In Vivo Imaging of Fluorescent Probes Linked to Antibodies Against Human and Rat Vascular Endothelial Growth Factor

Abstract

Purpose: Anti-VEGF therapy has improved functional outcome for many patients with neovascular AMD. A particular challenge in routine clinical application is to find the best treatment regimen as a high degree of interindividual variability of disease activity has been noted. The aim of the study was to investigate fluorescent probes linked to antibodies against VEGF for in vivo imaging in an animal model.

Methods: Bevacizumab, B20-4.1.1 and AF564 were covalently attached to the novel dye 6S-indocyanine green (ICG) maleimide. Binding and proliferation properties were assessed. In a rat model of laser-induced choroidal neovascularization, retinal uptake and topographic localization of antibody-conjugates were analyzed. Distribution and accumulation of the probes were determined by immunohistochemistry and flow cytometry analysis.

Results: Antibody-conjugates retained target binding affinity and showed no toxicity. In vivo imaging showed a strong fluorescence immediately following an intravenous or intravitreal injection. While accumulation within the laser lesions was visualized for all three antibody conjugates, the signal strength and the duration of fluorescence varied. In addition, distinct fluorescent spots were also recognized. Patterning and in-depth analyses including histology and flow cytometry data strongly suggest that the fluorescent spots represent labeled microglial cells and/or macrophages.

Conclusions: Pharmacokinetics of fluorescent-labeled bevacizumab, B20-4.1.1 and AF564 can be investigated in vivo. In this model, interpretation of long-term in vivo observations is difficult because of a possible rat-specific immune response and challenges to image localized binding of soluble VEGF. Further investigations in a primate model and the use of appropriate antibodies directed against the VEGF-receptor may represent alternative approaches.

Neovascular AMD is one of the leading causes of blindness in the Western world.¹ It is characterized by invasion of new and abnormal choroidal blood vessels at the posterior pole that penetrate through Bruch's membrane into retinal layers causing severe damage to neuronal tissue.² The development of anti-VEGF inhibitors (e.g., bevacizumab, ranibizumab, and aflibercept) and their repeated intravitreal injection represents a breakthrough in the management of patients with neovascular AMD.^3–6 

A particular challenge with VEGF-inhibitor treatment in neovascular AMD is the high degree of both inter- and intraindividual variability of the frequency of repeated injections over a long time period. Therefore, fixed treatment regimens are not recommended, and current practice is to perform retreatments when new disease activity is noted morphologically on optical coherence tomography (OCT) in vivo imaging, funduscopy, or functional decline.⁴ These signs are seen as a result of prior VEGF-hyperexpression in the outer retina. In addition, some patients do not have any or only little morphologic and or functional benefit from anti-VEGF treatment in real-life application (non- or poor responders).⁷ A major disadvantage of the anti-VEGF therapy is that it is currently not possible to visualize pharmacodynamics properties of the drug in vivo. Therefore, it is not possible to assess neither the VEGF-concentration at the site of the neovascular complex nor to conclude if the agent has actually reached its target. Furthermore, it is difficult to monitor drug dosage efficacy and resistance. Well-established diagnostic tools in ophthalmology including spectral-domain OCT and confocal scanning laser ophthalmoscopy (cSLO) allow for visualization of disease activity and assessment of treatment response in neovascular AMD has become clinical routine. However, all clinical markers are rather nonspecific and primarily address anatomic/morphologic features, while they are not able to detect preceding alterations on a molecular level. Overall, it appears to be conceivable that VEGF hyperexpression occur before morphologic evidence of damage to neuronal tissue as identified by established retinal imaging modalities. 

Molecular in vivo imaging using cSLO would be a promising and innovative approach to identify sites of VEGF hyperexpression at a very early stage, and thus allow not only a much earlier diagnosis but also a more individualized and refined treatment. The unique optical properties of the eye suggest that molecular imaging at high resolution may be successfully applied particularly in ophthalmology.^8–10 Previously, several applications of molecular retinal in vivo imaging using the confocal scanning laser ophthalmoscope have been reported. For example, Cordeiro et al.^11,12 have studied retinal cell apoptosis with fluorescent-labeled annexin V probes while Eter et al.¹³ investigated the inflammatory response of dendritic cells, macrophages and microglial cells upon laser coagulation. 

The laser-induced animal model of choroidal neovascularization (CNV) has been extensively used in various preclinical research projects.^14,15 It has been demonstrated that VEGF is upregulated in this model with a peak level of expression between 3 and 7 days following laser application.^16–18 Several therapeutic inhibitory antibodies or antibody fragments against VEGF are available but there is an ongoing controversial discussion in the literature about targeting of the humanized monoclonal anti-VEGF antibody bevacizumab to rodent VEGF isoforms. While some studies have reported no therapeutic effects in rats or inhibition of murine VEGF,^19–21 others have demonstrated a neutralizing effect.^22–24 

The aim of the current study was to design and investigate reporter molecules targeting VEGF for an in vivo imaging application. For this purpose, fluorescent dyes were coupled to VEGF antibodies, evaluated in vitro, and then tested in vivo in the animal model of laser-induced CNV. 

