Apoptosis in substantia nigra following developmental striatal excitotoxic injury

Proc. Nati. Acad. Sci. USA Vol. 91, pp. 8117-8121, August 1994 Neurobiology

Apoptosis in substantia nigra following developmental striatal excitotoxic injury (programmed cell death/quinolinate/dopaminergic neuron/Parkinson disease)

ALFONS MACAYA*t, FRANCINA MUNELL*, RUTH M. GUBITS*, AND ROBERT E. BURKEtt *Division of Pediatric Neurology and tDepartment of Neurology, Columbia University, New York, NY 10032 Communicated by Dominick P. Purpura, April 28, 1994 (received for review November 9, 1993)

in the size of the striatum, the target of these neurons, associated with an induced cell death event in SN. Two types of observation support this hypothesis. (i) Naturally occurring cell death is observed during postnatal development of the SN pars compacta (SNpc) (8). (ii) Many prior in vitro studies have shown effects of striatal preparations on the development of the SN (9-11). However, there are numerous other possible explanations for the observation of reduced dopaminergic SN neurons. Hypoxia-ischemia or excitotoxic injury to striatum may in some way directly injure SN cells, or following early striatal injury, with the resulting loss of a major y-taminobutyratergic projection to SN pars reticulata (SNpr), there may result an anterograde transynaptic degeneration of SNpr and SNpc "en cascade" (12). To differentiate among these possibilities, we have sought to identify the presence of cell death and its morphologic and biochemical characteristics in SN following a developmental striatal excitotoxic lesion made with quinolinic acid (QA). We have identified and quantified cell death in these structures by using a silver stain technique (13, 14). A frequent morphologic appearance of cells undergoing induced or natural cell death is that of apoptosis, which can be differentiated from cell necrosis due to direct excitotoxic injury at the ultrastructural level on the basis of several distinct features (15). Apoptosis can also frequently be distinguished from necrosis on the basis of genomic DNA fragmentation studies, which demonstrate a characteristic appearance on DNA gels of integral multimers of 180- to 200-bp fragments (a DNA "ladder") (16). We have, therefore, characterized the pattern of DNA fragmentation, as well as morphologic alterations, in SN following early striatal excitotoxic injury.

We have previously observed that an axonABSTRACT sparing injury to the developing striatum induced by the excitotoxin quinolinate results in a decrease in dopaminergic neurons in the substantia nigra pars compacta (SNpc) of the adult. This decrease occurs in the absence of direct injury to the SNpc. As the striatum is a major target for the SNpc dopaminergic system, we have hypothesized that a decrease in the size of the striatal target during development may result in an induced regressive event in the SNpc, similar to what has been described for many developing neural systems with peripheral targets. We have examined by morphologic and biochemical means the time course and character of cell death in SN following a unilateral striatal lesion with quinolinate in immature rats. The striatal lesion is associated with an induced cell death event in the Ipsilateral SN, observed first in SNpc and then in SN pars reticulata. The morphologic characteristics of the dying cells were typical of apoptosis. Immunotaining for tyrosine hydroxylase indicated that some of the apoptotic cells in the SNpc were dopmnergic. We conclude that developmental striatal excitotoxic injury is associated with induced apoptotic cell death in SN.

During development, many neural nuclei initially contain more neurons than are present in the mature animal. The ultimate number of neurons is determined by the magnitude of a regressive event in which natural cell death occurs (1). This regressive event is regulated in many instances by the contact of the nuclear group with its target structure (2-4). Experimental manipulations which alter this target-derived support, such as a decrease in target size, result in an augmented regressive event in the developing nucleus, with fewer neurons surviving into adulthood. Many of these concepts derive from studies of the prototypic growth factor, nerve growth factor (5). Most of the evidence supporting concepts related to target support derives from studies of neural structures which project to peripheral targets, particularly spinal motor and peripheral neural ganglia, and much less is known about neural groups with central projections. Studies of the isthmo-optic nucleus in chicken, which projects to the retina (3), and cerebellar granule cells, which project to Purkinje cells, support the concept of central target regulation of neural development (6), but relatively less is known about these mechanisms. We have previously observed that following axon-sparing injury of the striatum during development, due either to hypoxia-ischemia or intrastriatal injection of an excitotoxin, there is an associated decrease in the adult number of dopaminergic neurons in the substantia nigra (SN) (7). This decrease occurs in the absence of any morphologic evidence for direct injury to the SN. We have postulated that this decrease in dopaminergic neurons may be due to a reduction

