Why Swine Flu Is Resistant to Adamantane Drugs

ResearchBlogging.orgOn Wednesday, the CDC reported that influenza A H1N1 viruses from 13 patients with confirmed diagnoses of swine flu had been tested for resistance to a variety of antiviral drugs. The good news was that all of the isolates were susceptible to the antiviral drugs oseltamivir (Tamiflu) and zanamivir (Relenza). However, all 13 were resistant to adamantane-based drugs (amantadine and rimantadine). Resistance to adamantane drugs (which were developed first) has actually become quite widespread among flu viruses in general, so oseltamivir and zanamivir are commonly the drugs of choice.

i-1362f1de2f6fcb2b6256ee49389b2e99-M2.gifThe reason for the difference is that the adamantane drugs target different viral proteins from oseltamivir and zanamivir. The two major proteins on the surface of the influenza virus are hemagglutinin (the "H" in H1N1) and neuraminidase (the "N"). Due to the evolution of the influenza virus over time, these proteins come in a variety of different forms, and we label strains of flu by the specific class of hemagglutinin and neuraminidase the virus carries (H1N1, H3N2, etc.). These are the largest proteins on the surface of the virus, and they are the ones that our immune system generally reacts to. However, they aren't the only ones.

The surface of the influenza virus also has a much smaller protein called M2, which acts as a channel to let hydrogen ions pass across the virus' outer membrane. M2 plays an important role in viral function (more on that below), and, not surprisingly, so do hemagglutinin and neuraminidase. Specifically, hemagglutinin allows the virus to attach to a cell in order to infect it, and after the virus has replicated within the cell, neuraminidase allows the daughter viruses to detach from that cell and infect other cells. For a more in depth description of these proteins, check out this post at Effect Measure.

Getting back to antiviral drugs, oseltamivir and zanamivir work by inhibiting neuraminidase (for details, once again check out the above post at Effect Measure or this more recent one). The adamantane-based drugs, however, target M2. In January 2008, two research groups independently published atomic-resolution structures of the M2 channel bound to an inhibitor. Using NMR, Jason Schnell (in James Chou's lab at Harvard) solved the structure of rimantadine bound to M2 (shown above with the drug in red). Alternatively, Amanda L. Stouffer (in William DeGrado's lab in Pennsylvania) used x-ray crystallography to solve the structure of amantadine bound to M2.

The overall structures were similar, but the two methods detected totally different drug binding sites. Stouffer et al. found the drug directly blocking the pore through which hydrogen ions pass. Schnell et al., however, found the drug bound to the outside of the protein, suggesting an allosteric or more indirect mechanism. Rimantadine and amantadine are very similar in structure, and it is not expected that the two drugs would bind in different sites, so the likely explanation of this discrepancy is that one of the structures is "wrong" (i.e. the drug was binding in a site that was not physiologically relevant or the data was improperly interpreted).

Before I go any further, I should offer a disclaimer: Jason Schnell recently joined my department. I don't know him well, but we have met and conversed, and I have seen him present his work a few times. Largely because I am more familiar with his work, the rest of this discussion will be based on the Schnell/Chou model of M2 drug inhibition. Beyond that, though, I find the Schnell/Chou model more convincing. Although I am mostly an NMR scientist, I actually tend to believe that x-ray crystallography is a more powerful method for protein structure determination (in fact, I solved the structure of a protein complex by crystallography last summer, and my paper on that structure is currently in review). However, in this case, the NMR-based method that Schnell used to detect drug binding is much more specific and less open to interpretation than the crystallography-based method used by Stouffer. The NMR-based method is also less prone to artifacts and more physiologically relevant since it was done in solution. Schnell and Chou have also further validated their model (most recently in a paper authored by Rafal Pielak). However, Stouffer and DeGrado have also provided additional evidence for their model. I certainly cannot tell you definitively which model is correct, although I've told you which way I'm leaning. If you are so inclined, you should take a look at the original papers and decide for yourself. Regardless, the following discussion is based on the Schnell/Chou model.

The M2 channel is made up of four identical units. Each of these units has a central portion that crosses the viral membrane and two small portions on either side of the membrane. The pore that hydrogen ions travel down sits in the middle of the protein, surrounded by all four units. Normally, though, the M2 channel is closed. This is important for the influenza virus, because once the pore opens, it causes the pH to decrease inside the virus, initiating the disassembly of the virus, which is necessary for the virus to be able to replicate within the cell. So that this process won't happen prematurely outside of the cell, the M2 channel only opens when the virus is exposed to low pH.

