Redefining the concept of reactive astrocytes as
cells that remain within their unique domains
upon reaction to injury
Ulrika Wilhelmsson*, Eric A. Bushong†, Diana L. Price†, Benjamin L. Smarr†, Van Phung†, Masako Terada†,
Mark H. Ellisman†, and Milos Pekny*‡
*Department of Clinical Neuroscience and Rehabilitation, Institute of Neuroscience and Physiology, Sahlgrenska Academy, Go ¨teborg University, SE-405 30
Go ¨teborg, Sweden; and†National Center for Microscopy and Imaging Research, University of California at San Diego, La Jolla, CA 92093-0608
Edited by Pasko Rakic, Yale University School of Medicine, New Haven, CT, and approved September 15, 2006 (received for review April 7, 2006)
Reactive astrocytes in neurotrauma, stroke, or neurodegeneration
are thought to undergo cellular hypertrophy, based on their
morphological appearance revealed by immunohistochemical de-
tection of glial fibrillary acidic protein, vimentin, or nestin, all of
them forming intermediate filaments, a part of the cytoskeleton.
Here, we used a recently established dye-filling method to reveal
the full three-dimensional shape of astrocytes assessing the mor-
phology of reactive astrocytes in two neurotrauma models. Both in
the denervated hippocampal region and the lesioned cerebral
cortex, reactive astrocytes increased the thickness of their main
cellular processes but did not extend to occupy a greater volume
of tissue than nonreactive astrocytes. Despite this hypertrophy of
glial fibrillary acidic protein-containing cellular processes, interdig-
itation between adjacent hippocampal astrocytes remained mini-
mal. This work helps to redefine the century-old concept of
hypertrophy of reactive astrocytes.
astrocyte domains ? astrocyte hypertrophy
‘‘glia’’), astrocytes are receiving ever-increasing attention. The
rising recognition of their importance builds on knowledge of their
role in maintaining CNS homeostasis, providing nutrition for
were shown to control the number and function of neuronal
synapses (1, 2) and blood flow in the brain (3, 4).
Astrocytes exhibit an intricate bushy or spongiform morphology,
and their very fine processes are in close contact with synapses and
other components of brain parenchyma (5–8). These fine terminal
processes appear postnatally in the final stage of astrocyte matu-
ration. The subsequent elaboration of spongiform processes results
in the development of boundaries between neighboring astrocyte
domains, thus establishing exclusive territories for individual astro-
cytes, a phenomenon termed ‘‘tiling’’ (7, 9).
With earlier methods, including immunohistochemical detection
of astrocyte markers and impregnation techniques, the extent of
overlap between astrocyte territories was not amenable to investi-
gation. However, more recent staining methods, optical imaging
techniques, and 3D reconstruction paradigms using dye-filled as-
trocytes in semifixed tissue allowed the assessment of the bound-
aries of protoplasmic astrocyte territories in the CA1 area of the
uninjured rat hippocampus. Neighboring astrocytes invariably
touched each other but showed little interdigitation, basically tiling
to form unique domains (8).
are a hallmark of many neuropathologies. Reactive astrocytes are
characterized by high-level expression of glial fibrillary acidic
protein (GFAP), an intermediate filament protein, and by up-
regulation of intermediate filaments in the cytoplasm. Antibodies
against GFAP, the most frequently used astrocyte marker (10),
reveal the cytoskeletal structure but not the true cellular morphol-
ogy. Reactive astrocytes exhibit striking increases in GFAP immu-
nce considered to be merely a cellular layer filling the inter-
neuronal space and gluing neurons together (hence the term
noreactivity and in the number and length of GFAP-positive
processes. These findings have been interpreted as cellular hyper-
trophy (Fig. 1) (11–17), which might be expected to increase the
extent of interdigitation between astrocytes, normally in close
contact in the absence of brain injury.
tissue reached by individual cells and the interdigitation between
neighboring cells, we used a cell injection and 3D reconstruction
technique (7, 8, 18, 19) to assess the morphology of reactive and
nonreactive hippocampal and cortical astrocytes in mice. Contrary
to the widely accepted notion of hypertrophy of reactive astrocytes,
our findings suggest that these cells remain within their unique
domains while at the same time increasing the thickness of their
main cellular processes.
To assess the morphology of reactive astrocytes, we first subjected
adult mice to unilateral entorhinal cortex lesion. This injury partly
interrupts the perforant pathway, which innervates the molecular
layer in the ipsilateral dentate gyrus of the hippocampus (Fig. 1).