EXPERIMENTAL PROCEDURES Striatal Lesions with QA. Female rats 14-16 days pregnant were obtained from Charles River Breeding Laboratories. On postnatal day 7 (P7), pups of either sex received intrastriatal injection of 480 nmol of QA as described (7). Littermate pups were injected with phosphate-buffered saline alone as controls. Light Microscopy. At various times after QA infusion, pups were perfused with 4% paraformaldehyde in 0.1 M phosphate buffer (pH 7.1). After postfixation, the brain was rapidly frozen and serially sectioned at 30 pm through striatum and SN. Alternate sections were silver (8, 13) or Nissl stained. Quantitative Morphology. At least two silver-stained sections from each ofthree Paxinos-Watson (17) planes, 4.2, 3.7, and 3.2 (interaural), were analyzed by scanning the entire SN on both sides of the section with a x 40 objective. The average number of cells in each plane was added to provide a measure of the number of degenerating cells for that side of the SN for Abbreviations: Pn, postnatal day n; QA, quinolinic acid; SN, substantia nigra; SNpc, SN pars compacta; SNpr, SN pars reticulata; TH, tyrosine hydroxylase. tTo whom reprint requests should be addressed at: Neurological Institute, Box 67, 710 West 168th Street, New York, NY 10032.

The publication costs of this article were defrayed in part by page charge payment. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. §1734 solely to indicate this fact. 8117

8118

Proc. Natl. Acad. Sci. USA 91

Neurobiology: Macaya et al.

each animal. The cross-sectional areas of degenerating cells were determined on an Amersham RAS-3000 image analyzer. Immunostaining for Tyrosine Hydroxylase (TH). At 24 hr after striatal QA injury, rat pups were perfusion-fixed with 4% paraformaldehyde in 0.1 M phosphate buffer (pH 7.1). TH immunostaining was demonstrated with a mouse monoclonal anti-TH antibody (Boehringer Mannheim). The number of TH-positive neurons undergoing apoptosis was determined by scanning representative sections with a x60 objective and counting the number of cells with TH-positive cytoplasm and dark blue Nissl-stained chromatin clumps, characteristic of apoptosis (8). Electron Microscopy. At 24 or 48 hr after intrastriatal infusion of QA, rat pups were perfused intracardially with O.9o NaCl followed by 1% paraformaldehyde/1.5% glutaraldehyde/0.1 M phosphate buffer, pH 7.4, for 10 min. A tissue block encompassing the midbrain was taken and processed as described (18). DNA Fragmentation Analysis. At 2, 6, 12, 24, and 48 hr after striatal QA injection or at 2, 6, and 24 hr after vehicle injection, pups were sacrificed and the SN on each side was microdissected and frozen in liquid N2. DNA was extracted by standard procedures (ASAP kit, Boehringer Mannheim). Southern blot analysis was performed using the conditions of Wahl et al. (19). In Situ End Labeling. Free 3'-hydroxyl ends generated by endonuclease cleavage of genomic DNA during apoptosis were labeled with a commercial kit (ApopTag, Oncor) based on a method similar to that of Gavrieli et al. (20), but which utilizes digoxigenin-11-dUTP as label.

