Rapid desensitization determines the pharmacology of glutamate neurotoxicity

0028-3908(94)EOO42-P

Pergamon

Neuropharmacolog~ Vol. 33, No. 8, pp. 953-962, 1994 Copyright 0 1994 Elsevier Science Ltd Printed in Great Britain. All rights reserved 0028-3908/94 $7.00 + 0.00

Rapid Desensitization Determines the Pharmacology of Glutamate Neurotoxicity A. M. MOUDY,

K. A. YAMADA

and S. M. ROTHMAN*

Departments of Neurology and Anatomy and Neurobiology, Washington University School of Medicine and St Louis Children’s Hospital, St Louis, MO 63110, U.S.A. (Accepted 30 March 1994) Summary-Glutamate (Glu), the major excitatory neurotransmitter in the nervous system, is toxic to neurons when it accumulates at high concentrations in the extracellular space. Even though Glu is a mixed agonist, capable of activating N-methyl-D-aspartate (NMDA) receptors and non-NMDA receptors, in many preparations Glu neurotoxicity is prevented by selective blockade of NMDA receptors. In cultures of hippocampal neurons, treatment with 500 PM Glu for 30 min killed more than 90% of the neurons. The simultaneous addition of the selective NMDA agonist methyl-lo,1 I-dihydro-5-H-dibenzocyclohepten-5,10-imine (MK-801) reduced the cell loss to less than 30%. However, when Glu was combined with either diazoxide or cyclothiazide, two thiazides which dramatically diminish rapid Glu desensitization, MK-801 was no longer very protective and neuronal loss exceeded 80%. However, the non-NMDA antagonist 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX), in combination with MK-801, was able to prevent most Glu neurotoxicity in the presence of these thiazides. These experiments show that there are circumstances under which Glu neurotoxicity is produced by overactivation of non-NMDA receptors. Our observations offer a possible explanation for the recent finding that blockade of non-NMDA receptors is much more beneficial than NMDA receptor blockade in protecting the brain in some in uivo models of global ischemia. Keywords-Glutamate,

excitotoxicity,

NMDA,

cell death,

cerebral

ischemia,

calcium,

desensitization.

constant in the millisecond range (Kiskin et al., 1986; Mayer and Vyklicky, 1989). This desensitization alters the time course of synaptic currents and potentials (Trussel and Fischbach, 1989; Tang et al., 1991; Yamada and Rothman, 1992; Yamada and Tang, 1993). A variety of plant lectins can partially attenuate Glu desensitization at invertebrate and vertebrate Glu receptors (Mayer and Vyklicky, 1989; Thio et al., 1992). We recently found much more dramatic reduction of Glu desensitization with several thiazide compounds (Yamada and Rothman, 1992; Yamada and Tang, 1993). If Glu accumulates for a sufficiently long time in the extracellular space, it can reach concentrations that are neurotoxic (Olney, 1983). Recent in vitro studies have suggested that Glu neurotoxicity is largely produced by NMDA receptor activation. If NMDA receptors are selectively blocked, the toxicity of Glu is dramatically reduced (Choi et al., 1988; Michaels and Rothman, 1990). Because Glu toxicity may be an important factor in a variety of devastating neurological diseases, including stroke and anoxic neuronal damage (Choi and Rothman, 1990) we decided to investigate whether eliminating rapid Glu desensitization might change the pharmacological characteristics of Glu neurotoxicity.

There is now compelling evidence that glutamate (Glu) is the major fast excitatory neurotransmitter in the central nervous system (Collingridge and Lester, 1989). When released at synaptic terminals, Glu acts as a mixed agonist capable of activating at least four identifiable receptor subtypes: a-amino-3-hydroxy-5-methyl-4isoxazolepropionate (AMPA), N-methyl-D-aspartate (NMDA), kainate (KA), and metabotropic. The first three receptors are directly coupled to ion channels and named after the synthetic amino acid most specific for that receptor. The fourth type of receptor is coupled to intracellular second messenger systems and only indirectly gates ion channel opening and closing (Sugiyama et al., 1989). The AMPA and KA receptors are frequently referred to as “non-NMDA receptors” and are probably coupled to the same ionophore. One aspect of Glu receptor neurobiology that has only been recently appreciated is the rapid desensitization of AMPA receptors. In the sustained presence of Glu or AMP.l, inward currents rapidly diminish with a time *To whom correspondence should be addressed at: Department of Neurology, St Louis Children’s Hospital, One Children’s Place, Suite 12E25, St Louis, MO 63110, U.S.A. 953

A. M. Mouny i>f(li.

954

We found that removing desensitization results in an enormous amplification in the toxicity of Glu at the AMPA receptor, such that NMDA antagonists no longer protect from Glu neurotoxicity. AMPA receptor blockade was highly protective in these paradigms. Some of our results have been presented in preliminary form (Moudy ef al., 1992). METHODS

Tissue culture We prepared cultures of dispersed rat hippocampai using previously described techniques. neurons (Huettner and Baughman, 1986; Yamada er ul., 1989; Michaels and Rothman, 1990). For electrophysiology and toxicity experiments we used standard 35-mm tissue culture dishes (Falcon) coated with collagen and polylysine. For the intracellular free calcium determinations we grew the cultures on 25 mm diameter glass coverslips glued over holes drilled into the culture dishes.

