Foveal Ganglion Cell Loss Is Size Dependent in Experimental Glaucoma

Yoseph Glovinsky,* Harry A. Quigley,~f and Mary E. Peasef

Purpose. The purpose of this study was to study the pattern of foveal ganglion cell loss in experimental glaucoma. Methods. Retinal ganglion cell size and number in the foveal region of seven monkey eyes with experimental glaucoma was determined and compared to normal monkey eyes. Serial sections of macular retina were studied in two regions: the plateau of peak density of ganglion cells (800-1 100 M"I from the fovea), and within 500 /xm of the foveal center. Results. In normal eyes, cell densities were 37,900 ± 2700 in the foveal plateau and 17,200 ± 1800 cells/mm2 in the foveal center. There was selective loss of larger ganglion cells in glaucoma eyes. The degree of foveal ganglion cell loss was significantly correlated to the degree of nerve fiber loss in the temporal optic nerve of the same eye. Conclusions. Detection of early, central visual function loss in glaucoma could be enhanced by testing functions subserved by larger retinal ganglion cells. Invest Ophthalmol Vis Sci 1993; 34:395-400.

.Large ganglion cells located in the retinal midperiph- cells, this information could help to select tests for ery are selectively damaged in human and experimen- early glaucomatous damage. We therefore studied the tal glaucoma.1'2 Glaucoma selectively decreases the ax- foveal area in eyes of monkeys with experimental glau- onal How to the magnocellular layers of the lateral coma and in normal eyes. geniculate body subserving the retinal midperiphery.3 Previous reports45 did not resolve whether there was MATERIALS AND METHODS size-dependent damage in the fovea. This is important because foveal function tests are abundant and simple This investigation adhered to the principles of the to use.6 If there is selective injury to foveal ganglion ARVO Resolution on Use of Animals in Research and was approved and monitored by the Institutional Ani- mal Care and Use Committee of the Johns Hopkins From the ^Glaucoma Service and Dana Center for Preventive University School of Medicine. Ophthalmology, Wilmer Ophthalmological. institute, Johns Hopkins Experimental glaucoma was induced in one eye of University School, of Medicine, Baltimore, Maryland, and the cynomolgus monkeys (Macaca fascicularis) by argon *Goldsch.leger Eye Institute, Sadder School of Medicine, Tel-Aviv 7 University, Tel-Aviv, Israel. laser applications to the trabecular meshwork. After Supported in part by PHS Research Grants EY 02120 and 01765 several months of intraocular pressure elevation, (HA Q) and, unrestricted research support from. National Glaucoma various levels of glaucomatous optic nerve damage oc- Research, a program of the American. Health Assistance Foundation, Rochvillc, Maryland, and a grant from Mr. G. Sheba (YG). curred. The monkeys were then exsanguinated under Submitted for publication: December 24, 1991; accepted August 31, intravenous sodium pentobarbital anesthesia by per- 1992. fusion with cold saline followed by 4% paraformalde- Proprietary interest, category: N. Reprints requests: Harry A. Qui.gl.ey, Wilmer 120, Johns Hopkins hyde and 2% glutaraldehyde. (In two monkeys only 4% Hospital, 600 N. Wolfe, Baltimore, MD 21205. paraformaldehyde was used.) The eyes were enucle-

Downloaded from iovs.arvojournals.org on 09/27/2021 396 Investigative Ophthalmology & Visual Science, February 1993, Vol. 34, No. 2

