Disturbance of Hippocampal Long-TermPotentiation After Transient Ischemia inGFAP Deficient Mice

Hidenobu Tanaka,1 Akira Katoh,3 Keiji Oguro,1 Kuniko Shimazaki,2 Hiroshi Gomi,3

Shigeyoshi Itohara,3 Toshio Masuzawa,1 and Nobufumi Kawai2*1Department of Surgical Neurology, Jichi Medical School, Tochigi, Japan2Department of Physiology, Jichi Medical School, Tochigi, Japan3Behavioral Genetics Laboratory, RIKEN, Saitama, Japan

GFAP (glial fibrillary acidic protein) is an intermediatefilament protein found exclusively in the astrocytes of thecentral nervous system. We studied the role of GFAP inthe neuronal degeneration in the hippocampus after tran-sient ischemia using knockout mice. Wild-type C57Black/6 (GFAP�/�) mice and mutant (GFAP�/�) micewere subjected to occlusion of both carotid arteries for5–15 min. Hippocampal slices were prepared 3 daysafter reperfusion and the field excitatory postsynapticpotentials (fEPSP) in the CA1 were recorded. High fre-quency stimulation induced robust long-term potentia-tion (LTP) in GFAP�/�, as in GFAP�/� mice. After isch-emia, however, the LTP in GFAP�/� was significantlydepressed. Similarly, paired pulse facilitation (PPF) dis-played little difference between GFAP�/� and GFAP�/�,but after ischemia, the PPF in GFAP�/� showed a de-pression. Histological study revealed that loss of CA1and CA3 pyramidal neurons after ischemia was markedin GFAP�/�. MAP2 (dendritic) immunostaining in thepost-ischemic hippocampus showed little difference butNF200 (axonal) immunoreactivity was reduced inGFAP�/�. S100� (glial) immunoreactivity was similar inthe post-ischemic hippocampus of the GFAP�/� andGFAP�/�, indicating that reactive astrocytosis did notrequire GFAP. Our results suggest that GFAP has animportant role in astrocyte-neural interactions and thatischemic insult impairs LTP and accelerates neuronaldeath. © 2002 Wiley-Liss, Inc.

Key words: GFAP; knock out mouse; hippocampus;ischemia; long-term potentiation; S100�

Glial fibrillary acidic protein (GFAP) is the majorcytoskeletal protein of astrocytes in the brain (Dahl et al.,1986; Eng et al., 1995). Numerous studies have shown anincrease in GFAP immunostaining in the brain in varioustypes of brain injury including trauma, demyelination andbrain ischemia (Petito et al., 1990, 1993; Tanaka et al.,1991, 1992; Eng et al., 1994; Li et al., 1995). Recentevidence has shown that astrocytes play an active role in

partnership with neurons in protecting the CNS againstvarious kinds of insults, including cerebral ischemia andhypoxia (Vernadakis, 1996; Louw et al., 1998), or neuro-logical disorders (Tacconi, 1998). Immunohistochemicalstudies have demonstrated that after global forebrain isch-emia, the selective neuronal death that occurs in the CA1pyramidal cell layer of the hippocampus is accompanied bya reactive astrocytosis (Petito et al., 1990, 1993; Tanaka etal., 1992). These reactive changes are characterized byastrocytic proliferation and extensive hypertrophy of thecell body and intracellular processes including GFAP. Theactivated astrocytes have been reported to protect neuronssubjected to an ischemic insult in many laboratories(Tanaka et al., 1993; Louw et al., 1998; Gottlieb et al.,1999). Actually, astrocyte processes are intimately associ-ated with the synaptic cleft, where they may regulatesynaptic function through the uptake of neurotransmitters,buffering cations and pH, and present a barrier against thediffusion of calcium (Wenzel et al., 1991). It has also beenreported that the regulation of GFAP synthesis is altered indisorders of the central nervous system. For example,TGF-1, Interleukin-1 and TNF-�, which are released aspart of neuronal response to injury, are all known tomodulate the GFAP mRNA levels (Oh et al., 1993; Reillyet al., 1998; Krohn et al., 1999). The early GFAP mRNAincreases after middle cerebral artery (MCA) occlusioncould therefore be interpreted as a manifestation of astro-cytic participation in brain plasticity in this region (Ya-mashita et al., 1996).

Contract grant sponsor: Japanese Ministry of Education, Science and Cul-ture; Contract grant number: 11671383; Contract grant sponsor: SpecialCoordination Funds for promoting Science and Technology of the STA ofthe Japanese Government.

