CREB Giveth & Genetic Manipulation Taketh Away: The Story of Auditory Fear Memory in the Amygala

The idea that CREB levels can designate which neurons are recruited to the auditory fear memory trace is introduced in the 2009 Han et. al. paper. However, CREB plays a vast variety of roles within the cell, so the reason for this is unclear. It is the 2014 Yiu et. a. paper that finally gives an indication as to which role of CREB is important to this particular function. As it turns out, neuron excitability is the key.

After reading the first paper, I will admit that I had some burning questions left. For example- what role was CREB playing in this process? (thank you Yiu et. al) What happens if you knock out or knock down CREB function? Would it make it impossible for an auditory fear memory to be encoded at all? Is CREB necessary for this memory encoding process? Could this even be answered without making the affected neurons completely dysfunctional? Furthermore, could you allow the memory to be encoded, and temporarily erase it by simply inhibiting the CREB-cre cells instead of killing them all together? Could the memory be rescued by re-activating the same CREB-cre cells later on? After reading the second paper, I still wondered if there are other proteins in the cell that could play the same role in increasing neuronal excitability. Is CREB still necessary or simply some factor to increase excitability? Is CREB directly involved with the voltage dependent potassium channels examined in this paper, or are they independent mechanisms of excitability?

Something I have noticed in both of these papers is that the only measure of fear behavior is freezing. While freezing has been traditionally measured, it does not appear to be the only fear behavior displayed by some animals. An escape-like darting behavior has also been observed which would make it seem as though those animals were not freezing at all. Although the data in these papers seemed convincing, this is an important consideration to make when freezing is the only measurement used for fear behavior.

My final thoughts on these papers concerns the human disorder of post-traumatic stress disorder (PTSD). Although these methods are obviously not meant as a clinical treatment for humans, it may be interesting to consider what factors would create such an extremely strong fear memory to be encoded. The papers note that the size of the memory trace does not differ in size with the strength of the memory (ie. less neurons in the trace do not cause a weaker memory)- so what does cause a much stronger memory encoding? Is it possible that people who are susceptible to PTSD have a naturally higher level of CREB? Alternatively, are their amygdala neurons naturally more excitable? These are interesting questions to think about as field moves towards answering more questions about the human correlates of the brain regions and processes observed in mice.

10/17 Han and Yiu

            Han et al. deleted neurons activated during auditory fear memory to attempt to erase the memory. They knew that lateral amygdala (LA) neurons with relatively high CREB levels were preferentially activated during auditory fear memory. They labeled these neurons using HSV vectors in iDTR transgenic mice to elevate CREB levels in small populations of LA neurons. When the CREB-cre vector is injected into the mouse, the stop cassette loxP is excised, and DTR is expressed. When DT is administered, apoptosis occurs. The experiments run by Han et al. show that the cell death of the memory trace erases the fear memory.

            The reserachers proved that the deletion was chronic by testing mice up to twelve days after DT administration. On the twelfth day, the freezing levels were still as low as the second day. Furthermore, these mice were able to relearn the fear memory, which showed that the deletion was not an impairment of overall LA function. However, this made me wonder how many times or how often memory trace neurons could be ablated from the LA without any negative effects for the mouse. Since only a small amount of neurons were recruited and deleted, there were not any negative side effects seen in the mice after one time, but if it was repeatedly done, there may be negative effects. Further experiments could be done to investigate this by subjecting these mice to behavioral tests in the future.

            Yiu et al. sought to determine whether relatively higher LA neuronal excitability before training would preferentially allocate these neurons to a memory trace. They knew from previous research that increased CREB expression caused preferential recruitment on neurons to a fear memory, and ablation of these neurons caused the fear memory to be erased. They aimed to uncover a more detailed reason for the memory erasure. One result of increased CREB in a cell is increased excitability, so the researchers used an extensive amount of techniques to test neuronal excitability. The experiments went beyond using just neuronal excitability by overexpression of CREB; excitability through dnKCNQ2, CNO bound to hM3Dq, and ChR2 were also used. The results were consistent across methods.