Cultures were 15 days-in uitru when used for neurotoxicity experiments. Culture medium was replaced with minimal Essential Medium (MEM, Earle’ salts). For all of these experiments, we preincubated the cultures with the various Glu receptor antagonists, calcium channel blockers and thiazides for at least 2 min prior to adding Glu. All drugs and Glu were added to cultures from 100 x concentrates, unless speci~cally mentioned below in C~~~~cff~~.Cultures were returned to the incubator for 30min. They were then washed three times with Earle’s Basic Salt Solution (EBSS) lacking calcium but supplemented with 1 mM MgCl,. After this, they were placed back into MEM for another 18 h. We then added trypan blue (final concentration 0.4%) for 5 min, washed twice with MEM, and counted stained and unstained neurons in two areas of each culture dish. The counted areas were within two circles about 2 mm in diameter that had been etched into the bottom of the plastic dish. We identified areas to count prior to carrying out experiments to prevent introducing bias into our neurotoxicity measurements. We calculated percent viability as the fraction of unstained neurons in each field (Michaels and Rothman, 1990). We generally tested three dishes (six fields) in every individual paradigm and replicated all experimental conditions on at least two separate days. In a few situations, indicated below, we examined Glu neurotoxicity after an 18-h exposure. In these experiments, the cells were rinsed with EBSS, allowed to recover for 30 min in MEM, and then treated with trypan blue.

We used the whole-cell voltage clamp technique initially described by Hamill and colleagues (1981) and applied drugs using a rapid perfusion technique with

large bore (3OO~~m)pipettes (Yamada and Rothman, 1992). The extracellular bath solution contained (mM): IO glycine, 3 KCI, 1 MgCL, 140 NaCl, 2 CaC&. 0.001 tetrodotoxin, and 10 HEPES (to pH 7.3). The pipette solution contained (mM): 140 CsCl, 10 HEPES, 1.l EGTA, 2 Mg-ATPI and 4 glucose. During application of Glu and other drugs, we held the cells at -60 mV using our amplifier (Axoclamp 2A) in the continuoLls single electrode voltage clamp mode, Currents were filtered at 10 Khz, digitized at 0.6 Khz, and stored on hard disk for later analysis (pclamp; Axon Instruments). Intracellular calcium meawements We used the calcium-sensitive fluorescent dye fura(Grynkiewicz et al., 1985) to quantitate free intracellular calcium concentration (Caf”). We added 5 PM of the membrane permeable ester form of fura- (fura-2/AM) to the growth medium and incubated for 45-60 min. We then rinsed the cultures with MEM identical to the solution used in the neurotoxicity experiments, except for the substitution of bicarbonate by 15 mM HEPES to buffer pH to 7.3 in room air and the elimination of phenol red. The cultures were left in this medium and transferred to the heated stage (35°C) of an inverted fluorescence microscope and illuminated with a 75 W xenon lamp. We viewed fields containing several neurons with a 1.3 NA 40 x epifluorescent objective (Nikon). The cultures were alternately illuminated at 340 and 380 nm by having the excitation wavelength switched between two filters in a computer-controlled wheel. The 380 nm filter was placed in series with a 0.6 neutral density filter. The fluorescent signal was filtered at 510 nm and detected by a silicon intensifier tube camera (SIT; Dage model 68) connected to a personal computerbased image acquisition and analysis system (Image 1; Universal Imaging). We were able to do either on- or off-line analysis, since images could be stored to the hard drive or an optical “worm” drive. We measured the 340 and 380 nm fluorescence of a portion of the field lacking any cells to compensate for background fluorescence and subtracted this value from the experimenta images. In order to quantitate Caf+, we had the computer measure the background-corrected 340 and 380 nm fluorescence within regions that outlined the periphery of individual neurons. We used the ratio (R) of the 340/380 nm intensities to determine Ca:+ from the formula Ca’ = &[(R - R,i”)i(R,,, - R)](F,/F,), where the Kd of fura is 224 nM, R,,, and R,,, are fura ratios in zero and saturating concentrations of ionized calcium, and Fd/Fs is the ratio of 380 nm fluorescence intensity in both conditions (Grynkiewicz et al., 1985). We measured R,,,., R,,,, in solutions containing (mM): potassium acetate 130, potassium chloride 10, HEPES 10, EGTA I. 1, and K *-fura 0.02. For measurements of R,,,, an additional IO mM EGTA was added: For R,,,, 10 mM CaCl, was added. Droplets of these solutions on a glass coverslip were imaged with the same

Pharmacology

of glutamate neurotoxicity RESULTS

optical pathways used for the neuronal cultures. A blank lacking K+-furawas used to correct for background fluorescence in this configuration. Under these conditions, R,, = 0.72, R,,, = 23.6 and F,IF, = 20. The imaging was done in a static bath; drugs were added directly to the bath via pipette from concentrated (100 x ) stock solutions. We acquired images at 1 to 2 min intervals prior to drug addition and lo-30 set intervals afterwards, until steady state levels were reached.

