Induction of hippocampal LTD requires nitric-oxide-stimulated PKG activity and Ca2+ release from...

Chemically Induced, Activity-Independent LTD Elicited bySimultaneous Activation of PKG and Inhibition of PKA

LINDA SANTSCHI,1 MAGALI REYES-HARDE,1 AND PATRIC K. STANTON1,2

Departments of1Neuroscience and2Neurology, Albert Einstein College of Medicine, Bronx, New York 10461

Santschi, Linda, Magali Reyes-Harde, and Patric K. Stanton.Chemically induced, activity-independent LTD elicited by simulta-neous activation of PKG and inhibition of PKA.J. Neurophysiol.82:1577–1589, 1999. Although it is widely agreed that cyclic AMP isnecessary for the full expression of long-term potentiation of synapticstrength, it is unclear whether cyclic AMP or cyclic AMP-dependentprotein kinase (PKA) play roles in the induction of long-term depres-sion (LTD). We show here that two PKA inhibitors, H-89 (10mM)and KT5720 (1mM), are unable to block induction of LTD at Schaffercollateral-CA1 synapses in hippocampal slices in vitro. Rather, H-89enhancedthe magnitude of LTD induced by submaximal low-fre-quency stimulation. Raising [cGMP] with zaprinast (20mM), a se-lective type V phosphodiesterase inhibitor,reversiblydepressed syn-aptic potentials. However, coapplication of H-89 plus zaprinastconverted this to a robust LTD that depended critically on activationof cyclic GMP-dependent protein kinase (PKG). Chemically inducedLTD is activity-independent because it could be induced withoutstimulation and in tetrodotoxin (0.5mM). Additionally, chemical LTDdid not require activation ofN-methyl-D-aspartate or GABA receptorsand could be reversed by LTP. Stimulus-induced LTD occludedchemical LTD, suggesting a common expression mechanism. In con-trast to bath application, postsynaptic infusion of H-89 into CA1pyramidal neurons did not enhance LTD, suggesting a presynaptic siteof action. Further evidence for a presynaptic locus was supplied byexperiments where H-89 applied postsynaptically along with bathapplication of zaprinast was unable to produce chemical LTD. Thussimultaneous presynaptic generation of cyclic GMP and inhibition ofPKA is sufficientto induce LTD of synaptic transmission at Schaffercollateral-CA1 synapses.

I N T R O D U C T I O N

Homosynaptic long-term depression (LTD) is an input-spe-cific, long-lasting reduction in synaptic strength that can begenerated by prolonged low-frequency stimulation in a numberof cortical areas, including at Schaffer collateral/commissural-CA1 synapses in the hippocampus (for review, see Bear andAbraham 1996; Christie 1996). Although the precise, physio-logically relevant functions of LTD are still unclear, mostneural network models include both long-term increases anddecreases in synaptic strength to accomplish memory storage,information processing, and mnemonic classifications (Bienen-stock et al. 1982; Sejnowski 1977). It is a popular hypothesisthat memories may have their cellular correlate in precisespatiotemporal patterns of synaptic strengths. Since the dem-onstration of homosynaptic, associative LTD (Stanton andSejnowski 1989), much work has focused on the elucidation of

cellular mechanisms underlying both the induction and expres-sion of cortical LTD.

Cyclic nucleotide signal transduction cascades have been sug-gested to play a role in synaptic plasticity in a wide variety ofsystems. In mammalian hippocampus, it is now well establishedthat cyclic AMP-activated pathways play necessary roles in thefull expression of long-term potentiation (LTP) of synapticstrength (Abel et al. 1997; Blitzer et al. 1995; Frey et al. 1993;Hopkins and Johnston 1988; Matthies and Reymann 1993; Stan-ton and Sarvey 1985a,b; Stanton et al. 1989). Although most ofthese studies have focused on a late-phase of LTP involving bothactivation of postsynaptic kinases and alterations in gene expres-sion and protein synthesis, there is also evidence that supports apotential presynaptic role of cyclic AMP at mossy fiber terminalsin field CA3 (Huang et al. 1994) and at Schaffer collateral/commissural terminals in field CA1 of rat hippocampal slices(Chavez-Noriega and Stevens 1994).

Recent work in our and other laboratories (Gage et al. 1997;Izumi and Zorumski 1993; Reyes-Harde et al. 1999a,b) has linkedanother cyclic nucleotide second-messenger cascade, the nitricoxide (NO)-cyclic GMP system, to the induction of LTD. As afirst step toward identifying downstream effectors of cyclic GMP,recent evidence (Reyes-Harde et al. 1999a,b) implicates presyn-aptic cyclic GMP-dependent protein kinase (PKG) as a necessarycomponent in the LTD induction pathway. Thus cyclic AMP andcyclic GMP appear to be involved reciprocally in long-termregulation of synaptic strength. However, it also has been reportedthat cyclic AMP-dependent protein kinase (PKA) activity is nec-essary to permit the induction of LTD (Brandon et al. 1995).Given that PKA and PKG have reciprocal actions in many cells,(Inoue et al. 1995; Polanowska-Grabowska and Gear 1994; Wex-ler et al. 1998), it seemed surprising to us that PKA activationwould be necessary for both LTP and LTD. We set out to test ourhypothesis of a bidirectional regulation model in which these twocyclic nucleotides would have opposing actions on LTD. Weshow here that inhibition of PKA results in an enhancement ofstimulus-evoked LTD via a PKG-dependent mechanism. In addi-tion, we demonstrate that blockade of the hydrolysis of cyclicGMP, in conjunction with simultaneous inhibition of PKA, elicitsa sustained, chemically induced, activity-independent LTDthat appears to share mechanisms in common with stimulus-evoked LTD.

M E T H O D S

Transverse hippocampal slices (400-mm thick) were obtained fromSprague-Dawley rats of either sex (14–21 days old). Briefly, animalswere anesthetized and decapitated. The brain was removed rapidly,the hippocampus dissected and placed in 4°C artificial cerebrospinal

The costs of publication of this article were defrayed in part by the paymentof page charges. The article must therefore be hereby marked “advertisement”in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

15770022-3077/99 $5.00 Copyright © 1999 The American Physiological Society

fluid (ACSF; bubbled with 95%O2-5%CO2, pH 7.4), which comprisedin (mM) 126 NaCl, 26 NaHCO3, 1.25 NaH2PO4, 5 KCl, 2 CaCl2, 2MgCl2, and 10 D-glucose. Slices were cut, using an “egg-slicer,”which rapidly forced a parallel grid of 20-mm-diam wires spaced 400mm apart through the tissue. Individual slices were placed in aninterface recording chamber at 33°C, and perfused at a rate of 3ml/min with ACSF. When appropriate, one input served as a controlto verify the input specificity of LTD. Evoked population excitatorypostsynaptic potentials (fEPSPs) were recorded extracellularly in theapical dendritic field instratum radiatumfor a stable baseline periodof $30 min. Stimulation intensity was adjusted so that the amplitudeof each fEPSP was;50% of the maximum response amplitude beforegeneration of a population action potential ('2 mV). Experiments inwhich there was a drift in baseline of.5% were excluded fromfurther analysis. De novo LTD was induced by a 1 Hz/15 min (900stimuli) train of low-frequency stimulation (LFS). A “submaximal”LFS consisted of 400 pulses, also at a frequency of 1 Hz. For thereversal experiments, LTP was induced by a theta burst stimulation(TBS) paradigm where we gave 10 high-frequency bursts (100 Hz, 5pulses each) repeated at 5 bursts/s (200-ms interburst interval). ThisTBS was repeated four times spaced 15 s apart. The stimulus intensitywas increased during TBS so as to double the amplitude of the evokedEPSP. In experiments performed in the presence of GABAergicantagonists, [Mg21]o was raised to 4 mM, and area CA1 was isolatedby making a cut in stratum radiatum between fields CA3 and CA1. Forthe occlusion experiments, two separate inputs of Schaffer collateral-commissural axons were isolated by placement of stimulating elec-trodes on opposite sides of the recording site (extracellular electrodesfilled with 2 M NaCl), verified as separate inputs by a lack ofpaired-pulse interactions between them (50 ms interstimulus inter-vals), and alternately stimulated every 30 s using bipolar, stainlesssteel electrodes (150ms DC square pulses). Synaptic strength wasassessed by measuring the maximum slope of the falling phase of thefEPSP. Changes in synaptic strength after LFS and/or drug additionare expressed relative to the normalized pre-LFS baseline. The max-imum initial negative slopes of the fEPSPs were calculated using asix-point interpolation least-squares linear regression analysis. Sum-mary data are expressed as mean6 SE with significance preset to theP , 0.05 level, using Student’st-tests for unpaired data.