            If fear memory deletion was deemed ethical for humans and the technique was safe, it would be difficult to use in a clinical setting. The experiments proved that the vector must be injected before the memory occurs, so it wouldn’t be able to be used to erase a fear memory from a patient’s past. Unless the person knew that they were going to encounter something fearful, then it wouldn’t be useful. If there was a way to inject the vector before fear memory occured, memory deletion could be useful in treating PTSD and other mental illnesses.

Ramirez: Now and Then

      One of the things that I found the most interesting between the 2013 and 2015 Ramirez et. al papers is that in the first paper, the "false memory" is created with negative stimuli, where as in the second paper, the dentate gyrus is activated with a positive stimuli. The fact that this subset of engram neurons can code for both positive and negative memories show how widespread their control of memory formation must be. I find it interesting that it seems that these neurons and their pathways are responsible for any kind of memory- not fear specifically. This is especially important to note because the two papers use very different kinds of tests in order to explore the behavioral outcomes of planting "false memories". The 2013 paper measures differences in freezing levels, where as the 2015 paper measured open field tests, elevated plus maze tests, tail suspension tests, and sucrose preference tests. Measuring freezing levels is generally regarded as a test of fear, but poor performance on the other tests from the 2015 paper are indicators of anxiety.

      Another piece of information I found interesting was that the 2015 paper examined the relationship between the dentate gyrus and the nucleus accumbens. They measured the effects of optically reactivating DG cells that were previously labeled by a positive experience in mice that were given a dopamine blockade or a glutamate blockade drug. This is similar to the previous papers which examined the effects of this blockade in the Nacc when the VTA dopamine neurons were manipulated. From that previous paper, we learned that dopamine receptor blockades in the Nacc can be very detrimental on the motor functions of a mouse. I am curious as to why the Ramirez group chose to pursue this route anyways, even when it had the potential to throw off results as a result of decreased motor functioning. 

10/3 Ramirez 2013 and 2015

Ramirez et al. (2013) optogenetically activated memory engrams in the dentate gyrus (DG) of the hippocampus in c-fos-tTA mice. Optic fibers were implanted into the DG and an AAV encoding TRE-ChR2-mCherry was injected into the DG or CA1.

The mice were taken off dox after surgery and exposed to context A for labeling. While the animals were being fear conditioned in context B, the labeled cells from context A were optically stimulated.  When the ChR2-mCherry mice were placed in context A for the second time, they froze significantly more than the mCherry only mice. In neutral and novel context C, the freezing levels were not significantly different than the controls. This showed that the previously neutral context A was successfully linked with the fear memory from being shocked, creating a false memory. A part of the method that I questioned was that the mice were taken off dox for 42 hours to label neurons associated with context A, and then immediately put back on the dox diet. Since the mice were only in context A for the final 10 minutes of the 42 hours, it seems like it would be possible that other neurons active during the rest of the time might be labeled as well. It might be interesting to compare labeled neurons from the context A exposure to labeled neurons while the mice stayed in the home cage for 42 hours.

Ramirez et al. used the same technique to label memory engram cells in the 2015 paper. One of the experiments in the paper started with the c-fos-tTA rats on dox during surgery and recovery, and then off dox to label either a positive or neutral memory and immediately put back on dox afterwards. Then, all groups except the control were subjected to ten-day immobilization stress. The labeled DG cells were reactivated for zero, one, or five days before the TST and SPT, which are used to measure anhedonia. Chronic (5 day) reactivation of DG positive memory engram cells was shown to rescue depression-like behavior after chronic immobilization stress. The results of the TST and SPT showed no difference between the anhedonic behaviors of the non-stressed controls and the stressed chronic positive memory reactivated group. Furthermore, both groups had similar levels of neurogenesis in the DG. The neutral memory and one day positive memory stimulation groups did not exhibit a rescue effect. Exposing the mice to a real positive memory for five days also did not have a rescue effect or increase neurogenesis. It would be interesting to see how long the rescue effects lasted in the mice. If the effects are long lasting, it could be beneficial to find a way to use the technique in treating depression in humans.

