N-Glycans from Porcine Trachea and Lung: Predominant NeuAcα2-6Gal Could Be a Selective Pressure for Influenza Variants in Favor of Human-Type Receptor

It is known that pigs acted as “mixing vessels” for genesis of a new reassortant influenza strain responsible for pandemic H1N1 2009. However, the host factors driving the evolution of a reassorted virus in pigs to ‘jump species’ resulting in a human outbreak remain unclear. N-glycans derived from the porcine respiratory tract were enzymatically released, fluorescent labeled with 2-aminopyridine, separated according to charge, size and hydrophobicity, and structurally identified by a two-dimensional (size and hydrophobicity) HPLC mapping technique and MALDI-TOF mass spectrometry before and after exo-glycosidase digestion. We found a 3-, 5-, and 13-fold increases in NeuAcα2-6, a preferable human influenza receptor, over NeuAcα2-3, an avian influenza receptor, from upper and lower parts of the porcine trachea towards the porcine lung, a major target organ for swine virus replication. The large proportion of NeuAcα2-6 may exert selective pressure for selection of influenza variants with altered receptor preference for this human-type α2-6 receptor, a crucial first step for generating a human pandemic.


Introduction
Influenza A viruses cause local epidemics every year and occasionally cause worldwide pandemics, which have been considered to be major public health threats. Viruses in pandemic outbreaks acquire mutations that evade human immunity and efficiently transmit from human to human. Viruses carrying a hemagglutinin (HA) surface glycoprotein to which humans are immunologically naive can be derived from an avian virus or an avian-human reassortant virus [1]. Although there have been many reports on direct transfer of an avian virus, especially highly pathogenic H5N1, to humans [2,3], transmission between humans of avian viruses has been limited, inefficient and unsustained [4]. A number of factors, including the RNA polymerase PB2 subunit and HA activation proteases, may be involved in host range restriction and pathogenicity of influenza viruses [1,5]; however, HA plays a key role in the initial stage of infection and thus functions as a critical host range determinant [6]. HA recognizes host glycans with terminal sialic acids (Sia), which vary in structure among species and their tissues. Avian HA prefers host glycans with Siaa2-3Gal linkage, mainly found on the bird intestine and respiratory epithelia, whereas human HA prefers those with the Siaa2-6Gal linkage, which populate the human upper respiratory epithelia [7], a major site of its infection in humans [8]. To overcome this interspecies barrier, avian HA must switch its binding preference to Siaa2-6Gal. Three historical influenza pandemics, Spanish H1N1 in 1918, Asian H2N2 in 1957 and Hong Kong H3N2 in 1968 as well as the present pandemic H1N1 2009 have HAs with preferential recognition of Siaa2-6Gal, even though their HAs are of non-human origin [9]. Although it remains a mystery whether the 1918 pandemic virus was transmitted directly from avians or was the result of reassortment before the pandemic [10,11], the 1957, 1968 and 2009 pandemic viruses were reassortant viruses [12], and the 2009 pandemic was confirmed to be of swine origin and was hence named swine-origin H1N1 influenza viruses (S-OIVs) [12]. Although S-OIV causes mild disease, it has spread worldwide to more than 214 countries and has caused over 18,449 laboratory-confirmed deaths as of August 1, 2010 [13] due to relatively efficient transmission among humans.
Swines susceptible to both avian and human influenza viruses and possessing a trachea that expresses both Siaa2-3Gal (avian receptor) and Siaa2-6Gal (human receptor) [14] provide direct evidence supporting the theory of pigs as ''mixing vessels'' for the creation of reassortant viruses. However, receptor binding specificity data have shown that classical swines preferentially recognized NeuAca2-6Gal [14,15,16,17] and that avian-like swine switched their binding preference to NeuAca2-6Gal over time [14,15,16]. Analysis of amino acids has suggested that HA mutations are responsible for the increase in affinity of the virus for NeuAca2-6Gal [14,15,16]. However, why influenza viruses that have continued replication in pigs evolve HA receptor specificity to NeuAca2-6Gal is still puzzling. Types, structures and distribution of glycans on the host cell surface are thought to be associated with viral HA receptor specificity. Thus, we elucidated the structures of N-glycans required for influenza A infection [18] from the porcine trachea, an early site of influenza virus attack and replication, and porcine lung, a principal swine influenza replication site [19], by HPLC and matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF-MS) analyses.

