Why are there so many flu viruses?

The only thing constant in flu epidemiology is that it is always changing. New subtypes appear, old ones retreat; like a play where actors constantly change masks and costumes. Names are put forward in the press, such as the Mexican flu, which changed to swine flu, which changed to the new flu A/H1N1 (but, of course, the swine flu label stuck). The current evildoers in humans are H1N1 and H3N2. These are seasonal flu viruses, meaning that they circulate predominantly in humans, and only occasionally give infections in other animals. Both of them made the leap from another animal reservoir before becoming human flu viruses, and both, in turn, have once been avian influenza viruses.

Most readers will also remember the ‘bird flu’ virus H5N1. First of all: it still exists, endemic in parts of Asia, and in Egypt. It hasn’t left the scene. This virus is a highly-pathogenic avian influenza virus that cause rare, but often fatal infections in humans. The highly-pathogenic prefix means that it is an efficient poultry killer – with close to 100% mortality in infected chicken flocks. That’s like tossing in a mini nuke, closing the barn door and wait for the explosion. A mean virus, for a chicken.

However, the norm among avian influenza viruses is to be low-pathogenic, only causing mild infections in their hosts. For domestic poultry that equals a mild cold, in wild ducks even less so. Recently, yet another flu actor entered the scene: H7N9. This virus has caused a number of human infections and deaths in China, but contrary to H5N1 has not been associated with die-off of domestic poultry. New costume, new play, but still a deadly mix.

So, there is H1N1, H3N2, H5N1 and H7N9 out there – all with the capacity of infecting humans. Earlier flu pandemics have been caused by yet other viruses, and from studies of poultry workers and veterinarians we know that there are viruses with other H and N letters that can infected humans, but without leading to severe symptoms. Even if the list seems long, it is nothing compared to the total diversity of influenza A viruses. The H and the N are shorts for the two surface proteins hemagglutinin (responsible for attachment to cells, and to invasion) and neuraminidase (responsible for letting new virus progeny leave an infected cell). There are 16 H variants, and 9 N variants and as they are encoded on different genome segments, they can end up in any of 144 possible combinations, or subtypes. More than 100 of these subtypes have been found in ducks, and more than 70 of them have been found in our study population of Mallards at Ottenby, in SE Sweden. Thus, there are many, many more flu viruses out there lurking in the shadows.

But why are there so many viruses? And especially, why so many in Mallards?

In study published last week in PLOS Pathogens, we returned to this question and analyzed infection histories of more than 7,000 Mallards sampled at Ottenby during 8 years! Together these ducks were caught and sampled more than 18,000 times! The repeated capture and recaptures of ducks is a major benefit of our trapping scheme, as it allows us to follow the course of natural infections in different birds. This is a gospel I have been singing in two previous posts on individuals and reassortment, and a topic I am likely to return to. Predictable fellow, yes, yes. But let’s turn back to the subject.

What did we do? Well, we analyzed all cases where we had at least two characterized virus isolates from the same bird in the same season. Then we used this data to investigate how frequent reinfection with a particular subtype was given the first detected subtype and how this depended on time. This sounds rather simple, doesn’t it? In truth it was a rather large statistical undertaking, as the 25 supplementary files tells. The devil is in the detail – in this case in dealing with potential pseudo replication and test assumptions. Anyway, we leave the finer details of the stats for now and instead take a look at the table below. It is a contingency table, where rows and columns relate to H subtype at first and second infection, respectively. This means that the diagonal shows cases where the same subtype was isolated at both occasions. The colors highlight combinations that were either overrepresented (blue), or had a deficiency of cases (red) compared to the expected. The first thing to note is the diagonal, where very few cases of reinfections were noted. In other words, a bird infected with, let’s say, an H4 virus, will have a low probability of being infected with the same subtype again the same autumn. This is called homosubtypic immunity, and not different from what we want to achieve with vaccination in humans. Once you have had it you are immune (at least for some time…).

However, we also found a great degree of heterosubtypic immunity, meaning that an infection with one H subtype made reinfections with other related subtypes less frequent than expected. If you check the figure again, you can see that there are patterns to these cases of heterosubtypic immunity. In fact, they follow higher order clustering of hemagglutinin gene relationships, as can be seen in the next figure. H1, H2, H5 and H6 viruses belong to the H1 Clade, and a primary infection with any of these will make it less likely to be reinfected with other viruses of the same clade. The pattern was similar for other clades and was actually also detectable at the H Group level (the highest level of structure).

But what does this mean?

It is actually big business. It gives a very strong case for existing selection pressures for hemagglutinin gene diversification. Subtypes as a term predates the genomic area and is based on immune reactions. Typically, a subtype is defined as a group of viruses recognized by the same antisera (antibodies towards a particular virus). Subtypes are well resolved for hemagglutinin and neuraminidase, both in phylogenetic relationships and in responses to antisera. Things match. You would be tempted to think that virus subtypes have diversified until their antigenic properties are different enough for the immune system of the host animal to be unable to treat them with the same set of weapons. For instance, antibodies to an H1 virus shouldn’t interfere with an H2 virus infection.

Here, we show that heterosubtypic immunity is strong for hemagglutinin (but absent for neuraminidase), and that it follows genetic relationships. This means that there is ongoing warfare among hemagglutinin subtypes. If an individual is infected with one subtype, it then becomes harder for other related subtypes to enter and cause reinfections. The strength of this response, and its longevity, will be extremely important for infection dynamics at the population scale and drive which viruses that peak at different times. This is especially interesting in a migratory species like the Mallard, where viruses need to follow their hosts, not in only time, but also in geography. And it means that H subtypes are still diverging. The pace of this divergence would be very interesting to tackle, but will require good time series of influenza genomes (rest assured, we are sequencing like crazy and will return to this subject).

To conclude, our study provides evidence from the field on how natural selection in influenza A virus is driven by host immune processes and that it is evident for the most antigenic protein. The question ‘why’ is therefore dependent on disruptive selection. It also raises a bundle of additional questions. Is the diversification we see in influenza A virus the result of geographic allopatric processes, or through separation in different host species, or is there sympatric diversification going on?

More to do, more to do. This virus will keep us busy for sure.

Jonas Waldenström

Link to the article:

http://www.plospathogens.org/article/info%3Adoi%2F10.1371%2Fjournal.ppat.1003443