A Phage Display-Based Approach to Investigate Abnormal Neovessels of the Retina

Abstract

Purpose.: The goal of this project was to identify markers of abnormal neovascularization of the retinal vasculature, which is quintessential of pathologic angiogenesis that occurs in blinding diseases such as proliferative diabetic retinopathy.

Methods.: Abnormal retinal neovascularization was induced in rat pups by subjecting them to the 50/10 oxygen-induced retinopathy (OIR) protocol, which involves fluctuating levels of ambient oxygen. A peptide library (that was displayed on phage) was positively and negatively screened over the surface of retinas isolated from experimental and control rats, respectively. Binding of phage to retinal vessels was evaluated by confocal microscopy of retinal flat mounts decorated with fluorescently labeled phage. The topography of the inner limiting membrane was studied by scanning electron microscopy.

Results.: Screening a library of peptides displayed on phage over the surface of OIR retinas resulted in isolation of a particular phage (SH phage) that distinguished between abnormal neovessels and the normal vasculature. As expected, the recognition of abnormal neovessels relied on the unique peptide insert of SH phage. Abnormal neovessels consisted of at least three cell types that were present in the following order of descending abundance: endothelial > pericytes > macrophage/microglia. SH phage recognized both endothelial cells and macrophage/microglia. Finally, SH phage decorated abnormal neovessels at an early stage of their genesis.

Conclusions.: Abnormal development of neovessels is associated with early expression of distinct epitopes on the surface of cells within the pathologic vasculature. Screening phage display libraries is one approach to detecting such changes, and the resulting phage are potential imaging tools and/or drug delivery vehicles.

Introduction

Proliferative diabetic retinopathy (PDR) is a sight-threatening complication of diabetes that arises from a multitude of factors including hypoxia and changes in the level of vitreal angiomodulators. ¹ While the current standard of care is highly effective, it causes permanent damage to the neural retina. ² Further elucidation of PDR pathogenesis, especially at the molecular level, is likely to provide the insight necessary to develop effective treatments of PDR that result in fewer side effects. ³  

Owing to the paucity of rodent models of PDR, oxygen-induced retinopathy (OIR) models are typically used as a surrogate for investigating retinal angiogenesis. ^4–7 In the 50/10 rat model of OIR, newborn (P0) rats are subjected to a period of fluctuation oxygen concentration (50% and 10%), which interferes with normal development of the retinal vasculature and induces abnormal neovascularization. ⁸ The nature and timing of this angiogenic perturbation is highly predictable and constitutes one of the strengths of this model of hypoxia-induced abnormal neovascularization. 

Screening of peptide libraries displayed on the surface of an externally oriented phage protein has been successfully used in a variety of settings, including identification of peptide epitopes involved in protein–protein interactions, ⁹ probing molecular mechanisms of disease, ^10–12 and development of target-specific diagnostics or treatments ^13–19 (i.e., anti-angiogenic therapy). ^13,20 The general strategy is to devise a simple assay to screen the library for peptides with the desired characteristics, validate the behavior of the phage that contains the peptide of interest, and then design second-generation, phage-free reagents that are centered around the peptide of interest. 

Our goal was to identify markers of abnormal neovascularization of the retinal vasculature, which is quintessential of pathologic angiogenesis that occurs in blinding diseases such as PDR. While markers of angiogenic endothelium exist, they do not distinguish pathologic neovascularization from normal angiogenesis. We screened a peptide library displayed on phage over the surface of OIR retinas. We recovered phage that distinguished abnormal neovessels from the normal vasculature. It recognized a subset of cells in abnormal neovessels, namely, endothelial cells and macrophage/microglia. Potential applications of this phage include imaging and drug delivery. 

Materials and Methods

OIR

Newborn Sprague-Dawley rats were exposed to 14 successive 12-hour periods of 10% and 50% oxygen, followed by 6 days in room air (Fig. 1A). Control animals (RA) breathed room air for this entire period. Animals were euthanized either on postnatal day P20 or P16; eyes were enucleated and fixed for 45 minutes with 4% paraformaldehyde (PFA); retinas were extracted, fixed for 15 minutes with 4% PFA, and stored in PBS at 4°C until use. All animals were treated in accordance with the ARVO statement for the Use of Animals in Ophthalmic and Vision Research. 

