INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Multiple training trials are generally more effective for the production of robust learning and memory when they are spaced apart rather than when they are massed together, a phenomenon known as the trial spacing effect (Ebbinghaus 1885; Hintzman 1974; Lattal 1999; Rescorla 1988). Although the trial spacing effect has been studied extensively at the behavioral level, its underlying molecular mechanisms remain unclear (Eichenbaum 1997). In the fruit fly, Drosophila melanogaster, Tully et al. (1994) examined the protein synthesis requirements of long-term memory for conditioned odor avoidance following massed and spaced training. Flies trained with a temporally spaced protocol displayed improved long-term memory compared with flies trained with a temporally massed protocol, and this improvement was blocked by pharmacological inhibition of protein synthesis. Induction of an inhibitory cyclic AMP response element binding protein (CREB) transgene also blocked the improvement of long-term memory seen following spaced training (Yin et al. 1994). Furthermore, induction of an activator isoform of CREB allowed for effective long-term memory induction even after massed training (Yin et al. 1995). Thus the enhanced long-term memory in Drosophila following spaced training for conditioned odor avoidance appears to result from the selective recruitment of a CREB-modulated protein-synthesis-dependent pathway that was less effectively recruited by massed training.

Numerous studies have demonstrated a role for the mammalian hippocampus in certain types of learning and memory. Long-term memory for hippocampus-dependent contextual fear conditioning is critically dependent on protein synthesis (Abel et al. 1997; Bourtchouladze et al. 1998; Schafe et al. 1999). Similarly, in hippocampal area CA1, maintenance of the late phase of hippocampal long-term potentiation (L-LTP), a type of activity-dependent synaptic plasticity that may importantly regulate the expression of certain forms of long-term memory in the mammalian brain (Abel et al. 1997; see reviews by Bliss and Collingridge 1993; Martin et al. 2000), is also dependent on protein and RNA synthesis (Frey et al. 1988; Nguyen et al. 1994). Maintenance of L-LTP in hippocampal slices and in intact rodents may correlate with the persistence of some types of long-term memory (Abel et al. 1997; Barnes 1979; Doyere and Laroche 1992; Jones et al. 2001). However, to date, there exists no conjoint examination of the roles of protein synthesis in the expression of the trial spacing effect in behavioral long-term memory and L-LTP in rodents. Furthermore, few experiments have addressed the trial spacing effect at the cellular electrophysiological level by examining synaptic plasticity (but see Wu et al. 2001). Specifically, does the temporal spacing of behavioral training and synaptic stimulation critically regulate the expression of long-term memory and L-LTP in a protein-synthesis-dependent fashion? In mice lacking the alpha and delta isoforms of CREB, spaced training selectively rescues long-term memory (Kogan et al. 1997), suggesting that the molecular basis of the trial spacing effect may differ between Drosophila and mice. Because CREB-mediated transcription would give rise to those transcripts whose translation may mediate long-term memory storage, the observation that spaced training rescues the long-term memory deficits found in CREB mutant mice suggests that the temporal profile of behavioral training is important for engaging protein-synthesis-dependent mechanisms that are critical for expression of long-term memory.

Further, studies of the trial spacing effect in genetically modified mice examined tasks that were dependent on hippocampal function such as contextual fear conditioning and the Morris water maze (Kogan et al. 1997). In the present study, we examine the trial spacing effect on tasks that are dependent on the amygdala and the hippocampus. Both contextual and cued fear conditioning are sensitive to lesions of the amygdala, whereas only contextual fear conditioning is sensitive to hippocampal lesions (Holland and Bouton 1999; LeDoux 2000). Thus behavioral paradigms using contextual and cued fear conditioning can allow for the identification of aspects of the trial spacing effect that are hippocampus specific.

