The neurobiology of consolidations, or, how stable is the engram?

Key Words memory, consolidation, persistence, retrieval, reconsolidation

* Abstract Consolidation is the progressive postacquisition stabilization of long-term memory. The term is commonly used to refer to two types of processes: synaptic consolidation, which is accomplished within the first minutes to hours after learning and occurs in all memory systems studied so far; and system consolidation, which takes much longer, and in which memories that are initially dependent upon the hippocampus undergo reorganization and may become hippocampal-independent. The textbook account of consolidation is that for any item in memory, consolidation starts and ends just once. Recently, a heated debate has been revitalized on whether this is indeed the case, or, alternatively, whether memories become labile and must undergo some form of renewed consolidation every time they are activated. This debate focuses attention on fundamental issues concerning the nature of the memory trace, its maturation, persistence, retrievability, and modifiability.

 
CONTENTS 
 
INTRODUCTION 
ON THE TERMINOLOGY OF CONSOLIDATION 
GENERIC CRITERIA FOR CONSOLIDATION 
SYNAPTIC CONSOLIDATION 
THE STANDARD MODEL OF SYNAPTIC CONSOLIDATION 
SYSTEM CONSOLIDATION 
THE STANDARD MODEL OF SYSTEM CONSOLIDATION 
DOES SYSTEM CONSOLIDATION OCCUR ONLY 
  IN DECLARATIVE SYSTEMS? 
WHY CONSOLIDATE? 
RESERVATIONS CONCERNING THE CONSOLIDATION THEORY 
SELECTED RESPONSES TO THE RESERVATIONS 
DO MEMORIES RECONSOLIDATE? 
DIFFERENCES BETWEEN SYNAPTIC CONSOLIDATION 
  AND POSTULATED RECONSOLIDATION 
THE POSSIBILITY OF SYSTEM RECONSOLIDATION 
ON THE MULTIPLE VERSIONS 
  OF THE RECONSOLIDATION HYPOTHESIS 
RESERVATIONS CONCERNING THE STRONGER VERSIONS OF 
  THE RECONSOLIDATION HYPOTHESIS 
SELECTED RESPONSES TO THE RESERVATIONS 
ON THE DISTINCTION BETWEEN RECONSOLIDATION 
  AND DECONSOLIDATION 
EPILOGUE 

INTRODUCTION

In the domain of memory research and theory, consolidation (Latin for "to make firm"), or memory consolidation, refers to the progressive postacquisition stabilization of long-term memory, as well as to the memory phase(s) during which such presumed stabilization takes place (Dudai 2002a). It has long been suggested that fresh memories need time to stabilize, and that often, such traces are prone to interference by distracting stimuli, injuries, or toxins, which, however, lose their effectiveness with the passage of time. The first documented reference to memory consolidation is in the writings of Quintillian, the noted Roman teacher of rhetoric, who tunas his readers' attention to the "curious fact ... that the interval of a single night will greatly increase the strength of the memory," and raises the possibility that "... the power of recollection ... undergoes a process of ripening and maturing during the time which intervenes" (Quintillian first century A.D.). That such posttraining, time-dependent maturation process takes place was therefore probably known in the Middle Ages to orators and mnemonists who were well versed in the writings and mnemonotechniques of their Roman predecessors. The idea resurfaced again occasionally in different versions before the birth of experimental psychology (e.g., Hartley 1810).

The term "consolidation" is attributed to Muller & Pilzecker, who rediscovered, in a series of studies carried out in Gottingen between 1892 and 1900, that memory takes time to fixate, of undergo Konsolidierung (Muller & Pilzecker 1900). Their evidence was based on systematic search for the laws that govern the acquisition and retrieval of verbal material, a la Ebbinghaus (1885). Muller & Pilzecker found that correct recall of the target material improved during the first few minutes after training, and that if presented during the first minutes after training, intervening new stimuli tend to impair recall of the target material (a phenomenon they termed "retroactive inhibition"). They suggested that this reflects a posttraining interval during which associations consolidate in memory. Interestingly, as aptly noted by Lechner et al. (1999), though the Muller & Pilzecker study is frequently cited and referred to as the beginning of the modern era in the prolific field of memory consolidation, it has never been translated in full from German, and non-German readers must rely on abstracts and extracts (Lechner et al. 1999, McDougall 1901). In spite of several reports of failure to replicate the aforementioned Muller & Pilzecker's findings (e.g., McGeoch 1933, reviewed and analyzed in Wixted 2004), their conclusion, that the trace is still uncompleted when training is over, has been overall well consolidated in the collective memory of memory research.