10/3 Ramirez

Both Ramirez et al papers for this week involved optogenetically reactivating dentate gyrus cells, and both discussed several innovative and fascinating experiments to show what manipulating the hippocampus can do to downstream targets and behavior. Ramirez et all (2015) discussed how activating positive memory engrams suppresses depression-like behavior, which I found especially interesting because the past weeks have shown different mechanisms that suppress depression. All of these experiments are of equal importance because the mechanisms of depression are so complex and so many people are affected by it. Thus, the more experiments devoted to studying depression, the better. Targeting specific brain areas and/or cells, like in the second paper, is a step above many current antidepressants that affect the entire brain, and can lead to reduced side effects and greater efficacy.

Ramirez et al (2013), who “created a false memory in the hippocampus,” got me thinking to whether or not the researchers had actually created a false memory. While I thought the experiment itself was remarkable, and the results can be very important for future research, I asked myself if it was truly a false memory? Although the mice did freeze in a completely different context than where they got shocked (chambers A and B were very different in design), I don’t know if “false” is the most accurate word because the mice remembered they got shocked, which was a real memory. It’s just the context that was different. An example of what a real “false” memory could be is if the mouse was never shocked to begin with, and the researchers somehow manipulated their brains to make them think they got shocked. So, it was definitely a manipulated and artificial memory, but I am not sure that I would exactly call it a false memory.

10/3 - Ramirez

Human memory is notably susceptible to error and influence from external and internal factors. The constructive, step-by-step nature of memory formation and retrieval leaves ample opportunity for such error to occur. Ramirez et al (2013) present a unique method for manipulating these perceived “weaknesses” in memory for modeling the maladaptive behavioral symptoms associated with mental illness. By showing that an artificial memory can produce a fearful response when no fearful stimulus is present, this research opens the door for further research into behavioral manipulation via implementation of artificial memory.

Ramirez et all expanded on this research in 2015, showing that activating a positive memory can acutely rescue depression-like behavior in rodents. The study also attempted to rescue anxiety-like behavior with activation of positive memory, but saw no change in measures of anxiety. This dichotomy highlights a problem that persists in both research and clinical treatment of psychiatric disorders: that depression and anxiety can be effectively referred to a treated with blanket methods, despite clear neuropathological differences in the disorders. This study also adds to a growing list of potential acute treatments for depression and other psychiatric disorders, filling a huge gap in clinical treatment left by anti-depressants that take weeks to have an effect.
The study also found that repeated stimulation of neurons associated with a positive experience led to sustained reversal of stress-induced depression-like behavioral effects and promoted neurogenesis after 5 days. It would be interesting to see in further experiments whether these effects could be sustained for multiple weeks after stress has occurred. From a clinical standpoint, such treatment could be used in the window between onset of a depressive episode and onset of pharmacological effects of anti-depressants. For many people, this period of pharmacological treatment with absence of actual effects can be discouraging and can cause people to shy away from treatment in general. Acute treatments such as this could be incredibly meaningful for those suffering from acute psychiatric distress. 

Ramirez Papers - Joe

Ramirez papers 2013 and 2015

    The most compelling part of this paper, if not the interpretation of the data, was the novel genetic/behavioral approach that Ramirez et al. took to addressing their question: how can a subset of neurons that are active during a learning task be labelled and selectively activated? In other words, how can a memory be stored so that it’s accessible externally? To address this, they utilized the cfos-tTA transgenic mouse to be able to express a desired gene under the control of doxycycline. When this mouse is off dox (doxycycline is not in their diet), and a neuron is activated (the IEG, cfos is expressed), the tetracycline response element (TRE) is induced and the gene of interest is expressed (in this case ChR2). This provides temporal and spatial sensitivity to express ChR2 in only the neurons that are active during the off-dox period that have been infected with the AAV-ChR2. This population of cells that expresses ChR2 in the dentate gyrus of the hippocampus is considered the engram population; it is said that this population of cells holds the memory of what was learned during the off-dox period.

