Abstract
purpose. To investigate the passive bidirectional and active outward transport of
fluorescein through the blood–retina barrier (BRB) in diabetic
patients with clinically significant macular edema and in healthy
controls.
methods. The passive and active transport of fluorescein through the BRB was
quantitated by vitreous fluorometry. A previously developed method was
used to model passive transport. A new simulation model was developed
and evaluated for estimation of active transport. The study included 10
eyes of 5 healthy controls and 31 eyes of 20 diabetic patients with
clinically significant diabetic macular edema (CSME) in at least one
eye, totalling 25 eyes with CSME.
results. Passive permeability of fluorescein was increased by a factor of 12 in
eyes with edema compared to healthy controls (edema, 23.7 nm/sec;
healthy subjects, 1.9 nm/sec, P < 0.01), whereas
the active transport was doubled (edema, 84.1 nm/sec; healthy subjects,
43.5 nm/sec, P < 0.01). Unlike active transport,
passive permeability was related to the degree of retinopathy, in that
eyes with severe non-proliferative diabetic retinopathy had a passive
permeability that was significantly increased compared to moderate
retinopathy (32.1 nm/sec and 14.6 nm/sec, respectively, P < 0.05). The passive movement quantitated with
vitreous fluorometry was larger for diffuse and mixed leakage compared
to focal (P = 0.07).
conclusions. Insofar as the movement of fluorescein can be taken as a probe for the
movement of electrolytes and water, the pathogenesis of diabetic
macular edema seems to involve a disruption of the BRB, presumably its
inner component. The active resorptive functions of the blood–retina
barrier appear to be compensatorily increased to counteract edema
formation, although the increase is too small to prevent edema in the
face of severe leakage through the blood–retina
barrier.
The blood–retina barrier separates the neuroretina from
the blood. The barrier function is located at the retinal pigment
epithelium and the endothelium of the retinal vessels. The movement of
water through the blood–retina barrier appears to have two dominant
components: A passive (bidirectional) transport and an active transport
directed from the retina to the blood. Theoretically, macular edema
develops when the inflow of fluid into the retina exceeds the outflow.
Since the movement of water across the blood–retina barrier cannot be
traced noninvasively, clinical research has been confined to the use of
fluorescein as a surrogate marker of fluid and electrolyte
movement. Passive transport (permeability) of fluorescein has
been shown to increase in relation to the development of
retinopathy in the presence of macular edema.
1 2
In vitro studies of isolated retinal pigment–epithelium–choroid
preparations have shown that the outward active transport of
fluorescein is substantially larger than the passive transport and that
this transport is inhibited by metabolic (oubain) and competitive
inhibitors (probenecid).
3 4 5 In humans, the active
transport of fluorescein measured with vitreous fluorometry has also
been shown to be inhibited by probenecid.
6 A study of
patients with retinitis pigmentosa complicated with macular
edema
7 has shown an increase in active transport, whereas
the role of active transport in diabetic macular edema is unknown.
In the clinical study presented here, we have examined the passive and
the active transport of fluorescein through the blood–retina barrier
to evaluate the relative importance of these components in the
pathogenesis of diabetic macular edema. In addition, a simulation
method has been developed for the calculation of active transport.
Twenty-four diabetic patients with clinically significant
diabetic macular edema (CSME) in at least one eye were examined
consecutively. Both patients with IDDM and NIDDM were included.
Patients with proliferative retinopathy, cataract, pseudophakia, or
aphakia were excluded. Clinically significant macular edema was defined
according to the ETDRS criteria as retinal thickening within 500 μm
of the umbo, as hard exudates within the same 500 μm if associated
with retinal thickening, and as a >1 optic disc area of retinal
thickening if any part of the edematous area is within 1-disc diameter
from the umbo.
8
In eyes with posterior vitreous detachment or vitreous liquefaction
near the optical axis, vitreous fluorometry is unreliable as a tool to
assess the blood–retina barrier, because convection replaces the
otherwise steep preretinal diffusion gradient by a flat curve in front
of the retina.