*Correspondence to: Dr. Nobufumi Kawai, Department of Physiology,Jichi Medical School, Minamikawachi-machi, Tochigi-ken, 329-0498 Ja-pan. E-mail: kawainob@jichi.ac.jp

Received 18 December 2000; Revised 30 June 2001; Accepted 9 July 2001

Published online 19 November 2001

Journal of Neuroscience Research 67:11–20 (2002)

© 2002 Wiley-Liss, Inc.DOI 10.1002/jnr.10004

Recently, several groups have developed mice lack-ing GFAP to determine the function of GFAP (Gomi etal., 1995; Pekny et al., 1995; Liedtke et al., 1996; Galou etal., 1997; Wang et al., 1997). Gomi et al. (1995) reportedthat GFAP�/� mice exhibited normal development andno obvious anatomical abnormalities in the CNS. TheGFAP-deficient mice did not show any apparent clinicalproblems despite the absence of intermediate protein intheir astrocytes. On the other hand, Shibuki et al. (1996)demonstrated that long-term depression (LTD) at parallelfibers to Purkinje cell synapses in the GFAP-mutant micewas clearly deficient. They reported that GFAP is requiredfor communication between Bergmann glia and Purkinjecells for both LTD induction and maintenance. It was alsoshown that in the aging mice, absence of GFAP led toabnormal white matter architecture and impaired myeli-nation (Liedtke et al., 1996). Studies on synaptic transmis-sion in these GFAP-deficient mice indicate that althoughbasal nerve transmission is not affected, LTP is significantlyenhanced (McCall et al., 1996).

These results suggest that study of GFAP-deficient(GFAP�/�) mice can provide further insight regardingastrocytic functions in the cerebral insults. The presentstudy attempts to understand the role of GFAP as well asastrocytes after brain ischemia in GFAP�/� mice. A pre-liminary account has appeared elsewhere as an abstract(Tanaka et al., 1999).

MATERIALS AND METHODS

Production of GFAP Mutant Mice

Male C57 Black/6 mice (SLC Laboratories, Hamamatsu,Japan), GFAP�/� mice and GFAP�/� mice lacking GFAP geneexpression, aged 6–10 weeks and weighing 21–27 g were used.The animals were given free access to food and water ad libbefore surgery. The method used for targeted disruption of theGfap gene has been reported in detail previously (Gomi et al.,1995). The male C57 Black/6 GFAP�/� and GFAP�/� miceused were littermates intercrossed between male and femaleheterozygotes that had been backcrossed to C57BL mice for 10generations. The genotypes of the mice were determined bySouthern blot analysis of DNA prepared from the tails. TheGFAP�/� mice did not have any detectable GFAP expression asconfirmed by western blotting or immunohistochemistry. Theprocedures used on the laboratory animals were approved by theInstitutional Animal Care and Use Committee of Jichi MedicalSchool.

Brain Ischemia

General anesthesia was induced with 4% sevoflurane in70% N2O and 30% O2 and maintained with 1–2% sevofluranein 70% N2O and 30% O2 by means of an open facemask. Undera surgical microscope, both common carotid arteries (CCAs)were exposed through a midline incision in the neck andoccluded for 5, 10 or 15 min with small aneurysm clips, and thedistal portions of the arteries were inspected to confirm theabsence of blood flow. The body temperature was maintained at37.5°C with a heating lamp at the time of occlusion; the mice

were housed in a room at 22°C. On Day 3 after reperfusion, themouse brain was removed under ether anesthesia and used forelectrophysiological and histological studies.

Experimental Groups

The mice were subjected to 15 min of cerebral ischemia[C57 Black/6 GFAP�/�(12 mice) and GFAP�/� (7 mice)],10 min of cerebral ischemia [GFAP�/�(6 mice), GFAP�/�

(6 mice)], or 5 min of cerebral ischemia [GFAP�/� (5 mice),GFAP�/� (5 mice)]. Sham-operated controls underwent thesame surgical procedures as ischemic mice, except for the carotidocclusion [C57 Black/6 GFAP�/� (6 mice) and GFAP�/�

(6 mice)].The mice that showed motor deficit or abnormal spon-

taneous movements (four mice among 43 mice, 7%) werediscarded. We could not observe any difference of the move-ments in the cage between the experimental groups.

Electrophysiology

In electrophysiological experiments using hippocampalslices, no obvious change was seen at short period (within 1 day)after ischemia. Therefore, we made slice experiments on 3 daysafter ischemia, when overt neuronal death appeared (Kirino etal., 1982, 1992). Transverse sections (300–400 �m) of the hip-pocampus were prepared by a previously described method(Kirino et al., 1992; Tsubokawa et al., 1995, 1996; Kawai et al.,1998). Slices were maintained at room temperature in a sub-merged holding chamber for at least 1 hr before transfer to asubmersion-type recording chamber, continuously perfusedwith artificial cerebrospinal fluid (ACSF) at 3 ml/min, main-tained at the temperature 34 � 0.5°C.

The ACSF was composed of (mM) 127 NaCl, 1.5 KCl, 1.24KH2PO4, 2.4 CaCl2, 1.3 MgSO4, 26 NaHCO3 and 10 glucose,pH 7.4, and was bubbled with 95% O2/5% CO2. Field excita-tory postsynaptic potentials (fEPSPs) were recorded with glassrecording microelectrodes filled with 5� Krebs solution (ACSFsolution without Ca2�, Mg2� and glucose) and placed in thestratum radiatum of the CA1 subfield by stimulating Schaffercollateral/commissural fibers. To measure the maximal slope offEPSPs records were digitized at 10 kHz and analyzed by on-linecomputer. Evoked fEPSPs were digitally low-pass filtered andthe maximal rate of change in potential within a time windowselected around the rising phase was calculated. Control re-sponses were evoked every 10 sec and recorded for 10–30 min,until stable baseline responses were obtained. High frequencystimulation using 100 Hz trains for 1 second was applied toinduce long-term potentiation (LTP). Comparison of the degreeof potentiation was made between the average of 10 responsesimmediately preceding the tetanic stimulus and 10 responses at60 min after the tetanic stimulation. Statistical comparisonsbetween groups were made using Student’s paired t-test. Sig-nificant difference was assumed at P � 0.05.

