Cytosolic phospholipase A2 alpha mediateselectrophysiologic responses of hippocampalpyramidal neurons to neurotoxic NMDA treatmentYing Shen*†, Koji Kishimoto‡, David J. Linden*, and Adam Sapirstein‡§

*Department of Neuroscience, ‡Department of Anesthesiology and Critical Care Medicine, Johns Hopkins University School of Medicine,Baltimore, MD 21287; and †Department of Neurobiology, Neuroscience Institute, Zhejiang University School of Medicine,Hangzhou, Zhejiang 310058, China

Edited by Richard L. Huganir, Johns Hopkins University School of Medicine, Baltimore, MD, and approved February 7, 2007 (received for reviewJune 29, 2006)

The arachidonic acid-generating enzyme cytosolic phospholipaseA2 alpha (cPLA2�) has been implicated in the progression ofexcitotoxic neuronal injury. However, the mechanisms of cPLA2�

toxicity have yet to be determined. Here, we used a model systemexposing mouse hippocampal slices to NMDA as an excitotoxicinjury, in combination with simultaneous patch-clamp recordingand confocal Ca2� imaging of CA1 pyramidal neurons. NMDAtreatment caused significantly greater injury in wild-type (WT)than in cPLA2� null CA1 neurons. Bath application of NMDA evokeda slow inward current in voltage-clamped neurons (composed ofboth NMDA receptor-mediated and other conductances) that wassmaller in cPLA2� null than in WT slices. This was not due todown-regulation of NMDA receptor function because NMDA re-ceptor-mediated currents were equivalent in each genotype fol-lowing brief photolysis of caged glutamate. Current-clamp record-ings were made during and following NMDA exposure by elicitinga single action potential with a brief current injection. After NMDAexposure, WT CA1 neurons developed a spike-evoked plateaupotential and an increased spike-evoked dendritic Ca2� transient.These effects were absent in CA1 neurons from cPLA2� null miceand WT neurons treated with a cPLA2� inhibitor. The Ca-sensitiveK-channel toxins, apamin and paxilline, caused spike broadeningand Ca2� enhancement in WT and cPLA2� null slices. NMDAapplication in WT and arachidonate applied to cPLA2� null cellsoccluded the effects of apamin/paxilline. These results indicatethat cPLA2� activity is required for development of aberrantelectrophysiologic events triggered by NMDA receptor activation,in part through attenuation of K-channel function.

action potential � arachidonic acid � calcium � excitotoxicity � inhibition

The phospholipase A2 (PLA2) enzymes catalyze ester hydro-lysis of fatty acids from the second position of membrane

glycerophospholipids. In mammalian cells, because this sn-2position is highly enriched with arachidonic acid (AA), thePLA2s are considered the major regulated source of cellular AA.Once liberated by PLA2s, AA can be metabolized by a numberof enzymes to create the eicosanoids. These products and AAhave important functions in regulating cellular homeostasis andinflammation and have been implicated in a number of neuro-logic injuries (1, 2).

Of the PLA2s expressed in the brain, the enzyme cytosolicphospholipase A2 alpha (cPLA2�) has a unique set of biochemicalproperties that implicate it in AA signaling in response to neuronalactivity. In response to micromolar increases in intracellular Ca2�,cPLA2� translocates from the cytosol to cellular membranes, whereits phospholipid substrate resides (3). In addition, the enzymaticactivity of cPLA2� depends upon phosphorylation by p38 MAPK(4) and ERK-2 (5). cPLA2� has a marked preference for phospho-lipid substrates, with AA in the second position (6). Macrophagesand mast cells derived from cPLA2�-deficient mice (cPLA2��/�)

are unable to express prostaglandins and leukotrienes in responseto inflammatory stimuli (7, 8).

Several lines of investigation implicate cPLA2� in the poten-tiation of neurotoxicity. Transient global cerebral ischemia wasshown to persistently activate cPLA2� (9), and cPLA2��/� micewere significantly protected from transient focal cerebral isch-emia (1). A broader role for cPLA2� in neurologic injury wasdemonstrated in cPLA2��/� mice that suffered significantly lessdestruction of dopaminergic neurons following exposure to thetoxin 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP) inan experimental model of Parkinsonism (10). In rat organotypichippocampal cultures, the cPLA2� inhibitor arachidonyltrif lu-oromethyl ketone (AACOCF3) protected pyramidal neurons ofthe CA1 region from oxygen and glucose deprivation (11). Inaddition, mice treated with AACOCF3 suffered less neuronaldeath and had dramatically reduced induction of key cytokinesand chemokines in autoimmune encephalomyelitis (2).

Following cerebral ischemia and reperfusion, increased glu-tamate release activates NMDA receptors to increase neuronalCa2� f lux, resulting in excitotoxicity. Application of exogenousAA potentiates neuronal NMDA receptor-mediated currents(12) and NMDA-evoked Ca2� transients (13). The roles ofPLA2s in this process have not been clearly defined. We hypoth-esized that NMDA-triggered cPLA2� activation up-regulatesboth NMDA receptor-associated cation influx and its delayedelectrophysiologic sequelae. To test this hypothesis, we mea-sured both the neuronal toxicity and the Ca2� and electrophysi-ologic responses to NMDA in hippocampal neurons.

