The role of swine as “mixing vessel” for interspecies transmission of the influenza A subtype H1N1: A simultaneous Bayesian inference of phylogeny and ancestral hosts

Abstract

Condensing the information of a total of 1551 to 469 influenza A H1N1 isolates we investigated the frequency of host shifts among bird, human and swine. Phylogenies of hemagglutinin and neuraminidase as well as ancestral host reconstructions were simultaneously inferred in a Bayesian framework. The surface proteins had to be analyzed separately because of reassortment. Also the different tree topologies indicated the different evolutionary histories of these genes. The majority of interspecies transmissions involved isolates from swine confirming the role of pigs as “mixing vessel” for the influenza A virus. This was emphasized by the investigation of host specific amino acid positions. However, the simultaneous estimation of phylogeny and ancestral states resulted in considerable ambiguity in particular at deeper nodes and at the root cautioning against overstated conclusions. Our analysis highlights the urge of intensifying influenza surveillance programs for porcine hosts.

Introduction

Influenza is a seasonally occurring respiratory disease in humans. Worldwide, there are between three and five million infections per year, 10% of which end fatally. The viral infection caused by the influenza virus is classified into the three types A, B, and C, with A being mainly responsible for human infections. Influenza A is further classified into subtypes referring to the combination of the two surface proteins hemagglutinin (HA) and neuraminidase (NA) (WHO, 2009).

The natural host for all influenza A subtypes are wild birds. Despite the high number of possible combinations (15 HA types × 9 NA types = 135) only H1, H2 and H3 as well as N1 and N2 have successfully established in humans (Baigent and McCauley, 2003). Most relevant for seasonal outbreaks are the subtypes H1N1 and H3N2, in temperate zones preferentially in autumn and winter (Rambaut et al., 2008, WHO, 2009).

Besides birds and humans the virus can also occur in other mammals like swine, horse, or whales, but for physiological reasons interspecies transmissions are rare. However, successful transmission between two species can rapidly lead to a dispersion of the virus of pandemic dimension due to the new host's lack of antibodies (Baigent and McCauley, 2003). The first recorded and since then most devastating pandemic was the “Spanish Flu” in 1918 with more than 40 million deaths. Induced by the same subtype – H1N1 – that circulated until the fifties, the “Russian Flu” of 1977 still has an unexplained origin (Taubenberger and Morens, 2006).

The genome of the influenza virus contains eight pieces of segmented negative-sensed RNA with 13,600 bases all together that code for 10 viral proteins. The trimeric complex of hemagglutinin is encoded on the fourth segment and is responsible for virus binding and for adsorption to the host cell (Skehel and Wiley, 2000). The receptor binding site is host specific due to its configuration at certain codon positions (Baigent and McCauley, 2003). Binding takes place with sialic acid-α-2,3-galactose or sialic acid-α-2,6-galactose (Baigent and McCauley, 2003, Shen et al., 2009). As α-2,3 bound receptors are mainly found in birds, and α-2,6 bound receptors appear in the trachea of humans, a direct transmission is rarely successful but possible – as human cases of avian flu H5N1 demonstrate – because of the existence of avian specific receptors in the lower respiratory system. Swine has both receptor types and has therefore been called the “mixing vessel” for the influenza virus because it can function as an intermediate host (Castrucci et al., 1993, Scholtissek, 1994, Kida et al., 1994, Ma et al., 2008).

Neuraminidase, which is also indispensable for virus transmission, is encoded on segment six. This homotetrameric complex is functionally tightly linked with hemagglutinin and responsible for the virus release from the host cell, eventually leading to the infection of new cells. The release is again facilitated via host specific codon positions (Tamuri et al., 2009).

Phylogenetic reconstructions have been important for our understanding of virus evolution and epidemic pathways including geographical spread and host shifts (Rambaut et al., 2008, Tamuri et al., 2009, Lemey et al., 2009). Methodologically, tree reconstructions and reconstruction of ancestral states, such as geographical locations, hosts or codon positions (Wilson et al., 1991, Ronquist, 1994, Akashi et al., 2007) have been conducted in a parsimony framework. However, parsimony reconstructions do not account for statistical uncertainty both in the reconstruction of the phylogeny as well as the inference of ancestral states (Cunningham et al., 1998, Gubareva et al., 2002, Lemey et al., 2009, Haase et al., 2010, Hovmoller et al., 2010). In addition parsimony ignores branch lengths.

Probabilistic approaches have only recently been applied in the study of viral evolution. Haase et al. (2010), for example, reconstructed host shifts and ancestral ranges of H5N1 in a likelihood framework, but still using a single “optimal” tree. A fully probabilistic Bayesian approach allowing to express mapping and phylogenetic uncertainty simultaneously has been developed by Lemey et al. (2009). Most importantly probabilistic methods avoid overconfidence in seemingly unambiguous inferences usually resulting from maximum parsimony analysis.

Here we investigate the interspecies transmission of the H1N1 influenza A virus between the three hosts human, bird and swine based on all hemagglutinin and neuraminidase sequences available from the NCBI Flu-database as of October 31, 2009. We incorporated an ancestor state reconstruction into a Bayesian framework to study host shifts and additionally examined the human and avian specific amino acid positions in porcine sequences. Central to our analyses was the question about the role of swine as “mixing vessel” for H1N1.