7 Consequently, we used fluorometry scans
obtained 30 minutes after fluorescein injection to assess the
qualitative properties of the vitreous. If the vitreous curve was flat
immediately in front of the retina or throughout the posterior
vitreous, the eye was excluded. Fifteen eyes were excluded because of
such signs of vitreous liquefaction. Two additional eyes were excluded
because of a lid defect and vitreous hemorrhage. Of the remaining 31
eyes, 25 had CSME, whereas no CSME was found in 6 eyes
(Table 1) . The mean age was 56 years (range, 28–74), and the mean
diabetes duration was 10 years (range, 1–34). Metabolic control and
blood pressure had been determined every 3 months in the year before
examination. The mean arterial blood pressure value was 102 mm Hg
(range, 83–117) and the mean HbA
1c was 8.9
(range, 5.3–12.1).
Five healthy subjects (mean age, 24 years; range, 22–27) without eye
diseases, vitreous liquefaction, or known systemic disease were also
included.
The study was approved by the local medical ethics committee. All
participants gave their written informed consent after full information
according to the Helsinki declaration.
Stereoscopic color fundus photography and fluorescein angiography were
recorded using a Canon fundus camera (CF-60UV). Retinopathy was
graded on 60° fundus photographs using a procedure adapted to
the modified Arlie House description.
9 Presence of CSME
was evaluated on 40° fundus photographs by one ophthalmologist and
one technician. In case of disagreement, another ophthalmologist made
the final decision as to the grading level.
The fraction of leakage originating from microaneurisms versus more
diffuse leakage was evaluated on fluorescein angiography based on the
ETDRS system as focal, mixed or diffuse leakage (more than 67%,
between 33 and 67% and below 33% leakage from
microaneurisms).
10
Fluorescein and its metabolite fluorescein glucuronide were measured
using an ocular fluorometer (Fluorotron; OcuMetrics, San Jose, CA)
adapted to differential spectrofluorometry. With this method, the light
source is an argon laser changing rapidly between two different
excitation wavelengths (458 and 488 nm) allowing separate determination
of fluorescein and fluorescein glucuronide.
11 12 13 Pupillary dilation was induced by topical phenylephrine 10% and
cyclopentolate 0.5%. After a bolus injection of 14 mg/kg disodium
fluorescein, postinjection scans were performed at 30 and 60 minutes
for the calculation of the passive permeability of the blood–retina
barrier, and at 7, 8, 9, and 10 hours for the active transport (4 scans
at each session). Blood samples were obtained at 5, 15, 30, and 60
minutes and at the time of the late postinjection scans. Visual acuity
was measured with standard, retroilluminated ETDRS charts.
After intravenous fluorescein injection, the dye diffuses from the
retina into the vitreous. The concentration close to the retina is high
shortly after injection, decreasing drastically toward the center of
the eye
(Fig. 1) . In the anterior part of the eye, a high concentration of fluorescein
is seen in the anterior chamber, and a lower gradient builds up behind
the lens because fluorescein diffuses into the anterior vitreous from
the posterior chamber.
The passive transport or permeability (P
passive) of the
blood–retina barrier and the diffusion coefficient of fluorescein in
the vitreous (D) is calculated from the vitreous curve 1 to 5 mm in
front of the retina.
14 15 The model assumes a homogeneous
blood–retina barrier without active transport being involved in the
movement of fluorescein and an intact vitreous gel. The simplified
modeling of the barrier is justified, as the flux of fluorescein is
nearly unidirectional during the early phase, driven by the steep
concentration gradient from the plasma toward the vitreous. The model
corrects for variations in plasma concentration and absorption of the
lens.
In time, the flux of fluorescein from the plasma to the eye diminishes
and the net movement changes at the outward direction from the vitreous
to the blood. The preretinal gradient reverses so that the
concentration at the retina is lower than in the center of the vitreous
(Fig. 2) . Assuming equal electrical charges on both sides of the retina and
absence of bulk flow, the outward flux through the blood–retina
barrier equals the equidirectional flux in the nearby vitreous.