ated and immersed in the perfusion fluid for 24 hr. A cross-section of the retrolaminar optic nerve was postfixed in 1% osmium tetroxide and embedded in epoxy resin. Each eye was bisected at the equator and a 4 X 4 mm retina-choroid-sclera block with the foveola in its center was excised and postfixed in 2% osmium tetroxide. It was then placed in a mixture of equal parts of aged (2 wk at room temperature) saturated para- phenylenediamine (PPD) solution and 0.2 mol/l phos- phate buffer for 2 days at room temperature. A mix of one part 95% ethanol to two parts saturated PPD was next, and by adding small volumes of 95% ethanol every 15 min, the tissue was allowed to dehydrate grad- ually. The tissue was then immersed in a saturated solution of PPD in ] 00% ethanol for 1 hr, and for an- other hour in a saturated solution of PPD in propylene oxide. An equal volume of embedding resin was then added and shaken until it was dissolved. After 24 hr, the propylene oxide evaporated and the tissue was then embedded in the remaining blackened resin. Beginning at the nasal side of the block, semithin sections were cut until the center of the fovea was reached. Then, 50 serial 1-^rn sections were collected. At this area, the profiles of those retinal ganglion cells containing nucleoli were outlined under XI000 magnification with a camera lucida and a planimeter (Fig. 1). Cell diameter was calculated from their perimeter FIGURE 1. Retinal ganglion cells profiles in the plateau of (this was directly measured using the "length" option peak density. With the PPD block staining, cell borders are of the instrument), assuming that cells were circular. easily identified. (Original magnification XI000.) Cells of each fourth section were counted to allow a space of one nucleolus diameter (1.96 jum) between not. be used. Percentage cell loss was calculated for sections to avoid sampling a cell more than once. Cell each subgroup (bin) of cell diameter and was corre- density was calculated from the number of profiles lated to cell diameter. Selective damage of larger cells counted, section thickness, and average nucleolus size would be denoted by significant, positive slope of cell by the Abercombie formula.8 Cell density and diame- loss compared to diameter by linear regression analy- ter frequency distribution were studied in two areas: sis11 (P < 0.05). A similar analysis assessed the relation- (1) 0 to 500 jinn from the foveolar center (subserving ship of fiber loss to fiber diameter in the temporal the foveolar cones9) and (2) 800 to 1100 jum from the optic nerve sections of the same eyes. 9 foveolar center (the plateau of peak density ). Results The coefficient of determination (R2) was used to from equidistant areas above and below the foveolar estimate the amount of total variability in cell or nerve center were averaged. fiber loss that was accounted for by its linear relation- One-micron sections of the retrolaminar optic ship to cell or fiber diameter.'l For example, a correla- nerves were stained with toluidine blue. The number tion coefficient (R) of 0.7 between cell size and cell loss and diameter distribution of the fibers in the temporal means that 49% (0.72) of the variability in cell loss is half of each nerve were measured by the Zeiss IBAS related to the cell size, whereas 51% of the variability is image analysis system (Carl Zeiss, Inc., Thornwood, influenced by other factors. NY), as previously described.10 Twelve monkey eyes were studied: four normal RESULTS eyes, seven glaucomatous eyes, and one ocular hypertensive eye. Table 1 summarizes the clinical data of the In one normal eye, we measured the diameter of the seven monkey glaucomatous eyes. nucleolus in many cell profiles from vertical sections In each glaucomatous eye, cell density and diame- using a planimeter. The average nucleolus diameter ter distribution were compared to a control eye. This was 2.14 ± 0.15 ^m. This diameter did not vary among was the normal fellow eye in four monkeys, the ocular areas sampled every 100 pm between 400 to 1400 /xm hypertensive fellow eye in one, and an extensively stud- from the foveal center (Fig. 2). A tissue block of 12 ied normal eye (M7) in two eyes whose fellow eye could X 25 X 70 jum was then studied in detail. Each cell

Downloaded from iovs.arvojournals.org on 09/27/2021 Foveal Ganglion Studies 397

TABLE l. Clinical Data on the Monkey Glaucoma Eyes Duration* Percent Temporalf Monkey No. Mean IOP (±1 SD) (mo) Vertical C/D Ratio Nerve Fiber Loss

M. 26 ±2 0.6 M, 49 ± 5 0.8 M, 36 ± 5 0.7 21 M4 32 ± 6 0.3 15 M5 35 ± 1 0.8 63 Mr, 38 ±6 1.0 88 M7 45 ± 8 1.0

* Duration of elevated IOI*. t Axonal loss in the temporal half of the optic nerve. X Technical problem during post-fixation did not allow accurate counting of these optic nerve specimens. C/D, Cup/disc.