Immunohistochemistry

Mice were perfused with 0.9% NaCl under pentobarbitalanesthesia followed by perfusion with 4 % paraformaldehydedissolved in 0.1 M phosphate buffer (PB). The brains wereremoved and preserved overnight in a fixative, then transferred

12 Tanaka et al.

to 0.1 M PB containing 15 % sucrose. Sagittal sections of 50 �mthickness were made by a freezing microtome. The avidin-biotin-complex (ABC) method was used for the immunohisto-chemical study. Anti-microtubule associated protein2 (MAP2)and anti-neurofilament 200 (NF200) antibodies were purchasedfrom Chemicon International Inc. (Temecula, CA). Anti-S100� antibody was purchased from Sigma Chemical Company(St. Louis, MO). After incubation for 1 hr in 1.5% normal horseserum in phosphate-buffered saline containing Triton X-100 at0.1% (PBS-T), the sections were incubated with anti-MAP2antibody (1/1,000 dilution), anti-NF200 antibody (1/200 dilu-tion) or anti-S100� antibody (1/100 dilution) overnight at 4°C.Sections were washed in PBS and further incubated for 1 hrwith the biotin-conjugated anti-rabbit IgG (Vector Laborato-ries, Burlingame, CA) followed by washing in PBS and furtherincubated for 1 hr with an avidin-biotin complex (ABC EliteKit, Vector Laboratories). Then, after finally washing with PBS,the sections were developed in 3,3-diaminobenzidine for 5–10min until the staining became apparent. Staining was terminatedby washing two times for 5 min each in PB.

RESULTS

Long-Term PotentiationIn the hippocampal slice preparations of GFAP�/�

mice, field EPSPs (fEPSPs) in the CA1 subfield evoked bystimulation to Schaffer collateral/commissural input ap-peared to be similar to those in control, sham-operatedGFAP�/� mice. When high-frequency stimulation (100Hz for 1 sec) was applied, fEPSPs were greatly potentiated,exhibiting typical long-term potentiation (LTP) in theslices from both GFAP�/� and GFAP�/� mice (Fig. 1).The fEPSP slope in the GFAP�/� mice at 60 min aftertetanic stimulation was 201 � 23.8 % (13 slices, six mice)and that in the GFAP�/� mice was 199 � 17.2 % (nineslices, six mice). Therefore, no significant difference wasobserved between the hippocampal slices from GFAP�/�

and GFAP�/� mice in regard to the magnitude andduration of LTP (Fig. 2).

Next, we examined LTP induction in the hippo-campal slices of both the GFAP�/� and GFAP�/� miceafter ischemia. In the slices from GFAP�/� mice 3 daysafter ischemia for 15 min, tetanic stimulation induceddistinct LTP (Fig. 1A). The extent and duration of thepotentiation, however, were less than those in the controlmice. The fEPSP slope in the GFAP�/� mice at 60 minafter tetanic simulation was 151.7 � 9.2 % (nine slices, sixmice). LTP was severely impaired after ischemia in theslices from GFAP�/� mice and the fEPSP slope in theGFAP�/� mice at 60 min after tetanic simulation was109.5 � 13.5 % (10 slices, six mice), being significantlydifferent from that in the GFAP�/� mice (P � 0.05). Wealso compared the LTP in the hippocampal slices of thewild-type and mutant mice after ischemia for 10 min. Inthe GFAP�/� mice, no obvious difference was observedin the profile of LTP between the sham-operated miceand mice subjected to brain ischemia. In the GFAP�/�

mice, however, 10 min ischemia caused considerable sup-

pression of LTP. LTP was between that in the sham-operated mice and mice after ischemia for 15 min (Fig.1B). The fEPSP slopes at 60 min after tetanus in theGFAP�/� mice subjected to 10 min ischemia was165.7 � 20.7 % (13 slices, six mice). Figure 2 summarizesthe data of LTP in the hippocampal slices under thedifferent conditions. The average slopes of the fEPSPs at60 min after tetanic stimulation were compared among thegroups.

Input–Output RelationTo study the efficacy of synaptic transmission in the

GFAP�/� and GFAP�/� mice, we compared the input–output (I/O) relationships in the CA1 area. Low fre-quency stimulation (0.1 Hz) was applied with the intensity

Fig. 1. Changes in LTP in the hippocampal slices of GFAP�/� andGFAP�/� mice after ischemia. Time course of the maximal slope offield EPSPs (fEPSP) in the CA1 subfield are plotted. High frequencystimulation (100 Hz, 1 sec) was applied at 10 min after control lowfrequency (0.1 Hz) stimulation. A: Control GFAP�/� mice (E) andthose after ischemia for 10 min (�) and 15 min (‚). B: ControlGFAP�/� mice (F) and those after ischemia for 10 min (�) and 15 min(Œ). Each symbol indicates the average of 10 successive responses (sixmice, in each group). Error bars mean � SD. Inset records aresuperimposed traces of fEPSPs before (A) and after (B) high frequencystimulation.

