The Journal of Neuroscience, September 15, 1998, 18(18):71187126
Postsynaptic Complex Spike Bursting Enables the Induction of LTP by Theta Frequency Synaptic Stimulation
Mark J. Thomas,1 Ayako M. Watabe,2 Teena D. Moody,1 Michael Makhinson,1 and Thomas J. ODell2 Interdepartmental PhD Program for Neuroscience and 2Department of Physiology, School of Medicine, University of California, Los Angeles, Los Angeles, California 90095
Long-term potentiation (LTP), a persistent enhancement of synaptic transmission that may be involved in some forms of learning and memory, is induced at excitatory synapses in the CA1 region of the hippocampus by coincident presynaptic and postsynaptic activity. Although action potentials backpropagating into dendrites of hippocampal pyramidal cells provide sufcient postsynaptic activity to induce LTP under some in vitro conditions, it is not known whether LTP can be induced by patterns of postsynaptic action potential ring that occur in these cells in vivo. Here we report that a characteristic in vivo pattern of action potential generation in CA1 pyramidal cells At many excitatory synapses, coincident activity in the presynaptic terminal and postsynaptic cell induces long-term potentiation (LTP), a persistent enhancement of synaptic transmission thought to have a role in certain forms of learning and memory (Bliss and Collingridge, 1993). The signicance of coincident presynaptic and postsynaptic activity in the induction of LTP lies in the fact that it provides the simultaneous release of glutamate and postsynaptic membrane depolarization necessary for activation of NMDA-type glutamate receptors. C alcium inux through NMDA receptor ion channels in turn triggers a complex, protein kinase-dependent signaling pathway ultimately responsible for the modications that enhance synaptic transmission (Bliss and Collingridge, 1993). Until recently, the strong postsynaptic depolarization required for NMDA receptor activation and LTP induction was thought to arise primarily from the temporal and spatial summation of EPSPs during high-frequency stimulation of multiple presynaptic bers (Nicoll et al., 1988; Gustafsson and Wigstrom, 1990; De banne et al., 1996). However, recent ndings showing that action potentials initiated near the cell body back-propagate into pyramidal cell dendrites (Magee and Johnston, 1995; Spruston et al., 1995) suggest that dendritic action potentials may provide an additional means of achieving the postsynaptic depolarization needed to activate NMDA receptors and induce LTP. Indeed, EPSPs paired with back-propagating dendritic action potentials can induce LTP under some in vitro conditions (Scharfman and
Received April 3, 1998; revised June 18, 1998; accepted June 26, 1998. This work was supported by grants from the National Institute of Mental Health, the K lingenstein Fund, and the Pew Charitable Trusts to T.J.O. T.J.O. is a member of the University of C alifornia Los Angeles Brain Research Institute. We are gratef ul to D. V. Buonomano, D. Glanzman, and F. Krasne for comments on an earlier version of this manuscript. M.J.T. and A.M.W. made equal contributions to this work. Correspondence should be addressed to Dr. Thomas ODell, Department of Physiology, University of C alifornia Los Angeles School of Medicine, 53-231 Center for the Health Sciences, 10833 Le Conte Avenue, Los Angeles, CA 90095. Copyright 1998 Society for Neuroscience 0270-6474/98/187118-09$05.00/0
known as the complex spike burst enables the induction of LTP during theta frequency synaptic stimulation in the CA1 region of hippocampal slices maintained in vitro. Our results suggest that complex spike bursting may have an important role in synaptic processes involved in learning and memory formation, perhaps by producing a highly sensitive postsynaptic state during which even low frequencies of presynaptic activity can induce LTP. Key words: long-term potentiation; complex spike burst; hippocampus; pyramidal cells; synaptic transmission; learning and memory Sarvey, 1985; Magee and Johnston, 1997; Markram et al., 1997). Although these ndings suggest that dendritic action potentials may have an important role in the induction of LTP, it is not yet known whether patterns of action potential ring observed in CA1 pyramidal cells in vivo are sufcient and/or necessary for LTP induction. In vivo, hippocampal CA1 pyramidal cells generate action potentials either as single, isolated spikes or in high-frequency bursts of two or more action potentials that progressively decline in amplitude and increase in duration during the burst (Kandel and Spencer, 1961; Ranck, 1973; Fox and Ranck, 1975; Suzuki and Smith, 1985). This second mode of ring, known as the complex spike burst, is a dening electrophysiological signature of hippocampal pyramidal cells (Ranck, 1973; Fox and Ranck, 1975) and may represent an important form of information coding in the hippocampus (Lisman, 1997). Complex spike bursts may also have an important role in hippocampal synaptic plasticity because patterns of presynaptic ber stimulation that mimic complex spike bursting can, under certain conditions, induce LTP (Larson et al., 1986; Huerta and Lisman, 1993, 1995; Holscher et al., 1997). Because action potentials during complex spike bursts backpropagate into CA1 pyramidal cell dendrites in vivo (Buzaki et al., 1996), complex spike bursts might also have an important postsynaptic role in the induction of LTP by providing the postsynaptic activity needed for NMDA receptor activation. Thus, to determine whether EPSPs paired with patterns of postsynaptic spiking that occur in CA1 pyramidal cells in vivo can induce LTP, we investigated the postsynaptic role of both single action potentials and complex spike bursts in the induction of LTP during theta frequency synaptic stimulation.

MATERIALS AND METHODS

Transverse hippocampal slices, 400 m thick, were obtained from halothane-anesthetized male mice (C57BL /6) using standard techniques. Unless noted otherwise, all experiments were done using animals between 4 and 7 weeks of age. None of our ndings varied with animal age.
Thomas et al. Postsynaptic Complex Spike Bursting Enables LTP Induction
J. Neurosci., September 15, 1998, 18(18):71187126 7119
Although 5 Hz synaptic stimulation can induce long-term depression in slices from very young animals (Bolshakov and Siegelbaum, 1994; Oliet et al., 1997), it does not induce LTD in the CA1 region of hippocampal slices obtained from mice in the age range used in our experiments (Mayford et al., 1995; Thomas et al., 1996). Slices were maintained in an interface recording chamber (Fine Science Tools, Inc.) and perf used with an articial mouse C SF (AC SF) consisting of (in mM): 124 NaC l, 4.4 KC l, 25 NaHC O3 , 1 NaH2PO4 , 1.2 MgSO4 , 2 C aC l2 , and 10 glucose (gassed with 95% O2 and 5% C O2 ; temperature, 30C). Schaffer collateral and commissural bers were stimulated at 0.02 Hz with a bipolar nichrome wire stimulating electrode (0.02 msec duration pulses), and EPSPs evoked in CA1 pyramidal cells were recorded extracellularly in stratum radiatum with an AC SF-lled glass microelectrode (510 M). Glass microelectrodes (M) lled with 3 M K-acetate or 2 M K-methylsulfate were used in intracellular recordings. Only cells with resting membrane potentials more negative than 55 mV and with overshooting action potential amplitudes were used. Stimulation intensity was adjusted to evoke eld EPSPs (fEPSPs) that were 50% of the maximal fEPSP amplitude (strong intensity stimulation) or 20 25% of the maximal response (weak intensity stimulation). In two pathway experiments, independence of the bers activated by two stimulating electrodes was conrmed by the lack of paired-pulse facilitation when one pathway was stimulated 50 msec after the other. CA1 pyramidal cells were antidromically activated using a stimulating electrode placed in the alveus and stimulation intensities sufcient to evoke near-maximal antidromic population spikes (recorded with an extracellular electrode in stratum radiatum). Slices showing any evidence of contamination of the antidromic response by activation of bers in stratum oriens were not used. All stimulation protocols as well as data acquisition and analysis were performed using E xperimenters Workbench (Data Wave Technologies Inc.). Negative-going spikes appearing in fEPSP recordings were manually counted off-line by visually inspecting each trace recorded during 5 Hz stimulation. This analysis was done in a blind manner. All values are reported as mean SEM, and Students t tests were used to assess statistical signicance. Salts used in the AC SF were purchased from Sigma (St. L ouis, MO). All other compounds were purchased from Research Biochemicals (Natick, M A). Concentrated stock solutions of nifedipine and nimodipine (in dimethylsulfoxide) were prepared fresh daily under dim light, and experiments were done in the dark to minimize light exposure.

voltage-sensitive Na channel-dependent manner. The interspike intervals (57 msec), progressive decrease in spike amplitude, and progressive increase in spike duration during these 5 Hz stimulation-induced bursts are characteristic electrophysiological features of complex spike bursts recorded in vivo (Kandel and Spencer, 1961; Ranck, 1973; Fox and Ranck, 1975). In agreement with previous reports indicating that a highly TTX-sensitive, persistent Na conductance has a crucial role in burst generation in CA1 pyramidal cells (Azouz et al., 1996; Jensen et al., 1996), we found that a brief (10 min) application of TTX (250 nM) had little effect on excitatory synaptic transmission (see below) or on the generation of single action potentials but produced a near-complete suppression of complex spike bursting (Fig. 1C). During intracellular recordings from CA1 pyramidal cells we observed that when 150 pulses of 5 Hz stimulation were delivered in the absence of TTX, EPSPs elicited complex spike bursting (dened as two or more action potentials with interspike intervals of 10 msec) on 61 6% of stimulation pulses (n 9 cells). In contrast, although EPSPs reliably evoked single action potentials during 150 pulses of 5 Hz stimulation in the presence of 250 nM TTX (10 min application), complex spike bursts were observed on only 3 2% of the stimulation pulses (n 9 cells).
Five hertz synaptic stimulation induces an associative, NMDA receptor-dependent form of LTP
If complex spike bursts provide a sufcient level of dendritic depolarization to relieve the voltage-dependent Mg 2 ion block of the NMDA receptor, then the EPSP-evoked bursts that occur during 5 Hz stimulation should induce LTP. As shown in Figure 2 A, 30 sec of 5 Hz stimulation induced a persistent potentiation of synaptic transmission that was signicantly reduced when 5 Hz stimulation was delivered in the presence of the NMDA receptor blocker 2-amino-5-phosphonovaleric acid (100 M DL-APV). As is the case for certain high-frequency stimulation protocols (Grover and Teyler, 1990; Johnston et al., 1992; Ito et al., 1995), 5 Hz stimulation appears to induce LTP through both NMDA receptor-dependent and -independent signaling pathways because a small, but signicant ( p 0.01), potentiation was still evident 45 min after 5 Hz stimulation in APV. In experiments in which we doubled the effective concentration of APV by using 100 M D-APV a similar APV-resistant potentiation was also induced (fEPSPs were potentiated to 117.8 2.5% of baseline 45 min after 5 Hz stimulation; n 4), suggesting that the residual potentiation induced in the presence of APV is not attributable to incomplete NMDA receptor blockade. NMDA receptorindependent LTP of excitatory synaptic transmission in the CA1 region of the hippocampus is thought to be caused by Ca 2 inux through voltage-sensitive Ca 2 channels, because inhibitors of L-type (Grover and Teyler, 1990) or T-type (Ito et al., 1995, Magee and Johnston, 1997) Ca 2 channels can block this component of high-frequency stimulation-induced LTP. We found that 5 Hz stimulation-induced LTP was not inhibited by the L-type Ca 2 channel antagonists nifedipine and nimodipine [fEPSPs were potentiated to 151.4 11.3% of baseline (n 5) 45 min after 5 Hz stimulation in 10 M nifedipine and to 153.2 10.2% of baseline (n 5) after 5 Hz stimulation in 20 M nimodipine]. However, the T-type Ca 2 channel blocker Ni 2 signicantly reduced the amount of potentiation induced by 5 Hz stimulation [fEPSPs were 123.6 4.4% of baseline 45 min after 5 Hz stimulation in 50 M NiCl2 (n 8) compared with 153.7 7.7% of baseline in paired control experiments (n 8); t(14) 3.38; p 0.005]. In the presence of both 50 M NiCl2 and 100 M

