Intravitreal Concentrations of a Near-Infrared Fluorescence–Labeled Biotherapeutic Determined In Situ Using Confocal Scanning Laser Ophthalmoscopy

Abstract

Purpose.: The pharmacokinetics of ophthalmic biotherapeutics are difficult to determine in human vitreous humor. Because of the high transparency of living tissue to near-infrared (NIR) light, the temporal changes in vitreous concentrations of a biomolecule labeled with an NIR fluorescent probe can be monitored in situ with a scanning laser ophthalmoscope (SLO).

Methods.: A humanized IgG was labeled with the NIR probe IRDye800CW (CVX-4164). Rabbits were given CVX-4164 intravitreally, and NIR fluorescence intensity was measured in the central plane of the vitreous humor with an SLO. Fluorescence intensities were converted to concentrations by using standard curves.

Results.: Little background fluorescence was detected, and the minimum detectable concentration of CVX-4164 was <10 nM. Vitreal concentrations of CVX-4164 determined in situ declined with time, with C _max ≈ 1 μM and t _½ = 145 hours (112-μg dose). The t _½ of CVX-4164 was approximately three times greater than that of the IRDye800CW alone, whereas the vitreal clearance (CL) and volume of distribution (V _ss) of the native dye were approximately 2000- and 550-fold greater than that of the conjugate. CVX-4164 concentrations determined in situ were 2.6 to 4.4 times higher than those determined by ex vivo NIR fluorescence or ELISA in homogenized vitreous humor, reflecting the greater spatial resolution of in situ imaging. Moreover, vitreal concentrations determined in situ were >3 orders of magnitude greater than plasma concentrations of CVX-4164, as determined by ELISA, and had a different kinetic profile.

Conclusions.: This study demonstrates the feasibility of determining the pharmacokinetics of intraocular biotherapeutics labeled with NIR fluorescent probes by in situ monitoring.

The treatment of retinal diseases has been significantly advanced by the advent of biotherapeutics and the ability to safely deliver these agents within the eye by intravitreal (IVT) injection. Examples of the array of structural classes of ocular biotherapeutics includes the anti-VEGF agents ranibizumab, a humanized monoclonal IgG1κ Fab fragment, ¹ and pegaptanib, a pegylated oligoribonucleotide aptamer, ² both of which are approved for treating neovascular age-related macular degeneration. Further, the antisense oligodeoxynucleotide fomivirsen is used for the treatment of cytomegalovirus-induced retinitis. ³ The efficacy, high specificity, relative safety, and duration of action of these agents and the unique pharmacokinetic properties of the ocular compartment lend promise to the continuing development of biotherapeutics for new ocular indications. 

Despite these significant advances in ophthalmic pharmacology, the vitreal pharmacokinetics of therapeutics in general and biologicals in particular remain unclear. Such knowledge would better inform the choice of safe and effective dose regimens of these agents during the clinical development phase. Ultimately, this guidance would maximize the effectiveness of the therapy while reducing patient risk. Currently, the clinical pharmacokinetics of ophthalmics in human vitreous humor and retina are predicted from the results of extensive preclinical observations of drug concentrations measured from repeated blood samples and terminal sampling of ocular elements. These results are then extrapolated to humans on the basis of multiple measures of plasma concentrations over time. The effects of intravitreal diffusion, differences in regional accumulation, protein binding, and other factors unique to the ocular environment on the kinetics of these molecules are complex and difficult to account for in the models used. ^4,5 Because of the impracticality and risk of repeated sampling of vitreous humor in humans, direct measurements of vitreal drug concentrations are typically reserved for samples obtained by surgical intervention (e.g., vitrectomy). ^{6 –8} Recently, attempts have been made to estimate intravitreal concentrations of biotherapeutics by sampling the aqueous humor. ^{9 –11} This method represents a significant improvement in the determination of intraocular drug concentrations through direct observation, as drug concentrations in the aqueous humor parallel those of the vitreous. ¹² Nonetheless, the limitations of this technique are evident in the long intersample interval (1 month or greater), the limited number of samples that can typically be taken (two to three per eye), and the requirement for high-sensitivity assays for measuring drug concentrations in small sample volumes. These are usually taken at time points beyond the elimination half-life of the drug. ^11,13 Therefore, the requirement for repeated, direct measures of intravitreal concentrations of ophthalmic therapeutics remains unfilled. 

In addition to direct measures, labeling the molecule with a reporter agent is a commonly used aid in determining the kinetics, distribution, and metabolism of a drug. Radioactive (¹⁴C, ¹⁸F) or stable isotopes (²H), which can be incorporated into small molecules without significantly altering their kinetics, are some of the most commonly used tags. ^{14 –18} These involve powerful but costly labeling and measurement techniques. In contrast, it is possible to inexpensively label larger molecules, such as biologicals, with fluorescent reporters. This modality is well suited for tagging biologicals, because the relatively high molecular mass of the fluor (≈1 kDa) would not significantly alter the biological's molecular mass and therefore the pharmacokinetics of the biotherapeutic (20–150 kDa). Despite these advantages, monitoring the fate of fluorescently labeled molecules in vivo is not a routine process, perhaps because most fluorescent labels are excited by light at ultraviolet (UV) wavelengths. UV light is absorbed or scattered within a beam path of a few millimeters by the water in cells, ¹⁹ limiting access to tissue compartments and introducing additional variables to the quantitation of drug concentrations in those compartments. Similarly, the signal emitted by the fluor is typically in the visible wavelength, the intensity of which is also attenuated by water. Added to the signal attenuation are the impacts of fluorescence quenching by biologicals, the high background fluorescence of biological environments, the pH dependence of fluorescence intensity, photobleaching of the fluor, and potential tissue damage after repeated exposure to high-intensity UV light. Together, these factors have made monitoring molecules tagged with UV fluors untenable for pharmacokinetics studies requiring repeated sampling. 