Axonal degeneration triggers astrocyte activation, ?60% synaptic
loss, synapse remodeling in the molecular layer, and neurogenesis
in the dentate gyrus that is not directly affected by the trauma
(19–21). Astrocytes were visualized with antibodies against GFAP
4 days after the insult (Fig. 1), when astrocyte activation and
hypertrophy of cellular processes are maximal, as shown by GFAP
the uninjured contralateral dentate gyrus had slender GFAP-
positive cellular processes resembling those in noninjured mice. On
the lesioned side, astrocyte reactivity was prominent in the outer
and middle molecular layers of the dentate gyrus, the area partly
denervated by the lesion. Reactive astrocytes appeared to have
more main cellular processes containing GFAP intermediate fila-
ments and more intermediate filament bundles than nonreactive
astrocytes (Fig. 1).
Next, individual astrocytes in the middle and outer molecular
layers of the dentate gyrus on the lesioned and contralateral side
were loaded with Lucifer yellow or Alexa Fluor 568. Both reactive
and nonreactive astrocytes had a bushy morphology with many fine
terminal processes protruding from the main cellular processes
(Fig. 2A), as did astrocytes in noninjured mice (data not shown).
M.H.E., and M.P. designed research; U.W., E.A.B., D.L.P., B.L.S., V.P., and M.T. performed
The authors declare no conflict of interest.
This article is a PNAS direct submission.
Abbreviation: GFAP, glial fibrillary acidic protein.
‡To whom correspondence should be addressed at: Institute of Neuroscience and Physiol-
ogy, Department of Clinical Neuroscience and Rehabilitation, Go ¨teborg University, Box
440, SE-405 30 Go ¨teborg, Sweden. E-mail: firstname.lastname@example.org.
© 2006 by The National Academy of Sciences of the USA
November 14, 2006 ?
vol. 103 ?
no. 46 ?
Individual reactive and nonreactive astrocytes varied considerably
in shape and appearance. Many astrocytes had one or several
cellular processes terminating in glial end-feet surrounding blood
The main cellular processes appeared to be thicker in reactive
than in nonreactive astrocytes (Fig. 2A). In addition, we found
that the number of primary processes leaving the soma was
increased in reactive compared with nonreactive astrocytes
layer of the dentate gyrus on the injured side were reactive and all showed greater GFAP immunoreactivity (Center) than the nonreactive astrocytes on the
contralateral side (Left). The square in Right denotes the area corresponding to the images in Left and Center. EC, entorhinal cortex. (Scale bar, 25 ?m.)
Entorhinal cortex lesion triggers reactive gliosis in the hippocampus. Unilateral entorhinal cortex lesion triggers astrocyte activation in the outer and
(B and C) Quantification of main cellular processes leaving the soma (B) and processes visible 25 ?m from the cell soma (C). Thick processes were more numerous
in reactive astrocytes than in nonreactive astrocytes. Three-dimensional reconstruction (E) shows that reactive and nonreactive astrocytes access similar volumes
of tissue (D; unit y axis 103?m3). Error bars represent SEM.
Morphological assessment of reactive and nonreactive astrocytes in the hippocampus. (A) Maximum projections of dye-filled reactive and nonreactive
www.pnas.org?cgi?doi?10.1073?pnas.0602841103Wilhelmsson et al.
(6.1 ? 0.2 versus 5.2 ? 0.2; P ? 0.005; n ? 47 and 44 cells,
respectively) (Fig. 2B). On maximum projection images of
individual cells, reactive astrocytes also had more main processes
extending at least 25 ?m from the cell soma than nonreactive
astrocytes (3.9 ? 0.3 versus 2.1 ? 0.3; P ? 0.0001; n ? 47 and
42 cells, respectively) (Fig. 2C).
To determine whether hypertrophy of cellular processes
affects the volume of tissue reached by reactive astrocytes, we
compared the volume accessed by reactive astrocytes on the
injured side with that accessed by nonreactive astrocytes on
the uninjured side (Fig. 2 D and E). Single nonreactive and
reactive astrocytes accessed similar volumes of tissue (43.4 ?