ResultsThe toxicity of NMDA on hippocampal CA1 neurons wasdetermined in acute hippocampal slices derived from littermatewild-type (WT) and cPLA2��/� mice. Acute hippocampal sliceswere recovered for 1 h in artificial cerebral spinal f luid (aCSF)2. Following this recovery period, they were then switched toMg2�-free aCSF2 supplemented with the GABAA receptorantagonist GABAzine (5 �M) for 10 min, exposed to 10 �MNMDA or vehicle for 6 min, and finally replaced in aCSF for anadditional 20 min. Fluorescent Nissl and propidium iodide (PI)stains were used as markers of neuronal integrity and irreversible

Author contributions: Y.S., K.K., D.J.L., and A.S. designed research; Y.S. and K.K. performedresearch; Y.S., K.K., and D.J.L. analyzed data; and Y.S., D.J.L., and A.S. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Abbreviations: AACOCF3, arachidonyltrifluoromethyl ketone; cPLA2�, cytosolic phospho-lipase A2 alpha; PI, propidium iodide; PLA2, phospholipase A2.

§To whom correspondence should be addressed at: Johns Hopkins School of Medicine, 600North Wolfe Street, Halsted 842B, Baltimore, MD 21287. E-mail: .adam.sapirstein@jhmi.edu.

This article contains supporting information online at www.pnas.org/cgi/content/full/0605427104/DC1.

© 2007 by The National Academy of Sciences of the USA

6078–6083 � PNAS � April 3, 2007 � vol. 104 � no. 14 www.pnas.org�cgi�doi�10.1073�pnas.0605427104

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

20, 2

020

injury. In WT hippocampi, a small amount of PI uptake wasobserved in untreated pyramidal neurons, presumably reflectingdamage from slice preparation. This uptake was decreased incPLA2��/� neurons. However, NMDA treatment caused sig-nificant injury, demonstrated by an �20-fold increase in theneurotoxic index [the ratio of PI to Nissl staining in comparablesections; see supporting information (SI) Fig. 6]. In contrast,injury to cPLA2��/� hippocampal neurons following NMDAtreatment was 23-fold less than that in the WT (P � 0.01).NMDA did increase neuronal injury in cPLA2��/� slices 7.5-fold, as compared with untreated cPLA2��/�, indicating thatdeletion of cPLA2� does not completely eliminate NMDAtoxicity. Pretreatment of WT hippocampi with the cPLA2�inhibitor AACOCF3 (10 �M) before NMDA exposure alsoresulted in a significant reduction in the ratio of PI to Nisslstaining fluorescence intensity (P � 0.05). These results indicatethat cPLA2� enzymatic activity significantly amplifies the neu-rotoxicity of NMDA in acute hippocampal slices.

Voltage-clamp recordings were made from Cs�-loaded CA1pyramidal cells in hippocampal slices prepared from 3- to 5-week-old littermate WT and cPLA2��/� mice. A command potential of�70 mV was applied to the soma during a period of baselinerecording in the presence of a GABAA receptor antagonist. Thisbaseline recording was characterized by a small holding current(WT, �2.4 � 4.2 pA, n � 7; cPLA2��/�, �5.1 � 4.0 pA, n � 7),which was occasionally punctuated by a spontaneous excitatorypostsynaptic current. At t � 10 min, the perfusion was switched toaCSF supplemented with 10 �M NMDA and continued until t �16 min, when washout commenced (Fig. 1). NMDA evoked aninward current that reached a peak value of 152.3 � 24.8 pA in WTmice, but only 64.9 � 11.2 pA in cPLA2��/� mice (P � 0.01 byunpaired t test). Although the kinetics of the NMDA-evokedcurrent varied, the current generally had a slow onset (reflecting thespeed of bath solution exchange) and an inflection in the onsetperiod. The current peaked before washout commenced in bothgenotypes. The total charge transfer from NMDA application wasalso significantly greater in pyramidal cells of WT mice (30 � 7.8nC), as compared with cPLA2��/� mice (13 � 2.8 nC, P � 0.05).However, no significant differences were observed in the recoveryphase of the NMDA response, as indexed by the 50% decay time.Transmembrane current returned to baseline values in both groupsby t � 25 min (WT, �0.7 � 9.3 pA; cPLA2��/�, –2.9 � 3.5 pA).These results indicate that the total and peak cation flux into CA1pyramidal cells evoked by neurotoxic NMDA stimulation is signif-icantly attenuated in cPLA2��/� mice.

Does the altered electrophysiologic response of cPLA2��/�

mice to NMDA stem from the loss of cPLA2� enzymatic activityor from compensation for chronic cPLA2� loss? To address thisquestion, we repeated these experiments in WT slices, but added10 �M AACOCF3 to the bath at least 30 min before the

beginning of patch-clamp recording. Essentially, AACOCF3pretreatment completely replicated the effects of cPLA2� dele-tion. The peak current (86.7 � 9.7 pA, P � 0.05) and chargetransfer (16.9 � 1.5 nC, P � 0.05) evoked by bath NMDAapplication were significantly attenuated, compared with un-treated WT neurons. These results indicate that the slow inwardcurrent evoked by bath application of NMDA is attenuatedunder conditions of either chronic or acute cPLA2� deficiency.