In the present study, we tested the hypotheses that altering the temporal spacing of training and of synaptic stimulation leads to improved L-LTP and enhanced long-term memory in mice and that these improvements in synaptic plasticity and memory require protein synthesis. Our data show that temporally spaced synaptic stimulation in area CA1 of mouse hippocampal slices induced larger L-LTP than massed stimulation, and this enhancement was protein synthesis dependent. In parallel, temporally spaced training protocols elicited greater long-term memory for contextual (but not for cued) fear conditioning. This facilitative effect of temporal spacing on long-term memory for contextual fear was also blocked by pharmacological inhibition of protein synthesis. Our findings suggest that temporally spaced regimens more effectively recruit protein-synthesis-dependent pathways in the hippocampus, thus leading to more robust L-LTP and enhanced hippocampus-dependent long-term memory for contextual fear conditioning.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Electrophysiology

Transverse hippocampal slices (400-µm thickness) were prepared from C57BL/6J male mice (aged 8-16 wk; Jackson Laboratories, Bar Harbor, ME), and the slices were maintained in an interface chamber at 28°C (Abel et al. 1997). Slices were perfused with standard artificial cerebrospinal fluid (ACSF, 1 ml/min flow rate) with ionic composition as described previously (Nguyen et al. 2000). After a 1.5-h recovery period, a glass microelectrode (2-4 M resistance, filled with ACSF) was used to record field excitatory postsynaptic potentials (fEPSPs) from stratum radiatum of area CA1 during extracellular stimulation of the Schaffer collateral pathway with a bipolar nickel-chromium electrode (130-µm diam, A-M Systems, Carlsborg, WA) that was also positioned in stratum radiatum. Stimulation intensity (0.08-ms pulse width) was adjusted to give 40% of maximum evoked fEPSP amplitude, and baseline fEPSPs were elicited once per minute at this test stimulus intensity. For some experiments, fEPSPs were measured during stimulation (once per minute at test intensity) of a second independent pathway in stratum radiatum that did not receive high-frequency tetanic stimulation (Nguyen et al. 1994). L-LTP was induced by applying four 100-Hz bursts (1-s duration) separated by 5 min ("spaced" tetra-burst stimulation) or 20 s ("massed" tetra-burst stimulation) after an initial 20-min baseline period. Initial slopes of fEPSPs were measured and used as an index of synaptic strength. Data were analyzed with Excel (Microsoft) and GB-Stat (V 6.5, Dynamic Microsystems) and are presented as means ± SE. For statistical comparisons, data were analyzed every 5 min. A three-way ANOVA was used, and post hoc comparisons were made with a Newman-Keuls test.

Anisomycin (Sigma, St. Louis MO) was prepared as a concentrated stock solution in DMSO, and it was diluted in ACSF to obtain a final concentration of 25 µM. The final concentration of DMSO did not exceed 0.01%. Application of DMSO (0.01%) alone did not have any significant effect on basal synaptic transmission or on L-LTP (data not shown). Anisomycin was bath-applied for the entire 20-min pretetanization period and for 15 min following L-LTP induction. To ensure that slices were treated with anisomycin for identical periods of time, we terminated bath application of anisomycin immediately after high-frequency stimulation was completed for spaced tetraburst stimulation (100-Hz, 1-s duration, repeated 4 times at 5-min interburst intervals), whereas for massed tetraburst stimulation (same as spaced regimen, except for interburst interval, which was 20 s), drug application was stopped 15 min after the end of high-frequency stimulation.

Fear conditioning

We used experimentally naïve C57BL/6 male and female mice (ages 8-16 wk) that were bred in our colony from mice originally obtained from Jackson Laboratories. For behavioral studies, no gender differences were observed so that data from male and female mice were combined. Mice were maintained and bred under standard conditions consistent with National Institutes of Health guidelines for animal care and approved by the Institutional Animal Care and Use Committee of the University of Pennsylvania. Mice were maintained on a 12:12 light:dark cycle with lights on at 7 AM. Animals had free access to food and water in their home cage. All experiments were conducted during the light period between the hours of 9 AM and 5 PM.

