Genome-wide expression datasets of anti-VEGF and dexamethasone treatment of angiogenesis in the rat cornea

Therapeutics against pathologic new blood vessel growth, particularly those targeting vascular endothelial growth factor (VEGF) are of enormous clinical interest. In the eye, where anti-VEGF agents are in widespread clinical use for treating retinal and corneal blindness, only partial or transient efficacy and resistance to anti-VEGF agents are among the major drawbacks. Conversely, corticosteroids have long been used in ophthalmology for their potency in suppressing inflammation and angiogenesis, but their broad biological activity can give rise to side effects such as glaucoma and cataract. To aid in the search for more targeted and effective anti-angiogenic therapies in the eye, we present here a dataset comparing gene expression changes in dexamethasone versus anti-Vegfa treatment of inflammation leading to angiogenesis in the rat cornea. Global gene expression analysis with GeneChip Rat 230 2.0 microarrays was conducted and the metadata submitted to Expression Omnibus repository. Here, we present a high-quality validated dataset enabling genome-wide comparison of genes differentially targeted by dexamethasone and anti-Vegf treatments, to identify potential alternative therapeutic targets for evaluation.


Background & Summary
Pathological angiogenesis is a hallmark of many blinding diseases such as age-related macular degeneration, diabetic retinopathy, and corneal graft failure. In the eye, angiogenesis (or neovascularisation) is favoured by stimuli such as inflammation 1 , immunologic rejection, limbal stem cell deficiency, and hypoxia 2 . The choice of treatment for neovascularisation in the eye can be influenced by the stage of the neovascularisation process. While newly formed sprouts are usually leaky and susceptible to anti-VEGF treatment 3 , mature vessels require fewer angiogenic mediators 4 necessitating other angioregressive therapies 5 . In the cornea in particular, therapeutics are used off-label to manage corneal angiogenesis, given the current lack of approved treatments specific for the cornea. Corticosteroids and/or nonsteroidal anti-inflammatory drugs (NSAIDs) are used to target neovascularisation 6,7 , however, steroids may promote infection 8,9 , glaucoma 10 , cataract 11 , and herpes simplex recurrence 12,13 while NSAIDs are reported to cause corneal ulceration and melting 14 .
To date, targeting the VEGF family of pro-angiogenic factors is the most promising therapeutic approach as an alternative to steroid or NSAID use. Bevacizumab, a full-length recombinant humanized monoclonal antibody against VEGFA has been reported to treat corneal neovascularisation with some success 15,16 . Ranibizumab, a monoclonal antibody fragment also targeting VEGFA has similarly been shown to treat corneal neovascularisation 17 and corneal graft rejection 18 . Compared to bevacizumab and ranibizumab, pegaptanib (a 28-base ribonucleic aptamer) is of limited efficacy, since it is specific to the VEGF-165 isoform 19 . The VEGF trap (aflibercept), a decoy receptor for all VEGFA isoforms, is the latest therapeutic approach being tested for use in the cornea 20 . In spite of the benefits of these anti-VEGF targeted therapies, some side effects of VEGF targeting have been documented, for example impaired corneal epithelialization and wound healing, a pathology attributed to the possible adverse effects of VEGF deficiency on nerve growth and regeneration 21,22 . In addition, anti-VEGF therapies are reported to be only partially effective 23,24 . Likewise, an antisense oligonucleotide against the insulin receptor substrate-1, aganirsen, was recently reported to reduce corneal neovascularisation only partially (by 26%) in patients with keratitis in a phase III study 19 . In a previous study we reported that topical administration of anti-Vegf led to a 14% reduction in neovessels compared to a 90% reduction with dexamethasone treatment in a rat corneal model of angiogenesis 25 .
In summary, there is need for new targeted therapies that mimic the potent effects of corticosteroids, but without the side effects attributed to their broad activity. A better understanding of specific factors suppressed by steroids but not by VEGF-A blockade is of potential importance for the development of more effective therapies. Here, we provide a detailed comparative dataset describing whole-genome differences in the activity of the corticosteroid dexamethasone, versus anti-VEGF treatment in a rat model of inflammatory corneal angiogenesis. GeneChip Rat 230 2.0 microarrays were used to monitor the global gene expression changes, and all microarray files including associated controls are described here, along with detailed information on the conditions of their generation and instructions for their re-use.
The dataset described here (Data Citation 1) can be a valuable resource for investigating genes and pathways to identify novel factors to target in order to improve the management of inflammationassociated neovascularisation in the cornea. The dataset described here has been thoroughly analysed in our related manuscript 26 , where it was shown that dexamethasone treatment led to a suppression of inflammatory chemokines and other genes within known signaling pathways like PI3K-Akt and focal adhesion, while notably anti-Vegf treatment did not have this suppressive activity. Furthermore, dexamethasone treatment led to an unexpected activation of the classical complement pathway 26 .