Materials and Methods

Fluorescent Marker

Three different antibodies against VEGF were evaluated for fluorescence imaging: bevacizumab (humanized monoclonal antibody against human VEGF₁₆₅; Roche Pharma AG, Mannheim, Germany), B20-4.1.1 (monoclonal antibody against human and rat VEGF, a gift from Genentech, San Francisco, CA, USA), and AF564 antibody (polyclonal antibody against rat VEGF₁₆₄, AF564; R&D Systems, Wiesbaden-Nordenstadt, Germany). The antibodies were covalently attached to 6S-indocyanine green maleimide (mivenion GmbH, Berlin, Germany), a novel indocyanine green (ICG) dye derivative for protein labeling (Figs. 1A–C).^25,26 The labeling ratio was approximately 1:1 (antibody to dye) with a fluorescence quantum yield of 2%. Additionally, the following probes were tested: (1) human IgG (I4506; Sigma Aldrich Biochemie GmbH, Hamburg, Germany), (2) rat IgG (I4131, Sigma Aldrich Biochemie GmbH), both labeled with 6S-ICG maleimide (mivenion GmbH), (3) 6S-ICG-COOH (mivenion GmbH) as free dye label control (unconjugated dye-linker only), (4) conventional clinically approved ICG (Pulsion, Feldkirchen, Germany), and (5) PBS (sterile DPBS, P04-36500, Pan-Biotech GmbH, Aidenbach, Germany; Table). 

Figure 1

 

Chemical structures of ICG (A), free dye label (B), and antibody-6S-ICG conjugate with maleimide/2-iminothiolane connection (C). BIAcore assay was realized with labeled bevacizumab, AF564 and B20-4.1.1 (K_D = dissociation constant) (D). To follow cell proliferation by the MTT assay HUVEC cells and labeled as well as unlabeled monoclonal antibody (mab = bevacizumab) were cultured with increasing VEGF concentration ([E]: 0–50 ng), or fixed VEGF concentration ([F]: 20 ng/mL). In competition experiments antibodies were added at a concentration of 10 μg/ml.

Table

 

Characteristics of Fluorescent Samples.

 

For postmortem fluorescence microscopy, probes were labeled with the VIS analog indocarbocyanin (ICC, Cy3-type) instead of 6S-ICG-COOH due to technical difficulties of ex vivo fluorescence imaging in the near-infrared range.²⁷ 

Antibodies were labeled under sterile conditions. Briefly, antibodies were reacted with a 5-fold molar excess of 6S-ICG maleimide in the presence of a 10-fold excess of the thiolation reagent 2-iminothiolane. After incubation at 25°C for 24 hours, the reaction solution was purified by size-exclusion chromatography (Sephadex G-50 superfine, GE Life Sciences, Freiburg, Germany) using PBS as eluent. Successful removal of free dye excess was confirmed using thin layer chromatography (TLC) with Merck TLC RP C-18 plates (Merck KGaA, Darmstadt, Germany). After concentration to approximately 1 mg/mL using spin dialysis tubes (Centriprep YM-30, 30 kDa MWCO; Merck-Millipore, Merck KGaA, Darmstadt, Germany), solutions were lyophilized and subsequently stored at 2°C to 6°C. When needed, all labeled antibodies were reconstituted with distilled water to a concentration of 1 mg/mL for intravenous (volume: 0.2 mL) and of 5 μg/μL for intravitreal (volume: 1 μL) injection, respectively. 

Binding Properties

Binding affinities of recombinant rat (rr) VEGF₁₆₄ and recombinant human (rh) VEGF₁₆₅ to labeled bevacizumab, AF564 and B20-4.1.1 antibodies were performed using surface plasmon resonance (SPR) on a BIAcore X100 instrument (GE Healthcare, Freiburg, Germany). Briefly, VEGF of human or rat origin (R&D Systems) was immobilized on CM5 chips (GE Healthcare) to a density level of approximately 3000 resonance units (RU), following standard procedures. As analytes antibodies were diluted in running buffer (HBS-EP; GE Healthcare,) and injected at 30 μL/min at 25°C. Kinetic titration series (single cycle kinetics) up to 1000 nM in 10-fold dilution steps (1000, 100, 10, 1, 0.1 nM) followed by 600-second dissociation were performed. Response difference between ligand flow cell and mock treated reference flow cell was recorded, sensorgrams were blank subtracted and analyzed by plotting analyte concentration against response level at end of eject. 