Does cPLA2� inhibition directly attenuate NMDA receptorfunction? We addressed this question in two ways. First, cPLA2�deletion may down-regulate expression of NMDA receptors.Therefore, we performed Western analysis of the NR1 subunitof the NMDA receptor in posterior hippocampi protein fromWT and cPLA2��/� mice and found that the relative amounts ofNR1 protein were similar in both (�/�, 0.85 � 0.55, n � 3; �/�,1.0 � 0.5, n � 4; arbitrary units normalized to actin; P � NS).Second, to determine the functional status of NMDA receptors,we loaded slices with caged glutamate, 100 �M 4-methoxy-7-nitroindolinyl (MNI)-glutamate, and performed local photolysiswith 10-msec-long flashes of UV light. This tests the responsesof all NMDA receptors and not just those activated by synap-tically released glutamate. To restrict activation to NMDAreceptors on the recorded cell, slices were bathed in tetrodotoxin(to suppress Na� spiking), together with antagonists of AMPA,kainate, GABAA, and mGluR1/5 receptors. Test f lashes deliv-ered at 60- or 120-sec intervals evoked large, reasonably stableresponses that were completely blocked by the NMDA receptorantagonist D-AP5 (Fig. 2). In between cell population compar-isons, f lash-evoked NMDA currents in cPLA2��/� and WT cellswere similar as shown by the peak current amplitude (1083.5 �42.5 and 1114.2 � 45.1 pA, respectively; n � 32 and 29,respectively), charge transfer (206.6 � 18.2 and 242.6 � 19.4 pC,respectively), or kinetics. Following baseline recording, WT cellswere exposed to 10 �M AACOCF3 for �30 min (to mimic thepretreatment regimen used in Fig. 1), and flash-evoked NMDAcurrents were recorded and compared with baseline values (SIFig. 7). AACOCF3 treatment produced a small attenuation ofpeak NMDA current (77.4 � 1.4% of baseline, P � 0.05,compared with baseline values, n � 5), but also a small slowingof the decay phase. Thus, no significant change was observed inNMDA receptor-mediated charge transfer (82.2 � 5.9% ofbaseline, P � 0.05). Peak NMDA current following photolysiswas not attenuated in cPLA2��/�. Thus, the small effect ofAACOCF3 is unlikely to underlie the significant attenuation ofslow currents evoked by NMDA bath application in cPLA2��/�-and AACOCF3-treated WT neurons (Fig. 1).

Voltage-clamp recordings allow resolution of NMDA-evokedcurrents, but they do not illustrate the behavior of neurons undermore natural conditions. To this end, we performed current-clamp recordings by using a K�-based internal saline. CA1

Fig. 1. The slow inward current evoked by neurotoxic bath application of NMDA is attenuated under conditions of cPLA2� inhibition. (A) Cells were bathedin a GABAA receptor antagonist (GABAzine, 5 �M) and held at �70 mV. At t � 10 min, NMDA (10 �M) was applied to the bath for a duration of 6 min (as indicatedby the horizontal bar), after which recording continued until t � 60 min. Representative, unaveraged traces from single cells in slices are shown. The rapid inwardcurrents, which are superimposed on the slow responses, are spontaneous excitatory postsynaptic currents. (B) Population measures for the experiment shownin A. Bars indicate the means � SEM. cPLA2��/�, n � 7; WT, n � 7; WT, AACOCF3, n � 6. *, P � 0.05; **, P � 0.01 for comparisons to the �/� condition.

Shen et al. PNAS � April 3, 2007 � vol. 104 � no. 14 � 6079

NEU

ROSC

IEN

CE

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

20, 2

020

neurons rarely exhibited spontaneous action potentials in thisconfiguration. A stimulating pulse, delivered every 2 min, con-sisted of a constant 50-msec-long injection of depolarizingcurrent (range 0.3–0.6 nA), with the amplitude adjusted at theoutset to reliably evoke a single spike. This was preceded by asmall hyperpolarizing current injection to monitor Rinput (Fig.3A). The responses to this pulse could be stably recorded for �36min (Fig. 3 B and C) in WT neurons. When a separate group ofthese neurons was challenged with a 6-min-long bath applicationof NMDA, a small, transient depolarization was produced, butno significant alterations were observed in the evoked spikeduring NMDA application. However, following a delay of �15min, the properties of the evoked spike began to change: thefalling phase slowed, and a plateau potential developed at��20–0 mV for 4–30 msec, after which a falling phase returnedthe membrane voltage to the resting potential. The size of thisplateau potential, as indexed by measures of spike width (at 60%height) and the time-integrated depolarization, continued toincrease for the duration of the recording. The development ofthe plateau potential was not associated with significant changesin the resting membrane potential, Rinput, or the peak amplitudeof the spike. In cPLA2��/� neurons, exposure to NMDA pro-duced a smaller transient depolarization (Fig. 3C), consistentwith the reduced NMDA-evoked inward current seen in voltage-clamped cells (Fig. 1). Most importantly, cPLA2��/� neuronsnever developed spike-evoked plateau potentials (spike 60%width: WT, 13.4 � 2.8 msec at t � 36 min, n � 10; cPLA2��/�,2.8 � 0.3 msec, n � 9; P � 0.002).

