Abstract
purpose. To investigate the feasibility of transscleral drug delivery to the
choroid and retina.
methods. An osmotic pump was used to deliver IgG across the sclera of pigmented
rabbits, and levels were measured in the choroid, retina, vitreous
humor, aqueous humor, orbit, and plasma over 28 days. This method was
then used to deliver an anti–intercellular adhesion molecule-1
(ICAM-1) monoclonal antibody (mAb), and its effect on inhibiting
vascular endothelial growth factor (VEGF)–induced leukostasis in the
choroid and retina was determined by measuring tissue myeloperoxidase
(MPO) activity.
results. Levels of retinal and choroidal IgG were significantly higher than
baseline at all points up to 28 days (P ≤ 0.01). IgG
levels in the orbit, vitreous humor, aqueous humor, and plasma were
negligible (P > 0.05). MPO activity in the choroid
of eyes treated with anti–ICAM-1 mAb was 80% less
(P = 0.01) than in eyes receiving an equal rate of
delivery of an isotype control antibody. Inhibition of MPO activity in
the retina was 70% (P = 0.01). The plasma
concentration of anti–ICAM-1 mAb was 31,000-fold less than the
concentration in the osmotic pump.
conclusions. Minimally invasive transscleral delivery can be used to deliver
therapeutic levels of bioactive drugs to the choroid and retina with
negligible systemic absorption. This method of ocular drug delivery may
be used in the treatment of a variety of chorioretinal
disorders.
Severe vision loss in age-related macular degeneration and
diabetic retinopathy is due to pathologic ocular
neovascularization,
1 whereas glaucoma and retinitis
pigmentosa are characterized by the degeneration of retinal ganglion
cells and photoreceptors, respectively. Together, these diseases are
the principal causes of irreversible blindness
worldwide.
2 3 4 Although inhibitors of angiogenic factors
can inhibit ocular neovascularization,
5 6 and growth
factors can rescue retinal ganglion cells
7 and
photoreceptors,
8 delivering these drugs to the eye remains
difficult and potentially dangerous.
Systemic delivery of these agents in doses capable of achieving
therapeutic levels in the eye may alter the physiology of other organs.
Repeated long-term intravitreous delivery, as would be required for
these chorioretinal disorders, runs the risk of significant local
complications such as retinal detachment, endophthalmitis, vitreous
hemorrhage, and cataract formation. In addition, linear molecules
larger than 40 kDa and globular molecules larger than 70 kDa cannot
diffuse across the internal limiting membrane of the
retina,
9 10 11 12 precluding intravitreous delivery for many of
the antiangiogenic therapies under development. Topical eye drops
cannot deliver these biologics to the choroid or retina because of long
diffusional path length, counter-directional intraocular convection,
lacrimation, and corneal impermeability to large
molecules.
13 Photoactivated liposomes or caged molecules
may hold promise for selective delivery
14 15 ; however,
radiational and thermal damages associated with these modalities, as
well as the limited repertoire of drugs that can be enveloped, limit
the clinical utility of these approaches at present.
We hypothesized that transscleral permeation may be a viable
alternative because of the sclera’s large and accessible surface area,
high degree of hydration rendering it conducive to water-soluble
substances, hypocellularity with an attendant paucity of proteolytic
enzymes and protein-binding sites, and permeability that does not
appreciably decline with age.
16 17 18 19 Previous in vitro
experiments have demonstrated that the sclera is permeable to molecules
as large as 70 kDa,
19 20 and we have recently shown that a
150-kDa IgG molecule can diffuse across this tissue.
21
Existing methods of transscleral delivery are either nonselective or
destructive. Drugs injected into the subconjunctival space can reach
intraocular tissues,
22 23 24 but do so in part via systemic
absorption.
25 Transscleral iontophoresis can cause
unacceptable retinal necrosis and gliosis.
26 We have
developed a minimally invasive transscleral drug delivery modality that
can deliver therapeutic concentrations of bioactive proteins to the
choroid and retina without significant systemic absorption or tissue
damage. We tested the efficacy of this delivery strategy by inhibiting
growth factor–induced leukostasis with a monoclonal antibody (mAb)
directed against a cell adhesion molecule.
Vascular endothelial growth factor (VEGF), a 46-kDa homodimeric
globular glycoprotein, induces the expression of intercellular adhesion
molecule-1 (ICAM-1) in tumor and retinal vascular endothelium and
regulates leukocyte adhesion to endothelial cells.
27 28 Inhibition of ICAM-1 also decreases VEGF-induced leukostasis and
angiogenesis in the cornea.
29 Because ICAM-1 mediates
leukocyte endothelial adhesion and extravasation into surrounding
tissue,
30 myeloperoxidase (MPO) activity can be used to
quantify the tissue sequestration of leukocytes.
