terminals of GABAergic chandelier cells are lost at epileptic foci.

Axon terminals of chandelier cells were analyzed in monkeys with cortical focal epilepsy produced by alumina gel to determine if this type of GABAergic terminal is lost at epileptic foci. These terminals form a dense plexus with the axon initial segments of pyramidal neurons, especially those in layers II and III. Axon initial segments of pyramidal neurons were traced for at least 40 micron in serial thin sections and beyond this point were observed to become myelinated. In single sections, 10-15 axon terminals were found to form symmetric synapses throughout the entire length of the axon initial segments from nonepileptic preparations and were observed to synapse with only these structures and not adjacent dendrites or spines. In epileptic cortex, the axon initial segments of pyramidal neurons were apposed by glial profiles that contained clusters of filaments typical of reactive astrocytes. Only a few, small axon terminals were observed to form symmetric synapses with these axon initial segments. Thus, the chandelier cell axons appeared to degenerate in epileptic cortex. The highly strategic site of GABAergic inhibitory synapses on axon initial segments suggests that they exert a strong influence on the output of pyramidal cells. The near absence of these chandelier cell axons in epileptic foci most likely contributes to the hyperexcitability of neurons.


INTRODt r('TI()N
It is now apparent that in some varieties of epilepsy, a severe deficit occurs in the cortical GABAergic system. Numerous studies of experimental models that resemble post-traumatic epilepsy have shown a loss of GABAergic terminals in epileptic foei e-' ca. A quantitative assessment at the electron microscopic level has indicated that this loss is preferential for GABAergic, symmetric synapses2e. Biochemical studies also support this finding in that indices for other neurotransmitters were not reduced as severely as those for GABA I,e,.<. These findings in animal models have received corroboration from biochemical results of human epileptic loci >. Therefore, a loss of GABAergic terminals appears to be a hallmark of epileptic loci.
The initial inmmnocytochemical study on the mor- Ilil06-S993i85 $03.31i ~ IriS5 El~cxicr Science Publishers B.V /\rbib :~ first described this neuron in Golgi prepara-. tions. This neuron is characterized by a small soma with aspinous dendrites in layers I1 and 111. The chandelier cell's axon forms numerous vertical chains tha! appear to be apposed to pyramidal cell apical dendrites in light microscopic preparations Is,~3. Howe~cr, Somogyi > subsequently demonstrated in eleclron microscopic preparations that these axons form swnmetric synapses with the axon initial segment,~ of p~ramidal cells in layers 11 and 111. Further studies have confirmed these electron microscopic obserw~tions in numerous species anti cortical regions +'`>.> including the hippocampus ~1. The morphology of these lerminals and the localization of GAD-positive reaction product within these terminals s,>.et,~: have indicated that they are GABAergic. These data suggest that chandelier cell axon terminals exert a strong inhibitory influence on the output of pyramidal cells. A loss of such a plexus of axons might also contribute to ~hc hyperexcitability of neurons in an epileptic focus.

MATERIALS AND METHODS
The present study utilized specimens obtained from 3 of the 5 experimental monkeys from a prcvious study-'~. All of the monkeys had received alumina gel applications to the left cerebral hemispheres to produce seizure foci. Two of these monkeys (animals 3 and 5 from Ribak et al.>) received intracortical injections directly into both pre-and postcentral gyri. The remaining experimental monkey (animal 2 from Ribak et al. ~3) had an injection of alumina gel limited to the subarachnoid space in the area of the central sulcus. Electrocorticography of all experimental animals verified epileptic foci > and subsequently, the monkeys were fixed by intracardiac perfusions of a mixture of two aldehydes. The fixative solution contained 4% paraformaldehyde, 0.1% glutaraldehyde. and 0.002% CaCI~ in a 0.12 M phosphate buffer.
Blocks of cortical tissue were obtained from the epileptic focus and the homologous area in the contralateral nonepileptic cortex. All blocks were sectioned on a Sorvall TC-2 tissue sectioner at a thickness of 150 urn. Specimens that contained the entire cortical thickness were cut from these sections. These specimens were postfixed in 2% OsOa for I h. dehydrated in ethanol and embedded in Epon. For light microscopy, semithin 1 um sections were cut from the embedded specimens and stained with (I.05cf toluidine blue. Specimens were oriented in the ultramicrotome to obtain SeCtions thai yielded the longest segments of apical dendrites. The ~djacent thin sections from these specimens usually had the most number of identified axon initial segments These sections were then stained with uranyl acetate and lead citrate and examined ot~ ~ × 2 mm formvar, coated slot grids with the electron microscopc, RESUI:I-S All observations were obtained from electron microscopic preparations of monkey sensorimotor cortex. The analysis in the present study was limited to the axon initial segments of pyramidal cells in layers 11, IlI and V. Since the findings from the nonepiteptic hemisphere were similar to those described in a previous study of normal primate sensorimotor cortex>, they will be presented first. Then, the data obtained from epileptic cortex will be compared with the data from the nonepileptic cortex.

