Recall of Memory Sequences by Interaction of the Dentate and CA3: a Revised Model of the Phase Precession

Neural Networks 18 (2005) 1191–1201 www.elsevier.com/locate/neunet 2005 Special issue Recall of memory sequences by interaction of the dentate and CA3: A revised model of the phase precession

John E. Lisman a,*, Lucia M. Talamini b,1, Antonino Raffone c,2

a Department of Biology, Brandeis University, Volen Center for Complex Systems, Waltham, MA 02454, USA b Department of Psychonomics, University of Amsterdam, 1018WB Amsterdam, The Netherlands c Department of Psychology, University of Sunderland, Sunderland SR6 0DD, UK

Abstract

Behavioral and electrophysiological evidence indicates that the hippocampus has a special role in the encoding and recall of memory sequences. Importantly, the hippocampal phase precession, a phenomenon recorded as a rat moves through place ﬁelds, can be interpreted as cued recall of the sequence of upcoming places. The phase precession can be recorded in all hippocampal regions, but the role of each region has been unclear. Here, we suggest how the dentate and CA3 regions can work together to learn sequences, recall sequences, and generate the phase precession. Our proposal is constrained by information regarding synaptic plasticity rules, network connectivity, timing delays and theta/gamma oscillations. q 2005 Elsevier Ltd. All rights reserved.

Keywords: Dentate; Hippocampus; Memory sequences

The study of hippocampal function is now reaching the state Samsonovich & McNaughton, 1997; Sohal & Hasselmo, where the ﬁeld is a relatively mature one. There is general 1998; Wu & Yamaguchi, 2004). agreement that the main function of the hippocampus is the storage and recall of memories. Moreover, the underlying circuitry, the properties of individual cells and the rules of 1. Behavioral evidence that the hippocampus is involved synaptic plasticity have all been extensively studied with in memory of sequences considerable success. The hippocampus is thus a test-bed for one of the grandest efforts of neuroscience: to understand a Sequences evolve over time and it should therefore not be surprising that special network and synaptic processes are behaviorally relevant brain function in a reductionist way, needed to deal with the temporal issues posed by the storage spanning levels from the molecular, to the cellular, to the and retrieval of sequences. There is now a substantial body of network. Towards that end, we have previously proposed that data indicating that the hippocampus is required to memorize major properties of the dentate/CA3 circuitry can be under- the sequences of events that make up an ‘episode’. For stood as specialized for the storage and recall of memory instance, Honey and Good (Honey, Watt, & Good, 1998) sequences. This paper will review that work and outline a trained rats on tone-light sequences and then measured the rat’s revision of our thinking regarding which synapses store orienting to changes in the order of these sequences. Rats sequence information. This work is part of a larger effort by increased their orienting response about two-fold to sequence many laboratories to understand the theoretical basis of alterations, but only if their hippocampus was intact (Fig. 1). In sequence storage in the hippocampus (Levy, 1996; more speciﬁc tests of order memory (Fortin, Agster, & Eichenbaum, 2002), rats were presented with a series of odors; when presented with a pair of these odors, normal rats * C C Corresponding author. Tel.: 1 781 736 3145/3148; fax: 1 781 736 could reliably pick the one presented earlier, whereas rats with 3107. E-mail addresses: [email protected] (J.E. Lisman), ltalamini@fmg. hippocampal lesions could not. In contrast, if the rat was uva.nl (L.M. Talamini), [email protected] (A. Raffone). merely required to recognize whether a test item had been 1 Tel.: C31 20 5256853; fax: C31 20 6391656. previously presented (regardless of order), the hippocampus 2 Tel.: C44 191 5153889; fax: C44 191 5152308. was not required. 0893-6080/$ - see front matter q 2005 Elsevier Ltd. All rights reserved. It should perhaps be emphasized that while this paper doi:10.1016/j.neunet.2005.08.008 focuses on sequence memory, we do not mean to imply that 1192 J.E. Lisman et al. / Neural Networks 18 (2005) 1191–1201

3. Electrophysiological evidence for cued sequence recall during the awake theta state: properties of the ‘phase precession’

