Blood-Retinal Barrier Breakdown Investigated by Real-Time Magnetic Resonance Imaging After Gadolinium-Diethylenetriaminepentaacetic Acid Injection

Transcription

1 Investigative Ophthalmology & Visual Science, Vol. 32, No. 11, October 1991 Copyright Association for Research in Vision and Ophthalmology Blood-Retinal Barrier Breakdown Investigated by Real-Time Magnetic Resonance Imaging After Gadolinium-Diethylenetriaminepentaacetic Acid Injection Bruce A. Berkowirz,* Yukihiro Sato,f Charles A. Wilson,f and Eugene de Juan, Jrf Recent magnetic resonance imaging (MRI) studies show that gadolinium-diethylenetriaminepentaacetic acid (Gd-DTPA) entry into the vitreous space can be used as a qualitative marker of blood-retinal barrier (BRB) disruption. To determine if a more quantitative measurement of BRB breakdown could be obtained, the utility of acquiring real-time, Tl -weighted proton images was studied after Gd-DTPA injection. Two days before the MRI experiment, panretinal photocoagulation was done. The mean signal intensity over selected regions-of-interest (ROI) in the vitreous and anterior chamber was followed before and after (0,10,20,30,45, and 60 min) Gd-DTPA injection (1.0 mmol/kg, intravenously). At every laser power setting used in this study (0,200,400,600, and 800 mw), the change in the mean signal intensity could be approximated by a simple exponential equation. However, the time constants determined for these curves were too imprecise to be useful as correlates between laser power and BRB breakdown. The slope of the line fit to the data in the first 20 min postinjection (ie, an initial-rate analysis) was a more precise correlate between BRB breakdown and laser power. This slope represented the rate of change in mean signal intensity in the ROI as a result of the entry of Gd-DTPA, and it was called the "leakiness" parameter. The leakiness parameter reflected changes in the permeability surface area product of the BRB if the blood flow and the Gd-DTPA arterial concentration immediately after injection were approximately the same between animals. Invest Ophthalmol Vis Sci 32: ,1991 Gadolinium-diethylenetriaminepentaaceticn acid (Gd-DTPA) magnetic resonance imaging (MRI) is a promising method for investigating breakdown of the blood-retinal barrier (BRB). 12 When the BRB is disrupted, Gd-DTPA enters the vitreous space and strongly influences the surrounding proton relaxation rate ([Tl]" 1 ) in direct proportion to the Gd-DTPA concentration. 3 Thus, changes in the water proton relaxation rate (determined from MRI analysis) from the presence of Gd-DTPA may be used as a marker of BRB disruption. Compared with vitreous fluorophotometry (VF), Gd-DTPA MRI has two important advantages: (1) it does not require clear optical media (a From the *National Institute of Environmental Health Sciences, Research Triangle Park, and the fduke University Eye Center, Durham, North Carolina. Partially supported by National Institutes of Health research grant R01 EY07576, and Juvenile Diabetes Foundation (EdJ Jr.). Submitted for publication: January 29, 1991; accepted May 28, Reprint requests: Bruce A. Berkowitz, PhD, Mail-drop 4A-01, P.O. Box 12233, National Institute of Environmental Health Sciences, Research Triangle Park, NC major source of error in VF) and (2) it can produce two- and three-dimensional data compared with the one-dimensional data of VF. 2 Nevertheless, the information available from the MRI method described in the literature is limited because a long delay (greater than 1 hr) between the injection of the Gd-DTPA and the final image acquisition is necessary to optimize the precision of the Tl measurement. Thus, this approach primarily reveals the existence of BRB breakdown and, in general, does not provide the kinetic information necessary to derive important pathophysiologic parameters related to BRB disruption (eg, permeability surface area product). 4 By contrast with this method, which determines a single Tl value from several images, Tl-weighted images display in a single image T1 differences as variations in the signal intensity (ie, contrast). Changes in the relaxation rate with time then are reflected as signal-intensity changes and may be followed without having to determine precise Tl values. To estimate some of the pathophysiologic parameters associated with BRB breakdown, we investigated the utility of acquiring, during the first hour following Gd-DTPA injection, in vivo serial Tl-weighted proton images of 2854

