Over the years I’ve very much enjoyed being startled by the scientific literature, and there haven’t been many times when I’ve been more surprised than I was this morning. I’ve been making references on this blog for years about how we don’t even know how antidepressants work, but if this new paper is correct, then perhaps now we do. I’m amazed.
It’s from a multinational team led out of the University of Helsinki, and they have made some very interesting finds about the tyrosine kinase receptor-2 protein (TrkB). That has long been known to be a player in neuronal plasticity – TrkB is a high-affinity receptor for brain-derived neurotrophic factor (BDNF), which as its name implies, is one of the neurotrophic signaling proteins that can regulate the growth and branching of axons and synaptic density. These have been studied intensively since the 1980s as possible players (or treatments) in a number of CNS conditions – I vividly recall being on a (doomed) program that was investigating a related protein, nerve growth factor (NGF), for Alzheimer’s. Most of these work by binding to receptor tyrosine kinases – for example, NGF binds to one called TrkA.
The brain being what it is, it’s been hard to make headway on therapeutic uses of these systems, although they’re clearly extremely important. In recent years, a BDNF-centered hypothesis for depression has been on the rise. There are a lot of things about that idea that make sense, not least that it appears that BDNF levels seem to be correlated with features of depression itself in both animal models and humans, and it’s been shown that several known antidepressant drugs increase BDNF signaling through some unknown mechanism. The belief has been that this is an indirect effect of these drugs, because (after all) it’s already known they have high-affinity targets in the monoamine signaling system, e.g. the serotonin reuptake inhibitors like fluoxetine (Prozac). The same argument is made for the NMDA compounds, dopamine ligands, and the others, and it’s not a stupid idea by any means. If there’s one thing we’re sure of about these brain receptors, it’s that they have crosstalk with each other six ways from Sunday. So BDNF could well be a big factor, but there’s room to argue about all of this, as there is in most any important question about the functioning of the brain.
So one of the things this paper reveals is that a function of the TrkB protein is to serve as a sensor for cholesterol levels. I know that we’ve all been exposed to decades of public health stories to the effect that Cholesterol = Death, and when it coats the inside of your arteries that’s pretty accurate. But it’s also essential for life, as an irreplaceable component of every cell membrane. If anything, it’s even more important in the brain, where it’s been shown as yet another regulator of neuronal plasticity and is a big component of the myelin sheaths and the brain’s white matter. Out in the rest of the body, cholesterol is handled by a complex lipoprotein system involving the liver, the small intestine, and other organs that’s become famous over the years. But all this stops at the blood-brain barrier, and the brain has its own complicated system to import and distribute cholesterol to the neurons.
This paper identified a cholesterol recognition motif in the sequence of TrkB (which isn’t present in the related receptors). Addition of cholesterol to BDNF/TrkB systems had significant effects on signaling, and seemed to also promote the movement of TrkB protein from intracellular compartments up to the cell surface (which would also be expected to change the signaling landscape, naturally).
Now to the BDNF hypothesis. I used the phrase “unknown mechanism” above, and that’s exactly what this work may have cleared up. The authors show that when the TrkB protein forms a dimer in the cell membrane, a binding site for small molecules is formed at the interface. A whole list of known antidepressants (fluoxetine, imipramine, venlafaxine, moclobemide, ketamine, esketamine, and R,R-hydroxynorketamine) bind to this site at about 1 micromolar levels (and can displace each other in binding assays(, while a set of control CNS compounds like chlorpromazine, diphenhydramine, and indeed S,S-hydroxynorketamine do not. It will not be lost on those who’ve done research in the field that the antidepressant compounds listed above have been thought to work through completely different mechanisms.
But all of these compounds have similar effects on the TrkB-BDNF binding interaction – for example, they promote membrane trafficking of the TrkB protein itself. A very interesting tie-in is to the known effect of antidepressants on the number of AMPA receptors in the cell membrane, which has been linked to the antidepressant effects of ketamine and its derivatives. This effect is shut down by known TrkB inhibitors, so it appears to be downstream of the binding of the ketamine class to TrkB. It also appears that the shape and size of this small-molecule binding site is altered by the cholesterol-sensing functions of the TrkB transmembrane domain, and a prediction is that the binding would be most favorable in the most cholesterol-rich parts of the cell membrane. In what may not be a coincidence, that would the synaptic region in a neuron.
The authors do a good job of putting the hypotheses to the test by making various point mutations of the TrkB protein, and they show that they can disrupt the cholesterol-sensing function, the proposed small-molecule binding, and the BDNF effects by hitting key residues according to prediction. These effects are seen in binding assays and in super-resolution microscopy of the cell membrane, as well as in assays of neuronal plasticity itself. Finally, in a key experiment, it’s shown that such mutant forms of TrkB do in fact abrogate antidepressant drug effects in mice that express the altered receptor proteins. The results are very similar to those that you see in BDNF-deficient animals. There’s a lot of work that’s been put into this manuscript.
The cholesterol part of the story comes up in these experiments, too. The authors observed that the behavioral effects were also lost under acute pravistatin treatment (altering cholesterol homeostasis, and presumably having an effect on the TrkB small-molecule binding site as it senses a decrease in total membrane cholesterol). From what I can see, there no association has been found between statin use and depression (and it’s not like people haven’t looked!), but this is an interesting thing to consider. Perhaps if there is an effect it’s a transient one, as the brain then gets its cholesterol levels back up to a more acceptable range? Unknown.
All in all, this is an excellent and thought-provoking paper that has a lot of evidence backing it up. If these results stand up to further experimentation – and believe me, there’s going to be a lot of experimentation – then this is a major advance in our understanding of depression. Which would be very, very good news. I started off my work in the drug industry doing CNS medicinal chemistry, and was terrified about how little we really understood about such illnesses. It’s taken decades of work, but perhaps a key light switch has been turned on. Let’s hope so.