Ethics statement
Animal experiments were carried out according to the regulations and guidelines approved by the Animal Ethics Committee of the National Institute of Animal Health, Japan. The approved number is 08-058.

Porcine trachea and lungs
Porcine trachea and lungs were collected from a clinically healthy 5-year-old female LWD-pig, an offspring of a male Duroc (D) and a female F1 hybrid pig (a female Landrace (L) and a male Large White (W)).

Lectin detection of Siaa2-3Gal and Siaa2-6Gal receptors
The porcine trachea, which was divided into upper and lower parts, and the lungs were rinsed with ice-cold phosphatebuffered saline (PBS, pH 7.2) and then cut into 3-mm 3 cubes. Tissue blocks were fixed in 4% paraformaldehyde (PFA) in PBS overnight at 4uC and equilibrated in 30% sucrose in PBS overnight at 4uC. The blocks were embedded in OCT compound (Sakura Finetechnical Co. Ltd., Japan) and frozen at 280uC for cryosectioning. Cryostat sections (10 mm each) were air-dried and re-hydrated twice in PBS containing 0.1% Tween-20 (PBST) for 5 min each time at room temperature. Immunostaining was begun by adding 50 ml of a lectin mixture of biotin-conjugated Sambucus niga agglutinin (SNA; specific for NeuAca2-6Gal; 5 mg/mL; Vector Lab, Burlingame, CA) and digoxigenin (DIG)-conjugated Maackia amurensis agglutinin (MAA; specific for NeuAca2-3Gal; 10 mg/mL; Roche, Mannheim, Germany) to PBS containing 0.05% Tween-20 and 0.5% bovine serum albumin (BSA). After incubation for 30 min at 37uC, the sections were washed three times with PBST for 5 min each time. The sections were incubated with 50 ml of a solution of fluorescein-conjugated avidin (Zymed, USA) and rhodamine-conjugated anti-DIG antibodies (Roche) at a dilution of 1:100 for 30 min at 37uC. After three washes with PBST, stained sections were mounted on slides in buffered glycerol (pH 9.0) and examined using a fluorescence microscope (Olympus B1, Tokyo, Japan). Negative controls were done in the absence of lectin.

Preparation of porcine trachea and lungs for N-glycan analysis
The porcine tracheal tube was cut transversely into upper (14.14 g) and lower (10.48 g) parts. Each part was opened longitudinally, and its luminal surface was cleaned with cold 0.9% NaCl, removed by scraping, and cut into small pieces. The pieces were pooled, lyophilized (400 mg for the upper part and 560 mg for the lower part), and kept at -80uC until use.
The porcine lung (6,225 mg) was cleaned and cut into small pieces and then homogenized using a T10 Basic Ultra-Turrax homogenizer (IKA-Werke, Germany) followed by a glass-Teflon homogenizer (Eyela, Tokyo, Japan) in a cold buffer containing 20 mM Tris-HCl (pH 7.6), 150 mM NaCl, 0.3 M sucrose and protease inhibitor cocktail set 1 (Calbiochem, Sandiego, CA). The homogenate was then centrifuged at 1000 g for 10 min and the post-nuclear supernatant was ultracentrifuged at 100,000 g for 1 h (Beckman, SW28 rotor). The pellet (membrane fraction) was resuspended in 0.9% NaCl and recentrifuged. The washed pellet was dried by a lyophilizer (125 mg).

