Content uploaded by Jenny Harasty
Author content
All content in this area was uploaded by Jenny Harasty on Apr 07, 2018
Content may be subject to copyright.
N[URUSCHC[
ELSEVIER Neuroscience Letters 209(1996)29-32
LETTERS
Glial fibrillary acidic protein (GFAP) immunohistochemistry in human
cortex: a quantitative study using different antisera
Glenda M. Halliday ",c,*, Karen M. Cullen a, Jillian J. KriP, b, Antony J. Harding a, Jenny Harasty"
aNeuropathology Division, Department of Pathology, University of Sydney, Sydney, 2006, NSW, Australia
bDepartment of Pathology, Royal Prince Alfred Hospital, Camperdown, 2050, NSW, Australia
C Prince ~ Wales Medical Research Institute, Villa 2, Prince of Wales Hospital, High Street,
Randwick, 2031, NSW, Australia
Received 12 January 1996; revised version received 26 March 1996; accepted 27 March 1996
Abstract
Gliat fibrillary acidic protein (GFAP) is the principal marker for brain astrocytes. The present study aims to examine the variability
in GFAP immunohistochemistry in formalin-fixed human brain. Four commercially-available antisera were tested using standardised
protocols in the cerebral ce,rtex of three cases with prominent glial reactions and one control. GFAP immunoreactivity was largely con-
fined to the pial surface and white matter in control cortex, with the number of astrocytic cell bodies and processes as well as intensity
of staining markedly increased in damaged cortices. A dramatic difference in the pattern of GFAP staining using different antisera was
observed and may account for discrepancies between past studies. This variance has important practical implications for the interpreta-
tion of results using GFAP immunohistochemistry in human tissue.
Keywords:
Astrocytes; Cortex; Human; lmmunohistochemistry; Neuropathology; Quantitation
Astrocytes are involved in a number of vital homeo-
static mechanisms including maintenance of the blood-
brain barrier, extracellular buffering, microregulation of
cerebral blood flow and metabolism of glutamate and
ammonium [12,16]. In recent years the development of
antibodies to a number of glial-specific intracellular and
cell surface markers [6,29] has provided an important tool
for the study of astrocytes. The principal method for
identifying astrocytes is immunohistochemistry with an-
tisera raised against glialt fibrillary acidic protein (GFAP),
an intermediate filament present in the cell body and
processes of all astrocytes [6]. However, variations be-
tween studies have been described. Antisera from both
independent and commercial sources have been used to
describe the topography, morphology and reaction of as-
trocytes in normal and diseased conditions [2,8]. Factors
such as age [14], food restriction [20] and hormone levels
[10] have been shown to influence GFAP levels, with
upregulation in reactive astrocytes a highly specific phe-
nomenon [9,19]. It has been suggested that the two iso-
* Corresponding author. Tel.: +61 2 3822736; fax: +61 2 3822723.
forms of GFAP may be required in different circum-
stances (e.g. resting versus active cellular states [9]).
Factors influencing direct comparisons between different
studies need to be considered.
The present study aims to describe and quantify the
amount of GFAP visualised in human postmortem brain
tissue. Both control and pathological formalin-fixed hu-
man cortex was examined using several commercially
available antisera to GFAP. Three cases with a prominent
glial reaction were selected for study after examination of
haematoxylin and eosin stained sections from brain re-
gions affected. Routine brain autopsies were performed at
the Royal Prince Alfred Hospital and the New South
Wales (NSW) Institute of Forensic Medicine under the
NSW Transplantation and Anatomy Act. The brains were
removed and fixed by immersion in 15% buffered forma-
lin for 2 weeks. Case 1 had a cerebral abscess in the left
frontal lobe of approximately 7 days duration which had
ruptured to cause meningitis and ventriculitis; the abscess
wall and surrounding tissue were studied. Case 2 had a
history of infarction of part of the middle cerebral artery
territory five years prior to death; the right parietal lobe
0304-3940/96/$12.00 © 1996 Elsevier Science Ireland Ltd. All rights reserved
PII: S0304-3940(96) 1 2~<,92-X
30 G.M. Halliday et al. / Neuroscience Letters 209 (1996) 29-32
encompassing the infarction and its surrounds were stud-
ied. Case 3 had Alzheimer's disease with a 13 year his-
tory of progressive dementia and memory disturbance,
depression and loss of general living skills. Case 4
(control) was a 50 year old female with no history of neu-
rological disease and no neuropathological abnormalities.
