Neuronal expression of GFAP in patients with Alzheimer pathology and identification of novel GFAP splice forms

ORIGINAL RESEARCH ARTICLE

Neuronal expression of GFAP in patients with Alzheimerpathology and identification of novel GFAP splice formsEM Hol*,1, RF Roelofs1, E Moraal1, MAF Sonnemans1, JA Sluijs1, EA Proper2,3, PNE de Graan2,

DF Fischer1 and FW van Leeuwen1

1Netherlands Institute for Brain Research, Meibergdreef 33, 1105 AZ Amsterdam, The Netherlands; 2Rudolf Magnus Institutefor Neurosciences, UMCU, Universiteitsweg 100, 3584 CG Utrecht, The Netherlands

Glial fibrillary acidic protein (GFAP) is considered to be a highly specific marker for glia. Here,we report on the expression of GFAP in neurons in the human hippocampus. Intriguingly, thisneuronal GFAP is coded by out-of-frame splice variants and its expression is associated withAlzheimer pathology. We identified three novel GFAP splice forms: D 135 nt, D exon 6 and D164 nt. Neuronal GFAP is mainly observed in the pyramidal neurons of the hippocampus ofAlzheimer and Down syndrome patients and aged controls, but not in neurons of patientssuffering from hippocampal sclerosis. Apparently, the hippocampal neurons in patients withAlzheimer’s disease pathology are capable of expressing glia-specific genes.Molecular Psychiatry (2003) 8, 786–796. doi:10.1038/sj.mp.4001379

Keywords: glial fibrillary acidic protein; aberrant splicing; neurofibrillary tangles; Alzheimer’sdisease; Down syndrome; neurodegenerative disease

Alzheimer’s disease (AD) is the main cause of seniledementia1 and is accompanied by neuronal cyto-skeletal pathology, amyloid plaques and gliosis.Characteristic deposits of aberrant proteins, such asb-amyloid (A40/42),

2 hyperphosphorylated tau3 andþ1 proteins4 are observed in the neuropathologicalhallmarks of AD, that is, the plaques, tangles andneuropil threads. Furthermore, it is evident that theseverity of the AD pathology strongly correlates withthe density of the activated astrocytes.5 In these cells,the expression of glial fibrillary acidic protein (GFAP)is strongly upregulated.6,7

The þ 1 proteins that accumulate in brains of ADpatients are proteins with a wild-type N-terminus andan out-of-frame C-terminus.8 Owing to molecularmisreading, small deletions occur in mRNA in oradjacent to short repetitive sequences, resulting inframeshifted mRNA and consequently the translationof þ1 proteins.8,9 In a previous study, we have shownthat molecular misreading of two human genesoccurs, that is, amyloid precursor protein (APP) andubiquitin B (UBB). Both aberrant proteins, APPþ1 andUBBþ1, are present in aggregates and coexpressed inneurons in the hippocampus and cortex of ADpatients, Down syndrome (DS) patients and aged

nondemented controls.4 The mechanism of molecularmisreading is still elusive; however, we envisionthat the deletions are either introduced duringtranscription or might reflect post-transcriptionalmodifications.10 Apparently, high expression of agene facilitates the detection of molecular misreading,as we have shown earlier for vasopressin in thehomozygous Brattleboro rat11 and APP in DS patients.4

Therefore, the high level of GFAP expression inAD and DS patients prompted us to studywhether faulty GFAP mRNAs are expressed in thesepatients. The human GFAP gene is located onchromosome 1712 and encodes a class III intermediatefilament protein, which is expressed in astrocytes.13

Recently, it has been shown that mutations in theGFAP gene are associated with Alexander disease,14,15

which is a progressive degenerative neurologicaldisorder. The disease is characterized by cytoplasmicinclusions in astrocytes containing GFAP, the so-calledRosenthal fibers.16 Alexander disease is the firstexample of neuropathology caused by mutations inthe GFAP gene.

In this study, we focused on the human enthorinalcortex and hippocampus, since this area is the earliestand most affected area in AD.17,18 During the course ofour experiments, we observed that frameshifted GFAPis indeed expressed in the hippocampus of AD andDS patients. It turned out that this aberrant GFAPprotein was not the result of classical molecularmisreading, but the result of alternative splicing ofGFAP pre-mRNA. To our surprise, we observed thatthis GFAP protein is expressed in hippocampalneurons in AD and DS patients.

Received 05 August 2002; revised 17 April 2003; accepted 29April 2003

*Correspondence: Dr EM Hol, PhD, Netherlands Institute forBrain Research, Meibergdreef 33, 1105 AZ Amsterdam, TheNetherlands. E-mail: [email protected];3Present address: Department of Clinical Chemistry, Isala Klinie-ken, Weezenlanden, Zwolle, The Netherlands.

Molecular Psychiatry (2003) 8, 786–796& 2003 Nature Publishing Group All rights reserved 1359-4184/03 $25.00

www.nature.com/mp

Materials and methods

Human brain material

Post-mortem human brain material was obtained fromthe Netherlands Brain Bank in Amsterdam, theNetherlands (coordinator: Dr R Ravid). The hippo-campi were obtained at autopsy and fixed for 1 monthby immersion in 4% formalin in neutral phosphate-buffered saline (PBS, pH 7.4). All cases wereneuropathologically confirmed using convential his-topathological stainings including Bodian, methena-mine-silver and Congo red. In addition, hippocampaltissue was obtained by tissue biopsy from pharmaco-resistant temporal lobe epilepsy patients (Dr CWMvan Veelen and Dr PC van Rijen UMCU, Utrecht TheNetherlands) and was fixed in 4% buffered formalinfor 1 day.

Immunocytochemistry

Paraffin sections (6 mm thick) or 50mm thick vibra-tome sections of post-mortem hippocampi of 15 ADpatients, five DS patients, 25 nondemented controlsand epileptic tissue biopsy material of 12 pharma-coresistant temporal lobe epilepsy patients werestained for the presence of GFAPwt and GFAPþ1.The paraffin sections were deparaffinized in xyleneand graded ethanol dilution series and washed inTris-buffered saline (TBS; pH 7.6). These sections andthe vibratome sections were incubated with 100%formic acid on a rocking table for 30 min, rinsed indistilled water for 30 min and in PBS (pH 7.6) for30 min. The vibratome sections were treated withgraded methanols and finally with 20% methanol/0.3% H2O2 to inhibit pseudoperoxidase activity oferythrocytes. Subsequently, the sections were incu-bated with various polyclonal antibodies as describedpreviously.4 These antibodies included: GFAPwt (i)DAKO raised against bovine GFAP (1 : 300 paraffin,1 : 2000 vibratome; DAKO, Carpinteria, CA, USA), (ii)Sigma raised against human GFAP (1 : 1000) or (iii)Dahl19 (1 : 2000), GFAPþ 1 (epitope DRGDAGWRGH,1 : 1000, bleeding 010498), APPþ1 (1 : 250, bleeding020294), UBBþ 1 (1 : 400, Ubi2 bleeding 010994) andGFAPþ 1 preimmune serum (1 : 1000, bleeding270198).