Figure 1. 

 

The OIR protocol induced abnormal neovascularization at the surface of the retina. (A) The diagram illustrates the 50/10 OIR protocol; 21% O₂ is room air. A stack of confocal images of a representative flat-mounted retina isolated from either OIR (subjected to the OIR protocol) or RA (breathed room air) P20 rat pup, which was stained with IB4; the scale bar is 150 μm; arrows point to typical tufts. (B) Representative SEMs of the ILM (vitreal surface) of either RA or OIR retinas. Panels OIR-1 and OIR-2 are enlargements of the boxed regions of panel OIR. The scale bar in panel RA and OIR is 100 μm; 10 μm in panels OIR-1 and OIR-2. (C) The graph shows the expected downward trend in heterogeneity of the phage pool with increasing rounds of screening. The heterogeneity index reflects the number of phage-containing dissimilar peptide inserts. The inset shows the sequence of peptide inserts from individual phage that were randomly chosen and sequenced from the pool that was recovered at each round of screening. There was an increasing incidence of the sequence corresponding to the SH phage (highlighted in black) as a function of increasing rounds of selection.

Library Screening

The phage library (Ph.D.-C7C, New England BioLabs, Beverly, MA) was screened over the surface of flat-mounted retinas as described below. Retinas were obtained as described above, flat-mounted and glued (using liquid bandage, with inner limiting membrane [ILM] facing up) onto a single chamber within a four-chamber slide, and rinsed three times with PBS. In the first round of positive selection, the OIR retina was incubated for 20 minutes at room temperature with 2 × 10¹¹ plaque forming unit (pfu) of the library in a total volume of 100 μL PBS + 0.5% BSA, pH 7.4. Unbound phage was removed with six 5-minute 500-μL rinses of PBS + 0.5% BSA/0.05% Tween 20. Weakly bound phage was removed with a single, 8-minute exposure to 100 μL of 50 mM glycine-HCl, pH 2.2. The phage that remained after this step was eluted by the addition of fresh 100 μL of 50 mM glycine-HCl, pH 2.2, for 2 minutes. The eluate was neutralized with 200 μL of 100 mM Tris-HCl, pH 7.4. The resulting pool of phage was negatively selected (i.e., the unbound was retained) by exposing it to an RA retina for 20 minutes at room temperature in PBS + 0.5% BSA. This negative selection step was successively repeated with three additional RA retinas. The resulting pool of phage was amplified, titrated, and 10¹¹ pfu was subjected to four additional rounds of screening over the surface of OIR retinas. A subset of phage from each round of screening was chosen randomly, lysed, and subjected to DNA sequencing. 

DNA Sequencing and Analysis

Phage was used as a template for PCR, which was performed using Deep Vent polymerase (New England BioLabs, Ipswich, MA). In the presence of both the reverse (5′-ACACTGAGTTTCGCTACCA) and forward primers (5′-TCGGCGCAACTATCGGTATC), the resulting 360 base pair (bp) product encompassed the unique phage sequence. The resulting data were analyzed with Lasegene 7.2 software (DNASTAR, Madison, WI) and aligned using the ClustalV algorithm. A “family” consisted of at least two members, which were less than 200 (×100) nucleotide substitutions different. The number of families indicated the degree of heterogeneity of a phage pool. 

Peptide Synthesis

Peptides were synthesized by 21st Century (Marlboro, MA). The sequence of the SH and control peptides were ACSTEALRHCGGGS and ACAAAKAAACGGGS, respectively. Both peptides were an acetate salt and contained the disulfide bond between cysteines to mimic the structure displayed on phage. 

