Moore et al. vs Kellendonk et al. - Joe

Moore et al vs Kellendonk et al

Seminar in BioPsych

Fall 2016

Professor Shansky

Schizophrenia is a notoriously difficult mental illness to model, particularly because it it epigenetic and polygenic. Moore et al sought to create a reliable model for this disease in the rodent by carefully administering methylazoxymethanol acetate (MAM) at different stages in embryonic development. It had previously been shown that MAM administration on E15 led to development of a considerably smaller head; however, that is not typical of schizophrenic patients. The goal is to create a model that exhibits the anatomical, functional, and behavioral deficits of typical schizophrenic patients, and so Moore et al thought to try administering MAM two days later, on E17, when most proliferation of neurons has peaked in the cortex, so development would not be affected as strongly.

Fortunately, MAM-E17 mice did not exhibit the microcephaly that MAM-E15 mice did; their heads were similar in size to control mice. Moreover, brain region morphologies were similar to controls, aside from a slight reduction in cortical size. Very interestingly, they noticed that overall neuronal count was not affected; however, a reduction in density of neurons in one part of the brain was coupled with an increased density of neurons in another — to maintain the overall neuronal count. They tested a bunch of other measures to compare these mice to controls: electrophysiological recordings which showed that resting potentials in MAM-E17 mice were more depolarized than controls, but, interestingly, those same neurons were not more easily excited; behavioral assays which showed that experimental mice did not show gross motor abnormality, but did show a deficit in prepulse inhibition; pharmacological experiments which showed that MAM-E17 mice were more sensitive to increased schizophrenic stereotypies after PCP administration; and lastly, another pharmacological experiment which manifested a dichotomous responsiveness to amphetamine in prepubertal versus adult mice.

I’ve thought about what it would take to create a reliable model for a disease, or even just to create a transgenic mouse that behaves “normally” and I would imagine that it would require a plethora of intricate experiments with potentially shaky interpretations, because what does it really mean for a mouse to have schizophrenia — the florid symptoms can never be verified. Thus, I’m not sure I’m content with this series of experiments — not because of unruly data representation, but because I’m not sure how well a mouse model can truly represent the animal correlate of a schizophrenic patient. (It’s also interesting to note that it isn’t known how MAM methylation is preferentially targeted in cortical neurons.) Maybe there’s a more modest approach that may be more telling.

That brings us to Kellendonk et al, which was published just a month after Moore et al, and both papers came out of Columbia University. These experimenters thought: D2 receptor dysfunction has been implicated in schizophrenia in humans — specifically in the striatum —, and this dysfunction has a behavioral effect on working memory, which is known to be carried out largely by the prefrontal cortex (PFC). Thus, they thought, does overexpression of D2Rs in the striatum lead to cognitive deficits, and then, downstream, how is the PFC affected?

They adopted a tet-tag strategy to overexpress D2Rs in the striatum of transgenic mice. These mice were kept off dox to maintain overexpression; however, some of the experiments showed that introducing dox to return D2R expression back to normal did not reduce behavioral symptomatology — in fact, they may have exacerbated it. The first few experiments served to show that this mouse does indeed show dopamine activity level changes that parallel those of schizophrenic patients, and the next few experiments showed that locomotion, sensorimotor gating, and generalized anxiety levels were not vastly different from control animals. Then, the big experiment showed that these mice indeed showed a deficit in working memory, which is what they were hoping. So their next question was: what’s going on downstream of the striatum to lead to this behavioral deficit? They showed that lesioning the PFC led to a similar decline in ability to complete the working memory task, and this paralleled the decline in D2R transgenic mice. So the PFC is probably somehow involved in this deficit, but how? Well, DA innervation to the PFC seemed normal, as evidenced by density of TH-positive axon terminals in the PFC of experimental vs control mice. But, the levels of dopamine compared to its metabolites showed that dopamine was not getting biotransformed normally. Furthermore, administration of a selective agonist of excitatory DA receptors (D1R and D5R) showed a much stronger cfos activation in the PFC, but when the transgene was turned off, this cfos activation was much weaker! So overexpression of D2R probably has something to do with this downstream effect.