Confocal microscopy

To diminish the autofluorescence in the formalde-hyde-fixed hippocampi, the paraffin sections wereirradiated with light (Philips, Holland 40 W, 301 * CO,230 V, spoltineR63) for 48 h.20 Hereafter, the paraffinsections were deparaffinized in xylene and rehy-drated in graded ethanols. The sections were washedin TBS (pH 7.6) and heated in TBS in a pressurecooker for 20 min. After cooling down, the sectionswere washed twice in TBS and incubated for 48 h at41C with (i) a mouse monoclonal antibody recogniz-ing abnormal tau protein (MC-121), in combinationwith the rabbit polyclonal anti-GFAPþ 1 (bleeding010498; 1 : 100 MC-1/1 : 1000 GFAPþ 1) in supermix(0.05 M Tris, 0.9% NaCl, 0.25% gelatin and 0.5%

Triton X-100, pH 7.6) or (ii) a combination of MC-1(1 : 100) and GFAPwt (1 : 5000, DAKO, Carpinteria, CA,USA). Subsequently, the sections were washed inTBS and incubated for 1 h with anti-mouse-Cy2(Jackson ImmunoResearch Labs; 1 : 50)/anti-rabbit-Cy3 (Jackson ImmunoResearch Labs; 1 : 200)/TO-PRO-3 (Molecular Probes; 1 : 1000). Sections wereembedded in Mowiol (0.1 M Tris-HCl pH 8.5, 25%glycerol, 10% w/v Mowiol 4-88). Images wereacquired by confocal laser scanning microscopy(Zeiss 510).

In situ hybridizationNeuron-specific enolase (NSE) immunostaining wascombined with GFAP in situ hybridization on 6 mmparaffin sections of a 70-year-old female AD patient(NBB-83002). The sections were deparaffinized, re-hydrated and rinsed in RNAse-free TBS. Monoclonalanti-NSE (1 : 100 Novo-Castra in supermix) wasapplied to the sections for an overnight incubationat 41C. Subsequently, the sections were rinsed in TBSfollowed by an incubation with biotin-labeled horse-anti-mouse (Vector, Burlingame, CA, USA; 1 : 400 insupermix) at room temperature for 1 h. After anotherrinse in TBS, the sections were incubated with avidinbiotin complex (ABC, Vector Burlingame, CA, USA;1 : 1000 in supermix) at room temperature for 45 min.The sections were rinsed with TBS and stained withdiaminobenzidine in 0.05 M Tris buffer (pH 7.6). Aftera 2�2 min wash in PBS, the sections were deprotei-nized and processed for in situ hybridization asdescribed before.4 A 40-mer oligonucleotide probe(see Figure 1a) complementary to nt 2328–2367 ofGFAP mRNA (50-CTTAATTCCCACAATCCAGAGGC-CAGTGCAACTGGTCAC-30) was labeled using term-inal transferase and 35S-dATP. Control hybridizationswere performed with a sense probe 50-GTGAC-CAGTTGCACTTGGCCTCTGGATTG TGGGAATTA-AG-30.

Nonradioactive in situ hybridizationNonradioactive in situ hybridization was performedon formalin-fixed, paraffin-embedded human post-mortem hippocampus of a 66-year-old male ADpatient (NBB-88073). Digoxigenin-labeled (DIG)GFAP sense and antisense cRNA was generated byin vitro transcription using T3 or T7 RNA polymerase,respectively. The RNA was transcribed from humanGFAP cDNA nt 1121–1295 (see Figure 1a). The tissuesections (6 mm) were deparaffinized, rehydrated andtreated with proteinase K (10 mg/ml in 10 mM Tris-HCl, pH 7.5, 2 mM CaCl2) at 371C for 30 min. Thesections were postfixed in 4% buffered formalin inPBS for 5 min, followed by an incubation in ammo-nium chloride (1% in PBS) for 15 min. After rinsingtwice with PBS, the sections were incubated in 0.1%Triton X-100 in PBS for 20 min. The sections werehybridized with 200–400 ng/ml DIG-RNA in 200mlhybridization mix (50% formamid, 2�SSC, 10%dextran sulfate, 1� Denhardt’s, 5 mM EDTA, 10 mMphosphate buffer (pH 8.0), 0.5 mg/ml tRNA overnight

GFAP expression in neurofibrillary tanglesEM Hol et al

787

Molecular Psychiatry

at 551C). After hybridization, the sections werewashed in 2�SSC at room temperature for 30 min,2�SSC at 651C for 60 min and 0.1 SSC at 651C for60 min. Sections were transferred to buffer 1 (100 mMTris-HCl, 150 mM NaCl, pH 7.4). Before immunostain-ing, the sections were blocked with 5% FCS, 0.3%triton in TBS at room temperature for 1 h, followed byan overnight incubation with anti-DIG antiserum

conjugated to alkaline phosphatase (1 : 5000 in buffer1, 1% FCS, no triton) at room temperature. Thesections were washed 3� 5 min with buffer 1,equilibrated in buffer 2 (100 mM Tris-HCl, 50 mMMgCl2, 100 mM NaCl, pH 9.5) and the signal wasvisualized with NBT/BCIP overnight at RT. Stainingwas stopped by rinsing in TE pH 8 and slides wereshortly dehydrated in graded alcohols.

Detection of mutation by expression cloning andimmunoscreeningRNA was isolated from freshly frozen pieces ofautopsy brain tissue with Trizol (Gibco-BRL) accord-ing to the manufacturer’s protocol. Total RNA (2 mg)was primed with random hexamers and reversetranscribed with Expand reverse transcriptase at421C (Roche Molecular Biochemicals, Mannheim,Germany). A fragment of GFAP was amplified (nt864–1290, Figure 1a and b) using high-fidelity taqpolymerase (Roche). Primers used were: 50-TTGCGGGATCCCCAGTTGCAGTCCTTGACCT-30 (nt864–883) and 50-TCGAAGCTTGTGCTCCTGCTTG-GACT-30 (nt 1290–1271). For exact position of primerssee Figure 1a and b. Timing and temperatures were:551C for 30 s, 721C for 30 s and 921C for 30 s for 40cycles. The amplified product was isolated from theagarose gel and cloned in frame in the prokaryoticexpression vector PQE31 (Qiagen) or pBKS- (Strata-gene). After transformation, the protein productionwas induced by IPTG. Bacterial clones were copied tonitrocellulose filters. The filters were baked andstained for the GFAPþ1 protein. Several clones thatwere producing GFAPþ 1 immunoreactive proteinwere identified and their inserts were sequencedwith the Sequenase version 2 kit (USB, Cleveland,OH, USA).

