Canonical circuits in neuroscience

Gary Marcus, Adam Marblestone, and Thomas Dean have a nice perspective piece in Science this week on the atoms of neural computation (gated):

One hypothesis is that cortical neurons form a single, massively repeated “canonical” circuit, characterized as a kind of a “nonlinear spatiotemporal filter with adaptive properties”. In this classic view, it was “assumed that these…properties are identical for all neocortical areas.” Nearly four decades later, there is still no consensus about whether such a canonical circuit exists, either in terms of its anatomical basis or its function. Likewise, there is little evidence that such uniform architectures can capture the diversity of cortical function in simple mammals, let alone characteristically human processes such as language and abstract thinking…

What would it mean for the cortex to be diverse rather than uniform? One possibility is that neuroscience’s quarry should be not a single canonical circuit, but a broad array of reusable computational primitives—elementary units of processing akin to sets of basic instructions in a microprocessor—perhaps wired together in parallel, as in the reconfigurable integrated circuit type known as the field-programmable gate array.

Candidate computational primitives might include circuits for shifting the focus of attention, for encoding and manipulating sequences, and for normalizing the ratio between the activity of an individual neuron and a set of neurons. These might also include circuits for switching or gating information flow between different parts of cortex, and for working memory storage, decision-making, storage and transformation of information via population coding and the manipulation and encoding of variables, alongside machinery for hierarchical pattern recognition…

And so on. People have long proposed that ‘all cortex is the same’ or some such rubbish, it being all made of cortical columns that are the same from one bit of tissue to another. I’m not sure how many people really believe that, but you see the statement a lot (and it is a big reason people studying visual cortex claim they’re just interested in ‘cortex’).

Of course, on a smaller scale there has been a long interest in more primitive ‘microcircuits’. A few examples:

The Reichardt (motion) detector:

Reichardt detector

There is strong experimental evidence for this in drosophila (the fly), and is a repeated motif across visual space.

(see source for review)

Inhibitory circuit motifs:


(see source)

Circuit motif for flexible categorization:

flexible categorization motif

(see source)

Since we know all the circuitry of the worm C. elegans, you can look at which motifs are overrepresented:

c elegans 4 neuron circuit motifs

(see source)

Qian, J., Hintze, A., & Adami, C. (2011). Colored Motifs Reveal Computational Building Blocks in the C. elegans Brain PLoS ONE, 6 (3) DOI: 10.1371/journal.pone.0017013

Borst, A. (2007). Correlation versus gradient type motion detectors: the pros and cons Philosophical Transactions of the Royal Society B: Biological Sciences, 362 (1479), 369-374 DOI: 10.1098/rstb.2006.1964

Pfeffer, C. (2014). Inhibitory Neurons: Vip Cells Hit the Brake on Inhibition Current Biology, 24 (1) DOI: 10.1016/j.cub.2013.11.001

Mysore, S., & Knudsen, E. (2012). Reciprocal Inhibition of Inhibition: A Circuit Motif for Flexible Categorization in Stimulus Selection Neuron, 73 (1), 193-205 DOI: 10.1016/j.neuron.2011.10.037

Marcus, G., Marblestone, A., & Dean, T. (2014). The atoms of neural computation Science, 346 (6209), 551-552 DOI: 10.1126/science.1261661


A mechanics of depression

There are many reactions that can be taken in response to the world going crazy on you, and depression is one of these.  Even though it is (rightly) seen as perhaps not the greatest illness to have, there is a case to be made that depression is an energetically-efficient response to overwhelming stress; it can be better to shrink back and conserve your energy than fight it.  Think about it like this: you probably know some people who are super laid back, who take things as they come and don’t seem to stress out.  And you also probably know some people who freak out at stress, work really hard, and just seem to be stressed out all the time.  These are two different strategies for dealing with stress and one seems more likely to lead into depression.  At the same time that same strategy seems like the person is fighting harder to get out of the stressful situation.  How does the brain do something like that?

It is thanks to tools from the lab of Karl Deisseroth that we are finally able to begin to really, mechanistically, understand what is going on in the brain.  And fortunately, Deisseroth is both a research scientist and a psychiatrist who is interested in helping people with mental diseases.  There are three (!) papers published in Nature over the last month with his name on them, and they shed a lot of light on the mechanisms that are at work.

No one is really sure what it is about the brain that causes depression, although we have some hints: antidepressant drugs tend to work by modifying the release of the neuromodulators serotonin and dopamine (and norepinephrin).  We also know that an area called the prefrontal cortex (PFC) is highly linked to all sorts of psychiatric disorders; the PFC is an area that receives inputs from all over the place and then sends outputs right back out.  He’s the boss, the one that hears everything people have to say and then directs other areas in order to coordinate the brain to accomplish internal goals.  You can imagine what might happen if you have a bad boss: your brain is out of sync, things don’t get coordinated properly and then BAM, schizophrenia and depression.

