Abi-Rached: Could you tell us more about the recent discovery of ‘enzyme PKMzeta’? What new insights does this new finding bring to the neurobiology of memory?

Dudai: Until recently the basic conceptual framework in memory research was that memories are acquired and then kept until use, unless they fall into oblivion. For instance, I met you for the first time yesterday so I encoded your face and associated it with your name. That memory becomes consolidated over time unless something happens in between. For example, someone might say to me, ‘No, this is not the person you think it is.’ Once consolidated, the memory becomes ‘stored’— it is important to note that ‘storage’ is a metaphor—in a form that we dub long-term memory [LTM].You can then use that memory item or not. According to this textbook account and until a few years ago, people thought that the only way to prevent the memory from being kept in the long-term is to prevent its consolidation, which means that in the first few hours, or at most days, after you acquire the memory you do something to prevent it from going into long-term storage. In animal research we achieve this by administering various drugs. In humans, something similar may be achieved by using certain medications, such as certain tranquilizers of the benzodiazepines family, for example, Flunitrazepam.

However, few years ago an idea was revitalized in the field, mainly due the work of Susan Sara in Paris, Karim Nader and Joseph LeDoux in New York, and their colleagues. This intriguing idea is that items in memory are not consolidated just once per item and then become stable, but rather, such items become re-susceptible to interferences when you retrieve them. So if I go home and tell my friends that I met you in a café in Paris, at that moment of retrieval, when that memory becomes activated again, it reacquires for a short period some of the properties of a fresh memory, and you can disrupt it more or less similarly to the way you block consolidation of a new memory. By doing so you can change the old memory. In animal research this concept is called ‘reconsolidation’ because the post-retrieval phase recapitulates some—though not all—of the attributes of consolidation. In humans it might be much more difficult to do, but in principle one might think of methods to explore the possibility of interfering with reconsolidation in humans as well. Note that even if reconsolidation is possible, you still have to reactivate the memory, which means that even according to the new view the ‘trace’ (the physical imprint of the memory) or memory (I use both terms interchangeably here) exists in at least two alternating phases, active and inactive. When it is active it is more plastic and modifiable, and this happens when you acquire it and when you retrieve it. When you acquire it, it is active and when you retrieve it, it is re-activated. But still the memory has to be in an active mode, which means that it has to be in use, in order for you to intervene or try to intervene.

Now recent evidence focused on the PKM enzyme suggests that probably we do not even need to activate the trace in order to be able to modify it. Rather, there are opportunities which exist throughout the life of the memory, of which we still don’t know the physiological and functional meaning, in which you can disrupt an item in LTM. In other words, overt or explicit re-activation might not be necessary. This finding was originally discovered by neuroscientist Todd C. Sacktor in New York working on a cellular model called Long-Term Potentiation [LTP]. LTP is the ability to modify the response of a neuronal circuit by strong or repetitive stimulation. This can be done in vivo [in the brain] or in vitro [in a glass or plastic dish]. After stimulation in certain protocols, such as using a tetanic (repetitive) stimulation, you facilitate the ability of that same pathway to transmit information for a very long time. So this is considered a model for memory because a single event, the induction stimulus, alters the ability of a neuronal circuit to encode information for a long time. Using that model in a part of the brain called the hippocampus, which is long known to be important in certain types of memory, Sacktor together with his collaborators [Ling et al., Reference Ling, Benardo, Serrano, Blace, Kelly and Crary2002; Pastalkova et al., Reference Pastalkova, Serrano, Pinkhasove, Wallace, Fenton and Sacktor2006] have shown that if you take a chemical which is a selective inhibitor for a certain enzyme called PKMzeta, you can prevent not the formation of LTP, but its maintenance once the LTP memory has been formed. In other words, it seems like you can potentially erase memory without even bringing it to mind.