As we and others have previously reported, Glu can be very toxic to cultured central neurons (Choi et al., 1988; Michaels and Rothman, 1990). At 500 PM, over 90% of neurons died with either a 30 min or 18 h exposure. The uncompetitive NMDA antagonist MK801 (10 PM) protected most of the neurons from either the short or long duration Glu treatment [Fig. I.(A,B)] Thus, even when exposed to the mixed agonist Glu, blockade of only NMDA receptors was very effective in protecting neurons from injury for up to 18 h. Under these circumstances Glu activation of non-NMDA receptors seems to play a minor role in neuronal damage. The AMPA/KA antagonist CNQX provided little protection against Glu neurotoxicity (Fig. 1). Because we have recently found that certain thiazide derivatives can dramatically diminish rapid desensitization at the AMPA receptor (Yamada and Rothman, 1992; Yamada and Tang, 1993), we decided to investigate whether desensitization also influenced the pharmacology of Glu neurotoxicity. We found that both diazoxide (500 PM) and cyclothiazide (10 FM) had dramatic effects on neuronal death in our paradigms. When cultures were exposed to Glu and MK-801 for only 30 min in the presence of cyclothiazide, over 75% of the neurons died [Fig. 2, 3(A)]. Either diazoxide [Fig. 3(B)] or cyclothiazide [Fig. 2, 3(A,B)] had identical effects with 18 h-Glu exposures. In all three cases, the AMPA/KA

Chemicals We bought all chemicals from Sigma with the following exceptions: fura- came from Molecular Probes; MK-801 was a gift from Merck; 6-cyano-7-nitroquinoxaline-2,3-dione (CNQX) and 7-chlorokynurenate were purchased from Tocris Neuramin; cyclothiazide was generously provided by Lilly; and nimodipine and nifedipine were donated by Miles. The CNQX was dissolved in 5 nM NaOH to make a 5 mM stock that was diluted 1: 100 for the experiments; the nifedipine was dissolved in DMSO to make a 1000 x stock. Statistics All values in figures are means + 1 SD. We used ANOVA followed by the Student-Newman-Keuls test make multiple comparisons within groups.

30 minutes

A

100

Tij .g

60

-

5 .;

s cn ae

40

0

955

5 u.l q 0

-

COtl

GIU

GIU MU

GILI CNQX (20/rM)

GIU CNQX Whll

r

60

-

40

-

0

T

COll

GlU

GIU MU

Glu CNQX

Glu MU CNQX

Fig. 1. Pharmacology of short and long duration Glu neurotoxicity. (A) A 30-min exposure to 500 p M Glu kills over 90% of hippocampal neurons. This toxicity is almost completely eliminated by two minute preincubation with 1OpM MK-801 (MK). However, preincubation with the AMPA/KA antagonist CNQX failed to protect at either 20 or 50 PM (n = 12 for first and third groups, otherwise n = 6). “P < 0.05 compared to both control and all other experimental groups; bnot significantly different from each other. (B) Even with an 18-h 500 PM Glu exposure, the 10 PM MK-801 was still protective. CNQX (20 PM) did not improve survival on its own, but did add to the MK-801 protection (n = 44, 12, 38, 6 and 32 for groups 1-5, respectively). “P < 0.05 compared to all other groups; bnot significantly different from each other. Vertical bars in this and all other figures represent standard deviations.

956

A. M.

MOUDY

rt al.

__”

--___l-.

_ _-

Fig. 2. Cyclothiazide increases the neurotoxicity of Glu. (A) Sequential photographs of the same microscopic field before (Al) and after (A2.3) 30-min treatment with 500 PM Glu + 10 PM MK-801. Al and A2 are phase micrographs; A3 is bright field after trypan blue. There is very little neuronal loss. (B) Identical paradigm as A. except that 10 PM cyclothiazide has been added to the Glu + MK-801. There is now substantial neuronal loss, as indicated by the number of trypan blue stained cells. (C) Same paradigm as A and B, only both cyclothiazide and 50 PM CNQX have been added. C2 and C3 show much less neuronal death. Scale bar: 100 /Lrn A-C.

receptor antagonist CNQX dramatically reduced the neurotoxicity, indicating that it was due to overactivation at non-NMDA receptors. If removal of rapid desensitization augments Glu toxicity at the AMPA receptor/ionophore, then KA, at sufficient concentration and duration, should be neurotoxic on its own. We found that at 1 mM, a 30 min KA exposure killed over 80% of the neurons [Fig. 3(C)]. This injury was not substantially influenced by MK-801, but was blocked by CNQX. Cyclothiazide was relatively potent in our Glu neurotoxicity paradigm, with an IC,” of 3.8 PM (Fig. 4). On its own, cyclothiazide had no neurotoxicity, even with prolonged exposure [Fig. 3(B)]. In addition to potentiating the neurotoxicity of Glu in the presence of MK-801, cyclothiazide was able to reduce the threshold Glu concentration for producing AMPA-like neurotoxicity. A Glu concentration of

100 PM, ordinarily not toxic for 30 min in the presence of MK-801, killed over half of the neurons when combined with cyclothiazide (Fig. 5). We were concerned that the toxicity of Glu in the presence of cyclothiazide or diazoxide might be secondary to some nonspecific injury induced by all thiazide derivatives. However, when we tested two thiazides, chlorothiazide and benzthiazide, that are inactive in electrophysiology experiments on Glu desensitization (Yamada and Tang, 1993), we found no potentiation of Glu neurotoxicity [Fig. 6(A)]. We were also curious whether calcium entry through voltage-gated channels could explain the ability of the thiazides to potentiate Glu neurotoxicity. If this were the case, we might expect calcium channel antagonists to attenuate the neurotoxicity of Glyj’cyclothiazide. Neither nifedipine (at up to 100 PM) or nimodipine provided significant protection in these paradigms. providing no

