Functions of Mature Mammalian Astrocytes: A Current View

Abstract

Before the roles of normal, mature astrocytes in the mammalian CNS can be discussed, we first need to define these cells. A definition proposed here is that such a class is best defined as consisting of the protoplasmic and fibrous astrocytes of the gray and white matter, respectively, the Bergmann glia of the molecular layer of the cerebellum, and the Muller cells of the retina. It is concluded that the established properties and functions of these mature astrocytes are essential support for neuronal activity, in the sense of Claude Bernard's principle of maintaining "la fixité du milieu intérieur." This milieu would be the extracellular space common to astrocytes and neurons. More specialized roles, such as the recently described "light guides" for retinal Muller cells can also be viewed as support and facilitation. The ECS is also, of course, common to all other neural cells, but here, I limit the discussion to perturbations of the ECS caused only by neuronal activities and the resolution of these perturbations by astrocytes, such as control of increases in extracellular K(+), uptake of excitatory amino acids, and alterations in blood vessel diameter and therefore blood flow. It is also proposed how this fits into the current morphological picture for the protoplasmic astrocytes as having small cell bodies with up to 100,000 process endings that occupy separate territories on which the processes of neighboring astrocytes scarcely intrude.

Full text

http://nro.sagepub.com/
The Neuroscientist
http://nro.sagepub.com/content/16/1/79
The online version of this article can be found at:
DOI: 10.1177/1073858409342593
2010 16: 79Neuroscientist
Harold K. Kimelberg
Functions of Mature Mammalian Astrocytes: A Current View
Published by:
http://www.sagepublications.com
can be found at:The NeuroscientistAdditional services and information for
http://nro.sagepub.com/cgi/alertsEmail Alerts:
http://nro.sagepub.com/subscriptionsSubscriptions:
http://www.sagepub.com/journalsReprints.navReprints:
http://www.sagepub.com/journalsPermissions.navPermissions:
http://nro.sagepub.com/content/16/1/79.refs.htmlCitations:
at BROWN UNIVERSITY on November 21, 2010nro.sagepub.comDownloaded from

79
Functions of Mature Mammalian Astrocytes:
A Current View
Harold K. Kimelberg
The Neuroscientist
Volume 16 Number 1
February 2010 79-106
© 2010 The Author(s)
10.1177/1073858409342593
http://nro.sagepub.com
Before the roles of normal, mature astrocytes in the
mammalian CNS can be discussed, we first need to define
these cells. A definition proposed here is that such a class
is best defined as consisting of the protoplasmic and fibrous
astrocytes of the gray and white matter, respectively, the
Bergmann glia of the molecular layer of the cerebellum,
and the Muller cells of the retina. It is concluded that
the established properties and functions of these mature
astrocytes are essential support for neuronal activity, in the
sense of Claude Bernard’s principle of maintaining “la fixité
du milieu intérieur.” This milieu would be the extracellular
space common to astrocytes and neurons. More specialized
roles, such as the recently described “light guides” for retinal
Muller cells can also be viewed as support and facilitation.
The ECS is also, of course, common to all other neural
cells, but here, I limit the discussion to perturbations of the
ECS caused only by neuronal activities and the resolution
of these perturbations by astrocytes, such as control of
increases in extracellular K+, uptake of excitatory amino
acids, and alterations in blood vessel diameter and therefore
blood flow. It is also proposed how this fits into the current
morphological picture for the protoplasmic astrocytes as
having small cell bodies with up to 100,000 process endings
that occupy separate territories on which the processes of
neighboring astrocytes scarcely intrude.
Keywords: homeostasis; tripartite synapse; transmitter
uptake; transmitter release; potassium control; pH control;
astrocyte classi fication; aquaporins
Philosophical Considerations
Reviews, the Scientific Method, and
the Computer Revolution
“An account of the history of philosophy may proceed
in two ways. Either the story is purely expository, show-
ing what this man said and how that man was influenced.
Alternatively, the exposition may be combined with a
certain measure of critical discourse. This second
course has been adopted here” (Russell 1989).
In the biomedical sciences, reviews, as in the
Annual Review Series, used to be more like the first
alternative in the quote above from the foreword to Ber-
trand Russell’s illustrated shorter book, Wisdom of the
West, based on his well-regarded and nonillustrated
History of Western Philosophy published just after World
War II. These reviews were also parsimonious in put-
ting forward interpretations of the findings. Nowadays,
reviews in the biomedical sciences are much more in
the latter category of the quote, and interpretation is
quite generous. One reason for the changes is surely
the remarkable ease of access to the literature through
computerized search engines obviates the need for the
comprehensive list of publications that the older reviews
provided. Russell, who was born in 1872 and died in
1970, could not have foreseen the “Google” revolution
that now makes his first alternative, interpreted for
biology as published data and their interpretation,
largely redundant.
The computerized searches are, however, value
neutral, so an experienced scientist can always, hope-
fully, add valuable and objective commentary and put
things into perspective. I will therefore supply a “cer-
tain measure of critical discourse.” The relationships
between an idea and the preceding ones, a historical
analysis as in “how that man was influenced,” is also
something in which someone familiar with the field for
some time could be helpful, and because they may have
heard or even experienced what actually happened. I
have been active in the glial/astrocyte field for some 35
years now, having started with isolated cells and astro-
cyte cultures in 1973. My experience in doing science
started with my PhD studies in biochemistry, which I
completed in 1968, on “Studies of Cytochrome Oxidase
in Phosphorylating and Non-phosphorylating Systems”
and postdoctorally on reconstitution studies with cyto-
chromes and other membrane proteins with model
phospholipid membranes, before getting interested in
astroglia. At this stage of my career and life, a certain
memoir element inevitably creeps in, and I hope the
indulgent reader will humor me in this.
The criterion for the critical discourses will be to
try to give the simplest, valid interpretation of the pub-
lished data unless, by passage of time or error, the data
Ordway Research Institute, Albany, New York
Address correspondence to: Harold K. Kimelberg, Ordway Research
Institute, 150 New Scotland Avenue, Albany, NY 12208 Email:
hkimelberg@ordwayresearch.org
Review
at BROWN UNIVERSITY on November 21, 2010nro.sagepub.comDownloaded from

80 The Neuroscientist / Vol. 16, No. 1, February 2010
are seen to be flawed, which I may note as a warning
sign to those new to the field. This is in the spirit of the
13th century scholastic philosopher William of Ockham
that if a simple explanation will do, it is idle to seek a
complex one or one should not multiply entities beyond
what is necessary (Russell 1989). This was adopted as a
principle of the modern scientific method by such lumi-
naries as Isaac Newton (Taylor 1949; rule 1: we are to
admit no more causes of natural things than such as are
both true and sufficient to explain the appearances;
Feynman 1965), who realized that although all must
rest on observations and sound reasoning therefrom,
the basic principle of the philosophical school of empir-
icism (Russell 1989), one has no way of knowing if one
has a sufficient body of the needed data, and one cannot
anticipate the new findings, the essential limitation of
the empirical method. It has come to be known as
“Ockham’s razor.” Therefore, when a simple and correct
understanding emerges from an explanation, it is held
to support it, not so much a principle as a point of view.
Inferences from data are evaluated on the reason-
ableness of their interpretation of the observations.
This process is termed induction, which has a well-
known problem, illustrated by Russell’s inductivist
turkey, who, based on observations of regular feedings
at the same time every day over a long period of time,
hypothesized that he will be fed at 9 AM every day.
Unfortunately, the day after he thought he had enough
data to make this conclusion, he had his throat cut
because it was the day before Christmas, and he had
never observed Christmas. A hypothesis is an induction
from prior instances and is therefore always subject to
this type of error. The scientific method, therefore, adds
another step. Any useful hypothesis must have novel
implications. Therefore, to increase confidence in the
truth of the hypothesis, controlled experiments are
done to see if these predicted events occur. As with the
turkey feedings, when different processes, that is, fat-
tening the turkey from goodwill or for consumption,
share the same characteristics, one can be misled as to
what is going on. This may be a problem with the inter-
pretation of transmitter-filled vesicle release, considered
as a test for astrocytic control of synaptic transmission,
which shares a number of characteristics with other
cellular fusion processes that also involve SNARE pro-
teins (Sudhof and Rothman 2009).
Experimental verification is sought for in all the
scien ces, not just biology. For even in physics in which
reliable theory is much more possible, a theoretical
interpretation is only fully accepted if its predictions
conform to observed behavior (Poincare 1952). How-
ever, unlike physics in which the generalizations
(hypotheses) can be expressed mathematically because
of the homogeneity of the elementary phenomena con-
sidered, the biological sciences deal with phenomena at
a more complex level, and thus, precise mathematical-
type formulations are not possible (Poincare 1952). In
spite of Karl Popper’s insistence on falsifiable and least
probable hypotheses as the most fruitful approach
(Pavesi 1999), as a practical matter, we look for the most
likely hypotheses and always try to support them and
therefore limit what we find out (Poincare 1952). Ulti-
mately, of course, “the natural sciences, and in particular,
the most fundamental of them, physics, deal with sense
perceptions,” and “the object of all science is to coo-
rdinate our experiences and to bring them into a logical
system” (Einstein 1956).
Where Do Our Hypotheses of Astrocyte
Function Come from and How Do We Test
Them?
In spite of the logical limitations of the inductive nature
of hypotheses, in practice, they are widely used. I now
discuss where the hypothesized functions of astrocytes
have originated. In general, this question applies to any
cell. Early on, it was the morphology and the relations
of the astrocyte and their processes to other compo-
nents of the nervous system. Morphology still continues
to be instructive (e.g., Reichenbach and Wolburg 2009),
and the early ideas were quite prescient such as the
uptake of by-products of neuronal activity by the glial
processes surrounding articulations (i.e., synapses)
(Lugaro 1907), and this could be seamlessly extended
to transmitters, and the transfer of food from blood ves-
sels to neurons (Golgi 1885). Some were far more wide
ranging and speculative, such as control of sleep wake-
fulness by astrocyte processes intruding between
neurons by swelling and shrinkage and thereby modu-
lating interneuronal exchange as proposed by Schleich
in 1894 (Dierig 1994) and Cajal in 1895 (Garcia-Marin
and others 2007). However, at that time, experimental
tests of the likelihood of any of these properties were
not done, simply because they could not be. Continuing
such speculations without a good exp erimental founda-
tion has been a subject of criticism of the “glial” field
(Nicholls 1981a), and it would be unwise to repeat
them.
So, clues from morphology led to hypotheses of
function, which later were tested by methods that were
developing. These methods were originally histochemi-
cal for the cellular level and organ specific for the
analysis of biochemical processes and later, cell-specific
physiological studies. We are now seeing an explosion
in the type and complexity of measurements in biology
that are possible, so any ideas we have can now be
better examined as to how their features and their con-
sequences conform to how the system actually behaves.
The logical basis of the inferential und erpinnings, of
course, has not changed, and a lot of the controversy
arises from what are reasonable inferences to be drawn
from all the available observations, both the basic
observations and the experimental tests (both the latter
adjective and noun mean the same and derive from
at BROWN UNIVERSITY on November 21, 2010nro.sagepub.comDownloaded from

Functions of Mature Mammalian Astrocytes / Kimelberg 81
Latin: experimentum, a trial or test, and test is from
testum, an earthen pot for trying metals in).
How do we decide at any one time what specific
astrocyte functions to investigate? Individuals or groups
make guesses (Feynman 1965) as to function based on
the evidence available. So, other than our intuition, the
only reliable basis for our guesses, usually referred to by
the Greek word hypothesis (but do they not essentially
mean the same thing?), is the body of knowledge avail-
able. Once a guess is made, its implications are tested.
For biological processes, this is often done by using
inhibitors, of a type and at a concentration to confer
specificity, or by mimicking the activity using agonists.
However, a serious problem for astrocytes is that most
of their biochemical and physiological properties are
not exclusive to them. Therefore, investigators are
increasingly using cell-selective gene modification by
use of cell-specific promoters, such as GFAP for astro-
cytes. Hopefully, these are specific! This is clearly a
major advance in examining astrocyte function in vivo,
but at this early stage, there will be surprises and pit-
falls. A recent chapter from the McCarthy laboratory,
which has applied mouse transgenic approaches most
ext ensively to astrocytes, covers the problems and the
basic techniques employed and summarizes the results
so far published of altering specific genes in astrocytes
(Fiacco and others 2009). Some of the results were
unexpected and surprising. Major problems empha-
sized in this article are the pleiotropic effects of altering
one gene on many gene products, the difficulty of
achieving true cellular specificity, and the fact that the
gene needs to be altered late enough so that it does not
alter developmental processes. As already discussed,
what genes to alter is still a guess and personal prefer-
ence based on the randomly obtained extant data. At
present, these guesses seem pretty chancy in terms of
whether reasonably interpretable results are likely to be
achieved, especially in relation to the extensive costs
and difficulties of this approach and the large number
of potential genes and their variants that can be altered.
However, the rapid pace of technical advances is such
that many of these problems are likely to be ironed out.
Still, we should not simply assume that this is always
the best way of finding out what astrocytes do.
A feasible approach that can possibly provide a less
chancy way of choosing a property to alter or study is to
look at the whole of the messages expressed by the
microarray technique or use an array devoted to a par-
ticular class of genes. Both have very recently been
done for astrocytes (Cahoy and others 2008; Lovatt and
others 2007) and will be discussed in some detail in the
section “Recent Transcriptosome Data for Astrocytes.”
Also, refer to the recent perspective article by one of the
pioneers of this approach for astrocytes and other cells
of the CNS (Barres 2008). The highest expressed mes-
sages likely reflect the major proteins being made at
that time but this also depends a lot on relative rates of
synthesis and decay. But overall, the transcriptosome
data conform to what proteins are preferentially
expressed by astrocytes, and this approach should take
some of the guess work out of figuring out what pro-
teins and protein-dependent functions to study by gene
modification, effects of inhibitors, or other means.
Scientific Considerations
Astrocyte Classification Problem during
Development and in the Mature Brain
Astrocytes were the original neuroglia of the brain
(Kettenmann and Ransom 2005). I am only considering
in this review mature astrocytes, but they have essential
roles in development. They are migration pathways for
neurons, they support the growth of neurons, they are
critical to the development of neurons and synapses,
and they serve as progenitors for neurons and other
neural cells (Ransom and others 2003). Hence, it is not
surprising that their properties should change with
development as their functions change. Therefore,
much of the confusion, that is, contradictory data and
interpretations, in the astrocyte field is due to studying
astrocytes without regard for their developmental stage
and the type of preparation.
Therefore, although the title of this review was
originally suggested by the editor to be “The Astrocyte:
Current Views of Function,” I have changed this to
“Functions of Mature Mammalian Astrocytes: A (i.e.,
My) Current View.” I consider the totality of the func-
tions of developing and mature mammalian astrocytes
too large for a single, critical review, although Wang
and Bordey (2008) have very recently been brave
enough to undertake the task, and that review can be
usefully consulted for many of the topics omitted in the
current review, as well as, of course, a number of other
reviews and a large primary literature on astrocytes in
development (e.g. Ihrie and Alvarez-Buylla 2008).
These include the developmental origins of astrocytes
and the identification of GFAP(+) multipotential pro-
genitor cells (stem cells according to some definitions)
in the subventricular zone as astrocytes. However, as
Wang and Bordey (2008) pointed out, defining this
question on GFAP positivity alone does beg the ques-
tion of why these cells should be classified as astrocytes.
However, taking the 8 criteria noted in the next section
as diagnostic of astrocytes, they do possess all of them
(Liu and others 2006). Therefore, they are either stem
cells that show all the properties of astrocytes or astro-
cytes that also behave as multipotential (stem?) cells.
The criteria also exclude the NG2(+) cells, also known
as oligodendrocyte progenitor cells (or OPCs), which
were at one time thought to be astrocytes based on mor-
phology but are now known to show many differences
at BROWN UNIVERSITY on November 21, 2010nro.sagepub.comDownloaded from

