Important flu paper on immune response

Every day, it seems, we find out that what we thought we knew about flu isn't the case. As one noted flu expert said to me once, "I knew much more about flu 20 years ago than I do now." So it's good to remember that we are also finding out a lot about flu that we never knew or even thought we knew. A case in point is an extremely important new paper in PLoS Medicine ( Khurana S, Suguitan AL Jr., Rivera Y, Simmons CP, Lanzavecchia A, et al.(2009) Antigenic Fingerprinting of H5N1 Avian Influenza Using Convalescent Sera and Monoclonal Antibodies Reveals Potential Vaccine and Diagnostic Targets. PLoS Med 6(4): e1000049; online as of last night). This work makes a major advance in the science of antibody response to avian influenza/H5N1 ("bird flu"). The advance has two aspects. One is the information the work generated. Even more important is the second part: opening up specific new questions for further research.

Unlike much H5N1 work, this isn't based on experiments in mice, as important and fruitful as such work is and has been. Instead it examines the antibody response of victims of a 2004 bird flu outbreak in Vietnam. Of 18, 13 died. Blood samples were obtained from the survivors during their recoveries. These patients lived long enough to get a response from the part of their immune system that makes antibodies. What did that antibody response look like?

The question is a difficult one to answer. In analogy with how we recognize other people, you might think of the features of the virus, as seen by the immune system, to be like what clothes it's wearing, what style and color they are and what tools the virus is using. Just as we only use some features of another person for recognition -- for example, we usually don't pay attention to their elbows as a way to recognize them but do notice their hair color -- our immune system also notes only certain features of the virus. But which ones? There are many practical and scientific reasons for wanting to know this. The presence of these specific antibodies might be useful for a marker of past infection, for example, or may signal a vulnerable part of the virus to attack with drugs or vaccines. We know that the antibody response is not confined to a single recognition feature but may involve many. Not surprisingly some of the most prominent signals for antibody response are the ones prominently displayed on the surface of the virus (where they are easily "seen"), particularly on the hemagglutinin (HA) protein. As a result, this protein has a tendency to change rapidly as it comes under selective pressure from the immune system. That's why we need different flu vaccines each year and why one form of flu doesn't protect us from others or even last year's. Is the HA protein the only recognition feature? Most of what we think we know about the human response to H5N1 comes from mouse experiments or analogies with human responses to seasonal influenza. As a result of this paper we now know much more about human response to H5N1 and, as importantly, we are pointed in directions to look for more answers.

So what was done and how? The international team of researchers (US FDA/NIH, Oxford, Vietnam, Switzerland) took the antibody-response-containing serum of the H5N1 survivors and tested it against most of the proteins from the responsible viral isolate (A/Vietnam/1203/2004 (H5N1)), using whole-genome-fragment phage display libraries (GFPDL) to figure out the antibody targets. In addition they used a similar technique to explain why two pure (monoclonal) antibodies obtained from the blood of four of the same patients had different kinds of protection against other flu virus. Both protected against infection against closely related virus as the original infection (clade 1 on the family tree), but only one also protected against more distantly related bird flu virus from clade 2 (from Indonesia). Why?

Here's a brief description of GFDPL. Genetic material from all eight segments of the H5N1/A/Vietnam virus (that's the whole-genome part) was inserted into the genetic program of a bacteriophage. A phage is a virus that infects bacteria. When it gets into the bacteria it does the same thing that viruses do when they infect our cells: it hijacks the bacterium's protein making machinery for the sole purpose of making copies of itself. Because fragments of the flu virus's genome have been included in the coat making gene of the bacteriophage, tremendous amounts of bacteriophage bearing viral protein fragments on their surface are produced. Each bacteriophage virus has only one such fragment, but in totality they display a whole library's worth of fragments, one per bacteriophage virus. I'm making this sound too simple, of course. Separating the different library volumes (each bearing a specific flu virus protein fragment) is an arduous process (called "panning," as in panning for gold). It involves repeated enrichment of the specific fractions. The science behind this marvelous, but it is always important to remember implementing it is usually hard work at the bench, much of it tedious, boring and often unsuccessful. A lot can go wrong along the way of an experiment with possibly dozens of sequential actions. We don't see this in the finished paper. After the climb to the level of reliable results, the difficult experience on the ladder becomes invisible. Anyway, using this technique and a closely related one (random peptide libraries) the NIH researchers and their colleagues were able to do two things. One was explain why the monoclonal antibodies acted differently on the two clades of H5N1. The other was to identify many new recognition targets ("epitopes") used by the immune systems of the Vietnamese patients to make antibodies.

First the monoclonals. The antibody that only worked on clade 1 viruses targeted a portion of the HA (hemagglutinin) protein (L129) that differed in clade 1 and clade 2 viruses. That difference was enough to make the first monoclonal only work on clade 1. The second monoclonal, by contrast, targeted a portion of the HA that was identical in clade 1 and clade 2. In both cases the recognition sequences were on parts of the protein that were separated from each other in the protein sequence but which came close together when the HA was folded (think of a bunched up string that might have its two ends close together although they are on opposite ends). And while both of the recognition targets were in t receptor binding site, the second monoclonal bound to the viral protein target much more tightly than the first, suggesting that it is not only what an antibody recognizes that might be important but how tightly it binds to the recognition site. More research to do.

What about the repertoire of targets seen in the patients? In other words, what other antibodies were in their sera and what were they directed against? Studies on mice mainly involve the HA protein, but this work revealed still more HA targets not known from mice and targets on neuriminidase (NA), M2e, M1 and NP. In addition, there were antibodies against the reading frame variant from PB1-F2. This is a protein that is not part of the viral structure but only appears when the virus is infecting someone. While we know something about it in experimental systems, there had never been solid evidence that it was being expressed in infected people. Not only is it being expressed, but this study suggests it is seen by the immune system and is a target.

We still don't know if the antibodies in the survivors had anything to do with their survival. It may be that it is the other way around. The fact that they survived allowed enough time for their bodies to make antibodies. Do any of these targets involve viral clearance? Which ones? Which ones involve protection against the virus? Many of these antibody targets aren't seen in seasonal flu cases. Which ones are the result of the very severe pathology of bird flu rather than the virus itself? Can any of these antibodies be used to detect past infections that weren't apparent at the time?

It's clear this paper is hardly the last word on the subject. On the contrary, it is more like the first word. There is much I left out of interest (PLoS Medicine is open access so you can read the paper and the accompanying editorial; the Editor's explanation at the end is quite understandable), but of greater importance is the door to further research it opens.

This is very interesting research and tremendous fun to read. Fun or not, though, it is about a deadly serious problem. So any progress is welcome. And this is real progress.