profile in serial 1 fim sections of this block were mea- The ganglion cell densities in one normal eye at sured. Cell profiles that contained nucleoli were signif- various eccentricities are shown in Figure 4. As previ- icantly larger than those that did not (P < 0.000, t-test, ously reported,9 the peak density occurs 750 to 1100 Fig. 3). Also, when profiles of the same cell were stud- firm from the foveal center. Cell diameter distribution ied, those containing nucleoli were the largest ones measured every 100 jim showed either a distinct sec- and generally were located in the center of the cell. We ond peak of large cells (Fig. 5) or a skew toward larger therefore used the diameter of a profile containing the cell diameters. nucleolus as the best estimate of the true size of a The ganglion cell densities in the foveal center and cell.1213 In profiles of the same cell, we noted that in the plateau of peak density for four normal monkey small chromatin pieces in the nucleus might be con- eyes are listed in Table 2. The coefficient of variation fused with the nucleolus. To avoid this problem, we (standard deviation divided by mean) was 7% and 9.5% did not count cell profiles that contained only a dot- for the foveal plateau and the foveal center, respec- like edge of the nucleolus. The average diameter of tively. In one glaucomatous eye, the cell diameter dis- uncounted dot-like particles measured from a magni- tribution was measured twice in a masked fashion on 2 fied print of a tissue section was 0.87 yum. With a nu- 14 consecutive samples of 50 sections. The coefficient of cleolus diameter of 2.14 /im, the calculated thickness variation was 6%, and the cell diameter distributions of a nucleolus edge with a diameter of 0.87 yum is 0.09 from the two measurements were remarkably similar. fim. Because edges of both sides were ignored, the In seven glaucomatous eyes, we measured the per- average nucleolus size in the Abercombie formula was 14 centage cell loss in the plateau of peak ganglion cell therefore corrected to 1.96 nm (2.14 - [2 X 0.09]). density, compared to normal and estimated the size dependency of cell loss in this region by linear regies-

j.u-

CO w c 2.5- D o m o • D • • B B • • B 25- D 2.0- a u • B Q) 20- E 1.5-

eg profiles o 15- CO 1.0- "o

Downloaded from iovs.arvojournals.org on 09/27/2021 398 Investigative Ophthalmology & Visual Science, February 1993, Vol. 34, No. 2

60000 TABLE 2. Density of Ganglion Cells in Normal Monkey Retina Plateau Foveal Monkey No. Region* Region*

M7 41,300 15,000 M2 36,000 19,900 M3 34,700 16,300 M8 39,200 17,700 Average 37,900 17,200 Standard deviation 2700 (7%)f 1800 (9.5%)f * Cells/mm2 rounded to the nearest 100. f Coefficient of variation. -1500 -1000 -500 0 500 1000 1500 distance from foveal center (microns) In the temporal optic nerve, there was significant, FIGURE 4. Retinal ganglion cell density along the vertical reti- selective large fiber loss in four of five glaucomatous nal meridian of normal eye of M7. The areas of the foveal center (F) and the plateaus of peak density (P) are indicated specimens (P < 0.05). The averages of the ganglion by bars. Curve fittingwa s done with two runs of a kernel of cell size data in the foveal center and the foveal plateau 80% of the data point value and 10% of each two neighbor- were compared to the nerve fiber finding in the tem- ing points' values. poral optic nerve of the same eye. There was a good correlation (R = 0.94, P < 0.01) between the amount of cell loss in the fovea and the amount of nerve fiber sion analysis (Table 3). Larger cells were significantly loss in the temporal optic nerve. There also was excel- more likely to die in five of the seven monkeys. In eyes lent correlation between the variability in foveal gan- with advanced damage, the selectively greater loss of glion cell loss that was related to cell diameter, and large cells was less prominent. A representative com- temporal optic nerve fiber loss that was related to fiber parison of the cell diameter distributions of the nor- diameter (correlation between the coefficients of de- mal and glaucomatous eyes of monkey M is shown in 3 termination: R = 0.98, P < 0.01). Figure 6. The sample size of profiles containing nucleolus in each retinal area was 238 ±141 (range, 84- 511) in the normal eyes and 121 ± 98 (range, 35-353) DISCUSSION in the glaucomatous eyes. We found a selective loss of large ganglion cells in the In the foveal center, five monkeys had significant fovea of monkeys with experimental glaucoma. As in loss of ganglion cells (Table 4). In all five eyes there the monkey retinal midperiphery,2 the selective loss was a positive correlation between cell diameter and was evident even in eyes with mild atrophy of foveal cell loss, and this was significant by linear regression ganglion cells. This suggests that testing for functional analysis in three of the five. deficits of the large ganglion cells15 subserving either the central or midperipheral field may improve the detection of initial glaucomatous damage. Asai et al reported a selective large cell loss in the fovea of two