82 The Neuroscientist / Vol. 16, No. 1, February 2010
and have been proposed to be a fourth class of glia with
the suggested name of polydendrocytes (Nishiyama and
others 2009). The role of astrocytes in the development
of synapses and other aspects of astrocyte function has
recently been summarized by Barres (2008), echoing
many of the points just raised here.
Definition of Mature Astrocytes
There has been continued debate regarding what cells
should be considered as mature astrocytes (Barres
2003, 2008; Kimelberg 2004a, 2004b, 2009) and there-
fore, naturally, of whether they form a homogeneous
group. The most agreed-on and well-established criteria
for recognizing a mature astrocyte seem to be the
following:
1. First and foremost, nonexcitability. This is a
necessary but not sufficient criterion for an
astrocyte. Thus, if a cell is excitable, it is auto-
matically a neuron.
2. A very negative membrane potential deter-
mined by the transmembrane K+ gradient.
Again, a necessary but not sufficient criterion.
3. Uptake of the widespread excitatory amino
acid glutamate and GABA by astrocyte-
specific transporters. This and the remaining
criteria are more astrocyte specific, but like
most criteria, are not absolute. It is the com-
bination of a number of characteristics that
becomes more accurate or, in Bayesian terms,
more probable (Kimelberg 2004a, 2004b,
2009).
4. A large number of intermediate filament bun-
dles, which are the sites of the astrocyte-
specific protein GFAP.
5. Glycogen granules.
6. Processes from each cell surrounding blood
vessels.
7. Many more processes from each cell sur-
rounding synapses.
8. Linkage to other astrocytes by gap junctions
consisting of connexin 43 and 30.
These 8 criteria exclude, for example, the NG2
cells, which share some of the above criteria but can be
easily recognized on the basis of being NG2 positive
and negative by not exhibiting characteristics 3 to 8
(Nishiyama and others 2009). In the mature mamma-
lian brain, I would suggest that the term astrocyte under
physiological conditions be restricted to the white
(fibrous) and gray (protoplasmic) matter astrocytes and
that the Muller cells of the retina and the Bergmann
glia (actually the combined Golgi epithelial cells and
Bergmann fibers to be precise but unwieldy, so let us
stay with Bergmann glia; see Fig. 1). Using the
classification system for organisms for illustrative pur-
poses, we could have the following:
Normal mature astrocytes (class):
• Star-shaped astroglia (order)
1. Mature protoplasmic astrocytes in the gray
matter (species)
2. Mature fibrous astrocytes in the white matter
(species)
• Elongated astroglia (order)
3. Mature Muller cells spanning the retina (spe-
cies)
4. Mature Bergmann glia (BG) in the molecular
layer of the cerebellum (species)
To conform to the true Linnaean system, we would
need some Latin terms for the above, and this might be
a useful exercise. Obviously, we do not have the sine
qua non of species, sexual reproduction, but to carry
the analogy a little further, it seems that all the cells I
have noted as species only give rise to their own kind
by division. This would also eliminate the GFAP(+)
stem cells and would distinguish developing astrocytes
from mature astrocytes.
I exclude the hypertrophic, intensely GFAP(+)
astrocytes, which develop in response to different forms
of injury, as did Wang and Bordey (2008). This is a gen-
eral reaction of astrocytes to injury, and they are
therefore termed reactive astrocytes and should be per-
haps best included in a separate class termed mature
pathological astrocytes, which will then contain sepa-
rate orders and species. Their morphology is well
established (Wilhelmsson and others 2006), but there
is a relative paucity of information on how their proper-
ties and functions are altered. Therefore, I cannot
speculate at present on whe ther they are a class or not.
This is a fertile field, but the models used need to be
well founded. Currently, using culture models of reac-
tive astrocytes, as for normal astrocytes, would best be
avoided as it will introduce too much uncertainty into
the findings.
The radially oriented Bergmann glia (BG) resem-
ble the radial glia (RG) that give rise to starry astrocytes
in other regions during development and share with
the RG the important role of a scaffold for the migra-
tion of granule cells from out to in, the opposite
direction of the migration in the cortex (Rakic 1990).
They also give rise to astrocytes, ependymal cells, and
the GFAP(+) SVZ cells. The latter can also give rise to
another progenitor cell, which can give rise separately
to neuroblasts and astrocytes (Nishiyama and others
2009). The “radial glia,” or better radial astroglia
at BROWN UNIVERSITY on November 21, 2010nro.sagepub.comDownloaded from

Functions of Mature Mammalian Astrocytes / Kimelberg 83
(Morest and Silver 2003), can be viewed as the first
derivatives of undifferentiated epithelial cells that
divide the neural cylinder between a luminal (CSF
compartments) and an abluminal surface but are also
major multipotential progenitor cells in the early devel-
oping CNS. The other cell type I am considering as a
mature astrocyte, the Muller cells, remains in the adult
as epithelial-like cells but does not seem to function
for migration of the different retinal neurons (Morest
and Silver 2003). This leads to a concept that astro-
cytes can be viewed more like an epithelial cell, both
ontogenetically and functionally, than a neuron. In
their interposition between blood vessels and neurons
and synapses, they retain an epithelial-like cell rela-
tionship. This was proposed as a basis for the
segregation of ion transporters to effect pH control
(Kimelberg and others 1979; Kimelberg 1979). It has
now been clearly shown that 2 channels show such
segregation. For example, the AQP-4 water channel,
and the inwardly rectifying Kir4.1, K+ channel mole-
cule, are predominantly localized in the perivascular
astrocytic endfeet membranes (Amiry-Moghaddam
and Ottersen 2003). However, this presents a problem
because these very small localized areas cannot be
studied electrophysiologically, and other methods will
need to be used and developed. As a first step, perhaps,
we need a similar reliable description of the localiza-
tion of other channels and transporters.
Use of Astrocyte Cultures to Probe
Properties and Functions
The use of cultures is really to get enough viable cells to
study a process biochemically and for ease of access
and to directly assess the property of a class of cells
without the confounding problem of indirect effects
Figure 1. (A) Golgi-stained astrocytes from a 2-month-old human infant in the molecular and layers II and III of the cerebral cortex. A, B,
C, and D are cells in the first cortical lamina, and E, F, G, and H are cells in the second and third lamina. I and J are cells with endfeet
contacting blood vessels. V, blood vessel. From Ramón y Cajal 1913. (B) Muller cells in different species as indicated. Furthest right: Muller
cells as light guides. Courtesy of S. Skatchov from Franze and others (2007), cover page of that issue. (C) Different astrocyte cells in mature
mammalian cerebellum by Golgi staining (Golgi 1885). Working down from the cerebellar surface A at the top, represented by dotted line.
M, molecular; P, Purkinje; G, granule cell; W, white matter layers; b, Bergmann glia; s, protoplasmic astrocytes; v, velate astrocytes; f,
fibrous astrocytes. (D) Dye-filled astrocytes of rat hippocampus (personal communication, Eric Bushong).
at BROWN UNIVERSITY on November 21, 2010nro.sagepub.comDownloaded from

84 The Neuroscientist / Vol. 16, No. 1, February 2010
when investigating them in intact tissue. However, ease
of use is not a justification for using a model that is
invalid in terms of the questions being asked. Because
of ease of use, there have been a lot of studies that have
used primary GFAP(+) cultures prepared from different
regions of neonatal rodent brains. However, the fidelity
of this experimental short cut has been questioned
based on conflicting observations (Fiacco and others
2007; Kimelberg 1999, 2001; Cahoy and others 2008)
and is to be expected based on the plasticity of gene
expression, which is likely to alter when the astrocyte
precursors are grown for several weeks as a homoge-
neous cell culture. Indeed, Agulhon and others (2008)
recently wrote that “most investigators in the field
acknowledge that cultured astroglia represent a very
poor model for studying the functions of astrocytic
GqGPCRs in situ or in vivo.” And presumably for any-
thing else, unless the cultured astroglia express in all
particulars a property already shown for astrocytes in
situ or in vivo. In that case, like any cell culture system,
they can be used to examine aspects in a more detailed
manner. This may be particularly useful for primary
cultures made from transgenic animals in which the
effect of the genetic change can be looked at in a more
controlled manner. Also, the cultures can be more easily
used for siRNA techniques to look at the effect, for
example, of removing a protein when a good inhibitor is
not yet available. But what has now been shown to lead
to errors is where the cultures are used to predict what
properties astrocytes will have in situ. I tho ught this
was an important function for primary cultures 25 years
ago (Kimelberg 1983) but did emphasize that in situ
studies were always needed for confirmation. In view of
unpredictable differences between properties of astro-
cytes in culture and their properties in situ or in vivo,
theoretically to be expected from the plasticity of gene
expression, and also now demonstrated, I now think
that the direction has to be from what is known about
astrocyte properties and functions in situ, to those
properties and functions that cannot be examined pre-
cisely enough in situ, in primary cultures when they
robustly exhibit the same properties.
Functions of Mature Astrocytes
Exclusive dependence of astrocytic membrane potential
on K+ and the maintenance of [K+]o. The first dynamic
studies on glial cells were electrophysiological, as this
was the only cell-specific physiological technique that
could be used in situ, and were done in the amphibian
optic nerve in the mid-1960s. At that time, cultures
would have been frowned on! Only cell bodies could be
impaled, and this was actually an advantage as all the
cell bodies in the optic nerve are glia. Also, amphibian
tissue was easier to work with compared with mamma-
lian tissue. Two important findings emerged from these
studies. One was that these glia were found to be non-
excitable, and the second was that they had very nega-
tive membrane potentials determined essentially
exclusively by the transmembrane K+ gradient (Kuffler
and others 1966). These amphibian optic glia were
likely best described as astrocytes, but important later
work in cats showed that the same characteristics
applied to recorded astrocytes identified histologically
after recording in the adult mammalian cortex (Picker
and others 1981).
This led to the first seemingly experimentally sup-
ported function of glia as having a role in maintaining a
constant extracellular K+ concentration in the face of
neuronal activity, which would tend to increase it
(Orkand and others 1966). The process was termed K+
spatial buffering and involves redistribution of increased
[K+]o by a current loop set up by a membrane potential
difference that is due to locally increased [K+]o and
serves to dissipate the increased [K+]o to distant sites,
hence the term “spatial buffering” (Gardner-Medwin
1983; Kofuji and Newman 2004). It was an ingenious
speculation giving a function for the exclusive K+ con-
ductance of the astroglial membrane and seemingly
further supported by the organization of the cells into
syncytia, which could, in theory, carry the currents for
long distances. However, the evidence was actually very
indirect, and there is still debate as to the significance
of this function in vivo. None of the tests of this hypoth-
esis attempted from time to time, except perhaps for the
Muller astrocyte (Newman and others 1984), have
been even remotely conclusive. A major issue that
remains unresolved is how far the current can spread,
given the low resistance of the astrocyte, in relation to
the gradients of increased [K+]o that can be generated
by neuronal activity and how to demonstrate it really
occurs.
Later work also showed that astrocytes had voltage-
and time-dependent K+ channels and other ion
channels. This would degrade the efficiency of the K+
spatial buffering process. However, these findings were
usually seen in immature astrocytes and oligodendro-
cyte precursors, either in culture or in slices from young
animals (Zhou and others 2006), and could be due, for
example, to their involvement in the proliferation of
these cells at these earlier developmental times (Paez
and others 2008; Burg and others 2008). The identities
of the K+ channels present in mature astrocytes are still
unresolved. Equally important is where they are located
in terms of the intracellular segregation issue discussed
above. Currently, for mature protoplasmic astrocytes, it
is now known that the K+ channels will include Kir4.1
and some members of the 2 Pore (2P) K+ channels
(Skatchov and others 2006). The question of location
applies to all the membrane tra nsporters and channels
that will figure prominently in the support roles for
mature astrocytes proposed in this review. Such studies
at BROWN UNIVERSITY on November 21, 2010nro.sagepub.comDownloaded from

Functions of Mature Mammalian Astrocytes / Kimelberg 85
should be quite doable using techniques such as immu-
noelectron microscopy. This actually is “rocket science”
because “rocket science” is purely an engineering prob-
lem based on now 300-year-old discoveries by Newton
of the basic principles of motion and gravity.
In spite of unanswered issues regarding the original
K+ spatial buffering hypothesis, clearance of increasing
[K+]o due to neuronal activity became established in the
canon of astrocyte functions and was extended to
include other K+ uptake mechanisms. These included
K+ uptake by the (Na+ + K+) pump and cotransporters
belonging to the Slc12a gene family. These proposals
were even less well supported by the data. For example,
the (Na+ + K+) pump, due to the sites and the affinities
for Na+ and K+, responds most effectively to changes in
[Na+]i to pump out accumulating Na+, and its outside
K+ binding site is approximately 90% saturated at the
normal [K+]o of 3 mM. The NKC cotransporter
(Slc12a1) is also not particularly suited to dealing with
increases in [K+] much in excess of resting levels and
not expressed in astrocytes in the mouse transcripto-
some or was not present on the array (Cahoy and others
2008). However, if these uptake systems are active
enough to keep [K+]o within a few millimolars of its rest-
ing level, they should function well enough.
Another mechanism is by Donnan uptake of
increased [K+]o and Cl–, and this should not saturate at
any [K+]o value. However, this mechanism requires Cl–
channels in addition to K+ channels (Hodgkin and
Horowicz 1959), and there is little evidence for Cl–
channels under the mildly elevated [K+]o levels seen, for
example, during normal neuronal stimulation (Amzica
2002). There are likely to be specific K+ channels in the
astrocyte (Cahoy and others 2008), but it is not clear
that this is so for the (Na+ + K+) pump, and messages for
the other transporters mentioned above appear to be
absent (Cahoy and others 2008). A priori, there is no
reason K+ released from neurons cannot be controlled
by reuptake into the neuron, but this would be a delayed
phenomenon, and the astrocyte could take up K+ as
soon as it is released, but this by itself is not a compel-
ling reason. The (Na+ + K+) pump uses ATP directly
to maintain high [K+]i and low [Na+]i, and the transport-
ers use the resultant inward Na+ gradient to drive
the transport of obligatorily cotransported or exchange-
transported substances, a process referred to as secondary
active transport (Stein 2002). In the absence of free Na+
permeability, the (Na+ + K+) pump needs only to be acti-
vated to pump out Na+ that has entered the cell usefully
to effect some other needed transport, a situation that
has evolved beautifully in astrocytes. Thus, there is an
alternative, more reasonable function for the (Na+ + K+)
pump, and applying “Ockham’s razor” would favor that.
Ionic or metabotropic receptors on astrocytes and astro-
cyte-neuron signaling. The sensitive response of the
astrocyte membrane potential to changes in [K+] was
the first form of glial-neuronal signaling from neurons
to glia observed (Orkand and others 1966). At this early
stage in glial physiology, a specific feedback from the
astrocyte to neuron was not proposed. Such feedback is
now the focus of astrocytic modulation of synaptic
activity, and it is proposed that this is due to exocytotic
release of transmitters from astrocytes, secondary to a
rise in intracellular astrocytic Ca2+ (see section
“Modulation and Control of Synaptic Activity as in the
Tripartite Synapse Concept”), due to the stimulation of
astrocytic metabotropic receptors by the spillover of
released synaptic glutamate, ATP, or other substances.
Because there are no voltage-gated calcium channels
reliably shown in astrocytes in situ, influx of calcium is
not possible to achieve the high localized concentra-
tions of intracellular Ca2+ seen at neuronal sites of
release. In any case, the depolarization of the astrocyte
by the modest physiological rises of [K]o (Δ2-3 mM)
(Amzica 2002) is insufficient to depolarize the astro-
cyte enough to activate voltage-sensitive Ca2+ channels.
The modest depolarization of astrocytes is shown in
Figure 2 for spontaneous activity during anesthetic-
induced sleep in the cat cortex.
For some time, astrocyte depolarization was not
thought to occur by activation of ionotropic receptors,
as an early study had shown that depolarization of
astrocytes due to the addition of GABA was better
explained by electrogenic uptake of GABA because no
conductance changes could be measured, as expected if
ionotropic astrocytic receptors were being activated,
but this is an argument by omission and was not other-
wise directly shown (Krnjevic and Schwartz 1967).
Constanti and Galvan (1978) similarly concluded from
work on neuroglia in slices that depolarization due to
added glutamate was because of raised [K]o due to neu-
ronal activation. Thus, perhaps because of these and
other studies, the view arose that astroglia lacked recep-
tors and only responded to changes in neuronal activity
through depolarization of their membrane potentials in
response to raised [K]o. This also simplified neuronal
research because receptor-mediated effects of neu-
rotransmitters in intact tissue could then be attributed
only to neurons, so maybe it was accepted more readily
than it should have been. The major problem, however,
is that this view is wrong. It would be a very unusual
cell that lacked receptors, for all cells have receptors to
respond to their environments, albeit many of these are
metabotropic rather than ionotropic. The more rational
view is that astrocytes lack ionotropic receptors that
would make more biological sense, as astrocytes, being
nonexcitable, do not need to change their membrane
potentials for the same purposes as neurons. However,
this view was not borne out by early studies that showed
that GFAP(+) astrocytes in primary culture had gluta-
mate and GABA ionotropic receptors (Bowman and
at BROWN UNIVERSITY on November 21, 2010nro.sagepub.comDownloaded from