Assuming that the outwardly directed transport is mainly due to active
transport, a quantitative estimate can be derived from the equation:
Outward transport = (D ∗ dC/dx) ∗
C
r − 1, where dC/dx is the
preretinal fluorescein gradient, and
C
r − 1 is the fluorescein
concentration at the retina.
16 With this method, the
passive component of the outward transport is ignored and the active
transport is calculated from measurements at fixed time intervals, 7 to
10 hours after injection. A previous study attempted to account for the
non-negligible concentrations of fluorescein in the plasma at the time
of the late vitreous scans and their contribution to the net outward
flux at this time.
7 17 The effect of this correction is
small and does not solve the additional problem that there is a time
lag between the disappearance of fluorescein from the vitreous and the
plasma. In other words, it does not account for the non-steady–state
nature of the kinetics of fluorescein. This is of particular concern in
studies of macular edema, where considerable inter-eye variation is
seen in both the speed of diffusion in the vitreous (D) and in passive
and active transport. Consequently, a model is needed, where active
transport is calculated by taking the non-steady–state nature of the
system into consideration.
We made no attempt at determining the active transport of the
metabolite fluorescein glucuronide, because animal studies have shown
that fluorescein glucoronide movement across the blood–retina barrier
is predominantly passive and because the late-phase concentrations of
fluorescein glucuronide are close to the lower limit of the effective
range of the fluorometer.
The preretinal fluorescein curve was simulated using a compartment
model with two different types of geometry. In healthy subjects, a
spherical model of the eye was initially constructed as previously
reported.
3 18
The space between the retina and the center of the eye was divided into
approximately 300 conical cells
(Fig. 3) , the length of each cell being coupled to the diffusion coefficient of
fluorescein in the vitreous and defined from the Einstein equation for
diffusion in one direction, that is, along the optical axis of the eye
(cell length = √D ∗ 2 ∗ t), where D represents the diffusion
coefficient and t is time. The vitreous fluorescein concentration
calculations were reiterated once for each second elapsed between
injection and vitreous concentration measurement. Passive transport
from the blood into the first cell of the model was calculated from the
concentration difference between the observed plasma concentration
curve and the outermost cell multiplied by the passive permeability. An
initial estimate of outward transport was obtained using a previously
described simpler method (outward transport = D ∗ dC/dx ∗
C
r − 1). The diffusion
between cells was calculated using Fick’s law. The innermost cell in
the center of the eye was regarded as the point of symmetry. The method
for estimation of the active transport (T
active) employs
successive iterative model calculations comparing model and
experimental data using a least-quadrant method to obtain the best
possible fit between model and observations.
Simulated vitreous curves are shown in
Figure 4 , from 30 minutes to 10 hours. The time-related changes in the curves
are as expected from experimental data for healthy subjects.
The late-phase preretinal fluorescein concentration curves observed in
patients with marked diabetic macular edema generally did not fit the
curves predicted by the mathematical model
(Fig. 5) . To correct this discrepancy, the model was changed using a cubic cell
shape
(Fig. 6) with the same cell length as the conical cell that was used
previously. After this change of geometry, the model fitted well in
cases of marked edema.
The theoretical justification for the use of the cylindrical model is
the high degree of macular leakage that is not matched by a similar
increase in peripheral results in a preretinal concentration profile
similar to that obtained from a planar surface with a uniform leakage.
Since the 60° fluorescein angiograms used in the study did not show
marked peripheral leakage in any patient and because Dalgaard et al.
has shown that leakage from outside the macula does not contribute the
concentrations measured on the optical axis of the eye, we assume that
the sources of variation are to be found within the
macula.
