Acta Neuropathol (1995) 90:17-30 �9 Springer-Verlag 1995

Jos6 A. Rafols �9 Asif M. Daya �9 Brian J. O'Nei l Gary S. Krause �9 Robert W. Neumar �9 Blaine C. White

Global brain ischemia and reperfusion: Golgi apparatus ultrastructure in neurons selectively vulnerable to death

Received: 31 October 1994/Revised, accepted: 24 January 1995

Abs t rac t The neocortex and the hippocampus were ex- amined for lipid peroxidation products and ultrastructural alterations by fluorescence and electron microscopy, re- spectively, in rats subjected to 10 min of cardiac arrest or 10 min cardiac arrest and either 90 or 360 min reperfu- sion. Lipid peroxidation products were observed after 90 min reperfusion in the perikarya and proximal dendrites of neocortical pyramidal neurons and in the hippocampal hilar cells and CA 1, region; the fluorescence was most in- tense at the base of the apical dendrite, the region of the Golgi apparatus. After 90 min of reperfusion, the CA 1, showed considerable stretches of rough endoplasmic reticulum devoid of ribosomes and the Golgi cisternae were shorter and widely dilated. The neocortex showed similar endoplasmic reticulum changes, but no significant alterations to the Golgi were noted. In addition there were areas where strings of ribosomes appear to be detaching from the endoplasmic reticulum. After 360 min reperfu- sion in both the neocortex and the hippocampus, the dam- age appeared more severe. The Golgi was fragmented into vacuoles, membranous whorls had appeared, and dense aggregates of smooth vesicles were seen coalescing with each other and the vacuoles. These observations suggest that early Golgi involvement is a more important marker of lethal injury than ribosome release from the endoplas- mic reticulum. The areas of disturbed Golgi ultrastructure correspond to those areas that show evidence of lipid per- oxidation and imply that lipid peroxidation may be causally related to the disturbance in Golgi ultrastructure.

Key words Cerebral ischemia �9 Reperfusion injury - Golgi apparatus �9 Endoplasmic reticulum

J. A. Rafols (N~) �9 A. M. Daya Department of Anatomy and Cell Biology, Wayne State University School of Medicine, 540 E. Canfield, Detroit, MI 48201, USA Tel.: 1-313-577-1049; Fax: 1-313-577-3125

B. J. O'Neil �9 G. S. Krause �9 R. W. Neumar �9 B. C. White Department of Emergency Medicine, Wayne State University School of Medicine, Detroit, Michigan, USA

Introduction

Recent histological and biochemical evidence indicates that oxygen free radicals primarily damage lipids in selec- tively vulnerable neurons (SVNs) during brain reperfu- sion [44]. In particular, we found histochemical evidence in situ of an intense concentration of lipid peroxidation products at the base of the apical dendrites of pyramidal neurons in neocortex and hippocampus [43]; a recent in vitro study by Harris et al. [7] also observed this cellular localization of carbonyl groups generated by iron-induced damage in cultured hippocampal neurons. This area of the cell corresponds to the location of a conspicuous portion of the Golgi apparatus [21]. Indeed, in other models of damage by radicals, such as carbon tetrachloride intoxica- tion [24], the Golgi apparatus undergoes early damage, and it has been shown that products of lipid peroxidation inhibit the glycosylation functions associated with the Golgi [15].

The Golgi apparatus is a central component of the sys- tem that manages and maintains the plasma membrane. It assembles integrated membrane vesicles that then leave the Golgi and are transported to and incorporated into the plasma membrane [3, 39, 42]. Furthermore, smooth vesi- cles recycle plasma membrane segments to the Golgi ap- paratus where they are incorporated into the membranes of the Golgi [5]. In addition, vesicular-mediated resealing is a primary mechanism of repair of the plasmalemma [33]. Thus, if the lipid peroxidation products we have ob- served at the base of the apical dendrite of SVNs in reper- fused brains are concentrated in the Golgi, damage to this system by lipid peroxidation or its inhibition by accumu- lated lipid peroxidation products during early reperfusion could result in a critical block in the mechanism of mem- brane repair. Such damage would be expected to have ob- servable ultrastructural consequences early in the course of post-ischemic reperfusion. Thus, Petito and co-workers observed early post-ischemic transformation of SVN Golgi complexes into clustered vesicles without cisterns [23], accompanied by SVN loss of histochemical evi-

18

dence of thiamine pyrophosphatase activity normally ob- served in Golgi [22].

Experimental models of complete global ischemia fol- lowed by reperfusion have varied widely from multivessel occlusion to complete cardiac arrest followed by reperfu- sion. We have elected to pursue the latter model due to its replication of appropriate clinical events and also to avoid the effects of collateral circulation. Our time sequences are much different from those previously reported due to our desire to elucidate the earliest possible documentation of the ultrastructural changes associated with reperfusion injury. We also differ from other studies in that we have attempted to compare and contrast the sequential progres- sion of early subcellular damage in two selectively vul- nerable zones (i.e., neocortex and hippocampus) with dis- tinct spatial and temporal patterns of cellular involvement during reperfusion injury [43]. Hence, this electron micro- scopic study was undertaken to evaluate more thoroughly the ultrastructure in SVNs during brain ischemia and reperfusion with a specific orientation to observations that might provide insight into the mechanism involved in the accumulation of lipid peroxidation products and its possi- ble implications in neuronal cell survival and death.