1.4 versus 44.2 ? 1.5 103?m3; P ? 0.71; range of 26–69 and
30–70 103?m3; n ? 43 and 43 cells, respectively; with ? 0.05
and ? 0.2, a difference of 10% or higher would be detected)
(Fig. 2D). Thus, the increased thickness of cellular processes
shown in A. An asterisk indicates the necrotic area; the dotted line indicates the injury border. CC, corpus callosum. (Scale bar, 100 ?m.)
in layer I of cerebral cortex 4 days after cortical lesioning. Dye-filling reveals fine spongiform processes in reactive astrocytes comparable to those of nonreactive
astrocytes. Three-dimensional reconstruction shows that reactive and nonreactive astrocytes access similar volumes of tissue (D; unit y axis 103?m3). Error bars
Morphological assessment of reactive and nonreactive cortical astrocytes. (A) Maximum projections of dye-filled reactive and nonreactive astrocytes
Wilhelmsson et al.
November 14, 2006 ?
vol. 103 ?
no. 46 ?
does not alter the action radius of reactive astrocytes in the
We next investigated whether the lack of cellular hypertrophy
seen in reactive astrocytes in the deafferented dentate gyrus was
a response shared by astrocytes outside the hippocampus. We
assessed the morphological appearance of reactive astrocytes in
the cerebral cortex after electrically induced lesioning (24). This
injury paradigm induces extensive neuronal death and strong
astrocyte activation (Fig. 3C). Four days after the lesion, we
studied the morphology of reactive astrocytes in cortical layer I
by loading astrocytes 200–800 ?m from the injury border with
Lucifer yellow. Reactive astrocytes showed strongly GFAP-
positive cellular processes in the region surrounding the necrotic
area (Fig. 3A), whereas the same region in uninjured mice
showed only weakly GFAP-positive astrocytes (Fig. 3B). Exam-
ination of reactive astrocytes filled with Lucifer yellow revealed
a bushy morphology with many fine terminal processes protrud-
ing from the main cellular processes, similar to astrocytes filled
in uninjured mice. The main cellular processes appeared to be
thicker in reactive than in nonreactive astrocytes (Fig. 4A). The
number of primary processes leaving the soma was increased in
reactive compared with nonreactive astrocytes (6.2 ? 0.2 versus
5.1 ? 0.3; P ? 0.005; n ? 25 and 18 cells, respectively) (Fig. 4B).
On maximum projection images of individual cells, reactive
astrocytes had more main processes extending at least 15 ?m
from the cell soma than nonreactive astrocytes (6.6 ? 0.6 versus
3.4 ? 0.3; P ? 0.001; n ? 24 and 18 cells, respectively) (Fig. 4C).
As demonstrated above for hippocampal astrocytes, single non-
reactive and reactive cortical astrocytes accessed similar volumes
of tissue (26.1 ? 2.9 versus 22.7 ? 2.8 103?m3; P ? 0.41; range
? 0.05 and ? 0.2, a difference of 40% or higher would be
detected) (Fig. 4D).
Thus, the increased thickness of astrocyte processes did not
of the hippocampus or in the electrically lesioned cortex.
In the adult mammalian brain, the access domain of individual
hippocampal astrocytes shows minimal overlap and interdigita-
tion of fine cellular processes between adjacent astrocytes (8,
25). To compare the extent of interdigitation of reactive and
nonreactive astrocytes, we injected Lucifer yellow and Alexa
Fluor 568 into adjacent astrocytes in the dentate gyrus after
unilateral entorhinal cortex lesion and evaluated the overlap
between astrocyte territories on optical sections (Fig. 5). Inter-
digitation was also assessed on 3D images of dye-filled neigh-
boring astrocytes, and the overlap was highlighted in a pseudo-
color (Fig. 6A and Movies 1 and 2, which are published as
supporting information on the PNAS web site). Quantification
of the overlapping domains on series of optical sections showed
minimal interdigitation between reactive astrocytes and compa-
rable with that between nonreactive astrocytes (4.5 ? 0.8 versus
3.1 ? 0.6%; ranging from 0.06 to 12.9% for individual astrocytes;
n ? 20 and 18 cells, respectively). Thus, the hypertrophy of
cellular processes upon astrocyte activation in the hippocampus
did not affect the extent of overlap between the unique domains
of individual astrocytes.
This study shows that astrocytes in the denervated dentate gyrus
and lesioned cerebral cortex remain within their domains and
that the overlap between individual astrocyte domains in the
denervated dentate gyrus remains minimal. Although the main
cellular processes of reactive astrocytes containing GFAP inter-
mediate filaments become thicker, the overall size of these cells
was similar to that of nonreactive astrocytes. When visualized by
antibodies against intermediate filament protein GFAP, these
thicker processes appear longer, because they can be followed
over greater distances. However, the hypertrophy of cellular
processes did not affect the volume of tissue accessed by
individual astrocytes via their fine spongiform processes (Fig.
