GLIA 22:237–248 (1998)

Astrocytes Protect Neurons

From Neurotoxic Injury
by Serum Glutamate
ZU-CHENG YE AND HARALD SONTHEIMER*
Department of Neurobiology, The University of Alabama at Birmingham,
Birmingham, Alabama

KEY WORDS cell culture; glia; fetal calf serum; fetal bovine serum; horse serum;
glutamate transport; neurotoxicity

ABSTRACT Serum is used widely for culturing neurons and glial cells, and is
thought to provide essential, albeit undefined, factors such as hormones, growth factors,
and trace elements that promote the growth of cells in vitro. Moreover, serum can have
profound effects on cell proliferation, differentiation, and cell morphology, and may even
influence cell fate decisions. Despite the overall growth-promoting influence of serum on
cell culture, frequent media changes have been shown to be detrimental to neuronal
cultures, significantly reducing the yield of viable neurons. The reason for this loss of
neurons by frequent media changes has been puzzling.
We demonstrate that bovine and horse sera, the most popular serum complements for
CNS cell culture, are a significant source for glutamate, supplying glutamate at
concentrations sufficient to kill primary cultured hippocampal neurons. By using the
bioluminescence detection method, we determined the glutamate concentration [Glu] in
several batches of fetal bovine (calf ) sera (FBS) to be close to 1 mM, and that of horse sera
to be ,0.3 mM. Thus 10% serum supplement to culture media results in [Glu] of 30–100
µM due to serum alone.
We subsequently produced glutamate depleted media (GDM) by using primary
cultures of hippocampal astrocytes to absorb glutamate from media containing 10% FBS.
Within 3 h, astrocytes reduced the [Glu] in the medium from ,90 µM to less than 1 µM.
Sister cultures of hippocampal neuron that underwent frequent media changes with
GDM or GDM 1 partial untreated media demonstrated that GDM significantly increase
neuronal survival (10-fold at 21 DIV). Subsequent exposure to glutamate provided by
either untreated serum or by equivalent doses of exogenous glutamate added to GDM led
to dose-dependent neuronal cell death. The relative sensitivity of hippocampal neurons
to glutamate increased with increasing culture age from initial ED50 values of . 100 µM
(, 6 DIV) to ,6 µM in cultures maintained for 3 weeks or longer. The relative sensitivity
to exogenous glutamate was at least 2-fold higher in neurons cultured in GDM than in
sister cultures maintained in media containing untreated serum. The death of neurons
exposed to untreated media was blocked by the NMDA receptor antagonist MK-801.
These experiments suggest that the vulnerability of neurons to media changes can be
solely explained by excitotoxicity resulting from serum-borne glutamate. Moreover, we
propose that use of GDM may be advantageous for culturing hippocampal neurons and
may eliminate the possible selection for glutamate resistant neurons. The use of GDM
could be particularly important for studies of excitotoxicity; our study predicts that the

Contract grant sponsor: NIH; Contract grant numbers: R01 31234, P50 HD32901.
*Correspondence to: Dr. Harald Sontheimer, Department of Neurobiology, The University of Alabama at Birmingham, 1719 6th Avenue South, CIRC 545,
Birmingham, AL 25294.
Received 4 March 1997; Accepted 16 August 1997

r 1998 Wiley-Liss, Inc.

238 YE AND SONTHEIMER
ED50 for neuronal culture with regular serum will be artificially high and may not
adequately reflect the in vivo state. GLIA 22:237–248, 1998. r 1998 Wiley-Liss, Inc.
INTRODUCTION
Sera are frequently used as supplement for culturing
neuronal and glial cells (Buchhalter and Dichter, 1992;
Dichter, 1978; Juurlink and Hertz, 1992). The two most
commonly used are fetal bovine (calf ) and horse serum.
Serum provides hormones, growth factors, and trace
elements that are generally beneficial for cell growth
and long-term cell survival. The exact composition of
most sera is an enigma and significant species-to-
species and batch-to-batch variations in the effective-
ness of serum complements for cell culture are more
than anecdotal (Bottenstein, 1992).
Serum supplement not only affects long-term sur-
vival and cell proliferation but also may cause signifi-
cant phenotypic changes and can even determine cell
fate. For example, glial precursor cells maintained in
horse serum will give rise to oligodendrocytes, whereas
precursor cells exposed to fetal calf serum will develop
into astrocytes (Miller and Raff, 1984; Raff et al., 1983;
Raff, 1989). Withdrawal of serum from cortical astro-
cytes leads to a reversible stellation of flat, type-1 like
astrocytes (Lascola and Kraig, 1996) and serum starva-
tion (0.1% FCS) reversibly inhibits the proliferation of
cortical astrocytes (Langan et al., 1994).
While serum generally has beneficial effects on cell
growth, it has been documented that regular exchanges
of media can be harmful to neuronal cultures (Rosen-
berg and Aizenman, 1989). It has been speculated that
serum may contain factors that are toxic to neurons
(Buchhalter and Dichter, 1992). To reduce this neurotox-
icity, a number of investigators have experimented with
the use of serum-free defined media (Bottenstein, 1992;
Driscoll et al., 1993) or have significantly reduced the
concentration at which serum is supplemented to me-
dia for neuronal cell cultures (Harris and Rosenberg,
1993; Rosenberg, 1991). Others have chosen either
limited media changes or partial replacements (Buch-
halter and Dichter, 1992; Jahr and Stevens, 1987) or
chosen to eliminate media changes totally (Harris and
Rosenberg, 1993; Rosenberg and Aizenman, 1989).
These approaches have been variably successful in
enhancing neuronal survival.
Glutamate is the principal excitatory neurotransmit-
ter in the mammalian central nervous system. It is now
well established that elevated levels of glutamate can
result in wide-spread neurotoxicity (Choi, 1988; Tho-
mas, 1995). As a consequence, extracellular glutamate
is maintained at concentrations of ,1 µM by the
combined re-uptake into neurons and glial cells alike
(Nicholls and Attwell, 1990). This is accomplished by
several recently identified and cloned Na1-dependent
co-transport systems in neurons and glia, which are
believed to operate in parallel to sequester glutamate
and prevent excessive accumulations of glutamate in
the extracellular space (Kanai et al., 1993; Rothstein et
al., 1994). The importance of glial glutamate uptake in
the control of extracellular glutamate and therefore
protection of neurons from excitotoxic injury has been
suggested for some time. A recent study (Rothstein et
al., 1996), using knockout mutants in which the astro-
cytic glutamate transporter GLT-1 was selectively elimi-
nated, provided strong experimental evidence that as-
trocytic glutamate uptake is indeed essential for normal
brain function.