Methods

All animal experiments were approved by our institutional review board and were conducted following the "Principles of Laboratory Animal Care" (NIH publication No. 86-23, revised 1985). We studied four groups of rats: 1, normals; 2, 10-min KCl-induced car- diac arrest; 3, arrest, resuscitation, and 90-min reperfusion; and 4, 360-min reperfusion, using the rat model of cardiac arrest and re- suscitation we have described previously [31]. Brains were pre- pared either by the histochemical method that we have previously reported [43] for fluorescence microscopic study of lipid peroxida- tion products or for examination by electron microscopy.

Tissue preparation for electron microscopy

At the appropriate time each animal was anesthetized and perfused transcardially with isotonic saline (50 ml) followed by perfusion fixation with 250 ml cold paraformaldehyde (1.0%) and glu- taraldehyde (2.0%) in phosphate buffer. After perfusion fixation the brains were carefully resected and immersed in cold fixative overnight. The brains were sectioned (200 gm) with a vibratome, and single sections were collected in phosphate buffer. Under a dissection microscope, blocks of tissue containing the forelimb area of the sensorimotor cortex and the C A 1 region of the hip- pocampus were dissected and trimmed. We followed the stereo- taxic coordinates of Paxinos and Watson [20] to obtain consis- tently from the neocortical block a section passing 8.7 mm anterior to the interanral plane, whereas the section for Cornu Ammonis (CA1) was 5.7 mm anterior to the interaural plane. Each block was further trimmed of the adjacent white matter, washed in buffer, post-fixed in 2% OsO4, dehydrated, stained en bloc with uranyl ac- etate, and embedded flat in Embed-812 for purposes of orientation. Semithin (1 ~tm) and ultrathin sections were obtained with a Re- ichert Ultracut. Single semithin sections were stained with tolui- dine blue-methylene blue and visualized with a light microscope to determine final orientation prior to ultramicrotomy. Ultrathin sec- tions were stained with uranyl acetate and lead citrate and exam- ined with a Phillips EM-201 electron microscope.

Results

Fluorescence microscopy of the TBA-aldehyde adducts

At 90-min reperfusion, intense yellow-green granular flu- orescence (consistent with TBA-MDA emission at 553 nm) is observed in the perikarya and proximal dendrites of pyramidal neurons in numerous neocortical areas in the CA 1 of the hippocampus, as well as in many hippocampal hilar cells. In contrast, no fluorescence is detected in cells of the fascia dentata of the hippocampus (Fig. 1 A). Al- though fluorescent neurons are found in layers I I -VI of the neocortex, the largest number of these neurons in seen in layers III and V (Fig. 1 B). On closer examination fluo- rescent pyramidal neurons in layer V exhibit foci of in- tense fluorescence at the base of the apical dendrite (Fig. 1 B, arrows) as well as in other regions of the perikaryon surrounding the nucleus (Fig. 1 B, arrowhead). In the CA1 region most perikarya in the pyramidal cell layer exhibit clusters of granular fluorescence (Fig. 1 C, arrows) and sometimes single granules extend into the adjacent layer (e.g., stratum radiatum; Fig. 1 C, arrowheads). This is con- sistent with our previous report of localization in selec- tively vulnerable neurons of fluorescent products formed between TBA and aldehydic compounds generated during reperfusion [43]. The distribution of fluorescent granules in neurons of selectively vulnerable areas strongly suggest that in the region of the Golgi apparatus, or in another or- ganelle with widespread perikaryal distribution, there is specific subcellular localization of lipid peroxidation products formed during early reperfusion.

Electron microscopy

Ischemia of l O-min duration without recirculation

We observed essentially no ultrastructural differences be- tween the normal tissue and that obtained following only the 10-rain period of complete global brain ischemia by cardiac arrest without resuscitation. Thus, for purposes of brevity and to reflect the focus of our work on ultrastruc- tural alterations during reperfusion, we begin by reporting the ultrastructural results in the 10-min ischemic samples.

CA1. The pyramidal cell layer in CA 1 is readily identified by the high density of neuronal somatic profiles present in low-power electron micrographs (Fig.2A, labeled nu 1, nu2, nu3). Because of this density the cell membranes of adjacent neurons sometimes appose each other directly; however, most often they are separated by thin astroglia processes (Fig. 2 B, labeled as), glial perikarya, or various amounts of neuropil. At high magnification the perikaryon of these neurons exhibits its normal complement of or- ganelles including well-organized granular endoplasmic reticulum (RER), Golgi complexes (Figs.2B, C; 3 A, la- beled G), mitochondria, and cytoskeletal constituents. Characteristically, the external surfaces of the membranes

Fig. 1 A-C Photomicrographs of coronal sections through se- lectively vulnerable zones (SVZ) after cardiac arrest, re- suscitation, and 90-rain reper- fusion. A Fascia dentata of hippocampus showing no fluo- rescence indicative of MDA- TBA peroxidative products. B Layer 5 of sensorimotor cortex showing many cortical pyrami- dal ceils with foci of intense fluorescence at the bases of apical dendrites (arrows) or in the region of the perikaryon surrounding the nucleus (ar- rowhead). C CA 1 region of hippocampus showing granu- lar-like fluorescence in the perikarya of pyramidal ceils (arrows) and some granules extending into the adjacent layer (arrowhead). Bars A = 100gm, B=50gm, C = 1 0 gm