Shortly before Muller & Pilzecker introduced the term, the process of consolidation was also proposed based on clinical data. In "global," organic amnesia, memory of the recent past is commonly affected more than memory of the distant past; this observation is epitomized in Ribot's Law, or the Law of Regression: "Progressive destruction advances progressively from the unstable to the stable" (Ribot 1882). The idea was further elaborated a few years later by Burnham, who in a signal paper on amnesia integrated findings from experimental psychology and neurology, while at the same time emphasizing the dynamic nature of postexperience memory maturation: "There must be time for the processes of organization and assimilation (of memory) to take place. There must be time for nature to do her part.... Hurry defeats its own end" (Burnham 1903). It is noteworthy that Burnham's "time" actually refers to two different types of consolidation kinetics: fast, such as unveiled by the studies of Muller & Pilzecker, and slow, such as unveiled by the observations of residual premorbid memory in global amnesics. This temporal dichotomy suggests at the outset that the generic term "consolidation" conceals different types of processes and mechanisms. This is reflected in the title of this chapter and is further discussed below.

The quest for the neurobiological foundations of both slow and fast consolidation gained real momentum only in the second half of the last century. Quantitative, systematic studies of retrograde amnesia started to appear in the 1960s and 1970s (e.g., Sanders & Warrington 1971). These were accompanied by the development of animal models of human amnesia and attempts to identify brain substrates critical for slow consolidation (reviewed in Squire et al. 2001). In parallel, neuropharmacology, first systemic and later targeted to selected brain areas, began to unravel molecular candidates for the cellular machinery that subserves fast consolidation (Dudai & Morris 2000; McGaugh 1966, 2000). Cellular preparations and advanced molecular biology and neurogenetics have together revolutionized the field in the past two decades (Dudai 2002a, Milner et al. 1998). For the first time since Quintillian noted it, Ribot conceptualized it, and Muller & Pilzecker named it, we are now in a position to discuss the processes and mechanisms of consolidation at multiple levels of brain organization, from the molecular to the behavioral. Furthermore, the time is now ripe to reevaluate the status of the consolidation hypothesis: How valid is it? And what are the implications concerning the stability and retrievability of engrams and the nature of memory? In this review, I first discuss two main types of neuronal processes to which the term "consolidation" currently refers; then review data that have recently stirred anew a heated debate concerning the nature of the consolidated trace, and discuss their possible interpretation; and finally focus on selected issues that bear on our understanding of the nature of the engram and its persistence.

ON THE TERMINOLOGY OF CONSOLIDATION

The term "consolidation" is currently used in the neuroscience literature to refer to two types of processes, of a family of processes (Dudai 1996, Dudai & Morris 2000; Figure 1). One type is accomplished within the first minutes to hours after the encoding has occurred of practice ended. Ample evidence indicates that this relatively fast type of process takes place in local nodes in the neuronal circuit(s) that encode(s) the experience-dependent internal representation, i.e., the memory. Much attention has been devoted to processes and mechanisms of fast consolidation in synapses; therefore the phenomenon is commonly termed "synaptic consolidation." But it is now evident that this type of consolidation depends on cross talk between synapses and their cell body and nucleus (Dudai & Morris 2000, and see below). The terms "cellular" or "local consolidation" are therefore also appropriate. "Synaptic consolidation" will be preferred in the current discussion, because it focuses on a major site of use-dependent modification in neuronal circuits, and fits the prevailing dogma that synaptic plasticity underlies learning and memory.

[FIGURE 1 OMITTED]

Another type of consolidation process(es) takes weeks, months, or even years to be accomplished. It is believed to involve reorganization over time of the brain circuits, or the systems, that encode the memory, and in the course of this the trace may spread to new locations in the brain while at the same time relinquishing its dependence on parts of the circuits that have subserved its acquisition. This type of process is termed "system consolidation." The term "reorganization" has been proposed, but brain reorganization may occur under conditions other than memory consolidation, such as development, housekeeping, and homeostasis, and response to injury--though similar neural mechanisms might be involved. Moreover, shifting levels of analysis, reorganization applies to synaptic consolidation as well, involving remodeling of synaptic connectivity. "Slow consolidation" is sometimes used to refer to system consolidation, but is questionable because there might be cases in which system consolidation is accomplished within a time frame not so different from that of synaptic consolidation. "Early" and "late" consolidation is also occasionally used, which is fine as far as no implicit assumption is being made about when the process starts. And last, the terms "short-term" and "long-term" consolidation are also used to refer to synaptic and system consolidation, respectively, but this should be better avoided because these terms may connote short of long persistence of the trace, which is a different issue.

