sign in sign up
<<
Home , Fascia dentata, Granule cell

The Journal of Neuroscience, April 1993, f3(4): 151 l-l 522

Single Axonal Systems of Human Dentate Granule Cells Studied in Hippocampal Slices from Patients with Temporal Lobe

Masako Isokawa,’ Michel F. Levesque,2 Thomas L. Babb,ls and Jerome Engel, Jr.lv3 ‘Brain Research Institute and Departments of ‘Neurosurgery and 3Neurology, School of Medicine, University of California, Los Angeles, California 90024-1761

Previous histological and immunocytochemical studies sug- present findings are epilepsy associated. However, the pres- gest that reorganization of the dentate , ence of aberrant mossy fiber collaterals in the hippocampi the mossy fibers, can occur in epileptic human used in the present study has been confirmed by Timm’s (Sutula et al., 1989; Houser et al., 1990; Babb et al., 1991) staining and/or dynorphin immunohistochemistry in com- and in animal models of epilepsy (Tauck and Nadler, 1985; parison with nonepileptic autopsy material, indicating its re- Sutula et al., 1988; Cronin et al., 1992). However, neuroan- lation to epilepsy (Babb et al., 1991, 1992). At present, there atomical analyses of the trajectory and morphology of re- seems to be a consensus that the projection of mossy fiber organized axons are not yet available. The present study collaterals to the supragranular layer is a rare occurrence in was conducted to investigate single dentate granule cell normal rats (Lorento de No, 1934; Claiborne et al., 1988; axonal systems in human epileptic hippocampus. Individual Seress et al., 1991; present study), normal monkeys (Seress mossy fibers were directly visualized by injecting a tracer et al., 1991), and normal humans (Houser et al., 1990). Thus, (biocytin or Lucifer yellow) intracellularly in hippocampal we believe that the intracellularly stained aberrant col- slices prepared from temporal lobes that were surgically laterals reported here in the supragranular layer likely rep- removed from patients for treatment of intractable epilepsy. resent trajectories of reorganized mossy fibers associated Two major arborization patterns were identified: (1) the par- with medically intractable . ent axons extended to and coursed through the hilus toward [Key words: intracellular injections, granule cell axons, CA3, leaving collaterals along their paths in the hilus (N = biocytin, Lucifer yellow, sprouting, filopodia] 19 ); (2) in addition to the aforementioned axonal system, collateral(s) branched from the parent axon near the and projected to the granule cell layer and molecular Recent histological and immunocytochemical studies on epi- layer, forming an aberrant axonal pathway (N = 9 neurons). leptic human hippocampussuggest that the dentategranule cell These aberrant collaterals bore large boutons similar to those axons, the mossy fibers, in epileptic hippocampus have a col- of the hilar axons and formed extensive plexuses in the lateral arborization pattern that is different from that of normal granule cell layer and/or in the molecular layer. The summed mossy fibers in lower mammalsand in autopsy control human length of collaterals in the granular/molecular layers was tissues(Sutula et al., 1989; Houser et al., 1990; Babb et al., 1110.8 Frn on average, which was one-fourth of the total 1991). These aberrant collaterals were localized by Timm’s summed length of the mossy fibers (3898.5 pm on average). staining or dynorphin immunoreactivity as a band of reaction The size of the somata in neurons that had aberrant collat- product in the supragranularregion, especiallyin the inner mo- erals was significantly larger than that of neurons without lecular layer of the . Becausethe mossy fibersnor- such collaterals (p < 0.025). In four cases, filopodium-like mally extend their arbors to the polymorph (hilar) region of the fine processes were present near the and prox- dentate gyrus and CA3 field of the hippocampusproper, but do imal parts of the parent axon, suggesting that the aberrant not project to the molecular layer (Lorento de N6, 1934; Clai- collateral formation might be an ongoing process in these borne et al., 1986),epileptogenic dentate granule cellshave been tissues. The lack of control slices from normal living human thought to sprout axon collaterals beyond their normal arbo- hippocampus makes it difficult to assess to what extent the rization domain. The intensity of this sprouting showeda pos- itive correlation to the severity of seizure activities (Sutula et al., 1989) and the severity of hilar cell loss(Houser et al., 1990). Received June 9, 1992; revised Oct. 6, 1992; accepted Oct. 12, 1992. We thank Dr. F. E. Dudek for his assistance in settine uo our brain slice svstem. This finding in human epileptic hippocampuswas supportedby Dr. R. S. Fisher for letting us use his fluorescent micro&&e, Dr. D. M. Finch for several animal models of epilepsy. The kainate model (Tauck helpful comments and editorial assistance, and the members of the Clinical Neu- and Nadler, 1985), the kindling model (Sutula et al., 1988), and rophysiology Project for their clinical expertise. This research was conducted with permission of the UCLA Human Subjects Protection Committee and in accor- the pilocarpine model (Mello et al., 1990), all showeda dense dance with the guidelines of the Declaration of Helsinki. This work was supported band of Timm’s staining in the inner molecular layer, which bv NIH Grant NS02808. was indicative of reorganized mossy fiber paths. Physiological ~Correspondence should be addressed to Masako Isokawa, Ph.D., Brain Research Institute, CHS, UCLA, Los Angeles, CA 90024- 176 1. studiesin the kainate model suggestedthat sproutedfibers could Copyright 0 1993 Society for Neuroscience 0270-6474/93/13151 l-12$05.00/0 form recurrent excitatory circuits, constituting a possiblesub- 1512 lsokawa et al. * Single Mossy Fiber Axonal Systems in TLE Humans

