6 November 2006. γ-secretase may have been one of the first intramembrane proteases to be described, but since its discovery a host of these special cutters have been identified in organisms from bacteria on up. But their mechanism of action has been unclear, and controversial, because of the apparent incongruity of a water-dependent scission reaction proceeding in the hydrophobic environment of the cell membrane.
Some of the mystery has been solved with the publication last month of the first high-resolution crystal structure of an intramembrane protease. Yongchen Wang, Yingjiu Zhang, and Ya Ha from Yale School of Medicine, New Haven, Connecticut, solved the crystal structure of the core domain of the rhomboid protease GlpG from Escherichia coli. The six transmembrane segments enclose the catalytic site, with water, smack in the middle of the membrane. Their structure suggests that substrate enters laterally through a V-shaped opening between two of the transmembrane helices. The opening to the active site is blocked by a half-submerged loop, which could be a substrate gating feature.
Based on sequence comparisons, the GlpG structure is likely to be representative of rhomboid family members, but not of presenilins. As Matthew Freeman points out in a commentary accompanying the paper, which came out October 11 in Nature online, presenilins are not related to rhomboid proteins by mechanism, sequence, or evolution. For γ-secretase fans, the structure, then, is all about possibilities, about what presenilins might look like. In this vein, Ha and colleagues point out that presenilins have a highly conserved hydrophobic sequence, which might resemble the putative substrate gate in the rhomboid structure, and note that many FAD PS mutations are found in this region.
We talked with senior author Ya Ha about the rhomboid structure, and what it means for understanding presenilins. See below.—Pat McCaffrey.
Wang Y, Zhang Y, Ha Y. Crystal structure of a rhomboid family intramembrane protease. Nature. 2006 Oct 11;:1-5 [Epub ahead of print] Abstract
Freeman M. Structural biology: Enzyme theory holds water.
Nature. 2006 Oct 11; [Epub ahead of print] Abstract
Q&A with Ya Ha. Questions by Pat McCaffrey.
Q: You’ve solved the first structure of an intramembrane protease. What does the rhomboid structure tell us about this type of enzyme in general?
A: I think the main implication is, first of all, that intramembrane proteolysis really takes place in the membrane, as people have hypothesized. If you read Matthew Freeman’s News & Views commentary, he said, and a lot of other people have said similar things, “Now the doubters can be satisfied.” Why do people say that all the time? It’s because there are a lot of doubters out there.
I think the doubters have a very good reason to question the mechanism, since there is no such mechanism previously known. But now what we have shown is that the rhomboid protein has something that looks just like a protease active site, right in the middle of the membrane, embedded in the membrane. By seeing it, and correlating that with the previous biochemical and functional studies, to me it’s really very convincing that the cleavage takes place in the membrane.
So the interest in this structure comes not just from the medical value of some of these proteases, and not only because of their relevance to biology, but because these are also the first molecules to be studied in this particular paradigm, so they really are the pioneers of the field.
Then you can ask the question, Well if this one happens in the membrane, does the γ-secretase cleavage also takes place in the membrane?” We’d like to believe it is also happening in the membrane and that mutations do affect the mechanism of γ-secretase in Alzheimer disease.
Q: In the case of presenilins, what alternatives would be realistic if the cleavage does not take place in the membrane?
A: If you really get into the mechanistic aspect of the γ-secretase, there are a lot of mysteries, not only because cleavage happens in the membrane, but also because it appears to happen at different points in the membrane. If you imagine there’s one active site, and if that active site is fixed somewhere in the membrane, you’d ask yourself the question, How can it cleave APP at different sites? Why can it cleave Notch apparently at different sites? We know that it cleaves APP and Notch at different sites, at different depths in the membrane, and that’s very counterintuitive.
Well, maybe the protein is yanked out of the membrane—the worse case for the γ-secretase people is that the protease may not even be what we think. If it’s out of the membrane there are a lot of possibilities; other proteases could play a role. That’s just for example; I’m just making this up as an example. So there are those doubts.
