Relevance of synaptic tagging and capture to the persistence of long-term potentiation and everyday spatial memory

Abstract

Memory for inconsequential events fades, unless these happen before or after other novel or surprising events. However, our understanding of the neurobiological mechanisms of novelty-enhanced memory persistence is mainly restricted to aversive or fear-associated memories. We now outline an “everyday appetitive” behavioral model to examine whether and how unrelated novelty facilitates the persistence of spatial memory coupled to parallel electrophysiological studies of the persistence of long-term potentiation (LTP). Across successive days, rats were given one trial per day to find food in different places and later had to recall that day's location. This task is both hippocampus and NMDA receptor dependent. First, encoding with low reward induced place memory that decayed over 24 h; in parallel, weak tetanization of CA1 synapses in brain slices induced early-LTP fading to baseline. Second, novelty exploration scheduled 30 min after this weak encoding resulted in persistent place memory; similarly, strong tetanization—analogous to novelty—both induced late-LTP and rescued early- into late-LTP on an independent but convergent pathway. Third, hippocampal dopamine D1/D5 receptor blockade or protein synthesis inhibition within 15 min of exploration prevented persistent place memory and blocked late-LTP. Fourth, symmetrically, when spatial memory was encoded using strong reward, this memory persisted for 24 h unless encoding occurred under hippocampal D1/D5 receptor blockade. Novelty exploration before this encoding rescued the drug-induced memory impairment. Parallel effects were observed in LTP. These findings can be explained by the synaptic tagging and capture hypothesis.

Neurobiological theories of long-term memory (LTM) assert that strong events are remembered better than weak because they alone trigger “consolidation.” However, the memory of apparently unimportant things is an intriguing challenge to these accounts—particularly when these occur in association with surprising or emotionally significant events. The synaptic tagging and capture (STC) hypothesis of protein synthesis-dependent long-term potentiation (1, 2) may offer an explanation of this associative process, based on the idea that the neural mechanisms of initial long-term potentiation (LTP) expression (potentiation and tagging) can be dissociated from those regulating the availability of plasticity-related proteins (PRPs) that stabilize synaptic change. Thus, weakly induced LTP that is normally transient is sustained because PRPs associated with strong LTP on a separate pathway are captured by the synaptic tags set on the weakly tetanized pathway. As synaptic plasticity may be one component of the neural mechanisms of information storage (3–6), the persistence of memory should parallel the persistence of synaptic potentiation (7).

An important characteristic of ‘everyday memory’ is that we retain incidental information within LTM for only a short period, rarely creating an enduring memory (e.g., where we parked our car when out shopping). To model this kind of episodic-like memory, we have developed an analogous one-trial spatial memory task in an ‘event arena’ that depends on synaptic transmission and plasticity in the dorsal hippocampus (8, 9). The protocol continues everyday for months, with a new spatial memory encoded and generally forgotten each day. This is a closer analogy to everyday memory in humans than many current behavioral tasks studied in animals, and is likely subserved by temporary engrams mediated by distributed associative potentiation in the hippocampus.

Interestingly, unrelated novelty or surprise may stabilize the persistence of memories, even for inconsequential events that are normally forgotten. An example of this is the curious halo of incidental memories surrounding “flashbulb memories” (10, 11), such as what happened to us on the occasion of momentous events such as the terrorist attacks in 2001 in the 9/11 tragedy. It has recently been reported that memory for inhibitory avoidance by rats can also persist longer if exploration in a novel environment occurs around the time of learning (12). Similar findings have been shown for contextual fear conditioning, spatial object recognition and taste memory (13, 14).

