Cristina Alberini, Riccardo Brambilla, Jack (John) Byrne, Ronald (Ron) Davis (organiser), Yadin Dudai, Walter Gehring (organiser), Martin Heisenberg, Takeshi Ishihara, Pierre-Marie Lledo, Beat Lutz, Isabelle Mansuy, Karim Nader, Thomas Préat, Susan J. Sara, Scott Waddel
Ronald Davis and Walter Gehring
27 août-2 septembre 2003
Sitting in our offices, our minds are overwhelmed with memories of our workshop in Les Treilles. We remember the taste of the fresh figs on the path, the views over Provence, the sculpture garden, the warmth of the morning sun and the views of the stars and Mars at night. Memories that are truly vivid and alive. This is particularly apt given that our memories are inspiration and subject. Our Les Treilles workshop centered on how memories are processed by the brain. Over 100 years ago, the French clinician Ribot (1880) described in his treatise “Diseases of Memory” how new memories are more susceptible to being disrupted than older memories. Ebbinghaus (1885) and Müller & Pilzecker (1900) confirmed that memories are time-dependent phenomena – new memories are susceptible to disruption while older memories seemed to be ‘fixed’ in the brain. This time-dependence became the dominant view of memory processing with the addition that the time dependent step was the storage of the trace in the brain. We now refer to the time dependent process that stores memories as memory consolidation.
The insights of these 4 great minds have inspired generations of scientists to understand how the brain stores memories. Scientists have asked whether there is a single learning and memory system or multiple. We have found many. We have searched in simple invertebrate systems such as the nematode (Caenorhabditis elegans), the marine snail (Aplysia californica) and the fruit fly (Drosophila melanogaster) to mammals – most frequently rodents – for how memories are stored in the brain. Across systems the answer seems to be that learning is mediated by intracellular signaling cascades that precipitate cellular changes that are initially labile. If an animal is appropriately trained, long-term memories are formed by these same signaling cascades recruiting transcription factors, new mRNA and proteins. These proteins are then transported to the relevant synapses that are part of the neural network representing the ‘stored’ network. Longer-term memories are therefore protein synthesis dependent and are allegedly more stable involving long-lasting structural change. A partially satisfying answer but what are these proteins and what are the molecules that signal that information is being stored?
Intracellular signaling is activated by extracellular ligands. Some of the important transmitters are already known. Serotonin is a key transmitter in the snail model (as discussed by Jack Byrne). Martin Heisenberg presented data that suggests that, at least in fruit flies, different monoamine transmitter systems serve distinct roles in different forms of learning. Octopamine is required to learn when an odor is paired with punishment (ie. foot shock) and dopamine is required to pair odor with pleasure (food). Thomas Preat presented data that dopamine is also required during memory consolidation. In addition, information from nematodes and fruit flies suggests that peptide transmitters encoded by the HEN1 gene (Takeshi Ishihara) and the amnesiac gene (Scott Waddell) contribute to memory related signaling and perhaps consolidation (Scott Waddell).
Signaling by these molecules induces intracellular events and it is of great interest to decipher the nature of the downstream events. Many of these transmitters activate cyclic AMP (cAMP) synthesis and ras– Mitogen Activated protein kinase signaling–MAPK- (Ricardo Brambilla). Current dogma states that these kinase cascades are activated by, and are necessary for, learning. Furthermore, prolonged activation of cAMP signaling induces long-term memory. Cyclic AMP activates kinases and then phosphatases compete with kinases. Consistent with this simple model, calcineurin and protein phosphatase 1 inhibition improves memory in mouse models (Isabelle Mansuy).
If long-term memory is induced, the transcription factor CREB (cAMP response element binding protein) drives expression of new genes. Finding these genes is a major endeavor and some candidates were presented at the meeting. In Aplysia the protease Ap TBL-1 is an early-induced gene. TBL-1 cleaves the growth factor, TGFb. TGFb application alone causes some of the critical neuronal changes triggered by learning. CREB also induces another transcription factor –C/EBP. C/EBP in turn drives the expression of a number of genes including muscle specific tyrosine kinase –MuSK- (Cristina Alberini). Contrary to it’s name, MuSK is expressed in the mouse brain and appears to be induced by long-term memory training. In fruit flies a distant relative of MuSK encoded by the off-track gene is required for memory (Scott Waddell). Thomas Preat also presented the fruit fly crammer gene that is required for LTM and is also induced by LTM training.