Hippocampus in GFAP Knockout Mice After Ischemia 13

varying from 0.25 V to 2.0 V whereas the duration wasconstant (0.3 msec). The maximal slopes of the EPSPswere plotted against the stimulus intensity (Fig. 3). TheI/O curves in the sham-operated GFAP�/� andGFAP�/� mice showed no obvious difference. Threedays after ischemia for 15 min, however, the I/O curve inthe GFAP�/� mice showed a slight suppression; evenmore marked suppression was observed in the GFAP�/�

mice. The fEPSP slope in the GFAP�/� mice stayed at alow level and did not increase even for stronger stimulusintensities. These data suggest that activation of the inputfibers was saturated at low stimulus intensities in theGFAP�/� mice after ischemia.

Paired Pulse FacilitationTo study further the possible source of impairment of

LTP in the GFAP�/� mice, we examined paired-pulse fa-cilitation (PPF) of fEPSPs. Paired pulse stimulation with theinterval of 20–200 msec-induced facilitation of the fEPSPs ofCA1 neurons, which could be caused mainly by a presynapticmechanism. As shown in Figure 4, there was no significantdifference in the PPF profiles of slices from controlGFAP�/� mice, control GFAP�/� mice and GFAP�/�

mice after ischemia for 15 min. For an interpulse interval of40 msec, the ratio of the slope of the second to first fEPSPwas 1.56 � 0.07 (eight slices, five mice), 1.612 � 0.1 (sixslices, five mice) and 1.32 � 0.1(eight slices, six mice) incontrol GFAP�/�, control GFAP�/� and GFAP�/� miceafter ischemia, respectively. In contrast, the slices from theGFAP�/� mice after ischemia showed significant depression

in the PPF profile; the ratio of the slope of the second to firstEPSP for the interpulse interval of 40 msec was 1.09 � 0.1(six slices, five mice).

Neuronal Death After IschemiaWe performed histological studies on the hippocam-

pus of GFAP�/� and GFAP�/� mice after ischemia.Typical examples of hippocampal slices stained by cresylviolet at 3 days after ischemia for 10 min and 15 min arecompared (Fig. 5). In the GFAP�/� mice, ischemia for10 min did not cause any noticeable change (Fig. 5A),

Fig. 2. Comparison of LTP in the hippocampal slices of GFAP�/� andGFAP�/� mice after various duration of ischemia. Percentage increaseof the maximal slope of fEPSP at 60 min after tetanus was compared inGFAP�/� and GFAP�/� mice. Each group, n 6. Error bars mean � SD. Significant difference was observed between GFAP�/�

and GFAP�/� mice after ischemia for 15 min. Student’s t-test (P �0.05).

Fig. 3. Input-output relationships in GFAP�/� and GFAP�/� miceafter ischemia. The maximal slopes of fEPSPs are plotted against thestimulus intensity of the input fibers. A: Control GFAP�/� mice (E)and those after ischemia for 15 min (‚). B: Control GFAP�/� mice(F) and those after ischemia for 15 min (Œ). Each symbol indicates theaverage of the data from six slices, five mice.

14 Tanaka et al.

whereas ischemia for 15 min caused a small reduction inthe number of pyramidal neurons (Fig. 5C). In theGFAP�/� mice, considerable loss of the pyramidal neu-rons was observed after ischemia for 10 min (Fig. 5B), andmore severe damage was evident in the mutant mice after

ischemia for 15 min (Fig. 5D). It is of note that neuronaldeath was seen not only in the CA1 but also in CA3pyramidal neurons.

We made a quantitative study of neuronal death inthe hippocampus of GFAP�/� mice and GFAP�/� mice

Fig. 4. Paired pulse facilitation in GFAP�/� and GFAP�/� mice after ischemia. Ratios of the fEPSPslopes evoked by second stimulus to that by the first stimulus are plotted against interpulse intervals.A: Control GFAP�/� mice (E) and those after ischemia for 15 min (‚). B: Control GFAP�/� mice(F) and those after ischemia for 15 min (Œ). Each symbol indicates the average of the data from sixslices, five mice.

Hippocampus in GFAP Knockout Mice After Ischemia 15

after ischemia. In the sagittal sections (50 �m thickness),cell death was examined every 500 �m along long axis ofthe hippocampus and postischemic cell death was deter-mined by cresyl violet staining. Figure 6 summarized thedata of neuronal death in different conditions. Scoring asthe following criteria assessed neuronal damage (Pulsineliet al., 1982): 0, no neuronal damage; 1, minor neuronaldamage (less than 30% of the neurons are dead); 2, mod-erate damage (30% to 70% of neurons are dead); and 3,severe damage (more than 70% neurons are dead). It is ofnote that average score for GFAP�/� mice (broken line)is always larger than the average score for GFAP�/� mice(bar) irrespective of time of ischemia. Significant differ-ence of the average score between GFAP�/� andGFAP�/� mice was observed in CA1 and CA3 after15 min ischemia and in all the area after 20 min ischemia.