To further explore the relevance of STC, we conducted parallel behavioral and physiological studies. We occasionally scheduled brief unrelated novelty exploration to evaluate its impact on memory of individual subjects across successive conditions. Novelty exploration is known increase the firing of VTA dopamine neurons (15), hippocampal dopamine release (16) and the transcription of a number of activity-related genes (17) and so, like strong tetanization, may induce the PRPs that the STC hypothesis asserts as critical for memory persistence. The continuity of experience, memory and then forgetting through the day is similar to what happens with activity-dependent synaptic plasticity in vivo that is presumably triggered frequently (18) and for which protein synthesis-independent and protein synthesis-dependent mechanisms are variously engaged (19). This commonality is the basis of our conducting parallel studies.

Results

The rats were trained in an event arena to dig for food pellets that they carried back to the start boxes to eat (Fig. S1). The regular daily training consisted of a memory encoding trial and a later retrieval choice trial. In encoding, the start box door opened allowing entry into the arena containing a single sandwell (with one or three food pellets) at a changing place each day. This trial constituted an opportunity to incidentally encode where food was available on that day. In retrieval, 30–40 min after encoding, the same sandwell again contained food (three pellets), but there were now four other nonrewarded sandwells available (exacting controls for olfactory cues are given in SI Materials and Methods). The animal could use one-trial place memory from the earlier encoding trial to retrieve efficiently. The main study began once the animals were routinely making one error or fewer per retrieval trial. There followed a series of separate “conditions” conducted over 6 months consisting of analogous electrophysiological and behavioral components. Performance stability is shown in Fig. S2.

Condition 1: Decay of Early-LTP (E-LTP) and Weak Memory.

Electrophysiological studies of synaptic tagging and capture (STC) were conducted in hippocampal brain slices using the methods of Redondo et al. (20). We first examined the induction of LTP by weak tetanization in brain slices (Fig. 1 A and B). This caused LTP on an input pathway to CA1 at 30 min posttetanus that declined to baseline over 10 h with statistically significant decay over this period.

Fig. 1.

Time-dependent retention of synaptic potentiation and memory. (A) A weak tetanus (WTET) at one pathway (S1) elicits LTP at 30 min but not 10 h (n = 7). (B) Summary bar graph shows E-LTP, at 30 min and no L-LTP, at 10 h of this weakly tetanized pathway (open bars) and control pathway stability (filled bars) (WTET S1 vs. control at E-LTP, t₁₂ = 8.7, P < 0.01, at L-LTP, t₁₂ = 0.4, P > 0.05; WTET S1 E-LTP vs. L-LTP, t₆ = 3.9, P < 0.01). The asterisk indicates significant difference from control pathway (∼100%, P < 0.05); surrounding gray area indicates electrophysiology. (C) Rats dug in one open well (filled pink circle) for one food pellet followed by a nonrewarded probe trial with five open wells (open pink circles) at 30 min or 24 h (n = 16). (D) Spatial memory of one-pellet (1p) encoding at 30 min but not 24 h (correct > wrong digging at 30 min, t₁₅ = 4.5, P < 0.001; at 24 h, t₁₅ < 1, P > 0.5; 2 × 2 ANOVA, F_1,15 = 15.48, P < 0.005). The asterisk indicates significant differences between correct and wrong digging (P < 0.05). (E) Individual performance at 30 min and 24 h. Data are mean (±1 SEM in A, B, and D). Gray dashed line (in B and D) indicates the baseline or chance level.

The analogous “event arena” study evaluated retention, at 30 min or 24 h, of one-trial place memory rewarded by one food pellet only at encoding (weak encoding). We used a memory probe test in which five nonrewarded sandwells were presented for 60 s with time spent digging at each sandwell recorded. The order of the 30-min and the 24-h encoding-probe test pairs was counterbalanced, interleaved by 1 regular training day. This interleaved training, used throughout the study, was essential to sustain digging at the sandwells across days, as probe tests constituted extinction trials. There was good memory at 30 min but not at 24 h, with significant decay between 30 min and 24 h (Fig. 1D). Of note is the directly analogous “pattern” of the electrophysiological and behavioral bar graphs in Fig. 1 B and D. Given that this is a within-subjects procedure in which numerous tests are to be conducted on individual animals, Fig. 1E plots the performance of the 16 individual animals in each test.