TABLE 3. Correlation Between Ganglion Cell Size* and Cell Loss in the Plateau Region of Glaucomatous Retinas Percent Correlation Cell Loss (R) P M, 8 0.84 <0.01 My 16 0.97 <0.01 M3 33 0.92 <0.01 M4 40 0.86 <0.01 11 12 13 14 15 16 17 18 19 20 21 22 M, 43 0.74 <0.05 cell diameter (microns) M6 62 0.64 <0.1 M7 81 0.62 NS FIGURE 5. Frequency distribution of ganglion cell diameter in the plateau of peak density region of a normal retina. A * Cell size was defined as the average cell diameter. This was deter- skewed distribution toward larger cells is suggested. mined by dividing the measured cell perimeter by 3.1416.

Downloaded from iovs.arvojournals.org on 09/27/2021 Foveal Ganglion Studies 399

sity data between normal human and the normal ma- caque monkey eyes support the appropriateness of the monkey glaucoma model in the study of neural damage in human glaucoma. The coefficients of variation of cell density in the foveal center and the foveal plateau of the normal eyes were less than 10%. In the retinal midperiphery, however, we found that coefficients of variation varied from 12% at 3 mm eccentricity to 33% at 5 mm from the fovea.2 Curcio and Allen20 have reported similar findings in normal human retinas. These large individ- ual variabilities may be related to measurement error, but large variations among normal individuals may 11 12 13 14 15 16 17 18 19 20 21 22 cell diameter (microns) also be characteristic of ganglion cell density. If the latter is true, the variations could explain in part the FIGURE 6. Frequency distributions of ganglion cell diameters increase in differential light sensitivity in automated in the plateau region of a glaucoma eye (pluses) with 33% visual field testing with eccentricity from the fovea.21 cell loss, compared to die same area in its normal fellow eye (filled squares). Percentage cell loss (empty squares) is size We counted all cell profiles containing nucleoli dependent, as shown by the positive slope of the regression and did not attempt to avoid counting displaced ama- line, and significant correlation (R = 0.92, P < 0.001) be- crine cells. Displaced amacrine cells reportedly com- tween cell diameter and percentage cell damage. prise only 3% to 5% of the cells in the macular ganglion cell layer.9-20 It is therefore unlikely that remaining amacrines affected to a significant extent the cell human glaucoma eyes.16 We found selective large size distribution in the glaucomatous eyes. Further- pathway damage in areas corresponding to the retinal more, the good correspondence between the retinal midperiphery,1>3 but the methods used were not able and optic nerve selective effect lends more credence to to detect differential effects on ganglion cells from the the conclusion that the retinal data are a result of large foveal zone. In the optic nerve of monkeys with glau- ganglion cell susceptibility. coma4 and in human glaucomatous eyes,5 there is se- One may question the clinical relevance of selec- lective large nerve fiber damage in all regions. It ap- tive glaucomatous large cell loss in the fovea because it pears that larger retinal ganglion cells are more suscep- is commonly believed that foveal functions (ie, visual tible to death throughout the retina in experimental acuity) are preserved until late in the glaucomatous and spontaneous glaucoma. process. Stimulation of the retinal temporal midpe- Psychophysical studies also suggest early deteriora- riphery, for example, may therefore have a better tion of functions served by large cells in the fovea6-17'18 chance to detect early glaucomatous damage simply and in the retinal midperiphery18 of glaucomatous because of its greater regional susceptibility. Yet, the eyes. Knowledge of the characteristics of retinal gan- testing of foveal functions may have certain advan- glion cell subgroups (size, distribution, density, and tages over more peripheral visual field testing in glau- function) could guide the design of specific tests for coma. glaucoma. A test that preferentially stimulates a subgroup of ganglion cells with large cell bodies has a better chance of detecting glaucomatous damage, not TABLE 4. Correlation Between Ganglion Cell only because of their greater vulnerability, but also Size* and Cell Loss in the Foveal Region of because of their lower density. Glaucomatous Retinas The organization of the ganglion cell layer of the 919 Percent Correlation fovea is thought to be radially symmetrical. It is Cell Loss P likely, therefore, that our measurements in the vertical (R) meridia of normal eyes may represent the whole foveal M, 0 area. The peak density of ganglion cells in normal M2 28 0.54 NS monkeys has been reported to be between 30,000 to M3 29 0.89 <0.01 50,000 cells/mm2 with different histologic meth- M4 0 — — M, 63 0.41 NS ods.9-19 The peak density of 37,900 ± 2700 cells/mm2 M6 84 0.81 <0.01 reported here in the normal fellow eyes is consistent M7 88 0.81 <0.05 with these data. Curcio and Allen20 have reported a 2 * Cell size was defined as the average cell diameter. This was deter- peak density of 35,100 ± 2300 cells/mm at 1000 /im mined by dividing the measured cell perimeter by 3.1416. NS, not eccentricity in normal human retinas. The similar den- significant.