86 The Neuroscientist / Vol. 16, No. 1, February 2010
Kimelberg 1984; Kettenmann and others 1984), as well
as a variety of metabotropic receptors (Kimelberg
1988). The cultures, as discussed in the section “Use of
Astrocyte Cultures to Probe Properties and Functions,”
may, because of alterations in their gene exp ression,
express more receptors in culture than in situ (Kimel-
berg 1988), and the existence of ionotropic receptors in
astrocytes in situ, with a few exceptions, is particularly
uncertain (Zhou and Kimelberg 2001; Wang and Bordey
2008). One of these exceptions is the AMPA-type recep-
tors of the Ca2+ permeable variety found on Bergmann
glia in situ. In these cells, there are concentrations of
vesicles in the boutons of parallel fibers abutting BG.
These give fast inward currents in the BG due to the
release of glutamate from these vesicles, a process
termed ectopic release (Matsui and Jahr 2006). Other
than a role in maintaining excitatory synapses on the
Purkinje neurons for the Ca2+ influx associated with this
activation (Iino and others 2001), other roles have not
yet been described.
One problem for determining ionotropic receptor
activity is technical. Electrophysiological demonstra-
tion of active ionotropic receptors by the standard
method of showing conductance changes is very diffi-
cult in astrocytes because astrocytes in situ have
extremely low resistances (Zhou and others 2006, 2009).
This does not appear to be due to the syncytium but
rather a reflection of the high density of leak K+ chan-
nels (Schools and others 2006). The identification of
these channels, as well as their distribution in the plane
of the membrane, may begin to resolve, at last, the issue
of K+ spatial buffering. However, the extraordinarily low
input resistance of the astrocyte plasma membrane of
approximately 5 MΩ or so makes it impossible to study
Figure 2. (A) Model of spatial buffering (see text). (B) Normal intraneuronal electrical activity in the cat motor cortex during anesthetic-
induced sleep. It shows the widely seen continuous slow oscillation of the resting membrane potential between a hyperpolarized and
depolarized state differing by 10 to 15 mV, with action potential bursts occurring during the depolarized (up) state. (C) This shows the
membrane potential recorded with a second intracellular electrode, impaled into an astrocyte starting just after the second neuronal oscil-
lation cycle. The concurrent astrocyte and neuronal membrane potentials are expanded in the insert. The astrocyte membrane potential
oscillated in phase with the neuronal oscillations, and its peak depolarization (Δ2-5 mV) coincided with the end of the neuronal depolarized
state. The concurrent [K+]o increases were also measured (not shown), and the peak Δ values were approximately 1 mM. Modified from
Amzica 2002.
at BROWN UNIVERSITY on November 21, 2010nro.sagepub.comDownloaded from

Functions of Mature Mammalian Astrocytes / Kimelberg 87
the voltage-current relations across the astrocyte
plasma membrane and detect modest changes caused
by the activation of ionotropic receptors using the con-
tinuous single electrode, voltage-clamp technique, the
preferred method for studying the electrophysiology of
small cell bodies (Zhou and others 2009).
Control of [H+]o. An involvement of astrocytes in brain
pH control was first proposed based on a localization of
carbonic anhydrase (Car) in astrocytes at the blood
brain barrier, involving the transport of HCO3– linked
to the acceleration of intra-astrocytic CO2 hydration
(Tschirgi 1958). Subsequent studies showed Na+/H+
and Cl–/HCO3– exchangers and Car activity in pri-
mary astrocyte cultures (Kimelberg and others 1982;
Kimelberg and others 1979; Kimelberg 1981), and a
model of coupled Na+/H+ and Cl–/HCO3– exchange
was proposed to explain astrocytic swelling under
pathological conditions (Fig. 3). This model was, how-
ever, very difficult to test in situ. The all important cell-
specific measurements were made by acid-sensitive,
ion-specific microelectrodes (ISMs). Astrocytes in situ
were found to have normal pHi values of around 7.3,
but in ischemia, astrocytic pHi values of around 5.0
were found (Kraig and others 1985; Schmitt and others
2000). Astrocytes were therefore proposed to function
as proton sinks in ischemia. Such a low intracellular
pH is a surprising finding that has still not been well
explained. Other findings in primary astrocyte cultures
are an electrogenic sodium proton cotransporter (3
HCO3– to 2 Na+), which can acidify the extracellular
space when stimulated by an increased K+-dependent
depolarization and alkalinize the cell interior (Rose and
Ransom 1998), and sodium-dependent Cl–/HCO3–
exchange, as well as the sodium-independent exchanger
mentioned above (Schmitt and others 2000; Majumdar
and others 2008; Parker and others 2008). All these are
members of the bicarbonate transporter (BT) super-
family. Inspection of the trans criptosome list shows
that many of these transporters (gene symbols starting
with Slc), as well as other acid homeostasis-related
transporting systems, are preferentially expressed in
astrocytes (see Table 1 for some examples and supple-
mentary material in Cahoy and others 2008).
However, all these are correlational studies in the
sense that they show astrocytes have certain properties
consistent with a role in pH regulation, excluding the
ISM exp eriments noted above. A number of the studies
were done in culture and are flawed because of the
unpredictable correspondence between the properties
of such cells and the properties of astrocytes in situ.
The most recent studies address the obvious, yet neces-
sary, preliminary questions of which tra nsporters are on
astrocytes and which on neurons or if they are present
on both (Majumdar and others 2008). The next stage
would still be descriptive in that, as argued later for all
transporters and channels, the locations and likely seg-
regation of the transporters for the different regions of
individual astrocytes should be determined. This will
provide a good basis for hypotheses that can then be
tested by imaging with pH-sensitive dyes at the highest
resolution possible, and the effects of cell-specific and
conditional transgenics, if acute cell-specific effects
can be achieved. This is, in my opinion, a fertile field
for future work.
Another H+ transporting system is the lactate + H+
transporter, different isozymes of which are present in
astrocytes and neurons. These systems are responsible
for the efflux of lactate from astrocytes and uptake of
lactate into neurons as proposed in the astrocyte neuro-
nal lactate shuttle hypothesis (ANLSH) (Pellerin and
Magistretti 2004) (also see “Conclusions”). Thus, lac-
tate transport affects both pH levels, or viewed as the
converse, pH can affect lactate exchange and therefore
link the ANLS to changes in pHe and pHi. Note that
lactate transport is directly linked to pH because of the
H+ being cotransported with lactate. The end product
of glycolysis is the lactate anion and not lactic acid, so
it does not directly influence pH, as is often incorrectly
implied in the term lactic acidosis (Robergs and others
2004).
There are profound effects of changes in pH on
neuronal excitability, but to what extent regulation of
extracellular pH depends on astrocytes is really not
known. Astrocytes in situ contain high mRNA levels for
many pH-regulating enzymes and transporters (see
Table 1, gene symbols Car2, Slc4a4, Slc9a3r1). The
next step would be to determine where on the astrocyte
the membrane transporters are localized (Fig. 3) and
then to see how pH changes in the CNS are modified
when these are inhibited. Additionally, one could knock
out these transporters by astrocyte-specific genetic
ablation techniques, but these might be expected to
have far-ranging consequences. However, doing it any
other way is a problem because these pH transporters
are also seen on neurons, although there may be astro-
cytic specificity for some of the isoforms (Majumdar
and others 2008).
Control of extracellular glutamate, GABA, glycine, and
taurine concentrations by uptake. Uptake of the excit-
atory amino acid transmitter glutamate is arguably the
best established and important property of mature
astrocytes. Inactivation of glutamate then occurs by
conversion to glutamine by the astrocyte-specific, intra-
cellular enzyme glutamine synthetase, consuming ATP
and ammonia in the process. The glutamine is pro-
posed to leave the astrocyte, in the same region as it is
normally taken up, namely the perisynaptic processes,
by an amino acid carrier and is taken up by neighboring
neurons, where it is reconverted to glutamate via gluta-
minase (Hertz and others 2007).
at BROWN UNIVERSITY on November 21, 2010nro.sagepub.comDownloaded from

88 The Neuroscientist / Vol. 16, No. 1, February 2010
The uptake of glutamate into a small second com-
partment was first proposed based on the biochemical
evidence that the specific activity (SA) of radiolabeled
glutamine was greater than the SA of the radiolabeled
glutamate, its immediate precursor after injection into
the brain. Obviously, if the injected radiolabeled gluta-
mate had equilibrated with all the unlabeled glutamate
present in the brain, the SA of its product could not
have been greater (Berl and others 1961). By now, a
wealth of evidence has shown that specific EAA carriers
are specifically expressed in astrocytes (also shown by
high and specific levels of mRNA) (see Table 1) and
that their knockdown leads to increased glutamate
levels in the ECS (Rothstein and others 1996).
Mature astrocytes in situ are also known to contain
GABA transporters, specifically GAT1 and especially
GAT3. This has been done by a combination of light
and electron microscope immunocytochemistry
(Minelli and others 1995; Minelli and others 1996;
Ribak and others 1996; Gadea and López-Colomé
2001). However, these are not specific to astrocytes as
they are also found in neurons. The metabolic conse-
quences of GABA uptake can be quite varied (Haberg
and others 2001). There has been a large body of work
using the more easily studied primary astrocyte cul-
tures, but as already pointed out, these systems can be
misleading. Also, there are transporters for glycine
especially in the glycine-rich posterior regions of the
CNS. Of the 2 glycine transporters, GlyT-1 and -2,
Glyt-1 predominates in astrocytes and also seems to
have the more crucial role in lowering glycine levels in
the CNS (see Yang and Rothstein 2009 for a fuller
account of the neurotransmitter transporters in the
new book edited by Parpura and Haydon). Taurine is
also taken up by cultured astrocytes, but the role for
this transmitter, also thought to have a cell osmolyte
Na+
NT
HCO
K+
Na+
O2
CO2
H
CO2
+
Na+
K+
H2O
CO2
+
CA
Cl¯
Na+
Lactate
+ H+
+
HCO
Anoxia or
ischemia
Neuron
Astrocyte
nucleus
nucleus
mitochondrium
ATPase
Figure 3. Model of selective localization of pH and ion transporters in a mature astrocyte linking blood vessels to synaptic and neuronal
activity. Starting at the top and proceeding clockwise, the following transporters are depicted by circles with arrows to show direction of
transport: neurotransmitter (NT) + Na+ cotransporter; next 2 circles, different locations of the (Na+ + K+) pump; and then a Na+/H+ and Cl–/
HCO3
– exchanger. CA is carbonic anhydrase 2, and O2 and CO2 are shown as diffusing freely across membranes. Under conditions of
hypoxia or ischemia, a net efflux of lactate + H+ is shown coming from neurons. We now know these to be obligatorily cotransported via
a transporter (see text). Gap in astrocyte process leading to synapse indicates indeterminate distance. The increase in H+ + HCO3
– derived
from the increased CO2, facilitated within the astrocyte by C.A., was proposed to exchange for Na+ and Cl–, respectively, leading to cell
swelling. This would therefore increase under anoxic or ischemic conditions. Redrawn and modified from Kimelberg 1979.
at BROWN UNIVERSITY on November 21, 2010nro.sagepub.comDownloaded from

Functions of Mature Mammalian Astrocytes / Kimelberg 89
function in the brain because it is markedly released
upon astrocytic swelling, is not well elucidated (War-
skulat and others 2007).
Lack of transporters for control of extracellular mono-
amine transmitters. This is a good example of how an
activity clearly seen in primary astrocyte cultures
(Kimelberg and Katz 1985) has not been reproduced in
situ, in the sense that none of the well-characterized
monoamine transporters has been seen in astrocytes by
immunocytochemistry at the electron microscope level
or by a colocalization with an astrocyte marker at the
light microscope level. Examination of the supplemen-
tal table in Cahoy and others (2008) did not show any
significant levels for the different monoamine trans-
porter genes.
Control of cerebral blood flow (CBF). Astrocyte foot pro-
cesses surround all blood vessels in the brain including
the precapillary arterioles, which are the basic regula-
tors of blood flow (Reichenbach and Wolburg 2009).
Because of this, effects on blood flow were suggested
by Cajal in 1895 (Garcia-Marin and others 2007) and
have now been shown to occur (Zonta and others 2002;
Anderson and Nedergaard 2003). It is necessary for the
control to be exerted at the level of the arterioles as the
capillaries lack the smooth muscle needed to cause
contraction or relaxation to change the blood vessel
diameter (Traystman 1997; Edvinsson and others
1993). Note that there must always be some basic tone
to keep the arterioles at an intermediate diameter from
which they can constrict or relax (Traystman 1997).
The astrocytic ensheathment of CNS blood vessels
originates as blood vessels penetrate the brain paren-
chyma early in development from the arachnoid to
the brain parenchyma and carry with them the glia
limitans that consists of the endfeet of astrocytic pro-
cesses (Reichenbach and Wolburg 2009). Interactions
Table 1. Selected Transcriptosome Data from All Cell Types Expressing S-100β Promoter Driven EGFP
Activity (=Astrocytes?) and 2 Other Selected Cell Types with Different Promoter Activities
Cultured
Astro from Astro Astro Astro Astro Gray Neurons OPCs Gene
P1 Mice P1 P7-8 P17 P17 P16n P16 symbol Gene title
26,441 39,302 31,387 40,475 37,451 3411 13,873 Atp1a2 Na,K ATPase α2 polypeptide
11,386 27,557 22,812 30,532 28,898 1049 241 Gja1 Cx 43
16,201 15,463 9742 13,827 4710 31 13 Gfap Glial fibrillary acidic protein
2708 16,639 19,162 29,435 30,982 1673 195 Car2 Carbonic anhydrase 2
11,117 3167 8488 15,358 13,417 77 2197 Kcnj10 Kir4.1
5174 447 4762 21,120 23,463 4398 4350 Kcnk1 TWIK 1
13,089 22,326 22,900 24,431 20,901 195 192 Aqp 4 Aquaporin 4
4151 5422 5891 3484 2159 538 718 Pcx Pyruvate carboxylase
653 9631 15,563 19,568 24,913 440 64 Slco1c1 Solute carrier organic anion
transporter
4380 25,844 24,681 21,057 20,901 238 192 Slc1a3 GLAST
1013 1053 1252 1079 993 1722 226 Adrb1 AdrenergicRβ1
7776 19,256 19,553 29,041 26,916 452 1183 Slc4a4 Solute carrier family; anion
exchanger 4
6872 10,152 11,750 16,168 19,157 124 967 Slc9a3r1 Na/H exchanger
10,796 11,019 15,466 13,365 11,952 337 53 Aldh1L1 Aldehyde dehydrog 1,L1
804 477 1025 1170 1445 10,299 1430 Vamp2 Synaptobrevin
13,629 16,647 10,115 12,950 12,088 1203 14,167 Vamp3 Cellubrevin
351 610 521 667 783 10,409 1028 Stx1a Syntaxin 1A
36 249 295 344 277 21,916 2341 Snap25 Synaptosomal associated
protein 25
194 12,186 11,778 16,185 17,620 1594 140 Grm3 mGluR3
86 5218 3568 317 445 3384 2073 Grm5 mGlurR5
439 2174 1721 992 1361 8697 4403 Gria2 AMPA2
8672 2206 5991 12,833 10,914 8560 4636 GrinA NMDA
436 552 3500 9010 10,273 211 127 Slc1a2 GLT-1
5436 21,584 17,879 16,682 18,880 215 2232 Slc15a2 H+/peptide transporter
From supplemental Table S3a from Cahoy and others 2008. P refers to postnatal age of the animal from which the cells were obtained. The cultures are conven-
tional primary cultures from P1 mice. All astrocyte samples are from whole forebrains except for P17 gray, which is from cortical gray matter to study protoplasmic
astrocytes. They are sorted by being positive for EGFP driven by an S-100β promoter. This defines astrocytes on this basis; see text for limitations using this pro-
moter. P16n are neurons depleted of microglia, oligodendrocytes, astrocytes, and then residual endothelial cells by labeling with BSL1 lectin and panning purifica-
tion. See text for further details.
at BROWN UNIVERSITY on November 21, 2010nro.sagepub.comDownloaded from

