Virus preference for humans and birds in the 1918 virus

A paper published on Monday in the Proceedings of the National Academy of Sciences (PNAS) extends the work of a group of glycobiologists at MIT on unravelling why some flu virus likes bird cells and others like human cells. Glycobiology is the science that investigates the sugar studded proteins on the outside of cells. Like a suit of clothes, a cell's glycoprotein cover plays important functions in protecting the cell, identifying it and as a signal to interact with things outside of itself, such as hormones or immune cells. But other organisms have learned to use the same signals and can use glycoproteins as docking locaitons ("receptors") to gain access to the cell. The influenza virus does this with a specific kind of glycan (sugar) tip on the protein hairs that stick up from the cell membrane. For a long time we believed that the difference between a bird virus and a human virus was that the bird version looked for a specific sugar connected in one way while the human version looked for the same sugar connected in a slightly different way (see our previous posts on this arcane subject here, here and here). But there were always some things that didn't quite fit this picture. It appeared that human cells had plenty of the bird linkages in the upper respiratory tract but still weren't often infected by bird viruses and vice versa. It seemed that something besides the linkage of the sugar itself was involved.

Within the last several months a new idea emerged from the MIT group. It was not just the linkage. The paper couched what else was required in terms of the "topology" of the sugar receptor, calling the bird version cone shaped and the human version umbrella shaped and that is how it has been reported. Reading the papers, however, it is clear these descriptions should not be taken too literally. The main difference resides in the length of the sugar chain. The MIT group has identified long chain and short chain glycans. The key hypothesis is that the short chain version binds well to bird flu viruses and the long chain ones to human adapted viruses. The new PNAS paper takes another couple of steps to confirm this by showing that the protein on the surface of the influenza virus that docks to the cell (the hemagglutinin or HA protein) can be analyzed in terms of why the human ones like long chains and the bird ones short chains.

Specifically these researchers showed that the HA from a fatal case of 1918 flu that occurred in October 1918 in South Carolina binds strongly to cells with long sugar chains, cells that are found in the human upper respiratory tract. In particular these cells are mucus secreting cells called goblet cells. A single mutation in the South Carolina virus HA can alter the binding behavior, making it sufficiently weaker that the virus no longer transmits well in one of the main animal models for human influenza, the ferret model. While this virus still binds to long chain sugars, it does not do so as well because it has lost one of its contact points. Because there are many binding sites arrayed on the surface of the virus, this apparently makes the net binding considerably weaker. By making still one more mutation the binding to long chain sugars can be abolished and the virus now becomes more like a bird virus, binding preferentially the the short chain versions found in birds.

Exactly what is the clinical significance of all this is unclear. The resulting mutated South Carolina virus is identical to one isolated in almost the same week in 1918 from a rapidly fatal influenza case in New York. Other experiments have shown that the binding may have only to do with transmissibility, not virulence, but clearly the New York patient got the virus from someone else. So at the very least, viruses with variable binding to the virus were circulating simultaneously and killing people simultaneously.

We are learning, slowly, about the components of transmissibility but not enough to know how to spot a genetic change that signals increased transmissibility. The kinds and locations of changes in H1 HA that switch the binding from bird to human are not the same as those for the H3 HA which in turn are not the same as for the H5 HA (see Pappas et al.). Thus we can't yet say that this work will tell us when a mutation in H5N1 indicates that it has become more transmissible. That signal, alas, will almost certainly be epidemiological, not genetic.

Even though I am an epidemiologist, I regret that.