The Role of Punctuated Evolution in the Pathogenicity of Influenza Viruses

ABSTRACT Influenza is an acute respiratory disease caused by influenza viruses. Evolutionarily, all influenza viruses are zoonoses, arising in the animal reservoir and spilling over into the human population. Several times a century, one of these zoonotic events results in a new influenza virus lineage becoming established in humans and circulating for years or decades as an endemic strain. The worldwide pandemic that occurs shortly after the nascent virus becomes established can have a profound impact on morbidity and mortality. Because influenza viruses continually evolve and the illness they engender can vary considerably based on characteristics of the strain, the weather, other circulating or endemic pathogens, as well as the number of susceptible hosts, the impact of each season on human health is unpredictable. Over time, the general pattern is for pandemic strains to adapt and gradually take on characteristics of seasonal strains with lower virulence and a diminished synergism with bacterial pathogens. Study of this punctuated evolution yields a number of insights into the overall pathogenicity of influenza viruses.


THE PANDEMIC THREAT
Pandemic influenza represents a recurring threat to human health.Several times a century, novel influenza A viruses cross over from the animal reservoirs of the world and establish new, dominant lineages in humans.These newly endemic strains may cocirculate with other established lineages or may replace them.Invariably, zoonotic pandemic influenza viruses are more pathogenic than the better-adapted seasonal strain present upon their emergence.This is partly due to differences in primary viral virulence and partly mediated by an enhanced ability to facilitate bacterial superinfections, such as pneumonia (1).Over time, however, pandemic strains adapt and gradually take on characteristics of seasonal strains with lower virulence and a diminished synergism with bacterial pathogens.Study of this punctuated evolution yields a number of insights into the overall pathogenicity of influenza viruses.

BIOLOGY AND ECOLOGY OF INFLUENZA VIRUSES
Influenza viruses are divided into three types, termed A, B, and C. Influenza B and C viruses diverged evolutionarily from influenza A viruses long ago and are stably adapted to humans.Influenza A viruses are zoonoses that are constantly crossing over from animal reservoirs into humans, causing individual infections and occasionally restricted epidemics (2).Several times a century, one of these zoonotic strains establishes endemicity and circulates for, typically, several decades before being replaced by a new incursion.Human influenza virus lineages do not fully diverge over this evolutionarily short period and frequently cross back over into animals such as swine (3).Although rare interspecies transmission events have been documented with influenza B viruses (e.g., in harbor seals) (4), influenza B and C viruses are predominantly restricted to humans.

Influenza A Virus Biology
Influenza viruses are segmented, negative-strand RNA viruses.Structurally, they take many shapes, from roughly spherical to filamentous to amorphous virions, with a rough size range of 80 to 120 nm.Influenza A viruses have eight gene segments, which code for eight structural proteins, two well-characterized nonstructural proteins, and, variably, several potential accessory proteins (Table 1).These accessory proteins are encoded from alternate open reading frames, alternate in-frame start codons, or early terminal truncation of some gene segments resulting in complex expression patterns and regulation.Although a good deal of work has been done on the accessory protein PB1-F2 ( 5), the breadth of potential function of other accessory proteins is not clear.
Hemagglutinin (HA) is the primary attachment protein.Binding cell surface sialic acids as a receptor allows internalization by endocytosis (Fig. 1).Matrix protein 2 (M2) protein then acidifies the interior of the virion, triggering HA-mediated fusion of the virus envelope with the endosome and releasing viral RNAs (vRNAs) and the proteins that make up the RNA polymerase complex (PB2, PB1, and PA) into the cytoplasm.These move to the nucleus, where transcription of gene segments utilizing a (+)-strand copy RNA intermediate takes place, together with translation to mRNAs.Nascent (−)-strand vRNAs are incorporated with nucleoprotein (NP) into viral ribonucleoproteins (vRNPs) and exported from the nucleus via the nuclear export protein nonstructural protein 2 (NS-2) and matrix protein 1 (M1).mRNAs leave the nucleus and act as templates for synthesis of accessory proteins, structural proteins, and glycoproteins.Structural proteins, primarily M1, congregate at the cell membrane, initiating assembly and preparing for budding.The glycoproteins, HA and the neuraminidase (NA), are shunted through the endoplasmic reticulum (ER) and Golgi network for folding and posttranslational modification, including addition of glycans, prior to shuttling to the assembly site, where they are inserted into the cell membrane.Nonstructural and accessory proteins are typically not incorporated into the virus.Budding of mature virions is facilitated by the sialidase activity of the NA, which clears cell surface sialic acids from the HAs on the nascent virion (Fig. 1).S-1 has multiple functions within the cell, primarily in countering the antiviral response (6).Full-length PB1-F2, when present, targets mitochondria, inducing cell death and triggering an inflammatory response (7).The functions of multiple other accessory proteins are only now being elucidated, but many appear to impact viral and cellular gene regulation (Table 1).