Condition 2: Novelty-Enhanced Persistence of Weak Memory.

We then investigated whether unexpected novelty exploration would affect the persistence of memory. In the parallel electrophysiology study, we weakly tetanized one pathway (S1) and, 30 min later, applied strong tetanization (“novelty”) to a separate pathway, S2 (Fig. 2 A and B). This is theoretically appropriate as behavioral novelty and strong tetanization both up-regulate immediate early genes, necessary for PRP synthesis/availability; we return to a justification of this in the discussion. Strong tetanization not only led to 10-h LTP in S2, i.e., late-LTP (L-LTP), but also rescued the transient decay of E-LTP on S1, converting it to nondecaying L-LTP. This replicates a key phenomenon of STC (21).

Fig. 2.

The “weak-before-strong” phenomenon in synaptic potentiation and memory persistence. (A) WTET at one pathway (S1) elicited persistent LTP when a strong tetanus (STET) was applied to an independent pathway (S2) shortly afterward (n = 7). (B) Potentiation at 10 h shows no L-LTP for a weakly stimulated pathway (Left; from Fig. 1B, Right) but persistent L-LTP when WTET was coupled with STET (WTET S1 in Weak+Strong > Weak only: t₁₂ = 2.49, P < 0.05; In Weak+Strong, WTET S1 > control, t₁₂ = 3.08, P < 0.01). (C) Rats dug in one open well for one pellet. Exploration in a novel box (green grid square) 30 min after encoding. Nonrewarded probe trial with five open wells was conducted 24 h after encoding. (D) Spatial memory of one-pellet (1p) encoding, at 24-h delay, was enhanced by exploration in a novel box (No novel box: correct ∼ wrong digging, t₁₅ < 1, P > 0.5; with novel box: correct > wrong digging, t₁₅ = 4.35, P < 0.001; 2 × 2 ANOVA, F_1,15 = 6.94, P < 0.05). (E) Performance during probe tests after one-pellet encoding or one-pellet encoding + exploration. Data are means (±1 SEM in A, B, and D). Asterisks indicate significant differences between adjacent open and filled bars. Gray dashed line (in B and D) indicates baseline or chance level.

In the behavioral experiments, the same animals used in condition 1 (above) were given a weak encoding trial followed by a probe test 24 h later, with or without (in counterbalanced order) the opportunity for 5-min novelty exploration in a box placed within the arena 30 min after encoding (Fig. S1C and Fig. 2C). Without exploration, there was the usual indifferent place memory, whereas with exploration, memory was persistent (Fig. 2 D and E). Taken together, these findings show that strong tetanization/novelty can enhance the persistence of LTP/memory for an unrelated pathway/experience.

Condition 3: Timing, Location, and Measurement of Novelty Exploration.

To test whether there is a critical time window for novelty exploration, the 5-min exploration was either arranged to be 30 min after encoding or delayed to 6 h later. To maintain novelty, different substrates (e.g., mesh wire, small pebbles, etc) were placed on the floor of the box (Fig. S1C). Delaying the novelty by 6 h prevented the rescue of persistent memory for weakly encoded place memory at 24 h (Fig. S3A). Second, we found that exploration in a familiar context was not sufficient to induce the memory persistence (Fig. S3B). Third, we confirmed that the animals remembered the novel experience itself by showing locomotor habituation when a specific box and substrate was given twice (Fig. S3C). Fourth, we established that an allocentric spatial strategy was used to perform the task (Fig. S4).

Condition 4: Pharmacological Blockade of Novelty-Induced Persistence of Memory.