There is a lot beyond that which I can’t explain here, but for the sake of accuracy let me say a bit more about PKMzeta. This is a subtype of enzyme called protein kinase [PK] which adds a chemical group called ‘phosphoryl group’ [containing a phosphate] to other proteins, and by doing so it modifies the activity of those other proteins. This change is called phosphorylation and is the most common type of regulatory process in the cell. But the specificity of the phosphorylation depends on the type of enzymes and on their substrates, and there are many protein kinases. One of these subtypes or families is called PKC (A and B were found earlier). PKC is subdivided into other subtypes, one of which is PKCzeta. As with other protein kinases, PKCzeta is composed of two sub-units, a catalytic sub-unit that modifies the target by phosphorylating it as mentioned above, and a regulatory sub-unit. In real life this regulatory sub-unit blocks the catalytic one. You can generate an enzyme which is autonomous by cutting off the catalytic sub-unit. The protein kinase C which is autonomous is called protein kinase M, or PKM. PKCzeta which is autonomous is thus called PKMzeta. You can block that enzyme with various chemicals, some of which are rather non-specific, but one is highly selective, and this is that part of the regulative sub-unit protein which blocks the catalytic unit in real life. You can synthesize it; this is a 30-amino acid peptide. When it interacts with the catalytic sub-unit, it acts like a cork in a bottle.

Now that peptide, called the Zeta Inhibitory Peptide, abbreviated to ZIP, is water-soluble and cell permeable. This is a great advantage because it means that ZIP can freely go into the nerve cell. The disadvantage, which is still an advantage for us, is that because it is a peptide it is rapidly degraded so it acts only transiently. Thus Todd Sacktor blocked LTP by micro-infusing a solution containing ZIP into the rat hippocampus. Following that, we decided that we would like to use it on the cortex because, as we said earlier, this is the ultimate repository of most of our memories [Dudai, Reference Dudai2002]. We micro-infused a specific area of the rat cortex, the insular cortex, that stores taste associations, with ZIP and by doing so we successfully erased long-term, even very long-term, memories [Shema et al., Reference Shema, Sacktor and Dudai2007].

Abi-Rached: Soon after the publication of the discovery, the science news briefing of the National Institute of Mental Health [NIMH, 2007] hinted at the therapeutic implications, namely the potential clinical significance in PTSD. Cognitive enhancing medications might also emerge from these findings in the future [Shema et al., Reference Shema, Sacktor and Dudai2007]. Do you see this finding ultimately translating into clinical practice, or is it too early to say? And what are the hurdles facing the possible clinical translation of this breakthrough in memory research?

Dudai: The discovery indeed received much attention in scientific circles and made the headlines of many newspapers, but one should remember that the experiments were done on rats, and there is a long bumpy road from experiments in rats to potential applicability in humans. Clearly, this research identifies mechanisms that might be of great importance, but currently, those are only clues to a rich world that might yield better understanding of how memory persists, as well as novel pharmaceutical agents to treat memory problems. One of the current difficulties even in rats is that when you apply it to an area of the brain, all the memory in that area seems to be affected. So right now this is not a good model for treating PTSD, because there you wish to target specific associations. In that respect, research along the lines of reconsolidation, where only the reactivated memory is affected, might be more productive for the time being. But if you identify a mechanism you may end up developing something which is more specific. So this is a very important discovery, but this is only a ‘potential’ for more significant advantages. I cannot just drop a bit of ZIP in your coffee to make you lose some memory. First, because ZIP will very probably get degraded before it reaches your brain, and, second, why should I do that? Today there are already drugs available that can prevent or erase memory. These are commonly used in surgical and dental procedures and are much safer and more effective, though they act only for a short time.

The main issue, however, is that we cannot stop the progress of knowledge. So what we have to do is to engage with the meaning of the possibility that in the future we will have efficient methods to erase an item in memory, either in non-specific or specific ways. The use, abuse and misuse of technology are everywhere, and stopping research will not solve the emerging problem. So the idea is not to stop research but to know how to use the knowledge produced. Danger is everywhere, we live with it. If you cross the street you’re taking the risk of getting run over by a car. If you take an overdose of any painkiller that has been prescribed for relieving your pain, you might also die, and the same applies to many medications that we take. That’s why we have rules and regulations. So my premises are the following. First, you cannot stop producing knowledge. Second, knowledge in general has two facets, beneficial and potentially harmful. Third, in every case you have to evaluate what is reality vs. the ‘image of reality’ among the general public. In this case, the reality is that we’re in the very early stages of acquiring knowledge, but because the issue is so enticing and exciting it creates hype and interest, and people jump and protest: ‘Hey you’re erasing memory!’ The reality is, we can’t currently erase human memory unless we use harsh methods which have nothing to do with science—like a blow to the head.