Phage Labeling

Individual phage were amplified, concentrated by polyethylene glycol (PEG)/NaCl precipitation, and resuspended in PBS. The number of phage particles was determined spectrophotometrically, and the concentration was calculated using the equation: [virions/mL] = (A₂₆₉ − A₃₂₀) × 6 × 10⁶/number of viral DNA bp. ²¹ Approximately 2 × 10¹² virions were incubated with constant rotation for 1 hour at room temperature with 50 μg DyLight-650 Dye (ThermoScientific, Rockford, IL) in 100 μL of 50 mM borate buffer, pH 8.5. The volume was adjusted to 1 mL, unbound dye was removed by two rounds of phage precipitation with PEG/NaCl, and labeled phage was reconstituted in 50 μL PBS. The concentration and extent of labeling was determined spectrophotometrically using the equation [mol dye/mol phage] = (A₆₅₅ /e_fluor × 1000 × 6.022 × 10²³); e_fluor for DL-650 = 250,000. 

Staining Retina with Phage

Flat-mounted retinas were immobilized on four-chamber slides as described above and incubated for 1 hour with 1.7 × 10¹² virions of DL-650–labeled SH or control phage in 130 μL PBS + 1% BSA. The unbound phage was removed using ten 5-minute rinses of PBS + 0.1% Tween 20. The retinas were fixed with 4% PFA for 15 minutes at 4°C, washed three times with PBS, permeabilized with PBS + 1% Triton X-100 for 1 hour at room temperature, and incubated for 1 hour at room temperature with blocking solution (PBS + 0.5% Triton X-100, 3% nonfat dry milk, 0.1% BSA). The retinal vessels were co-stained overnight at 4°C with isolectin GS-IB4 (IB4) conjugated with Alexa Fluor 488 (Invitrogen, Carlsbad, CA) diluted in blocking solution. After washing with PBS containing 300 nM DAPI (Vector Laboratories, Burlingame, CA), slides were mounted with Vectashield mounting medium (Vector Laboratories) and examined under the confocal microscope. 

Competition of SH Phage Binding by Unlabeled Phage and Synthetic Peptides

Five times molar excess of the unlabeled SH or control phage was mixed with the labeled DL-650–labeled SH phage and incubated with a retina for 1 hour at room temperature. Synthetic peptides were used in the same way, except at a 1000 times molar excess. The remaining staining steps were exactly as described above. 

Staining of Retina to Identify Different Cell Types Present in Pathologic Neovessels

P20 OIR retinas were flat-mounted, permeabilized, and blocked using the phage staining protocol outlined above. The following antibodies/lectin, diluted in blocking solution, were used to stain pericytes, macrophage/microglia, astrocytes, and endothelial cells, respectively: polyclonal rabbit Ab anti-NG2 (Millipore, Billerica, MA), mouse monoclonal Ab anti-rat CD11b (Millipore), polyclonal rabbit Ab anti-GFAP (DACO, Carpinteria, CA) Von Willebrand factor (vWF) (DACO), and IB4 conjugated with Alexa Fluor 488. Tissue was incubated with antibody for at least 6 hours at room temperature. Following exposure to primary antibodies, retinas were washed three times for 5 minutes each with PBS + 0.5% Triton X-100. The appropriate secondary antibody, anti-rabbit-Cy3 or anti-mouse-DyLight 488 Ab (Jackson Immunoresearch, West Grove, PA), was diluted in PBS + 0.1 BSA/0.5% Triton X-100 and incubated overnight at 4°C. After they were washed three times for 5 minutes each with PBS, slides were mounted with Vectashield mounting medium containing DAPI and observed under confocal microscope. 

Confocal Microscopy

Z-stack pictures were taken throughout the retinal layers proceeding from the ILM toward the retinal pigment epithelium (RPE) with a Leica TCS SP-5 upright confocal laser scanning microscope - DM6000CS (Leica, Mannheim, Germany). The results are presented as a single z-layer or as a composite stack of layers. The movies begin at the ILM and move toward the RPE. 

Scanning Electron Microscopy (SEM)

Retinas isolated from P20 OIR and control animals were fixed in 1/2-strength Karnovsky's fixative (2% PFA; Sigma, St. Louis, MO; and 2.5% glutaraldehyde; Electron Microscopy Sciences, Hatfield, PA) for 24 to 48 hours at 4°C and then dehydrated in ethanol. After the samples were critical-point dried, they were mounted with the ILM facing up on SEM stubs, coated with chromium (150A) in the Gatan Ion Beam Coater and observed in JEOL Field Emission JSM-7401F Scanning Electron Microscope (JEOL USA Inc., Peabody, MA) at 5 kV. The surface of the ILM at the mid-periphery was photographed. 