If memory sequences are stored in the hippocampus they should be accessible during the awake state in order to guide behavior. Consistent with this assumption, the phase precession (Fig. 2) of hippocampal place cells (O’Keefe & Recce, 1993) can be interpreted as cued sequence recall (Jensen & Lisman, 1996; Skaggs et al., 1996; Tsodyks, Skaggs, Sejnowski, & McNaughton, 1996). For instance, the recall of a stored sequence of positions along a track can be initiated by an environmental cue (the current position), as illustrated in Fig. 3. During each theta cycle, the cue initiates the sequential recall of 5–6 upcoming positions along the track. What drives the precession is that on each successive theta cycle, the current place cue has progressed along the path due to the rat’s movement. Consistent with such a space-driven process, the velocity of the phase precession (phase change per unit time) is Fig. 1. Behavioral evidence for sequence storage by the hippocampus. Sham proportional to the velocity of the rat (Skaggs et al., 1996). operated rats orient to a change in the sequence of known stimuli (left). In the Moreover, when the rat’s movement in space is decoupled absence of the hippocampus (right), there is little orienting. Adapted from from the movement of its legs by placing the rat on a running Fig. 2c of Honey et al. (1998). wheel, no phase precession occurs (Hirase, Czurko, Csicsvari, this is the only function of the hippocampus. Indeed other & Buzsaki, 1999). This finding is particularly telling, but it functions include the formation of new associations as a result does not occur under all conditions: if the animal runs rapidly of combining lateral and medial perforant path inputs on the running wheel, some phase precession can still be (Talamini, Meeter, Elvevag, Murre, & Goldberg, 2005), the observed (Harris et al., 2002). One resolution of this apparently associations of items with context (Lisman, 1999) and novelty contradictory data may be that a weak form of phase precession detection (Meeter, Murre, & Talamini, 2004; see review in can occur independent of spatial cueing, simply as a result of Lisman & Grace, 2005). the interaction of theta-frequency inhibition with a slowly ramping excitation (Kamondi, Acsady, Wang, & Buzsaki, 1998; Magee, 2001). 2. Electrophysiological evidence for the involvement If, indeed, the phase precession reflects a recall process, it of the hippocampus in sequence replay during sleep should not be evident before learning occurs. Thus, when an animal is first exposed to a new environment, its movement Several laboratories (Lee & Wilson, 2002; Nadasdy, Hirase, should not produce phase precession. Since hippocampal Czurko, Csicsvari, & Buzsaki, 1999; Skaggs, McNaughton, Wilson, & Barnes, 1996) have observed that sequences of CA1 cell firing that occurred during the awake theta state were seen again during the sharp waves of slow wave sleep (SWS). This ‘replay’ during SWS is about 20 times faster than the sequence observed during awake theta (Lee & Wilson, 2002). In contrast, similar replay during REM sleep appears to be in real-time and involves temporal segments on the time scale of minutes. This dream-related process involves preservation of both the sequence of firing of different cells, and the time course of theta modulation that occurred during the waking state (Louie & Wilson, 2001). It is suspected that these forms of repetition during sleep are important for memory consolidation. Such sequence replay, in the absence of environmental input, strongly suggests that sequences can be retrieved in the hippocampus. Moreover, since it is known that sharp waves Fig. 2. The phase precession. This is a plot of every spike of a place cell in CA1, originate in CA3 (Buzsaki, 1986), it seems that the information as the animal repeatedly runs through the cell’s place field, from left to right. Each spike is assigned a phase relative to ongoing theta oscillation, as observed necessary for retrieval must be stored in the hippocampus itself in the hippocampal field potential. Adapted from Fig. 2A of Mehta et al. (2002). and is not simply passed on to the hippocampus from Note that other studies have found somewhat less phase variation in the right connected structures. half of the place field (e.g. Huxter, Burgess, & O’Keefe, 2003). J.E. Lisman et al. / Neural Networks 18 (2005) 1191–1201 1193

Fig. 3. Explanation of the phase precession. (a) As the rat runs through the place field of a of a place cell (positions A–G), firing of this cell occurs with progressively earlier phase on each successive theta cycle. The magnitude of the phase advance is one gamma period (the fast oscillation in the graph) per theta cycle. The inset (top right) is an intracellular recording from a pyramidal cell, showing the co-occurrence of gamma and theta in the hippocampus (taken from Soltesz & Deschenes, 1993). (b) Illustration of how the stored sequence A–M, when cued by current position, gives rise to sequence readout of subsequent positions. On the first theta cycle (#1), the cue, A, leads to readout of positions B to G. Thus, as the rat is moving along the sequence of positions, the recorded cell fires with progressively earlier phase, giving rise to the phase precession. It is important to note that in this model, the ‘true’ place field of the recorded cell is actually position G; firing at earlier positions predicts that G will be appearing sooner. (c) Illustration of the idea that different positions are encoded by subsets of cells in the network (distributed spatial code). function is characterized by fast learning, only a single position where a cat resides, it would be adaptive for the rat to exposure to the track may lead to the phase precession. stop before it gets there. Consistent with this scenario, Mehta (Mehta, Lee, & Wilson, 2002) found that the phase precession was not evident on the 4. Where is the phase precession generated? first passage of a rat around track (Fig. 4), but rapidly developed on subsequent runs. These results suggest that the The phase precession has been recorded in all hippocampal phase precession is indeed a result of learning. A further test subfields. It remains possible that the phase precession is of this hypothesis would be to block the synaptic plasticity already present in the entorhinal cortex layer 2/3 cells and is that is thought to underlie learning; in this case, development simply passed on to the hippocampus. Now that methods are of the phase precession should also be blocked. Many available to record with certainty from this class of cells (Brun hippocampal synapses display an NMDAR-dependent form et al., 2002), it will be important to test this possibility directly. of plasticity and it thus will be important to determine whether However, given the generation of sharp wave replay activity in the development of the phase precession can be blocked by CA3 (see earlier comments about sharp waves) and the interfering with this plasticity mechanism. However, it is necessity of the hippocampus proper for behavioral sequence important to recall that the LTP at some important synapses in memory, it seems likely that the hippocampus is directly involved in sequence generation during theta. Since there are the hippocampus (perforant path input to CA1 (Golding, Staff, no connections from CA1 back to the dentate or CA3, the phase & Spruston, 2002); mossy fiber inputs to CA3 (Zalutsky & precession cannot be generated exclusively in CA1. Since the Nicoll, 1990)) do not require NMDAR. Furthermore, even the precession is observed in the dentate, it could be generated ‘classic’ NMDAR-dependent synapse, the Schaffer collateral there and passed forward to CA3 and CA1. Alternatively, since input to CA1, can exhibit NMDAR-independent LTP when CA3 and the dentate are reciprocally connected, the phase stimulated at very high frequency (Morgan & Teyler, 1999). precession could occur through the interaction of these two The testing of the dependence of the phase precession on networks. This idea will be developed in the next sections. learning will have to take into consideration these complexities. 5. Requirements for accurate sequence recall; We hope the reader can now see the importance of the phase the roles of autoassociation and heteroassociation precession. If it indeed reflects cued sequence recall, it provides experimenters with a window onto a crucial hippocampal In early ideas about sequence recall, it was thought that memory function. The ability to deal with sequences allows the learning occurred exclusively by increasing ‘heteroassociative’ animal to store causal relations and use them to predict the weights, i.e. the weights of the synapses between groups of future. For instance, if a rat’s position predicts an upcoming cells encoding sequential memories. The overall sequence is 1194 J.E. Lisman et al. / Neural Networks 18 (2005) 1191–1201

