Shift Happens (Part I)
OK, I know what you're thinking: enough with the cryptic transmissions and vague allusions to a higher calling -- can he still bring the science? Yes we can! I don't intend to do this kind of thing much anymore, but I'm making an exception because I haven't been this jazzed about a new paper since I went ga-ga over Hawks et al (2007).
The paper in question is Andersen et al (2009), and the reason it's awesome -- aside from being a short, lucidly written, impeccable exemplar of good science -- is that its implications for neuroscience are huge. It ties together a lot of experimental work under a cohesive thesis in a really neat way. I have some friends who are going to resent this paper at first, but I'll aim to show why it's worth learning to love.
Many moons ago I started a series on basic neuroscience, which I then subsequently dropped, in part because because I got distracted by other questions I was led to in the process of learning the material. It's just as well, since this paper has convinced me that the standard model of neural activity I was basing my understanding on is wrong.* False. Incorrect. Erroneous. Inaccurate. This is an ex-paradigm.
The story begins with a simple observation, which actually antedates the Hodgkin-Huxley electrical model: in phase with the transient voltage spike traveling down an axon during an action potential, there's also a corresponding temperature spike -- heat is released and then reabsorbed by the membrane. This isn't what you should expect if your model of an action potential is based on current flows across the membrane: heat should be generated by the current flow and then slowly diffuse into the cell's environment, not spike back down in tandem with the voltage.
Instead what you see during each action potential is nearly zero net heat release. This has been noted occasionally in the literature but rarely addressed directly and never adequately explained. So the standard model makes an incorrect prediction. Likewise, there are other interesting observations of neural behavior that just aren't accounted for at all within it -- one being that the axon bulges outward laterally and shortens longitudinally when the neuron fires, and another being that you can induce action potentials by locally cooling or compressing a section of the neuron without any exogenous current added (in fact, turns out you can even do it when there are no sodium or potassium ions in the extracellular fluid at all).
So that's the empirical motivation for the alternative model being proposed by the authors, which combines these observations with one other under-appreciated fact: that cell membranes spend most of their time just barely on the fluid side of the fluid-gel phase line, and shift to the gel phase at temperatures slightly below "standard" body temperature. When this happens, the membrane thickens by about 16% and the volume increases by about 4%, which accords with the observations of mechanical deformation during an action potential. Moreover, these geometric changes alter the capacitance across the membrane in a way that happens to be just the right magnitude to account for the change in voltage during the impulse. (Incidentally, this also falsifies the assumption of the HH model that membrane capacitance is constant during the action potential: if the surface area and thickness change then the capacitance must too.) In experiments with fluorescent dyes, the changes in molecular flows across and around the membrane during an impulse are consistent with what you'd expect from such a transient phase change.
What the authors propose is a fairly straightforward thermodynamic interpretation of nerve pulses as pressure solitons -- wave packets akin to sound waves -- traveling down an axon, triggering a phase-shift which results in all the observed physical changes. On this view, the voltage gradient across the membrane is a peripheral aspect and the osmotic pressure gradients are what govern neural activity. Membranes are known to have the properties necessary to support pressure solitons, and this would neatly account for all the empirical observations mentioned above.
The authors also make a sexy theoretical point in contrasting their model with the current model: that there's a gross mismatch of scale between the phenomenon to be explained and the proposed current explanation for it. Action potentials propagate at rates ranging from 0.1-100 m/s depending on the type of neuron, and last for about 1-20 ms each, yielding a length range of a few millimeters to a few centimeters. But the protein pores in the membrane that transport ions are only about 5 nm in diameter and spaced out such that there are only at most a few thousand of them per square μm of membrane surface (and in myelinated axons, large spans where there are none of them whatsoever). There's about three to six orders of magnitude separating the phenomenon from the proposed loci of causation, and it's not at all clear how the ion pores function as an ensemble, or if they do at all. Contrariwise, there are several million lipid molecules per square micrometer which form a continuous ensemble whose macro-level properties are well understood.
From a biophysical "first principles" POV, the soliton model looks pretty appealing at this point. But the test is whether it can make a wider range of accurate predictions than the current model, and that's what I'll cover in the next post.
* As an aside, let this be a lesson to all of you: I am only as good as my inputs and assumptions, and if those are screwy then so is whatever I make of them. And most people are less epistemically vigilant than I am, so whatever hit your level of trust in what you read here just took, almost everything else you read should look just a shade less real right about now. For those who played it as a kid, now might be a good time to meditate on the broader implications of the "telephone" game.
