Wednesday, October 30, 2013

How we learn (and forget)



In the last few years, there has been a remarkable advance in the understanding of how we learn as infants.   The mechanisms of learning may also shed light on the nature of neurodegenerative disorders.    Paradoxically, the process of learning might be likened to a starting point of a very weak knowledge of a quasi-infinite number of undistinguished things where over time, the amount of knowledge and the connections gets pared back into a finite number of strong connections.   This is evident in a process called neural paring.

Let me speak to the process of neural paring in learning and also in neurodegenerative disorders. Neural paring is also known by the term synaptic paring.   Both terms mean a process where a relatively large number of neurons are present at birth, with a relatively small number of interconnections projecting from each.   The process of learning appears to be a process where the large initial neuron population gets pared back, and the remaining live neurons are strengthened, and the number of interconnections from these survivors increases.

The process of neural paring has been seen in a number of regions in the nervous system, but I’ll mention two in particular where the process has been observed:  the hippocampus and the neuromuscular system.

The hippocampus is in an older part of the brain called the limbic system – very much in the center.   It  is deeply involved in something called declarative memory, and also cognitive maps.   Declarative memory is the kind of memory that can be consciously recalled, like facts and individuals.   An example of a non-declarative memory might be the motor memory associated with a skill like bike riding.  

A second function of the hippocampus is to store cognitive maps, or a mental map of an individual’s surroundings.   Typically one doesn’t see creation of cognitive maps until an individual has undergone some maturity after birth.  

In a behavioral sense, we can see a distinction between young children and Alzheimer’s Disease victims on the one hand, and healthy mature adults on the other hand.   Much data have been collected on the behavior of lost persons.   Typically, both young children and Alzheimer’s Disease (AD) victims have poorly formed cognitive maps of their surroundings.   When they get lost, the typically are found very close to their last known position, not being able or wanting to wander very far.   Healthy mature adults, on the other hand possess a strong cognitive map, and the strength becomes something of a weakness.   The mature adults will try to find their way out of a situation where they’re lost and end up traveling much farther from their last known position than either young children or AD victims.   This has been born out in a large amount of data accumulated by search and rescue missions.

 Another part of the brain that has been studied in neural pairing is the neuromuscular interface.   Here nerve cells terminate on muscle cells.   When nerve cell fires next to a muscle cell, it releases a chemical across a junction called a synapse.   When the muscle cell senses the chemical, a neurotransmitter, it then reacts by contracting.   There are ways to foil this system.  One example is botox, or botulism toxin, which blocks the transmission of the neurotransmitter signals to the muscles.  

In both the hippocampus and the neuromuscular system, paring appears to occur as part of the maturation process.   One can broadly think of two kinds of interconnections between nerve cells or nerve in muscle cells.  There is one extreme case where there are a lot of neurons but each one has a relatively small number of connections.  In the other extreme case, there are a far smaller number of neurons running the show, but they are larger have a lot of connections projecting to other nerve or muscles cells.

The process of neural “growth” or “learning” appears to be associated with a phase after birth where there are at first lot of small neurons and many connections collectively, but a small number of connections from any individual neuron. As the individual matures, many nerve cells disappear, but the remaining ones are made healthier – larger and create many connections.

In a lot of ways, it’s like a growing forest – one starts out with a lot of weak seedlings, with a limited number of root and limb branches.   As the forest matures, most of the seedlings die, and the remaining ones are empowered to grow more roots and limbs

The process of this paring away of neurons and strengthening the remaining ones requires activation of the neural pathways for the pruning and strengthening process to take place.

You can look at the uncoordinated movements of an infant as a behavioral aspect of this process, where they slowly gain motor coordination over time.  Typically there is a factor of 10 die-off in nerve cells in a newborn, but the overall amount of nerve tissue remains the same or is even slightly increased over time as the surviving nerve cells grow more branches and larger trunks.

Recently, evidence has been found on the mechanism of how nerve cells are pared back during maturation.   It is a variation on the immune system response.   In the immune system outside the brain, there are two subsystems – the innate immune system and the adaptive immune system.   If there is an invading organism, say a bacterium, anti-bodies develop that recognize the foreign proteins on its cell membrane.  The anti-bodies themselves are part of the adaptive immune system, as different antibodies develop in response to different invaders.