Results

Isolation of Phage That Recognized Abnormal Neovessels

To induce abnormal retinal neovascularization, P0 Sprague-Dawley rats were subjected to the 50/10 OIR protocol,⁸ which consists of 14 successive 12-hour periods of 10% and 50% oxygen, followed by 6 days in room air (Fig. 1A). Control animals (RA) breathed room air for this entire period. Eyes were enucleated from either P16 (onset of neovascularization), or P20 (peak of neovascularization) rat pups, and whole retinas were flat mounted. Confocal-assisted visualization of retinal vessels stained with IB4 revealed the expected results²²; the vascular network in RA retinas extended to the outer edge of the retina, and all three layers (superficial, tertiary, and deep plexus) were interconnected (see Fig. 1A and Supplementary Movie 1). In contrast, the vasculature of OIR retinas displayed a plethora of abnormalities (Fig. 1A and Supplementary Movie 2); the superficial vasculature did not extend to the periphery of the retina, and vessels in tertiary and deep plexi were largely absent. Furthermore, superficial vessels lacked an arterial or venous phenotype, were irregular, engorged, and included disorganized clusters of cells (tufts; Fig. 1A OIR, arrows) that extended toward vitreous. 

The observation that abnormal vessels of the OIR retina extended into vitreous (Supplementary Movie 2) suggested that the ILM had been breached. Indeed, analysis of the ILM by SEM revealed a variety of OIR-induced aberrations. While there were regions that displayed the normal polygonal pattern characteristic for normal retina (Fig. 1B, RA), a variety of aberrant structures were observed in the mid-periphery (Fig. 1B, OIR). Taken together, the data indicate that the OIR protocol generates a variety of abnormal neovessels, including ones that are present on the vitreal surface of the retina. 

In order to isolate markers of abnormal neovessels, we screened a peptide library over the surface of the OIR retina, which exhibited these vessels. To this end, we used a phage display library in which peptides of the library are displayed on the surface of an externally oriented phage protein. ²¹ We performed a series of positive and negative panning steps over OIR and RA retinas, respectively (Supplementary Fig. S1). A randomly selected subset of the pool of phage that were selected by this screening protocol were sequenced. The resulting data were analyzed using the Lasegene 7.2 software, and the ClustalV algorithm was used to organize phage into families based on the similarity of the peptide insert. As shown in Fig. 1C, successive rounds of screening decreased the heterogeneity of the phage pool and resulted in enrichment of phage harboring a peptide insert consisting of STEALRH (Fig. 1C). We termed this phage “SH.” 

To test if SH phage recognized abnormal neovessels, we fluorescently labeled it and used it to stain OIR and RA retinas. As shown in Figures 2A and G, SH phage decorated the surface of abnormal vessels, whereas the morphologically normal vessels were less well labeled. Eighty percent of tufts (shown by arrows in Fig. 1A, OIR) were decorated by SH phage. The inability of SH phage to recognize normal vessels was reinforced by the observation that it poorly recognized RA retinas, which consisted of normal vessels (Figs. 2B, 2E). Finally, control phage, which does not have a peptide insert, typically failed to decorate the abnormal neovessels (Figs. 2C, 2F); only 11% of the tufts were recognized. 

Figure 2. 

 

SH phage recognized abnormal neovessels. Representative stacks of confocal images of flat-mounted retinas isolated from P20 OIR or RA pups were co-stained with vWF (D–F) and fluorescently-labeled SH or Control phage (A–C). (G–I) Merged pictures for vWF and phage staining. Scale bar is 75 μm. Very similar results were obtained when endothelial cells were visualized with IB4 instead of vWF (data not shown).

The fact that abnormal neovessels in OIR retinas were more accessible to phage than normal vessels within RA retinas raises the possibility that accessibility was the sole reason why SH phage was able to recognize the abnormal neovessels. However, regions of overtly normal vessels in close proximity to abnormal neovessels (Supplementary Movie 2) were typically not recognized by SH phage (Figs. 2, 3, 5). Furthermore, some of the abnormal neovessels that were recognized by SH phage were below the surface of the retina, indicating that phage could enter the OIR retina and thereby access vessels within the retina. Finally, SH phage preferred abnormal neovessels at an early stage of development (when they are physically and morphologically closest to normal vessels; see Fig. 6). While it is difficult to definitively address the extent to which accessibility contributed to the overt preference of SH phage for abnormal neovessels, we do not think that this is a major contributing factor. 