6B) and thus did not result in cellular hypertrophy.
Previously, the volume of tissue accessed by dye-filled non-
reactive astrocytes was measured in rat and mouse CA1 stratum
radiatum of the hippocampus (8, 25). Our findings suggest that
both reactive and nonreactive astrocytes in the molecular layer
of the dentate gyrus of the mouse hippocampus, and even more
so astrocytes in the cortical layer I, access a smaller volume of
tissue than astrocytes in stratum radiatum. Overlap of individual
astrocyte domains of nonreactive astrocytes in the stratum
neighboring cells (Lucifer yellow and Alexa Fluor 568). Maximum projections
of astrocytes in the molecular layer of the dentate gyrus on the lesioned and
contralateral sides show adjacent astrocyte territories (domains) on three
optical sections 4 ?m apart. Insets show territories with overlapping areas in
yellow. The extent of interdigitation between neighboring astrocytes, both
reactive and nonreactive, was limited and most prominent around blood
vessels (arrowheads). (Scale bar 25 ?m.)
Overlap between astrocyte territories assessed by dye-filling of
www.pnas.org?cgi?doi?10.1073?pnas.0602841103Wilhelmsson et al.
radiatum was very limited. In the molecular layer of the dentate
gyrus, we found a comparably small extent of interdigitation
between processes of neighboring astrocytes. Most importantly,
this territorial overlap was minimal in astrocytes reacting to an
Quantification of astrocyte cell density in rat cerebral cortex
(26) and calculations of average volume of cortical astrocytes
based on the length of GFAP-positive processes (27) indicated
a substantial overlap between neighboring astrocyte domains in
the rat cortex (28). Our experimental data on cortical astrocytes
reacting to electrically induced lesions showed that the volume
of tissue accessed by reactive and nonreactive astrocytes was
reactive astrocytes is comparable with that of nonreactive cor-
tical astrocytes. The extent to which this applies to other injuries
in the brain or in the spinal cord remains to be established,
because the morphological and functional responses of astro-
cytes may depend on the type of insult.
In the neonatal CNS, the main processes of neighboring imma-
ture astrocytes interdigitate extensively; gradually these cells de-
with only a limited overlap (7). Given their similarities with
immature astrocytes, reactive astrocytes might be expected to form
more extensive interdigitations between the domains of individual
astrocytes. However, our data indicate that this is not the case. The
limited overlap between both reactive and nonreactive astrocytes
tissue evenly and efficiently. The best way to achieve this may be
through contact spacing and pruning of interdigitating processes to
establish exclusive astrocyte territories.
Sufficiently stable cellular territories might be necessary for
the morphological and functional connection within the astro-
cyte syncytium and in situations such as axonal degeneration or
a direct neurotrauma can be essential for the astrocyte network
to carry out functions such as communication across gap junc-
tions or spacial buffering.
In conclusion, our findings suggest that the term ‘‘cellular
hypertrophy’’ frequently used to describe morphological changes
in reactive astrocytes may be misleading. Instead, reactive as-
trocytes show hypertrophy of their intermediate filament-rich
main cellular processes but seem to remain within their unique
Surgical Procedures. Unilateral entorhinal cortex lesioning was
performed as described (29) in six 5-mo-old females. Electrically
induced injury of the cerebral cortex was performed as described
(24) in six 10-mo-old female mice. All mice were on a mixed
genetic background (C57BL?6, 129Sv, 129Ola) and were main-
tained in a barrier animal facility.
Anesthetized mice were placed in a stereotactic frame, and a
hole was drilled through the skull. For entorhinal cortex lesion,
a retractable wire knife (Kopf Instruments, Tujunga, CA) was
lowered 1 mm down from the dura ?3.6 mm laterally and ?0.2
mm posterior to lambda. The wire knife was expanded 2 mm
horizontally and then lowered 2 mm twice at ?30° and ?135° to
avoid the hippocampal formation. For electrically induced lesion
of the cerebral cortex, a fine-needle electrode was inserted
through the skull 2.25 mm laterally at the level of bregma and
lowered 1.0 mm (measured from the meningeal level) into the
cortex of the right hemisphere. A second electrode was attached
to the root of the tail. By using Lesion Maker (Ugo Basile,
Comerio, Italy), a direct current of 5 mA was applied for 10 sec.