19

of RER cistemae are studded with rows of ribosomes (Fig. 2 B, C, arrows), although some stretches of the cis- ternae lack ribosomes. Most of the free ribosomes in be- tween the RER cisternae form polysomal arrangements which are particularly prominent in the regions of Nissl substance (Fig. 3 A, arrows). Golgi complexes (Figs. 2 B, C; 3 A) exhibit normal ultrastructure consisting of arrays of closely packed, flat and broad cisternae of smooth en- doplasmic reticulum and their associated tubules and vesi-

cles. All of the structural organelles, including mitochon- dria, microtubules, lysosomes, and the nuclear' envelope appear to be unaffected by the 10-rain global ischemia with no reperfusion.

Sensorimotor cortex. Examples of profiles of pyramidal neurons from layers II1 and V of the sensorimotor cortex are shown in Figs.3B and 4A, respectively. Details of portions of the perikaryon and nucleus of each of these

20

F ig .2A-C Electron micrographs of the CAa pyramidal cell layer of hippocampus from an animal subjected to 10-rain global is- chemia but no reperfusion. A Profiles of nuclei of pyramidal neu- rons (nul, nu2, nu3) are distinguished at relatively low magnifica- tion. Areas framed by rectangles, one in the upper and the other in the lower port ions of the micrographs are reproduced at higher magnification in B and C, respectively. B, C Note the arrayed cis- ternae of the Golgi complexes (G) in the perikarya of neurons nut (B) and nu3 (C), as well as the ribosome-studded cisternae of rough endoplasmic reticulum (arrows). In B, a thin astroglial sheath (as) is interposed between the plasma membranes of neu- rons nu t and nu 2. Bars A = 5 gm, B, C = 1 gm

neurons are reproduced at h igher magni f ica t ion in Figs. 3 C and 4 B. The normal ul t rastructure of smal l py ramida l neurons has been descr ibed p rev ious ly [21] and share s imilar i t ies with those i l lustrated here after 10-min is- chemia. General ly , the nuclear enve lope is smooth, except for an occas iona l shal low infolding, and surrounds a ho- mogeneous ly d ispersed ka ryop la sm ( F i g s . 3 C , 4B , la- be led nu). The Niss l bodies are represented by smal l ar- rays of R E R and their associa ted po lysomes . Complexes of the Golg i apparatus (Figs. 3 C, 4 B, l abe led G) par t ia l ly r ing the nucleus and extend into the base o f the apical dendr i te (Fig. 3 B, l abe led d). The concentr ic arrays o f

21

Fig.3 A CAI pyramidal cell layer of hippocampus; B, C, superfi- cial layer of sensorimotor cortex from an animal with 10-min global ischemia and no reperfusion. A Golgi complex (G) is inter- posed between arrayed cisternae of rough endoplasmic reticulum (arrows) representing the Nissl substance. B A portion of the nu- cleus (nu), adjacent perikaryon and apical dendrite (d) of a small

pyramidal neuron. Area framed by the rectangle is reproduced in C at higher magnification. C Golgi complexes (G) composed of arrayed flat and broad cisternae and their associated clathrin- coated vesicles (arrow), show normal ultrastructural features. Bars A-C = 1 [.tin

nar row and broad c i s t emae o f these complexes do not ap- pear to be a l tered at 10-min i schemia with no reperfusion. In addit ion, the inc idence of c la thr in-coa ted ves ic les (Fig. 3 C, arrows) assoc ia ted with these complexes is quali ta- t ive ly s imi lar to that found in normal mater ia l .

Ischemia of lO-min duration and 90-rain repelfusion

CA1. Al l the neuronal prof i les examined in the stratum py ramida l e exhib i ted s imi lar features fo l lowing i schemia and 90-min reperfusion. Two main al terat ions o f organel le ul trastructure, one re la ted to the R E R and the other to the

22

Fig.4A, B Deep layer of sensorimotor cortex from an animal with 10-rain global ischemia and no reperfusion. A Portion of the nucleus (nu) and perikaryon of a pyramidal neuron. Area framed by the rectangle is reproduced in B at a higher magnification. B Golgi complexes (G) and other perikaryal organelles display normal ultrastructural features

Golgi complexes, are observed at this stage. The RER is represented by single long and short cisternae lacking the characteristic arrays of the Nissl substance. On closer ex- amination, many of the outer surfaces of the cisternae have considerable stretches devoid of ribosomes, with

long strings of ribosomes appearing to be in the process of detaching from their adjacent membranes (Fig. 5 B, small arrows). In addition to free polysomal aggregates, many single ribosomes are also dispersed between the RER cis- temae and throughout the perikaryon. Alteration of the Golgi complex is more pronounced than that of the RER in that the ultrastructure of nearly each complex of the ap- paratus, which rings the nucleus, appears to be affected (Fig. 5 A, large arrows). Within each complex the integrity of the cisternal array is altered with single cisternae being shorter and widely dilated. Aggregates of smooth mi-