Will we be able to erase memory in specific ways one day? Definitely yes. And the solution to the potential misuse does not consist in regulating scientific development. Right now, all around the world, people are working in their labs on many things that could be potentially used for bad or good ends. The problem here is that people jump to science fiction, ‘Wow we can use that to erase memories!’ But it does not work like that. The process is slow and complex. I do agree that we have to think about and discuss how and when we would want to use memory erasure, before it becomes available.

Abi-Rached: Yet this belief that mental disorders are only brain disorders, that suffering from schizophrenia or bipolar disorder is like having hypertension or diabetes is not shared by everyone and seems to be more prevalent in the scientific community. Many would disagree with reducing ‘mental life’, such as our recollections, to mere brain processes. They would argue that there is something beyond the brain, something which has to do with personal identity, with individuality, with what Paul Ricœur calls a ‘capable person’, that is an individual with personal identity and social status.Footnote 1

Dudai: Interestingly, in the Bible the heart and kidneys rather than the brain were considered important organs. In fact the brain was not even mentioned once. At the time it was thought that people with guilt suffered from severe kidney colic and those who fell in love had heart aches. The brain didn’t seem to have a special status. But let’s return to our present time. Schizophrenia, manic depression and all mental disorders are organic diseases that have organic manifestations and can be treated with organic means, i.e. medications and other methods, but at the same time their aetiology is intimately interwoven with mental and social processes and they can be treated by mental and social means. I just saw a title of a talk in a psychiatric meeting that states that ‘psychotherapy is an excellent biological medication’. We intuitively tend to dissociate the physical from the mental, but the brain doesn’t seem to honour this distinction; therefore both the aetiology of the disease and its treatment can involve both physical and mental facets.

Abi-Rached: I am not disputing the claim that these are organic diseases. What is rather problematic is the reduction of mental life, its uniqueness and rich dimensions, to mere brain processes. Stated differently, I am suggesting that the whole (mental life) is not the sum of its parts (individual brain or neural processes). It seems from the discussion we’ve heard today at the Collège de FranceFootnote 2 that there is nothing called ‘mental life’, that this is an ordinary brain product … that to some extent mental life and the brain are homologous.

Dudai: I think that we have to distinguish here between reduction as a tool and reduction as a philosophy. As a tool, it’s very effective—you try to disassemble a piece of complex machinery by identifying and investigating each part at a time, and then proceed, bottom-up, to explore the interactions of the part in the higher level. Reductionism as a scientific tool is very successful—all our current knowledge about the working of genes and cells, for example, derives from this approach. I am not sure how many investigators who use reductionist tools in their research would claim that the lowest level of analysis is sufficient to explain the entire system. I believe very few think that way. Only many scientists probably do not devote much time to such contemplations because the success of reductionism is enticing and it yields intriguing data. So, following your question, we come to the issue of what is defined as a whole and what are the parts. I would claim that mental activity is included in the whole, which is by definition a sum of its parts, only that we do not yet know how the parts yield that manifestation of activity, mentality, which you define by exclusion as not being an ordinary brain product. It is, by definition, ordinary, we practise it right now while talking, only when we stop for a minute, we watch it in awe and don’t yet understand how it comes to be a product of all these various parts that we identify, because we do not yet understand the translation rules from one level to the other.