Many of these issues can be ameliorated by using fluorescent molecules excited by near infrared (NIR) wavelengths of light. ^19,20 Living tissue is highly transparent to photons in the NIR band of wavelengths (λ = 700–1300 nm). NIR light experiences reduced scattering in tissue and avoids water absorption of wavelengths shorter than 700 nm and hemoglobin absorption at wavelengths longer than 1300 nm. ^{19 –22} Moreover, the background fluorescence of tissue at NIR wavelengths is low, enhancing the potential signal-to-noise ratio for NIR fluors. The FDA-approved NIR dye indocyanine green has long been used to exploit this window of transparency. ²³ However, it has low fluorescence efficiency and physicochemical characteristics that render it unsuitable for labeling biotherapeutics. ^23,24 Recently, more efficient NIR dyes have been produced with chemical characteristics that are more amenable for labeling proteins and nucleotides. One of these, IRDye800CW (2-(3-{5-[7-(5-amino-1-carboxy-pentylcarbamoyl)-heptanoylamino]-1-carboxy-pentyl}-ureido)-pentanedioic acid; LICOR; Lincoln, NE), is water soluble with a λ_Ex of 778 nm and a λ_Em of 794 nm. ^25,26 The dye is biologically well tolerated, ²⁷ is manufactured to GMP (good manufacturing practice) criteria, ²⁴ and is otherwise suitable for labeling high-molecular-mass biologicals for in vivo imaging. ²⁸ In this study, we characterized the intravitreal pharmacokinetics of a humanized murine IgG (CVX-2000), used as a scaffold for delivering therapeutic peptide sequences, that was labeled with IRDye800CW. This was achieved by monitoring the NIR fluorescence of the molecule in situ using a confocal scanning laser ophthalmoscope (SLO) with a detector tuned to NIR wavelengths. The results of this investigation suggest that the pharmacokinetics of biotherapeutics tagged with NIR fluorescence labels may be determined by using repeated in situ monitoring of the compound in ocular compartments. 

Methods

Animals

These investigations were performed in male New Zealand White rabbits (Covance Research Products, Denver, PA), 3 to 5 months of age and 2 to 3 kg in body weight. The animals were housed individually in a psychologically enriched environment, with daily access to pelleted laboratory diet (PMI Certified 5322; PMI Nutrition International Inc., St. Louis, MO) and ad libitum access to water. Housing facilities were maintained at 15°C to 21°C in 30% to 70% humidity, with a 12-hour light–dark cycle. Veterinary care was available throughout the course of the study. The care and use of the animals were in accordance with the guidelines of the United States National Research Council, the Canadian Council on Animal Care, and the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. 

Reagent Preparation

The humanized murine IgG scaffold protein CVX-2000 was labeled with the IRDye800CW according to instructions provided in a kit (928-38040; LICOR, Lincoln, NE). ²⁶ The kit contains the N-hydroxysuccinimide form of the dye that reacts with free amino groups to form a stable conjugate. Care was taken to reduce the incorporation of too many dye molecules/IgG molecule, as this results in label quenching. The optimal number of dye molecules is two per protein molecule. It was established that a 2:1 molar ratio of CVX-2000:IRDye800CW to NHS resulted in 40% of the protein with one dye molecule attached, 23% with two, and 12% with three; 25% was unlabeled, yielding an average of 1.22 molecules of fluor/molecule of protein. This labeled protein is referred to as CVX-4164. After incubation for 3 hours at room temperature and desalting column clean-up, the labeled protein was diluted with sterile phosphate-buffered saline (PBS; pH 7.4), combined with histidine (10 mM), glycine (10 mM), and sucrose (58.4 nM), to final concentrations of 15 or 30 μM CVX-4164 for injection. In addition, the native dye IRDye800CW carboxylate (MW 1091; 929-09406; LICOR) ²⁶ was used. It was diluted to final concentrations of 30 or 90 μM in PBS for intravitreal injection. The group design, agent administration, and imaging schedules are summarized in Table 1  

Table 1.

 

Summary of the Experimental Design and Imaging Schedule

Treatment Group (n Animals)	Amount Injected (μg/Eye)	Test Agent		Eyes to Be Treated	Imaging Schedule
		Concentration of Injection Solution (μM)	Projected Final Concentration per Eye (μM)*
1. IRDye800CW carboxylate (4)	1.64	30	1	OU	1 Replicate measure/2 eyes/animal: 0.5, 1, 4, 24†; 24, 48†; 48, 72, 96‡; 96, 120‡
2. IRDye800CW carboxylate (4)	4.92	90	3	OU	1 Replicate measure/2 eyes/animal: 4, 24†; 24, 48†; 48, 96, 120‡; 144, 168‡
3. CVX-4164 (6)	112.5	15	0.5	OU	2 Replicate measures/2 eyes/animal: 12, 48, 96† h; 12, 24† h; 24, 96, 168, 240† h; 48, 168, 240† h.
4. CVX-4164 (6)	225	30	1	OU
5. CVX-2000 (8)	111.8	15	0.5	OS
	0	0	0	OD§
6. CVX-2000 (8)	224	30	1	OS
	0	0	0	OD§

 All eyes treated with a final volume of 50 μL. For IRDye800CW, pharmacokinetic values were determined using the concentrations based on the average of measures taken from two eyes per rabbit. For CVX-4164, values were determined using the concentrations from the average of two replicate measures per eye, with each eye treated individually. Groups 3–6 had plasma samples taken after each imaging session.

