THE JOURNAL OF COMPARATIVE NEUROLOGY 325~83-94 (1992)

Time-Course and Extent of Retinal Ganglion Cell Death Following Ablation of the Superior Colliculus in Neonatal Rats ALAN R. HARVEY AND DONALD ROBERTSON Departments of Anatomy and Human Biology (A.R.H.) and Physiology (D.R.), The University of Western Australia, Nedlands, Perth, Western Australia 6009, Australia

ABSTRACT This study has examined the deleterious effect of superior colliculus (SC) ablation on the viability of identified retinotectally projecting ganglion cells in the neonatal rat retina. The time-course and extent of lesion-induced retinal ganglion cell (rgc) death has been determined and an estimate obtained for the rate of clearance of individual dying neurons. In order to demonstrate the projection of rgcs to the SC and the subsequent death of these same neurons after SC lesions, the fluorescent dye diamidino yellow (DY) was injected into the left SC of anesthetized 2 day old Wistar rats (P2: day of birth = PO). DY retrogradely labels the nuclei of tectally projecting rgcs; if these identified rgcs subsequently die, their DY-labelled nuclei become pyknotic and can be visualized in retinal wholemounts. At P4 the rats were again anesthetized and the injected area, seen as ayellow patch in the SC, was removed by aspiration. Rats were perfused 2 to 336 hours after the lesion and retinal wholemounts of the right eye were prepared. Control rats received only DY injections and were perfused at times corresponding to the lesioned animals. In three sham-operated rats; the injected SC was reexposed at P4 but the tectal tissue was not removed. In each of the 42 rats that were analyzed, about 10% of the retina containing retrogradely labelled rgcs was counted; the number of pyknotic versus normally labelled rgcs was determined and changes in normal cell density were also assessed. Pyknotic rates in control and sham-operated rats were similar (average 0.896, n = 11). In SC-lesioned rats, the proportion of pyknotic DY-labelled rgcs increased to about 2.5% 4 to 8 hours postlesion (PL); the peak period of death occurred at 23 hours PL (8.0%).The amount of pyknosis decreased thereafter and most dying cells had been eliminated by 50 hours PL. Phagocytosis of dying cells was a common feature of retinae in SC lesioned rats. In the long-term (336 hours) rats, counts of normal DY-labelled rgcs in corresponding regions of control and lesioned rats revealed an average decrease in rgc density of 47.3%after P4 tectal ablation. Calculations suggest a clearance time of about 3 hours for dying neonatal rgcs. B 1992 Wiley-Liss, Inc. Key words: retinotectal projections, diamidino yellow, target removal, neurotrophic factors, apoptosis

The phenomenon of neuronal death during normal development has been extensively described in many species and in many regions of the maturing central and peripheral nervous systems (CNS and PNS) (see Cunningham, '82; Hamburger and Oppenheim, '82; Cowan et al., '84; Clarke, '85a, '90; Provis and Penfold, '88; Williams and Herrup, '88; Finlay and Pallas, '89; Oppenheim, '91 for reviews). It is widely held that this wave of naturally occurring cell death is related to regulation of cell number and the refining of morphology and connectivity in the developing CNS and PNS. Considerable evidence suggests that neurons compete for some form of target-derived factorb) during this period; experimental enlargement of target fields or application of appropriate growth factors can

o 1992 WILEY-LISS, INC.

decrease the amount of neuronal death, whereas target removal may result in the death of all cells projecting to a given region (e.g., Hamburger and Oppenheim, '82; Cowan et al., '84; Clarke, '85a, '90; Kuno, '90; Oppenheim, '91). The marked sensitivity of immature neurons to target ablation andlor axotomy, usually during the period when naturally occurring cell death is in progress, has been described in a number of regions of the CNS. These include the isthmo-optic nucleus (Sohal, '76; O'Leary and Cowan, '84; Catsicas and Clarke, '871, parabigeminal nucleus (Linden and Perry, '83; Linden and Pinon, '871, ventral horn of the spinal cord (Chu-Wang and Oppenheim, '78; Comans et Accepted July 4, 1992

84 al., '88; Kuno, 'go), and retina (Hughes and McLoon, '79; Udin and Schneider, '81; Perry and Cowey, '82; Dreher et al., '83; Carpenter et al., '86; Horsburgh and Sefton, '87; Finlay and Pallas, '89). The effects of neonatal lesions in the rat visual system have been the subject of particularly intensive study. For example, rat retinal ganglion cells (rgcs) are most sensitive to removal of the superior colliculus (SC) in the first postnatal week (Perry and Cowey, '79, '82; Dreher et al., '83; Carpenter et al., '86; Horsburgh and Sefton, '87). This period of vulnerability overlaps with the period of naturally occurring rgc death (Lam et al., '82; Potts et al., '82; Dreher et al., '83; Perry et al., '83; Crespo et al., '85; Horsburgh and Sefton, '87; McCall et al., '87) and corresponds to the phase when retinal axons are growing into the SC and establishing early retinotectal contacts (Lund and Lund, '71; Perry and Cowey, '82; Warton and McCart, '89). In most of these previous reports, analysis was limited by the fact that dying rgcs could not be distinguished from other neurons within the ganglion cell layer. The effect of SC lesions was therefore examinedpost hoe, by determining the number of rgcs that survived to adulthood. What were not studied were the dynamics of this in vivo situation, such as how soon do rgcs start to die, at what rate and for how long. This type of information is important because it might shed light on the nature of the mechanisms underlying rgc death. Such data would also serve as a baseline from which to study the effectiveness of pharmacological and other interventions in reducing or eliminating lesion-induced death of ganglion cells. In one report (Horsburgh and Sefton, '871, the number of pyknotic profiles in the ganglion cell layer was examined at various times after injections of kainic acid into the SC of 5 day old rats. Again, however, the nature of the dying cells (whether rgcs or amacrine cells) was not known. For a thorough analysis, a method is required that permits the identification of rgcs projecting to specific target sites in the brain and also allows the visualization of the death of these same neurons after target deprivation and/or axotomy. We have recently developed such a technique for studying naturally occurring cell death in the auditory (Robertson et al., '89) and visual systems (Harvey et al., '90). In the visual system, injection of the fluorescent dye diamidino yellow (DY) (Keizer et al., '83) into the SC of neonatal rats retrogradely labels the nuclei of tectally projecting rgcs. It has been shown that if these identified rgcs subsequently die, their DY-labelled nuclei become pyknotic and can be visualized in retinal wholemounts (Harvey et al., '90). With this simple method it is thus possible to demonstrate both the projection of rgcs to the SC and the death of these neurons during the course of normal development. This approach has now been used to analyze the timecourse and extent of rgc death after SC ablation in neonatal Wistar rats. The left SC was injected with DY at P2 (day of birth = PO) and the injected area, seen as a distinct yellow patch in the SC, was aspirated at P4. The 2 day delay between injection and lesion was chosen to ensure maximal retrograde labelling of ganglion cells in the contralateral retina. Lesions were carried out at P4 because it is close to the age when tectal ablation produces the greatest ganglion cell loss (Perry and Cowey, '82). In addition, most of the SC is still visible at this time, the occipital cortex not yet having grown over the midbrain. At various times after the SC lesion, rats were perfused and the proportion of pyknotic to normal DY-labelledrgcs was determined from retinal whole-

A.R. HARVEY AND D. ROBERTSON mounts and compared with the amount of death of rgcs in normal rats of corresponding age. Changes in normal rgc density were also assessed and an estimate has been made of the clearance time of cell degeneration in the retina. This work has been presented in abstract form (Harvey and Robertson, '91).

