamethasone-treated corneas. (F) The vascularized area of the cornea quantified from slit-lamp images at day 4 of repeat neovascularization indicated a lack ofsuppression of vascular invasion by dexamethasone despite a robust suppressive effect during initial angiogenesis. (G) The efficacy ofdexamethasone in suppressing initial angiogenesis was clinically apparent by slit-lamp examination while a lack of effect was observed duringrevascularization with a large vascularized area of the cornea persisting. (H) Quantification of vascularized areas expressed as a percentagereduction relative to nontreated corneas indicated dexamethasone efficacy of 91.2% in suppressing initial neovascularization and 23.7% insuppressing repeat neovascularization. Error bars represent SEM, with n¼ 4 samples per group. ANOVA with multiple comparisons was used forwithin-group comparison. Scale bars, (A–C) 50 lm.



human cases after antiangiogenic treatment, careful slit-lampimaging revealed that several thin, highly constricted vesselswith minimal flow persisted in the tissue. Vessel constriction isa known response during vascular regression,52 and here it wasobserved that such constriction severely limited the flow oferythrocytes. These persistent vessels rapidly (within 3 days)revascularized a large area of the tissue upon re-stimulation,much faster than would be possible by new vessel sprouting.Revascularization was principally due to hyperdilation of thepersistent vessels, already beginning at 24 hours after re-suture.New vessel sprouting was a secondary response that wasinitiated only 3 days after re-suture. These observations implythat a less than 100% effective treatment of corneal neovascu-larization will lead to only partially regressed vessels thatpersist in the tissue. Ending treatment or continued underlyingpathologic stimulus could then rapidly and forcefully reactivatethese vessels to mount a significant repeat neovascularizationresponse.

In summary, repeat corneal neovascularization is a rapidprocess characterized by a stronger inflammatory and vasodi-lation response mediated by activation of numerous proin-flammatory signaling pathways and genes, as well as inhibitionof anti-inflammatory signaling. Rapid tissue reperfusion origi-nates from incompletely regressed vessels and immune-mediated processes that may prime the corneal tissue for amore aggressive neovascular phenotype after repeated insult,injury, or removal of antiangiogenic treatment. This highlightsthe necessity of achieving as complete vessel regression aspossible, confirmation of complete regression, and develop-ment of strategies targeting an enhanced and more aggressiveinflammatory and neovascular response in a repeat phase.

Acknowledgments

The authors thank the contribution Camilla Hildesjo from thedepartments of Clinical Pathology and Clinical Genetics, RegionOstergotland (Linkoping, Sweden) for technical assistance withtissue embedding and sectioning.

Supported by a grant from the Swedish Research Council (GrantNo. 2012–2472; Stockholm, Sweden), and a stipend from theOgonfonden 2019 (Enskede, Sweden).

Disclosure: A. Mukwaya, None; P. Mirabelli, None; A. Lenni-kov, None; M. Thangavelu, None; L. Jensen, None; B. Peebo,None; N. Lagali, None

References

1. Roshandel D, Eslani M, Baradaran-Rafii A, et al. Current andupcoming therapies for corneal neovascularization. Ocul Surf.2018;16:398–414

2. Chang J-H, Gabison EE, Kato T, Azar DT. Corneal neovascu-larization. Curr Opin Ophthalmol. 2001;12:242–249

3. Cursiefen C, Wenkel H, Martus P, et al. Impact of short-termversus long-term topical steroids on corneal neovasculariza-tion after non-high-risk keratoplasty. Graefes Arch Clin Exp

Ophthalmol. 2001;239:514–521.

4. Kim SW, Ha BJ, Kim EK, Tchah H. The effect of topicalbevacizumab on corneal neovascularization. Ophthalmology.2008;115:e33–e38.

5. Bahar I, Kaiserman I, McAllum P, Rootman D, Slomovic A.Subconjunctival bevacizumab injection for corneal neovascu-larization. Cornea. 2008;27:142–147.

6. Ferrari G, Dastjerdi MH, Okanobo A, et al. Topical ranibizu-mab as a treatment of corneal neovascularization. Cornea.2013;32:992–997.

7. Dastjerdi MH, Al-Arfaj KM, Nallasamy N, et al. Topicalbevacizumab in the treatment of corneal neovascularization:results of a prospective, open-label, noncomparative study.Arch Ophthalmol. 2009;127:381–389.

8. Cursiefen C, Viaud E, Bock F, et al. Aganirsen antisenseoligonucleotide eye drops inhibit keratitis-induced cornealneovascularization and reduce need for transplantation: the I-CAN study. Ophthalmology. 2014;121:1683–1692.