We then addressed whether novelty-induced persistence of memory is mechanistically similar to STC in brain slices. A key idea is that dopamine activation of D1/D5 receptors is critical for the PRP synthesis necessary for L-LTP (22), the origin of these fibers coactivated by the strong tetanus being the VTA the dopamine neurons of which can be activated by novelty (2, 15, 23). In brain slices (Fig. 3 A and B), we extend previous observations (24) with the finding that the rescue of persistent potentiation after weak tetanization of a pathway was not observed if subsequent strong tetanization was conducted in the presence of SCH23390. Thus, D1/D5 receptors are critical for induction of the persistence that characterizes L-LTP. In addition, we confirmed that anisomycin blocks L-LTP (Fig. 3 C and D). Maintenance of potentiation of the strongly tetanized pathway was impaired by anisomycin, but also there was no potentiation in the weakly tetanized pathway at 10 h.

Fig. 3.

The role of hippocampal D1/D5 receptors and protein synthesis in facilitation of memory persistence in the weak-before-strong model. (A) SCH23390 (10 μM) at STET of pathway S2 blocked S2 LTP persistence at 10 h. An independent pathway (S1) with WTET before both drug infusion and STET of S2 also failed to maintain S1 potentiation for 10 h (n = 6). (B) Summary bar graph shows L-LTP of a WTET (in Weak+Strong) pathway (from Fig. 2B, Right) and impaired L-LTP of the WTET (in Weak+Strong) pathway when STET was applied under SCH23390 (at 10 h: WTET S1 vs. control, t₁₀= 1.3, P = 0.22; WTET S1 in Weak+Strong alone vs. in Weak+Strong+SCH23390, t₁₁ = 2.26, P < 0.05). (C) Anisomycin (25 μM) present during STET of pathway S2 blocked LTP maintenance at 10 h. An independent pathway (S1) given WTET before both drug infusion and STET in S2 also failed to maintain LTP (n = 7). (D) Summary bar graphs shows L-LTP of a WTET pathway (in Weak+Strong, from Fig. 2B, Right) and impaired L-LTP of WTET pathway when STET to the other pathway was under anisomycin (at 10 h: WTET S1 vs. control, t₁₂ < 1, P = 0.68; WTET S1 in Weak+Strong alone vs. in Weak+Strong+Aniso, t₁₂ = 2.48, P < 0.05). (E and F) (Top) Summary of behavioral procedures. (E) Without novelty, hippocampal infusions of vehicle (Veh) or D1/D5 receptor antagonist, SCH23390 (1 μg/μL) did not affect performance (Left: similar chance level correct and wrong digging at 24 h, P > 0.5; 2 × 2 ANOVA, F_1,10 < 1). Exploration in a novel box 30 min after one-pellet encoding enhanced memory persistence (Midright: correct > wrong digging, t₁₀ = 4.05, P < 0.01). This was impaired by SCH23390 infusion before exploration (Right: 2 × 2 ANOVA F_1,10 = 9.95, P < 0.01). (F) Novelty exploration after one pellet encoding facilitated memory retention unless novelty was followed immediately by hippocampal anisomycin (125 μg/μL. Left: correct > wrong digging in Veh t₇ = 4.74, P < 0.01; in Ani, t₇ < 1, P > 0.5; 2 × 2 ANOVA, F_1,7 = 10.29, P < 0.01). In contrast, a 6-h–delayed anisomycin infusion did not block novelty-enhanced memory (Right: correct > wrong digging in Veh, t₇ = 2.66, P < 0.05; in Ani, t₇ = 3.82, P < 0.01; 2 × 2 ANOVA, F_1,7 < 1, P > 0.5). Data are mean ±1 SEM. Asterisks indicate significant differences between adjacent open and filled bars. Gray dashed line in B, D, E, and F indicates baseline or chance level.