RESULTS Quantitation of Cell Death and Light Microscope Observadons. Silver and Nissl staining of striatal sections revealed a unilateral lesion in the striatum, although some cell death was also observed in overlying cortex and in the medial septal region. Following unilateral lesion of the striatum at P7 with QA, an increase in the number of dying cells revealed by silver stain was observed in both SNpc and SNpr (Fig. 1). In SNpc, a clear increase was observed 12 hr after lesion, and by 24 hr there was a statistically significant increase in comparison to the contralateral side (P < 0.0001, ANOVA). This difference persisted at 48 and 72 hr in comparison to both the contralateral side and vehicle-injected controls. In SNpr, a difference between injected and contralateral sides did not become apparent until 24 hr and did not achieve significance until 48 hr. In absolute terms, the magnitude of cell death was about 2-fold greater in SNpc than in SNpr. However, for this developmental period, the relative increase in degenerating cells in the two structures was about the same, =10-fold. Morphometric analysis of all of the silver-stained degenerating cells in a representative section from three of the four rats examined 48 hr postlesion revealed no difference in the mean cross-sectional area of degenerating cells on the lesioned side (36.3 ± 1.3 Am2; n = 110) in comparison to the contralateral control side (36.7 ± 3.4 "m2; n = 14). A typical example of induced developmental cell death in SNpc and SNpr is shown in Fig. 2A on the side of the striatal lesion in comparison to the contralateral, nonlesioned side in Fig. 2B. At the light microscope level, the morphologic appearance of the degenerating cells on the lesioned side was typical of that for apoptosis (21), with the formation of a shrunken, rounded cytoplasm and multiple, rounded, dark-staining chromatin clumps (Fig. 2C). The degenerating cells on the lesioned side were identical in appearance to the few degenerating cells on the intact side and to cells previously described during natural developmental cell death in rat SNpc (8).

(1994)

A .

z C

Control side, QA Experimental QA

aRiHht 'de, PBSside, . Lef side, PBS P < 0.0001, ANOVA *

* & ~ I

70

o 50

*8 30

E z

10

0

1

2

3

4

5

6

7

8

9 10 11 12 13 14

Post-Lesion Day

B

50

* * Z 40

h

30

0

20

E

0

1

2

3

4

5

6 7 8 9 10 11 12 13 14 Post-Lesion Day

FIG. 1. (A) Number of pyknotic cells in SNpc at various times following a striatal lesion made with 480 nmol of QA on P7. Thirty-eight rats were studied (n = 3 or 4 at each time point). Each point represents the mean ± SEM. The difference between the injured side and the contralateral side first became significant at 24 hr and remained so until % hr. Control rats received vehicle without QA (phosphate-buffered saline, PBS). (B) Number of pyknotic cells in SNpr at different time points following a striatal lesion. A significantly increased number of degenerating cells was first observed at 48 hr.

Numerous neurons in SNpc on the side of striatal injury were TH-positive and undergoing apoptosis (Fig. 3). Among five animals at 24 hr following QA injection there were 3.4 ± 0.9 apoptotic TH-positive neurons per section on the ipsilateral side and 0.4 ± 0.2 on the contralateral side (P = 0.02, Wilcoxon). Ultrastructural Observations. Ultrastructural analysis of multiple degenerating cells in SNpc and SNpr on the side of striatal injury at 24 and 48 hr after the lesion revealed features typical of those described for apoptosis (21) (Fig. 4). Such cells were about 10-fold more numerous on the side of striatal injury than on the nonlesioned side. DNA Fragmentation Patterns. At multiple time points following intrastriatal injection of QA or vehicle, DNA fragmentation analysis demonstrated a ladder of discrete nucleosome-sized DNA multimers characteristic of apoptosis in some, but not all, animals on both the lesioned and contralateral control sides (data not shown). Detection of the multimeric fragments in these samples required prolonged exposure of radiolabeled Southern blots (Fig. 5C). The pattern was rarely visible in brain tissue samples by UV illumination of ethidium bromide-stained gels (Fig. SA), or after brief exposures of Southern blots (Fig. SB). We did not observe a consistently increased intensity of nucleosomal fragments on the lesioned side in comparison to either the contralateral control side (Fig. 5) or in comparison to vehicleinjected controls (data not shown). Thus, while the qualitative pattern of DNA fragmentation on the injured side was like that observed in the presence of basal levels of developmental cell death, we did not observe a quantitative alteration in the intensity of the banding pattern.