Results

Designing the antibody against frameshifted GFAPThe sequence of the human GFAP gene can be foundin GenBank as part of the ‘homo sapiens chromosome17 working draft sequence segment’ (accession num-ber NT 010765, GFAP gene nt 189463–199332). Wededuced the exon–exon boundaries by comparing theknown exon–exon transitions in the mouse GFAPmRNA (GenBank accession number X02801) with thehighly homologous (87%) human GFAP mRNA(GenBank accession number J04569).12 The humanGFAP mRNA is 3017 bp in length and consists of nineexons; exon 1 nt 1–464, exon 2 nt 465–525, exon 3 nt526–621, exon 4 nt 622–783, exon 5 nt 784–909, exon6 nt 910–1130, exon 7 nt 1131–1174, exon 8 1175–1260 and exon 9 1261–3017. The coding region startsat nt 4 and stops early in exon 9 at nt 1302 (Figure 1a).

The initial focus of our studies was to find outwhether molecular misreading of the GFAP geneoccurs. From our earlier studies, we know that short-repetitive sequences, like GAGAG, are hotspots formolecular misreading.4,22 The mRNA sequence ofGFAP includes 10 GAGAG-motifs and six of these are

Figure 1 (a) Glial fibrillary acidic protein. The humanGFAP gene contains nine exons, the boundaries of whichare drawn in this schematic diagram of GFAP mRNA.Within the GFAP mRNA, 10 GAGAG-motifs are present, sixmotifs are comprised within the coding sequence (CDS),which is the GFAPa mRNA. A combined approach of PCRand immunoscreening was applied to search for possiblemutations near the fifth GAGAG motif (see Materials andmethods). The position of the (i) 408 bp PCR product (nt864–1290 of GenBank accession number J04569) amplifiedfor the immunoscreening approach, (ii) the riboprobe (nt1121–1295) used for the nonradioactive in situ hybridiza-tion and (iii) the oligoprobe (nt 2328–2367) used for theradioactive in situ hybridization are indicated. (b) Part ofthe human GFAP coding sequence. The sequence betweenthe forward and reverse PCR primers was amplified andcloned into a prokaryotic expression vector for the im-munoscreening approach. The GFAPþ1 antibody was raisedagainst the very C-terminus of the frame-shifted GFAPprotein, that is, DRGDAGWRGH (bold amino acids). Thispeptide was chosen based on an anticipated dinucleotidedeletion in the GAGAG motif at position 922–926, whichwould give rise to a mutant GFAP protein, that is, GFAPþ 1,with an extended out-of-frame C-terminus.


788


within the coding region (Figure 1a). Classicalmolecular misreading4 would result in a dinucleotidedeletion in one of these motifs or adjacent to such asequence. Since only a dinucleotide deletion in thefifth motif would result in an extended frameshiftedC-terminus (Figure 1b), we decided to raise a rabbitpolyclonal antibody against this putative frameshiftedGFAP protein, that is, DRGDAGWRGH, as describedbefore.4 An extensive BLAST search was performed toexclude homology of the 10 amino-acid long C-terminal sequence DRGDAGWRGH to other proteins.Furthermore, this sequence was compared to all sixreading frames of the nonredundant nucleotidedatabase in GenBank to check for homology. Neitherthe protein BLAST nor the nucleotide BLASTrevealed other proteins or genes, except for GFAP,that show homology to DRGDAGWRGH. We termedthe partly frameshifted GFAP as GFAPþ1, in accor-dance with APPþ 1 and UBBþ 1 proteins reportedearlier.4 The GFAPþ1 antibody does not crossreactwith purified human GFAPwt protein (ICN Biomedi-cals; Western blot: results not shown).

Frameshifted GFAP protein is expressed in neurons inthe hippocampus of patients with AD pathologyTo test whether molecular misreading of the GFAPgene can occur, we started by analyzing the expres-sion of the frameshifted GFAP protein, GFAPþ1, inthe human hippocampus. The group of subjectsconsisted of 20 nondemented controls (age 34–90years, 13 male, seven female), 15 AD patients (age 40–92 years, six male, nine female) and five DS patients(age 58–67 years, one male, four female). Paraffinsections of hippocampi of these patients wereimmunostained for the presence of GFAPþ 1 andGFAPwt (Table 1). Because GFAP is believed to be acell-specific marker for astrocytes, we expectedGFAPþ 1 protein expression to be confined to astro-cytes. In nondemented controls, without AD neuro-pathology, no GFAPþ1 immunoreactivity wasobserved (Figure 2a). Surprisingly, the GFAPþ 1

immunopositive cells observed in elderly controlsand AD patients were primarily neurons (Figure 2cand e, arrowheads), although some sporadic astro-cytes were immunopositive as well. GFAPþ1 immu-noreactivity was observed in the hippocampus of 12AD patients, and also in five DS patients (Table 1),who suffer from age-dependent Alzheimer-type neu-rodegeneration.23 Although GFAPþ 1-expressing neu-rons were mainly restricted to AD and DS patients, afew isolated GFAPþ1 immunopositive neurons werefound in the entorhinal cortex and CA1 area of fourout of 21 nondemented controls (Figure 2c, arrow-head). In the hippocampi of these nondementedcontrols, neuropathological changes, that is, plaquesand tangles, were observed as well. Three of theGFAPþ 1 immunopositive controls were older than 70years (Table 1). Immunostaining with the polyclonalantibody of DAKO, directed against bovine GFAPwt,revealed the expected presence of GFAP in astrocytesin each case (Figure 2b, d and f, arrows). Careful

examination of the paraffin sections stained forGFAPwt confirmed that pyramidal neurons in thehippocampus indeed were GFAP immunopositive(Figure 2f, arrowheads), but only in AD and DSpatients and in some nondemented controls withinitial neuropathology (Table 1). The GFAPþ1 pre-immune serum showed no staining at all (results notshown). The GFAPwt expressing neurons in theparaffin sections showed a discrete, but weak,immunopositive signal compared to the astrocyticstaining.

