Fluorescent Vesicle System

Fluorescent Vesicle System A New Technique for Measuring Blood Flow in the Retina Bahram Khoobehi, PhD, Gholam A. Peyman, MD Purpose: To measure blood flow in the retinal circulation and optic nerve head capillaries with an innovative fluorescent vesicle system. Methods: Carboxyfluorescein was encapsulated into liposomes; the vesicles ranged from 0.1 to 2 J,Lm in diameter. After intravenous injection of the liposome suspension, the fundus was viewed using a scanning laser ophthalmoscope. Images of the fundus showing circulating liposomes were stored on videotape. An image analyzing system was used to digitize the captured video frames and transfer them to a computer's hard disk for permanent storage. Software developed in the authors' laboratory allowed them to overlay multiple video frames to create a single image that provided a visible record of the path taken by a particular vesicle in a given time period. The information on this image was used to calculate the velocity of the vesicle, and hence the velocity of the blood flow in the vessel. Results: With this system, individual liposomes as small as 100 nm were visible in all retinal vessels (arteries, capillaries, and veins). Quantitative analysis of vesicle movement in the major retinal vessels of the cynomolgus monkey yielded an average velocity of 9.33 ± 1.67 rnrn/second in a large vein (diameter, 130 J,Lm) and 16.10 ± 5.7 mm/ second in a large retinal artery (diameter, 64 J,Lm). The average velocity in the macular capillaries was 0.76 mrn/second (range, 0.45-1.33 rnm/second), whereas the average velocity in the optic nerve head capillaries was 1.39 rnrn/second (range, 0.96-2.25 mm/ second). Conclusion: The fluorescent vesicle system can be used for simultaneous measurement of blood flow in the retinal arteries, veins, and capillaries of the macula and optic nerve. Ophthalmology 1994;101:1716-1726

Current methods for measuring blood flow in the ocular vasculature include dye dilution.l" laser Doppler velocimetry.v " the blue-field entoptic phenomenon.P'!" laOriginally received: October 6, 1993. Revision accepted: June 14, 1994, From the LSU Eye Center, Louisiana State University Medical Center School of Medicine, New Orleans. Presented at the Association for Research in Vision and Ophthalmology Annual Meeting, Sarasota, May 1993. Supported in part by U.S. Public Health Service grants EY07541, EY08137, and EY02377 from the National Eye Institute, National Institutes of Health, Bethesda, Maryland, The authors have a proprietary interest in this technology. Reprint requests to Bahram Khoobehi, PhD, LSU Eye Center, 2020 Gravier St, Suite B, New Orleans, LA 70112-2234.

1716

beled microspheres, 17-20 selective angiography or targeted dye delivery,21-25 and the scanning laser technique.r'v" In laser Doppler velocimetry, the instruments must be carefully calibrated and precisely oriented before use and the data must be extensively analyzed to determine the phase shift of the laser light. As a result of its limitations, this method is capable of measuring blood flow only in major retinal vessels. The laser Doppler flowmeter is an extension of this systerrr'? that quantitates optic nerve head and choroidal blood flow by measuring the flux of erythrocytes in a vascular network embedded in tissue. Flux is defined as the number of cells times velocity. This system is based on the theory that laser light directed into tissue will scatter from stationary tissue and moving erythrocytes; this scattered light directed to a photodetector shows a Doppler

Khoobehi and Peyman . Fluorescent Vesicle System broadening over a spectrum of frequencies. The reason for ha ving a broadening frequencies spectrum is that photons scattered by moving erythrocytes are frequencyshifted in proportion to their velocity, whereas the photons scattered from tissue remain at their original frequency. Doppler-broadening spectrum frequencies can be processed to obtain the erythrocyte flow." The blue-field entoptic technique depends on the subject's ability to indicate that he/she sees his/her own leukocytes moving through the vessels. Hence, this method is capable of measuring blood flow of the macula in humans, but cannot be used experimentally in animal models. In addition, the responses of the retinal macrocirculation and the macular microcirculation to any physiologic or pathologic changes cannot be investigated simultaneously by either laser Doppler velocimetry or the bluefield entoptic technique. With the dye dilution technique, which is an extension of traditional fluorescein angiography, the injected dye bolus must be recorded as it first passes through the retinal vasculature and before all the vessels are infused with the dye. The time course of the dye intensity is measured at two points on a retinal vessel, and the desired data then are calculated from the resulting curves. There are several difficulties with this technique: (I) the angiogram cannot be repeated on the same da y; (2) the background fluorescence of the choroid interferes with accurate densitometric measurements of the dye front ; (3) the systemic injection of free dye and recirculation of the dye result in a poorly defined dye front; and, most importantly, (4) recirculation of the dye begins before the first dye front passage is completed . Thus, when the retinal circulation time is abnormally long, its measurement may be impossible with this procedure. One other method involves intravenous injection of radioactively labeled or unlabeled microspheres, killing the animal, and then counting the number of microspheres in the various retinal and optic nerve head tissues and/or choroidal vessels. From this information, the retinal blood flow velocity can be interpolated. Obviously, this method is unsuitable for human application. The selective angiography method, also called the targeted laser dye delivery system, avoids the problems associated with the dye dilution technique and provides superior visualization of the fundus without choroidal interference. In this system , temperature-sensitive liposomes2 1- 25 containing a highly concentrated solution of fluorescent dye are injected intravenously. The liposomes with encapsulated dye circulating through the retina and choroid are invisible to the fundus camera because the fluorescence of the dye is quenched at the high concentrations encapsulated in the vesicles. When the liposomes are exposed to a heat pulse generated by an electromagnetic energy source such as a laser, the dye inside the liposome absorbs the energy, the vesicle breaks down , and the dye is released into the bloodstream, where it is diluted and becomes intensely fluorescent. This type of targeted system releases dye only in the particular vessel or area of tissue on which the laser beam impinges. Two disadvantages of this method are: (I) it requires encapsulation