Few studies have noticed the glutamate contamina-
tion in cell culture media or serum complements (Choi
et al., 1987; Cambier et al., 1995). One study reported
that glutamate may be derived from glutamine impuri-
ties in the medium (Driscoll et al., 1993) or due to the
neuronal conversion of glutamate from glutamine
(Driscoll et al., 1993; Rosenberg, 1991). However, it has
been documented that astrocytes promote neuronal
survival in vitro and that astrocytes can protect neu-
rons from excitotoxic cell death after exposure to gluta-
mate (Harris and Rosenberg, 1993; Rosenberg and
Aizenman, 1989).
The objectives of the present study were to character-
ize glutamate contents in commonly used animal sera
and study their potential neurotoxic effects on neuronal
cultures. We demonstrate that the glutamate content in
most sera is sufficient to cause significant neurotoxicity.
This toxicity can be eliminated by exposure of media to
confluent astrocytes, which rapidly deplete glutamate
from the medium and reduce it from ,90 µM to , 1 µM
within 3 h. Our results suggest that media-induced
neurotoxicity can be solely explained by the relatively
high glutamate content in most sera.
MATERIALS AND METHODS
Materials
NAD(P)H:FMN oxidoreductase, glutamate dehydro-
genase were purchased from Boehringer Mannheim
(Indianapolis, IN) and Myristyl aldehyde was obtained
from Aldrich (Milwaukee,WI). Cell culture supplies
were mainly obtained from Becton Dickinson (Franklin
Lakes, NJ) and Corning (Corning, NY). Enzymes and
chemicals were purchased from Sigma Chemical Co.
(St. Louis. MO) unless mentioned otherwise.
Cell Culture and Glutamate Depleted Media
Primary cultures of rat postnatal hippocampal astro-
cytes were prepared from Sprague-Dawley rat pups
(P1–P3), as we have previously described (Ye and
Sontheimer, 1996). Confluent hippocampal astrocytes
cultured in 75 cm2 flasks were utilized after 7–20 days
 NEUROTOXICITY BY SERUM GLUTAMATE
in vitro (DIV) for production of glutamate depleted
media (GDM), in which [Glu] were typically , 1 µM.
Briefly, media in the flasks was collected and replaced
by 30–40 ml fresh warmed media every 6–10 h Fresh
culture media was glutamine free Earle’s Minimum
Essential Medium (Gibco, Grand Island, NY), supple-
mented with 10% FBS (Hyclone, Logan, UT), 20 mM
glucose, 10 U/ml penicillin, and 10 µg/ml streptomycin.
The GDM was filtered through 0.2 µm bottom top filter
and either stored at 4°C and used within 2 weeks or
stored at 220°C for later usage. To eliminate possible
temperature and pH differences, GDM was used after it
had been allowed to equilibrate CO2 and temperature
for 2 h in the incubator in a ventilated flask.
Primary cultures of rat hippocampal neurons were
prepared from E17–18 embryos removed from timed-
pregnant Sprague-Dawley rats (Charles River Labora-
tories, Wilmington, MA) narcotized by CO2 followed by
cervical dislocation. Briefly, hippocampi were dissected
and freed of meninges under a low-power dissection
microscope. Harvested tissues were rotated 15 min at
37°C in pH 7.2 oxygenated enzyme solution containing
20 units/ml papain (Worthington, Freehold, NJ) and (in
mM): NaCl 137, KCl 5.3, MgCl2 1, Glucose 25, HEPES
10, and CaCl2 3, supplemented with 0.5 mM EDTA and
0.2 mg/ml L-cysteine. Tissues were then pelleted and
triturated with fire-polished Pasteur pipettes in GDM
supplemented with 1.5 mg/ml trypsin inhibitor and
bovine serum albumin. Cell suspensions were centri-
fuged again and suspended with GDM and plated at a
density of 0.5–1.0 3 105 cells/cm2 into 24 well-plates or
35 mm Petri dishes (Corning, NY) previously coated
with poly-ornithine and maintained in a 5% CO2/95%
air-humidified incubator (Lab-Line, Melrose Park, IL).
Culture media was changed 2 days after plating and
subsequently every 4–5 days with GDM. This proce-
dure resulted in mixed cultures of neurons and glia. For
comparison with previously published work, some cells
were cultured in 10% calf/horse serum (Gibco, Grand
Island, NY) and/or glutamate depleted calf/horse se-
rum.
Cell Identification
Cultures were stained with antibodies recognizing
either glial fibrillary acidic protein (GFAP; Incstar,
Stillwater, MN) or neurofilament protein (Boehringer
Mannheim, Indianapolis, IN) to confirm morphological
cell identification as astrocyte or neuron, respectively.
Cells cultured on glass coverslips were fixed for 10 min
in 4% ice-cold paraformadehyde and then washed in
PBS, followed by fixation in 70% ethanol at 4°C. Both
primary and secondary antibodies were prepared in a
solution containing (w/w) NaCl 0.9%, glycine 2%, and
(v/v) FBS 5% in Dulbecco’s PBS (Gibco, Grand Island,
NY). Cells were incubated at 4°C for 24 h with primary
antibodies, and then coverslips were washed three
times with D-PBS and incubated with FITC-conjugated
goat anti-rabbit antibody for 2 h at room temperature.
239
Astrocytes used for production of GDM were .95%
GFAP positive.
Quantitative Determination of Glutamate
Content in Sera and Media
We used the bioluminescence method for measuring
glutamate concentration [Glu] in solution as described
by Fosse et al. (1986) with minor modifications. Briefly,
the glutamate-specific reagent mixture contained potas-
sium phosphate 25 mM (pH 7.0), Triton X-100 40 µg/ml,
dithiothreitol 100 µM, myristyl aldehyde 40 µM, b-
NAD 2 mM, ADP 250 µM, FMN 2.5 µM, luciferase 40
µg/ml, NAD(P)H:FMN oxidoreductase 400 mU/ml, and
glutamate dehydrogenase 0.5 mg/ml. All experiments
were performed with a Monolight 2010 luminometer
(Analytical Luminescence Lab., Ann Arbor, MI).
