Report

Dopamine Controls Persistence of Long-Term Memory Storage

See allHide authors and affiliations

Science  21 Aug 2009:
Vol. 325, Issue 5943, pp. 1017-1020
DOI: 10.1126/science.1172545

Abstract

The paradigmatic feature of long-term memory (LTM) is its persistence. However, little is known about the mechanisms that make some LTMs last longer than others. In rats, a long-lasting fear LTM vanished rapidly when the D1 dopamine receptor antagonist SCH23390 was injected into the dorsal hippocampus 12 hours, but not immediately or 9 hours, after the fearful experience. Conversely, intrahippocampal application of the D1 agonist SK38393 at the same critical post-training time converted a rapidly decaying fear LTM into a persistent one. This effect was mediated by brain-derived neurotrophic factor and regulated by the ventral tegmental area (VTA). Thus, the persistence of LTM depends on activation of VTA/hippocampus dopaminergic connections and can be specifically modulated by manipulating this system at definite post-learning time points.

Relevant environmental changes trigger transient dopamine-dependent states in the hippocampus that favor memory encoding and synaptic potentiation, perhaps by attaching motivational connotations to experiences (14). The ventral tegmental area (VTA) is critical for assessing the significance of punishments and rewards (5), and it has been postulated that a VTA/hippocampus dopaminergic loop controls the entry of information into long-term memory (LTM) (6). Dopamine regulates the expression of proteins essential for the establishment of lasting neuronal plasticity, such as brain-derived neurotrophic factor (BDNF) (7). In rats, we recently demonstrated that in order to persist, fear LTM requires BDNF expression in the hippocampus 12 hours after training (8). We have now investigated the role of hippocampal dopamine and the VTA on LTM persistence.

The effect of hippocampal dopamine on neuronal plasticity and LTM formation is mainly mediated by D1 receptors (4, 9). To test whether D1 receptors are also involved in LTM persistence, we used a one-trial, step-down, inhibitory avoidance task (IA) in rats (10). IA training can be attuned to induce short- or long-lasting LTM. Training with a weak footshock generated a LTM lasting 2 days or less. Conversely, training with a strong footshock induced a persistent LTM, lasting over 14 days [Fig. 1A; t(18) = 4.40, t(18) = 12.16, t(18) = 6.60, P < 0.001]. Immediately or 9 or 12 hours after strong training, rats received the D1 receptor antagonist SCH23390 (1.5 μg per side) in the CA1 region of the dorsal hippocampus. LTM was tested 2, 7, or 14 days later. SCH23390 had no effect on LTM when given immediately or 9 hours after training. However, when administered 12 hours after training, SCH23390 induced amnesia 7 and 14 but not 2 days later [Fig. 1B; t(14) = 5.30, P < 0.001, and t(14) = 3.59, P < 0.01]. In accordance with this result, when injected 12 hours after weak training, the D1 receptor agonist SKF38393 (12.5 μg per side) enhanced LTM 7 and 14 but not 2 days later [Fig. 1C; t(14) = 3.84 and t(14) = 3.03, P < 0.01]. Intra-CA1 infusion of SKF38393 immediately or 9 hours after weak training did not affect LTM. Co-infusion of SCH23390 12 hours after weak training antagonized the facilitatory effect of intra-CA1 SKF38393 (fig. S1).

Fig. 1

Activation of D1 receptors in dorsal CA1 12 hours after training modulates LTM persistence. (A) Animals were trained in IA with a weak (0.4 mA for 2 s) or a strong (0.8 mA for 2 s) footshock. (B to E) Animals were trained in IA with a strong [(B) and (E)] or a weak [(C) and (D)] footshock and immediately or 9 or 12 hours later received SCH23390 (B), SKF38393 (C), 8Br-cAMP (D), or PKI (E) into CA1. VEH, vehicle. (F) Animals trained in IA with a strong footshock received SKF38393 plus PKI into CA1 12 hours after traning. LTM was assessed either 2, 7, or 14 days after training. n = 8 to 10 rats per group.

Hippocampal D1 receptors are coupled to cyclic adenosine monophosphate (cAMP)–dependent protein kinase (PKA) activation. Thus, we investigated the role of this kinase on LTM persistence. Intra-CA1 infusion of 8-bromoadenosine-3′,5′-cyclic monophosphate (8Br-cAMP) immediately after weak training increased LTM at all post-training times analyzed. Application of 8Br-cAMP (2.5 μg per side) 12 hours after training enhanced retention 7 and 14 but not 2 days later [Fig. 1D; t(14) = 2.40 and t(14) = 2.68, P < 0.05]. Similarly, intra-CA1 infusion of the PKA inhibitor PKI (6.5 μg per side) immediately after strong training hindered LTM 2, 7, and 14 days after training, but when infused 12 hours after training it only hampered retention 7 and 14 days thereafter [Fig. 1E; t(18) = 11.48 and t(18) = 4.57, P < 0.001]. Co-infusion of SKF38393 did not prevent the amnesia induced by PKI [Fig. 1F; t(18) = 4.08, P < 0.001].

