Sites of Neocortical Reorganization Critical for Remote Spatial Memory

See allHide authors and affiliations

Science  02 Jul 2004:
Vol. 305, Issue 5680, pp. 96-99
DOI: 10.1126/science.1098180

This article has a correction. Please see:


The hippocampus is crucial for spatial memory formation, yet it does not store long-lasting memories. By combining functional brain imaging and region-specific neuronal inactivation in mice, we identified prefrontal and anterior cingulate cortex as critical for storage and retrieval of remote spatial memories. Imaging of activity-dependent genes also revealed an involvement of parietal and retrosplenial cortices during consolidation of remote memory. Long-term memory storage within some of these neocortical regions was accompanied by structural changes including synaptogenesis and laminar reorganization, concomitant with a functional disengagement of the hippocampus and posterior cingulate cortices. Thus, consolidation of spatial memory requires a time-dependent hippocampal-cortical dialogue, ultimately enabling widespread cortical networks to mediate effortful recall and use of cortically stored remote memories independently.

Memory consolidation is the process whereby recently acquired information gradually transforms from an initially labile state into enduring stable memories (13). The hippocampus is crucial for the formation of new declarative memories (2, 4, 5). However, its role as a consolidation organizing device is time-limited (2, 6, 7); after a time, brain regions, not yet identified but presumably neocortical sites, independently mediate retrieval of remote memories (2, 68).

To identify these extrahippocampal regions involved in processing remote spatial memories, we mapped the regional expression of the inducible immediate early genes (IEGs) zif268 and c-fos during memory processing in mice (fig. S1). These IEGs are required for synaptic plasticity and memory formation (911) and are used as an index of neuronal activation (911). Independent groups of C57BL/6 mice were trained to locate the invariant position of a baited arm in a five-arm maze (12, 13). Either 1 day (recent) or 30 days (remote) after memory acquisition, all groups were tested by using a single trial of memory retention (13). In order to control for nonmnemonic aspects (e.g., contextual arousal, locomotor activity) of the spatial reference memory procedure that could affect gene expression, we used additional paired control mice exposed to the same stimuli but not confronted with any arm choice during testing. Gene expression specifically induced by memory processing was evaluated by normalizing IEG counts from experimental animals with respect to paired controls tested either 1 or 30 days after acquisition. We found a significant group × time × brain region interaction [Zif268 staining: F(13,504) = 5.48, P < 0.001; Fos staining: F(20,756) = 5.08, P < 0.001], indicating that region-specific changes in neuronal activation occurred as a function of the retention intervals. Between recent (1 day later) and remote (30 days later) memory testing, a significant increase in the number of Zif268 immunoreactive neurons was seen in widespread cortical regions including the prefrontal, anterior cingulate, and retrosplenial cortices (Fig. 1, A and B). Increases were also observed when Fos expression was examined in the prefrontal and anterior cingulate cortices (fig. S2), indicating that the observed changes in Zif268 expression generalized to another activity-dependent gene involved in memory formation (11). Response accuracy in animals tested on either day 1 or 30 was similar [80.0 ± 13.3% versus 70.0 ± 15.3, respectively, F(1,18) = 0.24, P > 0.62; fig. S3], which showed that the observed changes in gene expression were not simply related to the level of memory performance but reflected an increased functional implication of specific neocortical regions in long-term memory storage and retrieval.

Fig. 1.

Cortical reorganization during remote spatial memory processing. (A) Zif268 counts relative to paired controls in the prefrontal, anterior cingulate, posterior cingulate, and retrosplenial cortices after testing for recent (day 1) or remote (day 30) memory retention. (B) Photomicrographs showing increased Zif268 (top) and GAP-43 (bottom) labeling in anterior cingulate cortex (aCC) on day 30 as compared with day 1. Nuclei were stained with 4′,6′-diamidino-2-phenylindole (DAPI) (blue); GAP-43 labeling appears in green. (C) A time-dependent redistribution of Zif268 immunopositive nuclei within parietal layers II, III, IV, and VI occurred in experimental animals (top left) between day 1 (white bars) and day 30 (black bars). This laminar redistribution was not observed in paired control subjects (top right). Corresponding photomicrographs of Zif268 staining within parietal cortical layers of experimental animals tested on day 1 or day 30 are shown below. *P < 0.01 versus respective controls (100% line); †P < 0.01 versus day 1; n = 10 mice per group. Scale bars, (B), top, 100 μm; bottom, 10 μm; (C), 100 μm.

