Investigative Ophthalmology & Visual Science, Vol. 33, No. 13, December 1992Copyright © Association for Research in Vision and Ophthalmology

In Vivo Imaging of Breakdown of the Innerand Outer Blood-Retinal Barriers

Harsha A. Sen,* Bruce A. Berkowirz,f Noburo Ando,* and Eugene de Juan, Jr*

Real-time contrast-enhanced magnetic resonance imaging (MRI) was used to distinguish betweenexperimentally induced breakdown of the vascular (inner) and retinal pigment epithelial (RPE; outer)blood-retinal barrier (BRB) in vivo. Pigmented rabbits were treated with intravenous sodium iodate 30mg/kg, (a specific RPE cell poison), intravitreal N-ethylcarboxamidoadenosine (NECA) 10~3 mol/1(which specifically disrupts the vascular BRB), or retinal diode laser photocoagulation. Coronal T,-weighted proton images were acquired in a timed sequence after intravenous injection of gadoliniumdiethylenetriaminepentaacetic acid (Gd-DTPA). Images were analyzed to localize leakage of Gd-DTPA and determine the permeability surface area product normalized per unit area (PS')- The patternof enhancement observed in eyes treated with sodium iodate differed clearly from that in eyes treatedwith NECA. PS' values were significantly higher in eyes treated with sodium iodate than with NECA.Simultaneous leakage from the outer and inner BRB in eyes treated with dense retinal laser photo-coagulation could be localized and quantitated independently. Invest Ophthalmol Vis Sci 33:3507-3512,1992

The blood-retinal barrier (BRB) is a selective physi-ologic barrier to the free diffusion of molecules fromthe bloodstream to the retina. It is made up of theretinal pigment epithelium (RPE), or outer BRB, andthe endothelium of the retinal vessels, or inner BRB.1

Through such shared characteristics as intercellulartight junctions and facilitated and active transportprocesses,1'2 the two epithelial layers maintain thestable intercellular microenvironment necessary forproper retinal function. However, they are anatomi-cally and physiologically distinct entities that can beaffected differently by disease processes. Vascular dis-eases such as vasculitis and vaso-occlusive disease ap-pear to disrupt primarily the inner BRB,3 while choroi-dal infarction and RPE damage can lead to break-down of the outer BRB in toxemia of pregnancy and

From the *Duke University Eye Center, Durham, North Caro-lina, and the fNational Institute of Environmental Health Sciences,Research Triangle Park.

This project was partially supported by NIH research grant R01EY07576, a Juvenile Diabetes Foundation grant, and a Heed Foun-dation Fellowship (Dr. Sen).

Dr. Berkowitz is a recipient of a Research to Prevent BlindnessCareer Development Award and is now with the Department ofOphthalmology, Southwestern Medical Center, Dallas, Texas. Dr.de Juan is now at The Wilmer Institute, Johns Hopkins Hospital,Baltimore, Maryland.

Submitted for publication: April 10, 1992; accepted July 13,1992.

Reprint requests: Eugene de Juan, Jr., The Wilmer Ophthalmo-logical Institute, 600 N. Wolfe Street, Maumenee 721, Baltimore,MD 21287-9277.

malignant hypertension.4-5 The individual contribu-tion of the inner or outer BRB to leakage can be moredifficult to delineate in complex processes such as ocu-lar inflammation, diabetic retinopathy, or prolifera-tive vitreoretinopathy,36 where anatomic changessuch as retinal detachment or neovascularization7

and chemical mediators8'9 may interact to produceBRB compromise. The development of logical andeffective therapeutic alternatives in these conditionsrequires that the role of each component of the BRBbe better understood.

Precise in vivo localization of leakage to the inneror outer BRB using existing techniques for evaluatingBRB breakdown often is difficult. The diffuse distri-bution of small vessels over the holangiotic retina andtheir proximity to the RPE render fluorescein angiog-raphy and vitreous fluorophotometry unable to distin-guish between the two with certainty. Histopathologicstudies using tracer substances such as horseradishperoxidase or immunohistochemical staining of na-tive albumin10 can localize leakage to specific sites butrequire that the eye be removed for tissue fixation,sectioning, and microscopy. This precludes in vivostudies of the physiologic kinetics of breakdown orlongitudinal studies in the same eye.

