Last year, I wrote about a scientific controversy over the structure of the influenza M2 proton channel, particularly over the protein's binding site for adamantane type anti-flu drugs. The Schnell/Chou model, based on solution NMR, had the drug binding to the outside of the channel, within the membrane (at a 4:1 drug:protein ratio). On the other hand, the Stouffer/DeGrado model had the drug binding inside the channel (1:1 ratio), based on X-ray crystallography studies.
A new study was recently published in Nature (the same journal that published the original two competing papers), based this time on solid state NMR. This study, by lead author Sarah Cady in Mei Hong's lab, finds evidence for both binding sites, but argues that the pore (Stouffer/DeGrado model) is the higher affinity site. This subject involves a legitimate scientific controversy, and has clearly been in need of a third party to help resolve it. Unfortunately, the DeGrado group was also involved with this one, so it's not quite the impartial judgment that this subject needed. Regardless, the new study does offer additional evidence to support the Stouffer/DeGrado model, although I don't think it's quite the final word.
I'm an NMR scientist, but I use solution state NMR specifically. However, solid state NMR is a whole other animal--in some ways, at least--so I can't write in great detail about the methods used in the new paper (and that would make for a very opaque and rather boring blog post anyway!). Therefore, I will just provide a few comments about the paper and what I think its ramifications are for this controversy.
Cady et al. make the case that adamantane drugs bind within the M2 pore and that the site on the outside of the paper is largely irrelevant. Certainly the new paper provides compelling evidence in favor of the Stouffer/DeGrado model, but the fact that the other binding site was observed is also significant. The site inside the pore of the M2 protein has higher affinity, but the affinity is only "at least 40-fold greater", according to the paper. Although this is a clear difference, in biological terms this isn't so enormous. They also note that in order to observe the membrane binding site, the drug must make up a large proportion of the total membrane fraction. But, that's not necessarily an argument against the membrane binding site, since in vivo the drug would also naturally partition into the membrane.
Overall, the Cady et al. paper appears to be technically solid to me, although I don't have the expertise to evaluate the methods rigorously. Certainly, they demonstrate that the drug binds with high affinity within the M2 protein pore, and they can measure distances between atoms in the protein and the drug with great precision. However, when the authors write that they have solved the structure of the drug binding site, it's important to emphasize that they have only solved the structure of the binding site at high resolution, and not the rest of the M2 protein. Because the M2 protein consists of four separate subunits, a very important part of solving such a structure is measuring distances between these different subunits. Cady et al. have only measured four distances between subunits (really 1 distance that occurs four times in the structure, since the four subunits are identical). They also have 60 (15x4) orientational measurements, plus some additional distance measurements within the same subunit (and it is my understanding that all of these measurements are from their earlier studies).
What I thought was somewhat misleading about this study, though, was that the authors write "For comparison, the previous [Schnell/Chou] solution NMR M2 structure ensemble was constrained by 12 interhelical NOEs and 18 amide residual dipolar couplings for the TM region." I call this misleading, because if they were going to count these the same way they counted their own measurements, these would be 48 (12x4) distance measurements between subunits and 72 (18x4) orientational measurements. These are just the measurements within the transmembrane portion, and the Schnell/Chou structure has additional measurements outside of this region, since they used a longer construct (plus hundreds of additional measurements within the same subunit). (I'll give Cady et al. that it can be much more painstaking to make these measurements by solid state NMR, so this is certainly an impressive structure that was probably very difficult to achieve, but it's important that we make all of these comparisons on the same terms.)
One aspect of the new study that I think is particularly good, though, is that these authors used lipid bilayers, as opposed to the detergent micelles used in the Schnell/Chou study. Basically, this means that in this study the protein is in an environment that is more similar to its natural environment. However, as in the Stouffer/DeGrado study, Cady et al. used a short protein construct--as mentioned above--only consisting of the transmembrane segments of the M2 protein. The Schnell/Chou study, on the other hand, used a longer construct that is probably more biologically relevant. The transmembrane portion of the M2 protein is clearly the most important and relevant part, but the other portions may also affect the structure and function of the protein.
So, what does all of this mean in the end? Well, I think it's too early to say. Personally, I'm starting to believe that both sites may have some biological relevance, and that could explain how both sides of the controversy have been able to amass such a significant amount of data in favor of their respective sides. Another factor here that I haven't mentioned yet in this post is that these studies have used two different--although very similar--drugs: rimantadine in the Schnell/Chou study and adamantane in the others. I would be surprised if this turns out to be an important distinction, but it's something that should be explored further nonetheless.
Therefore, while the new study provides compelling new evidence in favor of adamantane drugs inhibiting M2 by binding within the pore, in my mind this still isn't a closed case.
Cady, S., Schmidt-Rohr, K., Wang, J., Soto, C., DeGrado, W., & Hong, M. (2010). Structure of the amantadine binding site of influenza M2 proton channels in lipid bilayers Nature, 463 (7281), 689-692 DOI: 10.1038/nature08722
Hat tip to Revere of Effect Measure.
This is very interesting. There is precedent for high-potency hydrophobic drug binding to ion channels in the membrane phase based on a two-stage process of non-specific partitioning in the membrane and then binding to a low affinity binding site on the channel protein. As you mention, the two dimensionality of the drug diffusion space in the plane of the membrane results in the high potency. This is the most likely scenario for dihydropyridine drug binding to L-type voltage-gated calcium channels.
One important aspect that you missed in reading this latest paper is that selective labeling allowed the authors to identify the drug in the structure. (see methods) In fact they claim that Chou might have had the drug in the channel, but did not see it without selective labeling of the amino acid involved. This was only possible because the DeGrado group synthesizes the peptides.
Also, being at high pH, this structure is different from the previous crystal structure. Remember that this particular protein lives between different pH environments.
Actually, I didn't miss that Cady et al. used selective labeling. As I mentioned in the post, I wasn't going to go into great detail about his paper, and that's not a detail that I thought was particularly relevant to the discussion here. In fact, there would not have been a great advantage to selectively labeling M2 for the solution state NMR studies on this system. Schnell et al. performed the experiments with universally isotopically labeled and deuterated protein, and with protonated drug, allowing them to be able to specifically identify any resonances coming from the drug. Maybe you're referring to this passage from Cady et al.: "Thus, the drug may have been present in the lumen in the solution NMR sample but not observable without isotopic labeling." I don't totally understand this point, since Schnell et al. were observing proton-proton NOEs--the most robust method for measuring atomic distances in solution NMR--so isotopically labeling the drug would not have assisted with this.
You are correct that Cady et al. solved the structure at neutral pH, as did Schnell et al.
I think this is a nice summary of just how difficult it is to know with a structure what is "real" sometimes - this isn't the only structure where there is heated debate about which is right and which is wrong, partly it's a case of NMR vs crystallography but there are so many factors to take into account. And now we're beginning to see subtle changes in protein structures that we weren't picking up before. Ithink we're in for interesting (read: difficult) times!
I don't really see it as one structure is right and the other is wrong since they not even looking at the same system.
Different drugs and protein truncations could mean that they are both correct (or wrong).
There has also been a "paper" in trends in biochemical sciences on the topic (TIBS 35, page 471). It uses a computational approach to identify the hotspot for drug binding. If any of you has the time, I'd like to know what you think about it... It confirms the existence of both binding sites, but highlights the pore one as the primary drug binding site.