Figure 3. 

 

The unique peptide insert was a key specificity determinant of SH phage. Representative stacks of confocal images of flat-mounted retinas isolated from OIR P20 pups that were stained with fluorescently labeled SH phage in the absence of competitor (A). (B, C) The staining was done with a 5-fold excess of unlabeled control or SH phage, respectively. (D, E) The staining was done in the presence of a 1000-fold molar excess of either control or SH peptide, respectively. Scale bar is 50 μm.

Figure 4. 

 

Abnormal neovessels were composed of endothelial cells, pericytes, and macrophages. Representative stacks of confocal images of flat-mounted retinas isolated from OIR P20 pups were co-stained with IB4 and the indicated antibody (A–C). DAPI was used to indicate the nuclei. Merged images are shown in the right-hand column. Scale bar is 50 μm.

Figure 5. 

 

SH phage recognized endothelial cells and macrophage/microglia within abnormal neovessels. Representative stacks of confocal images of flat-mounted retinas isolated from OIR P20 pups that were co-stained with fluorescently labeled SH phage and the indicated antibody or lectin (A–D). DAPI was used to mark the nuclei. White and yellow arrows denote examples of endothelial cells and macrophages that were recognized by SH phage, respectively. Scale bar is 75 μm.

Figure 6. 

 

SH phage identified early pathology in P16 retina. Representative stacks of confocal images of flat-mounted retinas isolated from OIR and RA P16 pups were co-stained with IB4 and the either SH or control phage. The panels in the right-hand column are merged enlargements of boxed areas from the left-hand column. DAPI was used to indicate the nuclei. Scale bar is 50 μm for low and 25 μm for high magnifications.

We consistently observed that SH phage decorated the outside of the vessels. This may indicate that the SH-phage binding partner is a molecule expressed on the surface of the cells within the neovessels. A BLAST search of the peptide insert of the SH phage (CSTEALRHC) revealed similarity to IL-4 receptor, which suggests that the phage binding partner could be a chemokine expressed by cells present within the pathologic neovessels. This observation is reminiscent of the finding that a chemokine receptors (CCR3) is expressed on endothelial cells of choroidal neovessels from patients with age-related macular degeneration. ²³  

In conclusion, we developed a novel and feasible application of phage display. Furthermore, the SH phage distinguished between abnormal neovessels and normal vasculature and thereby constitutes a potential imaging/targeting tool. 

Recognition of Abnormal Neovessels Was Dependent on the Unique Peptide Insert

We investigated the specificity determinants of the SH phage in the following series of experiments. Competition studies revealed that at a 5-fold excess, phage lacking a peptide insert did not impede binding of SH phage, whereas SH phage did (Figs. 3A–C). These results suggested that the peptide insert of SH phage was an essential determinant of specificity, and this idea was supported by the observation that a 1000-fold excess of the synthetic SH peptide (ACSTEALRHCGGGS) ablated binding of SH phage, whereas the control peptide (ACAAAKAAACGGGS) did not (Figs. 3D, 3E). Lower concentrations of the peptide did not compete effectively (data not shown), which is likely due to a lower avidity of a peptide as compared with phage. ²¹ We proceeded to test if the synthetic, cyclic SH peptide was sufficient to recognize abnormal neovessels and found that it was not (data not shown). Longer SH peptides or a GST protein engineered to display the SH peptide on an exposed face of the protein also failed to distinguish abnormal neovessels from the normal vasculature (data not shown). We conclude that the determinants of specificity of SH phage include the unique peptide insert and other, as yet undetermined, elements. 