We recorded intracellular evoked EPSPs in CA1 pyramidal neurons(RMP 5 261 6 1.6 mV; RN 5 55.86 2.4 MV) impaled with sharpmicroelectrodes (90–130 MV, 2 M KAcetate). Maximum initial pos-itive slope of the EPSP (V/s) was measured to assess changes inexcitatory synaptic strength, and each experiment normalized to itspre-LFS baseline value. In most experiments, simultaneous field re-cordings were made to monitor recording condition stability. Stocksolutions of H-89 (Biomol), KT5720 (LC Labs/Alexis Corp), KT5823(Biomol), picrotoxin (Sigma), and zaprinast (Sigma) were dissolvedin dimethyl sulfoxide (DMSO) as 1,0003 stocks and diluted either inACSF for bath application or in 2 M KAcetate for intracellularinfusion. D-2-amino-5-phosphonopentanoic acid (D-AP5; TocrisCookson) was dissolved in aqueous NaOH.Nv-L-nitro-arginine(Sigma) was dissolved in 1 M HCl. Tetrodotoxin (Sigma) and CGP35348 (kind gift of Ciba-Geigy Pharmaceuticals) were dissolved indH2O to a stock concentration of 1 mM and 0.5 M, respectively.Forskolin (Sigma) was dissolved in DMSO to a 50 mM stock con-centration. Intracellular and extracellular control experiments usedequal concentrations of DMSO vehicle alone, and expression of LTDwas unaffected by this vehicle at any concentration employed.

R E S U L T S

PKA inhibitors do not block the induction ofhomosynaptic LTD

To test the necessity for PKA activation for the induction ofLTD at Schaffer collateral-CA1 synapses, we employed two

selective, cell-permeant PKA inhibitors, H-89 (a synthetic in-hibitor) (Chijiwa et al. 1990) and KT5720 (a secondary me-tabolite-derived inhibitor) (Kase et al. 1987). Figure 1A illus-trates the induction of LTD by LFS (1 Hz/900 s; solid bar) inthe presence of bath-applied H-89 (10mM; open bar). LTD inthese slices was indistinguishable from control LTD (236.662.9%;n 5 16, data not shown) in either magnitude or duration(LTD in H-89 5 235.1 6 6.0%; P . 0.20; Student’st-testcompared with controls;n 5 9). Similarly, the second PKAinhibitor, KT5720 (1mM; open bar; Fig. 1B), was also unableto block the induction of LTD by LFS (LTD in KT57205235.26 6.4%;n 5 6). Taken together, these experiments leadto the conclusion that PKA activation is not a necessary step inthe induction of homosynaptic LTD at Schaffer collateral-CA1synapses.

Homosynaptic LTD is enhanced by inhibiting PKA

The preceding observations, and previous studies showingthat PKA activity is necessary for the full expression of LTP(Frey et al. 1993; Matthies and Reymann 1993; Stanton andSarvey 1985a,b; Stanton et al. 1989), raise the question ofwhether preventing activation of PKA in fact might enhance orunmask the induction of a reciprocal modification of synapticstrength, LTD. To test this hypothesis, we used a shorter,submaximal LFS stimulation, 1 Hz for 400 s, which inducedonly a small amount of LTD in control slices (Fig. 1C;211.56 2.9%;n 5 11). In a second group of slices (Fig. 1D;n 5 14), we bath-applied H-89 (10mM; open bar) 45 minbefore submaximal LFS of Schaffer collateral/commissuralaxons (1 Hz/400 s; solid bar). In contrast to control slices,submaximal LFS elicited virtually maximal LTD when PKAwas inhibited (235.4 6 4.1%, P , 0.05; Student’st-testcompared with controls 60 min post-LFS). This intriguingresult prompted further investigation into the synaptic locus ofaction of H-89 as well as studies aimed at testing the possibilityof a co-requirement for an NO-cyclic GMP-PKG pathway(Gage et al. 1997) in this “unmasked” LTD.

Postsynaptic PKA inhibition does not enhance LTD

In a recent study (Gage et al. 1997), we presented evidencethat at least one form of homosynaptic LTD requires activationof a presynaptic cyclic GMP-mediated cascade triggered by thegaseous intercellular messenger NO. To test the possibility thatPKA counteracts this cascade directly, it was necessary todetermine whether H-89 acts pre- and/or postsynaptically. Toaccomplish this, we used sharp intracellular microelectrodes tofill single CA1 pyramidal neurons with H-89. Intracellularelectrodes were backfilled with a solution containing a 50-foldhigher concentration of H-89 (500mM) than that which waseffective in enhancing LTD when applied extracellularly.

Figure 2A illustrates the effect of submaximal Schaffer col-lateral LFS (1 Hz/400 s; solid bar) on intracellularly recordedEPSPs in control, untreated CA1 pyramidal neurons (n 5 7).The left insetshows examples of single EPSPs before and aftersubmaximal LFS in a control cell. Theright inset illustratesburst-induced afterhyperpolarization (AHPs;10.5–1.0nA/100ms injected current) before and after bath application of theadenylate cyclase stimulant forskolin (100mM), which eliciteda well-characterized PKA-mediated block of the late AHP(Madison and Nicoll 1986).

1578 L. SANTSCHI, M. REYES-HARDE, AND P. K. STANTON

Figure 2B illustrates the effects of identical submaximal LFSon CA1 pyramidal neurons filled with H-89 to selectivelyinhibit postsynaptic PKA. CA1 pyramidal neurons were im-paled with microelectrodes filled with 500mM H-89 (in 2 MK1 acetate1 0.5% DMSO; Re 5 90–130 MV). At least 45min were allowed for leakage of H-89 into neurons (open bar),after which submaximal LFS (1 Hz/400 s; solid bar) wasapplied to one Schaffer collateral input, while a second inputserved as control (not shown). In contrast to extracellularapplication of H-89, selective inhibition of postsynaptic PKAdid not statistically enhance the magnitude of LTD evoked bysubmaximal LFS (LTD in control5 210.4 6 6.7%, H-89treated5 217.4 6 4.3%, n 5 8; P . 0.20, Student’st-testcompared with control intracellular EPSPs 30 min post-LFS).The left inset illustrates representative EPSPs before and aftersubmaximal stimulation.

To verify that postsynaptic PKA in fact was blocked underour recording conditions, we tested the ability of the adenylatecyclase stimulant forskolin to block the burst-evoked AHP inthree of these cells. Theright inset illustrates substantial butnot complete prevention of forskolin’s actions in cells infusedwith H-89 compared with control cells, confirming blockade of

PKA. In fact, H-89 (10 mM) produced virtually identical('80%) block of forskolin’s actions when applied extracellu-larly or intracellularly. Taken together, these experiments sug-gest that the majority (but perhaps not all) of the enhancementof LTD produced by inhibition of PKA is most likely due to apresynaptic site of action.

NO synthase is necessary for the form of LTD unmasked byinhibiting PKA

We have shown recently that full LFS-evoked homosynapticLTD consists of both NO-guanylyl cyclase-dependent and-independent components (Gage et al. 1997). However, it wasnot clear which component(s) contribute to the LTD unmaskedby inhibiting PKA in our current experiment. Therefore, weused the cell-permeant NO synthase inhibitorNv-nitro-L-ar-ginine (L-NA) to determine whether NO production is, in fact,a necessary component of LTD evoked under either condition.Figure 3A (n 5 6) illustrates that, in agreement with previousreports (Izumi and Zorumski 1993; Otani and Connor 1995),bath application of L-NA (100mM; open bar) partially reducedthe magnitude of LTD produced by LFS (1 Hz/900 s; solid bar;

FIG. 1. Inhibition of cyclic AMP-dependent protein kinase (PKA) enhances the magnitude of long-term depression (LTD).A:time course of LTD of Schaffer collateral/commissural-CA1 synaptic transmission in hippocampal slices (n 5 8) pretreated by bathapplication of the PKA inhibitor H-89 (10mM, open bar). After a 30-min baseline period, H-89 was bath applied for 45 min beforelow-frequency stimulation (LFS), with wash-out commencing 30 min post-LFS. LFS (1 Hz/900 s; solid bar) evoked significant,homosynaptic LTD of synaptic strength.Inset: single Evoked population excitatory postsynaptic potentials (fEPSPs) duringbaseline acquisition (1) and 40 min after (2) LFS (calibration bars: 2 mV/5 ms; each point is the normalized mean6 SE).B: bathapplication of a different PKA inhibitor, KT5720 (1mM, open bar), also failed to block the induction of LTD by LFS (1 Hz/900s; solid bar). KT5720 was bath applied 30 min before LFS and washed out 15 min post-LFS.Inset: single fEPSPs during baselineacquisition (1) and 40 min after induction of LTD (2) (calibration bars: 2 mV/5 ms; each point is the normalized mean6 SE ofresponses from 6 slices).C: time course of submaximal LTD elicited in 11 slices by a shorter LFS (1 Hz/400 s; solid bar). Incontrast to the 900s LFS, this stimulus paradigm did not elicit maximal LTD.Inset: representative fEPSPs during baselineacquisition (1) and 40 min after (2) submaximal LFS (each point is the normalized mean6 SE).D: in contrast to control slices,slices (n 5 14) pretreated with H-89 (10mM, open bar) exhibited markedly and significantly larger LTD both during and aftersubmaximal LFS (1 Hz/400 s).Inset: representative fEPSPs during baseline acquisition (1) and 40 min after (2) submaximal LFS.