    Then, by tagging neurons while learning a context (cfos expression while off dox), the memory trace of the then-neutral context is presumably “saved” by expression of ChR2 — presumably because it cannot be clear whether synchronous optogenetic activation of this entire population should mimic the endogenously emanating retrieval of the memory. That is an apparent confound of this approach, and of in vivo optogenetics in general: does optogenetic stimulation of an ensemble of neurons simulate the activity that those neurons exhibit when stimulated endogenously?

    Whether that question is yes or no, they very interestingly found that optogenetically stimulating this ensemble while fear conditioning a mouse in a completely different context that they were naive to was enough to elicit a CS-US link between the first context (the labelled ensemble) and the fear. This fear was also context-dependent — that is, freezing behavior was not observed in a third context, to which they were still naive. That stood as a control experiment to validate their experimental group’s data. This is an important point to bring up; this genetic, behavioral, time-dependent, spatially-dependent approach demands many controls to account for all the potential misinterpretations that is inevitable with this approach. This becomes more evident in Ramirez et al 2015 when specific circuits are probed using the same tet-tag technique. However, a great control that could have very easily been overlooked was when the behavioral paradigm was flipped so that context A was replaced with context C, to show that there was nothing inherent to context A that was causing this change in behavior.

    Lastly, the entire experiment was repeated with conditioned place avoidance in place of contextual fear conditioning. This was an interesting way to show that this artificially induced memory association could be applied to other behavioral paradigms.

    In Ramirez et al 2015, a more clinically translational question was addressed: can recollection of a positive memory lead to amelioration of depressive symptoms? This question was tackled using the same general experimental design as the aforementioned one: tag neurons off dox during a pleasant experience and then express ChR2 in those neurons to later be able to activate that ensemble. Interestingly, the pleasant experience of choice was a male encounter with a female. Firstly, this implies that the entire experiment was carried out using male mice. And secondly, that seems like an arbitrary classification of a pleasant encounter; it was not clear what the basis of their decision was.

    Despite this, the retrieval tests used were measures of both anxiety- and depression-like symptoms, because the distinction can be variable. Furthermore, the control groups utilized were extensive, and tackled a lot of potential misinterpretations of the data. The 6 group all underwent the tail suspension test (TST) and sucrose preference test (SPT) as measures of anhedonia, open field test (OFT) and elevated plus maze test (EPMT) as measures of anxiety, and novelty-suppressed feeding (NSF) as a measure of motivation. In mice that had undergone stress, activation of the pleasant memory engram was enough to rescue the depressive symptoms exhibited in the TST and the SPT. This effect was acute, which was interesting, because it indicates that perhaps perseverant activation of a positive memory trace is necessary to create a strong positive motivational state following a traumatic experience (this was addressed later). There was no change in behavior in the anxiety-related tests, which is an indication of what sorts of pathological states a positive memory might be able to amend.     All of these mice underwent four retrieval tests after the chronic stress they received, and so it is also possible that some learning went on during that time, especially if the engram was activated during each test. I don’t think the order of the retrieval tests was varied, to account for this possibility.

    The most interesting part of this series of experiments was probing of the potentially important circuits that culminate in the manifestation of this behavior rescue. It was made clear that the behavioral rescue due to activation of the memory trace in the DG can be blocked by pharmacologically inhibiting NAc neurotransmission. And moreover, this effect can be blocked by optogenetically inhibiting the axon terminals of BLA projections to the NAc. This is important because it says that there is something important about the connectivity of the BLA and the NAc to exhibit a positive behavioral state following retrieval of a positive memory. It was never shown that DG->BLA connectivity directly plays a role in this behavior, but previous literature has shown that these three regions are crucial for limbic function.

    Lastly, they wanted to show that this acute effect could become sustained if the memory engram were activated multiple times after stress. This is a great example of the Hebbian synapse concept — that multiple innervations of one neuron by another increases the strength of that synapse, which is the basis of neural plasticity. A great follow up to this experiment would be to see if this sustained rescue of behavior would not diminish over days (or even weeks) after retrieval.

    This novel technique is a very clever way to approach questions of the memory engram; however, it is not clear whether activation of a population of neurons simultaneously mimics endogenous retrieval of a memory. Thus, it is hard to say what neurons these actually encode.