DL-APV, fEPSPs were 107.5 2.7% of baseline 45 min after 5 Hz stimulation (n 8) compared with 151.1 5.8% of baseline in paired control experiments (n 6). Thus, activation of both NMDA receptors and T-type C a 2 channels appears to contribute to the induction of LTP by 5 Hz stimulation. Nifedipine, nimodipine, Ni 2, and APV did not block 5 Hz stimulationinduced complex spike bursting, suggesting that activation of L -type and T-type C a 2 channels as well as NMDA receptors is not required for complex spike bursting. We f urther characterized the potentiation induced by 5 Hz stimulation by investigating the effects of presynaptic ber stimulation strength on the induction of LTP. Although strongintensity 5 Hz stimulation induced LTP, weak-intensity stimulation failed to induce both complex spike bursting and LTP (Fig.
2 B). LTP could be induced in a group of synapses activated by weak-intensity presynaptic stimulation, however, if these synapses were coactivated during 5 Hz stimulation with a strongly stimulated independent group of synaptic inputs that elicited complex spike bursting (Fig. 2 B). Thus, like high-frequency stimulationinduced LTP (Nicoll et al., 1988, Gustafsson and Wigstrom, 1990; Bliss and Collingridge, 1993), 5 Hz stimulation induces an associative form of LTP whereby independent groups of synapses can interact in a positive manner to induce LTP.
Complex spike bursting is required for the induction of LTP by 5 Hz stimulation
To determine whether complex spike bursts are required for the induction of LTP during 5 Hz stimulation, we examined whether
J. Neurosci., September 15, 1998, 18(18):71187126 7121
Figure 2. Complex spike bursts enable LTP induction during 5 Hz stimulation. A, A 30 sec train of 5 Hz stimulation was delivered (at time 0). In control experiments ( lled symbols) fEPSPs were potentiated to 148.6 5.3% of baseline (n 27) 45 min after 5 Hz stimulation, whereas fEPSPs were 117.3 4.8% of baseline after 5 Hz stimulation in 100 M DL-APV (open symbols; n 7; signicantly less than control, p 0.01). B, Weak-intensity stimulation of a single pathway (open symbols; n 9) failed to induce a persistent change in synaptic transmission. In contrast, LTP was induced when a pathway activated with weak-intensity stimulation ( lled symbols; n 5) was coactivated during 5 Hz stimulation with an independent pathway stimulated at a high intensity. For clarity the results from activation of the strong pathway are not shown (EPSPs in the pathway activated by strong-intensity 5 Hz stimulation were potentiated to 166.7 11.0% of baseline; n 5). C, A 10 min bath application of 250 nM TTX (indicated by the bar) blocks the induction of LTP by 5 Hz stimulation. Forty-ve minutes after 5 Hz stimulation in TTX ( lled circles), fEPSPs were 102.4 4.2% of baseline (n 7), whereas synaptic transmission was potentiated to 147.3 8.5% of baseline in control experiments (no TTX; open circles; n 10). Bath application of TTX alone had no effect on synaptic transmission (diamonds; n 5). Traces are fEPSPs recorded at the end of 5 Hz stimulation in control and TTX experiments. C alibration: 2 mV, 10 msec. D, TTX does not inhibit the induction of LTP by two trains of 100 Hz stimulation (1 sec duration; intertrain interval 10 sec, delivered at time 0). fEPSPs were potentiated to 187.0 12.8% of baseline 60 min after tetanus in control experiments (open symbols; n 7) and were potentiated to 188.3 9.1% of baseline after 100 Hz stimulation in 250 nM TTX (indicated by the bar, lled symbols; n 7).

pharmacologically suppressing complex spike bursting could prevent the induction of LTP by strong-intensity 5 Hz stimulation. Because T-type calcium channels appear to be involved in the induction of LTP by 5 Hz stimulation, we did not attempt to use intracellular electrodes containing the Na channel blocker QX314 in these experiments because this compound also inhibits T-type C a 2 channels (Talbot and Sayer, 1996). Instead, because our earlier experiments conrmed previous ndings that low concentrations of TTX inhibit bursting with little effect on the generation of single action potentials (Azouz et al., 1996), we examined whether LTP induction was blocked when complex spike bursting during 5 Hz stimulation was suppressed by TTX. Although a short application of 250 nM TTX had no effect on
baseline synaptic transmission, it strongly suppressed complex spike bursting during 5 Hz stimulation and blocked the induction of LTP (Fig. 2C). In contrast, 250 nM TTX had no effect on high-frequency stimulation-induced LTP (Fig. 2 D). Thus, TTX does not have generalized effects on synaptic transmission that prevent the induction of LTP, suggesting that the TTX block of 5 Hz stimulation-induced LTP is caused by the suppression of complex spike bursting. Moreover, because these concentrations of TTX have little effect on the ability of EPSPs to evoke single postsynaptic action potentials (Fig. 1C), the TTX block of 5 Hz stimulation-induced LTP suggests that bursts, rather than single action potentials, are specically required for LTP induction. We also examined the need for complex spike bursting in 5 Hz
7122 J. Neurosci., September 15, 1998, 18(18):71187126
Figure 3. A, Twenty-ve pulses of 5 Hz synaptic stimulation alone has little persistent effect on synaptic transmission (open symbols; fEPSPs were 109.7 2.8% of baseline 45 min after 5 Hz stimulation; n 22) but induces LTP when paired with complex spike burst-like antidromic stimulation ( lled symbols; fEPSPs were potentiated to 152.9 9.63% of baseline; n 14; signicantly greater than synaptic stimulation alone, p 0.005). Inset, Response during pairing of synaptic and antidromic burst stimulation (calibration: 5 mV, 5 msec). These experiments were done using slices from animals between 3 and 4 weeks of age (n 14 for 5 Hz stimulation alone; n 8 for 5 Hz stimulation paired with antidromic stimulation) and between 5 and 7 weeks of age (n 8 for 5 Hz stimulation alone; n 6 for 5 Hz stimulation paired with antidromic stimulation). The results from these two groups of animals were not different and have been combined. B, Complex spike burst-like antidromic stimulation induces LTP when paired with 50 pulses of 2.5 Hz synaptic stimulation ( lled circles; n 9), whereas pairing EPSPs with a single antidromic stimulation pulse has no lasting effect on synaptic strength (open circles; n 6; EPSPs were 103.2 4.4% of baseline). Fifty pulses of 2.5 Hz stimulation alone had no effect on synaptic strength ( lled triangles; n 8). C, Mean number of spikes in fEPSP recordings evoked by each of 75 pulses of 5 Hz stimulation during the continuous train (right; SEM values ranged from 0 to 0.46) and during ve trains of 15 pulses (lef t; intertrain interval, 7 sec; SEM values ranged from 0 to 0.41). D, Seventy-ve pulses of 5 Hz stimulation induced LTP when delivered as a continuous train (open symbols; fEPSPs were potentiated to 142.7 11.9% of baseline; n 8) but had little effect on synaptic transmission when delivered as ve trains of 15 pulses ( lled symbols; EPSPs were 110.72 5.2% of baseline; n 8; signicantly less than potentiation induced by continuous 75 pulse train, p 0.025). Data from the same experiments shown in C.

stimulation-induced LTP by determining whether LTP could be induced by pairing EPSPs with antidromically stimulated bursts of action potentials that mimicked complex spike bursting (three actions potentials at 200 Hz, beginning 7 msec after the start of the EPSP; Fig. 3A, inset). Although strong-intensity 5 Hz trains that terminated before complex spike bursting began (5 Hz for 5 sec) induced little persistent change in synaptic strength, signicant LTP was induced when this short train of 5 Hz stimulation was paired with antidromic stimulation that mimicked complex spike bursting (Fig. 3A). Antidromic stimulation alone had no effect on synaptic transmission (n 3; data not shown). It has been reported (Jester et al., 1995) that pairing antidromic action potentials with orthodromic synaptic stimulation fails to induce
LTP in the CA1 region of the hippocampus. There are numerous methodological differences between our experiments and those described in this report, especially with respect to the patterns and numbers of antidromic and orthodromic stimulation pulses used during pairing. Although we have not investigated which of these variables account for these different results, we believe it is signicant that in our experiments LTP was induced by stimulation protocols specically designed to mimic EPSP-evoked complex spike bursting. We also observed that LTP could be induced by pairing complex spike-like antidromic stimulation with another low-frequency synaptic stimulation protocol (50 pulses at 2.5 Hz) that alone failed to induce complex spike bursting and had no lasting effect on synaptic transmission (Fig. 3B). In these experi-
J. Neurosci., September 15, 1998, 18(18):71187126 7123
ments fEPSPs were 105.6% of baseline 45 min after 2.5 Hz synaptic stimulation alone (n 8) and 149.5 10.2% of baseline after pairing 2.5 Hz synaptic stimulation with antidromic bursts (n 9). Consistent with the results from our experiments with TTX that suggest that EPSPs paired with single postsynaptic action potentials are not sufcient for LTP induction, a single pulse of antidromic stimulation paired with 2.5 Hz synaptic stimulation (delivered 7 msec after the EPSP) failed to induce LTP (Fig. 3B). Finally, because EPSPs evoked complex spike bursts with a characteristic delay after the start of a 5 Hz stimulation train, we also investigated whether delivering 5 Hz stimulation as a continuous train that elicited complex spike bursting or as several short groups of stimulation pulses that were not long enough for complex spike bursting to begin could affect the induction of LTP. We found that although 75 pulses of 5 Hz stimulation delivered as a continuous train induced complex spike bursting and LTP, 75 stimulation pulses delivered as ve trains of 15 pulses (intertrain interval, 7 sec) failed to induce complex spike bursting and had little lasting effect on synaptic transmission (Fig. 3C,D).