As you might imagine, these things are all interconnected in the brain: the PFC talks to the serotonin and dopamine areas, and the serotonin and dopamine areas talk to the PFC.  And these connections are particularly important.  Take the connection between the PFC and an area that releases serotonin, the dorsal raphe nucleus (DRN).  This connection is required to motivate an animal to avoid escapable stress.  A mouse that is in a position to escape from stress will clearly do so.  However, if you inactivate the PFC the mouse will not escape from stress and its release of serotonin will look the same as if it were in a stressful situation it can’t escape from.  If you disable the PFC in a stressful situation that it can’t escape from?  No change: the PFC seems to control motivated behavior from escapable situations only, and without it you can’t.

That’s exactly what one of the recent Deisseroth papers examined.  They were able to directly activate only the PFC neurons that send information to the DRN and by doing so they found a way to escape from a learned kind of helplessness.  Rats that are stuck in a cup of water will struggle for a while, attempting to escape.  After a while they learn that struggling isn’t getting them anywhere and they just kind of give up.  But if you activate the PFC connections to the DRN?  The rats launch back into the struggle again!  But this doesn’t happen if you just activate all the neurons of the PFC or all the neurons of the DRN: there is a specific pathway through both of these brain regions that motivates an escape from helplessness.

Release of dopamine can help motivate escape from a helpless condition as well, although it is released from a different part of the brain.  The ventral tegmental area (VTA) is one of the main release sites of dopamine in the brain and is the signal of ‘pleasure’ in the brain, although it is perhaps more accurate to say that it is the primary signal of motivation.  And if you stimulate the neurons in the VTA you get an increase in motivation to escape a depressing circumstance, just as you’d expect from that area.  And specifically, this is because of a release of dopamine from the VTA to another area of the brain, the nucleus accumbens (NAcc).  What is likely to be happening is that the VTA is sending a motivating signal to an action and learning center of the brain (the NAcc), and that center of the brain helps decide what to do next, and what to do next is to get the heck out of there.  Something exciting happens here: if you now go and record the neurons in the NAcc after additional dopamine is released from the VTA, they now respond to different things.  The whole way that an action is represented in the brain changes, and in a way that emphasizes escape.

But this ability to learn escape can have a negative side.  Take the example of another method of stressing out mice, chronic social defeat.  What you do here is force mice to get defeated in battle again and again.  Yes, this is actually a commonly studied behavior; these poor guys are basically given PTSD.  But it turns out that some mice are resilient to this stress, they can withstand it and not get depressed.  If you look at the neurons in the VTA, the susceptible animals show an increased amount of bursts of activity (technically: phasic firing) during stress while the resilient animals just hummed along with no change of activity in the VTA at all!  This natural increase in firing can be simulated in the resilient animals by artificially increasing VTA firing.  Then, when you test whether they have acquired PTSD?  Well, it turns out that they have.  This makes a certain kind of sense: dopamine reinforces behavior, so susceptible animals are seeing more dopamine and hence more reinforcing of defeat than are the resilient animals.  Again, though, different internal pathways have different effects in the brain: if you activate the neurons that send information to the ‘pleasure center’, the nucleus accumbens then you are more vulnerable to stress.  And if you inhibit the activity of neurons that send information to the PFC, then you also become more vulnerable to stress.

The same sets of neurons that can help you escape stress (the VTA to NAcc connection) are the ones that will cause you to be more depressed in the future.  This suggests that there might be a tradeoff in life: you can be stressed out but really motivated to escape stress, or you can put up with a whole bunch of stress and be laid back about it in the future.  But it seems like it might be hard to be both.  There are of course in-betweens: the PFC, the boss, has specific circuits dedicated to telling the VTA and the DRN what to do, and can tell them to do opposite things.  And of course, the more you try to escape stress and fail, the more you learn it is futile to escape stress, triggering a terrible feedback cycle.  But if you want to be learning, you better be trying.


Chaudhury, D., Walsh, J., Friedman, A., Juarez, B., Ku, S., Koo, J., Ferguson, D., Tsai, H., Pomeranz, L., Christoffel, D., Nectow, A., Ekstrand, M., Domingos, A., Mazei-Robison, M., Mouzon, E., Lobo, M., Neve, R., Friedman, J., Russo, S., Deisseroth, K., Nestler, E., & Han, M. (2012). Rapid regulation of depression-related behaviours by control of midbrain dopamine neurons Nature, 493 (7433), 532-536 DOI: 10.1038/nature11713
Tye, K., Mirzabekov, J., Warden, M., Ferenczi, E., Tsai, H., Finkelstein, J., Kim, S., Adhikari, A., Thompson, K., Andalman, A., Gunaydin, L., Witten, I., & Deisseroth, K. (2012). Dopamine neurons modulate neural encoding and expression of depression-related behaviour Nature, 493 (7433), 537-541 DOI: 10.1038/nature11740
Warden, M., Selimbeyoglu, A., Mirzabekov, J., Lo, M., Thompson, K., Kim, S., Adhikari, A., Tye, K., Frank, L., & Deisseroth, K. (2012). A prefrontal cortex–brainstem neuronal projection that controls response to behavioural challenge Nature DOI: 10.1038/nature11617
Amat, J., Baratta, M., Paul, E., Bland, S., Watkins, L., & Maier, S. (2005). Medial prefrontal cortex determines how stressor controllability affects behavior and dorsal raphe nucleus Nature Neuroscience, 8 (3), 365-371 DOI: 10.1038/nn1399

Photo from