Simultaneous confocal Ca2� imaging was performed on theCA1 pyramidal neurons that were challenged with NMDA. Aregion of interest was defined in the proximal dendrite (�25 �mfrom the soma). Following a 20-min waiting period to allow dyediffusion in the dendrite, both resting and spike-evoked dendriticCa2� were stably recorded in WT neurons over the 36-minrecording period (Fig. 4). NMDA treatment produced a smallincrease in the spike-evoked Ca2� transient that was stable for�5 min. Following NMDA washout, the spike-evoked Ca2�

transient increased further, reaching a level of 240 � 54% of

baseline integrated Ca2� at t � 36 min, the last time pointrecorded. Interestingly, application of NMDA did not producea measurable change in resting Ca2� concentration, likely re-f lecting that small changes in Ca2� are not easily detected byFluo-4 (Kd for Ca2� of �500 nM) in neurons. When thesemeasurements were repeated by using cPLA2��/� neurons,spike-evoked Ca2� slightly increased during NMDA application.Following washout, the spike-evoked Ca2� transients returned topre-NMDA levels and did not change during the recording(cPLA2��/�, 127 � 10% of baseline integrated Ca2� at t � 36min; P � 0.05). Similar changes were seen when the peak of theCa2� transient was measured (WT, 201 � 21% of baseline;cPLA2��/�, 97 � 4% of baseline at t � 36 min; P � 0.0005). Thedelayed spike-evoked plateau potentials and the associated

Fig. 2. Rapid NMDA currents evoked by glutamate photolysis are normal inCA1 pyramidal cells derived from cPLA2��/� mice. Hippocampal slices werebathed in Mg2�-free aCSF supplemented with MNI-glutamate and antago-nists of GABAA, AMPA, kainate, and mGluR1/5 receptors. In these conditions,10-msec-long flashes of UV light produced glutamate photolysis, whichevoked an NMDA receptor-mediated inward current. The current tracesshown are the averages of five consecutive responses. The lower trace illus-trates the duration of the UV flash. These responses were completely blockedby the NMDA receptor antagonist D-AP5. Bar graphs show population com-parisons. No significant differences were observed between �/� and �/�groups (P � 0.10 for all). n � 29 and 32 cells, respectively.

Fig. 3. Development of a spike-evoked plateau potential triggered byNMDA treatment is blocked in CA1 pyramidal cells from cPLA2��/� mice. (A) Asample trace illustrates the stimulation and recording protocol. Current-clamprecordings were combined with dendritic Ca2� imaging; the latter data areshown in Fig. 4. Initially, a brief injection of hyperpolarizing current wasdelivered to allow for the measurement of Rinput; 500 msec later, a 50-msec-long injection of depolarizing current was given. This current was calibratedat the outset of the experiment to reliably evoke a single spike and was heldconstant thereafter. Current-clamp recording continued until t � 0.8 sec tocapture the evoked spike, and image acquisition continued until t � 1.2 sec tocapture the spike-evoked Ca2� transient. (B) Representative single traces ofevoked spikes taken at the points indicated on the time-course graph in C.Time a (dark blue line) is at the start of the recording period. Time b (greenline) occurs 12 min from the start of the experiment, which is after 4 min ofNMDA exposure. Time c (light blue line) is 36 min after the start of recordingand is the last time point in the experiment. The �/� group was a control, inwhich cells from WT mice were recorded without NMDA challenge. (Scalebars: 10 msec.) (C) Time-course graphs showing population measures ofelectrophysiologic parameters. �/� NMDA, n � 10; �/�, n � 5; �/� NMDA,n � 9. AP, action potential.

6080 � www.pnas.org�cgi�doi�10.1073�pnas.0605427104 Shen et al.

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

20, 2

020

increases in spike-evoked Ca2� transients are likely to be earlymanifestations of NMDA-induced excitotoxic injury that occurin WT, but not in cPLA2��/� mice.

Simultaneous current-clamp and Ca2� imaging experimentswere repeated by using WT mice pretreated with 10 �MAACOCF3 or its inactive analog, AACOCH3 (Fig. 5 and SI Fig.8). Remarkably, AACOCF3 pretreatment completely preventedthe development of the spike-evoked plateau potential followingNMDA challenge (60% spike width, 3.0 � 0.6 msec; n � 10, t �36 min; Fig. 5), thereby replicating the effect of cPLA2� geneticdeletion. When WT neurons were pretreated with the controlcompound AACOCH3, they developed spike-evoked plateaupotentials following NMDA challenge in a manner similar tountreated WT neurons (25.9 � 14.0 msec; n � 5). AACOCF3,but not AACOCH3, prevented the delayed increase in spike-evoked Ca2� transients following NMDA challenge, whichwas reflected in measures of both the Ca2� transient peak(AACOCH3, 176 � 30% of baseline at t � 36 min; AACOCF3,

98 � 5% of baseline; P � 0.01) and the Ca2� transient integral(AACOCH3, 175 � 21% of baseline; AACOCF3, 115 � 5% ofbaseline; P � 0.01). These results suggest that either chroniccPLA2� gene deletion or acute inhibition protects CA1 pyrami-dal neurons from the electrophysiologic sequelae of neurotoxicNMDA challenge.