Pharmacology

A

of glutamate neurotoxicity B

30 minutes

18 hr

100

100

a

a

b

80

i

a .z

957

60

> 5 cl3

40

a9

20

0

Con

Glu MK Cycle

Glu

Glu MK Cycle CNQX (20//M)

MK Diaz

MK Cycle CNQX (50pM)

MK Diaz CNQX

MK Cycle

MK Cycle CNQX

*

KA

C

b

.; ’

60-

2 0 8

40 -

T

i

20 -

0

Con

KA

KA CNQX

KA MK

KA CNQX MK

Fig. 3. Cyclothiazide changes the neuropharmacology of Glu neurotoxicity. (A) The addition of 10 PM cyclothiazide (Cycle) to Glu (SOOpM) for 30min removes the protective effect of MK-801 (MK;lOpM). Protection is restored in a concentration-dependent manner by CNQX (n = 12 for groups 1 and 2; n = 6 for groups 3 and 4). "P < 0.05 compared to cultures lacking CNQX. (B) With an 18-hr Glu exposure in the presence of MK-801, either 500 PM diazoxide (Diaz) or IO PM cyclothiazide increase neurotoxicity dramatically. In both cases CNQX (SOpM) was almost completely protective. Additionally, note that an 1%hr exposure to cyclothiazide alone fails to injure cells (n = 44,8, 8, 12, 12 and 6 for groups 1-6, respectively). "P < 0.05 compared to cultures lacking CNQX; bnot significantly different from control. (C) KA (1 mM) also was neurotoxic in our 30-min paradigm and this injury was blocked by CNQX as well (n = 6). "P < 0.05compared with KA alone; bP < 0.05compared to KA + MK-801.

direct evidence for this particular form of neurotoxicity [Fig. 6(B)]. More likely, the limited Glu desensitization allowed the entry of calcium directly through AMPAgated channels (Pruss et al., 1991). Less likely, the prolonged depolarization seen in cyclothiazide and Glu could be detrimental. Another concern raised by the design of our experiments, was that CNQX protected neurons by providing additional blockade of NMDA receptor-gated channels, beyond the MK-801 antagonism. We and others (Lester et al., 1989; Yamada et al., 1989) have previously shown

that CNQX can block NMDA responses by binding to the glycine site. In order to investigate this possibility, we tested two additional NMDA antagonists in the presence of MK-801 to see whether they attenuated the neurotoxicity of Glu plus cyclothiazide. When either aminophosphonovalerate or 7-chlorokynurenate were combined with MK-801, there was negligible reduction in neuronal death [Fig. 6(C)]. In sister cultures CNQX almost totally protected the neurons. We were curious how well electrophysiological recording would correlate with the neurotoxicity findings. We

A. M. M~UDY et al.

95x 100

80

A

P---T I

F .it L J

[

r

100

\I

60

IC,, = 3&M

r

i;

20 -

0

fCyclo1 in flwl

Fig. 4. Concentration-response curve for cyclothiazide (Cycle) in Glu neurotoxicity. As the concentration of cyclothiazide is increased, the 30-min survival in Glu {SO0PM) + MK-801 (IO PM progressively decreases. Cell counts are done at 18 h.

GIU MK Chloro

GIU MK Benz

B

examined currents produced by rapid application of 500pM Glu alone, Glu + MK-801 (10 PM), and Glu + MK-801 + cyciothiazide + CNQX (20 PM). MK801 does nothing by itself to Glu currents at these concentrations, likely because the concentration of Glu is quite high and the use-dependent block by MK-801 has not had time to develop (Yamada, unpublished). We compared the peak, rapidly desensitizing current and the steady-state current at the end of three different drug combinations in nine neurons (Fig. 7). In the presence of cyclothiazide, the peak Glu current was potentiated by only about one third, but the steady-state current increased by two and a half times [Fig. 7(A)]. When CNQX was added to the cyclothiazide, the peak GIu current actually dropped below control value and steady-state current was 109 + 46% of control value (not significant) [Fig. 7(B)]. Since these experiments were -No

Glu

Con

GlU MK

CYCIO

GlU MK

Cycl0 CNOX

GIU

GIU

MK

MK

MK

CYCIO

Cycle

CYCIO

Glu

NIF NtF (3Ojdb41 wO&4t

NIM

Glu

Glu

cycle

CIII + cycto 100

80 Con ’ .g

GIU

GIU

Glu

60

5 3

40

Con MK

!