Downloaded from iovs.arvojournals.org on 09/27/2021 400 Investigative Ophthalmology 8c Visual Science, February 1993, Vol. 34, No. 2

First, although the vertical poles of the optic nerve damages large optic nerve fibers. Invest Ophthalmol Vis are more susceptible to glaucomatous damage, histo- Sci. 1987; 28:913-920. logic evidence from optic nerves and retinas of human 5. Quigley HA, Dunkelberger GR, Green WR. Chronic glaucomatous eyes does not support complete sparing human glaucoma causing selectively greater loss of of foveal zone ganglion cells in early glaucoma. Fibers large optic nerve fibers. Ophthalmology. 1988;95:357- 363. die throughout the nerve, although fewer in the hori- 6. Adams AJ, Heron G, Husted R. Clinical measures of zontal compared to the vertical poles.4 Ganglion cells central vision function in glaucoma and ocular hyper- are lost in the foveal region as well as in the retinal tension. Arch Ophthalmol. 1987; 105:782-787. midperiphery; however, detection of sensitivity loss by 7. Quigley HA, Hohman RM. Laser energy levels for tra- automated perimetry requires the loss of twice as becular meshwork damage in the primate eye. Invest many cells in the central 9° compared to the more Ophthalmol Vis Sci. 1983; 24:1305-1307. 1 peripheral retina. Each foveal cone is subserved by 8. Abercombie M. Estimation of nuclear populations two to three ganglion cells,920 presumably to enhance from microtome sections. Anal Rec. 1946; 94:239- detection of complex visual stimuli in the central field. 247. The consequently greater number of ganglion cells in 9. Wassle H, Grunert U, RoherbbeckJ, Boycott BB. Cor- the fovea allows substantial redundancy in detection tical magnification factor and the ganglion cell density of simple stimuli. Therefore, despite a significant fo- of the primate retina. Nature. 1989; 341:643-646. veal ganglion cell loss, one remaining ganglion cell per 10. Sanchez RM, Dunkelberger GR, Quigley HA. The number and diameter distribution of axons in the cone may suffice to detect a spot of light or high-con- monkey optic nerve. Invest Ophthalmol Vis Sci. trast letters. Different stimuli, however, that use func- 1986;27:1342-1350. tions subserved by larger ganglion cells, may detect 1722 11. Daniel WW. Simple linear regression and correlation. early glaucomatous damage in the foveal area. In: Daniel WW, ed. Biostatislics: A Foundation for Analy- Second, the normal variability of cell density in the sis in the Health Sciences. New York: John Wiley & Sons; central retina is less than in more peripheral areas. If a 1991:366-438. stimulus involves the response of most ganglion cells 12. Palmer RJ, Holland GR. Nucleolar eccentricity in tri- in an area for its processing, tests in the fovea may geminal ganglion neurons. J Anal. 1988; 157:163- better detect abnormality than those in the midperiph- 168. 