Nonepileptic cortex
A previous study 22 from this laboratory described the features of nonepileptic monkey sensorimotor cortex in electron microscopic preparanons. Descnptions of layer V pyramidal neurons and adjacenl neuropil regions were provided. Briefly, layer V pyramidal neurons have multiangular-shaped somata formed by an array of basal dendrites, a single apical dendrite and a single axon usually located between the basal dendrites. Much of the soma is occupied by a large, rounded nucleus wh~se nucleoplasm contains a relatively homogeneous sprinkling of electronopaque chromatin and a centrally located nucleolus ]~.22. Pyramidal neurons contain numerous organelles in the perikaryal cytoplasm including both free ribosomes and cisternae of granular endoptasmic reuculum. Terminals thai form axosomatic s}napses with these neurons make only ,~ymmetnc synapses. Pyramidal neurons in layers II and III are somewhat smaller than those in layer V. bul they share the same ultrastructural characteristics, Since our prewous ultrastructural studv 22 provided details on the somata and dendrites of pyramidal neurons and the axon terminals thai formed synapses with these structures, the present study will involve primarily an analysis of the synaptic contacts \`` itll aXOil initial Scglnenls of pyranlidal netlrons.
The axon initial segmeill {irises from the axon hillock region of the neuronal cell hod\'. Peters ct ill. > provide a COlllplctc description of the feattll-es of ~lxon initial segments. Such features will bl-icfh' bc describcd lind attention will bc given to those slttlCttlres and rehiiionships th{it appe:,lr differcnt in tilL' epileptic cortex.
Axon initial segments usually appear al tile base of file p_vranlidai cell bed) (Pig. 1). The botlnd{lr\ be-tweeI1 those two portions of tile lleuroll is :.lpp{ii+ellt lit 1o+`` magnification because tile cisternao of the ~rantlhlr endoplasmic rcticulum or Nissl bodies that reside within the soma do i1Ot enter the ~iX,Oll. The {iXOll hillock region is coniciil. \vhcl+eas the axon initial segillel]l has ii const{inl cvlmdrical shape that is main- In addition, lhe aXOll hillock lcgion of these same nemOllS displa>s i/shnilar loss of ternlinals that form s\nlnlotric S VilapSes. lhe smaller p> iamidal neurons in layers 11 and Ill of ten have no terminals that form synapses wth their axon initial segments. However, the larger pyramidal cells in layer V including two examined Betz cells have the highest number of initial segment terminals and this finding probably results from the much larger size of their axon initial scgmellts.
The second difference observed in these epileptic preparatives is the increased number of astrocvtic processes that lie adjacent and orient parallel to the axon mitial segments. Most of these astrocytic processes are packed with nunlerous filaments (Fig. 7). Such processes are found adjacent to the somata of these same pyramidal cells as described previously -~-'. Often, the same astrocytic process will appose a portion of the cell bodx, the axon hillock and a l(I l;m length of the axon inilial segment (Figs. 5 and ~3). Although this orientation preference for the hmgitudinal axis of the axon initial segment is observed most frequentl,v (Fig. 7), somc glial processes arc oriented transverse to this axis (double arrow in Fig. 4). This apposition of glial processes is usually only one process thick. However, inultiple layers of glial processes are found adjacent to the larger axons of Betz cells.