To consolidate these data, we successively immu-nostained 50 mm thick vibratome sections of a 92-year-old female AD patient (NBB-96115). Both, astrocytesand pyramidal neurons in the hippocampus wereclearly stained with the DAKO polyclonal GFAPwt

antibody (Figure 3a and c). In contrast, mainlyneurons, including tangle-shaped neurons, expressGFAPþ 1 (Figure 3b). An occasional GFAPþ 1 immu-nopositive astrocyte is observed as well (Figure 3d).Staining with three different polyclonal GFAPwt

antisera, that is, Sigma (result not shown), DAKO(Figure 4a) and Dahl19 (Figure 4c), gave similar resultsand strengthened the initial results of GFAP expres-sion in neurons and astrocytes in the hippocampus ofAD patients. These antibodies stained the sameneuronal population as the GFAPþ 1 antibody in ADpatient NBB-96115 (Figure 4b), indeed confirming theobservation that neurons can express GFAP.

GFAP mRNA is expressed in neuronsWe validated the presence of neuronal GFAP proteinby studying the expression of GFAP mRNA inneurons by two different in situ hybridizationmethods, with either a 35S-labeled oligonucleotideor a digoxigenin-labeled riboprobe (see Figure 1a forprobe positions). With the 35S-labeled 50 mer GFAPoligonucleotide hybridizations, we observed GFAPexpression in cells that were NSE immunopositive(Figure 5a and b, arrows). In agreement with theincreased expression of GFAP mRNA in activated,intense clusters of silver grains were also observed inNSE-negative cells (Figure 5a and b, arrowheads). Thehybridization with the sense probe showed no signal.The nonradioactive in situ hybridization showedconclusively that GFAP mRNA is expressed in tangles(Figure 5c, arrow). Many GFAP-expressing astrocyteswere observed as well (Figure 5d, e and f, arrow-heads). Additionally, we observed that GFAP mRNA(Figure 5e) and GFAP protein (Figure 5f) coexist inneurons and glia in consecutive sections. The resultsof the in situ hybridization confirm the protein data,thereby corroborating the finding that GFAP isexpressed in these neurons.

Neuronal, but not astroglial, GFAPþ 1 expression isassociated with AD pathologyTo analyze whether GFAPþ1 immunostaining isassociated with AD and DS, and not merely the resultof gliosis, GFAP and GFAPþ1 stainings were per-formed on hippocampal sections of 12 patients


789


suffering from pharmaco-resistant temporal lobeepilepsy (Table 2). This hippocampal tissue isobtained by tissue biopsies and is fixed in formalinfor a short period (1 day), compared to the 1-monthfixation of post-mortem AD and DS patients’ hippo-campi. Therefore, five hippocampi from nondemen-ted control patients with a fixation time of 1 day wereincluded as controls for this study. As an additionalcontrol, a hippocampus from AD patients with 1-day

fixation time was stained for GFAPþ 1 expression.Pyramidal neurons in this hippocampus were posi-tive for GFAPþ 1, indicating that the short fixationtime did not influence immunostaining of the tissue.The epilepsy patients are classified into two groups:patients with severe hippocampal sclerosis (HS) andwithout sclerosis (non-HS). Especially, patients withsevere HS are known to display an increase in GFAPimmunoreactivity.24 In none of the epileptic patients

Table 1 GFAPwt and GFAPþ 1 immunostaining in hippocampi of nondemented controls, AD patients, and Down syndrome patients

Patient # NBB Age (years) Sex Neuropathology GFAPwt GFAPþ1

Plaques Tangles Astrocytes Neurons Astrocytes Neurons

Nondemented controls89003 34 M � � þ � � �97162 38 M � � þ � � �94060 45 M � � þ � � �97159* 48 M � � þ � � �98169 49 M � � þ � � �98006* 50 M � � ND ND � �94125 51 M � � þ � � �94119 51 F � � þ � � �87051 51 F � � þ � � �86003 53 M þ þ þ � þ þ82015 57 F � � þ � � �88037 58 M � þ (some) þ � � �88008 60 M � � þ � � �91125 61 M þ þ (some) þ � � �90079 72 M þ þ þ þ � þ91026 80 F þ þ þ � � �91027 82 F � þ þ þ � þ90080 85 M þ þ þ � � �90083 90 F � þ (some) þ � � �81007 90 F þ þ þ þ � þ

Alzheimer’s disease89057* 40 M þ þ þ �/þ � þ90102 49 M þ þ þ � � �/þ91092 54 F þ þ þ � � þ88073 66 M þ þ þ þ � þ93047 70 M þ þ þ þ � �83002 70 F þ þ þ þ � þ91094 73 F þ þ þ þ � þ þ90118 77 M þ þ þ þ � þ93045 83 F þ þ þ þ � þ88028 85 F þ þ þ þ � þ91081 85 F þ þ þ þ � þ90117 86 M þ þ þ � � �96114 88 F þ þ þ �/þ � þ86002 90 F þ þ þ þ � �96115* 92 F þ þ þ þ � þ þ

Down syndrome patients92080 58 F þ þ þ �/þ � þ89055* 59 F þ þ þ þ þ þ93161 62 F þ þ þ þ � þ94058 64 M þ þ þ �/þ � þ93028* 67 F þ þ þ þ � þ

ND, not done; NBB, Netherlands Brain Bank, autopsy material.*Patient’s RNA used for immunoscreening.


790


neuronal GFAPþ1 staining was observed. GFAPþ 1

staining was observed in a few astrocytes in three HSpatients and one non-HS patient; only two or threeGFAPþ 1 immunopositive astrocytes per section werefound, which is comparable to the number of GFAPþ 1

immunopositive astrocytes in AD and controlpatients. Therefore, our data indicate that the occur-rence of the GFAPþ1 protein in neurons is associatedwith AD pathology. In fact, additional immunostain-ings on the temporal cortex showed that GFAP is alsoexpressed in tangle-shaped neurons in the temporalcortex of patients with AD pathology (data notshown).

Subsequently, a confocal study on hippocampalsections of AD patients was performed with anti-bodies against GFAPþ1 or GFAPwt and an antibodydirected against abnormally phosphorylated tau, thatis, MC-1. This double-labeling experiment showedthat GFAPþ1 is indeed coexpressed with abnormallyphosphorylated tau (p-Tau; MC-1 antibody) in a

number of tangle-shaped neurons in the hippocam-pus of patients with AD pathology (Figure 6a). Someneurons also coexpressed GFAPwt and p-Tau (Figure6b). In Figure 6b, clearly cells can be identified thatnot only express GFAP wt (arrowhead) or p-TAU(asterisk), but also cells that coexpress GFAPwt andp-Tau (arrows).