SH Phage Recognized a Subset of Cell Types Present in Abnormal Neovessels

To determine which cell types were recognized by SH phage, we first considered the composition of abnormal neovessels. In addition to endothelial cells (IB4 staining in Fig. 4), pericytes (NG2-positive cells) were abundant in neovessels (Fig. 4A). Like pericytes, CD11b-positive cells (macrophage/microglia) were embedded within abnormal neovessels; however, they were much less abundant than pericytes (Fig. 4B). The shape of these macrophage/microglia suggested that they were activated ²⁴ (Fig. 4B and Supplementary Fig. S2). Finally, astrocytes (GFAP-positive) surrounded abnormal neovessels, whereas they were integral to overtly normal vessels (Fig. 4C). 

Co-staining OIR retinas with SH phage and markers of various cell types revealed that SH phage recognized endothelial cells and macrophage/microglia (Fig. 5). Examples of co-localization of SH phage and endothelial cells are highlighted with white arrows in Fig. 5A, whereas yellow arrows in Fig. 5B point out co-localization of SH phage with macrophage/microglia. SH phage routinely recognized activated macrophage/microglia, not all of which were localized in abnormal neovessels (Fig. 5B). SH phage never decorated pericytes or astrocytes (data not shown). We concluded that SH phage recognizes both the endothelial cells and macrophage/microglia in abnormal neovessels. 

Unique Cell Surface Markers Appear at an Early Stage in the Development of Abnormal Neovessels

To assess whether SH phage can recognize abnormal neovessels early in their development, we analyzed retinas from OIR pups harvested at P16. Although the extent of abnormal neovascularization was modest and not apparent in most samples, vessels were readily detected by SH phage (Figs. 6A–C). Control phage failed to decorate the vessels of P16 retinas (Figs. 6D–F), and RA retinas were not stained by SH phage (Figs. 6G–I). These findings indicate that the extracellular epitopes recognized by SH phage are displayed at an early stage in the development of abnormal neovessels. 

Discussion

Herein we describe a novel application of phage display library screening, and learn that ischemia-induced development of abnormal neovessels involves early expression of unique epitopes on the surface of a subset of the participating cell types. In addition, we describe a specific phage that has the potential to image and/or deliver payloads to abnormal neovessels. 

The utility of an imaging/delivery tool could be potentially increased by learning whether it distinguishes normal from abnormal neovessels. Our experimental approaches did not generate this information. While normal neovessels were present in RA retinas harvested at P16, these vessels were deep within the retina and potentially unavailable to phage. Developing an imaging reagent that is smaller (and therefore unequivocally able to penetrate the retina) is one tactic to determine whether the tools being developed see all neovessels or only those that develop abnormally. 

There are a variety of applications for a reagent that targets neovessels. For instance, as an imaging tool, it could be used to assess the likelihood that patients with nonproliferative diabetic retinopathy will progress to the proliferative stage of the disease. Similarly, such an imaging tool would be invaluable to assess the efficacy of anti-angiogenic therapies. A reagent that targets neovessels could also be used as a vehicle to deliver cytotoxic agents with the overall intent of eliminating the neovessels. In the case of patients with proliferative diabetic retinopathy, a method to selectively and safely ablate neovessels would be an advance over the current standard of care, which permanently damages the neural retina. ³ In the examples cited above, the reagent would not need to distinguish between normal and pathologic neovessels since all neovascularization in the adult retina are pathologic. 

The observation that the surface of abnormal neovessels expresses epitopes that are absent from normal neovessels suggests a viable approach to molecularly characterize the angiogenic program. What directs the orderly progression of cells through the various phases of the angiogenic program is difficult to address without a means to identify and isolate cells committed to migration versus proliferation versus quiescence. Developing a set of cell surface-based tools to isolate cells executing distinct phases of the angiogenic program will bring us closer to a molecular appreciation of this process. 

Supplementary Materials

pdf S3 

mov S1 

mov S2 

Acknowledgments

We thank Chiara Gerhardinger, Donald Pottle, and Anton Komar for their critical input to the project and/or manuscript, and Kimberly Kelly for advice with phage display. 

Footnotes

 Supported by grants from Alcon (AK), Knights Templar Eye Foundation, and One Sight Foundation (MS).

Footnotes

 Disclosure: M. Staniszewska, Alcon (F), P; X. Gu, Alcon (E), P; C. Romano, Alcon (E), P; A. Kazlauskas, Alcon (F), P