The parallel behavioral studies continued with the now cannulas-implanted rats which, with further interspersed training days, were given weak place memory encoding days that were either with or without later novelty exploration conducted under the influence of SCH23390 or vehicle (n = 11, due to blocked cannulas in two rats). We observed no 24-h memory with weak encoding after either vehicle or SCH23390 infusions (Fig. 3E, Left). Persistent place memory over 24 h was seen when weak encoding was followed by novelty exploration, but this was blocked by intrahippocampal SCH23390 during exploration (Fig. 3E, Right). The behavioral tests conducted with anisomycin (n = 8) are presented here but, in practice, were done toward the end of all other experimentation because of the potential toxicity of this drug (Table S1). Intrahippocampal infusion of anisomycin immediately after novelty exploration blocked the rescue of 24-h memory by novelty after weak encoding of place memory (Fig. 3F, Left). Delaying the infusion by 6 h allowed rescue of 24-h memory to occur normally (Fig. 3F, Right). These findings suggest that the critical time window for the PRP synthesis and/or local availability is within 6 h of exploration.

Condition 5: Rescue of D1/D5-Dependent Memory Through Novelty Exploration.

In classical studies of sensitization in Aplysia, it has been well established that a single exposure to neuromodulation by 5HT induces short-term forms of presynaptic facilitation and behavioral sensitization, whereas multiple exposures result in long-term changes (25, 26). Similarly, multiple trials of training give rise to more persistent memory in many tasks, e.g., sensitization in Aplysia (27) and fear conditioning in rats (28). Changing to strong encoding in the event arena (three-pellet reward instead of one pellet) should induce 24-h memory, which it did in an NMDA- receptor–dependent manner (Fig. S5). The persistence of this strong memory encoding was sensitive to D1/D5 receptor blockade with a 3.3- but not a 1-μg/μL infusion of SCH23390 (Fig. 4D).

Fig. 4.

Novelty exploration rescues the memory impairment induced by hippocampal D1/D5 receptor blockade. (A) When strong tetanization (STET) of S2 occurred under SCH23390 after STET of another independent pathway (S1), L- LTP was present on both pathways (n = 5). (B) Summary at 10 h shows SCH23390 failed to block L-LTP in S2 when coupled with predrug STET in S1 (S2 > control in vehicle: t₈ = 4.1, P < 0.01 and SCH23390: t₈ = 6.5, P < 0.01; and no difference between conditions: t₈ = 0.8, P > 0.05). (C) Exploration sometimes scheduled 1 h before three-pellet encoding that during or without hippocampal D1/D5 receptor blockade. (D) Retention at 24 h of three-pellet encoding was impaired by a higher dose of SCH23390 (3.3 μg/μl) (correct = wrong digging: t₈ < 1, P > 0.3), but not 1 μg/μl (correct > wrong digging t₈ = 3.2, P 0.01) (2 × 3 ANOVA (correct/wrong digging by Veh/SCH low dose/SCH high dose) F_2,16 = 4.59, P < 0.05). (E) Novelty exploration 1 h before encoding completely rescued 24-h memory despite encoding during 3.3 μg/μl dose of SCH23390 (similar correct digging in Veh vs. SCH, F_1,8 < 1; correct > wrong digging; Veh: t₈ = 5.27, P < 0.001; SCH: t₈ = 4.94, P < 0.01). Data are means ± 1 SEM. Asterisks indicate significant differences between adjacent open and filled bars. The grey dashed line indicates baseline or chance level.

We then investigated whether this forgetting of strongly encoded place memory could be rescued by prior novelty. To avoid the drug having an effect on the downstream effects of novelty, it was necessary to present the novelty before the memory encoding (Fig. 4C). On the critical tests, the animals were given exploration of the novel box 1 h before “strong” place memory encoding with, 15 min before the memory-encoding trial, intrahippocampal infusions of SCH23390 (3.3 μg/μL) or vehicle. Memory was measured, as usual, with a nonrewarded probe test 24 h after encoding (Fig. 4C). Remarkably, memory over 24 h was not only observed after the preencoding vehicle infusions (Fig. 4E, Left), but also after preencoding SCH23390 infusions (Fig. 4E, Right). As the STC hypothesis uniquely predicts, dopamine-dependent memory encoded in the presence of D1/D5 receptor blockade is rescued by prior novelty.