On the training day, a mouse was placed in the conditioning chamber (Med Associates, Georgia, VT) for 2 min before the onset of the conditioned stimulus (CS; white noise at 85-dB intensity) that lasted for 30 s. The last 2 s of the CS were paired with a single unconditioned stimulus (US; 1.5-mA footshock for 2 s) with the noise and shock coterminating. The mouse then remained in the chamber for 30 s. Mice in both the massed and spaced training groups were then placed in their home cages between trials. All mice were returned to the conditioning chamber, and they were given two more training sessions identical to the first training session for a total of three training sessions. For the massed group, the intershock interval was 3.5 min, and for the spaced group, the intershock interval was 1 h. Thus the massed and spaced groups differed only in the time elapsed between training sessions.

Conditioning was assessed by measuring freezing behavior, defined as complete lack of movement except for respiration, in intervals of 5 s (Abel et al. 1997). Testing for contextual fear conditioning occurred 24 h after training. Contextual conditioning was assessed for five consecutive minutes in the chamber in which the mice were trained. Cued conditioning was assessed by placing the mice in a novel context that differed in color, shape, and odor for 2 min (pre-CS test) after which they were exposed to the CS for 3 min (CS test). Our studies have shown that the order of testing for contextual and cued fear conditioning does not affect long-term memory or the sensitivity of long-term memory to anisomycin (data not shown). Therefore we tested the mice for cued fear conditioning at 25 h, 1 h following the test for contextual fear. The freezing values observed for the experiments in this study were well within limits that we have observed in other experiments carried out in our laboratory. The freezing values were not excessively high or low. It therefore appears unlikely that our results can be interpreted on the basis of floor or ceiling effects.

For behavioral experiments, anisomycin (Sigma) was dissolved in phosphate-buffered saline, and the pH was adjusted with 1 N HCl to 7.4. Previous experiments demonstrated that subcutaneously injected anisomycin selectively impairs long-term memory for fear conditioning without altering short-term memory in mice (Abel et al. 1997; Bourtchouladze et al. 1998; Davis and Squire 1984). Furthermore, similar impairments of long-term memory are observed whether injections of anisomycin are administered immediately pretraining or immediately posttraining (Abel et al. 1997; Bourtchouladze et al. 1998) suggesting that the suppression of freezing observed 24 h after training in mice given pretraining injections of anisomycin is not due to state-dependent effects of anisomycin at the time of training.

Mice were subcutaneously injected with either 150 mg of anisomycin/kg of body weight or an equivalent volume of saline. Mice were given a single injection immediately before the first training session. At this dosage, anisomycin inhibits cerebral protein synthesis in mice by >90% for at least 2 h after injection (Flood et al. 1973), thus inhibiting protein synthesis throughout training for both massed and spaced protocols. Our experiments have shown that mice injected with anisomycin 2 h prior to training with three CS-US pairings showed impairments in long-term memory for contextual fear conditioning (Abel and Scharf, unpublished observations).

Different groups of mice (massed and spaced, drug-treated and control) were trained and tested in a balanced manner. Experimenters were unaware of treatment at the time of testing. Data were analyzed with Excel (Microsoft) and GB-Stat (V 6.5, Dynamic Microsystems). Two-way analyses of variance (ANOVA) were used and post hoc comparisons were made using a Newman-Keuls test. Data are presented as means ± SE.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Protein synthesis dependent enhancement of L-LTP by temporally spaced synaptic stimulation