Influenza A Virus Ecology
Influenza A viruses are zoonotic viruses endemic to wild bird populations of the world, particularly migratory waterfowl, such as ducks and shorebirds.There is tremendous diversity in the viruses in these reservoirs, since each of the eight gene segments can be phylogenetically stratified into multiple distinct lineages, and reassortment of segments between different viruses creates an extremely large number of possible combinations.These viruses make constant incursions into other permissive species with which they have contact, most notably domestic poultry (e.g., chickens, turkeys, and ducks) and pigs (Fig. 2).Epidemics of novel viruses in domestic poultry are common, bringing these viruses into close contact with humans (8).Epidemics also occur in swine, with some strains establishing endemic lineages over decades, similar to the process that occurs in humans (3).Reassortment events within epidemics and between epidemic and endemic strains and viruses in other reservoirs are common, increasing the diversity of the extant strains.Viruses in domestic swine and poultry reservoirs frequently cross over to humans, causing individual infections or epidemics, some symptomatic and some inapparent.In rare cases, a virus is able to establish endemicity in humans and cause a pandemic with worldwide spread.

INFLUENZA A VIRUS EVOLUTION AND ADAPTATION IN HUMANS Pandemics
Influenza A viruses are subtyped by serologic characterization of the two surface glycoproteins, HA and NA.Historically, this nomenclature was adopted because HA and NA are the major antigens of the virus.There have been 18 HA subtypes and 10 NA subtypes identified thus far in nature.Four major lineages have caused pandemics and established themselves as endemic viruses in humans in the last century (Fig. 3).An H1N1 strain caused the most severe pandemic on record in 1918-1919, resulting in more than 50 million deaths.This lineage was maintained in humans until 1956, when it was replaced by a reassortant H2N2 strain with gene segments derived from both the circulating H1N1 strain which it replaced and an avian virus.The H2N2 strain was, in turn, replaced by an H3N2 virus, which again was a reassortant of the previously circulating strain.A version of the earlier H1N1 lineage reemerged, likely as a laboratory escape, in 1977 and cocirculated for several decades with the H3N2 viruses.This strain was replaced in 2009 with a novel swine-derived H1N1 strain that currently cocirculates with the H3N2 lineage (Fig. 3).

Adaptation
Variation in the influenza virus genomes results from a high mutation rate, which is an effect of having a low-fidelity polymerase and lacking a proofreading mechanism to correct base pair mismatches.The rapid cycle time and high mutation rate generate a large pool of quasispecies variants that interact together,   segments which are a better fit for the present evolutionary pressure.Influenza viruses tend to adapt very rapidly (on a scale of years) to positive selection pressure and more slowly (on a scale of decades) to lose or modify undesirable traits through negative selection.

Rapid Evolution
At least four common scenarios are known to drive evolution of influenza A viruses through positive selection.The most striking is establishing a lineage in a new host through a cross-species jump.In humans, this occurs immediately following the introduction of a new pandemic strain.The virus evolves rapidly for several years, then slows, and eventually (after decades) reaches an equilibrium such that in the absence of other sources of positive selection described below, most genes are not changing except through gradual negative selection for improved fitness in the host (9).
As an example, an analysis of swine lineages demonstrated that a recently introduced H3 lineage had a more rapid rate of evolutionary change in each gene than an H1 lineage which had been established for several decades (10).
The most commonly discussed selection pressure is from antibodies induced either by infection or by vaccine campaigns.This largely impacts HA and NA as the primary targets of the immune system.For this reason, HA and NA tend to have higher amino acid substitution rates than the other genes.These common changes are a combination of mutations around the receptor binding site of HA that alter antigenicity and bystander or compensatory structural changes.Interestingly, although changes in glycosylation status have been proposed as evolutionarily advantageous via shielding of antigenic sites from antibodies, one recent analysis suggests that glycosylation sites on HA are not under positive selection in the context of antigenic drift (11).
Antiviral drug use is another commonly encountered selective pressure.Rapid evolution of NA can be seen in sites related to NA activity following introduction of NA inhibitor antivirals into a population (12).Finally, reassortment is a common event, both within human virus populations and between human and animal strains.The presence of gene products from different influenza A virus lineages in a newly reassorted virus causes functional mismatches that drive selection of genes coding for variants of proteins that work better together.A classic example is the need for a balance of HA binding to sialic acids to enable entry and NA sialidase activity to facilitate budding.Reassortment-induced mismatches are corrected by evolutionary change (13).