High levels of complex spike bursting are required for LTP induction during long trains of 5 Hz stimulation
If the induction of LTP by low-frequency synaptic stimulation depends solely on EPSP and complex burst coincidence, then LTP should be reliably induced by all trains of stimulation that produce at least the minimal or threshold number of coincident EPSP and complex spike bursts needed to activate the NMDA receptor-dependent signaling pathways responsible for LTP. However, when we examined the effect of train duration on both complex spike bursting and LTP induction, the results clearly indicated that LTP was not simply induced after the occurrence of some threshold number of EPSP-evoked complex spike bursts. As shown in Figure 4 A, during a 3 min long train of 5 Hz stimulation (900 pulses), complex spike bursting was prominent during the rst 150 stimulation pulses and then gradually declined, terminating on average by the 300th stimulation pulse. When we examined the effects of different duration 5 Hz trains on the induction of LTP, we observed that when LTP was induced by 75 and 150 pulse trains of 5 Hz stimulation, shorter (25 pulses) or longer (300 and 900 pulses) trains had little persistent effect on synaptic strength (Fig. 4 B). These results indicate that short trains of 5 Hz stimulation (25 pulses) that terminate before complex spike bursting begins fail to induce LTP and that by 75 pulses of 5 Hz stimulation a sufcient number of EPSP-evoked complex spike bursts (27 5, mean SEM) have occurred to induce LTP. Yet, when 5 Hz stimulation was continued for longer periods (300 and 900 pulses) no LTP was induced, even though the total number of EPSP-evoked complex spike bursts was six to nine times greater during longer trains of 5 Hz stimulation than during the 75 pulse train (Fig. 4 B, inset). Thus, although EPSPevoked complex spike bursting can activate the NMDA receptordependent processes required for LTP induction, it appears that other cellular processes that inhibit LTP induction are also activated during longer trains of 5 Hz stimulation. Indeed, previous studies have shown that NMDA receptor activation can activate processes that increase the threshold for LTP induction and/or reverse previously established LTP (Fujii et al., 1991; Huang et al., 1992; ODell and Kandel, 1994). Based on previous observations that protein phosphatase inhibitors enhance the induction of LTP by long, but not short, trains of 5 Hz stimulation (Thomas et al., 1996) and prevent the reversal of previously established LTP by 5 Hz stimulation (ODell and Kandel, 1994), it seems likely that

protein phosphatase activation importantly contributes to the inhibition of LTP induction during long trains of 5 Hz stimulation. Modest levels of NMDA receptor activation during lowfrequency synaptic stimulation are thought to activate protein phosphatases by producing modest increases in intracellular calcium (Mulkey and Malenka, 1992; Lisman, 1994; Cummings et al., 1996). In contrast, more intense NMDA receptor activation is thought to elicit a larger increase in intracellular calcium that both activates the protein kinases directly responsible for inducing LTP (Bliss and Collingridge, 1993) and inhibits the activity of protein phosphatases that might oppose the induction of LTP (Lisman, 1994; Blitzer et al., 1995; Thomas et al., 1996). Because strong NMDA receptor activation can activate a signaling pathway that inhibits protein phosphatases, we investigated whether increasing levels of NMDA receptor activation by increasing the number of EPSPs that evoke complex spike bursts during long 5 Hz stimulation trains could enable the induction of LTP. Although we have not investigated the cellular mechanisms that regulate complex spike bursting, EPSPs evoked complex spike bursts in a highly stereotypic manner during 5 Hz stimulation, and we found that different stimulation patterns could be used to manipulate the total number of coincident EPSP and complex spike bursts evoked by a given number of 5 Hz stimulation pulses (Figs. 3C,D, 4). When 900 pulses of 5 Hz stimulation were delivered as six trains of 150 pulses (intertrain interval, 20 sec) there was a more than twofold increase in complex spike bursting in 9 of 12 experiments (Fig. 4C; the percentage of EPSPs that evoked complex spike bursts was increased to 70% during patterned stimulation compared with 27% during a continuous train). Consistent with the hypothesis that increasing the amount of EPSP and complex spike burst coincidence should enable the induction of LTP during long trains of 5 Hz stimulation, this pattern of stimulation induced signicant LTP (Fig. 4 D). In the three experiments in which patterned stimulation failed to enhance complex spike bursting (19% of the EPSPs evoked complex spike bursts), no LTP was observed (fEPSPs were 109.6 5.2% of baseline). We also examined the effects of delivering 300 pulses of 5 Hz stimulation as two 150 pulse trains (intertrain interval, 10 sec). Here the EPSP and complex spike burst coincidence was increased (73% of the EPSPs evoked complex spike bursts compared with 52% for a continuous train), and now 300 pulses of 5 Hz stimulation induced LTP (fEPSPs were potentiated to 153.8 18.2% of baseline; n 6; data not shown).

complex spike bursting begins during 5 Hz stimulation observed in our experiments. Although decreased inhibitory synaptic transmission may contribute to complex spike burst generation, disinhibition alone cannot account for the pattern of complex spike bursting observed in our experiments, because complex spike bursting typically declined after 30 sec of 5 Hz stimulation and ceased altogether after 1 min of stimulation (Fig. 4 A), whereas inhibitory synaptic potentials remain depressed for the duration of 5 Hz stimulation (data not shown). Synaptic stimulation could also contribute to complex spike bursting by activating postsynaptic NMDA receptors, which have been proposed to contribute to burst ring in hippocampal pyramidal cells (Abraham and Kairiss, 1988; Poolos and Kocsis, 1990; Pongracz et al., 1992). However, a high concentration of the NMDA receptor antagonist
J. Neurosci., September 15, 1998, 18(18):71187126 7125
APV (100 M) did not block complex spike bursting in our experiments, suggesting that NMDA receptor activation is not required for EPSP-evoked complex spike bursting during 5 Hz stimulation. C learly, much remains to be discovered regarding the synaptic and cellular mechanisms responsible for the activitydependent pattern of complex spike bursting elicited by 5 Hz trains of synaptic stimulation. However, the ability of 5 Hz synaptic stimulation to reliably elicit complex spike bursts in the hippocampal slice preparation should facilitate a more in-depth analysis of this phenomenon.
Complex spike bursts enable the induction of LTP during 5 Hz stimulation
Recent studies investigating phenomena ranging from the elementary properties of synaptic transmission to the behavioral correlates of single-unit ring in vivo suggest that the complex spike mode of action potential ring in hippocampal pyramidal cells represents an informationally rich form of neuronal activity (for review, see Lisman, 1997). For instance, hippocampal pyramidal cells are often preferentially activated when animals enter specic locations with the environment (OKeefe and Dostrovsky, 1971; OKeefe, 1976; Muller, 1996), and the region in which a given cell is maximally activated, called its place eld, is more precisely dened by complex spike bursting than by single spikes (Otto et al., 1991). Complex spike bursting in ensembles of hippocampal neurons may thus generate a more accurate internal representation of position in the external environment than single spikes (Lisman, 1997). In our experiments we found that (1) LTP is induced by strong-intensity 5 Hz synaptic stimulation that evokes postsynaptic complex spike bursting but not by weakintensity 5 Hz stimulation that fails to evoke complex spike bursts; (2) LTP is blocked by low concentrations of TTX that have little effect on the generation of single postsynaptic action potentials but block complex spike bursting; (3) strong-intensity 5 Hz stimulation trains that terminate before complex spike bursting begins fail to induce LTP; (4) LTP is induced by pairing synaptic potentials with simulated complex spike bursts evoked by antidromic stimulation but not by pairing EPSPs with single antidromic action potentials; (5) the induction of LTP is inhibited by patterning stimulation protocols to reduce complex spike bursting during short trains of 5 Hz stimulation; and (6) patterned stimulation protocols that increase complex spike bursting during long trains of 5 Hz stimulation enable LTP induction. Our results thus indicate that, in addition to their proposed role in information coding, postsynaptic complex spike bursts also have an important role in synaptic plasticity. Although the patterns of neuronal activity that induce LTP in vivo are unknown, previous studies have shown that patterns of presynaptic ber stimulation that mimic complex spike bursting can induce LTP. For instance, LTP can be induced in vitro by presynaptic bursts of high-frequency stimulation delivered at the theta frequency (Larson et al., 1986) or in phase with carbacholinduced theta frequency-like activity (Huerta and Lisman, 1993, 1995). Likewise, bursts of high-frequency presynaptic stimulation applied during the positive phase of the theta rhythm in vivo induce robust LTP (Holscher et al., 1997). Our results indicate that in addition to this potential presynaptic role for complex spike bursting, postsynaptic complex spike bursts enable the induction of LTP in the absence of presynaptic bursting. One possibility suggested by our ndings is that back-propagating action potentials during postsynaptic complex spike bursting in vivo (Buzaki et al., 1996) produce a state of heightened sensitivity

during which even low frequencies of presynaptic activity can readily induce LTP. Because action potentials can be initiated in pyramidal cell dendrites under some conditions (Spencer and Kandel, 1961; Wong et al., 1979; Poolos and Kocsis, 1990; Turner et al., 1991; Spruston et al., 1995), our results do not rule out the possibility that dendritically initiated, rather than backpropagating, complex spike bursts produce such an effect. Because brief trains of 5 Hz stimulation (75 and 150 pulses) induce LTP, our results indicate that unpotentiated synapses that are coactive with postsynaptic complex spike bursts will readily undergo LTP, even when relatively few EPSPs evoke complex spike bursts. In contrast, little LTP was induced by longer trains of continuous 5 Hz stimulation (300 and 900 pulses), even though the total number of EPSP-evoked complex spike bursts was several times greater than that elicited by shorter trains of 5 Hz stimulation. LTP could be induced by long trains of 5 Hz stimulation when patterned stimulation was used to increase the number of EPSP-evoked complex spike bursts. Thus, the amount of coincident presynaptic and postsynaptic activity (EPSP-evoked complex spike bursts) required to induce LTP increases during 5 Hz stimulation in a manner similar to that predicted by the BCM (Bienenstock, Cooper, Muro) model of synaptic plasticity (Bear et al., 1987), which proposes that the threshold level of coincident synaptic activity needed to induce persistent increases in synaptic strength increases with the induction of LTP. Because previous studies have shown that protein phosphatase inhibitors enable the induction of LTP during long trains of 5 Hz stimulation (Thomas et al., 1996), an activity-dependent activation of protein phosphatases may underlie the need for increased levels of EPSP and complex spike burst coincidence to induce LTP during long trains of 5 Hz stimulation. Physiologically, this need for higher levels of coincident EPSP and complex spike bursts to induce LTP as stimulation extends beyond that minimally required for LTP induction may act to restrict LTP to only those synapses that reliably evoke complex spike bursting (Fig. 4, compare A,C and D). Thus, LTP induced by EPSP-evoked complex spike bursting during lowfrequency stimulation is truly Hebbian (Hebb, 1949); it not only requires postsynaptic action potentials but, at least for longduration trains, also requires that EPSPs consistently contribute to postsynaptic bursting. Hippocampal pyramidal cells often re complex spike bursts when animals occupy specic locations within the environment (Ranck, 1973; Muller, 1996). Although place-specic ring of pyramidal cells is probably generated endogenously via locomotor cues that provide information about position in the external environment, synaptic inputs that convey place-specic sensory information about the environment are thought to become linked, or bound, to ring of a particular place cell through a process such as LTP (McNaughton et al., 1996). Our results, which show that synaptic inputs active during complex spike bursting undergo robust LTP, suggest how this latter process may occur in vivo and provide a cellular basis for understanding how the theta frequency complex spike bursting observed in CA1 pyramidal cells during behaviors associated with learning (Otto et al., 1991) might contribute to memory formation.