The mice were kept in heated cages until they recovered from
Dye-Filling of Astrocytes. Cells in fixed tissue were filled with dye
after lesioning, the mice were deeply anesthetized with Nem-
butal (10 mg?100 g of body weight) and transcardially perfused
with oxygenated Ringer’s solution (37°C) (0.79% NaCl?0.038%
KCl?0.02% MgCl2?6H2O?0.018% Na2HPO4?0.125% NaHCO3?
0.03% CaCl2?2H2O?0.2% dextrose?0.02% xylocaine) and then
with 4% paraformaldehyde in PBS (pH 7.4, 37°C) for 10 min.
The brain was extracted and postfixed in ice-cold 4% parafor-
maldehyde for 1 h and cut with a vibratome into 75- to
100-?m-thick slices. The slices were stored in PBS at 4°C until
The slices were placed under a ?60 water objective (N.A. of
1.4) and observed with an Olympus (Center Valley, PA)
BX50WI microscope with infrared differential interference con-
trast optics. Astrocytes were identified by the shape and size of
their somata. Glass micropipettes (o.d., 1.00 mm; i.d., 0.58 mm;
resistance, 100–400 M?) were pulled on a vertical puller (Kopf
Instruments) and backfilled with 5% aqueous Lucifer yellow
(Sigma-Aldrich, St. Louis, MO) or 10 mM Alexa Fluor 568
(Molecular Probes, Eugene, OR) in 200 mM KCl. Astrocytes in
the outer and middle molecular layer of the dentate gyrus of the
hippocampus or in layer I of cerebral cortex were impaled and
iontophoretically injected with the respective dye by using 1-sec
pulses of negative current (0.5 Hz) for 1–2 min. After several
cells were filled, the slices were placed in ice-cold 4% parafor-
maldehyde overnight and then coverslipped in Gelvatol (32).
Immunohistochemistry. The slices were placed in ice-cold 4%
paraformaldehyde overnight at 4°C. The next day, the slices were
repeatedly washed in PBS and permeabilized for 1 h at room
temperature in PBS containing 1% BSA, 0.25% Triton X-100,
and 3% normal donkey serum, followed by incubation with
guinea pig antibodies against GFAP (Sigma-Aldrich; 1:100) for
48 h at 4°C in PBS containing 1% BSA, 0.1% Triton X-100, and
0.3% normal donkey serum. After several washes in PBS, the
the dentate gyrus. The yellow zone shows the border area where cellular
processes of two adjacent astrocytes interdigitate. (B) Reactive astrocytes stay
within their domains, but their main cellular processes get thicker, making
them visible over a greater distance (illustrated here by the circles).
The domains of nonreactive and reactive astrocytes—a concept. (A)
Wilhelmsson et al.
November 14, 2006 ?
vol. 103 ?
no. 46 ?
slices were incubated overnight at 4°C with donkey anti-guinea
pig antibodies conjugated with RRX (1:300; Jackson Immu-
noResearch, West Grove, PA) and mounted in Gelvatol.
Image Acquisition and Analysis. Confocal z-series were acquired
(Bio-Rad, Hercules, CA) attached to a Nikon (Kanagawa,
Japan) E600FN microscope and an Olympus FluoView 1000
laser-scanning confocal microscope system attached to an Olym-
pus IX81 microscope, both equipped with a ?60 oil-immersion
objective (Plan Apo N.A. of 1.4 and 1.42, respectively). Images
were visualized and analyzed with Imaris 4.0.4 (Bitplane, Zurich,
Switzerland) and ImageJ (National Institutes of Health, Be-
thesda, MD). The number of cellular processes leaving the soma
was assessed on the stack of optical sections from individual
astrocytes. Main cellular processes extending more than 25 or 15
?m from the soma (see Fig. 2C), for hippocampal or cortical
astrocytes, respectively, were counted on maximum projections
of dye-filled astrocytes. The volume of tissue reached by dye-
filled astrocytes was measured on 3D reconstructions of the
astrocytes by thresholding the images and then creating a volume
of the thresholded voxels (see Fig. 2E). Interdigitation between
adjacent hippocampal astrocytes was evaluated on optical sec-
tions through adjacent astrocytes filled with different dyes
(Lucifer yellow and Alexa Fluor 568). For quantification of
overlapping astrocyte domains, each astrocyte territory was
manually delineated and pseudocolored, the areas of overlap-
ping territories on optical sections were measured and expressed
as a percentage of the total astrocyte area. Interdigitation was
also visualized on 3D reconstructed images of dye-filled astro-
indicating areas with neighboring astrocyte processes in close
proximity, were then selected and highlighted on the original 3D
U.W. was supported by Swedish Medical Society Grant 16850, Wilhelm
och Martina Lundgrens stiftelsen, and Hja ¨rnfonden and Åhle ´n-
stiftelsen; M.P. was supported by Swedish Research Council Grant
11548, the Swedish Stroke Association, ALF Go ¨teborg, Hja ¨rnfonden,
Trygg-Hansa, and Torsten och Ragnar So ¨derbergs stiftelser; and M.H.E.
was supported by National Institutes of Health Grants RR004050,
NS046068, NS014718, and CA4084314.