*  Assumes a vitreous volume of 1.5 mL.

†  Terminal sampling of vitreous occurred at these time points.

‡  Replicates 3 and 4 from groups 3 and 4 consisted of only one animal each.

§  In groups 5 and 6, OD received the injection vehicle.

In Situ Imaging

Administration of Test and Control Agents.

Rabbits were anesthetized for IVT administration of test substances, which was performed by a board-certified veterinary ophthalmologist. Topical gentamicin ophthalmic solution was applied to the injected eyes on the day before the injection, immediately after, and on the following day. Approximately 30 minutes before injection, 1% tropicamide drops were applied, and the rabbits were subsequently sedated with ketamine/glycopyrrolate/xylazine IM. Anesthesia (isoflurane/oxygen) was administered during the procedure as necessary. The conjunctiva was then flushed with benzalkonium chloride in sterile water (1:10,000, vol/vol). After imaging, a bland ophthalmic ointment was applied, except at 24 hours, when the animals would have received gentamicin ointment from the previous day's dose. 

All test substances were sterile filtered before injection. A fresh syringe with a 29- or 30-gauge needle was used for each IVT injection. All test agents were administered in a final volume of 50 μL/eye. The eyes were closed to avoid dryness or were hydrated with saline irrigation during anesthesia, as necessary. During anesthesia, the rabbits were hydrated with IV infusions of lactated Ringer's solution at 5 mL/kg/h. 

In Situ Measures of Near-Infrared Fluorescence.

Before each imaging session, the SLO (model HRA2; Heidelberg Engineering, Heidelberg, Germany) was calibrated according to NIR fluorescence intensity standards representing known concentrations of either IRDye800CW or the CVX-4164 batch used for injection. The wide-field lens was not used for these calibrations. Readings were taken in ICGA mode (λ_ex = 820 nm; λ_em =790 nm) at 40%, 50%, 60%, 70%, 80%, and 90% detector sensitivity. 

NIR fluorescent images of rabbit vitreous humor injected in situ with IRDye800CW, CVX-4164, CVX-2000, or saline were obtained with the SLO at the indicated times after injection (Table 1). The SLO was initially focused, with the standard (30°) lens, on the retina as a plane of reference, and an IR mode image of the retinal surface was captured (Fig. 1). The point of focus was retracted several millimeters (+8 D) into the vitreous, and images were acquired at 40% to 90% sensitivities with a supplemental wide-field (65°) lens. Subsequently, images were obtained at one sensitivity (50%) at +2 D through +42 D, into and through the vitreous, to characterize the volume of the injected material. Calibration standards were not used for these analyses. 

Figure 1.

 

NIR fluorescence images taken within the eye 1, 4, and 24 hours after intravitreal administration of 1.64 μg IRDye800CW. The focal plane of the SLO was adjusted to the retina and various depths within the vitreous humor using +4- and +8-D lenses. Images were acquired from the left eye of rabbit 103, with the detector sensitivity set at 50% of maximum. The circular area in the center of some images is an artifact of the rabbit lens. Other image artifacts result from hydrating perfusate drops on the optics.

After each imaging session, 0.5 mL of blood was collected into tubes containing K₂EDTA from the auricular artery or a vein. The plasma was separated and retained for subsequent determinations of CVX-2000 or -4164 concentrations with an enzyme-linked immunosorbent assay (ELISA). 

Images were saved in JPEG format and evaluated for fluorescence intensity using ImageJ (version 1.42q; developed by Wayne Rasband, National Institutes of Health, Bethesda, MD; available at http://rsb.info.nih.gov/ij/index.html) image-analysis software. 

Ex Vivo Measures of Compound Concentrations

At the end of the imaging series, the animals were euthanatized and samples (∼1.5 mL) of vitreous humor obtained. The sample from each eye was homogenized (Bio 101 FastPrep; Savant Instruments, Hicksville, NY) separately, and 200 μL from each supernatant was transferred into a clear-bottomed, 96-well plate for fluorescence intensity measurements in a calibrated spectrofluorometer (SpectraMax M₂ with SoftMax PRO, version 5.0.1; Molecular Devices, Sunnyvale, CA). 

CVX-2000 or -4164 concentrations in vitreous humor and plasma were determined by ELISA with an anti-CVX-2000 idiotype monoclonal antibody used for capture and a horseradish peroxidase (HRP)–labeled goat anti-human antibody (H+L; Bethyl Laboratories, Inc., Montgomery, TX) used for detection. For each assay, a standard curve was included, using CVX-2000 or -4164, of known concentrations ranging from 0.003 to 30 nM in a matrix of pooled rabbit vitreous humor (Pel-Freez Biologicals, Rogers, AR) or rabbit K₂EDTA plasma (Bioreclamation, Hicksville, NY). Samples were diluted so as to be within the linear range of the assay using pooled rabbit vitreous humor or rabbit K₂EDTA plasma. Before addition to the plate, a 20-fold dilution was performed (Super Block; Scytek Laboratories, Logan, UT), resulting in a final concentration of 5% vitreous humor or plasma and 95% assay diluent. After the initial incubation for 2 hours at room temperature with agitation and washing steps, the bound CVX-2000 or -4164 was detected with horseradish peroxidase (HRP)–labeled goat anti-human antibody (H+L). The substrate 3,3′,5,5′-tetra-methylbenzidine (Kirkegaard and Perry Laboratories, Gaithersburg, MD) was added to the plate and cleaved by the HRP. The reaction was stopped with a 2-M sulfuric acid solution, and the plate was read at 450 to 650 nm on an absorbance microplate reader (SpectraMax; Molecular Devices). The concentration of samples was determined by interpolation from a standard curve fitted using a five-parameter logistic equation calculated by the accompanying software (Softmax Pro5.2; Molecular Devices). 