90 The Neuroscientist / Vol. 16, No. 1, February 2010
between the astrocytic sheath and the vascular endo-
thelial cells are thought to be responsible for the forma-
tion of the interendothelial tight junctions that form
the BBB, which occurs at around the end of the first
trimester in humans (Abbott 2002). However, if the
role of the astrocyte is purely dev elopmental, why does
the vascular ensheathment persist through adulthood?
Presumably, they are then converted to physiological
roles, or a constant astrocytic influence is needed to
maintain the BBB. Clearly, the former can be the lac-
tate shuttle hypothesis (see below), control of ingress of
compounds such as glucose and amino acids or egress
of waste metabolites (Pardridge 1997), and control of
[K+]o and blood flow (Newman 1987; Filosa and others
2006). The result of these 2 different fates would be a
“naked neuron” interacting directly with blood vessels
or interactions filtered through astrocytes, as shown in
Figure 4. A reasonable view would be that the basic
reason astrocytes persist in the mature state is that they
are transducers of changes in neuronal activity affect-
ing blood vessel (arteriole) diameter and therefore flow.
It has been recently suggested that one of the major
roles of the mGluR-related increases in intracellular
[Ca2+] in astrocytes is to activate phospholipase 2 to
generate arachidonic acid and then prostaglandin 2 via
COX-1 to dilate vascular smooth muscle (Gordon and
others 2008; Koehler and others 2009). The astrocytes
can also synthesize vasodilating epoxyeicosatrienoic
acids (EETs) (Koehler and others 2009). This interpo-
sitioning of the astrocyte will also lead to it being a
major conduit for glucose and other substances from
the blood, as discussed later in the section “Astrocyte
Neuronal Lactate Shuttle Hypothesis (ANLSH) and
Metabolic Functions of Astrocytes.” However, the
effect on blood flow is exerted at the level of the arteri-
oles, so the persistence of astrocyte endfeet covering
the whole vasculature would be best rationalized as
serving as a conduit and to maintain, as well as initiate,
the endothelial tight junctions comprising the BBB.
Astrocytes could perhaps better act as transducers
summing inputs of neuronal activity via activation of
receptors to result in an output of vasodilatation or
constriction. Nevertheless, this task could also be per-
formed by neuronal axons, and there are such innerva-
tions of blood vessels with presumed activity in
regulating blood vessel diameter (Girouard and Iadecola
2006). Also, see Koehler and others (2009) for a recent
summary of the possible neuronal, smooth muscle, and
endothelial cell, as well as astrocyte, influences on vas-
cular smooth muscle tone.
The details of astrocyte regulation of blood vessel
dia meter and flow are now being worked out in both
brain slices (Gordon and others 2008) and by 2-photon
imaging in vivo (Takano and others 2006). Differences
have been observed between the 2 experimental sys-
tems, which may depend on the degree of normoxia of
the tissue versus the slices (Gordon and others 2008).
In this regard, the intact animal would appear to, by
definition, provide the normal oxygenation, with the
caveat of it being anestheti zed, although a slice, of
course, provides better opportunity for experimental
interventions. Recently, an in vivo study has shown
fine tuning of astrocyte Ca2+ responses to responses of
contiguous neurons in the visual cortex of isoflurane-
anesthetized ferrets in response to visual stimulation
(Schummers and others 2008), which could then affect
vessel diameter by Ca2+-dependent mechanisms, as
reported in the recent studies just mentioned. The
astrocyte response was much more sensitive than neu-
rons to increasing levels of isoflurane, and this was used
to show that the map of the hemodynamic response
depended on the astrocyte Ca2+ response. The fine
tuning of astrocytes to the activities of neurons will be
discussed in more detail in the section “A General Sup-
port Theory of Astrocyte Function.”
Water transport and aquaporins (AQPs). Recent work
has shown that the intramembranous particles forming
orthogonal arrays, first revealed by freeze fracture elec-
tronmicroscopy to be predominantly localized at the
perivascular membranes of astrocytes around both
capillaries and arterioles, and to be a defining charac-
teristic of these processes (Landis and Reese 1981),
contain a specific isoform of the recently discovered
water channels or aquaporins, namely AQP4 (Amiry-
Moghaddam and Ottersen 2003). This is shown in
Figure 5 by immunostaining for both AQP-4 and GFAP
from Simard and others (2003). The intramembranous
proteins that form the arrays were originally supposed
to be K+ channels, in keeping with the K+ spatial buffer-
ing or siphoning concept, where increased [K+]o due to
increased neuronal activity would be removed from the
CNS by efflux from the astrocyte to the blood (Newman
1986). This concept was criticized on the basis that it
would seem counterproductive to lose K+ as it would
need to be regained, but this could occur by reversal of
the entire process once [K+]o in the CNS returned to its
normal levels. Recent work has shown (also see section
“Exclusive Dependence of Astrocytic Membrane
Potential on K+ and the Maintenance of [K+]o”) that the
Kir4.1 channel, whose RNA (see also Table 1) and pro-
tein (Olsen and others 2005) are well expressed in
astrocytes, is also part of the assemblies (Amiry-
Moghaddam and Ottersen 2003).
The reasons for the AQP-4 channels remain
unclear. Obviously, it should increase water transport
across the astrocyte membranes and would presumably
be a major route for water transport into and out of the
brain. However, water still has to get across the endo-
thelial cells, which contain much fewer water channels
(Amiry-Moghaddam and Ottersen 2003). Also, a need
for AQPs in water transport assumes that water
at BROWN UNIVERSITY on November 21, 2010nro.sagepub.comDownloaded from

Functions of Mature Mammalian Astrocytes / Kimelberg 91
transport across the same membrane without water
channels is rate limiting. We simply do not know the
rate of water transport in the unmodified membranes of
the endothelial cells and astrocytes, although there are
suggestions that the water transport of some neurons is
very slow (Aitken and others 1998). What seems to be
generally overlooked is that unmodified phospholipid
membranes, the basic component of all biological mem-
branes, have relatively high rate constants for water
transport of 0.005 cm.s–1 (Kimelberg 2004b), and based
on the fluid mosaic model of cell membranes, might be
expected to be the one path for water transport. It may
be that this rate is not fast enough relative to the fast
transport of substrates such as glucose and K+ in the
astrocyte and is needed to avoid increases in osmotic
pre ssure inside the cell, limiting the further transport of
these substances. The AQP4 therefore allows rapid
water transport, which could then be regulated,
although it is not yet known how these channels can be
modulated.
Primary astrocyte cultures prepared from AQP-4
knockout mice have a 7-fold reduced water permeabil-
ity, an example of how these cultures can be useful.
These mice also showed neuroprotection after cerebral
ischemia and reduced edema (Amiry-Moghaddam and
Ottersen 2003). Why, you might ask, if AQP-4 is astro-
cyte specific, not use a specific inhibitor of AQP-4 to
determine an astrocyte-specific effect? The reason is
that there are, as yet, no specific inhibitors of AQPs.
Further, not knowing the magnitude of the reduced
astrocytic water transport in vivo and how this affects
brain function means that the protective effects cannot
be unequivocally attributed to the AQP-4 KO. Also, we
must take into the account the extraordinary broad
pleiotropic effects of, in this case, the knockout of Cx
43, reported by Iacobas and others (2004).
Astrocyte neuronal lactate shuttle hypothesis (ANLSH)
and metabolic functions of astrocytes. Magistretti and
colleagues proposed an important hypothesis termed
the astrocyte neuron lactate shuttle hypothesis
(ANLSH) in which glucose enters the CNS via the
astrocytic processes and is there converted by aerobic
glycolysis to lactate, which then serves as the principal
food for neurons (Magistretti and others 1994; Pellerin
and Magistretti 2003; Voutsinos-Porche and others
2003). Like regulation of blood vessel diameter and
thence flow (see section “Control of Cerebral Blood
Flow (CBF)”), the underlying morphological character-
istic was recognized since the times of Golgi (1885)
and Ramón y Cajal (1913). Namely, blood vessels in
the CNS are surrounded by astrocytic processes, and
we now know that this can be close to 100% (Virgintino
and others 1997; Reichenbach and Wolburg 2009).
This fundamental fact has given rise to many hypothe-
ses ranging from the development of the BBB due to
signals derived from the astrocytes as the CNS devel-
ops, to the ANLSH. It seems that all material that does
not diffuse between the astrocytic processes will have
to pass initially through the thin astrocytic processes
surrounding the blood vessel with the option of then
being transported out or diffusing through the entire
astrocyte. The spaces between the astrocytic processes
are of the order of a few hundreds of angstroms wide,
so this could still form a major short circuit path as the
membranes would be a major resistance pathway for
polar substances unless there are sufficient carriers on
the astrocytic membranes.
THE ASTROGLIAL CLOTHED NEURON AND
BLOOD VESSELS
THE NAKED NEURON AND BLOOD VESSELS
~60µm
Figure 4. The naked versus the astrocyte-clothed neuron. In the latter case, the astrocyte can act as a transducer and conduit between
neurons and blood vessels. The 60 µm refers to the average distance between brain capillaries (see Fig. 5), which gives rise to the Krogh
tissue cylinder concept for the brain, where diffusion of oxygen related to consumption gives rise to exponential decreases in oxygen
levels into the tissue cylinder (Lübbers 1977). See text for further discussion.
at BROWN UNIVERSITY on November 21, 2010nro.sagepub.comDownloaded from

92 The Neuroscientist / Vol. 16, No. 1, February 2010
The sequence for the ASLNH is thus the
following:
glucose → endothelial cells → astrocytes → lactate →
neuron (1).
Pellerin and Magistretti (2003) point out that this
seq uence is not obligatory, and parallel pathways can
coexist. But then, it will be very difficult to test this
hypothesis as there are no limits to the degree to which
deviations may be explained away. Further, Hertz and
others (2007) have recently emphasized that astrocytes
seem as active as neurons for oxidative metabolism.
Although one could argue with some of the basic
assumptions they make, this is still very much an open
question. They do, however, point out that the finest
astrocyte processes (also called filopodia, lamellipodia,
or peripheral astrocytic processes or PAPs; see also
Reichenbach and Wolburg 2009) are too small to con-
tain mitochondria, but these also “account for more of
the astrocytic volume than the cell body, larger pro-
cesses and endfeet.” Thus, the true oxidative metabolic
potential of single astrocytes is a mixture of distinct
compartments.
Most of the supporting observations have been
made in primary astrocyte cultures for the usual rea-
sons, but these are convincing only for the cultures and
actually will not have the constraints because of the
morphological complexity just mentioned. The primary
cultures form a flat, contact-inhibited monolayer, which
can be made to form processes, but these are fewer and
fatter than in situ. Some of the differences in gene
expression in astrocyte cultures will surely be those that
adapt the cells to a lower oxygen partial pressure due to
the limited solubility of oxygen in solution than seen at
the blood vessels, according to the Krogh tissue cylin-
der concept (Lübbers 1977). Astrocytes in situ have the
LDH isoform that favors the reductive production of
lactate, and neurons the form that favors oxidation of
lactate. However, although astrocytes have the glucose
(Glu)-1 carrier, there is also a high density of high affin-
ity Glu-3 transporters on neuronal membranes (Leino
and others 1997). The distribution of the 14 monocar-
boxylic transporters (MCT) isoforms is yet to be
Figure 5. AQP-4 and GFAP labeling of astrocytes in the rat cortex. (A) GFAP immunolabeling of astrocytes in the cortex. Individual astro-
cytes are star shaped and distributed symmetrically, with minimal contact with neighboring astrocytes. Astrocytic processes to blood
vessels differ from other processes by being straight, unbranched, and of wider diameter (red arrowheads). The surfaces of large to
medium-size vessels were densely covered by GFAP+ astrocytic endfeet. Inset, an astrocyte with 2 vascular processes. (B) Double immu-
nolabeling of AQP-4 (red) and GFAP (green). Aquaporin-4 immunolabeling reveals that the entire network of vessels, including capillaries,
is covered by astrocytic processes, but some are GFAP negative. Smaller vessels and capillaries are mostly GFAP negative but display
intense labeling against the astrocyte-specific channel AQP-4. Scale bar, 60 µm. From Simard and others 2003.
at BROWN UNIVERSITY on November 21, 2010nro.sagepub.comDownloaded from

Functions of Mature Mammalian Astrocytes / Kimelberg 93
consistently correlated with the export of lactate from
astrocytes and uptake into neurons. These are all cor-
relational studies that speak to what may happen in situ
or in vivo, and the scientific method ultimately requires
direct tests of what actually happens. These, again, are
difficult to do and therefore few and far between. The
study by Sibson and others (1998) is quoted as a direct
test in vivo, but the results can be interpreted in several
different ways.
A recent transcriptosome paper focused on enzymes
and other proteins related to those processes principally
involved in producing metabolic energy in astrocytes.
The astrocytes were labeled by pGFAP-driven GFP exp-
ression stained for GLT-1 after isolation to also cover a
large pool of cells that were not GFAP-GFP positive,
and then were FACs sorted. The pGFAP-GFP(–)/GLT-
1(+) mice had only 2-fold less of the reliable astrocyte
marker AQP-4 than the cells that were positive for both
markers, but 236-fold more than Thy1(+) neurons, as
measured by qPCR. The cells were isolated from adult
10- to 12-week-old mice using papain (Lovatt and
others 2007). It was found that astrocytes contained
transcripts for enzymes involved with glycolytic conver-
sion of glucose to lactate, particularly >13-fold more
lactic dehydrogenase b, which converts pyruvate to lac-
tate, relative to neurons. The LDHa isozyme that
principally converts lactate to pyruvate was 8-fold
enriched in neurons. This latter transcript had to be
measured by qPCR, as it was not present on the micro-
array chip used. The absence of a good probe for an
mRNA is an important factor to be taken into account
in all microarray studies. However, there were also high
amounts of all the enzymes involved in the TCA cycle in
astrocytes, and mass spectrometric measurements
showed that these cycles were active. Electron micro-
graphs were also presented that showed high density of
mitochondria in GFP(+) astrocytic processes around or
close to blood vessels. In terms of the reproducibility of
the transcriptosome studies, of which there are at pres-
ent only 2 for isolated astrocytes, the other study (Cahoy
and others 2008) showed equivalent expression of
LDHa in isolated neurons, astrocytes, and even primary
cultures. However, the LDHb form was several-fold
higher in isolated astrocytes as compared to neurons
and also quite high in the primary cultures. Plausible
differences are that the oldest mice in the Cahoy and
others (2008) study were only 17 days old, while, as just
mentioned, Lovatt and others (2007) used 10- to
12-week-old mice, or that Cahoy and others (2008)
used S100β promoter, a less selective astrocyte marker,
to label the cells.
Recently, a study in vivo and in situ in slices has
att empted to show that the ANLSH operates as pro-
posed (Erlichman and others 2008). The investigators
studied the medullary retrotrapezoid nucleus (RTN),
which contains chemosensory neurons that increase
the respiratory rate when pHe decreases. They added
4-CIN to specifically block the monocarboxylate lactate
plus H+ transporter type 2 (MCT2), held to be respon-
sible for lactate plus H+ uptake into neurons, and found
a decrease in pHe. They found that neurons increased
the uptake of glucose when 4-CIN was present. Thus, it
appears that the sequence (1) above is not obligatory,
and neurons will also take up glucose when lactate + H+
is unavailable. This does not seem surprising. Astro-
cytes can also metabolize glucose to both generate
lactate by reduction of pyruvate for export to neurons
and oxidize pyruvate to water and carbon dioxide them-
selves (Hertz and others 2007). Presumably, future
work should focus on what controls the relative activi-
ties of each path, including and in addition to the
relative availability of the 2 substrates. It needs to be
determined that any physiological division of labor has
significance more than an adaptation to morphologi-
cally imposed constraints on what parts of the astrocytes
are large enough to accommodate mitochondria.
In a recent commentary, Pellerin and Magistretti
(2003) wrote that “the important point here is not so
much to decide, based upon the actual pieces of evi-
dence, whether an hypothesis is right or wrong but
rather to point out what is heuristically valid in it, what
we have learned, what remains to be assessed, what
new hypotheses can be proposed, and which experi-
ments are critical for this.” This is reminiscent of
Newton’s comment “that hypotheses should not be the
measure of things but are useful in that they may fur-
nish experiments” (Kimelberg 2004a). The ASLNH has
performed admirably in this regard, but this is, as
always, hopefully a temporary period, and the true state
of affairs will be resolved in the not too distant future.
Antioxidant functions. The mammalian CNS is particu-
larly subject to the damaging effects of reactive oxygen
species (ROSs) because of its high rate of oxidative
metabolic activity and its high fatty acid content in the
large quantities of myelin and other membranes. The
unsaturated carbon-carbon bonds, needed to ensure
sufficient fluidity of the fatty acid side chains of the
phospholipids, are the ones most susceptible to oxida-
tive damage by ROSs (Halliwell and Gutteridge 1985,
1999). ROSs are the unpaired electron versions of
atomic and molecular oxygen that cause the breakdown
of a large number of lipid and proteins in an autocata-
lytic manner. Because the full reduction of molecular
oxygen by the respiratory chain will never be complete,
these ROSs need to be neutralized, and antioxidants
and enzymes such as catalase and peroxidases are pres-
ent to neutralize these. However, as usual, under
pathological conditions, these protective pathways are
overwhelmed, and ROS-induced damage becomes a
major source of cellular injury, for example, in the
reperfusion phase of cerebral ischemia.
at BROWN UNIVERSITY on November 21, 2010nro.sagepub.comDownloaded from