An electrophysiological analogy is the frequent use of multiple trains of high-frequency tetanization in experiments on L-LTP. Such tetanization of one pathway in the presence of SCH23390 blocked L-LTP. However, when this was preceded by strong tetanization of second independent pathway before application of SCH23390 (Fig. 4 A and B), both pathways maintained LTP for 10 h. This “strong-before-strong” protocol indicates that SCH23390 blocks LTP maintenance by blocking the synthesis of PRPs rather than affecting the setting of synaptic tags (20, 24).

Condition 6: Meta-Analysis of Good, Poor, and Rescued Memory Conditions.

Our within-subjects design offers the unique opportunity to compare individual subjects across conditions and so to provide further tests of the relevance of the STC theory to memory. The entire study was conducted over 6 mo; we have noted that not all animals completed all procedures of the full study, but it was still possible to pool the probe test data into a “meta-analysis” that offered the following three categories: (i) the average performance of individual animals in that set of training protocols in which good memory was predicted, such as the 24-h memory of a strong, three-pellet encoding trial and 30-min memory of a weak, one-pellet trial; (ii) those in which poor memory was predicted, such as 24-h memory after weak encoding; and (iii) those in which memory rescue was predicted, such as 24-h memory after weak encoding followed by novelty exploration (full compilation of conditions in Table S2). We found that 14 of the 16 animals complied with our prediction in showing a good–poor–rescued “V-shaped” function (χ² = 9, P < 0.005; overall ANOVA of the three categories, F_2,30 = 22.73, P < 0.001; the quadratic component, reflecting this V-shape, was highly significant, F_1,15 = 49.88, P < 0.001) (Fig. 5). In addition, we analyzed whether there would be any correlation between the levels of performance of individual subjects and their sensitivity to novelty. Although the STC hypothesis makes no specific predictions, we nonetheless observed that the relative decline in performance between the good and poor categories was correlated with the relative increase in memory persistence between the poor and the rescued categories (Fig. S6).

Fig. 5.

Meta-analysis of individual animal data. Individual performance shows V-shaped pattern of 24-h memory across the good, poor, and rescued categories. Examples are as follows: strong encoding that renders memory persistence (good); weak encoding that renders forgetting (poor); and conditions in which novelty exploration enhances poor memory persistence (rescued). Of the 16 animals, 14 displayed a V-shaped function (described in text).

Discussion

Incidental memory occurs automatically in the course of our day-to-day stream of activities and is sometimes vital for their successful accomplishment. Although such inconsequential information enters LTM, there is little value in retaining it for any length time—leading to one of the beneficial “seven sins” of memory, namely, transience (29). However, if something surprising happens, not only do we remember the surprise itself, but it may induce (or appear to induce) better memory of surrounding events. We have developed an everyday, one-trial, allocentric spatial memory paradigm in rats, and have established that closely timed but unrelated novelty can extend the persistence of weakly encoded place memory. As the STC theory of L-LTP (1) offers a unique explanation of this remarkable phenomenon, it was essential to conduct parallel, in vitro electrophysiological studies mimicking our behavioral experiments. These confirmed the rescue of weakly induced LTP by strong tetanization of another pathway and that strong tetanization rescues persistent LTP induced on another pathway in the presence of SCH23390 (24), and revealed the findings that rescue of L-LTP on a weakly tetanized pathway by subsequent strong tetanization is sensitive to both SCH23390 and anisomycin. Our behavioral data directly parallel these LTP findings: (i) intrahippocampal blockade of D1/D5R receptors and inhibition of protein synthesis at the time of novelty exploration prevented novelty from enhancing memory persistence; (ii) strong encoding of the daily place memory enabled persistence over 24 h, and the blockade of 24 h memory by the D1/D5 antagonist SCH23390 could be rescued by novelty exploration shortly before training; and (iii) a meta-analysis of good, poor, and rescued memory revealed that the pattern shown by 14 of the 16 individual subjects complies with predictions of the STC hypothesis relevant to behavior. Our unusual use of long-time course experiments on LTP in vitro (most L-LTP studies end at ∼3 h) reveals parallels between the physiological and behavioral domains, and offers further support for the synaptic plasticity and memory hypothesis (5).