REFERENCES

Abraham WC, Kairiss EW (1988) Effects of the NMDA antagonist 2AP5 on complex spike discharge by hippocampal pyramidal cells. Neurosci Lett 89:36 42. Azouz R, Jensen MS, Yaari Y (1996) Ionic basis of spike afterdepolarization and burst generation in adult rat hippocampal CA1 pyramidal cells. J Physiol (L ond) 492:211223.
7126 J. Neurosci., September 15, 1998, 18(18):71187126
Bear MF, Cooper LN, Ebner FF (1987) A physiological basis for a theory of synapse modication. Science 237:42 48. Bland BH (1990) Physiology and pharmacology of hippocampal formation theta rhythms. Prog Neurobiol 26:154. Bliss TVP, Collingridge GL (1993) A synaptic model of memory: longterm potentiation in the hippocampus. Nature 361:3139. Blitzer RD, Wong T, Nouranifar R, Iyengar R, Landau EM (1995) Postsynaptic cAMP pathway gates early LTP in the hippocampal CA1 region. Neuron 15:14031414. Bolshakov VY, Siegelbaum SA (1994) Postsynaptic induction and presynaptic expression of hippocampal long-term depression. Science 264:1148 1152. Buzaki G, Penttonen M, Nadasdy Z, Bragin A (1996) Pattern and inhibition-dependent invasion of pyramidal cell dendrites by fast spike in the hippocampus in vivo. Proc Natl Acad Sci USA 93:99219925. Cummings JA, Mulkey RM, Nicoll R A, Malenka RC (1996) C a 2 signaling requirements for long-term depression in the hippocampus. Neuron 16:825 833. Davies CH, Davies SN, Collingridge GL (1990) Paired-pulse depression of monosynaptic GABA-mediated inhibitory postsynaptic responses in rat hippocampus. J Physiol (L ond) 424:513531. Debanne D, Gahwiler BH, Thompson SM (1996) Cooperative interac tions in the induction of long-term potentiation and depression of synaptic excitation between hippocampal CA3-CA1 cell pairs in vitro. Proc Natl Acad Sci USA 93:1122511230. Fox S, Ranck JB (1975) L ocalization and anatomical identication of theta and complex spike cells in dorsal hippocampal formation of rats. Exp Neurol 49:299 313. Fujii S, Saito K, Miyakawa H, Ito K , Kato H (1991) Reversal of longterm potentiation (depotentiation) induced by tetanus stimulation of the input to CA1 neurons of guinea pig hippocampal slices. Brain Res 555:112122. Grover LM, Teyler TJ (1990) Two components of long-term potentiation induced by different patterns of afferent activation. Nature 347:477 479. Gustafsson B, Wigstrom H (1990) Basic features of long-term potentia tion in the hippocampus. Semin Neurosci 2:321333. Hebb DO (1949) The organization of behavior. New York: Wiley. Holscher C, Anwyl R, Rowan MJ (1997) Stimulation on the positive phase of hippocampal theta rhythm induces long-term potentiation that can be depotentiated by stimulation on the negative phase in area CA1 in vivo. J Neurosci 15:6470 6477. Huang Y-Y, Colino A, Selig DK , Malenka RC (1992) The inuence of prior synaptic activity on the induction of long-term potentiation. Science 255:730 733. Huerta PT, Lisman JE (1993) Heightened synaptic plasticity of hippocampal CA1 neurons during a cholinergically induced rhythmic state. Nature 364:723725. Huerta PT, Lisman JE (1995) Bidirectional synaptic plasticity induced by a single burst during cholinergic theta oscillation in CA1 in vitro. Neuron 15:10531063. Ito K-I, Miura M, Furuse H, Z hixiong C, Kato H, Yasutomi D, Inoue T, Mikoshiba K, Kimura T, Sakakibara S, Miyakawa H (1995) Voltagegated Ca 2 channel blockers, -AgaIVA and Ni 2, suppress the induction of theta-burst induced long-term potentiation in guinea-pig hippocampal CA1 neurons. Neurosci Lett 183:112115. Jensen MS, Azouz R, Yaari Y (1994) Variant ring patterns in rat hippocampal pyramidal cells modulated by extracellular potassium. J Neurophysiol 71:831 839. Jensen MS, Azouz R, Yaari Y (1996) Spike after-depolarization and burst generation in adult rat hippocampal CA1 pyramidal cells. J Physiol (Lond) 492:199 210. Jester JM, Campbell LW, Sejnowski TJ (1995) Associative EPSP-spike potentiation induced by pairing orthodromic and antidromic stimulation in rat hippocampal slices. J Physiol (L ond) 484:689 705. Johnston D, Williams S, Jaffe D, Gray R (1992) NMDA receptorindependent long-term potentiation. Annu Rev Physiol 54:489 505. Kandel ER, Spencer WA (1961) Electrophysiology of hippocampal neurons. II. After-potentials and repetitive ring. J Neurophysiol 25:243259. Larson J, Wong D, Lynch G (1986) Patterned stimulation at the theta frequency is optimal for the induction of hippocampal long-term potentiation. Brain Res 368:347350. Lisman JE (1994) The C aM kinase II hypothesis for the storage of synaptic memory. Trends Neurosci 17:406 412. Lisman JE (1997) Bursts as a unit of neural information: making unreliable synapses reliable. Trends Neurosci 20:38 43.

Magee JC, Johnston D (1995) Synaptic activation of voltage-gated channels in the dendrites of hippocampal pyramidal neurons. Science 268:301304. Magee JC, Johnston D (1997) A synaptically controlled, associative signal for Hebbian plasticity in hippocampal neurons. Science 275:209 212. Markram H, L ubke J, Frotscher M, Sakmann B (1997) Regulation of synaptic efcacy by coincidence of postsynaptic APs and EPSPs. Science 275:213215. Mayford M, Wang J, Kandel ER, ODell TJ (1995) C aMKII regulates the frequency-response f unction of hippocampal synapses for the production of both LTD and LTP. C ell 81:891904. McC arren M, Alger BE (1985) Use-dependent depression of IPSPs in rat hippocampal pyramidal cells in vitro. J Neurophysiol 53:557571. McNaughton BL, Barnes CA, Gerrard JL, Gothard K, Jung MW, Knierim JJ, Kudrimoti H, Qin Y, Skaggs W E, Suster M, Weaver KL (1996) Deciphering the hippocampal polyglot: the hippocampus as a path integration system. J E xp Biol 199:173185. Mulkey RM, Malenka RC (1992) Mechanisms underlying induction of homosynaptic long-term depression in area CA1 of the hippocampus. Neuron 9:967975. Muller R (1996) A quarter of a century of place cells. Neuron 17:813 822. Nicoll R A, Kauer JA, Malenka RC (1988) The current excitement in long-term potentiation. Neuron 1:97103. ODell TJ, Kandel ER (1994) L ow-frequency stimulation erases LTP through an N DM A receptor-mediated activation of protein phosphatases. Learn Mem 1:129 139. OKeefe J (1976) Place units in the hippocampus of the freely moving rat. E xp Neurol 51:78 109. OKeefe J, Dostrovsky J (1971) The hippocampus as a spatial map. Preliminary evidence from unit activity in the freely moving rat. Brain Res 34:171175. Oliet SHR, Malenka RC, Nicoll R A (1997) T wo distinct forms of longterm depression coexist in CA1 hippocampal pyramidal cells. Neuron 18:969 982. Otto T, Eichenbaum H, Wiener SI, Wible CG (1991) Learning-related patterns of CA1 spike trains parallel stimulation parameters optimal for inducing hippocampal long-term potentiation. Hippocampus 1:181192. Pacelli GJ, Su W, Kelso SR (1991) Activity-induced decrease in early and late inhibitory synaptic conductances in hippocampus. Synapse 7:113. Pongracz F, Poolos N P, Kocsis JD, Shepherd GM (1992) A model of NMDA receptor-mediated activity in dendrites of hippocampal CA1 pyramidal neurons. J Neurophysiol 68:2248 2259. Poolos N P, Kocsis JD (1990) Dendritic action potentials activated by NMDA receptor-mediated EPSPs in CA1 pyramidal cells. Brain Res 524:342346. Ranck JB (1973) Studies on single neurons in dorsal hippocampal formation and septum in unrestrained rats. Part I. Behavioral correlates and ring repertoires. E xp Neurol 41:461531. Scharfman H E, Sarvey JM (1985) Postsynaptic ring during repetitive stimulation is required for long-term potentiation in hippocampus. Brain Res 331:267274. Spencer WA, Kandel ER (1961) Electrophysiology of hippocampal neurons. IV. Fast prepotentials. J Neurophysiol 24:272285. Spruston N, Schiller Y, Stuart B, Sakmann B (1995) Activity-dependent action potential invasion and calcium inux into hippocampal CA1 dendrites. Science 268:297300. Suzuki SS, Smith GK (1985) Burst characteristics of hippocampal complex spike cells in the awake rat. E xp Neurol 89:90 95. Suzuki SS, Smith GK (1988) Spontaneous EEG spikes in the normal hippocampus. III. Relations to evoked potentials. Electroencephalogr C lin Neurophysiol 69:541549. Talbot MJ, Sayer RJ (1996) Intracellular QX-341 inhibits calcium currents in hippocampal CA1 pyramidal neurons. J Neurophysiol 76:2120 2124. Thomas MJ, Moody TD, Makhinson M, ODell TJ (1996) Activitydependent -adrenergic modulation of low frequency stimulation induced LTP in the hippocampal CA1 region. Neuron 17:475 482. Thompson SM, Gahwiler BH (1989) Activity-dependent disinhibition. I. Repetitive stimulation reduces I PSP driving force and conductance in the hippocampus in vitro. J Neurophysiol 61:512533. T urner RW, Meyers DER, Richardson TL, Baker JL (1991) The site for initiation of action potential discharge over the somatodendritic axis rat hippocampal CA1 pyramidal neurons. J Neurosci 11:2270 2280. Wong RK S, Prince DA, Basbaum AI (1979) Intradendritic recordings from hippocampal neurons. Proc Natl Acad Sci USA 76:986 990.

The Journal of Neuroscience, December 24, 2008 28(52):1403114041 14031
Development/Plasticity/Repair
Postsynaptic Action Potentials Are Required for NitricOxide-Dependent Long-Term Potentiation in CA1 Neurons of Adult GluR1 Knock-Out and Wild-Type Mice
Keith G. Phillips, Neil R. Hardingham, and Kevin Fox
Cardiff School of Biosciences, Cardiff University, Cardiff CF10 3AX, United Kingdom
Neocortical long-term potentiation (LTP) consists of both presynaptic and postsynaptic components that rely on nitric oxide (NO) and the GluR1 subunit of the AMPA receptor, respectively. In this study, we found that hippocampal LTP, induced by theta-burst stimulation in mature (8-week-old) GluR1 knock-out mice was almost entirely NO dependent and involved both the splice variant of NO synthase-1 and the NO synthase-3 isoforms of NO synthase. Theta-burst induced LTP was also partly NO-dependent in wild-type mice and made up 50% of the potentiation 2 h after tetanus. Theta-burst stimulation reliably produced postsynaptic spikes, including a high probability of complex spikes. Inhibition of postsynaptic somatic spikes with intracellular QX314 or local TTX application prevented LTP in the GluR1 knock-out mice and also blocked the NO component of LTP in wild types. We conclude that theta-burst stimulation is particularly well suited to producing the postsynaptic somatic spikes required for NO-dependent LTP. Key words: plasticity; potentiation; nNOS; eNOS; working memory; long-term memory