94 The Neuroscientist / Vol. 16, No. 1, February 2010
Astrocytes have a number of antioxidant systems
such as the GSSG-GSH system (Aschner 2000; Ander-
son and Nedergaard 2003). Isolated astrocytes also
contain messages for the enzymes superoxide dismutase
(Sod) and catalase (Cat), but these are not specifically
enriched in astrocytes but are found in neurons at rea-
sonably high levels and other neural cells too (Cahoy
and others 2008). However, the astrocytes clearly show
high levels, which can function both to limit their own
production of ROSs and also take care of those pro-
duced by other cells. It may, actually, be more of the
former as these species are short lived and therefore
will have limited diffusion path lengths.
Muller cells as living “light guides.” This is an example
of a recently discovered function of the Muller astro-
cytes of the retina in addition to a number of more
general astrocyte properties described for these cells,
such as glutamate and K+ uptake and pH control. As
pointed out in the paper describing this phenomenon
(Franze and others 2007), the histology of the mam-
malian eye has always been paradoxical in that the
photoreceptors are located at the back of the retina,
where the light’s transmission would be degraded by
scattering. The Muller cells’ orientation and low scat-
tering make them able to conduct the light like optical
fibers to the interior of the retina so that it falls on the
photoreceptors with less degradation (see rightmost
image in Fig. 1C).
This property can be best viewed as a support role,
allowing light to reach the photoreceptors, where it
starts the complex process of being transformed to
vision. It is a vital role given the structure of the retina
(why the retina is built this way is another question)
and would seem to be best described as facilitative; that
is, helps the eye’s neuronal photoreceptors do their job
of performing the initial step of translating light into
vision.
This finding also raises another question as to how
we do science in terms of obtaining direct evidence for
this property. This is not the first description of such a
function for living cells, but it is for the Muller cells. It
is, at present, a convincing proposal because it uses
their ordered structure, which reduces light scattering,
to clarify an old puzzle, but more direct evidence will be
required to firmly establish this function. However, the
light focusing and conducting property depends on the
total structure of the cell, and it is not clear that this
property can be interfered with without affecting any-
thing else. Would it be possible to make these cells
opaque without interfering with their other functions?
This will need a conditional genetic change because if
the cells are opaque throughout the development of the
retina, it may well have secondary effects on the devel-
opment of the photoreceptors and other components
that will affect the light-sensing properties of the retina
in other ways. Removing the Muller cells would be even
more drastic and will likely have a wide spectrum of
developmental and physiological effects. So, the defini-
tive experiments are difficult to envisage.
Modulation and control of synaptic activity as in the
tripartite synapse concept. Recently, a role has also been
proposed for astrocytes that could perhaps be best
described as an “information-conveying, instructional”
role. The concept is based again on the long known fact
that the ends of astrocyte processes surround synapses.
It proposes exocytotic release of glutamate or ATP from
these astrocyte processes, a process referred to as “glio-
transmission.” As usual, the initial supporting data
were from primary astrocyte cultures (Parpura and oth-
ers 1994), and cultures are still used to analyze details
(Montana and others 2004). Diligent work has now
established some evidence for vesicular release from
astrocytes in situ influencing synaptic activity (Zhang
and others 2004). The major technique used to show
exocytotic release of a transmitter from astrocytes in
vivo is a genetic engineering approach involving inhib-
iting exocytotic release from astrocytes by overexpres-
sion of the vesicular SNARE protein, synaptobrevin 2
(VAMP2), targeted specifically to astrocytes via the
GFAP promoter (Pascual and others 2005; Halassa and
others 2009; Fiacco and others 2009). There is the
question of whether the genetic approach will work in
all astrocytes, such as in the GFAP(–) ones in the cor-
tex (the definition problem again). The overall concept
has come to be known as the tripartite synapse concept
with the perisynaptic astrocytic process as the third
active member of the triumvirate (Araque and others
1999).
The data to date provide convincing evidence that
membrane fusion–related release of glutamate from
astrocytes can be obtained under certain conditions,
leading to slow inward currents and increased sEPSC
frequency in granule cell neurons in the dentate of the
rat hippocampus, or induction of a slow inward current
that increases action potential firing in the target
neuron in the so-called up state (i.e., a depolarized neu-
ronal Vm; see Fig. 2) in the nucleus accumbens. In both
cases, the effect was attributed to glutamate from astro-
cytes acting on NR2B subunit–containing extrasynaptic
NMDA receptors (D’Ascenzo and others 2007; Fellin
and others 2004). Currently, there is more emphasis on
release of ATP with its conversion to adenosine in the
extracellular space (Halassa and others 2009). How-
ever, although del Castillo and Katz first proposed the
vesicular hypothesis in 1957 (Katz 1966), individual
quantal vesicular release at the NMJ was only generally
accepted as the cause of mepps about 20 years later,
after extensive corroborative electrophysiological and
EM work (Kuffler and Nicholls 1976). It has been 15
years or so since gliotransmission by astrocytes was first
at BROWN UNIVERSITY on November 21, 2010nro.sagepub.comDownloaded from

Functions of Mature Mammalian Astrocytes / Kimelberg 95
proposed, and the pace of research today is faster, but
its confirmation has to be in a more complex system
than the NMJ. What evidence still needs to be found?
As shown in Figure 6 taken from a recent review (Agul-
hon and others 2008), the existence of GPCR-mediated
Ca2+ increases in astrocytes has by now been convinc-
ingly shown, first in cultures and then in situ. The
intracellular Ca2+ increase is not surprising because this
is what these G protein–coupled receptors do. What is
yet to be established, as depicted in the right panel of
Figure 6, are the consequences of this widespread cell
response in astrocytes and whether they occur under
physiological conditions in the mature mammalian
brain. One consequence proposed is the release of
transmitters by the fusion of transmitter-filled vesicles
with the astrocyte membrane. One problem with “glio-
transmission” as a consequence is the generality of the
membrane fusion process. “Life in eukaryotes depends
on the fusion of membraneous organelles. Every vital
process depends on the orderly execution of membrane
fusion . . . . SNARE and SM proteins have long been
known to be required for fusion” (Sudhof and Rothman
2009). Therefore, how does one determine that only
the release of transmitter-filled vesicles is being tar-
geted? There is now evidence that, in mouse primary
astrocyte cultures, the Ca2+-dependent release is from
SNARE protein–dependent fusion of lysosomes with
the plasma membrane (Zhang and others 2007). Lyso-
somes contain a plethora of compounds, including ATP
and adenosine that figure prominently in the com-
pounds measured or whose receptors are responsible
for measured activity, after conditions proposed to lead
to vesicular rel ease are instituted. Therefore, it seems
that the Ca2+-dependent release phenomena, studied
first in cultures and then in situ through genetic engi-
neering, could be artifactual because it is difficult to
imagine a physiological release process due to lyso-
somal exocytosis. There are also other concerns. Release
in cultures is produced by mechanical stimulation but
not by mGLUR activation (Li and others 2008). A
genetically engineered astrocyte-specific Ca2+-stimulat-
ing receptor does not cause any neuronal changes when
stimulated in vivo by an agonist not present in the CNS
(Fiacco and others 2007). Let us also not lose sight of
the fact that there appears to be 10 times less vesicle
density in astrocytes compared to presynaptic neuronal
terminals (Jourdain and others 2007), so the magnitude
of the response is likely to be much less and therefore
more difficult to study.
Recently, Haydon and his colleagues have shown
cha nges in the dominant negative (dn) synaptobrevin 2
transgenic animals for LTP (Pascual and others 2005),
and an aspect of the EEG during anesthetized-induced
sleep termed sleep pressure, but not time of sleep (Hal-
assa and others 2009). The intervening steps up to such
complex behavioral changes will need to be clarified to
understand why what occurs does occur. Numerous
controls will need to be done to avoid purely nonglio-
transmission effects. A number were done in the most
recent study (Halassa and others 2009), but space
limitations do not allow a point-by-point analysis of
the details, especially as what constitutes an adequate
number of controls is always debatable. Whether,
overall, the controls are convincing is a matter of peer
review, and time, as usual, will determine whether
they are truly adequate. Reassuringly, the sleep effect
was stated to be reversible, which could be done
because the tet-off system was used. This argues a
developmental effect, which would seem difficult to
reverse.
What is the gliotransmission/tripartite synapse con-
cept really saying in terms of how we view the way
synapses behave? It is very difficult to distinguish
between changes due to effects on supportive functions
or a direct instructional role for astrocytes. For exam-
ple, glutamate release versus uptake: If astrocytic
uptake of glutamate released at synapses is reduced in
a test situation, it would be the same as increased
release from astrocytes. This was indeed shown in the
rat hypothalamic supraoptic nucleus (SON) (Oliet and
others 2001). In virgin rats, astrocyte coverage of neu-
rons is extensive but retracts in lactation, resulting in
reduced uptake of glutamate. Changes in the EPSCs
corresponded to the expected changes in [glutamate]o.
Conceptually, how can we distinguish between
astrocy tes affecting the circuits involved in LTP in a
manner that is critical to whether the change occurs or
allowing neuronal circuits to generate LTP behavior or
sleep? In the latter case, the astrocyte involvement
would seem to be permissive. This is a difficult distinc-
tion to make, and in the latter case, I would take this to
mean that the astrocyte modulates the process taking
place but in no way initiates it. Alternatively, the role of
any transmitter released from astrocytes could be in
response to changed synaptic activity to produce a
homeostatic response in the neuron to avoid some
excessive change that might prevent the synapse doing
what it had been instructed to do as part of a neuronal
circuit. The same reasoning would apply to the knock-
out of, say, H+ transporting systems such as Cl–/
HCO3
– or Na+/H+ exchangers involved in maintaining
[H+]o because of its tendency to increase because of
neuronal activity (Chesler 2003). Could we logically
conclude from such observations that astrocytes have
an instructional role in information processing? This
does become a question of semantics, but precisely
stating what we mean, or think we mean, is essential for
sense.
There is also a lot of evidence for alternative path-
ways for transmitter release from astrocytes via
exocytosis of a dedicated population of transmitter-
filled vesicles. Most of these observations, for practical
at BROWN UNIVERSITY on November 21, 2010nro.sagepub.comDownloaded from

96 The Neuroscientist / Vol. 16, No. 1, February 2010
reasons, have been made in primary astrocyte cultures,
which while providing a seductive experimental sim-
plicity are (to repeat this point once again) not reliable
indicators of what happens in situ. So far, these alterna-
tive paths are release by reversal of EAA transporters,
likely only occurring under pathological conditions,
swelling-activated anion channels also likely only rele-
vant in pathology, hemichannels, and P2X receptors for
ATP. See recent papers in a special issue of Glia (Deit-
mer and others 2006) for discussions of these different
routes. As already noted, some examples of fusion-
dependent release may be of lysosomes. Clearly, all
these different pathways will have to be subjected to the
same criteria for establishing their functional relevance,
and if astrocytes modulate synaptic activity in the
normal course of events by transmitter release, why
should these other routes not be involved as well as
vesicular release unless they are only pathological?
Also, the question of how the transmitter is normally
released from astrocytes has to be put into the context
that the amount and speed seen for neuronal synaptic
release are likely not required. It also would seem that
there should only be one main route; otherwise, one
might expect a lot of noise in the process.
I have taken quite a bit of space to deal with glio-
transmission as an astrocyte function because it has
been a major topic in the astrocyte field for the past
decade or so, with the tantalizing aim of showing astro-
cytes would be directly involved in information processing
and therefore are not “just support cells” (Volterra and
Meldolesi 2005). However, the need for this activity, that
is, a theoretical framework comparable to excitable neu-
ronal circuits, in which this idea would fit, has not yet
been made explicit. We have cells that both take up glu-
tamate and also release it in a physiologically regulated
way, and inhibition of the former and activation of the
latter would both increase [glutamate]o around the syn-
apse. What would astrocytic modulation/control add to
synaptic transmission and neuronal circuits? One plau-
sible idea would be to integrate different inputs onto the
astrocyte, which would provide some extra control and
integration; otherwise, feedback control from released
transmitters acting on presynaptic receptors to inhibit
release would seem adequate. This was, in fact, first pro-
posed by Winder and others (1996), and I reproduce
their Figure 8 as Figure 7 to illustrate the concept. The
evidence for glial (astroglial) involvement was based on
inhibition of the cAMP levels and its effects by an astro-
cyte toxin in hippocampal slices. Some of the proposals
are dated as noted in the legend. However, the overall
concept is similar to recent proposed examples of ATP
release modulating synaptic events in the hippocampus
according to the tripartite synapse hypothesis (Pascual
and others 2005; Halassa and others 2009). The major
difference is that in the later studies the release is pro-
posed to be via regulated exocytosis.
Variation in astrocyte ensheathment affects neuronal
outputs. One additional function of astrocytes implied
in the preceding discussions is that they surround and
Figure 6. Illustrates well-established increases in intracellular Ca2+ in a mature perisynaptic astrocyte due to GPCR activation by spillover
of transmitters such as glutamate for mGluRs (left), contrasted with the unresolved questions of the consequences of the rise in astrocyte
Ca2+, as represented on the right. From Agulhon and others 2008.
at BROWN UNIVERSITY on November 21, 2010nro.sagepub.comDownloaded from

Functions of Mature Mammalian Astrocytes / Kimelberg 97
form limiting boundaries between individual neurons,
groups of neurons, and for the entire brain itself such
as in the glial limitans covering the surface of the brain.
In certain regions, it has been shown that retraction of
such ensheathment leads to increased interaction
between formerly astrocyte process–separated neuro-
nal surface membranes. This can also lead to the
removal of diffusion barriers facilitating the release of
neuronal secretions. One well-studied system is the
supraoptic nucleus, where the large oxytocin and
vasopressin-releasing neurons are surrounded by astro-
cytes that retract at the end of pregnancy to induce
lactation. Axons of these neurons terminate in the pitu-
itary, where what is considered astroglial homologs, the
pituicytes, also change shape to control the release of
OX and VP (Hatton 1997). These could be viewed as a
second-order function in which removal of glial pro-
cesses allows otherwise impeded effects to be pro-
longed, such as persistence of neuron-released
glutamate in the ECS, leading overall to an increase in
neuronal excitability. These are longer terms effects
occurring over minutes and hours.
Recent Transcriptosome Data for Astrocytes
In Table 1, I show some of the messages for gene mRNA
products from the more general of the 2 recent pioneer-
ing studies on gene expression mentioned earlier to
illustrate the amazing amount of unbiased data that can
be obtained from the microarray approach for astro-
cytes to compare to other neural cells, as was done in
Cahoy and others (2008). The messages I have selected
are those that I think are thought to be important in
astrocytes. The numbers represent the relative normal-
ized fluorescent intensities above an excluded baseline.
However, there seems no way of telling whether a gene
product is not listed because its levels fell below the
baseline criteria or whether the appropriate probes
were not on the chip. Also, all probes are not equally
efficient in binding to the same mRNA, so it was
deemed that 3 different probes for each mRNA were
needed at a minimum. To make these measurements,
populations of one type of acutely dissociated cells from
the mouse forebrain were depleted of other cell types by
immunopanning. Astrocytes were further purified by
FACs sorting using S100Δ-EGFP. In all, 12,416 unique
genes were identified (see supplementary information
in Cahoy and others 2008 for details). Another recent
microarray study focused on certain genes associated
with intermediary metabolism (Lovatt and others
2007), which has been useful in the topic on the lactate
shuttle and met abolic functions as discussed in the sec-
tion “Astrocyte Neuronal Lactate Shutte Hypothesis
(ANLSH) and Metabolic Functions of Astrocytes.”
These data gain considerable interest from prior
studies on astrocytes but also give interesting new
emphases. Extensive “mining” of the messages from the
original table can stimulate novel ideas and new per-
spectives. To note, a few interesting things: there are
numerous representatives of the different solute carrier
(Slc) families, from glutamate transporters to bicarbon-
ate anion exchange, proton sodium exchange, and also
high levels of carbonic anhydrase 2 (Car2). The high
level of expression of Car2 is of interest to me. There
was an old question as to whether carbonic anhydrase 2
was a marker for oligodendrocytes or also in astrocytes,
as it was found in primary astrocyte cultures (Kimelberg
and others 1982) and in astrocytes as well as oligoden-
drocytes in situ by immunocytochemistry (Cammer and
Zhang 1991). From the table, this message (gene
symbol Car2) can be seen to be high in isolated astro-
cytes and about 10-fold lower in primary astrocyte
cultures. Not shown in Table 1 is that Car2 is almost as
high (24,030) in myelinating oligodendrocytes and
lower in nonmyelinating oligodendrocytes (4376), as
previously found for the protein by immunocytochemis-
try in situ (Cammer and Zhang 1991). Thus, the
question of whether Car2 was in astrocytes or present
in the cultures due to oligodendrocyte contamination is
now better resolved, so the old controversy seems set-
tled. This, together with the bicarbonate chloride
exchangers and proton sodium exchanger, fits the
model of coupled exchange accelerated by Car2 that we
originally proposed (Kimelberg and others 1979), which
could explain bicarbonate-dependent increased Na+
i
and astrocytic swelling. An important novel finding was
that the high and astrocyte-specific expression of
Aldh1L1 (aldehyde dehydrogenase 1 family, member L
1) message, which was also supported by immunocyto-
chemistry presented in the paper, is a more general
marker for all astrocytes than GFAP.
As expected from basic cell physiology, there is a
ubiquitous expression of the Na+, K+ ATPase (Na+
pump), and the complete table contains a number of
other general housekeeping proteins generally distrib-
uted across all cell types. There is high message
expression for Kir and the 2 pore domain channel
TWIK. As expected from studies on astrocytes in situ to
date, there is expression of AQP4 and GFAP messages
selectively in astrocytes. The GFAP message in gray
matter P17 astrocytes is 3-fold lower than for the mes-
sage from the P17 mixed white-matter/gray-matter
forebrain in the preceding column. It has long been
known from immunocytochemistry (Bignami and Dahl
1974) that gray matter (protoplasmic) astrocytes far
more weakly for GFAP than do white matter astrocytes
(fibrous astrocytes). This is a remarkable example of the
reliability of the microarray system and how mRNA can
correspond to protein. There is high mGluR3, except in
the cultured astrocytes, which differed in a number of
other aspects. High expression of only mGlur3, and to a
lesser extent 5, message was found by us by single cell
at BROWN UNIVERSITY on November 21, 2010nro.sagepub.comDownloaded from