of high concentrations of dye into the liposomes, which is technically difficult; and (2) the laser beam used to lyse the heat-sensitive liposomes must pass through the cornea, the anterior chamber, the lens, and the vitreous, resulting in some scattering. The scanning laser technique is another method that allows direct quantitative measurement of erythrocyte flow velocities in perifoveal capilJaries. This method provides superior resolution of retinal vessels and the blood flowing through them, compared with traditional fluorescein angiography and videoangiography techniques. In this procedure, sodium fluorescein is injected into the cubital vein of the patient's arm. The dye then travels to the eye where it is visualized using the scanning laser ophthalmoscope. Current techniques have concentrated on the small capillaries because the uniformity of blood flow allows for greater visualization of individual boluses of dye. Digital imaging programs are used to analyze areas of hyperfluorescence and hypofluorescence to determine relative dye concentrations with in the various vessels of the retina. By tracking a plug ofhyperfluorescence through a particular capillary over several consecutive frames, the blood flow velocity in the capillary beds can be measured directly. Although the scanning laser technique allows for highresolution scanning and anal ysis of the retina, it does have some of the same limiting factors seen in more conventional angiography techniques. Its principal drawback is background interference as a result of choroidal filling. Investigators have found the optimum viewing time to be the first 10 seconds after the dye has reached the retina; after that, dye in the background interferes with visualization of surface vessel flow and results in decreased res0Iution. 26- 28 Background fluorescence also limits the technique to capillary-size vessels; in larger vessels, too much dye is present to allow a plug of dye to be tracked accurately. Another factor impeding visualization is the fluorescence of the leukocytes, platelets, and the plasma within the vessels themselves. Tracking a particular plug of hyperfluorescence is more difficult because it gets lost among all the other fluorescing substances. This problem has led some investigators to track dark spaces in the capillaries.?" However, this approach is beset with the same problems: the dark spots get lost among the areas of hyperfluorescence, making tracking difficult. We describe here a new system for measuring blood flow that overcomes the limitations of current techniques. In our fluorescent vesicle system , fluorescent liposomes, scanning laser ophthalmoscopy, and a digital image analysis system are used to measure blood flow in the retina, optic ner ve head capillaries, and choroid. In addition, unlike most other approaches, our system is capable of measuring flow simultaneously in both the retinal macrocirculation and the microcirculation of the choriocapillaris, macula, and optic nerve head.

Materials and Methods All lipids used in this study were obtained from Avanti Polar Lipids (Pelham, AL). Th e cholesterol was obtained

1717

Ophthalmology

Volume 101, Number 10, October 1994

from Sigma Chemical (St. Louis, MO). Both the lipids and cholesterol were used without further purification. Sephadex LH-20 (lipophilic sephadex) was obtained from Sigma. Carboxyfluorescein was obtained from Molecular Probes (Junction City, OR). All chemicals were reagent grade or better. One cynomolgus monkey, one African green monkey, and six pigmented rabbits were used for this study.

Preparation of Liposomes Oifferent compositions of lipids with and without cholesterol were used. Liposomes were prepared by combining dipalmitoylphosphatidylcholine (OPPC) and dipalmitoylphosphatidylgiycerol (OPPG) in a 4: 1 mol ratio or phosphatidylcholine (PC) and phosphatidylgiycerol (PG), also in a 4: 1 ratio. The PC-PG mixture was stabilized with cholesterol; the lipid:cholesterol ratio was 6:4 mol. In addition, liposomes were prepared from PC, cholesterol (chol) .and distearylphosphatidylethanolamine (OSPE) covalently linked to polyethylene glycol (PEG) (OSPEPEG 2000 ; average molecular weight = 2000 daltons) in a 5:4.5:0.5 mol ratio of PC:chol:OSPE-PEG 2000. Because OPPC is a fully saturated phospholipid that may increase the stability of the preparation, these liposomes also were made with OPPC in place of PC,

The Three Formation Methods Used 1. Large unilamellar vesicles (average diameter, 250 nm) with encapsulated carboxyfluorescein were prepared by the reverse evaporation method;" which we have used in the past. 2 I.22 The method was identical to that previously reported, except for the occasional substitution of the phospholipid OSPE-PEG 2000 formulation. Preparations containing liposomes with a maximum diameter of220 nm were obtained by passing the large unilamellar vesicles through a sterile Millex-G V 0.22-~m filter unit (Millipore, Bedford, MA). 2. Large multilamellar vesicles were prepared by means of a freezing and thawing technique. Briefly, this method consisted of drying the phospholipid under vacuum to remove all traces of organic solvent. Then the dried lipid mixture was rehydrated with the aqueous solution to be encapsulated. The lipid and aqueous solution (carboxyfluorescein) mixture was incubated at 50° C and then mixed for approximately 30 minutes with periodic vortex until a suspension formed. Next, the suspension was frozen in liquid nitrogen and thawed three times. This procedure produced multilamellar vesicles in a broad range of sizes from 10 to 20 nm to 1 to 2 urn in diameter. Finally, dialysis was used to remove unencapsulated dye. 3. To generate "small" liposomes of approximately 100 nm average diameter, we used an extrusion technique that has been shown to generate homogeneous populations of defined vesicle size." Li1718

posomes of a broad range of sizes were filtered ten times through two stacked O.I-~m polycarbonate filters in a high-pressure extrusion device and then dialyzed to remove unencapsulated dye.

In Vitro Visualization of the Fluorescent Vesicle System in a Model Eye

To demonstrate that it is the dye-containing vesicles that are fluorescing, an in vitro model eye was created. The model consisted of capillary tubing, a double aspheric lens, and an aqueous solution continuously flowing through the capillary tubing. The capillary tubing was suspended between two fasteners and placed against a black background. The double aspheric lens was located one focal length from the tubing. The syringe in which the aqueous solution was placed was connected in series with a Sage s-ring pump (Sage Instruments, Orion Research, Cambridge, MA) and the tubing. Vesicles of different diameters were encapsulated with carboxyfluorescein as described above and added to the aqueous solution; both a buffer solution and whole heparinized blood were used as the aqueous carrier solutions. The circulating vesicles were visualized with a Rodenstock (Oanbury, CT) scanning laser ophthalmoscope aimed through the double aspheric lens.