1579CHEMICALLY INDUCED, ACTIVITY-INDEPENDENT LTD

223.7 6 2.9% compared with236.6 6 2.9% in controls).NOS inhibition did not, however, completely block LTD,consistent with the existence of an NO-independent componentof LTD (Gage et al. 1997).

Next, we coapplied L-NA (100mM) plus H-89 (10mM) 45min before giving submaximal LFS (1 Hz/400 s; solid bar).Figure 3B (n 5 8) illustrates that the robust LTD previouslyseen in the presence of H-89 was blocked almost completely byinhibition of NOS activity. The existence of a small, residualamount of depression (217.46 5.5%) suggests that a part of

the observed H-89 effect also may be NO-independent, per-haps acting through a postsynaptic cascade involving proteinphosphatases (Mulkey et al. 1993).

Cyclic GMP accumulation alone is not sufficientto induce LTD

One physiological function of presynaptic nitric oxide is tobind to the heme moiety of soluble guanylyl cyclase (sGC),causing the synthesis of cGMP. However, it was unclearwhether an elevation in [cGMP] alone could be sufficient toelicit LTD. Isozyme-specific phosphodiesterase (PDE) inhibi-tors are useful pharmacological tools for examining the phys-iological roles of cyclic nucleotides in a variety of systems(Beavo 1995). In particular, it is the type V PDE that isprimarily responsible for the selective breakdown of cyclicGMP. Therefore, a selective PDE V inhibitor, such as zapri-nast, markedly and selectively elevates intracellular [cGMP](Gillespie and Beavo 1989).

FIG. 3. Inhibition of nitric oxide synthase (NOS) partially blocks induction ofLTD and fully prevents the ability of H-89 to enhance LTD.A: time course ofexperiments (n 5 6) where the selective NOS inhibitorNv-nitro-L-arginine(L-NA; 100 mM; open bar), was bath applied 30 min before full LFS (1 Hz/900s; solid bar). Magnitude of LTD was reduced significantly in L-NA compared withuntreated slices (each point is the normalized mean6 SE). B: time course ofexperiments (n 5 8) in which L-NA (100mM) and H-89 (10mM) were coappliedin the bath (open bar) 45 min before submaximal Schaffer collateral LFS stimu-lation (1 Hz/400 s; solid bar). Inhibiting NOS markedly reduced the LTD inducedin H-89 by submaximal LFS (each point is the normalized mean6 SE).

FIG. 2. Presynaptic PKA inhibition is necessary to enhance LTD.A: timecourse of control experiments (n 5 7) plotting intracellular Schaffer collateral-evoked EPSP slopes from single CA1 pyramidal neurons. Submaximal Schaf-fer collateral LFS (1 Hz/400 s; solid bar) evoked only a small LTD, similar toextracellular recordings (Fig. 1C). Left inset: intracellular EPSPs before (1) and30 min after (2) onset of submaximal LFS (calibration bars: 4 mV/5 ms).Rightinset: burst-induced afterhyperpolarizations (AHP) in a CA1 pyramidal cellbefore (Control) and 15 min after (Forskolin) bath application of the adenylatecyclase stimulant forskolin (100mM), which blocked the late AHP (actionpotentials truncated; calibration bars: 4 mV/300 ms).B: time course of aparallel set of experiments (n 5 8) where CA1 pyramidal neurons wereimpaled with microelectrodes containing 500mM H-89 (hatched bar). Afterallowing$45 min for H-89 to diffuse into the cell, submaximal LFS (1 Hz/400s; solid bar) was applied to 1 Schaffer collateral/commissural synaptic input,while a 2nd input served as an unstimulated control. With postsynaptic PKAblockade, submaximal LFS resulted in markedly less LTD than that observedin slices where H-89 was bath applied (B). Insetsare as inA, where thesignificant prevention of PKA-dependent blockade of the AHP by forskolin(100 mM) confirmed perfusion of H-89 into the postsynaptic neuron.


Figure 4A shows that bath application of zaprinast (5mM;open bar;n 5 6) caused a transient depression of synapticEPSPs that fully reversed within 30 min of drug washout. Thusraising [cGMP] by this method, in the absence of any stimulustrain, produced only a transient, rapidly reversible short-termdepression, consistent with previous reports (Boulton etal.1994). Similarly, Fig. 4B demonstrates that zaprinast (openbar; n 5 6), added along with coincident submaximal LFS (1Hz/400 s; solid bar), produced a depression that, though muchlarger in magnitude, was still fully reversible on drug washout.Using a higher concentration of zaprinast (20mM), Fig. 4C(n 5 8) shows that, even at this concentration, zaprinast-induced depression was still rapidly and completely reversibleon drug washout. Thus, the reversibility of zaprinast’s effect isprobably not attributable to some threshold level in the [cGMP]achieved. Zaprinast at this concentration (20mM), paired withsubmaximal stimulation, also produced only a reversible de-pression (n 5 2, data not shown). Taken together, these ex-periments indicate that simply elevating [cGMP] alone is in-sufficient to elicit fully saturated, stable LTD of synapticstrength.

Simultaneously raising [cyclic GMP] while inhibiting PKAinduces LTD

Studies in our and other laboratories (Boulton et al. 1994;Gage et al. 1997) support a role for cGMP in synaptic depres-sion at CA1 synapses in rat hippocampal slices. Because rais-ing cGMP alone is insufficient to produce a long-lasting de-pression, but we do find a marked enhancement in themagnitude of LTD when another signaling system (PKA) isinhibited, we hypothesized that these two pharmacologicalmanipulations in concert might supply the factors needed toproduce a long-lasting depression of synaptic strength.

In contrast to the effect of zaprinast alone, Fig. 4D illustrates

the induction of an apparently saturated LTD when zaprinast(20 mM; solid bar;n 5 8) and H-89 (10mM; open bar) werecoapplied. This “chemical LTD” remained stable for$2 hfollowing washout of both drugs (241.8 6 3.8%,P , 0.05,Student’st-test, 60 min postapplication compared with preap-plication baselines). These data support our hypothesis thatthese two cyclic nucleotide second messengers mediate oppos-ing biological signals and that simultaneous elevation of[cGMP] and inhibition of PKA is, in fact,sufficientto elicitLTD at Schaffer collateral-CA1 synapses.

Chemical LTD is not associated with alteration in paired-pulse facilitation

Our intracellular data with postsynaptic infusion of H-89(Fig. 2B) suggests that the synaptic locus of this effect maybe presynaptic. A common, but not universal, feature of apresynaptic site of action is an alteration in paired pulsefacilitation (PPF) (Zucker 1989). PPF is an enhancement inthe response magnitude to a second stimulus when precededat a short interstimulus interval (10 –100 ms) by a condi-tioning stimulus. PPF is believed, though not without de-bate, to result from residual-free Ca21 in the presynapticterminal, which is retained near sites of transmitter releasefor a period of time after the conditioning pulse (Katz andMiledi 1968; but see also Bertram et al. 1996).

We compared PPF ratios before and after the induction ofchemical LTD (Fig. 5). The pooled chemical LTD that waselicited is illustrated in Fig. 5A (n 5 13) and was similar toprevious experiments. Associated PPF was measured at twointerstimulus intervals, 50 ms (Fig. 5B; n 5 8) and 15 ms (Fig.5C; n 5 5). PPF at either paired-pulse interval was not con-sistently altered during chemical LTD.

FIG. 4. Raising intracellular [cGMP] causes only re-versible depression of synaptic transmission, but whencombined with inhibition of PKA, elicits a long-lastingLTD. A: time course of experiments (n 5 6) in which theselective type V isozyme-specific phosphodiesterase(PDE) inhibitor zaprinast (5mM) was bath applied for 40min (open bar). Zaprinast caused a transient depressionthat rapidly returned nearly to baseline within 30 min ofdrug washout.B: when submaximal Schaffer collateralLFS (1 Hz/400 s; solid bar) was given during bath appli-cation of zaprinast (5mM, open bar;n 5 6), it evokedsignificantly larger acute depression than with zaprinastalone, but synaptic strength still recovered to the sameextent and with similar kinetics as in zaprinast alone (eachpoint is the normalized mean6 SE).C: bath applicationof a higher concentration of zaprinast (20mM; open bar;n 5 8) evoked a marked, but again reversible, depressionof Schaffer collateral-CA1 synaptic strength. Time courseof recovery was significantly longer than in 5mM zapri-nast (Fig. 4A). Inset: single fEPSPs during baseline ac-quisition (1) and 70 min after (2) washout of zaprinast(calibration bars: 2 mV/5 ms; each point is the normalizedmean6 SE). D: in contrast to zaprinast alone, coappli-cation of zaprinast (20mM; solid bar; n 5 8) in thepresence of the PKA inhibitor H-89 (10mM; open bar)now evoked an LTD that stably persisted for.2 h of drugwashout.Inset: single fEPSPs during baseline acquisition(1) and 70 min after (2) washout of both H-89 andzaprinast. (Calibration bars: 2 mV/5 ms. Each point is thenormalized mean6 SE).