Coexpression with other frameshifted proteinsEarlier, we have reported on the expression of twoframeshifted proteins in AD neurons, that is, APPþ 1

and UBBþ1.4 To characterize the neurons that doexpress GFAPþ1 further, we stained three consecutive6-mm-thick paraffin sections of a 70-year-old femaleAD patient (NBB-83002) for the expression of UBBþ 1

(Figure 7a and d), GFAPþ 1 (Figure 7b and e) andAPPþ1 (Figure 7c and f). We observed a clearcolocalization of GFAPþ1, APPþ1 and UBBþ1 inseveral neurons in the hippocampus of this ADpatient (Figure 7), suggesting a common pathwayresulting in the accumulation of frameshifted pro-teins.

Splice variants of GFAPIn order to find out whether molecular misreadingunderlies the expression of the GFAPþ 1 protein inneurons, we examined the occurrence of frameshiftedtranscripts expressed in the human brain. Total RNAwas isolated from two nondemented controls (NBB-97159, NBB-98006), two AD patients (NBB-89057,NBB-96115) and two DS patients (NBB-89055, NBB-93028) (footnote * in Table 1). GFAP was amplified by

Figure 2 GFAPþ1 and GFAPwt expression in the humanhippocampus. GFAPþ1 (a, c and e) and GFAPwt (b, d and f)immunoreactivity in paraffin sections of nondementedcontrols (NBB-94119 and NBB-91027) and an AD patient(NBB-96115). In the hippocampus of an AD patient,numerous neurons express GFAPþ 1 (arrowheads, e). Incontrast with the nondemented controls that either showedno GFAPþ1 immunoreactivity (a) or only a few GFAPþ1

expressing neurons (arrowhead, c). GFAPwt is expressed inastrocytes (arrows) in the nondemented controls (b and d)and AD case (f). The asterisks in (e) mark nonspecificstaining of capillaries, bar¼100mm.

Figure 3 GFAPwt and GFAPþ1 in neurons and astrocytes inAD. In 50 mm-thick vibratome sections of the hippocampusof an AD patient (NBB-96115) distinct GFAPwt (a) andGFAPþ 1 (b) expressing neurons, including tangles, wereobserved. Astrocytes clearly are GFAPwt immunopositive(c, asterisk). A minority of these astrocytes also expressGFAPþ 1 (d, asterisk). Bar in (c) (same magnification as in aand b) and in (d)¼50 mm.


791


polymerase chain reaction (PCR) using Taq polymer-ase with proofreading capacity to avoid introductionof mutations. We amplified only part of the mRNA,that is, nt 864–1290 (GenBank accession numberJ04569), which includes the predicted frameshiftmutation in the fifth GAGAG motif (Figure 1b). Theresulting PCR product was cloned in a prokaryoticexpression vector, protein expression was inducedand GFAPþ1 immunonegative and immunopositivecolonies were sequenced. All six patients, AD, DS aswell as the nondemented controls, revealed immuno-positive clones, indicating transcripts with frameshiftmutations. The immunonegative colonies were alleither GFAPwt, or contained a PCR-product with a135 nt in-frame deletion of GFAPwt (D nt 910–1044).The D 135 variant was found in the followingpatients, NBB 98057 (AD), 89052 (DS) and 93028(DS). In contrast to dinucleotide deletions that wefound earlier in APP and UBB mRNA,4 we found twoadditional GFAP out-of-frame splice forms (Figure 8a)that accounted for the expression of GFAPþ1 protein.Either exon 6 was spliced out (D nt 910–1130), or wefound a deletion of 164 nt (D nt 976–1139). The D

Figure 4 Different GFAP antibodies label pyramidalneurons in the hippocampus of an AD patient. Immunos-tained 50 mm vibratome sections of an AD patient (NBB-96115) clearly show that polyclonal GFAPwt antibodies fromdifferent sources (a) DAKO and (c) Dahl label pyramidalneurons (arrow heads) as well as astrocytes in thehippocampus. Note the clear hippocampal gliosis, due tothe AD pathology, in (a) and (c) (arrows). In the samehippocampal area, many GFAPþ1-immunopositive neuronsare observed (b, arrow heads). In (a) and (b) highermagnifications of immunopositive neurons are depicted,bar¼100mm.

Figure 5 Neurons express GFAP mRNA. A few NSE-immunopositive neurons in the hippocampus of an ADpatient (NBB 83002) express GFAP mRNA (a and b, arrows).The arrowheads indicate GFAP-expressing astrocytes. Thesections were counterstained with hematoxylin. With DIG-labeled riboprobes, tangles (c, arrow) and astroycytes (d,arrow heads) were observed in the hippocampus of an ADpatient (NBB: 88073). The GFAP mRNA (e) colocalizes withGFAPwt immunoreactivity (f). A tangle-shaped neuron(arrow) and astrocytes (arrowheads) express GFAP mRNA(e) as well as protein (f). The inset shows a highermagnification of the tangle-shaped neuron. NSE-IR¼neuron-specific enolase immunoreactivity, bar¼50mm.


792


exon 6 splice form was found in NBB 98006 (CON)and NBB 96115 (AD) and the D 164 form was found inNBB 89052 (DS) and NBB 93028 (DS). The D 135 formresults in the translation of a shorter GFAPwt protein,which lacks part of coil 2B. The two out-of-framesplice forms, that is, D exon 6 and D164, are translatedin GFAP proteins, which lack part of coil 2B and theentire tail region. The frameshifted C-terminus ofthese proteins contain the complete epitope thatis recognized by the GFAPþ 1 antibody (Figures 8band 1b).

Discussion

Our study provides clear evidence that GFAP is notexclusively expressed by glial cells, as has been theconsensus for many years. Remarkably, we alsoobserved the expression of GFAP mRNA and proteinin hippocampal neurons in AD and DS patients andaged controls with early AD type of pathology. Thisneuronal GFAP, GFAPþ1, is an out-of-frame variant ofGFAPwt and is the product of two novel splice formsof GFAP mRNA. In total, three novel splice formswere found, which are expressed in nondementedcontrols as well as AD and DS patients. Theexpression of the GFAPþ1 protein in the hippocam-pus, however, is confined to a population of pyrami-

dal neurons in areas with AD pathology, in AD andDS patients and in elderly nondemented controls.