The firing pattern of midbrain dopamine neurons changes across the sleep/wake cycle (11), and LTM persistence may depend on the circadian reactivation of PKA (12). However, the effects of SCH23390, SKF38393, PKI, and 8Br-cAMP on LTM were independent of the time of the day at which training was performed (fig. S2).

We next analyzed whether late post-training activation of hippocampal D1 receptors controlled the increase in BDNF required for LTM persistence (8). Strong but not weak IA training increased BDNF levels in dorsal CA1 12 hours after training [Fig. 2A; F(2,11) = 14.41, P < 0.001; q = 4.51]. When given into CA1 11.5 hours after strong training, SCH23390 (1.5 μg per side) blocked the training-induced increase in BDNF [Fig. 2B; F(2,10) = 8.30, P < 0.01; q = 3.75]. Concurring with this, intra-CA1 infusion of SKF38393 (12.5 μg per side) 11.5 hours after weak training increased BDNF levels [Fig. 2C; F(2,12) = 7.12, P < 0.01; q = 3.59]. Intra-CA1 administration of BDNF (0.25 μg per side) 12 hours after weak training increased LTM persistence, whereas infusion of function-blocking antibody to BDNF (1 μg per side) at the same time after strong training induced its rapid decay (fig. S3). Co-infusion of BDNF reversed the amnesic effect of SCH23390 and PKI given into CA1 12 hours after strong training [Fig. 2D; F(4,36) = 24.21, P < 0.001; t = 5.81 and t = 5.21], whereas co-infusion of function-blocking antibody to BDNF impaired the increase in persistence caused by SKF38393 and 8Br-cAMP given 12 hours after weak training [Fig. 2E; F(4,34) = 13.73, P < 0.001; t = 5.47 and t = 3.98], indicating that the action of hippocampal BDNF on LTM persistence is downstream D1 receptors. Amnesia caused by intra-CA1 administration of function-blocking antibody to BDNF 12 hours after strong training was unaffected by co-infusion of SKF38393 or 8Br-cAMP. Neither SCH23390 nor PKI affected the facilitation induced by BDNF infused into CA1 12 hours after weak training (fig. S3).

Fig. 2

Late post-training activation of D1 receptors in dorsal CA1 controls the increase in BDNF required for LTM persistence. (A) Animals trained in IA with a weak or a strong footshock were killed 12 hours thereafter; NAI, naïve. (B and C) Rats trained in IA with a strong (B) or a weak (C) footshock received intra-CA1 SCH23390 (B) or SKF38393 (C) 11.5 hours after training and were killed 30 min later. In all cases, the dorsal CA1 was dissected out, and total homogenates were analyzed by immunoblotting with antibodies to BDNF. n = 4 or 5 rats per group. (D and E) Animals were trained in IA with a strong (D) or a weak (E) footshock and 12 hours later received SCH23390 plus BDNF (D), PKI plus BDNF (D), SKF38393 plus antibody to BDNF (E), or 8Br-cAMP plus antibody to BDNF (E) into dorsal CA1. LTM was assessed 14 days after training. n = 7 to 11 rats per group.

Tyrosine hydroxylase (TH) catalyzes the rate-limiting step in catecholamine biosynthesis. Phosphorylation at Ser40 (pSer40) increases TH activity to allow for replenishment of neurotransmitter stores during increased catecholamine signaling (13). To investigate whether LTM maintenance is indeed associated with activation of the hippocampal dopaminergic system late after acquisition, we measured TH phosphorylation. Strong but not weak IA training increased pSer40-TH in dorsal CA1 12 hours after training and this increase was blocked by intra-VTA infusion of lidocaine (10 μg per side) 11.5 hours after training (fig. S4).