Clearly there is considerable complexity in the molecular architecture of memory and it seems the same will be true of turning memories off. We can learn association and then learn to inhibit those associations (memory extinction). The endogenous cannabinoid (CB1) system is a critical component of the mechanisms involved in memory extinction but interestingly not in memory acquisition (Beat Lutz). CB1 mutant mice could be fear conditioned but these fearful memories could not be extinguished.
Beyond the molecules, there is also the likelihood that the cellular make-up of the brain is not as static as previously believed. Neurogenesis accounts for the appearance of 80, 000 new neurons per day in the rodent olfactory bulb (Pierre-Marie Lledo). That’s an amazing 1% of the total population and suggests this is truly a dynamic brain region. These new neurons travel for four days in migratory streams following guidance molecules. In an odor rich environment more new cells survive and strikingly, this improves the ability to recognize familiar versus novel odors! Therefore it is conceivable that the neurons that comprise circuits may also be changed in response to memory. There are 9, 000 new cells /day in the rodent dentate gyrus – a region more commonly associated with memory processes. Thus, neurogenesis may also contribute to the formation and storage of memories.
Invertebrate systems like snails, flies and worms that are readily amenable to molecular manipulation, would be even richer if neuronal signaling could be visualized while the animal learns. Using a reporter system that senses synaptic vesicle release, Ron Davis monitored synaptic events in the antennal lobe while the fly was trained to associate an odor with an electric shock. Strikingly, an additional glomerulus -VA7 – appears to be activated by coincident odor and shock cues. This glomerulus is not activated by either cue alone.
Also leading off from his seminal studies using the fruit fly, Walter Gehring presented a fascinating story on the conserved role of PAX6 gene homologs in the development of evolutionarily and morphologically distinct eye structures. PAX genes regulate eye development from mammals to animals like jellyfish – that don’t even have a brain! This led Walter Gehring to conclude that the brain likely evolved after an ability to sense light because to paraphrase: “I can’t see any evolutionary advantage of losing your brain.”
We are working within the traditions of memory consolidation attempting to explain how memories become fixed over time, we have also embarked on deciphering new properties of memories. In 1932 Bartlett pointed out that memories are not snapshots passively read out by the brain, but are instead dynamic and reconstructive in nature. It has always seemed difficult to comprehend how ‘fixed’ memories can at the same time be dynamic or reconstructed. However, we can now reconcile these two conceptually distant traditions of memory fixation with reconstruction. To this end, we need to invoke the property of consolidated memories to return to a sensitive labile state when they are reactivated, a phenomenon first demonstrated by Donald Lewis (1968). In this way, a fixed memory can return to a labile state during which time it can be weakened, strengthened or changed (a reconsolidation step). There are many issues concerning reconsolidation such as what the relationship is between consolidation and reconsolidation? Is reconsolidation simply a recapitulation of consolidation? Do all memories undergo reconsolidation or not? How does reconsolidation interact with other memories? Already there are many novel and exciting findings.
Susan Sara, one of the original people working on issues of reconsolidation reviewed her classic work using the maze originally used by Donald Lewis, discussed how reactivation of a consolidated memory made the memory sensitive to nor-adrenergic modulation, thus demonstrating that reconsolidation is sensitive to mono-aminergic stimulation. The mechanism that is engaged by emotional memories to strengthen memory consolidation is thus re-engaged during reconsolidation. Karim Nader extended the work on reconsolidation to show the universal requirement of protein synthesis in consolidation of memories is recapitulated during reconsolidation. Indeed, direct challenge of protein synthesis inhibitors into either the amygdala or hippocampus blocks the reconsolidation of emotional and contextual memories, respectively. These findings demonstrate that protein synthesis requirements for reconsolidation generalize across learning systems and types of memories.
Yadin Dudai then discussed the first boundary condition for this new phenomenon. He demonstrated that when the reactivation of consolidated memories induces significant extinction, extinction and not reconsolidation, is the process most affected by an amnesic treatment. Interestingly, when the reactivation session does not induce significant extinction, reconsolidation is blocked. This pattern of results was demonstrated in rodents and fish.
We have been successful in unveiling some of the mechanisms that confer memories with their time dependent properties. We are beginning to grasp how and when memories become “unfixed” so that they can be reconstructed. In so doing we are extending our scientific inquiry into a more dynamic component of memory that was until recently only in the domain of cognitive psychologists.