This approach to finding a neurobiological correlate of schizophrenia was much more modest and, in my opinion, more successful, than that of Moore et al. Instead of administering a drug whose drug action is not entirely understood and attempting to run after all the possible relations to human schizophrenia that it may have, Kellendonk et al took a much more targeted and controlled approach, by looking at DA receptor modulation, particularly because many schizophrenia treatments are based on DA modulation. I don’t know very much (anything) about schizophrenia, but I agree much more with the approach of Kellendonk et al than with that of Moore et al.

10/24 Herry et al. and Courtin et al.

            Herry et al. discovered that different groups of neurons in the BA are involved in fear learning and extinction. They used electrophysiology to isolate these different groups of neurons. When recording these neurons during extinction, they discovered that once extinction neurons began to increase in activity, fear neurons began to decrease one extinction block afterwards. After the neuronal changes, the behavioral changes began to occur; freezing began to decrease. The changes were measured with the change point analysis algorithm. I had never heard of the change point analysis algorithm before, and I thought it was an interesting approach to see if the neuronal changes actually corresponded with the behavioral changes. After fear renewal, extinction neurons were active 7 days after extinction. It would be interesting to test these neurons after a longer period of time to see if the activation of extinction neurons was more chronic. Also, after fear renewal, the extinction resistant neurons were not shown. I was curious why the authors chose not to show the data. Some additional data I would like to see is the raster plot for the rest of the experiments in the data set. It was present in the first figure but not for the rest of the experiments in the paper.

            Courtin et al. used an auditory fear conditioning paradigm that was similar to Herry et al., but used different techniques to look at populations of neurons. They recorded interneurons and principal neurons in the dmPFC that were responsive to CS+. The researches silenced type 2 (PVINs) interneurons using a cre-recombinase AAV. They were shown to have a large role in fear behavior. Optical silencing of PVINs by ChR2 before fear conditioning transiently induced freezing. The researchers also discovered that principal neurons (PNs) disinhibited during CS+ presentations preferentially targeted the BLA. This is because theta phase reseting by PVINs synchronizes PNs after CS+ presentations, and dmPFC PNs preferentially target the BLA to drive fear responses.

Fear Expression / Fear Extinction - Joe

Herry et al vs Courtin et al

Seminar in Biopsych

Fall 2016

Professor Shansky

It is a longstanding finding that cued fear conditioning leads to a marked increase in fear expression in mice, as indicated by a strong increase in freezing behavior (along with coupled autonomic responses, such as increased corticosterone, increased heart rate, etc.). Furthermore, it’s known that continuous presentation of the conditioned stimulus without the unconditioned stimulus is enough to dissociate the association between the CS and the US. Lastly, it’s been shown that reintroducing a contextual cue (the initial context that a mouse was conditioned in) is enough to recover the fear response that was extinguished. This is a good model for PTSD in humans, the most popular treatment for it — exposure therapy —, and its lack of efficacy: most people that undergo exposure therapy experience spontaneous recovery. What circuits mediate this behavior and what about the nature of the circuit leads to this strange series of behavioral changes? In other words, how do fear memories get stored and accessed, and when we try to forget them, why are they not efficiently forgotten?

Herry et al tried to unveil the inner workings of this fear circuit by taking a look at the basolateral amygdala (BLA), a brain structure previously implicated in fear memory consolidation and expression. By using an electrophysiological approach to record from these neurons, neurons with firing patterns correlated specifically with certain stimulus presentations were able to be dissected out. They segregated two groups of neurons that were responsive to the CS+ (that was originally paired with the US) either after fear conditioning or after fear extinction; these neurons will collectively be referred to as the fear ensemble and the extinction ensemble, respectively. Furthermore, they showed that the pattern of activity of either ensemble preceded the respective behavioral changes (freezing increase/decrease). Interestingly, they observed that fear renewal was coupled with a decrease in the activity of the extinction ensemble and a preferential recruitment of the fear ensemble, once again. This is in line with the hypothesis that this collection of “fear neurons” directly control the switch to a fearful state. An interesting question to ask is what the “extinction-resistant” population of neurons represents, and how they contribute to the modulation of this behavior, if at all. The BLA is a very heterogeneous brain region, so maybe it has nothing to do with fear at all. 

