Chronic Ischemia Induces Regional Axonal Damage in Experimental PrimateOptic Neuropathy FREE

Vascular abnormalities are thought to contribute to the developmentof glaucomatous optic neuropathy and have been investigated for more than150 years. A variety of experimental and clinical observations demonstratecirculatory aberrations of the anterior optic nerve, the peripapillary region,1-3the choroid,^2,4- 5 andthe retrobulbar vasculature^6- 11 ineyes with glaucomatous optic neuropathy.^7,11- 14 Theseapparent alterations in the circulation of the glaucomatous optic nerve implicatedysfunction of vascular regulatory mechanisms. Further support for vasculardysregulation has been suggested by the finding of elevated levels of endothelin-1(ET-1), a potent vasoconstricting peptide, in the plasma^10,15- 17 andaqueous humor^18- 19 of patientswith glaucoma.

To further investigate the consequences of chronic ischemia, we developeda model of sustained optic nerve ischemia. This model uses the continuousadministration of ET-1 to the perineural region of the optic nerve in theretrobulbar space and has been described in rabbits,²⁰ rats,²¹ and nonhuman primates.²² Theseexperiments allow the direct in vivo observation of optic nerve structureand function during periods of chronic ischemia. Comparisons to human glaucomatousoptic neuropathy and experimental optic neuropathy following chronic elevationof intraocular pressure (IOP) are also possible. In the rabbit model, a reductionof blood flow to the anterior optic nerve of approximately 38% was achieved,resulting in loss of optic nerve axons, loss of neuroretinal tissues of theoptic nerve, and enlargement of the optic nerve cup.^22- 23 Similarly,the nonhuman primate model showed approximately 36% blood flow reductionsto the anterior optic nerve²² and significantloss of optic nerve axons after 3 to 6 months of continuous ischemia.^24- 25 Although axonal loss in these initialstudies appeared in all regions of the optic nerve, visual inspection of theoptic nerve sections revealed regional differences in the loss of axons. Thisfinding has led to a more thorough examination of regional axonal changes.Because nonhuman primates have a vascular architecture and tissue propertieswithin the anterior optic nerve that are similar to the human eye,^26- 27 primate models are often used toprovide meaningful comparisons to human glaucoma.

This study examined the susceptibility of the optic nerve retinal ganglioncell axons to chronic ischemia in the nonhuman primate eye. Axonal loss withinthe optic nerve is assessed histologically and compared between eyes withchronic ischemia from ET-1 treatment and untreated contralateral control eyes.We further evaluate changes of axonal density in various subregions of theoptic nerve following experimental chronic ischemia to examine regional differencesin the axonal loss.

Twelve adult female rhesus (Macaca mulatta)monkeys, weighing 4 to 8 kg and between the ages of 8 and 19 years, were usedin accord with the Association for Research in Vision and Ophthalmology Statementon the use of animals in ophthalmic and vision research. At the beginningof the study, animals underwent a complete eye examination, which includedslitlamp examination, IOP measurements, and dilated funduscopic examinationfollowed by stereoscopic optic nerve photography. For these examinations,animals were sedated with ketamine hydrochloride (Ketaset; Fort Dodge AnimalHealth, Fort Dodge, Iowa) administered intramuscularly (15 mg/kg) and wereintubated. Anesthesia of the animals is achieved with inhalation of 3% isofluranein oxygen. Blood oxygen saturation and pulse rate are monitored via peripheraloxymetry and core body temperature was maintained with a heating pad.