Human umbilical vein endothelial cells (HUVEC; PromoCell GmbH, Heidelberg, Germany) were cultured in endothelial cell growth medium with supplement-mix (PromoCell GmbH) as previously described.¹⁵ This study complies with the principles outlined in the Declaration of Helsinki for medical research involving human subjects (2008). Cells were seeded in medium at 2 × 10⁴ cells/mL, cultured at 37°C with 5% CO₂ up to 80% of confluence, and then split 1:4. The human hematopoietic cell line U937 was routinely propagated in Roswell Park Memorial Institute (RPMI) medium, with 10% fetal calf serum (FCS) and penicillin/streptomycin (PAN-Biotech GmbH, Aidenbach, Germany). Cells were seeded in medium at 1 × 10⁵ cells/mL, cultured at 37°C with 5% CO₂, and split 1:30 two times a week. To follow cell proliferation by MTT [3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide] assay, 1 × 10⁴ HUVEC cells/well were cultured for 48 hours either with increasing VEGF concentrations (0–50 ng; PeproTech GmbH, Hamburg, Germany) or a fixed VEGF concentration (20 ng/mL). In control experiments, antibodies (bevacizumab and IgG) were added at a concentration of 10 μg/mL and medium was used as the untreated control (100% value). To test cytotoxic effects of antibody conjugates, 1 × 10⁴ U937 cells/well were cultured with increasing concentrations of antibody conjugates. After 2 days of culture, 10 μL MTT (5 mg/mL in PBS; Calbiochem, San Diego, CA, USA) were added to each well and the plates were incubated for 4 hours. The resulting formazan product was dissolved with acid isopropanol and the absorbance at a wavelength of 570 nm was read on a microplate spectrophotometer (Anthos Reader HT 2; anthos Microsystems GmbH, Krefeld Germany). 

Animals

All investigations were performed on male adult Dark Agouti rats, each weighting 200 to 250 g, and CX₃CR1^GFP/+ mice (provided by Christian Kurts, University of Bonn, Bonn, Germany), weighting approximately 20 g. Rats were anesthetized for all procedures by intraperitoneal injection of ketamine (60 mg/kg body weight) and medetomidine hydrochloride (0.5 mg/kg body weight).¹⁵ Anesthesia of rats was reversed by intraperitoneal injection of a 20% atipamezol (1 mL/kg) solution. Mice were anesthetized by intraperitoneal injection of a mixture of ketamine (50 mg/kg body weight) and xylazin (5 mg/kg body weight). All animal procedures were approved by local authorities (Landesamt für Natur, Umwelt und Verbraucherschutz Nordrhein-Westfalen, Germany) and complied with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. 

Animals received a topical administration of 0.5% tropicamide (Mydriaticum Stulln, Pharma Stulln, Stulln, Germany) eye drops for pupillary dilation of both eyes before in vivo imaging and laser treatment. All probes were injected intravenously or intravitreally 7 days (D7) following laser treatment (Table). For the latter, a Hamilton needle system with a 32-G needle (Hamilton Company, Bonaduz, Switzerland) was used. In every case, only a single injection was done per animal (0.2 mL) or eye (1 μL), respectively (Table). Retinal uptake of the probes was recorded for up to 200 days. 

Laser Photocoagulation

Retinal laser lesions in age-matched rats and mice were placed around the optic nerve head with an argon laser (Novus2000; Coherent, Dieburg, Germany) to induce CNV formation at day 0 (D0) as previously described.^15,26 The following settings were used: 514-nm excitation, 0.1-second pulse duration, 150-mW laser power, and 100-μm spot size for rats as well as 514-nm excitation, 0.1-second pulse duration, 200-mW laser power, and 50-μm spot size for laser treatment in mice. 

In Vivo Imaging

In vivo near-infrared reflectance and fluorescence imaging of the ocular fundus of the rodent eye was performed using confocal scanning laser ophthalmoscopy (HRA2; Heidelberg Engineering, Heidelberg, Germany) in 51 animals (Table) as described previously.¹⁵ The laser power was set at 100% and the detector sensitivity at 94%, respectively. Fluorescence images with different focus settings (0 till +20 diopter) were taken before dye injection (baseline) and at repeatedly predefined time points following dye injection (24 hours, 7, 14, 21, 28, 35, 42, 49, 56, 70, 84, 98, 119, 147, 175, and 203 days). The confocal approach allows for the acquisition of sectional scans through the rodent retina and investigates the depth location of fluorescent signals.¹² Eye drops (Oculotect fluid 50 mg/mL; Novartis Pharma, Nürnberg, Germany) were placed on the eyes every 1 to 2 minutes. 

In Vivo Image Analysis

For assessment of the pharmacokinetics of different samples, fluorescence at respective time points was qualitatively compared following intravenous and intravitreal application. For the more detailed quantitative analysis of the fluorescence intensity following intravitreal application within the laser lesions was assessed as pixel intensity from predefined circles (circle size: 26000 pixels) within all laser lesions using Photoshop CS5 software (Adobe Systems, Inc., San Jose, CA, USA) subtracting the averaged pixel values from the mean zero gray value.^15,28,29 The overall fluorescence intensity (background signal) was assessed accordingly from randomly selected well-illuminated choroidal area outside the laser lesions, which did not contain major blood vessels, followed by subtracting the averaged pixel value from the mean gray value. Furthermore, the accumulation index was calculated by the pixel intensity within the laser lesions and background ratio for quantitative assessment of the signal strength.²⁸ The specific binding activity of the three anti-VEGF antibodies was evaluated in terms of the signal-to-noise ratio of the individual signal strength to the species-specific background IgG antibody. 

Finally, the number of fluorescent spots was determined applying ImageJ software (http://imagej.nih.gov/ij/; provided in the public domain by the National Institutes of Health, Bethesda, MD, USA). Images were converted into 8-bit images and adjusted to a lower and upper pixel threshold. Using the analyze particle tool, the number of fluorescent spots was counted automatically. 