Glu 30#M MK

Glu 70@4 MK

Glu

lOO#M MK

Fig. 5. Cyclothiazide (Cycle) reduces the threshold for Glu neurotoxicity. After 2 min preincubation with 10 PM cyclothiazide, Glu with 1OpM MK-801 (MK) was toxic at concentrations as low as 70 PM (n = 6 for Con + Cycle and both 70 PM Glu groups; otherwise n = 12). “No significant difference from cultures lacking cyclothiazide; “P < 0.05 compared to cultures lacking cyclothiazide

Fig. 6. Other drug effects on Giu neurotoxicity. (A) A 2 min preincubation with the “inactive” thiazides, chlorothiazide (500 PM; Chloro) and benzthiazide (250 PM; Benz), had no significant effect on the toxicity of 500 PM Glu + 10 PM MK-801 (MK) (n = 6 for all groups in A-C). (B) A two minute preincubation with CNQX (50pM) was much more effective than either nifedepine (Nif) or nimodipine (30pM; Nim) in reducing the cyclothiazide (Cycle) augmentation of Glu + MK-80t toxicity. “P < 0.05 compared to all groups except control. C. Addition of CNQX (2OpM), but not 7-chlorokynurenate (30pM; 7CK) or aminophosphonovalerate (500 p M; APV), diminished the 18 h toxicity of cyclothiazide with Glu + MK-801. “P -=ZY 0.05 compared to all groups except control.

Pharmacology

i C 3

200

‘E

100

n m

of glutamate neurotoxicity

peak steady-state

13 8

0 GIU MK CyCfO

GIU MK CYCIO

CNQX

Fig. 7. Physiological interactions between Glu (500~~), cyclothiazide (10 PM), and CNQX (20 PM). A. Whoie cell patch recordings show the small increase in peak Glu-induced current produced by cyclothiazide (cyclo) and the enormous increase in steady-state current (compare Al and A2). CNQX blocks both peak and steady-state current enhancement by cyclothiazide (compare A2 and A3). (B) Data from nine cells demonstrating that cyclothiazide produces a much more marked enhancement of steady-state Glu current than peak current and that CNQX removes this augmentation.

done with CNQX at only 20 p M, we would expect even greater attenuation of inward steady-state currents with the 5OpM CNQX concentration used in most of the toxicity experiments. These observations are compatible with the hypothesis that CNQX, at the concentrations employed in the neurotoxicity paradigms, can significantly attenuate the enhancement of Glu currents, especially steady-state currents, produced by cyclothiazide. The most economical hypothesis to explain our observations about thiazides potentiating Glu neurotoxicity is that these compounds enhance calcium entry and consequently produce neurotoxic elevations of Caf + . In order to evaluate this hypothesis, we measured neuronal Caz+ using the fluorescent calcium indicator fura-2lAM (Grynkiewicz et al., 1985). We first examined whether the addition of cyclothiazide to cells already exposed to Glu + MK-801 would eievate Ca’$- . Cyclothiazide

959

caused a small increase above the steady value of CA:’ produced by prolonged exposure to Glu + MK-801 (Figs 8,9). Surprisingly, though, Ca” did not reach the peak level of Ca”+ produced immediately after the addition of Glu to cultures already containing MK-801 (Fig. 9). In addition, the peak values for Caf + in cultures preincubated with cyclothiazide + MK-801 and then exposed to Glu were actually lower than the peak Cat-@ levels stimulated by Glu in cultures incubated in MK801 alone (Moudy and Rothman, unpublished). Therefore, the increased neurotoxicity associated with Glu + cyclothiazide was not reflected in higher Caf +. This lack of direct correlation between Ca’+ and neuronal death has been observed before by us and others (Michael and Rothman, 1990; Dubinsky and Rothman, 1991; Tymianski et al., 1993). We were interested in dete~ining whether our observations on desensitization and Glu neurotoxicity had relevance to the pathophysiology of ischemic neuional damage. Cerebral ischemia is accompanied by a lactic acidosis (Nedergaard et al., 1991), which might limit any Glu neurotoxicity acting through the NMDA receptor (Tang et al., 1990). If, however, rapid Glu desensitization were removed, AMPA receptors might make a substantial contribution to neuronal injury and explain the protection produced in global ischemia by nonNMDA receptor antagonists (Sheardown et al., 1990; Buchan et al., 1991; Nellgard and Wieloch, 1992). We simulated this condition by dropping the extracellular pH in our medium to 6.4 by direct addition of HCI to HEPES-buffered medium. A 30 min exposure to this medium was not very toxic to neurons, even when cyclothiazide or Glu (100 PM) were in&vidualEy added (Fig. 10). However, the combination of both killed over 90% of the neurons. CNQX blocked this neuronal injury and returned viability back to control levels. Therefore, at pH values seen in some stroke models, Glu neurotoxicity is dominated by AMPA receptor activation, when desensitization is eliminated. DISCUSSION The expe~mental results described above provide striking documentation of Glu neurotoxicity mediated by AMPA receptors. They agree with earlier results

Fig. 8. Ratio imaging of Cat+ after cyclothiazide. (A) Gray scale control ratio image of Ca,2+ . (B) Ratio image in the presence of Glu (500 PM) and MK-801 (10 FM). (C) Addition of cyclothiazide (10 ,uM) slightly increases the Ca” * . Scale bar: 25 pm A-C.