The surgical procedures for the implantation of the ET-1 minipumps inprimates has been described previously.²⁰ Theuntreated contralateral eye served as the control. In brief, an osmotic minipump(Alzet model 2004; Durect Corp, Cupertino, Calif) preloaded with approximately250-µL ET-1 (Peptides International, Louisville, Ky) solution (55µM/mL)was implanted subcutaneously superior to the lateral rim of the orbit of theright eye. The outlet of the pump was connected to one end of a silicone deliverytube. The tube was passed subcutaneously into the superotemporal orbit andunder the bulbar conjunctiva and the Tenon capsule. The tube was placed beneaththe superior rectus muscle, and the end of the tube was secured in the retrobulbarspace adjacent to the superonasal aspect of the optic nerve. The tube wasfixed in place using a 9-0 nylon suture through the sclera. The minipump continuallydelivered the ET-1 solution to the retrobulbar region at a dose of 0.34 µg/dwith a constant flow rate of 6 µL/d over 4 weeks. At the end of each4-week period, the minipump was replaced with a new ET-1 loaded pump at theoriginal site of placement; the patency of the tube was ensured by irrigationof balanced salt solution (Alcon Laboratories Inc, Fort Worth, Tex), priorto the placement of the new pump. Tubes with a blocked opening due to scartissue growth were replaced. Otherwise, the tube was left in place and thenew pump attached. Sham implantation of minipumps loaded with vehicle withoutET-1 demonstrated no short-term effect on blood flow and other factors examinedin a previous report.^22,28 Anadditional sham animal was included in the present study, but unfortunately,it died of unrelated causes 3 months following pump implantation; tissue wasunavailable for analysis. In addition, both optic nerves of a control animalwith normal eyes (no pump implantation and normal findings on ocular examination)were analyzed to assess the normal intereye variability of axonal density.

The duration of ET-1–induced ischemia was approximately 6 months(mean [SD], 23 [2] weeks) in 8 monkeys, 9 months (38 weeks) in 1 monkey, and12 months (52 [1] weeks) in 3 monkeys (Table 1). During this period, a variety of in vivo measurementswere made in each animal at baseline prior to pump implantation and bimonthlyduring the study. These tests included IOP measurements in both eyes measuredwith tonometry (Tonopen; Pat Leonard Surgical Inc, Shawnee, Kan) immediatelyafter the animals were anesthetized with 15 mg/kg of intramuscular ketaminehydrochloride. Three measurements were made for each eye and a mean IOP wasrecorded. Optic nerve appearance was monitored with stereoscopic optic nervephotography (model 3-DX; Nidek Co Ltd, Tokyo, Japan). Photographs of the opticnerves were evaluated in a masked fashion for indications of acute anteriorischemic optic neuropathy, such as focal or global pallor, edema, or hemorrhages.

To test the chemical stability of ET-1 in the minipump during the 4-weekimplantation cycles, 4 minipumps were preloaded with the same ET-1 solutionas described earlier and placed in an isotonic saline solution at 37°Cfor 1 through 4 weeks. At the end of each time point, the ET-1 solution waswithdrawn from the pumps. A sample of 5 µL was obtained from each minipumpand compared with a freshly prepared control ET-1 solution (3 tests for eachsample). Using high-performance liquid chromatography (model 1100; HewlettPackard, Palo Alto, Calif) the test samples and control samples were analyzed.Compared with the control sample, the percentage of ET-1 detected in the samplesat the end of 1 through 4 weeks were 110%, 83%, 83%, and 74%, respectively.

At the end of the experiments, the animals were euthanized with an intravenousinjection of 100 mg/kg pentobarbital sodium and phenytoin sodium (Euthasol;Delmarva Laboratories Inc, Midlothian, Va). Perfusion fixation was accomplishedusing a 4% formaldehyde solution via the carotid arteries. Both eyes wereimmediately enucleated and fixed in a 4% formaldehyde solution for an additional2 to 3 hours. The retrobulbar optic nerves were transected (approximately2-3 mm behind the globe), and a 0.5-mm-thick section was obtained and processedfor resin sectioning. All of the optic nerve tissue sections were fixed in5% glutaraldehyde in a phosphate buffer (pH 7.4) for 1 hour and then rinsedin phosphate-buffered saline solution before being postfixed in 2% osmiumtetroxide for 3½ hours. The tissue was rinsed again, dehydrated inan ethanol-acetone series, and embedded in epoxy resin (Epon 812; Epon-LKBInstrument, Gaithersburg, Md). Semithin sections (1 µm) were cut andmounted on slides. The adjacent stump of each optic nerve still attached tothe globe immediately opposite to the cut surface was marked with coloredtissue dye (Tissue Marking Dyes; Triangle Biomedical Sciences Inc, Durham,NC) to preserve the orientation in situ for later anatomical orientation ofeach section.