Is the increase in the spike-evoked Ca2� transient followingNMDA treatment in WT neurons solely a consequence of theparallel development of spike-evoked plateau potentials? Thetime courses of the spike and Ca2� responses suggest that this isnot the case. For example, in WT neurons treated with NMDA,at t � 24 min, spike-evoked Ca2� responses are augmented (Fig.4), but the plateau potential has yet to develop (Fig. 3). Toanalyze this further, we created a scatter plot to compare thechange in the Ca2� integral to the change in the spike integralat t � 24 min (SI Fig. 9A). This revealed no significant corre-lation in either the cPLA2�-intact (�/� and AACOCH3) orcPLA2�-deficient (�/� and AACOCF3) groups (for �/� andAACOCH3 groups, Pearson r correlation value � 0.02; P � 0.96;for �/� and AACOCF3 groups, Pearson r correlation value �0.08; P � 0.73). However, in a scatter plot constructed for a latertime point, 36 min (SI Fig. 9B), a highly positive correlationemerged in the cPLA2�-intact groups (Pearson r correlationvalue � 0.95; P � 0.0001), but not in the cPLA2�-deficient

Fig. 4. Development of enhanced spike-evoked Ca2� transients triggered byNMDA treatment is blocked in CA1 pyramidal cells from cPLA2��/� mice. (A)(Left) Typical dye-filled CA1 pyramidal cell with an attached patch pipetteprotruding to the right of the region of interest (ROI) in the proximal apicaldendrite. This is a projected z-stack of confocal images. The excited fluoro-phore is Alexa-Fluor 594 hydrazide. (Scale bar: 20 �m.) (Right) Raw Fluo-4fluorescence images from the ROI. The letters correspond to times in thecourse of the experiment, as shown in B; and the numbers correspond to timepoints in individual spike-evoked Ca2� transients, as shown in C. (B) Time-course graphs showing normalized population measures of spike-evoked Ca2�

transient parameters. �/� NMDA, n � 10; �/�, n � 5; �/� NMDA, n � 9. (C)Representative single traces of spike-evoked Ca2� transients taken at thepoints indicated on the time-course graph in B. No filtering or averaging wasapplied to these traces. The frame acquisition speed was 50 Hz. (Scale bars: 250msec, 25% �F/F.)

Fig. 5. Development of a spike-evoked plateau potential triggered byNMDA treatment is blocked in CA1 pyramidal cells treated with the cPLA2�

inhibitor AACOCF3. (A) Representative single traces of evoked spikes taken atthe points indicated on the time-course graph in B. (Scale bars: 10 msec.) (B)Time-course graphs showing population measures of electrophysiologic pa-rameters. �/� NMDA AACOCF3, n � 10; �/� NMDA AACOCH3 (a controlcompound that does not inhibit cPLA2�), n � 5.

Shen et al. PNAS � April 3, 2007 � vol. 104 � no. 14 � 6081

NEU

ROSC

IEN

CE

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

20, 2

020

groups (P � 0.17). This suggests that, following NMDA chal-lenge, multiple factors contribute to the increase in spike-evokedCa2� transients. In an early phase, these increases appear to beindependent of spike broadening. However, somewhat later, att � 36 min, spike broadening is a strong predictor of the increasein the spike-evoked Ca2� transient. It is likely that NMDA treat-ment triggers at least two different mechanisms that impact spike-evoked Ca2� transients, both of which are cPLA2�-dependent.

If consequences of NMDA treatment include a cPLA2�-dependent defect in spike repolarization and a related increasein the spike-evoked Ca2� transient, then perhaps pharmacologicdisruption of spike repolarization could produce a similar in-crease in spike-evoked Ca2� transients. A number of conduc-tances contribute to spike repolarization in CA1 pyramidalneurons, including Ca2�-sensitive K channels of the BK and SKtypes. We simultaneously blocked BK and SK channels with 100nM apamin and 1 �M paxilline, respectively, and found that thissignificantly increased both the spike integral and the associatedCa2� integral in WT cells (spike integral, 177 � 9% of baseline;Ca2� integral, 202 � 7% of baseline, n � 6). Furthermore,NMDA treatment occluded the effects of subsequent apamin/paxilline application on both measures (NMDA spike integral,239 � 53% of baseline; Ca2� integral, 241 � 54% of baseline, n �10; NMDA plus apamin/paxilline spike integral, 278 � 45% ofbaseline; Ca2� integral, 252 � 28% of baseline, n � 7), suggest-ing that NMDA treatment may act in part by attenuating SKand/or BK channels (SI Fig. 10).