Recently, T,-weighted magnetic resonance imaging(MRI) in conjunction with intravenous administra-tion of the paramagnetic contrast agent gadoliniumdiethylenetriaminepentaacetic acid (Gd-DTPA) hasbeen used to quantitate breakdown of the blood-retinal barrier in the rabbit eye in vivo.111213 The anat-omy of the rabbit eye is unusual among mammals in

3507

3508 INVESTIGATIVE OPHTHALMOLOGY & VISUAL SCIENCE / December 1992 Vol. 30

that the retinal vessels are limited to a stripe of myelin-ated tissue that radiates horizontally from the opticnerve head, while the remainder of the retina is avas-cular (Fig. 1 A). A coronal MRI slice through the headcan be chosen that shows the eye bisected longitudi-nally from the optic nerve to the cornea perpendicularto the myelin wing (Fig. IB). In the resulting section,the vascular structures are segregated in the region ofthe optic nerve (solid arrow), and the avascular retinaextends superiorly and inferiorly (open arrowheads)to meet the ciliary body anteriorly. This study investi-gated the ability of Gd-DTPA-enhanced MRI to takeadvantage of these anatomic characteristics to differ-entiate between experimentally induced breakdown

BFig. I. (A) Fundus photograph of a rabbit eye. Retinal vessels are

limited to the myelin wing radiating horizontally from the opticnerve. (B) Diagram depicting a vertical section through the rabbiteye. Vascular structures associated with the myelin wing are indi-cated by the solid arrow, while avascular retina is indicated by openarrowheads.

of the vascular (inner) and pigment epithelial (outer)BRB in the rabbit eye.

Materials and Methods

The 2-3 kg minilop rabbits used in this study weretreated in accordance with institutional guidelinesand the ARVO Resolution on the Use of Animals inResearch. Experimental defects in the outer or innerBRB were created according to the following.

Outer BRB Defect

One group of rabbits received an intravenous injec-tion of sodium iodate 30 mg/kg via the marginal earvein 24 hr before measurement. A second group ofrabbits was anesthetized with subcutaneous ketamine35 mg/kg and xylazine 5 mg/kg and their pupils weredilated with phenylephrine 2.5% and cyclopentolate1% eye drops. Laser photocoagulation was applied tothe inferior retina of both eyes via an indirect diodelaser delivery system (810 nm, Oculight; Iris Medical,Mountain View, CA). One eye was treated with 300spots in a dense pattern (half burn-widths apart) andthe fellow eye was treated with 150 burns in a loosepattern (one-and-a-half to two burn-widths apart),with the first row of spots beginning about 1 mm be-low the optic nerve head. Burn duration was 300 msand power was titrated to achieve a medium-whiteburn, usually between 230 and 280 mW. Control ani-mals were exposed to the aiming beam for the sameamount of time required to perform the laser treat-ment.

Inner BRB Defect

N-ethylcarboxamidoadenosine (NECA) was ob-tained from Sigma Chemical Co., (St. Louis, MO) andwas made up in a concentrated stock solution of di-methylsulfoxide (DMSO; Fisher Scientific, FairLawn, NJ), filter-sterilized, and diluted to 10~3 mol/1in sterile phosphate-buffered saline solution. Animalswere anesthetized with subcutaneous ketamine andxylazine, and their pupils dilated as described above.Injections were administered by inserting a 30 G nee-dle 2 mm posterior to the limbus and slowly injecting0.1 ml of NECA or vehicle alone into the mid-vitreous cavity under visualization by indirect oph-thalmoscopy, 6 hr before measurement. NECA alsowas injected in identical manner into one eye of someanimals that had been treated with intravenous so-dium iodate.

All in vivo MRI experiments were performed aspreviously described,12 on a 4.7 T GE CSI horizontalbore system (General Electric, Freemont, CA) using awhole head, on-edge, split capacitance Helmholtz coil