Another possibility is raised by the S2P system [another intramembrane protease that cleaves the sterol regulatory element binding protein]. There was a paper that described a mechanism where somehow the substrate moved out of the membrane to be cleaved, so this is not really a radical idea.
Q: Can you use the rhomboid structure to model any features of presenilin?
A: No. They are so different. That’s the short answer, but let me explain a little.
I can say, “No, no, no, we cannot do that, don’t even try, it’s impossible.” But I can also say, if you just look at the mechanism, there must be some commonalities. And you can use these common themes to start to experiment, to design experiments based on general principles. But in terms of modeling one structure on the other to generate something to really describe the structure of presenilin, or even to guess how the mutations would affect its activity, that’s impossible at this time.
Q: What about the recently presented low-resolution structures of presenilins? Can you make a comparison to your rhomboid structure?
A: The implications of the two structures are very different. Let me just make a careful statement on my interpretation of the two structures.
In our structure, the active site is buried inside among the transmembrane helices. It’s a very small site—a few loops may move and open up enough space for a peptide to go in; that’s what’s implicated in our structure. In the presenilin structure, there appears to be a fairly large hole in the middle of the protein which is very big and which I would say can almost fit a protein. You can fit a lot of things in there. What they mention in their paper has some similarity to a proteasome, which is a soluble protease. It’s like a soccer ball with a big cavity inside into which proteins get pulled in and that’s how we degrade proteins in the cell. So they compare the presenilin to the proteasome system, whereas our protein structure is more like a regular protease structure, where there is just a little trough in the protein active site, just wide enough to fit a peptide.
Also, in the presenilin structure, though they cannot see very well, they look at the features they have and there appear to be portals or small openings in the shell that would allow the protein to go into the big hydrophilic chamber. That has totally different mechanistic implications. Our structure does not exclude the possibility that presenilin may work in a different way, in other words, by creating a hydrophilic chamber in the middle of the helices. If that’s true, the mechanistic implications of the two structures are certainly different.
Q: For the AD researcher reading this, what’s the take-home message from your work?
A: Besides the implication I mentioned earlier, that proteolysis can happen in the membrane, I think our structure might get them to ask themselves this question: how does the substrate get to the active site?
For a soluble protease that’s not a problem—soluble proteases are in water, the active site is hydrophilic, it likes water. For a protease in the aqueous environment, it’s really not a big problem—the peptide will just bind to the active site, which is open. But once the active site is in the middle of the membrane, the active site is enclosed by the protein structure. Somehow, the transmembrane substrate has to get at it, and that was another thing people had doubts about. If the active site is already closed, how can a substrate get in? It seems impossible.
But now our structure suggests ways a transmembrane structure may go through the protease structure, through some sort of a lateral mechanism with a gate that controls access to active site. I’ve heard this idea mentioned in regard to γ-secretase, and now I think researchers should feel more confident about pursing the idea further that the substrate can actually go through between the transmembrane helices of the protein and into the active site. That lateral movement idea could definitely be pursued further.
Q: You’ve done structural work on APP. Are you still working on that? And what about presenilin?
A: Yes, we’re still working on APP. Presenilin is on our radar screen.
Q: The rhomboid protease is a single polypeptide, while the γ-secretase is a multi-protein complex. Will it be amenable to solving the structure by a similar approach?
A: I would say definitely yes. A lot of people in this field like to say this is a difference, where some are single proteins and some are multi-protein complexes. But from a structural biology point of view, there’s really no difference. Sometimes multi-component complexes actually appear as single proteins in some species because they are transcribed as single polypeptide chains with multiple domains. In other species, each domain is a separate protein. Once they are together, it doesn’t matter if it’s multiple proteins or a single protein. You may have a different level of regulation, but at the bottom line, at the structural level, there is really no difference. I think there are very amenable, similar approaches.
I doubt a structure of isolated presenilin would be successful. The evidence appears quite convincing right now that this protein can only be functional in the presence of other components. The reasons for that may be manifold, but the simplest possibility is that without other proteins, the protease may not even fold correctly. So you need them together. But bigger is not necessarily worse—it doesn’t make it more difficult. A small protein can be just as difficult in terms of getting the structure.