Memory Formation: The Sequence of Biochemical Events in the Hippocampus and Its Connection to Activity in Other Brain Structures

Transcription

1 NEUROBIOLOGY OF LEARNING AND MEMORY 68, (1997) ARTICLE NO. NL Memory Formation: The Sequence of Biochemical Events in the Hippocampus and Its Connection to Activity in Other Brain Structures Ivan Izquierdo* and Jorge H. Medina,1 *Departamento de BioquıB mica, Instituto de Ciências Básicas da Saúde, Universidade Federal do Rio Grande do Sul, Ramiro Barcellos 2600, Porto Alegre, RS, Brazil; and Laboratorio de Neurorreceptores, Instituto de BiologıB a Celular, Facultad de Medicina, Universidad de Buenos Aires, Paraguay 2155, 3er Piso, Buenos Aires, Argentina Recent data have demonstrated a biochemical sequence of events in the rat hippocampus that is necessary for memory formation of inhibitory avoidance behavior. The sequence initially involves the activation of three different types of glutamate receptors followed by changes in second messengers and biochemical cascades led by enhanced activity of protein kinases A, C, and G and calcium calmodulin protein kinase II, followed by changes in glutamate receptor subunits and binding properties and increased expression of constitutive and inducible transcription factors. The biochemical events are regulated early after training by hormonal and neurohumoral mechanisms related to alertness, anxiety, and stress, and 3 6 h after training by pathways related to mood and affect. The early modulation is mediated locally by GABAergic, cholinergic, and noradrenergic synapses and by putative retrograde synaptic messengers, and extrinsically by the amygdala and possibly the medial septum, which handle emotional components of memories and are direct or indirect sites of action for several hormones and neurotransmitters. The late modulation relies on dopamine D 1, b-noradrenergic, and 5HT1A receptors in the hippocampus and dopaminergic, noradrenergic, and serotoninergic pathways. Evidence indicates that hippocampal activity mediated by glutamate AMPA receptors must persist during at least 3 h after training in order for memories to be consolidated. Probably, this activity is transmitted to other areas, including the source of the dopaminergic, noradrenergic, and serotoninergic pathways, and the entorhinal and posterior parietal cortex. The entorhinal and posterior parietal cortex participate in memory consolidation minutes after the hippocampal chain of events starts, in both cases through glutamate NMDA receptor-mediated processes, and their intervention is necessary in order to complete memory consolidation. The hippocampus, amygdala, entorhinal cortex, and parietal cortex are involved in retrieval in the first few days after training; at 30 days from training only the entorhinal and parietal cortex are involved, and at 60 days only the parietal cortex is necessary for retrieval. Based on observations on other forms of hippocampal plasticity and on memory formation in the chick brain, it is suggested that the hippocampal chain of events that underlies memory formation is linked to long-term storage elsewhere through activity-dependent changes in cell connectivity Academic Press 1 This work was supported by PRONEX, Brazil, and University of Buenos Aires, Argentina. Address correspondence and reprint requests to Dr. Ivan Izquierdo, Departamento de BioquıB mica, Instituto de Ciências Básicas da Saúde, Universidade Federal do Rio Grande do Sul, Av. Ramiro Barcellos 2600, Porto Alegre, RS, Brazil. Fax: /97 $25.00 Copyright 1997 by Academic Press All rights of reproduction in any form reserved.

2 286 IZQUIERDO AND MEDINA Step-down inhibitory (passive) avoidance learning in the rat triggers biochemical events in the hippocampus that are necessary for the retention of this task. The events are similar in many ways to those described for different types of long-term potentiation (LTP) and other forms of neural plasticity (Baudry, Bi, & Tocco, 1996; Bliss & Collingridge, 1993; Collingridge & Bliss, 1993; Izquierdo & Medina, 1995, 1997; Maren & Baudry, 1995; Martin & Kandel, 1996; Reymann, 1993). They are triggered by glutamate receptor activation and involve at least four different biochemical cascades led by different protein kinases: Protein kinase G (PKG) (Bernabeu, Schröder, Quevedo, Cammarota, Izquierdo, & Medina, 1997c), protein kinase C (PKC) (Bernabeu, Izquierdo, Cammarota, Jerusalinsky, & Medina, 1995a; Bernabeu, Cammarota, Izquierdo, & Medina, 1997b), calcium calmodulin-dependent protein kinase II (CaMKII) (Bernabeu et al., 1997b; Wolfman, Fin, Dias, Bianchin, Da Silva, Schmitz, Medina, & Izquierdo, 1994), and protein kinase A (PKA) (Bernabeu, Bevilaqua, Ardenghi, Bromberg, Schmitz, Bianchin, Izquierdo, & Medina, 1997a; Bevilaqua, Ardenghi, Schröder, Bromberg, Schmitz, Schaeffer, Quevedo, Bianchin, Walz, Medina, & Izquierdo, 1997) (Fig. 1). Several steps of these cascades have been implicated in other forms of learning that also involve the hippocampus (Abel, Nguyen, Barad, Deuel, Kandel & Bourchuladze, 1997; Bliss & Collingridge, 1993; Guzowski & McGaugh, 1997; Izquierdo, Da Cunha, Rosat, Jerusalinsky, Ferreira, & Medina, 1992; Izquierdo & Medina, 1995; Mayford, Bach, Huang, Wang, Hawkins, & Kandel, 1996; Noguès, Micheau, & Jaffard, 1994; Tan & Liang, 1996; Tocco, Devgan, Hauge, Weiss, Baudry, & Thompson, 1991; Wilson & Tonegawa, 1997), in another form of inhibitory avoidance in the chick brain (Anokhin, Mileusnic, Shamakina, & Rose, 1991; Rose, 1995a,b; Zhao, Polya, Wang, Gibbs, Sedman, & Ng, 1995), and in other forms of neural plasticity (Abel et al., 1997; Bartsch, Ghirardi, Skehel, Karl, Herder, Chen, Bailey, & Kandel, 1995; Beninger & Nakonechny, 1996; Bliss & Collingridge, 1993; Carew, 1996; Huang, Colley, & Routtenberg, 1992; Reymann, 1993; Yin & Tully, 1996). Most of the evidence for the role of these processes in memory formation comes from studies on the effect on memory of infusions of specific receptor agonists and antagonists or enzyme inhibitors at various times after inhibitory avoidance training into the hippocampus and elsewhere and from detailed biochemical or histochemical analysis of these receptors and enzymes in the same structures, also at various times after training (Izquierdo & Medina, 1995, 1997). Several of the findings have been corroborated in other tasks using the same techniques (Izquierdo & Medina, 1995, 1997), and, more recently, by second-generation gene knockout studies in mouse hippocampus (Mayford et al., 1996; Wilson & Tonegawa, 1997). Step-down inhibitory avoidance involves learning not to step down from a platform in order to avoid a mild footshock. It is usually acquired in one single trial, which makes it ideal for studying processes initiated by training, uncontaminated by prior or further trials, rehearsals, or retrievals (Gold, 1986; Izquierdo, 1989). Step-down avoidance involves the specific repression of the natural tendency of rats to explore beyond the platform, without affecting the performance of exploratory behavior while on the platform, repeated approximations to its border, or abortive step-down responses; for these reasons we prefer the term inhibitory to passive (Netto & Izquierdo, 1985). There are many variants of inhibitory avoidance: In rodents, not to step through a door into a compartment where they receive footshocks; in flies, not to enter a foul-