A. M.

960

MOUDY

et 01.

provides a plausible explanation for the focus NMDA receptor’s involvement in cell death (Mayer et al., 1987). Excessive accumulation of intracellular calcium has been related to cellular injury in general, and neuronal damage specifically, for over a decade (Schanne et al., 1979; Simon et al., 1984a; Choi, 1987). The protective effects of NMDA antagonists in various in vitro and in z’ivo models of brain hypoxia/ischemia further established the pathophysiological importance of the NMDA class of Glu receptors (Simon et al., 1984b: Goldberg et al., 1987; Rothman et cd., 1987). The results here are in general agreement with past reports from our laboratory and others’. Even extended exposure (one day) to high concentrations of extracellular Glu was tolerated by hippocampal neurons as long as MK-801 was present to block NMDA channel opening. However, as soon as we removed Glu desensitization with either diazoxide or cyclothiazide, MK-801 no longer was very protective. Furthermore, the threshold concentration for inducing AMPA receptor-mediated Glu toxicity was lowered by cyclothiazide. The most likely mechanism responsible for Glu neurotoxicity with cyclothiazide or diazoxide is enhanced activation of the AMPA receptor/ionophore complex. The block of toxicity by CNQX but not additional NMDA antagonists makes any other proximate explanation untenable. The precise explanation for the injury is not so clear. Removal of desensitization should enhance calcium entry through voltage-gated channels. The failure of either nifedipine or nimodipine to protect from neurotoxicity argues against the involvement of L-type calcium channels in this process, although other types of calcium channels could still be playing a role. A more compelling argument against the toxicity of voltage-gated calcium entry is the observation that deionophore

on the

5

34 I--~/

A

10

20

30 40 50 Time Imin)

60

70

80

12091

I 1694

Fig. 9. Effect of cyclothiazide (cycle) on Caf + (A) Caf + in single neuron, initially incubated in MEM + 10 PM MK-801 (1) and followed sequentially. Addition of 500 PM Glu (2) produced a large elevation in Caf + beyond that observed in 10 PM MK-801 alone. Caf+ gradually diminished to a steadystate level (3), which rose slightly after addition of 10 PM

cyclothiazide (4). There was a subsequent reduction in Caf+ towards a steady-state (5). (B) Summary of Caf + measurements from 58 neurons treated identically to the cell in A. Cyclothiazide had only a small influence on Caf + Numbers on abscissa legend correspond to points marked in A, indicating drug additions. Groups (2) and (4) are peak values; the others are steady state. “Not significantly different from the groups 3 and 5.

showing that block of desensitization potentiates either quisqualate or AMPA neurotoxicity in vitro (Thio et al., 1992; May and Robinson, 1993). The present report provides more details of the pharmacology and pathophysiology of this phenomenon. We have documented that AMPA/KA antagonists attenuate this form of Glu toxicity and quantitated neuronal injury one day after the Glu exposure. This eliminates the possibility of our simply looking at exaggerated cell swelling. Much of the prior research on the pharmacology of Glu neurotoxicity had emphasized the dominant role of the NMDA receptor/ionophore in producing the neuronal damage associated with Glu (Rothman and Olney, 1987; Choi et al., 1988; Michaels and Rothman, 1990). The relatively high calcium permeability of the NMDA

pH6.4

CYCIO

GlU

GIU Cycle

GIU Cycle CNQX

Fig. 10. Interaction of cyclothiazide (cycle) and extracellular acidosis. In this set of experiments, neurons were exposed for 30 min to a pH of 6.4 in cyclothiazide (IO PM) and Glu (100pM) alone and then combined. The combination increased death to over 90%, which was prevented by CNQX (50 /*M) (n = 18 for all groups). “P < 0.05 compared to all other groups.

Pharmacology

of glutamate neurotoxicity

polarization with elevated extracellular potassium is not particularly damaging to cultured neurons (Michaels and Rothman, 1990; Tymianski et al., 1993). There are circumstances when the neurotoxicity of excitatory amino acids is attenuated by blockers of voltage-gated calcium channels (Weiss et al., 1990). However, this only occurs when the insult is a prolonged exposure to a very low concentration of amino acid. A probable explanation for the cyclothiazide enhancement of Glu toxicity is that the prolonged opening of the AMPA ionophorc allows increased entry of extracellular calcium. This would agree with the observation of Pruss and colleagues (1991) that KA stimulates cobalt entry, a marker of divalent cation permeability. Murphy and Miller (1989) have also described KA-induced Caf + rises not mediated by voltage-gated calcium channels. The most direct proof of this would be to remove calcium from the extracellular medium and observe attenuation of neurotoxicity. We have tried several variations of this experiment, with inconclusive results. Our neurons are damaged in the absence of extracellular it di~cult to perform the appropriate

calcium,

making

ion substitutions

(McCaslin and Smith, 1990). We expected that cyclothiazide would substantially elevate Ca’+ beyond values seen with Glu plus MK-801, but instead observed an insignificant change in neuronal Ca:+ when cyclothiazide was added to our cultures (Fig. 9). Despite the small increment in CaF+, neuronal loss dramatically increased. There are at least two possible explanations for this. First, Caf + poorly reflects ongoing calcium entry once levels are already increased above baseline. Second, Caf + correlates poorly with neuronal death in a variety of preparations. Results from our lab and others have recently stressed that site of calcium entry is relatively more important than absolute magnitude of Caf + in controlling cell viability (Michaels and Rothman, 1990; Dubinsky and Rothman, 199 1; Tymianski et al., 1993). We also recognize that our current imaging technology does not adequately resolve the exact location of the Cat + , its site of entry, or even its detailed time course. There is already compelling biochemical evidence that all Cat+ is not equivalent (Lerea et al., 1992)‘ We are still trying to establish the potential relevance of our experimental results to naturally occurring neurotoxicity. It is convenient to believe that one “teleological role” for Glu desensitization is to protect central neurons from overactivation and injury. The combination of desensitization and acidosis, a frequent concomitant of cerebra1 ischemia, may be first line, endogenous protective mechanisms against hypoxiciischemic neuronal damage. These “autoprotective devices” could fail after tens of seconds, eventually resulting in neuronal death. There is now precedent for Glu receptor alterations accompanying brain ischemia (Pellegrini-Giampietro et al., 1992). If these altered receptors desensitize slowly, neuronal death could increase dramatically and become much more dependent upon Glu acting at the KA