Ex Vivo Measures and Histology

For immunohistochemical analyses, both eyes of eight animals received an intravitreal injection of bevacizumab-6S-ICC after 7 days of initial laser treatment (D0). Four animals were used to prepare frozen sections and the eyes of four animals were required for flatmount preparations. Following in vivo imaging at day 14, all animals were euthanized applying 100% carbonic acid gas. Eyes were immediately enucleated and processed for retinal and RPE/choroid/sclera flatmounts or frozen sections and examined by fluorescence microscopy as previously described.¹⁵ In brief, eyes were fixed in 4% paraformaldehyde followed by cryoprotection by soaking in 30% sucrose at 4°C overnight. Eyes were frozen in methylbutane, stored at −80°C for 24 hours and embedded in OCT medium (Tissue Tek; Sakura Finetek Europe B.V., Alphen aan den Rijn, the Netherlands) on dry ice. Frozen sections of 10 μm were mounted on a glass slide, fixed with 100% methanol and labeled with DAPI (DAPI BioChemica, 4′,6-Diamidino-2-phenylindole dihydrochloride; AppliChem GmbH, Darmstadt, Germany), CD68 (mouse anti CD68 IgG1 Clone ED1; Acris Antibodies GmbH, Herford, Germany), and Iba1 (polyclonal anti Iba1, rabbit; Wako Chemicals GmbH, Neuss, Germany). 

For flatmount preparation, following enucleation and fixation, the anterior part of the eye was removed and retina as well as choroid was separated. Flatmount sections were incubated with a 1:200 dilution of CD68 (mouse anti CD68 IgG1 Clone ED1; Acris Antibodies GmbH) and Iba1 (polyclonal anti Iba1, rabbit; Wako Chemicals GmbH) antibody at 4°C overnight. Thereafter, segments were incubated with secondary antibodies (Alexa Fluor488 goat anti mouse IgG or Alexa Fluor488 goat anti-rabbit IgG; both Invitrogen Molecular Probes, Eugene, Oregon, USA). Fluorescent images were made with a Zeiss Axiovert (200 M) microscope equipped with an ApoTome and an AxioCam MRc set-up (Carl Zeiss Microscopy GmbH, Göttingen, Germany) using a ×10 and ×20 objective and a FITC (excitation filter wavelength, 480/40 nm; emission filter wavelength, 535/50 nm) and a Cy3 filter set (appropriate for the ICC dye; excitation filter wavelength, 564 nm; emission filter wavelength, 600/40 nm). 

CX₃CR1^GFP/+ Mice

In order to further investigate distinct hyperfluorescent spots, five CX₃CR1^GFP/+ mice were imaged following intraperitoneal injection of the free dye label (6S-ICG-COOH). Using the cSLO, fluorescence of mononuclear phagocytes (endogenous biomarker) was imaged in vivo by blue laser light excitation (488 nm), while fluorescence of 6S-ICG-COOH as free dye label control was detected by near-infrared light excitation (780 nm). 

Flow cytometry analysis was performed in laser-treated and untreated CX₃CR1^GFP/+ mice following intravitreal injection of bevacizumab-6S-ICG, ICG or DPBS at day 7. At day 14, animals were euthanized using 100% carbonic acid gas. Eyes were immediately enucleated and dissected. Segments were conveyed into digestion medium, which contained a mixture of 0.5 mg/mL collagenase and 100 μg/mL DNase I in RPMI⁺ medium (Invitrogen, Karlsruhe, Germany) as described recently.³⁰ Retinal and choroidal segments were incubated with the digestion medium at 37°C for 20 and 40 minutes on a shaker. Afterwards, cells were washed and filtered through a 100-μm nylon mesh. Cells were centrifuged and washed several times with buffer solution. Fc receptors were blocked with medium containing human IgG (monoclonal antibody; BD Biosciences, San Jose, CA, USA) prior to surface staining for 20 minutes at 4°C. Cells were stained in medium with anti-Gr1-eFluor450 (1:400; clone RB6-8C5), anti-CD11c-BV605 (1:400; clone N418), anti-CD45-PE (1:400; clone 30-F11), and anti-CD11b-PerCP-Cy5.5 (1:1000; clone M1/70; all from BioLegend Inc., San Diego, USA). Afterward, cells were washed with buffer solution and centrifuged. Finally, buffer solution and a predefined number of PE- and APC-labeled microbeads (BD Biosciences) were added for cell number calculation. Retinal and choroidal cells were analyzed on a LSRFortessa cytometer (BD Biosciences) and data processing performed with the FlowJo software (Tristar, Ashland, TN, USA). 

The classification of immune cells by flow cytometry results was performed in accordance to published criteria and validated cell surface markers.^13,30–33 Detected immune cells were analyzed by identifying the established phagocyte cell surface marker CD11b and CD11c. Further subclassification of ICG⁺GFP⁺ phagocytes was performed based on the expression of cell surface markers Gr1 and CD45. Gr1 is a cell surface marker for inflammatory macrophages. Microglial cells were classified as CD45^low and macrophages as CD45^high as previously described.^30,34 

Statistical Analysis

Statistical analyses were performed by using the SPSS software package Statistics 22 (SPSS, Inc., Chicago, IL, USA). A statistical analysis of pixel intensities and analysis of fluorescent spots were performed with log transformed values and compared with PBS control group over the complete time period of 203 days by a linear mixed-model. Flow cytometry results were analyzed by using Student's t-test. All results were expressed as mean values ± SD and P values less than 0.05 were considered as statistically significant. Significance levels are indicated using * for P less than or equal to 0.05; ** for P less than or equal to 0.01, and *** for P less than or equal to 0.001. 