Abi-Rached: Since we mentioned reductionism or what you call ‘fallacies’ [Dudai, Reference Dudai, Rees and Rose2004], let us now tackle the problem of animal models. Although animal studies have been and are still essential for medical progress and our understanding of normal physiology, anatomy and development, they fail to fully account for human psychopathologies such as depression, bipolar disorder, obsessive compulsive disorder and so on. One of the reasons is the difference between behaviour and syndrome [Insel, Reference Insel2007]. A rat’s fear does not compare to the embodied traumatic memories that human beings experience. This is because a mental disorder does not result from the excess or deficiency of a single feature but a panoply of different variables at the intersection of behaviour, cognition and emotion [Insel, Reference Insel2007]. What are the limitations of the rodent model [mouse/rat] in memory research? And on what basis can we extrapolate the findings in erasing a particular LTM in the rat to the more complicated model of human traumatic experience?

Dudai: The philosopher Thomas Nagel [1974] wrote a widely cited paper entitled ‘What is it like to be a bat?’ My view is that we will never really know because we will never have a real intentionality for the subjective feeling of another species. So I do not know how much fear conditioning mimics traumatic experience at the subjective level, and I have no doubt actually that, because of the dense associations and connotations of the human mind, it will always be different in humans. But first, the basic molecular and cellular mechanisms and maybe the basic circuits are believed to be similar, and the physiological measures corroborate this assumption. Second, in science, as we just said earlier, we must have simpler models, and those in rodents are the best we have. Therefore, even though nobody would probably claim that rat PTSD is human PTSD, and the rat’s LTM is the human LTM, the basic building blocks are considered to be shared. This assumption requires verification, of course, when a certain line of research culminates in a discovery worth trying on human volunteers according to the proper medical and ethical rules. It is important to note that we very much depend on animal models—nobody would allow interventions in human brain and behaviour in an exploratory manner without data from simpler system—but we have to note what a model means. A model in biology is actually a different system, hence it never really mimics all the properties of the modelled system.

Using animal models, but considering at the same time the richness and ramifications of the nets of personal memories that we have, what the field attempts to do at the end of the day is to erase not necessarily a single complete chunk of memory but to be able to blunt or erase the extremely annoying emotional burden part of the traumatic experience. So you will still remember the traumatic event and be able to process it and work to integrate it productively in your personal mental narrative, but the intensity of the emotion will be blunted. The goal of some investigators in the field is to be able to activate the memory and then give you a medication that mellows the re-enacted trauma. A group of medications are under consideration for this purpose, such as beta-blockers, which are relatively safe. But new and more specific families of pharmacological agents might emerge from current studies, such as those on PKMzeta.

Abi-Rached: Could you elaborate more on the memory of an animal model, what does ‘memory’ in a rat or rodent amount to? How can you mimic trauma in a rat?

Dudai: Let’s take again the case of PTSD. There are three commonly used major models for traumatic memories in animal research. Two of them are variants of what is called ‘fear conditioning’, in which you present a rat or mouse with a stimulus, a tone or flash of light, usually a tone. Then you couple the tone with a mild foot shock, which is not life-threatening but unpleasant. And then the rodent, like a Pavlovian dog, learns to associate the tone with the shock and fears, and avoids the tone. So the trauma is the shock, the cue that brings the shock (the reminder) is the tone, and the memory is manifested in the behavioural signature of fear. And this is an effective training because the animal remembers that for many months. A variant of this model is to provide the shock, either with a tone or without it, in a certain context, like in one type of cage, then you test the animal in another cage. So if the animal is tested in the same cage in which it is shocked, it associates the context with the trauma. This shows how in trauma some of the cues are not the direct sensory cues but more complex contextual cues. And there are diverse variants of fear conditioning. There is another useful model, which is called conditioned-taste aversion. This is the association of a taste with transient malaise (intestinal distress).You present the rat or mouse with a new taste, an hour later inject them with a chemical that mimics transient intestinal malaise, and after 2–3 hours the physiological effect goes away completely but they remember to associate the taste with the feeling and avoid it later. And this is a very robust experimental protocol, which unfortunately many of us become familiar with when we taste a novel food and feel sick afterwards. This is a prime example of convergent evolution in neural and behavioural mechanisms.

Abi-Rached: But to what extent can these models inform us about the complexity of human traumatic memories? How can we speak of the ethical and social implications of these memory erasers if rodent models and human models are radically different, if not on the molecular basis, at least on many other levels and dimensions?