The sections on the slides were stained with 1% phenylenediamine (in1:1 methanol/isopropanol) for 20 minutes, rinsed 2 to 3 times with isopropanol,and air-dried. If staining was incomplete with phenylenediamine, at a temperatureof approximately 80°C, adjacent sections were stained with 1% toluidineblue in the phosphate-buffered solution (pH 7.0-7.4) for 3 minutes followedby adding a few drops of Sorensen's buffer for another 2 minutes. The slideswere rinsed with distilled water and air-dried. The orientation of the opticnerve sections was determined by matching certain landmarks (such as bloodvessels and contour lines of axonal bundles) of the sections with the adjacentoptic nerve stumps from where the sections were cut.

The methodological details of the retinal ganglion cell (RGC) axonaldensity quantification are described in a previous publication.²⁹ Inbrief, with an image analysis system (Bioquant; R&M Biometrics Inc, Nashville,Tenn), a composite of the entire area of the optic nerve cross section wascreated. The area was then divided with a grid such that each square of thegrid had a size of 2025 µm² (45 × 45 µm). Eightpercent of the grid squares in untreated controls and 20% in ET-1–treatedoptic nerves were randomly selected. All axons within each of the randomlyselected grid squares were counted using the image analysis system. Thesepercentages of the total area to be counted were chosen to minimize variancewhile maximizing efficiency.²⁹ The densityof axons in each of the grid squares was calculated by dividing the numberof axons by the area of the grid square, and the density of each region wasestimated by averaging the densities of all of the counted grid squares withinthe region of interest. This technique allowed for the variance to be calculatedfor the mean axonal densities in each region and for statistical comparisonof the individual regions. In addition, the area of the entire optic nervewas automatically calculated by the image analysis system by thresholdingthe outer edges of the optic nerve sections. This was done to access for swellingor shrinkage of the optic nerves and its effect on density calculations.

For statistical analysis and assessment of regional axonal loss, theoptic nerves were divided into 16 subregions consisting of an inner and outersegment within each of 8 wedge-shaped, radial sectors as shown in Figure 1 (ST, TS, TI, IT, IN, NI, NS, SNwith T indicating temporal; N, nasal; S, superior; and I, inferior). All 16subregions had approximately equal areas. As mentioned earlier, the axonaldensity for each region was calculated as the average of all grid-square sampleswithin the region. Similarly, the total overall axonal density for an entireoptic nerve was calculated as the average of all density samples (ie, allcounted grid squares) for each optic nerve.

The effect of ET-1–induced chronic ischemia was evaluated by comparisonof ET-1–treated and control eye group averages for overall axonal densityusing a 1-tailed paired t test (n = 12 pairs). Two-way,matched-pair analysis of variance (ANOVA)(treatment × animal) was usedto further evaluate the effect of ET-1 treatment on the overall axonal density,while controlling for variation due to differences between individual animals.Bonferroni-corrected post hoc tests were used to evaluate the significanceof ET-1 effects for individual animals (pairs of eyes). Potential regionaleffects of ET-1–induced ischemia were first evaluated across the wholegroup (n = 12) using 2-way, matched-pair ANOVA (treatment × region).Regional effects of ET-1 treatment were further explored within individualanimals using a similar 2-way, matched-pair ANOVA (treatment × region).This analysis was possible because the variance of each observation (meanof each optic nerve region) could be calculated based on the numerous samplesobtained (grid squares counted) in each optic nerve region. For this regionalanalysis of each individual animal, the effect of ET-1 treatment was consideredstatistically significant at P<.0042; thus, αwas adjusted for 12 comparisons. If there was also a significant interactionbetween ET-1 treatment and optic nerve region, Bonferroni-corrected post hoctests were used to evaluate the significance of axonal density differencesamong individual pairs of regions (ET-1 vs control eye). Finally, the opticnerve areas were compared using a paired t test toassess for optic nerve swelling or shrinkage. Intraocular pressure measurementsfrom baseline, the midpoint, and the end of the experiment were analyzed withrepeated-measures ANOVA.