strate of epileptogenicity (Tauck and Nadler, 1985; Cronin et Materials and Methods al., 1992). However, cellular anatomy to elucidate the paths and Patient selection. Hippocampalslices were obtained from 23 patients morphology of the sprouted fibers has not been reported in either who hadmedically intractable temporal lobe epilepsy and entered the human epileptic hippocampus or experimental models of epi- epilepsysurgery program for treatmentof their seizuredisorders. In all of them,there was evidence of a medialtemporal origin of epileptogenic lepsy. discharges.Patient selection was based on a seriesof evaluationsthat In the present study, we used a method of intracellular injec- includedEEG monitoring,neurological and cognitive testing, positron tion of tracers (biocytin or Lucifer yellow) into living single emissiontomography, magnetic resonance imaging, and depth electrode dentate granule cells in human epileptic hippocampus main- implants(Engel, 1987). Neuroanatomical examinations ofthe surgically tained in vitro, and directly visualized their axonal morphology removedspecimens included cell countsto assessneuronal death and hippocampalsclerosis in epileptogenicregions (Babb et al., 1984).All and arborization patterns. The advantage of using this method of thesewere performed on every hippocampalspecimen as part of is (1) only a single is stained at any one time per slice, routineexaminations for the identificationof epilepticfocal pathology. so that all fibers visualized can be consideredto originate in the The resultsof pathologicalexaminations were made available from the stainedneuron. This overcomesthe disadvantageof using the database of the UCLA EpilepsySurgery Program. Brain tissues. The hippocampaltissue for slicepreparation was ob- Golgi staining, which stains multiple neurons simultaneously tained directly from neurosurgeonsat the University of Californiaat and yet capriciously stains only a subpopulation of total fibers. LosAngeles. The resectedhippocampus was immediately immersed in (2) In caseswhere stained neurons have aberrant fibers, the oxygen-containingice-cold artnicial cerebrospinal fluid censistingof (in relationship between the parent axon in the hilus and the ab- mM)124 NaCl, 3 KCI, 2.4CaCI,. 26 NaHCO,. 1.3 M&O,. 1.24NaH,PO,. errant collateralsin the molecular layer can be qualitatively and and.10 glucose, pH 7.4. Slices’werecut perpendicular-tothe ant&or- posterioraxis of the hippocampusat a thicknessof 500pm (Vibroslicer, quantitatively studied. (3) Injecting a tracer in living cells may World PrecisionInstruments), and placed on the rampof a liquid-gas minimize the possibility of obscuringfine axonal branchesthat interfacerecording chamber (Haas et al., 1979)at 34 + 1°C.As a control might occur in autopsy or biopsy tissuesprepared for patho- comparison,normal rat hippocampalslices were prepared (10 males, logical examinations. 150-200am: CharlesRiver). The dentategranule cells were iniected ” with the sametracers and with the sameexperimental protocols using From the nature of this study, which usesliving human hip- our humanbrain slicechamber. This procedure provided (1) technical pocampalspecimens, our findingscannot be directly compared confirmationon intracellulardye fillingin humanepileptic neurons and to normal controls that would be obtained from disease-free (2) a guidelinefor criteria in selectingneurons with satisfactorytracer living human hippocampus.Ethical considerationspreclude the injections. use of suchtissue. In the present study, we made the following Intracellular injection of tracers. Either biocytin (Sigma)or Lucifer yellow(in lithiumchloride salt, Sinma) was used. Intracellular electrodes efforts to establish alternative control comparisons in inter- wereprepared from borosilicateglass pipettes, and filled with 2 M po- preting our data. First, we compared the present human hip- tassiumacetate with 2%biocytin (Horikawaand Armstrong, 1988) or pocampal slice data with normal rat hippocampal slice data. 0.1 M lithium chloridewith 5% Luciferyellow (Stewart,1978). Elec- The experiments were performed using the same techniques, trodeswere positioned at the dentategranule cell layer and lowered by a hydraulicmicromanipulator (Narishige, WR-88). The penetrationof equipment, and slice chambersin both groups, usingthe same a cell wasdetermined by measuringthe restingmembrane potential experimental protocol. This method provides a certain degree (-50 to -65 mV). In thosecases in whichcells were held 30 min or of support in contrastingdifferences in cell morphology between longerwith biocytin-tilledelectrode, biocytin was allowed to diffuse into an epileptic condition and a normal condition. It also rules out the intracellular space passively. Otherwise, direct negative current puls- sometechnical problemsthat might arisein interpreting human es (l-2 nA, 300 msec in duration; 0.5 Hz) were applied through an electrodefor severalminutes to inject the tracers.Lucifer yellow-in- results when we completely relied on others for control data. jected slices were fixed by 4% paraformaldehyde and mounted on glass However, this method risks underestimating speciesdifferences slides. The injected cells were examined and photographed under a in dentategranule cell morphology. For example,basal fluorescence microscope (Nikon Microphoto-FXA). Biocytin-injected have beenidentified as normal dendritic morphology of dentate slices were first fixed by 4% paraformaldehyde, and then sectioned into 30 Frn thickness and processed according to the method of Horikawa granulecells in monkeysand humans(Seress and Mrzljak, 1987) and Armstrong (1988). Biocytin-filled neurons were detected by a re- but not in rats (Seressand Pokorny, 1981; Desmondand Levy, action product of horseradish peroxidase and diaminobenzidine. Cell 1982). Thus, careful comparative studies in mossy fiber mor- morphology was examined and photographed under a light microscope phology in the phylogenetic tree cannot be neglectedin inter- (Zeiss Universal microscope), and traced by using a camera lucida. preting our data. At present,there seemsto be a consensusthat There were no noticeable differences between Lucifer yellow-injected neurons and biocytin-injected neurons in visualizing cell morphology the presenceof mossyfiber collaterals in the molecular layer is for the purposeof the presentstudy, so the datawere pooled for the a rare occurrence in normal rats (Seresset al., 199I), normal analysis.When measurements were made for quantification,they were monkeys (Seresset al., 1991) and normal humans (Houser et done on camera lucida drawings or directly from the processed speci- al., 1990). Second,the present findings of aberrant collaterals mens by light microscopy. in the supragranulelayer are supported by neuropathological examinations that assessthe location of mossy fibers in the Results human epileptic hippocampus.These examinations were made Seventy-two neuronsin 23 hippocampi from 23 patients were in the hippocampi usedin the presentstudy. The adjacentblocks intracellularly injected and visualized with either biocytin or of the hippocampusof the sameindividual patients usedin the Lucifer yellow. Among them, 19 neurons(11 neuronswith bio- presentstudy were examinedfor (1) cell density and (2) Timm’s cytin and 8 neuronswith Lucifer yellow) in 10 patients (1544 staining and/or dynorphin immunoreactivity by neuroanato- years of age;2 malesand 8 females)showed satisfactory injection mists in the UCLA Epilepsy Surgery Program. These neuro- of the tracers and were selectedfor analysis.Selection criteria pathological results were compared with normal autopsy spec- werebased on (1) the extent of stainedmossy fibers that showed imens, which were diseasefree. The presenceof cell loss and the arborization beyond the parent axon and primary collaterals mossyfiber reorganization (i.e., collateralsin the supragranular in the hilus, and (2) the presenceof clearly visible varicosities layer) in these patients has been attributed to temporal lobe along the fibers. When neuronsshowed densely stained axonal epilepsy (Babb et al., 1991, 1992). arbors, they also showed discretely stained somata and well- The Journal of Neuroscience, April 1993, 134) 1513