3 HIPPOCAMPAL BIOCHEMISTRY AND MEMORY 287 odored area; in chicks, not to peck a bitter bead; in humans, to refrain from putting the fingers in the electrical outlet or from crossing a street without looking. Obviously, this task represents one of the major determinants of survival behavior in all species (Gold, 1986), and the brevity of its acquisition should not mislead into thinking that it is inborn or implicit. Indeed, there is a seldom-measured implicit component of this task (bradycardia) that, unlike its explicit component (increased avoidance latency), is insensitive to electroconvulsive shock (Hine & Paolino, 1970). The declarative component of this task, like that of many others (Calderazzo Filho, Cavalheiro, & Izquierdo, 1977; Eichenbaum, 1996; Gabrielli, Brewer, Desmond, & Glover, 1997; Izquierdo et al., 1992; Izquierdo & Medina, 1991, 1995; Matthies, 1982, 1989; Squire, 1992; Tocco et al., 1991; Vnek & Rothblat, 1996), crucially involves the hippocampus (Izquierdo et al., 1992; Izquierdo & Medina, 1991, 1995, 1997; Lorenzini, Baldi, Bucherelli, Sacchetti, & Tassoni, 1996; O Connell, O Malley, & Regan, 1997). Carew (1996) recently observed that typically, the encoding of a lasting memory entails considerable practice and rehearsal, but...rare events in our life...are so powerful as to be instantly recorded and remembered for a lifetime. An afferent volley such as that used in order to induce LTP in rats, or the association of a given experience with a footshock or a strong taste, such as that used for inhibitory avoidance conditioning in rats and chicks, respectively, or a strong odor such as that used for training in the fly, or a jet of 5HT in the mollusk ganglion certainly constitute rare and powerful events and induce pronounced biochemical changes. It is presumed that similar changes occur to a lesser degree, or in a smaller number of synapses, in less dramatic memories (Carew, 1996; Izquierdo & Medina, 1997; Yin & Tully, 1996). The presumption is based on a large body of evidence showing that the effects of many of the drugs that alter the memory of inhibitory avoidance and many of the hippocampal biochemical changes are also seen following exploratory habituation and other tasks (Izquierdo et al., 1992; Izquierdo & Medina, 1995; Matthies, 1982, 1989; Rose, 1995a; Tan & Liang, 1996; Wolfman, Da Cunha, Jerusalinsky, Levi de Stein, Viola, Izquierdo, & Medina, 1991). THE HIPPOCAMPAL BIOCHEMICAL CASCADES OF INHIBITORY AVOIDANCE LEARNING The Initial Role of Glutamate Receptors The biochemical events underlying memory formation of the inhibitory avoidance task in the hippocampus involve, first, an activation of NMDA, AMPA, and metabotropic glutamate receptors, and a short-lasting increase of NMDA1 levels and a longer lasting increase of GluR1 levels (see below) (Izquierdo & Medina, 1995, 1997; Medina & Izquierdo, 1995). Immediate but not delayed ( min) intrahippocampal infusion of the NMDA antagonist aminophosphonopentanoic acid (AP5) (Izquierdo et al., 1992; Jerusalinsky, Ferreira, Da Silva, Bianchin, Ruschel, Medina, & Izquierdo, 1992) or of the glutamate metabotropic receptor antagonist methyl-carboxyphenyl glycine (Bianchin, Da Silva, Schmitz, Medina, & Izquierdo, 1994) is amnestic. Immediate posttraining intrahippocampal infusion of glutamate or of the metabotropic agonist aminocyclopentane dicarboxylate causes retrograde memory facilitation (Bianchin et al., 1994; Izquierdo et al., 1992). Intrahippocampal adminis-

4 288 IZQUIERDO AND MEDINA tration of the AMPA receptor antagonist cianonitroquinoxaline-dione (CNQX) causes amnesia for the inhibitory avoidance task when given up to 3 h after training (Jerusalinsky et al., 1992), which indicates that hippocampal excitation mediated by these receptors is necessary for memory formation during that period. The role of hippocampal NMDA receptors in several forms of learning, particularly those that involve distant visual cues, has been studied by several authors (see references in Bliss & Collingridge, 1993; Izquierdo & Medina, 1995, 1997; Reymann, 1993). Actually, the first authors to suggest this were Morris, Anderson, Lynch, and Baudry (1986), studying the effect of chronic intracerebroventricular AP5 release by a micropump on spatial learning in a water maze. Recently, selective knockout of the gene that encodes for the NMDA1 receptor subunit in mouse CA1 was shown to disrupt both place cell ensemble activity and memory of spatial memory (Wilson & Tonegawa, 1997). A similar previous finding had been obtained using first-generation (whole body) deletion of the NMDAR1 gene (Sakimura, Kutsuwada, Ito, Manabe, Takayama, Kushiya, Yagi, Alzawa, Inoue, Sugiyama, & Mishina, 1995). The role of hippocampal NMDA receptors in memory formation has been discussed by numerous authors (see Izquierdo & Medina, 1995, 1997, for references). The role of hippocampal AMPA receptors in memory formation was discussed by Jerusalinsky et al. (1992) and by Cammarota, Bernabeu, Izquierdo, and Medina (1996). The role of metabotropic glutamate receptors in memory formation in this and other tasks has been recently discussed by Reymann (1993) and Riedel (1996). Early Involvement of Putative Retrograde Messengers Evidence suggests a participation of the putative retrograde messengers that regulate glutamate release, nitric oxide (NO), carbon monoxide (CO), and the platelet activating factor (PAF) in the early phase of memory formation of the step-down task in the hippocampus (Medina & Izquierdo, 1995). The role is much clearer than the one proposed for these substances in LTP induction (see Bliss & Collingridge, 1993; Medina & Izquierdo, 1995; Son, Hawkins, Martin, Kiebler, Huang, Fishman, & Kandel, 1997). The immediate but not the delayed (30 min) intrahippocampal infusion of the NO releaser S-nitroso- N-aminopenicillin (Bernabeu, Levi de Stein, Fin, Izquierdo, & Medina, 1995b) or of a soluble form of the platelet activating factor (mcpaf) (Izquierdo, Fin, Da Silva, Jerusalinsky, Quillfeldt, Ferreira, Medina, & Bazan, 1995) causes memory facilitation. The activity of NO synthase (Bernabeu et al., 1995b), and of heme oxygenase, the enzyme that produces CO (Bernabeu, Levi de Stein, Princ, Fin, Juknat, Batlle, Izquierdo, & Medina, 1995c), sharply increases immediately after training and returns to normal a few minutes later. Immediate but not delayed (30 min) infusion into CA1 of a PAF antagonist (Izquierdo et al., 1995), of a NO synthase inhibitor (Bernabeu et al., 1995b), or of a heme oxygenase inhibitor (Bernabeu et al., 1995c) causes retrograde amnesia for this task. Early Modulation by GABA A and Other Receptors In the first few minutes after training, memory formation is very sensitive to inhibition by GABA A receptors (when given into CA1, muscimol is amnestic and picrotoxin facilitates memory) and to modulation by cholinergic muscarinic