The culture media samples (10% serum) were centri-
fuged and diluted 1:20–100 with distilled water and
stored at 220°C until measurement. Original serum
samples were diluted 1:200–1,000 with distilled water.
Glutamate standards were prepared in distilled water
and/or in 1:20–100 diluted GDM media. Each serum
sample was measured at least twice.
Time Lapse Video Microscopy
Neuronal cultures in 35 mm Petri dishes were placed
in a Leiden chamber (Ince et al., 1983) maintained
within a micro-CO2-incubator (Nikon, Japan) mounted
on top of a Nikon Diaphot microscope. Cells were
maintained in a humidified environment at 37°C, sup-
plied with H2O saturated 5%/95% CO2/O2 mixture
through the gas duct of the chamber. To ensure that no
significant evaporation of medium occurred throughout
these experiments, the osmolarity of the medium was
determined at various times throughout several experi-
ments. Changes were always , 5% after 48 h. Cell
images were obtained at 2003 magnification under
phase-contrast using a CCD camera (Edmund Scien-
tific Co., Barrington, NJ) and were captured at 1 s
intervals with a time-lapse VCR (Panasonic, Japan).
Images were digitized off-line from videotapes for analy-
sis and documentation, using a commercial frame-
grabber (Intel, Santa Clara, CA) attached to a microcom-
puter. Representative results are shown in Figures 3
and 6.
Neuronal Survival
Hippocampal neurons could be readily distinguished
from the bed-layer of astrocytes by their characteristic
cell morphology. Neurons had phase bright, rounded-up
cell bodies with few but long processes (see, for ex-
ample, Figs. 3 and 6). In phase-contrast images, neuro-
nal cell bodies were usually surrounded by a halo. The
length and the number of cell processes increased in the
240 YE AND SONTHEIMER

first 2 weeks in vitro. Neurons older than 2 weeks TABLE 1. Glutamate concentration in commercial sera
frequently associated in groups and their processes Serum typea and source Lot number [Glutamate] µM
intertwined in a net-like fashion (see Figs. 3 and 6).
FBS (Hyclone) 11112526 891.9
Astrocytes in these mixed cultures, by contrast, were FBS (Hyclone) AFB4919 830.1
mainly flat and non-process bearing. We confirmed our FBS (Hyclone) AFC 4965 853.2
FBS (Hyclone) 11152508 808.2
morphological identification of neurons on sister cul- FBS (Hyclone) 11112589 874.6
tures in which we stained cells with anti-neurofilament FCS (Atlanta) 4000C 1195.7
antibodies. FBS (Summit) 21L12 1067.5
HS (Hyclone) 21S12171 434.4
To compare the toxicity of serum-glutamate, exog- HS (Gibco) 32K1563 309.7
enous glutamate and GDM, viable neurons were counted HS (Gibco) 40P9062 315.2
in 6–10 random fields per well under 1603magnifica- HS (Gibco) 46K6762 290.2
HS (Sigma) 24H0982 273.7
tion in each condition. Experiments were carried out HS (Gibco) 33P4333 319.1
simultaneously on cultures from the same preparation HS (Equitech) SE30-215 359.4
BCS (Hyclone) 21S2595 316.8
plated at the same density in 24-well plate. Each CS (Gibco) 39N0942 202.6
treatment group consisted of a minimum of two wells aFBS, fetal bovine serum; FCS, fetal calf serum; HS, horse serum; BCS, bovine
(2–6). Culture media was changed to new GDM (0.5 ml calf serum; CS, calf serum. Average: FBS(FCS): 931.6 6 58.9 µM, n 5 7; HS:
per well) supplemented with defined [Glu] either from 328.8 6 21.9 µM, n 5 7.
serum or from glutamate stock. Cell counts were ob-
tained 24–48 h after glutamate treatment; at this time,
the neuronal degeneration was readily visible. Trypan obtain culture medium in which glutamate was de-
blue excluded, process-bearing neurons are considered pleted by uptake into astrocytes. In a recent develop-
as viable. mental study we showed that the relative rate of
glutamate uptake in hippocampal astrocytes varies as a
function of pre- and post-natal development (unpub-
RESULTS lished data) and that these differences persisted even if
Sera Used as Culture Supplement Contains cells were cultured for several weeks. Specifically we
Glutamate at High Concentrations found that hippocampal astrocytes obtained from post-
natal 2–7 days pups had the highest glutamate uptake
Most investigators who culture CNS neurons supple- capability. Consequently we used confluent cultures of
ment their culture medium with 5–20% serum, most P2 hippocampal astrocytes grown in 75 mm2 flasks for
commonly 10% (v/v), and use either fetal calf (bovine) or 1–3 weeks to obtain glutamate depleted medium. To
horse serum. By using the bioluminescence assay (Fosse determine the rate of glutamate depletion from the
et al., 1986), which allows the reliable detection of medium, 30 ml fresh medium containing 10% FBS
sub-µM concentrations of glutamate, we determined (Hyclone, Logan, UT) was applied to two flasks, and
the free [Glu] in several samples of sera from various subsequently 0.1 ml media was sampled for glutamate
batches obtained from Hyclone (Logan, UT), Gibco content after various time periods ranging from 5 min
(Grand Island, NY), Sigma Chemical Co. (St. Louis, to 10 h. The resulting [Glu] were plotted as a function of
MO), Atlanta Biological (Norcross, GA), Summit Bio- time after addition of media to the flask (Fig. 1). [Glu] in
technology (Ft. Collins, CO), and Equitech-Bio (Ingram, the medium showed a rapid initial decline, and within
TX). Thus, we determined glutamate concentration in 30 min, concentrations dropped from 85 µM to 20 µM.
seven different fetal bovine (calf ) sera (FBS) and seven After 2–3 h, [Glu] reached values below 1 µM. By
different horse sera (HS). [Glu] in these sera are listed contrast, the [Glu] of media maintained in cell free
in Table 1. Mean [Glu] (6SE) were FBS (931.6 6 58.9 flasks under otherwise identical conditions did not
µM), n57 batches, ranging from 808.2 µM to 1195.7 µM change after this period and remain unchanged even
and HS (328.8621.9µM), n57 batches, ranging from after 2 days, the longest time tested. By using this
273.7 µM to 434.4 µM. For each batch, measurements procedure, we effectively used astrocytes to produce
were obtained at least in duplicates. Interestingly, glutamate-depleted media (GDM).