13. Boss BD, Peterson GM, Cowan WM. On the number ery because of the small variability in central cell den- of neurons in the dentate gyrus of the rat. Brain Res. sity. 338:144-150. Third, foveal tests are abundant and simple to 14. Konigsmark BW. Methods for the counting of neu- 6 use. Most tests of peripheral functions involve map- rons. In: Nauta WJH, Ebbesson SOE eds. Con- ping of large retinal areas and require expensive temporary Research Methods in Neuroanatomy. Berlin: equipment and significant patient cooperation to Springer-Verlag; 1970:325. maintain fixation. Foveal tests, on the other hand, in- 15. Kaplan E, Lee BB, Shapley RM. New views of primate volve presentation of a stimulus at the point of maxi- retinal functions. Progress in Retina, Research. mum interest. 1990;9:273-336. 16. Asai T, Katsumori N, Mizokami K. Retinal ganglion Key Words cell damage in human glaucoma: I. studies on the so- mal diameter. Folia Ophthalmologica Japonica. 1987; cell diameter, experimental glaucoma, fovea, ganglion cells, 8:701-711. monkey. 17. Atkin A, Bodis-Wollner I, Wolkstein M, Moss A, References Podos SM. Abnormalities of central contrast sensitivity in glaucoma. AmJ Ophthalmol. 1979;88:205-211. 1. Quigley HA, Dunkelberger GR, Green WR. Studies of 18. Joffe KM, Raymond JE, Crichton A. Motion perime- retinal ganglion cell atrophy correlated with auto- try in glaucoma (abstract). Invest Ophthalmol Vis Sci. mated perimetry in human eyes with glaucoma. AmJ 1991;32(suppl):1103. Ophthalmol. 1989; 107:453-464. 19. Perry VH, Cowey A. The ganglion cell and cone distri- 2. Clovinsky Y, Quigley HA, Dunkelberger GR. Retinal butions in the monkey's retina: implications for cen- ganglion cell loss is size dependent in experimental tral magnification factors. Vision Res. 1985; 25:1 795- glaucoma. Invest Ophthalmol Vis Sci. 1991; 32:484- 1810. 491. 20. Curcio CA, Allen KA. Topography of ganglion cells in 3. Dandona L, Hendrickson A, Quigley HA. Selective human retina. J Comp Neurol. 1990; 300:5-25. effects of experimental glaucoma on axonal transport 21. Heijl A, Lindgren G, Olsson J. Normal variability of by retinal ganglion cells to the dorsal lateral geniculate static perimetric threshold values across the central nucleus. Invest Ophthalmol Vis Sci. 1991; 32:1593- visual field. Arch Ophthalmol. 1987; 105:1544-1549. 1599. 22. Tyler CW. Specific deficits of flicker sensitivity in glau- 4. Quigley HA, Sanchez RM, Dunkelberger GR, L'Her- coma and ocular hypertension. Invest Ophthalmol Vis nault NL, Baginski TA. Chronic glaucoma selectively Sci. 1981; 20:204-212.

Downloaded from iovs.arvojournals.org on 09/27/2021