Fluorescence Data Analysis

Standard curves comprising five standards in duplicate were used to convert fluorescence intensity from either images in grayscale or detector readouts to molar concentrations. The fluorescence intensity of standard concentrations (0.01, 0.1, 0.5, 1, 2, 3, 5, 7, and 10 μM) of either the native IRDye800CW carboxylate or CVX-4164 in PBS was measured before every imaging session. For tabletop spectrofluorometric detection, a λ_Ex of 780 nm, λ_Em of 800 nm, and λ_cutoff of 695 nm were used. Calibration cuvettes had a beam path 4 mm in length. For in situ imaging with the SLO, duplicate fluorescent images (ICGA mode) were obtained at each concentration using detector sensitivities ranging from 40% to 90% of maximum. Once the saturation level was determined, that maximum concentration was prepared and fluorescent images taken with neutral-density filters (Wratten [Kodak, Rochester, NY] and/or glass filters varying from 1% to 50% light transmission) at multiple detector sensitivities. Plotting fluorescence intensities at each level of detector sensitivity demonstrated a linear relationship between concentration and fluorescence intensity. The highest concentration of IRDye800CW in the linear range was 30 μM under the tested conditions, with higher concentrations saturating the detector. The lowest concentration of IRDye800CW reliably detected by SLO was <10 nM. 

The first image acquired with the SLO was taken at the retina in IR mode with the wide-field lens, followed by a second image at +8 D in IR mode, to position the sample plane. Duplicate images were taken at each level of detector sensitivity (40%–90%) at +8 D in fluorescence mode with the laser set for the ICG wavelength. Once the session was complete, images are opened in Image J for determination of grayscale values over the imaging region of interest (ROI). The ROI encompassed the entire field of fluorescence acquired at 90% sensitivity (≈1,493,962 pixels) and was superimposed over all fields acquired from that subject. Grayscale values were then converted to concentrations using the standard curve acquired before the session. Calibration curves were fitted to the data using nonlinear regression techniques (Prism; GraphPad Software, San Diego, CA). 

Pharmacokinetic parameters of IRDye800CW, CVX-2000, and CVX-4164 in vitreous humor and plasma were calculated from concentration–time data by using noncompartmental analyses (WinNonLin ver 5.2; Pharsight, Cary, NC) when data from three or more time points were available. Although examination of the vitreous concentration data suggested that the test agents could be characterized by one-compartment pharmacokinetics with monoexponential elimination, the pharmacokinetic parameters were determined using noncompartmental analyses, because no baseline conditions are assumed. The sparse-sampling model option was used to allow for estimation of error associated with exposure parameters for nonserial sampling of data sets. ²⁹ To allow for estimation of vitreal clearance (CL) and volume of distribution (V _ss), doses were converted to units of nanomoles based on an MW of 152,000 and 1166 g/mol for CVX-4164 and IRDye800CW, respectively. Two-way ANOVA followed by a Bonferroni-corrected post hoc comparison matrix (Prism; GraphPad Software) was used to assess the significance of differences in datasets where appropriate. 

Results

Initial imaging parameters, such as optimal focal plane, detector sensitivity, and dose were determined using the native IRDye800CW carboxylate. Both rabbit eyes were injected IVT with either 1.64 or 4.92 μg of dye in a volume of 50 μL (Table 1). Images were acquired shortly after injection at a detector sensitivity of 50%, with a range of focal points to optimize the imaging plane relative to the location of the injection (Figs. 1, 2). A range of fluorescence intensities were observed, depending on the diopter strength of the lens used for imaging and the dose of fluor, such that the maximum fluorescence intensity was obtained at +10 D after administration of the 1.64-μg dose of IRDye800CW, whereas +16 D provided the highest intensity for the 4.92-μg dose. However, fluorescence intensities were similar over a broad range of diopters regardless of dose, such that a uniformly high level of intensity could be obtained using a +8-D lens for both doses (Fig. 2), placing the focal plane approximately midway along the anterior–posterior axis of the eye. 

Figure 2.

 

Relationship between the NIR fluorescence intensity after administration of 1.64 or 4.92 μg of IRDye800CW and the ocular plane of focus. Consistently high intensities of fluorescence were acquired with the +8-D lens, regardless of dose. It was unclear whether the plane of focus was still in the vitreous humor at higher diopters (> +20 D). Detector sensitivity was set at 50% of maximum.

Although the +8-D lens was subsequently used to acquire most of the images in the study, adjusting the focal plane at early (1–24 hours) time points after administration indicated that the injected material required time to diffuse uniformly throughout the volume of the vitreous. Evidence of uneven dispersion was apparent at 1 and 4 hours after dye administration (Fig. 1, center column), with a more uniform fluorescence pattern appearing by 24 hours (Fig. 1, bottom row). Images of vitreal NIR fluorescence were captured using the +8-D lens over the 40% to 90% detector sensitivity range from other rabbits at 24 to 120 hours after administration of either 1.64 (Fig. 3) or 4.92 μg of IRDye800CW. Fluorescence intensity was dose- and time-dependent (Fig. 4). 

Figure 3.

 

Examples of NIR fluorescence images obtained from rabbit eyes at various detector sensitivities and times after IVT administration of IRDye800CW (1.64 μg). Percentages represent detector sensitivities. Labeling to the left indicates the subject number, time after administration that the images were taken, and the eye imaged. All images were acquired with a +8-D lens.