From cases 3 and 4, 3 mm blocks of the temporal lobe,
including the hippocampus, and the superior frontal cor-
tex were embedded in paraffin or cut on a freezing mi-
crotome.
Four primary antisera were selected for comparison in
this study: (I) Sigma G3893 mouse anti-pig GFAP, (II)
Sigma G9269 rabbit anti-human GFAP, (III) DAKO
M761 mouse anti-human GFAP, and (IV) DAKO Z334
rabbit anti-cow GFAP. Antisera were selected so that (i)
both polyclonal antisera and monoclonal antibodies were
tested, (ii) both human and non-human (porcine and bo-
vine) GFAP were used as the antigen, and (iii) the antis-
era were in common use in several laboratories. Primary
antisera were diluted in 0.1 M Tris-HCl buffer (pH 7.4),
detected using routine avidin-biotin-peroxidase proce-
dures, and optimal concentrations determined to avoid
non-specific staining of tissue elements (Fig. 1A) as pre-
viously described [5]. Additional sections were treated to
confirm the specificity of the immunohistochemical reac-
tion. No peroxidase precipitation was observed if the pri-
mary antisera were omitted. In adjacent sections, a neu-
ronal marker (monoclonal anti-parvalbumin, Sigma
P3171, diluted 1:10 000) served as a positive control [25].
Two series of paraffin-embedded, 10/~m, slide-mounted
sections or frozen, 50/zm, free-floating sections were
immunohistochemically stained with GFAP antisera as
follows: section 1, antibody I; section 2, antiserum II;
section 3, antibody III; section 4, antiserum IV; section 5,
silver stain; section 6, antibody I and Cresyl violet; sec-
tion 7, antiserum II and Cresyl violet; section 8, antibody
III and Cresyl violet; section 9, antiserum IV and Cresyl
violet; section 10, haematoxylin and eosin. The areal
fraction of the sections occupied by GFAP-positive astro-
cytic cell bodies was determined using a point counting
technique [1]. A grid of 110 points within an eyepiece of
an Olympus BH-2 microscope was used at a final magni-
fication of 200x. In the cortex, 4-5 strips perpendicular
to the pial surface and spanning the entire cortical ribbon
were counted. In the white matter, 4-5 randomly placed
grids were counted. Repeated measures gave a 3.7% er-
ror.
Areas with significant pathology contained high
amounts of GFAP immunoreactivity (cases 1, 2 and 3),
while control material (case 4) contained less immunore-
activity (Fig. 1B) in a restricted anatomical pattern. In the
control case, antibody I revealed the least staining (areal
fraction of 0%, antiserum II areal fraction of 0.08%, anti-
body III areal fraction of 0.04%), while antiserum IV
demonstrated the greatest amount of staining (areal frac-
tion of 0.34%, at least four times more astrocytes visual-
A 0.51 f e•- l ~ ~"antib°dy~'t• I
o i~o looo ;o" io'
o~,
5 100 1000 104 1~
°tl O" !0oo ;,o,
'I'--'t ant,..=,V
o 100 1000 104 10 s
dilution
0
Casel CaseZ Case3 Case4
Fig. 1. (A) Titration curves for each of the four antisera determined on
both slide-mounted, paraffin-embedded and free-floating, frozen sec-
tions of
the cortex
from case 3. The amount of GFAP-immunoreactivity
in cortex (small circles) and white matter (small boxes) was quantified
and plotted against antiserum concentration (logarithmic scale). The
GFAP areal fraction was not significantly different for tissue processed
using different methods (t-test, P > 0.05). The maximal areal fraction of
GFAP-immunoreactivity in the absence of non-specific, or background,
staining was taken as optimal staining (indicated by the line). Sub-
optimal titre revealed few or no immunoreactive astrocytes. (B) Bar
chart of the areal fraction of cortical GFAP-immunoreactivity in the
four cases studied using each of the four different antisera in paraffin-
embedded sections.
ised than with the other antisera). Polyclonal antisera re-
vealed more astrocytic elements than did monoclonal
antibodies.
In contrast to the control tissue, the pathological tissue
sampled revealed large numbers of GFAP-immunore-
active astrocytes within the cortical mantle (Fig. 2B,E,H).
Despite this, no increase in the number of cortical astro-
cytes was seen in serial Nissl-stained sections in any case
(40_+5 in case 1, 37_+ 5 in case 2, 40_+4 in case 3,
40 _+ 4 in case 4, per 0.152 mm 2, n = 10 fields in each).
Reactive astrocytes were most obvious in the parenchyma
surrounding the areas of pathology (Fig. 2). All cases had
a normal distribution of subpial astrocytes (Fig. 2A,D,G).