We then asked whether this cortical reorganization during memory consolidation was also accompanied by structural changes within neuronal networks. We thus examined, as a function of the retention interval, the cortical expression of growth-associated protein 43 (GAP-43), a presynaptic protein that controls axon growth and sprouting (14) and that is used as a marker of newly formed synapses (14, 15). We found that animals from the 30-day memory retention group exhibited increased GAP-43 labeling in the anterior cingulate cortex as compared with the 1-day retention group [42.58 ± 3.31 versus 22.71 ± 3.59, respectively; F(1,13) = 16.61, P < 0.01; Fig. 1B], which positively correlated with the number of Zif268 immunopositive neurons [r = 0.94; F(1,7) = 48.48, P < 0.001; fig. S4]. The enhanced Zif268 expression observed after 30 days may thus be attributed, at least in part, to synaptogenesis, which increases the complexity and extent of cortical networks involved in memory storage. Indeed, increased GAP-43 labeling was also observed in the prefrontal cortex on day 30 [18.50 ± 2.38 versus 9.0 ± 2.27 on day 1, F(1,6) = 7.08, P < 0.05], although changes in synaptic remodeling were not analyzed systematically in all cortical regions that showed increased IEG expression. In addition to promoting experience-dependent synaptic growth during the 30-day retention interval, the maintenance of elevated levels of GAP-43 proteins observed at the time of remote memory testing may have facilitated synaptic function via increased neurotransmitter release and enhanced long-term potentiation within the reorganized neocortical neuronal networks (14, 15).

Current connectionist models of memory consolidation predict that remote memory storage via strengthening of cortico-cortical connections could be accomplished without necessarily requiring any additional neuronal activity (6, 8). In accordance with this prediction, we observed, in the parietal cortex, a shift in the distribution pattern of neuronal activation from deep cortical layers V and VI to superficial layers II, III, and IV in experimental, but not in control, subjects retrieving remote versus recent memory [Zif268 staining: delay × layer interaction, F(3,72) = 46.72, P < 0.001, Fig. 1C; Fos staining: F(3,72) = 81.30, P < 0.001; fig. S4]. Although based solely on correlative data (i.e., altered levels of IEG expression), this laminar reorganization is consistent with the suggested implication of laminae II and III in catalyzing the formation of cortico-cortical neural assemblies and thereby constituting major sites of storage of encoded information (16). The laminar reorganization of activity within the parietal cortex, together with the recruitment of the retrosplenial cortex, may thus reflect functional participation of these regions in the associative processing of complex visuospatial representations (17), including the gradual establishment of extrahippocampal spatial maps (17, 18).

Consistent with a transitory role of hippocampal formation in memory consolidation (2, 68, 10), the reorganization in cortical regions was associated with a concomitant decrease of Zif268 immunoreactivity induced by remote memory testing in the dorsal [F(1,36) = 34.56, P < 0.001)] and ventral hippocampus [F(1,36) = 25.43, P < 0.001], as well as the entorhinal cortex [F(1,36) = 6.12, P < 0.05] (Fig. 2, A and B; fig. S5). It should be noted that Zif268 expression in the hippocampus was significantly lower than that of paired control subjects (Fig. 2A), which raises the possibility that inhibitory influences may ultimately control the level of engagement of the hippocampal formation in memory consolidation (7). A similar decrease was also observed in the posterior cingulate cortex [Fig. 1A; F(1,36) = 6.70, P < 0.05], which suggests a conjoint involvement of hippocampal and certain cortical networks during early memory processing (7). Hippocampal disengagement was specifically related to memory consolidation, since mice tested over the same time period in a working memory paradigm in which information changes from trial to trial (and is thus stored only temporarily) did not show decreased Zif268 expression in the hippocampus [Fig. 2C and fig. S5, memory type × time interaction: F(1,34) = 31.67, P < 0.001].