23

Fig.SA, B CA 1 pyramidal cell layer of the hippocampus from an animal with 10-min global ischemia followed by 90-min reperfu- sion. A Portion of the nucleus (nu) and the perikaryon of a pyra- midal neuron. Note the absence of arrays of cisternae in each of the Golgi complexes (arrows) at this reperfusion time. Nucleolus (ncl). B Another portion of the nucleus (nu) and adjacent perikaryon of the neuron in A, illustrated at a higher magnification. The centrally placed Golgi complex (G) is characterized by widely dilated and seemingly shorter cistemae than those in animals with no reperfusion. Some curved cistemae (large arrow) are also ob- served. Aggregates of smooth vesicles are also seen in association with the dilated cistemae, with some vesicles fusing (small arrow- head) with the membrane of the dilated sacs. The other morpho- logical feature at this reperfusion time is the apparent stripping of ribosomes (arrows) from the membranes of rough endoplasmic cisternae. Bars A, B = 1 gm

crovesicles are found between the dilated cisternae, some- times fringed by a curved RER cisterna (Fig. 5 B, large ar- row). In some instances the membrane of some vesicles dist inctly fuses with that of the dilated cistemae (Fig. 5 B, arrowhead), indicating that the vesicle is either being added or budding off the wall of the cisterna.

Sensorimotor cortex. Perikarya of many pyramidal neu- rons in layers III and V of the sensorimotor cortex exhibit s imilar alterations of the RER and Golgi complex as those described for CA 1 . In this regard, alterations similar to those described for the CA~, are consistently seen at the RER; however, in some pyramidal neurons, very little al-

24

Fig.6A, B Deep layer of sensorimotor cortex from an animal with 10-rain global ischemia followed by 90-rain reperfusion. A Portion of the nucleus (nu) and adjacent perikalyon of a pyramidal neuron. Note that most of the rough endoplasmic reticulum is com- posed of long "stringy" cisternae (arrows). Area framed by rectan- gle is reproduced in B at a higher magnification. (ncl nucleolus). B In contrast to the Golgi complex of the CAx neuron in Fig. 5 B, the

concentric arrays of the Golgi complex (G) are preserved, although in this case numerous clathrin-coated vesicles (arrows) are seen in association with the complex. A long "stringy" cisternae of rough endoplasmic reticulum has portions almost denuded of ribosomes (between arrowheads). Note also that a long string of ribosomes (double arrowhead) appears to strip off its associated cisterna. Bars A, B= l Bm

terat ion of the Golg i complex is observed. A n example of such a cel l is shown in F i g . 6 A , B; this cell exhibi ts the long "s t r ingy" R E R c i s t emae ( F i g . 6 A , arrows) , long stretches o f c is ternae devoid of r ibosomes ( F i g . 6 A , ar- rowheads) , and the long strings o f po ly r i bosome appar- ent ly de tached f rom these cis ternae (Fig. 6B, double ar-

rowheads) . In this same cell , the Golg i complex (G) mor- pho logy appears normal , except that it is associa ted with increased numbers of c la thr in-coated ves ic les adjacent to the cisternal array.

25

F ig .7A-D Deep layer of sensorimotor cortex from an animal subjected to 10-min global ischemia followed by 360-min reperfu- sion. A, B Portions of the nucleus (nu), adjacent perikaryon, and apical dendrite (d) of two pyramidal neurons. Areas framed by rec- tangles are reproduced in C and D at higher magnifications. C Pro- nounced microvesiculation (arrow) in between dilated Golgi cis-

ternae is observed at this reperfusion time. D Membranous whorls are also observed at this time with numerous vesicles (small ar- rows) seen fusing with membranes of the whorl. Sometimes a core of cytoplasm (large arrow) containing ribosomes, vesicles, and other organelles is found within the whorl. Bars A-D = l~xm

Ischemia o f l O-rain duration and 360-rain reperfusion

At longer reperfus ion t imes the ul t rastructural features o f py ramida l cells in both C A 1 h ippocampus and sensor imo- tor cor tex share many s imilar i t ies in that in both areas the ma jo r subcel lu lar ev idence of damage is seen in the R E R

and the Golg i complexes . Here we only show examples of sensor imotor cort ical cel ls because the py ramida l cel ls of the C A 1 region represent ident ica l features.

Al though the degree of invo lvement in sJingle cells varies considerably, many py ramida l neurons exhibi t greater ul t ras t ructural a l terat ion of the Golg i complexes

26

Fig.8 Low-power micrograph showing portions of the nucleus (nu) and adjacent perikaryon of a pyramidal neuron in the deep layer of the sensorimotor cortex of an animal with 10-min global ischemia followed by 360-min reperfusion. Portions of an adjacent astroglia (as) is also shown. Note that each of the Golgi complexes

(arrowheads) shows pronounced microvesiculation. Large arrows point at a continuous, highly distended perinuclear cisternae of smooth endoplasmic reticulum. Some of the rough endoplasmic cisternae in the astrocyte also appear distended (arrows). (ncl nu- cleolus). Bar = 1 gm

after 360-min reperfusion than after 90-rain reperfusion. Here the dilated Golgi cisternae appear further frag- mented into vacuoles of various sizes. This is evident both at the base of the apical dendrite (Fig. 7 A, B, labeled d) as well as in other regions of the perikaryon. In between these vacuoles, dense aggregates of smooth vesicles are

seen coalescing with each other and with the vacuoles (Fig. 7 C, large arrow). Aggregates of this type occupy large areas of the perikaryon (Fig. 8, arrowheads), coinci- dent with the location of the Golgi complexes observed at earlier reperfusion times. In addition membranous whorls (Fig. 7 D, 9) are also observed in association with the ag-