5 HIPPOCAMPAL BIOCHEMISTRY AND MEMORY 289 and b-noradrenergic receptors (when given into CA1, norepinephrine and oxotremorine facilitate memory, scopolamine is amnestic, and timolol alters sensitivity to GABAergic modulation; Izquierdo et al., 1992). Changes in Glutamate Receptor Properties In synaptic membranes extracted from CA1 there is a short-lived increase of NMDA1 levels measured by immunoblot at 30 min from training, and a slow rise of the levels of GluR1 that starts about 30 min after training and peaks at approximately 3 h (Bernabeu et al., 1997c). NMDA1 is a specific subunit of NMDA receptors, and GluR1 is a specific subunit of AMPA receptors. The GluR1 increase occurs concomitantly with a slow rise of B max of AMPA to AMPA receptors in synaptic membranes extracted from all hippocampal subregions (Cammarota, Izquierdo, Wolfman, Levi de Stein, Bernabeu, Jerusalinsky, & Medina, 1995) that lasts about 3 h in CA1 and CA2, but as much as 168 h in CA3 and in the dentate gyrus (Cammarota et al., 1996). Increased AMPA binding has been reported in the hippocampus after LTP by Tocco, Maren, Shors, Baudry, and Thompson (1992) and by Tocco et al. (1991) following eye-blink conditioning. Sergueeva, Fedorov, and Reymann (1993) have reported increased electrophysiological sensitivity of hippocampal AMPA receptors to AMPA in the first 90 min after LTP. The significance of such changes is not known, but no doubt they reflect increased excitability of hippocampal neurons, which may underlie plastic changes in their output to projection sites. The increase of GluR1 levels measured by immunoblot in CA1 synaptosomal membranes seen after training may correlate with phosphorylation of the subunit by CaMKII and PKA (see below). As will be commented further on, the activity of these two enzymes in hippocampus increases right after training. Incubation of synaptic membranes under conditions of optimum phosphorylation by CaMKII and/or PKA increases not only phosphorylation of the subunit, but also its B max for AMPA and the actual amount of GluR1 measurable by immunoblot (Bernabeu et al., 1997b). The regulation of AMPA (and NMDA) receptors by PKA- and CaMKII-mediated phosphorylation has been well established in various synapses of the rat brain (McGlade-McCulloh, Yamamoto, Tan, Brickey, & Soderling, 1993; Pasqualotto & Shaw, 1996; Roche, O Brien, Mammen, Bernhardt & Huganir, 1996). Involvement of Extrahippocampal Structures Several of the findings mentioned above (effect of AP5, CNQX, picrotoxin, muscimol, oxotremorine, scopolamine, norepinephrine, timolol, and PAF) have been also observed studying immediate posttraining infusions into the amygdala and medial septum (see Izquierdo et al., 1992; Izquierdo & Medina, 1995; Walz, Da Silva, Bueno e Silva, Medina, & Izquierdo, 1992). This led to the supposition that LTP-like phenomena in these other structures might be involved in memory consolidation, alongside those in the hippocampus (Izquierdo et al., 1992; Jerusalinsky et al., 1992), a supposition that later evidence proved erroneous (see below and Izquierdo & Medina, 1997). Ample evidence suggests that the amygdala (Bevilaqua et al., 1997; Cahill & McGaugh, 1990, 1996; Gray, 1982; Izquierdo & Medina, 1991, 1997; Walz et al., 1992) and to an extent the medial septum (Gray, 1982; Izquierdo & Medina, 1991, 1997; Walz et al., 1992; Wolfman et al., 1991) are involved in the processing of anxiety, alertness, or aversiveness.

6 290 IZQUIERDO AND MEDINA FIG. 1. This and all the following figures show biochemical data of hippocampal CA1 regions of rats sacrificed at various times after step-down inhibitory avoidance training with a 0.3-mA footshock, and test session performance data of rats that received bilateral drug infusions into CA1 at various times after training. Time of sacrifice or infusion is shown in the abscissae, in min. The biochemical data in the ordinates are expressed as means { SEM % of shocked controls (animals placed on the grid of the apparatus and exposed just to the footshock without avoidance training). The behavioral data are means { SEM test minus training step-down latencies expressed as a % of those from saline- or vehicle-treated controls. Asterisks indicate significant differences from controls in Newman Keuls tests: *p õ.01 and **p õ.05. This figure shows data relevant to the cgmp/pkg cascade: cgmp levels measured by radioimmunoassay in CA1 (tilted solid squares), PKG activity in CA1 measured using peptide G-5611 as substrate (solid squares), and behavioral effects of LY83583 (2.5 mg/side) or 8-Br-cGMP (1.25 mg/side) infused into CA1. Both cgmp levels and PKG activity rise immediately after training and return to normal 30 min later. LY83583 causes retrograde amnesia, and 8-Br-cGMP causes retrograde memory facilitation, when given 0 but not 30 min posttraining. Data are from Bernabeu et al. (1996, 1997c). These findings point to an early, rapid, and necessary intervention of cgmp and PKG in memory consolidation. The cgmp/pkg Cascade NO and CO activate soluble guanylyl cyclase, the enzyme that synthesizes cyclic guanosyl monophospate (cgmp) (Zhuo, Hu, Schultz, Kandel, & Hawkins, 1994a). NO, CO, and cgmp may induce long-lasting synaptic potentiation in hippocampal CA1 (Zhuo, Kandel, & Hawkins, 1994b) and the cgmp protein kinase (PKG) appears to be crucial for CA1 LTP (Zhuo et al., 1994a). The immediate but not the delayed ( min) infusion into CA1 of 8-Br- GMP causes memory facilitation of step-down avoidance (Bernabeu, Schmitz, Faillace, Izquierdo, & Medina, 1996), whereas that of the guanylyl cyclase inhibitor LY83583 is amnestic (Bernabeu et al., 1997c). Guanylyl cyclase activity increases immediately after training and returns to normal 30 min later (Bernabeu et al., 1997c) (Fig. 1). Endogenous cgmp levels in rat CA1 increase markedly, but very briefly, soon after inhibitory avoidance training and return to normal levels between 30 and 60 min later (Bernabeu et al., 1996). Thus,