In Vivo Visualization of the Fluorescent Vesicle System in Monkey and Rabbit Eyes

Before the in vivo experiments, monkeys were anesthetized using an intramuscular injection of ketamine (30 rug/kg) mixed with xylazine (3 rug/kg) ; rabbits were anesthetized using an intramuscular injection ofketamine (50 mg/kg) and xylazine (5 mg/kg), The eyes were dilated with 1.5% phenylephrine HCl and 0.5% tropicamide. The suspension of fluorescent vesicles was injected into the saphenous vein of the monkey or an ear vein of the pigmented rabbit. The suspension consisted of liposomes of various diameters in a dose of 0.03 ml/kg, corresponding to a carboxyfluorescein dose of approximately 70 ~g/kg and a lipid dose of I mg/kg. For the in vivo studies of blood flow, large unilamellar vesicles of various sizes were filtered as described in the Materials and Methods section, above, so that the largest vesicles visualized were 220 nm in diameter. The concentration of the dye encapsulated in the liposome was chosen to yield the most efficient liposome fluorescence in the bloodstream. Immediately after the liposome suspension was injected, the fundus was viewed through the scanning laser ophthalmoscope, and images of the fundus with liposomes circulating in all vessels were captured on videotape. The scanning laser ophthalmoscope was used to obtain video images of the retina in the form of a National Television Standards Committee-type video signal which was recorded on a video cassette recorder or digitized by a realtime video digitizer. The video signal was sampled at 30 frames per second. Every frame was digitized into a 512 X 512-bit mapped image and stored on the digitizer's

Khoobehi and Peyman . Fluorescent Vesicle System memory. The image sequences were either accessed directly from the digitizer's memory by a computer analysis program or were transferred to permanent storage devices, such as a computer hard disk, digital tape, or CD-ROM, for future analysis and reference. The image analysis system (HRX Digital Imaging Processor, Amtronics, New Orleans, LA), which is capable ofdigitizing and storing digital sequences up to 8 seconds long in realtime, was used to digitize the captured video images. Under the control of a personal computer, the HRX processor was able to capture, digitize, and store up to 256 frames of live video (512 X 512 X 8-image format for every frame) while displaying the digital video on a television monitor. After the images were stored in the HRX memory, they were transferred to the computer's hard disk for permanent storage. Development of Registration Technique to Compensate for Shifts in Video Frames Caused by Eye Movements Because it is sometimes necessary to capture up to 180 consecutive frames, which means a recording time of 6 seconds, absolute immobility of the eye cannot always be maintained. The possibility of translational (x.y), rotational, and scaling shifts from one frame to the next, caused by movement of the eye, had to be taken into account. To obtain images in perfect spatial register'? (i.e., with the frames aligned such that a given feature of the fundus corresponds to the same pixel coordinate in all the frames), we developed a registration technique to correct for positional shifting superimposed on the time-related changes of successive images. The software program we developed to correct for differences between a pair of images performed the following analysis: 1. A subarea of the image such as an area around the optic nerve was selected; the first image was used as a reference image, and the second image as a template. 2. A rectangular area of the template was compared to a rectangle of the same size in the search area by calculating the cross-correlation of the two rectangles. The normalized cross-correlation function is a measure of similarity of these two rectangular areas. 3. The template then was moved one pixel at a time over the search area and the cross-correlation value was obtained at each location. The location of maximum correlation between the template and the reference image was considered to provide the amount of translational shift between the two images. 4. The above process was repeated for each successive pair of frames (2 and 3, 3 and 4, and so on). Once all movement parameters were obtained, the images were transformed, resulting in a sequence free from movement.

In all of our in vivo studies of the eyes of anesthetized animals, the observed deformations from frame to frame have been of the translational type. Although small « ±IOO) rotational shifts could potentially occur, none was seen in these experiments. Because all of the images were taken during the same session and the distance from the eye to the camera was held constant, scaling differences were the least likely and were not detected. Measurement of Blood Flow Velocity To determine blood flow velocity, we developed a software program that overlaid multiple video frames on a single image to produce a visible record of the path taken by a particular vesicle in a given time period. The path was used to measure the velocity of the vesicle, and hence the velocity of the blood flow in the vessel. For every location on the image, the software program compared the pixel intensities ofall the frames to be overlaid and selected the maximum (brightest) value as the pixel value of the final image. Because the luminosity of the vesicle was much higher than that ofany other feature of the retina (at times, as much as twice the pixel intensity of the other pixels), any vesicle that appeared in a frame was visible on the overlaid image. The result was an image that displayed the frame-by-frame movement of the vesicle. The diameter of the vessels was measured by taking a red-free video of the fundus using the scanning laser ophthalmoscope in the green mode only. To calculate the vessel diameter, five measurements were made and averaged to determine the vessel diameter at the single point where the liposome was located at that particular moment. For distance measurements, at least ten liposomes at one location were traced. The results then were averaged, and the velocity of the liposomes was obtained by dividing the average by the time period of vesicle movement (33 mseconds for a successive pair of video frames). For major arteries and veins, the velocities ofvesicles were measured at the center of the vessels (maximum centerline velocity). However, for capillary blood flow velocity, this consideration was unnecessary.

Results In Vitro Model The in vitro model eye, when scanned by the scanning laser ophthalmoscope, showed fluorescent vesicles flowing through the tubing in both the buffer and the heparinized blood solutions. Figure 1 shows filtered large unilamellar vesicles with a maximum diameter of 220 nm traveling through l80-jIm diameter capillary tubing. The optimal concentration of encapsulated carboxyfluorescein was found to be 20 mM. Vesicles of uniform 100-nm size or smaller could be visualized. Different concentrations of liposome suspension were tested; dilutions as low as 1/ 25,000 (calculated totalliposome volume/buffer or blood volume) were video recorded.

1719

Ophthalmology

Volume 101, Number 10, October 1994

Figure 1. Liposorne s with a maximum diameter of 220 /lm d iluted 1: 25.000 in who le blood are seen moving from left to right th rough capillary tubing. The inner d iameter of th e tubing is 280 /l m . The brightest liposom es (large arro ws) are tra veling closest to th e viewe r.

Neither the buffer nor the heparinized blood fluoresced when exposed to the scanning laser ophthalmoscope in the absence of fluorescent vesicles. It has been postulated by others that the spots visualized by this fluorescent vesicle system in vivo are actually the result of either staining of leukocytes that have phagocytosed the liposomes or possibly an immunologic reaction involving aggregation or complement activation. This theory has been conclusively put to rest by the in vitro model , in that fluorescent vesicles were clearly visible in the circulating buffer solution although this solution was entirely devoid of leukocytes, immunoreactant particles, and/or other cell-mediated intermediates of the immune response.