Methods
The methods described here are an expansion of those described in our related work 26 .

Suture model of inflammatory corneal angiogenesis
The model of suture induced cornea neovascularisation was used as previously described 1,25,27 . In summary, two nylon sutures were placed temporally to induce an inflammatory response that leads to new sprouting from limbal vessels after 2-3 days 27 . As a follow up of our previous findings 25 , here eye drops were administered topically immediately after suture placement i.e., during the pre-sprouting phase, but with high inflammation present in the cornea. At 48 h, in vivo confocal microscopy (IVCM) and slit lamp data was collected and corneal tissue was harvested for RNA processing. Global gene expression changes were assayed using microarrays. An illustration of the experimental design is shown in Fig. 1.
Treatment regime. Three groups of six rats each were treated with one of three topical treatments: IgG (Cat. No. 108-C, R&D Systems) at 20 μg ml −1 , anti-Vegf (Cat. No AF 564, a neutralizing rat-specific goat polyclonal pan-VEGFA antibody, R&D Systems, Minneaplois MN, USA) at 20 μg ml −1 or dexamethasone (Opnol, Clean Chemical, Sweden AB, Borlänge, Sweden) at 1 mg ml −1 . The treatments were administered topically, 4 times daily, starting immediately following suture placement, and continuing up to 48 h ( Fig. 1). Additionally, a group of four non-sutured and non-treated corneas was harvested to serve as the control. Sample collection and RNA extraction. After 48 h of topical treatment, animals were euthanized and a standardized region of tissue between the sutures and limbus which indicated a strong inflammatory response by in vivo confocal microscopy 26 was carefully dissected out. The tissue was harvested from all groups (i.e., IgG, anti-Vegf, dexamethasone and non-sutured/non-treated control), six animals per treatment group and then immediately immersed in RNA later (Qiagen) and temporarily stored at 4°C. Total RNA was then extracted from all the six samples using Qiagen mini RNA preparation kit (Qiagen). RNA concentration was then determined using a NanoDrop spectrophotometer (Thermo Scientific), and RNA quality was determined using the Agilent bioanalyzer 2100 (Agilent Technologies). RNA integrity number (RIN) ≥ 7 was set as the cut-off for sample inclusion for downstream processing for microarray analysis. Of the six RNA samples extracted per treatment group, four RNA samples per treatment group (with highest RIN) were selected for downstream processing.
Microarray target preparation. GeneChip Rat 230 2.0 microarrays were used for global gene expression analysis. Four microarray chips were used per experimental group, including non-sutured and non-treated control corneas. Each microarray chip represented a single corneal sample (16 microarrays in total) i.e., no pooling of cornea tissue was performed. As per the protocol (Manual Target Preparation for GeneChip 3′ Expression Arrays-Affymetrix), a total of 100 ng of high quality RNA per sample was mixed with poly-A spike-in Controls, and the mixture was used for first-strand synthesis to yield single-stranded cDNA with the T7 promoter sequence at the 5′ end. Next, using a DNA polymerase and RNase H, the single-stranded cDNA was converted to double-stranded cDNA to serve as a template for in vitro transcription using T7 RNA polymerase. This resulted in the synthesis of labelled complementary RNA (cRNA). The labelled cRNA was then purified, and then

Data Records
Following data acquisition, the CEL files were normalised, and CHP files were generated using Expression Console Software (Affymetrix. Inc). The CEL and CHP files were submitted to Gene Expression Omnibus repository (Data Citation 1). The samples used in the study are summarised in Table 1.

Technical Validation
Following suture placement, randomisation was performed in assigning the rats to different treatment groups. For the purpose of statistical inference, a total of four rats was used per treatment to correspond to four microarray chips per treatment group. ANOVA was used to compare effects of the given treatments across treatment groups, with Tukey multiple comparison test used to isolate pairwise differences. Efficacy of the anti-Vegf treatment was verified by Western blot analysis prior to the beginning of the study to confirm the expected effect on the gene target 26 . Suture placement was performed by the same experienced surgeon for all treatment groups, to reduce the chances of technical variation. In addition, a similar phenotypic appearance within each group was confirmed by slit lamp biomicroscopy and in vivo confocal microscopy 26 .