Antisera II, III and IV all revealed similar areal fractions
of GFAP-immunoreactivity in cases 1 and 2 (the infarct
G.M. Halliday et al. / Neuroscience Letters 209 (1996) 29-32 31
Fig. 2. Photomicrographs of paraffin-embedded pathological tissue
using antiserum IV from the subpial (A,D,G), parenchymal (B,E,H) and
white matter (C,F,I) regions. Scale bar in (1) is equivalent for all pho-
tomicrographs. (A-C) is from case 1 with a cerebral abscess, (D-F) is
from ease 2 with an old cerebral infarct and case 3 is from a patient
with Alzheimer's disease. Significant upregulation of GFAP staining is
seen in the parenchyma ancl white matter in each patient. Subpial
staining is similar to that seen in the control. Upregulation was most
obvious in areas surrounding the site of damage, (B) adjacent to the
abscess, (E) in the cortex for some distance around the infarct site, and
(H) predominantly associated with plaque formation in Alzheimer's
disease (layers IIl and V).
and abscess, respectively; Fig. 1B), whereas antiserum IV
was superior in both the control (case 4, Fig. IB) and the
case with Alzheimer's disease (case 3, Fig. 1B).
We have demonstrated that application of a consistent
and uniform immunohistochemical technique reveals dif-
ferent numbers and patterns of astrocytes depending on
the antiserum used, i.e. not all commercially available
GFAP antisera detect reactive changes to the same de-
gree. The variation between antisera means that consider-
able variation in results is likely to occur between labora-
tories. These findings have practical implications for the
interpretation and reporting of results using GFAP stain-
ing for the identification of astrocytes. GFAP-negative
cells cannot be assumed to be non-astrocytic.
It has been suggested that degradation of GFAP occurs
within 24 h of death [6], although this is not supported by
the findings of this and other studies [11]. While postmor-
tem changes to GFAP may occur, the variability between
antisera was constant despite various postmortem delays
before fixation. In fact, GFAP staining was still promi-
nent after a postmortem delay of 72 h (see Fig. 2D-F). As
our fixation method is routine for postmortem human
tissue, the results will be generally applicable to the ma-
jority of studies in humans. Thus, differences in postmor-
tem delay, fixation [18], or tissue processing cannot ac-
count for the demonstrated variability in GFAP staining.
Both frozen and paraffin sections were directly compared
using sequential blocks. Our protocol is a standardised
technique used by many authors with the detection
method used in our laboratory previously tested and op-
timised [5]. The differences between antisera were consis-
tent within these protocols. Although there are numerous
techniques for revealing antigenic sites and enhancing the
visualisation of the reaction product, our results show that
the staining pattern achieved is dependent on the primary
antisera. Further optimisation of the immunohistochemi-
cal method is unlikely to eliminate these differences.
Four different commercially available antisera were
used in this study. The greatest differences in staining
were seen between the monoclonal and polyclonal antis-
era. This finding is to be expected as polyclonal antisera
recognise a greater number of epitopes. The antisera cho-
sen were developed in various host species against GFAP
purified from different tissue sources. While GFAP ap-
pears to be highly conserved and consistently expressed
in astrocytes [6,22], the detection of different binding
sites by specific antibodies may contribute to discrepan-
cies in staining patterns. Surprisingly, the monoclonal
antibody developed against human brain (antibody III)
was not the most sensitive for GFAP immunohistochemis-
try in formalin-fixed human brain. Differences between
polyclonal antisera and monoclonal antibodies have been
noted previously, with polyclonal antisera revealing more
astrocytes [17,30]. Such differences can be marked, with
ischaemic lesions in the rat showing increased staining
with a polyclonal antiserum and a loss of staining with a
monoclonal antibody [24]. Many laboratories use antisera
produced 'in house' which, despite the validation of
specificity using immunoblotting, may produce different
staining patterns in situ. It would therefore seem prudent
to use polyclonal antisera if maximal detection of GFAP
is required, although it should be noted that each poly-
clonal antiserum will be unique and may have a unique
staining pattern. We do not mean to recommend a particu-
lar antiserum for use in human brain, but highlight a vari-
able which has not been emphasised before, i.e. that under
identical conditions, different commercially available
GFAP antisera stain different numbers of astrocytes in
routinely fixed human brain.