The mean (SD) IOP for the 12 ET-1–treated eyes at baseline (pretreatment)was 15.5 (4.3) mm Hg. At the midpoint of each animal's experimental period,the average IOP in ET-1–treated eyes was 17.1 (3.0) mm Hg. At the finalexamination in the ET-1–treated eyes the IOP was 14.9 (4.5) mm Hg. Theaverage IOP in the group of untreated contralateral control eyes was 15.1(3.6) mm Hg at baseline, 16.7 (3.8) mm Hg during the experimental period,and 14.6 (4.0) mm Hg at the final examination. There was no significant differencebetween the IOPs in the control and ET-1–treated eyes (P = .61). Likewise, the IOP did not differ significantly between baselineand final measurements in either the control eyes or the ET-1–treatedeyes (P>.05, ANOVA).

The stereophotographs of the optic nerves were examined and comparedbetween baseline and subsequent examinations. There was no evidence of pallor,edema, or other changes commonly associated with anterior ischemic optic neuropathyin any of the photographs (Figure 2).Owing to the few animals and the high variability associated with clinicalassessment of stereoscopic optic nerve photographs for progressive change,we have excluded such an analysis from this article. However, automated opticnerve head analysis (Heidelberg Retina Tomographs; Heidelberg Engineering,Dossenheim, Germany) was performed and is being analyzed.

Under the light microscope, the optic nerves of the ET-1–treatedeyes showed a variety of morphological changes in the RGC axons and connectivetissues compared with the untreated contralateral eyes (Figure 3). These changes included axonal demyelination, axonal swelling,axonal shrinkage, and axonal fragmentation. These changes may represent thevarious stages in the continuum of axonal degeneration. Connective tissuechanges included increased thickness of the perineurium and enlarged extra-axonalspaces associated with the loss of axons. In areas with more severe damage,neuroglial cells appeared to be larger and increased in number and axons appearedto be significantly larger. Figure 3 showscomparisons between 2 different optic nerve regions in 2 different individualanimals (monkeys 2 and 6). The optic nerve subregions were chosen becausethey represent areas of significant axonal loss in the ET-1–treatedeyes compared with the untreated contralateral eyes (see the "Regional RGCAxonal Density Comparison" subsection). The mean and standard deviation ofthe axon sizes is given for each region in Figure 3. On average, in regions of axonal damage in the ET-1–treatedeyes, more of the RGC axons appeared swollen, the variability of the axonsize was greater, and the optic nerves were larger. The mean area of the opticnerves in the ET-1–treated eyes was 6.27 (0.96) mm² whilethe area of the untreated contralateral control eyes was 5.91 (0.89) mm². This represents an overall increase of 6% in the size of the opticnerves in the ET-1–treated eyes (P = .03, 1-tailedpaired t test). However, the contribution of opticnerve swelling to axonal density calculations is not uniform, as some ET-1–treatedoptic nerves were larger and some were unchanged. In addition, axon sizesappeared larger overall in the ET-1–treated eyes, meaning that axonalenlargement and extracellular space enlargement both contribute to the overallincrease in the size of the optic nerves. As an example, monkey 6 (Figure 3) had significant overall, as wellas regional, axonal loss without optic nerve swelling (ET-1–treatedeye optic nerve size, 7.3 mm²; untreated contralateral controleye, 7.2 mm²), while monkey 2 had significant regional and overallaxonal loss with optic nerve enlargement (ET-1–treated eye optic nervesize, 7.4 mm²; untreated contralateral control eye, 6.3 mm²).

Following 6 to 12 months of continuous ET-1 administration in 1 eye,the group mean axonal density for the entire optic nerve among the 12 ET-1–treatedeyes was 208 310/mm², while the untreated contralateral controleye group mean was 220 661/mm² (t =2.04; P = .03 1-tailed paired t test; Table 1). On average,this change represents a 5.1% (10.0%) decrease of axonal density owing toET-1–induced chronic ischemia. However, Table 1 also shows that the effect of ET-1 treatment was not uniformacross all 12 animals; individual animals varied greatly in response to chronicischemia. There was no significant effect of duration of exposure to ET-1treatment (Kruskal-Wallis statistic = 0.04, P = .98).Admittedly, the number of animals treated for 9 and 12 months is small andmay have prevented detection of an effect. Axonal density also varied substantiallyamong untreated contralateral control eyes ranging from 170 622 to 286 825axons/mm² (220 661 [35 855] axons/mm²). Therefore,a 2-way ANOVA (treatment × animal) was applied and Bonferroni-correctedpost hoc tests were used to evaluate the effect of ET-1 on overall axonaldensity for the whole experimental group and for each individual, respectively(Table 1). This allowed for controlof the variation owing to differences between individual animals. As expected,axonal density varied significantly between individual animals (F = 129.9; P<.001). The ET-1 treatmenteffect was significant for the group as a whole (F =311.2; P<.001), and there was a significant interactionbetween treatment and animal effects (F = 31.4, P<.001). Bonferroni-corrected post hoc testing revealedthat 9 of the 12 animals had significant axonal density differences betweentheir ET-1–treated and untreated contralateral control eyes, but 2 ofthese had a greater density in the ET-1–treated eye (Table 1). Significant axonal loss was seen in 7 animals and theloss varied from 4% to 21%, with a mean (SD) loss of 11.6% (6.7%).