Can the product of cPLA2� activity, AA, produce spikebroadening and related increases in spike-evoked Ca2� tran-sients? We bath-applied 10 �M AA by using the same timingprotocol as that used for NMDA (a 6-min-long exposure,followed by washout) in cPLA2��/� neurons to determinewhether AA mediated effects downstream from cPLA2� (SI Fig.11). AA treatment produced a delayed spike-broadening andCa2�-transient increase (spike integral, 124 � 4% of baseline;Ca2� integral, 175 � 31% of baseline, n � 12), although this wasa somewhat smaller effect than that evoked by NMDA in WTneurons. When the apamin/paxilline mixture was applied toneurons that had previously been treated with AA, there was asmall but significant additional increase in the spike integral(152 � 8% of baseline, n � 12; P � 0.05) and a small butinsignificant increase in the spike-evoked Ca2� transient (210 �32% of baseline; P � 0.05). When apamin/paxilline treatmentwas applied to a separate group of cPLA2��/� cells, the increasesin spike and Ca2� integrals were not significantly different fromthose produced by AA plus apamin/paxilline (spike integral,148 � 13% of baseline; Ca2� integral, 185 � 13% of baseline, n �6; P � 0.10 for both measures). This suggests that at least aportion of the action of AA on spike broadening and the increasein the spike-evoked Ca2� transient is mediated by inhibition ofSK and/or BK channels.

DiscussionThe main finding of these studies is that cPLA2� activity isrequired for the immediate electrophysiologic events that lead toneurotoxicity in hippocampal CA1 pyramidal cells. We foundthat expression of cPLA2� was highly correlated to the onset ofirreversible CA1 pyramidal neuronal injury, as demonstrated byPI uptake (SI Fig. 6). Deletion of the cPLA2� gene or inhibitionwith AACOCF3 resulted in a significant reduction in the slow,inward current triggered by NMDA treatment (Fig. 1). Thisalteration in NMDA-triggered current was not due to changes inthe conductance of NMDA receptors in the cPLA2��/� micebecause the currents produced by glutamate photolysis in thepresence of inhibitors of other glutamate receptors were equiv-alent in the WT and cPLA2��/� neurons (Fig. 2). Likewise,expression of the NR1 subunit of the NMDA receptor wasequivalent in both genotypes. Coincident measurement of spike-

evoked Ca2� transients and action potentials demonstrated thatcPLA2� expression is essential for the appearance of spike-evoked plateau potentials and increases in spike-evoked Ca2�

transients that follow NMDA exposure (Figs. 3 and 4). Takentogether with the toxicity findings, these events are likely toconstitute an early portion of the excitotoxic response to NMDA.Importantly, the cPLA2� inhibitor AACOCF3, but not theinactive compound AACOCH3, replicated the effect of genedeletion (Fig. 5 and SI Fig. 8). The concordance of the effect ofthe inhibitor with that of gene deletion indicates that the findingsin cPLA2��/� mice were not due to genetic compensation orchronic loss of cPLA2� activity. These experiments are the firstto reveal electrophysiologic steps linking cPLA2� to excitotox-icity. Understanding the mechanism(s) by which cPLA2� medi-ates neurotoxicity is important for developing therapeutic strat-egies for stroke and other diseases.

We suggest that bath NMDA application evokes a Ca2� f luxin both the recorded neuron and its neighbors, and that theseCa2� transients activate cPLA2� and initiate cPLA2�-dependentintracellular events that ultimately augment the peak and totalNMDA-evoked current and lead to cPLA2� enhancement ofearly neuronal injury (14). AA has been shown by others topotentiate NMDA receptor-mediated currents (12) and NMDA-evoked Ca2� transients (13) in cerebellar granule cells andhippocampal pyramidal cells, respectively.

The effect of NMDA on membrane current appears to berelatively brief: NMDA-evoked membrane current peaks duringNMDA application and returns to baseline within a few minutesof washout. We believe that, during this time, NMDA treatmentinduces Ca2� transients and cPLA2�-dependent AA productionthat set in motion at least two events that ultimately impact Ca2�

signaling. The first involves increases in spike-evoked Ca2�

signals that are independent of changes in spike shape, as seensoon after NMDA washout (t � 24 min; Figs. 3 and 4 and SI Fig.9A). The second is a cascade that will ultimately result in thedelayed appearance of spike-evoked plateau potentials (Fig. 3)and correlated increases in spike-evoked Ca2� transients (Fig. 4and SI Fig. 9B). The increases in spike-evoked Ca2� transientsare likely an early trigger of excitotoxic processes (15, 16). Thesesequelae of NMDA treatment are dependent upon cPLA2�activity and are therefore abolished in CA1 neurons of bothcPLA2��/�- and AACOCF3-treated slices.

Both NMDA treatment in WT neurons and AA treatment incPLA2��/� neurons produced spike broadening and increases inspike-evoked Ca2� transients that occluded the effects ofapamin/paxilline. This suggests that cPLA2� activity triggered byNMDA treatment produces AA that attenuates BK and/or SKchannels that underlie spike repolarization. cPLA2� might alsoinhibit the voltage-gated transient K� current (IA) mediated byKv channels because AA suppresses this current when applied byintracellular pipette to CA1 neurons (17) and has been shown toattenuate the Kv1.4 channel (18).

Some caveats should be considered in the present studies. First,AACOCF3 is not the most potent or selective inhibitor of cPLA2�(19), but it has been used successfully in CNS models of disease (2),and it offers the availability of the negative control compoundAACOCH3. At the 10-�M dose, its inhibition of cyclooxygenase isminor. Metabolism did not limit its efficacy because we continuallysuperfused the drug during the experiment.