Figure 4.

 

Change in fluorescence intensity with time after the administration of IRDye800CW. (A) The intensity in GSVs of NIR fluorescence of rabbits treated with 1.64 μg of IRDye800CW declined with time. Higher sensitivities (90%) resulted in detector saturation (GSV >200), particularly at early time points (<96 hours). All images were acquired with a +8-D lens. Values represent the mean ± SD of observations from both eyes of two to three rabbits injected IVT with 1.64 μg IRDye800CW. (B) Changes in IRDye800CW concentrations determined after IVT administration of 1.64 or 4.92 μg, as determined using in situ and ex vivo fluorescence detection. In situ fluorescence intensity was acquired using the SLO, and ex vivo fluorescence of homogenized vitreous humor was measured with a tabletop spectrofluorometer. Conversions from GSVs to concentrations were made with calibration curves derived from images acquired with the SLO at 70% detector sensitivity for both doses. Data represent the mean ± SD of results from both eyes of four rabbits. (C) Pharmacokinetics of IRDye800CW in the vitreous humor, as determined from the in situ fluorescence data by noncompartmental analysis (WinNonLin, ver 5.2; Pharsight, Cary, NC). The line was fitted to the data from the 1.64-μg dose, with r ² = 0.88 and λ_z = 39.7 hours. Four points were used in the calculation. Point X was declared an outlier by the program and excluded from analysis. The line was fitted to the data from the 4.92 μg dose with r ² = 0.95 and λ_z = 34.1 hours. Six points were used in the calculation.

An example of the temporal dependency of the fluorescence intensity of 1.64 μg of IRDye800CW in the vitreous at various sampling sensitivities is indicated in Figure 4A. The detectors are saturated (grayscale values >200) throughout most of the sampling period when the highest sensitivity setting (90%) was used. However, fluorescence intensity was fairly linear over the 24 to 120 hours after administration at the 50% to 80% sensitivity levels, without evidence of detector saturation at the 24-hour time point. Converting grayscale intensity values acquired in situ at 70% sensitivity to dye concentrations provided data suitable for noncompartmental analysis of dye pharmacokinetics (Table 2). IRDye800CW resided in the vitreous humor with a t _½ of 28 to 34 hours and a C _max that increased with dose by approximately 2.5-fold. A tabletop spectrofluorometer was also used to measure NIR fluorescence intensity in ex vivo samples of vitreous at 24, 48, 96, and 120 hours after administration (Fig. 4B). Concentrations of IRDye800CW, measured ex vivo in vitreous samples taken from rabbits injected with 4.92 μg of dye, ranged from 1.5- to 3.0-fold higher than the corresponding concentrations determined from in situ measures (Fig. 4B) and were similar in amplitude and range to that observed after the administration of 1.64 μg (1.5- 6.7-fold). Noncompartmental analysis of IRDye800CW indicated that the pharmacokinetics were linear across doses, that the dye reached calculated maximum concentrations (C _max) of 1359 and 3860 nM at 24 hours after administration of 1.64 and 4.92 μg, respectively (Fig. 4C, Table 2), and that the dye had a t _½ of 30 to 40 hours. Moreover, the volume of distribution (V _ss) was approximately 400 mL. 

Table 2.

 

Summary of the Pharmacokinetic Parameters of IRDye800CW and CVX-4164 in Rabbit Vitreous Humor

Agent	Dose, μg (nmol)	C max (nM)	AUC0-last (nM/h)	AUC0-∞ (nM/h)	AUC Extrapolated (%)	t 1/2 (h)	CL (mL/h)	V ss (mL)
IRDye800CW	1.64 (1.41)*	1,350 ± 260	91.0 ± 7.2	109	17	40	7.87	457
	4.92 (4.22)*	3,860 ± 120	320 ± 9.1	325	3	34	7.97	363
CVX-4164	112 (0.76)†	954 ± 5	151,000 ± 103,00	201,000	25	125	0.0038	0.67
	2258 (1.52)‡	1,740 ± 7	263,000 ± 4,760	389,000	32	145	0.0039	0.82

 Pharmacokinetic parameters determined by analyzing vitreous humor concentration data obtained by in situ fluorescence, using noncompartmental techniques (WinNonlin ver 5.2, Pharsight, Cary, NC). Values represent the mean ± SEM.

  T _max = 24 hours in all cases.

*  70% detector sensitivity.

†  60% detector sensitivity.

‡  50% detector sensitivity.

These acquisition parameters (+8-D lens; imaging initiation time, ≥12 hours) were used to inform the measurement of an IgG labeled with IRDye800CW (CVX-4164) in the vitreous humor and plasma of rabbits. The doses are representative of currently used doses of ophthalmic therapeutics (e.g., 300 or 500 μg/eye for ranibizumab ^30,31 and 300 μg/eye for pegaptanib sodium ² ), with the duration of the sampling period spanning one t _½ of a typical ophthalmic biological. As observed with IRDye800CW, the fluorescence intensity of CVX-4164 was optimal when using a +8-D lens and declined monoexponentially with time (Fig. 5). The distribution of fluorescence was not always uniform before 24 hours, particularly at the lower detector sensitivities. Nonetheless, signal strength was high, with the NIR fluorescence intensity after the 112-μg dose saturating the detectors at 80% and 90% sensitivity (Fig. 6A) and the 225-μg dose at 60% to 90% sensitivity (Fig. 6B) for much of the sampling period. The intensity of background fluorescence, as measured in the presence of injection vehicle or up to 224 μg of the unlabeled CVX-2000, was negligible at all time points (<13 GSV; Figs. 5, 6). 