No. 10 MRI EVALUATION OF BLOOD-RETINAL BARRIER BREAKDOWN / Sen er ol 0509

(diameter 9 cm) tuned to 200 MHz. Briefly, each ani-mal was induced with subcutaneous ketamine andxylazine and maintained on a continuous ketamine/xylazine intravenous infusion14 for the duration of theexperiment. Mechanical ventilation via an endotra-cheal tube was accompanied by monitoring of heartrate, blood pressure, blood gases, and rectal tempera-ture. The anesthetized animal was positioned in anonmagnetic cradle such that a single coronal slicewent through the center of both eyes and was orientedperpendicular to the myelin wing. Data were collectedusing a spin-echo sequence with TR = 450 ms and TE= 60 ms, and slice thickness of 3 mm for sodiumiodate experiments and 6 mm for all other experi-ments. The field of view varied between 57 X 57 mmand 75 X 75 mm depending on the head size. Imageswere obtained with 128 phase encode steps, 256 com-plex data points, and two acquisition/phase encodesteps. Approximately 3 min was needed to acquireeach image. The experimental protocol consisted ofcollecting a control image before the contrast agentwas infused. Gd-DTPA (0.5 or 1.0 mmol/kg, Magne-vist, Berlex Labs, Wayne, NJ) diluted with an equalvolume of normal saline then was injected via themarginal ear vein over a 3 or 6 min period, respec-tively. Images were collected such that with the mid-point of the injection as time zero (t = 0), the zerophase encode gradient of each image came at t = 3.0,6.5, 11.5, 16.5, and 21.5 min for the 3 min injection,and t = 4.5, 8, 13, 18, and 23 min for the 6 mininjection.

Image analysis was performed on a Macintosh FXII computer (Apple Computer, Cupertino, CA) usingthe program Image (W. Rasband, NIH, version 1.41).The analysis consisted of defining a region of interest(ROI) in one image, obtaining a mean signal intensityover that ROI, and applying the same ROI to theother images in the set. In each eye, the ROI for thevitreous was hand drawn and included the entire areaof Gd-DTPA leakage. An external vial (1 cm internaldiameter) containing Gd-DTPA-doped water was in-cluded in each image and served as a spatial calibra-tion and intensity standard. The change in ROI meansignal intensity over time was used to calculate thepermeability surface area product normalized perunit area (PS') using the method of Berkowitz et al.15

Statistical analysis was performed using Student's t-

ResultsIntravenous administration of sodium iodate 30

mg/kg produced leakage of Gd-DTPA across the en-tire avascular retina at 24 hr (Fig. 2). This area ofenhancement appeared to include the epithelium ofthe ciliary body and posterior iris in some animals. PS'

Fig. 2. Representative coronal magnetic resonance images ob-tained 24 hr after intravenous injection of sodium iodate 30 mg/kg.The image is oriented "upside-down," ie caudal structures are supe-rior and crania] ones are inferior. (A) Before infusion of Gd-DTPA.(B) Six-and-a-half minutes after Gd-DTPA infusion. (C) B-A sub-traction image. Arrowheads indicate leakage of Gd-DTPA acrossthe avascular retina. Broad arrow indicates epithelium of the ciliarybody and posterior iris. Short arrow indicates discontinuity in Gd-DTPA enhancement at the point of entry of the optic nerve andvessels.

for this group was 6.43 ± 0.8 X 10 4 cm/sec (mean± standard error of the mean; n = 10).

Magnetic resonance images obtained 6 hr afterNECA 10"3 mol/1 was injected into the vitreous cavitydemonstrated leakage of Gd-DTPA in the posteriorsegment limited to the location of the myelin wing(Fig. 3). The avascular retina remained unenhancedin all injected eyes. Mean PS' for these eyes was 2.48± 0.8 X 10"4 cm/sec (n = 10), significantly less thanthat for eyes treated with sodium iodate (P = 0.001).

3510 INVESTIGATIVE OPHTHALMOLOGY & VI5UAL SCIENCE / December 1992 Vol. 03

Fig. 3. Images obtained 6 hr after the intravitreal injection of 0.1ml of NECA 10"3 mol/l in both eyes. (A) Before infusion of Gd-DTPA. (B) Twenty three minutes after Gd-DTPA infusion wascompleted. Arrowheads indicate leakage of Gd-DTPA from the my-elin wing. Arrow indicates leakage from the ciliary body.

Although Gd-DTPA appears in the aqueous of nor-mal or control eyes after intravenous injection,12 leak-age from the ciliary body seemed enhanced in some ofthe eyes treated with NECA, and there appeared to beincreased posterior diffusion of tracer into the vit-reous.