Fig. 4. Evidence that the phase precession is learned. (a) Phase of each spike is plotted as a function of position in the place field on the first lap on a track (on that day). The phase precession is not evident (red dots; see also c); in contrast, phase precession is evident on the last lap (blue dots; see also d). (b) The development of phase precession with lap number; the plot shows how the correlation of phase with position on the track increases with the number of laps. Taken from Fig. 4 of Mehta et al. (2002). thus encoded by connecting cells that represent memory A to the correct form, B (Fig. 5 inset). It was, thus, proposed that the cells that represent memory B, which then connect to memory concatenation of error could be avoided if each heteroassocia- C, etc.; see Fig. 5 inset). The recall process can then be tive step in a recall chain (e.g. A to B0) was followed by an triggered by the presentation of A, which acts as a memory cue autoassociative step (B0 to B). This would then provide an to evoke the rest of the sequence; specifically, the firing of A accurate starting point for the next heteroassociative step in the cells would excite B cells which excite C cells, and so on. chaining process. In this way, the entire sequence could be However, this kind of simple chaining process has a major accurately recalled (Fig. 5 inset). problem: the noise in neuronal recruitment can give rise to Several critical questions about such interplay of associative errors in the excitation of B, such that some cells that are part of processes were left unexplored by early models. In particular, it B will not fire, whereas others that are not part of B will. This was unclear what type of network architecture would support corruption can be signified as B0. In the next step of the recall such a process and how autoassociative and heteroassociative process, the B0 cells will excite the C cells. However, because weights could be selectively stored by different synapses. of the corruption of B and because of additional errors Moreover, no attempt was made to relate this idea to actual produced in this chaining step, C cells will be more corrupted memory circuits. We have sought to map this dual function onto than B0, signified as C00. Thus, as chaining proceeds, memories the hippocampal memory networks and to provide a physio- will become progressively more corrupted. This is termed the logically plausible basis for the selective storage of different concatenation of error. types of weights at different synapses. An initial attempt was Two papers have suggested similar, theoretical solutions to made in (Lisman, 1999) which we now revise here. the concatenation of error problem (Kleinfeld, 1986; Sompo- linsky & Kanter, 1986); both are based on the error correction 6. The dentate and CA3 networks are reciprocally properties of autoassociative networks. Here, autoassociation connected refers to the strengthening of synaptic connections between cells encoding the same memory (e.g. B). Through such Several anatomical and electrophysiological observations connections, a corrupted memory, B0, may be transformed to provided the starting point for our thinking about the operations J.E. Lisman et al. / Neural Networks 18 (2005) 1191–1201 1195

Fig. 5. Hippocampal dentate and CA3 regions and their excitatory pathways. Inset: schemes for how autoassociation, heteroassociation and their combination can contribute to sequence recall. Note that the number of primes indicates the degree of corruption of a memory item. of the dentate and CA3 in the storage and recall of sequences. sequence recall. A previous model (Lisman, 1999) proposed Dentate granule cells have axons called mossy fibers, which that the autoassociative step might occur first, and thus in the powerfully excite CA3 cells. What is less well appreciated is dentate, and that the heteroassociative step might therefore that there is also a pathway for flow of information back from occur in CA3. The arguments for this assignment were not very CA3 to the dentate (reviewed in Lisman, 1999). CA3 axons, in strong, and our view of this matter has now changed to the addition to exciting other CA3 cells and providing Schaffer reverse, in line with some recently reported experiments. collateral input to CA1, also send collaterals back to the hilus Specifically, genetic disruption of synaptic plasticity in CA3 of the dentate (Fig. 5). These collaterals most likely target has been shown to block pattern completion, the hallmark of ‘mossy cells’, which, in turn, provide massive numbers of autoassociative function (Nakazawa et al., 2002). There is thus excitatory synapses to granule cells, forming the principal input good reason to believe that the recurrent collaterals of CA3 in the inner third of the molecular layer (Jackson & Scharfman, cells form an autoassociative network. Moreover, we have 1996; Scharfman, 1994). In the ventral hippocampus, CA3 developed new ideas about the fundamental importance of axons also synapse onto dentate granule cells directly (Li, network delays in determining how specific synapses can Somogyi, Ylinen, & Buzsaki, 1994). selectively encode heteroassociative weights. As the reader The feedback pathways from CA3 to dentate have received will see, these ideas provide a principled argument that relatively little investigation and warrant further study. heteroassociation should occur in the dentate. Although there is suggestive evidence for the connection of Before describing these ideas, it is necessary to state the CA3 cells to mossy cells, this needs to be confirmed by paired following premises (for more detailed review of these ideas see recordings of identified cells. The fact that sharp waves, Jensen & Lisman, 2005; Lisman, 2005). generated in CA3, are propagated back to the dentate, provides the strongest evidence to date that feedback is indeed operative (1) We assume that each ‘item’ in a sequence is already (Buzsaki, 1986; Penttonen, Kamondi, Sik, Acsady, & Buzsaki, ‘chunked’, meaning that even though an item, such as a 1997). spoken word, has temporal aspects, at the level of the medial temporal lobe it has come to be represented by a 7. A revised model: network delay determines why purely spatial code. autoassociation occurs in CA3, whereas heteroassociation (2) Sequential items are assumed to arrive at the hippocampus occurs in the dentate with a temporal separation of about 20 ms. This time corresponds to the period of gamma frequency (w40 Hz) The reciprocally connected dentate and CA3 networks oscillatory population dynamics observed in (para)hippo- could provide the kind of interaction between an autoassocia- campal circuits. The importance of such gamma oscil- tive and a heteroassociative network needed for accurate lations relates to ideas about a role of a multi-item working 1196 J.E. Lisman et al. / Neural Networks 18 (2005) 1191–1201