98 The Neuroscientist / Vol. 16, No. 1, February 2010
RT-PCR on acutely isolated astrocytes (Schools and
Kimelberg 1999). The data on transmitter storing vesi-
cles, and fusion proteins for their exocytosis, are mixed,
but the authors concluded that “the astrocyte transcrip-
tosome, however, does not include mRNAs for vesicle
glutamate transporters or synaptic proteins including
Vglut, Vglut2, synapsin 1, and synaptotagmin found in
neurons.” It seems that the vesicle RSNARE protein
cellubrevin (Vamp3), but not synaptobrevin (Vamp2), is
an exception, but it needs a number of other proteins to
effect fusion (Scales and others 2000; Jahn and Scheller
2006; Sudhof and Rothman 2009) and at present does
not support using the dn synaptobrevin 2 transgenics
for in vivo gliotransmission studies, as discussed in the
section “Modulation and Control of Synaptic Activity as
in the Tripartite Synapse Concept.”
A General Support Theory of Astrocyte
Function
What would a current general support theory of
astrocyte function look like? Bushong and others
(2002) emphasized in their combined immunohisto-
chemical and dye-filling studies in the CA1 region of
1-month-old Sprague-Dawley rats that the dye-filled
astrocytes were morphologically homogenous and their
highly bushy processes occupied separate territories of
approximately 70,000 µm3. This is an important finding
and concept. Note that they cut 100-µm sections from
the brains of paraformaldehyde-perfused rats, presum-
ably to avoid rapid postmortem, prefixation changes, to
avoid dye mixing by cell-cell coupling. In a later study
(Bushong and others 2004), they showed that these
features develop with age and resemble the mature
form at ≥PN 21. This is also after the glutamatergic
synaptic system matures (~PN14) and corresponds
with when we see the development of the purely elec-
trophysiologically passive astrocyte in the CA1 (Zhou
and others 2006). Related events seem to be going on
here. They did not do any recordings from their astro-
cytes because they were fixed.
Each astrocyte soma can have up to 100,000 pro-
cess endings, as estimated by the volume occupied by
the entire astrocyte domain and the number of synapses
in this volume, assuming one process ending contacts
each synapse. Much fewer processes contact blood ves-
sels. The process endings of separate astrocytes overlap
only about 10%, and that gap junction and synaptic and
Figure 7. Proposed integrative roles for perisynaptic astrocyte modulation of synaptic activity at a hippocampal glutamatergic synapse
(from Winder and Conn 1996). Both spillover activation of the mGluRs on astrocytes and activation of adrenergic β receptors by release of
catecholamines from terminals of another neuron were proposed to produce maximum intracellular levels of cAMP for release and its
subsequent conversion to adenosine in the ECS to act on the presynaptic adenosine A1 receptor. The representation of the astrocyte is
quite diagrammatic and should represent the end of a perisynaptic astrocytic process. In this drawing, the adrenergic input seems depicted
as a synapse, but there is no evidence for these on astrocytes. Actually, catecholaminergic neurons are known not to have discrete post-
synaptic specializations but release transmitter from varicosities and are considered to produce a more diffuse transmission through the
ECS, as in the concept of “volume transmission” (Descarries and others 1991). However, to my knowledge, this represents the first pos-
tulation of an effect on synaptic activity by release of a neuroactive agent with some experimental support and further proposed integration
of >1 input.
at BROWN UNIVERSITY on November 21, 2010nro.sagepub.comDownloaded from

Functions of Mature Mammalian Astrocytes / Kimelberg 99
domain maturity all occur at the same time in the hip-
pocampus, (Bushong and others 2002, 2004) can be
incorporated into a general theory of astrocyte function
in the mature brain, at least in the hippocampus (Fig.
8). The vascular and synaptic endfeet will function as
autonomous units responding to specific events at these
loci, as independent entities driven by local feedback
signals. Experiments supporting independent domains
for intracellular [Ca2+] increases were proposed a few
years back and experimentally supported for the Berg-
mann glia lamellae and filipodia enwrapping the
synapses that the Purkinje dendritic spines make with
the parallel fibers (Grosche and others 1999). Also, see
Reichenbach and Wolburg (2009) for a discussion on
microdomains to macrodomains based on current
detailed morphological understanding. This then raises
the question of how astrocytes specialize for their syn-
aptic versus vascular properties, and the simplest
hypothesis would be segregation of functional mem-
brane proteins within the plasma membrane with
presumably different transporters, channels, and other
membrane protein classes on astrocyte endfeet, lamel-
lipodia, and filipodia contacting blood vessels and
synapses.
In terms of synchronizing or otherwise integrating
neuronal activity, there has also been no demonstration
of any polarity in regard to these populous processes, so
it would require that a signal from one or several syn-
apses would be transmitted randomly to all other
synapses touched by processes of the same cell, limited
by the rate of diffusion of an intracellular messenger or
the electrotonic decay of a membrane potential depo-
larization. However, it has also been suggested that
each astrocytic domain integrates all the synaptic activ-
ity within it with vascular activity (Nedergaard and
others 2003), and this could certainly occur within one
astrocytic domain with a sufficiently large intracellular
signal that could spread to all the processes of one cell,
especially the larger diameter processes that go to the
arterioles. Such a process seems to have been recently
shown in the visual cortex of the isoflurane-anesthe-
tized ferret, where Ca2+ increases in single astrocyte
soma closely followed the responses in neighboring
neuronal soma (Schummers and others 2008). Almost
simultaneously, another group reported that locomotor
activity in the awake mouse caused hundreds of astro-
cytes in the cerebellum to increase their Ca2+ levels
(Nimmerjahn and others 2009). Smaller spontaneous
res ponses and artificially stimulated locomotor
responses were markedly inhibited by isoflurane anes-
thesia. Thus, the full extent of the astrocytic Ca2+
responses may only be revealed in the absence of anes-
thesia, and activation can be very extensive, presumably
when a large field of neurons is activated to elicit a
widespread increase in intracellular astrocyte Ca2+
levels leading to a generalized vasodilation. As noted in
relation to Figure 4, is this the major function of mature
astrocytes? Is this Ockham’s grail?
The role of the gap junctions between touching
astrocyte process tips could, at the minimum, be to mark
the domain, although slow diffusion events such as the
diffusion of glucose, changing in response to increased
neuronal activity, have been shown to occur (Rouach
and others 2008). This diffusion will be increased if
there is segregation of glucose transporters with concen-
tration at, say, the perivascular endfeet where the glucose
initially enters and less on the processes and astrocyte
soma. This limits the leakage of glucose out of the astro-
cyte nonspecifically, a problem that seems associated
with the K+ spatial buffering hypotheses, as dealt with in
the section “Ionic or Metabotropic Receptors on Astro-
cytes and Astrocyte-Neuron Signaling.” In keeping with
the ANLSH, we might anticipate the astrocytic form of
the lactate transporter to be quite widely distributed
along the astrocyte membranes because it is proposed
that it supplies lactate generally to neurons. However,
here, as in other roles for astrocytes, the segregation of
transporters and channels can be very complex because
of the extensive branching of the processes. As previously
noted, there have been studies showing this for a few
membrane proteins, but the issue has yet to be addressed
systematically and for complete clarity will need to be
done at the EM level.
This support model of the mature mammalian
astrocyte is well grounded in cell and biological princi-
ples. The autonomous ends of the many processes
cannot exist independently of a cell nucleus (classic cell
theory) to synthesize and export to the processes’
replacement proteins or new ones because of a change
in conditions. The only enucleated cells in mammals
are the red blood cells that have a limited life span of
approximately 1 month! For the CNS, the cells of the
lens are also enucleated but form a crystalline structure
and are longer lived (Franze and others 2007). Such
transport will need a transporting system to carry pro-
teins and remove them for degradation in lysozymes of
the cell body, or partly they may be degraded in lyzo-
somes present in the process ends. Because especially
the ends of the perisynaptic processes (fine lamellae
and filipodia that arise from the processes emanating
from the astrocytic cell body) are too small to contain
mitochondria (Hertz and others 2007), some of the
ATP needed will have to diffuse from mitochondria in
the soma and larger diameter processes more proximal
to the soma. Alternatively, the energy requirements at
the process ends could be met completely by glycolysis,
which is much less efficient. However, a recent paper
by Lovatt and others (2007), in which GFAP promoter-
driven GFP was used to identify the small astrocyte
perisynaptic endfeet, found that the number of mito-
chondria in the perivascular endfeet was as great or
greater than other profiles in the surrounding neuropil.
at BROWN UNIVERSITY on November 21, 2010nro.sagepub.comDownloaded from

100 The Neuroscientist / Vol. 16, No. 1, February 2010
However, Chao and others (2002), using electron
microscopy and S100Δ-immunoreactivity to identify the
processes, had previously noted that the fine lamellae
and filipodia were, unlike the astrocytic cell body and
thicker proximal processes, devoid of organelles and
cytoskeletal elements. Therefore, whe ther the endfeet
supply sufficient energy by oxidative metabolism for
their needs, and this would diminish the amount of lac-
tate available for export to neurons, as envisaged in the
ANLSH, remains an open question.
The other general biological principle is economy
of space. Up to 100,000 synaptic targets, and a fewer
number of blood vessels, can be served by one cell soma
(Fig. 8) and thus conserve space, which in the mam-
malian brains is at a premium and has also, for example,
led to myelination to reduce axonal diameters to con-
serve space for a given action potential velocity (Hille
1992). One could, of course, have more astrocytic cell
bodies with fewer processes, for the astrocytic soma are
quite small, approximately 10 µm in diameter. Presum-
ably, the packing that has evolved is optimal. Very
recently, Oberheim and others (2009) have shown that
the domain of protoplasmic astrocytes in the human
cortex is 2.6-fold larger than in rodents. At the mini-
mum, this may be the optimal packing relation for
astrocytes in the much larger human brain and would
therefore be consistent with the same support roles as
in the rodent rather than related to the greater behav-
ioral complexity produced by the human brain as they
suggested. The cell bodies of the human astrocytes
appear to be the same as the rodent, approximately 10
µm diameter, and thus, one would expect a lower den-
sity of astrocytic soma in the human brain compared to
the rodent. One problem with this study is that the
descriptions were based not on filling the cells with dye,
as was done by Bushong and others (2004) for their
initial “domain” studies, but on GFAP staining, which,
as noted before, stains only a small proportion of the
processes of rodent astrocytes. One reason the human
cortical astrocytes might stain more extensively for
GFAP is that, of course, the human tissue was obtained
surgically, in this case, from epileptic subjects. Thus,
Glutamate uptake and
conversion to glutamine. Also
release of glut, ATP by
exocytosis(?). Specific
localizations of other transport
systems?
1. Glucose uptake
2. Release of H+, K+ to blood
3. Known channels; Kir4.1, AQP4,
4. pH and other transporters?
Soma; Protein synthesis
and energy production.
Very negative
membrane potentials
are measured here.
Gap Junctions. Cx43,30.
Astro-astro communication.
Movement of K+ Ca2+ IP3
etc. Dilution of taken up
substances.
1,000 - 10,000
total processes
Figure 8. Model of mature mammalian protoplasmic astrocyte showing major relations and sites of functions for astrocytes (see text). There
is as yet no evidence that any of the populous processes (1000 to 10,000), mainly to synapses with fewer and larger diameter ones to
blood vessels or to gap junctions, have any polarity, both morphologically or functionally, beyond these different targets (see Fig. 5 legend),
as would be expected if a change at one or a few synapses could specifically affect the activity of other synapses through signals con-
ducted through a single astrocytic cell or a syncytium (see Kimelberg 2007). With stronger stimuli, one could envisage a single astrocyte
or an entire syncytium if an entire neuronal field is involved (Nimmerjahn and others 2009) as a transducer for activity to become a vascu-
lar input with the synaptic processes being the inputs and the fewer and larger processes to mainly arterioles being the outputs, to increase
blood flow and maintain the gradients for increased supply of oxygen and glucose according to the Krogh vascular supply cylinder concept
(Lübbers 1977). This will likewise facilitate increased efflux of deleterious metabolites. Modified from Kimelberg 2007.
at BROWN UNIVERSITY on November 21, 2010nro.sagepub.comDownloaded from

Functions of Mature Mammalian Astrocytes / Kimelberg 101
there is no way of knowing to what extent the human
astrocytes were mildly reactive.
In summary, the domain concept for populous, pro-
cess-bearing mature astrocytes seems more consistent
with autonomous homeostatic endfeet functions rather
than integrative activity for neuronal networks and has
arisen for simpler biological reasons. The next question
seems to be to determine, in a systematic way, the
nature and location of different transporters and chan-
nel proteins in the membranes based on well-established
pro perties plus the new trasncriptosome data rather
than to more randomly guess various functions and
then try to experimentally support them.
Conclusions
Current views of the functions and roles of mature
astrocytes range from long and well-established support
roles, such as uptake of glutamate at synapses and the
surprisingly still unclarified role of control of [K+]o,
through recent proposals that they function to modu-
late and control synaptic activity through regulated
exocytotic release of transmitters such as glutamate and
ATP (Pascual and others 2005; Haydon and Carmi-
gnoto 2006), thus adding a further component to the
already difficult and unresol ved problem of relating
neuronal circuitry and synaptic activity to information
processing and cognitive and emotional behavior. The
evidence supporting the latter view is still in its prelimi-
nary phase, and some of the observations used to
support the concept are now in dispute, as discussed in
the section “Modulation and Control of Synaptic Activ-
ity as in the Tripartite Synapse Concept.” Otherwise,
many of the current studies are applying modern tech-
niques to establish the mechanisms of functions first
put forward a hundred years or so ago (see section
“Where Do Our Hypotheses of Astrocyte Function
Come from and How Do We Test Them?”).
As already noted several times, because of their
extraordinarily low Rm, mature astrocytes are very dif-
ficult to study in relation to a likely highly complex
pattern of localization of different channels in the plane
of the membrane. This will also apply to electrogenic
transporters that will also be difficult to study function-
ally by electrophysiology because of the small effects
seen for charge transfer across the low Rm membranes.
These cells may have what is viewed as “simple” support
functions, but because of their low resistance and very
numerous processes, they are immensely difficult to
study. We therefore have the very unsatisfactory situa-
tion of absence of evidence not being evidence of
absence. However, in the scientific method, we can
only use evidence of presence to make advances.
To distinguish between supportive and more active
roles is difficult, and the problem becomes one of logic
and philosophical argument in defining these as
opposing views. This distinction has also recently been
discussed by Tekkök and Ransom (2004). They suggest
that the distinction between a “homeostatic” and a “sig-
naling” interaction can be based on whether substances
like K+ or glutamate are only taken up or whether they
also convey “information” to astrocytes. On this basis,
astrocytes are certainly involved in information transfer,
but this seems to me far more simple than what is meant
by information transfer in the brain. Information is con-
veyed to all cells in the body via receptors and other
means so they can respond to varying conditions to
effect responses essentially involved in the Claude Ber-
nard concept of maintaining “la fixité du milieu
intérieur,” an enlightening, basic physiological concept
that was later given a Greek-derived name, homeostasis.
Many scientists seem to make a distinction between
philosophy and doing science and seem far more com-
fortable only making increasingly complicated
measurements, utilizing the latest techniques and
equipment, to show possible relationships rather than
making some reasoned judgment as to what those rela-
tionships really are in terms of the larger physiological
picture. They always propose that, from the detailed,
the general will emerge, but does this justify any and all
detail, and to what extent? This distinction between
making measurements and thinking about what they
signify was made some years ago in the foreword to the
posthumously published book by David Keilin, the dis-
coverer of the cytochrome respiratory chain in 1925
(Keilin 1925) and the PhD mentor of my PhD mentor,
Peter Nicholls (see below). “Nowadays, an increasingly
large number of papers are published on intracellular
respiration. There are several reasons for this; for exam-
ple (1) the advent of recording spectrophotometers,
which can be manipulated by technical assistants, of
high-speed centrifuges and of apparatus for moving-
boundary and paper electrophoresis and for
chromatography; (2) the commercial availability of
highly purified components of active biological systems,
such as enzymes, co-enzymes, proteins, peptides and
amino-acids; (3) the existence of numerous up-to-date
reviews of the subject and symposia which introduce
the young research worker to the latest developments
and the jargon which describes them. All this enables
him to acquire fundamental knowledge in any particu-
lar field quickly and to make useful contributions by
finding new facts. However, the great complexity of
living cells and organisms leaves ample opportunity for
those not provided with the most modern facilities to
make fundamental contributions by postulating new
problems and discovering new paths of approach. There
will always be room for such individual workers how-
ever highly organized research becomes” (Keilin 1966).
How much have such changes grown in the 50
years or so since these lines were penned? Keilin was
very much a hands-on scientist, so his concerns must
at BROWN UNIVERSITY on November 21, 2010nro.sagepub.comDownloaded from