961

receptor-ionophore complex. Conversely, this scenario provides us with some therapeutic optimism. If Glu, acting at non-desensitized receptors, is responsible for a component of neuronal loss, we should be able to use appropriate antagonists to prevent cell death, Ackno~~ledgements.--We would like to thank Nancy Lancaster

for preparing the hipp~campal cuftures and Brenda McCall for typing the manuscript. This work was supported by NIH RQI 19988 (SMR) and K08 01443 (KAY).

REFERENCES Buchan A. M., Li H., Cho S., and Pulsi~lelli W. A. (1991) Blockade of the AMPA receptor prevents Ca, hippocampal injury foIlowing severe but transient forebrain ischemia in adult rats. Neurosci. Lett. 132: 255-258. Choi D. W. (1987) Ionic dependence of glutamate neurotoxicity in cortical cell culture. J. Neurosci. 7: 380-390. Choi D. W. (1990) Cerebral hypoxia: some new approaches and unanswered questions. J. Neurosci. 10: 2493-2501. Choi D. W. and Rothman S. M. (1990) The role of glutamate neurotoxicity in hypoxic-is~hemi~ neuronal death. A. Rev. Neurosci. 13: 171-182. Choi D. W., Koh J. and Peters S. (1988) Glutamate neurotoxicity in cortical cell culture: attenuation by NMDA antagonists. J. Neurosci. 8: 1855196. Collingridge G. L. and Lester R. A. S. (1989) Excitatory amino acid receptors in the vertebrate central nervous system. Pharmac. Rev. 40: 143-210.

Dubinsky J. M. and Rothman S. M. (1991) Intracellular calcium concentrations during “chemical hypoxia” and excitotoxic neuronal injury. f. Neurosci. 11: 2545-2.55 1. Goldberg M. P., Weiss J. H., Pham P. C. and Choi D. W. (1987) N-methyl-o-aspartate receptors mediate hypoxic neuronal injury in cortical culture. J. Phnrmac. Exp. Ther. 243: 784-91.

Grynkiewicz G., Poenie M. and Tsien R. Y. (1985) A new generation of Ca’-’ indicators with greatly improved fluorescence properties. J. Biol. Chem. 260: 3440-3450. Hahn J. S., Aizenman E. and Lipton S. (1988) Central mammalian neurons normally resistant to glutamate toxicity are made sensitive by elevated extracellular Ca2 * : toxicity is blocked by the N-methyl-D-aspartate antagonist MK-801. Proc. Natn. Acad. Sci. U.S.A. 85: 6556-6560.

Hamill 0. P., Marty A., Neher E., Sakmann B. and Sigworth F. J. (1981) Improved patch-clamp techniques for high resolution current recording from cells and cell-free membrane patches. P$uegers Arch. 391: 85-100. Huettner J. E. and Baughman R. W. (1986) Primary culture of identified neurons from the visual cortex of postnatal rats. J. Neurosci. 6: 3044-3060. K&kin N. I. and Krishtal 0. A. and Tsyndrenko A. Y. (1986) Excitatory amino acid receptors in hippocampal neurons: kainate fails to desensitize them Neurosci. L&t. 63: 225-230. Lerea L. S., Butler L. S. and McNamara J. 0. (1992) NMDA and non-NMDA receptor mediated increase of c-fos mRNA in dentate gyrus neurons involves calcium influx via different routes. J. Neurosci. 12: 2973-298 1. Lester R. A., Quarum M. L., Parker J. D., Weber E. and Jahr C. E. (1989) Interaction of 6-cyano-7-nitroquinoxa~ine-7,3-