Okay, so there are two discrete populations of neurons that are more highly active during fearful and “safety” states, but where do these properties of these neurons come from? Are they anatomically different? Herry et al showed that the hippocampus preferentially innervates the fear neurons in the BLA, and those neurons go on to project to the mPFC, while there is a reciprocal connectivity between the BLA extinction neurons and the mPFC. This kind of makes sense, considering the hippocampus has been implicated in memory acquisition, and the mPFC has been implicated in decision making. Lastly, they showed that inactivation of the BLA does not change the current fear state of the mouse, but removes the ability of a change in the emotional state of the mouse.

When Cyril Herry finished his post-doc, he continued to try to probe this fear circuit to understand what is leading to this differential emotional state after behavioral (or neuronal) manipulations. It’s been shown that the mPFC has many different neuronal subtypes, one of which includes parvalbumin interneurons. PV interneurons are a subtype of interneurons characterized by their expression of parvalbumin; they are fast-spiking in nature, and exhibit perisomatic influence on principal neurons. Using a combination of electrophysiological recordings (multi-unit and LFP) and optogenetics (activation and inactivation), Courtin et al was able to show that PV interneurons are necessary and sufficient for fear expression in mice (I hear necessary and sufficient gets you Nature papers). To gain control of just the parvalbumin neuron population in the mPFC, they used PV-cre mice and injected cre-dependent AAVs for optogenetic manipulation. They showed, in several ways, that manipulating the activity of the PVIN population altered the activity pattern of the principal neurons (PN) in the mPFC to, in turn, affect fear expression. They showed that PVIN inactivation reset the theta rhythm in the mPFC, which has previously (and since) been shown to be important for fear expression), and preferentially activated the PN population, to lead to an increase in fear behavior. Great, so there is a modulation of activity in the mPFC that relates to fear expression, but that doesn’t tell us how it fits into the grand scheme of the fear circuit. So lastly, Courtin et al showed that those mPFC PN neurons that were activated during PVIN inactivation targeted the BLA, as shown by antidromic activation of efferents to the BLA with concurrent recording of the mPFC neurons.

These two papers provided deep and thorough insight into the mechanistic nature of the fear circuit; however, how the fear circuit works in its entirety is yet to be unveiled. Herry et al found two segregated populations of neurons in the BLA that showed dichotomous firing patterns paired with dichotomous behaviors, while Courtin et al showed that PV interneuron activity modulates the communication of the mPFC to the BLA. But, where do the PV interneurons in the mPFC receive their projections from, and which BLA neurons get innervated by mPFC projections? It’s in the works!

Both papers ended with just about the exact same closing paragraph: the translational goal of this work, which is to help understand what goes on in anxiety disorders where fear expression regulation goes awry.

Electrophysiology. is. Everywhere.

Electrophysiology can often be a mystifying area of research for anyone who is unfamiliar. I hope that the discussion of these two papers will be enlightening, as I myself still have gaps in my understanding of the methods. However confusing it may be, electrophysiology is a powerful tool for decoding specific circuitry within the brain. Both papers have used this technique to identify key populations of principle neurons and interneurons that play a role in fear conditioning and extinction. These methods heavily relied on mathematical analysis of ephys data to identify neurons as opposed to staining of cell types like many of the previous papers we have read. To me, this raises a question of which methodology is more reliable.

After reading the first paper by Herry et al (2008), I was left wishing that they had done a few more experiments. They determined that there were two populations of neurons controlling fear and extinction. Although they did a cFos study at one point in the paper to show which neurons were recently active, it did not appear that they did any sort of staining to identify what types of neurons these were. They could have stained for NeuN to identify principal neurons and somatostatin for GABAergic interneurons. These could have provided an added level of characterization of the neurons that they were examining if these markers were co-localized with the cFos. I found the second paper to be more comprehensive in its characterization of subgroups of neurons within the dmPFC-BA-hippocampus circuitry.