memory buffer that may provide input to the hippocampus feedback connections between CA3 pattern A and dentate (reviewed in Jensen & Lisman, 2005). The need for such a pattern B. Thus, according to these principles, the feedback buffer follows from the following consideration. Most synapses in the dentate are well suited to selectively encode the models of hippocampal learning have assumed that heteroassociative links that underlie sequence memory. sequence learning is driven by hippocampal inputs that An important remaining question is the identity of the directly reflect the timing of sensory input. In this case, dentate synapses that mediate heteroassociation. One candidate however, learning can only occur between items that arrive is the set of feedback synapses of CA3 cells onto mossy cells. within less than 50 ms because this is the duration of the Since mossy cells also receive feedforward excitatory synapses LTP window (see definition below); the broad class of from dentate granule cells, the coincidence of feedforward and situations where items arrive at longer temporal separation feedback excitation, referred to in the above paragraph, could could thus not be handled by this class of models. be occurring at mossy cells. A second possible site for the A solution to this problem may be that the cortex contains coincidence is the mossy cell synapse onto dentate granule a multi-item buffer in which sequential items that are cells. According to this view, the CA3–mossy cell pathway presented at long temporal separations, become active with should be viewed as a conduit of information from CA3 back to a short temporal separation within the buffer. A specific the dentate cells; the critical coincidence then occurs at the model of such a buffer has been developed based on the granule cell dendrites when they receive feedforward input theta (5–10 Hz) and gamma frequency (30–100 Hz) from cortex and feedback information from CA3 via mossy oscillations observed in the medial temporal lobe (Lisman cells. Given the large number of granule cells (106) compared & Idiart, 1995). According to this model, successively to mossy cells (104) and the enormous number of mossy cell presented items become active in the buffer in successive synapses onto the granule cell population (O109), these gamma subcycles of theta oscillations (irrespective of the synapses would seem to be an appropriate site for hetero- actual interval at which they were presented to the associative connections. organism), and this pattern is repeated on each theta cycle. Successive gamma cycles occur at an interval 8. Understanding additional properties of the phase of 10–30 ms. As we will see, this interval roughly matches precession the delay produced by transmission from dentate to CA3 and back to the dentate. This, in turn, has important 8.1. The role of gamma in chaining: predicting the consequences for how heteroassociative weights may be magnitude of the phase precession stored. (3) A final assumption is that heteroassociation is encoded by As we emphasized previously, the recall of sequences is an LTP-like process and should therefore be consistent proposed to occur through a cue-initiated chaining process. In with what is known about this process. LTP requires that principle such chaining could be very rapid, with delays limited postsynaptic depolarization occur after the presynaptic only by conduction delays, synaptic delays and postsynaptic spike. There can, however, be some delay; plots of LTP vs. integration time. Such delays are short and cumulatively in the delay decay with a time constant of about 50 ms, defining range of several milliseconds. Indeed, such rapid chaining is what is termed the ‘LTP window’. thought to underlie the so-called ‘synfire’ chains in cortex and sharp waves in the hippocampus (Abeles, Hayon, & Lehmann, Given this as background, we can now put forward our 2004; Buzsaki, 1986). However, we believe that the chaining model of sequence learning in dentate–CA3 circuits (illustrated process that underlies the phase precession is much slower and in Fig. 6). It is proposed that encoding of autoassociation and that the slowing results from gamma frequency oscillatory heteroassociation at different synapses results simply from population dynamics. Gamma oscillations arise from a delays in the looping of information between dentate and CA3 negative feedback loop between principle cells and inter- circuits. Measurements in rats suggest that the time it takes to neurons that provide fast feedback inhibition (Fisahn, Pike, go around this loop will be about 20 ms; 7–10 ms for spikes to Buhl, & Paulsen, 1998; Mann, Suckling, Hajos, Greenfield, & be evoked in CA3 as a result of firing in dentate granule cells Paulsen, 2005). Whereas fast feedback is weak during sharp (Henze, Wittner, & Buzsaki, 2002; Yeckel & Berger, 1990), waves, it is very powerful during the awake state, leading to and another 7–10 ms for the activation of granule cells by the prominent gamma oscillation superimposed on the theta wave firing of CA3 cells (Scharfman, 1994; 1996; Wu, Canning, & (Fig. 3, inset) (Bragin et al., 1995; Buzsaki, Buhl, Harris, Leung, 1998). As a result of these delays, feedback information Csicsvari, Czeh and Morozov, 2003). The slowing of the triggered by a memory A will arrive at synapses on the dentate chaining process by gamma can be understood as follows: after cells that encode A about 20 ms after they first fired. Such pre a group of cells fires (representing a particular place), they after post timing will not evoke LTP; if anything LTD will excite inhibitory interneurons which shut down all remaining occur. On the other hand, the same feedback from A will arrive pyramidal cells; these cells are unlikely to fire again until this at the dentate at approximately the same time that a pattern inhibition has decayed. Thus, the chaining process is turned representing B is excited by input from cortex. This coincident into a series of discrete epochs by gamma (the utility of this timing will produce pre- and postsynaptic firing that falls discreteness is that it allows downstream networks to detect within the LTP window and lead to the strengthening of groupings using a coincidence detection mechanism). J.E. Lisman et al. / Neural Networks 18 (2005) 1191–1201 1197

Fig. 6. Scheme showing how network delay can lead to selective strengthening of heteroassociative linkages in CA3-to-dentate feedback synapses, based on a learning rule (top right) that, at other synapses (e.g. in recurrent collaterals in CA3), leads to autoassociation.