Results

In Vitro Experiments

Results of SPR affinity analysis demonstrated different nanomolar affinities of fluorescent-labeled bevacizumab to rr VEGF₁₆₄ (K_D = 540 nM) and rh VEGF₁₆₅ (K_D = 116 nM), with a comparable lower affinity to the rat protein (Fig. 1D). Binding affinity of AF564-6S-ICG to the recombinant rat protein rr VEGF₁₆₄ resulted in an apparent K_D value of 147 nM (Fig. 1D). In this particular case, multivalent binding of polyclonal antibodies cannot be excluded. In addition, monoclonal antibody B20-4.1.1-6S-ICG was able to bind VEGF from both species with K_D values of 207 and 222 nM for the rat and human protein. 

Furthermore, in vitro studies revealed no toxic effects of all tested antibodies and antibody conjugates with respect to viability of hematopoietic U937 cells. Cell proliferation of HUVEC, as monitored by the MTT assay, was enhanced dose-dependently by VEGF and bevacizumab neutralized this effect (Fig. 1E). Antibody function was retained, neither chemical coupling of fluorescence dyes, nor formulation by lyophilization disabled VEGF targeting (Fig. 1F). 

In Vivo Imaging and Pharmacokinetics

At initial in vivo imaging (D7, 0 minutes) no remarkable autofluorescence was detectable (Fig. 2 and Supplementary Fig. S1). Following intravenous application, signals in the retinal vessels were always immediately visible for all tested conjugates and intensity only slowly declined over time. At 24 hours (D8), postintravenous probe application an accumulation of fluorescence within the laser lesions and the appearance of multiple hyperfluorescent spots were always observed (Fig. 2 and Table). Following intravitreal injection, a strong homogenous fluorescence was detectable at the injection site quadrant (Fig. 2). No fluorescence was visible within retinal blood vessels at any time, while a variable fluorescence signal was seen within the laser lesions after 24 hours (D8) and fluorescent spots appeared. Distinct spots especially accumulated around the optic disk and at the wall of major retinal veins (Fig. 2 and Supplementary Fig. S1). Some of the in vivo detectable fluorescent spots may be stationary while others are mobile (Supplementary Fig. S2). Further analysis using different focus settings revealed that fluorescent spots mainly localized within the inner retina, while a more blurred pattern was detectable upon focusing the laser lesions (Supplementary Fig. S3). Fluorescence as well as number of fluorescent spots decreased over time (Fig. 2). 

Figure 2

 

In vivo fluorescence imaging over time in the rat eye of dye labeled anti-VEGF antibodies for both intravenous and intravitreal applications, respectively. Time points are indicated at the bottom of respective images (D = day following laser treatment).

Evaluating differences between the three different intravitreal injected-labeled antibody probes (Table), bevacizumab-6S-ICG consistently showed the highest number and most sustained intensity of fluorescent spots (P < 0.0001; Fig. 3). The lowest count and intensity of hyperfluorescent spots was found for the rat specific AF564-6S-ICG probe (P < 0.0001), while B20-4.1.1-6S-ICG (P < 0.0001) ranked in between as illustrated in Figure 3 in detail (all P values were compared with the unlabeled PBS control over the total time period). Similar to bevacizumab-6S-ICG, the control experiment with unspecific human IgG-6S-ICG (P < 0.0001) and free dye label (P < 0.0001) also resulted in a high number of sustained fluorescent spots following intravitreal injection. Lower numbers were observed following an application of rat IgG-6S-ICG (P < 0.0001). Only very few spots of different shapes appeared following an intravitreal injection of ICG. They were only visible 24 hours following intravitreal application (P = 0.636). Finally, intravitreal application of PBS control produced a barely detectable fluorescence within the laser lesions. 

Figure 3

 

Number of fluorescent spots following intravitreal application (mean and SD in a linear mixed-model) of: bevacizumab-6S-ICG (n = 9), B20-4.1.1-6S-ICG (n = 6), AF564-6S-ICG (n = 8), human IgG-6S-ICG (n = 4), rat IgG-6S-ICG (n = 4), ICG (n = 8), free dye label (n = 5), and PBS (n = 5) at given time points before and following injection at day 7 following laser treatment. Over time, number of fluorescent spots decreased. Following the intravitreal injection of free dye label, the fluorescence intensity was very bright and no reasonable fluorescence measurement was possible in the first 24 hours following the application. Hence, the analysis of the number of fluorescent spots is missing for free dye label at D8.