Chemical LTD requires PKG activity

While cGMP can exert its biological effects through a num-ber of effector cascades, it is unclear which of these down-stream pathways are required to induce LTD. A primary actionof cGMP is the stimulation of PKG, which in many systems is

considered the primary intracellular receptor protein forcGMP. Recent data from our laboratory (Reyes and Stanton1997) demonstrate that PKG activation is necessary for theinduction of LTD. This result, together with our present datainvolving PKA inhibition, led us to hypothesize a role forreciprocal regulation of long-term synaptic plasticity by thetwo cyclic nucleotide-activated protein kinases. To determineif the LTD observed in H-89 also requires the coincidentactivation of PKG, we used the selective, cell-permeant PKGinhibitor KT5823. Figure 6A illustrates that, when H-89 andKT5823 (10 and 1mM respectively; open bar) were co-bathapplied 45 min before submaximal LFS (1 Hz/400 s; solid bar),the H-89 unmasking of LTD was blocked completely (24.361.9%; n 5 10). This shows that the unmasking of LTD byinhibiting PKA does require the unopposed activity of PKG.Just as significantly, this result also ensured that the LTD-enhancing effect seen with H-89 was, in fact, due exclusivelyto PKA inhibition and not attributable to cross-inhibitionof PKG.

Because LTD in H-89 was prevented by simultaneous ad-dition of KT5823, we investigated whether PKG activity alsois needed for chemically induced LTD. Figure 6B illustratesthat blockade of PKG could, indeed, prevent the expression ofchemical LTD produced by raising [cGMP] while inhibitingPKA (60 min postwash;26.16 9.7%;n 5 4). These data lendstrong support to the contention that PKG is the primaryphysiological target for cyclic GMP generated in response tozaprinast and that PKG activity unopposed by PKA is what isrequired to yield stable, long-lasting LTD.

Both PKG and PKA function presynaptically to regulatechemical LTD

Before considering possible substrates for PKG, it was firstnecessary to determine the synaptic locus of PKG activityrequired for LTD. To address this, we dissolved KT5823 (50mM) at 50 times the extracellularly effective concentration in 2M potassium acetate plus 0.5% DMSO and backfilled micro-electrodes for intracellular perfusion. Intracellular evoked EP-SPs in single CA1 pyramidal neurons were recorded for abaseline period of$45 min to allow KT5823 to leak into thepostsynaptic neuron. H-89 (10mM) then was bath appliedalone for 10 min, followed by 30-min bath application of H-89plus zaprinast (20mM). Figure 6C illustrates the results ofthese experiments in which postsynaptic infusion of KT5823was unable to block the expression of chemical LTD (243.161.2%;n 5 6). This result supports the hypothesis that presyn-aptic, not postsynaptic, PKG is necessary for chemical LTD.

Our previous experiment (Fig. 2) suggests a presynapticlocus for PKA inhibition in the enhancement of LTD; however,the site of action of H-89 in the expression of chemical LTDcould be different. To answer this question, we backfilledelectrodes with H-89 (500mM) for intracellular infusion ofCA1 pyramidal neurons. After 45 min of infusion and baselinerecording, we bath applied zaprinast (20mM) for 30 min,followed by drug washout. Figure 6D shows that chemicalLTD could no longer be elicited when H-89 was appliedpostsynaptically (210.5 6 8.5%; n 5 6), indicating that thenegative role for PKA important to the induction of LTD isprobably also presynaptic.

FIG. 5. Paired-pulse facilitation (PPF) is unaffected by chemical LTD.A:time course of LTD of fEPSP slopes in all experiments (n 5 13) where PPFwas monitored before and after the induction of chemical LTD (each point isthe normalized mean6 SE). B: time course of paired-pulse facilitation,measured as the ratio of fEPSP slopes of the 2nd to 1st response (P2/P1),measured at an interstimulus interval of 50 ms (n 5 8). Chemical LTD failedto produce a change in the average PPF ratio (each point is the mean6 SE).C: time course of experiments (n 5 5) in which the PPF ratio was monitoredbefore and after chemical LTD induction, using 15-ms interstimulus interval.At this shorter interval, PPF was similarly unaffected by chemical LTDinduction (each point is the mean6 SE).


Chemical LTD is reversed by HFS and occluded bystimulus-evoked LTD, but is activity-independent

Associated with studies of LTD is an ongoing concern thatdepression of synaptic potentials may reflect irreversible syn-aptic damage rather than a reversible physiological phenome-non. Figure 7A demonstrates that, similar to stimulus-inducedLTD, chemical LTD can be reversed by the induction of LTP.After inducing stable chemical LTD (233.36 0.76%), high-frequency stimulation (HFS; theta burst stimulation; 4 trains of100 Hz/5 pulse burst times 10, interburst interval 200 ms; seeMETHODS ) was applied 45 min after drug washout (TBS indi-cated by arrows), which evoked robust LTP (1208% changefrom chemically depressed baseline). fEPSPs remained stablypotentiated for 1 h, at which time a second TBS was givenwhich resulted in a slight additional potentiation. A third TBSgiven 45 min later elicited no further potentiation, indicatingthat LTP was saturated at this level. This demonstration ofreversibility argues against irreversible damage underlyingchemical LTD.

These data also hint that stimulus-induced and chemical

LTD may share some common mechanisms because both canbe reversed by the same HFS. We set out to test the hypothesisthat these may not be separate and distinct forms of LTD bydetermining whether saturation of stimulus-induced LTD atone set of synapses could selectively occlude chemical LTD.As shown in Fig. 7B, we alternately stimulated each of twoindependent Schaffer collateral pathways once every 30 s (S1and S2) and recorded the evoked fEPSPs in stratum radiatumof field CA1. After the baseline period, the S1 input (solidcircles) received repeated LFS trains (4–5 trains, 1 Hz/600 s),spaced 10 min apart, until saturated, homosynaptic LTD wasinduced. Ten minutes after the last train, we bath applied H-89(10 mM; open bar) for 30 min, followed by coapplication ofH-89 plus zaprinast (20mM; solid bar) for an additional 30min. In the presence of both inhibitors, both inputs did, indeed,decrease markedly in strength. However, on drug removal, thesaturated input (S1) returned to its former level of depression(S1 saturated, prechemical LTD:232.86 4.7%; postchemicalLTD: 236.2 6 5.6%). In contrast, the naive input (S2, opencircles) demonstrated chemical LTD that was indistinguishable

FIG. 6. Presynaptic cyclic GMP-dependent protein kinase (PKG) activity and PKA inhibition are both necessary for chemicalLTD expression.A: time course of experiments (n 5 10) where the selective PKG inhibitor KT5823 (1mM) plus H-89 (10mM)were co-bath-applied 30 min before submaximal LFS (1 Hz/400 s; solid bar). KT5823 completely prevented the abilty of H-89 toenhance LTD (each point is the normalized mean6 SE).B: time course of experiments (n 5 4) in which KT5823 (1mM) and H-89(10 mM; open bar) were co-applied 30 min before zaprinast addition (20mM; closed bar). Zaprinast caused equal short-termdepression of synaptic potentials, but inhibition of PKG completely blocked chemically induced LTD (each point is the normalizedmean6 SE). C: time course of experiments (n 5 6) in which CA1 pyramidal neurons were impaled with microelectrodescontaining 50mM (hatched bar) KT5823. A 15-min period of stabilization was allowed before the onset of baseline acquisition.A 30-min baseline recording period was followed by a 10-min bath application of H-89 (10mM; open bar), then a 30-mincoapplication of zaprinast (20mM; solid bar). Postsynaptic infusion of KT5823 failed to prevent the expression of chemicallyinduced LTD, supporting a presynaptic locus of chemical LTD.Inset: intracellular EPSPs at onset of H-89 application (1) and 30min after drug washout (2) (calibration bars: 4 mV/15 ms; each point is the normalized mean6 SE). D: time course of 6experiments in which CA1 pyramidal neurons were impaled with microelectrodes containing 500mM H-89 (hatched bar). After15-min stabilization and a 30-min baseline period, zaprinast (20mM; solid bar) was applied for 30 min. Postsynaptic infusion ofH-89 combined with bath application of zaprinast failed to produce a long-lasting depression of synaptic strength, again consistentwith a presynaptic site for chemical LTD.Inset: intracellular EPSPs before zaprinast addition to the bath (1) and 40 min after (2)zaprinast washout (calibration bars: 4 mV/15 ms; each point is the normalized mean6 SE).


in magnitude from stimulus-induced LTD (242.0 6 6.4%,P . 0.20, Student’st-test comparing S1 to S2). The occlusionof chemical LTD by stimulus-evoked LTD further supports thehypothesis that they share at least some common expressionmechanisms.