2 No. 11 MPJ STUDY OF BRB BREAKDOWN / Berkowirz er ol 2855 the rabbit eye that previously had undergone panretinal photocoagulation (PRP). Materials and Methods Animal Preparation The 2-3-kg male or female minilop rabbits used in this study were treated in accordance with institutional guidelines and the ARVO Resolution on the Use of Animals in Research. Two days before the MRI experiment, the animals were anesthetized with a mixture of ketamine HC1 (35 mg/kg) and xylazine HC1 (5 mg/kg) intramuscularly and photocoagulated with an Aurora system argon ion blue-green laser (Model ; Lexel, Palo Alto, CA). Approximately 400 lesions were produced in either the left or right eye with laser exposures of 200 msec, and each eye was treated with a different laser power setting (0, 200, 400, 600, or 800 mw). On the day of the experiment, each animal was anesthetized initially as described. Anesthesia was maintained by intravenous administration of ketamine (20-40 mg/kg/hr) through the auricular vein; xylazine was administered intravenously as needed. The animal's heart rate and blood pressure were monitored continually from a femoral arterial catheter as previously described. 5 The animal then was ventilated artificially through a tracheotomy. While on the respirator, pancuronium bromide was administered intravenously (1.2 mg/kg/ hr). Blood-gas monitoring (Micro 13; Instrumentation Laboratories, Lexington, MA) was done periodically to ensure proper status of the animal. The rectal temperature of the animal was maintained at C with a circulating water blanket connected to a constant temperature bath. The animal was placed gently in a home-built nonmagnetic cradle and secured in the prone position. At the end of the experiments, the animals were killed by KC1 injection intravenously. MRI Procedure All in vivo experiments were done on a 4.7-T General Electric CSI horizontal-bore system (General Electric, Freemont, CA) using a whole-head, on-edge, split-capacitance Helmholtz coil (diameter, 9 cm) tuned to 200 MHz. Coronal images were obtained using a combination spin- and gradient-recalled echo sequence with a repetition time of 500 msec and echo time of 28 msec. Slices were selected using a 2-msec sine pulse, and the slice thickness was 3 mm. The field of view was varied between 60 X 60 mm and 85 X 85 mm depending on the head size. The images were obtained with 128 phase-encode steps, 256 complex data points, and two acquisition per phase-encode steps. Each image took approximately 2 min to acquire. The Gd-DTPA (1.0 mmol/kg, Magnevist; Berlex Lab, Wayne, NJ) was infused over a 6-8-min period into the ear vein, and images were obtained before (control), immediately after, and at 10, 20, 30, 45, and 60 min after Gd-DTPA intravenous injection. The images were processed with minimal trapezoidal apodization (Tl = T2 = 10 Hz) followed by zero filling to a final matrix size of 256 X 256 using the software resident on the machine. The images then were transferred by a serial line to a Macintosh II data station (Apple Computer, Inc., Cupertino, CA). Images were analyzed using the program Image (version 1.29; W. Rashban, National Institutes of Health, Bethesda, MD) and consisted of defining a region of interest (ROI) in one image, obtaining a mean signal intensity over that ROI, and then applying that same ROI to the other images in the set. In each eye, the ROI for the vitreous was a hand-drawn semicircle that conformed to the retinal surface over the area of leakage. The semicircle was closed with a chord drawn through the posterior vitreous. For the anterior chamber, the ROI was the whole anterior chamber. The temporal profile of the mean signal intensities then was analyzed. The mean signal intensity (I(t)) time course was approximated by a simple exponential of the form I(t) = ((af- ao)*(l - exp(-t/k)) + ao); where af is the intensity after obtaining steady state (ie, exp(-t/k) «=< 0), ao is the baseline ROI mean signal intensity, and