Isolation and identification of N-glycan structures
The lyophilized porcine trachea (upper and lower parts) and lungs were delipidated by extraction solvents, 80% ethanol (three times), 100% ethanol, chloroform/methanol (2:1,v/v) and chloroform/methanol/H 2 O (1:2:0.8, v/v/v), in that order and used as the starting material for structural analysis. The conditions for enzymatic release, fluorescent labeling, separation and structural identification of N-glycans were the same as those reported previously [20,21,22]. In brief, the delipidated extract (20 mg) was proteolyzed with pepsin plus glycoamidase A and the resultant peptides were further digested with pronase to amino acids. The reducing ends of the released N-glycans after purification by Bio-Gel P-2 chromatography were labeled with fluorescent 2-aminopyridine. After removal of excess reagents by Sephadex G-15 chromatography, the pyrimidylamino (PA)glycans were separated by a TSK-gel diethylaminoethyl (DEAE)-5PW column (Tosoh, Tokyo) based on charge. Each separated fraction was further applied to a Shim-pack HRCoctadecyl silica (ODS) column (Shimadzu, Kyoto) for separation on the basis of hydrophobicity, and the elution time of each peak was expressed as glucose unit (GU) value. The molecular mass of each PA-glycan fraction was analyzed by MALDI-TOF-MS. Fractions containing two kinds of glycans were further subfractionated according to molecular sized by a TSK-gel Amide-80 column (Tosoh) and the elution positions were calibrated in GU values. Sample PA-glycans were mapped on the basis of on their GU and molecular mass values and their coordinates were compared with those of reference PA-glycans in the GALAXY database [23]. Sample PA-glycans showing no agreement with any PA-glycans in the database were sequentially trimmed with exoglycosidases: a-galactosidase (coffee bean, Oxford Gly-coScience, Bedford, MA); b-galactosidase (jack bean, Seikagaku Kogyo, Tokyo, Japan); a-sialidase (Arthrobacter ureafaciens, Nacalai Tesque, Kyoto, Japan); and a2,3-sialidase (Salmonella typhimurium LT2, Takara Bio Inc, Otsu, Japan) under conditions described previously [21,22,24]. Each trimmed PA-glycan at each step was analyzed until its elution and mass coordinates became identical with the known reference on the map. The structures were confirmed by co-chromatography of the sample and reference PA-glycans.

Distribution of Siaa2-3Gal and Siaa2-6Gal receptors in the porcine trachea (upper and lower parts) and lungs
Double-labeling of sections of the porcine trachea and lungs ( Figure 1) revealed that both MAA lectin, specific for Siaa2-3Gal linkages, and SNA lectin, specific for Siaa2-6Gal linkages, were stained in both the upper and lower parts of the trachea and in the lung with SNA binding being more dominant than MAA binding. On the ciliated pseudostratified columnar epithelium of the upper and lower parts of the trachea, MAA reacted predominantly with ciliated cells, whereas SNA also reacted with goblet cells. Both bronchiole and alveoli were stained by SNA and MAA. However, SNA was observed as a fine line at the bronchial epithelial border, whereas MAA was not.