Second, we chose a low dose of NMDA (10 �M, 6 min) thatis consistent with delayed excitotoxic injury in the hippocampus(20) and allowed reliable patch-clamp recording throughout thecourse of the experiment. Larger doses of NMDA or longerexposure times could potentially show a different cPLA2�dependence, but resulted in less stable recording conditions.Indeed, 30 �M NMDA applied to mixed neuronal-glial culturesresulted in sustained Ca2� elevation that was cyclooxygenase-1-dependent (21). When acutely isolated CA1 neurons were

6082 � www.pnas.org�cgi�doi�10.1073�pnas.0605427104 Shen et al.

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

20, 2

020

exposed to 100 �M NMDA for 10 min, a large, progressive,postexposure Na�/Ca2� conductance developed that was insen-sitive to later blockade of NMDA receptors and that appearedto trigger cell injury (22). However, in our model (whole-cell-patched CA1 neurons in an acutely prepared hippocampal slice),a 6-min exposure with 10 �M NMDA did not trigger a progres-sive postexposure current (Fig. 1).

We initiated these studies to determine whether cPLA2� hada direct effect on electrical and Ca2� signaling that is believed tobe part of the excitotoxic cascade. Our findings directly relatecPLA2� activity to electrophysiologic events of excitotoxicity.Future experiments are likely to demonstrate the potential ofcPLA2� inhibition as a clinical therapy for prevention of earlyneurologic injury in a wide variety of diseases.

Materials and MethodsMice. All studies were conducted with the approval of The JohnsHopkins Animal Care and Use Committee. In these studies, weused WT and cPLA2��/� mice (23) that had been backcrossed onthe BALB/C strain for �10 generations. PLA2 activity was signif-icantly reduced in cPLA2��/� hippocampi (see SI Methods).cPLA2� mice were the gift of Joseph V. Bonventre (Brigham andWomen’s Hospital, Boston, MA).

Slice Preparation. Hippocampi were removed from 3- to 5-week-old WT and cPLA2��/� mice after decapitation. Slices (250 �m)were cut with a vibrating tissue slicer (VT1000S; Leica, Wetzlar,Germany) in ice-cold aCSF1 (in mM) 110 NaCl, 2.5 KCl, 1.2NaH2PO4, 25 NaHCO3, 0.5 CaCl2, 7 MgCl2, 2.4 pyruvate, 1.3ascorbic acid, 20 glucose, adjusted to pH 7.4, and oxygenatedwith 95% O2/5% CO2. Slices were then kept in aCSF2 (in mM)125 NaCl, 2.5 KCl, 1 NaH2PO4, 26.2 NaHCO3, 2.5 CaCl2, 1.3MgCl2, 20 glucose (pH 7.4) for at least 1 h at room temperature.

Whole-Cell, Patch-Clamp Recording. Slices were placed in a sub-merged chamber and perfused with aCSF2 at 1 ml/min. Mg2�

was omitted from and 5 �M GABAzine was added to theexternal saline. Whole-cell recordings were obtained undervisual guidance from pyramidal neuronal somata by using in-frared disseminated intravascular coagulation (DIC) micros-copy. For voltage-clamp recordings, electrodes were filled witha saline containing (in mM) 130 Cs-methanesulfonate, 10 CsCl,2 MgCl2, 10 Hepes, 0.2 EGTA, 4 Na2-ATP, and 0.4 Na-GTP (pH7.3), yielding a final resistance of 3–5 megaohms. For currentclamp recording, the internal saline contained 130 K-methanesulfonate, 10 NaCl, 2 MgCl2, 10 Hepes, 4 Na2-ATP, and0.4 Na-GTP (pH 7.3). Cells were voltage-clamped at �70 mV.Current and voltage traces were filtered at 2 kHz, digitized at 10kHz, and acquired with pCLAMP 9 software. Only cells withstable Rinput throughout the recording were used. Recordings

from neurons with spontaneous burst firing, unstable restingmembrane voltages, unstable resting membrane currents, andresting membrane potential more positive than �65 mV werediscarded. Group data were expressed as means � SEM. Sta-tistical comparisons used Student’s t test.

Glutamate Photolysis. To selectively and broadly activate theNMDA receptors on a CA1 pyramidal cell, slices were perfusedwith Mg2�-free aCSF2 supplemented with MNI-glutamate (100�M), GABAzine (10 �M), tetrodotoxin (0.5 �M), and Trolox C(40 �M). Glutamate photolysis was achieved by directing theoutput of a 100-W Hg burner lamp into the epifluorescence portof the upright microscope and reflecting light from a 400-nmdichroic mirror to the objective for full-field illumination. Amechanical shutter was used to deliver 10-msec-long test f lashesat 60- to 120-sec intervals. Experiments were performed in thepresence of NBQX (40 �M) and UBP-301 (70 �M) to blockAMPA and kainate receptors, respectively, and in the presenceof CPCCOEt (80 �M) and MPEP (1 �M) to block mGluR1 andmGluR5, respectively.