[Glu] were consistently higher in fetal bovine (calf ) To ensure that the time-course of this glutamate
serum than in horse serum and calf serum. depletion was reproducible and a reliable means to
produce GDM, we studied repeated applications of
glutamate in the same cell preparation. As described
Serum Glutamate Can Be Effectively Depleted above, the initial glutamate challenge came from serum-
by Cultured Astrocytes containing medium. The two subsequent challenges
(Fig. 2, arrows) used exogenous glutamate from stock
It is well established that astrocytes possess effective solutions. As illustrated in Figure 2, glutamate deple-
transport mechanisms for the uptake of glutamate from tion was highly reproducible. At the time points indi-
the extracellular space (Rothstein et al., 1996). Most of cated by arrows, 0.27 ml of a 10 mM glutamate stock
this transport occurs through a Na1-dependent co- solution were applied to the glutamate depleted flask
transporter (Kanai et al., 1993). We thus set out to followed by a 10 s mixture. The media was immediately
 NEUROTOXICITY BY SERUM GLUTAMATE
Fig. 1. Depletion of serum glutamate by astrocytes. Culture media
(30 ml) containing 10% fetal bovine serum was added into two 75 cm2
flasks containing confluent hippocampal astrocytes from P2 rats.
Samples of media (0.1 ml) were taken at the time points indicated and
the glutamate content determined. [Glutamate] dropped below 1 µM
within 2–3 h. The inset of Figure 1 expands the part of the curve
(. 1 h) that corresponds to the most physiologically relevant gluta-
mate concentrations (mean 6 SD, representative of seven experi-
ments). No [Glutamate] changes were observed in cell free control
flasks.
sampled and the concentration-time curve was plotted.
The time course for depletion of subsequent addition of
exogenous glutamate was actually identical to the
initial depletion of serum-borne glutamate. These data
demonstrate that production of GDM by astrocytic
culture occurs in a reliable and stereotypic manner.
Moreover, a time period of 2–3 h was sufficient for
absorbing essentially all glutamate supplied to the
media by a 10% serum complement. Considering the
variants in astrocytic glutamate uptake capacity and
possible heterogeneity in [Glu] within a flask (which
may have escaped our sampling procedure), we usually
used 6–10 h intervals for harvesting GDM.
Effects of Serum Glutamate and
Neuronal Survival
We used time-lapse video microscopy whereby im-
ages were captured once every second to assess chronic
and acute effects of serum glutamate on neuronal
morphology and survival. For these experiments we
started out by culturing mixed neuronal/glial cultures
in GDM. Several representative hippocampal neurons
241
Fig. 2. Glutamate depletion is reproducible and independent of
glutamate source. Ten percent FBS media (30 ml) was applied to a
flask of confluent P2 astrocytes. Subsequently, two additional applica-
tions were made (arrows) from a stock glutamate solution (0.27 ml 3
10 mM) at 6 h and 12.5 h, respectively. Astrocytes depleted exogenous
glutamate in the same manner as serum glutamate.
growing on a bed-layer of astrocytes for 14 days in the
presence of GDM are visible in Figure 3A. This culture
was monitored in the time-lapse system for 25 h
without any noticeable neuronal damage (Fig. 3B).
Media was then replaced with fresh GDM (2.5 ml) for
another 24 h (Fig. 3C,D), still without any signs of
neurotoxicity. Subsequently, the media was switched to
2.5 ml untreated 10% FBS media (UM) (previously
maintained in the incubator over 2 h to equilibrate pH
and temperature). As visible in Figure 3E, 5 min after
this media change, some morphological changes were
already visible in the neuronal soma (arrows). These
changes, indicative of neuronal swelling, become more
noticeable over the subsequent hour of incubation (Fig.
3F), and most neurons showed pronounced swelling
and a diminishing number of processes. These signs of
neuronal damage progressed (Fig. 3G) with longer-
term exposure (8 h) and almost all neurons had com-
pletely disintegrated and disappeared by 24 h (Fig. 3H).
In contrast to neurons, we did not observe changes in
the viability of the bed-layer astrocytes under the same
experimental conditions.
In order to further characterize the difference be-
tween GDM and untreated media (UM) on neuronal
viability, the noncompetitive NMDA receptor antago-
nist MK801 was utilized (Fig. 4). A plate of 18 DIV
mixed cultures similar to those shown in Figure 3 was
242 YE AND SONTHEIMER

Fig. 3. Time lapse microscopy recordings of neurons before and
after exposure to GDM and regular 10% FBS media. A,B: A 14 DIV
culture was maintained under time-lapse observation in GDM for 24
h; C,D: Fresh GDM was applied for another 24 h; E–H: 10% FBS
media was added (E: 5 min after, arrows point to neurons with visible
morphology changes; F: 67 min after; G: about 8 h after; H: 24 h after).
No neurotoxicity was observed while cells were maintained in GDM,
but cells rapidly died after exposure to serum-containing media (E–H).
A–H: Scale bar: 100 µm.
 NEUROTOXICITY BY SERUM GLUTAMATE
Fig. 4. MK-801 protects against neurotoxicity from media changes.
An 18 DIV culture received either no media exchange (control) or
complete media exchange by glutamate depleted media (GDM), un-
treated media (UM), or UM 1 5 µM MK801 (mean 6 SE, three wells
per group, fields520). MK801 protection suggests an NMDA receptor-
mediated neurotoxicity. ***P,0.001, ANOVA.
switched to either GDM, UM, or UM 1 MK801 (5 µM)
and cell viability determined 24 h later. Neurons that
received a GDM media change did not differ from
non-media changed control as assessed by morphologi-
cal evidence of damage (see above). However, almost all
neurons that received untreated media died. Those that
receive MK801 simultaneously with untreated media
completely escaped from the toxic effect of untreated
media. This demonstrates that in untreated media
glutamate is the primary toxic factor and that neurotox-
icity is mediated by MK801-sensitive NMDA receptors.