It has been proposed that VTA/hippocampus connections gate relevant information into LTM through a D1/D5-dependent mechanism (6). Dopamine modulates late long-term potentiation (LTP) in CA1 (14), whereas VTA N-methyl-d-aspartate receptors (NMDARs) regulate the development of dopamine-mediated long-term changes in the brain reward circuitry, control the establishment of persistent behaviors, and cause burst discharge of dopamine neurons (1517). Therefore, we examined whether LTM persistence requires NMDAR activation in the VTA late after training. Intra-VTA infusion of the NMDAR antagonist AP5 (1 μg per side) 12 hours after strong training hindered LTM 14 but not 2 days later [Fig. 3A; t(18) = 5.15, P < 0.001]. This was prevented by co-infusion of SKF38393 (12.5 μg per side) or BDNF (0.25 μg per side) into dorsal CA1 [Fig. 3B; F(3,38) = 10.93, P < 0.001; t = 4.36 and t = 4.92]. Amnesia by intra-VTA AP5 was not reversed by co-infusion of SKF38393 into the nucleus accumbens 12 hours after training (fig. S5). NMDA (0.1 μg per side) given into the VTA 12 hours after weak training enhanced LTM retention 14 but not 2 days after training [Fig. 3C; t(16) = 3.59, P < 0.01], an effect blocked by co-injection of SCH23390 (1.5 μg per side) or antibody to BDNF (1 μg per side) into CA1 [Fig. 3D; F(3,38) = 8.86, P < 0.001; t = 3.74 and t = 3.90]. The increase in CA1 BDNF and pSer40-TH caused by strong IA training was blocked by AP5 given into the VTA 11.5 hours after training [Fig. 3E; F(2,11) = 7.90, P < 0.01; t = 3.58 and F(2,9) = 6.42, P < 0.05; t = 3.20], whereas intra-VTA infusion of NMDA 11.5 hours after weak training increased BDNF and pSer40-TH in dorsal CA1 30 min later [Fig. 3F; F(2,11) = 10.97, P < 0.01, t = 4.41; and F(2,8) = 15.36, P < 0.01, t = 3.41]. Thus, NMDAR activation in the VTA around 12 hours after training up-regulates the hippocampal dopaminergic system which, through a D1-dependent mechanism, controls the expression of BDNF required for LTM persistence.

Fig. 3

LTM persistence requires NMDAR activation in the VTA late after training. Animals submitted to strong (A and B) or weak (C and D) IA training received intra-VTA infusions of AP5 (A); intra-VTA AP5 plus intra-CA1 infusions of vehicle, SKF38393, or BDNF (B); intra-VTA NMDA (C); or intra-VTA NMDA plus intra-CA1 infusions of vehicle, SCH23390, or function-blocking antibody to BDNF (D) 12 hours after training. LTM was assessed 2 or 14 days later. n = 8 to 11 rats per group. Animals submitted to strong (E) or weak (F) IA training received intra-VTA infusions of AP5 (E) or NMDA (F) 11.5 hours after training and were killed 30 min later. The dorsal CA1 was dissected out, and total homogenates were analyzed by immunoblotting with antibodies to BDNF or pSer40-TH. n = 4 or 5 rats per group.

A strong footshock is more salient than a weak one. Footshock strength correlates with LTM persistence (Fig. 1A). Because the VTA modulates the fear-arousing properties of footshock (18), we asked whether VTA inactivation immediately after training also affected LTM maintenance. AP5 (1 μg per side) given into the VTA right after strong training hampered LTM retention 14 but not 2 days after training [Fig. 4A; t(18) = 3.58, P < 0.01] and blocked the increase in BDNF and pSer40-TH that happens in dorsal CA1 12 hours later [Fig. 4B; F(2,11) = 9.55, P < 0.01, t = 3.95; and F(2,9) = 16.14, P < 0.01, t = 5.55]. Conversely, intra-VTA infusion of NMDA (0.1 μg per side) immediately after weak training enhanced LTM 14 but not 2 days after training [Fig. 4C; t(16) = 2.12, P < 0.05] and increased BDNF and pSer40-TH in dorsal CA1 12 hours thereafter [Fig. 4D; F(2,11) = 3.95, P < 0.05, t = 2.45; and F(2,9) = 9.39, P < 0.01, t = 4.12]. Neither AP5 nor NMDA affected LTM persistence when given into the VTA 9 hours after training (fig. S6). Intra-VTA infusion of the AMPAR antagonist 6-cyano-7-nitroquinoxaline-2,3-dione (0.07 μg per side) or the γ-aminobutyric acid type A receptor agonist muscimol (0.05 μg per side) 9 hours after strong training had no effect on retention (fig. S7). Thus, early post-training activation of the VTA is also essential to determine the duration of LTM storage. However, this role is different from that played by the VTA 12 hours later, because it does not involve hippocampal D1 receptors (Fig. 1).