To determine whether chemical LTD requires any synapticactivity at all for its generation, we performed experiments whereall synaptic stimulation was suspended during the period of drugapplication and not resumed until 30 min postwashout. As shownin Fig. 7C, virtually identical chemical LTD was elicited in theabsence of electrical stimulation (237.1 6 3.1%, n 5 5). Tocompletely silence both evoked and spontaneous synaptic activity,we also performed the same experiment in the presence of theNa1 channel blocker TTX (0.5mM). As Fig. 7D shows, chemicalLTD was induced in the presence of TTX, after which all threedrugs were washed out. Although considerable time was neededfor TTX to completely wash out of these slices ('80 min), oncefEPSPs stabilized, the magnitude of LTD was the same as incontrols (236.26 4.3%;n 5 4). Thus, chemically induced LTDdoes appear to converge on some of the same mechanisms asstimulus-induced LTD while bypassing the need for any synapticstimulation.

Induction of chemical LTD does not require either NMDA orGABA receptor activation

An initial trigger for one form of stimulus-induced LTD isactivation of theN-methyl-D-aspartate (NMDA) class of glu-tamate receptors (Dudek and Bear 1992). One possible phys-iological cascade for raising presynaptic [cGMP] begins withthe activation of postsynaptic NMDA receptors, Ca21 influx,and activation of the Ca21/calmodulin-dependent enzyme NOsynthase (NOS). NO then diffuses readily across membranes,allowing it to act on neighboring cells and/or presynapticterminals (Boulton et al.1994). One of the targets activated byNO is soluble guanylate cyclase (sGC), elevating [cGMP]. Itseems likely that chemical LTD, by directly elevating [cGMP],bypasses the need for NMDA receptor activation. To testwhether NMDA receptor activation is required for chemicalLTD, we attempted to elicit chemical LTD in the presence ofthe NMDA receptor blocker D-AP5. Figure 8A shows thatchemical LTD still could be evoked despite NMDA receptorblockade (50mM AP5; hatched bar;n 5 8). These resultsdemonstrate that effectively raising [cGMP] in response tozaprinast is enough to overcome the need for NMDA receptor-

FIG. 7. Chemically induced LTD is reversed by high-frequency stimulation (HFS), occluded by stimulus-induced LTD, but isactivity-independent.A: representative experiment in which repeated trains of high-frequency theta burst stimulation (HFS, seeMETHODS) were given at 3 time points after drug washout (45, 105, and 150 min postdrug application). HFS reversed chemicallyinduced LTD. Schaffer collateral-evoked EPSPs remained stably potentiated until recording was terminated.B: 2 independentSchaffer collateral pathways (S1 and S2; seeinset) were stimulated alternately each 30 s (n 5 5). After 30 min of baselinerecording, 1 input (S1, closed circles) received repeated trains of LFS (4–5 trains; 1 Hz/600 s) at 10-min intertrain intervals untilsaturated LTD was evoked. Ten minutes after the last LFS train, H-89 (10mM; open bar) was bath applied for 30 min; after whichtime zaprinast (20mM; solid bar) was coapplied for an additional 30 min. Significant chemical LTD was evoked only at the naivesynapses, but not those where stimulus-induced LTD had been saturated (each point is the normalized mean6 SE).C: in a separategroup of 5 slices, all electrical stimulation was terminated immediately before drug application (10mM H-89; 20 mM zaprinast,open and solid bars respectively) and resumed 30 min after drug washout. Expression of chemical LTD was indistinguishable fromstimulated slices (each point is the normalized mean6 SE). D: TTX (0.5 mM) was coapplied throughout the period of H-89addition (10mM, open bar;n 5 4) to block all action potential generation, and electrical stimulation suspended during drugapplication and the first 30 min of drug washout. Under these conditions, chemical LTD was still elicited (each point is thenormalized mean6 SE).


gated Ca21 influx, provided an opposing cAMP-activated sig-naling pathway is inhibited concomitantly.

In considering the synaptic locus of chemical LTD, alter-ations in inhibitory circuitry that modulate excitatory transmis-sion are also potential contributors, even though blockade ofGABAergic inhibition has been shown rather to enhance stim-ulus-evoked LTD (Wagner and Alger 1995). Trudeau et al.(1997) have shown that hippocampal inhibitory postsynapticpotentials (IPSPs) are unusually sensitive to PKA activation,raising the possibility that the effect of PKA inhibition couldinclude an indirect effect on inhibitory tone. Therefore, wetested whether chemical LTD still could be elicited underconditions of GABA receptor blockade. Experiments wereconducted in the presence of both the GABAA receptor antag-onist picrotoxin (10mM), and the GABAB receptor blocker,

CGP 35348 (400mM; hatched bar). Figure 8B (n 5 8) showsthat the expression of chemically induced LTD does not re-quire intact GABAergic modulation, supporting a direct effectof chemical LTD on monosynaptic excitatory transmission.

D I S C U S S I O N

While there has been some progress in elucidating cellularmechanisms underlying LTD, we are far from a completeunderstanding of either the molecular processes or cognitivefunction(s) of this form of plasticity. Interestingly, many of thesame second-messenger cascades appear to be involved in bothLTP and LTD, leading one to question where the critical pointsof divergence in the two phenomena may be. Here, we supplythe first evidence of reciprocal regulation of LTD by cyclicGMP-(LTD enhancing) and cyclic AMP-(LTD suppressing)dependent protein kinases in presynaptic nerve terminals.

The fact that PKA plays an important role in the persistenceof LTP is well established (Frey et al. 1993; Hopkins andJohnston 1988; Stanton and Sarvey 1985a,b; Stanton et al.1989). If PKA-mediated phosphorylation of protein substratesfavors the expression of LTP, we predicted that inhibition ofPKA should shift phosphorylation states in favor of LTD. Inthis study, we demonstrate a marked enhancement of themagnitude of LTD when PKA is inhibited. This result is incontrast to the observations of Brandon et al. (1995), whoreported a complete block of LTD in control slices by PKAinhibition. The reasons for these differences are unclear. Oneexplanation could be that effects of PKA blockade may be ageand/or genus dependent. In our study, we used juvenile rats(14–21 days), whereas older mice (4–6 wk) were used byBrandon et al. (1995).

Intracellular infusion of H-89 into postsynaptic pyramidalneurons produced far less enhancement of LTD than bathapplication, suggesting that the effect involves, at least in part,a presynaptic site of action. However, because the delivery ofsubstances from sharp microelectrodes is difficult to ensure, weused a 50-fold higher concentration of H-89 than for bathapplication and allowed ample diffusion time ($45 min) be-fore stimulation. We previously have found these methods tobe effective in achieving intracellular activity (Reyes and Stan-ton 1996; Stanton and Gage 1996), but we further verified thepresence of inhibition directly by testing the effect of H-89infusion on a well-characterized postsynaptic PKA-dependentresponse, suppression of the AHP (Madison and Nicoll 1986).In control cells, PKA-dependent inhibition of the AHP by theadenylate cyclase stimulant forskolin was complete. Forsko-lin’s block was reduced.80%, but not completely prevented,in cells impaled with electrodes containing 500mM H-89.

Although postsynaptic infusion of H89 as well as KT5823support a presynaptic site for LTD, our PPF data does not. PPFis believed to reflect enhanced transmitter release due to resid-ual presynaptic [Ca21]i, fostering the idea that manipulationsthat change release probability should alter PPF. It is morequestionable whether PPF alterations are a sensitive measure ofpresynaptic plasticity. Early field potential studies in field CA1failed to show changes in PPF after induction of LTP (Mullerand Lynch 1989), whereas some later studies have reportedsignificant decreases in average PPF ratio (Kleschevnikov et al.1997). It has been suggested that intracellular EPSPs may bemore accurate in assessing PPF because they are less contam-

FIG. 8. Chemically induced LTD does not requireN-methyl-D-aspartate(NMDA) receptor activation or intact GABAergic inhibition.A: NMDA re-ceptor antagonistD22-amino-5-phosphonopentanoic acid (AP5; 50mM,hatched bar;n 5 7) was bath applied 15 min before H-89. H-89 and zaprinast(open and solid bars respectively) then were applied as previously, in thecontinued presence of AP5. Chemical LTD still was elicited, despite NMDAreceptor blockade (each point is the normalized mean6 SE). B: GABAA

receptor antagonist picrotoxin (10mM) and the GABAB blocker CGP 35348 (1mM; hatched bar;n 5 8) were bath applied 30 min before H-89. Theseexperiments were peformed in 4 mM Mg21 to reduce hyperexcitability, andarea CA1 was isolated by cutting Schaffer collateral axons in stratum radiatum,separating them from CA3 neurons. Chemical LTD still was expressed duringGABA receptor blockade.Inset: fEPSPs at the onset of picrotoxin plus CGP35348 application (1), 20 min in H-89 application (2), and 40 min afterwashout of all compounds (3) (calibration bars: 2 mV/5 ms; each point is thenormalized mean6 SE).


inated by polysynaptic events or population action potentials.Unfortunately, intracellular results have proven just as equiv-ocal; some showing changes during LTP (Voronin and Kuhnt1990) and LTD (Bolshakov and Sieglebaum 1994), others donot (Hjelmstad et al. 1997; Manabe et al. 1993). Interestingly,Bertram et al. (1996) recently supplied support for an alterna-tive model of PPF that, by depending on Ca21 binding domainsdirectly on release apparatus proteins, could allow for LTP andLTD in the presynaptic terminal that would not alter PPF. Ourdata may point to presynaptic targets insensitive to PPF, ordamage the presynaptic hypothesis. Other methods of evaluat-ing transmitter release will be needed to resolve this question.