Figure 5.

 

Examples of NIR fluorescence images obtained from rabbit eyes over a range of times and detector sensitivities after IVT administration of labeled protein CVX-4164 (112 μg) or unlabeled protein CVX-2000 (112 μg). Percentages represent detector sensitivities. Labeling to the left indicates the agent administered, subject number, time after administration that the images were taken, and the eye imaged. All images were acquired with a +8-D lens. The NIR fluorescence intensity of the eyes injected with CVX-4164 declined with time. In contrast, the fluorescence intensity of the CVX-2000-treated eyes remained at background levels, regardless of the detector sensitivity.

Figure 6.

 

The temporal relationship between detector sensitivity and NIR fluorescence intensities of images obtained in situ by SLO after the administration of 112 (A) or 225 (B) μg of CVX-4164. NIR fluorescence intensities associated with administration of the unlabeled protein CVX-2000 (*, 112 or 224 μg) were superimposable on background fluorescence observed after administration of vehicle and did not vary with time or detector sensitivity. Results obtained using the highest detector sensitivity (90%) are shown. Measured grayscale intensities for CVX-2000 and vehicle were approximately 13 at all time points and doses. A curve was fitted to the data acquired at 40% detector sensitivity in a monoexponential decay model (Prism; GraphPad Software, San Diego, CA). The use of 50% (A) or 40% (B) sensitivity settings resulted in the least amount of detector saturation at the earliest time points. Values represent the mean ± SD of observations in six rabbits (both eyes), except for the CVX-2000- and vehicle-treated subjects, where n = 8 (single eyes).

The NIR fluorescence intensity values obtained in situ were subsequently converted to concentrations and analyzed using noncompartmental pharmacokinetics analysis (Fig. 7, Table 2). The choice of detector sensitivity providing the fluorescence measures used for conversion to concentrations was based on the degree of detector saturation at the earliest time points measured. Therefore, intensity measures obtained at 60% sensitivity after administration of 112 μg and 50% sensitivity after administration of 225 μg were used for conversion to concentration values. The maximum estimated intraocular concentration of CVX-4164 (C _max) was 954 nM after the 112-μg dose and 1740 nM after the 225-μg dose. The intraocular t _½ ranged from approximately 135 to 145 hours, and the V _ss was 0.7 to 0.8 mL. The concentrations of CVX-4164 based on in situ fluorescence of the 112-μg dose at 24, 96, and 240 hours after administration were 2.6 to 4.4 times greater than those determined using ex vivo fluorescence measures (Fig. 7A). Similarly, CVX-4164 concentrations determined by in situ fluorescence after administration of the 225-μg dose were 2.6- to 3.4-fold higher than those determined using ex vivo fluorescence measures (Fig. 7B). CVX-4164 concentrations determined by ELISA were similar to those determined by ex vivo fluorescence, but were consistently lower than the concentrations determined by in situ fluorescence, regardless of dose. In the cohort given 112 μg of CVX-4164, concentrations determined by in situ fluorescence were approximately 5.8-fold greater than those determined by ELISA. Moreover, CVX-4164 concentrations determined by in situ fluorescence after the 225-μg dose were three- to fivefold greater than the levels determined by ELISA. 

Figure 7.

 

Temporal changes in vitreal concentrations of CVX-4164 after IVT administration. (A) Vitreal concentrations of CVX-4164 (112 μg IVT) declined with time, but are significantly higher than vitreous concentrations determined ex vivo, with either a spectrofluorometer or ELISA at the 24- and 96-hour time points. In situ fluorescence intensity was acquired using the SLO, and ex vivo fluorescence of homogenized vitreous humor was measured with a tabletop spectrofluorometer. Conversions from GSVs to concentrations were made with calibration curves from images acquired with the SLO using the +8-D lens at 50% detector sensitivity. Values represent the mean ± SD of results from replicate measures taken from each eye of six rabbits. (B) Vitreal concentrations of CVX-4164 (225 μg), as determined in situ by SLO (in situ NIRF), were significantly higher than levels determined ex vivo with either a spectrofluorometer (ex vivo NIRF) or ELISA at 24, 96, or 240 hours. Concentration determinations were made using calibration curves and images acquired with the SLO using the +8-D lens at 40% detector sensitivity. Values represent the mean ± SD of results from both eyes of six rabbits. (C) The pharmacokinetics of CVX-4164 (112 and 224 μg) in the vitreous humor, as determined from the in situ fluorescence data using noncompartmental analysis (WinNonLin, ver 5.2; Pharsight, Cary, NC). A line was fitted to four points from the 112-μg dose dataset, with r ² = 0.87 and λ_z = 125 hours. Four points from the 225-μg dataset were used in the calculation of a line, with r ² = 0.89 and λ_z=145 hours. **Significantly different from contemporaneous CVX-4164 concentrations determined by different measurement techniques (ex vivo, in situ). P < 0.01, two-way ANOVA and Bonferroni's post hoc comparison.

Significant differences were observed between the concentrations of CVX-4164 in the vitreous humor and in the plasma (Figs. 7, 8, Tables 2, 3). As would be expected after direct administration into the ocular compartment, the T _max for CVX-4164 in the plasma lagged behind that of the vitreous by approximately 96 hours. The vitreous C _max values were approximately 1300- to 2200-fold greater than those of plasma. However, the plasma t _½ of CVX-4164 was comparable to that in the vitreous humor (∼6 days). 

Figure 8.