After influenza binds to the outside of the cell, it is encircled by part of the cell membrane, which then breaks off inside the cell, forming a compartment called an endosome. The low pH of the endosome then activates the M2 channel, initiating virus disassembly. Low pH opens up the M2 channel by loosening the connections between the individual units, causing the protein to become dynamic enough that hydrogen ions can squeeze through the central pore.

Adamantane-based drugs, however, bind to the outside of M2 and cause it to be locked in the closed state. This means that in the presence of these drugs, the pH inside the virus never lowers and the virus is unable to replicate.

Swine flu, however, has a single mutation (S31N) that makes it resistant to adamantane drugs. This is actually a relatively common mutation found in flu viruses in general and the main source of all flu adamantane resistance. Over time, widespread use of this relatively older class of drugs has put significant evolutionary pressure on influenza, giving viruses with this mutation a selective advantage, leading to increasing rates of adamantane resistance in recent years.

It turns out that the S31N mutation does not directly affect the part of the protein that the drug binds to, but it actually has a more indirect effect: it makes the M2 channel slightly more dynamic and the connections between the individual units slightly looser. The channel still remains closed at higher pH, so M2 still functions properly, but the mutation makes the connections between the subunits loose enough that adamantane drugs are unable to lock M2 into the closed state.

This figure from Pielak et al. illustrates this phenomenon schematically:

i-f8aed99c67a6d96e7cc9aa9787668764-drug-resistance.gif

Overall, these findings are interesting for two reasons. Firstly, they explain an important medically-relevant phenomenon--influenza drug resistance--which has become all the more timely with the current outbreak of swine flu and the threat of pandemic. Secondly, this is a poignant illustration of the central role that protein flexibility and dynamics play in biology, a concept that scientists have only just begun to explore, but one that you'll likely hear much more about in the future, due in part to powerful advances in the field of protein NMR


Schnell, J., & Chou, J. (2008). Structure and mechanism of the M2 proton channel of influenza A virus Nature, 451 (7178), 591-595 DOI: 10.1038/nature06531

Pielak, R., Schnell, J., & Chou, J. (2009). Mechanism of drug inhibition and drug resistance of influenza A M2 channel Proceedings of the National Academy of Sciences DOI: 10.1073/pnas.0902548106


Stouffer, A., Acharya, R., Salom, D., Levine, A., Di Costanzo, L., Soto, C., Tereshko, V., Nanda, V., Stayrook, S., & DeGrado, W. (2008). Structural basis for the function and inhibition of an influenza virus proton channel Nature, 451 (7178), 596-599 DOI: 10.1038/nature06528

Stouffer, A., Ma, C., Cristian, L., Ohigashi, Y., Lamb, R., Lear, J., Pinto, L., & DeGrado, W. (2008). The Interplay of Functional Tuning, Drug Resistance, and Thermodynamic Stability in the Evolution of the M2 Proton Channel from the Influenza A Virus Structure, 16 (7), 1067-1076 DOI: 10.1016/j.str.2008.04.011


For more on flu, check out Effect Measure and Aetiology.


Update: Check out Pharyngula, where PZ Myers uses this post to take on some creationist nonsense.

Also, check out Discovering Biology in a Digital World for a discussion about M2 channel structures and mechanism of adamantane inhibition (using the Stouffer/DeGrado model).

More like this

The downside is that we always believe we are smarter than the virus or bacteria. The replication rates of these viruses and bacteria are exponential and trying to keep up with the constant mutations are staggering. Are focus should be on the overall health of the community and the world. Are processing of food should also be reconsidered. With few people getting a nourishing diet, they are more susceptible to diseases. When you process 1000's of chicken and cattle in the same facility, it is amazing how viruses and bacteria get to infect the entire batch. Food for thought.

Excellent post.

Please tell me I'm not the only person who saw the title of this post and thought it was going to have something to do with Wolverine's claws.

you were not :)

Please tell me I'm not the only person who saw the title of this post and thought it was going to have something to do with Wolverine's claws.

I'm curious now. Someone please explain!

Adamantane -> Adamantium. The comic book character Wolverine's skeleton is made of this supposedly indestructible fictional metal alloy.

Thanks for the post on anti-viral action. It was interesting.

By Nils Ross (not verified) on 01 May 2009 #permalink

Wolverine has adamantiun claws!

By John Yates (not verified) on 01 May 2009 #permalink

Wolverine can heal flu with his adamantane claws!

Nice post and quite helpful (even for a non-scientist, like me).

I have to say, though, that the figure used for schematic illustration wasn't particularly helpful, at least insofar as the only difference between the two rows appeared to me to be that the bottom row was fuzzier, and it wasn't clear what work the fuzziness was doing. :) And it wasn't clear what the black block and the white block were, though I think I got it after a moment.