31 We
therefore investigated whether transscleral delivery of a mouse
anti-human ICAM-1 mAb, which inhibits rabbit neutrophil adhesion
through cross-reactivity to rabbit ICAM-1, could inhibit VEGF-induced
leukostasis in the choroid and retina by measuring MPO activity in
these tissues.
The experiments adhered to the ARVO Statement for the Use of
Animals in Ophthalmic and Vision Research and guidelines developed by
the Animal Care Committee of the Massachusetts Eye and Ear Infirmary.
Dutch-belted rabbits were anesthetized with intramuscular ketamine (40
mg/kg; Abbott, N. Chicago, IL) and xylazine (10 mg/kg; Bayer, Shawnee
Mission, KS). Osmotic pumps (ALZET; ALZA, Palo Alto, CA) were
loaded with drug and incubated at 37°C before implantation. The
osmotic pump was implanted subcutaneously between the scapulae and
connected to a brain infusion kit (ALZA), which was modified so that
the tip could be secured to, and face, the orbital surface of the
sclera with a single biodegradable polyglactin 910 suture (Ethicon,
Somerville, NJ) in the superotemporal quadrant of the eye, 14- to 16-mm
posterior to the limbus (near the equator;
Fig. 1 ). Care was taken to make a partial thickness pass through the sclera.
If uvea, blood, or vitreous was observed during the procedure, the
experiment was terminated.
Blood was collected by cardiac puncture before surgical
enucleation of the eyes under deep anesthesia. Aqueous humor of each
eye was collected using a 30-gauge needle. Vitreous humor, retina,
choroid, and orbital tissue of both eyes were dissected and isolated
under a microscope. The choroid of the treated eye was separated into
two hemispheres: proximal (in which the tip of the pump was centered)
and distal to the tip of the pump. Animals were killed with
intracardiac pentobarbital (100 mg/kg; Vortech, Dearborn, MI).
ALZET 2ML4 osmotic pumps (4 weeks, 2.5 μl/h) containing
fluorescein isothiocyanate (FITC)–conjugated rabbit IgG (15.5 mg/ml;
Sigma, St. Louis, MO) were implanted in one eye of each animal. Animals
were killed at 3, 5, 13, 20, and 28 days after implantation, and
fluorescence was measured in ocular tissues and plasma. Clearance of
FITC–IgG was determined by implanting osmotic pumps (model 2001D;
ALZET; 24 hours, 8 μl/h) in one eye of each animal for 1 day and
measuring fluorescence in ocular tissues at 1, 3, 5, and 9 days after
explantation.
Choroid and retina obtained from eyes in which osmotic pumps
containing FITC–IgG were implanted were subjected to protein
precipitation with 20% trichloroacetic acid (Sigma),
32 and the supernatants were assayed for residual fluorescence, which
would suggest cleavage of FITC from IgG. As a confirmatory test, tissue
homogenates were placed in a diffusion chamber separated by fresh
virgin sclera
21 to determine diffusion kinetics of
fluorescent molecules in the tissue, which was compared with diffusion
kinetics of FITC–IgG, because significant differences between the
diffusion of tissue homogenate fluorescence and the diffusion of native
FITC–IgG also would suggest cleavage of FITC from IgG. Briefly, an in
vitro transscleral diffusion apparatus was constructed by
attaching fresh sclera to 2 spectrophotometry polystyrene cuvettes
(Sigma), each with a 5 × 10 window fashioned 2 mm from the
bottom, with a small amount of cyanoacrylate tissue adhesive (Ellman
International, Hewlett, NY). Transscleral diffusion of fluorescent
molecules at 37°C in a 5% CO
2 atmosphere was
determined by sampling every 30 minutes over 3 hours.
Fluorescence was measured (excitation, 495 nm; emission, 525 nm)
using a MPF 44A fluorescence spectrophotometer (Perkin–Elmer, Newton,
MA) in a right-angle geometry with 3-nm bandwidths. Corrections were
made for tissue autofluorescence using the values for the nonimplanted
eye.
ALZET 2001D osmotic pumps, one containing mouse anti–ICAM-1 IgG2a
mAb (2 mg/ml) from clone BIRR0001 and one containing mouse nonimmune
IgG2a mAb (2 mg/ml) (R&D Systems, Minneapolis, MN), were implanted in
the superotemporal quadrant of each eye. The surgeon was masked to the
identity of the two pumps. Six hours after implantation, animals were
anesthetized, and 0.5% proparacaine (Alcon, Ft. Worth, TX) and 0.3%
ofloxacin (Allergan, Hormigueros, PR) eye drops were topically applied.