The actions of zaprinast appear somewhat different fromthose of the NO donor,S-nitroso-N-acetyl penicillamine(SNAP). We previously have observed that combining sub-maximal LFS with SNAP elicits a robust, stable LTD (Gage etal. 1997; Reyes-Harde et al. 1999b). In the present study,zaprinast plus submaximal LFS produced only reversible de-pression, implying that SNAP, and, hence, NO, possess addi-tional properties that zaprinast does not. Although nonspecificactions of SNAP are possible, the concentration we used (100mM) is below those reported to act directly on sulfhydrylmoieties. An alternative explanation could involve differencesin subcellular localization of PDE, PKA, and PKG. De Venteet al. (1996) have shown that both the magnitude and local-ization of NO-mediated [cGMP] accumulation in hippocampusis influenced by isozyme-specific PDE inhibition. Kinase ac-tivity can be restricted spatially, at least for PKA (for review,see Coghlan et al. 1993) and also possibly for PKG (Vo et al.1998). Shakur et al. (1993, 1995) have shown membranecompartmentalization of a specific PDE4 isoform, while Wha-lin et al. (1988a,b) have shown the same for cyclic GMP-stimulated PDE2. In pyramidal cells, it is possible that inhib-iting PDE V might cause local increases in [cGMP] thatactivate only a subset of the substrates necessary for LTD. Incontrast, NO-activated GC may either activate a different pool

of cyclic GMP or PKG-dependent events and/or stimulatePDE2-mediated hydrolysis of cAMP.

Perhaps of greatest interest, our investigations have culminatedin the discovery of conditions that are both necessary and suffi-cient to elicit sustained depression of synaptic efficacy that com-pletely bypasses the requirement for electrical stimulation,NMDA, or GABA receptor activation. Interestingly, in the den-tate gyrus, Wu et al. (1998) recently have reported a long-lastingdepression of synaptic strength produced by zaprinast alone,which was, however, activity-dependent and required metabo-tropic glutamate receptor (mGluR) activation. Group II mGluRsare coupled negatively to adenylate cyclase, providing a potentialmeans of lowering [cAMP]. It is tempting to speculate that groupII mGluRs might effectively serve the same purpose for LTD inthe dentate as H-89 does in our chemical LTD in CA1.

The demonstration that saturating stimulus-induced LTD oc-cludes chemically induced LTD indicates that these two forms ofsynaptic depression either act via the same mechanisms or, atsome level, converge on shared pathways. Previous studies in ourlaboratory have shown that stimulus-induced LTD requires PKGactivity, and the demonstration here that the PKG inhibitorKT5823 blocked chemically induced LTD provides additionalsupport for a convergence. In earlier studies, we showed thatpresynaptic ryanodine-sensitive Ca21 stores (Reyes and Stanton1996) and activation of presynaptic Ca21/calmodulin-dependentprotein kinase II (Stanton and Gage 1996) are both needed toinduce LTD. We recently have found that antagonists of thecGMP-stimulated messenger cyclic ADP ribose, which releasescalcium from stores in sea urchin eggs (Galione et al. 1991), alsocan block the induction of LTD (Reyes-Harde et al. 1999a,b). Therole of this cascade in chemical LTD, and how it modulatesglutamate release, remain to be determined.

In contrast to chemical LTD, the initial reversible depressioninduced by zaprinast was not blocked by KT5823. Boulton etal. (1994) reported this action of zaprinast to be presynaptic,suggesting there must be a cyclic GMP-mediated mechanism

FIG. 9. Potential presynaptic mecha-nisms by which chemical LTD could modu-late glutamate release. A1, type 1 adenosinereceptors; CAMKII, calcium/calmodulin de-pendent protein kinase; cADP ribose, cylicADP ribose; cGMP, 39,59-cyclic GMP;CNG, cyclic nucleotide gated channel; H-89,N-[2-(p-Bromocinnamylamino)ethyl]-5-iso-quinolinesulfonamide 2HCL.


for depressing transmitter release that is PKG-independent.Based on the literature, we propose the following. Elevation of[cGMP] activates cyclic GMP-stimulated (type II) PDE, whichhydrolyzes cyclic AMP and reduces a tonic cyclic AMP-mediated enhancement of glutamate release. Studies support-ing this include demonstrations by Doerner and Alger (1988)that cyclic GMP can depress hippocampal Ca21 currentsthrough a PKG-independent mechanism and by Broome et al.(1994), who found that activation of presynaptic A1 adenosinereceptors, which tonically reduce [cyclic AMP] and suppressglutamate release, is necessary for zaprinast’s actions. Anotherpossible target is a cyclic nucleotide-gated cation channel inhippocampus whose tonic activation requires cyclic AMP(Bradley et al. 1997). Whether reversible cyclic GMP-medi-ated depression is distinct from LTD, or PKG just adds phos-phorylation events that increase the duration of this existingmechanism, is unknown.

There are multiple mechanisms whereby PKG and PKAmight affect glutamate release (see Fig. 9). Both kinases canphosphorylate presynaptic channels known to alter transmitterrelease. PKA can enhance glutamate release by phosphorylat-ing either presynaptic Ca21 channels (Hell et al. 1995) orpresynaptic kainate receptors (GluR6) (Chittajallu et al. 1996).Conversely, PKG can directly phosphorylate and open K1

channels in hippocampal neurons (Furukawa et al. 1996) andindirectly activate protein phosphatase 2A, which opens K1

channels in pituitary tumor cells (White et al. 1993). Simulta-neously reducing Ca21 conductance through PKA inhibitionand activating K1 channels via PKG could elicit markedreductions in transmitter release.

Phosphorylation of synaptic vesicle proteins provide anotherway kinases might influence transmitter release. An attractive,but untested, hypothesis is that PKG may phosphorylate anddownregulate a protein vital to synaptic vesicle mobilization(Sistiaga et al. 1997). Wang and Robinson (1995) have shownthe existence of.40 PKG substrate proteins that remain un-characterized and of unknown function. Several proteins in-volved in vesicle transport have been identified as potentialpresynaptic targets for PKA phosphorylation in vitro. Theseinclude synapsin1,}-SNAP (} soluble NSF attachment pro-tein), and rabphilin 3A. Injection of dephosphorylated synap-sin1 into squid axons inhibits transmitter release (Llinas et al.1985), but phosphorylated synapsin has no converse effect, andsynapsin1 knockout mice show no deficits in LTP (Spillane etal. 1995). Phosphorylation of rabphilin 3A is required forinteraction with another protein, rab3A, allowing its associa-tion with the fusion apparatus and recruitment of vesicles forexocytosis (Fykse et al. 1995). Interestingly, in rab3A knock-out mice, all electrophysiological parameters are normal inarea CA1 except for an increase in synaptic depression evokedby short stimuli (Geppert et al. 1994). In vivo phosphorylationof these proteins has yet to be shown, leaving the physiologicalrelevance of in vitro studies uncertain.

Recent studies (Kameyama et al. 1998; Lee et al. 1998) de-scribe another chemical means for inducing LTD that used a brief(3 min) bath application of NMDA to hippocampal slices. It wassuggested that this form of LTD is critically dependent on selec-tive dephosphorylation of postsynaptic GluR1 AMPA receptorsubunits at a PKA-sensitive site and shares mechanisms withstimulus-induced LTD. It is becoming clear that there are multiplemechanistically independent forms of LTD (Bolshakov and

Siegelbaum 1994; Gage et al. 1997; Oliet et al. 1997). Ourchemical LTD is activity-independent, bypasses the need forNMDA-receptor activation, and appears to depend on presynapticPKG activation and PKA inhibition. Therefore, we hypothesizethat our chemical LTD activates a presynaptic, cyclic-GMP- andPKG-dependent form of LTD, whereas that described by Lee etal. (1998) is a postsynaptic, dephosphorylation-mediated form.

In summary, our data support the hypothesis that in theinduction of homosynaptic LTD, cyclic GMP, acting via PKG,is involved in promoting cellular events that are antagonistic tothose mediated by activating PKA. The notion that these twocyclic nucleotide second-messenger systems mediate opposingbiological signals has been demonstrated in many other tissues.The idea that a specific form of synaptic plasticity depends onthe bidirectional control of kinase activity presents an attractive“push-pull” model for the amplification of signal transductioncascades. In this regard, chemically induced LTD should proveuseful for the study of the biochemical cascades underlyingstimulus-induced LTD.

We thank A. Kyrozis, T. Opitz, A. Peinado, and S. Siegelbaum for helpfuldiscussions.

This work was supported by Whitehall Foundation Grant A98-32 to P. K.Stanton and National Institutes of Health Medical Scientist Training GrantF31GM-16379 to M. Reyes-Harde.

Address for reprint requests: P. K. Stanton, Dept. of Neuroscience, AlbertEinstein College of Medicine, 1300 Morris Park Ave., Bronx, NY 10461-1602.