GENERIC CRITERIA FOR CONSOLIDATION

Because the assumption that some form or another of consolidation does take place has already attained the status of tenet in the neurobiology of memory (admittedly, with lingering opposition, as noted below), it is prudent to review the criteria for demonstrating that consolidation has indeed taken place in a given system or preparation. By definition, consolidation is progressive postacquisition stabilization of memory. Hence to demonstrate consolidation, a limited time window of susceptibility of long-term memory to an amnesic agent must be proven, following cessation of experience or training. This agent should be devoid of significant effects on sensorimotor faculties required to execute the task, as well as on acquisition per se and on short-term memory, and should not induce state-dependent memory in the protocol used. Monotonous effectiveness of a blocking agent over time suggests that this agent impairs the maintenance, retrieval, or expression of memory, not its consolidation. The aforementioned criterion is necessary and sufficient. Studies that report consolidation merely on the basis of time-dependent postacquisition changes in brain activity use a correlative approach that might supplement the aforementioned criterion, but by themselves do not prove that consolidation has taken place, as the observed changes might not necessarily be causally related to the stabilization or functional reorganization of memory.

SYNAPTIC CONSOLIDATION

Synaptic consolidation is universal; it has been described in all the species, preparations, and memory tasks investigated to date, so far as the task results in long-term memory. Long-term memory, in the context of discussion of synaptic consolidation, is conventionally defined as memory that lasts more than 24 hours, except in the study of long-term potentiation, a popular model of learning-related synaptic plasticity, in which even one hour is considered long. Listed below are selected common themes that emerge from the experimental data on synaptic consolidation (Dudai & Morris 2000):

a. Within a short time after training, new memories become resistant to interferences, or agents, which otherwise are capable of truncating the formation of long-term memory. These types of interferences or blockers include behavioral distractors, drugs, seizures, and anatomical lesions. The time window of susceptibility depends on the task and type of interference or blocker, and ranges from seconds to minutes (e.g., electroconvulsive shock in conditioning, Duncan 1949, McGaugh 1966), to hours (distractor tasks in motor skills, Shadmehr & Holocomb 1997; macromolecular synthesis inhibition in many types of tasks, see below).

b. The stabilization process is not a step-function, but rather various drugs or mutations can be used to dissect it into what appears to be intermediate phases in consolidation (e.g., DeZazzo & Tully 1995, Ghirardi et al. 1995, Grecksch & Matthies 1980, Rosenzweig et al. 1993, Winder et al. 1998). The time course of at least some of these phases is not a given, and could be altered by experimental manipulations, which may mimic in vivo stimuli and processes (Frey & Morris 1997). It is also unclear whether phases of consolidation must take place in a prescribed order for the consolidation to complete successfully.

c. The application of RNA or protein synthesis inhibitors during or immediately after training blocks the formation of long-term memory (Agranoff & Klinger 1964, Davis & Squire 1984, Freeman et al. 1995, Montarolo et al. 1986, Rosenblum et al. 1993). At least in the behaving animal, massive reduction in protein synthesis (>90%) is required to achieve the effect. Usually, a similar transient reduction in macromolecular synthesis does not significantly affect perception, short-term memory, or either the retention or the retrieval of long-term memory once it has been established, with the single (important) exception of application of these inhibitors immediately after retrieval, which is discussed below.

d. Intracellular signal transduction cascades (1), and particularly the cyclic adenosine monophosphate (cAMP) response element (CRE)-mediated modulation of gene expression by such eascades, are thought to play an important role in the consolidation of short- into long-term memory. This has been established in neuronal models of plasticity as well as in behaving animals (e.g., Deisseroth et al. 1996, Frank & Greenberg 1994, Kaang et al. 1993, Lamprecht et al. 1997, Yin et al. 1994). A prominent example is the cAMP cascade, which involves activation of a …