The antibodies bond to the membrane of the invader – this triggers the innate part of the immune system, which recognizes the only “back-side” of the antibody, which is not specific to a particular foreign invader – it’s a kind of universal.   The innate system has a cascade of chemical reactions called the complement system where the presence of the antibody on the cell membrane triggers the body to attack the membrane of the invader.   This cascade of interactions eventually leads to a protein complex that creates a hole in the cell membrane of the invader.   Once that hole, or multiple holes open up, the invader dies as pressures across the hole cause the cell to self-destruct.

The recent insight into neuron-paring is the existence of the same immune complement system inside the brain in mammals as they undergo maturation and “learning”.   The chemicals associated with the complement system in immune response are found in large quantities in mammals just after birth when the neural paring is the largest.  

The precise mechanism is unknown as to how a nerve cell gets tagged for destruction, but there is a “punishment” theory that suggests that one nerve cell wins out over another by firing more often than the other.   Then, a kind of immune response cell in the nervous system called a microglial cell tags the nerve cell destined for destruction, which then initiates the complement system cascade that ensures the death and clearance of the tagged cell. 

It appears that a similar neural paring occurs in patients who suffer from Alzheimers and other neurodegenerative diseases.   Rather than a learning process, this is more of an autoimmune response, where cells are pared, but there doesn’t seem to be any maturation and growth of remaining cells.   In Alzheimer’s disease, nerve cells appear to be tagged by the presence of amyloid placques, which initiate the complement cascade system, leading to the cell’s destruction.   As such, it is not part of the entire process of neural paring, but only the destruction part.

Refs:

A. Stephan, B. Barres, B. Stevens, The Complement System: An Unexpected Role in Synaptic Pruning During Development and Disease, Ann. Rev. Neuroscience, 2012. 35: 369-89.

J. Tapia et al. Pervasive synaptic branch removal in the mammalian neuromuscular system at birth, Neuron, 2012.  74.5: 816-129

B. Ojo et al.  Age-Induced Loss of Mossy Fibre Synapses on CA3 Thorns in the CA3 Stratum Ludicum, Neuroscience Journal (2013) ID 839535. 


6 comments:

  1. Hi, John:

    Apologies for taking your fascinating article as a launch point for a polemic, but I guess I can’t resist. So here goes…

    Neuroscience of higher order mental function (memory, learning, mental mapping, etc.) has indeed made remarkable progress, perhaps several orders of magnitude, in the past couple decades. However, we have to remember that that may have taken it from 10e-30 to 10e-27 in terms of truly understanding mind by studying the physical brain. In other words, we still know essentially nothing about it.

    Theories are full of huge gaps and are regularly questioned, invalidated and revised, replaced or abandoned (often silently); negative results are left in the “file drawer”; crucial technological and mathematical tools in neuroscience are frequently found to have significant flaws that invalidate rafts of published studies; strong claims are made on the basis of weak studies with minuscule n’s; complexities, exceptions, anomalies and new unknowns arise faster than answers; weasel terms such as "suggests", "involves", "is implicated" abound; metaphors that would be laughed out of hard-core biology and physiology, not to mention physics and chemistry are flung around with abandon and acquire the patina of actual science; myriads of conflicting, contradictory and complicating data are winked away in linking brain regions with mental functions; mathematical constructs are projected as entrancing pictures of brains lighting up with activity, when in fact no such picture exists anywhere except in the fevered computations of a computer; scientists mix hardcore neuroscience data with high-level mental functions to imply that they are linked when in fact that is an unsupportable leap. And the popular press -- including articles by many scientists -- merrily ignores all these cautions, limitations, complexities and qualifications and takes a thimble of data as an ocean of progress to declare that we are within sight of fully grasping the mind as a physiological phenomenon and thereby somehow understanding "what it means to be human".

    In fact, I'd venture that science is no more in a position to posit a path to understanding mind and mental function by the neuroscience route than a day hiker trekking toward Mt. Everest with the moon hanging over a ridge (http://bit.ly/17xWGRH) can envision a practical way to the summit, much less to the moon. In other words, there still may be insuperable obstacles in the way, mostly unknown today, and there may not even be such a path. In fact, on the basis of several mathematical, scientific and philosophical principles as well as tracking the neuroscience literature at several levels, I personally think there is no such path, and that we are only hauling gear to base camp with the Khumbu icefall still ahead. I may not live long enough to see that proven and widely accepted, but I bet that within 10-15 years the current flood of optimism (to switch metaphors) about the neuroscience of high-level mentation will have receded, leaving behind isolated swirls of actual scientific results among the rocks.