27

Fig. 9 A, B Deep layer of sensorimotor cortex from an animal sub- jected to 10-min global ischemia followed by 360-min reperfusion. A Profile of the nucleus (nu) and sunounding perikaryon of a pyra- midal neuron. Note the extent of the large perinuclear cisternae (arrows). Area framed by rectangle is reproduced in B at a higher magnification. B In addition to dilated cistemae of the Golgi com- plex and pronounced microvesiculation (large arrow), many vesi- cles (small arrows) are observed fusing with the membranes of a forming whorl, as well as those of the dilated cisternae (small ar- rowhead). The integrity of the nuclear envelope and its associated nuclear pore complexes (large arrowheads) does not appear to be compromised even at this late reperfusion stage (nu nucleus). Bars A , B = 1 ~Lm

gregates at this time. Many smooth vesicles (F ig .7D, small arrow) are seen fusing with the membranes of the whorl, whose concentr ical ly disposed membranes often encircle a core of membrane -bound cytoplasm which in- cludes r ibosomes, vesicles, and other organelles (Fig. 7 D, large arrow). Because the smooth vesicles forming an ag- gregate often fuse with the membranes of dilated vacuoles and with those of small whorls, it would appear they con- tribute to the formation of these structures. It is not un- usual to see various developmental stages of the whorls in the same neuron. Some cells also exhibit cont inuous and greatly distended cisternae of smooth endoplasmic reticu- lum of variable lengths. Whereas in some neurons it oc-

28

cupies a sector of the perikaryon in close proximity to the nucleus (Fig. 8 large arrows), in other cells it surrounds the nucleus and extends to more ectoplasmic regions of the perikaryon (Fig. 9 A, large arrows).

Discussion

Our ultrastructural data demonstrate major alterations of morphology in the RER and Golgi complexes of selec- tively vulnerable neurons during reperfusion. The Golgi exhibit cisternal "budding" after 90-min reperfusion, and after 6-h reperfusion grossly abnormal membranous whorls appear that include microtubule fragments and ri- bosomes. These morphological characteristics are seen in selectively vulnerable neurons from the same areas where there is early fluorescence histochemical evidence of con- centrated lipid peroxidation products. We did not see ul- trastructural alterations in other organelles such as mito- chondria, lysosomes, lipofuscin, or cytosketetal struc- tures, which might readily be correlated to the histochem- ical observations.

RER alterations are consistently observed in pyramidal neurons of both the CA1 and sensorimotor cortical areas. The RER is largely denuded of polyribosomes, and polyribosome chains appear to be detaching from the RER. As noted above, ultrastructural alterations in Golgi morphology are also observed in selectively vulnerable neurons during early post-ischemic reperfusion; however, there appear to be important differences between the CA1 neurons, where extensive Golgi involvement is uniformly observed, and the cortical neurons, where Golgi involve- ment ranges from extensive to negligible. Because the in- sult of ischemia and reperfusion utilized here is known to be nearly 100% lethal to hippocampal CA~ neurons [4], but less uniformly lethal to cortical pyramidal neurons, our observations suggest that early Golgi involvement is a more important marker of lethal injury than ribosome re- lease from the RER.

The morphological coincidence of the areas where neurons show histochemical evidence of lipid peroxida- tion with disturbed Golgi ultrastructure and the associa- tion of both these observations with the known SVNs sug- gests that the process of lipid peroxidation may be causally related to the disturbance in Golgi ultrastructure. Previous ultrastructural studies have shown disruption of neuronal Golgi complexes [22, 23] and RER [37] early in the course of post-ischemic reperfusion, and reversibility of these ultrastructural changes could be critical for neu- ronal survival after an ischemic insult. Petito et al. [22, 23] using a rat four-vessel occlusion model, observed an associated reappearance of Golgi cisternae together with RER-bound polyribosomes in reversibly injured neurons after 120-min reperfusion, whereas these structures re- mained disrupted with progressive dilation of Golgi vesi- cles in irreversibly injured neurons [22, 23]. Furuta et al. [4] demonstrated polyribosomal disaggregation and sup- pression of protein synthesis in CA1 neurons after 3-h reperfusion following 5 min of carotid occlusion in ger-

bils; without protective preconditioning, these ultrastruc- rural changes and suppressed protein synthesis persisted at least 24 h and were associated with 92% neuronal demise at 7 days. However, in animals protected by is- chemic preconditioning, polyribosomes reassembled and protein synthesis rates returned to normal by 24 h; this ul- trastructural and functional recovery was associated with 100% neuronal survival at 7 days. These studies indicate that during early reperfusion both reversibly and irre- versibly injured neurons exhibit disaggregation of polyri- bosomes and alterations in Golgi ultrastructure; the asso- ciation between resolution of these abnormalities and neu- ronal survival suggests that such resolution is an essential part of recovering from ischemic injury.