S100� ImmunostainingS100� protein, which belongs to a family of Ca2�-

binding proteins involved in cytoskeletal reorganization, isknown to be co-localized with GFAP in astrocytes. Im-munostaining using S100�antibody was compared in thehippocampus of wild and GFAP�/� mice after ischemiafor 15 min (Fig. 7). In the GFAP�/� mice, distinctimmunoreactivity was observed over a broad area of thehippocampus (Fig. 7A,C,E,G). Interestingly, increase inS100� immunostaining was also clearly observed in theGFAP�/� mice (Fig. 7B,D,F,H). These data indicate thatastrocytosis takes place in the brain of GFAP-deficientmice subjected to ischemia similar to that observed in wildtype mice. Although the number of S100� positive astro-

cytes showed no great difference between GFAP�/� andGFAP�/� mice after ischemia, more marked staining inarborization of S100�-containing cells was observed inGFAP�/� mice (Fig. 7F,H).

MAP 2 and NF 200We further studied the post-ischemic changes in the

expression of MAP2 and NF 200 in the GFAP�/� andGFAP�/� mice. We used MAP2 as a marker of dendritesand NF 200 as a marker of axons. MAP2 immunostainingin the CA1 area at 3 days after 15 min ischemia in theGFAP�/� and GFAP�/� mice are compared (Fig. 8).Although the immunoreactivity was much more reducedin the GFAP�/� mice than GFAP�/� mice due to theloss of pyramidal neurons, heavy staining of the dendritesof surviving neurons are obviously seen even in GFAP�/�

mice (Fig. 8D,H) . NF 200 immunostaining, whichmainly marks Schaffer collateral axons in the CA1 subfield,was substantially reduced in the GFAP�/� mice (Fig.9D,H). This could be explained by the degeneration ofCA3 pyramidal neurons in the GFAP�/� mice.

DISCUSSIONThe present study has demonstrated that LTP was

more strongly suppressed in the CA1 pyramidal neurons ofthe GFAP�/� mice after ischemia than that in GFAP�/�

mice. There are a number of possible explanations for thestronger suppression of LTP after ischemia in theGFAP�/� mice. First, it could be due to severe degener-ation or functional impairment of the pyramidal neuronsafter ischemia. We have previously reported LTP induc-tion of CA1 pyramidal neurons after ischemia in the gerbilmodel (Kirino et al., 1992; Tsubokawa et al., 1995, 1996).The induction of LTP could be produced by survivedneurons although considerable neurons were dead. In thepresent study, neuronal death was more marked inGFAP�/� mice than GFAP�/� mice after ischemia asshown in Figure 6. Decrease in magnitude and duration ofLTP (Fig. 1) was closely correlated with the scores ofneuronal death. Increased susceptibility to ischemia inGFAP-deficient mice was also reported by Nawashiro etal. (2000) after middle cerebral artery occlusion. The sec-ond possibility is reduction in presynaptic input fibers orig-inating from the CA3 pyramidal neurons. In GFAP�/�

mice, not only CA1 but also CA3 pyramidal neuronsshowed degeneration after ischemia (Fig. 6). Ischemia for15 min caused damage of in CA3 neurons in 86% (six outof seven animals) of the mutant mice. Therefore, Schaffercollateral/commissural fibers originating from CA3 pyra-midal neurons might be degenerated, resulting in thereduction of presynaptic impulses from CA1 neuronscould be reduced. Depression of paired pulse facilitationand input-output relations in the post-ischemic mutantmice may reflect the degeneration of CA3 pyramidalneurons. As the third possibility, it is considered thatfunctions of astrocytes, one of which is to promote syn-aptic transmission are not adequate to maintain LTP inGFAP�/� mice when they were subjected to brain isch-

Fig. 5. Neuronal death in the hippocampal slices of GFAP�/� andGFAP�/� mice after ischemia. Cresyl violet staining 3 days afterischemia. A: GFAP�/� mice, ischemia for 10 min. B: GFAP�/� mice,ischemia for 10 min. C: GFAP�/� mice, ischemia for 15 min.D: GFAP�/� mice, ischemia for 15 min.

16 Tanaka et al.

emia. There are many reports suggesting the role of astro-cytes or glial cells in the maintenance of synaptic plasticity,through glia–neuron intercommunication and alsothrough glia-to-glia signaling including extracellular ma-trix and perineuronal nets (Wenzel et al., 1991; Mennericket al., 1994; Hansson et al., 1995; Vernadakis, 1996;Bergles et al., 1997; Pfrieger et al., 1997; Araque et al.,1998). Wenzel et al. (1991) reported the influence of LTPon the spatial relationship between astrocytic processes andpotentiated synapses in the rat brain. They found that theastrocytes near the potentiated synapses exhibited a higherdegree of ramification and enlargement of surface area.They also observed that astrocytic processes established acloser spatial contact to the potentiated axo-spinodendriticsynapses and tended to cover the surface of synaptic bou-tons or spines more extensively as well as limit the synapticcleft laterally. As a fourth possibility, enhanced toxicity ofglutamate by dysfunction of the glutamate transportersmight be occurred in GFAP�/� mice. It has been re-ported that loss of glial glutamate transporters GLAST orGLT-1 induced elevation of extracellular glutamate levelsand neurodegeneration (Rothstein et al., 1995, 1996).Although GLAST and GLT-1 are expressed by morpho-logically distinct GFAP-positive protein in the hippocam-pal culture (Perego et al., 2000) and our data show that thenumber of the S100� positive astrocytes was increasedboth in GFAP�/� and GFAP�/� mice after ischemia, thepossibility remains that functional deficiency in glial glu-

tamate transporters takes place more profoundly inGFAP�/� mice after ischemia.