Several studies have demonstrated regional susceptibility within theoptic nerve in glaucoma and other forms of optic neuropathy in human, as wellas in experimental glaucoma in nonhuman primates and rodents. In visual evaluationof most optic nerve cross sections, it was immediately evident that varioussubregions sustained more damage than others. To evaluate the potential regionalsusceptibility in this experimental model of chronic ischemia, the axonaldensity was also assessed for 16 optic nerve subregions (see the "Methods"section). Figure 4 shows regionaldifferences in axonal densities for the group of 12 experimental animals,ET-1–treated vs untreated contralateral control eyes. Regional analysisshowed a significant group effect of ET-1–induced ischemia (F = 19.9; P<.001), but there was no significantinteraction observed between treatment and region (F =1.2; P>.05). As most of the variance (71%) in axonaldensity was attributable to differences between individual animals, comparisonsbetween the 2 eyes (ET-1–treated and untreated contralateral control)of each animal would allow a better measure of regional axonal density change.

Table 2 lists that regionallosses in axonal density were observed in 8 of 12 ET-1–treated eyesrelative to the corresponding region of the optic nerve from each animal'suntreated contralateral control eye. Two of the 12 (monkeys 3 and 4) experimentaleyes had differences in regional axonal density where the density was actuallysignificantly greater in the ET-1–treated eye. The remaining 2 experimentaleyes (monkeys 5 and 7) had no significant effect of ET-1 treatment and, thus,analysis of regional differences (treatment × region interactions) wasnot applicable.

Figure 5 shows the spectrumof regional effects observed across the group. The pair of pie charts in eachpanel shows data for a single animal. The numbers within each region in theleft column represent the axonal density in the ET-1–treated eye relativeto the density in the corresponding region of the optic nerve from the fellowcontrol eye (ET/control). The right column for each pair shows the probabilityplot determined by 2-way ANOVA with Bonferroni-corrected post hoc testingof the observed regional differences between the ET-1–treated eye anduntreated contralateral control eye for each individual animal. There werefew regions where the axonal density in the ET-1–treated eye was significantlygreater than the density in corresponding region of the untreated contralateralcontrol eye. For example, panels D through G demonstrate significantly loweraxonal densities within single subregions of the untreated contralateral controleyes (gray-shaded subregions, ET-1 axonal density>control). In contrast, regionswhere the axonal density in the ET-1–treated eye was significantly reducedcompared with the axonal density in the corresponding region of the untreatedcontralateral control eye was common. Panels A through F each show that significantloss of axonal density occurred in multiple subregions of ET-1–treatedeyes. Most common areas of axonal loss were found in clusters of contiguoussubregions. In panel B, there is significant loss in both the superonasaland inferotemporal regions, whereas panels A and C through F show more focaldeficits in clusters of 2 to 7 subregions.

To further investigate the variability of axonal density in healthyeyes and between the 2 eyes of a healthy primate, 2 eyes of a single monkeywithout minipump implantation were examined and analyzed in an identical statisticalfashion. Figure 6 shows the relativeaxonal densities in the various regions between 2 normal eyes of the singlemonkey. The probability plot derived by identical statistical methods as wasused for the ET-1–treated monkeys. As expected, there was some variabilityof the intereye ratio of axonal density for various optic nerve regions, althoughnone of the intereye regional differences were statistically significant (F = 0.1, P = .79). The effectof region was statistically significant (F = 11.3, P<.001), but the effect of eye was not statisticallysignificant (F = 0.1, P =.79). The global axonal density difference between these 2 healthy eyes wasless than 1% and not significantly different (OD, 233 118 axons/mm²; and OS, 221 470 axons/mm²; P =.78, paired t test).