    ReplyDelete
    Replies
    1. Hi, David -

      I'm not sure if you're directing this at the above post or it's a general comment.

      On the post itself, it is fairly well established that the complement cascade system is correlated with neural paring. You might want to read the review article that I posted at the bottom. This is a great summary of what is known about the cascade system and microglia. It is also well known that the amyloid-beta peptide binding to C1q activates the classical arm of the complement cascade system. So on both counts the above is on fairly solid ground.

      I mention both the hippocampal paring and the neuromuscular paring because these seem to be the best studied examples. I'd recommend reading Lichtman's article if you have any doubts about the paring - it's a remarkable study using a 3 dimensional reconstruction of axonal branching and synaptic connectivity in the nerve-muscle interface in mice.

      In terms of claims made, I tried to limit myself to a summary of the three papers. In ad-libbing, I think the forest analogy is a reasonable one and makes it easier to grasp what is going on.

      In more general terms, there's a fair amount that is known about the roles of the hippocampus and entorhinal cortex in forming cognitive maps, although the interplay of the two is not understood as of yet. Certainly there are conjectures about it, but I think it's beyond question that these play a major role. How the spatial information relates to the role of the parietal lobe is an interesting question and, to my knowledge, hasn't been articulated, although I have my own personal conjectures.

      Finding correspondences with neural functions and their roles and even how individual neurons encode spatial information is a pretty advanced topic, so I think we actually have made some significant progress.



      Best,

      John

      Delete
    2. This comment has been removed by the author.

      Delete
    3. Part II

      Having said that, a *full* picture of the human brain is a pretty ambitious topic and I don't think that there is anyone who claims it's a cakewalk. You should check out Lichtman's TED talk at CalTech where he outlines the challenges of the problem. One can even do back-of-the-envelope calculations to capture the magnitude of the issues and put just the connectome piece of it into perspective.

      Even if we understand the connectome bits, which may be decades away, there is also the spatio-temporal organization, which is very challenging. We're at the point where we can track maybe 100 neurons in close proximity using electrode studies. Going beyond this requires a major advance in technology.

      Still, I think it's a worthwhile endeavor, given the crippling nature of neurodegenerative disorders alone.

      On the issue of science writing - there is a necessary element that writers have to simplify material for the lay reader. The above piece was an attempt to help a writer-colleague understand the material about neural paring and the role of the immune system. I sent him the references above, but he couldn't wade through them, so this was a Cliff-notes version.

      I think one needs to articulate the goals of any study. I'm not sure that I can even articulate what it means to "understand the mind". I think we can understand pieces of it. Shakespeare, Homer, and Virgil all did a pretty good job at articulating what it means to be human - so maybe it's better to look to them for the time being.

      I do agree that science writers and even scientist to make leaps. It's up to the individual to judge whether that's a good or a bad thing. Here's one example - in an article in Science about grid cells in the entorhinal cortex, the author made a reference to Kant's concept of a priori knowledge. I knew what the author was talking about. The study didn't really validate this concept from Kant, but it was an interesting slant on it. So, I didn't see it in the light of an extravagant claim, but just the art of writing for a more general public. One has to filter the metaphors a bit to appreciate what the author is doing.

      In terms of progress, I'd say we're approaching Camp 1, and are above the Khumbu Icefall. Some of us climb for the joy of climbing. Summiting is a secondary consideration for some (but not all!!).

      Delete
    4. Sorry -- have not had time to digest all this and reply. Meanwhile, here is a funny interlude on the subject (or, I hope you find it funny)...

      http://blogs.discovermagazine.com/neuroskeptic/2013/11/10/handwavers-guide-brain/#.Un__oeLjWmM

      --David

      Delete
    5. Pretty funny, yeah, that about summarizes it. Actually I think evolutionary psychology is even worse. The actual journal articles are much more careful than the pop-science articles in my experience - they're very straightforward - e.g. Lichtman's on neural paring in the neuromuscular junction.

      Delete