MATERIALS AND METHODS Two day old (P2) neonatal Wistar rats were anesthetized with ether and a small bone flap was opened over the left SC. The flap was cut on three sides but remained attached medially; it was folded back under the skin and was easily repositioned over the SC after completion of the DY injection. At P2, the caudal 75% or so of the tectum is accessible, lying posterior to the transverse sinus. Approximately 0.1 to 0.2 ~1 of a 2% aqueous suspension of DY (in distilled water) was pressure injected into the left SC through a glass micropipette. To optimize retrograde retinotectal labelling, care was taken to ensure that the tip of the micropipette remained relatively superficial. Excess DY which leaked out was carefully wiped away, the bone flap was pushed back over the SC, and the skin was sutured with 610 silk. At P4, rats which were to be lesioned were reanesthetized and the bone flap was located and reopened. The injection site, seen as a distinct yellow patch in the left SC, was then aspirated by gentle suction. Particular attention was given to removing all tissue containing the fluorescent material. A piece of Gelfoam was placed in the cavity and the wounds were closed as above. At times ranging from 2 to 336 hours after the tectal lesion (2,4,8,12,16,20,23,28,50,and 336 hours PL), rats were deeply anesthetized with sodium pentobarbitone (Nembutal, IP) and were perfused through the heart with 6% formaldehyde in phosphate buffer (pH 7.4). In each animal, to ensure that retinal orientation was maintained, a small cut was made nasally in the right eye before the eye was removed from the cranium. After the retinae were detached from the eye cup, a deep slit was made nasally and the retinae were mounted (ganglion cell layer uppermost) on gelatin-coated slides. After drying, retinal wholemounts were coverslipped in nonfluorescent immersion oil (Olympus) and were examined by fluorescence microscopy (peak excitation wavelength 405 nm and barrier cut-off at 455 nm). Fourteen rats which received tectal DY injections at P2 were not lesioned; these rats served as controls. Three of the animals were sham-operated, that is, at P4 they were reanesthetized, and the bone flap was opened and then closed without aspiration of the underlying SC. Control and sham-operated animals were perfused at times corresponding to the lesioned animals and retinal wholemounts were prepared as above. Regions of the wholemounted retinae containing the brightest DY labelling were chosen for photographic analysis (Fig. 1B). The ganglion cell layer of each retina was photographed through a 40 x oil immersion objective onto Ektachrome daylight or tungsten color slide film (Fig. 3C-F). In general 10 (but sometimes 11 or 12) fields were randomly selected within the area of brightest rgc label, with the proviso that, on occasion, fields were slightly adjusted to ensure that a flat area of retina was photographed. In most retinae, a total area of 0.476 mm2 was photographed and subsequently counted. Color slides were assigned a number and then randomly sorted so that the

TIME-COURSE OF RAT RETINAL GANGLION CELL DEATH

85

Fig. 1. A: Dorsal view of the midbrain showing the location of a DY injection in the left superior colliculus (SC) in a long-term (336 hour) control rat. DY injections were generally angled caudal to rostral; in this animal a DY track could he seen running anteriorly and deeper into the SC. The approximate location of the visual fields, 30" azimuth and horizontal meridian (HM) (from Sefton and Dreher, '85) are drawn onto this camera-lucida tracing. B: Wholemount of the right retina from the same animal. A cut was made in the nasal retina prior to

removal of the eye. The oval region around the optic nerve head (ON) delineates the total extent of DY label. The approximate area which was photographed and counted is indicated by the hatched region ventronasal to ON; 0.476 mm2 of retina was analyzed. C: Schematic diagram of the visual fields on the right retina; nasal (N)is to the right (after Sefton and Dreher, '85). IC, inferior colliculus; In, lower nasal field; un, upper nasal field; It, lower temporal field; ut, upper temporal field. Scale bar for both A and B = 2 mm.

observers (A.R.H. and D.R.) did not know the origin of each retinal photomicrograph. The slides were then projected onto a white screen and counts of normal and pyknotic rgcs, as well as unclassified profiles, were made. In a few fields there was some ruffling of the retina and parts of the ganglion cell layer were therefore not in focus. These regions were excluded from analysis. After photography and analysis of the fluorescent label, the coverslips were removed from the long-term (336 hour) retinae of control and lesioned rats and the retinae were stained with cresyl violet. These retinae were then recoverslipped and photographed with conventional light microscopic techniques.

visual field (Fig. 1A). Label was found in the corresponding part of the contralateral retina (Fig. lB,C). The oval region surrounding the optic nerve (ON) head in Figure 1B shows the total extent of DY labelling in this retina. Label was faint towards the periphery of this zone and brightest ventronasal to ON. This was the area that was photographed (hatched area in Fig. 1B).

RESULTS Fifty-nine P2 rats received left tectal DY injections; the right retina of 42 of these animals was photographed and the data quantitatively analyzed. Twenty-eight of the 42 rats were lesioned, 11 served as controls, and 3 were sham-operated. Data from the remaining 17 animals were not collected, either because of poor DY injections or because the retinae were not flat or contained unacceptable levels of autofluorescence. An example of a DY injection into the left SC of a control rat, and the resulting pattern of retrograde labelling in the right retina, is shown schematically in Figure 1. This animal was a long-term control, perfused at P18 (i.e., 336 hours after P4). The injection, still visible after perfusion as a yellow patch in the SC, was made in a region of the SC processing information from part of the upper temporal

Overall distribution of rgc label and sample size The majority of DY injections were made in the middle or towards the caudomedial quadrant of the SC. The pattern of rgc labelling in the retinae was generally consistent with this placement (Fig. 1A,C).Thus in 45% of cases, label was predominantly found in a wedge spreading out from around the ON towards nasal retina. In 20% of retinae, label was found in a n area surrounding the ON head and in 7.5% of retinae labelled rgcs were mostly confined to ventral retina. These patterns are consistent with DY injections and spread of label in the parts of the SC containing upper temporal visual field representation (Fig. lA,C). In 10% of rats label was primarily seen in temporal retina, indicating a more rostral SC injection. Only in 17.5% of cases was label biased towards dorsal retina (lateral SC injection). Even in these rats, counts were generally made towards the ON head. This is important for estimates of cell density (see later) since in P4 rats rgc densities are uniform except for a small area of lower density close to the superior edge of the retina (McCall et al., '87). Thus, despite some interanimal

A.R. HARVEY AND D. ROBERTSON

86

Fig. 2. Fluorescence photomicrograph of DY-labelled structures in a retinal wholemount, 20 hours postlesion. The majority of these fluorescent profiles are nuclei of retrogradely labelled rgcs. Five pyknotic rgcs can be seen (large arrows); a phagocytic cell (arrowhead) appears to be

engulfing one of these dying neurons. Other unclassified structures, probably phagocytic cells, are also visible (small arrows). Scale bar = 10

differences in the location of rgc label, comparison of DY-labelled rgc density at different times postlesion and, in particular, between control and lesion groups can be made. The proportion of retina containing DY-labelled rgcs ranged from 15 to 50%. However, in the vast majority of cases, only about 20 to 25% of the total retinal area contained retrogradely labelled cells (Fig. 1B). Retinal area (excluding the long-term control and lesioned rats) was between 22 to 28 mm2(cf. McCall et al., '87); thus about 5 to 7 mm2 of retina contained DY label. Since, in almost all cases, 0.476 mm2 of retina was photographed and counted, it can be concluded that from 7 to 10% of the DY-labelled parts of the retina were quantitatively analyzed. Moreover, much of this DY label was faint, particularly around the periphery of the labelled zone. Thus the proportion of each retina containing the brightest rgc label (presumably corresponding most closely to the injection site in the SC), which was photographed and analyzed, was considerably more than 10%in many instances.