Figure 1. a, Dentate granule cell axons (the mossy fibers) were stained in the hilus by intracellular injection of Lucifer yellow. The vertically oriented parent axon extruded many branching collaterals. The soma and basal dendrites are seen at the top of the photomicrograph (out of focus). b, The fusiform-shaped varicosities were observed along the mossy fibers (stained by Lucifer yellow). c and d, The large, irregularly shaped varicosities (biocytin). e, A floweret-like terminal was observed along the fibers. Fusiform-shaped varicosities were also stained. f, Axonal lacunae (arrows) were present in addition to varicosities. These lacunae were larger in size than varicosities, and only the contour was stained. Scale bar: a, 30 pm; b, 13.5 wrn; c-e, 12 pm; 5 18 pm.

developed dendritic arbors with detailed spine structuresas re- hilar region. Cameralucida drawings of horizontal and vertical ported elsewhere(Isokawa and Levesque, 1991). arborsare shownin Figure 2, A and B, respectively. An example of a normal rat mossy fiber is shown in Figure 2C as a com- Mossyjbers in the hilus parison. In the human mossy fibers of the present group of Figure la showsa photomicrograph of mossy fiber arbors in specimens,there were a greaternumber of local collateralscom- the hilus stainedby Lucifer yellow. In this neuron, axons showed pared with normal rats, and these collaterals formed complex a vertically oriented arborization pattern. The collaterals arborization patterns. In six neurons, densely formed plexuses branched out from both proximal and distal parts of the parent were observed in some collaterals. An example is shown in axon. The vertically oriented arborization pattern wasobserved Figure 4C (arrow 5). in eight neurons. In the remaining neurons (N = 1 l), the ar- Varicosities werepresent along the axon collaterals.They were borization pattern was predominantly oriented horizontally, densely stained by both biocytin and Lucifer yellow. Two dif- showingwell-developed collateralscovering a wide rangeof the ferent shapeswere identified: the fusiform shapewith a length