Introduction

The mechanisms underlying hippocampal long-term potentiation (LTP) have been studied extensively since its original discovery (Bliss and Lomo, 1973). Recently, efforts have concentrated on the postsynaptic mechanisms of LTP, which involve insertion of AMPA receptors into the postsynaptic membrane (Malinow and Malenka, 2002). However, presynaptic components of hippocampal LTP have also been documented (Malinow and Tsien, 1990), most recently by direct imaging (Stanton et al., 2005; Bayazitov et al., 2007) and less is known of the incipient mechanisms involved in this form of LTP. Because LTP can be induced postsynaptically (Malenka et al., 1989) and yet is partly expressed presynaptically (Stanton et al., 2005; Bayazitov et al., 2007), some retrograde factor must be involved in coordinating presynaptic and postsynaptic components of transmission strength (Lisman and Raghavachari, 2006). Nitric oxide (NO) became an early candidate for this retrograde factor (Haley et al., 1992; Kantor et al., 1996; Son et al., 1996), but research in this area has slowed partly due to the difficulty of reproducing findings in different laboratories on the role of NO in LTP and memory [for review, see Holscher (1997)]. Recent evidence from the neocortex has shown that layer II/III cells exhibit LTP that can be separated into presynaptic and postsynaptic components by manipulating GluR1 and NOS
Received Aug. 20, 2008; revised Oct. 2, 2008; accepted Oct. 18, 2008. This work was supported by the Medical Research Council (UK) and the National Institute of Mental Health (Conte Center). We thank them for supporting this work and Rob Malenka for critically reading a previous version of this manuscript. Correspondence should be addressed to Kevin Fox, Cardiff School of Biosciences, Museum Avenue, Cardiff University, Cardiff CF10 3AX, UK. E-mail: foxkd@cardiff.ac.uk. DOI:10.1523/JNEUROSCI.3984-08.2008 Copyright 2008 Society for Neuroscience 0270-6474/08/2814031-11$15.00/0
(Hardingham and Fox, 2006). While GluR1 is responsible for the postsynaptic component of LTP in the neocortex, the presynaptic component is dependent on postsynaptic NO synthase (NOS) activation (Hardingham and Fox, 2006). In the neocortex, LTP cannot be abolished entirely by blocking either NOS or GluR1, but blocking both simultaneously eliminates LTP. The first reports of GluR1-dependent LTP in the hippocampus indicated that LTP was completely absent in GluR1 knockouts (Zamanillo et al., 1999), but it was later discovered that GluR1 dependent LTP was present in younger animals and required a spike-timing protocol to induce it (Hoffman et al., 2002; Jensen et al., 2003). This raises the possibility that this residual component of LTP is also NO dependent in the hippocampus, as it is in the neocortex. Therefore, we looked at LTP in the hippocampus of GluR1 null mutants (Zamanillo et al., 1999) to test whether LTP is NOdependent. We also studied the isoforms of NOS involved using the endothelial NOS (NOS-3) knock-out and the neuronal NOS (NOS-1) knock-out mice. The NOS-1 knock-out shows a 94.5% reduction in catalytic activity (Huang et al., 1993) and lacks the major splice variant but not the and gamma splice variants of NOS-1 (Eliasson et al., 1997). The splice variant contains a PDZ domain which links NOS-1 to PSD-95 and hence to the postsynaptic density, whereas the and gamma isoforms do not and so are cytoplasmic (Eliasson et al., 1997). Deletion of the synaptically located NOS-1 isoform therefore makes the NOS-1 knock-out particularly well suited to studying synaptic deficits. Our studies reveal that both major NOS isoforms play a role in hippocampal LTP and that postsynaptic spikes are necessary for the induction of the NO component of LTP in both GluR1 knock-outs and wild-type mouse hippocampus.

14032 J. Neurosci., December 24, 2008 28(52):1403114041
Phillips et al. Nitric-Oxide-Dependent Hippocampal LTP

Materials and Methods

Animals. Subjects were mice aged P(weeks) for the intracellular experiments and P(weeks) for the extracellular experiments. AMPA receptor subunit 1 (GluR1) knock-out mice, NO synthase isoform 1 (NOS-1) knock-out mice, NO synthase isoform 3 (NOS-3) knock-out mice and wild-type littermates were bred into a C57BL/6 background and maintained in the colony as heterozygotes. Experimental null mutants and wild-type littermates were bred from heterozygote crosses (cousin mating). Double knock-out animals were created by breeding heterozygous single knock-outs until double heterozygous males and females were produced. Double knock-outs were produced by mating double heterozygous animals, or on a few occasions by mating GluR1/ NOS/ mice with double heterozygotes. The GluR1 knock-out mice were kindly supplied by Rolf Sprengel (Max Planck Institute for Medical Research, Heidelberg, Germany) via the Rawlins laboratory at Oxford University. The NOS-1 and NOS-3 knock-outs were obtained from The Jackson Laboratory. We genotyped the animals used in this study by PCR using primers ordered from MWG. The following primer sequences were used, for NOS-1: (oIMR13) 5 CTT GGG TGG AGA GGC TAT TC 3; oIMR14 5AGG TGA GAT GAC AGG AGA TC 3; (oIMR406) 5 TCA GAT CTG ATC CGA GGA GG 3; (oIMR407) 5 TTC CAG AGC GCT GTC ATA GC 3. For NOS-3: (oIMR94) 5 TGG CTA CCC GTG ATA TTG CT 3; (oIMR1823) 5 ATT TCC TGT CCC CTG CCT TC 3; (0IMR1824) 5GGC CAG TCT CAG AGC CAT AC 3. Jackson Laboratories supplied both NOS-1 and NOS-3 primer sequences. For the GluR1 knock-outs we used (1005) 5 AAT GCC TAG TAC TAT AGT GCA CG 3; (MH60) 5 CAC TCA CAG CAA TGA AGC AGG AC 3; (3Int3) 5 CTG CCT GGG TAA AGT GAC TTG G 3. Rolf Sprengel supplied primer sequences for the GluR1 knock-outs. Slice preparation. Mice were killed via cervical dislocation and decapitated. Brains were quickly removed and immersed into ice-cold artificial CSF (aCSF) [composition (in mM): 124 NaCl, 2.3 KCl, 2 CaCl2, 1.3 MgSO4, 1 NaH2PO4, 26 NaHCO3, and 11 D-glucose] constantly bubbled with 95%O2/5%CO2 to maintain the pH at 7.4. Coronal sections (400 m) were cut with a vibratome and incubated for at least 1 h in a submersion chamber kept at 32C. Extracellular field potentials. Slices were transferred to a submerged recording chamber perfused with aCSF at 32C. Extracellular field potentials were recorded in the stratum radiatum of the CA1 region of hippocampus using carbon fiber electrodes. Responses were evoked in control and test pathways using a 20 s square voltage step applied at 0.05 Hz through two monopolar electrodes located in stratum radiatum test (S1) and control (S2) pathways. The S1 electrode was placed approximately equidistant from the molecular and pyramidal layers on the CA3 side of the recorded cell. To ensure pathway independence, the stimulating electrodes were placed at slightly different depths in the stratum radiatum. The S2 electrode was placed either higher or lower than S1 (in alternate experiments) and was always located on the subiculum side of the recorded cell. If any effect on the S2 control pathway was observed after tetanus given to the S1 pathway, the recording was discarded. Input/ output (I/O) curves were produced by gradual increases in stimulus strength at the beginning of each experiment, until a stable baseline of evoked response was reached. The test stimulus pulse was then adjusted to produce a field EPSP (fEPSP) whose slope and amplitude was 40% that of the maximum possible fEPSP and was kept constant throughout the experiment. The negative going slope of each fEPSP was measured over the 20 80% range of the peak amplitude. Responses were amplified (Axoclamp 2B), digitized [Cambridge Electronic Design (CED) 1401], and recorded using Signal (CED). Dual extracellular and whole-cell patch-clamp recordings. Recordings were made in a submerged chamber perfused with aCSF at 32C. To enable the results of the intracellular recordings to be compared directly to the extracellular recordings it was important to keep the stimulus strength constant between studies. This was achieved by recording an extracellular I/O curve before the intracellular recording was made. Intracellular recordings were then obtained from CA1 cells that were directly above the fEPSP recording electrode, perpendicular to the stratum

J. Neurosci., December 24, 2008 28(52):1403114041 14033
maximal value measured from the I/O curves plotted at the start of the experiment. Theta-burst stimulation at the same pulse width (20 s) or 100 Hz stimulation at an increased stimulus strength, (produced by doubling the width of the stimulus pulse from 20 to 40 s) were also ineffective at inducing LTP in the GluR1 knock-outs (Fig. 1 B, C). However, thetaburst stimulation in combination with the double pulse-width stimulus produced LTP of similar magnitude both in wild types and GluR1 knock-outs (173 7% in wild types vs in GluR1 knockouts measured at 60 min after tetanus) (Fig. 1 D). LTP was significant in both cases here ( p 0.001, Bonferroni corrected post hoc t test). Over the first 20 min, LTP in the GluR1 knock-outs increased more slowly than LTP in the wild types. Figure 1 D (bottom, solid line) illustrates the difference in the time course of the potentiation by subtracting the potentiation seen in the GluR1 knock-outs from the potentiation seen in wild types; it is very similar to the LTP described by Hoffman et al. (2002) and Jensen et al. (2003). However, two factors were different in the present study; first LTP was produced purely by orthodromic stimuli and did not require postsynaptic current injection to ensure spike pairing. Second, the animals were at least 8 weeks Figure 1. High-intensity theta-burst stimulation produces GluR1- independent LTP that depends on NMDAR and CaMKII. A, C, of age and therefore the LTP was not reIn GluR1 / mice (E), 100 Hz stimulation (100 pulses at 100 Hz, repeated three times at 0.05 Hz, delivered at arrow) at either stricted to immature synases. low intensity (A) (test pulse-width 20 s) or high intensity (C) (double test pulse-width 40 s), produces no significant potenWe found that the induction of LTP in tiation of the fEPSP, compared with a highly significant potentiation in wild-type mice (F). B, Low-intensity theta-burst stimuthe GluR1 knock-outs depended not only lation [four pulses at 100 Hz repeated 10 times at 5 Hz (theta) repeated three times at 0.05 Hz] also produces no fEPSP potentiation in the GluR1 / mice (E), whereas significant potentiation is seen in wild-type mice (F). D, High-intensity theta-burst on the intensity of stimulation but also on stimulation produces a slowly rising form of potentiation in GluR1 / mice (E) that is indistinguishable from the potentiation the parameters of the tetanus protocol. in wild-type mice (F) at 60 min. Each point plots the average amplitude of four successive fEPSPs normalized with respect to the Neither theta-burst stimulation with a 20 baseline and expressed as mean SEM. Insets are representative traces taken at time points indicated by the bars (red, control s stimulus pulse-width (Fig. 1 B) nor 100 period; black, min) with the symbols identifying individual experimental conditions. Calibration: 1 mV, 10 ms. E, 50 M Hz stimulation with a 40 s stimulus / D-AP5 or 5 M AIP completely block LTP in wild-type (filled bars) and GluR1 mice (gray bars). F, I/O curves for wild-type (F) pulse-width (Fig. 1C) reliably induced and GluR1 / (E) mice show no difference in baseline transmission [p 0.05, not significant (NS)]. G, Fiber volley, I/O curves LTP in our hands. The dependence of LTP for wild-type (F) and GluR1 / (E) mice also show no difference in baseline transmission (p 0.05, NS). WT, Wild type. on stimulus intensity in the GluR1 knockouts could not be accounted for by lower levels of synaptic transmission (when malized the mean amplitude and CV2 values to the control period, and compared with wild types), as the I/O curves were not signifiplotted values for the two time periods (Malinow and Tsien, 1990). Paired-pulse facilitation (PPF) was measured (interstimulus interval cantly different between the two genotypes (Fig. 1 F) (Bonferroni 75 ms) during the control period and 40 min after the induction of LTP. corrected t test, p 0.05). To analyze this result further, the I/O PPF was expressed as a ratio, i.e., the amplitude of the second EPSP was function was also assessed relative to the size of the fiber volley. divided by the amplitude of the first. The average PPF during the control The fiber volley amplitude is proportional to the number of axperiod was then compared with the PPF ratio 40 min after the tetanus. ons activated, allowing for an independent measurement of input The change in the PPF ratio (PPF) was then calculated by subtracting strength and compensating for any small differences in stimulatthe post tetanus PPF ratio from the control PPF ratio in each individual ing and recording electrode placement between experiments. experiment and then averaged (Hardingham and Fox, 2006). However, when the I/O response was plotted against the fiber volley we still found no difference between wild types and GluR1 Results LTP can be induced by orthodromic theta-burst stimulation knock-outs (Bonferroni corrected t test, p 0.05) (Fig. 1G). As in adult mice mentioned above, the stimulus intensity was routinely set at 40% A tetanic stimulus applied at 100 Hz to the schaeffer collateralof maximum response saturation corresponding to a mean value CA1 pathway produced robust LTP in wild-type mice (mean of 10V (Fig. 1 F). As can be seen from the I/O curve the response SEM 153 9%) but not in GluR1 knock-out mice (96 4%) averages for the two genotypes at the 40% setting are very similar (Fig. 1 A). The stimulus intensity was routinely set at 40% of the (Fig. 1 F, G).