19
Tests of the cubic model on healthy subjects and patients with minor
degrees of diabetic macular edema demonstrated that the cubic model
failed to fit the scans from these subjects. Our presumptive
explanation is that at 6 to 8 hours after the injection of fluorescein,
the concentration of fluorescein in the vitreous is determined by
diffusion from both the retina and from the anterior segment of the
eye. The relative leakage varies markedly between the groups of
subjects in the present study. In healthy subjects and diabetic
patients with minor degrees of leakage through the blood–retina
barrier, the fluorescein concentration in the anterior chamber is
approximately 5 times the concentration 1 mm in front of the retina at
one hour after injection. It is apparent at this point that fluorescein
from the posterior chamber and the ciliary body has reached the
anterior vitrous behind the lens
(Fig. 1) . Given 5 to 7 hours more time
to diffuse from here to the rest of the vitreous, it is obvious that
retrolental fluorescein will have moved into the posterior vitreous.
Consequently, the conventional assumption of spherical symmetry used in
the modeling of posterior vitreous fluorescein kinetics had to be
abandoned. Instead, the center of symmetry of the model was moved from
the geometrical center of the posterior vitreous hemisphere to a
position closer to the retina and the magnitude of this movement was
adjusted according to the observed difference in concentration between
the anterior chamber and the preretinal vitreous. Using such a model,
acceptable agreement with the observed vitreous curves was found both
in healthy subjects and in diabetic patients with various degrees of
macular edema
(Fig. 7) . The results confirm earlier studies,
4 18 which have
shown that diffusion in both directions between the anterior and the
posterior part of the eye is important when vitreous fluorometry is
performed many hours after injection.
To evaluate the final model, the calculated values for fluorescein
concentration at the retina (Cr) were compared with the
observed curves for all subjects. The correlation between calculated
and simulated values was very high (r = 0.90),
indicating that the model fitted well over the range of healthy
controls and diabetic patients.
Active transport as calculated by the final simulation model was
linearly correlated with the previous method
6 16 (
r = 0.91) with an overall increase in
T
active of 65%. The reason for the increase compared to
the previously used method is probably that the model calculates the
active transport right at the retina and not by backward extrapolation
of the preretinal curve observed from 1 to 5 mm into the vitreous. The
latter leads to underestimation of the gradient at the retina, because
the gradient theoretically must be steeper the closer it is to the
retina.
Passive fluorescein permeability was lowest for edema primarily
due to microaneurisms (focal edema), intermediate for mixed, and
largest for diffuse edema, where less than 33% of the leakage is
estimated to originate from microaneurisms.
Ppassive for focal, mixed, and diffuse edema was
found to 14.7 ± 18.5 (n = 8), 26.3 ± 17.5
(n = 9) and 29.9 ± 22.2 nm/sec (n = 8), respectively. The difference was not statistically significant
(ANOVA P = 0.07). Tactive was
found to 80.9 ± 51.7, 87.6 ± 52.2, and 83.6 ± 40.2
nm/sec, respectively.
Using fluorescein as a marker, we have examined the passive
permeability and active outward transport through the blood–retina
barrier with the aim of elucidating the pathogenesis of diabetic
macular edema.
Passive permeability was significantly increased in macular edema by a
factor of 12 compared to healthy subjects, with a considerable
variation that was associated with the variation in extent of the
edema. The more severe levels of passive transport were found in eyes
with advanced retinopathy and large areas of diffuse leakage on
fluorescein angiograms. Moderately increased passive transport was seen
in cases with leakage originating from microaneurisms and type I
patients were typically in this group.
The passive permeability of fluorescein and the larger metabolite
fluorescein glucuronide did not differ in spite of a large difference
in octanol/water partition coefficient between these compounds. Studies
of postmortem retinas from patients with diabetic retinopathy have
demonstrated transendothelial channels and changes in the tight
junctions
20 which could lead to equal transport of
fluorescein and fluorescein-glucuronide and the present study is in
agreement with previous studies on diabetic patients and macular edema
in retinitis pigmentosa
3 4 7 indicating similarities in
the breakdown of the blood–retina barrier despite a different stage of
disease or a different pathology.
Compared to healthy eyes, active transport was increased significantly
by a factor of approximately 2, indicating a stimulation of outward,
active transport concomitant with the increase in passive permeability.