The Journal of Neuroscience, April 1993, f3(4) 1515 of 2-5 Km (Fig. 1b) and the irregular shape, 4-5 pm in size (Fig. somata were dislocated to the molecular layer (Houser, 1990). lc,d). Fusiform varicosities were regularly present along the Finally, in the remaining three neurons, single collaterals were fibers at approximately 7-18 pm intervals in most cases, but identified in the molecular layer. We did not observe plexuses occasionally at 35-50 pm intervals. The irregular varicosities in these collaterals. An example is shown in Figure 3d. In this were larger in size and present along the fibers and at endings. photomicrograph, a single collateral branched out from the par- They were more sparsely located than the fusiform varicosities. ent axon at a proximal level and extended back towards the The locations of the large, irregular varicosities in axonal arbors molecular layer. Figure 4C shows another example of a single were plotted in the camera lucida drawings whenever they were aberrant collateral in the molecular layer. In this case, the entire clearly visible with low magnification (see Figs. 2,4). These two length of the collateral could not be traced; thus, a branching types of varicosities were also common in the control rat mossy point of this collateral from the parent axon was not identified. fibers. The locations of the large, irregular varicosities in one of No plexuses were observed in this collateral. our control rat dentate granule cells are depicted in Figure 2C. As shown in Figure 3b, aberrant collaterals bore large irregular In addition, in human mossy fibers, round varicosities were varicosities. They were especially packed in plexuses. Fusiform- observed at branching points. The diameter of these varicosities shaped varicosities were also present (Fig. 3~). The sizes of these were 2-3 pm. In two neurons, one stained with biocytin and varicosities were not different from those observed in the hilus. the other stained with Lucifer yellow, terminals shaped like a Although, on occasion, large and clearly visible lacunae were floweret were observed. These terminals looked like five to eight present near branching points of aberrant collaterals (Fig. 3d), round varicosities with short stems extruding from the same there were rarely lacunae observed in the collaterals located in spot of the collateral and forming a blooming flower petal ap- the molecular layer. In addition, filopodium-like fine processes pearance (Fig. le). Some of the irregular varicosities, especially were observed in four neurons. These processes were located larger ones, could be irregularly aggregated flowerets. In addition near the axon hillock (Fig. 5A) and at the proximal part of the to the varicosities, lacunae were observed as large swellings parent axon below the hillock (Fig. 5B). In 75% of the cases in along the fibers (arrows in Fig. lf). The shape of the lacunae which neurons showed these filopodia (three out of four neu- was mostly round, and the size was considerably larger (5-7 pm rons), they also showed mossy fiber collaterals in either the in diameter) than that of the varicosities. In contrast to the dentate gyrus granular layer or the molecular layer. We observed varicosities, which were stained dark, the inside of lacunae were terminal enlargement in some of these filopodia, which could pale when they were labeled with biocytin or Lucifer yellow. be a putative growth cone. An example is shown in Figure 5. The extent of aberrant collaterals in individual neurons was Axon collaterals in the granule and molecular layers estimated quantitatively by calculating the summed lengths of In addition to arbors in the hilus, mossy fiber collaterals were axon branches in the dentate gyrus. Due to a technical limitation observed in the dentate granule cell layer and/or molecular layer of bleaching Lucifer yellow during the observation, this calcu- in nine human dentate granule cells. It was a common feature lation was done only in biocytin-filled neurons on camera lucida for these aberrant collaterals to branch out from the parent axon drawings. The average length of summed mossy fibers in the at a proximal point, sometimes even before the initial branching hilus was 3698.5 pm f 46 1.9 SEM in 10 neurons. On the other point of the hilar collaterals. A representative example of a hand, the average length of summed mossy fibers in the granule biocytin-stained aberrant mossy fiber collateral is shown in Fig- layer and molecular layer was 1 I 10.8 pm f 345.9 SEM in four ure 3a. The neuron in this figure showed two collaterals that neurons. This indicates that, on average, a quarter of the total extended to the molecular layer by branching out from the par- mossy fiber lengths became aberrant collaterals (Table 1). In ent axon near the soma. These aberrant collaterals passed through addition, the size of somata in neurons with and without ab- the granule layer and formed a dense plexus with many terminal errant collaterals was compared. This was done in all neurons boutons in the molecular layer (Fig. 3b,c). One collateral ex- stained by biocytin or Lucifer yellow by measuring the longer tended farther beyond the plexus and reached its own dendritic and shorter dimensions of ovoid-shaped somata. As shown in domain (Fig. 3a, upper left). A camera lucida drawing of this Table 2, the soma size was significantly greater in neurons with neuron is shown in Figure 4A. The aberrant collateral plexus is aberrant collaterals than in neurons without those collaterals (p indicated by arrow 1. A dense plexus and intensive collateral < 0.025, t = 2.19; the longer dimension is called length and the formation in the molecular layer were observed in two of nine shorter dimension is called width in the table). neurons. In another four of nine neurons, mossy fiber collaterals were mostly limited to the granule cell layer (Fig. 4B). They Discussion formed dense plexuses in the granule cell layer, and the somata By using an intracellular labeling technique, we have elucidated were surrounded by fine axon collaterals (arrow 2 in Fig. 4B). details of morphology and arborization patterns of the mossy Some of them extended to the molecular layer (arrow 3 in Fig. fibers in epileptic human hippocampus. Extensively ramified 4B). As the somata of these neurons were located in the granule axon collaterals of dentate granule cells were visualized, as pre- cell layer, the presence of aberrant collaterals around the somata viously shown by Golgi studies (Scheibel et al., 1974). However, was not due to granule cell dispersion in which granule cell in addition to cell morphology with Golgi-like appearance, our t Figure 2. Cameralucida drawings of two human dentate granule cells (A and B) and a rat dentate granule cell (C). A, A well-developed axon arbor covered a wide horizontal domain in the hilus. The large, irregular varicosities are illustrated along the fibers whenever they were obvious with low magnification. Floweret-like terminal boutons were observed in proximal and distal branches (arrows). B, A vertically oriented parent axon was accompanied by the collaterals that branched out from various portions of the parent axon. C, A dentate granule cell in normal rat hippocampus. A well-developed axon arbor and clearly identifiable varicosities were observed. The distal ends of most of the dendrites were clearly visualized as bending tips staying at the outer edge of the molecular layer without crossing the hippocampal fissure (asterisks). These features indicate that the dye was satisfactorily transported throughout the dendrites. Scale bar: A and B, 100 brn; C, 45.5 pm.

The Journal of Neuroscience, April 1993, 13(4) 1517

Table 1. Lengths of summed axon branches (mossy fibers) in dentate gyrus

Locations Average t SEM Range Hilus 3698.5 pm t- 461.9 SEM 2 193-6496 pm (N = 10) Dentate gyrus (GL and ML) 1110.8 pm & 345.9 SEM 620-2133 pm (N=4) Calculations were done on neurons stained by biocytin only. GL, granule cell layer; ML, molecular layer.

Table 2. Size of somata in neurons with or without aberrant collaterals of the mossy fiber

Cells [Average length (L) + SEMI x [Average width (w) + SEMI With aberrant collaterals [25.0(L) + 2.11 x [14.0(w) f 1.31 pm* (N=9) Without aberrant collaterals [19.1 (L) ? 1.61 x [12.8 (lV) ? 1.41 pm (N = 10) * Measurement was done in all the neurons stained by either biocytin or Lucifer yellow. The difference was significant by Student’s t test at the level of p < 0.025, t = 2.19.

Table 3. Relationships between mossy fiber arborization patterns and other neuropathological findings in the dentate gyrus