IMPACT OF ADAPTATION ON PATHOGENICITY Hemagglutinin
A virus must satisfy three criteria in order to be considered the cause of a pandemic (14).It must have an antigenically novel HA, spread worldwide, and cause disease.The HA gene is therefore the one gene that must adapt after all pandemics, since a change in hosts is required to meet these criteria.Because of its central role in virulence and immunity, the HA of a nascent pandemic strain is under intense positive selection pressure.HAs which have recently emerged from the avian reservoirs typically have residual specificity for α-2,3-sialic acids, the predominant form in birds, while gaining specificity for α-2,6-sialic acids (15), which are the predominant receptor for influenza A viruses in the human respiratory tract.The binding affinity for α-2,3-sialic acids tends to weaken or disappear over time during adaptation.The HA must also adapt to fit any new gene products from segments that reassorted from a different strain, e.g., to maintain functional balance with a new NA.Several stabilizing mutations occur to facilitate growth in the new host, including changes to the pH of fusion specific to replication within the human lung (16).
Over a longer time scale, the HAs of strains which achieve endemicity in humans gain glycans on the head region of the protein (17).N-linked glycosylation is a common posttranslational modification of viral glycoproteins (18).Glycans are added during transit of the ER and Golgi apparatus at the glycosylation site motif Asn-X-Ser/Thr, where X may represent any amino acid except proline (19).A diverse repertoire of glycans can result from this process, and these modifications may have substantial effects on the biology of the proteins.Some of these glycosylation sites, primarily in the stalk region, are indispensable to the proper folding and conformation of the HA molecule (20).For many years, the prevailing dogma in the field has been that these glycans serve as an antigenic shield from antibody surveillance in an immune population (21).Although experimental data demonstrate that the presence of glycans on the head of the HA does alter antibody recognition (22), a lack of positive selection (11) suggests that this may not be the dominant effect driving addition of glycans during adaptation.Paradoxically, too much shielding may have harmful consequences to the hosts, as failure to neutralize virus in the setting of T-cell activation during reinfection may enhance immune pathology (22).
More recently, a role for glycosylation in modulation of innate immune surveillance has been recognized.

The Role of Punctuated Evolution in the Pathogenicity of Influenza Viruses
Influenza virus glycoproteins are flagged in the ER as "non-self" when they are poorly glycosylated, triggering the unfolded protein response and ER stress (23).Mediated by IRE1α, signal transduction downstream of the c-Jun N-terminal kinase drives inflammatory responses leading to acute lung injury.As influenza A viruses adapt and gain glycosylation, this innate sensing mechanism no longer recognizes the HA as foreign, the cascade is not activated, and considerably less lung pathology occurs during the infection (23).It is likely that selection of glycosylation variants is therefore an adaptation to avoid innate sensing in the ER, as well as the adaptive immune system in the form of antibodies that recognize epitopes on the head region of the HA.This escape comes at a cost, however, as heavily glycosylated viruses are more easily neutralized by collagenous lectins, such as surfactant protein D (17,21).

Neuraminidase
The role of NA in virulence is as a complement to HA, facilitating budding by cleaving sialic acids and disrupting sialic acid-HA binding.This requires maintenance of a balance between HA binding affinity and NA activity; viruses with too high NA activity relative to HA affinity have difficulty binding and establishing an infection, while viruses with too little relative NA activity have difficulty budding.Reassortant strains with mismatched HA-NA pairs undergo rapid adaptation, with the NA often evolving to match HA and regain this balance (13).The NA has a major role in secondary bacterial infections such as pneumonia (24).Sialic acid cleavage uncovers receptors for bacteria (25) and releases sialic acids into the extracellular medium, where they can be used as a carbon source, facilitating bacterial growth (26).The strength of the cleavage activity of the NA correlates broadly with the severity of epidemic influenza; circulating viruses with high NA activity map to some of the most severe seasons of the last century (27).It is likely that changes in NA activity either to match HA affinity changes during adaptation or in response to drugs (28) have a downstream impact on pathogenicity through change in support for secondary bacterial infections.Loss of NA activity as a compensatory mechanism to escape NA inhibitors in a population with frequent treatment (29) should diminish pathogenicity in this manner.

PB1-F2
PB1-F2 is a small, multifunctional accessory protein of influenza A viruses which is evolutionarily conserved in avian strains (30).It contributes to pathogenicity by triggering cell death through mitochondrial interactions and through support of secondary bacterial infections (7,31).Its role in the life cycle of the virus is dispensable in mammalian hosts, as nonfunctional forms are negatively selected over decades during adaptation in both humans and swine.Expressed from an alternate start codon in the +1 reading frame of the PB1 gene segment, PB1-F2 functions in endemic virus lineages are lost through truncation to shorter forms that lack the active sites in the C-terminal portion of the protein (typically to 11 or 56 amino acids, reduced from the full length of 87 to 90 amino acids [30]), or through loss of the start codon itself or mutation of the active site amino acids to a neutral or even antibacterial configuration (32).Strains lacking the cytotoxic functions supported by the C-terminal domain do not support secondary bacterial infections very efficiently, and in the human H3N2 lineage, the antibacterial effect appears to be a form of viral-bacterial warfare supporting the virus in its host niche (32,33).Overall, the clear pattern, as seen with changes in HA glycosylation, is evolution away from variants that cause inflammation and immunopathology (7).