Behavioral Model of Everyday Memory.

At a time when striking observations in molecular cell biology are having an impact on neuroscience, it is important to recognize the value of innovative behavioral paradigms to realize a deep understanding of information processing in the brain. Standard tasks involving fear conditioning, inhibitory avoidance, and many others remain analytically useful, but they do not capture a key aspect of how memory functions with respect to the continuity of experience. The “discrete” nature of so many protocols has misled us into thinking there is an inexorable sequence of encoding, storage and then consolidation for all aspects of LTM. In our new event arena, the animal learns new information each day but does so against a background of other events happening before or after, such as unexpected novelty, a phenomenon analogous to inducing weak LTP around the same time as strong tetanization of independent afferents. The persistence of memory and of synaptic potentiation is dramatically affected. This protocol is unique in being characterized by the continuity of daily one-trial learning coupled to occasional neuromodulatory events.

One limitation of human studies of “flashbulb memory” has been the difficulty of establishing whether enhanced memory for associated inconsequential events is due to repeated recollection with friends (30) or to neural activity around the time of memory encoding. Because surprising events such as the 9/11 tragedy cannot be predicted, human studies are necessarily restricted to memory retrieval and cannot study the encoding process. Building an animal model of flashbulb memory is fraught with uncertain assumptions, but the value of even limited animal work is that prospective studies can be designed and thus memory encoding investigated. Our animal model, together with previous studies (12, 13), establishes that repeated memory retrieval is not necessary for this rescue of incidental memory in animals. It follows that neural activity at or around encoding is critical. Furthermore, our use of a within-subjects design with repeated experiences across days revealed a consistency in individual subjects across conditions in which good, poor, and rescued memory was predicted. This behavioral protocol will therefore be ideal, in the way that cross-sectional paradigms are not, for detailed analysis of underlying molecular events using inducible gene targeting or optogenetics (31, 32). We are presently adapting the event arena task for mice.

Theoretical Account in Terms of STC.

Novelty triggers dopamine neuromodulation of synaptic plasticity in the hippocampus (15, 33). The synergistic role of dopamine and NMDA receptors for persistent LTP points to the necessity of D1/D5 receptor activation for the availability of PRPs (23). Novelty exploration up-regulates immediate early genes (17, 34, 35), and consequently the synthesis and distribution of PRPs. Our results suggest that this time window is at least between 1 h before and 30 min after encoding, consistent with other studies (12).

The novelty-enhanced memory persistence echoes studies in freely moving animals revealing that novelty exploration during and after LTP induction promotes the persistence of synaptic potentiation (36–38). In addition, Li et al. have separately shown that novelty exploration lowers the threshold for LTP induction occurring 5 min later (39). However, the mechanisms for facilitating induction are likely to differ from those for enhancing persistence. This is because the facilitation of LTP induction by novelty fails to occur when novelty is presented 30 min before LTP induction but the enhancement of LTP persistence still occurs (38). Moreover, the effect of novelty on LTP persistence is also seen when novelty is scheduled after LTP induction (37).

Dopamine neuron activity has also been shown to be critical in several instances of reward learning (40). Previous work from our laboratory has shown that, whereas LTM is impaired, short-term spatial memory (20–30 min) is unaffected by hippocampal D1/D5 receptor blockade during encoding in both the watermaze (41) and event arena (42). Our data from the current study add to this in two ways: First, variation in reward magnitude (one vs. three pellets) affected memory persistence in a D1/D5-receptor–dependent manner. A relatively high dose of SCH23390 was required to block 24-h memory of spatial location after three pellets. Second, novelty exploration after one-pellet memory encoding induced LTM, and this could be blocked by a low dose of the antagonist. When LTM after three-pellet encoding was blocked by a high-dose of SCH23390 at encoding, LTM could be rescued by 5-min novelty exploration. These data imply that the PRPs required for memory persistence can be supplied by other neural events happening within the novelty time window.