GFAP is a classical marker for glia cells, includingastrocytes. An extensive literature search revealedtwo papers describing possible GFAP expression inneurons. (i) In tuberous sclerosis patients, the typicalgiant cells in the brain show neuronal as well as glialcell-like features. These giant neuronal-like cellsexpress GFAP.25 (ii) In AD patients, it has beendescribed that occasionally hippocampal neurofibril-lary tangles were GFAP immunopositive, using amonoclonal GFAP antibody.7 It has been discussedbefore that GFAP-immunopositive tangles in ADbrains are actually GFAP-immunopositive fine pro-cesses of astrocytes that penetrate the bundles ofpaired helical filaments in ghost tangles.26,27 Our data,however, provide strong evidence that GFAP isexpressed in neurons that are no ghost tangles, sincethe GFAPwt and GFAPþ1 antibodies stain neurons inwhich a nucleus is visible. Furthermore, the presenceof GFAP mRNA in neurons (Figure 5a, c and e) provesthat the GFAP immunostaining observed in theneurons is indeed the result of GFAP expression.However, we cannot exclude the fact that GFAPwt isalso expressed by these cells, since the probe used forin situ hybridization recognizes the novel GFAPsplice forms and the full-length GFAPwt mRNA.Furthermore, the GFAPwt antibody of DAKO willstain GFAPwt protein as well as GFAPþ1 protein.GFAP immunostaining is routinely performed onparaffin- or formalin-fixed frozen sections of thehuman brain. Our experience is that neuronal GFAPstaining in paraffin sections is much weaker com-pared to the staining in the thicker vibratomesections, which might explain why other researchershave never appreciated the neuronal staining as agenuine staining. As mentioned before, GFAP-immu-nopositive neurons have been reported by others, butthis staining has been explained as astrocytic pro-cesses engulfing ghost tangles.27 By staining 50 mmthick vibratome sections, the intraneuronal expres-sion of GFAP could be observed, indicating that thesecells are not ghost tangles.

By expression cloning, we observed that thepresence of the GFAPþ1 protein was caused by theexpression of novel GFAP out-of-frame splice forms.The neuronal GFAPþ1 protein is either the transla-tional product of the D 164 mRNA or of the D exon 6GFAP mRNA. In addition to the out-of frame spliceforms, we also observed a novel in-frame GFAP spliceform, that is, D 135. Alternative splicing of GFAP hasbeen reported by others,28–31 but the splice forms ofthe present study have not been described before.Since we were specifically aiming at studying theoccurrence of molecular misreading around theGAGAG motif in exon 6, we amplified by PCR thepart of the mRNA that starts in exon 5 and includesthe coding region of the final C-terminus of theGFAPwt protein. Further studies are needed toinvestigate the full-length cDNAs of the splice formsdescribed in this study. Our approach, though, clearly

Table 2 GFAPwt and GFAPþ 1 immunostaining in hippo-campi of controls, epileptic patients without hippocampalsclerosis and with hippocampal sclerosis

Patient #UMCU

Age(years)

Sex GFAPwt GFAPþ 1

Astro-cytes

Neurons Astro-cytes

Neurons

Controls – Autopsy1 44 M þ � � �2 48 F þ � � �3 54 M þ � � �4 56 F þ � � �5 75 F þ � � �

Epilepsy patients without hippocampal sclerosis – Biopsy1 11 F þ � � �2 24 F þ � � �3 36 F þ � � �4 40 M þ � � �5 40 M þ � � �6 44 M þ � þ �

Epilepsy patients with hippocampal sclerosis – Biopsy1 31 F þ � þ �2 32 M þ � � �3 35 M þ � � �4 37 M þ � þ �5 38 M þ � þ �6 45 M þ � � �

UMCU: University Medical Center Utrecht.


793


showed that three novel splice forms are expressed inthe human brain and two of these splice formsexplain the presence of the GFAPþ1 protein inhippocampal neurons.

Eukaryotic pre-mRNA splicing is regulated byconsensus sequences at the intron boundaries andbranch site.32 According to the GT–AG splicing rule,the splice acceptor site consensus sequence isGTAAGT and the splice donor site consensus se-quence is Py12NCAG. The GFAP D 135 form is an in-frame alternative splice form, since the 5’splice site isthe splice donor of intron 5 and the 3’splice site is a

splice site consensus sequence (Py12NCAG) in exon 6,that is, ggCCCgCCaCTTGCAG. The GFAP D exon 6form is an out-of-frame alternative splice form and the

Figure 6 Colocalization of GFAPþ 1 and GFAPwt withabnormally phosphorylated tau. Confocal pictures of dou-ble-labeled paraffin sections of the hippocampus of two ADpatients (NBB 96115 and NBB 83002) clearly show thatGFAPþ 1 colocalizes in neurons that express abnormallyphosphorylated tau (MC-1/p-Tau) (a). The double labelingwith GFAPwt and MC-1 (b) clearly shows neurons thatexpress both proteins (arrows) and neurons that onlyexpress p-Tau (*) and cells that only express GFAPwt

(arrowhead). The large panels (a and b) show the mergedred (GFAPþ1 or GFAPwt), green channel (MC-1) and bluechannel (TO-PRO; nuclear dye). The separate green and redchannels of (a) and (b) are depicted in the smaller panels tothe right.

Figure 7 þ 1 Proteins are coexpressed. The frameshiftedproteins UBBþ1 (a, d), GFAPþ1 (b, e) and APPþ1 (c, f)coexist in pyramidal neurons in the hippocampus of ADpatients. Note the three capillaries that are present in eachsection: *, # and .. The boxed area in (a, b and c) is shownin a higher magnification in (d, e and f). The proteins areclearly colocalized in the neurons indicated with 1, 2 and 3.Patient NBB-83002, bar (a, b, c)¼100mm, bar (d, e, f)¼50mm.

Figure 8 GFAP splice variants and protein isoforms. (a)Sequence analysis of the three found splice forms. The D135 nt form is an alternative splice variant and results in aGFAP protein with a shortened coil 2B. The D exon 6 form isan alternative splice variant and the D 164 nt form is anaberrant splice variant, both resulting in a GFAP proteinwith a shortened coil 2B and an out-of-frame C-terminus. (b)Schematic drawing representing the four GFAP isoformsthat are the result of the described splice forms. All spliceforms result in a shortened coil 2B . The D exon 6 and D164 nt form result in the out-of-frame þ 1 C-terminus ’ L1,L1/2 and L2 are the linker sequences in between the coilregions.


794


GFAP D 164 form is an out-of-frame cryptic spliceform, since both the splice acceptor (50-CAG-GAGGCGCT- - - - - -) as well as the splice donor(- - - - - - - -GATCCACCAT-30) do not resemble splicesites’ consensus sequences.