Another item that I felt was missing from this first paper was in terms of the behavior in the section "BA inactivation prevents behavioural transitions". They found that "inactivation of the BA before extinction training prevented the acquisition of extinction". I was wondering if there was any way for them to have shown that this effect is specific to fear learning. Perhaps they could have completed another behavioral test/ paradigm that would have shown that the animal's ability to learn at all was still intact after being exposed to muscimol.

10/24 fear papers

This week’s papers further discussed fear expression and expanded on the topic by performing experiments revolving around switching it on and off by neuronal circuits and how some PVIN interneurons shape activity to drive fear expression.

The Courtin et al. 2014 paper in particular left me with a few questions. First of all, they kind of lost me when they started discussing theta phase resetting and oscillations in Figure 4 and Figure 5. Maybe it’s just my lack of knowledge in that particular subject matter, but I feel like they could have introduced it and discussed it better since it appears to be a vital aspect of their experiments. If they just touched upon the basics then that would have made it a lot easier for readers like me to follow.

In that same paper, I was a bit confused on Figure 2d. They stimulated PVINs via ChR2, which would ultimately inhibit freezing since PVINs (such as Type 2 PVINs) are inhibited during freezing. The results show that there was a “significant” decrease in freezing in mice injected with ChR2 over the GFP control mice, which makes sense, but in those ChR2 mice, there was still a 60% freezing rate (in the Post FC graph). Not to mention, when you shine no light (thus you do not see effects of ChR2), it’s still about a 60% freezing rate. When you compare that number to Figure 3c, in which PVINs were stimulated with the inhibiting ArchT which would increase freezing, the numbers aren’t too far off. I would have liked to see, however, another graph in Figure 3c showing Post FC (like in Figure 3d) to have a better comparison.

Han et al vs Yiu et al - Joe

Han et al vs Yiu et al

Seminar in Biopsych

Fall 2016

Rebecca Shansky

Diphtheria toxin receptors (DTRs) are not endogenously expressed in mice, and neither is diphtheria toxin (DT). Thus, inducing expression of DTRs in the LA in a cre-dependent manner allows for specific ablation of neurons in the LA through injection of DT anytime after ample infection of the virus. This is what Han et al did. Transgenic mice (iDTR mice) were injected stereotaxically with a cre-dependent virus. In the population of cells that the virus infected, cre was expressed, which means that DTR was also expressed. Thus, (intraperitoneal?) injection of DT anytime after the cre had had enough time to be expressed would induce apoptosis in the susceptible (DTR-positive) population. Up to this point, all the cells in the LA injection site expressed the DTR; however, if the cre virus also encoded cDNA for CREB, only the LA neurons that expressed high levels of CREB would also be infected with cre, thus inducing DTRs (introducing apoptosis susceptibility). This is interesting because they mentioned that the CREB-cre group and the control-cre group showed similar levels of cell death — if the CREB-cre group is supposed to infect a narrower subset of neurons and the same volume of virus is injected into both groups, one would imagine that DT injection would kill a smaller population of cells in the CREB-cre group.

It was convincing that there were several groups of controls in this paper. Firstly, there was the experimental group (CREB-cre with DT in iDTR mice), and the subsequent iDTR mouse controls (control-cre with DT and CREB-cre with vehicle). Another control could have been control-cre with vehicle, but if there was no experimental effect in the control-cre with DT group, the control-cre with vehicle probably would not have been very informative. Furthermore, there were wild type controls as a proof of principle that the CREB-cre virus was infecting cells that were preferentially active during fear conditioning (the cells that were encoding the memory). With all the controls, only the experimental group showed a significant change before and after DT administration. However, this entire paper is based on cell-type specific apoptosis and concurrent freezing behavior change, and no control was done to account for crude locomotor changes. The LA is just medial to the cortex, and CREB is expressed throughout the entire brain, and so ablation of cells expressing high levels of CREB can precipitate a number of confounds outside of the LA. This is without considering that probably not every cell in the LA that expresses high levels of CREB is recruited in a particular memory trace. Still, it controls for impairments in learning after ablation and impairments in other memories prior to ablation, which are very convincing. Still, this change in behavior could be due to a change in locomotor activity around ablation, which is accounted for by the brain days later.