Membrane whorls

The formation of whorl patterns of endoplasmic reticulum has been reported in neuronal [8, 18] and non-neuronal [17, 19] cell types under a variety of normal and patho- logical conditions; however, it is unclear whether whorl bodies represent membrane regeneration or cellular de- generation. Hwang et al. [10] studied the in vivo effects on rat liver of various inhibitors of protein synthesis. They found that massive membrane whorls were produced when membrane-bound, but not cytosolic, ribosomes were inhibited. Furthermore, there was marked suppres- sion in protein synthesis with increased lipid synthesis during the formation of whorls, leading them to suggest that the formation of membrane whorls was linked to a dissociation in the synthesis of the lipid and protein com- ponents of membranes. Our ultrastructural evidence in a cell culture model (unpublished) demonstrates that radical damage can induce membrane whorls, but we have not yet examined whether this ultrastructural phenomenon is associated with a radical-induced inhibition of protein synthesis.

Reperfusion-induced detachment of polyribosomes from the RER

The arrows in Fig.5B and 6B point at strings of ribo- somes, presumably connected by an mRNA, in the process of either attaching or detaching from the RER. The former is unlikely to be true since these strings are not seen in normal brains. Moreover, they are not seen during ischemia and, therefore, we believe that these strings represent detachment of the ribosomes from the RER as a consequence of reperfusion. Molecular mecha- nisms that might underlie this observation include disrup- tion of microtubules or disruption of the RER-ribosome binding apparatus.

Microtubules, through interaction with various micro- tubule-associated proteins (MAPs) [2, 11, 25, 27, 28, 34], play a role in ribosomal binding to the RER, and micro- tubule disruption occurs, particularly in dendrites of SVNs, following cerebral ischemia and reperfusion [22,

37, 45]. Walker and Whitfield [40] demonstrated that dis- ruption of microtubules by colchicine resulted in a pro- gressive decrease in the size of membrane-bound polyri- bosomes consistent with inhibition of initiation. However, their results suggest dissociation of poly(A +) mRNA from membrane-bound polyribosomes without dissociation of the ribosomal monomers from the membrane. This pat- tern is not consistent with our observation of RER detach- ment of polyribosomes still bound to mRNA being medi- ated by cytoskeletal dissolution.

Contact between the ribosome and the RER membrane surface is believed to be a function of two receptor-medi- ated mechanisms. Proteins destined for secretion or inclu- sion in membranous organelles contain a sequence of amino acids that signal the protein is to be synthesized on the RER. As soon as this signal peptide emerges from the ribosome, it binds to a signal recognition particle (SRP) [1], causing a pause in translation. The SRP-ribosome complex migrates to the RER [13, 14, 41] and binds to a SRP receptor in the RER. In a poorly understood se- quence, the ribosome binds to a ribosome receptor on the RER, the SRP is released, a protein translocation pore is formed, and translation resumes.

The binding of the SRP-ribosome complex to the RER could be disrupted in several ways. Both the SRP and the ribosome receptors are proteins, and proteolysis is known to disrupt the ability of these receptors to bind their re- spective ligands [12]. Although proteolysis is not thought to be a major factor in the post-ischemic brain, this does not rule out selective or small-scale proteolysis [9]. Alter- ations in the peptide undergoing synthesis are not likely to be a factor in maintaining ribosomal attachment to the RER; dissociation of nascent peptides from the ribosome- pore complex does not result in release of ribosomes from microsomes [30]. These receptors are also sensitive to the surrounding lipid milieu. The ribosome receptor, for ex- ample, is inactivated in vitro when microsomes are ex- posed to phospholipase C or phospholipase A 2 [12]. This may have significance in view of the accumulation of lipid peroxidation products that we previously reported in the Golgi, and the increased microsomal phospholipase A 2 activity seen after ischemia and reperfusion [26]. In summary, intriguing suggestions can now be made re- garding potential mechanisms by which the interrelations of the RER membrane and its ribosome or SRP receptors are likely involved in polyribosome release during reper- fusion, although the precise causality for this phenome- non remains to be determined.

Reperfusion-induced lipid peroxidation and alterations in Golgi morphology

The role of the Golgi apparatus in membrane assembly and recycling of membrane components is of particular interest after cerebral ischemia and reperfusion. Rapid re- sealing of the plasmalemma is achieved by vesicle deliv- ery, docking, and fusion in a mechanism similar to that for the exocytosis of neurotransmitters [33]. This suggests

29

that membrane repair by Golgi-derived vesicles may be a critical process during post-ischemic reperfusion and pro- vides a rationale for our ultrastructural observation of ap- parent "budding" from the trans-cisterna in the animal model. This is supported by our cell culture model in which we observed dilated Golgi cistemae surrounded by numerous vesicles that appear clathrin coated after 6-h ex- posure to the radical insult (unpublished).

The concentration of lipid peroxidation products in the Golgi may reflect an attempt to recycle peroxidized plas- malemma membrane. Membrane recycling by the Golgi apparatus was seen by Sleight and Pagano [32], who demonstrated that fluorescent phosphatidylcholine (the major phospholipid component of neuronal plasma mem- brane) incorporated into the plasma membrane of cultured fibroblasts specifically accumulated at the Golgi appara- tus under normothermic conditions by a mechanism that is energy and temperature dependent. Similarly, in sympa- thetic neurons, labeled plasma membrane segments are in- corporated into the Golgi apparatus [5]. The mechanism of recycling membrane from the plasma membrane to the Golgi apparatus is thought to be mediated by endocytotic vesicles. Although, Thibaut-Vercruyssen et al. [35] have observed evidence of a significant decrease in receptor- specific endocytosis following free radical injury in cul- tured endothelial cells, more general studies of membrane recycling after radical damage have still to be performed.