The paper in question is Andersen et al (2009), and the reason it's awesome -- aside from being a short, lucidly written, impeccable exemplar of good science -- is that its implications for neuroscience are huge. It ties together a lot of experimental work under a cohesive thesis in a really neat way. I have some friends who are going to resent this paper at first, but I'll aim to show why it's worth learning to love.
Many moons ago I started a series on basic neuroscience, which I then subsequently dropped, in part because because I got distracted by other questions I was led to in the process of learning the material. It's just as well, since this paper has convinced me that the standard model of neural activity I was basing my understanding on is wrong.* False. Incorrect. Erroneous. Inaccurate. This is an ex-paradigm.
The story begins with a simple observation, which actually antedates the Hodgkin-Huxley electrical model: in phase with the transient voltage spike traveling down an axon during an action potential, there's also a corresponding temperature spike -- heat is released and then reabsorbed by the membrane. This isn't what you should expect if your model of an action potential is based on current flows across the membrane: heat should be generated by the current flow and then slowly diffuse into the cell's environment, not spike back down in tandem with the voltage.
Instead what you see during each action potential is nearly zero net heat release. This has been noted occasionally in the literature but rarely addressed directly and never adequately explained. So the standard model makes an incorrect prediction. Likewise, there are other interesting observations of neural behavior that just aren't accounted for at all within it -- one being that the axon bulges outward laterally and shortens longitudinally when the neuron fires, and another being that you can induce action potentials by locally cooling or compressing a section of the neuron without any exogenous current added (in fact, turns out you can even do it when there are no sodium or potassium ions in the extracellular fluid at all).
So that's the empirical motivation for the alternative model being proposed by the authors, which combines these observations with one other under-appreciated fact: that cell membranes spend most of their time just barely on the fluid side of the fluid-gel phase line, and shift to the gel phase at temperatures slightly below "standard" body temperature. When this happens, the membrane thickens by about 16% and the volume increases by about 4%, which accords with the observations of mechanical deformation during an action potential. Moreover, these geometric changes alter the capacitance across the membrane in a way that happens to be just the right magnitude to account for the change in voltage during the impulse. (Incidentally, this also falsifies the assumption of the HH model that membrane capacitance is constant during the action potential: if the surface area and thickness change then the capacitance must too.) In experiments with fluorescent dyes, the changes in molecular flows across and around the membrane during an impulse are consistent with what you'd expect from such a transient phase change.
What the authors propose is a fairly straightforward thermodynamic interpretation of nerve pulses as pressure solitons -- wave packets akin to sound waves -- traveling down an axon, triggering a phase-shift which results in all the observed physical changes. On this view, the voltage gradient across the membrane is a peripheral aspect and the osmotic pressure gradients are what govern neural activity. Membranes are known to have the properties necessary to support pressure solitons, and this would neatly account for all the empirical observations mentioned above.
The authors also make a sexy theoretical point in contrasting their model with the current model: that there's a gross mismatch of scale between the phenomenon to be explained and the proposed current explanation for it. Action potentials propagate at rates ranging from 0.1-100 m/s depending on the type of neuron, and last for about 1-20 ms each, yielding a length range of a few millimeters to a few centimeters. But the protein pores in the membrane that transport ions are only about 5 nm in diameter and spaced out such that there are only at most a few thousand of them per square μm of membrane surface (and in myelinated axons, large spans where there are none of them whatsoever). There's about three to six orders of magnitude separating the phenomenon from the proposed loci of causation, and it's not at all clear how the ion pores function as an ensemble, or if they do at all. Contrariwise, there are several million lipid molecules per square micrometer which form a continuous ensemble whose macro-level properties are well understood.
From a biophysical "first principles" POV, the soliton model looks pretty appealing at this point. But the test is whether it can make a wider range of accurate predictions than the current model, and that's what I'll cover in the next post.
* As an aside, let this be a lesson to all of you: I am only as good as my inputs and assumptions, and if those are screwy then so is whatever I make of them. And most people are less epistemically vigilant than I am, so whatever hit your level of trust in what you read here just took, almost everything else you read should look just a shade less real right about now. For those who played it as a kid, now might be a good time to meditate on the broader implications of the "telephone" game.