Intra- Age at Mossy fiber cellularly Mossy fibers in: Pa- &e seizure onset Granule layer reorganization stained tient (yr) Sex (yr) Hilar cell loss cell loss detected by: neurons Hilus GL ML Filopodia A 36 F 21 0% 2 l-32% Timm’s staining #l (B) x (Pl) (27 t 5.5 SEM) #2 (B) X B 24 F 4 5% 3846% Timm’s staining #l (W X (42 t- 4.0 SEM) #2 (W X #3 (W x (Pl) C 44 M 31 0% 35-48% Dynorphin staining #l 0-Y) X x (Pl) (42 f 6.5 SEM) D 29 F 2.5 50% 4246% Dynorphin staining #l O-Y) X (44 + 2.0 SEM) 10 months). E 38 F 13 43-50% 30-52% Dynorphin staining #l U-Y) X X (4.7 + 3.5 SEM) (38 k 5.1 SEM) #2 O-Y) X X X F 19 F 1.2 65% 50-51% Dynorphin staining #l 0-Y) X (51 + 0.5 SEM) #2 (B) x (PI) G 29 F 5d 75% 58-66% Timm’s and dynorphin #l (B) X (forceps (62 k 4.0 SEM) staining #2 (B) x (P)) x (PI) delivery, #3 U-9 x (PI) x (PI) seizure at X birth) H 25 M 10 24-75% 53-78% Dynorphin staining #l (W X x (PI) (50 + 25.5 SEM) (62 + 5.6 SEM) #2 (LY) X X X I 15 F 3 75% 68-73% No data #l U-Y) X (febrile (62 2 9.0 SEM) #2 U-Y) X X seizure, 1.3 yr) J 30 F 1.5 No data No data No data #l (B) x (PI) x (PI) Nemopathological information was made available from the data base for UCLA Epilepsy Surgery program. When cell count was done in more than two locations, the range and the average with SEM are shown. GL, granule cell layer; ML, molecular layer; pl, plexus; B, biocytin-filled neuron; LY, Lucifer yellow-filled neurons. t Figure 3. Aberrant mossy fiber collaterals in the human epileptic hippocampus. a, The collaterals formed a dense plexus in the molecular layer (biocytin). One collateral reached to its own dendritic domain (upper left). The area photographed here is shown by a polygon in Figure 4A. b and c, High-magnification photomicrographs that show parts of plexuses formed by aberrant collaterals. Large, irregular varicosities (b) and fusiform- shaped or round varicosities (c) were observed in these plexuses. d, A single mossy fiber collateral branched from a proximal part of the parent axon, and extended back to the molecular layer. Axonal lacunae were visible near the branching point (Lucifer yellow). This granule cell had a basal . Scale bar: a, 17 pm; b, 7.5 pm; c, 10 pm; d, 30 pm. Figure 4. Camera lucida drawings of human dentate granule cells with aberrant collaterals. A, Mossy fiber collaterals formed a well-developed, dense plexus in the molecular layer (a polygon indicated by arrow I shows the area photographed in Fig. 3~). Arrowheads indicate the distal ends of dendrites at the outer edge of the molecular layer. B, Mossy fiber collaterals formed a plexus around a soma (fine lines shown by arrow 2). Some of those fibers extended to the molecular layer (fine lines shown by arrow 3). Dendrites were drawn with thicker lines. C, An example of a single collateral observed in the molecular layer (arrow 4). It had a few branching fibers. Arrow 5 indicates a mossy fiber plexus in the hilus. The varicosities were not all obvious with low magnification. The granule cell layer (CL) is indicated by dashed lines. Scale bar: A-C, 160 pm. The Journal of Neuroscience, April 1993, f3(4) 1519 technique of staining only one cell at a time offered an advantage in showing that all the fibers stained in a given slice originated from a single neuron. This allowed us to perform qualitative and quantitative analyses of arborization patterns and domains of single mossy fiber axonal systems in human epileptic dentate granule cells, providing cellular evidence for aberrant mossy fiber paths and their detailed morphology. Neurological and neuropathological conditions, examined in individual patients used in the present study, are discussed in relation to our data of intracellularly stained aberrant collaterals. This clinical information was made available from the CNP data base for the UCLA Epilepsy Surgery Program (see Table 3). Cell loss in the hippocampi and seizure histories of individual patients used in the present study were similar to those of pre- vious reports on mossy fiber reorganization in human temporal lobe epilepsy (Sutula et al., 1989; Houser et al., 1990; Babb et al., 1991). In addition, the presence of aberrant mossy fiber collaterals was confirmed in the present patients by Timm’s staining and/or dynorphin immunohistochemistry in compar- ison with autopsy control (Babb et al., 199 1, 1992). These find- ings support our conclusion that the intracellularly stained ab- errant axon collaterals in the granule/molecular layers in the present study probably represent trajectories of reorganized mossy fibers that are associated with medically intractable hu- man temporal lobe epilepsy.