In the present experiments, the magnitude and du-ration of LTP in sham-operated mice was not significantlydifferent between GFAP�/� and GFAP�/� mice. Shibukiet al. (1996) also reported that the difference in the ampli-tude of LTP between GFAP�/� mice and GFAP�/�

mice was not significant. McCall et al. (1996) reported thatGFAP-null mice showed enhanced LTP as compared tocontrol mice. Although they showed a significant differ-ence in the amplitude of population spikes between mu-tant and control mice, the difference of the EPSP slope didnot reach statistical significance. In our experiments, weconsistently measured the slope of fEPSPs and used stim-ulus intensities below those evoking population spikes,because of the amplitude of population spikes sometimescomplicates the analysis.

Histological examination by cresyl violet stainingshowed that in the mouse ischemia model, neuronal losswas seen not only in CA1 but also in CA3 pyramidalneurons in both GFAP�/� mice and the wild-type mice.The immunohistochemical study showed apparent de-crease in NF200 staining in the CA1 area in GFAP�/�

mice after ischemia for 15 min, indicating loss of axonalbranches originating from CA3 neurons (Fig. 9). It isnoteworthy that astrocytosis, as determined by S100�labeling, was observed after ischemia in both GFAP�/�

mice and the wild-type mice. The results indicate that

Fig. 6. Quantification of ischemia-induced neuronal death. Neuronal damage score for GFAP�/�

mice (open dot) and GFAP�/� mice (closed dot) in CA1, CA3 and dentate gyrus (DG) at various timefor ischemia (5, 10, 15, and 20 min). One dot represents an animal. The average of damage score isindicated by bar (GFAP�/� mice) and by broken line (GFAP�/� mice). Significance: *P � 0.05,**P � 0.01.

Hippocampus in GFAP Knockout Mice After Ischemia 17

reactive astrocytosis after ischemia could take place in theabsence of GFAP. It has been reported that GFAP-deficient mice showed a normal abundance of astrocytes inthe hippocampus but that these are completely lackingintermediate filaments (Pekny et al., 1995). These dataimply that although both GFAP and S100� are specificmarkers of astrocytes, S100� can be expressed indepen-dently of GFAP.

There are several reports indicating the role ofGFAP in synaptic plasticity and neurodegeneration(Fahrig, 1994; Valentim et al., 1999). Fahrig (1994)examined the changes in the solubility of GFAP afterischemic brain damage and suggested that GFAP did not

only passively reflect the reactive state of astrocytes butwas actively involved in the functional transformationof these cells. More recently, Valentim et al. (1999)reported the effects of transient ischemia on GFAPphosphorylation and immunoreactivity in the rat hip-pocampus. They concluded that changes in the phos-phorylation of GFAP might be essential for the plasticresponse of astrocytes to neuronal damage, as neuronsand astrocytes act as functional units involved in ho-meostasis, plasticity and neurodegeneration. As anotheraspect, it is of note that ischemia-induced glutamatecauses astrocytic swelling by Na� coupled cotransport-ers because glutamate is transported by such tranporters

Fig. 7. S100� immunostaining in the hippocampal slices of GFAP�/�

(A,C,E,G) and GFAP�/� (B,D,F,H) mice. A,B: Control; (C–H) 3days after ischemia for 15 min. E–H: Enlarged photographs of (A–D),respectively.

Fig. 8. MAP 2 immunostaining in the hippocampus of GFAP�/�

(A,C,E,G) and GFAP�/� (B,D,F,H) mice. A, B: Control; (C–H) 3days after ischemia for 15 min. E–H: Enlarged photographs of (A–D)respectively.

18 Tanaka et al.

in the astrocytes (Rothstein et al., 1994; Lehre et al.,1995). The astroglial swelling may affect plasticity of thesynaptic function after ischemia (Kimberg, 1999). Re-cent studies have suggested that some cytokines inducedby neuronal injury or imidazoline-I2-receptor mayhave a functional role in regulating the expression ofGFAP (Oh et al., 1993; Regunathan et al., 1993; Reillyet al., 1998; Krohn et al., 1999). These mechanismscontrolling GFAP expression may be worth evaluatingfurther in the future to clarify the functions of GFAP aswell as astrocytes in the face of cerebral insults.

ACKNOWLEDGMENTSThe authors thank Mrs. K. Koeda, Mr. T. Mizui for

their technical assistance. This study was supported by theJapanese Ministry of Education, Science and Culture(grant 11671383) and Special Coordination Funds forpromoting Science and Technology of the STA of theJapanese Government.

REFERENCESAraque A, Parpura V, Sanzgiri RP, Haydon PG. 1998. Glutamate-

dependent astrocyte modulation of synaptic transmission between cul-tured hippocampal neurons. Eur J Neurosci 10:2129–2142.

Bergles DE, Jahr CE. 1997. Synaptic activation of glutamate transporters inhippocampal astrocytes. Neuron 19:1297–1308.