9. Gupta D, Illingworth C. Treatments for corneal neovascular-ization: a review. Cornea. 2001;30:927–938.

FIGURE 9. Graphic depiction of phenotypical and regulatory changes during repeat neovascularization. Incomplete regression leaves behindnarrow persistent vessels in the tissue that upon re-stimulation rapidly dilate leading to a hyperperfusion of the persistent vessels. Subsequently, denovo sprouting from these vessels is evident by day 4 after repeat stimulation. Many inflammatory cells also infiltrate the corneal stroma starting onday 1. Pathways, such as leukocyte extravasation signaling, production of NO and ROS signaling, and LPS/IL-1–mediated inhibition of RXR function,are activated to enhance revascularization, and at the same time pathways, such as LXR/RXR activation and inhibition of matrix metalloproteasessignaling, are inhibited. Upregulation of pro-inflammatory genes, such as Ccl2, Il1b, Ccl3, Cxcl2, further promotes and sustains the inflammatoryresponse. Upstream regulators, such as VEGFA, Il1b, and Il6, are potential early mediators of the repeat neovascularization response.



10. Martin DF, Maguire MG, Ying GS, Grunwald JE, Fine SL, JaffeGJ; CATT Research Group. Ranibizumab and bevacizumab forneovascular age-related macular degeneration. N Engl J Med.2011;364:1897–1908.

11. Wong WL, Su X, Li X, et al. Global prevalence of age-relatedmacular degeneration and disease burden projection for 2020and 2040: a systematic review and meta-analysis. Lancet Glob

Health. 2014;2:e106–e116.

12. Adrean SD, Chaili S, Grant S, Pirouz A. Recurrence rate ofchoroidal neovascularization in neovascular age-related mac-ular degeneration managed with a treat–extend–stop proto-col. Ophthalmol Retina. 2018;2:225–230.

13. Potter MJ, Szabo SM. Recurrence of choroidal neovascularisa-tion after photodynamic therapy in patients with age-relatedmacular degeneration. Br J Ophthalmol. 2007;91:753–756.

14. Gain P, Jullienne R, He Z, et al. Global survey of cornealtransplantation and eye banking. JAMA Ophthalmol. 2016;134;167–173.

15. Di Zazzo A, Kheirkhah A, Abud TB, Goyal S, Dana R.Management of high-risk corneal transplantation. Surv

Ophthalmol. 2017;62:816–827.

16. Peebo BB, Fagerholm P, Traneus-Rockert C, Lagali N. Cellularlevel characterization of capillary regression in inflammatoryangiogenesis using an in vivo corneal model. Angiogenesis.2011;14:393–405.

17. Kelly T-L, Coster DJ, Williams KA. Repeat penetrating cornealtransplantation in patients with keratoconus. Ophthalmology.2011;118:1538–1542.

18. Hou Y, Le VNH, Clahsen T, Schneider AC, Bock F, Cursiefen C.Photodynamic therapy leads to time-dependent regression ofpathologic corneal (lymph) angiogenesis and promotes high-risk corneal allograft survival. Invest Ophthalmol Vis Sci.2017;58:5862–5869.

19. Hou Y, Le VNH, Toth G, et al. UV light crosslinking regressesmature corneal blood and lymphatic vessels and promotessubsequent high-risk corneal transplant survival. Am J

Transplant. 2018;18:2873–2884.

20. Mirabelli P, Peebo BB, Xeroudaki M, Koulikovska M, Lagali N.Early effects of dexamethasone and anti-VEGF therapy in aninflammatory corneal neovascularization model. Exp Eye Res.2014;125:118–127.

21. Mukwaya A, Peebo B, Xeroudaki M, et al. Factors regulatingcapillary remodeling in a reversible model of inflammatorycorneal angiogenesis. Sci Rep. 2016;6:32137.

22. Lennikov A, Mirabelli P, Mukwaya A, et al. Selective IKK2inhibitor IMD0354 disrupts NF-jB signaling to suppresscorneal inflammation and angiogenesis. Angiogenesis. 2018;21:267–285.

23. Mirabelli P, Mukwaya A, Lennikov A, et al. Genome-wideexpression differences in anti-Vegf and dexamethasonetreatment of inflammatory angiogenesis in the rat cornea.Sci Rep. 2017;7:7616.

24. Mukwaya A, Mirabelli P, Lennikov A, et al. Genome-wideexpression datasets of anti-VEGF and dexamethasone treat-ment of angiogenesis in the rat cornea. Sci data. 2017;4:170111.

25. Mukwaya A, Lennikov A, Xeroudaki M, et al. Time-dependentLXR/RXR pathway modulation characterizes capillary remod-eling in inflammatory corneal neovascularization. Angiogen-

esis. 2018;21:395–413.

26. Mukwaya A, Lindvall JM, Xeroudaki M, et al. A microarraywhole-genome gene expression dataset in a rat model ofinflammatory corneal angiogenesis. Sci Data. 2016;3:160103.

27. Cursiefen C, Colin J, Dana R, et al. Consensus statement onindications for anti-angiogenic therapy in the management ofcorneal diseases associated with neovascularisation: outcomeof an expert roundtable. Br J Ophthalmol. 2012;96:3–9.

28. Zhong W, Montana M, Santosa SM, et al. Angiogenesis andlymphangiogenesis in corneal transplantation–a review. Surv

Ophthalmol. 2018;63:453–479.