Aberrant collaterals, cell loss, and seizure histories Animal studies demonstrated that extensive sprouting of the mossyfibers occurred in the supragranularlayer following deaf- ferentation of the dentate granule cells by the lesion of hilar projections (Laurberg and Zimmer, 1981). In human temporal lobe epilepsy, pathology often includes hilar cell loss. Table 3 summarizesthe arborization patterns of the mossyfibers in 19 neurons in 10 patients in relation to seizure histories and se- lective cell lossin the dentate gyrus. A large variability of cell lossamong patients wasobserved in the hilus (5-75% in seven patients and no cell lossin two patients). In contrast, the cell losswas more consistent in the granule cell layer (27-62% in nine patients). When the severity of cell loss was in the range of Xl-75% in the hilus and 62% in the granule cell layer on average, most of the cells that were sampledfor intracellular staining showedaberrant collaterals in the granule cell layer or in the molecular layer (patients G, H, and I). Considering an extremely low samplingprobability of intracellular staining in a given hippocampal slice, this finding allowed us to estimate that a high proportion of the dentate granule cells could have aberrant collaterals in an advanced stageof cell lossin the pres- ent patient group. This estimatewas supportedby a report that these patients showedstrong Timm’s staining and/or positive immunoreactivity for dynorphin in the molecular layer (Babb Figure 5. Filopodium-like fine processes were observed in the human et al., 1991, 1992; T. Babb, unpublished observations). Houser dentate granule cells in epileptogenic hippocampi. They were present et al. (1990) also reported a positive correlation of the extent of in the vicinity of the axon hillock (arrows in A) and in the parent axon aberrant collaterals to the severity of hilar cell loss. However, immediately beneath the hillock (B). Both pictures were taken from the same neuron stained by Lucifer yellow. Terminal enlargement was ob- in caseswhere the degree of cell loss was lessintense in the served in one of these filopodia, which could be a putative growth cone granule cell layer or in the hilus, we did not see any specific (pointed by the bottom arrow in A). Scale bar: A and B, 10 pm. correlation between hilar cell lossand the presenceof aberrant collaterals that were stained by our intracellular injection tech- nique (Table 3). The lack of intracellularly stained aberrant eralsand the amount of supragranular Timm’s staining. In one collaterals in neurons of the specimensthat showedless severe case,it was possibleto observe well-developed mossyfiber col- cell losscould reflect the small samplesize in intracellular tracer laterals in the granule cell layer when cell loss was observed injections. Further studiesare needed to correlate the number only in the granule cell layer, with no detectablecell lossin the of neurons that showed intracellularly stained aberrant collat- hilus (patient C). Although the type of neuronslost in the granule 1520 lsokawa et al. * Single Mossy Fiber Axonal Systems in TLE Humans cell layer is not as well understood as in the hilus (deLanerolle mine when aberrant collaterals were formed and whether they et al., 1989; Sloviter et al., 199 I), disruption of local circuits for are still in the processof growth. However, the presenceof the granule cells (Ribak et al., 1982) might potentiate the for- filopodium-like fine processesin the samemossy fibers that bear mation of supragranular mossy fibers. aberrant collateralssuggests that different degreesand stagesof Among the patients who showed aberrant collaterals in the collateral formation might be present in our epileptic hippo- present study, half of them (three of six) had histories of early campal specimens.An active regenerationof axonal collaterals childhood seizures. The clearest example of extensively devel- might be an ongoing processin medically intractable epileptic oped aberrant collaterals in the supragranular layer was iden- human hippocampus. tified in a patient who had experienced tonic-clonic seizures since birth (patient G in Table 3). Thus, as reported by Houser et al. (1990) early childhood seizures contribute to mossy fiber Arborization patterns of aberrant collaterals collateralization. However, this observation was not generalized Aberrant mossy fiber collaterals were observed in the dentate to the rest of the patients in the present study. This is because granule cell layer and in the molecular layer in 9 of 19 total extended arbors of aberrant collaterals were found in only half neuronsstained in the presenthuman specimens.The extent of of the present patients who had febrile seizures or early seizure arborization of the aberrant collaterals in these nine neurons onset. The rest of the patients experienced their first seizures at varied. Some of the neurons showedextensive plexuses,and ages 13-3 1 even though the sampled neurons for intracellular others showedsingle collaterals with a few short branches.One injection in these patients showed extensively developed aber- possibleexplanation for this variability may be technical, per- rant collaterals. We cannot rule out the possibility that these hapsdue to insufficient transport of injected tracers. However, results reflect limited or biased sampling of intracellular stain- the fiber lengths that were stained by intracellular labeling av- ing. However, it appears that a history of early age seizures is eraged3698.5 pm in the present study (seeTable 1). This mea- not enough by itself to be considered as a determining factor of surement was 50% longer than the summed length of mossy aberrant collateral formation. fiber collaterals reported in the rat dentate granule cells (2300 pm) (Claibome et al., 1986).Thus, it is unlikely that our tracer Varicosities, lacunae, andfilopodia transport was insufficient. Second, if the aberrant collaterals Intracellular injections of tracers revealed two types of varicos- tended to have a smaller diameter compared to normally ex- ities alongthe fibers: small fusiform-shapedswellings and larger isting mossy fiber collaterals, they might not be satisfactorily irregularly shapedexpansions. Both were present in all 19 neu- stained.Indeed, someregenerated axon collateralswere reported rons regardlessof the presenceof aberrant fibers. Similar shapes to have smaller diameters in certain instances(Frotscher and of varicositieswere observed in our control rat dentate granule Zimmer, 1983). However, in our case,the diametersof aberrant cells, as well as in normal rat mossyfibers that were stained by collaterals were not necessarilysmaller than that of hilar col- the Golgi method (Blackstad and Kjaerheim, 1961) and by in- laterals. As shown in Figure 3, aberrant collateralswere some- tracellular HRP injections (Claibome et al., 1986). In humans, times larger in diameter than hilar axons. Third, the absenceof however,a previousGolgi study identified only irregularly shaped aberrant collaterals in some neurons could result from a pos- expansions as varicosities in normal autopsy hippocampus sibility that those collateralswere not in the samehippocampal (Scheibelet al., 1974):fusiform-shaped swellings were reported slice in which the somatawere located; that is, they could have as being pathological and unique to epileptic hippocampus. beenlocated more rostrally or caudally. This concern arosefrom Scheibelet al. (1974) reported that fusiform-shaped swellings reports on the rat indicating that hilar neurons projected to becameenlarged and formed stringsof beadsin advancedpatho- dentate granule cells at different levels in the anterior-posterior logical stages. axis (Amaral and Insausti, 1989) and in the horizontal axis In addition to varicosities, there were lacunaein the present (Isokawa-Akessonand Finch, 1989). However, in contrast to human specimens.As shown in Figures lf and 3d, the shapeof hilar cells, mossyfibers in the rat hilus were reported to have a lacunaewas often round, and they were larger than varicosities lamellar organization in which individual lamellae were con- and looked hollow inside. The contours of lacunae were not fined within the thicknessof the slices(500 Mm) (Claibome, et evenly stainedbut had one or a few denselystained spots. Some al., 1986). Although this evidence does not rule out the possi- of them looked like abnormally swollen varicosities. It needs bility that human granule cell axons, especiallyaberrant collat- to be considered whether the lacunae are pathological. They erals,may travel substantially alongthe transverseaxis, we have resemblednodules in preterminal fibersof epileptic tissue.Their evidence that indirectly rejects this possibility. Our data dem- shapesand sizeswere similar to pathological nodulesdescribed onstratethat the somataof neuronsshowing aberrant collaterals in Golgi-stained epileptic granule cells (Scheibel et al., 1974). were significantly larger than neurons that did not show such Considering that our tissue is from epileptic patients, it is not collaterals. This suggeststhat an increasedfield of innervation surprisingthat we observed such pathologically deformed var- and increasedmetabolic demandson the cell may produce so- icosities.On the other hand, formation of lacunaewas reported ma1 hypertrophy (Grafstein and McQuarrie, 1978). Such an to be a sign of rapid growth of the cytoplasm in developing example has beenreported in the regeneratingtrochlear human cerebralcortical neurons(Purpura, 1975). In theseneu- in adult mammalian CNS (Murphy et al., 1990). Thus, if we rons, lacunaewere often observed with filopodia, and they were had missedsubstantial numbers of aberrant collateralsthat were consideredas morphologic featuresof cell growth. Thus, lacunae hypothesized to be located outside of our slice specimens,we may not be simply regardedas pathologically swollen varicos- would not have found a positive correlation between enlarged ities. The determination of whether the lacunae are normally somata and the presenceof aberrant collaterals. Accordingly, functioning synaptic varicosities needsto await electron micro- we assumethat most of the aberrant collateralswere observed scopic study to verify the presenceof a pre- and postsynaptic in the present study. Basedon this evidence, we conclude that apparatus. It is beyond the scopeof the present study to deter- our findings of variable amounts of aberrant mossy fiber col- The Journal of Neuroscience, April 1993, 13(4) 1521