After pump placement, 2 μg of human recombinant VEGF
(VEGF165), diluted in 100 μl of pyrogen-free Dulbecco’s phosphate-buffered saline (PBS; Sigma),
was injected into the vitreous body through the inferonasal pars plana
of each eye with a 30-gauge needle. To normalize intraocular pressure,
100 μl of aqueous humor was removed with a 30-gauge needle. Animals
were killed 24 hours after implantation and MPO activity was measured
in ocular tissues. To ensure that the intravitreous injection did not
provide an intraocular conduit for the antibodies, 2 animals were
implanted with ALZET 2001D osmotic pumps containing FITC–IgG, and
a 30-gauge needle was used to perforate the inferonasal sclera. The
fluorescence in ocular tissues, 24 hours later, was compared with that
in animals without the perforation.
Myeloperoxidase was extracted by freezing, thawing, and sonicating
tissue in 50 mM potassium phosphate buffer, pH 6.0 (Sigma), containing
0.5% hexadecyltrimethylammonium bromide (Sigma), three times. MPO
activity in supernatants was measured by the change in absorbance at
460 nm that resulted from decomposition of 0.0005% hydrogen peroxide
in the presence of 0.167 mg/ml
O-dianisidine
(Sigma)
31 and compared with the activity of 1 unit of MPO
(Sigma), using a microplate reader (model MR4000; Dynatech, Chantilly,
VA). The assay was performed in masked fashion.
A 96-well plate was coated with 10 mg/ml goat anti-mouse
IgG2a-affinity purified antibody (Bethyl, Montgomery, TX) in 0.05%
sodium carbonate, pH 9.6 (Bethyl), covered with parafilm, and stored
overnight at 4°C. The plate was washed twice with washing buffer (50
mM Tris with 0.05% Tween-20, pH 8.0; Sigma), and nonspecific
adsorption sites were blocked by incubation in postcoat buffer (50 mM
Tris with 1% bovine serum albumin, pH 8.0; Sigma) for 30 minutes at
room temperature (25°C). After two washings with washing buffer,
mouse IgG2a reference standard (Bethyl), and unknown samples of blood
diluted in postcoat buffer containing 0.05% Tween-20 (Bethyl) were
added to the wells. The plate was incubated for 1 hour at room
temperature before it was washed twice with washing buffer. Horseradish
peroxidase–conjugated goat–anti-mouse IgG2a (Bethyl) was diluted
(1:20,000) in postcoat solution containing 0.05% Tween-20 and added to
the wells. After 1 hour of incubation at room temperature, the plate
was washed twice with washing buffer. The 3,3′,5,5′-tetramethyl
benzidine peroxidase substrate solution and peroxidase solution B (both
from Kirkegaard and Perry, Gaithersburg, MD) were quickly mixed and
added to the wells. The reaction was carried out for exactly 10 minutes
in the dark at room temperature and then stopped with 2 M
H2SO4 (Sigma). The
absorbance was measured at 450 nm.
Tissue concentrations of FITC–IgG were compared by standard
linear ANOVA, and the paired Student’s t-test was used to
compare MPO levels between eyes. All probability values were
two-tailed. An α level of 0.05 was used as the criterion to reject
the null hypothesis of equality of means.
FITC–IgG was delivered to the superotemporal scleral surface at a
rate of 2.5 μl/h for 28 days via an osmotic pump. Levels of retinal
and choroidal fluorescence, a quantitative marker of IgG concentration,
were significantly higher than baseline at all times (
Fig. 2 ;
n = 4 per time,
P ≤ 0.01 for each
time). Levels in the orbit, vitreous humor, and aqueous humor were
negligible (
Fig. 3 ;
n = 4 per time,
P > 0.05 for each
time). No fluorescence was detected in the plasma at any time. The
concentration of IgG in the choroid in the hemisphere proximal to the
pump, which reached a plateau of 6% of the concentration in the
osmotic pump, was roughly 50% greater than in the distal hemisphere
and 50% greater than the overall retinal concentration. The
elimination of fluorescence from the choroid and retina followed
first-order kinetics, with half-lives of approximately 3 days (
Fig. 4 ;
n = 4 per time).
To confirm the continued linkage of FITC to IgG, protein precipitation
of tissue homogenates at various times was performed. Virtually all
fluorescence was protein-bound (99.6% in retina and 99.8% in choroid,
at 28 days;
n = 3), indicating that the IgG molecule
crossed the sclera intact and did not undergo significant cleavage over
the time studied. Additionally, the in vitro transscleral diffusion of
fluorescence from retinal tissue homogenates (mean permeability
coefficient = 6.2 × 10
−6 cm/sec) and
choroidal homogenates (mean permeability coefficient = 5.6 ×
10
−6 cm/sec) was not significantly different
(
P > 0.05 for both comparisons) from that of FITC–IgG
(mean permeability coefficient = 4.6 ×
10
−6 cm/sec),
21 indicating that the
tissue fluorescence emanated from intact FITC–IgG. Iatrogenic
perforation of the sclera at the injection site did not result in
increased intraocular delivery (
Table 1 , indicating that lateral surface diffusion did not play a significant
role in transscleral entry.