Pyknotic rgcs. Figure 2 also shows five frankly degenerating cells (large arrows) undergoing nuclear pyknosis. The morphology of these brightly labelled pyknotic structures is identical to that seen using conventional histological techniques (see Harvey et al., '90). Classically, pyknosis-or apoptosis (Wyllie et al., '80)-involves the condensation of nuclear material into a single, homogenous sphere which subsequently breaks up into smaller spherical fragments (e.g., Hughes, '61; Silver and Hughes, '73; Hamburger et al., '81).This type of degeneration is typical of dying retinal neurons and has been described in many species either during normal development or after injury (e.g., Hughes and McLoon, '79; Cunningham et al., '81; Dunlop and Beazley, '84; Horsburgh and Sefton, '87; Provis and Penfold, '88). Of the five brightly fluorescent pyknotic rgcs shown in Figure 2, two possess a single spherical condensation (4-5 km in diameter) of DY-labelled nuclear chromatin, and the remaining three cells are in various stages of fragmentation. Retinal ganglion cells with fragmental nuclei were counted as one degenerating cell if the fragments were close together and were contained within an area similar to that of nearby nondegenerating DY-labelled nuclei (Fig. 2). This approach is the same as that used by others studying cell death using conventional staining techniques (Hughes and McLoon, '79; Linden and Pinto, '85; Dunlop and Beazley, '87; Horsburgh and Sefton, '87). Phagocytes and unclassified material. In control retinae (unlesioned animals) about 2% of the total population of DY-labelled structures could not be characterized. Some of this label may have been debris; for example, bright, but isolated and very small (1-2 km) profiles were sometimes

Appearance and classification of DY-containing profiles Normal rgcs. An example of the retrogradely labelled fluorescent structures seen in neonatal rat retinae after tectal lesions is shown in Figure 2. The majority of these profiles are DY-labelled nuclei of normal retinotectally projecting ganglion cells. The nuclei were circular or oval in shape and ranged in size from 6.5 to 11.2 km in diameter. DY was distributed more or less evenly throughout the nuclei, although the intensity of label varied from one rgc to another.

w.

TIME-COURSE OF RAT RETINAL GANGLION CELL DEATH seen but could not be classed as either normal or frankly pyknotic rgcs. Other unclassified profiles appeared more cellular; they exhibited weak sometimes blotchy fluorescence and possessed pleiomorphic, often elongated contours. Again these were not classed as normal or dying rgcs. The number of unclassified profiles in the plane of focus of the ganglion cell layer was greater in lesioned animals and increased with longer PL survival times (up to 50 hours). The proportion of the total population of DY profiles which could not be classified was about 6% at 23-28 hours and about 15% a t 50 hours PL. This increase reflected a greater amount of fluorescent debris as well as an increase in the number of pleiomorphic, often elongated cellular-like profiles. Many or most of these nonganglion cell fluorescent structures were probably phagocytic cells. Examples are shown in Figure 2 (a 20 hour PL retina). One presumed phagocytic cell (arrowhead) appears to be engulfing a pyknotic rgc, while other pleiomorphic cellular profiles can be seen (small arrows). The origin and nature of these presumed phagocytes is considered further in the Discussion.

Time-course and extent of rgc death In the following analysis, only unequivocally DY-labelled normal rgcs and frankly pyknotic profiles are considered; the unclassified fluorescent material and presumed phagocytes described in the preceding paragraph have been excluded from the data base. Control animals. A total of 11 control (DY injection only) and 3 sham-operated (DY injection plus sham lesion) rats was analyzed. Eight of the control rats were fixed on P4 or P5; the remaining three long-term controls were fixed at P18. The amount of pyknosis in the three sham-operated animals was not significantly different from the other eight short-term controls and the data have therefore been considered together. An example of a DY injection into the left SC of a 2 hour control rat is shown in Figure 3A; most of the fluorescent dye was located in the superficial layers (stratum griseum superficiale and stratum opticum). Retrograde rgc label in a 23 hour control is shown in Figure 3C. As can be seen in Figure 3C, the amount of pyknosis in the control retinae was low. On average about 3,000 DY-labelled rgcs were counted in the sample area of each control retina, of which 20-30 were clearly pyknotic. For each animal, the percentage of pyknotic rgcs was obtained by dividing the number of pyknotic cells by the total number of identified DY-labelled rgcs (normal and pyknotic) and multiplying by 100. These data are shown as the open circles in Figure 4A. The proportion of frankly pyknotic profiles ranged from 0.45 to 1.22%, the number decreasing with increasing postnatal age. A regression line is drawn through these points ( r = -0.621). The average amount of pyknosis in the P4 to P5 control and shamoperated rats (n = 11)was 0.81%. In most control retinae, DY-labelled rgcs were counted from regions around or just nasal to the ON head. There was no obvious association between the retinal area sampled and the amount of naturally occurring cell death (cf. Horsburgh and Sefton, '87). The density of normal DY-labelled rgcs was calculated from the photographed retinal sample. These data are shown for 10 of the 11control animals in Figure 4C (open circles). One retina was excluded from this analysis because of abnormal shrinkage during processing which gave rise to a very high density count. Cell densities ranged from 5,521

87

to 7,293 cells/mm2,with a mean of 6,286 cells/mm2. Cell density decreased with increasing age (the regression line in Fig. 4C has an r value of -0.73). No pyknotic profiles were seen in the three long-term (336 hour) control rats. In these older animals, perfused 16 days after the P2 DY injections, nuclear DY label in rgcs was still strong. Small, weakly fluorescent profiles were sometimes seen in the ganglion cell layer and faint label was also sometimes now seen in deeper retinal layers. Nonetheless it was possible to identify and count accurately rgcs in the wholemount material. By P18, retinal area has increased and circular isodensity lines typical of the adult rat have developed (McCall et al., '87). Because of this, we endeavoured to count DY-labelled rgcs around, or nasal to, the optic nerve head; densities ranged from 3,462 to 3,862 cells/mm2 (mean 3,694 cells/mm2) (Fig. 4C). These estimates are similar to those described by McCall et al. ('87) in corresponding regions of P20 rat retinae. The long-term control retinae were subsequently stained with cresyl violet. The typical appearance of the ganglion cell layer in these retinae is shown in Figure 3G. Lesioned animals. Rats were perfused 2 to 336 hours after removal of the DY-injected region of the SC. An example of a 2 hour PL tectum is shown in Figure 3B. Note the almost complete absence of fluorescent material in the remaining SC. Examples of DY label in the retinae of lesioned rats are shown in Figure 3D-F. Qualitatively, a slight increase in the number of pyknotic profiles was evident a t 4 hours PL (Fig. 3D), and large numbers were seen by 23 hours PL (Fig. 3E). By 50 hours PL (Fig. 3F) the number of pyknotic profiles had decreased dramatically, the density of normal rgcs had also fallen, and there was now an increase in the number of unclassified pleiomorphic profiles (presumed phagocytes-see earlier). The data for the 28 lesioned animals are presented in Figure 4 (filled circles). Twenty-five rats were perfused within 50 hours PL and three were perfused 336 hours PL (long-term lesion animals) (Fig. 4C). The percentage of pyknotic DY-labelled rgcs found at various times after tectal ablation is shown in Figure 4A. There was no association between the area of retina sampled and the postlesion age of the animals. No pyknotic profiles were identified in the three long-term lesioned rats. In Figure 4A the mean values at each time PL are joined together by a continuous line. Note an early increase in the rate of pyknosis at 4 to 8 hours PL, followed by a later phase of even greater cell death at 20 to 28 hours PL. The absolute number of pyknotic profiles counted in the photographed sample area of each lesioned retina is shown in Figure 4B. In some fields, because of unevenness in the retina, not all the photographed area was in focus and therefore could not be analyzed. The data in Figure 4B have thus been normalized and give the number of pyknotic profiles in a standard 0.476 mm2of retina. The density of normal, DY-labelled rgcs in lesioned rats was calculated from the photographed material (Fig. 4C). The densities were similar to those seen in control animals up to 8 hours PL but appeared to fall during the next 12 hours. In the three long-term (336 hours PL) rats, as in the long-term control animals, rgcs were still brightly labelled with DY but now there were also greater numbers of more weakly fluorescent profiles in the ganglion cell layer. Densities of identified rgcs counted around, or nasal to, the optic nerve head ranged from 1,625 to 2,288 cells/mm2 (mean 1,946 cellsimm2) (Fig. 4C). This was less than the densities

Fig. 3. A-F: Fluorescence photomicrographs. A: Low-power view showing a DY injection site in the left superior colliculus (SC) (2 hour control). The approximate depth of the retinorecipient layers (stratum griseum superficiale and stratum opticum) is indicated by the joined arrows. B: Low-power view of left SC 2 hours postlesion (PL), showing ablated region of the target. The tectal midline (m) is arrowed in both A and B. C-F Examples of rgc labelling in retinal wholemounts. Counts of normal and pyknotic profiles were made from fields such as these, photographed at this magnification (area of each photograph = 270 km x 180 km). C: 23 hour control. D: 4 hours PL. E: 23 hours PL. F:

50 hours PL. Note the increasing number of pyknotic profiles in the hours following target ablation. By 50 hours, the number of obviously pyknotic cells has decreased substantially, but there is now considerable debris and there are many unclassified pleomorphic profiles, most of which are likely to be phagocytic cells. G-H: Cresyl violet-stained retinal wholemounts from a long-term (336 hours) control ( G ) and lesioned (H) animal. Note the dramatic decrease in cell density in the rgc layer in the lesioned rat. Scale bars: A,B = 500 pm; C-F = 50 km; G,H = 100 km.