7 HIPPOCAMPAL BIOCHEMISTRY AND MEMORY 291 perhaps as in LTP (Zhuo et al., 1994a), the evidence indicates that the cgmp/ PKG cascade in CA1 is crucial early on for the long-term plasticity of inhibitory avoidance (Kim, 1996). There is so far no indication of what PKG substrates may be involved in memory processing in the first few minutes that follow acquisition. In LTP, presynaptic guanylyl cyclase is supposedly a target of NO and CO action, and PKG activation might help in the mobilization of glutamatergic synaptic vesicles toward the synaptic cleft (see Zhuo et al., 1994a, and Medina & Izquierdo, 1995, for references). These possibilities await further study. The CaMKII Cascade CaMKII activity is regulated by the increase of intracellular Ca 2/ that follows activation of NMDA receptors or the entry of Ca 2/ through voltagedependent channels that trigger LTP in CA1 and in CA3, respectively, or in other areas of the brain (Bliss & Collingridge, 1993). Not surprisingly, CaMKII has been suggested early on to play a key role in the early postinduction phase of LTP (Reymann et al., 1988), which was confirmed by later studies using specific inhibitors (Ito, Hidaka, & Sugiyama, 1991) and enzyme assays (Fukunaga, Stoppini, Miyamoto, & Müller, 1993), as well as firstgeneration (whole body; Gordon, Cioffi, Silva, & Stryker, 1996) and secondgeneration transgenic studies (localized to CA3, Mayford et al., 1996, respectively). The time course of the intervention of CaMKII in hippocampal CA1 or CA3 LTP more or less overlaps with that of PKC, except that the peak of CaMKII activity occurs earlier and the increase of activity lasts less than that of PKC (see Reymann, 1993). CaMKII mediates the phosphorylation of a variety of proteins of importance in synaptic plasticity, including the ionotropic glutamate receptors (see above), CREB (Ferrer, Blanco, Rivera, Carmona, Ballabriga, Olivé, & Planas, 1996), and neurofilaments and other structural proteins (Schröder, de Mattos-Dutra, de Freitas, Lisboa, Zilles, Pessoa-Pureur, & Izquierdo, 1997). In all these cases, it acts concomitantly with PKA; however, since the sites of each of these proteins that are phosphorylated by the two enzymes are different, the functions subserved by CaMKII and PKA are also different. For reviews of these aspects, see Ferrer et al. (1996), Pasqualotto and Shaw (1996), Roche et al. (1996), and Schröder et al. (1997). The first evidence for an involvement of CaMKII in memory processing in hippocampus and amygdala was provided by Wolfman et al., (1994), who infused the specific inhibitor of this enzyme, KN62, into CA1 or into the amygdala at different times after step-down inhibitory avoidance training. A very strong amnestic effect was obtained when the inhibitor was infused immediately posttraining into hippocampus or amygdala. Intrahippocampal infusion of the drug 30 min after training had only a partial amnestic effect, and infusions 2 4 h after training into either structure were ineffective. This suggested that CaMKII plays an essential role in the hippocampus (and amygdala) in the first few minutes after training, as it does in LTP in the first few minutes after induction (Ito et al., 1991; Reymann, Brodemann, Käse, & Matthies, 1988). Subsequently, Tan and Liang (1996) reported similar findings on the amnestic effect of CaMKII inhibitors in a spatial learning task and showed a posttraining increase of CaMKII activity following training in that task. We were

8 292 IZQUIERDO AND MEDINA FIG. 2. The CaMKII cascade. Ca 2/ -dependent (shaded squares) and Ca 2/ -independent (crosses) CaMKII activity in CA1 measured using Syntide-2 as substrate, CA1 levels of the AMPA receptor subunit, GluR1 measured by immunoblot (open squares) and GluR1 phosphorylation (tilted crosses), [ 3 H]AMPA binding to AMPA receptors in CA1 (bars), and behavioral effects of CNQX (1.25 mg/side, solid triangles) and of the CaMKII inhibitor KN62 (3.6 mg/side; open triangles) infused into CA1. CaMKII activity increased in the first half hour after training; the Ca 2/ -dependent activity increased more and remained high for a longer time than the Ca 2/ -independent activity. GluR1 levels, GluR1 phosphorylation, and AMPA binding increased gradually from 30 min after training on. Both CNQX and KN62 caused retrograde amnesia; the former, when given up to 180 min posttraining, the latter only when given up to 180 min posttraining, the latter only when given immediately after training, and only partially when given 30 min later. Data from Jerusalinsky et al. (1992), Wolfman et al. (1994), Cammarota et al. (1996, 1997), and Bernabeu et al. (1997b). The data show a necessary participation of CaMKII in memory consolidation in the first few min after training and of AMPA receptors from 0 to 3 h after training. Various data in the literature suggest that the latter depends on the former (see Bernabeu et al., 1997b). able to confirm and complement these findings using the step-down inhibitory avoidance task (Fig. 2) (see Bernabeu et al., 1997b). Ca 2/ -dependent a-camkii activity increased markedly right after training, decreased slightly but still remained high at 30 min, and returned to normal in 120 min (Bernabeu et al., 1997b). These data are the mirror image of those that were obtained by Wolfman et al. (1994) using Kn62. In addition, we found that Ca 2/ -independent b-camkii activity increased slightly immediately after training and returned to normal in 30 min. Autophosphorylation of the enzyme (i.e., conversion from the active, Ca 2/ -dependent to the inactive, Ca 2/ -independent form) was maximal immediately after training (Bernabeu et al., 1997b). Changes in the phosphorylation levels of neurofilament proteins of rat hippocampus, which is mediated by CaMKII and by PKA, have been observed 60 min after either step-down inhibitory avoidance training or the habituation of exploration of the training apparatus (Schröder et al., 1997).