Some of the staining differences observed between an-
tisera may be due to the detection of different GFAP iso-
forms [9]. Theoretically, GFAP should be contained in all
astrocytes [6], yet in practice some cells may not contain
enough GFAP of a particular isoform to be detected im-
munohistochemically by light microscopy. For instance,
with antibody IV, typical fibrous astrocytes in large fibre
tracts were readily detected, as were astrocytes along the
pial surface. This distribution is more extensive than that
commonly described in control human cortex [3,7,15,23,
28] but similar to that described in rat cortex [13,24,27].
In rats, GFAP reaction product is not always seen in as-
32
G.M. Halliday et al. / Neuroscience Letters 209 (1996) 29-32
trocytic cell bodies, but only within their fine processes
[4] where mRNA is found [26]. Thus, it cannot be as-
sumed that a lack of detectable staining necessarily means
an absence of GFAP within the cell.
The present study shows that only some antisera detect
all reactive astrocytes in pathological material. A recent
quantitative study showed that the total number of astro-
cytes (both GFAP-positive and -negative) did not change
with AIDS infection although the proportion of reactive
GFAP-positive astrocytes markedly increased [28].
Whether an increase in GFAP-immunoreactivity is in-
dicative of the cytoplasmic hypertrophy which defines a
reactive astrocyte [21], or of changing GFAP isoforms (as
recently proposed [9]), or even of a change in the anti-
genic profile of the protein, remains to be established.
The concept of astrocytic reactivity may need to be more
clearly defined.
This project was funded by the National Health and
Medical Research Council of Australia. We thank Prof
Clive Harper and Dr Roger Pamphlett, Department of
Pathology (Neuropathology Division), University of Syd-
ney and Royal Prince Alfred Hospital, Camperdown for
the diagnostic evaluation of the cases presented. We are
grateful to the staff of the histopathology laboratory who
prepared all the diagnostic material for analysis and, in
particular, Lyndall Baum who cut the paraffin sections
used in this study.
[1] Aherne, W.A. and Dunnhill, M.S., Morphometry, Amold, 1982.
[2] Baumann, N., Baron-Van Evercooren, A., Jacque, C. and Zalc, B.,
Glial biology and disorders, Curr. Opin. Neurol. Neurosurg., 6
(1993) 27-33.
[3] Beach, T.G., Walker, R. and McGeer, E.G., Patterns of gliosis in
Alzheimer's disease and aging cerebrum, Glia, 2 (1989) 420-436.
[4] Boyes, B.E., Kim, S.U., Lee, V. and Sung, S.C., Immunohisto-
chemical co-localization of S-100fl and the glial fibrillary acidic
protein in rat brain, Neuroscience, 17 (1986) 857-865.
[5] Ellis, J. and Halliday, G., A comparative study of avidin-biotin-
peroxidase complexes for immunohistochemical detection of an-
tigen in neural tissue, Biotech. Histochem., 67 (1992) 367-371.
[6] Eng, L.F., Glial fibrillary acidic protein (GFAP): the major pro-
tein of glial intermediate filaments in differentiated astrocytes, J.
Neuroimmunol., 8 (1985) 203-214.
[7] Esiri, M.M. and Morris, C.S., Immunocytochemical study of
macrophages and microglial cells and extracelhilar matrix com-
ponents in human CNS disease, 2: non-neoplastic diseases, J.
Neurol. Sci., 101 (1991) 59-72.
[8] Frederickson, R.C., Astroglia in Alzheimer's disease, Neurobiol.
Aging, 13 (1992) 239-253.
[9] Galea, E., Dupouey, P. and Feinstein, D.L., Glial fibrillary acidic
protein mRNA isotypes: expression in vitro and in vivo, J. Neuro-
sci. Res., 41 (1995) 452-461.
[10] Garcia-Segura, L.M., Luquin, S., Parrducz, A. and Naftolin, F.,
Gonadal hormone regulation of glial fribrillary acid protein im-
munoreactivity and glial ultrastructure in the rat neuroendocrine
hypothalamus, Glia, 10 (1994) 59--69.
[1 I] Harpin, M.L., Dela~re, P., Javoy-Avid, F., Bock, E., Jacque, C.,
Delpech, B., Villarroya, H., Duykaerts, C., Hauw, J.J. and Bau-
man, N., Glial fibrillary acidic protein and flA4 protein deposits in
temporal lobe of aging brain and senile dementia of the AIz-
heimer-type: relation with the cognitive state and with quantita-
tive studies of senile plaques and neurofibrillary tangles, J. Neu-
rosci. Res., 27 (1990) 587-594.