Theoretically, a decrease in active transport may be involved in the
development of macular edema, but this theory cannot be confirmed by
this study. The significance of the increase compared to the changes in
passive permeability is not known and has to be investigated with
interventional studies.
The active, reabsorptive transport was larger than the passive
permeability in absolute values. Our knowledge of the transport systems
in the blood–retina barrier is limited and neither the stoichiometric
relation between passive and active transport nor the natural substrate
of the system that actively transports fluorescein are known. Our
results do not indicate whether the increase is caused by an increased
number of transport carrier sites or changes in affinity.
A small fraction of patients with macular edema demonstrated low values
for both passive and active transport. In four such eyes, the passive
permeability averaged 3.0 nm/sec, equal to the mean value in healthy
subjects + 2 SD. The active transport in these eyes averaged 31.8
nm/sec, which is below the mean value in healthy subjects. By
comparison, in the four eyes with the largest values for passive
permeability (51.2 nm/sec), the active permeability averaged 114.5
nm/sec. It remains to be investigated whether the spontaneous prognosis
and the response to photocoagulation treatment differs from that seen
in the other patients.
In conclusion, the present study indicates that the major change in
transport through the blood–retina barrier in diabetic macular edema
is due to an increase in passive leakage through a damaged barrier,
whereas the active transport, located at the retinal pigmentary
epithelium, is intact and probably responsible for the increased
resorptive activity that seems to arise in response to the disruption
of the inner blood–retina barrier. By a combination of angiographic,
clinical, and experimental evidence, we conclude that the defect is
located in the inner blood–retina barrier and pharmacological agents
should focus on preventing or reversing the defect. Whether a medical
stimulation of the active transport is feasible remains to be
investigated.
Supported by grants from The Danish Diabetes Association and Danish Eye Health Society.
Submitted for publication November 2, 1999; revised April 14 and September 14, 2000; accepted November 8, 2000.
Commercial relationships policy: N.
Corresponding author: Birgit Sander, Department of Ophthalmology, Herlev Hospital, University of Copenhagen, DK 2730 Herlev, Denmark.
bsander@idea.dk
Table 1. Clinical Characteristics and Visual Acuity (VA)
Table 1. Clinical Characteristics and Visual Acuity (VA)
| Type of Diabetes I/II | VA logMAR | HbA1c | Retinopathy |
| CSME, n = 25 | 5/20 | 0.0 | 8.8* | Moderate NPDR, † : 13 |
| | | | Severe NPDR, ‡ : 12 |
| No CSME, n = 6 | 5/1 | 0.1 | 9.4* | Moderate NPDR, † : 6 |
Table 2. Passive Permeability (Ppassive) and Active Transport
(Tactive) in Diabetic Patients
Table 2. Passive Permeability (Ppassive) and Active Transport
(Tactive) in Diabetic Patients
| Ppassive nm/sec | Tactive nm/sec | Tactive/Ppassive |
| CSME, n = 25 | 23.73 ± 19.7* , † | 84.18 ± 46.5* | 6.55 ± 7.1* , † |
| No CSME, n = 6 | 2.34 ± 0.8 | 77.52 ± 33.8* | 35.55 ± 12.0 |
| Healthy, n = 10 | 1.94 ± 0.4 | 43.45 ± 21.9 | 25.25 ± 19.5 |
The authors thank clinical photographer Hans-Henrik Petersen for
skillful assistance and nurse Anni Nielsen for photographic readings.
Claus Engler, Peter Dalgaard, and Peter Koch Jensen contributed with
insightful remarks.
Krogsaa B, Lund-Andersen H, Mehlsen J, et al. The blood–retina barrier permeability in diabetic patients. Acta Ophthalmol
. 1981;59:689–694.
[CrossRef] Bursell S, Delori F, Yoshida A, et al. Vitreous fluorometric evaluation of diabetics. Invest Ophthalmol Vis Sci
. 1984;25:703–710.
[PubMed]Yoshida A, Ishiko S, Kojima M. Outward permeability of the blood–retina barrier. Graefe’s Arch Clin Exp Ophthalmol
. 1992;230:84–87.