Using an established model of optic nerve ischemia that is based onET-1 administration to the retrobulbar space in the primate eye, we demonstratethat chronic ischemia results in significant damage of the optic nerve withoutelevation of IOP. Eyes in which ischemia was induced for 6 to 12 months showedsignificant decrease in axonal density. While the axonal loss was significantoverall for the population, the damage represents early loss with an averageof approximately 5%. However, some individuals have global loss exceeding15% to 20%, with even greater amounts of regional loss. This ischemia-inducedaxonal loss was commonly focal in nature, but the location of loss withinthe optic nerve varied among individual animals. The axonal loss was oftenseen in clusters of contiguous subregions of the optic nerve of an individualanimal. Clinical findings typical for anterior ischemic optic neuropathy,such as generalized or focal optic nerve pallor, optic nerve and retinal edema,and telangiectatic vessels, were not seen. Although, on average, the retrobulbaroptic nerve area was increased in the ischemic eyes owing to axonal enlargementand extracellular space enlargement from axonal loss, there was no clinicalevidence of anterior optic nerve swelling. This finding may imply a differencebetween severe acute ischemia associated with anterior ischemic optic neuropathyand milder, chronic ischemia associated with the hemodynamic alterations producedin our model.

Variability in the response to ischemia was seen both between animalsand among regions of the optic nerve. These findings suggest that optic nervesdiffer in their susceptibility to ischemia and that individual optic nerveshave varied regional susceptibility. Regional axonal loss and structural damagewithin the optic nerve is found in various forms of optic neuropathy, includingglaucoma.^30- 31 The findings ofthe present study pose an important question: Why are certain optic nervesand certain optic nerve regions of an individual optic nerve more susceptibleto damage from noxious stimuli?

Presumably, disparate insults such as elevated IOP and chronic ischemiamay both result in regional or focal axonal injury. However, before consideringpotential hypotheses for the regional loss of axons in this model, one shouldconsider if the model itself is responsible for a particular pattern of loss.The delivery tubes for the ET-1 within the retrobulbar space were locatedto the nasal or superonasal side of the optic nerve in all animals. Althoughthere is no way to know if the decrease in blood flow to the optic nerve wasuniform, the highly varied susceptibility to damage is somewhat surprising.If the location of the delivery of the ET-1 is an important factor in thepattern of axonal loss, one would anticipate a greater loss of axons in thesuperonasal region owing to the relatively greater exposure. However, thispattern was not uniformly seen. Instead,Figure3 and Figure 5 demonstratethat regional susceptibility was varied among individuals. Varying the placementof the delivery tube in future experiments will provide additional informationregarding this matter.

Preferential superior and inferior regional damage of the optic nervein human glaucoma is seen clinically as the development of localized neuroretinalrim notches, optic disc hemorrhages, and visual function defects. Jonas etal³⁰ and Mok et al³² showedglaucomatous optic nerve loss occurs in all neuroretinal rim sectors of theoptic nerve, but in early glaucoma the most pronounced loss is in the inferotemporalrim. Preferential regional damage in glaucoma has traditionally been attributedto differential mechanical forces and explained by anatomical features atthe level of the anterior lamina cribrosa.^33- 35 Thatis, in the superior and inferior regions, there is less connective tissuearound lamina cribrosa pores and the pores are large compared with the nasaland temporal regions.^33- 34 Inaddition, lamina cribrosa pores in the peripheral region are longer than thecentral pores.³⁵ Recent biomechanical modelingof posterior segment of the eye and optic nerve suggests disproportionatestress and strain within the optic nerve may contribute to regional loss ofaxons.³⁶ These anatomical features may causethe axons passing through the lamina cribrosa within these regions to be morevulnerable to mechanical stress, such as an increase in IOP.