Magnetic resonance images from eyes treated withsodium iodate and NECA (n = 2) were similar tothose obtained in eyes treated with sodium iodatealone. Gd-DTPA appeared to enter the eye across allblood-retinal interfaces, but inner and outer BRBbreakdown could not be separated on the image.Mean PS' for these eyes was 8.24 ± 1.7 X 10~4 cm/sec(n = 2). There was no significant difference betweenthis value and that for eyes treated with sodium iodatealone (P = 0.24).

Graphic comparison of PS' values between the dif-ferent groups is presented in Fig. 4. All treatmentswere significantly different from control (PS' = 8.8± 6.9 X 10"6 cm/sec; n = 7) with P = 0.027 for NECAalone and P < 0.0001 for sodium iodate or sodiumiodate with NECA.

Eyes that received diode laser photocoagulationshowed the expected leakage of Gd-DTPA from thearea of the laser burns (n = 4) (Fig. 5). However, ineyes that received 300 burns in a dense pattern (one-

control NalO Na 10 + NECA

Fig. 4. Graph of permeability surface area product normalizedper unit area (PS') for eyes treated with NECA, sodium iodate(NaIO3), or both together (NaIO3 + NECA) compared with control.Values are cm/sec and error bars represent ± standard error of,themean. N signifies number of eyes measured in each treatmentgroup. *P = 0.027. **P < 0.0001 by Student's unpaired t-test fordifference from control.

half burn-width apart), we observed additional leak-age from the vessels of the myelin wing that could belocalized and analyzed separately from that arisingfrom the lasered area. The corresponding area of themore lightly treated fellow eye (150 burns one-and-a-half to two burn-widths apart) did not demonstratethis leakage. Quantitative analysis confirmed this dif-ference with PS' for the myelin wing area (excludingleakage from treated avascular retina) nearly an orderof magnitude higher in the more heavily treated eyes(1.54 x 10"4± 0.19 cm/sec versus 1.69 x 10~5±0.19cm/sec, P = 0.018).

Fig. 5. Image obtained 48 hr after treatment with 300 burns ofdiode laser photocoagulation in a dense pattern (one-half burnwidth apart) below the optic nerve head in the eye on the left and 23min after intravenous infusion of Gd-DTPA. Leakage of Gd-DTPAin the area of treatment can be clearly seen (arrowheads). Addi-tional leakage can be seen from the vessels of the myelin wing (longarrow). This additional leakage is not seen in the fellow eye, whichreceived 150 burns in a looser pattern (one-and-a-half to two burnwidths apart).

No. 13 MRI EVALUATION OF BLOOD-RETINAL BARRIER BREAKDOWN / Sen er ol 3511

Discussion

The anatomic separation of the various compo-nents of the blood-ocular barrier in the rabbit eye facil-itate their differentiation in an magnetic resonanceimage. A vascular BRB—analogous to the inner BRBin humans in that both are made up of a tight nonfe-nestrated endothelium that restricts the passage ofsmall molecules such as albumin and fluorescein intothe vitreous—exists in the vessels of the myelinwing.16 The remainder of the retina is avascular. It isnourished by the choroidal circulation, from which itis separated by the tight junctions of the RPE in amanner analogous to the outer BRB in the human.The ciliary body is separated from the posterior com-ponents of the blood-ocular barrier by virtue of itsanterior location. The images presented in this studydemonstrate that contrast-enhanced MRI can takeadvantage of these anatomic characteristics to localizeleakage that occurs into the eye to any of these struc-tures.

Three different models of experimental BRB break-down were compared. The uniform breakdown of theouter BRB produced by sodium iodate, a specificRPE cell poison,17 differed in appearance from break-down of the inner BRB caused by NECA, a potentadenosine agonist that, when injected intravitreally,opens tight junctions between vascular endothelialcells, but not between the RPE.16 In eyes treated withNECA, Gd-DTPA leakage occurred in the area of themyelin wing and not from the outer BRB. The abilityof this technique to show separate contributions of thetwo barriers simultaneously was demonstrated by theimages taken in eyes that received 300 burns of laserphotocoagulation in a dense pattern. The unexpectedleakage from the myelin wing could be analyzed sepa-rately from the leakage from the treated area despitetheir proximity in the eye.