Characteristics of Liposome Activity in Vivo Liposomes with different phospholipid compositions and different survival times were recorded in vivo in the rabbit and monkey. When high concentrations ofliposomes were injected and remained in circulation for long periods of time , their movement through the vessels in the fundu s resembled the lights of heavy traffic at night on a fastmoving freeway during rush hour. When low concentrations of liposomes were injected, the vesicles traveling through the fundus looked like falling stars against a dark sky. In the retina as in the in vitro model, individual liposomes as small as 100 nm were visible. Measurement of blood flow velocity was not restricted to vessels of any particular size in the retina or optic nerve head; the fluorescent vesicles were visible in major vessels as well as in small capillaries, including the macula and optic nerve head , allowing objective simultaneous measurement of blood flow velocity in the entire retinal macrocirculation, macular microcirculation, opt ic nerve head capillaries, and choroidal circulation. For indi vidual liposomes to be visible in a vessel, the concentration of the encapsulated fluorescent dye had to be low enough to allow the dye to fluoresce within the vesicles. Too high a concentration of encapsulated dye resulted in quenching of the fluorescence. Again, in the retina as in the in vitro model , the optimal concentration of carboxyfluorescein in the liposomes was found to be 20 mM . Thi s value was determined from pixel intensity corresponding to the liposome in the registered, digitized frame. Liposomes of three different sizes showed differences in circulation times and potential for adherence to the

1720

vessel walls. The multi lamellar vesicles, which appeared large (range, 1-2 JIm), cleared from the circulation very quickly, with a maximum half-life of 20 minutes. Occasionall y, these large vesicles appeared to stick to vessel walls, more often in choroidal vessels than in retinal vessels. Smaller liposomes made by either the reverse-evaporation technique (large unilamellar vesicles with an average diameter of 250 nm) or the extrusion technique (range of sizes, 10-100 nm) had a longer half-life, approximately 90 minutes, and showed no stickiness ifused with in a few weeks of preparation. When suspensions of smaller vesicles made of DPPC and DPPG or PC, PG , and cholesterol and refrigerated at 4 0 C for several months were used in pigmented rabbits, some of the vesicles adh ered to the walls of the choroidal vessels, although they eventually detached and re-entered the circulation. Because the diameter of the liposomes was approximately 50 times smaller than that of the smallest capillary, it was unlikel y that these vesicles could cause a blockage. However, to prevent the vessel wall adherence seen in these stored batches of liposomes, we modified the liposome formulation using the method of Allen et a1 33,35,36 and Gabizon and Papahadjopoulos," which involved the inclusion of DSPE-PEG 2000. The new fluorescent vesicles showed no adherence even when they were stored for 3 months before being injected into rabbits and monkeys, The vesicles made with DSPE-PEG 2000 could be visualized for several hours after injection into the animal host; individual vesicleswere visible for as long as 4 hours, However, the concentration of fluorescent vesicles did not remain constant; immediately after injection, thousands of fluorescent vesicles were present within each vessel, whereas after 4 hours onl y small groupings of vesicles could still be seen.

Quantitative Analysis of the Hemodynamics of the Major Retinal Vessels in the Monkey Using Fluorescent Vesicles With the computer program we developed, multiple video frames were overlaid on a single image, making visible the path that a particular vesicle traveled over a given period of time. The temporal resolution of the system was 33 mseconds. From these images, blood flow velocities in the retinal vessels ofa monkey eye were calculated . Two ofthe retinal veins, designated A and B, had diameters of 138 ± 8 and 95 ± 7 JIm, respectively (mean ± standard deviation). In these two vessels, the velocity of the liposomes was measured at the center ofthe vessel. For vesselA, the following velocities were obtained: 10.21, 11.97, 9.41, 7.74, 6.38, 9.57, and 10.05 mm/second; the mean velocity (± standard deviation) was 9.33 ± 1.67 rum/second. In vessel B, the following velocities were obtained: 10.05, 7.82, 6.91, 6.62, and 5.89 mm/second; the mean velocity (± standard deviation) was 7.64 ± 1.44 rum/second. Retinal vein C had a diameter of 130 ± 9 JIm. Liposome velocity within the vein was calculated for vesicles moving at various distances from the vesselwall (Table 1). Plotting

Khoobehi and Peyman

Table 1. Velocity of Blood in a Central Retinal Vein of a Primate Distance from Vessel Wall" (ILm)

Velocity (mm/sec)

20 30

2.0 4.2

40

5.5

55 58 62 65

6.8

7.2 8.0

9.8

• Vessel diameter = 130 ± 9 urn.

the velocity of the liposome versus the distance from the vessel wall showed that velocity increases with increasing distance from the vessel wall, and that the highest velocity is attained in the middle of the vessel at the maximal distance from the walls (Fig 2). These findings are consistent with the concept oflaminar flow, in which the flow within a vessel is characterized by a parabolic front, with the leading edge situated in the middle of the channel of flow. Thus, centerline velocity must be calculated from the path of a vesicle traveling in the center of a vessel. The retinal artery had a diameter of 64 ± 7 /-Lm. The following center-of-vessel liposome velocities were observed: 11.65, 16.12,27.45, 17.55, 13.56, and 10.21 mm/ second. The average velocity of fluid through the artery was calculated as 16.10 ± 5.7 mm/second. However, in this context, average velocity has no real meaning when one considers the range from a low of 10.21 rum/second to a high of 27.45 mm/second, representing almost a threefold difference. This variation is likely caused by pulsatile flow differences in the artery arising from the relatively high speed immediately after ventricular systole to the relatively slow speed of ventricular diastole. To test this hypothesis, we would have to repeat the experiment with a simultaneous electrocardiogram recording. The data from veins A and B are more uniform; the absence of a large variation in velocity values, as was seen in the artery, is the result of the increased compliance of the veins, which absorb the increased pressure of systole by expanding their walls. Thus, in the venous system, velocity is averaged over the cardiac cycle. The velocity in the larger diameter vein is greater than that in the smaller diameter vein. This finding is to be expected because resistance decreases as diameter increases, leading to increased velocity. Figures 3 and 4 show two typical composite images that were constructed by overlaying several consecutive video frames, 33 mseconds apart. In Figure 3, a vesicle was followed as it traveled through an artery. Because eight frames correspond to 266 mseconds of video time, this single image displays the path that this particular vesicle traveled in 266 mseconds.