Fig. 4

Early post-training activation of the VTA determines LTM duration. (A) Animals submitted to strong IA training received intra-VTA vehicle or AP5 immediately after training. LTM was tested 2 or 14 days later. (B) Animals were trained and treated as in (A) but killed 12 hours later. The dorsal CA1 was dissected out, and total homogenates were immunoblotted with antibodies to BDNF or pSer40-TH. (C) Animals submitted to weak IA training received intra-VTA vehicle or NMDA immediately affter training. LTM was tested as in (A). (D) Animals trained and treated as in (C) were killed 12 hours later, and total homogenates were immunoblotted as in (B). Animals submitted to strong (E and G) or weak (F and H) IA training received AP5 [(E) and (G)] or NMDA [(F) and (H)] into the VTA immediately after training and 12 hours later were given intra-CA1 SKF38393 or BDNF (E), intra-CA1 SCH23390 or function-blocking antibody to BDNF (F), intra-VTA NMDA (G), or intra-VTA AP5 (H). LTM was tested 14 days after training. n = 4 or 5 or 8 to 10 rats per group for the biochemical and pharmacological experiments, respectively.

Is there any functional relationship between the early post-training activation of the VTA and the delayed VTA/CA1 dopaminergic interaction relevant for LTM persistence? Intra-CA1 infusion of SKF38393 or BDNF 12 hours after strong training reversed the amnesic effect of the immediate post-training intra-VTA administration of AP5 [Fig. 4E; F(3,33) = 5.19, P < 0.01; t = 3.33 and t = 3.53]. Moreover, when given into dorsal CA1 12 hours after weak training, SCH23390 or function-blocking antibody to BDNF hindered the promnesic action of the immediate post-training intra-VTA administration of NMDA [Fig. 4F; F(3,28) = 5.12, P < 0.01; t = 3.01 and t = 3.29]. Intra-VTA infusion of NMDA 12 hours after strong training did not reverse the decrease in LTM persistence induced by the immediate post-training administration of AP5 into the VTA [Fig. 4G; F(2,24) = 11.13, P < 0.001; t = 4.18 and t = 3.98]. Intra-VTA infusion of AP5 12 hours after weak IA training did not affect the increase in LTM persistence induced by the immediate post-training administration of NMDA into the VTA [Fig. 4H; F(2,21) = 6.96, P < 0.01; t = 2.63 and t = 3.60].

Most findings suggest that dopamine modulates LTM encoding and consolidation during or early after training (4, 9). Our data indicate that dopamine also controls the maintenance of LTM storage through a late post-acquisition mechanism involving D1 receptors signaling in dorsal CA1. This regulates the expression of BDNF necessary for LTM persistence and requires NMDAR activity in the VTA immediately and 12 hours after training. Pharmacological activation of NMDAR in the VTA at these critical time points induces BDNF expression in CA1 and converts a rapidly decaying LTM into a persistent one. Moreover, the fact that the reduction in LTM persistence caused by AP5 in the VTA is reversed by intra-CA1 infusion of SKF38393 or BDNF 12 hours after training indicates that long-term maintenance of the learned response depends on a hierarchic sequential process involving VTA-hippocampus dopamine connections.

We propose that the VTA, or the VTA/hippocampus loop (6), is specifically activated during motivationally relevant experiences and, through a process that culminates with a D1-dependent increase of BDNF in the dorsal hippocampus, determines LTM duration. The two periods during which the VTA is required for LTM persistence are linked by a mechanism that does not require the continuous participation of its main excitatory and inhibitory inputs. NMDAR-dependent phenomena that initiate delayed and long-lasting modifications in the efficacy of excitatory synapses on VTA dopamine neurons, probably underlying persistent behaviors, have been described (19, 20), and evidence indicates that activation of D1/D5 receptors in the hippocampus is necessary for the persistence but not induction of synaptic potentiation (2124).

The idea that the entrance of new information into LTM depends on dopamine mechanisms controlled by VTA/hippocampus interactions has been postulated on the basis of theoretical considerations (2, 6) and is partially supported by empirical data (4). Our findings suggest that the role of these interactions is broader than previously thought. This is important for understanding the neural basis of persistent behaviors and for the development of treatments able to modulate not merely the entry but also maintenance of learned information into LTM.

Supporting Online Material

www.sciencemag.org/cgi/content/full/325/5943/1017/DC1

Materials and Methods

Figs. S1 to S7

References

References and Notes

  1. This work was supported by the Conselho Nacional de Desenvolvimento Científico e Tecnológico and Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (Brazil) and the Consejo Nacional de Investigaciones Científicas y Técnicas and Agencia Nacional de Promoción Científica y Tecnológica (Argentina).
View Abstract

Stay Connected to Science

Navigate This Article