Dahl D, Bjorklund H, Bignami A. 1986. Immunological markers in astro-cytes. In: Ferdorroff S, Vernadakis A, editors. Astrocytes, vol 3. Orlando:Academic Press. p 1–25.

Eng LF, Ghirnikar RS. 1994. GFAP and astrogliosis. Brain Pathol 4:229–237.

Eng LF, Lee Y. 1995. Intermediate filaments in astrocytes. In: KettenmannH, Ransom BR, editors. Neuroglia. Oxford: Oxford University Press. p650–667.

Fahrig T. 1994. Changes in the solubility of glial fibrillary acidic proteinafter ischemic brain damage in the mouse. J Neurochem 63:1796–1801.

Galou M, Gao J, Humbert J, Mericskay M, Li Z, Paulin D, Vicart P. 1997.The importance of intermediate filaments in the adaptation of tissues tomechanical stress: evidence from gene knockout studies. Biol Cell 89:85–97.

Gomi H, Yokoyama T, Fujimoto K, Ikeda T, Katoh A, Itoh T, Itohara S.1995. Mice devoid of the glial fibrillary acidic protein develop normallyand are susceptible to scrapie prions. Neuron 14:29–41.

Gottlieb M, Matute C. 1999. Expression of nerve growth factor in astro-cytes of the hippocampal CA1 area following transient forebrain ischemia.Neuroscience 91:1027–1034.

Hansson E, Ronnback L. 1995. Astrocytes in glutamate neurotransmission.FASEB J 9:343–350.

Kawai K, Nakagomi T, Kirino T, Tamura A, Kawai N. 1998. Precondi-tioning in vivo ischemia inhibits anoxic LTP and functionally protectsCA1 neurons in the gerbil. J Cereb Blood Flow Metab 18:288–296.

Kimberg HK. 1999. Cell swelling in cerebral ischemia. In: Waltz W, editor.Cerebral ischemia: molecular and cellular pathophysiology. New Jersey:Humana Press. p 45–67.

Kirino T. 1982. Delayed neuronal death in the gerbil hippocampus follow-ing ischemia. Brain Res 239:57–69.

Kirino T, Robinson HPC, Miwa A, Tamura A, Kawai N. 1992. Distur-bance of membrane function preceding ischemic delayed neuronal deathin the gerbil hippocampus. J Cereb Blood Flow Metab 12:408–417.

Krohn K, Rozovsky I, Wals P, Teter B, Anderson CP, Finch CE. 1999.Glial fibrillary acidic protein transcription responses to transforminggrowth factor-beta1 and interleukin-1beta are mediated by a nuclearfactor-1-like site in the near-upstream promoter. J Neurochem 72:1353–1361.

Lehre KP, Levy LM, Ottersen OP, Storm-Mathisen J, Danbolt NC. 1995.Differential expression of two glial glutamate transporters in the rat brain:quantitative and immunocytochemical observations. J Neurosci 15:1835–1853.

Li Y, Chopp M, Zhang ZG, Zhang RL. 1995. Expression of glial fibrillaryacidic protein in areas of focal cerebral ischemia accompanies neuronalexpression of 72-kDa heat shock protein. J Neurosci 128:134–142.

Fig. 9. NF 200 immunostaining in the hippocampal slices of GFAP�/�

(A,C,E,G) and GFAP�/� (B,D,F,H) mice. A,B: Control; (C–H) 3days after ischemia for 15 min. E–H: Enlarged photographs of (A–D),respectively.

Hippocampus in GFAP Knockout Mice After Ischemia 19

Liedtke W, Edelmann W, Bieri PL, Chiu FC, Cowan NJ, Kucherlapati R,Raine CS. 1996. GFAP is necessary for the integrity of CNS white matterarchitecture and long-term maintenance of myelination. Neuron 17:607–615.

Louw DF, Masada T, Sutherland GR. 1998. Ischemic neuronal injury isameliorated by astrocyte activation. Can J Neurol Sci 25:102–107.

McCall MA, Gregg RG, Behringer RR, Brenner M, Delaney CL, Gal-breath EJ, Zhang CL, Pearce RA, Chiu SY, Messing A. 1996. Targeteddeletion in astrocyte intermediate filament (Gfap) alters neuronal physi-ology. Proc Natl Acad Sci USA 93:6361–6366.

Mennerick S, Zorumski CF. 1994. Glial contributions to excitatory neu-rotransmission in cultured hippocampal cells. Nature 368:59–62.

Nawashiro H, Brenner M, Fukui S, Shima K, Hallembeck M. 2000. Highsusceptibility to cerebral ischemia in GFAP-null mice. J Cereb BloodFlow Metab 20:1040–1044.

Oh YJ, Markelonics GJ, Oh TH. 1993. Effects of interleukin-1 beta andtumor necrosis factor-alpha on the expression of glial fibrillary acidicprotein and transferrin in cultured astrocytes. Glia 8:77–86.

Pekny M, Leveen P, Pekna M, Eliasson C, Berthold CH, Westermark B,Betsholtz C. 1995. Mice lacking glial fibrillary acidic protein displayastrocytes devoid of intermediate filaments but develop and reproducenormally. EMBO J 18:1590–1598.