As I guess you've figured out, the black box represents the adamantane drug. The fuzziness represents flexibility. So, when the pH lowers, the channel becomes flexible and dynamic enough for hydrogen ions to move through. The mutation also makes the channel more flexible (hence fuzzier)--although not flexible enough that hydrogen ions can move freely through, but still flexible enough that the drug can't actively inhibit it.

dmv, thanks for raising your hand, I was too busy scratching my head...

Nick, after reading your explanation and going back and forth to the diagram (only 6 times he sheepishly adds) me, myself and I, as a curious laypeople,were finally able to understand the process, and will now be able to wow our friends, geeks that they are....
Thanks for posting this explanation of the process and making it understandable for those of us that like to know the how's as well as the why's.

I got here from effectmeasure and will now add your blog to my daily perusals

Nick, many thanks for taking the time and effort to post the explanation(s) and graphics. Although I am a non-scientist, as a journalist it's increasingly important for people like myself to make an attempt to try to understand the science and 'technical' aspects of major issues. Even if a person doesn't follow every single thing, fathoming the essence of the latest research and what you're getting at is most important. And, given the subject matter, your writing and the 'flow' of it is very clear.

I'm glad you're enjoying the post, and I'm happy to hear some of you will be coming back. I think that this issue in particular--the binding of adamantane-based drugs to the influenza M2 channel--is quite exciting, as there is a real, legitimate scientific controversy involved, and it'll be interesting to see how this plays out in the future.

One thought (prior to grasping the mechanics of the diagram) did pop into my head which you could probably answer,

"This is important for the influenza virus, because once the pore opens, it causes the pH to decrease inside the virus, initiating the disassembly of the virus, which is necessary for the virus to be able to replicate within the cell. So that this process won't happen prematurely outside of the cell, the M2 channel only opens when the virus is exposed to low pH."

Could facilitating the "premature disassembly" by opening the channel and lowering the ph help to "cripple" or just weaken the virus's ability to reproduce by spewing it contents outside the cell?

This is getting pretty far out of my expertise now, but certainly having a channel that opens prematurely would be a major disadvantage to the virus. I don't know if anyone has developed a way to cause that to happen, but I imagine scientists are looking into that. In theory, though, that could be one way for a novel antiviral to target the virus.

Maybe a more concise question would be ,
What would be the consequences of premature dissaembly ?

Sorry I posted too quickly, thanks.

Presumably, it would kill the virus or at least inhibit its ability to infect cells (or possibly make it difficult for it to stably assemble). This is all just speculation on my part, though.

Got it, thank you, I won't hold you to your answers... ; )

Very informative article. But I have two questions:

(×1) How influenza RNP's survive the low pH and RNAse activity in the endosome?

(2) How RNP's move out from the endosome to the cytoplasm?

Thank's.

Good questions, Ariel, although they're probably more appropriate for a virologist. I can't speak to question #2, but with respect to #1, I can give a couple of thoughts. The influenza RNA stays bound to a variety of proteins, which presumably protect it from RNase activity. Also, RNA is stable in acid in general, although proteins vary in their acid stability. Plenty of proteins are stable under mildly acidic conditions, though, so I would imagine this is true for the proteins protecting the influenza RNA.

Please keep up with the good work. Stop the Jenny, the Jenny, the Jenny McCarthy "ACTIVISM". Your writings and those of the other science bloggers help to keep us all honest and reinforce the dangers of stupidity.

By Dr.ShawnaMurrayMD (not verified) on 05 May 2009 #permalink

I actually work in the Degrado lab and we are currently discussing the conflicting structures. Some things that I wanted to add that may not be obvious to a casual observer:

1. Rimantidine has been shown by multiple independent methods to stoichiometrically inhibit the M2 proton channel at one molecule per channel, which is consistent with the x-ray structure as opposed to the NMR structure.
2. The NMR structure was created with an overwhelming excess of rimantidine), which would favor non-specific binding of it to non-physiologically-relevant areas on the protein.
3. The NMR structure was unable to detect energy transfer from the side chains lining the pore to any rimantidine atoms inside the pore, however the chemical shift of rimantidine inside a cavity will be substantially different from that in bulk solution and the paper was similarly unable to detect water inside the proposed cavity either (water inside the channel also has a chemical shift different from bulk water). Thus, it cannot eliminate the possibility of a rimantidine bound inside the channel.

Thanks for the comments, Gabriel. Having read through the papers from both groups, I can say that it looks like a pretty legitimate scientific controversy to me, and it will certainly be interesting to see how it resolves itself in the future.