9 HIPPOCAMPAL BIOCHEMISTRY AND MEMORY 293 An elegant general confirmation of the importance of a-camkii in learning processes has been obtained using regulated expression of the CaMKII transgene localized to the CA3 region of the hippocampus (Mayford et al., 1996). As is known, retrograde amnesia for the inhibitory avoidance task can be induced by a variety of agents, including some that are clearly modulatory, like systemically administered b-endorphin (see Izquierdo, 1989; McGaugh, 1989). Recently (L. A. Izquierdo, Schröder, Ardenghi, Quevedo, Bevilaqua, Netto, I. Izquierdo, & Medina, 1997b), we have observed that the retrograde amnesia caused by systemic b-endorphin administration or by electroconvulsive shock can be reversed by pretest administrations of adrenocorticotropin or vasopressin (see Bohus, 1994), whereas the amnesia induced by the CaMKII inhibitor KN62 given intrahippocampally 0 h after training or that induced by the PKA inhibitor KT5720 given intrahippocampally 3 h after training cannot be reversed by the hormones. This further illustrates that CaMKII and PKA (see next section) participate in the core mechanism, and not in a modulatory mechanism, of memory in the hippocampus. The PKC Cascade PKC is activated by various transmitters, including glutamate and acetylcholine (Bliss & Collingridge, 1993; Jerusalinsky, Kornisiuk, & Izquierdo, 1997), and is essential for the continuation of LTP beyond the induction phase (Ben-Ari, Anikstejn, & Bergestovski, 1992; Colley & Routtenberg, 1993; Huang et al., 1992). In LTP, first the postsynaptic g isoform of PKC is activated and then the presynaptic b isoform, which phosphorylates the presynaptic protein GAP-43 (Routtenberg, Lovinger & Stewart, 1985). PKC inhibitors given early after induction block LTP (Huang et al., 1992; Reymann et al., 1988). One important feature of PKC is that, like Ca 2/, it stimulates adenylyl cyclase and therefore camp synthesis (Yoshimura & Cooper, 1993), which, as will be seen below, plays a crucial role in the activation of PKA and the triggering of another cascade of great importance in the long-term persistence of memory and other plastic events. In view of the many similarities between the hippocampal biochemistry of memory and that of LTP (Bliss & Collingridge, 1993; Izquierdo & Medina, 1995, 1997; Maren & Baudry, 1995; Reymann, 1993), the suggestion arose that PKC may be involved in memory formation. PKC is redistributed from cytoplasm to membrane in rat hippocampal CA3 after discrimination learning (Olds, Golski, McPhie, Olton, Mishkin, & Alkon, 1990) and eye-blink conditioning (Scharenberg, Olds, Schereurs, Craig & Alkon, 1991). Sheu, McCabe, Horn, and Routtenberg (1993) reported increased PKC substrate phosphorylation in specific regions of the chick brain after inhibitory avoidance. Serrano, Benistan, Oxonian, RodrıB guez, Rosenzweig, and Bennett (1994) reported on the amnestic effect of PKC and other protein kinase inhibitors administered into the chick brain in this task. Actually, abundant proof that memory formation in the chick depends on membrane-bound protein kinase C was provided several years earlier by Bourchuladze, Potter, and Rose (1990). Noguès et al. (1994) detected increased PKC activity in mouse hippocampus following a spatial task. Transgenic mice with a knockout of the gene that encodes for g- PKC in the whole organism (i.e., first-generation knockouts; see Wilson & Tonegawa, 1997) showed impaired hippocampal LTP (Abeliovich, Chen, Goda, Silva, Stevens & Tonegawa, 1993a) and spatial and contextual learning (Abeli-

10 294 IZQUIERDO AND MEDINA FIG. 3. The PKC cascade. PKC activity measured in CA1 by in vitro phosphorylation using histone IIIS as substrate (squares), phosphorylation of the PKC substrate GAP-43 in CA1 (triangles), and behavioral effect of the PKC inhibitor CGP41231 (2.5 mg/side) infused into CA1. PKC levels rise immediately after training, reach a peak at 30 min, and return to normal at 120 min. GAP-43 phosphorylation reaches a peak at 30 min. CGP41231 is fully amnestic when given 0 or 30 min posttraining into CA1, and partially amnestic when given at 120 min. A similar effect has been reported for another PKC inhibitor, staurosporin (Jerusalinsky et al., 1994b). Data from Cammarota et al. (1997) and Jerusalinsky et al. (1994b). The data point to a necessary role of PKC activity in CA1 in memory consolidation in the first 30 min after training. ovich, Paylor, Chen, Kim, Wehner, & Tonegawa, 1993b), two declarative tasks that, like most others, depend on the hippocampus (Eichenbaum, 1996; Eichenbaum, Schoenbaum, Young, & Bunsey, 1996). Criticisms of first-generation transgenic experiments (developmental variables, adaptive changes, alterations all over the body) can be found in Maren and Baudry (1995), Routtenberg (1995), and Wilson and Tonegawa (1997); the LTP and learning changes seen in g-pkc mutants, however, correlate well with those predicted from pharmacological or biochemical experiments (see below). We observed that the intrahippocampal (Jerusalinsky, Quillfeldt, Walz, Da Silva, Medina & Izquierdo, 1994b), intra-amygdala, or intraentorhinal (Jerusalinsky et al., 1992) infusion of two different PKC inhibitors, staurosporin and CGP41231, within the first 120 min after step-down inhibitory avoidance training, blocked memory expression of this task measured 24 h later (Fig. 3). The amnesia was complete when the drugs were given into hippocampus or amygdala up to 60 min after training, and partial when the infusion was delayed for 120 min. Subsequently, we measured posttraining PKC activity by two different methods: One less specific (phorbol dibutyrate binding; Bernabeu et al., 1995a), and another one very specific (using a synthetic peptide substrate; Cammarota, Paratcha, Levi de Stein, Bernabeu, Izquierdo, & Medina, 1997; Fig. 3). Both experiments revealed a very large increase of membrane-bound PKC activity in hippocampus relative to shocked and naive controls, starting immediately after step-down avoidance training, reaching a peak 0.5 h later, and returning to normal values over the next 2.0 h. A similar