[CrossRef] Cunha-Vaz J, Maurice D. The active transport of fluorescein by the retinal vessels and the retina. J Physiol
. 1967;191:467–486.
[CrossRef] [PubMed]Koyano S, Araie M, Eguchi S. Movement of fluorescein and fluorescein glucuronide across retinal pigment epithelium-choroid. Invest Ophthalmol Vis Sci
. 1993;34:531–538.
[PubMed]Engler C, Sander B, Koefoed P, et al. Probenecid inhibition of the outward transport of fluorescein across the human blood–retina barrier. Acta Ophthalmol (Copenh)
. 1994;72:663–667.
[PubMed]Moldow B, Sander B, Larsen M, et al. The effect of acetazolamide on passive and active transport of fluorescein across the blood-retina barrier in retinitis pigmentosa complicated by macular edema. Graefe’s Arch Clin Exp Ophthalmol
. 1998;236:881–889.
[CrossRef] Early Treatment Diabetic Retinopathy Study Research Group. Photocoagulation for diabetic macular edema. Early treatment diabetic retinopathy study report number 1. Arch Ophthal
. 1985;103:1796–1806.
[CrossRef] [PubMed]Early Treatment Diabetic Retinopathy Study Research Group. ETDRS report number 10. Grading diabetic retinopathy from stereoscopic color fundus photographs—an extension of the modified Arlie House classification. Ophthalmology
. 1991;98:786–807.
[CrossRef] [PubMed]Early Treatment Diabetic Retinopathy Study Research Group. ETDRS report number 11. Classification of diabetic retinopathy from fluorescein angiograms. Ophthalmology
. 1991;98:807–822.
[CrossRef] [PubMed]Larsen M, Dalgaard P, Lund-Andersen H. Differential spectrofluorometry in the human vitreous: blood–retina barrier permeability to fluorescein and fluorescein glucuronide. Graefe’s Arch Clin Exp Ophthalmol
. 1991;229:350–357.
[CrossRef] Larsen M, Dalgaard P, Lund-Andersen H. Determination of spatial coordinates in ocular fluorometry. Graefe’s Arch Clin Exp Ophthalmol
. 1991;229:358–362.
[CrossRef] Larsen M, Lund-Andersen H. Lens fluorometry: light attenuation effects and estimation of total lens transmittance. Graefe’s Arch Clin Exp Ophthalmol
. 1991;229:363–370.
[CrossRef] Lund-Andersen H, Krogsaa B, La Cour M, Larsen J. Quantitative vitreous fluorophotometry applying a mathematical model of the eye. Invest Ophthalmol Vis Sci
. 1985;26:698–710.
[PubMed]Dalgaard P, Larsen M. Fitting numerical solutions of differential equations to experimental data: a case study and some general remarks. Biometrics
. 1990;46:1097–1109.
[CrossRef] [PubMed]Engler C, Sander B, Larsen M, et al. Fluorescein transport across the human blood-retina barrier in the direction vitreous to blood. Acta Ophthalmol. 1994;72:655–662.
Moldow B, Larsen M, Sander B, Lund-Andersen H. The effect of acetazolamide on passive and active transport of fluorescein transport across the normal blood–retina barrier. Invest Ophthalmol Vis Sci
. 1999;40(8)1770–1775.
[PubMed]Ogura Y, Tsukahara Y, Saito I, Kondo T. Estimation of the permeability of the blood–retina barrier in normal individuals. Invest Ophthalmol Vis Sci
. 1985;26:969–976.
[PubMed]Dalgaard P, Barker VA, Lund-Andersen H. Vitreous fluorometry: mathematical analysis of the effect of peripheral leakage on axial scans. Invest Ophthalmol Vis Sci
. 1989;30:1522–1526.
[PubMed]Ishibashi T, Inomata H. Ultrastructure of retinal vessels in diabetic patients. Br J Ophthalmol
. 1993;77:574–578.
[CrossRef] [PubMed]