To investigate the role of temporal spacing in regulating the expression of L-LTP in the hippocampus, we explored the protein synthesis dependence of CA1 L-LTP induced by four 1-s, 100-Hz trains spaced either 5-min apart (spaced stimulation) or 20-s apart (massed stimulation). In each of these stimulation protocols, slices received the same total number of stimuli (400) at an identical intra-burst frequency of 100 Hz. We varied only the time interval between consecutive bursts of stimuli. We kept the total amount of imposed activity constant because different durations and frequencies of stimulation can induce LTP or long-term depression (LTD) (see Bienenstock et al. 1982; Mayford et al. 1995). Also, we selected 5-min and 20-s interburst intervals for our spaced and massed tetraburst protocols, respectively, because these temporal patterns of stimulation have been shown to elicit long-lasting forms of L-LTP in area CA1 of C57BL/6J mouse hippocampal slices (Abel et al. 1997; Bach et al. 1999; Nguyen et al. 2000; Woo et al. 2000a). However, it should be noted that the protein synthesis dependence of L-LTP induced in mouse hippocampal slices by these two particular temporal patterns of stimulation has not been assessed and compared with each other.

Long-lasting L-LTP was elicited following either spaced or massed stimulation. An ANOVA demonstrated a main effect of temporal spacing [F(1,599) = 6.4; P < 0.05], indicating that spaced stimulation elicited greater LTP than massed stimulation (Fig. 1). This temporal spacing enhancement was still intact 2 h posttetanus (272 ± 22% for spaced stimulation and 200 ± 19% for massed stimulation, P < 0.05, Figs. 1 and 2). Thus spaced tetraburst stimulation induced more robust potentiation of mean fEPSP slopes in area CA1 of mouse hippocampal slices than massed stimulation, providing a demonstration of the trial spacing effect on the cellular level.

View larger version (32K): [in this window] [in a new window]   Fig. 1. Long-term potentiation (LTP) following spaced tetanic stimulation is more sensitive to anisomycin than LTP following massed tetanic stimulaton. A: spaced tetanic stimulation induced robust LTP that was attenuated by anisomycin. Slices were stimulated using four 100-Hz bursts (1-s duration) at 5-min interburst intervals. Anisomycin substantially impaired LTP, with mean field excitatory postsynaptic potential (fEPSP) slopes reaching pretetanization baseline levels at approximately 1 h after high-frequency tetraburst stimulation. Anisomycin did not affect transmission in a 2nd adjacent pathway that did not experience tetraburst stimulation. n represents the number of mice followed by the total number of slices. B: massed tetanic stimulation induced LTP that was impaired by anisomycin. Slices were stimulated using 4 100-Hz bursts (each lasting 1 s) at 20-s interburst intervals. Anisomycin significantly reduced mean fEPSP slopes beginning at approximately 90 min posttetanus. Anisomycin did not affect transmission in a 2nd adjacent pathway that did not experience tetraburst stimulation. n represents the number of mice followed by the total number of slices.

View larger version (11K): [in this window] [in a new window]   Fig. 2. Larger late phase of hippocampal LTP (L-LTP) was observed following spaced tetanic stimulation than after massed tetanic stimulation. Two hours following tetanic stimulation, the spaced group (n = 9) showed a greater mean fEPSP slope than the massed group (n = 7; P < 0.05; compare ) and anisomycin significantly reduced mean fEPSP slopes following both massed (n = 6) and spaced (n = 7) tetanic stimulation (P < 0.01). At this 2-h time point, there was no significant difference between the 2 drug-treated groups (P > 0.05; compare ).

