Revision time! Earlier this month, I wrote about a really interesting protein structure, the aggregated form of TDP-43 that’s found in diseases like ALS. That post (at least the way it reads now, before I go back and add to to) talks about how TDP-43 aggregates are found in neurons of people with that disease, and how such aggregates are also found in several other neurodegenerative disorders such as frontotemporal dementia (FTD).
Not so fast. There’s a paper just out that has taken a closer look at these precipitated proteins in that disease (more accurately, frontotemporal lobar degeneration or FTLD, which is a larger umbrella term). The authors obtained samples from four of the five recognized clinical subtypes of this condition, and in every case they found that the aggregated fibrils are not made out of TDP-43 at all. Instead, they are made out of transmembrane protein 106B (TMEM106B), which is also known to be involved (somehow) with the disease, but whose role is completely unclear. In some mouse models of the disease, reduced levels of this protein seem to be protective, but not in others (which probably says as much about the mouse models as it says about FTLD – the rodent models in this area are generally not much good, with endless arguments about their translatability to human disease).
I should be more exact, though: these fibrils are actually just made out of a particular 135-amino-acid piece from the C-terminal of TMEM106B. This paper, like the other one I wrote about earlier, uses cryoelectron microscopy to work out the protein structure, and this team naturally enough started fitting their model with TDP-43 protein, since that’s what these things are made of, right? And while that assumption works great with ALS-derived fibrils, it became clear very quickly that these FTLD fibrils just could not be made to fit. They just looked over the data at that point and assembled two possible protein sequences (one in one direction and one in the other), based residue-by-residue on which amino acids seemed to fit the electron densities the best. And both of the immediately rang the chimes for the C-terminus of TMEM106B as by far the best match. And it’s a very good match indeed – for example, they see extra electron density off the ends of several asparagine residues, and these fit exactly the ones that are known to be glycosylated in the protein.
There’s the structure of the aggregated protein as they see it – you can compare it to the TDP-43 structure in that earlier post, and while of course they’re different, you can see the sorts of things that lead to insoluble aggregation in both of them. Look at how those side chains mesh together like the teeth of gears, for example. What you can’t see are things like positive and negative charges matching up – although it has to be noted that this one has a few uncompensated negative charges left over, which is a bit odd. The authors speculate that this may be happening because TMEM106B is a protein in the lysosomal membrane, and the C-terminal region of it is facing the inside. Lysosomes are very acidic, and it may well be that the amino acid side chains are protonated under these conditions and aren’t really carrying negative charges at all. Overall, there are 18 beta strands in the structure, leading to the name “golf course fold” for the whole thing.
One of the ideas coming out of these studies is that we may end up classifying aggregation diseases as much by their protein folds as by the names of the individual proteins involved. As was seen in that recent TDP-43 work, when you take samples from different patients (and even sometimes from different subtypes of the disease, as here) you see the same damn protein fold across all of them. That seems to be the key feature, and as structural biology marches on we will assemble a library of toxic aggregate folds, with some of them having more than one protein able to land in them. I suspect that this one will remain unusual because of those buried (pseudo)negative charges, though.
At any rate, this result is going to change the direction of research in this disease, and that’s a good thing. Are TMEM106B fibrils pathogenic in themselves? Or are they just a sideshow, a waste product of some other pathogenic process? If they are direct causes of trouble, is that caused by loss of the functional protein itself, or do the aggregates have some direct toxic effect of their own? All these questions and more are now in play, because we finally are looking at the right protein.