To more quantitatively compare the effect of the
serum-born glutamate and exogenous glutamate, we
established glutamate toxicity dose-response curves on
sister cultures treated with known concentrations of
either serum glutamate or stock glutamate. A represen-
tative example obtained from neurons maintained in
culture for 20 days in GDM is shown in Figure 5.
Increasing glutamate concentrations resulted in the
dose-dependent killing of neurons without any notice-
able difference between serum glutamate and exog-
enous glutamate. These results further suggest that
serum-born glutamate is the principal neurotoxic factor
in serum containing media.
Partial media change and/or reduced percentage
serum supplement have been frequently used in main-
taining neuronal culture (Black et al., 1995; Buchhalter
and Dichter, 1992; Jahr and Stevens, 1987; Juurlink
and Hertz, 1993). Based on our data (above), we would
expect that the efficacy of this approach is due to
reduced concentrations of glutamate delivered by par-
243
Fig. 5. Serum glutamate toxicity is indistinguishable from toxicity
by exogenous glutamate. Sister cultures of neurons (20 DIV) received
glutamate either from serum or from glutamate stock solution.
Neurotoxicity was determined 24 h after receiving glutamate at the
concentrations indicated. With increasing [Glu], viable neuronal cell
counts declined. Glutamate from these two sources were equally toxic.
ED50 5 10 µM. Mean 6 SE, fields 5 14–16.
tial serum supplement. To better compare our studies
to those published, we examined the effects of partial
media changes on mixed hippocampal culture. We used
partial media changes ranging from 25 to 75% and
included cultures maintained for 1–3 weeks in these
studies. Not unexpectedly, time-lapse microscopy re-
vealed that neuronal survival was dependent on the
quantity of glutamate delivered by the media ex-
changes. A representative example of a 21 DIV culture
that received 30 µM serum-born glutamate due to a
partial media change is shown in Figure 6. Twenty-four
hours later (Fig. 6B), most neurons in the field disinte-
grated but at least one survived (arrow). Some viable
neurons were also found in other areas on the same
dish (Fig. 6C,D, different fields from Fig. 6A,B). Cells
that disintegrated in response to glutamate challenges
(Fig. 6B) typically had complicated networks of pro-
cesses, while those that were more glutamate resistant
had relatively simple processes and fewer contacts to
adjacent cells. Interestingly, the extent to which media
changes affected neuronal survival also varied with
culture age with young cultures being less sensitive to
media changes than older cultures. Figure 7 summa-
rizes data from cultures of different age that received
either GDM or partial media changes (70% GDM 1 30%
UM) every 3 days after plating. Neuronal number in
cultures receiving partial media changes were signifi-
cantly lower than those receiving GDM only. Most
244 YE AND SONTHEIMER
Fig. 6. Serum glutamate selectively eliminates subpopulation of
neurons. A: A 21 DIV culture 2 min prior to 30 µM glutamate from
serum; B: 24 hours later, the soma and processes of most neurons were
disintegrated; the arrow points to an intact neuron. The few surviving
notably, at 21 DIV, glutamate from partial UM (30%)
was sufficient to kill the majority of neurons, whereas
cultures received GDM had 10-fold more viable neu-
rons.
Cells Cultured in Serum Show Higher ED50
Values for Glutamate Than Cells Cultured
in Glutamate Depleted Medium
Since we found that even the glutamate content
delivered by partial media changes was sufficient to kill
neurons in culture, we must assume that media changes
eliminate subpopulations of neurons and consequently
lead to the selection of neurons that show increased
tolerance to glutamate. Therefore, we would expect the
apparent ED50 values for glutamate toxicity deter-
mined in such cultures should differ from those ob-
tained in cultures maintained in GDM. We thus deter-
mined the relative toxicity of glutamate in 16 DIV sister
cultures that were either maintained in GDM through-
out, or that received 70% GDM 1 30% UM, or GDM
125 µM stock glutamate every 3 days starting at 5 DIV
(Fig. 8a). GDM-treated cultures showed the highest
neurons (C,D) found in the same dish after 24 h glutamate exposure
showed less elaborate processes and appeared to make fewer contacts
with each other.
glutamate susceptibility with ED50 values of 30 µM. We
observed a shift in the glutamate toxicity dose-response
curve towards higher [Glu] with cultures that received
either partial UM (GDM 1 30% UM) or added gluta-
mate in media changes (GDM 1 25 µM glutamate). The
latter two conditions did not differ and both had ED50
values close to 60 µM, twice the value observed for
GDM cultures.
In our glutamate analysis we found that horse or calf
serum, used by some investigators for culturing neu-
rons (Black et al., 1995; Choi et al., 1987; Rosenberg
and Aizenman, 1989; Yool et al., 1988), contains signifi-
cantly lower [Glu] than fetal calf serum (Table 1),
suggesting that these sera may be a less toxic culture
supplement. To test whether cell viability and subse-
quent susceptibility to glutamate is enhanced if cells
are cultured in these sera, we compared relative gluta-
mate toxicity in sister cultures after they had been
maintained for 22 DIV in media supplemented with
either glutamate depleted calf serum (CSGDM) or
untreated calf serum (CSM). Neurons maintained in
untreated calf serum showed significantly higher ED50
values for glutamate than cells maintained in gluta-
 NEUROTOXICITY BY SERUM GLUTAMATE
Fig. 7. Partial media changes lead to lower neuron yields than
GDM. Sister cultures at the same density received GDM or partial
untreated media (70% GDM 1 30% UM) every 3 days after 5 DIV.
Groups receiving GDM 1 UM had fewer viable neurons. (Mean 6 SE,
neurons counted from 11–13 random fields.) *P,0.05, **P,0.01,
***P,0.001, Student’s t-test for GDM vs. GDM 1 UM.
mate depleted calf serum (Fig. 8b). Similar results were
obtained with horse serum (data not shown).