TIME-COURSE OF RAT RETINAL GANGLION CELL DEATH

89

0

8

-A-

*

’\ -8 - - - - - - -

I

I

I

I

I

1

1

-0 1

/A fP

/ /. .

B 200 100-

*

:

i

0 0

7-

€6000

8

8

’t t

I

t

/A/-

1

0

_ _ _ _----A0

0 -

0

0

E Q

-

8

0

C - - - 0 - -- -

-

8

1

I

8000

0

f

0

b-

0

- _- - - _ - - - -0- - _ 0

0

c z

0

0,4000

O L

0 0

-80

-0-

>

n I

I

2

4

I

1

8

12

16

20

23

28

Wk6

HOURS POST-LESION Fig. 4. Summary of the data from the 42 retinae used in this study; filled circles, lesioned animals; open circles, control and sham-operated rats. A Percentage of identified diamidino yellow (DY)-labelledretinal ganglion cells (rgcs) which were pyknotic. The short horizontal lines show the mean percentage of dying rgcs in lesioned rats at each postlesion time. The dashed line is the regression line for control and sham-operated rats ( r = -0.621). B: Absolute number of pyknotic profiles in lesioned animals at different times postlesion (for further

details see text). C: Density of normal DY-labelled rgcs in control and lesioned rats. Short horizontal lines show the mean values; the dashed line is the regression line for control and sham-operated animals (r = -0.73). In B and C, one 23 hour postlesion animal and one 8 hour control rat have been omitted due to excessive shrinkage of the retinal wholemount. This resulted in abnormally high pyknotic counts and cell densities. Details about the statistical analyses are given in the text.

A.R. HARVEY AND D. ROBERTSON

90 seen in control retinae and represented, on average, a 47.3% reduction in the number of surviving DY-labelled retinotectally projecting ganglion cells. After photography and analysis of the fluorescent material, the retinae of the 336 hour PL rats were stained with cresyl violet. The retinae appeared disorganized and the number of cells in the ganglion cell layer was clearly reduced (Fig. 3H). Although not quantified, soma sizes of the remaining rgcs in the retinae of long-term lesioned rats appeared to be relatively large (cf. Carpenter et al., '86).

Statistical analysis The errors inherent in our method of scoring pyknotic profiles were small and were generally much less than the observed interanimal variability. Assuming a random distribution of normal and pyknotic cells within the fields that were counted, the absolute error for each animal is defined by sampling statistics as the square root of the number of normal or dying cells that were counted. Thus, for the large numbers of normal DY-labelled cells the errors of estimate were small, being in the worst case only 3.5%of the number counted. For the smaller number of pyknotic profiles, the errors were of necessity larger and ranged from 7% in a 23 hour PL animal to 30% in a 16 hour control animal. When the amount of pyknosis was expressed as a percentage of the total number of DY-labelled cells (Fig. 4A), the worst case was 0.45 t 0.14% and in the best case the error of estimate was 8.2 0.58%. The data indicate an early increase in pyknosis and a later phase of much higher rates of pyknosis, peaking at about 23 hours PL (Fig. 4A). When the pooled 4 hour and 8 hour data were compared with the corresponding shams and controls, the difference was highly significant (onetailed Student's t-test, t = 4.5'7, df = 8, 0.0005 < P < 0.005). Rates of pyknosis at 2 hours PL were not significantly different from control (one-tailed t-test, t = 1.54, df = 5, 0.1 < P < 0.2). A similar comparison between the pooled 4-8 hour PL and pooled 20-23 hour PL animals showed that the difference between the rates of pyknosis in these two time windows was highly significant (t = 6.91, df = 10, P < 0.0005). Similar significant differences between control and 4-8 hour PL data (t = 3.41, P < 0.005), and 4-8 hour PL and 20-23 hour PL data (t = 5.58, P < 0.0005) were obtained when the absolute numbers of pyknotic profiles (Fig. 4B) were compared. Comparison of the density of surviving DY-labelled rgcs in control and lesioned rats revealed no significant differences at 4-8 hour PL; however, differences in cell density between controls and pooled 16, 20, and 23 hour PL rats were significant (one-tailed t-test, t = 2.71, df = 8,0.025 < P < 0.01).

*

DISCUSSION This paper documents the dynamic changes which occur in the rat retina in the first 50 hours after removal of the contralateral SC. The new and critical element in this analysis has been the use of the retrogradely transported fluorescent tracer DY, which selectively labels cell nuclei (Keizer et al., '83; Harvey et al., '90). By injecting DY into the SC at P2, and then removing the injected region 2 days later, it has been possible to label a cohort of retinotectally projecting ganglion cells and to determine, in a narrow time window, the proportion of identified rgcs that become pyknotic and die as a result of the lesion.

Previous developmental studies have shown that death of neurons in the rgc layer is almost always associated with nuclear condensation and fragmentation (pyknosis or apoptosis) followed by phagocytosis; importantly, this type of degeneration is seen during both naturally occurring and lesion-induced cell death (e.g., Hughes and McLoon, '79; Cunningham et al., '81; Miller and Oberdorfer, '81; Cunningham, '82; Sengelaub and Finlay, '82; Perry et al., '83; Dunlop and Beazley, '84, '87; Young, '84; Jenkins and Straznicky, '86; Beazley et al., '87; Horsburgh and Sefton, '87; Provis and Penfold, '88). However until now, estimates of the rates of pyknosis have almost always involved counting the number of pyknotic profiles in the ganglion cell layer without being able to distinguish between rgcs or displaced amacrine cells (Cunningham et al., '83; Perry et al., '83; Beazley et al., '87; Horsburgh and Sefton, '87).Only in the young quokka has it been shown that some of the pyknotic cells in the ganglion cell layer are rgcs with axons that grow into the brain (Dunlop and Beazley, '84). Use of the nucleophilic DY tracer is a simple way of specifically identifying rgcs and determining the proportion of these cells that are undergoing pyknosis. It is possible that injection of dye into the SC might in itself cause an increase in rgc death in the contralateral eye; however, comparison of the rates of pyknosis in the control DY-injected animals with those obtained in earlier reports suggests that any deleterious effect is minimal. In the present study, the average amount of pyknosis in the P4 to P5 control rats was 0.81%. This compares with total rates of pyknosis in the ganglion cell layer (using conventional staining techniques) of 0.6 to 1.3%in P5 rats (Cunningham et al., '81), 0.4 to 0.7%in P3 to P5 rats (Perry et al., '83; Beazley et al., '871, and 0.9 to 1.03%in P4 to P5 rats (Horsburgh and Sefton, '87). It is important to note here that the rates of pyknosis in the three rats sham-operated at P4 were not significantly different from control rats that received only DY injections. Thus the repeated anesthesia and surgery did not of itself increase rgc death.