Is L-LTP induced by spaced and massed stimulation differentially sensitive to protein synthesis inhibition? An ANOVA revealed a reliable main effect of anisomycin [F(1,599) = 600.3; P < 0.0001] and an interaction between anisomycin and temporal spacing [F(1,599) = 201.9; P < 0.0001]. These results suggest that although L-LTP elicited by either spaced or massed stimulation was impaired by anisomycin, L-LTP was more sensitive to protein synthesis inhibition following spaced stimulation than massed stimulation. One hour after the end of spaced stimulation, mean fEPSP slopes were 276 ± 20 and 110 ± 15% for control and anisomycin-treated groups, respectively (Fig. 1A). The drug-treated mean fEPSP slope value in the spaced group was comparable to pretetanus baseline values (Fig. 1A, ). In contrast, 1 h after the end of massed stimulation, mean fEPSP slopes were 206 ± 14 and 165 ± 19% for control and anisomycin treatments, respectively (Fig. 1B,  and ). Thus 1 h after the end of tetraburst stimulation, anisomycin caused mean fEPSP slopes that were initially potentiated by spaced stimulation to return to pretetanus baseline levels, whereas fEPSP slopes potentiated by massed stimulation in the presence of anisomycin were not reliably different from controls (P > 0.05). Two hours after massed stimulation, fEPSP values were 200 ± 19 and 129 ± 13% for control and anisomycin treatments, respectively. Hence, spaced tetraburst stimulation in area CA1 of mouse hippocampal slices elicited a larger potentiation that was more poorly maintained in the presence of protein synthesis inhibition as compared with massed tetraburst stimulation. These data suggest that the maintenance of L-LTP induced by spaced tetraburst stimulation is more sensitive to disruption by inhibition of protein synthesis than the maintenance of L-LTP elicited by a more massed temporal pattern of tetraburst stimulation.

In summary, like L-LTP elicited by spaced stimulation, L-LTP induced by massed stimulation was dependent on protein synthesis. However, the time courses of protein synthesis dependence were different between the massed and spaced groups. L-LTP elicited by massed stimulation in the presence of anisomycin approached pretetanus baseline levels at approximately 2 h after the end of massed tetraburst stimulation (Fig. 1B). In contrast, temporally spaced tetanic stimulation produced a form of L-LTP that was more sensitive to attenuation by protein synthesis inhibition (Fig. 1A): mean fEPSP slopes in drug-treated slices were at pretetanus baseline values at approximately 1 h after the end of spaced tetraburst stimulation. Also, at 2 h posttetanus, spaced stimulation induced significantly larger L-LTP in the control group than in the massed control group (Fig. 2, compare , P < 0.05). These data suggest that the improvement of L-LTP observed by increasing the temporal spacing of synaptic stimulation is dependent on protein synthesis.

Increased temporal spacing between training trials enhances hippocampus-dependent long-term memory in a manner that requires protein synthesis

Although some studies have examined the trial spacing effect at the behavioral level (Kogan et al. 1997; Tully et al. 1994), it is unclear whether altering the temporal spacing of behavioral training in mice can improve long-term memory in a manner that requires protein-synthesis-dependent processes. Contextual and cued fear conditioning are two forms of associative learning that allow for effective memory induction after a single training session, making them ideally suited for experiments aimed at examining the protein synthesis dependence of the trial spacing effect. We asked the following questions: is spaced training more effective for inducing long-term memory than massed training, and if so, does the facilitative effect of spaced training on long-term memory require protein synthesis?

We trained mice with one series of three, single-shock trials, with an intershock interval of 3.5 min (massed) or 1 h (spaced). In between trials, mice were removed from the training chamber and returned to their home cages. Thus all animals were given identical training trials, with the same amounts of exposure to the training context and to the conditioning stimulus (CS), along with identical handling. As such, spaced and massed training groups differed only in the time (i.e., temporal spacing) imposed between training trials.

We found that temporally spaced trials improved long-term memory for contextual fear conditioning, but not cued fear conditioning (Fig. 3). For contextual fear conditioning, a two-way ANOVA revealed a reliable main effect of anisomycin on long-term memory [F(1,52) = 50.4; P < 0.0001], a reliable main effect of temporal spacing [F(1,52) = 9.5; P < 0.01], and an interaction between anisomycin and temporal spacing [F(1,52) = 5.6; P < 0.05; n = 14 for all groups]. The spaced anisomycin-treated group showed significantly less contextual freezing than the spaced control group (7 ± 3 and 35 ± 5%, respectively; P < 0.01; Fig. 3A), and the massed anisomycin-treated group displayed significantly less contextual freezing than the massed control group (5 ± 1 and 19 ± 3%, respectively; P < 0.01; Fig. 3A). More importantly, the spaced control group showed significantly greater contextual freezing than the massed control group (compare  in Fig. 3A; P < 0.01), and this enhancement of long-term memory for contextual fear was blocked by anisomycin (compare  of Fig. 3A; P > 0.1 for comparison between drug-treated massed and drug-treated spaced groups).