Susceptibility to Glutamate-Neurotoxicity
Increases With Development In Vitro
In our time-lapse studies, we consistently observed
that older cultures were more sensitive to glutamate
than younger cultures. To determine whether pro-
longed culturing makes neurons more susceptible to
neurotoxic injury, we used sister cultures from the
same preparation and repeated the glutamate toxicity
experiments described above on cultures ranging from
6 to 42 DIV and obtained complete dose-response
curves for glutamate induced cell death. For each age
group, we determined the relative ED50 value for
glutamate and plotted these values as a function of in
vitro age (Fig. 9). These data clearly suggest a develop-
mental change in the sensitivity to glutamate such that
neurons are much less susceptible to glutamate-
induced cell death during the first week after plating
but become highly susceptible in long-term cultures.
Indeed, during the first 4 days in culture, neurons
consistently were resistant to serum-borne glutamate.
Exposure of these immature neurons to 10% serum
media or equal concentrations of glutamate (e.g., 80–
100 µM) did not significantly damage these cells (data
not shown). The ED50 values rapidly declines in the
second week in vitro, and after 3 weeks or longer, most
cells were highly susceptible to glutamate with mean
245
ED50 values around 6 µM. Similar developmental
changes in the susceptibility of neurons to glutamate
have been documented for other preparation (Mattson
et al., 1991a,b; Choi et al., 1987).
DISCUSSION
Most laboratories that study neurons in culture rely
on the use of sera from either bovine (calf ) or horse to
supply growth factors and trace elements that benefit
cell growth in vitro. However, frequent media changes
containing serum have been shown to be detrimental to
neuronal growth and cell viability. We determined the
glutamate content in sera from the most commonly
used sera for neuronal and glial cell cultures and
assayed several batches of bovine (calf ) and horse
serum from different sources. We found that [Glu] were
remarkable high, ranging from 273.7 µM to 1195.7 µM,
concentrations that would be expected to be neurotoxic
if used as a 10% media supplement. Indeed, our own
assessment of the viability of hippocampal neurons
suggested that these serum-glutamate concentrations
were highly toxic and resulted in significant loss of
neurons. As shown by others before (Choi et al., 1987;
Oka et al., 1993), the survival of astrocytes that are also
present in our cultures was not affected by serum-
glutamate. Since astrocytes are capable of sequestering
extracellular glutamate by Na1-dependent uptake, we
explored their use as ‘‘glutamate sponges’’ to absorb
serum-glutamate and thus detoxify the serum. We
achieved this by using astrocyte cultures exposed to
serum-containing medium for several hours, resulting
in a highly stereotypic depletion of serum-borne gluta-
mate to levels less than 1 µM, a concentration similar to
that reported for CSF (Nicholls and Attwell, 1990).
Interestingly, 3 h exposures of 10% serum contained in
40 ml media to confluent astrocytes in a 75 mm2 flasks
was sufficient to reduce glutamate from ,90 µM
to , 1 µM.
We since cultured hippocampal neurons in GDM and
found excellent long-term survival, despite frequent
media changes. Moreover, subsequent treatment of
hippocampal cultures with defined [Glu] from either a
stock solution or serum-glutamate resulted in marked
and concentration dependent neuronal death. The rela-
tive sensitivity of these neurons to glutamate was
dependent on their in vitro age, with ED50 values of
about 6 µM for long-term cultures (. 28 DIV) and ED50
values of . 16 µM for cells , 14 DIV.
A large body of literature now supports the notion
that most CNS neurons are highly sensitive to gluta-
mate with low µM concentrations being sufficient to kill
most neurons (Choi et al., 1987; Choi, 1988; Michaels
and Rothman, 1990). It is well established that CSF
[Glu] in the healthy brain is maintained close to 1 µM
(Nicholls and Attwell, 1990). Moreover, much of the
control of extracellular glutamate is attributed to glial
rather than neuronal uptake. It is thus not surprising
that in vitro, astrocytes would reduce glutamate to
246 YE AND SONTHEIMER
Fig. 8. Serum-glutamate in the culture media reduces glutamate
sensitivity. A: The glutamate dose-toxicity relationship of 16 DIV
hippocampal neurons that had been cultured (changing media every 3
days) either in GDM, 70% GDM 1 30% untreated media (UM) or
GDM 1 25µM glutamate illustrates that prior exposure to glutamate
reduces its neurotoxic effect. Serum-born glutamate or exogenous
glutamate were equally effective in producing a shift of the curve to
Fig. 9. Vulnerability of neurons to glutamate increases with in vitro
age. Relative toxicity of glutamate was assessed in sister cultures of
neurons from the same preparation maintained in vitro for the times
indicated. The ED50 for glutamate toxicity was determined for each
age group and plotted as a function of in vitro age. Vulnerability to
glutamate developed in the first 20 DIV.
higher levels (cells counted from 12–16 random fields). B: Glutamate
dose-toxicity curve of 22 DIV hippocampal neurons cultured in 10%
calf serum media (CSM) or glutamate depleted calf serum media
(CSGDM) with an ED50 around 60 µM and 30 µM, respectively. Data
were expressed as Mean 6 SE, fields515–18. *P,0.05, ** P,0.01,
***P,0.001. Comparisons were made between groups exposed to
same [Glu]. A, ANOVA; B, Student’s t-test.
concentrations comparable to that in CSF. However, in
the brain, the extracellular space is very small compris-
ing , 20% of the total volume. By comparison, in our in
vitro situation where extracellular volume is much
greater than the cell volume, it is remarkable that
astrocytes were able to continually sequester large
amounts of glutamate in a matter of hours. The ‘‘spong-
ing’’ of glutamate from the media is likely to be compan-
ioned by the cytoplasmic conversion of glutamate to
glutamine, a-ketoglutarate and other metabolites (Hertz
and Schousboe, 1986).