DIS(t SSI()N
The major finding of this electron microscopic study is that most of the axon terminals which form symmetric synapses with axon initial segments of pyramidal netirons are lost at epileptic foci in monkeys. Although previous studies have demonstrated a loss of axosomatic synapses in epileptic loci 3-7,2-~, 35. this report is the first to document a loss of these initial segment symmetric synapses that are formed by terminals mainly derived from the chandelier cell (Fig. 8).
Previous studies that utilized a combined Golgielectron microscopic method have shown that axon terminals forming synapses with laver II-I11 pyranlidal neuron axon initial seglnents arise from a specific cortical neuronal type, the chandclicr cell <>.>,>.
Although a similar study' of the chandelier cell has not been made in the monkey scnsorimotor cortex, it is likely that this cell exists in this region because Somogyi et al. x<> have dcmonstratcd chandelier cells in the motor cortex of cat alld in the monkevs visual cortex. In addition, the distribution of axon tcrrninals alongside the axon initial segments of pyramidal cells in laycrs 11 and III of the monkey sensorimotor cortex es is very simihtr to the distribution of chandelier cell axons. Taken together, these dala mdicate that most of the axon terminals that synapse with axon initial segments of pyramidal neurons arise from chandelier cells.
Recent resuhs from immunocvtochen~ical studies which localize lhe GABA synthesizing enzyme, GAD, have indicated that chandelier cells tire GABAergic because most terminals that form symmetric synapses with axon initial segments contain GAD imn3unoreactivitv s.il,>. In fact. Freund et al. s have provided direct cvideilce for this notion by combining Golgi impregnation with immuriocvtochemistry m the same preparation. In one fortuitous case from cat visual cortex, they demonstrated that 11 Golgi-imprcgnated bout(ms derived from a single chandelier cell contained OAD-posilive reaction product. The highly,' strategic site of these GABAergic synapses on axon initial segments suggests that they exert a strong inhibitory effect on the pyramidal neurons which they contact. The large loss of these GABAergic synapses in epileptic loci indicates a significant reduction of inhibition because (iABA has an inhibitory action tm cortical neuronsi4 i-. Thereforc, a probable result of this loss is a hypercxcitability of neurons at the epileptic focus. Thcse findings are consistent with our previous results that dcmonstrated a severe loss of GAI)-positive axon terminals tit sites of alumintlnl gel-educed epilepsy3L In addition, lhey coniplemeilt ~i previous electron microscopic studv of ¢ixosonlalic and axe-  loss of axosonlatic symmetric synapses and a 5IV<; loss of axodendritic synapses in cortical layer V of epileptic loci. Similar reductions in these two types of synapses were also obscrved in the superficial cortical layers where the present study was undertaken. It is interesting to note that the loss of symmetric synapses with axon initial seglncnts was lnorc similar to lhc larger reduction of axosomatic synapses than the smaller loss of axodendritic symlnetric synapses.
Thercf(tre, tile most severe loss of GABAergic. symmetric synapses at epileptic foci suggests a degeneration of two cortical (]ALIA cell types, basket and chandelier cells. The terminals of basket and chandelier cells have SOIllC similarities. They send their axons to the soma and axon initial segment, respectively, of pyramidal neurons, alld these two sites are most strategic for the control of action potential generatiOl-l. Thus, both cell types can dramatically influence the activity and output of the cortical proiection neurons. Second, these two types of terminals average more than onc mitochondria per terminal. This fact has hecn doennlented in nlanv previous studies for chandelier cell terminals (sec also Fig. 3) and in a quantitative study for the basket cell terminals -~-~. Thus, the chandelier cell axonal plexus that forms initial segment Svlnnletric synapses lnay provide a tonically active inhibition of cortical pnljcction neurons in a way proposed origimlllx for the pcriccllular basket cell axonal plcxus :-~. Furthcr support for this similar function of thesc txvo cell types arises fronl their similar degree of malfunction in epileptic foci. Although a thor(/ugh quantitative assessment was not made in the present stud\, one can conchlde from the electron microscopic data of initial segment synapses in epileptic foci that they are reduced m lnagnitudc to a similar amount as the axOSOlllatic synapses as prc~ iously reportede-L 257 It is likely that the chandelier cells ha',e degeneratcd in these epileptic foci because most of their axon terminals are not present. Other cvMcnce to support this notion is derived from a report that dcmonstr:.tted neuronal degeneration in Fink -Hcimer preparalions of alumina gel-treated epileptic i11Ollkevs'.   Another clcctron micrograph froin epileptic cortex that shows a layer II1 pyramidal soma, proximal apical dendritc (AP) and axon initial segment (arrows). The soma contains a large nucleus (N) and a perikaryal cytoplasm that contains the typical organclles as well as :m abundance ol lipofuscm granules. The initial segment is apposed to a gliaI profile that also contacts the axon hillock and soma. × 13.000  Fig. 7. En]argcment of a distal portion of another axon initial segment (IS) from a layer It[ pyramidal neuron from an epileptic preparation. 1his axon displays fasciculation of microtubules (M) and a dense coating beneath the axolemma (arrov,). Profiles of reactive astroQtes (A) that contain filaments lie adjacent to this axon initial segment. Axon terminals that form symmetric synapses are not present. × 2g,()O0.