Received 11 January 1999; accepted in final form 19 April 1999.

REFERENCES

ABEL, T., NGUYEN, P. V., BARAD, M., DEUEL, T. A., KANDEL, E. R., AND

BOURTCHOULADZE, R. Genetic demonstration of a role for PKA in the latephase of LTP and in hippocampus-based long-term memory.Cell 88:615–626, 1997.

BEAR, M. F. AND ABRAHAM, W. C. Long-term depression in hippocampus.Annu. Rev. Neurosci.19: 437–462, 1996.

BEAVO, J. A. Cyclic nucleotide phosphodiesterases: functional implications ofmultiple isoforms.Physiol. Rev.75: 725–748, 1995.

BERTRAM, R., SHERMAN, A., AND STANLEY, E. F. Single-domain/bound calciumhypothesis of transmitter release and facilitation.J. Neurophysiol.75: 1919–1931, 1996.

BIENENSTOCK, E., COOPER, L. N., AND MUNRO, P. Theory for the developmentof neuron selectivity: orientation specificity and binocular interaction invisual cortex.J. Neurosci.2: 32–48, 1982.

BLITZER, R. D., WONG, T., NOURANIFAR, R., IYENGAR, R., AND LANDAU, E. M.Postsynaptic cAMP pathway gates early LTP in hippocampal CA1 region.Neuron15: 1403–1414, 1995.

BOLSHAKOV, V. Y. AND SIEGELBAUM, S. A. Postsynaptic induction and pre-synaptic expression of hippocampal long-term depression.Science264:1148–1152, 1994.

BOULTON, C. L., IRVING, A. J., SOUTHAM, E., POTIER, B., GARTHWAITE, J.,AND

COLLINGRIDGE, G. L. The nitric oxide-cyclic GMP pathway and synapticdepression in rat hippocampal slices.Eur. J. Neurosci.6: 1528–1535, 1994.

BRADLEY, J., ZHANG, Y., BAKIN , R., LESTER, H. A., RONNETT, G. V., AND ZINN,K. Functional expression of the heteromeric “olfactory” cyclic nucleotide-gated channel in the hippocampus: a potential effector of synaptic plasticityin brain neurons.J. Neurosci.17: 1993–2005, 1997.

BRANDON, E. P., ZHUO, M., HUANG, Y. Y., QI, M., GERHOLD, K. A., BURTON,K. A., KANDEL, E. R., MCKNIGHT, G. S.,AND IDZERDA, R. L. Hippocampallong-term depression and depotentiation are defective in mice carrying atargeted disruption of the gene encoding the RI beta subunit of cAMP-dependent protein kinase.Proc. Natl. Acad. Sci. USA92: 8851–8855, 1995.

BROOME, M. R., COLLINGRIDGE, G. L., AND IRVING, A. J. Activation of theNO-cGMP signaling pathway depresses hippocampal synaptic transmissionthrough an adenosine receptor-dependent mechanism.Neuropharmacology33: 1511–1513, 1994.


CHAVEZ-NORIEGA, L. E. AND STEVENS, C. F. Increased transmitter release atexcitatory synapses produced by direct activation of adenylate cyclase in rathippocampal slices.J. Neurosci.14: 310–317, 1994.

CHIJIWA, T., MISHIMA, A., HAGIWARA, M., SANO, M., HAYASHI, K., INOUE, T.,NAITO, K., TOSHIOKA, T., AND HIDAKA , H. Inhibition of forskolin-inducedneurite outgrowth and protein phosphorylation by a newly synthesizedselective inhibitor of cyclic AMP-dependent protein kinase,N-[2-(p-bromo-cinnamylamino)ethyl]-5-isoquinolinesulfonamide (H-89), of PC12D pheo-chromocytoma cells.J. Biol. Chem.265: 5267–5272, 1990.

CHITTAJALLU , R,, VIGNESM M,, KUMLESHM K., BARNES, J. M., COLLINGRIDGE,G. L., AND HENLEY, J. M. Regulation of glutamate by presynaptic kainatereceptors in the hippocampus.Nature379: 78–81, 1996.

COGHLAN, V. M., BERGESON, S. E., LANGEBERG, L., NILAVER, G., AND SCOTT,J. D. A-kinase anchoring proteins: a key to selective activation of cAMP-responsive events?Mol. Cell Biochem.127–128: 309–319, 1993, 1993.

CHRISTIE, B. R. (Editor). Hippocampus forum: long-term depression.Hip-pocampus6: 1–95, 1996.

DE VENTE, J., HOPKINS, D. A., MARKERINK-VAN ITTERSUM, M., AND STEIN-BUSCH, H. W. Effects of the 39,59-phosphodiesterase inhibitors isobutyl-methylxanthine and zaprinast on NO-mediated cGMP accumulation in thehippocampus slice preparation: an immunocytochemical study.J. Chem.Neuroanat.10: 241–248, 1996.

DOERNER, D. AND ALGER, B. E. Cyclic GMP depresses hippocampal Ca21

current through a mechanism independent of cGMP-dependent proteinkinase.Neuron1: 693–699, 1988.

DUDEK, S. M. AND BEAR, M. F. Homosynaptic long-term depression in areaCA1 of the hippocampus and effects of N-methyl-D-aspartate receptorblockade.Proc. Natl. Acad. Sci. USA89: 4363–4367, 1992.

FREY, U., HUANG, Y. Y., AND KANDEL, E. R. Effects of cAMP simulate a latestage of LTP in hippocampal CA1 neurons.Science260: 1661–1664, 1993.

FYKSE, E. M., LI, C., AND SUDHOF, T. C. Phosphorylation of rabphilin-3A byCa21/calmodulin- and cAMP-dependent protein kinases in vitro.J. Neuro-sci. 15: 2385–2395, 1995.

FURUKAWA, K., BARGER, S. W., BLALOCK, E. M., AND MATTSON, M. P.Activation of K1 channels and suppression of neuronal activity by secretedb-amyloid-precursor protein.Nature379: 74–78, 1996.

GAGE, A. T., REYES, M., AND STANTON, P. K. Nitric oxide-guanylyl-cyclase-dependent and -independent components of multiple forms of long-termsynaptic depression.Hippocampus7: 286–295, 1997.

GALIONE, A., LEE, H. C., AND BUSA, W. B. Ca21-induced Ca21 release in seaurchin egg homogenates: modulation by cyclic ADPribose.Science253:1143–1146, 1991.

GEPPERT, M., BOLSHAKOV, V. Y., SIEGELBAUM, S. A., TAKEI, K., DE CAMILLI ,P., HAMMER, R. E., AND SUDHOF, T. C. The role of Rab3A in neurotrans-mitter release.Nature369: 493–497, 1994.

GILLESPIE, P. G.AND BEAVO, J. A. Inhibition and stimulation of photoreceptorphosphodiesterases by dipyridamole and M&B 22,948.Mol. Pharmacol.36:773–781, 1989.

HELL, W. J., YOKOYAMA , C. T., BREEZE, L. J., CHAVKIN , C., CATTERALL, W. A.Phosphorylation of presynaptic and postsynaptic calcium channels bycAMP-dependent protein kinase in hippocampal neurons.EMBO J. 14:3036–3044, 1995.

HJELMSTAD, G. O., NICOLL, R. A., AND MALENKA , R. C. Synaptic refractoryperiod provides a measure of probability of release in the hippocampus.Neuron19: 1309–1318, 1997.

HOPKINS, W. F. AND JOHNSTON, D. Noradrenergic enhancement of long-termpotentiation at mossy fiber synapses in the hippocampus.J. Neurophysiol.59: 667–687, 1988.

HUANG, Y. Y., LIM, X. C.,AND KANDEL, E. R. cAMP contributes to mossy fiberLTP by initiating both a covalently mediated early phase and macromolec-ular synthesis-dependent late phase.Cell 79: 69–79, 1994.

INOUE, A., HIRUMA, Y., HIROSE, S., YAMAGUCHI, A., AND HAGIWARA, H.Reciprocal regulation by cyclic nucleotides of the differentiation of ratosteoblast-like cells and mineralization of nodules.Biochem. Biophys. Res.Commun.215: 1104–1110, 1995.

IZUMI, Y. AND ZORUMSKI, C. F. Nitric oxide and long-term synaptic depressionin the rat hippocampus.Neuroreport4: 1131–1134, 1993.

KAMEYAMA , K., LEE, H. K., BEAR M. F., AND HUGANIR R. L. Involvement ofa postsynaptic protein kinase A substrate in the expression of homosynapticlong-term depression.Neuron21: 1163–1175, 1998.

KASE, H., IWAHASHI, K., NAKANISHI , S., MATSUDA, Y., YAMADA , K., TAKA -HASHI, M., MURAKATA , C., SATO, A., AND KANEKO, M. K-252 compounds,novel and potent inhibitors of protein kinase C and cyclic nucleotide-

dependent protein kinases.Biochem. Biophys. Res. Commun.142: 436–440,1987.