Hello All,

I use to work in this field. An important point to note that in both in electrophysiological experiments (from several groups), drug binding measurements using biophysical techniques and cell culture most experiments suggest a 1:1 ratio of amantadine/rimantadine binding to an M2 tetramer. Therefore when the current scientific literature is assessed the Degrado model provides a better fitin this respect. However an important point to note that the Degrado team only used a TM peptide in their experiments. I have done full length M2/M2 inhibitor binding measurements using biophysical techniques which also suggest this one to one ratio. Note Tim Cross et al also suggest on amantadine molecule in the channel causes a structural change thereby providing inhibition.

Therefore I lean on the side of the crystal structure as a more likely scenario. However it is important to note several issues with the degrado work:
1. If I remember rightly it was done on Octylglucoside a detergent, not lipid. Different detergents can cause helical variations especially in peptides and certainly in M2 (Salom et al 2001).

2. Crystal structure is a peptide of the TM domain. Other groups have shown structure in the C-terminus so a full length protein may look very different especially in tilt.

3. Can't remember how much drug the peptide was soaked in but anything over 50uM is a lot.

Therefore I am not sure either model provides a definitive explanation on the two major points:

1. How does the M2 pass protons?

2. How exactly do M2 inhibitors work? and how does teh virus produce resistance?

By Philip Spearpoint (not verified) on 07 May 2009 #permalink

Sorry to say this sentence is probably wrong: "widespread use of this relatively older class of drugs has put significant evolutionary pressure on influenza, giving viruses with this mutation a selective advantage, leading to increasing rates of adamantane resistance in recent years". Resistance to amantadine spread because it was associated with another mutation giving selective advantadge. It was not because the use of amantadine.

And in this sentence: "It turns out that the S31N mutation does not directly affect the part of the protein that the drug binds to, but it actually has a more indirect effect" ... you should mention that the S31N mutation is in the binding site proposed by the X-ray structure (and 99% of the articles on this topic, including biochemical, biophysical, electrophysiological, NMR, computational... studies)

I did say in the post that I would be basing my discussion on the Schnell et al. model. Despite that, even though the S31N mutation is in the binding site proposed in the Stouffer/DeGrado model, it does not provide a good structural explanation for how a Ser-to-Asn mutation would specifically disrupt contacts between the protein and the drug.

Actually, the very fact that S31N mutation occurs in the binding site is an overwhelmingly obvious structural explanation for how Asn at this position precludes drug binding. It both sterically hinders drug binding and changes the hydrophobic nature of the binding site to be more hydrophillic.
To address the next argument coming...
It is well known, in the field of membrane proteins, that a Ser in the transmembrane region is more often hydrogen bonded to its own backbone amide. With its H-bond satisfied, Ser behaves as Ala. Ser does not interact unfavorably with the adamantane drugs bound to the binding site. In the crystal structure, Ser does not coordinate or come within 2.15 angstroms of the drug as Chou Sr. accused (throw a surface on the drug bound crystal structure and see for yourself).
Also, mutating a Ser to Ala should have little effect, if any at all, on the conduction, drug binding, and assembly of the channel. A S31A mutation is extremely conservative and the fact that it binds drug should under no circumstances be construed to negate the pore binding site.

By Amanda Stouffer (not verified) on 09 Jun 2009 #permalink

Thanks for your comment. Clearly, both groups have presented quite a bit of data in favor of their respective models, and as an outsider I'm certainly in no position to say definitively which one is correct. Still, for me it comes down to the the fact that the way the binding site was identified in the NMR structure (by NOEs) is more definitive, although certainly that doesn't mean that's the end of the story. Either way, though, I don't think the controversy will be resolved unless one of the groups presents some new slam-dunk data or a third party conducts his/her own studies to evaluate the structures.

sir pls tell me drug target of swine flu virus

By rahul singh jasrotia (not verified) on 23 Jun 2009 #permalink

please , I want understand the relations between the symptoms eg. fever , sore throat , diarrhea , headache , dry cough etc .. thank you

Well I guess the latest paper by the DeGrado group pretty much puts a nail on the coffin in their favor. They show binding to the spurious site (Chou site) only at the insane concentration of 7% amantidine/lipid ratio. That site is an artifact of the use of micelles in the NMR structure.

Nature. 2010 Feb 4;463(7281):689-92.
Structure of the amantadine binding site of influenza M2 proton channels in lipid bilayers.
Cady SD, Schmidt-Rohr K, Wang J, Soto CS, Degrado WF, Hong M.

I am in no way trying to discount the importance of the pore binding site; the new study provides pretty convincing evidence that it exists. But, from what I can tell, the evidence is not nearly strong enough to totally discount a role for the outer binding site.