Neurobiology: Macaya et A

Proc. Natl. Acad. Sci. USA 91 (1994)

8119

FIG. 2. (A) Silver stain of SN 48 hr after ipsilateral striatal injection of 480 nmol of QA. Numerous silver-stained, degenerating cells are observed in SNpc and SNpr. (Bar = 10 pum.) A single cell (arrow) is shown in higher magication in C. (B) Contralateral side of the same SN section shown in A. A single degenerating cell is observed in SNpc (arrow). (C) A single representative cell in SN on the side of injury shows features characteristic of apoptosis: a rounded, shrunken cytoplasm and multiple, rounded, dark-staining chromatin clumps. (Bar = 10 ,um.) (D) A similar cell is observed on the contralateral control side.

In Situ End Labeling. Striatal injury with QA at P7 was associated with an increased number of labeled cells in SN as assessed by the in situ end-labeling technique. In six SN sections (two from each of three lesioned pups) there were 15.1 ± 5 labeled cells ipsilateral and 1.5 ± 1.3 labeled cells contralateral to the side of striatal injury (P = 0.01, Wilcoxon). The morphology of the labeled cells was typical for that observed in apoptosis: labeled chromatin material often appeared as discrete clumps as shown in Fig. 6A.

DISCUSSION Our analysis indicates that there is increased cell death in both SNpc and SNpr following a striatal QA lesion at P7. The relative magnitude of the augmentation in cell death was approximately equal at its maximum in SNpc and SNpr, at

FIG. 3. Immunoperoxidase staining for TH, with Nissl counterstain, in the SNpc at 24 hr following striatal QA lesion. A single TH-positive neuron is seen with four Nissl-stained chromatin clumps in the nucleus, typical of apoptosis (large arrow). The blue Nisslstained, rounded structures (small arrow) represent normal stained nuclei of TH-negative cells. (Bar = 10 pxm.)

10-fold above basal levels of ongoing natural cell death. It is unknown whether the longer duration and greater absolute magnitude of cell death in SNpc translate into a greater relative loss of cells in comparison to SNpr. Such a conclusion would be possible only if we knew that the period oftime for which dying cells take up the silver stain and persist in tissue was the same in SNpc and SNpr. To determine the effect of early striatal injury on the ultimate number of living neurons in the two regions will require direct quantification ofthe final populations. The increase in number of dying cells in SNpc and SNpr which we have observed was not due to an increase in their size, with a resulting increase in double counting error (22). By both morphologic and biochemical criteria, the cell death in SN following striatal injury appears to be apoptosis (21). We consider the term apoptosis to be a designation for a specific morphologic pattern of death observed in programmed cell death, and it is only one of several morphologies observed (23). We consider morphologic criteria for apoptosis to be primary because historically the term was used to refer to a specific morphologic pattern (21) and because biochemical criteria for apoptosis [185-bp fragmentation of DNA (24)] are not always fulfilled even in its presence (25). The morphologic characteristics of induced cell death in SN at the light microscope level included a shrunken, rounded appearance of the cytoplasm and the formation of multiple, rounded, dark-staining chromatin clumps (21). This identical appearance was observed on the contralateral, noninjected side (see Fig. 2) in controls and in normal rats during development (8). The appearance of apoptosis was confirmed at the ultrastructural level in dying SN cells on the side of striatal injury. The major features included (i) formation of rounded, electron-dense chromatin clumps; (ii) preservation of both the nuclear and cellular membranes; (iii) preservation of intracellular organelles; and (iv) phagocytosis of apoptotic bodies by surrounding healthy cells. Immunostaining for TH demonstrated that some of the dying cells in the SNpc were dopaminergic.