Perego C, Vanoni C, Bossi M, Massari S, Basdev H, Longhi R, Pietrini G.2000. The GLT-1 and GLAST glutamate transporters are expressedmorphologically distinct astrocytes and regulated by neuronal activity inprimary hippocampal co-cultures. J Neurochem 75:1076–1084.

Petito CK, Morgello S, Felix JC, Lesser ML. 1990. The two patterns ofreactive astrocytosis in post-ischemic rat brain. J Cereb Blood Flow Metab10:850–859.

Petito CK, Halaby IA. 1993. Relationship between ischemia and ischemicneuronal necrosis to astrocyte expression of glial fibrillary acidic protein.Int J Dev Neurosci 11:239–247.

Pfrieger FW, Barres BA. 1997. Synaptic efficacy enhanced by glial cells invitro. Science 277:1684–1687.

Regunathan S, Feinstein DL, Reis DJ. 1993. Expression of non-adrenergicimidazoline sites in rat cerebral cortical astrocytes. J Neurosci Res 34:681–688.

Reilly JF, Maher PA, Kumari VG. 1998. Regulation of astrocytes GFAPexpression by TGF-beta1 and FGF-2. Glia 22:202–210.

Rothstein JD, Martin L, Levey AI, Dykes-Hoberg M, Jin L, Wu D, NashN, Kuncl RW. 1994. Localization of neuronal and glial glutamate trans-porters. Neuron 13:713–725.

Rothstein JD, Dykes-Hoberg M, Pardo CA, Bristol LA, Jin L, Kuncl RW,Kanai Y, Hediger MA, Wang Y, Schielke JP, Welty DF. 1996. Knockoutof glutamate transporters reveals a major role for astroglial transport inexcitotoxicity and clearance of glutamate. Neuron 16:675–686.

Rothstein JD, Van Kammen M, Levey AI, Martin LJ, Kuncl RW. 1995.Selective loss of glial glutamate transporter GLT-1 in amyotropic lateralsclerosis. Ann Neurol 38:73–84.

Shibuki K, Gomi H, Chen L, Bao S, Kim JJ, Wakatsuki H, Fujisaki T,Fujimoto K, Katoh A, Ikeda T, Chen C, Thompson RF, Itohara S. 1996.Deficient cerebellar long-term depression, impaired eyeblink condition-ing, and normal motor coordination in GFAP mutant mice. Neuron16:587–599.

Tacconi MT. 1998. Neuronal death: is there a role for astrocytes? Neuro-chem Res 23:759–765.

Tanaka H, Araki M, Masuzawa T. 1991. Differential response of threeastrocyte-specific proteins to fibrial transection of the rat brain: immuno-histochemical observations with antibodies against glial fibrillary acidicprotein, glutamine synthetase and S-100 protein. Acta Histochem Cyto-chem 24:11–19.

Tanaka H, Araki M, Masuzawa T. 1992. Reaction of astrocytes in the gerbilhippocampus following transient ischemia: immunohistochemical obser-vations with antibodies against glial fibrillary acidic protein, glutaminesynthetase, and S-100 protein. Exp Neurol 116:264–274.

Tanaka H, Masuzawa T. 1993. Cerebral ischemia and astrocytes—withparticular reference to a role of glutamine synthetase in astrocytes. AdvNeurol Sci 37:630–637.

Tanaka H, Katoh K, Shimazaki K, Gomi H, Itohara S, Masuzawa T, KawaiN. 1999. Disturbed hippocampal functions of GFAP knockout mice aftertransient ischemia. Soc Neurosci Abstr 25:2107.

Tsubokawa H, Oguro K, Masuzawa T, Nakajima T, Kawai N. 1995.Effects of spider toxin and its analog on glutamate-activated currents inthe hippocampal CA1 neuron after ischemia. J Neurophysiol 74:218–225.

Tsubokawa H, Oguro K, Robinsdon HPC, Masuzawa T, Kawai N. 1996.Intracellular inositol 1,3,4,5-tetrakisphosphate enhances the calcium cur-rent in hippocampal CA1 neurons of the gerbil following transientforebrain ischemia. J Physiol 497:67–78.

Valentim LM, Michalowski CB, Gottardo SP, Pedroso L, Gestrich LG,Netto CA, Salbego CG, Rodnight R. 1999. Effects of transient cerebralischemia on glial fibrillary acidic protein phosphorylation and immuno-content in rat hippocampus. Neuroscience 91:1291–1297.

Vernadakis A. 1996. Glia-neuron intercommunications and synaptic plas-ticity. Prog Neurobiol 49:185–214.

Wang X, Messing A, David S. 1997. Axonal and nonneuronal cell responsesto spinal cord injury in mice lacking glial fibrillary acidic protein. ExpNeurol 148:568–576.

Wenzel J, Lammert G, Meyer U, Krug M. 1991. The influence of long-term potentiation on the spatial relationship between astrocyte processesand potentiated synapses in the dentate gyrus neuropil of rat brain. BrainRes 560:122–131.

Yamashita K, Vogel P, Fritze K, Back T, Hossmann KA, Wiessner C.1996. Monitoring the temporal and spatial activation pattern of astrocytesin focal cerebral ischemia using in situ hybridization to GFAP mRNA:comparison with sgp-2 and hsp70 mRNA and the effect of glutamatereceptor antagonists. Brain Res 735:285–297.

20 Tanaka et al.