14034 J. Neurosci., December 24, 2008 28(52):1403114041
We investigated whether the LTP seen in the GluR1 knock-outs depended on the same receptors and signaling cascades as LTP in the wild types (Fig. 1 E). We conclude that induction of LTP in the GluR1 knock-outs by theta-burst stimulation applied using a 40 s stimulus pulse was dependent on NMDA receptors because it was blocked by 50 M D-AP5 applied extracellularly (101 4%; significantly different from control, p 0.001; t(27) 4.2) and dependent on CaMKII because it was blocked by 5 M autocamtide inhibitory peptide (AIP) applied intracellularly (103 6, p 0.001, t(27) 4.1). Efficacy of LTP protocols is strongly correlated with spike production To understand more about the differences between the theta-burst and 100 Hz stimulation protocols, we recorded intracellularly from postsynaptic cells during LTP induction. We found that theta-burst stimulation only produced a significant number of postsynaptic action potentials at the higher stimulation intensity (40 s pulse-width) as shown in Figure 2 B, D. To Figure 2. Number of postsynaptic spikes differs significantly between induction protocols. Example traces of intracellular our surprise, we found that in our hands, recording during (A), 100 Hz stimulation at low intensity; B, theta-burst stimulation at low intensity; C, 100 Hz stimulation at high 100 Hz stimulation was not at all effective intensity; and D, theta-burst stimulation at high intensity. Left (panels) are the first 100 ms of each burst. E, Bars indicate the in producing postsynaptic spikes. Action number of spikes produced per stimuli in the train (total number of spikes in train/total number of stimuli given in the train). The potentials did not follow the high rate of probability of generating a spike is significantly greater during high-intensity (gray bars) theta-burst stimulation compared with low-intensity (white bars) 100 Hz stimulation, high-intensity 100 Hz stimulation and low-intensity theta-burst stimulation. F, The stimulation and rapidly failed over time, total number of spikes per train is significantly greater during high-intensity (gray bars) theta-burst stimulation compared with either due to depolarization block or per- low-intensity (white bars) 100 Hz stimulation, high-intensity 100 Hz stimulation or low-intensity theta-burst stimulation. haps due to spike accommodation (Fig. 2 A, C). somatic action potentials in LTP induction in the GluR1 knockWe quantified these effects and found that the spike probabilouts, we applied the specific sodium channel blocker TTX extraity (per stimulus) was 40-fold greater during theta-burst stimcellularly via a micropipette carefully positioned close to the ulation (44 8% in wild-type, 53 6% in GluR1 knock-outs) soma of the cell being recorded from, under visual control (Fig. than for 100 Hz stimulation (1 6% in wild-type, 2 7% in 4 A). The bath aCSF flowed from dendrites to soma to further GluR1 knock-outs) (Fig. 2 E, F ), using the same intensity of stimlocalize TTX to the soma. Using this technique it was possible to ulation in each case (40% of maximum, 40 s pulse-width). Conpressure eject TTX onto the soma and reversibly block action sequently, applying more presynaptic stimuli during a 100 Hz potentials (Fig. 4 B). We found that blockade of somatic action protocol produced many fewer spikes than with a theta-burst potentials had little effect on the degree of EPSP summation protocol. Although it is possible that other experimenters procaused by the stimulus (Control 22.0 0.7 mV, TTX 19.0 duced action potentials using 100 Hz stimulation, we were not 2.1 mV) (Fig. 4C,D). able to do so, and, as described below, this made 100 Hz stimuBlocking somatic action potentials with TTX did not prevent lation a useful tool for some of the experiments in these studies. LTP in wild types (mean 155 17%) (Fig. 4 E), but it did prevent LTP in GluR1 knock-out mice (109 5%) (Fig. 4 F). Blocking somatic spikes prevents LTP in GluR1 knock-outs Experiments with TTX were interleaved with control experiTo determine the importance of somatic spikes in LTP induction ments. A two-way ANOVA showed an interaction between TTX we recorded from CA1 pyramidal neurones using electrodes contreatment and genotype (F(1,64) 5.08, p 0.03) and post hoc taining the sodium channel blocker QX314 and used the thetatests revealed that this was because the GluR1 knock-out only burst LTP protocol (40 s duration pulses). We found that after showed significant LTP without TTX (t(35) 3.1, p 0.01), while breaking into the cell QX314 rapidly abolished action potentials wild types showed LTP with or without TTX (t(23) 1.68, p (Fig. 3A). Although spikes were eliminated during the theta-burst 0.05). On several occasions we were able to hold the postsynaptic tetanus, the degree of EPSP summation with QX314 was almost cell long enough to reverse the effects of TTX and recover normal identical to control levels (Control 22.0 0.7 mV; QX314 action potentials (Fig. 4 B) (n 3). LTP was not induced when 23.0 1.1 mV). (Fig. 3 B, C). We found that QX314 prevented action potentials were blocked, but subsequently could be ininduction of LTP in the GluR1 knock-out mice (Fig. 3D) but had duced when action potential firing was restored (supplemental no effect on wild-type LTP. Fig. 1, available at www.jneurosci.org as supplemental material). It was possible that QX314 acted by eliminating somatic or We further analyzed the data from the experiments described dendritic orthodromic spikes or by affecting targets other than above to see if the depolarization level produced by the induction sodium channels. Therefore, as a more specific test of the need for

J. Neurosci., December 24, 2008 28(52):1403114041 14035
LTP both with and without spikes. On average, theta-burst stimulation using a double pulse-width produced more than one spike per train (mean 1.75 spikes/train) and produced a number of complex spikes in the postsynaptic cells (Fig. 2 D). From the 44 cells recorded, there was on average four complex spikes per theta-burst train and all cases showed at least one complex spike. Single pulse-width theta-burst stimulation produced far fewer spikes per train (mean 0.22 spikes/train) and rarely produced complex spikes (2 from 16 cells), which might explain why it was less effective in producing LTP in the GluR1 knock-out mice. Spike-dependent LTP in GluR1 knock-outs is largely NO dependent In barrel cortex, a large part of the LTP expressed in GluR1 knock-outs is dependent on NO (Hardingham and Fox, 2006). To determine whether a similar dependency exists in the CA1 region of the hippocampus, we perfused alternate GluR1 knockout slices with the NOS inhibitor L-NNA (Fig. 6 A). With extracellular L-NNA, LTP was reduced to 115 11% at 60 min after a theta-burst tetanus compared with 172 16% in untreated controls, which was a highly significantly reduction (Fig. 6 D p 0.001). The small amount of residual LTP present with L-NNA application was however still significantly different from the untetanised control pathway [using a paired t test (t(10) 3.29, p 0.05, Bonferroni corrected)]. Application of L-NNA did not decrease the probability of spike induction. Spike probability was 0.64 0.09 in untreated GluR1 knock-outs and 0.56 0.06 in L-NNA treated GluR1 knock-outs (F(1,29) 1.16, p 0.56). This data therefore implies that NOS is significantly involved in hippocampal LTP. This conclusion was corroborated by evidence from double knock-out mice in which LTP was reduced in both GluR1/NOS-1 and GluR1/NOS-3 double knock-out animals (Fig. 6 B, C). In both cases, application of L-NNA further reduced LTP in the double knock-outs, indicating that both isoforms of NOS (endothelial (NOS-3) and neuronal (NOS-1) are involved in LTP in the schaeffer collateral CA1 pathway. The residual component of LTP present in the GluR1/NOS-1 knock-outs (142 8%) was significantly different from the untetanised control pathway (t(28) 3.92, p 0.001). Treatment of the double knock-outs with L-NNA (an unspecific NOS inhibitor) reduced but did not totally block LTP (mean 117 11%, t(9) 1.7, p 0.05) (Fig. 6D, summary bars). As was the case with L-NNA application to wild-type slices, reduction of LTP in the double knock-outs was not due to an inability to produce action potentials in the theta-burst tetanus, while the I/O curves were again indistinguishable from those of the single GluR1 knock-outs (Fig. 6 E, F ). These results therefore imply that NOS is involved in a substantial component of LTP in the GluR1 knock-outs. Spike-dependent LTP in wild types is partly NO dependent As the LTP in GluR1 knock-outs requires action potentials and is also largely NO dependent, we hypothesized that the same is true of a component of wild type LTP. We therefore again used two LTP induction protocols, one that caused consistent spike production (theta-burst) and one that in our hands only sparingly produced spikes (100 Hz). Both protocols produced LTP in the wild-type mice (Fig. 7). However, application of L-NNA reduced only the LTP produced by theta-burst stimulation and not that produced by 100 Hz stimulation. The level of LTP induced by theta-burst stimulation was almost halved by application of

Evidence regarding the presynaptic origin of NO-dependent LTP NO signaling has been implicated in the presynaptic modulation of transmitter release in LTP (ODell, 1991). Since LTP in the GluR1 knock-outs is almost fully blocked by NOS inhibition, while wild-type LTP is only partly blocked, one might predict that the locus of expression of the LTP would also be almost entirely presynaptic in the GluR1 knock-outs and a mixture of presynaptic and postsynaptic in the wild types. To test this, we monitored PPF before and 40 min after LTP induction. In the wild types, there was no overall change in PPF after LTP induction (PPF 0.02 0.07, t(19) 0.68, p 0.5) (Fig. 8C). There was a large variability in PPF between individual recordings; six cells showed an increase, six cells show no change, and eight show a decrease in PPF after LTP (Fig. 8 A). However, in the GluR1 knock-outs there was far lower variability in the PPF, with only one cell showing a substantial increase, six remaining unchanged while 14 cells showed a decrease in the PPF ratio after LTP (Fig. 8 B). Using a paired t test and comparing each cell before and after LTP we found a significant decrease in PPF for the GluR1 knock-outs following LTP (PPF 0.28 0.04, t (20) 2.9, p 0.01). Consequently, PPF was significantly different in wild types and GluR1 knock-outs (t(32) 2.7, p 0.05) (Fig. 8C). The initial PPF ratio has been shown to be inversely related to the magnitude and sign of PPF following LTP in the hippocam-
Figure 4. Somatic spikes are required for plasticity in GluR1 / mice. A, Schematic diagram of the experimental setup. Action potential generation and propagation can be blocked by local pressure application of TTX (10 M) to the soma. The slice is positioned so that TTX does not perfuse on the stratum (s.) radiatum. B, Example trace illustrating how spikes generated by a depolarizing current injection (2.5 nA, 500 ms) can be reversibly blocked by the local somatic application of TTX (10 M). C, Spiking during high-intensity theta-burst stimulation (black line) is blocked if TTX is perfused on the soma (red line). D, EPSP summation during the theta-burst stimulation is unaffected by somatic TTX application. E, LTP was induced after a 5 min control period by a high-intensity theta-burst stimulation at t 0 (arrow). Somatic TTX application (E) has a small effect on wild-type (WT) LTP at 45 min (F). F, The LTP observed in GluR1 / mice (F) is completely abolished when somatic spikes are blocked with local TTX application (E). Each point plots the average amplitude of eight successive EPSPs normalized with respect to the baseline and expressed as mean SEM. Insets are representative traces taken at time points indicated by the bars (red, control period; black, min) with the symbols identifying the experimental conditions.