The vascular leakage observed in eyes treated withdense photocoagulation was not seen in the morelightly treated eyes and may have been the result of aninflammatory effect. This may be similar to the clini-cal experience with diabetic patients undergoingpanretinal photocoagulation, who sometimes showworsening of macular edema if they are treated tooheavily.18 That the retinal edema occurs in an arearemote from the area of treatment in these patientsmay mean that it results not from leakage from thelaser burns themselves, but from the macular capil-lary bed (inner BRB) because of the effects of diffus-ible inflammatory mediators.

In this technique, it appears that at least some de-gree of outer barrier integrity must remain to visualizecoexisting vascular or focal leakage. An intact RPEbarrier provides an unenhanced area in the image

against which such leakage can be seen. For example,if the outer BRB is only partially disrupted, as in theeyes treated with laser, simultaneous leakage from theretinal vessels is easily visualized. Complete disrup-tion of the outer barrier, however, tends to obscureother leakage, as in eyes treated with both sodiumiodate and NECA. We could not distinguish betweeninner and outer barrier leakage in these eyes, probablybecause Gd-DTPA entered rapidly and the tracercrossing each barrier coalesced to form a single diffu-sional front. Our technique incorporated a 3 or 6 mininfusion time for the tracer to minimize physiologicperturbations, as well as a 3 min period to acquireeach image. This technique would not have been sen-sitive to very early separation of this leakage.

The results of this study support the findings ofother investigators regarding the use of sodium iodateand NECA in producing experimental BRB compro-mise. Sorsby19'20 and Noell17 studied the effect of so-dium iodate on the eye using histopathologic tech-niques. They concluded that the RPE was the sole celltype affected. Our study confirms this effect on theouter BRB and extends the observation by includingthe epithelium of the ciliary body and posterior iris.Neither Sorsby, Noell, nor other investigators21-22

have reported examination of this area, so additionalstudy is required to determine whether the pigmentedor nonpigmented layers, or both, are affected. Be-cause the tight junctions found between the cells ofthe nonpigmented epithelium are thought to serve aspart of the blood-ocular barrier,23 sodium iodatelikely disrupts this layer, at least in part.

Vinores et al examined the effect of adenosine,NECA, and prostaglandin E, on the blood-retinalbarrier using an immunohistochemical stain for na-tive albumin to localize BRB breakdown.16 At theelectron microscopic level, they demonstrated open-ing of the vascular endothelial tight junctions, withalbumin staining extending the length of the openjunction from the luminal to abluminal side. Exami-nation of the RPE failed to reveal evidence of openingof junctions or albumin leakage. Our results supporttheir observations and suggest that the effect of NECAis limited to the vascular endothelium, with no effecton the outer BRB. In addition, although not quanti-tated in this study, our images indicate an increasedpermeability of the ciliary body and blood-aqueousbarrier in eyes treated with NECA.

MRI studies using Gd-DTPA as an intravasculartracer allow serial, noninvasive evaluation of ana-tomic and kinetic parameters of breakdown of theBRB. This study demonstrates, for the first time, an invivo technique to separate breakdown of the innerand outer BRB in a common experimental model. Bycombining imaging capability with determination of

3512 INVESTIGATIVE OPHTHALMOLOGY & VISUAL SCIENCE / December 1992 Vol. 33

PS, the technique also can be used to quantitativelyassay and localize the effect of intraocularly injectedputative vasoactive agents and other chemical media-tors on specific parts of the blood-ocular barrier. Theability to independently evaluate the components ofthis barrier in vivo may lead to a better understandingof their individual modulation and roles in diseaseprocesses.

Key words: blood-retinal barrier, Gd-DTPA, in vivo, MRI,permeability

Acknowledgments

The authors thank Alicia Gaitan, Alex Funk, and GlenToney for their technical assistance, Jeannie Peterson forassistance in preparing the manuscript, and Dr. RobertLondon for his support.

References

1. Bito LZ and DeRousseau CJ: Transport functions of the blood-retinal barrier system and the microenvironment of the retina.In The Blood-Retinal Barriers, Cunha-Vaz JG, editor. NewYork, Plenum Press, pp. 133-163, 1980.

2. Cunha-Vaz JG and Maurice DM: The active transport of fluo-rescein by the retinal vessels and the retina. J Physiol 191:467,1967.