962

A. M,

MOUDY

drone with the N-methyl-n-aspartate receptor-associated glycine site. Molec. Phurmac. 35: 565-570. May P. C. and Robison P. M. (1993) Cyclothiazide treatment unmasks AMPA excitotoxicity in rat primary hippocampal cultures. J. ~~~~uc~e~?z 60: 1171-I 174. Mayer M. L. and Vyklicky L. (1989) Concanavahn A selectivity reduces desensitization of mammalian neuronal quisqualate receptors. Proc. Natn. Acad. Sri. U.S.A. 86: 141 I-1415. Mayer M. L., MacDermott A. B., Westbrook G. L,, Smith S. J. and Barker J. L. (1987) Agonistand voltage-gated calcium entry in cultured mouse spinal cord neurons under voltage clamp measured using arsenazo III. J. Neurosci. 7: 3230-3244. McCaslin P. P. and Smith T. G. (1990) Low calcium-induced release of glutamate results in autotoxicity of cerebellar granule cells. Brain Res. 513: 280-285. Michaels R. L. and Rothman S. M. (1990) Glutamate neurotoxicity in vitro: antagonist pharmacology and intracellular calcium concentrations. J. Neurosci. 10: 283-292. Moudy A. M., Yamada K. A. and Rothman S. M. (1992) Desensitization changes the pharmacology of glutamate neurotoxicity. Sot. Neurosci. Abstr. 18: 757. Murphy S. N. and Miller R. J. (1989) Regulation of Cat + influx into striatal neurons by kainic acid. J. Pharm. Expt. Left. 249: 1844193. Nedergaard M., Kraig R. P., Tanabe J. and Pulsinelli W. A. (1991) Dynamics of interstitial and intracellular pH in evolving brain infarct. Am. J. Physiol. 260: PR581p588. Nellgard B. and Wieloch T. (1992) Postischemic blockade by AMPA but not NMDA receptors mitigates damage in rat brain following transient severe cerebral &hernia. J. Cereb. Blood Flow Metab. 12: 2-l 1. Olney J. W. (1983) Excitotoxins: an overview. In: Excitotoxins (Fuxe K., Roberts P. and Schwartz R., Eds), pp. 82-96. MacMillan, London. Pellegrini-Giampietro D. E., Zukin R. S., Bennett M. V., Cho S. and Pulsinelli W. A. (1992) Switch in glutamate receptor submit gene expression in Car subfield of hippocampus following global ischemia in rats. Proc. Nutn. Acad. Sci. U.S.A. 89: 10499910503. Pruss R. M., Akeson R. L., Racke M. M. and Wiburn J. L. (1991) Agonist-activated cobalt uptake identifies divalent cation-permeabIe kainate receptors on neurons and gliai cells. ~eurun 7: 5099518. Rothman S. M. and Olney J. W. (1987) Excitotoxicity and the NMDA receptor. h?nds Neurosci. 10: 299-302. Rothman S. M., Thurston J. H., Hauhart R. E., Clark G. D. and Soloman J. S. (1987) Ketamine protects hippocampal neurons from anoxia in vitro. Neuroscience 21: 673-678. Schanne F. A., Kane A., Young E. and Farber J. (1979)

rt al.

Calcium dependence of toxic cell death: a final common pathway. Science 206: 699-700. Sheardown M. J., Nielson E. O., Hansen A. J., Jacobsen P. and Honore T. (1990) 2,3-Dihydroxy-4-nitro-7-sulfamoyl-~nzo(F)quinoxaline: a neuro protectant for cerebral ischemia. Science 247: 57 l-574. Simon R. P., Griffiths T., Evans M. C., Swan J. H. and Meldrum B. S. (1984a) Calcium overload in selectivity vulnerable neurons of the hippocampus during and after ischemia: an electron microscopy study in the rdt. J. Cereb. Blood F~OVVMetab. 4: 350-361. Simon R. P., Swan J. H., Griffiths T. and Meldrum B. S. (1984b) Blockade of n-methyl-D-aspartate receptors may protect against ischemic damage in the brain. Science 226: 850-852. Sugiyama H., Ito I. and Watanabe M. (1989) Glutamate receptor subtypes may be classified into two major categories: a study on xenopus oocytes injected with rat brain mRNA. Neuron 3: 1299132. Tang C.-M., Dichter M. and Morad M. (1990) Modulation of the iv-methyl-D-aspartate channel by extracellular H + hoc. Xatrt. Acad. Sci. U.S.A. 87: 6445-6449. Tang C.-M., Shi Q.-Y., Katchman A. and Lynch G. (1991) Modulation of the time course of fast EPSC’s and glutamate channel kinetics by ani~dcetam Science 254: 288290. Thio L. L., Clifford D. B. and Zorumski C. F. (1992) Blockade of ionotropic quisqualate receptor desensitization by wheat germ agglutin~n in cultured postnatal rat hippocampal neurons J. ~e~ro~~~sioi. 68: 1917-I 929. Trussell L. 0. and Fischbach G. D. (1989) Glutamate receptor desensitization and its role in synaptic transmission. Neuron 3: 209-218. Tymianski M., Charlton M. P., Carlen P. C. and Tator C. H. (1993) Source specificity of early calcium neurotoxicity in cultured embryonic spinal neurons. J. Neurosci. 13: 2085-2104. Weiss J. H., Hartley D. M., Koh J. and Choi D. W. (1990) The calcium channel blocker nifedipine attenuates slow excitatory amino acid neurotoxicity. Science 247: 1474-1477. Yamada K. A. and Rothman S. M. (1992) Diazoxide blocks glutamate desensitization and prolongs excitatory postsynaptic currents in rat hippocampal neurons. J. Physiol. 458: 4099423. Yamada K. A., Dubinsky J. M. and Rothman S. M. (1989) Quantitative physiological characterization of a quinoxalinedione non-NMDA receptor antagonist. J. Neurosci. 9: 3230-3236. Yamada K. and Tang C. M. (1993) Benzothiazides inhibit rapid glutamate receptor desensitization and enhance glutamate synaptic currents. J. Neurosci. 13: 3904-3915. Zorumski C. F., Thio L. L., Clark G. D. and Clifford D. B. (1990) Blockade of desensitization augments quisqualate excitotoxicity in hippocampal neurons. Neuron 5: 61-66.