Discussion
Our results for double lectin immunostaining agree with previous observations that there is dual expression of Siaa2-6Gal (human-type) and Siaa2-3Gal (avian-type) receptors on the surface of porcine tracheal and lung epithelia, while Siaa2-6Gal is dominant [14,25,26]. Both human and avian influenza viruses have been isolated from pig populations, indicating that transmission of whole human or avian viruses to swines has occurred in nature [27]. Although SNA and MAA lectins are useful for detection of the localized distribution of a2-6 and a2-3 linkages, respectively, between Sia and Gal, influenza viruses can distinguish not only glycosidic linkages but also types of Sias, which influence the influenza virus host range [25,28]. HPLC with MALDI-TOF-MS techniques have been continuously developed over the past years for determination of precise N-glycan structures and quantitative analysis of those N-glycans; influenza virus was reported to bind to but unable to be internalized into cells deficient in N-linked glycan, in spite of no deficiency in O-linked glycan, suggesting the requirement of N-linked glycoprotein for successful endocytosis of influenza virus into its host [18]. By using these techniques, we found that N-glycans of the trachea and lungs from a 5-year-old pig were composed of 38 kinds, of which 17, 6, 5 and 10 (tracheal upper and lower parts) and 19, 9, 5 and 5 (lung) were neutral, monosialyl-NeuAc, monosialyl-NeuGc and disialyl, respectively, at neutral/acidic molar ratios of 64:36 (upper trachea), 62:38 (lower trachea) and 45:55 (lung). A few neutral glycans are regarded as virus receptors; however, a triple reassortant H1N1 human isolate, A/Iowa/1/06, has recently been shown to be able to bind to a complex-type N-glycan with terminal LacNAc [16]. These neutral N-glycans were found in the porcine tracheal upper and lower parts and the porcine lungs with molar ratios of 8.9, 7.4 and 4.1, respectively. In contrast to neutral glycans, Sia with a negative charge is a major receptor and a host range determinant of influenza viruses [25]. Two major entities that influence infectivity of influenza viruses are type and linkage of Sia [1,29]. Two prevalent Sias (NeuAc and NeuGc) found in mammalian cells were identified, but NeuAc was the predominant form in  [28] and NeuGc accounting for more than 90% of Sia in epithelial cells of the horse trachea [25], whereas normal human tissues possess only NeuAc as they have a non-functional hydroxylase to produce NeuGc [30]. The absence of NeuGc may protect humans from infection with some pathogens, such as enterotoxigenic Escherichia coli K99 [30], but not from influenza viruses. Overall, both avian and human influenza A viruses appear to exhibit preference for NeuAc rather than NeuGc glycoconjugates. The third notuncommon Sia, 9-O-Ac-NeuAc, a primary receptor determinant of influenza C virus for infection of host cells [31], was not detected, indicating that this N-glycan either is not synthesized in the porcine trachea and lung or is present in other tissue. This is in agreement with the fact that influenza C viruses cause mild infection in the upper respiratory tract [32], although the presence of 9-O-Ac-NeuAc in the porcine trachea and lungs should not rule out the possibility that O-linked glycoproteins or glycolipids carry 9-O-Ac-NeuAc.
The type of glycosidic linkage between Sia and Gal on the host cell surface is clearly the principal determinant of the host range restriction of influenza viruses: a2-3-linkage is avian virus preference, while a2-6-linkage is human virus preference. We found that there are gradually increased molar ratios of a2-6linked sialyl glycans compared to those of a2-3-linked sialyl glycans, 3.2-, 4.9-and 13.2-fold for NeuAc and 1.8-, 2.7-and 5.9fold for NeuGc in the upper trachea, lower trachea and lungs of the pig, respectively. Our data can explain why influenza viruses, both avian-like [14,15], classical swine and triple reassortant swine influenza viruses [9], replicated in pigs have changed in their receptor binding preference to human a2-6 receptor and have occasionally been isolated from humans. Our data also provide an explanation of why avian influenza virus before genetic change produced lower virus titers with limitation of spread throughout the porcine respiratory tract in a comparison experiment with a swine virus [33]. Thus, pigs not only serve as 'mixing vessels' facilitating reassortment of viral gene segments to produce new influenza strains but also possess abundant proportions of humantype glycan receptors, providing a selective pressure to select/ evolve the virus with a receptor preference for this receptor. A selective pressure on the receptor specificity of the viral HA to the human host seems to be a prerequisite for the generation of a virus with pandemic potential in accordance with historical data that most infections by swine viruses cause only limited human-tohuman transmission [9]. To become established in a human population, additional factors are likely to be required for optimization of its host-cell tropism. By an integrated biochemical, analytical