11 HIPPOCAMPAL BIOCHEMISTRY AND MEMORY 295 pattern of changes was observed, with less intensity, in the amygdala and in the frontal, parietal, and entorhinal cortex, and to a lesser degree in cerebellum (Bernabeu et al., 1995a). No changes were observed in other brain structures. The coincidence of the time course of the change of hippocampal PKC activity with that of the amnestic effects of the PKC inhibitors (see above) points to a clear and crucial involvement of this enzyme in the posttraining memory processing of this task. The function of hippocampal PKC in memory is as unclear as it is in LTP (see Colley & Routtenberg, 1993; Reymann, 1993). As found by Routtenberg et al. (1985) following LTP, we detected 30 min after avoidance training a brief peak of enhanced PKC-mediated phosphorylation of the presynaptic protein, GAP-43, which coincided with the PKC activity peak and was blocked, like the latter, by diverse PKC inhibitors (Cammarota et al., 1997). GAP-43 has been proposed to be involved in the generation of activity-dependent synaptic morphological changes (Benowitz & Routtenberg, 1997), a function which may be important for long-term memory (see below). The camp/pka/creb-p Cascade Like LTP in rat hippocampus (Grant & Silva, 1994; Huang & Kandel, 1995, 1996; Huang, Li, & Kandel, 1994), 5HT-induced facilitation in the mollusk Aplysia (Bartsch et al., 1995), odor conditioning in the fruit fly Drosophila (Tully, 1996; Yin & Tully, 1996), long-term neuromuscular junction plasticity in Drosophila (Davis, Schuster, & Goodman, 1996), spatial learning in the mouse (Bourchuladze, Frenguelli, Blendy, Cioffi, Schutz, & Silva, 1994), and, at least in part, inhibitory avoidance in the chick (Rose, 1995a,b; Zhao et al., 1995), the biochemical events triggered by inhibitory avoidance learning in rat hippocampus, involve the late intervention of a cyclic adenylyl monophosphate (camp)/pka/creb (camp response element-binding protein) signaling pathway (Bernabeu et al., 1997a c; Bevilaqua et al., 1997). CREB is a family of transcription factors that regulate the synthesis of a number of proteins, including inducible transcription factors, when phosphorylated (CREB-P) (Davis et al., 1996; Ferrer et al., 1996). PKA-mediated CREB-P activation modulates gene activation and protein synthesis critical for persistence of all the forms of plasticity mentioned above beyond 3 or 4 h (Bourchuladze et al., 1994; Carew, 1996; Deisseroth, Bito, & Tsien, 1996; Martin & Kandel, 1996; Tully, 1996; Yin & Tully, 1996). Like LTP (Frey, Huang, & Kandel, 1993), memory of step-down inhibitory avoidance is enhanced by the intrahippocampal administration of the 8-Br analog of camp or by stimulators of adenylyl cyclase 3 or 6 h after training (Bernabeu et al., 1996, 1997a,c; Bevilaqua et al., 1997). Hippocampal camp levels slowly increase beginning 60 min after step-down inhibitory avoidance training, reaching a peak min after training; the increase is not accompanied by changes in the activity of camp-specific phosphodiesterase, suggesting that it is due to enhanced adenylyl cyclase activity (Bernabeu et al., 1996, 1997a,b). PKA activity increases after training, in two peaks: The first, immediately after training, and the second, higher, 3 to 6 h after training (Fig. 4). The second peak correlates with the maximum rise of camp levels after training and may be triggered by it (Bernabeu et al., 1997a). It is not known what triggers the earlier peak of PKA activity (Bernabeu et al., 1997b).

12 296 IZQUIERDO AND MEDINA FIG. 4. The camp/pka/creb-p cascade. CA1 levels of camp measured by radioimmunoassay (gray squares), PKA activity measured using Kemptide as a substrate (solid squares), and CREB- P (gray squares) and c-fos (tilted crosses) measured by immunohistochemistry, and effect of posttraining infusion into CA1 of 8-Br-cAMP (1.25 mg/side; open triangles) and of the PKA inhibitor KT5720 (2.5 mg/side; solid triangles) on retention of the avoidance task. camp levels rise slowly after training, reached a peak at 180 min, remained high at 360 min, and returned to normal at 540 min. There were two peaks of PKA activity and CREB-P: the first 0 min after training, and a second, higher peak at min; both PKA values and CREB-P levels returned to normal at 540 min from training. There was a late increase of c-fos levels which accompanied the second PKA/CREB-P peak. Posttraining 8-Br-cAMP enhanced memory when given 180 or 360 min after training. The PKA inhibitor was amnestic when given 0, 180, or 360 min after training, but not 60 or 540 min after training. Data from Bernabeu et al. (1996, 1997b) and Bevilaqua et al. (1997). The data point to a crucial role in memory consolidation of a camp/pka/creb-p cascade in CA1, accompanied by a c-fos increase, as well as to an early posttraining involvement of PKA (presumably using endogenous basal camp levels) and CREB-P in this process. There is, thus, a major difference between the PKA cascade and those initiated by PKG, PKC, or CaMKII. The PKG cascade lasts less than 30 min; the CaMKII cascade lasts little more than 60 min; the PKC cascade extends for about 120 min; and the PKA increase occurs in two peaks, the first immediately and the second several hours after training (Fig. 4). The data fit with those that have been described for hippocampal LTP (see Bliss & Collingridge, 1993; Huang et al., 1994; Huang & Kandel, 1995, 1996; Izquierdo & Medina, 1995; Kim, 1996; Reymann, 1993; Zhuo et al., 1994a,b). CREB-P levels in the CA1 region also increase after step-down avoidance training, in two peaks: One, immediately after training, and another one, larger, 3 to 6 h after training (Bernabeu et al., 1997a,b) (Fig. 4). Both CREP- P peaks correlate with those of increased PKA activity in the hippocampus (Fig. 4). The second wave of PKA and CREB-P is also coincident in time