laterals in the granule/molecular layers represent differing de- leptic hippocampus: evidence for feedforward excitation. Dendron grees of arborization. 1:7-25. Bekkers JM, Stevens CF (1991) Excitatory and inhibitory autaptic Aberrant collaterals appear to form . Electron mi- currents in isolated hippocampal neurons maintained in cell culture. croscopy has identified that zinc-containing terminals formed Proc Nat1 Acad Sci USA 88:7834-7838. synapses on a dendrite of an unidentified neuron in human Blackstad TW, Kjaerheim A (1961) Special axo-dendritic synapses in epileptic dentate gyrus (Babb et al., 1991). However, whether the hippocampal cortex: electron and light microscopic studies on the aberrant collaterals terminated on their own dendrites, on neigh- layer of mossy fibers. J Comp Neurol 117: 133-l 59. Claibome BJ, Amaral DG, Cowan WM (1986) A light and electron boring granule cell dendrites, or on inhibitory in microscopic analysis of the mossy fibers of the rat dentate gyrus. J the granule/molecular layers cannot be addressed in the present Comp Neurol 246:435458. study. This is because we did not stain any target cells of those Cronin J, Obenaus A, Houser CR, Dudek FE (1992) Electrophysiology aberrant collaterals. In our study, in four of nine neurons that of dentate granule cells after kainate-induced synaptic reorganization of the mossy fibers. Brain Res 573:305-310. showed aberrant collaterals, their somata and proximal den- deLanerolle N, Kim JH, Robbins RJ, Spencer DD (1989) Hippocam- drites were surrounded by these collaterals and their plexuses. pal loss and plasticity in human temporal lobe epilepsy. Furthermore, two neurons were found to have aberrant collat- Brain Res 495:387-395. erals that reached their own dendritic domains and formed ex- Desmond NL, Levy WB (1982) A quantitative anatomical study of tensive plexuses. Whether these collaterals made synapses on the granule cell dkndritic~fields of the rat dentate gyrus using a novel orobabilistic method. J Coma Neurol 2 12: 13 1-145. their own dendrites, that is, autapses, could not be determined Engel J Jr (1987) Approaches to localization ofthe epileptogenic lesion. under light microscopic observations. However, in hippocampal In: Surgical treatment of the (Engel J Jr, ed), pp 75-96. culture, formation has been reported in neurons grown New York: Raven. by themselves in isolation (Bekkers and Stevens, 1991; Segal, Frotscher M, Zimmer J (1983) Lesion-induced mossy fibers to the molecular layer of the rat : identification of postsynaptic 1991) and in neurons that were grown as a pair, also forming granule cells by the Golgi-EM technique. J Comp Neurol 215:299- mutual synapses (C. Stevens, unpublished observations). This 311. evidence suggests the possibility that dentate granule cells form Grafstein B, McQuarrie IG (1978) The role of the cell body in presynaptic terminals on themselves through aberrant collat- axonal regeneration. In: Neuronal plasticity (Cotman CW, ed), pp 155-195. New York: Raven. erals. On the other hand, a physiological study in an animal Haas HL, Schaerer B, Vosmansky M (1979) A simple perfusion cham- model of hippocampal epilepsy indicated a possibility that ab- ber for the study of slices in vitro. J Neurosci Methods errant collaterals on inhibitory intemeurons in the den- 1:323-325. tate gyrus (Sloviter, 1992). Horikawa K, Armstrong WE (1988) A versatile means of intracellular Studies in an experimentally produced aberrant mossy fiber labeling: injection of biocytin and its detection with avidin conjugates. system (Cronin et al., 1992) and in slice preparations from ep- J Neurosci Methods 25: l-l 1. Houser CR ( 1990) Granule cell dispersion in the dentate gyrus of ileptic patients (Masukawa et al., 1989) suggested that slightly humans with temporal lobe epilepsy. Brain Res 535: 195-204. decreased inhibition might be a mechanism for producing syn- Houser CR, Miyashiro JE, Swartz BE, Walsh GO, Rich JR, Delgado- chronous dentate gyrus activity. However, in human epileptic Escueta AV (1990) Altered patterns of dynorphin immunoreactivity hippocampus, GABAergic neurons are preserved (Babb et al., suggest mossy fiber reorganization in human hippocampal epilepsy. 1989). During the depolarization of mossy fiber terminals for J Neurosci 10:267-282. Howell GA, Welch MG, Fredrichson CJ (1984) Stimulation induced the release of glutamate, they corelease zinc, an antagonist of a uptake and release of zinc in hippocampal slices. Nature 325:522- subclass of excitatory amino acid receptors (Assaf and Chung, 525. 1984; Howell et al., 1984), and dynorphin, an inhibitory mod- Isokawa M, Levesque MF (199 1) Increased NMDA responses and ulator (Siggins et al., 1986; Tenian et al., 1988). Intracellular dendritic degeneration in human epileptic hippocampal neurons in recording from epileptic dentate granule cells in humans showed slices. Neurosci Lett 132:2 12-2 16. that many neurons possess quite normal membrane properties Isokawa M, Avanzini G, Finch DM, Babb TL, Levesque MF (1991) Physiologic properties of human dentate granule cells in slices pre- and synaptic transmission (Isokawa et al., 1991). Further in- pared from epileptic patients. Epilepsy Res 9:242-250. vestigations are required to determine whether the excitatory- Isokawa-Akesson M, Finch D (1989) Neurophysiological character- inhibitory balance of neuronal circuits in the dentate gyrus is ization of commissurally projecting dentate neurons in the rat. Brain permanently altered toward enhanced excitation in the anatom- Res 480:92-104. ically reorganized, seizure-susceptible hippocampus. Laurberg S, Zimmer J (198 1) Lesion-induced sprouting of hippocam- pal mossy fiber collaterals to the fascia dentata in developing and adult rats. J Comp Neurol200:433-459. References Lorento de No R (1934) Studies on the structure of the . II. continuation of the study of the ammonic system. J Psycho1 Neurol Amaml DG, Insausti R (1990) The . In: The 46:113-177. human (Paxinos G, ed), pp 71 l-755. New York: Masukawa LM, Higashima M, Kim JH, Spencer DD (1989) Epilep- Academic. tiform discharges evoked in hippocampal brain slices from epileptic Assaf SY, Chung S (1984) Release of endogenous Zn*+ from brain patients. Brain Res 493: 168-174. tissue during activity. Nature 308:734-736.- Mello LEAM, Cavalheiro EA, Babb TL, Kupfer WR, Pretorius JK, Tan Babb TL. Brown WJ. Pretorius J. Davennort C. Lieb JP. Crandall PH AM, Finch DM (1990) Pilocarpine-induced status epilepticus leads (1984) Temporal ‘lobe volumetric ceil densities in temporal lobe to chronic seizures and mossy fiber sprouting in the hippocampus. epilepsy. Epilepsia 25:729-740. Epilepsia 3 1:675. Babb TL, Pretorius JK, Kupfer W, Crandall PH (1989) Glutamate Murphy EH, Brown J, Lannuzzelli PG, Baker R (1990) Regeneration decarboxylase-immunoreactive neurons are preserved in human ep- and soma size changes following axotomy of the trochlear nerve. J ileptic hippocampus. J Neurosci 9:2562-2574. Comp Neurol292:524-536. Babb TL, Kupfer WR, Pretorius JK, Crandall PH, Levesque MF (199 1) Purpura DP (1975) Dendritic differentiation in human cerebral cortex: Synaptic reorganization by mossy fibers in human epileptic fascia normal and aberrant developmental patterns. In: Advances in neu- dentata. Neuroscience 42:351-363. rology, Vol 12, physiology and pathology of dendrites (Kreutzberg Babb TL, Pretorius JK, Kupfer WR, Mathern GW, Crandall PH, Lev- GW, ed), pp 91-116. New York: Raven. esque MF (1992) Aberrant synaptic reorganization in human epi- Ribak CE, Bradbume RM, Harris AB (1982) A preferential loss of 1522 lsokawa et al. * Single Mossy Fiber Axonal Systems in TLE Humans