Finally, modifications in microtubular organization could also play an important role in the loss of cisternal organization seen after 6 h of reperfusion. Organized vesicular transport [29] and the structural integrity of the Golgi apparatus [36] rely on the presence and organiza- tion of microtubules in the cytoplasm. Neuronal micro- tubule disruption, particularly in the apical dendrite, has been well documented following cerebral ischemia and reperfusion [37, 38, 45], which may be an effect of mem- brane lipid peroxidation. Alkenal products of lipid perox- idation appear to interact directly with tubulin mad inhibit microtubule polymerization [16]. Furthermore, treatment with a radical scavenger up to 1 h post-ischemia induces dramatic preservation of neurons and dendrite MAP2 im- munoreactivity in the CA I hippocampus [6]. Further elec- tron microscopic-imnmnocytochemical studies using spe- cific markers to cytoskeletal components addressing these issues are currently being carried out in our laboratory.

Acknowledgements This work is supported in part by National Institutes of Health Grants (NS-01585 and GM-08167). Dr. Neu- mar is supported in part by a grant from the Emergency Medicine Foundation/Genentech.

References

1. Adelman MR, Sabatini DD, Blobel G (1973) Ribosome-mem- brane interaction. Nondestructive disassembly of rat liver rough microsomes into ribosomal and membrane component. J Cell Biol 56 : 206-229

2. Bryan J, Nagle BW, Doenges KH (1975) Inhibition of mbulin assembly by RNA and other polyanions: evidence for required protein. Proc Natl Acad Sci USA 75 : 3570-3574

30

3.Farquhar MG (1978) Traffic of products and membranes through the Golgi complex. In: Silverstein KSC (ed) Transport of macromolecules in cellular systems. Dahlem Konferenzen, Berlin, pp 341-362

4.Furuta S, Ohta S, Hatakeyama T, Nakamura K, Saburo S (1993) Recovery of protein synthesis in tolerance-induced hip- pocampal CA1 neurons after transient forebrain ischemia. Acta Neuropathol 86 : 329-336

5. Gonatas NK, Steiber A, Kim SU, Graham DI, Avrameas S (1975) Internalization of neuronal plasma membrane ricin re- ceptors into the Golgi apparatus. Exp Cell Res 94 : 426-431

6. Hara H, Kogure K (1990) Prevention of hippocampus neuronal damage in ischemic gerbils by a novel lipid peroxidation in- hibitor (quinazoline derivative). J Pharmacol Exp Ther 255: 906-913

7.Harris ME, Carney JM, Leedle RA (1994) Detection of oxida- tion products in individual neurons by fluorescence mi- croscopy. Exp Neurol 129:95-102

8.Hirano A (1978) Changes of the neuronal endoplasmic reticu- lnm in the peripheral nervous system in mutant hamsters with hind leg paralysis and normal controls. J Neuropathol Exp Neurol 37 : 75-84

9.Hong SC, Goto Y, Lanzino G, Soleau S, Kassell NF, Lee KS (1994) Neuroprotection with a calpain inhibitor in a model of focal cerebral ischemia. Stroke 25 : 663-669

10.Hwang KM, Yang LC, Carrico CK, Schulz RA, Schenkman JB, Sartorelli AC (1974) Production of membrane whorls in rat liver by some inhibitors of protein synthesis. J Cell Biol 62: 20-31

11. Jessus C, Huchon D, Friedrich E, Francon J, Ozon R (1984) In- teraction between rat brain microtubule associated proteins (MAPs) and free ribosomes for Xenopus oocyte: a possible mechanism for the in vivo distribution of MAPs. Cell Differ 14 : 295-301

12.Jothy S, Bilodeau JL, Simpkins H (1975) The role of mem- brane proteins and phospholipids in the interaction of ribo- somes with endoplasmic reticulum membranes. Can J Biochem 53 : 1039-1045

13.Krieg UC, Walter P, Johnson AE (1986) Photocrosslinking of the signal sequence of nascent preprolactin to the 54 kD polypeptide of the signal recognition particle. Proc Natl Acad Sci USA 83 : 8604-8608

14. Kurzchalia TV, Wiedmann M, Girshovich AS, Bochkareva ES, Bielka H, Rapaport TA (1986) The signal sequence of nascent preprolactin interacts with the 54 kD polypeptide of the signal recognition particle. Nature 320: 634-636

15.Marinari UM, Pronzato MA, Cottalasso D, Rolla C, Biasi F, Poli G, Nanni G, Dianzani MU (1987) Inhibition of liver Golgi glycosylation activities by carbonyl products of lipid peroxida- tion. Free Radic Res Commun 3:319-324

16.Miglietta A, Gabriel L, Gadoni E (1987) Microtubular protein impairment by pentanal and hexanal. Cell Biochem Funct 5: 189-194

17. Muraoka Y, Yahara I, Nara H, Watanabe H (1981) Steroid-in- duced concentric membrane whorls in dog liver. Experientia 37 : 389-390

18. Naftolin F, Bmhlmann-Papazyan M, Baetens D, Garcia-Segura M (1985) Neurons with whorl bodies have increased number of synapses. Brain Res 329 : 289-293