Our data, together with those of others (11, 12), therefore provide strong evidence that STC is relevant to memory. Behavioral data that are potentially discrepant with this idea showed that prior exposure to a novel taste was insufficient to rescue an impairment in latent inhibition for conditioned taste aversion caused by the inhibition of protein synthesis in the gustatory cortex (14). This could be because the unknown PRPs up-regulated by the novel taste may not have lasted long enough for the learning taste to capture 100 min later, or because the novel and learning tastes engage only partially overlapping populations of neurons (43).

Alternative Account of Novelty-Induced Memory Enhancement.

A possible alternative account of our data might be in terms of a novelty-associated modulation of memory consolidation (44). Memory modulation theory proposes that, during postlearning consolidation, neurotransmission in the amygdala (e.g., via β-adrenergic receptors) or stress hormones (e.g., via glucocorticoid receptors) can modulate the persistence of declarative memory in other brain areas including the hippocampus (45, 46). This theory might be extended to include novelty and thus, although developed in the context of emotional learning protocols, might still be relevant to our data. A key difference between this and our STC approach is that it does not require the concept of synaptic tags (1). We recognize that novelty exploration may up-regulate neurotransmission in the amygdala sufficient to modulate the consolidation of memory encoded elsewhere (e.g., in hippocampus). However, memory modulation theory does not predict two key features of our data that may be better explained by STC theory. First, intrahippocampal SCH23390 during novelty exploration blocked the persistence of place memory. As dopamine fibers to the hippocampus originate in the VTA, not the amygdala, it is more natural to see our data as fitting the Lisman and Grace theory of novelty-associated dopamine neuromodulation (15). Second, when exploration was scheduled to occur before memory encoding, a rescue of memory blocked by SCH23390 was still observed. Given the emphasis on postencoding events in the modulation account (44), it cannot coherently be extended to explain events occurring before memory encoding. We therefore reemphasize the symmetry feature of STC theory with respect to whether the up-regulation of PRPs occurs before or after the event associated with the setting of synaptic tags (47).

Conclusion

Through conjoint use of directly analogous physiological and behavioral studies, we have revealed a similarity between STC and what Moncada and Viola (11) have aptly called “behavioral tagging” at both the phenomenonological level (Figs. 1 and 2 and Fig. S3) and the level of biological mechanisms (Figs. 3 and 4). Behavioral tagging will influence theories of memory processing because animal experiments in which individual training experiences occur in isolation are artificial. Events, and the cognitive processes set in train by their encoding as memory traces, take place seamlessly against a background of other neural events that may affect the fate of such memories via the interplay of consolidation mechanisms. Our unique everyday memory task captures this essential and ubiquitous feature of memory, and will enable further investigation of the diverse determinants and triggers of the neural mechanisms of memory consolidation.

Materials and Methods

For full details, see SI Materials and Methods.

Subjects, Infusions, and Procedures.

Adult male Lister Hooded rats were used for behavior studies and Wistar rats were used for electrophysiology. All procedures followed the UK Home Office regulations. Rats were chronically implanted with cannulas aiming at the dorsal hippocampus (Fig. S7). SCH23390 (1 μg/μl or 3.3 μg/μl), Anisomycin (125 μg/μl), D-AP5 (5.9 μg/μl) or vehicle were infused at 0.25 μl/min (by R.L.R. and blind to S.-H.W., who ran the behavior). Training consisted of habituation, pretraining, and one-trial spatial task (Fig. S2). The six conditions, described in the main text, were then conducted (Tables S1 and S2).