and data mining approach, it has recently been shown that long a2-6 sialylated glycans with umbrella-like topology (glycans containing multiple repeating lactosamine units, such as NeuAc(a2-6)[Gal(b1-4)GlcNAc (b1-3)]n) may be required for sufficient viral transmission between humans: human-adapted HAs bind with high affinity to umbrella-like topology (long a2-6 glycans), whereas avian and swine HAs preferentially recognize cone-like topology (a2-3 or short a2-6) [29,34]. The N-glycan profile of human bronchial epithelial cells showed the presence of a2-6 long branches [29]. Previous N-glycans derived from amniotic membrane (AM) cells of chicken embryonated eggs used for growing viruses isolated from human hosts consist of short a2-6 trisaccharide branches, NeuAc(a2-6)Gal(b1-4)GlcNAc-, but not a long lactosamine structure [35]. Similar to N-glycans isolated from AM cells, N-glycans with a long branch could not be detected in the porcine trachea and porcine lungs; nevertheless, the possibility that long branches may be present on O-linked a2-6 cannot be excluded. However, the recent emergence of S-OIVs with swift human-to-human transmission has confirmed that pigs are the source of the generation of influenza viruses with pandemic potential. . Structures and relative quantities of neutral, mono-and di-sialyl PA-oligosaccharides derived from the porcine upper trachea, lower trachea and lungs of a pig. a, mol % was calculated from the peak area in Figure 3 by comparison with total N-glycan content in each porcine tissue. b, Structures of PA-oligosaccharides are represented by symbols as follows: red diamond, NeuAca2-6; purple diamond, NeuAca2-3; green diamond, NeuGca2-6; light blue diamond, NeuGca2-3; yellow circle with a, a-galactose(Gal); yellow circle, galactose; blue square, Nacetylglucosamine (GlcNAc); green circle, mannose (Man), red triangle, fucose (Fuc). c, Units of glucose (GU) were calculated from the elution times of the peaks obtained from the ODS column in Figure 3. doi:10.1371/journal.pone.0016302.g004 Other glycan modifications, such as fucosylation and sulfation, may be involved in the receptor binding activity of viral HAs [17,29,36,37]. Fucosylation and/or sulfation at Gal or GlcNAc/ GalNAc on position 2 or 3, respectively, of the terminal trisaccharide (Sia-Gal-GlcNAc/GalNAc-) appear to affect receptor binding activity of some influenza viruses, such as increase in binding affinity of H5N1 and H7N1 chicken viruses to a2-3-linked Sia [37]. Fucose residue detected in N-glycans of the porcine trachea and lungs was found only at the initial GlcNAc of the N-glycan core connected to asparagine. Sulfate residue was not detected in porcine trachea and lung N-glycans.
It should be noted that glycoconjugate profiles of the porcine respiratory tract, which are involved in mediation and regulation of many physiological and pathological processes, may vary among different species, ages, sex and lifestyle; however, the finding in this study that Siaa2-6Gals are dominant along the respiratory epithelial tract of a 5-year-old female LWD-pig is in agreement with recently reported lectin-binding profiles of 4-8week-old healthy post-weaned male United Kingdom pigs [26]. All age groups of pigs can be infected by influenza A viruses. During a farrow-to-finish operation, whereas growing-finishing pigs are replaced almost every 6 months for their flesh, mother pigs (sows) remain in a farm for breeding. Influenza infection in a sow herd can cause abortion and can be transmitted to a vulnerable population including their piglets. The changing nature of influenza virus could allow re-infection of sows and, thus, sows could experience several seasons of influenza during their lives and are considered to be involved in the mechanism for maintenance of swine influenza viruses in a farm [38]. Immune evasion from naïve sows results in mutant influenza viruses with antigenic changes and introduction of influenza viruses from other species into this reservoir displaying NeuAca2-6 predominance may result in generation of new strains that are efficiently transmitted to humans.
In summary, the most striking finding in the present study was the presence of a greater abundance of N-glycans carrying a2-6 over a2-3 linkage type, especially di-NeuAc(a2-6) bi-antennary complex type, at the sites of influenza virus replication. This finding explains why influenza viruses that have continuously circulated in pigs displayed an increased affinity for human a2-6 sialylated receptor and indicate the necessity to enhance global surveillance for the emergence of a new variant influenza virus in pigs with the ability to bind to a human-type receptor and the need to establish preparation plans for the next pandemic, not only for  Table 1. Amounts of a2,3and a2,6-sialyl receptors in comparison with those of the other outer residues present on N-glycans derived from the porcine trachea (upper and lower) and lungs of a pig. H1 and H3 viruses, circulating in swine populations, but also H5 and H7 avian influenza viruses, especially highly pathogenic H5N1 virus experimentally able to infect pigs [39].