19.Nickerson PA, Curtis JC (1969) Concentric whorls of rough endoplasmic reticulum in adrenocortical cells of the Mongolian gerbil. J Cell Biol 40 : 859-862

20.Paxinos G, Watson C (1986) The rat brain in stereotaxic coor- dinates, 2nd edn. Academic Press, New York

21.Peters A, Palay SL, deWebster H (1991) The fine structure of the nervous system: neurons and their supporting ceils, 3rd edn. Oxford University Press, New York, pp 21-22

22. Petito CK, Lapinski RL (1986) Post-ischemic alterations in ul- trastructural cytochemistry of neuronal Golgi apparatus. Lab Invest 55 : 696-702

23. Petito CK, Pulsinelli WA (1984) Sequential development of re- versible and irreversible neuronal damage following cerebral ischemia. J Neuropathol Exp Neurol 43 : 141-153

24. Pronzato MA, Cottalasso D, Domenicotti C, Tenca C, Traverso N, Nanni G, Marinari UM (1990) Effects of G e l 4 poisoning on metabolism of dolichol in rat liver microsomes and Golgi ap- paratus. Free Radic Res Commun 11 : 267-277

25. Ris H (1985) The cytoplasmic filament system in critical point- dried whole mounts and plastic-embedded sections. J Cell Biol 100 : 1474-1487

26.Rordorf G, Uemura Y, Bonventre JV (1991) Characterization of phospholipase A2 (PLA2) activity in gerbil brain: enhanced activities of cytosolic, mitochondrial, and microsomal forms after ischemia and reperfusion. J Neurosci. 11 : 1829-1836

27. Schr6der HC, Bemd A, Zahn RK, Mtiller WEG (1982) Interac- tion of polyribosomal components and polynucleotides with microtubule proteins. Mol Biol Rep 8 : 233-237

28. Schr6der HC, Bernd A, Zahn RK, Mtiller WEG (1984) Binding of polyribonucleotides and polydeoxyribonucleotides to bovine brain micrombule protein: age-dependent modulation via phos- phorylation of high-molecular weight microtubule-associated proteins and tan proteins. Mech Ageing Dev 24:101-117

29. Schubert P, Kreutzberg GW, Lux HD (1972) Neuroplasmic transport in dendrites: effect of cholchicine on morphology and physiology of motoneurons in the cat. Brain Res 47:331-343

30. Simon SM, Blobel G (1991) A protein-conducting channel in the endoplasmic reticulum. Cell 65:371-380

31.Skjaerlund JM, Krause GS, Feldman DM, White BC (1991) The effect of EMHP on post-cardiac arrest survival of rats. Re- suscitation 22:139-149

32. Sleight GR, Pagano RE (1984) Transport of a fluorescent phos- phatidylcholine analog from the plasma membrane to the Golgi apparatus. J Cell Biol 99 : 742-751

33. Steinhardt RA, Bi G, Alderton JM (1994) Cell membrane re- sealing by a vesicular mechanism similar to neurotransmitter release. Science 263 : 390-393

34. Suprenant KA Tempero LB, Hammer LE (1989) Association of ribosomes with in vitro assembled microtubules. Cell Motil Cytoskel 14:401415

35.Thibaut-Vercruyssen R, Poumay Y, Ronveaux-Dupal MF (1992) Influence of a short oxidative stress on the LDL endo- cytosis by human endothelial cells: an ultrastructural study. J Submicrose Cytol Pathol 24:61-73

36.Thyberg J, Moskalewski S (1985) Microtubules and the orga- nization of the Golgi complex. Exp Cell Res 159:1-16

37.Tomimoto H, Yanagihara T (1992) Electron microscopic in- vestigation of the cerebral cortex after cerebral ischemia and reperfusion in the gerbil. Brain Res 598 : 87-97

38.Tomimoto H, Yanagihara T (1992) Immunoelectron micro- scopic study of tubulin and microtubule-associated proteins af- ter transient cerebral ischemia in gerbils. Acta Neuropathol 84: 394-399

39. Voelker DR (1990) Lipid transport pathways in mammalian cells. Experientia 46 : 569-579

40. Walker PR, Whitfield JF (1985) Cytoplasmic microtubules are essential for the formation of membrane-bound polyribosomes. J Biol Chem 260 : 765-770

41.Walter P, Blobel G (1981) Translocation of proteins across the endoplasmic reticulum. III. Signal recognition protein (SRP) causes signal sequence-dependent and site-specific arrest of chain elongation that is released by microsomal membranes. J Cell Biol 91:557-561

42.Whaley WG, Danwalder M (1979) The Golgi apparatus, the plasma membrane, and functional integration. Int Rev Cytol 58 : 199-245

43. White BC, Daya A, DeGracia DJ, O'Neil B J, Skjaerlund JM, Trumble S, Krause GS, Rafols JA (1993) Fluorescent histo- chemical localization of lipid perocidation during brain reper- fusion following cardiac arrest. Acta Neuropathot 86:1-9

44.White BC, Grossman LI, Krause GS (1993) Brain injury by global ischemia and reperfusion: a theoretical perspective on membrane damage and repair. Neurology 43:1656-1665

45. Yamamoto K, Morimoto K, Yanagihara T (1986) Cerebral is- chemia in the gerbil: transmission electron microscopic and immunoelectron microscopic investigation. Brain Res 384: 1-10