An interesting question to ask is if the title of the paper suits the findings. The most convincing experiments in this paper were the controls for inabilities to learn after ablation and impairments to learn prior to ablation. This showed that if ablation is targeting a memory trace, it is not affecting other memories learned prior to the auditory fear learning, and it also did not actually damage the memory-making system in the brain for future use. Still, it is unclear whether the neuronal population annihilated specifically encoded that memory.

Yiu et al attempted to follow up the findings of Han et al and others to conclusively say how neurons in the LA are recruited to be a part of a particular memory trace. To test the claim that LA neurons are recruited based on intrinsic properties (their current excitability), Yiu et al employed similar methods to those that Han et al did; however, instead of completely eliminating a particular subset of neurons in the LA, Yiu et al modulated their neuronal excitability through various viral techniques. The best part about these first few experiments was that before they began to show how excitability ties into the recruitment of neurons for the memory trace, they did a series of proof of concept experiments showing that their methods indeed increase/decrease excitability reliably — good controls.

What was interesting about this paper was that the same claim was proven three times, using three different approaches; however, all three approaches told the same story: dnKCNQ infection increases neuronal excitability which preferentially recruits infected neurons into the memory trace; excitatory DREADD infection coupled with CNO administration increases neuronal excitability which preferentially recruits infected neurons into the memory trace; and lastly ChR2 infection with light administration increases neuronal excitability which preferentially recruits infected neurons into the memory trace. The only change between the experiments was the increase in temporal sensitivity of the experiments. Were they all necessary to drive that point home? I’m not sure. Nonetheless, the finding is awesome! Simply activating a subset of neurons before a learning task “primes them” and they subsequently are more likely to be a part of the memory trace, as measured by neuronal activity during a retrieval task (coupled with an in situ measuring arc translation). The experiments were well-controlled; they accounted for the active timeframe of HSV by doing behavior several days after injection of the virus, and also accounted for overall anxiety behavior by showing that increasing excitability in the LA doesn’t just alter locomotor behavior — it’s context specific, among several other crucial controls. Overall, it seems convincing that increasing neuronal excitability increases probability that the given neuron will become recruited into the memory trace. And it’s also crazy to get that insight into the underlying mechanisms of neuronal recruitment during learning.

Han et al had a question regarding what happens if the neurons that are recruited into a memory trace are eliminated from the mouse’s memory repertoire after fear conditioning; can they still recall the memory? Yiu et al took it a step back and said, “well what is it about these neurons that are recruited that allows them to be recruited; it could be any collection of neurons that becomes a memory trace, but if you induce excitability, can you artificially recruit those neurons in a memory trace?” Apparently, you can.

10/17: Fear Memories

This week’s papers were both related to the effects of neuronal excitability on fear memory expression. Han et al., 2009 showed that increased CREB levels during fear learning are critical to the stability of that memory by deleting neurons showing overexpressed CREB levels, which blocked expression of fear memories by decreasing transgenic mice’s freezing levels. I thought the researchers did a very impressive job in finding the neurons in the LA that are actually involved in the fear learning process because instead of being in a small cluster they are widely spread out. There could have been more experiments in disrupting CREB, however, to further show that it is indeed involved.

Yiu et al., 2014 used the knowledge from the first paper to show that increasing neuronal excitability biases recruitment into the memory trace, enhancing memory formation. I thought this experiment was more comprehensive than the first one because they not only showed the involvement of CREB, but also numerous other molecules such as dnKCNQ2, hM3Dq + CNO, and ChR2 in neuronal excitability and memory formation enhancement. Furthermore, I like how they used Kir2.1 to show opposite effects in mice. They also used a wide-ranging set of control experiments, including the anxiety test and the anatomical specificity test.

From a clinical aspect, these experiments can be very valuable for the treatment of fear disorders such as PTSD. Clearly, these experiments are not exactly feasible in humans, but knowing that you can manipulate neurons in your lateral amygdala during fear memory expression is very important. These papers show that you need to excite neurons prior to training the fear memory, but perhaps researchers can next dive into ways you can manipulate fear memories after it has been encoded.