The beneficial role of astrocytes to neuronal cultures
has been known for some time. Rosenberg et al. (1989)
determined that the neuronal susceptibility to gluta-
mate-toxicity differed depending on the presence or
absence of glial cells. They determined an ED50 values
for glutamate in 3–5-week-old astrocyte-rich neuronal
cultures (ED50 194 6 43 µM) that is much higher than
the results of present study. These astrocyte-rich neuro-
nal cultures received three media changes per week
with medium containing 10% calf serum. It is interest-
ing to note that horse serum and calf serum contains
several fold lower glutamate and may thus be a better
supplement than fetal bovine (calf ) serum. Indeed, a
number of labs utilized horse serum instead of fetal calf
(bovine) serum for maintaining neuronal cultures (Choi
et al., 1987; Black et al., 1995; Morioka et al., 1995;
 NEUROTOXICITY BY SERUM GLUTAMATE
Reif, 1993; Yool et al., 1988). When we repeated experi-
ments in which we cultured hippocampal astrocytes in
either GDM or calf serum, we found that neurons and
glial cells fed with calf serum established different
morphology from those fed by GDM. In phase-images,
neurons in GDM showed more elaborate processes and
appeared to float on the astrocytic bed-layer. By compari-
son, neurons grown in CSM stayed individual, had
shorter process, and integrated much more in the
astrocyte layer (data not shown). Harris and Rosenberg
(1993) established mixed culture in calf serum and
suggested that astrocytes may cover synapses, thereby
protecting them from media or glutamate exposure;
thus those tighter association of astrocytes and neurons
could have accounted for the high apparent resistant of
those neurons to glutamate. However, in addition, our
data (Fig. 8b) would suggest the involvement of selec-
tion for more glutamate tolerant neurons by the re-
peated exposure to untreated calf serum, which will
eliminate those neurons that are more sensitive to
glutamate. Similarly, positive selection for more gluta-
mate resistant populations of neurons may have re-
sulted from media changes using untreated horse se-
rum (Black et al., 1995; Morioka et al., 1995; Reif, 1993;
Yool et al., 1988) and may have contributed to the
relatively high ED50 values of 50–100 µM as reported
(Choi et al., 1987).
Glial conditioned medium has been used in some
studies arguing that astrocytes promote neuronal sur-
vival by the release of neurotrophic factors (Dugan et
al., 1995). Undoubtedly, astrocytes are a rich source for
trophic factors (Raju et al., 1994; Schwartz and Nishi-
yama, 1994) and likely are an important source for
neurotrophic factors in vivo. However, our studies
would suggest that the primary beneficial effect of
culturing neurons in astrocyte-conditioned serum con-
taining medium results from the glutamate detoxifica-
tion, as within 2–3 h, astrocytes can sequester . 99% of
glutamate contained in the culture medium of a rela-
tively large volume.
CONCLUSION
The present study demonstrates that serum supple-
ments contain significant concentrations of glutamate,
sufficient to kill most mature neurons. Serum-gluta-
mate is the principal toxic factor to neurons and can
explain the reduced neuronal viability in cell cultures
repeatedly exposed to serum-containing media. Neu-
rons cultured in serum-containing media will result in
selection of subpopulations that are more glutamate
resistant. If used for pharmacological studies evaluat-
ing neurotoxcicity or protection from neurotoxic injury,
these cells will probably not adequately represent in
vivo conditions. Glutamate-depleted media can be
readily prepared by a 3–6 h incubation of serum-
containing media in culture flasks with astrocytic mono-
layers, resulting in media that contains glutamate at
247
concentrations similar to that in cerebro-spinal fluid. If
grown in GDM, neuronal survival was much enhanced
and promoted establishment of complex processes. In
addition, cells exhibited a much higher sensitivity to
exogenous glutamate. Such cultures may be more repre-
sentative of neurons in vivo and may provide an
alternative model for future neurotoxicity studies.
ACKNOWLEDGMENTS
We thank Dr. Michael J. Friedlander, Felecia Hester,
and Carolyn D. Gancayco in the Department of Neuro-
biology for their help with bioluminescence measure-
ments of glutamate.
REFERENCES
Black, M.A., Tremblay, R., Mealing, G., Ray, R., Durkin, J.P., Whit-
field, J.F., Blosser, J., and Morley, P. (1995) N-methyl-d-aspartate- or
glutamate-mediated toxicity in cultured rat cortical neurons is
antagonized by fpl 15896ar. J. Neurochem., 65:2170–2177.
Bottenstein, J.E. (1992) Environmental influences on cells in culture.
In: Practical Cell Culture Techniques. A.A. Boulton, G.B. Baker, and
W. Walz, eds. Totowa, New Jersey: The Human Press, Inc., pp.
63–85.
Buchhalter, J.R. and Dichter, M.A. (1992) Neurons. In: A.A. Boulton,
G.B. Baker, and W. Walz, eds. Practical Cell Culture Techniques.
Totowa, New Jersey: The Human Press, Inc., pp. 241–268.
Cambier, D., Aı̈t-Ikhlef, A., Parvy, P., Hantaz-Ambroise, D., Murawsky,
M., and Rieger, F. (1995) Excess extracellular and low intracellular
glutamate in poorly differentiating wobbler astrocytes and astrocyte
recovery in glutamine-depleted culture medium. J. Neurochem.,
65:1199–1204.
Choi, D.W., Maulucci-Gedde, M., and Kriegstein, A. (1987) Glutamate
neurotoxicity in cortical cell culture. J. Neurosci., 7:357-368.
Choi, D.W. (1988) Glutamate neurotoxicity and diseases of the nervous
system (Review). Neuron, 1:623–634.
Choi, D.W., Maulucci, G.M. and Kriegstein, A.R. (1987) Glutamate
neurotoxicity in cortical cell culture. J. Neurosci., 7:357–368.
Dichter, M.A. (1978) Rat cortical neurons in cell culture: Culture
methods, cell morphology, electrophysiology, and synapse forma-
tion. Brain Res., 149:279–293.
Driscoll, B.F., Deibler, G.E., Law, M.J., and Crane, A.M. (1993)
Damage to neurons in culture following medium change: Role of
glutamine and extracellular generation of glutamate. J. Neuro-
chem., 61:1795–1800.
Dugan, L.L., Bruno, V.M.G., Amagasu, S.M., and Giffard, R.G. (1995)
Glia modulate the response of murine cortical neurons to excitotox-
icity: Glia exacerbate AMPA neurotoxicity. J. Neurosci., 15:4545–
4555.
Fosse, V.M., Kolstad, J., and Fonnum, F. (1986) A bioluminescence
method for the measurement of l-glutamate: applications to the
study of changes in the release of l-glutamate from lateral genicu-
late nucleus and superior colliculus after visual cortex ablation in
rats. J. Neurochem., 47:340–349.