Time-courseof lesion-induced rgc death An increase in the proportion of DY-labelledrgcs that was pyknotic (about 2.5 times control) was already apparent 4 to 8 hours PL, indicating a remarkably rapid response to target removal and damage to the terminal portions of the axons. There was a later, relatively large increase in pyknosis between 20 to 28 hours PL, peaking at about 23 hours. The proportion of pyknotic to normal DY-labelled rgcs was about 8 times the control level at this time. The lesion-induced death of rgcs was almost complete by 50 hours PL. Related changes in the density of normal DYlabelled rgcs became evident 16 to 20 hours PL. In other developing systems, the response of neurons to target removal or axotomy has a similar time-course. After amputation of the frog hind limb, Hughes ('61) described increased degeneration in the ventral horn between 1 and 2.5 days after the operation. In chick embryos death of both dorsal root ganglion cells and motorneurons is maximal about 24 hours after peripheral axotomy (Oppenheim et al., '90). In the neonatal rat visual system, cell death in the parabigeminal nucleus peaks 24 hours after bilateral tectal ablation (Pinon and Linden, '901, and intraretinal lesions cause an increase in neuronal death in the ganglion cell layer which peaks at 24 hours PL and is complete by 48 hours PL (Miller and Oberdorfer, '81). Optic nerve (ON) section at birth results in a large increase in pyknosis after

TIME-COURSE OF RAT RETINAL GANGLION CELL DEATH 24 hours and the major loss of cells in the ganglion cell layer is prior to 3 days PL (Beazley et al., '87). Finally, after tectal kainic acid injections in 5 day old rats, pyknosis in the rgc layer is maximal 8 to 12 hours after the injection and slowly declines to control levels by 48 hours (Horsburgh and Sefton, '87). This rapid response to injury seen in neonatal rats is different from that seen after axonal damage in the adult visual system (Allcutt et al., '84). After ON section, death of rgcs occurs over a much more protracted period of time; degeneration is first apparent after 8 days (Thanos, '91) and 90 to 95% of cells have died by 30 days PL (Richardson et al., '82; Villegaz-Perez et al., '88; Sievers et al., '89; Maffei et al., '90). Effects appear to be even slower after intracranial nerve crush (Misantoneet al., '84). This may be related to the fact that the injury site was some distance from the rgc somata (Lieberman, '74; Villegaz-Perez et al., '88). It is possible that, in adults, rgcs are less dependent on targetderived factors and can be temporarily supported by mature intraretinal glia or even by afferent input from other retinal neurons. Miiller cell involvement seems unlikely, since conditioned media from these cells does not significantly enhance survival of rgcs 6 days or older (Raju and Bennett, '86). The role that afferents can sometimes play in cell survival has been documented (e.g., Cunningham, '82; Linden and Perry, '82; Clarke, '85b; Furber et al., '87; Linden and Pinon, '87; Oppenheim, '91) and it may be significant that synapses in the inner plexiform layer of the rat retina are not seen before P6 but develop at or just before eye-opening (P14) (Horsburgh and Sefton, '87).

Clearance and phagocytosis A long-standing issue in developmental neurobiology is the rate of clearance of individual dying neurons; what is the period of time between the first sign of apoptotic changes and the final removal of the cellular debris by phagocytes? Estimates have been made in other systems. For example, a clearance time of 1.4 hours has been estimated for fetal mouse motorneurons (Harris-Flanagan, '69) and 3.2 hours for dying neurons in the tadpole ventral horn (Hughes, '61); the latter estimate assumed a mean, nonfluctuating rate of cell death after limb removal. More recently, Horsburgh and Sefton ('87) suggested a clearance time of 4 hours or less for neurons in the neonatal rat retina, assuming all pyknotic profiles were ganglion cells and that clearance times were similar for all ganglion cell classes. As pointed out by these authors, clearance times may also change if there is rapid and massive cell death (as happens after target lesions); in this situation availability of phagocytes could become a limiting factor (cf. Gliicksmann, '51). With these provisos in mind we have estimated the clearance time for pyknotic, DY-labelled rgcs after tectal lesions. An assumption has to be made that clearance rates are similar for all rgcs, regardless of whether they are dying as a consequence of normal development or as a result of target ablation. It is also assumed that pyknotic rgcs do not revert back to healthy rgcs with a normal appearance. At P18,16 days after the DY injections and 14 days after tectal lesions, the density of surviving DY-labelled rgcs within the labelled zone was, on average, 52.7% of that seen in comparable parts of control retinae. This represents a relative loss of about 47.3% of the ganglion cells, over and above the cell loss associated with naturally occurring cell death. Others have reported a loss of 40% and 67% of cells

91

in the ganglion cell layer after tectal ablation at P3 or P5, respectively (Perry and Cowey, '82). The proportion of pyknotic rgcs was measured at 4 hourly intervals up to 28 hours PL; for the following analysis approximate values for 32,36,40,44, and 48 hours PL have been extrapolated from the experimental data (see Fig. 4A). Because of the existence of rgc death in normal neonatal rat retinae, at each sample time the percent pyknosis in control animals (see regression line, Fig. 4A) was subtracted from the mean percent pyknosis observed in lesion animals. Performing a kind of "hazard analysis," the proportion of dying cells at each sample time has then been determined from the population remaining at the previous sample time. Thus at 4 hours PL, 1.5%of cells are pyknotic, leaving a population of 98.5%; at 8 hours PL, 1.7% of the remaining 98.5% of healthy cells are pyknotic, and so on. If the clearance time is 4 hours, coincident with the experimental sample time, then these sequential calculations should lead to an end-point of 47.3% cell death. If rgcs remain pyknotic for longer than 4 hours they will have been sampled (at least) twice and the calculations will lead to an overestimate in the number of dying rgcs. On the other hand, if the clearance time is less than 4 hours then a population of rgcs will have become pyknotic and have been removed between observation times; the calculated final proportion of dying rgcs will therefore be less than the observed 47.3%. When the analysis of the 4 hour sample times is carried out, the summed proportion of dying rgcs comes to about 37%. Such an underestimate would be expected if the clearance time for rgcs after tectal ablation was approximately 3 hours. Extrapolation of the data in Figure 4A to imaginary 3 hour observation points results in an estimated loss of 44%, which is close to the observed percentage reduction in rgcs. There are, of course, a range of assumptions and approximations in the foregoing analysis; nonetheless the data suggest strongly that pyknotic rgcs are cleared in about 3 hours, a figure similar to that reported by Hughes ('61) and Horsburgh and Sefton ('87). In normal neonatal rats, at P4 there are about 125,000 ganglion cells (Potts et al., '82; McCall et al., '87) which reduces to the adult figure of approximately 110,000 (Potts et al., '82; Perry et al., '83).Assuming a naturally occurring pyknosis rate of 1.0% at P4, reducing to about 0.3% a t P6 (extrapolation of regression line in Fig. 4A) and a clearance time of 3 hours, an analysis similar to that described above using 3 hour sample times gives a remarkably accurate estimate of the final, adult number of rgcs. Phugocytosis. Unclassified pleiomorphic structures exhibiting weak, sometimes blotchy fluorescence were seen in all retinal wholemounts, but particularly in the retinae of lesioned animals. These cellular-like structures often had an elongated shape and resembled phagocytic cells. Close association between phagocytes and degenerating pyknotic neurons has been extensively described in the developing retinae of many species using conventional staining techniques (e.g., Cunningham et al., '81; Miller and Oberdorfer, '81; Perry et al., '83; Dunlop and Beazley, '84, '87; Young, '84; Beazley et al., '86; Horsburgh and Sefton, '87). Immunohistochemical evidence in the mouse during normal development suggests that these phagocytes are macrophages which migrate from the vascular supply and subsequently differentiate into microglia (Hume et al., '83). After intraretinal section of axons, however, both macrophages and Miiller cells have been reported to contain degenerating material (Miller and Oberdorfer, '811. In the present study,

A.R. HARVEY AND D. ROBERTSON

92 the nature of the phagocytes seen in the retinae of rats with tectal lesions is not known. Immunohistochemical studies are to be carried out to investigate this issue further. Whatever their origin, the phagocytic response is extremely rapid and efficient; almost half of the labelled ganglion cell population is cleared, or is in the process of being removed, 48 hours PL. The ultimate fate of these phagocytes and how many might remain within the eye into adulthood is not known; however, there is evidence that during naturally occurring cell death many phagocytes are transformed into microglia and become resident within the adult rat retina (Thanos, '9 1).