102 The Neuroscientist / Vol. 16, No. 1, February 2010
be viewed seriously and because they are magnified
many fold today. His recording spectrophotometers,
purified reagents, and reviews have now expanded to
many more very expensive new types of equipment from
sequencers to 2-photon confocal microscopes, ready-
to-use kits, and recipes for the most complex reactions,
to PubMed databases and a continuous calendar of
meetings. Yet, because of the great complexity of living
cells, such as astrocytes, their function still eludes us,
and we need fruitful concepts as much as we need new
measurements.
As already noted several times, the distinguishing
feature of science, in distinction to all the other
branches of philosophy, is that postulates are tested not
only by what seems reasonable and logical in terms of
the existing database but also whether they conform to
new realities implied by specific hypotheses. Whether
the new studies, in the light of current knowledge, do
conform to, and only to, the specific hypothesis postu-
lated, or whether they are consistent with other
reasonable viewpoints, is at the heart of the interpreta-
tive aspect of the scientific method. This predictive
element of the scientific method is critical, for it is far
easier, and far less convincing, to retrospectively ratio-
nalize than to prospectively predict.
To distinguish between having only support or also
active roles is immensely difficult because neurons and
astrocytes occupy the same spaces, and the unraveling
of cause and effect has long defied the experimental
tools available. In science, we accept an idea when it
satisfies the data and when it is fruitful, that is, leads to
new clarifications and a simplification of what seemed
incomprehensibly complex. The concept of the astro-
cyte as a supportive component of the CNS is surely
well established, but we lack resolution of some of the
details of already established mechanisms and likely the
uncovering of other support mechanisms until we have
near to a complete picture. This is basically knowledge
for knowledge’s sake while the current emphasis is on
translational research. But as Robert Boyle noted
(“Some Considerations Touching the Usefulness of
Experimental Natural Philosophy” [2nd ed. Oxford,
1664]) in a quote obtained from an article, by my PhD
mentor, Peter Nicholls (Nicholls 1981b), “Though that
famous distinction, introduced by the Lord Verulam
(Francis Bacon), whereby experiments are sorted into
luciferous and fructiferous may be of commendable use
(for politeness!), yet there are few fructiferous experi-
ments which may not become luciferous to the attentive
considerer . . . . and those experiments whose more
obvious use is to detect the nature or causes of things
may be exceedingly fructiferous.” The new ideas of
astrocyte function are provocative but seem at present
to have too little foundation, both in the reasons for
postulating them and experimental support. If they are
too unreasonable and are basically not the way
astrocytes function, they will lead us into false and
therefore nonfructiferous directions. We need to be
very cautious and critical in this regard to determine in
what directions the considerable mental and experi-
mental efforts, together with the essential financial
support, are expended.
Acknowledgments
I thank NSF (IOS-0642818) and an endowment from
M. Frank Rudy and Marjorie C. Rudy to Ordway
Research Institute for financial support during the
preparation of this article and Min Zhou for helpful dis-
cussions and help with some of the figures. I also thank
all the authors and the copyright holders for permission
to use their work.
References
Abbott NJ. 2002. Astrocyte-endothelial interactions and
blood-brain barrier permeability. J Anatomy 200:629–38.
Agulhon C, Petravicz J, McMullen AB, Sweger EJ, Minton
SK, Taves SR, and others. 2008. What is the role of astro-
cyte calcium in neurophysiology? Neuron 59:932–46.
Aitken PG, Borgdorff AJ, Juta AJA, Kiehart DP, Somjen GG,
Wadman WJ. 1998. Volume changes induced by osmotic
stress in freshly isolated rat hippocampal neurons. Pflug
Arch 436:991–8.
Amiry-Moghaddam M, Ottersen OP. 2003. The molecular
basis of water transport in the brain. Nat Rev Neurosci
4:991–1001.
Amzica F. 2002. In vivo electrophysiological evidences for
cortical neuron-glia interactions during slow (<1 Hz) and
paroxysmal sleep oscillations. J Physiol Paris 96:209–19.
Anderson CM, Nedergaard M. 2003. Astrocyte-mediated
control of cerebral microcirculation. Trends Neurosci
26:340–4.
Araque A, Parpura V, Sanzgiri RP, Haydon PG. 1999.
Tripartite synapses: glia, the unacknowledged partner.
Trends Neurosci 22:208–15.
Aschner M. 2000. Neuron-astrocyte interactions: implica-
tions for cellular energetics and antioxidant levels.
NeuroTox 21:1101–7.
Barres BA. 2003. What is a glial cell? Glia 43: 4–5.
Barres BA. 2008. The mystery and magic of glia: a perspective
on their roles in health and disease. Neuron 60:430–40.
Berl S, Lajtha A, Waelsch H. 1961. Amino acid and protein
metabolism. VI: cerebral compartments of glutamic acid
metabolism. J Neurochem 7:186–97.
Bignami A, Dahl D. 1974. Astrocyte-specific protein and neu-
roglial differentiation: an immunofluorescence study with
antibodies to the glial fibrillary acidic protein. J Comp
Neurol 153:27–38.
Bowman CL, Kimelberg HK. 1984. Excitatory amino acids
directly depolarize rat brain astrocytes in primary culture.
Nature 311:656–9.
Burg ED, Remillard CV, Yuan JX. 2008. Potassium channels
in the regulation of pulmonary artery smooth muscle cell
at BROWN UNIVERSITY on November 21, 2010nro.sagepub.comDownloaded from

Functions of Mature Mammalian Astrocytes / Kimelberg 103
proliferation and apoptosis: pharmacotherapeutic impli-
cations. Br J Pharmacol 153 Suppl 1:S99–111.
Bushong EA, Martone ME, Ellisman MH. 2004. Maturation
of astrocyte morphology and the establishment of astro-
cyte domains during postnatal hippocampal development.
Int J Dev Neurosci 22:73–86.
Bushong EA, Martone ME, Jones YZ, Ellisman MH. 2002.
Protoplasmic astrocytes in CA1 stratum radiatum occupy
separate anatomical domains. J Neurosci 22:183–92.
Cahoy JD, Emery B, Kaushal A, Foo LC, Zamanian JL,
Christopherson KS, and others. 2008. A transcriptome
database for astrocytes, neurons, and oligodendrocytes: a
new resource for understanding brain development and
function. J Neurosci 28:264–78.
Cammer W, Zhang H. 1991. Comparison of immunocyto-
chemical staining of astrocytes, oligodendrocytes, and
myelinated fibers in the brains of carbonic anhydrase
II-deficient mice and normal littermates. J Neuroimmunol
34:81–6.
Chao TI, Rickmann M, Wolff JR. 2002. The synapse-
astrocyte boundary: an anatomical basis for an integrative
role of glia in sysnaptic transmission. In: Volterra A,
Magistretti P, Haydon PG, editors. The tripartite synapse:
glia in sysnaptic trasnmission. Oxford: Oxford University
Press. p 3–23.
Chesler M. 2003. Regulation and modulation of pH in the
brain. Physiol Rev 83:1183–221.
Constanti A, Galvan M. 1978. Amino acid evoked depolariza-
tion of electrically inexcitable (neuroglial?) cells in the
guinea pig olfactory cortex slice. Brain Res 153:183–7.
D’Ascenzo M, Fellin T, Terunuma M, Revilla-Sanchez R,
Meaney DF, Auberson YP, and others. 2007. mGluR5
stimulates gliotransmission in the nucleus accumbens.
Proc Natl Acad Sci U S A 104:1995–2000.
Deitmer JW, McCarthy KD, Scemes E, Giaume C. 2006.
Information processing and transmission in glia: calcium
signaling and transmitter release. Glia 54:639–41.
Descarries L, Seguela P, Watkins KC. 1991. Nonjunctional
relationships of monoamine axon terminals in the cere-
bral cortex of adult rat. In: Fuxe K, Agnati LF, editors.
Volume transmission in the brain: novel mechanisms for
neural transmission. New York: Raven. p 53–62.
Dierig S. 1994. Extending the neuron doctrine: Carl Ludwig
Schleich (1859-1922) and his reflections on neuroglia at
the inception of the neural-network concept in 1894.
TINS 17:449–52.
Edvinsson L, MacKenzie ET, McCulloch J. 1993. Vascular
smooth muscle reactivity in vitro and in situ. In: Cerebral
blood flow and metabolism. New York: Raven Press. p
113–41.
Einstein A. 1956. The meaning of relativity. Princeton (NJ):
Princeton University Press.
Erlichman JS, Hewitt A, Damon TL, Hart M, Kurascz J, Li A,
Leiter JC. 2008. Inhibition of monocarboxylate transporter
2 in the retrotrapezoid nucleus in rats: a test of the astrocyte-
neuron lactate-shuttle hypothesis. J Neurosci 28:4888–96.
Fellin T, Pascual O, Gobbo S, Pozzan T, Haydon PG,
Carmignoto G. 2004. Neuronal synchrony mediated by
astrocytic glutamate through activation of extrasynaptic
NMDA receptors. Neuron 43:729–43.
Feynman R. 1965. The character of physical law. Cambridge
(MA): MIT Press.
Fiacco T, Casper K, Sweger E, Agulhon C, Taves S, Kurtzer-
Minton S, McCarthy KD. 2009. Molecular approaches
for studying astrocytes. In: Parpura V, Haydon PG, edi-
tors. Astrocytes in (patho)physiology of the nervous sys-
tem. New York: Springer. p 383–405.
Fiacco TA, Agulhon C, Taves SR, Petravicz J, Casper KB,
Dong X, and others. 2007. Selective stimulation of astro-
cyte calcium in situ does not affect neuronal excitatory
synaptic activity. Neuron 54:611–26.
Filosa JA, Bonev AD, Straub SV, Meredith AL, Wilkerson
MK, Aldrich RW, Nelson MT. 2006. Local potassium
signaling couples neuronal activity to vasodilation in the
brain. Nature Neurosci 9:1397–403.
Franze K, Grosche J, Skatchkov SN, Schinkinger S, Foja C,
Schild D, and others. 2007. Muller cells are living optical
fibers in the vertebrate retina. Proc Natl Acad Sci U S A
104:8287–92.
Gadea A, López-Colomé AM. 2001. Glial transporters for
glutamate, glycine, and GABA. II: GABA transporters. J
Neurosci Res 63:461–8.
Garcia-Marín V, García-López P, Freire M. 2007. Cajal’s con-
tributions to glial research. Trends Neurosci 30:479–87.
Gardner-Medwin AR. 1983. Study of the mechanisms by
which potassium moves through brain tissue in the rat. J
Physiol 335:353–74.
Girouard H, Iadecola C. 2006. Neurovascular coupling in the
normal brain and in hypertension, stroke, and Alzheimer
disease. J Appl Physiol 100:328–35.
Golgi C. 1885. Sulla fina anatomia degli organi centrali del
sisterma nervoso. Riv Sper Fremiat Med Leg Alienazione
Ment 11:72–123.
Gordon GR, Choi HB, Rungta RL, Ellis-Davies GC, MacVicar
BA. 2008. Brain metabolism dictates the polarity of astro-
cyte control over arterioles. Nature 456:745–50.
Grosche J, Matyash V, Moller T, Verkhratsky A, Reichenbach
A, Kettenmann H. 1999. Microdomains for neuron-glia
interaction: parallel fiber signaling to Bergmann glial
cells. Nat Neurosci 2:139–43.
Haberg A, Qu H, Saether O, Unsgard G, Haraldseth O,
Sonnewald U. 2001. Differences in neurotransmitter
synthesis and intermediary metabolism between gluta-
matergic and GABAergic neurons during 4 hours of
middle cerebral artery occlusion in the rat: the role of
astrocytes in neuronal survival. J Cereb Blood Flow
Metab 21:1451–63.
Halassa MM, Florian C, Fellin T, Munoz JR, Lee SY, Abel T,
and others. 2009. Astrocytic modulation of sleep homeo-
stasis and cognitive consequences of sleep loss. Neuron
61:213–9.
Halliwell B, Gutteridge JMC. 1985. Oxygen radicals and the
nervous system. TINS 8:22–6.
Halliwell B, Gutteridge JMC. 1999. Free radicals in biology
and medicine. Oxford: Oxford University Press.
Hatton G. 1997. Function-related plasticity in hypothalamus.
Ann Rev Neurosci 20:375–97.
Haydon PG, Carmignoto G. 2006. Astrocyte control of syn-
aptic transmission and neurovascular coupling. Physiol
Rev 86:1009–31.
Hertz L, Peng L, Dienel GA. 2007. Energy metabolism
in astrocytes: high rate of oxidative metabolism and
spatiotemporal dependence on glycolysis/glycogenolysis. J
Cereb Blood Flow Metab 27:219–49.
at BROWN UNIVERSITY on November 21, 2010nro.sagepub.comDownloaded from

104 The Neuroscientist / Vol. 16, No. 1, February 2010
Hille B. 1992. Ionic channels of excitable membranes.
Sunderland (MA): Sinauer Associates Inc.
Hodgkin AL, Horowicz P. 1959. Influence of potassium and
chloride ions on the membrane potential of single muscle
fibres. J Physiol 148:127–60.
Iacobas DA, Scemes E, Spray DC. 2004. Gene expression
alterations in connexin null mice extend beyond the gap
junction. Neurochem Int 45:243–50.
Ihrie RA, Alvarez-Buylla A. 2008. Neural stem cells disguised
as astrocytes. In: Parpura V, Haydon PG, editors.
Astrocytes in (patho)physiology of the nervous system.
New York: Springer. p 27–47.
Iino M, Goto K, Kakegawa W, Okado H, Sudo M, Ishiuchi S,
and others. 2001. Glia-synapse interaction through
Ca2+-permeable AMPA receptors in Bergmann glia
Science 292:926–9.
Jahn R, Scheller RH. 2006. SNAREs: engines for membrane
fusion. Nat Rev Mol Cell Biol 7:631–43.
Jourdain P, Bergersen LH, Bhaukaurally K, Bezzi P, Santello
M, Domercq M, and others. 2007. Glutamate exocytosis
from astrocytes controls synaptic strength. Nat Neurosci
10:331–9.
Katz B. 1966. Nerve, muscle and synapse. New York:
McGraw-Hill Inc.
Keilin D. 1925. On cytochrome, a respiratory pigment, com-
mon to animals, yeast and higher plants. Proc R Soc
London B 98:312–39.
Keilin D. 1966. The history of cell respiration and cyto-
chrome. London: Cambridge University Press.
Kettenmann H, Backus KH, Schachner M. 1984. Aspartate,
glutamate and gamma-aminobutyric acid depolarize cul-
tured astrocytes. Neurosci Letts 52:25–9.
Kettenmann H, Ransom B. 2005. The concept of neuroglia:
a historical perpspective. In: Neuroglia (Kettenmann H,
Ransom B, eds), pp 1–16. New York: Oxford University
Press.
Kimelberg HK. 2009. Astrocyte heterogeneity or homogeneity.
In: Parpura V, Haydon PG, editors. Astrocytes in (patho)
physiology of the nervous system. New York: Springer.
p 1–26.
Kimelberg HK. 1979. Glial enzymes and ion transport in
brain swelling. In: Popp AJ, Bourke RS, Nelson LR,
Kimelberg HK, editors. Neural trauma. New York: Raven
Press. p 137–53.
Kimelberg HK. 1981. Active accumulation and exchange
transport of chloride in astroglial cells in culture. Biochem
Biophys Acta 646:179–84.
Kimelberg HK. 1983. Primary astrocyte cultures: a key to
astrocyte function. Cell Molec Neurobiol 3:1–16.
Kimelberg HK, editor. 1988. Glial cell receptors. New York:
Raven Press.
Kimelberg HK. 1999. Current methods and approaches to
studying astrocytes: a forum position paper. NeuroTox
20:703–12.
Kimelberg HK. 2001. Glia-neuronal culture models: do we
need to change the paradigms? Trends Neurosci 24:205.
Kimelberg HK. 2004a. The role of hypotheses in current
research, illustrated by hypotheses on the possible role of
astrocytes in energy metabolism and cerebral blood flow:
from Newton to now. J Cereb Blood Flow Metab 24:
1235–9.
Kimelberg HK. 2004b. Water homeostasis in the brain: basic
concepts. Neurosci 129:851–60.
Kimelberg HK. 2007. Supportive or information-processing
functions of the mature protoplasmic astrocyte in the
mammalian CNS? A critical appraisal. Neuron Glia Biol
3:181–9.
Kimelberg HK, Biddlecome SM, Bourke RS. 1979. SITS-
inhibitable Cl- transport and Na+-dependent H+ produc-
tion in primary astroglial cultures. Brain Res 173:111–24.
Kimelberg HK, Katz DM. 1985. High-affinity uptake of sero-
tonin into immunocytochemically identified astrocytes.
Science 228:889–91.
Kimelberg HK, Stieg PE, Mazurkiewicz JE. 1982.
Immunocytochemical and biochemical analysis of car-
bonic anhydrase in primary astrocyte cultures from rat
brain. J Neurochem 39:734–42.
Koehler RC, Roman RJ, Harder DR. 2009. Astrocytes and the
regulation of cerebral blood flow. Trends Neurosci
32:160–9.
Kofuji P, Newman EA. 2004. Potassium buffering in the cen-
tral nervous system. Neurosci 129:1045–56.
Kraig RP, Pulsinelli WA, Plum F. 1985. Hydrogen ion buff-
ering during complete brain ischemia. Brain Res 342:
281–90.
Krnjevic K, Schwartz S. 1967. Some properties of unrespon-
sive cells in the cerebral cortex. Exp Brain Res 3:306–19.
Kuffler SW, Nicholls JG. 1976. From neuron to brain.
Sunderland (MA): Sinauer Assoc.
Kuffler SW, Nicholls JG, Orkand RK. 1966. Physiological
properties of glial cells in the central nervous system of
amphibia. J Neurophysiol 29:768–87.
Landis DMD, Reese TS. 1981. Membrane structure in mam-
malian astrocytes: a review of freeze fracture studies on
adult, developing, reactive and cultured astrocytes. J Exp
Biol 95:35–48.
Leino RL, Gerhart DZ, van Bueren AM, McCall AL, Drewes
LR. 1997. Ultrastructural localization of GLUT 1 and
GLUT 3 glucose transporters in rat brain. J Neurosci Res
49:617–26.
Li D, Ropert N, Koulakoff A, Giaume C, Oheim M. 2008.
Lysosomes are the major vesicular compartment undergo-
ing Ca2+-regulated exocytosis from cortical astrocytes. J
Neurosci 28:7648–58.
Liu X, Bolteus AJ, Balkin DM, Henschel O, Bordey A. 2006.
GFAP-expressing cells in the postnatal subventricular
zone display a unique glial phenotype intermediate
between radial glia and astrocytes. Glia 54:394–410.
Lovatt D, Sonnewald U, Waagepetersen HS, Schousboe A,
He W, Lin JH, and others. 2007. The transcriptome and
metabolic gene signature of protoplasmic astrocytes in
the adult murine cortex. J Neurosci 27:12255–66.
Lübbers DW. 1977. Quantitative measurement and descrip-
tion of oxygen supply to the tissue. In: Jöbsis FF, editor.
Oxygen and physiological function. Dallas (TX):
Professional Information Library. p 254–76.
Lugaro E. 1907. Sulle funzioni della nevroglia. Riv D Pat
Nerv Ment 12:225–33.
Magistretti PJ, Sorg O, Naichen Y, Pellerin L, De Rham S,
Martin J-L. 1994. Regulation of astrocyte energy metabo-
lism by neurotransmitters. Renal Physiol Biochem
17:168–71.
at BROWN UNIVERSITY on November 21, 2010nro.sagepub.comDownloaded from