KATZ, B. AND MILEDI, R. The role of calcium in neuromuscular facilitation.J. Physiol. (Lond.)195: 481–492, 1968.

KLESCHEVNIKOV, A. M., SOKOLOV, M. V., KUHNT, U., DAWE, G. S., STEPHEN-SON, J. D.,AND VORONIN, L. L. Changes in paired-pulse facilitation correlatewith induction of long-term potentiation in area CA1 of rat hippocampalslices.Neuroscience76: 829–843, 1997.

LEE, H. K., KAMEYAMA , K., HUGANIR, R. L., AND BEAR, M. F. NMDA induceslong-term synaptic depression and dephosphorylation of the GluR1 subunitof AMPA receptors in hippocampus.Neuron21: 1151–1162, 1998.

LLINAS, R., MCGUINNESS, T. L., LEONARD, C. S., SUGIMORI, M., AND GREEN-GARD, P. Intraterminal injection of synapsin I or calcium/calmodulin-depen-dent protein kinase II alters neurotransmitter release at the squid giantsynapse.Proc. Natl. Acad. Sci. USA82: 3035–3039, 1985.

MADISON, D. V. AND NICOLL, R. A. Cyclic adenosine 39,59-monophosphatemediates beta-receptor actions of noradrenaline in rat hippocampal pyrami-dal cells.J. Physiol. (Lond.)372: 245–259, 1986.

MANABE, T., WYLLIE , D.J.A., PERKEL, D. J.,AND NICOLL, R. A. Modulation ofsynaptic transmission and long-term potentiation: effects of paired-pulsefacilitation and EPSC variance in the CA1 region of the hippocampus.J. Neurophysiol.70: 1451–1459, 1993.

MATTHIES, H. AND REYMANN, K. G. Protein kinase A inhibitors prevent themaintenance of hippocampal long-term potentiation.Neuroreport4: 712–714, 1993.

MULKEY, R. M., HERRON, C. E., AND MALENKA , R. C. An essential role forprotein phosphatases in hippocampal long-term depression.Science261:1051–1055, 1993.

MULLER, D. AND LYNCH, G. Evidence that changes in presynaptic currents arenot responsible for long-term potentiation in hippocampus.Brain Res.479:290–299, 1989.

OLIET S. H., MALENKA R. C., AND NICOLL R. A. Two distinct forms oflong-term depression coexist in CA1 hippocampal pyramidal cells.Neuron18: 969–982, 1997.

OTANI, S.AND CONNOR, J. A. Long-term depression of naive synapses in adulthippocampus induced by asynchronous synaptic activity.J. Neurophysiol.73: 2596–2601, 1995.

POLANOWSKA-GRABOWSKA, R. AND GEAR, A. R. Role of cyclic nucleotides inrapid platelet adhesion to collagen.Blood 83: 2508–2515, 1994.

REYES, M. AND STANTON, P. K. Induction of hippocampal long-term depressionrequires release of Ca21 from separate presynaptic and postsynaptic intra-cellular stores.J. Neurosci.16: 5951–5960, 1996.

REYES, M. AND STANTON, P. K. Induction of LTD via a presynaptic nitricoxide-mediated cascade requires release of calcium from ryanodine-sensi-tive stores.Soc Neurosci Abstr.23: 1175, 1997.

REYES-HARDE, M., EMPSON, F., POTTER, B.V.L., GALIONE, A., AND STANTON,P. K. Evidence of a role for cyclic ADP-ribose in long-term synapticdepression in hippocampus.Proc. Natl. Acad. Sci. USA96: 4061–4066,1999a.

REYES-HARDE, M., POTTER, B.V.L., GALIONE, A., AND STANTON, P. K. Induc-tion of hippocampal LTD requires nitric-oxide-stimulated PKG activity andCa21 release from cyclic ADP-ribose-sensitive stores.J. Neurophysiol. 82:1569–1576, 1999b.

SEJNOWSKI, T. Storing covariance with nonlinearly interacting neurons.J.Math. Biol. 4: 303–321, 1977.

SHAKUR, Y., PRYDE, J. G.,AND HOUSLAY, M. D. Engineered deletion of theunique N-terminal domain of the cyclic AMP-specific phosphodiesteraseRD1 prevents plasma membrane association and the attainment of enhancedthermostability without altering its sensitivity to inhibition by rolipram.Biochem. J.292: 677–686, 1993.

SHAKUR, Y., WILSON, M., POOLEY, L., LOBBAN, M., GRIFFITHS, S. L., CAMP-BELL, A. M., BEATTIE, J., DALY , C., AND HOUSLAY, M. D. Identification andcharacterization of the type-IVA cyclic AMP-specific phosphodiesteraseRD1 as a membrane-bound protein expressed in cerebellum.Biochem. J.306: 801–809, 1995.

SISTIAGA, A., MIRAS-PORTUGAL, M. T., AND SANCHEZ-PRIETO, J. Modulation ofglutamate release by a nitric oxide/cyclic GMP-dependent pathway.Eur.J. Pharmacol.321: 247–257, 1997.

SPILLANE, D. M., ROSAHL, T. W., SUDHOF, T. C., AND MALENKA , R. C.Long-term potentiation in mice lacking synapsins.Neuropharmacology34:1573–1579, 1995.

STANTON, P. K. AND GAGE, A. T. Distinct synaptic loci of Ca21/calmodulin-dependent protein kinase II necessary for long-term potentiation and depres-sion.J. Neurophysiol.76: 2097–2101, 1996.


STANTON, P. K., MODY, I., AND HEINEMANN, U. A role forN-methyl-D-aspartatereceptors in norepinephrine-induced long-lasting potentiation in the dentategyrus.Exp. Brain Res.77: 517–530, 1989.

STANTON, P. K. AND SARVEY, J. M. Depletion of norepinephrine, but notserotonin, reduces long-term potentiation in the dentate gyrus of rat hip-pocampal slices.J. Neurosci.5: 2169–2176, 1985a.

STANTON, P. K. AND SARVEY, J. M. Blockade of norepinephrine-inducedlong-lasting potentiation in the hippocampal dentate gyrus by an inhibitor ofprotein synthesis.Brain Res.361: 276–283, 1985b.

STANTON, P. K. AND SEJNOWSKI, T. J. Associative long-term depression in thehippocampus induced by hebbian covariance.Nature339: 215–218, 1989.

TRUDEAU, L. E., EMERY, D. G., HAYDON, P. G. Direct modulation of thesecretory machinery underlies PKA-dependent synaptic facilitation in hip-pocampal neurons.Neuron17: 789–797, 1996.

VO, N. K., GETTEMY J. M., AND COGHLAN, V. M. Identification of cGMP-dependent protein kinase anchoring proteins.Biochem. Biophys. Res. Com-mun.246: 831–835, 1998.

VORONIN, L. L. AND KUHNT, U. Long-term potentiation affects facilitation ratioof Epsps recorded from CA1 pyramidal cells in the guinea pig hippocampalslice.Neurosci. Res. Commun.6: 149–155, 1990.

WAGNER, J. J.AND ALGER, B. E. GABAergic and developmental influences onhomosynaptic LTD and depotentiation in rat hippocampus.J. Neurosci.15:1577–1586, 1995.

WANG, X. AND ROBINSON, P. J. Cyclic GMP-dependent protein kinase sub-strates in rat brain.J. Neurochem.65: 595–604, 1995.

WEXLER, E., STANTON, P. K., AND NAWY, S. Nitric oxide depresses GABAA

receptor function via coactivation of cGMP-dependent kinase and phospho-diesterase.J. Neurosci.18: 2342–2349, 1998.

WHALIN , M. E., GARRETT, R. L., JR., THOMPSON, W. J., AND STRADA, S. J.Correlation of cell-free brain cyclic nucleotide phosphodiesterase activitiesto cyclic AMP decay in intact brain slices.Second Messengers Phospho-proteins12: 311–325, 1988a.

WHALIN , M. E., STRADA, S. J.,AND THOMPSON, W. J Purification and partialcharacterization of membrane-associated type II (cGMP-activatable) cyclicnucleotide phosphodiesterase from rabbit brain.Biochim. Biophys. Acta972:79–94, 1988b.

WHITE, R. E., LEE, A. B., SHCHERBATKO, A. D., LINCOLN, T. M., SCHONBRUNN,A., AND ARMSTRONG, D. L. Potassium channel stimulation by natriureticpeptides through cGMP-dependent dephosphorylation.Nature 361: 263–266, 1993.

WU, J., WANG, Y., ROWAN, M. J,AND ANWYL, R. Evidence for involvement ofthe cGMP-protein kinase G signaling system in the induction of long-termdepression, but not long-term potentiation, in the dentate gyrus in vitro.J. Neurosci.18: 3589–3596, 1998.

ZUCKER, R. S. Short-term synaptic plasticity.Annu. Rev. Neurosci.12: 13–31,1989.


Induction of hippocampal LTD requires nitric-oxide-stimulated PKG activity and Ca2+ release from...

Documents

Transcript of Induction of hippocampal LTD requires nitric-oxide-stimulated PKG activity and Ca2+ release from...