 

Temporal relationship between CVX-4164 and -2000 concentrations in the vitreous humor and plasma, as determined by ELISA. (A) Contemporaneous concentrations of CVX-4164 (112 and 225 μg) and CVX-2000 (112 μg and 224 μg) in the vitreous humor, as determined by ELISA, did not differ significantly between themselves. (B) Plasma levels after administration of 112 μg of CVX-2000 were significantly higher than those after 112 μg of CVX-4164 at 24, 48, 168, and 240 hours after administration. Similarly, plasma concentrations of CVX-2000 (224 μg) were significantly higher than those of CVX-4164 (225 μg) at 168 and 240 hours after administration. Regardless of the analytical modality used (in situ or ex vivo fluorescence, ELISA), concentrations of CVX-4164 in the vitreous humor were two to three orders of magnitude greater than in the plasma. Values represent the mean ± SD of results in four to six rabbits. *Significantly different from contemporaneous CVX-4164 concentrations. P < 0.05, respectively, two-way ANOVA and Bonferroni's post hoc comparison matrix.

Table 3.

 

Summary of the Pharmacokinetic Parameters of CVX-2000 and CVX-4164 in Rabbit Plasma

Agent	Dose (μg)	T max (h)	C max (nM)	AUC0-last (nM · hr)	t 1/2 (h)
CVX-2000	112	168	1.52 ± 0.07*	254 ± 11*	ND
	224	168	3.34 ± 0.20*	522 ± 25*	ND
CVX-4164	112	96	0.71 ± 0.06	109 ± 160	142
	225	96	1.53 ± 0.11	277 ± 22	214

 Parameters determined using noncompartmental analysis of concentration data determined by ELISA.

 Values represent the mean ± SEM of results from four to six rabbits. ND, t _1/2 for CVX-2000 could not be determined because of an incomplete absorption phase.

*  Significantly different from the corresponding value for CVX-4164-treated animals; P < 0.01, two-way ANOVA followed by Bonferroni-corrected post hoc comparison matrix.

The results of the ELISA allowed an examination of the pharmacokinetics of the unlabeled CVX-2000 in vitreous humor and plasma (Table 3, Fig. 8). Concentrations of CVX-2000 in the vitreous humor after IVT administration of 112 or 224 μg were similar to those observed after treatment with the corresponding doses of CVX-4164 (Fig. 8A). However, plasma concentrations of CVX-2000 after either the 112-μg (24, 48, 168, or 240 hours) or 225-μg (168 or 240 hours) doses were significantly higher than those of CVX-4164 at multiple time points. This was also reflected in significantly (approximately twofold) higher C _max and AUC values for CVX-2000 than for CVX-4164. The prolonged absorption phase (slow appearance of compound in the plasma compartment after IVT injection) and insufficient sampling duration precluded the characterization of CVX compound elimination from the plasma compartment. 

Discussion

The difficulty of obtaining samples of vitreous humor with high temporal and spatial resolution for determining the pharmacokinetics of intraocular biotherapeutics underlies this attempt to monitor vitreal concentrations of biologicals in situ. Previous investigations using noninvasive techniques, such as contrast-enhanced MRI, ^32,33 have imaged the ocular globe and otherwise inaccessible structures with high resolution. Nonetheless, signal sensitivity (low millimolar versus nanomolar), scan durations (5–10 minutes vs. <1 second for optical techniques), and the unknown kinetics of biologicals labeled with contrast-enhancing agents in the eye remain as unresolved issues for the use of this modality. In contrast, labeling a biological with an NIR fluorescent ligand and administering it by IVT injection should allow repeated measures of intravitreal concentrations of the agent over time using an SLO tuned to operate at NIR wavelengths. In the present study, we investigated the temporal changes in NIR fluorescence intensity of both a small molecule (the native IRDye800CW carboxylate) and a biological of high molecular weight (the IRDye800CW-labeled IgG, CVX-4164) after injection into the vitreous humor. 

Using the SLO, it was possible to monitor in situ the NIR fluorescence of a 1.64-μg dose of IRDye800CW shortly after its injection into the vitreous humor of rabbits. Within the first hour after the injection, fluorescence was localized within a relatively small, highly intense volume in the vitreous. The fluorescence intensity saturated the SLO detector, preventing accurate reading of its concentration, but it is unlikely that it was greater than the 30-μM concentration of the injected solution. With time, the fluorescence was more evenly distributed throughout the vitreous, with a fairly uniform field of intensity observed by 24 hours after administration. Repeated measures over time allowed the vitreal pharmacokinetics of IRDye800CW to be modeled, yielding an approximate t _½ of 1 to 1.5 days, which is consistent with the t _½/molecular weight relationships of other small-molecule agents administered intravitreally to rabbits. ^34,35 The vitreal pharmacokinetics of the larger molecule CVX-4164 differed significantly from that of IRDye800CW. CVX-4164 diffused more slowly through the vitreal medium, such that it was not evenly distributed throughout the imaging field in some subjects for as long as 96 hours after administration. The t _½ of CVX-4164, as determined by noncompartmental analyses, was 5 to 6 days, which is three to six times longer than that of IRDye800CW. Because CVX-4164 is a humanized antibody administered to a rabbit, its clearance is more rapid than in humans. ³⁶ Indeed, the observed t _½ of CVX-4164 is similar to the range of t _½ previously reported for humanized antibodies administered IVT to rabbits (4.88–5.6 days). ^12,37 In summary, these observations are consistent with previous reports that vitreal clearance rates for therapeutics depend on molecular weight and, thus, diffusion, as well as ocular geometry, and in the case of some biologicals, species. ^{38 –41}  