Fig. 2.

Time-limited role of hippocampus in remote spatial memory storage and retrieval. (A) Zif268 counts relative to controls in dorsal (top) and ventral (bottom) hippocampus after testing for recent (day 1) versus remote (day 30) memory retention. Zif268 expression was elevated in these structures after testing recent memory but, in contrast, was decreased below control levels after testing remote memory. (B) Photomicrographs of Zif268 labeling in dorsal hippocampus (CA1d) after testing for spatial reference memory, recent (top) as compared to remote (bottom). (C) Hippocampal disengagement, shown in (B), was not observed in animals tested for working memory. *P < 0.01 versus respective controls (100% line); †P < 0.01 versus day 1; n = 10 mice per group. Scale bars, 50 μm.

We next infused the anesthetic lidocaine into selected brain regions before testing for memory retention to transiently silence neuronal activity and thereby to minimize any possible compensatory mechanisms within memory systems associated with irreversible lesions (13). Inactivation of hippocampus or posterior cingulate cortex disrupted recent, but not remote, memory retrieval (Fig. 3, A and B; fig. S6), which indicated that these two interconnected brain regions (19) mediate information processing in parallel during early stages of memory consolidation (7). In contrast, silencing neuronal activity in prefrontal or anterior cingulate cortex selectively disrupted retrieval of remote memories (Fig. 3, A and B; fig. S6). These findings provide evidence for a critical requirement for specific neocortical regions in remote spatial memory retrieval and indicate that the hippocampus does not play a permanent role for as long as memories remain viable (68, 20). They are consistent with two observations: that functional disruption of cortical plasticity preferentially disrupts the establishment of enduring memories in CaMKII knockout mice (21, 22) and that in semantic dementia in humans, neocortical atrophy selectively disrupts the retrieval of remote memories (23). Our data further reveal the progressive establishment of a cortical memory store during the consolidation process, in which spatial information is slowly embedded within distributed but structured cortical networks. Conscious recollection and offline reactivation of hippocampalcortical networks during phases of sleep (24) have been proposed as possible mechanisms ensuring memory processing and reconsol-idation of memory traces (25), which could thereby enable the updating of preexisting cortical memory representations. The recruitment of prefrontal, anterior cingulate, and retrosplenial cortices we report here may reflect the integrative and evolving roles of these areas in memory storage (22), effortful recall (26), performance monitoring (27), and use of consolidated remote memories (22, 28, 29).

Fig. 3.

(A) Effects of neuronal inactivation of the prefrontal, anterior or posterior cingulate cortices, or dorsal hippocampus by lidocaine (black bars) as compared with vehicle [artificial cerebrospinal fluid (aCSF), white bars] on recent (day 1) and remote (day 30) memory retrieval. (B) Coronal diagrams [adapted from (30)] showing the location of injection sites in prefrontal cortex (top) and dorsal hippocampus (CA1d, bottom) plotted onto a single plane (AP +1.7 mm and –2.0 mm, respectively). Efficacy of lidocaine inactivation was verified by assessing the area exhibiting blockade of Zif268 staining. Injection sites for subjects receiving vehicle (open triangles) are plotted on the left; the injection sites (black triangles) and maximum extent (shaded blue area) of lidocaine inactivation are shown on the right of each section. Animals were included in the study only if cannula tips were correctly located within targeted structures and complete IEG inactivation was circumscribed to the region of interest with minimal diffusion spread along the guide cannula track. *P < 0.05 and **P < 0.01 versus vehicle-injected controls; n = 7 to 10 mice per group. Scale bar, 100 μm.

In summary, our findings point to a shift in hippocampal and neocortical roles during the course of memory consolidation that ultimately enables specific cortical areas to assume responsibility for remote spatial memory retrieval. The key cortical sites we have identified constitute primary targets for unveiling the precise cellular and molecular bases of the changes occurring in cortical synaptic connectivity as memories consolidate.

Supporting Online Material

Materials and Methods

Figs. S1 to S6

References and Notes

References and Notes

View Abstract

Stay Connected to Science

Navigate This Article