13 HIPPOCAMPAL BIOCHEMISTRY AND MEMORY 297 with the amnestic effect of intrahippocampal KT5720 (Bevilaqua et al., 1997; Izquierdo et al., 1997b; Fig. 4). CREB-P induces synthesis of the transcription factor c-fos, among other proteins (Ferrer et al., 1996). C-fos increases have been reported not only after LTP (Dragunow, Abraham, Goulding, Mason, Robertson, & Faull, 1989) but also, sometimes very rapidly, after a great variety of forms of stimulation (Kaczmarek, 1992). We did in fact observe an increase of c-fos levels 3 6 h after training in CA1 (Fig. 4), coincident with the second peak of CREB-P. Posttraining c-fos increases are seen in CA2, CA3, entorhinal cortex, and posterior parietal cortex, preceded in those cases by a lower early (0 h) peak; no c- fos changes were detected in striatum or thalamus (Bernabeu et al., 1997b). The first wave of CREB-P and c-fos might be related to what Frey and Morris (1997) have recently defined as a synaptic tag ; i.e., a short-lasting molecular event that sets the stage for the development of a later, proteindependent phase of LTP. It is possible that the second wave of PKA, CREB-P, and c-fos that occurs in the hippocampus after training is functionally more important than the first, since it lasts several hours and is modulated by substances acting on dopaminergic, noradrenergic, and 5HT1A receptors (see below). Several years ago, based upon pioneering findings of his group on two posttraining peaks of RNA, protein, and glycoprotein synthesis (see below), one right after training and the other 5 8 h later, Matthies (1982, 1989) elaborated on the possible function significance of these two metabolic waves, long before CREB and its implications for plastic processes were discovered. A recent experiment by Guzowski and McGaugh (1997) underlines the key importance of hippocampal CREB for the persistence of memory for more than 4 h (as had been shown for hippocampal LTP; see Huang & Kandel, 1995, 1996; Huang et al., 1994). The administration of an oligodeoxynucleotide against CREB mrna into rat hippocampus prior to training hinders memory of water maze learning measured 48, but not 4, h after training; a randomsequence oligodeoxynucleotide has no effect (Guzowski & McGaugh, 1997). The Significance of Posttraining Biochemical Events in the Hippocampus The early events (glutamate receptor activation, etc.) and the PKG, PKC, CaMKII, and camp/pka/creb-p cascades are sequentially articulated in LTP (Baudry et al., 1996; Bliss & Collingridge, 1993; Deisseroth et al., 1996; Frey et al., 1993; Huang et al., 1994; Izquierdo, 1997; Izquierdo & Medina, 1995, 1997; Malinow, 1994; Maren & Baudry, 1995; Martin & Kandel, 1996; Mayford et al., 1996; Reymann, 1993) and maintain glutamatergic transmission enhanced for at least 6 h. It is to be presumed that they subserve a similar role in the hippocampus in memory consolidation, inasmuch as they represent major pathways used by the hippocampus to generate long-lasting synaptic plasticity of a facilitatory type. The postulation that memory may rely on LTP (Izquierdo et al., 1992; Lynch & Baudry, 1984; Matthies, 1989) has been very heuristic over the years, in particular, because it led to the demonstration of the main steps involved in memory formation in the rat hippocampus (see above) and in other forms of plasticity both in the hippocampus and elsewhere (Carew, 1996), which are very similar to those of the various forms of LTP (Izquierdo & Medina, 1995, 1997; Maren & Baudry, 1995).

14 298 IZQUIERDO AND MEDINA Discussion of this issue, however, has now become a bit idle (Barnes, 1996). It should be expected that in most situations in which the hippocampus is required to maintain synaptic function enhanced for a long time it will use similar biochemical processes, as other systems do (see Bartsch et al., 1995; Carew, 1996; Davis et al., 1996; Rose, 1995a,b; Yin & Tully, 1996). In the hippocampus or cerebellum, slight departures from these sequences (for example, a predominance of phosphatase over CaMKII activity) generate instead long-lasting depression (Bindman, Christofi, Murphy, & Nowicki, 1991; Tsumoto, 1993). In learning situations, the hippocampus is supposed to receive the stimuli relevant to the training experience from collaterals of the sensory system, from the prefrontal cortex, and from the entorhinal cortex (Damasio, 1995; Green, 1964; Hyman, van Hoesen & Damasio, 1990; Izquierdo et al., 1992; Willner, Bianchin, Walz, Bueno e Silva, Zanatta, & Izquierdo, 1993; Witter, Groenewegen, Lopes da Silva, & Lohman, 1989). These stimuli may or may not arrive in volleys such as are needed to provoke clear-cut, long-lasting LTP (Bliss & Collingridge, 1993; Huang et al., 1994). There are now many known types of LTP, including some that are NMDA-independent, some that need and others that do not need the activation of metabotropic receptors, some that involve presynaptic changes more than others, some that are and others that are not regulated by retrograde synaptic messengers, some whose late proteindependent phase is modulated by D 1 receptors and others in which this phase is modulated by b-adrenoceptors, etc. (Bliss & Collingridge, 1993; Collingridge & Bliss, 1995; Huang & Kandel, 1995, 1996; Kleschevnikov, Sokolov, Kuhnt, Dawe, Stephenson, & Voronin, 1997; Malinow, 1994; Nicoll & Malenka, 1995). It is possible, even likely, that the hippocampus will use biochemical sequences similar to those of one or another type of LTP depending on the task and on the subarea that plays a primary role (CA3 or CA1, the subiculum, etc.; see Gabrielli et al., 1997; Mayford et al., 1996). At this stage, it is wise to simply admit that the biochemical events of memory formation in rat hippocampus are very much like those of LTP (Bernabeu et al., 1997a,b; Bevilaqua et al., 1997; Carew, 1996; Izquierdo & Medina, 1995, 1997; Figs. 1 4) and to refrain from proposing causal relationships between the two sets of phenomena. That there may in many cases be an overlap, and, for example, repetitive aversive training may block CA1 LTP (Izaki & Arita, 1996), is the least that can be expected from processes that share the same cell population and many of the same mechanisms. However, many such saturation studies have failed (see Barnes, 1996), even when successive LTPs were used in the same neuron field (Frey, Schollmeier, Reymann, & Seidenbecher, 1995), let alone when the effect of learning on LTP or vice versa was investigated (Barnes, 1996; Bliss & Richter-Levin, 1993). RELATION OF HIPPOCAMPAL EVENTS WITH THOSE IN OTHER REGIONS OF THE BRAIN As mentioned above, some of the early events of memory formation in the hippocampus also occur in amygdala, medial septum, and entorhinal cortex: Early sensitivity to NMDA and AMPA receptor blockers, GABA A receptor agonists, cholinergic muscarinic and noradrenergic agonists and antagonists, and PKC or CaMKII inhibitors (Davis, 1992; Izquierdo et al., 1992; Jerusalinsky et al., 1992, 1993; Jerusalinsky, Quillfeldt, Walz, Da Silva, Bueno e Silva,