8120

Neurobiology: Macaya et al.

u,.0Ks A}^

Proc. Natl. Acad. Sci. USA 91

(1994)

.

.;

1)

V.-

1.1 1,

.j

,

.N

i.)

X

,,,

i;.

s >

I:

_ i At

*.wS 8,,

And..

t >

r

k

%.

www

O'AL

i,

I

m

Ir ru

IS

1;

E ('*.'T He

IS

.-

A, is

J

jo

.".

" :.

6

_t:s frJ ;;' 4

D A_

B ^. ,.

a

$ * * tab

i..: 1...

FIG. 4. Ultrastructure of degenerating cells in SNpc and SNpr following striatal lesion with 480 nmol of QA on P7. (A) A degenerating cell in SNpc at 24 hr following striatal lesion. The cytoplasm of the cell is rounded and relatively electron dense. The nuclear membrane is intact and shows marked infolding on either side. There is a single, rounded, electron-dense chromatin clump. (B) The same cell in A is shown at higher magnification. Two arrows point out intact mitochondria with normal-appearing cristae; one is marked by an arrowhead on the right. Intact endoplasmic reticulum is also observed. (C) A second degenerating cell in SNpc at 24 hr following striatal lesion is shown with three rounded, electron-dense chromatin clumps. (D) An electron-dense apoptotic body is observed phagocytized by a perineuronal microglial cell 48 hr after striatal injury. (Bar in B = 1 ,pm; bar in D = 1 Aum in A, C, and D.)

DNA fragmentation studies revealed bands corresponding to multimers of 180-200 bp in microdissected SN tissue from both sides of the brain in vehicle-injected animals at P7 and P8 (data not shown) and on both sides of the SN following striatal excitotoxic injury. We interpret this finding to mean that the qualitative pattern of DNA fragmentation is similar on both sides and is typical of that for apoptosis. We were unable to demonstrate a quantitative increase in the intensity of nucleosomal fiagments on the side of induced cell death, possibly because the increase in apoptotic fragmentation taking place in SN may be minimal in relation to the amount of fragmented DNA in surrounding tissue, which is not affected by the striatal lesion. In addition, while the gel is useful for qualitative demonstration of DNA fragmentation, it is not necessarily quantitative. Use of the in situ endlabeling technique for the demonstration of free 3'-hydroxyl ends generated by endonuclease cleavage permitted us to demonstrate not only the characteristic morphology of apoptosis by a biochemical means but also to show quantitatively using this biochemical criterion that there is an increase in apoptotic cell death in SN. These results support our initial hypothesis that the decrease in the number of SN neurons observed in adulthood following striatal excitotoxic injury in development is mediated by induced cell death. This induced death event appears to be identical by morphologic criteria to the natural developmental cell death which occurs in SN, and, specifically, conforms to the morphologic pattern of apoptosis. Our re-

sults exclude any role played by excitotoxicity either directly due to QA, or indirectly following the loss of the striatonigral yaminobutyratergic projection to SNpr. Since some of the dying neurons were demonstrated to be dopaminergic, and since SNpc neurons are postmitotic at P7 (26), it is likely that this induced death event results in fewer SNpc neurons surviving into adulthood, as we have previously observed (7). Given that the type of cell death in SN is apoptosis and that this morphologic pattern is typical for induced developmental neuron death following deprivation of target-derived developmental support (27), it seems likely that the induced cell death in SN is in some way related to loss of developmental "support" from the striatum. The nature of this support is not addressed by the current study, and it may differ for SNpc and SNpr, although the morphology of cell death appeared similar in both. The striatum is a major target of SNpc dopaminergic neurons, so it is possible that striatal injury results in the loss of a target-derived, retrogradely transported trophic factor(s), according to classic concepts (5). However, there exists a striatonigral pathway emanating primarily from striatal striosomes which may also carry support in anterograde fashion to SNpc. It is unlikely that induced SNpc neuron death is due to direct injury to dopaminergic terminals after QA injection, as Schwarcz et al. (28) have shown that higher doses of QA do not affect TH activity or damage axons. Unlike SNpc, SNpr does not project to striatum; rather, it receives a major striatonigral projection which uses -ami-