k is the time constant. The product (t/k) was found to be 0.5 or less during the initial 20 min after injection. Thus, this exponential equation could be approximated by a linear equation (ie, an initial-rate analysis) I(t) = ((af - ao)(t/k) + ao) to within an error of 20% or less. The data were analyzed both byfittingfirst the exponential equation and then by initial-rate analysis. The data were analyzed statistically with student t-test, with P < 0.05 signifying statistical significance. In three animals, blood samples were obtained in heparinized syringes during the control period, immediately after the Gd-DTPA injection, and 5, 10, 20, 40, and 60 min postinjection. These samples were stored on ice. At the end of the MRI experiments, the samples were centrifuged, and the plasma fraction was obtained for later nuclear magnetic resonance analysis. A simple inversion recovery Tl experiment was done at 200 MHz on the water signal of the plasma fraction to estimate the Gd-DTPA blood time course. The Tl rate was converted to a Gd-DTPA concentration using a calibration curve obtained from a separate phantom study of various known amounts of Gd-DTPA in water at 25 C (data not shown). No attempt was made to correct for the effects of plasma on the relaxation rate of the Gd-

3 2856 INVESTIGATIVE OPHTHALMOLOGY b VISUAL 5CIENCE / October 1991 DTPA3 because only the relative time course was important in our study. Results Fundus photographs of pigmented rabbit eyes that had undergone PRP at laser power settings of 200, 400,600, and 800 mw are shown in Figure 1. In these experiments, the 200 mw setting produced barely discernible lesions compared with the lesions produced at the higher laser powers. At 400 mw and greater, lesions grew progressively more severe. Figure 2 illustrates sequential coronal T1 -weighted proton images of a rabbit that had both eyes treated by similar PRP protocol except at different laser power settings (800 mw or 400 mw). The presence of GdDTPA in the vitreous could be followed by the increase in brightness in these images. In eyes treated at 400 mw and greater, the signal intensity in the vitreous increased over time. Untreated eyes and eyes treated with a laser power of 200 mw only showed a minor increase in signal intensity over time. Mean signal intensity from the ROIs then were plotted against time for all animals. As shown by the time course in Figure 3, each curve could be fit to a simple exponential function with a characteristic time constant. However, the precision in these time constants was poor because they were too long relative to the overall experimental time (60 min). For the 800mW power setting, the time constant was 62 ± 17 min (mean ± standard deviation, n = 6); for 600 mw, it was 86 ± 46 min (n = 3); for 400 mw, it was 560 ± 867 min (n = 2); for 200 mw, it was 132 ± 219 min (n = 3); and for 0 mw, it was 78 ± 32 min (n = 3). Because of the large error bars, these values are statistically the same (P > 0.3). The time constant for eyes that received the same laser power treatment contained large standard deviations, and the exponential time constant did not provide a good basis with which to compare animals in these experiments. Long term, the functional form of the signal intensity change is expected to be somewhat more complex than an exponential function, and so it was advantageous to do an initial-rate analysis on each curve (Fig. 3). The slope of the line determined with this latter analysis correlated better with the signal-intensity change and the laser power setting (vide infra). In this analysis, the mean slope of the line was determined using the first 20 min of data. This slope represents the rate of change in mean signal intensity over the ROI from Fig. 1. Photographs of pigment rabbit fundus 2 days after argon laser PRP procedure. Rabbit eyes were treated with the laser power set to (A) 200 mw, (B) 400 mw, (C) 600 mw, and (D) 800 mw. These photographs were obtained just prior to MRI examination. Vol. 32