15 HIPPOCAMPAL BIOCHEMISTRY AND MEMORY 299 Bianchin, Zanatta, Ruschel, Schmitz, Paczko, Medina, & Izquierdo, 1994a; Wolfman et al., 1994). The late camp/pka-dependent phase is absent in the amygdala (Bevilaqua et al., 1997). Most of the biochemical changes listed above have been observed in CA1, but some have been observed in other hippocampal subareas as well (see above). This should not be surprising, given the direct synaptic connection between CA3 and CA1 (Hyman et al., 1990), and the diverse indirect connections that exist among the various hippocampal subregions (Green, 1964; Iijima, Witter, Ichikawa, Tominaga, Kajiwara, & Matsumoto, 1996; Witter et al., 1989). It is possible (Jerusalinsky et al., 1992, 1994a) that short or abortive forms of LTP-like phenomena (Bliss & Collingridge, 1993; Reymann, 1993) may participate in early posttraining memory processing by the amygdala or medial septum. Involvement of these areas depends on the task. In inhibitory avoidance all of them are necessary; in less emotional tasks, like habituation to a novel environment, the amygdala and the septum are not involved (Izquierdo et al., 1992; Izquierdo & Medina, 1991, 1995; Wolfman et al., 1991); in a highly emotional task (conditioned fear), the amygdala plays a major role (Davis, 1992) and apparently this role includes CaMKII (Mayford et al., 1996), as it does in inhibitory avoidance (Wolfman et al., 1994). The timing of the onset of the initial NMDA-dependent, muscimol-dependent phase of memory in hippocampus (and amygdala), entorhinal, and parietal cortex is sequential, which indicates that these structures operate sequentially and concertedly in the formation of memories (Izquierdo, Quillfeldt, Zanatta, Quevedo, Schaeffer, Schmitz, & Medina, 1997a; Zanatta, Schaeffer, Schmitz, Medina, Quevado, Quillfeldt, & Izquierdo, 1996). This phase extends for a few minutes after training in hippocampus and amygdala, starts 30 min after training in entorhinal cortex, and 60 min after training in posterior parietal cortex. The time in which memory is sensitive to the amnestic effect of AP5 or muscimol is less than 30 min in hippocampus or amygdala, but lasts up to 270 min in the entorhinal or parietal cortex (Zanatta et al., 1996). The possible pathways involved should be searched for among those reviewed by Hyman et al. (1990) and Witter et al. (1989). In addition to its late role in memory processing, other data point to an early intervention of the entorhinal cortex in learning, as a station through which signals eventually reach the hippocampus so that this can play its role in consolidation (Squire, 1992; Jerusalinsky et al., 1994a; Willner et al., 1993). MODULATORY INFLUENCES IN THE IMMEDIATE POSTTRAINING PERIOD The early events (õ30 min after training) are subject to inhibition by GABA A receptors (Brioni, 1993; Izquierdo, 1997; Izquierdo et al., 1992, 1997a) modulated by endogenous benzodiazepine-like substances (Izquierdo & Medina, 1991; Wolfman et al., 1991) and are, in addition, modulated by cholinergic muscarinic and b-noradrenergic synapses (Izquierdo et al., 1992). A variety of central and peripheral modulatory systems also influence memory in the early posttraining period: b-endorphin, cholinergic nicotinic receptors, serotonin, adrenocorticotropin, vasopressin, oxytocin, glucocorticoids, epinephrine, norepinephrine, and glucose (see Bohus, 1994; Brioni, 1993; Gold, 1995; Izquierdo, 1989; Izquierdo & Medina, 1995, 1997; McGaugh, 1989; McGaugh, Cahill, Parent, Mesches, Coleman-Mesches, & Salinas, 1995; Roo-

16 300 IZQUIERDO AND MEDINA zendaal & McGaugh, 1996). Most of these substances are in fact released during many forms of behavioral training, which signifies that many or most memories are actually formed and often retrieved while under their influence (Izquierdo, 1989). Corticosterone also modulates memory consolidation of inhibitory avoidance in the chick (Rose, 1995a). All these substances affect hippocampal LTP (see Gold, 1995; Izquierdo & Medina, 1997). Cholinergic muscarinic agents upregulate NMDA receptors, activate the phosphoinositide cascade, and alter protein kinase C activity (see Jerusalinsky et al., 1997). Catecholamines enhance camp levels and thereby protein kinase A activity (Bevilaqua et al., 1997). Peripherally administered catecholamines influence central catecholamine levels and actions (Gold, 1995; McGaugh et al., 1995), vasopressin and corticotropin influence brain catecholamine levels (see Bohus, 1994; Gold, 1995, for references), and corticotropin, corticosteroids, and opioid agonists and antagonists influence brain norepinephrine effects (see Izquierdo, 1989; McGaugh et al., 1995). Most of these agents act primarily through interactions with brain noradrenergic mechanisms in the amygdala (Cahill & McGaugh, 1996; Gallagher, Kapp, Musty, & Driscoll, 1977; McGaugh, 1989; McGaugh & Cahill, 1997; McGaugh et al., 1995; McGaugh, Introini-Collison, Juler, & Izquierdo, 1986). Their main influence on the hippocampal mechanisms of memory formation is therefore indirect and must consist of interferences with direct or indirect amygdala hippocampus connections (Hyman et al., 1990; Witter et al., 1989) early after training. Peripherally administered opioids and opioid antagonists alter memory also partly through actions on the medial septum (Bostock, Gallagher, & King, 1988; Izquierdo, 1989), which is of course directly linked to the hippocampus (Gray, 1982). Glucocorticoids may also act through specific receptors in hippocampus (De Kloet, Rosenfeld, Van Eekelen, Ratka, Joëls, & Levine, 1989). Peripheral epinephrine, norepinephrine, and corticoids regulate glycemia, and hyperglycemia mimics the memory-facilitatory effects of these substances and enhances hippocampal LTP (Gold, 1995). It is generally agreed that the amygdala is the main brain center that adds emotional colors or tinges to memories at the time of encoding (Cahill & McGaugh, 1996; Gallagher et al., 1977; McGaugh & Cahill, 1997; McGaugh et al., 1995) and at the time of retrieval (see Bechara, Tranel, Damasio, Adolphs, Rockland, & Damasio, 1995; Izquierdo, 1989; Izquierdo et al., 1997a; Rogan & Le Doux, 1996; Scott, Young, Calder, Hellawell, Aggleton, & Johnson, 1997). Some evidence suggests that the medial septum/diagonal band complex may play a similar role (see Cahill & McGaugh, 1996; Gray, 1982; Heimer, Harlan, Alheid, Garcia, & De Olmos, 1997; Izquierdo et al., 1992; Wolfman et al., 1991). Throughout onto- or phylogeny, the corticomedial prefrontal cortex may take over some of these functions (Damasio, 1995). Despite earlier claims (e.g., Bianchin, Walz, Ruschel, Zanatta, Da Silva, Bueno e Silva, Paczko, Medina, & Izquierdo, 1993), most current evidence indicates that the amygdala is not a site of storage (Bevilaqua et al., 1997; Izquierdo et al., 1997; McGaugh & Cahill, 1997; McGaugh et al., 1995; see below). The same may be said of the medial septum/diagonal band area. This structure is believed to influence memory through its direct connection with the hippocampus (Izquierdo & Medina, 1997), and this connection has a timelimited role in memory formation (Kim, Clark, & Thompson, 1995; Cahill & McGaugh, 1996). Hormones and neuromodulators have been proposed to alter memory by