3. Goldberg MF: Diseases affecting the inner blood-retinalbarrier. In The Blood-Retinal Barriers, Cunha-Vaz JG, editor.New York, Plenum Press, pp. 309-363, 1980.

4. Fastenberg DM, Fetkenhour CL, Choromokos E, and ShochDE: Choroidal vascular changes in toxemia of pregnancy. AmJ Ophthalmol 89:362, 1980.

5. Klein BA: Ischemic infarcts of the choroid (Elschnig apots). Acause of retinal separation in hypertensive disease with renalinsufficiency. A clinical and histopathologic study. Am J Oph-thalmol 66:1069, 1968.

6. Sen HA, Robertson TJ, Conway BP, and Campochiaro PA:The role of breakdown of the blood-retinal barrier in cell-injec-tion models of proliferative vitreoretinopathy. Arch Ophthal-mol 106:1291, 1988.

7. de Juan Jr E, Wilson D, and Hatchell D: Breakdown of theblood-retinal barrier in a model of retinal neovascularization.Invest Ophthalmol Vis Sci 28:1108, 1987.

8. Campochiaro PA and Sen HA: Adenosine and its agonists

cause retinal vasodilation and hemorrhages: Implications forischemic retinopathies. Arch Ophthalmol 107:412, 1989.

9. Miyake K: Prevention of cystoid macular edema after lens ex-traction by topical indomethacin: II. A controlled study in bilat-eral extractions. Jpn J Ophthalmol 22:80, 1978.

10. Vinores SA, Gadegbeku C, Campochiaro PA, and Green WR:Immunohistochemical localization of blood-retinal barrierbreakdown in human diabetics. Am J Pathol 134:231, 1989.

11. Plehwe WE, McRobbie DW, Lerski RA, and Kohner EM:Quantitative MRI in assessment of the blood-retinal barrier.Invest Ophthalmol Vis Sci 29:663, 1988.

12. Berkowitz BA, Sato Y, Wilson CA, and de Juan Jr E: Bloodretinal barrier breakdown investigated by real-time MRI follow-ing Gd-DTPA injection. Invest Ophthalmol Vis Sci 32:2854,1991.

13. Sato Y, Berkowitz BA, Wilson CA, and de Juan Jr E: Blood-retinal barrier breakdown caused by diode and argon laser en-dophotocoagulation: A comparative MRI study. Arch Ophthal-mol 110:277, 1992.

14. Wyatt JD, Scott RAW, and Richardson ME: The effect of pro-longed ketamine-xylazine intravenous infusion on arterialblood pH, blood gases, mean arterial blood pressure, heart andrespiratory rates, rectal temperature and reflexes in the rabbit.Lab Anim Sci 39:411, 1989.

15. Berkowitz BA, Sen HA, Ando N, de Juan Jr E, and Tofts PS:Validation of using real-time MRI following gd-DTPA injec-tion to measure the permeability surface area product of blood-retinal barrier breakdown. Invest Ophthalmol Vis Sci 33:1124,1992.

16. Vinores SA, Sen HA, and Campochiaro PA: An adenosine ago-nist and prostaglandin E, cause breakdown of the blood retinalbarrier by opening tight junctions between vascular endothelialcells. Invest Ophthalmol Vis Sci 33:1870, 1992.

17. Noell WK: Experimentally induced toxic effects of structureand function of visual cells and pigment epithelium. Am J Oph-thalmol 36:103-116, 1953.

18. Sorsby A: Experimental pigmentary degeneration of the retinaby sodium iodate. Br J Ophthalmol 25:58, 1941.

19. Sorsby A: The nature of experimental degeneration of the ret-ina. Br J Ophthalmol 25:62, 1941.

20. Anstadt B, Blair NP, Rusin M. Cunha-Vaz JG, and Tso MOM:Alteration of the blood-retinal barrier by sodium iodate: Ki-netic vitreous fluorophotometry and horseradish peroxidasetracer studies. Exp Eye Res 35:653, 1982.

21. Peyman GA, Vlchek JK, and Seth NV: Chorioretinal diffu-sional processes following pigment epithelial degeneration. In-vest Ophthalmol 14:58, 1975.

22. Streeten B: Ciliary Body. In Biomedical Foundations of Oph-thalmology, Duane TD and Jaeger EA, editors. Philadelphia,J.B. Lippincott, 1988, vol. 1, pp. 1-29.