GFAP is the major component of the intermediatefilaments in mature astrocytes,33 and is important forproviding mechanical strength to astrocytes. Theprecise function of GFAP, however, is still not wellunderstood. For instance, GFAP knockout miceappear to be normal and show no obvious anatomicalabnormalities in the CNS.34 However, an additionallack in vimentin in GFAP-/- astrocytes results inproblems with the formation of intermediate fila-ments.35 The 50 kDa human GFAP protein is, likeother intermediate filaments, composed of a head, rodand tail region.13 The novel splice forms described inthe present study all lack part of the coil 2B and theout-of-frame splice forms completely lack the tail. Therod region plays an important role in the formation of10 nm filaments and the tail region has been im-plicated to be involved in the interaction of the GFAPwith other proteins.36 Therefore, the novel GFAPisoforms might affect the structure of the cell, since itmight interfere with the neuronal intermediate fila-ments.

The main question is why hippocampal neurons inAD patients start expressing a glia-specific transcript?It has been reported before that multipotent astrogliacan differentiate in either astrocytes or neurons.37 Theimmature astrocytes are mainly GFAP immunoposi-tive, but are capable of changing their expressionpattern and start expressing neuron-specific proteins,like neuron-specific markers b-III tubulin, L1 andmicrotubule-associated protein 2. In addition, it hasbeen shown that oligodendrocyte precursor cells canrevert to multipotential neural stem cells, which candifferentiate in neurons and astrocytes.38 The hall-mark of neuronal stem cells is their capacity togenerate new cells that will differentiate in eitherglia or neurons. It has been shown by severallaboratories that in AD patients the neurons at risk,including neurons containing intracellular tangles,express cell cycle-related proteins, such as P16 andCDK4,39 cdc2/cyclin B1,40 and nucleolin.41 Recently,it was reported that hippocampal pyramidal neuronscomplete a full S-phase, demonstrating that DNAreplication occurs in neurons before neuronal death.42

Therefore, we propose that the data on the initiationof the cell cycle in combination with our data on theinduction of GFAP in AD hippocampal neuronsreflect a retro-differentiation of these neurons. Itmight be a defense mechanism of the neuron, tosurvive and eventually, if survival is successful,differentiate in neurons again.

In the neurons that express the GFAPþ 1 proteinother þ 1 proteins, APPþ1 and UBBþ1, were alsofound. This finding implicates that neurons that doaccumulate these proteins might have a generalfailure in the degradation of aberrant proteins. Thisis likely to be caused by the presence of UBBþ1 in

these neurons, since UBBþ 1 is known to inhibitproteasomal degradation43,44 and eventually causeneuronal death.45 Therefore, aberrant proteins, in-cluding GFAPþ 1, might become apparent in neuronsin the hippocampus, due to an initiation of GFAPgene expression in combination with a failure in theproteasomal degradation machinery, which can beinitiated by molecular misreading of the UBB gene.

Acknowledgements

We thank the Netherlands Brain Bank team (coordi-nator R Ravid) for the human post-mortem brainmaterial, CWM van Veelen and PC van Rijen for thebiopsy material and P Davies for supplying the MC1antiserum. The research is supported by HFSP(RG0148/1999-B), NWO-MW (903-51-179) andNWO-memory processes and dementia (970-10-029and 970-10-002), EU 5th framework (QLRT-02238),Hersenstichting Nederland (H00.06). EA Proper wassupported by the Epilepsy Fund of the Netherlands(96-04).

References

1 Ott A, Breteler MM, van Harskamp F, Claus JJ, van der Cammen TJ,Grobbee DE et al. Prevalence of Alzheimer’s disease and vasculardementia: association with education. The Rotterdam study. BMJ1995; 310: 970–973.

2 Selkoe DJ. The cell biology of beta-amyloid precursor protein andpresenilin in Alzheimer’s disease. Trends Cell Biol 1998; 8:447–453.

3 Alonso AC, Grundke-Iqbal I, Iqbal K. Alzheimer’s diseasehyperphosphorylated tau sequesters normal tau into tangles offilaments and disassembles microtubules. Nat Med 1996; 2: 783–787.

4 Van Leeuwen FW, De Kleijn DPV, Van den Hurk HH, Neubauer A,Sonnemans MAF, Sluijs JA et al. Frameshift mutants of b amyloidprecursor protein and ubiquitin-B in Alzheimer’s and Downpatients. Science 1998; 279: 242–247.

5 Muramori F, Kobayashi K, Nakamura I. A quantitative study ofneurofibrillary tangles, senile plaques and astrocytes in thehippocampal subdivisions and entorhinal cortex in Alzheimer’sdisease, normal controls and non-Alzheimer neuropsychiatricdiseases. Psychiatry Clin Neurosci 1998; 52: 593–599.

6 Hanzel DK, Trojanowski JQ, Johnston RF, Loring JF. High-throughput quantitative histological analysis of Alzheimer’sdisease pathology using a confocal digital microscanner. NatBiotechnol 1999; 17: 53–57.

7 Beach TG, Walker R, McGeer EG. Patterns of gliosis in Alzheimer’sdisease and aging cerebrum. Glia 1989; 2: 420–436.

8 Van Leeuwen FW, Burbach JPH, Hol EM. Mutations in RNA: a firstexample of molecular misreading in Alzheimer’s disease. TrendsNeurosci 1998; 21: 331–335.

9 Van Leeuwen FW, Fischer DF, Kamel D, Sluijs JA, SonnemansMAF, Benne R et al. Molecular misreading: a new type oftranscript mutation expressed during aging. Neurobiol Aging2000; 21: 879–891.

10 Hol EM, Neubauer A, De Kleijn DPV, Sluijs JA, Ramdjielal RDJ,Sonnemans MAF et al. Dinucleotide deletions in neuronaltranscripts: a novel type of mutation in non-familial Alzheimer’sdisease and Down syndrome patients. Prog Brain Res 1998; 117:379–395.

11 Evans DAP, Van der Kleij AAM, Sonnemans MAF, Burbach JPH,Van Leeuwen FW. Frameshift mutations at two hotspots invasopressin transcripts in post-mitotic neurons. Proc Natl AcadSci USA 1994; 91: 6059–6063.


795


12 Brenner M. Structure and transcriptional regulation of the GFAPgene. Brain Pathol 1994; 4: 245–257.

13 Reeves SA, Helman LJ, Allison A, Israel MA. Molecular cloningand primary structure of human glial fibrillary acidic protein. ProcNatl Acad Sci USA 1989; 86: 5178–5182.

14 Brenner M, Johnson AB, Boespflug-Tanguy O, Rodriguez D,Goldman JE, Messing A. Mutations in GFAP, encoding glialfibrillary acidic protein, are associated with Alexander disease.Nat Genet 2001; 27: 117–120.