Fluorescent Vesicle System In Figure 4, a vesiclewas followed as it traveled through an artery, entered and exited a capillary to an interconnecting vein, and finally exited the retina through a major vein. This single image represents 1600 mseconds of video and was constructed by overlaying 48 digitized frames. The vesicle was visible in the artery in 8 frames (266 mseconds), in the capillary in 20 frames (666 mseconds), in the interconnecting vein in 6 frames (200 mseconds), and in the major vein in 14 frames (466 mseconds). Note that the intervals between successive positions of the vesicle are unequal in the artery, but uniform in the major vein. The differences in the artery are the result of pulsatile flow caused by the heartbeat. Note also that the intervals in the major vein are very small; this is a result of the position of the vesicle near the wall of the vein, where blood flow velocity is relatively slow. Vesicles traveling in

10

-

8

0

Q)

C/)

---E

-E

-

6

~

°

0 0

Q)

>

4

2

20

30

40

50

60

Distance (urn) Figure 2. Plotting the velocity of the vesicle against the distance of the moving vesicle from the vessel wall demonstrates the laminar nature of the flow.

1721

Ophthalmology

Volume 101, Number 10, October 1994

Figure 3. A single vesicle is fol1owed through a monkey artery (A). This image, obtained by overlaying eight video frames, highlights the path (arrow) that this particular vesicle traveled in 266 mseconds.

the center of the major vein would show greater spacing, indicating higher velocity blood flow.

Measurement of Macular and Optic Nerve Head Blood Flow Velocity in the Monkey with the Fluorescent Vesic1e System Using the scanning laser ophthalmoscope and the fluorescent vesicle system in monkeys, we were able to measure the capillary blood flow velocity simultaneously in the macular and optic nerve head regions. The experiment was performed on one cynomolgus monkey and repeated on one African green monkey. The fluorescent vesicle suspension was administered intravenously to the anesthetized monkeys. The remainder of the process was performed as described above, except that the system was focused on capillaries. The angiograms were videotaped and digitized for analysis. In the cynomolgus monkey, measurements of the macular capillaries showed the following velocities (mean ± standard deviation): 0.77 ± 0.07, 0.88 ± 0.1, 0.76 ± 0.09, 1.33 ± 0.1, 0.58 ± 0.1, 0.62 ± 0.08, 0.72 ± 0.06, 0.73 ± 0.1, 0.83 ± 0.05, 0.80 ± 0.06,0.45 ± 0.12, 0.56

Figure 4. A vesicle is seen traveling through a monkey artery (A), entering and exiting a capillary (C) to an interconnecting vein, and exiting the retina through a major vein (V). The direction of vesicle movement is shown with arrows. The image represents 1600 mseconds of video and was constructed by overlaying 48 digitized frames.

1722

Figure 5. Two vesicles (thick arrow) are fol1owed as they travel in an arteriole (A), move through a prefoveal capil1ary, and exit through a venule (V). The image was constructed by overlaying 44 video frames, corresponding to 1.47 seconds of video time. Transit time in the prefoveal capillary was 20 frames or 667 mseconds. The paired arrows indicate the successive positions of the leading vesicle (a) and the trailing vesicle (b). As the two vesicles enter the capillary, the trailing vesicle can be seen to catch up with the leading vesicle and the space between them nearly disappears. As the leading vesicle enters the vein, it accelerates and again there is visible space between the two vesicles. The apparent doubling of individual vesicles (large arrow) is an artifact (see the Discussion section for additional details). F = fovea.

± 0.11, 0.93 ± 0.09, and 0.74 ± 0.1 mm/second. The average macular capillary velocity was 0.76 mm/second (range, 0.45-1.33 rnm/second). In Figures 5 and 6, two vesicles are shown moving through macular capillaries in the monkey retina. The

Figure 6. Two vesicles are seen moving (single arrows) in an arteriole (A), then through a prefoveal capillary, and exiting through a venule (V). The image was constructed by overlaying 49 video frames, corresponding to 1.6 seconds of video time. Transit time in the capillary was 21 frames or 700 mseconds. The paired arrows indicate the successive positions of the leading vesicle (a)and the trailing vesicle (b). Notice that the separation between the two vesicles is greater in the artery than in the vein, because velocity is higher in the artery. F = fovea.

Khoobehi and Peyman . Fluorescent Vesicle System image in Figure 5 was constructed by overlaying 44 video frames, corresponding to 1.47 seconds of video time. In each frame, two vesicles were visible moving in an arteriole, then through a prefoveal capillary, and exiting through a venule. Transit time in the prefoveal capillary was 20 frames or 667 mseconds. As the vesicles entered the capillary, the trailing vesicle caught up with the leading vesicle and the space between them nearly disappeared. As the leading vesicle entered the vein, it accelerated and again there was visible space between the two vesicles. In Figure 6, the image was constructed by overlaying 49 video frames, corresponding to 1.6 seconds of video time. In each frame, two vesicles were visible moving in an arteriole, then through a prefoveal capillary, and exiting through a venule. Transit time in the prefoveal capillary was 21 frames or 700 mseconds. Note that, again, the separation between the vesicles is greater in the artery than in the vein, because velocity is higher in the artery. In the African green monkey, the following velocities were measured in the prefoveal capillaries (mean ± standard deviation, in rum/second): 0.4 ± 0.1, 0.93 ± 0.09, 0.95 ± 0.11, 0.89 ± 0.08, and 0.91 ± 0.09. The average was 0.82 ± 0.16 mm/second, The average velocity in the macular capillaries in this animal was 0.7 ± 0.38 mm/ second. Comparing individual pairs of vessels (one prefoveal capillary, the other a premacular capillary, branched from the same arteriole) showed that the velocity in the premacular capillary was significantly slower than the velocity in the prefoveal capillary (P < 0.001). Measurements at the optic nerve head in the cynomolgus monkey yielded the following velocities (mean ± standard deviation, in mm/second) for efferent arterioles: 1.5 ± 0.2, 1.36 ± 0.21, 1.8 ± 0.23, 1.79 ± 0.28, and 2.25 ± 0.3 and the following velocities for efferent venules: 0.99 ± 0.13, 0.98 ± 0.09, 0.96 ± 0.1, 1.09 ± 0.16, 1.21 ± 0.1. The average velocity in the optic nerve head capillaries (both arterioles and venules) was 1.39 mm/second, with a range from 0.96 to 2.25 mm/second. Figure 7 shows an overlay representation of vesicles as they move through an optic nerve head capillary at various points in time. The image was constructed by overlaying 21 digitized video frames, corresponding to 700 mseconds of video time. On the left side of the photograph, a pair of vesicles is visible moving in a precapillary arteriole, then through a capillary, and exiting the capillary to a postcapillary venule. The precapillary arteriole is barely visible in the original video; the capillary itself is not at all visible. Only the path of the vesicle movement is visible in the constructed image. Transit time was 3 frames (100 mseconds) in the precapillary arteriole, 16 frames (533 mseconds) in the capillary, and 2 frames (66 mseconds) in the postcapillary venule. On the right side of the photograph, another vesicle is shown moving through the capillary and exiting through a postcapillary venule.