View larger version (13K): [in this window] [in a new window]   Fig. 3. Temporal spacing improves long-term memory for contextual but not for cued fear conditioning. A: in control groups, spaced training induced more robust long-term memory than massed training for contextual fear conditioning (compare ). Inhibition of protein synthesis by anisomycin impaired expression of long-term memory elicited by both massed and spaced training (). Note that anisomycin also blocked the enhancement of long-term memory by spaced training, as compared with levels of freezing achieved following massed training. Thus the 2 anisomycin-treated groups did not differ significantly. n = 14 for each group. B: spaced training and massed training induced similar levels of long-term memory for cued fear conditioning (compare ). Anisomycin disrupted memory following both temporally massed and spaced training (). n = 14 for each group.

For cued fear conditioning, a two-way ANOVA revealed a reliable main effect of anisomycin [F(1,52) = 24.9; P < 0.0001], an unreliable main effect of temporal spacing [F(1,52) = 0.2; P > 0.05], and no interaction between anisomycin and temporal spacing [F(1,52) = 0.2; P > 0.05; n = 14 for all groups]. The spaced anisomycin group displayed significantly less CS-evoked freezing than the spaced control group (16 ± 4 and 37 ± 5%, respectively; P < 0.01), and the massed anisomycin group displayed significantly less CS-evoked freezing than the massed control group (12 ± 3 and 37 ± 6%, respectively; P < 0.01). Pre-CS values were 0.3 ± 0.3 and 6 ± 2% for the temporally spaced anisomycin and control groups, respectively, and the pre-CS values were 0.0 ± 0.0 and 0.0 ± 0.0% for the temporally massed anisomycin and control groups, respectively. The levels of freezing in the two drug-treated groups ( of Fig. 3B) were not significantly different from each other (P > 0.1). Also, more importantly, the spaced control group did not show significantly greater CS-evoked freezing than the massed control group (P > 0.05; compare  in Fig. 3B). Thus in contrast to contextual fear conditioning, long-term memory for cued fear conditioning was not enhanced by increasing the temporal spacing between training sessions.

In summary, for contextual fear conditioning, increasing the temporal spacing between training trials significantly improved long-term memory in the control group (compare  in Fig. 3A). However, increased intertrial temporal spacing did not significantly improve long-term memory in the anisomycin-treated groups (compare  in Fig. 3A). These data suggest, in a striking parallel to our L-LTP experiments, that increased intertrial spacing improves long-term memory in a manner that is dependent on protein synthesis. Furthermore, because contextual and cued fear conditioning are sensitive to lesions of the amygdala, and contextual fear conditioning is also sensitive to hippocampal lesions (reviewed in Holland and Bouton 1999; LeDoux 2000), our observation that increased temporal spacing improved long-term memory for contextual, but not for cued, fear conditioning suggests that this improvement of long-term memory for contextual fear is largely, but not necessarily exclusively, mediated by the hippocampus.