15 Aoki Y, Haginoya K, Munakata M, Yokoyama H, Nishio T, TogashiN et al. A novel mutation in glial fibrillary acidic protein gene in apatient with Alexander disease. Neurosci Lett 2001; 312: 71–74.

16 Messing A, Goldman JE, Johnson AB, Brenner M. Alexanderdisease: new insights from genetics. J Neuropathol Exp Neurol2001; 60: 563–573.

17 Braak H, Braak E. Neuropathological stageing of Alzheimer-relatedchanges. Acta Neuropathol 1991; 82: 239–259.

18 Thal DR, Rub U, Orantes M, Braak H. Phases of A beta-depositionin the human brain and its relevance for the development of AD.Neurology 2002; 58: 1791–1800.

19 Dahl D, Bignami A. Immunogenic properties of the glial fibrillaryacidic protein. Brain Res 1976; 116: 150–157.

20 Neumann M, Gabel D. Simple method for reduction of autofluor-escence in fluorescence microscopy. J Histochem Cytochem 2002;50: 437–439.

21 Jicha GA, Bowser R, Kazam IG, Davies P. Alz-50 and MC-1, a newmonoclonal antibody raised to paired helical filaments, recognizeconformational epitopes on recombinant tau. J Neurosci Res 1997;48: 128–132.

22 van Den Hurk WH, Willems HJ, Bloemen M, Martens GJ. Novelframeshift mutations near short simple repeats. J Biol Chem 2001;276: 11496–11498.

23 Seidl R, Cairns N, Lubec G. The brain in Down syndrome. J NeuralTransm Suppl 2001; 61: 247–261.

24 Proper EA, Oestreicher AB, Jansen GH, Veelen CW, van Rijen PC,Gispen WH et al. Immunohistochemical characterization of mossyfibre sprouting in the hippocampus of patients with pharmaco-resistant temporal lobe epilepsy. Brain 2000; 123: 19–30.

25 Bender BL, Yunis EJ. Central nervous system pathology of tube-rous sclerosis in children. Ultrastruct Pathol 1980; 1: 287–299.

26 Probst A, Ulrich J, Heitz PU. Senile dementia of Alzheimer type:astroglial reaction to extracellular neurofibrillary tangles in thehippocampus. Acta Neuropathol 1982; 57: 75–79.

27 Yamaguchi H, Morimatsu M, Hirai S, Takahashi K. Alzheimer’sneurofibrillary tangles are penetrated by astroglial processes andappear eosinophilic in their final stages. Acta Neuropathol 1987;72: 214–217.

28 Feinstein DA, Weinmaster GA, Milner R. Isolation of cDNA clonesencoding rat glial fibrillary acidic protein: expression in astrocytesand in Schwann cells. J Neurosci Res 1992; 32: 1–14.

29 Zelenika D, Grima B, Brenner M, Pessac B. A novel glial fibrillaryacidic protein mRNA lacking exon 1. Brain Res Mol Brain Res1995; 30: 251–258.

30 Condorelli DF, Nicoletti VG, Barresi V, Conticello SG, Caruso A,Tendi EA et al. Structural features of the rat GFAP gene andidentification of a novel alternative transcript. J Neurosci Res1999; 56: 219–228.

31 Nielsen AL, Holm IE, Johansen M, Bonven BJ, Jorgensen P,Jorgensen AL. A new splice form of glial fibrillary acidic protein,GFAP epsilon, interacts with the presenilin proteins. J Biol Chem2002; 277: 29983–29991.

32 Mount SM. Genomic sequence, splicing, and gene annotation. AmJ Hum Genet 2000; 67: 788–792.

33 Eng LF. Glial fibrillary acidic protein (GFAP): the major protein ofglial intermediate filaments in differentiated astrocytes. J Neu-roimmunol 1985; 8: 203–214.

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

35 Pekny M, Johansson CB, Eliasson C, Stakeberg J, Wallen A,Perlmann T et al. Abnormal reaction to central nervous systeminjury in mice lacking glial fibrillary acidic protein and vimentin.J Cell Biol 1999; 145: 503–514.

36 Inagaki M, Nakamura Y, Takeda M, Nishimura T, Inagaki N. Glialfibrillary acidic protein: dynamic property and regulation byphosphorylation. Brain Pathol 1994; 4: 239–243.

37 Laywell ED, Rakic P, Kukekov VG, Holland EC, Steindler DA.Identification of a multipotent astrocytic stem cell in the immatureand adult mouse brain. Proc Natl Acad Sci USA 2000; 97: 13883–13888.

38 Kondo T, Raff M. Oligodendrocyte precursor cells reprogrammedto become multipotential CNS stem cells. Science 2000; 289:1754–1757.

39 McShea A, Harris PL, Webster KR, Wahl AF, Smith MA. Abnormalexpression of the cell cycle regulators P16 and CDK4 inAlzheimer’s disease. Am J Pathol 1997; 150: 1933–1939.

40 Vincent I, Jicha G, Rosado M, Dickson DW. Aberrant expression ofmitotic cdc2/cyclin B1 kinase in degenerating neurons ofAlzheimer’s disease brain. J Neurosci 1997; 17: 3588–3598.

41 Dranovsky A, Vincent I, Gregori L, Schwarzman A, Colflesh D,Enghild J et al. Cdc2 phosphorylation of nucleolin demarcatesmitotic stages and Alzheimer’s disease pathology. Neurobiol Aging2001; 22: 517–528.

42 Yang Y, Geldmacher DS, Herrup K. DNA replication precedesneuronal cell death in Alzheimer’s disease. J Neurosci 2001; 21:2661–2668.

43 Lam YA, Pickart CM, Alban A, Landon M, Jamieson C, Ramage Ret al. Inhibition of the ubiquitin-proteasome system in Alzheimer’sdisease. Proc Natl Acad Sci USA 2000; 97: 9902–9906.

44 Lindsten K, de Vrij FM, Verhoef LG, Fischer DF, van Leeuwen FW,Hol EM et al. Mutant ubiquitin found in neurodegenerativedisorders is a ubiquitin fusion degradation substrate that blocksproteasomal degradation. J Cell Biol 2002; 157: 417–427.

45 De Vrij FMS, Sluijs JA, Gregori L, Fischer DF, Hermens WTJMC,Goldgaber D et al. Mutant ubiquitin expressed in Alzheimer’sdisease causes neuronal death. FASEB J 2001; 15: 2680–2688.


796


Documents

Neuronal expression of GFAP in patients with Alzheimer pathology and identification of novel GFAP splice forms