Regarding fluorescence of VEGF-targeting antibody conjugates at laser lesion sites (Fig. 4A) the observed fluorescence in the background fluorescence and signal-to-noise ratio was high. Overall, bevacizumab-6S-ICG generated the highest and longest lasting signal (P < 0.0001), which was 28.2- and 11.8-fold compared with the PBS control and calculated at 1 and 7 days following the intravitreal injection. Signal strength of B20-4.1.1-6S-ICG (10.2- and 3.3-fold and P = 0.023) was in between and fluorescence of AF564-6S-ICG showed the lowest fluorescence within the laser lesions (8- and 2.9-fold and P = 0.225). Also the control free dye label (51.1- and 23-fold) and nonspecific antibody control human IgG-6S-ICG (37.5- and 16.8-fold) had an even stronger fluorescence and even longer lasting fluorescence signal (both: P < 0.0001) than bevacizumab-6S-ICG (Fig. 4A). The fluorescence intensity of unspecific rat IgG-6S-ICG (8.6- and 3.7-fold and P = 0.053) within laser lesions was comparable to signals produced by the specific AF564-6S-ICG. Indocyanine green fluorescence (28.8- and 4.9-fold and P ≤ 0.002) peaked at day 8 and then rapidly decreased following further observation. The course of background fluorescence of individual probes (Fig. 4B) was comparable but lower than the signal within the laser lesions. Assessment of the accumulation index (Fig. 4C) revealed a stronger accumulation for the applied human-specific antibodies (bevacizumab-6S-ICG and human IgG-6S-ICG) compared with the rat-specific antibodies (AF564-6S-ICG and rat IgG-6S-ICG). Accumulation following the intravitreal injection of B20-4.1.1-6S-ICG was in between (Fig. 4C). One week postintravitreal application, accumulation of conventional ICG peaked and fluorescence decreased following further observations rapidly underling the described high accumulation within the laser lesions and almost no detectable background fluorescence. Interestingly, accumulation index of free dye label only slowly increased at the beginning of the measurement and changed only moderate over time compared with the other tested probes. The differences in the signal strength for laser lesions and the background fluorescence between the different probes continued to decrease following day 8 (see Fig. 4). 

Figure 4

 

Pixel intensities within the laser lesions (A) and background level (B) following intravitreal injection of fluorescent labeled antibodies over time (mean values ± SD). Intravitreal bevacizumab-6S-ICG (n = 9), B20-4.1.1-6S-ICG (n = 6), and AF564-6S-ICG (n = 8), as well as humanIgG-6S-ICG (n = 4), ratIgG-6S-ICG (n = 4), conventional ICG (n = 8), free dye label (n = 5), and PBS as control group (n = 5) at different time points before and following injection at day 7 following laser treatment. Pixel intensity within the laser lesions and the background increased initially for all tested probes following the intravitreal injection. Accumulation index (C) was calculated as the ration of pixel intensity within the laser lesion and background signal. Over time, decreased fluorescence intensity at the site of CNV and in the background was measured. Finally, the signal strength (D) of three anti-VEGF antibodies against their species-specific background at different time intervals was evaluated.

In a further analysis, we compared the signal strength of three samples against their species-specific background at different time intervals (i.e., comparing bevacizumab-6S-ICG against human IgG-6S-ICG, B20-4.1.1-6S-ICG divided by human and rat IgG-6S-ICG and AF564-6S-ICG by rat IgG-6S-ICG). As shown in Fig. 4D, this analysis of the signal-to-noise ratio did not show a clear specific accumulation of anti-VEGF samples at the site of laser lesions or background at any time point as detected by in vivo imaging. 

In vivo imaging in the genetically modified CX₃CR1^GFP/+ mice revealed an accumulation of green fluorescent mononuclear phagocytes within the laser lesions. Interestingly, the additional intraperitoneal injection of free dye label gave a similar pattern and resulted in a high degree of colocalization (Supplementary Fig. S4). 

Histology and Immunohistochemistry

Immunohistochemistry on choroidal and retinal flatmounts was performed (Fig. 5). Choroidal flatmounts showed an accumulation of bevacizumab-6S-ICC within the laser lesions and only a small number of freely distributed fluorescent spots. Almost no fluorescence was detectable within the laser lesions of retinal flatmounts, however, distinct fluorescent spots were localized outside. Analyses with cell-specific markers demonstrate a colocalization with macrophages and microglial cells. Hence, this might indicate cellular uptake of the antibody-conjugate by phagocytosis. The insert of the retinal flatmount picture (Fig. 5, enlarged picture with arrow) indicated an amoeboid microglial cell, containing four, red-stained vesicular structures. 

Figure 5

 

Choroidal and retinal flatmount staining prepared 7 days following intravitreal injection of bevacizumab-6S-ICC (Cy3-type ICC = indocarbocyanine, red). Staining with CD68 (green, first two rows) showed a colocalization of antibody-conjugates and macrophages, especially within the laser lesions. Staining with Iba1 (green, last two rows) showed also a colocalization with microglial cells. Arrow shows an amoeboid Iba1-positive cell with four small bevacizumab-6S-ICC spots inside (inserted detailed figure).

Immunohistology of frozen cross-sections also showed an accumulation of bevacizumab-6S-ICC within the laser lesions (Fig. 6A) and the choroidal layer; and again we found a colocalization with the microglia marker Iba1 (Fig. 6B), but even more so with the macrophage marker CD68 (Fig. 6C). 