1. Ullian EM, Sapperstein SK, Christopherson KS, Barres BA (2001) Science
2. Christopherson KS, Ullian EM, Stokes CC, Mullowney CE, Hell JW, Agah A,
Lawler J, Mosher DF, Bornstein P, Barres BA (2005) Cell 120:421–433.
3. Mulligan SJ, MacVicar BA (2004) Nature 431:195–199.
4. Zonta M, Angulo MC, Gobbo S, Rosengarten B, Hossmann KA, Pozzan T,
Carmignoto G (2003) Nat Neurosci 6:43–50.
5. Grosche J, Matyash V, Moller T, Verkhratsky A, Reichenbach A, Kettenmann
H (1999) Nat Neurosci 2:139–143.
6. Kosaka T, Hama K (1986) J Comp Neurol 249:242–260.
7. Bushong EA, Martone ME, Ellisman MH (2004) Int J Dev Neurosci 22:73–86.
8. Bushong EA, Martone ME, Jones YZ, Ellisman MH (2002) J Neurosci
9. Haber M, Murai K (2006) Neuron Glia Biol 2:59–66.
10. Eng LF, Ghirnikar RS, Lee YL (2000) Neurochem Res 25:1439–1451.
11. Kimelberg HK, Norenberg MD (1989) Sci Am 260:66–72, 74, 76.
12. Eddleston M, Mucke L (1993) Neuroscience 54:15–36.
13. Ridet JL, Malhotra SK, Privat A, Gage FH (1997) Trends Neurosci 20:570–577.
14. Fawcett JW, Asher RA (1999) Brain Res Bull 49:377–391.
15. McGraw J, Hiebert GW, Steeves JD (2001) J Neurosci Res 63:109–115.
16. Silver J, Miller JH (2004) Nat Rev Neurosci 5:146–156.
17. Sofroniew MV (2005) Neuroscientist 11:400–407.
18. Bushong EA, Martone ME, Ellisman MH (2003) J Comp Neurol 462:241–251.
19. Wilhelmsson U, Li L, Pekna M, Berthold CH, Blom S, Eliasson C, Renner O,
Bushong E, Ellisman M, Morgan TE, et al. (2004) J Neurosci 24:5016–5021.
20. Matthews DA, Cotman C, Lynch G (1976) Brain Res 115:1–21.
21. Steward O, Vinsant SL (1983) J Comp Neurol 214:370–386.
22. Steward O, Torre ER, Phillips LL, Trimmer PA (1990) J Neurosci 10:2373–
23. Rose G, Lynch G, Cotman CW (1976) Brain Res Bull 1:87–92.
24. Enge M, Wilhelmsson U, Abramsson A, Stakeberg J, Kuhn R, Betsholtz C,
Pekny M (2003) Neurochem Res 28:271–279.
25. Ogata K, Kosaka T (2002) Neuroscience 113:221–233.
26. Distler C, Dreher Z, Stone J (1991) Glia 4:484–494.
27. Rohlmann A, Wolff J (1996) in Gap Junctions in the Nervous System, eds Spray
D, Dermietzel R (Landes, New York), pp 175–192.
28. Chao T, Rickmann M, Wolff J (2002) in The Tripartite Synapse: Glia in Synaptic
Transmission, eds Volterra A, Magistretti P, Haydon P (Oxford Univ Press,
New York), pp 3–23.
29. Stone DJ, Rozovsky I, Morgan TE, Anderson CP, Finch CE (1998) J Neurosci
30. Belichenko PV, Dahlstrom A (1995) J Neurosci Methods 57:55–61.
31. Buhl EH (1993) Microsc Res Tech 24:15–30.
32. Harlow E, Lane D (1988) Antibodies: A Laboratory Manual (Cold Spring
Harbor Lab Press, Cold Spring Harbor, NY).
www.pnas.org?cgi?doi?10.1073?pnas.0602841103Wilhelmsson et al.