Realization that both mechanical insults preferentially damage certainregions of the optic nerve and that a chronic ischemic insult produces regionaldamage, forces a reexamination of the topic. Several hypothetical explanationsbased on regional differences in the anatomical, metabolic, and/or vascularprofiles of the anterior optic nerve can be considered. First, regional anatomicaldifferences that exist within the lamina cribrosa, such as differences inthe constitutive glial cell populations and in the retinal ganglion cell types,may account for varied susceptibility. Second, the metabolic demand may varybetween different regions of the optic nerve. Under normal physiological conditionsthe metabolic demands of the certain regions of the optic nerve are accommodatedfor, but when perturbed, the system is overwhelmed and high metabolic regionswould be more vulnerable to insults, such as perfusion deficits or localizedmechanical stress. Third, differences in the vascular anatomy, such as thenumber and caliber of supply arteries or the distribution of capillaries withinthe optic nerve, may play a role in regional variations. Perhaps the totalnumber of capillaries subserving a set number of RGC axons in various regionsof the optic nerve is borderline but becomes inadequate and exhibits differentialsusceptibility when stressed owing to differential regional autoregulatorycapabilities.^37- 38 Regional functionalinadequacy may not be accompanied by a structural decrease in the vasculature,as demonstrated previously.³⁹ Instead of ameasurable decrease in the number of capillaries, loss of autoregulatory capacitymay be the underlying deficiency.

The existence of regional watershed zones has been suggested to be importantin the development of regional damage in the human glaucomatous optic nerve.^40- 42 Approximately 60%of these watershed zones pass through the temporal half of the optic nervein eyes with glaucoma.⁴³ Because the area withina watershed zone is comparatively less perfused, especially during hemodynamicstress, these regions are more vulnerable to ischemic damage. Given the similaroptic nerve structure and vasculature of nonhuman primates to the human eye,the varied severity and regional location of axonal damage seen in this experimentmay be owing to the result of uneven perfusion of the optic nerve caused bywatershed zones. This hypothesis might also explain the variability in theresponse between different animals. In the future, the addition of serialin vivo angiography before and after the onset of endothelin treatment wouldallow testing the hypotheses in this primate model.

Fourth, differential regional response or sensitivity to the vasoconstrictiveeffects of ET-1 treatment could also explain the findings of this study. Thismodel assumes that the axonal damage is the direct result of ischemia, whichwas induced by ET-1.^20,22 Endothelinsare a family of short-chain polypeptides that have been identified and describedover the last 2 decades. As the most potent vasoconstrictor known, ET-1 wasselected in this model to create significant vasoconstriction over a prolongedperiod. This peptide affects the vasculature through 2 major receptors, ETA and B receptors. Recent studies have indicated that the ET B receptor, whichis largely located on glial cells in the central nervous system, can alsoparticipate in the pathological mechanism of neuronal damage.^44- 47 Thisconcept may be particularly interesting because patients with primary open-angleglaucoma may have higher ET-1 levels in the plasma and vitreous humor, comparedwith healthy controls.^{10,15- 16,18- 19} Thus,ET-1 may cause blood flow insufficiencies in the optic nerve owing to itspotent vasoconstrictor effect or, on the other hand, may cause neuronal damagemediated by ET B receptor on the glial cells via mechanisms not yet fullyunderstood. Examining the distribution of ET receptors and their activityin the optic nerve may provide insight into these issues.

We have demonstrated a significant loss of RGC axons in the primateoptic nerve following prolonged, chronic ischemia. This ischemia-induced,axonal loss was commonly focal in nature, but variable in location withinthe optic nerve of individual animals. As well, there was a significant variabilityin the response of different animals to the ischemic insult. Preferentialregional loss of optic nerve axons may result from regional differences inthe anatomy, metabolism, or vasculature of the optic nerve. These observationswarrant further investigation leading to a more complete understanding ofthe mechanisms of ischemic neuronal damage and their potential role in thedevelopment of glaucomatous optic neuropathy.

Corresponding author: George A. Cioffi, MD, Discoveries in Sight,Devers Eye Institute, 1040 NW 22nd Ave, Portland, OR 97210 (gacioffi@discoveriesinsight.org).

Submitted for publication May 14, 2004; accepted July 14, 2004.

This study was supported by grants RO1-EY05231 from the National Institutesof Health, Bethesda, Md (Dr Cioffi); and the Legacy Good Samaritan Foundation,Portland, Ore (Dr Cioffi).