Functions of Mature Mammalian Astrocytes / Kimelberg 105
Majumdar D, Maunsbach AB, Shacka JJ, Williams JB, Berger
UV, Schultz KP, and others. 2008. Localization of electro-
genic Na/bicarbonate cotransporter NBCe1 variants in
rat brain. Neurosci 155:818–32.
Matsui K, Jahr CE. 2006. Exocytosis unbound. Curr Opin
Neurobiol 16:305–11.
Minelli A, Brecha NC, Karschlin C, DeBiasi S, Conti F.
1995. GAT-1, a high-affinity GABA plasma membrane
transporter, is localized to neurons and astroglia in the
cerebral cortex. J Neurosci 15:7734–46.
Minelli A, DeBiasi S, Brecha NC, Zuccarello LV, Conti F.
1996. GAT-3, a high-affinity GABA plasma membrane
transporter, is localized to astrocytic processes, and it is
not confined to the vicinity of GABAergic synapses in the
cerebral cortex. J Neurosci 16:6255–64.
Montana V, Ni Y, Sunjara V, Hua X, Parpura V. 2004.
Vesicular glutamate transporter-dependent glutamate
release from astrocytes. J Neurosci 24:2633–42.
Morest DK, Silver J. 2003. Precursors of neurons, neuroglia,
and ependymal cells in the CNS: what are they? Where
are they from? How do they get where they are going?
Glia 43:6–18.
Nedergaard M, Ransom B, Goldman SA. 2003. New roles for
astrocytes: redefining the functional architecture of the
brain. Trends Neurosci 26:523–30.
Newman EA. 1986. High potassium conductance in astro-
cyte endfeet. Science 233:453–4.
Newman EA. 1987. Does the release of potassium from
astrocyte endfeet regulate cerebral blood flow? Science
237:896–8.
Newman EA, Frambach DA, Odette LL. 1984. Control of
extracellular potassium levels by retinal glial cell K+
siphoning. Science 225:1174–5.
Nicholls J. 1981a. Perspectives on the cell biology of glia. J
Exp Biol 95:3–5.
Nicholls P. 1981b. Liposomes as artificial organelles, topo-
chemical matrices, and therapeutic carrier systems. Int
Rev Cytol Suppl 12:327–88.
Nimmerjahn A, Mukamel EA, Schnitzer MJ. 2009. Motor
behavior activates Bergmann glial networks. Neuron
62:400–12.
Nishiyama A, Komitova M, Suzuki R, Zhu X. 2009.
Polydendrocytes (NG2 cells): multifunctional cells with
lineage plasticity. Nat Rev Neurosci 10:9–22.
Oberheim NA, Takano T, Han X, Xu Q, Wyatt JD, Pilcher W,
and others. 2009. Uniquely hominid features of adult
human astrocytes. J Neurosci 29:3276–87.
Oliet SH, Piet R, Poulain DA. 2001. Control of glutamate
clearance and synaptic efficacy by glial coverage of neu-
rons. Science 292:923–6.
Olsen ML, Higashimori H, Campbell SL, Hablitz JJ,
Sontheimer H. 2005. Functional expression of K(ir)4.1
channels in spinal cord astrocytes. Glia 53:516–28.
Orkand RK, Nicholls JG, Kuffler SW. 1966. Effect of nerve
impulses on the membrane potential of glial cells in the
central nervous system of amphibia. J Neurophysiol
29:788–806.
Paez PM, Fulton D, Colwell CS, Campagnoni AT. 2008.
Voltage-operated Ca(2+) and Na(+) channels in the oligo-
dendrocyte lineage. J Neurosci Res 87:3259–66.
Pardridge WM. 1997. Drug delivery to the brain. J Cereb
Blood Flow Metab 17:713–31.
Parker MD, Bouyer P, Daly CM, Boron WF. 2008. Cloning
and characterization of novel human SLC4A8 gene
products encoding Na+-driven Cl-/HCO3(-) exchanger
variants NDCBE-A, -C, and -D. Physiol Genomics
34:265–76.
Parpura V, Basarsky TA, Liu F, Jeftinija K, Jeftinija S, Haydon
PG. 1994. Glutamate-mediated astrocyte-neuron signal-
ling. Nature 369:744–7.
Pascual O, Casper KB, Kubera C, Zhang J, Revilla-Sanchez
R, Sul JY, and others. 2005. Astrocytic purinergic signal-
ing coordinates synaptic networks. Science 310:113–6.
Pavesi M. 1999. Induction versus popper: substance versus
semantics. Int J Epidemiol 28:360–1.
Pellerin L, Magistretti PJ. 2003. Food for thought: challeng-
ing the dogmas. J Cereb Blood Flow Metab 23:1282–6.
Pellerin L, Magistretti PJ. 2004. Neuroenergetics: calling
upon astrocytes to satisfy hungry neurons. Neuroscientist
10:53–62.
Picker S, Pieper CF, Goldring S. 1981. Glial membrane
potentials and their relationship to [K+]o in man and
guinea pig. J Neurosurg 55:347–63.
Poincare H. 1952. Science and hypothesis. New York: Dover
Publications Inc.
Rakic P. 1990. Principles of neural cell migration. Experientia
46(9):882–91.
Ramon y Cajal S. 1913. Contribucion al conocimento de la
neuroglia del cerebro humano. Trab Lab Invest Biol Univ
Madrid 11:255–315.
Ransom B, Behar T, Nedergaard M. 2003. New roles for
astrocytes (stars at last). Trends Neurosci 26:520–2.
Reichenbach A, Wolburg H. 2009. Structural association of
astrocytes with neurons and vascualture: defining territo-
rial boundaries. In: Parpura V, Haydon PG, editors.
Astrocytes in (patho)physiology of the nervous system.
New York: Springer. p 251–86.
Ribak CE, Tong WM, Brecha NC. 1996. GABA plasma mem-
brane transporters, GAT-1 and GAT-3, display different
distributions in the rat hippocampus. J Comp Neurol
367:595–606.
Robergs RA, Ghiasvand F, Parker D. 2004. Biochemistry of
exercise-induced metabolic acidosis. Am J Physiol Regul
Integr Comp Physiol 287:R502–16.
Rose CR, Ransom BR. 1998. pH regulation in mammalian
glia. In: Kaila K, Ransom BR, editors. pH and brain func-
tion. New York: Wiley-Liss. p 253–76.
Rothstein JD, Dykes-Hoberg M, Pardo CA, Bristol LA, Jin L,
Kuncl RW, and others. 1996. Knockout of glutamate
transporters reveals a major role for astroglial transport in
excitotoxicity and clearance of glutamate. Neuron
16:675–86.
Rouach N, Koulakoff A, Abudara V, Willecke K, Giaume C.
2008. Astroglial metabolic networks sustain hippocampal
synaptic transmission. Science 322:1551–5.
Russell B. 1989. Wisdom of the west. New York: Crescent
Books.
Scales SJ, Chen YA, Yoo BY, Patel SM, Doung YC, Scheller
RH. 2000. SNAREs contribute to the specificity of mem-
brane fusion. Neuron 26:457–64.
Schmitt BM, Berger UV, Douglas RM, Bevensee MO,
Hediger MA, Haddad GG, Boron WF. 2000. Na/HCO3
cotransporters in rat brain: expression in glia, neurons,
and choroid plexus. J Neurosci 20:6839–48.
at BROWN UNIVERSITY on November 21, 2010nro.sagepub.comDownloaded from

106 The Neuroscientist / Vol. 16, No. 1, February 2010
Schools GP, Kimelberg HK. 1999. mGluR3 and mGluR5
are the predominant metabotropic glutamate receptor
mRNAs expressed in hippocampal astrocytes acutely iso-
lated from young rats. J Neurosci Res 58:533–43.
Schools GP, Zhou M, Kimelberg HK. 2006. Development of
gap junctions in hippocampal astrocytes: evidence that
whole cell electrophysiological phenotype is an intrinsic
property of the individual cell. J Neurophysiol 96:
1383–92.
Schummers J, Yu H, Sur M. 2008. Tuned responses of astro-
cytes and their influence on hemodynamic signals in the
visual cortex. Science 320:1638–43.
Sibson NR, Shen J, Mason GF, Rothman DL, Behar KL,
Shulman RG. 1998. Functional energy metabolism:
in vivo 13C-NMR spectroscopy evidence for coupling of
cerebral glucose consumption and glutamatergic neuro-
nal activity. Dev Neurosci 20:321–30.
Simard M, Arcuino G, Takano T, Liu QS, Nedergaard M.
2003. Signaling at the gliovascular interface. J Neurosci
23:9254–62.
Skatchkov SN, Eaton MJ, Shuba YM, Kucheryavykh YV,
Derst C, Veh RW, and others. 2006. Tandem-pore domain
potassium channels are functionally expressed in retinal
(Muller) glial cells. Glia 53:266–76.
Stein WD. 2002. Cell volume homeostasis: ionic and non-
ionic mechanisms. The sodium pump in the emergence
of animal cells. Int Rev Cytol 215:231–58.
Sudhof TC, Rothman JE. 2009. Membrane fusion: grap-
pling with SNARE and SM proteins. Science 323:
474–7.
Takano T, Tian GF, Peng W, Lou N, Libionka W, Han X,
Nedergaard M. 2006. Astrocyte-mediated control of cere-
bral blood flow. Nat Neurosci 9:260–7.
Taylor FS. 1949. Science past and present. London: William
Heinemann Ltd.
Tekkök SB, Ransom BR. 2004. Glial-neuronal interactions
and signaling: an introduction. In: Hatton G, Parpura V,
editors. Glial neuronal signaling. Norwell (MA): Kluwer
Academic Publishers. p 1–20.
Traystman RJ. 1997. Regulation of cerebral blood flow by
carbon dioxide. In: Welch KMA, Caplan LR, Reis DJ,
Siesjo BK, Weir B, editors. Primer on cerebrovascular
diseases. San Diego (CA): Academic Press. p 55–8.
Tschirgi RD. 1958. The blood-brain barrier. In: Windle WF,
editor. Biology of neuroglia. Springfield (IL): Charles C.
Thomas. p 130–8.
Virgintino D, Monaghan P, Robertson D, Errede M, Bertossi
M, Ambrosi G, Roncali L. 1997. An immunohistochemi-
cal and morphometric study on astrocytes and microvas-
culature in the human cerebral cortex. Histochem J
29:655–60.
Volterra A, Meldolesi J. 2005. Astrocytes, from brain glue to
communication elements: the revolution continues. Nat
Rev Neurosci 6:626–40.
Voutsinos-Porche B, Bonvento G, Tanaka K, Steiner P,
Welker E, Chatton JY, and others. 2003. Glial glutamate
transporters mediate a functional metabolic crosstalk
between neurons and astrocytes in the mouse developing
cortex. Neuron 37:275–86.
Wang DD, Bordey A. 2008. The astrocyte odyssey. Prog
Neurobiol 86:342–67.
Warskulat U, Heller-Stilb B, Oermann E, Zilles K, Haas H, Lang
F, Haussinger D. 2007. Phenotype of the taurine transporter
knockout mouse. Methods Enzymol 428:439–58.
Wilhelmsson U, Bushong EA, Price DL, Smarr BL, Phung V,
Terada M, and others. 2006. Redefining the concept of
reactive astrocytes as cells that remain within their
unique domains upon reaction to injury. Proc Natl Acad
Sci U S A 103:17513–8.
Winder DG, Conn PJ. 1996. Roles of metabotropic glutamate
receptors in glial function and glial-neuronal communi-
cation. J Neurosci Res 46:131–7.
Winder DG, Ritch PS, Gereau RW, Conn PJ. 1996. Novel
glial-neuronal signalling by coactivation of metabotropic
glutamate and β-adrenergic receptors in rat hippocam-
pus. J Physiol (Lond ) 494:743–55.
Yang Y, Rothstein JD. 2009. Specialized neurotransmitter
transporters in astrocytes. In: Parpura V, Haydon PG, edi-
tors. Astrocytes in (patho)physiology of the nervous sys-
tem. New York: Springer. p 69–105.
Zhang Q, Pangrsic T, Kreft M, Krzan M, Li N, Sul JY, and
others. 2004. Fusion-related release of glutamate from
astrocytes. J Biol Chem 279:12724–33.
Zhang Z, Chen G, Zhou W, Song A, Xu T, Luo Q, and others.
2007. Regulated ATP release from astrocytes through
lysosome exocytosis. Nat Cell Biol 9:945–53.
Zhou M, Kimelberg HK. 2001. Freshly isolated hippocampal
CA1 astrocytes comprise two populations differing in
glutamate transporter and AMPA receptor expression.
J Neurosci 21:7901–8.
Zhou M, Schools GP, Kimelberg HK. 2006. Development of
GLAST(+) astrocytes and NG2(+) glia in rat hippocam-
pus CA1: mature astrocytes are electrophysiologically
passive. J Neurophysiol 95:134–43.
Zhou M, Xu G, Xie M, Zhang X, Schools GP, Ma L, and others.
2009. TWIK-1 and TREK-1 are potassium channels con-
tributing significantly to astrocyte passive conductance in
rat hippocampal slices. J Neurosci 29:8551–8564.
Zonta M, Angulo MC, Gobbo S, Rosengarten B, Hossmann
KA, Pozzan T, Carmignoto G. 2002. Neuron-to-astrocyte
signaling is central to the dynamic control of brain micro-
circulation. Nat Neurosci 6:43–50.
For reprints and permissions queries, please visit SAGE’s Web site at http://www.sagepub.com/journalsPermissions.nav.
at BROWN UNIVERSITY on November 21, 2010nro.sagepub.comDownloaded from