The presence of strong gamma means that firing may not occur The clocking of the hippocampus by theta and gamma, at the moment excitation arrives; rather inhibition has to decay during learning and recall, leads to a simple explanation of the before firing occurs, thereby slowing the chaining process. magnitude of the phase advance. Consider a rat running at its One line of experiments that directly demonstrates the average velocity during learning (VAL); the core assumption is importance of theta phase allowed testing of just how finely that a working memory buffer absorbs information about the theta phase can be meaningfully subdivided. The idea of rat’s current position at every theta cycle, and compresses the discrete phase coding predicts that optimal decoding schemes information, so that sequential positions on the rat’s trajectory should be dividing theta into a number of phase bins equal to are placed in sequential gamma bins of one theta cycle. Thus, the number of gamma cycles within a theta cycle; yet finer when the buffer is full (let us say its capacity, as determined by phase bins should provide no additional information. This the number of gamma cycles within theta, is w7), it will prediction was tested on a large data set from groups of neurons contain the current position in the last gamma cycle, and the recorded simultaneously as rats moved along a track (Jensen & previous six positions in the preceding gamma cycles. This set Lisman, 2000). Ensemble data sorted by different phase criteria of information will be repeated for many theta cycles, allowing were used to predict the animal’s position along the track. The LTP-like processes to incorporate the information into difference between the predicted and actual position allowed hippocampal long-term memory. Now, during recall, when objective comparison of different binning processes. It was the animal enters the initial place in this sequence, the gamma- found that increasing the number of phase bins within theta frequency chaining process leads to the prediction of the cycles increased the accuracy of position prediction up to about upcoming six positions. If the rat is running at its average five phase bins (Fig. 7). Increasing the number of phase bins velocity VAR during recall, the rat’s position will have moved further did not produce further improvement and appears to to the second position by the onset of the next theta cycle and it have been deleterious. These results are therefore consistent is this second position that now serves as a cue. As explained with the idea that information is parceled out in the different previously, it is exactly this updated cueing that leads to the gamma subcycles of a theta cycle. phase precession. However, one can now see an additional 1198 J.E. Lisman et al. / Neural Networks 18 (2005) 1191–1201

Fig. 7. Evidence that increasing the number of phase bins within theta to about 5–6 provides additional information with which to reconstruct the rat’s position from the firing pattern of a large population of pyramidal cells. From Fig. 4 of Jensen and Lisman (2000). On the y-axis is the improvement in position reconstruction accuracy (predicted position minus actual position) achieved by taking phase information into account. The x-axis is the number of phase bins within theta that were used for the analysis. A and B utilize slightly different analytical methods. prediction; as the rat is approaching a given position on the position is imposed on CA3/CA1 by layer 2/3 of the entorhinal familiar trajectory, the firing response of a cell encoding that cortex, while the chaining process is somehow inhibited (e.g. position will fire predictively during six theta cycles; on the by inhibiting the feedback relay through mossy cells or by seventh cycle, the rat is at that position, and fires as a response inhibiting granule cells). Then, in the next period of theta, the to direct sensory input (presumably from entorhinal cortex). On entorhinal input ceases and the intrahippocampal chaining the eighth cycle, the rat is beyond the place field and the cell is begins. It is this chaining process t hat gives rise to the phase silent. This argument makes it clear that if, during recall, the rat dependent part of the curve in Fig. 2. The assumption of an runs through the place field at velocity VAL, the place cell oscillatory EC input is in line with observations of theta- should fire on a number of theta cycles that equals the number frequency population dynamics in this structure (e.g. Kocsis, of gamma cycles within a theta cycle.Toagoodfirst Bragin, & Buzsaki, 1999). Furthermore, entorhinal excitation approximation this is the case (Skaggs et al., 1996). More to the hippocampus is known to be accompanied by a powerful rigorous testing of this prediction could be achieved by inhibitory influence, which may contribute to inhibition of the determining VAL during learning, correcting phase precession chaining process (Behr, Gloveli, & Heinemann, 1998; Buzsaki, during recall for actual velocity relative to VAL,and 1984; Empson & Heinemann, 1995). The explanation of determining directly the number of gamma cycles within a bimodality can now be formulated as follows: as the rat enters theta cycle as the animal moves. the place field the neuron fires with late phase as a result of the intrahippocampal chaining; on subsequent theta cycles, it fires 8.2. Bimodality of the phase precession with earlier phase and earlier phase because it is earlier in the chaining process (the rat’s movement has updated the cue). There are several intriguing properties of the phase Finally, as the rat approaches the end of the place field, the cell precession that should further constrain possible models. As is now driven by the entorhinal cortex rather than by seen in Fig. 2, the phase precession in CA1 begins as the animal intrahippocampal connections. The phase distribution is now enters the place field of the recorded cell from the left. It is in broad because the entrohinal cortex drives the hippocampus this region of the place field that the phase distribution is over a broad phase range (which however overlaps little with relatively narrow and changes systematically as the animal the phase range over which the precession occurs). moves (the precession). However, as the rat reaches the end of the place field, the phase distribution becomes wide and is 8.3. Place field expansion and asymmetry distributed over a range of phases that overlaps little with the phase-range that occurs during the precession (Fig. 2). On the As a rat repeatedly traverses a track, graphs of firing rate vs. basis of such observations, it has been suggested that the cell’s position gradually change in several ways: they become firing response within the place field is bimodal, with a phase- asymmetric and they expand to cover a larger area (Mehta, dependent part and a less phase-dependent part (Yamaguchi, Barnes, & McNaughton, 1997). Place field expansion is Aota, McNaughton, & Lipa, 2002). Interestingly, the phase- blocked by NMDA antagonists, consistent with the idea that independent firing is not as evident in the dentate; rather the phenomenon arises from an LTP-like process (Ekstrom, granule cells are simply quiet during this period. There are Meltzer, McNaughton, & Barnes, 2001). Place field expansion currently no reports about the pattern in CA3. lasts for hours, but appears to decay entirely within a day Fig. 8 shows a potential explanation of the bimodality. The (Mehta et al., 1997). Several models of how this occurs have central assumption is that in the first, phase-independent, half been put forward, having in common the idea that asymmetric of the theta cycle the sensory stimulus representing current LTP could allow the network to enhance its predictive J.E. Lisman et al. / Neural Networks 18 (2005) 1191–1201 1199