[12] Holash, J.A., Noden, D.M. and Stewart, P.A., Re-evaluating the
role of astrocytes in blood-brain barrier induction, Dev. Dyn., 197
(1993) 14-25.
[13] Kelley, M.S. and Steward, O., The role of neuronal activity in
upregulating GFAP mRNA levels after electrolytic lesions of the
entorhinal cortex, Int. J. Dev. Neurosci., 11 (1993) 105-115.
[14] Kohama, S.G., Goss, J.R., Finch, C.E. and McNeill, T.H., In-
creases of glial fibrillary acidic protein in the aging female mouse
brain, Neurobiol. Aging, 16 (1995) 59--67.
[15] Mancardi, G.L., Linwnicz, B.H. and Mandybur, T.I., Fibrous
astrocytes in Alzheimer's disease and senile dementia of the Alz-
heimer's type, Acta Neuropathol., 61 (1983) 76-80.
[16] Martin, D.L., Synthesis and release of neuroactive substances by
glial cells, Glia, 5 (1992) 81-94.
[17] McLendon, R.E., Burger, P.C., Pegram, C.N., Eng, L.F. and
Bigner, D.D., The immunohistochemical application of three anti-
GFAP monoclonal antibodies to formalin-fixed, paraffin-
embedded, normal and neoplastic brain tissues, J. Neuropathol.
Exp. Neurol., 45 (1986) 692-703.
[18] Mitchell, J., Best, N., Sundstrom, L.E. and Wheal, H.V., The use
of sodium sulphide-fixed brain tissue for immunocytochemical
staining of activated microglia and reactive astrocytes, Histo-
chemistry, 99 (1993) 91-94.
[19] Neary, J.T., Whittemore, S.R., Zhu, Q. and Norenberg, M.D.,
Destabilization of glial fibrillary acidic protein mRNA in astro-
cytes by ammonia and protection by extracellular ATP, J. Neuro-
chem., 63 (1994) 2021-2027.
[20] Nichols, N.R., Finch, C.E. and Nelson, J.F., Food restriction de-
lays the age-related increase in GFAP mRNA in rat hypothala-
mus., Neurobiol. Aging, 16 (1995) 105-110.
[21] Norenberg, M.D., Astrocyte responses to CNS injury, J. Neuropa-
thol. Exp. Neurol., 53 (1994) 213-220.
[22] Onteniente, B., Kimura, H. and Maeda, T., Comparative study of
the glial fibrillary acidic protein in vertebrates by PAP immuno-
histochemistry, J. Comp. Neurol., 215 (1983) 427-436.
[23] Schecter, R., Yen, S.-H.C. and Terry, R.D., Fibrous astrocytes in
senile dementia of the Alzheimer type, J. Neuropathol. Exp. Neu-
rol., 40 (1980) 95-101.
[24] Schmidt-Kastner, R., Wietasch, K., Weigel, H. and Eysel, U.T.,
Immunohistochemical staining for glial fibrillary acidic protein
(GFAP) after deafferentation or ischemic infarction in rat visual
system: features of reactive and damaged astrocytes, Int. J. Dev.
Neurosci., 11 (1993) 157-174.
[25] Sloviter, R.S., Sollas, A.L., Barbaro, N.M. and Laxer, K.D., Cal-
cium-binding protein (calbindin-D28K) and parvalbumin immu-
nocytochemistry in the normal and epileptic human hippocampus,
J. Comp. Neurol., 308 (1991) 381-396.
[26] Steward, O. and Banker, G.A., Getting the message from the gene
to the synapse: sorting and intracellular transport of RNA in neu-
rons, Trends Neurosci., 15 (1992) 180-186.
[27] Takamiya, Y., Kohsaka, S., Toya, S., Otani, M. and Tsukada, Y.,
Immunohistochemical studies on the proliferation of reactive as-
trocytes and the expression of cytoskeletal proteins following
brain injury in rats, Dev. Brain Res., 38 (1988) 201-210.
[28] Weis, S., Huag, H. and Budka, H., Astroglial changes in the cere-
bral cortex of AIDS brains: a morphometric and immunohisto-
chemical investigation, Neuropathol. Appl. Neurobiol., 19 (1993)
329-355.
[29] Wilkin, G.P., Marriott, D.R. and Cholewinski, A.J., Astrocyte
heterogeneity, Trends Neurosci., 13 (1990) 43-45.
[30] Wittchow, R. and Landas, S.K., Glial fibrillary acidic protein
expression in pleomorphic adenoma, chordoma, and astrocytoma,
Arch. Pathol. Lab. Med., 115 (1991) 1030-1033.