Cause of death The nature of the lesion may have an important bearing on the extent and time-course of rgc death in the young animal. It has been reported that injections of kainic acid into the neonatal rat SC result in the death of tectal neurons but do not significantly affect retinotectal axons (Carpenter et al., '86; Horsburgh and Sefton, '87). Ganglion cells are maximally sensitive to this neurotoxin if it is applied in the first 2 postnatal days (Carpenter et al., '86). Unfortunately, the time-course of rgc sensitivity to kainic acid injection is blurred to some extent by the fact that it is not known how long the toxin remains active in the brain and furthermore, immature cells may not all develop the appropriate receptors at the same time. Nonetheless this early sensitivity to kainic acid injection does differ from ablation experiments, which involve both target removal and direct physical damage to the axons. In these latter experiments, maximal cell death (67% of total rgc population) is seen after P5 lesions (Perry and Cowey, '82). Comparison of the time-course of rgc death after P5 kainic acid lesions (Horsburgh and Sefton, '87) and after P4 tectal ablation (present study) reveals similarities and also some dissimilarities. Although rgcs were not specifically labelled in the kainic acid experiments, in both studies the amount of pyknosis was approximately 2 to 2.5 times control levels 8 to 12 hours PL. However, after tectal ablation there was a later period (20 to 28 hours PL) during which pyknotic rates were significantly higher (Fig. 4A,B). This increase was not evident in the kainic acid lesion data; indeed, in that study the number of pyknotic profiles was maximal 12 hours postinjection and slowly declined over the next 24 hours (Horsburgh and Sefton, '87). To establish to what extent this difference is related to the nature of the initial target lesion it will be necessary to measure rates of rgc death in rats which receive P2 DY tectal injections followed by injection of kainic acid into the DY-labelled part of the SC at P4. Recently, possible mechanisms by which neurons die when deprived of target-derived factors (either by competition, axotomy, or target removal) have been described. Loss of such factors, particularly during development, results in the expression of so-called suicide genes which trigger a cascade of changes resulting in the production of proteins that initiate neuronal degeneration and death (Oppenheim, '91). This death can be prevented by inhibiting protein or RNA synthesis (Martin et al., '88; Oppenheim et al., 'go), or by providing agents that increase cyclic AMP levels (Rydel and Greene, '88; Martinet al., '90) or maintain intracellular calcium concentrations (Koike et al., '89). Interestingly, after growth factor deprivation in tissue culture there is a 24 hour delay before onset of cell death (e.g., Martin et al., '88, '90). Such a delay period is consistent with the time of

maximal cell death in the present in vivo study (cf. Oppenheim et al., '90; Pinon and Linden, '90); however, an increase in pyknosis was seen as early as 4 hours PL, in both this study and after P5 kainic acid SC injections (Horsburgh and Sefton, '87). At the time of the SC ablation (P4), all rgc axons are unmyelinated and are of uniformly small diameter (Sefton and Lam, '84). The length of retinotectal axons in a P4iP5 rat is about 8 or 9 mm. Rates of transport of growth factors and other molecules have been estimated to be 2-3 mmihr (e.g., Hendry et al., '74; Schwab et al., '77; Johnson et al., '78; Ferguson et al., '90). Thus after tectal lesions, assuming continuation of retrograde axonal transport mechanisms, ganglion cells in the retina might be expected to register the loss of some central target factor about 2.5 to 3 hours later. The increase in pyknosis 4 hours PL indicates a very rapid response, and perhaps suggests a cause of death different from the cascade of gene expression described above. In this regard it will clearly be important to determine how intraocular application of inhibitors of protein or RNA synthesis affects rgc death after neonatal tectal lesions. Analysis of the comparative effects of these inhibitors on the early and late phases of rgc death will be of particular interest. A neurotrophic factor has been isolated from the rat SC which enhances the in vitro survival of rgcs (Schulz et al., '90; cf. Lehwalder et al., '89). Other growth factors, such as fibroblast growth factor (FGF) (Bahr et al., '89) and brain-derived neurotrophic factor (BDNF) (Johnson et al., '86; Johnson, '89; Thanos et al., '89) also promote rat rgc survival for limited periods of time in tissue culture. In vivo, in adult rats, a number of approaches have been used to promote rgc survival after ON section. These include 1) suturing a peripheral nerve to the ON stump (Berry et al., '88; Villegas-Perez et al., '88); 2) injecting Schwann cells intraocularly (Maffei et al., '90); 3) applying fetal neural tissue to the sectioned nerve (Sievers et al., '89; Gravina et al., '90); 4)using FGF (Sievers et al., '87) or nerve growth factor (NGF) (Carmignoto et al., '89); and 5 ) injecting protease inhibitors into the vitreous of the eye (Thanos, '91). Testing the effects of these methods in the young rat, especially the effects of in vivo application of purified growth factors (FGF, NGF, BDNF) and factors isolated from the SC itself (Schulz et al., '90; cf. Cunningham et al., '87), may also help to elucidate the mechanisms underlying the observed pattern of rgc death after neonatal tectal target ablation.

Ganglion cell survival after tectal lesions Finally, it is notable that just over half the DY-labelled rgcs appeared normal 14 days PL and there was no obvious pyknosis in the retinae at this age. These retrogradely labelled cells must have sent axons into (or perhaps through) the injected and then ablated region, yet they survived the trauma for at least 2 weeks. Thus it is clearly not the case, as has been suggested by others (Perry and Cowey, '82; cf. Udin and Schneider, '811, that the survival of rgcs after tectal lesions in these young rats is because the axons of these cells have not yet grown into the SC. Indeed pyknotic and normal DY-labelled rgcs were generally sampled and analyzed in the area of brightest label, presumably corresponding closely to the location of the initial tectal injection. Thus the majority of rgcs would have been projecting to visuotopically appropriate sites in the SC. The most parsimonious explanation is that surviving DY-labelled rgcs had already established, or were able to develop, sustaining

TIME-COURSE OF RAT RETINAL GANGLION CELL DEATH collaterals to other visual target areas in the pretectum or diencephalon or even to adjacent nonlesioned tectal tissue (cf. Fry and Cowan, '72; Lieberman, '74; Carpenter et al., '86). Perry and Cowey ('82) also invoked the presence of collateral branches to explain why, in rats P30 or older, there was no rgc death after complete tectal ablation. However, it has been estimated (Carpenter et al., '86) that only 40% of retinotectally projecting ganglion cells have axon collaterals to other nontectal target areas and thus collaterization is unlikely to be the sole factor influencing rgc survival after tectal ablation in the developing and adult brain.

ACKNOWLEDGMENTS We are indebted to Professor Ann Sefton for her helpful and constructive comments on the manuscript. We wish to thank Catharine Knights (Dept. of Behavioural Physiology, Institute of Animal Physiology and Genetics Research, Babraham, Cambridge) for her excellent secretarial assistance. We thank Dr. G. Yates and Dr. R. Owens for their advice on statistics and mathematics. The work was supported by grants from the NH & MRC and The University of Western Australia.