Overall, our electrophysiological experiments on L-LTP in hippocampal slices and our behavioral data from fear conditioning of mice suggest that protein synthesis is required for both the improvement of long-term memory and the enhanced maintenance of L-LTP seen following temporally spaced patterns of behavioral training and synaptic stimulation. Our findings suggest that a temporally spaced regimen more effectively recruits protein-synthesis-dependent pathways in the hippocampus than a more temporally massed protocol, thus leading to more robust maintenance of L-LTP and enhanced hippocampus-dependent long-term memory for contextual fear conditioning.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Our experiments suggest that temporally spaced patterns of synaptic stimulation and behavioral training improve L-LTP and enhance long-term memory by more effectively recruiting protein-synthesis-dependent mechanisms. Ours is the first study to conjointly examine the protein synthesis requirements of spaced and massed patterns of behavioral training and synaptic stimulation at both the behavioral and cellular electrophysiological levels in mice. We found that spaced stimulation was more effective in inducing larger and longer-lasting L-LTP than massed stimulation and that this improvement was protein synthesis dependent. Similarly, in our behavioral studies, increased temporal spacing improved long-term memory for contextual fear conditioning in a protein-synthesis-dependent manner. The spaced protocol used here improved long-term memory for contextual but not for cued fear conditioning. This suggests that this improvement is largely mediated by the hippocampus. Our findings, therefore, constitute behavioral and cellular electrophysiological evidence that temporal spacing plays a key role in modulating hippocampal memory function and synaptic plasticity and that temporally spaced training protocols for contextual fear conditioning may be well-suited for elucidating the roles of hippocampal information processing in specific types of long-term memory. Furthermore, although we do not know whether L-LTP induced by spaced stimulation directly contributes to the improved long-term memory seen after spaced contextual fear conditioning, our results suggest an intriguing correlation between these synaptic and cognitive processes. Our studies also highlight a parallel between the enhanced maintenance of L-LTP and the improved long-term memory following spaced stimulation or training: both require protein synthesis.

The different intertrial time intervals used for our behavioral and electrophysiological experiments were necessitated by practical considerations. However, the intertrial interval for spaced behavioral training was approximately 17-fold longer than the intertrial interval for massed training. The interburst interval for spaced synaptic stimulation was 15-fold longer than the interburst interval for massed synaptic stimulation. Thus the spaced-to-massed intertrial and interburst ratios were similar. Would varying these temporal ratios modulate the degree, and the protein synthesis-dependence, of the enhancement of L-LTP and long-term memory in mice? Further work is needed to resolve this question and to clearly define the actual patterns of spike activity that occur in the hippocampus during different intertrial training intervals.

Our results are consistent with the idea that the protein-synthesis-dependent molecular machinery underlying long-term memory storage following contextual fear conditioning is saturated rapidly (Eichenbaum 1997) and thus additional training trials administered within a very short period of time (massed trials) may be of limited benefit. By spacing trials farther apart, additional intertrial time allows for the optimal recruitment of protein-synthesis-dependent molecular pathways that ultimately lead to the emergence of improved long-term memory (Tully et al. 1994; Yin et al. 1994).

When administered around the time of training, protein-synthesis inhibition induced by anisomycin selectively impaired long-term memory in mice, whereas immediate learning and short-term memory were spared (Abel et al. 1997; Davis and Squire 1984). An anisomycin dosage of 150 mg/kg, as used in the present study, is known to cause more than 90% inhibition of protein synthesis within the brain for at least 2 h postinjection (Flood et al. 1973). Notably, mice given a spaced training protocol, in which the last training session was carried out 2 h postinjection, showed marked deficits in contextual and cued fear conditioning. To date, levels of protein synthesis inhibition caused by systemic administration of anisomycin in different regions of the brain have not been measured. Additionally, while it is widely suspected that protein synthesis is necessary in the hippocampus, the effect of protein synthesis inhibition in the hippocampus per se on contextual fear conditioning has not been assessed. Our data do show, however, that increased temporal spacing improves long-term memory for contextual, but not for cued fear conditioning, and that this improvement requires protein synthesis. This suggests that the improvement of long-term memory following spaced training may result from the recruitment of protein-synthesis-dependent pathways in the hippocampus.