4 No AARl STUDY OF BRB BREAKDOWN / Derkowirz er ol Fig. 2. Representative real-time, T l weighted, coronal proton images acquired during (A) the control period prior to Gd-DTPA injection. (B) immediately after the injection, (C) 10 min post-injection, and (D) 20 min post-injection. Additional images (not shown) were also obtained at 30, 45, and 60 min post-injection. In this animal both eyes were treated with a similar PRP procedure, except the laser power was set to 800 mw for the left eye in the Figure and 400 mw for the right eye. By 10 min post-injection the different extent of leakiness between these eyes is clearly visible. Note that the rate of change in signal intensity over the anterior chamber is similar for each eye. the entry of Gd-DTPA, and we referred to it as the "leakiness" associated with a particular power setting. The relationship between the slope from the initialrate analysis and the pathophysiologic parameters associated with BRB breakdown will be discussed subsequently. The results of the initial-rate analyses at the various power settings used in this study are presented in Figure 4. The leakiness parameter of eyes treated with 0 mw (control) supports the conclusion from the literature that Gd-DTPA normally does not enter the vitreous space.1'6 The leakiness parameter determined in control eyes (0 mw) and eyes treated at the 200 mw setting were similar (P = 0.08), whereas eyes treated at higher power settings than 200 mw had values that were statistically different from control eyes (P < ). It is not surprising that the 200 mw treated eyes did not appear to have a disrupted BRB 2 days after treatment, given the mild clinical effect observed in these eyes (Fig. 1). Although the leakiness parameter of the eyes treated with laser powers of 600 and 800 mw were not statistically different (P = 0.06), the increase in retinal leakiness with increasing laser powers of 400, 600, and 800 mw agreed qualitatively with the ophthalmoscopic observation of a corresponding increase in lesion diameter (Fig. 1). The leakiness parameter for the aqueous humor did not differ between untreated eyes and treated eyes at any of the various laser powers. To understand better the underlying physical mechanisms associated with BRB breakdown, it is necessary to verify that the plasma time course of the GdDTPA in these experiments could be fit to a sum of exponential functions. The plasma Gd-DTPA time course could be followed by taking advantage of the fact that the relaxation rate (Tl)~' of water is related linearly to the Gd-DTPA concentration if all other conditions are constant.3 Figure 5 is a plot of the serum Gd-DTPA concentration versus time under this injection protocol. In this case, the temporal profile was well approximated by the sum of two exponentials.7 Discussion This study investigated the utility of acquiring serial, Tl-weighted proton images of rabbit eyes during the first hour after Gd-DTPA injection to estimate certain kinetic parameters associated with BRB breakdown. It was found that an initial-rate analysis of the influx of Gd-DTPA into the vitreous space provided a reasonably precise correlate of leakiness with BRB perturbation. The blood-aqueous barrier normally is permeable to Gd-DTPA,8 and no correlation

5 2858 INVESTIGATIVE OPHTHALMOLOGY & VISUAL SCIENCE / Ocrober 1991 Vol. 32 between the rate of Gd-DTPA influx into the anterior chamber and laser power was observed. The contrast we obtained was improved by administering a higher dose of Gd-DTPA (1.0 mmol/kg) compared with the clinical dose (0.1 mmol/kg). 9 Possible toxic effects of using the higher Gd-DTPA concentration are minimal because this dose is still an order of magnitude lower than the lethal dose for 50% of animals given intravenous Gd-DTPA (10 mmol/kg). 9 However, to avoid possible adverse physiologic effects, the Gd-DTPA was injected over a 6-8-min period. In general, after the Gd-DTPA injection, no change in the heart rate, blood pressure, or blood-gas parameters (ph, Po 2, and Pco 2 ) was observed. Our key finding was that the leakiness parameter is a relatively precise correlation of BRB disruption in the rabbit eye. However, this is an empiric observation and does not give us insight into the underlying perturbed physiology. To understand the physical parameters associated with the leakiness parameter better, we adapted the equations of others, 4 developed to model the leakage of Gd-DTPA