GABAergic, symmetric synapses in epileptic foci: a quantitative ul- organization in the dentate gyrus-_ of kainate-treated rats. Neurosci trastructural analysis of monkey neocortex. J Neurosci 2: 1725-l 735. Leit 137:91-96. Scheibel ME, Crandall PH, Scheibel AB (1974) The hippocampal- Sloviter RS, Lossas AL, Barbaro NM, Laxer KD (199 1) Calcium- dentate complex in temporal lobe epilepsy. Epilepsia 1555-80. binding protein (Calbindin-D28K) and narvalbumin immunocvto- Segal MM (199 1) Epileptiform activity in microcultures containing chemis& in the normal and epileptic human hippocampus. J Camp one excitatory hippocampal neuron. J Neurophysiol 65:761-770. Neurol308:38 l-396. Seress L, Mrzljak L (1987) Basal dendrites of granule cells are normal Stewart W (1978) Intracellular marking of neurons with a highly flu- features of the fetal and adult dentate gyrus of both monkey and orescent naphthalimide dye. Cell 14:741-759. human hippocampal formations. Brain Res 405: 169-174. Sutula T, Xiao-Xian H, Cavazos J, Scott G (1988) Synaptic reorga- Seress L, Pokorny J (198 1) Structure of the granular layer of the rat nization in the hippocampus induced by abnormal functional activity. dentate gyrus: a light microscopic and Golgi study. J Anat 133: 18 l- Science 239: 1147-l 150. 195. Sutula T, Cascino G, Cavazos J, Parada I, Ramires L (1989) Mossy Seress L, Bakay RAE, Ribak CE (199 1) Postsynaptic targets of Timm- fiber synaptic reorganization in the epileptic human temporal lobe. labeled mossy fibers in rats and monkey hippocampus: a light and Ann Neural 26:321-330. electron microscopic study. Sot Neurosci Abstr 17: 127. Tauck DL, Nadler JV (1985) Evidence of functional mossy fiber Signs GR, Henriksen SJ, Chavkin C, Gruol D (1986) Opioid peptides sprouting in hippocampal formation of kainic acid-treated rats. J and in the limbic system: cellular mechanisms. In: Neurosci 5:1016-1022. Advances in neurology, Vol 44, Baiic mechanisms of the epilepsies Terrian DM, Johnston D, Claibome BJ, Ansah-Yiadom R, Strittmatter

(Delgado-Escueta AV. Ward AA, Woodburv _ DM. Porter RJ. eds).II WJ, Rea MA (1988) Glutamate and dynorphin release from a sub- pp 50 l-5 12. New York: Raven. ’ cellular fraction enriched in hippocampal mossy fiber synaptosomes. Sloviter RS (1992) Possible functional consequences of synaptic re- Brain Res Bull 2 1:343-35 1.