29. Dvorak HF, Brown LF, Detmar M, Dvorak AM. Vascularpermeability factor/vascular endothelial growth factor, micro-vascular hyperpermeability, and angiogenesis. Am J Pathol.1995;146:1029–1039.

30. Yoshida S, Yoshida A, Matsui H, Takada Y, Ishibashi T.Involvement of macrophage chemotactic protein-1 andinterleukin-1b during inflammatory but not basic fibroblastgrowth factor–dependent neovascularization in the mousecornea. Lab Invest. 2003;83:927–938.

31. Lu P, Li L, Wu Y, Mukaida N, Zhang X. Essential contribution ofCCL3 to alkali-induced corneal neovascularization by regulat-ing vascular endothelial growth factor production by macro-phages. Mol Vis. 2008;14:1614–1622.

32. Addison CL, Daniel TO, Burdick MD, et al. The CXCchemokine receptor 2, CXCR2, is the putative receptor forELRþ CXC chemokine-induced angiogenic activity. J Immu-

nol. 2000;165:5269–5277.

33. Palmer RM, Ashton D, Moncada S. Vascular endothelial cellssynthesize nitric oxide from L-arginine. Nature. 1988;333:664–666.

34. Kenyon BM, Voest EE, Chen CC, Flynn E, Folkman J, D’AmatoRJ. A model of angiogenesis in the mouse cornea. Invest

Ophthalmol Vis Sci. 1996;37:1625–1632.

35. Deutsch TA, Hughes WF. Suppressive effects of indomethacinon thermally induced neovascularization of rabbit corneas.Am J Ophthalmol. 1979;87:536–540.

36. Alom-Ruiz SP, Anilkumar N, Shah AM. Reactive oxygen speciesand endothelial activation. Antioxid Redox Signal. 2008;10:1089–1100.

37. Kalogeris T, Bao Y, Korthuis RJ. Mitochondrial reactive oxygenspecies: a double edged sword in ischemia/reperfusion vspreconditioning. Redox Biol. 2014;2:702–714.

38. Mittal M, Siddiqui MR, Tran K, Reddy SP, Malik AB. Reactiveoxygen species in inflammation and tissue injury. Antioxid

Redox Signal. 2014;20:1126–1167.

39. Joseph SB, Castrillo A, Laffitte BA, Mangelsdorf DJ, Tontonoz P.Reciprocal regulation of inflammation and lipid metabolismby liver X receptors. Nat Med. 2003;9:213–219.

40. Walcher D, Kummel A, Kehrle B, et al. LXR activation reducesproinflammatory cytokine expression in human CD4-positivelymphocytes. Arterioscler Thromb Vasc Biol. 2006;26:1022–1028.

41. Conrady CD, Zheng M, Stone DU, Carr DJ. CD8þ T cellssuppress viral replication in the cornea but contribute toVEGF-C–induced lymphatic vessel genesis. J Immunol. 2012;189:425–432.

42. Mor F, Quintana FJ, Cohen IR. Angiogenesis-inflammationcross-talk: vascular endothelial growth factor is secreted byactivated T cells and induces Th1 polarization. J Immunol.2004;172:4618–4623.

43. Lee CG, Link H, Baluk P, et al. Vascular endothelial growthfactor (VEGF) induces remodeling and enhances T H 2-mediated sensitization and inflammation in the lung. Nat

Med. 2004;10:1095–1103.

44. Camelo S. Potential sources and roles of adaptive immunity inage-related macular degeneration: shall we rename AMD intoautoimmune macular disease? Autoimmune Dis. 2014;2014:532487.

45. Cursiefen C, Chen L, Borges LP, et al. VEGF-A stimulateslymphangiogenesis and hemangiogenesis in inflammatoryneovascularization via macrophage recruitment. J Clin Invest.2004;113:1040–1050.

46. Peebo BB, Fagerholm P, Traneus-Rockert C, Lagali N. Cellular-level characterization of lymph vessels in live, unlabeled



corneas by in vivo confocal microscopy. Invest Ophthalmol

Vis Sci. 2010;51:830–835.

47. Kelley PM, Connor AL, Tempero RM. Lymphatic vessel

memory stimulated by recurrent inflammation. Am J Pathol.

2013;182:2418–2428.

48. Maruyama K, Ii M, Cursiefen C, et al. Inflammation-induced

lymphangiogenesis in the cornea arises from CD11b-positive

macrophages. J Clin Invest. 2005;115:2363–2372.

49. Gurung HR, Carr MM, Carr DJ. Cornea lymphatics drive theCD8þ T-cell response to herpes simplex virus-1. Immunol

Cell Biol. 2017;95:87–98.

50. Kersey J, Broadway D. Corticosteroid-induced glaucoma: areview of the literature. Eye. 2006;20:407–416.

51. Urban RC Jr, Cotlier E. Corticosteroid-induced cataracts. Surv

Ophthalmol. 1986;31:102–110.

52. Ausprunk D. The sequence of events in the regression ofcorneal capillaries. Lab Invest. 1978;38:284–294.



repeat corneal neovascularization is characterized by more...

Documents