VEGF-induced leukostasis in the retina and choroid, as measured by
MPO activity, was markedly inhibited by the delivery of anti–ICAM-1
mAb
(Fig. 5) . Myeloperoxidase activity in the choroid of the eye treated with
anti–ICAM-1 mAb (2 mg/ml delivered at 8 μl/h) was 80% less
(
P = 0.01) than in the eye receiving an equal rate of
delivery of an isotype control antibody (
n = 5).
Inhibition of MPO activity in the retina was 70% (
P =
0.01;
n = 5). As expected, the diffusion of MPO, whose
molecular weight is 70 kDa, into the vitreous humor was minimal in both
groups of eyes. The plasma concentration of anti–ICAM-1 mAb, 64.5 ± 73.4 ng/ml, was 31,000-fold less than the concentration in the
osmotic pump.
We have demonstrated that antibodies can be delivered across the
sclera into the choroid and retina without significant proteolytic
degradation, loss of bioactivity, or systemic absorption. The target
selectivity (choroid/retina to plasma) was on the order of 700:1. The
low levels of fluorescence in the nonimplanted eye and the spatial
concentration gradient observed in the choroid of the implanted eye
also indicate no significant systemic absorption. We confirmed that the
fluorescence detected in the choroid and retina of the implanted eye
was not due to fluorescein that may have been cleaved from the
FITC–IgG complex by subjecting the tissue extracts to protein
precipitation and by analyzing the diffusion kinetics of the
fluorescent molecules in the tissues. The extremely low levels of
fluorescence in the vitreous humor also suggest that the antibody was
not cleaved from FITC, which would have readily diffused across the
retina.
11 The retained bioactivity of the anti–ICAM-1 mAb
also confirmed intact transscleral delivery of the protein.
Transscleral delivery of anti–ICAM-1 mAb resulted in significant
suppression of VEGF-induced leukostasis. The site through which VEGF
was injected into the vitreous is unlikely to have served as a conduit
for the mAb because experiments with FITC–IgG revealed no significant
increase in intraocular concentration of fluorescence resulting from
the creation of a scleral perforation at the pars plana, 20-mm distant
from the pump tip. Furthermore, even if the perforation resulted in
increased vitreous levels of mAb, it is unlikely to have had any impact
on the retinal or choroidal vasculature, owing to the diffusion barrier
of the internal limiting membrane of the retina.
9 10 11 12
We speculate that the high ocular selectivity is due to the sponge-like
nature of the sclera. Although bolus injections of drugs into the
subconjunctival space may overwhelm the absorptive capacity of the
sclera, leading to systemic absorption,
25 the gradual
delivery used in these experiments may permit more complete scleral
absorption because the drug-sclera contact time is greater than the lag
time to steady state flux.
33 A long-term transscleral
delivery device may be clinically feasible because the human eye is
remarkably tolerant of foreign bodies overlying the sclera, such as
scleral buckles used in treating retinal detachment, even for
years.
34 Moreover, human sclera is
hypocellular
16 and has a large surface area (16.3
cm
2),
18 both of which would
facilitate diffusion.
Angiogenesis plays an integral role in the pathogenesis of a variety of
posterior segment disorders. Because laser photocoagulation, the only
proven therapy for intraocular neovascularization,
35 36 is
quite destructive, there has been a rapid growth in the development of
drugs targeting VEGF receptors, integrins, matrix metalloproteinases,
and other molecules. Although these potentially effective therapies are
under development, there is a parallel need to develop drug delivery
systems to the choroid and retina that are safe, effective, and
feasible for long-term use. We have described a nondestructive and
minimally invasive method for the targeted delivery of high molecular
weight compounds to the posterior segment of the eye. This drug
delivery method exhibits linear kinetics of absorption and elimination,
with the potential to deliver constant doses of medication. It is
robust because it is not limited to delivering antiangiogenic drugs,
but can be extended to neuroprotective agents or vectors for gene
transfer, which hold promise in the treatment of
glaucoma
7 37 and other chorioretinal
degenerations.
8 38
The authors thank Robert Rothlein at Boehringer Ingelheim
(Ridgefield, CT) and Napoleone Ferrara at Genentech (San Francisco, CA)
for their generous gifts of anti–ICAM-1 mAb and VEGF, respectively;
Tiansen Li at the Massachusetts Eye and Ear Infirmary for the use of a
sonicator; and Ambati M. Rao and Balamurali K. Ambati for critical
review of this article.