pus, for instance by Schulz et al. (1994). In agreement with these data, we also saw a negative correlation between the initial PPF ratio and the PPF after LTP both in wild types (r 0.47, n 19, p 0.05) and GluR1 knock-outs (r 0.55, n 21, p 0.01) (supplemental Fig. 3A, available at www.jneurosci.org as supplemental material). In the GluR1 knock-outs there was also a negative correlation between the PPF ratio and the magnitude of LTP at 45 min (r 0.52, n 21, p 0.03) and a positive correlation between the control PPF ratio and the magnitude of LTP (r 0.47, n 21, p 0.01) (supplemental Fig. 3 B, C, available at www.jneurosci.org as supplemental material); however, neither of these correlations were apparent in the wild types. The initial PPF and PPF failed to predict the magnitude of LTP in wild types, presumably due to additional postsynaptic mechanisms involving GluR1, while in the GluR1 knock-outs these postsynaptic mechanisms are not available so presynaptic mechanisms dominate. Increased dependence of LTP on presynaptic mechanisms in GluR1 knock-outs has also been reported in the barrel cortex (Hardingham and Fox, 2006) The presynaptic locus of LTP in GluR1 knock-outs was further corroborated by normalized mean CV2 analysis (Malinow and Tsien, 1990). Purely postsynaptic changes would produce a plot with a horizontal trajectory (Fig. 8 D), whereas changes in N
J. Neurosci., December 24, 2008 28(52):1403114041 14037
ual experiments and the average slope and error was calculated for wild types and GluR1 knock-outs, wild-type slope 1.07 0.05, GluR1 knock-out slope 1.30 0.07, t test t(34) 2.5, p 0.05), indicating that in GluR1 knock-outs the locus of LTP expression is more presynaptic than wild types. The conclusion of the CV2 analysis seems consistent with the paired-pulse analysis in that they both suggested a predominantly presynaptic component of LTP in the GluR1 knock-outs and a mixed locus of LTP expression in the wild types. In wild types, theta-burst stimulation produces an additional component of LTP to 100 Hz stimulation These experiments suggest that two mechanistically distinct components of LTP are generated in wild types, dependent on the induction protocol used and on whether somatic spikes are produced during the tetanus. If this is true one might predict that theta-burst stimulation, which induces both GluR1- and NOdependent forms of LTP should occlude subsequent LTP induced by 100Hz stimulation, whereas the converse would not be true. One would predict that since 100 Hz stimulation does not in our hands, induce NO-dependent LTP, it would be possible to produce additional NO-dependent LTP with theta-burst stimulation following the 100 Hz stimulation. We tested this hypothesis in studies where we induced LTP with a strong stimulation protocol (3 theta-burst or Hz) using the 40 s stimulus pulse width. Thirty minutes after LTP the stimulus intensity was turned down to return the field EPSP to its control value and we then tried to induce LTP a second time. We found that theta-burst stimulation produced LTP that occluded further LTP induced by 100 Hz stimulation (Fig. 9A). Transient potentiation (STP) was produced by the second tetanus but it fell back to baseline within 30 min (101 2%) (Fig. 9C). However, if we swapped the order of the stimulus protocols so that the 100 Hz tetanus occurred before the theta-burst tetanus, a small LTP was observed (Fig. 9B). There was no clear post-tetanic potentiation episode and the potentiation rose then remained at a steady state level of (117 7%) for the 50 min we followed it (Fig. 9C). Statistical analysis showed that the thetaburst LTP was significantly different from baseline ( p 0.05). This experiment supports the hypothesis that postsynaptic action potentials (recruited using theta-burst stimulation) activate a mechanistically different and additional component of LTP to that induced by depolarization without postsynaptic action potentials (produced, in our hands, using 100 Hz stimulation). It also suggests that the LTP component produced only by thetaburst is smaller than the LTP component common to both protocols.

Figure 5. The level of potentiation in the GluR1 / mice is correlated to the number of spikes observed in the burst and not to the level of depolarization in the burst. A, Magnitude of LTP in the GluR1 / depends on the total number of spikes during the burst. Individual experiments show a correlation between the number of spikes in a burst and the increase in EPSP observed with LTP. B, Magnitude of LTP in the GluR1 / is not correlated to the average amplitude of summated EPSPs during the theta-burst stimulation. C, Same data as shown in B with control theta-burst stimulation data excluded as spiking will be related to the level of depolarization. Data are pooled from LTP experiments generated by high-intensity 100 Hz stimulation (white circle), theta-burst stimulation (red circle), theta-burst stimulation with 0.2 mM QX 314 (black circle), and theta-burst stimulation with 10 M TTX (blue circle).

Discussion

The main findings of this study are that postsynaptic action potentials are necessary for the NO-dependent component of hippocampal LTP in both GluR1 knock-out and wild-type mice. In GluR1 knock-out mice, almost all the LTP is NO sensitive, while in wild types the later stages of LTP are NO sensitive. In wild types, NO-dependent LTP accounts for 50% of the potentiation 2 h post-tetanus. Presumably, the remaining component of LTP in wild types is GluR1 dependent, which would account for the large difference in the size of the NO-dependent component between the two genotypes. Comparison with previous studies on the role of NO in LTP Previous studies on the role of NO in LTP have investigated the source of discrepant results in different labs. One of the primary
or Pr would cause more vertical trajectories. This is because CV2 is proportional to NPr (1 Pr) 1 and is therefore not dependent on Q, whereas the mean amplitude is proportional to NPrQ and is therefore proportional to Q (in which N is the number of release site, Pr is the probability of release, and Q is the quantal size). In wild types, the trajectory of the CV2 plot was approximately diagonal, indicative of a mixed locus of potentiation [consistent with the work of Hardingham and Fox (2006)] (Fig. 8 D). In GluR1 knock-outs, the trajectory of the CV2 plot was significantly steeper than in wild types (linear fits were made to individ-
14038 J. Neurosci., December 24, 2008 28(52):1403114041
factors appears to be the differing levels of NOS present in different rat (Holscher, 2002) and mouse strains (Blackshaw et al., 2003). The wild-type and mutant mice used in these studies are from a C57/Black 6 background in which NOS-1 is expressed in CA1 cells at higher levels than 129sv mice or rats, but at a level not dissimilar to humans (Blackshaw et al., 2003). The present study exposes two further sources of possible confusion when investigating the NO-dependent component of LTP; first, LTP can occur in the hippocampus despite inhibition of NOS and second, activation of the NO mechanism depends on postsynaptic spike production during the tetanus, which is rarely monitored in extracellular field studies (which comprise practically all studies on the role of NOS in LTP). Both factors could lead to underestimating the role that NO plays in hippocampal LTP. To take the first of these factors; the GluR1 component of LTP would still be present even in cases where NOS activity was completely pharmacologically or genetically inactivated. This explains why many studies have found only a partial block of LTP with NOS inhibition (ODell et al., 1994; Son et al., 1996; Holscher, 2002). The second factor concerns the production of postsynaptic ac- Figure 6. Comparison of LTP in GluR1 / single-knock-out and GluR1 /NOS-1 / and GluR1 / NOS-3 / / (E) mice is significantly reduced by a 5 min tion potentials; it is certainly our experi- double- mutant mice with or without NOS inhibitor. A, LTP in the GluR1 application of 100 M L-NNA (F). B, LTP in the GluR1 / NOS-1 / double-mutant mice (U) is reduced when compared ence that increasing stimulus strength / single-mutant mice (E). The remaining LTP in double-mutant mice is further reduced by 100 M L-NNA (F) tends to inactivate sodium channels and with the GluR1 and is similar to LTP observed in single GluR1 / with 100 M L-NNA. C, LTP in the GluR1 / NOS-3 / double-mutant mice reduce spike production during a 100 Hz (U) is also slightly reduced compared with the GluR1 / single-mutant mice (E) and is comparable with the LTP in GluR1 / tetanus. Absence of postsynaptic spikes NOS-1 / double-mutant mice. The remaining LTP in the double-mutant mice can be further reduced by 100 M L-NNA eliminates the NO component of LTP. (F). Each point plots the average amplitude of three successive fEPSPs normalized with respect to the baseline and expressed as However, the GluR1 component of LTP mean SEM. Insets are representative traces taken at time points indicated by the bars (red, control period; black, min) does not rely on action-potentials and with the symbols identifying the experimental conditions. Calibration: 1 mV, 10 ms. D, The average levels of LTP at min are therefore an increased stimulus strength plotted for the S1 and S2 pathways in the three genotypes. Note the similarity of the LTP in the presence of L-NNA for all three does not affect it in the same way. The genotypes. E, F, I/O curves for GluR1 / single-mutant (F) and GluR1 / NOS-1 / double-mutant (E) and GluR1 / / double-mutant (F) (E) mice show no differences in baseline transmission (p 0.05, NS). combined effect of increasing the stimulus NOS-3 strength is therefore to decrease the NOshown that GluR1 knock-outs have normal levels of whisker dependent component of LTP relative to the GluR1-dependent evoked responses in layers II/III, IV and V of the barrel cortex and component. This explains several reports in the literature that normal levels of synaptic response in the layer IV to II/III and increasing stimulus intensity reduces the NO-dependent compoII/III to V pathway (Wright et al., 2008). However, synaptic scalnent of LTP (Gribkoff and Lum-Ragan, 1992; Chetkovich et al., ing is known to require GluR1 containing AMPA receptors in the 1993; Haley et al., 1993; ODell et al., 1994). hippocampus, which might predict a reduction in distal synaptic In our hands we found that theta-burst stimulation produced currents in the GluR1 knock-outs (Andrasfalvy et al., 2003). It spikes more readily than 100 Hz stimulation. However, this is not may be that presynaptic plasticity mechanisms are able to comto say that it is impossible to produce NO-dependent LTP with pensate for the lack of postsynaptic scaling. 100 Hz stimulation. In fact, some of the pioneering studies on the Earlier studies on GluR1 knock-outs concluded that LTP rerole of NO in LTP found that 100 Hz stimulation produced NOlied on spike pairing protocols that were effective in younger but dependent LTP, provided that the stimulus used was of a weak not older animals (Jensen et al., 2003) or that it required pairing intensity (ODell et al., 1991,1994). We assume that in these cases a burst of postsynaptic spikes with presynaptic stimulation the stimulus was weaker than we used in our studies and that it (Hoffman et al., 2002). Here we show that spikes are essential for more successfully produced postsynaptic spikes during the induction of LTP in the GluR1 knock-outs but that it is sufficient tetanus. that they are produced naturally from orthodromic stimulation. Previous studies have shown that somatic action potentials are Comparison with previous studies on GluR1 knock-outs usually generated by prior dendritic spikes during theta-burst We found no differences in baseline levels of synaptic efficacy stimulation (Golding et al., 2002). Here we found that thetabetween GluR1 knock-out animals and wild types, consistent with previous studies (Zamanillo et al., 1999). It has also been burst stimulation reliably evoked postsynaptic action potentials

14040 J. Neurosci., December 24, 2008 28(52):1403114041
nitrosylation of NSF and hence affecting insertion of AMPA receptors into the membrane (Huang et al., 2005). While we cannot rule out some postsynaptic action of NO in these studies, the paired pulse and CV 2 analysis in the present data suggest that NO mainly acts presynaptically, in common with the conclusions of several other studies (ODell et al., 1991; Hawkins et al., 1998). Finally, one further study has also reported that postsynaptic spikes are necessary during theta-burst LTP induction for the persistence of LTP (Raymond, 2008) again suggesting that the slower developing, NO-dependent component of LTP induced by theta-burst stimulation requires postsynaptic spikes. In conclusion, the recent discovery of different temporal components of memory formation is paralleled by the discovery of different temporal components of LTP. This study and previous studies suggest that the early and late components have different presynaptic and postsynaptic loci. In this study, we further show that the later component relies strongly on NO, which in turn relies on postsynaptic spike production and may provide a means for dissecting different components of hippocampus-dependent memory in the future.