Figure 6

 

Detection of bevacizumab-6S-ICC (red) in frozen cross sections 7 days following intravitreal injection. In the DAPI stained cross section (A), fluorescent bevacizumab-6S-ICC spots were visible, especially in the area of the laser lesion. Staining for Iba1 shows a moderate colocalization of antibody-conjugates and microglial cells ([B], yellow). A distinct positive double staining with bevacizumab-6S-ICC and CD68 positive macrophages is obvious at the site of CNV (C).

Classification of Immune Cells by Flow Cytometry

Following laser treatment, CX₃CR1^GFP/+ mice showed an increasing number of phagocytes, especially inflammatory macrophages and activated microglial cells. Analysis by flow cytometry identified cell populations positive for GFP and ICG (Fig. 7). These GFP and bevacizumab-6S-ICG–positive phagocytes were identified in retinal and choroidal cell subsets and subclassified for abundance of CD45. Retinal GFP⁺ phagocytes were identified mainly (81%) as microglial cells (CD45^lowGr1^lowCX₃CR1^high) and 54% were identified as bevacizumab-6S-ICG positive (Fig. 7). Macrophages (CD45^highGr1^lowCX₃CR1^high) were almost absent in the retina (8%). In the choroid, both macrophages (32%) and microglial cells (28%) were identified as ICG⁺ and GFP⁺ (Fig. 7). 

Figure 7

 

For flow cytometry analysis CX₃CR1^GFP/+ mice were used. Animals were injected intravitreally with bevacizumab-6S-ICG (n = 7) 7 days following laser treatment. Isolated cells were characterized by flow cytometry (mean values ± SD) and subclassified for bevacizumab-6S-ICG–positive cells and the marker CD45. CD45^low cells were identified as microglial cells and CD45^high cells as macrophages.

Discussion

This study was initiated to develop a more sensitive approach to determine VEGF activity and treatment response of VEGF inhibitors for the management of patients with neovascular AMD. The first steps of this endeavor were achieved in the current study, namely the successful fluorescent-labeled binding of three different mono- and polyclonal antibodies against VEGF (bevacizumab, B20-4.1.1 and AF564), the demonstration of preserved binding affinity by in vitro analysis and the ability to visualize the uptake and topographic distribution by in vivo retinal imaging at multiple time points. Furthermore, the qualitative and quantitative analyses of in vivo imaging as well as the postmortem analysis in the rat model of laser-induced CNV revealed a strong fluorescence signal at the sites of laser lesions. The stronger fluorescence signal at laser lesion sites following intravitreal as compared with intravenous injection for each individual probe would indicate that local as compared with systemic application would result in improved detection of fluorescent dyes in the eye as systemic absorption and/or degradation of fluorescent labeled antibodies before reaching the eye does not happen (to the same extent) via the intravitreal route. However, no matter of the route of administration, overall no specific binding, as compared to IgG-antibodies, was visualized for the three-labeled VEGF inhibitors by in vivo imaging at any time point. 

In addition to localized fluorescence accumulation at the sites of CNV, multiple fluorescent spots were clearly detected for all tested labeled probes. The observation of fluorescent spots both along with and without fluorescent probes and also using different excitation and emission wavelengths in different disease models has been reported before.^35–41 While the visualization of a few spots in the native rodent model appears to be a common finding that is also correlated to age, fluorescent spots in the context of endogenous and exogenous fluorescent probes has been correlated to single cells. For example, migration of microglial cells have been monitored in laser-treated CX₃CR1^GFP/GFP mice in combination with intravitreal ICG application.³⁹ In the current study, we also observed a small numbers of spots following intravitreal ICG application, visible up to 24 hours after probe application. However, with all other tested samples, the number of fluorescent spots was relatively high, indicating that the occurrence of these spots is not a result of nonspecific binding of ICG alone. As the number of these spots was particularly high using the free dye label, it is conceivable that the appearance of fluorescence spots is caused by the labelling with 6S-ICG-COOH itself. The appearance of spots seems to be independent of the laser treatment as post bevacizumab-6S-ICG injection these fluorescent spots were also observed in eyes without any prior laser treatment (data not shown). 

The higher number of spots following intravitreal as compared with intravenous application would suggest that local versus systemic probe application was correlated with the number of spots. The results of the quantitative analysis indicate that the number of spots was associated with the species-specific origin of the probes and both, the appearance and the degradation of these spots are reflected by an immune response toward nonself epitopes. It would be conceivable that a more pronounced immune response was caused by the human-specific, following by the combined human-murine and then followed by the murine-specific antibody as the highest number and longest duration of spots were observed post bevacizumab-6S-ICG treatment, followed by B20-4.1.1-6S-ICG and finally by AF564-6S-ICG application. This assumption would be underscored by similar observations made for control dyes, revealing that human IgG had a higher number and longer duration of spots as the rat IgG. Possibly the observation of longer visibility of fluorescence accumulation within laser lesions of bevacizumab-6S-ICG, followed by B20-4.1.1-6S-ICG, and then followed by AF564-6S-ICG treated eyes may also reflect an immune response, due to the differences of the species-specific origin of the antibodies.^42–46 