Neurobiology: Macaya et al. A

2

1

B

3

4

5

2 3 4

S

6

_

1

6

C

Sth. FIG. 5. Ethidium bromide-stained gel and Southern blots of genomic DNA extracted from SN region following unilateral striatal QA lesion at P7. (A) Ethidium bromide-stained gel photographed under UV illumination. Eighteen micrograms of DNA was loaded in lanes 1-4. Lane 1: SN, experimental side, 6 hr after striatal QA; lane 2: SN, control side, same rat as lane 1; lane 3: SN, experimental side, 24 hr after striatal QA; lane 4: SN, control side at 24 hr; lanes 5 and 6: DNA (0.64 and 6.4 ,ug, respectively) from hypoxic neonatal rat liver. A clear "ladder" of 180- to 200-bp multimeric fragments is observed in lane 6. No bands are observed in other lanes. (B) Southern blot of gel shown in A, following exposure to x-ray film with two intensifying screens for 5 hr. Multiple nucleosomal fragments are observed in lanes 5 and 6 (liver). (C) Same Southern blot as shown in B, exposed to x-ray film with two intensifying screens for 48 hr. The nucleosomal fragments in lanes 5 and 6 are overexposed, but the mono- and dinucleosomal fragments are now visible in lanes 1-4 (arrows).

nobutyrate and substance P as transmitters. Presumably, striatal injury leads to induced cell death in SNpr on the basis of a loss of some anterograde influence on SNpr. It is unlikely that SNpc cell death occurs as a consequence of an "en cascade" degeneration of SNpr following striatal injury (12), because SNpc cell death clearly preceded that in SNpr. One of the implications of this study is that, following injury to the developing brain by an excitotoxic mechanism, the ensuing cell death occurs not only because of direct excitotoxic injury, but also at remote sites because of apo-

a

A

a

FIG. 6. (A) In situ end labeling of DNA is detected in multiple cells in SNpc at 48 hr, ipsilateral to striatal injury. Three show labeled nuclear chromatin clumps (small arrows). One shows labeled chromatin in a ring associated with the nuclear membrane (large arrow). (B) In SNpc contralateral to the side of striatal injury, no labeled cells are observed in a field corresponding to the one shown in A. (Bar = 10 ,um.)

Proc. Natl. Acad. Sci. USA 91 (1994)

8121

ptosis. The importance of this observation is that apoptosis is a distinct death mechanism, with specific regulatory features, and it can possibly be modified by methods quite different from those which may influence excitotoxicity. Thus, there may be novel approaches for treating neuronal death which occurs in the setting of a variety of cerebral insults which are thought to be mediated initially by an excitotoxic mechanism, including hypoxia-ischemia, hypoglycemia, seizures and head trauma. It is also possible that the mechanisms underlying induced cell death in SNpc, related either to expression of trophic factors or to the genes which mediate programmed cell death, may play a role in the idiopathic degeneration of these neurons late in life, such as that occurring in Parkinson disease. We are grateful to Ms. Pat White for superb secretarial assistance. Our deepest thanks go to Ms. Evelyn Dean and Dr. Robert Sloviter, who performed the ultrastructural analysis. We are grateful to Drs. Ann-Judith Silverman, Joan Witkin, and Virginia Tennyson for their expert advice in interpreting the electron micrographs. We thank Ms. Tinmarla Oo for technical assistance in performing in situ end labeling. This work was supported by National Institutes of Health Grant NS26836 (R.E.B.), United Cerebral Palsy (R.E.B.), the Parkinson's Disease Foundation (R.E.B. and A.M.), the Colleen Giblin Foundation for Pediatric Neurology Research (R.M.G. and A.M.), and Fondo de Investigaciones Sanitarias, Spain (F.M.).