References

Andrasfalvy BK, Smith MA, Borchardt T, Sprengel R, Magee JC (2003) Im paired regulation of synaptic strength in hippocampal neurons from GluR1-deficient mice. J Physiol 552:35 45. Bayazitov IT, Richardson RJ, Fricke RG, Zakharenko SS (2007) Slow presynaptic and fast postsynaptic components of compound long-term potentiation. J Neurosci 27:11510 11521. Blackshaw S, Eliasson MJ, Sawa A, Watkins CC, Krug D, Gupta A, Arai T, Ferrante RJ, Snyder SH (2003) Species, strain and developmental variations in hippocampal neuronal and endothelial nitric oxide synthase clarify discrepancies in nitric oxide-dependent synaptic plasticity. Neuroscience 119:979 990. Bliss TV, Lomo T (1973) Long-lasting potentiation of synaptic transmission in the dentate area of the anaesthetized rabbit following stimulation of the perforant path. J Physiol 232:331356. Chetkovich DM, Klann E, Sweatt JD (1993) Nitric oxide synthaseindependent long-term potentiation in area CA1 of hippocampus. Neuroreport 4:919 922. Eliasson MJ, Blackshaw S, Schell MJ, Snyder SH (1997) Neuronal nitric oxide synthase alternatively spliced forms: prominent functional localizations in the brain. Proc Natl Acad Sci U S A 94:3396 3401. Golding NL, Staff NP, Spruston N (2002) Dendritic spikes as a mechanism for cooperative long-term potentiation. Nature 418:326 331. Gribkoff VK, Lum-Ragan JT (1992) Evidence for nitric oxide synthase inhibitor-sensitive and insensitive hippocampal synaptic potentiation. J Neurophysiol 68:639 642. Haley JE, Wilcox GL, Chapman PF (1992) The role of nitric oxide in hippocampal long-term potentiation. Neuron 8:211216. Haley JE, Malen PL, Chapman PF (1993) Nitric oxide synthase inhibitors block long-term potentiation induced by weak but not strong tetanic stimulation at physiological brain temperatures in rat hippocampal slices. Neurosci Lett 160:85 88. Hardingham N, Fox K (2006) The role of nitric oxide and GluR1 in presynaptic and postsynaptic components of neocortical potentiation. J Neurosci 26:73957404. Hawkins RD, Son H, Arancio O (1998) Nitric oxide as a retrograde messenger during long-term potentiation in hippocampus. Prog Brain Res 118:155172. Hoffman DA, Sprengel R, Sakmann B (2002) Molecular dissection of hippocampal theta-burst pairing potentiation. Proc Natl Acad Sci U S A 99:7740 7745. Holscher C (1997) Nitric oxide, the enigmatic neuronal messenger: its role in synaptic plasticity. Trends Neurosci 20:298 303. Holscher C (2002) Different strains of rats show different sensitivity to block of long-term potentiation by nitric oxide synthase inhibitors. Eur J Pharmacol 457:99 106. Huang PL, Dawson TM, Bredt DS, Snyder SH, Fishman MC (1993) Targeted disruption of the neuronal nitric oxide synthase gene. Cell 75:12731286.

Figure 9. LTP induced by theta-burst stimulation occludes subsequent 100 Hz stimulation induced LTP, whereas LTP induced by 100 Hz stimulation does not completely occlude subsequent LTP induced by theta-burst stimulation. A, LTP was initially induced in the S1 pathway by high- intensity theta-burst stimulation; after 20 min of further recording, the baseline was reset to the initial control level by decreasing the stimulus intensity. A further 10 min baseline period was then recorded before a second high-intensity 100 Hz stimulation was applied to the same pathway. The potentiation induced by the 100 Hz stimulation was transient (STP) and returned back to baseline within 45 min. B, The same dual LTP protocol treatment as in A but performed in reverse: first, high-intensity 100 Hz stimulation was given to S1 followed by high-intensity theta-burst stimulation after resetting the baseline again. The potentiation induced by the high-intensity theta-burst stimulation did not return to baseline and was significantly different from the S2 pathway at 60 min after tetanus ( p 0.05). Each point plots the average amplitude of three successive fEPSPs normalized with respect to the initial baseline and is expressed as mean SEM. Insets are representative traces taken at time points indicated by the bars. C, Bars represent mean levels of potentiation as indicated taken 60 min after the second tetanus.
A compound LTP with an early postsynaptic component and later presynaptic component has recently been described in the hippocampus (Bayazitov et al., 2007). Presynaptic function was monitored directly using a transgenic mouse strain expressing a pH sensitive fluorescent VAMP2 marker in neurons. The presynaptic component of LTP was only induced by theta-burst stimulation and not by 100 Hz stimulation (Bayazitov et al., 2007). Our study suggests that only theta-burst activity evokes presynaptic LTP in the Bayazitov study because it produces postsynaptic spikes more effectively than 100 Hz stimulation and hence enables the NO-dependent component of LTP. There is evidence that NO can play a postsynaptic role in LTP by producing direct

Phillips et al. Nitric-Oxide-Dependent Hippocampal LTP Huang Y, Man HY, Sekine-Aizawa Y, Han Y, Juluri K, Luo H, Cheah J, Lowenstein C, Huganir RL, Snyder SH (2005) S-nitrosylation of N-ethylmaleimide sensitive factor mediates surface expression of AMPA receptors. Neuron 46:533540. Ishida A, Shigeri Y, Tatsu Y, Uegaki K, Kameshita I, Okuno S, Kitani T, Yumoto N, Fujisawa H (1998) Critical amino acid residues of AIP, a highly specific inhibitory peptide of calmodulin-dependent protein kinase II. FEBS Lett 427:115118. Jensen V, Kaiser KM, Borchardt T, Adelmann G, Rozov A, Burnashev N, Brix C, Frotscher M, Andersen P, Hvalby O, Sakmann B, Seeburg PH, Sprengel R (2003) A juvenile form of postsynaptic hippocampal long-term potentiation in mice deficient for the AMPA receptor subunit GluR-A. J Physiol 553:843 856. Kantor DB, Lanzrein M, Stary SJ, Sandoval GM, Smith WB, Sullivan BM, Davidson N, Schuman EM (1996) A role for endothelial NO synthase in LTP revealed by adenovirus-mediated inhibition and rescue. Science 274:1744 1748. Lisman J, Raghavachari S (2006) A unified model of the presynaptic and postsynaptic changes during LTP at CA1 synapses. Sci STKE 2006:re11. Mack V, Burnashev N, Kaiser KM, Rozov A, Jensen V, Hvalby O, Seeburg PH, Sakmann B, Sprengel R (2001) Conditional restoration of hippocampal synaptic potentiation in Glur-A-deficient mice. Science 292:25012504. Malenka RC, Kauer JA, Perkel DJ, Mauk MD, Kelly PT, Nicoll RA, Waxham MN (1989) An essential role for postsynaptic calmodulin and protein kinase activity in long-term potentiation. Nature 340:554 557. Malinow R, Malenka RC (2002) AMPA receptor trafficking and synaptic plasticity. Annu Rev Neurosci 25:103126. Malinow R, Tsien RW (1990) Presynaptic enhancement shown by wholecell recordings of long-term potentiation in hippocampal slices. Nature 346:177180. ODell TJ, Hawkins RD, Kandel ER, Arancio O (1991) Tests of the roles of two diffusible substances in long-term potentiation: evidence for nitric oxide as a possible early retrograde messenger. Proc Natl Acad Sci U S A 88:1128511289. ODell TJ, Huang PL, Dawson TM, Dinerman JL, Snyder SH, Kandel ER,

J. Neurosci., December 24, 2008 28(52):Fishman MC (1994) Endothelial NOS and the blockade of LTP by NOS inhibitors in mice lacking neuronal NOS. Science 265:542546. Raymond CR (2008) Different requirements for action potentials in the induction of different forms of long-term potentiation. J Physiol 586:1859 1865. Sanderson DJ, Good MA, Seeburg PH, Sprengel R, Rawlins JN, Bannerman DM (2008) The role of the GluR-A (GluR1) AMPA receptor subunit in learning and memory. Prog Brain Res 169:159 178. Schmitt WB, Deacon RM, Seeburg PH, Rawlins JN, Bannerman DM (2003) A within-subjects, within-task demonstration of intact spatial reference memory and impaired spatial working memory in glutamate receptor-Adeficient mice. J Neurosci 23:39533959. Schmitt WB, Sprengel R, Mack V, Draft RW, Seeburg PH, Deacon RM, Rawlins JN, Bannerman DM (2005) Restoration of spatial working memory by genetic rescue of GluR-A-deficient mice. Nat Neurosci 8:270 272. Schulz PE, Cook EP, Johnston D (1994) Changes in paired-pulse facilitation suggest presynaptic involvement in long-term potentiation. J Neurosci 14:53255337. Son H, Hawkins RD, Martin K, Kiebler M, Huang PL, Fishman MC, Kandel ER (1996) Long-term potentiation is reduced in mice that are doubly mutant in endothelial and neuronal nitric oxide synthase. Cell 87:10151023. Stanton PK, Winterer J, Zhang XL, Muller W (2005) Imaging LTP of pre synaptic release of FMfrom the rapidly recycling vesicle pool of Schaffer collateral-CA1 synapses in rat hippocampal slices. Eur J Neurosci 22:24512461. Wright N, Glazewski S, Hardingham N, Phillips K, Pervolaraki E, Fox K (2008) Laminar analysis of the role played by GluR1 in experiencedependent and synaptic depression of sensory responses in barrel cortex. Nat Neurosci 11:1140 1142. Zamanillo D, Sprengel R, Hvalby O, Jensen V, Burnashev N, Rozov A, Kaiser KM, Koster HJ, Borchardt T, Worley P, Lubke J, Frotscher M, Kelly PH, Sommer B, Andersen P, Seeburg PH, Sakmann B (1999) Importance of AMPA receptors for hippocampal synaptic plasticity but not for spatial learning. Science 284:18051811.