Nonstructural Protein 1
NS-1 is another multifunctional nonstructural protein produced by influenza viruses.It performs a number of functions in the life cycle of influenza A viruses (6), chiefly as an antagonist of interferon and the antiviral response.Influenza A virus NS-1 contains three potential src homology binding domains, one SH2bm and two SH3bm, which are conserved in many avian lineages.Most avian strains, excluding H5N1 and H9N2 variants, share a consensus sequence from amino acids 212 to 217 defining an SH3(II)bm (34).This domain allows binding of NS-1 to c-Abl and subsequent inhibition of its functions in lung homeostasis.The end results of this blockade are acute lung injury and greatly increased pathogenicity of viruses that carry the motif (35).The 1918 pandemic strain expressed an avian NS-1 that contained this domain; however, the motif was mutated during adaptation of the H1N1 lineage to humans to a form that could not bind c-Abl and did not enhance acute lung injury (35).As the more pathogenic variant of NS-1 is conserved in many avian lineages, viruses that contain it represent a particularly dangerous pandemic threat.Another SH2bm which is able to bind to the p85 subunit of phosphatidylinositol 3-kinase (PI3K) and activate PI3K signaling with multiple downstream effects on immune function is conserved throughout avian and human influenza viruses (36).The reasons why this src homology domain is not lost during adaptation while the SH3(II)bm was selected against are not clear.However, these adaptive changes in NS-1 reinforce the pattern discussed above of selection for strains that do not activate the immune system or trigger inflammatory responses, as these are likely to be negative for both the virus and the host.Canonical functions important for virus survival that do not impact inflammatory responses such as inhibition of interferon are preserved.

CONCLUSIONS
The punctuated nature of zoonotic incursions into humans and subsequent establishment of endemicity provide a recurring cycle of high-and low-pathogenicity influenza A viruses.Most pandemic strains express pathogenic variants of multiple difference virulence factors and cause significant morbidity and mortality relative to those caused by seasonal strains (1).These strains adapt over time, consistently modifying or losing virulence factors so that seasonal strains are relatively less pathogenic upon the inevitable emergence of a new pandemic virus.A virus similar to the 1918 pandemic strains possessing multiple such factors, or indeed ones that have not yet been elucidated, could cross over from an animal reservoir at any time.The threat from the avian reservoir appears much more dangerous than that from swine, since adaptation of influenza A viruses in swine parallels that in humans in many ways, with frequent mutation and loss of these pathogenic gene variants.As an example, PB1-F2 of the H1N1 2009 pandemic strain was lost during evolution in swine prior to crossing over to humans, likely reducing the overall virulence of this virus and its ability to support secondary bacterial pneumonia.The underlying reasons for conservation of many of these virulence traits in avian species are unclear but are likely related to competition and survival in the differing host niche in birds, the gut.
Pandemic preparation and surveillance of influenza A viruses in animal reservoirs should focus on strains with these (and potentially other) molecular signatures of virulence, as strains carrying multiple such factors are likely to be the most pathogenic prospects for a severe pandemic.This will allow a more rational, targeted approach to vaccine and antiviral design against the worst-case scenario.In addition, multiple animal coronaviruses and parainfluenza viruses appear capable of making jumps into humans and likely use similar mechanisms to cause disease.Investigation of the virulence of these agents in the context of broad themes related to the pathogenicity of zoonotic agents should be a priority for the research community.

FIGURE 1
FIGURE 1 Influenza A virus life cycle in cells.

FIGURE 2
FIGURE 2 Influenza A virus ecology.The wild-bird reservoir is the source of all zoonotic influenza A viruses.These viruses cross over into humans through intermediate species, such as domestic poultry and swine.Farm animal silhouettes by Otutor, used under License CC BY 3.0 (https://creativecommons.org/licenses/by/3.0/us/);oyster catcher silhouette courtesy of Rachison Alexandra; human silhouettes courtesy of Mackey Creations.

FIGURE 3
FIGURE 3 Pandemic timeline.Four major lineages of influenza A virus have established endemicity in humans in the last century.The 1957 and 1968 pandemic viruses were reassortants which included genes from the previously circulating viruses which they replaced.The 1918 and 2009 pandemic strains came directly from animal reservoirs.The seasonal H1N1 lineage which circulated early in the 20th century was replaced in 1957 but reemerged in 1976 and cocirculated for 32 years with seasonal H3N2 strains.

TABLE 1
Influenza A virus gene functions