LITERATURE CITED Allcutt, D., M. Berry, and J. Sievers (1984) A qualitative comparison of the reactions of retinal ganglion cell axons to optic nerve crush in neonatal and adult mice. Dev. Brain Res. 16.231-240. B a r , M., J. Vanselow, and S. Thanos (1989) Ability of adult rat ganglion cells to regrow axons in. uitro can be influenced by fibroblast growth factor and gangliosides. Neurosci. Lett. 96;197-201. Beazley, L.D., J.E. Darby, and V.H. Perry (1986) Cell death in the retinal ganglion cell layer during optic nerve regeneration for the frog Rana pipiens. Vision Res. 26543-556. Beazley, L.D., V.H. Perry, B. Baker, and J.E. Darby (1987) An investigation into the role of ganglion cells in the regulation of division and death of other retinal cells. Dev. Brain Res. 33:169-184. Berry, M., L. Rees, S. Hall, P. Yiu, and J. Sievers (1988) Optic axons regenerate into sciatic nerve isografts only in the presence of Schwann cells. Brain Res. Bull. 20:223-231. Carmignoto, G., L. Maffei, P. Candeo, R. Canella, and C. Comelli (1989) Effect of NGF on the survival of rat retinal ganglion cells following optic nerve section. J. Neurosci. 9:1263-1272. Carpenter, P., A.J. Sefton, B. Dreher, and W-L. Lim (1986) Role of target tissue in regulating the development of retinal ganglion cells in the Albino rat: Effects of kainate lesions in the superior colliculus. J. Comp. Neurol. 251:240-259. Catsicas, S., and P.G.H. Clarke (1987) Abrupt loss of dependence of retinopetal neurons on their target cells, as shown by intraocular injection of kainate in chick embryos. J. Comp. Neurol. 262523-534. Chu-Wang, I.-W., and R.W. Oppenheim (1978) Cell death of motoneurons in the chick embryo spinal cord. I. Light and electron microscopic study of naturally occurring and induced cell loss during development. J. Comp. Neurol. 177:33-58. Clarke, P.G.H. (1985a) Neuronal death in the development of the vertebrate nervous system. Trends Neurosci. 8:345-349. Clarke, P.G.H. (1985b) Neuronal death during development in the isthmooptic nucleus of the chick: Sustaining role of afferents from the tectum. J. Comp. Neurol. 234:365-379. Clarke, P.G.H. (1990) Developmental cell death: Morphological diversity and multiple mechanisms. Anat. Embryol. 181:195-213. Comans, P.E., I.S. McLennan, R.F. Mark, and LA. Hendry (1988)Mammalian motoneuron development: Effect of peripheral deprivation on motoneuron numbers in a marsupial. J. Comp. Neurol. 27O:lll-120. Cowan, W.M., J.W. Fawcett, D.D.M. O'Leary, and B.B. Stanfield (1984) Regressive events in neurogenesis. Science 225:125%1265. Crespo, D., D.D.M. O'Leary, and W.M. Cowan (1985) Changes in the numbers of oDtic nerve fibres durine late Drenatal and oostnatal development in the albino rat. Dev. Brain Res. 19:129-134.

93

Cunningham, T.J. (1982) Naturally occurring neuron death and its regulation by developing neural pathways. Int. Rev. Cytol. 74:163-185. Cunningham, T.J., I.M. Mohler, and D.L. Giordano (1981) Naturally occurring neuron death in the ganglion cell layer of the neonatal rat: Morphology and evidence for regional correspondence with neuron death in superior colliculus. Dev. Brain Res. 2203-215. Cunningham, T.J., F. Haun, and P.D. Chantler (1987) Diffusible proteins prolong survival of dorsal lateral geniculate neurons following occipital cortex lesions in newborn rats. Dev. Brain Res. 37:133-141. Dreher, B., R.A. Potts, and M.R. Bennett (1983) Evidence that the early postnatal reduction in the number of rat retinal ganglion cells is due to a wave of ganglion cell death. Neurosci. Lett. 36255-260. Dunlop, S.A.,and L.D. Beazley (1984) Development of the area centralis and visual streak in mammals. In J. Stone, B. Dreher, and D.H. Rapaport (eds): Development of Visual Pathways in Mammals. New York: Alan R. Liss, pp. 75-88. Dunlop, S.A., and L.D. Beazley (1987) Cell death in the developing retinal ganglion cell layer of the wallaby Setoniz brachyurus. J. Comp. Neurol. 264: 14-23. Ferguson, I.A., J.B. Schweitzer, and E.M. Johnson, Jr. (1990) Basic fibroblast growth factor: Receptor-mediated internalization, metabolism, and anterograde axonal transport in retinal ganglion cells. J. Neurosci. 10:2 176-2 189. Finlay, B.L., and S.L. Pallas (1989) Control of cell number in the developing mammalian visual system. Prog. Neurobiol. 32207-234. Fry, F.J., and W.M. Cowan (1972) A study of retrograde cell degeneration in the lateral mamilliary nucleus of the cat, with special reference to the role of axonal branching in the preservation of the cell. J. Comp. Neurol. 144: 1-24. Furber, S., R.W. Oppenheim, and D. Prevette (1987) Naturally occurring neuron death in the ciliary ganglion of the chick embryo following removal of preganglionic input: Evidence for the role of afferents in ganglion cell survival. J. Neurosci. 7:1816-1832. Glucksmann, A. (1951) Cell deaths in normal vertebrate ontogeny. Biol Rev. 26: 59-86. Gravina, A,, L. Domeniei, N. Berardi, L. Galli, and L. Mafei (1990) Transplant of embryonal nervous tissue preserves the responses of rat retinal ganglion cells after section of the optic nerve. Exp. Brain Res. 80:631-634. Hamburger, V., and R.W. Oppenheim (1982) Naturally occurring neuronal death in vertebrates. Neurosci. Comment. 1:39-55. Hamburger, V., J.K. Brunso-Bechtold, and J.W. Yip (1981) Neuronal death in the spinal ganglia of the chick embryo and its reduction by nerve growth factor. J. Neurosci. 1:60-71. Harris-Flanagan, A.E. (1969) Differentiation and degeneration in the motor horn of the fetal mouse. J. Morphol. 129:281-306. Harvey, A.R., and D. Robertson (1991) Time-course of death of identified retinal ganglion cells following ablation of the superior colliculus in neonatal rats. Eur. J. Neurosci. Suppl. 4.292. Harvey, A.R., D. Robertson, and K.S. Cole (1990) Direct visualization of death of neurons projecting to specific targets in the developing rat brain. Exp. Brain Res. 80:213-217. Hendry, LA., K. Stockel, H. Thoenen, and L.L. Iversen (1974) The retrograde axonal transport of nerve growth factor. Brain Res. 68:103-121. Horsburgh, G.M., and A.J. Sefton (1987) Cellular degeneration and synaptogenesis in the developing retina of the rat. J. Comp. Neurol. 263:553566. Hughes, A. (1961) Cell degeneration in the larval ventral horn of Xenopus laeuis (Daudin). J. Embryol. Exp. Morph. 9:269-284. Hughes, W.F., and S.C. McLoon (1979) Ganglion cell death during normal retinal development in the chick: Comparisons with cell death induced by early target field destruction. Exp. Neurol. 66:587-601. Hume, D.A., V.H. Perry, and S. Gordon (1983) Immunohistochemical localization of a macrophage-specific antigen in developing mouse retina: Phagocytosis of dying neurons and differentiation of migroglial cells to form a regular array in the plexiform layers. J. Cell Biol. 97.253-257. Jenkins, S., and C. Straznicky (1986) Naturally occurring and induced ganglion cell death: A retinal whole-mount autoradiographic study in Xenopus. Anat. Embryol. 174t59-66. Johnson, E.M., Jr., R.Y. Andres, and R.A. Bradshaw (1978) Characterization of the retrograde transport of nerve growth factor (NGF) using high specific activity 112511NGF.Brain Res. 150:319-331. Johnson, J.E. (1989) Retinal cultures used to assay CNS neurotrophic factors. In R.A. Rush (ed): Nerve Growth Factors. New York: John Wilev &Sons, pp. 111-133