Additional insight into the diffusion characteristics of compounds in the vitreous was provided by measuring the V _ss. The V _ss of IRDye800CW in the vitreous humor was greater than the physiological volume of 1.5 to 2 mL, consistent with the extensive distribution expected of a small molecule over the first 24 hours and its potential interaction with ocular components. ⁴² Consistent with this observation, the V _ss of CVX-4164 in the vitreous humor (0.7–0.8 mL) was less than the physiological volume of the vitreous. This reflects the slower diffusion and limited distribution of the large molecule within the vitreous and contrasts with the measures observed using ELISA or NIR fluorescence of ex vivo samples. The lower concentrations detected in ex vivo samples was probably not related to alteration of the protein epitopes by the fluorescent labeling procedure, as the concentrations determined by ex vivo fluorescence of CVX-4164 were similar to those identified by the ELISA. Moreover, after administration of similar doses of CVX-2000 and -4164, the IVT concentrations of these agents, as determined by ELISA, were consistent. A more likely explanation for the discrepancies between the vitreal concentrations of these agents may be found in the different ways that the samples were processed for in situ and ex vivo analyses. Concentration measures by both ex vivo fluorescence and ELISA were made by using terminal samples, where the vitreous humor was removed and homogenized before sampling for assay. In this process, the fluorescent material is homogeneously distributed throughout the vitreous, eliminating any variations in local concentrations. In contrast, the in situ fluorescence measures made in this study were spatially constrained, despite attempts to measure vitreal fluorescence at time points when the material was uniformly distributed. Therefore, concentration gradients were manifested as regionally higher levels of fluorescently labeled material than would be observed in a more homogeneous preparation. The choice between determining the local or more general concentration may be provided in future assessments by acquiring fluorescence intensities in multiple focal planes and reconstructing the images into a three-dimensional image, or z-stack. ^43,44 This method will allow not only regional variations in concentration to be assessed, but also the boundaries and therefore volume of the area of fluorescence, resulting in better agreement between concentrations determined using different technologies. 

Not surprisingly, substantial differences between the vitreal and plasma pharmacokinetics of CVX-4164 were observed. Plasma concentrations (C _max) of CVX-4164 were less than 0.1% of those measured in situ in the vitreous, suggesting that local concentrations of biotherapeutics may remain in an efficacious range longer than suggested by estimates based on plasma levels. Although additional studies of the systemic pharmacokinetics of CVX-4164 must be performed to determine the K _a, and thus, whether intraocular CVX-4164 behaves according to “flip-flop” kinetics, ^45,46 the similar t _½ of CVX-4164 in the plasma and vitreous suggests that transfer of CVX-4164 from the eye into the plasma may be the rate-limiting step for its clearance. Plasma pharmacokinetics of CVX-4164 and -2000 as determined by ELISA also differed greatly. Although the vitreous concentrations of both agents were similar, plasma concentrations of the unlabeled compound were much higher than those of the labeled agent, suggesting that peripheral clearance of the labeled agent was more rapid. Plasma samples should be taken at later time points after administration, to better characterize the t _½ of CVX-2000 and to determine whether the peripheral clearance of CVX-4164 is more rapid than the unlabeled protein. If the differential clearance rates are sustained, it may be the result of rapid accumulation of fluorescently labeled protein in the liver and kidney. ⁴⁷ This may have a salutary effect on imaging by reducing background fluorescence associated with nonspecific interactions of a labeled IgG, thereby improving the signal-to-noise ratios for systemic imaging procedures. 

In summary, this study provides preliminary evidence that the intravitreal concentrations of a biotherapeutic agent labeled with an NIR fluorescent probe may be repeatedly monitored in situ by SLO. This technique is sensitive (<10 nM), with a workable dynamic range (100×, which may be further extended), low background fluorescence, and superior spatial resolution compared with more invasive techniques. In addition, living subjects were repeatedly scanned over a 10-day period without overt evidence of acute ocular toxicity. Because the near-infrared fluorescent (NIRF) dye used in this study has undergone toxicologic characterization and is currently in clinical trials, this technique may become a valuable modality for monitoring the clinical pharmacokinetics of biotherapeutic agents in various ocular structures, including the anterior chamber and possibly the cornea and tear film. Moreover, biologicals labeled with NIRF probes may become useful for monitoring ocular biomarkers or as diagnostics. Using techniques with greater spatial resolution, such as adaptive optics coupled to SLO ⁴⁸ linked with Fourier-domain OCT, ⁴⁹ should allow the visualization of individual cells fluorescently labeled with appropriate reagents. For example, the Fab fragment of a monoclonal antibody directed against annexin and labeled with an NIRF probe could allow visualization of apoptotic cells in the retina, which could be used as an endpoint in the development of retinoprotective agents. Moreover, the combination of SLO/OCT ⁵⁰ could allow the anatomic registration of an NIRF signal in the retina, as well as aid in the volumetric reconstruction of the vitreal injection site of an NIRF-labeled biotherapeutic. Thus, the potential applications of this molecular imaging technique for solving problems of ocular clinical pharmacology, developing biomarker surrogates of ocular function, and diagnosing ocular disease are considerable. 

Footnotes

 Supported by Pfizer Inc.

Footnotes

 Disclosure: A.S. Basile, Pfizer Inc. (E); G. Glazier, Charles River Laboratories (E); A. Lee, Pfizer (E); L.-Y. Jiang, Pfizer Inc. (E); T.R. Johnson, Pfizer Inc. (E); M.J. Shields, Pfizer Inc. (E); M. Vezina, Charles River Laboratories (E); V.R. Doppalapudi, Pfizer Inc. (E)