The transcription factor CREB has been identified as a potential modulator of long-term memory. Studies by Yin et al. (1994, 1995) in Drosophila have implicated CREB as a "molecular switch" responsible for regulating expression of the protein-synthesis-dependent component of long-term memory. These authors suggest that spaced training is particularly effective in inducing the protein-synthesis-dependent component of long-lasting memory because of differential deactivation rates of CREB repressor and activator isoforms, although these rates of deactivation have not been measured. In knockout mice lacking the alpha and delta isoforms of CREB, deficits in long-term memory for contextual and cued fear conditioning were evident following a single training trial (Bourtchouladze et al. 1994). When trained with multiple training trials, these CREB mutant mice showed long-term memory deficits following massed, but not spaced, training for contextual fear conditioning (Kogan et al. 1997). These experiments, however, did not use equal numbers of training sessions between the massed and spaced groups, and handling between trials was not identical in the two groups. Therefore those experiments did not directly address the issue of trial spacing. Additionally, genetic background appeared to be a determinant of whether CREB mutant mice displayed memory deficits (Gass et al. 1998; Graves et al. 2002). Furthermore, 15-20% of CREB binding activity is intact in these mice (J. A. Blendy, unpublished observations), and levels of other transcription factors are altered (Blendy et al. 1996). Thus although CREB may be involved in modulating the trial spacing effect in mice, defining its exact role will require the development of genetically modified mouse strains in which CREB function is more substantially impaired in a regionally and temporally controlled fashion (Gass et al. 1998; Pittenger et al. 2000; but see Rammes et al. 2000).

Our finding that increased intertrial spacing for contextual fear conditioning in mice leads to more effective memory induction that is dependent on protein synthesis parallels the finding in Drosophila that improved long-term memory observed 1 day after spaced training is attributable to the recruitment of protein-synthesis-dependent mechanisms (Tully et al. 1994). Furthermore, a similar parallel to the CREB dependence of spaced training in Drosophila is observed in rats trained in a fear-potentiated startle paradigm. Rats display enhanced long-term memory for this task following spaced training, and this enhancement can be artificially mimicked following massed training by overexpression of CREB in the amygdala (Josselyn et al. 2001). Thus there appears to be an evolutionary conservation of CREB and protein-synthesis-dependent mechanisms responsible for regulating the expression of some forms of learning and memory, and these mechanisms are critically regulated by the temporal spacing between behavioral training sessions.

Recent studies have indicated that cAMP-dependent protein kinase (PKA) is critical for long-term memory and L-LTP in area CA1 in mice (Abel et al. 1997) and that CREB phosphorylation is mediated by PKA and by mitogen-activated protein kinases (MAPK) (Impey et al. 1998; Roberson et al. 1999). Genetically modified mice that express an inhibitory form of a regulatory subunit of PKA in hippocampal neurons showed impairments in CA1 L-LTP elicited by spaced stimulation (Abel et al. 1997), but L-LTP elicited by massed theta-burst stimulation was normal in these mice (Woo et al. 2000a,b). These PKA mutant mice also showed impairments in long-term memory for contextual fear conditioning following spaced, but not massed, training (Woo et al. 2000b). MAPK also plays a critical role in long-term memory (Atkins et al. 1998) and LTP (English and Sweatt 1997). Spaced, but not massed, stimulation elicited persistent MAPK activation and the protrusion of new dendritic filipodia in cultured hippocampal neurons (Wu et al. 2001). Our results do not directly show that protein synthesis is differentially modulated by trial spacing, and they do not identify which specific subcellular processes are engaged more effectively following spaced trials. In particular, it is not known whether trial spacing can effectively modulate the efficacies of the PKA and MAPK pathways in mice. Nonetheless, given our data showing that temporal spacing leads to improved hippocampal long-term memory and L-LTP by recruitment of protein-synthesis-dependent mechanisms, the apparent sensitivity of PKA and MAPK pathways to trial spacing suggests that the regulation of these signaling cascades may mediate the trial spacing effect by critically modulating gene induction and protein synthesis.