RNA quality check
The quality of the RNA was checked prior to downstream analysis, using the Agilent Bioanalyser 2100 (Agilent Technologies), using the 2100 expert_Eukaryote Total RNA Nano settings. The analysis showed clear, defined 28 and 18 s rRNA peaks, an indication of high quality RNA. Moreover, low noise between the peaks, and minimal low molecular weight noise was detected. In Fig. 2a an elctrophoregram is presented illustrating the peaks for the 18 and 28 s rRNA, and in Fig. 2b the sample RIN values are given.

Quality control of microarray data
The raw CEL files were converted into expression measures, background-corrected, and data-normalized using the RMA method 28 , with the help of Expression Console Software (Affymetrix. Inc). In Fig. 2c the relative log expression Signal-RMA is given. In Fig. 2d are boxplots representing four of the Affymetrix Spike-In controls, lys-M, phe-M, thr-M and dap-M illustrating the range of signal (log2) over the experiment. In Fig. 2e is the PCA-Probe Cell Intensity Data. The PCA1 (67.0%) represents anti-Vegf and IgG treated groups in blue and red colours, respectively. PCA2 (11.7%) represents dexamethasone treated group in the orange colour, while PCA3 (4.2%) displays the control samples in green colour. Note the partial overlap of anti-Vegf and IgG groups and clear separation from the dexamethasone treated corneas.
The CHP files were used for gene expression analysis using Transcriptome Analysis Console (TAC) Software (Affymetrix, Inc.). Using the obtained linearized signal intensity values from the TAC analysis, a correlation analysis between samples within the same treatment groups was performed. The resulting scatter plots with their corresponding correlation coefficient (R 2 ) are presented in Fig. 3a-d. There was strong correlation within samples of the same treatment group. The linearized values can then be log transformed and used for functional annotation analysis. As an example, STRING 29 pathway enrichment analysis was performed using the differentially expressed genes in the dexamethasone treated group. Selected pathways from this analysis are presented in Fig. 3e, and the genes involved in one of the pathways (PI3K-Akt signalling pathway) are represented as an example in Fig. 3f.

Gene fold change by qPCR
To validate microarray 30 gene fold change (Fig. 4a), here we used a separate set of rats (3 per treatment group) for qPCR analysis. Rats were sutured and treated identically as for microarray samples. At the 48 h time point, animals were sacrificed and cornea tissue was harvested and used for RNA extraction, without pooling. Following RNA extraction, cDNA was synthesised using superscript cDNA synthesis kit (Invitrogen), and used for qPCR analysis using TaqMan primers. Genes Fgf7 and Serpinb2 were assayed as exemplary genes, with results shown in Fig. 4b.

Localisation of translated genes
Cornea samples from the respective treatment groups were harvested and fixed in paraformaldehyde for 24 h, and then stabilised in paraffin blocks. Then 5 μm cross-sections of the tissue were made. Prior to staining, the sections were deparaffinised and rehydrated. The sections were probed with primary antibody for Fgf7 and Serpinb2, and detected with fluorescently (Alexa 488) labelled secondary antibody Fig. 4c. Images were acquired using a laser scanning confocal fluorescent microscope (Zeiss LSM700).

Usage Notes
The raw CEL files can be normalised using Expression console, and gene fold change can be obtained using Transcriptome Analysis Console. Both software packages can be freely accessed from Affymetrix (Affymetrix Inc).
Transcriptome Analysis Console (Affymetrix Inc) can be used for pathway enrichment analysis, where both pathway P-value and significance can be obtained. The genes involved in a given pathway can also be mapped within a pathway of interest.
Also using other publicly available software for bioinformatics analysis such as STRING 29 , Venny 31 , DAVID 32 , PANTHER 33 , etc., and those commercially available like Ingenuity pathway analysis (IPA-Qiagen) (http://www.ingenuity.com/), the differentially expressed genes can be used to deduce pathway and biological processes of interest, and to further isolate the genes involved in them.
In combination with our recently published Data Descriptor 34 , genes important for endogenous restoration of corneal avascularity during either an angiogenic sprouting or regression phase can be compared to their behaviour observed with either dexamethasone or anti-Vegf treatment in the current dataset, to gain an understating of how these treatments modulate these factors.