94 Johnson, J.E., Y-A. Barde, M. Schwab, and H. Thoenen (1986) Brain-derived neurotrophic factor supports the survival of cultured rat retinal ganglion cells. J. Neurosci. 6:3031-3038. Keizer, K.. H.G.J.M. Kuypers, A.M. Huisman, and 0. Dann (1983) Diamidino yellow dihydrochloride (DY.2HCL): A new fluorescent retrograde neuronal tracer, which migrates only very slowly out of the cell. Exp. Brain Res. 51r179-191. Koike, T., D.P. Martin, and E.M. Johnson, J r . (1989) Role of Ca2+channels in the ability of membrane depolarization to prevent neuronal death induced by trophic factor deprivation: Evidence that levels of Cay+ determine nerve growth factor dependence of sympathetic ganglion cells. Proc. Natl. Acad. Sci. USA 86:6421-6425. Kuno, M. (1990) Target dependence of motoneuronal survival: The current status. Neurosci. Res. 9:155-172. Lam, K., A.J. Sefton, and M.R. Bennett (1982) Loss of axons from the optic nerve of the rat during early postnatal development. Dev. Brain Res. 3:487-491. Lehwalder, D., P.L. Jeffrey, and K. Unsicker (1989) Survival of purified embryonic chick retinal ganglion cells in the presence of neurotrophic factors. J. Neurosci. Res. 24:329-337. Lieberman, A.R. (1974) Some factors affecting retrograde neuronal responses to axonal lesions. In R. Bellairs and E.G. Gray (eds): Essays on the Nervous System. Oxford: Clarendon Press, pp. 71-105. Linden, R., and V.H. Perry (1982) Ganglion cell death within the developing retina: A regulatory role for retinal dendrites? Neuroscience 7:28132827. Linden, R., and V.H. Perry (1983) Retrograde and anterograde-transneuronal degeneration in the parabigeminal nucleus following tectal lesions in developing rats. J. Comp. Neurol. 218r270-281. Linden, R., and L.H. Pinto (1985) Developmental genetics of the retina: Evidence that the pearl mutation in the mouse affects the time course of natural cell death in the ganglion cell layer. Exp. Brain Res. 60:79-86. Linden, R., and L.G.P. Pinon (1987) Dual control by targets and afferents of developmental neuronal death in the mammalian central nervous system: A study in the parahigeminal nucleus of the rat. J. Comp. Neurol. 266:141-149. Lund, R.D., and J.S. Lund (1971) Modifications of synaptic patterns in the superior colliculus of the rat during development and following deafferentiation. Vision Res. (Suppl.) 3:281-298. Maf€ei, L., G. Carmignoto, V.H. Perry, P. Candeo, and G. Ferrari (1990) Schwann cells promote the survival of rat retinal ganglion cells after optic nerve section. Proc. Natl. Acad. Sci. USA 87:1855-1859. Martin, D.P., R.E. Schmidt, P.S. DiStefano, O.H. Lowry, J.G. Carter, and E.M. Johnson, J r . (1988) Inhibitors of protein synthesis and RNA synthesis prevent neuronal death caused by nerve growth factor deprivation. J. Cell Biol. 106:829-844. Martin, D.P., T.L. Wallace, and E.M. Johnson, J r . (1990) Cytosine arabinoside kills postmitotic neurons in a fashion resembling trophic factor deprivation: Evidence that a deoxycytidine-dependent process may be required for nerve growth factor signal transduction. J. Neurosci. 10: 184-193. McCall, M.J., S.R. Robinson, and B. Dreher (1987) Differential retinal growth appears to be the primary factor producing the ganglion cell density gradient in the rat. Neurosci. Lett. 79:78-84. Miller, N.M., and M. Oberdorfer (1981) Neuronal and neuroglial responses following retinal lesions in the neonatal rats. J. Comp. Neurol. 202493504 Misantone, L.J., M. Gershenhaum, and M. Murray (1984) Viability of retinal ganglion cells after optic nerve crush in adult rats. J. Neurocytol. 13:449465. O'Leary, D.D.M., and W.M. Cowan (1984) Survival of isthmo-optic neurons after early removal of one eye. Dev. Brain Res. 12.293-310. Oppenheim, R.W. (1991) Cell death during development of the nervous system. Annu. Rev. Neurosci. 14:453-501. Oppenheim, R.W., D. Prevette, M. Tytell, and S. Homma (1990) Naturally occurring and induced neuronal death in the chick embryo irt uiuo requires protein and RNA synthesis: Evidence for the role of cell death genes. Dev. Biol. 138:104-113. Perry, V.H., and A. Cowey (1979) The effects of unilateral cortical and tectal lesions on retinal ganglion cells in rats. Exp. Brain Res. 3585-95. Perry, V.H., and A. Cowey (1982) A sensitive period for ganglion cell degeneration and the formation of aberrant retino-fugal connections following tectal lesions in rats. Neuroscience 7:583-594.

A.R. HARVEY AND D. ROBERTSON Perry, V.H., Z. Henderson, and R. Linden (1983) Postnatal changes in retinal ganglion cell and optic axon populations in the pigmented rat. J. Comp. Neurol. 219:356-368. Pinon, L.G.P., and R. Linden (1990) Interactive effects of deafferentation and target removal on cell death in the parabigeminal nucleus of developing rats. Brazilian J. Med. Biol. Res. 23:659-662. Potts, R.A., B. Dreher, and M.R. Bennett (1982)The loss of ganglion cells in the developing retina of the rat. Dev. Brain Res. 3:481486. Provis, J.M., and P.L. Penfold (1988)Cell death and the elimination of retinal axons during development. Prog. Neurobiol. 31:331-347. Raju, T.R., and M.R. Bennett (1986) Retinal ganglion cell survival requirements: A major but transient dependence on Miiller glia during development. Brain Res. 383:165-176. Richardson, P.M., V.M.K. Issa, and S.Shemie (1982) Regeneration and retrograde degeneration of axons in the rat optic nerve. J. Neurocytol. 11:949-966. Robertson, D., A.R. Harvey, and K.S. Cole (1989) Postnatal development of the efferent innervation of the rat cochlea. Dev. Brain Res. 47r197-207. Rydel, R.E., and L.A. Greene (1988) CAMP analogs promote survival and neurite outgrowth in cultures of rat sympathetic and sensory neurons independently of nerve growth factor. Proc. Natl. Acad. Sci. USA 85:1257-1261. Schulz, M., T. Raju, G. Ralston, and M.R. Bennett (1990) A retinal ganglion cell neurotrophic factor purified from the superior colliculus. J. Neurochem. 55832-841. Schwah, M., Y. Agid, J. Glowinski, and H. Thoenen (1977) Retrogradeaxanal transport of "51-tetanus toxin as a tool for tracing fiber connections in the central nervous system; connections of the rostra1 part of the rat neostriatum. Brain Res. 126:211-224. Sefton, A.J., and K. Lam (1984) Quantitative and morphological studies on developing optic axons in normal and enucleated albino rats. Exp. Brain Res. 57t107-117. Sefton, A.J., and B. Dreher (1985) Visual system. In G. Paxinos (ed):The Rat Nervous System. Sydney: Academic Press, pp. 169-221. Sengelaub, D.R., and B.L. Finlay (1982) Cell death in the mammalian visual system during normal development I. Retinal ganglion cells. J. Comp. Neurol. 204:311-317. Sievers, J., B. Hausmann, and M. Berry (1989) Fetal brain grafts rescue adult retinal ganglion cells from axotomy-induced cell death. J. Comp. Neurol. 281:467478. Sievers, J., B. Hausmann, K. Unsicker, and M. Berry (1987) Fibroblast growth factors promote the survival of adult rat retinal ganglion cells after transection of the optic nerve. Neurosci. Lett. 76:157-162. Silver, J., andA.F.W. Hughes (1973) The roleofcell death duringmorphogenesis of the mammalian eye. J. Morphol. 140:159-170. Sohal, G.S. (1976) Effects of deafferentation on the development of the isthmo-optic nucleus in the duck. Exp. Neurol. 50:161-173. Thanos, S. (1991) The relationship of microglial cells to dying neurons during natural neuronal cell death and axotomy-induced degeneration of the rat retina. Eur. J. Neurosci. 3:1189-1207. Thanos, S.,M. Bahr, Y-A. Barde, and J. Vanselow (1989) Survival and axonal elongation of adult rat retinal ganglion cells. In uitro effects of lesioned sciatic nerve and brain derived neurotrophic factor. Eur. J. Neurosci. I : 19-26. Udin, S.B., and G.E. Schneider (1981) Compressed retinotectal projection in hamsters: Fewer ganglion cells project to tectum after neonatal tectal lesions. Exp. Brain Res. 43t261-269. Villegas-Perez, M.P., M. Vidal-Sanz, G.M. Bray, and A.J. Aguayo (1988) Influences of peripheral nerve grafts on the survival and regrowth of axotomized retinal ganglion cells in adult rats. J. Neurosci. 8265-280. Warton, S.S., and R. McCart (1989) Synaptogenesis in the striatum griseum superficiale of the rat superior colliculus. Synapse 3: 136-149. Williams, R.W., and K. Herrup (1988) The control of neuron number. Annu. Rev. Neurosci. 11:423453. Wyllie, A.H., J.F.R. Kerr, and A.R. Currie (1980) Cell death: The significance of apoptosis. Int. Rev. Cytol. 68:251-306. Young, R.W. (1984) Cell death during differentiation of the retina in the mouse. J. Comp. Neurol. 229:362-373.