across the bloodbrain barrier, to our study. This analysis assumes that the perturbation of the BRB is distributed uniformly along the retina. Therefore, the initial change in signal intensity (d(s(t))/dt) after Gd-DTPA administration is related to the initial signal intensity [s(0)] and the extraction fraction of Gd-DTPA from the choroidal circulation (E, where E = 1 - exp(-ps/f), where PS is the permeability surface area product (ml/g/sec)), F is the choroidal blood flow (ml/g/sec), a is the initial arterial concentration, and 8 is an unknown local tissue influenced constant 4 : 100 ' Time (min) Fig. 3. Representative time course of the mean signal intensity from the vitreous ROI of the left ( ) and right ( ) eyes of the animal in Figure 2. The dotted lines are the best fit to a simple exponential function (see Materials and Methods). In this case, the time constant for the left eye was 52.6 min and for the right eye was 71.4 min. An initial rate analysis of thefirst20 min following Gd-DTPA injection is presented as the solid lines. For this animal the slope for the left eye was 5.5 and for the right eye it was 2.0. The units for these slopes are change in signal intensity (arbitrary units)/min. The slope determined in the initial rate analysis is called the leakiness parameter since it reflects the amount of Gd-DTPA entering the vitreous space through the disrupted retina. t(n-3 d(s(t)/dt) = s(0) + (EFa/5). [1] The leakiness parameter we found appears to depend on several factors (Equation 1). The value of the extraction fraction of Gd-DTPA from the choroid, and thus the BRB permeability surface area product, is difficult to obtain because 8 is unknown. Nonetheless, this unknown is expected to remain relatively constant between animals because similar pathophysiology is being compared. If the blood flow and starting arterial Gd-DTPA concentration also are constant between animals, relative changes in the BRB permeability surface area product then can be assessed quantitai(n Laser Power (mw) Laser Power (ran) Fig. 4. The relationship between the leakiness parameter and the laser power setting for the (A) vitreous, and (B) anterior chamber. As expected, no discernible correlation was found in the anterior chamber. The larger error bars in (B) probably reflect the variable rate of aqueous humor flow. B

6 No. 11 MRI STUDY OF BRB BREAKDOWN / Derkowirz er ol 2859 so 30 Time (min) Fig. 5. This is a plot of the plasma Gd-DTPA concentration vs. time after Gd-DTPA injection. The data represented by the filled circles were obtained from three animals, and the datum represented by thefilledsquare was obtained from a single animal. This curve was wellfitto a sum of two exponentials, taking into account the baseline Tl rate. The time constants for this curve were a, = 0.45 and a 2 = min" 1. tively (Fig. 4). This technique may be applied to evaluate BRB breakdown under conditions other than PRP treatment. It is well known that BRB breakdown is the earliest physiologic manifestation of diabetic retinopathy. 2 The usefulness of this technique in the investigation of breakdown of the BRB in humans remains to be determined. Analysis of our data in terms of Equation 1 is complicated by several additional factors. First, the Gd- DTPA entering the vitreous space is not distributed immediately and homogeneously, but because of the slow diffusion of Gd-DTPA in the vitreous, it is distributed in a heterogeneous manner (Fig. 2). To account for this nonuniform Gd-DTPA vitreous distribution, a single mean signal intensity was determined over a large ROI in the vitreous. Thus, signal-intensity variations within the ROI tend to cancel each other, producing a more homogeneous-appearing signal-intensity distribution. In addition, by analyzing the data in this fashion, the results are not influenced as strongly by the entry of Gd-DTPA from outside the selected slice. A second complication is that the proper application of Equation 1 to our data would require that changes in the signal intensity of the preretinal vitreous be monitored selectively over time. However, this volume is defined poorly, and its rate of change in the signal intensity is fairly rapid compared with the image-acquisition time (data not shown). These problems also are addressed by the method of data analysis we described because the rate at which the mean signal intensity changes over time for a large, well-defined portion of the vitreous space will, in general, be slower than the rate of change of the preretinal signal intensity over the same time. This buffering effect can be understood as follows: the useful upper Gd-DTPA concentration is reached when, for a given set of dataacquisition parameters, additional Gd-DTPA does not improve the contrast (ie, saturation is reached). The preretinal volume initially will contain the highest level of Gd-DTPA, and the signal intensity in this volume will reach saturation faster than other regions in the vitreous. However, by obtaining the mean signal intensity over a larger ROI, contributions from the high preretinal signal intensity will be suppressed by other regions in the vitreous that have a lower signal intensity. Thus, the overall rate at which the signal intensity over the larger ROI changes will be slower. A disadvantage of this data-analysis scheme is that although it greatly improves the precision of the measurement by significantly suppressing the effects of signal-intensity variation, it introduces a slight random error into the mean signal-intensity value because the ROI is hand-drawn and thus somewhat different from eye to eye. However, the contributions of the random error to the final mean signal intensity are not cumulative and probably only minimally perturb the final mean values. In conclusion, the utility of real-time, Tl-weighted MRI was evaluated to compare quantitatively the breakdown of the BRB between animals. A correlation (the leakiness parameter) of BRB breakdown was determined that is related to the permeability surface area product (through the extraction term). This parameter was obtained by measuring the mean signal intensity rate of change (ie, the slope) during the first 20 min postinjection and assuming that the blood parameters (flow and initial arterial Gd-DTPA concentration) were relatively constant between animals. The image contrast we produced was improved through the use of relatively high concentrations of injected Gd-DTPA compared to previous studies. Key words: magnetic resonance imaging, blood-retinal barrier, gadolinium, panretinal photocoagulation, rabbit Acknowledgments The authors thank Drs. Diane Hatchell and Bob London for their ongoing support and Alex Funk for continuing technical support. References 1. Frank JA, Dwyer AJ, Girton M, Knop RH, Sank VJ, Gansow OA, Magerstadt M, Brechbiel M, and Doppman JL: Opening of blood-ocular barrier demonstrated by contrast-enhanced MR imaging. J Comput Assist Tomogr 10:912, 1986.

7 2860 INVESTIGATIVE OPHTHALMOLOGY 6 VISUAL SCIENCE / Ocrober 1991 Vol Plehwe WE, McRobbie DW, Lcrski RA, and Kohner EM: Quantitative magnetic resonance imaging in assessment of the blood-retinal barrier. Invest Ophthalmol Vis Sci 29:663, Gadian DG, Payne JA, Bryant DF, Young IR, Carr DH, and Bydder GM: Gadolinium-DTPA as a contrast agent in MR imaging: Theoretical projections and practical observations. J Comput Assist Tomogr 9:242, Larsson HBW, Stubgaard M, Frederiksen JL, Jensen M, Henriksen O, and Paulson OB: Quantitation of blood-brain barrier defect by magnetic resonance imaging and gadolinium- DTPA in patients with multiple sclerosis and brain tumors. Magn Reson Med 16:117, Berkowitz BA, Wilson CA, and Hatchell DL: Oxygen kinetics in the vitreous substitute perfluorotributylamine: A I9 F NMR study in vivo. Invest Ophthalmol Vis Sci 32:2382, Runge VM, Clanton JA, Price AC, Wehr CJ, Herzer WA, Partain CL, and James AE Jr: The use of Gd DTPA as a perfusion agent and marker of blood-brain barrier disruption. Magn Reson Imaging 3:43, Tofts PS and Kermode AG: Measurement of the blood-brain barrier permeability and leakage space using dynamic MR imaging: 1. Fundamental concepts. Magn Reson Med 17:357, Cheng H-M, Kwong KK, Xiong J, and Chang C: GdDTPA-enhanced magnetic resonance imaging of the aqueousflowin the rabbit eye. Magn Reson Med 17:237, Weinmann H-J, Brasch RC, Press WR, and Wesby GE: Characteristics of gadolium-dtpa complex: A potential NMR contrast agent. AJR 142:619, 1984.