Ca2� Imaging. For pyramidal neuronal Ca2� imaging, 400 �MFluo-4 was added to internal saline, and 300 �M Alexa Fluor 594hydrazide was included to reveal dendritic morphology. We chosefluo-4, as opposed to a higher affinity indicator, because we wereconcerned that a high-affinity indicator would introduce temporaldistortion into the Ca2� transients (25) and might also buffer Ca2�

transients sufficiently to attenuate excitotoxic processes. Afterobtaining the whole-cell configuration, the dyes were allowed todiffuse into cells for at least 20 min. Ca2� transients were elicitedby brief current injections (10–50 msec; sufficient to reliably evokea single spike) every 2 min and were recorded by using a Zeiss (CarlZeiss, Thornwood, NY) Pascal scanning confocal microscopeequipped with a 40 water-immersion objective lens. Fluo-4 wasexcited with the 488-nm line of an Argon ion laser, and emitted lightwas collected through a 505-nm long-pass filter. Alexa Fluor 594hydrazide was excited with the 543-nm line of an He-Ne laser, andthe emitted light was collected through a 560-nm long-pass filter.Fluo-4 images were recorded in frame-scan mode with 4 4binning of the defined region of interest, which allowed a frame rateof 50 Hz. For analysis, foreground pixels were determined bymanually thresholding the peak spike response image, and theseforeground pixels were then spatially averaged to calculate �F/F0for each frame.

The authors thank Roland Bock and Sarah J. Texel for technicalassistance, members of the Linden laboratory for constructive critique,and Tzipora Sofare for editorial assistance. This work was supported bythe Develbiss Fund (D.J.L.), the National Natural Science Foundationof China Grant 30600168 (to Y.S.), and the Zhejiang Natural ScienceFoundation of China Grant R206018 (to Y.S.).

1. Bonventre JV, Huang Z, Taheri MR, O’Leary E, Li E, Moskowitz MA,Sapirstein A (1997) Nature 390:622–625.

2. Kalyvas A, David S (2004) Neuron 41:323–335.3. Evans JH, Leslie CC (2004) J Biol Chem 279:6005–6016.4. Kramer RM, Roberts EF, Um SL, Borsch-Haubold AG, Watson SP, Fisher MJ,

Jakubowski JA (1996) J Biol Chem 271:27723–27729.5. Nemenoff RA, Winitz S, Qian NX, Van Putten V, Johnson GL, Heasley LE

(1993) J Biol Chem 268:1960–1964.6. Clark JD, Lin LL, Kriz RW, Ramesha CS, Sultzman LA, Lin AY, Milona N,

Knopf JL (1991) Cell 65:1043–1051.7. Sapirstein A, Spech RA, Witzgall R, Bonventre JV (1996) J Biol Chem

271:21505–21513.8. Fujishima H, Sanchez Mejia RO, Bingham CO, III, Lam BK, Sapirstein A, Bonventre

JV, Austen KF, Arm JP (1999) Proc Natl Acad Sci USA 96:4803–4807.9. Rordorf G, Uemura Y, Bonventre JV (1991) J Neurosci 11:1829–1836.

10. Klivenyi P, Beal MF, Ferrante RJ, Andreassen OA, Wermer M, Chin MR,Bonventre JV (1998) J Neurochem 71:2634–2637.

11. Arai K, Ikegaya Y, Nakatani Y, Kudo I, Nishiyama N, Matsuki N (2001) EurJ Neurosci 13:2319–2323.

12. Miller B, Sarantis M, Traynelis SF, Attwell D (1992) Nature 355:722–725.13. Richards DA, Bliss TV, Richards CD (2003) Eur J Neurosci 17:2323–2328.14. Bazan NG (2003) J Lipid Res 44:2221–2233.15. Tymianski M, Charlton MP, Carlen PL, Tator CH (1993) J Neurosci 13:2085–2104.16. Randall RD, Thayer SA (1992) J Neurosci 12:1882–1895.17. Angelova P, Muller W (2006) Eur J Neurosci 23:2375–2384.18. Danthi S, Enyeart JA, Enyeart JJ (2003) J Membr Biol 195:147–164.19. Ghomashchi F, Stewart A, Hefner Y, Ramanadham S, Turk J, Leslie CC, Gelb

MH (2001) Biochim Biophys Acta 1513:160–166.20. McCullough L, Wu L, Haughey N, Liang X, Hand T, Wang Q, Breyer RM,

Andreasson K (2004) J Neurosci 24:257–268.21. Kawano T, Anrather J, Zhou P, Park L, Wang G, Frys KA, Kunz A, Cho S, Orio

M, Iadecola C (2006) Nat Med 12:225–229.22. Chen QX, Perkins KL, Choi DW, Wong RK (1997) J Neurosci 17:4032–4036.23. Bonventre JV (1997) J Lipid Mediat Cell Signal 16:199–208.24. Yang HC, Mosior M, Ni B, Dennis EA (1999) J Neurochem 73:1278–1287.25. Sabatini BL, Regehr WG (1998) Biophys J 74:1549–1563.26. Clark JD, Milona N, Knopf JL (1990) Proc Natl Acad Sci USA 87:7708–7712.

Shen et al. PNAS � April 3, 2007 � vol. 104 � no. 14 � 6083

NEU

ROSC

IEN

CE

Dow

nloa

ded

by g

uest

on

Sep

tem

ber

20, 2

020