Discussion In early studies,">" we used highly concentrated fluorescent dye encapsulated into heat-sensitive liposomes which

Figure 7. An overlay representation of vesicles as they move through an optic nerve head capillary at various points in time. On the left side of the photograph, a vesicle is visible moving in a precapillary arteriole (A) and a capillary (B)and exiting the capillary to a postcapillary venule. Only the path of the vesicle movement is visible in the constructed image. The direction of movement is indicated by arrows. On the right side of the photograph, a second vesicle is shown moving from a vessel (C) through the capillary (D) and exiting through a postcapillary venule.

were injected intravenously. When these liposomes were exposed to the heat pulse generated by a laser, hemoglobin and the highly concentrated dye inside the liposome absorbed the heat pulse, causing the dye to be released into the bloodstream, where it was diluted and fluoresced. Then we traced the released dye front to measure the blood flow velocity.P The advantages of this method are (I) direct retinal circulation time measurements can be obtained; (2) any vascular defects can be observed easily; (3) several repeated readings can be made from one injection with no harm to the test animal; and, perhaps most importantly, (4) the macular microcirculation and retinal macrocirculation can be investigated simultaneously with the targeted dye delivery system; therefore, changes in both circulations in response to a treatment regimen can be detected simultaneously. There are several disadvantages to this technique. I. To achieve a sharp dye front by lysing temperaturesensitive liposomes with a laser heat pulse, we have to encapsulate a high concentration of dye. However, the amount of the dye encapsulated into the liposomes is limited by the fact that the osmolarity (osmotic pressure) of dye cannot exceed 250 to 300 mOsm or the contents ofthe liposome will leak out. Encapsulation of up to 100 mM is theoretically possible; practically, however, the limit is closer to 80 mM (due to leakage during the period ofstorage), which is equivalent to 280 mOsm. This corresponds to approximately 28 mg/ml of dye. To achieve the traditional angiogram concentration (7 mg/kg, used in humans) of dye in living primates weighing 5 to 6 kg on average, an injection of (5 kg) X (7 mg/kg) X (1/28 ml/rng) = 1.25 ml would be necessary. Because only one fourth of the total volume of the li1723

Ophthalmology

Volume 101, Number 10, October 1994

posome suspension is actually vesicles, the volume of the injection would have to be 1.25 ml X 4 = 5 ml. Although we have achieved satisfactory results with one half that amount (approximately 2 ml) in nonhuman primates, the amount needed for an average person weighing 60 kg would still be 60 kg X 2 ml/ 7 kg = 17 ml; this is approximately 0.5 g lipid. With our fluorescent vesicle system, only one tenth of that amount is needed. In fact, the fluorescent vesicle system sharply reduces the amount of dye and lipid that must be injected compared with both conventional fluorescein angiography, as well as the targeted dye delivery system. With our fluorescent vesicle technique, we only used 0.1 to 0.2 ml of the liposome suspension in monkeys weighing 7 to 10 kg in most of the experiments; this represents 0.5 to 1 mg/kg of lipids and 10 to 20 ~g/kg of dye. 2. To release enough dye from the liposomes, a spot approximately 0.6 mm in diameter must be heated. In other words, the heat-sensitive liposomes require a "heated path" at least 0.6 mm long for sufficient exposure to release the dye efficiently. In a typical experiment, repetitive delivery of 50 to 60 laser pulses resulted in visible damage to the pigment epithelium and choroid. The consequences of such a mild injury outside the macular area could be debated; however, it might have clinical importance when the lesion is near the fovea. The scanning laser technique currently is used clinically to perform video fluorescein angiography of the retina. This method allows continuous recording of high-resolution retinal pictures. The optical quality of the refractive media of the eye and laser spot size (scan) are the only two factors that limit the spatial resolution of the retinal images. In practice, this resolution is approximately 10 to 15 ~m.26-28 However, with our system, we are able to visualize circulating vesicles that are 100 times smaller than the resolution of the scanning laser ophthalmoscope; this is possible because the photons from fluorescing vesicles reach the detector. One artifact produced by this apparent increase in resolution of the scanning laser ophthalmoscope is the phenomenon of liposome pairs. In Figures I and 3 to 6, the fluorescent vesicles are seen as closely paired bright spots, one on top of the other. While it is possible that the fluorescent vesicles may occasionally travel in pairs, the consistency of this finding suggests otherwise. By its very nature, the scanning laser ophthalmoscope creates an image by scanning horizontally across the field. When the beam encounters an area ofhyperfluorescence, it is recorded as a dot. After resetting and rescanning the field, the same fluorescent vesicles again will be recorded if the area of hyperfluorescence on the retina is larger than the distance between consecutive scan lines. If this is the case, two dots will appear, one on top of the other, as is evident in Figures I and 3 to 6. This is additional evidence that the fluorescent vesicle system increases the resolving capacity of the scanning laser ophthalmoscope. Two early considerations involving the fluorescent vesicle system were the circulation life of the fluorescent