1. Cowan, W. M., Fawcett, J. W., O'Leary, D. D. M. & Stanfield, B. B. (1984) Science 225, 1258-1265. 2. Cunningham, T. J. (1982) Int. Rev. Cytol. 74, 163-186. 3. Clarke, P. G. H. (1985) Trends Neurosci. 8, 345-349. 4. Oppenheim, R. W. (1991) Annu. Rev. Neurosci. 14, 453-501. 5. Barde, Y. (1989) Neuron 2, 1525-1534. 6. Herrup, K. & Sumter, K. (1987) J. Neurosci. 7, 829-836. 7. Burke, R. E., Macaya, A., De Vivo, D., Kenyon, N. & Janec, E. (1992) Neuroscience 50, 559-569. 8. Janec, E. & Burke, R. E. (1993) Mol. Cell. Neurosci. 4, 30-35. 9. Prochiantz, A., di Porzio, U., Kato, A., Berger, B. & Glowinski, J. (1979) Proc. Natl. Acad. Sci. USA 76, 5387-5391. 10. Hemmendinger, L. M., Garber, B. B., Hoffmann, P. C. & Heller, A. (1981) Proc. Natd. Acad. Sci. USA 78, 1264-1268. 11. Tomozawa, Y. & Appel, S. H. (1986) Brain Res. 3"9, 111-124. 12. Krammer, E. B. (1980) Brain Res. 196, 209-221. 13. Gallyas, F., Wolff, J. R., Bottcher, H. & Zaborsky, L. (1980) Stain Technol. 55, 299-306. 14. Sloviter, R. S., Sollas, A. L., Dean, E. & Neubort, S. (1993) J. Comp. Neurol. 330, 324-336. 15. Kerr, J. F. R. & Harmon, B. V. (1991) in Apoptosis: The Molecular Basis of Cell Death, eds. Tomei, L. D. & Cope, F. 0. (Cold Spring Harbor Lab. Press, Plainview, NY), pp. 47-60. 16. Arends, M. J., Morris, R. G. & Wyllie, A. H. (1990) Am. J. Pathol. 136, 593-608. 17. Paxinos, G. & Watson, C. (1982) The Rat Brain in Stereotaxic Coordinates (Academic, San Diego). 18. Sloviter, R. S., Dean, E. & Neubort, S. (1993) J. Comp. Neurol. 330, 337-351. 19. Wahl, G. M., Stern, M. & Stark, G. R. (1979) Proc. Natl. Acad. Sci. USA 76, 3683-3687. 20. Gavrieli, Y., Sherman, Y. & Ben-Sasson, S. A. (1992) J. Cell Biol. 119, 493-501. 21. Kerr, J. F. R., Wyllie, A. H. & Curie, A. R. (1972) Br. J. Cancer 26, 239-257. 22. Abercrombie, M. (1946) Anat. Rec. 94, 239-247. 23. Clarke, P. G. H. (1990) Anat. Embryol. 181, 195-213. 24. Wyllie, A. H., Morris, R. G., Smith, A. L. & Dunlop, D. (1984) J. Pathol. 142, 67-77. 25. Zakeri, Z., Quaglino, D., Latham, T. & Lockshin, R. (1993) FASEB J. 7, 470-478. 26. Lauder, J. M. & Bloom, F. E. (1974) J. Comp. Neurol. 155, 469-482. 27. Chu-Wang, I.-W. & Oppenheim, R. W. (1978) J. Comp. Neurol. 177, 33-58. 28. Schwarcz, R., Whetsell, W. 0. & Mangano, R. M. (1983) Science 219, 316-318.