Dudai: Let me reiterate again that the applicability of memory erasers was/is a bit exaggerated in the general press. We are far away from being able to do it in humans, even if we were willing to do so. But still, I utterly agree that the discussions should refer to the potential ethical implications once ‘memory erasers’ become a reality. The idea beyond the research is twofold. First, to understand how the memory machinery in the brain works; erasers in this context are either tools to probe the system, or the outcome of the findings of the research. Second, very importantly, to develop tools to alleviate the sufferings of PTSD patients on the one hand, and of demented patients, such as Alzheimer’s patients, on the other – whereby in the latter case the idea is to use chemicals that enhance the machinery targeted by the erasers, rather than jam it.

I am interested in the basic mechanisms and the questions, but not necessarily in the applications. I do take responsibility for what I do. I take potential applicability into account, and I even initiate discussions of it in the scientific community and in other circles. The goal of my lab is to understand how multiple systems of memory work, particularly, how single experiences are recorded and retained in the brain. And over two-thirds of my lab activity has to do with human episodic memory and not with pharmacological studies on memory erasers. We’re interested in properties and mechanisms of episodic memory in humans, we use stimuli which are real-life stimuli, such as movies, we even produced special narrative movies to be used in our research. Movies are excellent stimuli for this purpose because they can present episodes in a controlled and reproducible manner. So we do most of our research on real-life or real-life-like memory.

Abi-Rached: Could you tell us more about the implications of the possible ‘surgical targeting’ of the enzymatic pathway of LTM that may emerge from the findings of PKMzeta on the rationale of the pathogenesis of PTSD but also on therapeutic approaches vis-à-vis traumatic memories?

Dudai: I do not think in terms of wonder drugs. I am thinking in terms of a toolkit, which includes many tools, such as beta blockers, glutamate receptor blockers and new pharmacological agents. In due time there may be some that target the PKMzeta-regulated machinery. These will permit the care provider to consider which combination of agents should be used, ranging from behavioural therapy to medications, and probably a combination of both. The advantage will be that using the right chemicals, the success of the therapy might be higher and faster.

Abi-Rached: Could you tell us more about the role of epigenetics in memory formation and synaptic plasticity, more specifically post its manipulation like with the potential use of long-term targeted ‘memory erasers’? I am asking the question because most of the ethical and normative concerns assume that memory can be easily manipulated.

Dudai: Epigenetics refers to changes in gene expression caused by mechanisms other than changes in the underlying DNA sequence. In other words, it refers to the influences of the environment on the potential of the expression of the genetic material in our tissues, brain included. For example, if a pregnant woman suffers from PTSD, that woman has a reasonable chance that some aspects of her child’s development will be affected not only by genetics but also by the fact that the hormones of stress that are highly activated in her body will have an impact on the potential of the genetic expression of the body systems, including of course the brain, in her child later in life. Memory systems are affected by the internal and external milieu, and many factors such as stress, the amount of sugar in your blood and so on. In addition you affect a specific memory in two other ways. First, by the particular state of that memory ‘encoding’, say you are waiting to meet somebody and you start building a narrative around that meeting on the basis of what you already knew beforehand and your mood at that moment. The discrepancy with reality is therefore inherent in that memory formation. Second, you affect a specific memory by the particular state of the subject during the ‘retrieval’ of memory. What happens in the first few minutes or hours of memory retrieval depends on many contextual and environmental factors and will affect the repositioning of that item of memory in your memory ‘store’. That specific item in memory is unlikely to stay the same; the mere experience and context of retrieval is bound to change it forever, where ‘forever’ amounts to the time period until the next retrieval. It sounds almost absurd, but because of this malleability of memories upon retrieval, one can say that the most faithful memory is the one which is never used. We have rather good molecular, cellular and behavioural evidence that this modulation of memory items after their retrieval indeed takes place, and those of us who study methods to modify traumatic memories, exploit this post-retrieval time window. But we also know from many studies that the way your memory is retrieved is determined to a large degree by the context, cues and bodily states already encoded in that memory. Hence encoding is a sort of miniature ‘big bang’ that determines the fate of memory for a lifetime to come.