1724

vesicles and the tendency of some vesicles to adhere to the vessel wall before re-entering the circulation. Numerous combinations oflipids and cholesterol were tested, as detailed in the Results section. For any given lipid and cholesterol combination, circulation life and adherencetendency were dependent on liposome size, which was controlled by the extrusion plates. Smaller liposomes circulated for longer times and were less adherent, compared with larger liposomes of the same composition. In fact, there was a continuum from the large multilamellar vesicles (the largest, most adherent, and shortest-lived) to the large unilamellar vesicles to the 100-nm fluorescent vesicles. The 100-nm fluorescent vesicles showed virtually no adherence, regardless of the liposome composition. One compositional factor that influenced both circulation time and adherence was the presence of polyethylene glycol. When polyethylene glycol was added to any existing lipid composition, adherence virtually was eliminated and circulation time was increased. Circulation times of 4 hours were achieved using 100-nm fluorescent vesicles containing polyethylene glycol. These results confirm other reports of polyethylene glycol decreasing adherence and increasing circulation times in vitro. The fluorescent vesicle technique allows the simultaneous visualization and quantification of the retinal macrocirculation, macular, and optic nerve head microcirculation. This study represents the first time that optic nerve head capillary velocities have been directly and objectively studied. Values obtained with our fluorescent vesicle technique in the macular capillaries of cynomolgus monkeys are comparable to those obtained by Riva and Petrig" using the blue-field entoptic effect in humans. Recently, Wolf et al,27 using traditional scanning laser ophthalmoscopy on humans, reported average velocity values in the macular capillaries of 3.2 mm/second, The difference could be the result of interspecies differences (human versus cynomolgus monkey) and/or because of the anesthetics used to sedate the monkeys for the study. Additional studies are needed to determine the accuracy and reproducibility of these measurements. As described above, the fluorescent vesicle technique is capable of recognizing pulsatile flow in the retinal vessels. This capability is most easily appreciated by overlaying consecutive photographs ofa single vesicle at equal time intervals as it passes through a vessel. If the diameter of the vessel remained unchanged and flow were nonpulsatile, one would expect to see an equal distance between vesicles for each consecutive frame. Pulsatile flow, iffollowed for a long enough period of time (several seconds), will show a repeating pattern oflonger and longer distances between vesicles, followed by a period ofshorter and shorter distances between vesicles. We have seen distinctly pulsatile flow in the retinal arteries, but not in the retinal veins. In Figures 5 and 6, however, it appears that there is a slight discrepancy in the spacing between consecutive vesicles, suggesting the possibility that there may be pulsatile flow in the macular and optic nerve head capillaries. Interestingly, the standard deviation of velocity in veins and capillaries measured by the fluorescent vesicle system

Khoobehi and Peyman . Fluorescent Vesicle System was small. The increase in the standard deviation of velocity in the artery partially is related to the pulsatile nature of the flow in the arteries. The flow of vesicles in a large-caliber tube produces different velocities, depending on the location of the vesicles (Fig 2). With the technology used in this study, it is possible to locate vesicles that are moving in the center of a large vein or artery in two dimensions; however, determining the vesicle's three-dimensional location is difficult. Therefore, it is advantageous to calculate the mean velocity of the vesicles in the major arteries and veins rather than a maximum centerline velocity. To calculate mean velocity of the vesicles in a major vessel, one measures the velocity of a large number of vesicles and averages them to determine the mean vesicle velocity in this vessel. Weare in the process of refining a technique using the HRX digital imaging processor which will take 8 seconds (240 consecutive frames) of recorded video from a single vessel and measure the velocity (at the same point in the cardiac cycle) of every fluorescent vesicle that passes through the vessel; in a typical retinal vessel with the fluorescent vesicle concentration used, this can result in a database of between 100 and 1000 velocities. An average velocity of fluorescent vesicles within the vessel will be calculated, providing a reasonable estimate of the mean velocity of flow through the vessel. Capillaries, having a smaller caliber, did not show a significant difference in velocity between vesicles traveling closer to or farther from the walls of these vessels. In summary, the fluorescent vesicle system has most of the characteristics of an ideal flow measurement method. The ideal method should be able to: 1. Measure the blood flow in all areas of the retina. The fluorescent vesicle system is not limited even by the resolution of the scanning laser ophthalmoscope. The vesicle can be traced, and the exact transit time can be calculated in any size capillary. 2. Measure retinal, macular, and optic nerve head circulation simultaneously. Circulation in the macula is significantly different from that of the rest of the retina, as well as that of the optic nerve head. Objective and simultaneous measurement of the response of the retinal macrocirculation and the macular and optic nerve head microcirculation to change is not possible by any other technique. With the fluorescent vesicle system, we can simultaneously demonstrate changes in these three circulations in response to any physiologic or pharmacologic stimuli. It is also possible to measure choroidal circulation at the same time by introducing vesicles containing indocyanine-green. We have successfully encapsulated indocyanine-green and visualized these fluorescent vesicles in the pigmented rabbit and primate models using the scanning laser ophthalmoscope in the infrared mode; a separate study is currently being prepared to present the data. 3. Measure blood flow at various intervals during the cardiac cycle. The fluorescent vesicles have been used so far to measure venous flow during all phases

of the cardiac cycle. Additionally, we have noted that the system also appears to produce information about pulsatile flow as manifested by the unequal spacing of the vesicles on the tracking figures. For further investigation of this phenomenon, we propose to digitize an analog output of the electrocardiogram and record it with the scanning laser ophthalmoscope signal output to correlate the blood flow velocity with the pulsations in the arteries. 4. Measure flow without observer or subject bias. With the blue-field entoptic method, the subject must be able to match his/her flow to the patterns on a computer screen. The subject may introduce some error by reacting to the pattern slightly differently. In the targeted dye delivery system,21-25 the operator need only aim the beam correctly and expose the blood to the laser energy. However, with the fluorescent vesicle system, there is no operator involved to make errors, except in calculations. Thus, with the use of this fluorescent vesicle system, we can measure volumetric blood flow in all arteries and veins in the retina, and blood flow velocity in all capillaries in the retina, specifically, in the macula, the optic nerve head, and the choriocapillaries. The pathophysiologies of all these circulations are different, but with this system, any alterations induced in any or all of these circulations by a pharmacologic agent can be monitored simultaneously.