Harris, K.M. and Rosenberg, P.A. (1993) Localization of synapses in
rat cortical cultures. Neuroscience, 53:495–508.
Hertz, L. and Schousboe, A. (1986) Role of astrocytes in compartmenta-
tion of amino acid and energy metabolism. In: Astrocytes: Biochemis-
try, Physiology, and Pharmacology of Astrocytes. S. Fedoroff and A.
Vernadakis, eds. Orlando: Academic Press, Inc.
Ince, C., Ypey, D.L., Diesselhoff-Den, D., Viser, J.A.M., De, V., and Van,
F. (1983) Micro-CO2-incubator for use on a microscope. J. Physiol.
(London), 60:269–275.
Jahr, C.E. and Stevens, C.F. (1987) Glutamate activates multiple
single channel conductances in hippocampal neurons. Nature, 325:
522–525.
Juurlink, B.H. and Hertz, L. (1993) Ischemia-induced death of astro-
cytes and neurons in primary culture: Pitfalls in quantifying
neuronal cell death. Dev. Brain Res., 71:239–246.
248 YE AND SONTHEIMER
Juurlink, B.H.J. and Hertz, L. (1992) Astrocytes. In: Practical Cell
Culture Techniques, A.A. Boulton, G.B. Baker, and W. Walz, eds.
Totowa, New Jersey: The Human Press, Inc., pp. 269–321.
Kanai, Y., Smith, C.P., and Hediger, M.A. (1993) The elusive transport-
ers with a high affinity for glutamate. (Review). Trends Neurosci.,
16:365–370.
Langan, T.J., Slater, M.C., and Kelly, K. (1994) Novel relationships of
growth factors to the G1/S transition in cultured astrocytes from rat
forebrain. Glia, 10:30–39.
Lascola, C.D., and Kraig, R.P. (1996) Whole-cell chloride currents in
rat astrocytes accompany changes in cell morphology. J. Neurosci.,
16:2532–2545.
Mattson, M.P., Rychlik, B., You, J.S., and Sisken, J.E. (1991a) Sensitiv-
ity of cultured human embryonic cerebral cortical neurons to
excitatory amino acid-induced calcium influx and neurotoxicity.
Brain Res., 542:97–106.
Mattson, M.P., Wang, H., and Michaelis, E.K. (1991b) Developmental
expression, compartmentalization, and possible role in excitotoxic-
ity of a putative NMDA receptor protein in cultured hippocampal
neurons. Brain Res., 565:94–108.
Michaels, R.L. and Rothman, S.M. (1990) Glutamate neurotoxicity in
vitro: Antagonist pharmacology and intracellular calcium concentra-
tions. J. Neurosci., 10:283–292.
Miller, R.H. and Raff, M.C. (1984) Fibrous and protoplasmic astrocytes
are biochemically and developmentally distinct. J. Neurosci., 4:585–
592.
Morioka, M., Fukunaga, K., Nagahiro, S., Kurino, M., Ushio, Y., and
Miyamoto, E. (1995) Glutamate-induced loss of Ca21/calmodulin-
dependent protein kinase II activity in cultured rat hippocampal
neurons. J. Neurochem., 64:2132–2139.
Nicholls, D. and Attwell, D. (1990) The release and uptake of excita-
tory amino acids [see comments]. (Review). Trends Pharmacol. Sci.,
11:462–468.
Oka, A., Belliveau, M.J., Rosenberg, P.A., and Volpe, J.J. (1993)
Vulnerability of oligodendroglia to glutamate: Pharamacology,
mechanisms, and prevention. J. Neurosci., 13:1441–1453.
Raff, M.C. (1989) Glial cell diversification in the rat optic nerve.
Science, 243:1450–1455.
Raff, M.C., Miller, R.H., and Noble, M. (1983) A glial progenitor cell
that develops in vitro into an astrocyte or an oligodendrocyte
depending on culture medium. Nature, 303:390–396.
Raju, T.R., Rao, M.S., Nagaraja, T.N., Meti, B.L., and Schulz, M. (1994)
Retinal ganglion cell survival and neurite regeneration in vitro after
cell death period are dependent upon target derived trophic factor
and retinal glial factor(s). Brain Res., 664:247–251.
Reif, D.W. (1993) Delayed production of nitric oxide contributes to
NMDA-mediated neuronal damage. Neuroreport, 4:566–568.
Rosenberg, P.A. (1991) Accumulation of extracellular glutamate and
neuronal death in astrocyte-poor cortical cultures exposed to gluta-
mine. Glia, 4:91–100.
Rosenberg, P.A. and Aizenman, E. (1989) Hundred-fold increase in
neuronal vulnerability to glutamate toxicity in astrocyte-poor cul-
tures of rat cerebral cortex [published erratum appears in Neurosci.
Lett., 1990, 116:399]. Neurosci. Lett., 103:162–168.
Rothstein, J.D., Dykes-Hoberg, M., Pardo, C.A., Bristol, L.A., Jin, L.,
Kuncl, R.W., Kanai, Y., Hediger, M.A., Wang, Y.F., Schielke, J.P., and
Welty, D.F. (1996) Knockout of glutamate transporters reveals a
major role for astroglial transport in excitotoxicity and clearance of
glutamate. Neuron, 16:675–686.
Rothstein, J.D., Martin, L., Levey, A.I., Dykes-Hoberg, M., Jin, L., Wu,
D., Nash, N., and Kuncl, R.W. (1994) Localization of neuronal and
glial glutamate transporters. Neuron, 13:713–725.
Schwartz, J.P. and Nishiyama, N. (1994) Neurotrophic factor gene
expression in astrocytes during development and following injury.
Brain Res. Bull., 35:403–407.
Thomas, R.J. (1995) Excitatory amino acids in health and disease.
(review). J. Am. Geriatr. Soc., 43:1279–1289.
Ye, Z.C. and Sontheimer, H. (1996) Cytokine modulation of glial
glutamate uptake: A possible involvement of nitric oxide. Neurore-
port, 7:2181–2185.
Yool, A.J., Dionne, V.E., and Gruol, D.L. (1988) Developmental changes
in K1-selective channel activity during differentiation of the Pur-
kinje neuron in culture. J. Neurosci., 8:1971–1980.