Fig. 8. Possible explanation of the bimodality of the phase precession. As a rat moves along a familiar track, the initial firing of a cell that codes for an upcoming position (G) is driven by an intrahippocampal chaining process (such spikes are designated by solid lines). During this type of excitation, firing occurs relatively late in the theta cycle and is relatively phase-specific. When the rat finally reaches position G, the cell now gets direct entorhinal input, which is distributed over a somewhat broad phase range (spikes in the entrohinal cortex or hippocampal spikes driven by direct cortical input are designated by dashed lines). The entorhinal driven spiking occurs during the first half of the theta cycle and does not stimulate the chaining process. The reason for lack of chaining is unclear, but presumably has something to do with other actions of the entorhinal input (see text). One may suppose that the sharp onset of the chaining process at the offset of entorhinal input would provide a discrete enough signal to ensure that subsequent steps were also discrete and phase-specific. (prospective) ability—i.e. cells further ahead of the current frequency is not fixed, but rather arises dynamically through position will become active in the process of expansion (Abbott the interplay of excitation and negative feedback inhibition. As & Blum, 1996; Mehta, Quirk, & Wilson, 2000). Jensen and the excitation that mediates intrahippocampal chaining Lisman (1996) showed through simulation that gamma becomes stronger through repeated experience, postsynaptic

Fig. 9. Simulations showing how place field expansion occurs as a consequence of the increased heteroassociative weights resulting from repeated traversal of a track. (a) On right, plots of cued recall after various number of laps. Plots show firing pattern during a theta cycle as a result of cued recall (arrows at top mark presentation of the cue: first two items in the sequence). Three plots are shown after 1, 10, and 20 laps (top to bottom). Firing of each cell occurs within a gamma cycle; note the increase in the number of gamma cycles as a function of number of laps. Synaptic matrices (left) show that the synapses have become stronger. (b) With increasing lap number, the location of firing of cell #7 becomes earlier on the track. (c) There is also an increase in the size of the place field. Velocity of rat: VZ 100.0 cm/s. Taken from Fig. 5 of Jensen and Lisman (1996). 1200 J.E. Lisman et al. / Neural Networks 18 (2005) 1191–1201 cells fire earlier, and lead to an increase in gamma frequency. A Acknowledgements consequence is that more gamma cycles can fit within a theta cycle (Fig. 9). Given that more distant positions can be This work has been supported by grant 1 R01 NS50944-01 predicted with every additional gamma cycle contained within as part of the NSF/NIH Collaborative Research in Compu- a theta cycle, the furthest predicted position increases, thereby tational Neuroscience Program, and NIH grants 1 R01 accounting for the expansion of the place field. MH61975 and 1 R01-NS-27337. We thank Sridhar Ragha- vachari and Rod Rinkus for reading the manuscript and we gratefully acknowledge the support of the W.M. Keck 9. Concluding remarks Foundation.