References I. Bulpitt CJ, Oollery CT. Estimation of retinal blood flow by measurement ofthe mean circulation time. Cardiovasc Res 1971;5:406-12. 2. Hill OW, Griffiths JO, Young S. Retinal blood flow measured by fluorescence angiography. Trans Ophthalmol Soc U K 1973;93:325-32. 3. Riva CE, Ben-Sira I. A two-point fluorophotometer for the human ocular fundus. ApplOptics 1975;14:2691-3. 4. Hill OW, Young S. Arterial inflow studies of the cat retina using high-speed cine angiography. Exp Eye Res 1976;23: 35-45. 5. Riva CE, Feke GT, Ben-Sira I. Fluorescein dye-dilution technique and retinal circulation. Am J Physiol 1978;234: H315-22. 6. Preussner PR, Richard G, Oarrelmann 0, et al. Quantitative measurement of retinal blood flow in human beings by application of digital image-processing methods to television fluorescein angiograms. Graefes Arch Clin Exp Ophthalmol 1983;221: 110-12. 7. Favilla I, Barry WR, Turner 11. Video and digital fluorescein angiography. Aust N Z J Ophthalmol 1986;14:229-34. 8. Riva CE, Grunwald JE, Sinclair SH, O'Keefe K. Fundus camera-based retinal LOV. Appl Optics 1981;20:117-20. 9. Hill OW, Pike ER, Gardner K. Laser doppler velocimetry of the retinal blood flow. Trans Ophthalmol Soc U K 1981;101:152-5. 10. Green GJ, Feke GT, Goger OG, McMeel JW. Clinical application of the laser Doppler technique for retinal blood flow studies. Arch Ophthalmol 1983;10 I:971-4.

1725

Ophthalmology

Volume 101, Number 10, October 1994

II. Sebag J, Delori FC, Feke GT, et al. Anterior optic nerve blood flow decreases in clinical neurogenic optic atrophy. Ophthalmology 1986;93:858-65. 12. Feke GT, Goger DG, Tagawa H, Delori Fe. Laser Doppler technique for absolute measurement of blood speed in retinal vessels. IEEE Trans Biomed Eng 1987;34:673-80. 13. Feke GT, Tagawa H, Deupree DM, et al. Blood flow in the normal human retina. Invest Ophthalmol Vis Sci 1989;30: 58-65. 14. Sebag J, Delori FC, Feke GT, Weiter JJ. Effects of optic atrophy on retinal blood flow and oxygen saturation in humans. Arch Ophthalmol 1989; 107:222-6. 15. Riva CE, Petrig B. Blue field entoptic phenomenon and blood velocity in the retinal capillaries. J Opt Soc Am 1980;70: 1234-8. 16. Fallon TJ, Maxwell DL, Kohner EM. Autoregulation of retinal blood flow in diabetic retinopathy measured by the blue-light entoptic technique. Ophthalmology 1987;94: 1410-15. 17. Rosen DA, Marshall J, Kohner EM, et al. Experimental retinal branch vein occlusion in rhesus monkeys. II. Retinal blood flow studies. Br J Ophthalmol 1979;63:388-92. 18. Aim A, Bill A. The oxygen supply to the retina. I. Effects of changes in intraocular and arterial blood pressures, and arterial Paz and PC02 on the oxygen tension in the vitreous body of the cat. Acta Physiol Scand 1972;84:261-74. 19. Sperber GO, Bill A. Blood flow and glucose consumption in the optic nerve, retina and brain: effects of high intraocular pressure. Exp Eye Res 1985;41:639-53. 20. Small KW, Stefansson E, Hatchell DL. Retinal blood flow in normal and diabetic dogs. Invest Ophthalmol Vis Sci 1987;28:672-5. 21. Khoobehi B, Niesman MR, Peyman GA, Oncel M. Repetitive, selective angiography of individual vessels of the retina. Retina 1989;9:87-96. 22. Khoobehi B, Peyman GA, Niesman MR, Oncel M. Measurement of retinal blood velocity and flow rate in primates using a liposome-dye system. Ophthalmology 1989;96:90512. 23. Zeimer RC, Khoobehi B, Peyman GA, et al. Feasibility of blood flow measurement by externally controlled dye delivery. Invest Ophthalmol Vis Sci 1989;30:660-7.

1726

24. Khoobehi B, Schuele KM, Ali OM, Peyman GA. Measurement of circulation time in the retinal vasculature using selective angiography. Ophthalmology 1990;97:1061-70. 25. Khoobehi B, Char CA, Peyman GA, Schuele KM. Study of the mechanisms oflaser-induced release ofliposome-encapsulated dye. Lasers Surg Med 1990;10:303-9. 26. Tanaka T, Muraoka K, Shimizu K. Fluorescein fundus angiography with scanning laser ophthalmoscope. Visibility of leukocytes and platelets in perifoveal capillaries. Ophthalmology 1991;98: 1824-9. 27. Wolf S, Arend 0, Toonen H, et al. Retinal capillary blood flow measurement with a scanning laser ophthalmoscope: Preliminary results. Ophthalmology 1991;98:996-1000. 28. Bertram B, Wolf S, Fiehofer S, et al. Retinal circulation times in diabetes mellitus type 1. Br J Ophthalmol1991;7 5: 462-5. 29. Riva CE, Harino S, Petrig BL, Shonat RD. Laser Doppler flowmetry in the optic nerve. Exp Eye Res 1992;55:499506. 30. Szoka F Jr, Papahadjopoulos D. Comparative properties and methods of preparation of lipid vesicles (liposomes). Annu Rev Biophys Bioeng 1980;9:467-508. 31. Hope MJ, Bally MB, Webb G, Cullis PRo Production of large unilamellar vesicles by a rapid extrusion procedure. Characterization of size distribution, trapped volume and ability to maintain a membrane potential. Biochim Biophys Acta 1985;812:55-65. 32. Pratt WK. Digital Image Processing, 2nd ed, New York: John Wiley & Sons, 1991;662-5. 33. Allen TM, Chonn A. Large unilamellar liposomes with low uptake into the reticuloendothelial system. FEBS Lett 1987;223:42-6. 34. Gabizon A, Papahadjopoulos D. Liposome formulations with prolonged circulation time in blood and enhanced uptake by tumors. Proc Nat! Acad Sci USA 1988;85:694953. 35. Allen TM, Hansen C. Pharmacokinetics of stealth versus conventional liposomes: effect of dose. Biochim Biophys Acta 1991; 1068:133-41. 36. Allen TM, Hansen C, Martin F, et al. Liposomes containing synthetic lipid derivatives of poly(ethylene glycol) show prolonged circulation half-lives in vivo. Biochim Biophys Acta 1991;1066:29-36.