Abi-Rached: So after the traumatic memory has been ‘inscribed’, encoded and stored, the therapeutic window to intervene, modulate and manipulate that memory could indeed be large. In other words, even a therapeutically manipulated memory (such as with ‘memory erasers’) is still plastic and apt to be modulated.

Dudai: Yes. Take again the example of a PTSD patient—and here I am not talking from a clinical perspective because I am not a clinician. Theoretically you can still intervene and change the memory of the event. In the first few minutes or hours of the traumatic event the ability to intervene is large. But the ability to change and modulate memory decreases as time passes. So time plays a big role. A window of opportunity may open again after retrieval, but its effectiveness compared to the immediate post-event time window is yet unclear and is under investigation.

Abi-Rached: In that case, how relevant and practically feasible is the idea of using such potential memory pills say in the rapid rehabilitation and integration of convicts, particularly children and adolescents, based on the justification that a ‘plastic’ brain can rapidly adapt post-manipulation [Illes, Reference Illes2007]?

Dudai: I think that practically it is currently far from being relevant. But again, this should not prevent us from discussing the social and ethical coordinates once, in due time, medications become available that might provide benefits in pathological situations at the expense of memory modification. In general, and importantly, this also calls for great caution in determining what ‘pathological’ and ‘normal’ is. The slippery slope to totalitarian society, or at least to a homogeneous boring one, is always around the corner. I basically think that when behaviour counts, there are no magic pills, and that only a spectrum of methods, ranging from behavioural to clinical, might help.

Abi-Rached: Lastly, if you had to summarize the key questions, problems and issues in memory research today what would they be?

Dudai: First, to understand the code of the system, in other words, what is the ‘machine language’ of the brain, how parts of the brain talk to themselves and to each other, and how is specific information encoded in the internal language(s) of the brain. We do not yet know much about the meaning of information we intercept eavesdropping on the conversation of brain cells and circuits with each other. We do know quite a lot about the molecular and cellular building blocks that permit such internal hidden conversations to take place. If you make the analogy to language, it will look like this: we know quite a lot about the letters, we know a little bit about the punctuation, we start to understand some words, but we do not know how to read well or understand sentences, let alone paragraphs. Basically we lack the Champollion or Rosetta Stone. The minute we find it, we’ll be able to start understanding how information is encoded. We probably need many Rosetta stones, one will not suffice. There are beginnings in simple systems or in parts of complex systems but the major problem in brain research is the same all over, namely how to read information in the brain. And because that information is complex we need teams of people to work on it, individuals from different disciplines and backgrounds, including mathematicians, computer scientists, biologists and so on.

Second, we need to understand how memory changes are stored over time and what is actually stored. Of course ‘stored’ is used as a metaphor [Dudai, Reference Dudai2002]. Descartes, for instance, used the hydraulic system as a metaphor, Plato used etched wax tablets as a metaphor, we use the computer metaphor. But is the computer metaphor any better than the previous metaphors? I doubt it is. It is only fancier though it might still be rather primitive.

Third, we need to solve the mystery of how it is that I can retrieve very complex information within only a fraction of a second and how I can tell you that information is available though I cannot retrieve it. This is related to phenomena like the so-called tip-of-the-tongue, feeling-of-knowing or intuition. We don’t know much about how these occur and what their underlying mechanisms are. There are psychoanalytic explanations for some of these phenomena but these are not mapped onto mechanistic explanations.

For me, comprehending things deeply makes sense only if we can replicate our findings. Giambattista Vico, the seventeenth to eighteenth-century philosopher who is said to have never left Naples but still knew a great deal about the world, formulated a maxim that was used in early scholastic philosophy, verum ipsum factum. This can be translated as the true is the made. And this has immense implications for the researcher. I can understand only if I can imitate or replicate. Some researchers claim that we understand language, for example, because we can talk. Vico took this maxim from theology and moved it to secular philosophy. In theology it was used as an argument that only God can understand the world because He made it. Similarly we can argue that we can understand the mechanism of memory only if we can build a biological memory system. That is my own philosophical take on it … and I believe in it.