Several results of the current study suggest that these hyperfluorescent spots are spatially correlated to microglial cells and/or macrophages: (1) in vivo experiments in CX₃CR1^GFP/+ mice showed a colocalization of free dye label (6S-ICG-COOH) with CX₃CR1^GFP/+ cells. According to Combadière et al.,⁴⁷ CX₃CR1^GFP/+ cells are mainly microglial cells, (2) staining of retinal flatmounts and frozen sections disclosed a colocalization with fluorescent spots of both macrophages and microglial cells, suggesting an uptake of fluorescent labeled antibodies by these immune cells, (3) data of cell sorting revealed a main uptake of ICG by (retinal) microglial cells and macrophages. These observations suggest that fluorescent-labeled antibodies were mainly phagocytized by microglial cells in the retina, while phagocytosis of ICG in the choroid was mainly due to both, macrophages as well as microglial cells, (4) the appearance of hyperfluorescent spots in retinal areas at later time points, which had initially not shown any spots, may indicate migrating cells such as macrophages and microglial cells,¹³ and (5) cyanine dyes, like ICG, have a strong affinity for membrane phospholipids and uptake by neuronal and microglial cells has been previously described, also resulting in the appearance of fluorescent spots as seen by in vivo imaging.^39,48 

The visualization of fluorescence intensities within laser lesions and fluorescent spots, particularly at later time points did not necessarily reflect the presence of the individual probe. It would be also conceivable that the full-length fluorescent-labeled antibodies might be decomposed. Indocyanine green would then have remained undecomposed in immune cells and only been slowly degraded over time. Hence, phagocytic cells are labeled indirectly. The use of ICC for immunohistochemistry as labelling dye instead of ICG (as for in vivo imaging) represents a potential confounder for interpretation of the postmortem analysis. This approach had to be chosen due to technical difficulties for ex vivo imaging of near-infrared dyes at high resolution. As both dyes are cyanine dyes with similar chemical properties, we believe that the use of ICC for ex vivo analysis instead of ICG is a rational approach. 

Vascular endothelial growth factor levels have not been directly measured. However, the localization and the time course of VEGF in the laser-induced CNV animal model have been previously reported.^16,17 As the spatial-temporal fluorescence signals of fluorescent-labeled B20-4.1.1 or AF564 were similar to the nonspecific rat IgG at the site of laser lesions, no clear evidence for the antibody accumulation at sites of presumed VEGF hyperexpression could be demonstrated. This was in contrast to the initial in vitro findings with fluorescent-labeled bevacizumab in human cell cultures and may be related to various reasons. These include the different species, the in vivo setting, the time course of image acquisition, and the applied concentration of samples. An additional approach for future experiments could address VEGF receptor targeting, assuming that specific accumulations of VEGF activity could be rather observed for a more stationary receptor as compared with free-floating VEGF. 

Also, a rat model was used whereby rodents do not have a macula and the model of laser-induced CNV differs from neovascular AMD in the human eyes. In addition, translation of results from the rodent model to human is limited. Of note, it remains unclear if a presumed similar reaction with the visualization of fluorescent spots may also occur in the latter. 

In conclusion, the first steps for molecular in vivo imaging of CNV applying fluorescent-labeled samples directed against VEGF are achieved in the current study. The basic principle of imaging fluorescent-labelled antibodies in the rat model of laser-induced CNV could be demonstrated. However, no specific binding of antibodies in comparison to controls was clearly established. In addition, the occurrence of hyperfluorescent spots that are likely caused by an immune response made the interpretation of in vivo observations challenging. Alternative approaches would include the use of VEGF antibodies directed against the VEGF receptor instead of soluble VEGF. Furthermore, the CNV model in monkeys may be less sensitive to specifies-specific effects and may give more insights into the pharmacodynamics of fluorescent labeled-VEGF antibodies, closer simulating the situations in humans. These experiments appear to be essential prior to a possible application in humans. Molecular imaging as investigated in this study may open the door for earlier diagnosis and more refined individualized anti-VEGF therapy in AMD patients in order to optimize functional outcome. 

Acknowledgments

The authors thank Torsten Krause for help with flow cytometry performance; Genentech for providing the antibody B20-4.1.1; Claudine Strack for excellent technical assistance in histology, Suzan Hunt for critical review of the manuscript, and Christian Kurts for providing CX₃CR1^GFP/+ mice. 

Supported by the German Ministry of Education and Research (BMBF), FKZ 13N10349 (Berlin, Germany). 

Disclosure: J.H. Meyer, Heidelberg Engineering (F), Carl-Zeiss Meditec AG (F), Genentech (F); A. Cunea, Heidelberg Engineering (F), Carl-Zeiss Meditec AG (F), Genentech (F); K. Licha, mivenion GmbH (E), P; P. Welker, mivenion GmbH (E); D. Sonntag-Bensch, Heidelberg Engineering (F), Carl-Zeiss Meditec AG (F), Genentech (F), P. Wafula, None; J. Dernedde, None; R. Fimmers, None; F.G. Holz, Acucela (C, F), Allergan (C, F), Bayer (C, F), Boehringer Ingelheim (C), Carl Zeiss Meditec AG (F), Genentech (C, F), Heidelberg Engineering (C, F), Merz (F), Novartis (C, F), Optos (F), Roche (C); S. Schmitz-Valckenberg, Heidelberg Engineering (C, F, R), Optos (F, R), Carl-Zeiss Meditec AG (F), Genentech (C), Novartis (C, F, R), Roche (C, F)