The current account of phase precession, based on the References bidirectional interaction of the dentate and CA3, provides the first explanation of why phase precession should be seen in Abbott, L. F., & Blum, K. I. (1996). Functional significance of long-term both these structures. It also assigns a function to the feedback potentiation for sequence learning and prediction. Cereb Cortex, 6, connections from CA3 to dentate, for which no purpose has 406–416. been previously proposed. It should be emphasized that this Abeles, M., Hayon, G., & Lehmann, D. (2004). Modeling compositionality by dynamic binding of synfire chains. Journal of Computational Neuro- bidirectional interaction has not yet been simulated. Such science, 17, 179–201. simulation, including the roles of gamma, feedback delays and Behr, J., Gloveli, T., & Heinemann, U. (1998). The perforant path projection plasticity rules, would provide an important test of our model from the medial entorhinal cortex layer III to the subiculum in the rat and clarify some of the population dynamics underlying the combined hippocampal-entorhinal cortex slice. European Journal of proposed mechanism. The model of sequence memory, as we Neuroscience, 10, 1011–1018. Bragin, A., Jando, G., Nadasdy, Z., Hetke, J., Wise, K., & Buzsaki, G. (1995). have developed it here, leads to many further predications that Gamma (40–100 Hz) oscillation in the hippocampus of the behaving rat. are summarized in Table 1. It is hoped that through an interplay Journal of Neuroscience, 15, 47–60. of theory and experimental tests, we will converge upon a well- Brun, V. H., Otnass, M. K., Molden, S., Steffenach, H. A., Witter, M. P., Moser, constrained model of how the hippocampus stores and recalls M. B., et al. (2002). Place cells and place recognition maintained by direct memories. entorhinal-hippocampal circuitry. Science, 296, 2243–2246. Buzsaki, G. (1984). Feed-forward inhibition in the hippocampal formation. Table 1 Progress in Neurobiology, 22, 131–153. Testable predictions of the current model Buzsaki, G. (1986). Hippocampal sharp waves: Their origin and significance. Brain Research, 398, 242–252. Generation of phase precession by dentate/CA3 interaction Buzsaki, G., Buhl, D. L., Harris, K. D., Csicsvari, J., Czeh, B., & Morozov, A. Prediction: A key test of this idea is that the precession NOT be observed in (2003). Hippocampal network patterns of activity in the mouse. layer 2 of the entorhinal cortex, the layer that provides input to the dentate. The Neuroscience, 116, 201–211. precession in CA1 is assumed to arise secondarily, due to the input from CA3. Ekstrom, A. D., Meltzer, J., McNaughton, B. L., & Barnes, C. A. (2001). Thus, cutting the CA3 to CA1 pathway should prevent or at least diminish the NMDA receptor antagonism blocks experience-dependent expansion of phase precession. hippocampal ‘place fields’. Neuron, 31, 631–638. Reciprocal interaction between dentate and CA3 on each gamma cycle Empson, R. M., & Heinemann, U. (1995). The perforant path projection to Prediction: During the phase precession, dentate granule cells and CA3 hippocampal area CA1 in the rat hippocampal-entorhinal cortex combined pyramidal cells should fire out of phase with each other on the time scale of slice. Journal of Physiology, 484(Pt 3), 707–720. gamma Fisahn, A., Pike, F. G., Buhl, E. H., & Paulsen, O. (1998). Cholinergic The pathway from CA3 to dentate: role of mossy cells induction of network oscillations at 40 Hz in the hippocampus in vitro. Prediction: The ability of CA3 cells to excite granule cells is most clearly Nature, 394, 186–189. demonstrated during sharp waves. These originate in CA3 and drive granule Fortin, N. J., Agster, K. L., & Eichenbaum, H. B. (2002). Critical role of the cells. Although it is strongly suspected that CA3 cells do this through a two-step hippocampus in memory for sequences of events. Nature Neuroscience, 5, process in which the first step is CA3-mediated excitation of mossy cells, this 458–462. pathway needs to be confirmed by cell pair recording. If indeed, mossy cells are Golding, N. L., Staff, N. P., & Spruston, N. (2002). Dendritic spikes as a vital for this feedback flow of information, deleting them should block mechanism for cooperative long-term potentiation. Nature, 418, 326–331. sequence learning and recall. One caveat here is that there is a population of Harris, K. D., Henze, D. A., Hirase, H., Leinekugel, X., Dragoi, G., Czurko, A., CA3 cells in the ventral hippocampus that make direct contact onto granule et al. (2002). Spike train dynamics predicts theta-related phase precession cells, a pathway that could possibly compensate for mossy cell destruction in hippocampal pyramidal cells. Nature, 417, 738–741. Basis of the biomodality of the phase precession Henze, D. A., Wittner, L., & Buzsaki, G. (2002). Single granule cells reliably Prediction: It is suggested that layer 2 cells of entorhinal cortex are active discharge targets in the hippocampal CA3 network in vivo. Nature predominately during the half-cycle of hippocampal theta over which the phase Neuroscience, 5, 790–795. precession does NOT occur (see Fig. 7) Hirase, H., Czurko, H., Csicsvari, J., & Buzsaki, G. (1999). Firing rate and Role of high frequency firing in dentate and the resulting facilitation of mossy fiber synapses theta-phase coding by hippocampal pyramidal neurons during ‘space European Journal of Neuroscience 11 Prediction: The delay important for heteroassociative learning (Fig. 6)is clamping’. , , 4373–4380. achieved in part by the requirement for high frequency firing of dentate cells; Honey, R. C., Watt, A., & Good, M. (1998). Hippocampal lesions disrupt an this produces a gradual facilitation of mossy fiber synapses into CA3, associative mismatch process. Journal of Neuroscience, 18, 2226–2230. eventually allowing a single granule cell to fire its CA3 targets (Henze et al., Huxter, J., Burgess, N., & O’Keefe, J. (2003). Independent rate and temporal 2002). The prediction is that lower dentate firing rates or disabling the coding in hippocampal pyramidal cells. Nature, 425, 828–832. enormous facilitation properties of mossy fiber synapses would prevent Jackson, M. B., & Scharfman, H. E. (1996). Positive feedback from hilar mossy sequence learning (i.e. prevent the development of the phase precession) cells to granule cells in the dentate gyrus revealed by voltage-sensitive dye and microelectrode recording. Journal of Neurophysiology, 76, 601–616. J.E. Lisman et al. / Neural Networks 18 (2005) 1191–1201 1201

Jensen, O., & Lisman, J. E. (1996). Hippocampal CA3 region predicts memory Morgan, S. L., & Teyler, T. J. (1999). VDCCs and NMDARs underlie two sequences: accounting for the phase precession of place cells. Learning and forms of LTP in CA1 hippocampus in vivo. Journal of Neurophysiology, Memory, 3, 279–287. 82, 736–740. Jensen, O., & Lisman, J. E. (2000). Position reconstruction from an ensemble of Nadasdy, Z., Hirase, H., Czurko, A., Csicsvari, J., & Buzsaki, G. (1999). hippocampal place cells: Contribution of theta phase coding. Journal of Replay and time compression of recurring spike sequences in the Neurophysiology, 83, 2602–2609. hippocampus. Journal of Neuroscience, 19, 9497–9507. Jensen, O., & Lisman, J. E. (2005). Hippocampal sequence-encoding driven by Nakazawa, K., Quirk, M. C., Chitwood, R. A., Watanabe, M., Yeckel, M. F., a cortical multi-item working memory buffer. Trends in Neurosciences, 28, Sun, L. D., et al. (2002). Requirement for hippocampal CA3 NMDA 67–72. receptors in associative memory recall. Science, 297, 211–218. Kamondi, A., Acsady, L., Wang, X. J., & Buzsaki, G. (1998). Theta oscillations O’Keefe, J., & Recce, M. L. (1993). Phase relationship between hippocampal in somata and dendrites of hippocampal pyramidal cells in vivo: Activity- place units and the EEG theta rhythm. Hippocampus, 3, 317–330. dependent phase-precession of action potentials. Hippocampus, 8, Penttonen, M., Kamondi, A., Sik, A., Acsady, L., & Buzsaki, G. (1997). Feed- 244–261. forward and feed-back activation of the dentate gyrus in vivo during dentate Kleinfeld, D. (1986). Sequential state generation by model neural networks. spikes and sharp wave bursts. Hippocampus, 7, 437–450. Proceedings of the National Academy of Sciences of the United States of Samsonovich, A., & McNaughton, B. L. (1997). Path integration and cognitive America, 83, 9469–9473. mapping in a continuous attractor neural network model. Journal of Kocsis, B., Bragin, A., & Buzsaki, G. (1999). Interdependence of multiple theta Neuroscience, 17, 5900–5920. generators in the hippocampus: A partial coherence analysis. Journal of Scharfman, H. E. (1994). Evidence from simultaneous intracellular recordings Neuroscience, 19, 6200–6212. in rat hippocampal slices that area CA3 pyramidal cells innervate dentate Lee, A. K., & Wilson, M. A. (2002). Memory of sequential experience in the hilar mossy cells. Journal of Neurophysiology, 72, 2167–2180. hippocampus during slow wave sleep. Neuron, 36, 1183–1194. Scharfman, H. E. (1996). Conditions required for polysynaptic excitation of Levy, W. B. (1996). A sequence predicting CA3 is a flexible associator that dentate granule cells by area CA3 pyramidal cells in rat hippocampal slices. learns and uses context to solve hippocampal-like tasks. Hippocampus, 6, Neuroscience, 72, 655–668. 579–590. Skaggs, W. E., McNaughton, B. L., Wilson, M. A., & Barnes, C. A. (1996). Li, X. G., Somogyi, P., Ylinen, A., & Buzsaki, G. (1994). The hippocampal Theta phase precession in hippocampal neuronal populations and the CA3 network: An in vivo intracellular labeling study. The Journal of compression of temporal sequences. Hippocampus, 6, 149–172. Comparative Neurology, 339, 181–208. Sohal, V. S., & Hasselmo, M. E. (1998). GABA(B) modulation improves Lisman J. (2005) The theta/gamma discrete phase code that occurs during the sequence disambiguation in computational models of hippocampal region hippocampal phase precession may be a more general brain coding scheme. CA3. Hippocampus, 8, 171–193. Accepted to the Special Theta Issue of Hippocampus. Soltesz, I., & Deschenes, M. (1993). Low- and high-frequency membrane Lisman, J. E. (1999). Relating hippocampal circuitry to function: Recall of potential oscillations during theta activity in CA1 and CA3 pyramidal memory sequences by reciprocal dentate–CA3 interactions. Neuron, 22, neurons of the rat hippocampus under ketamine-xylazine anesthesia. 233–242. Journal of Neurophysiology, 70, 97–116. Lisman, J. E., & Grace, A. A. (2005). The Hippocampal-VTA loop: Controlling Sompolinsky, H., & Kanter, I. I. (1986). Temporal association in asymmetric the entry of information into long-term memory. Neuron, 46, 703–713. neural networks. Physical Review Letters, 57, 2861–2864. Lisman, J. E., & Idiart, M. A. (1995). Storage of 7 C/K 2 short-term memories Talamini, L. M., Meeter, M., Elvevag, B., Murre, J. M., & Goldberg, T. E. in oscillatory subcycles. Science, 267, 1512–1515. (2005). Reduced parahippocampal connectivity produces schizophrenia- Louie, K., & Wilson, M. A. (2001). Temporally structured replay of awake like memory deficits in simulated neural circuits with reduced parahippocampal ensemble activity during rapid eye movement sleep. Neuron, hippocampal connectivity. Archives of Genernal Psychiatry, 62, 485–493. 29, 145–156. Tsodyks, M. V., Skaggs, W. E., Sejnowski, T. J., & McNaughton, B. L. (1996). Magee, J. C. (2001). Dendritic mechanisms of phase precession in hippocampal Population dynamics and theta rhythm phase precession of hippocampal CA1 pyramidal neurons. Journal of Neurophysiology, 86, 528–532. place cell firing: a spiking neuron model. Hippocampus, 6, 271–280. Mann, E. O., Suckling, J. M., Hajos, N., Greenfield, S. A., & Paulsen, O. (2005). Wu, K., Canning, K. J., & Leung, L. S. (1998). Functional interconnections Perisomatic feedback inhibition underlies cholinergically induced fast network between CA3 and the dentate gyrus revealed by current source density oscillations in the rat hippocampus in vitro. Neuron, 45, 105–117. analysis. Hippocampus, 8, 217–230. Meeter, M., Murre, J. M., & Talamini, L. M. (2004). Mode shifting between Wu, Z., & Yamaguchi, Y. (2004). Input-dependent learning rule for the storage and recall based on novelty detection in oscillating hippocampal memory of spatiotemporal sequences in hippocampal network with theta circuits. Hippocampus, 14, 722–741. phase precession. Biological Cybernetics, 90, 113–124. Mehta, M. R., Barnes, C. A., & McNaughton, B. L. (1997). Experience- Yamaguchi, Y., Aota, Y., McNaughton, B. L., & Lipa, P. (2002). Bimodality of dependent, asymmetric expansion of hippocampal place fields. Proceed- theta phase precession in hippocampal place cells in freely running rats. ings of the National Academy of Sciences of the United States of America, Journal of Neurophysiology, 87, 2629–2642. 94, 8918–8921. Yeckel, M. F., & Berger, T. W. (1990). Feedforward excitation of the Mehta, M. R., Lee, A. K., & Wilson, M. A. (2002). Role of experience and hippocampus by afferents from the entorhinal cortex: Redefenition of the oscillations in transforming a rate code into a temporal code. Nature, 417, role of the trisynaptic pathway. Proceedings of the National Academy of 741–746. Sciences of the United States of America, 87, 5832–5836. Mehta, M. R., Quirk, M. C., & Wilson, M. A. (2000). Experience-dependent Zalutsky, R. A., & Nicoll, R. A. (1990). Comparison of two forms of long-term asymmetric shape of hippocampal receptive fields. Neuron, 25, 707–715. potentiation in single hippocampal neurons. Science, 248, 1619–1624.