Changes in H5N1 Influenza Virus Hemagglutinin Receptor Binding Domain Affect Systemic Spread

Changes in H5N1 influenza virus hemagglutinin receptor binding domain affect systemic spread

Hui-Ling Yena, Jerry R. Aldridgea, Adrianus C. M. Boona, Natalia A. Ilyushinaa, Rachelle Salomona, Diane J. Hulse-Posta, Henju Marjukia, John Franksa, David A. Boltza, Dorothy Bushb, Aleksandr S. Lipatova, Richard J. Webbya, Jerold E. Rehgb, and Robert G. Webstera,1

aDivision of Virology, Department of Infectious Diseases and bDepartment of Pathology, St. Jude Children’s Research Hospital, Memphis, TN 38105

Contributed by Robert G. Webster, November 6, 2008 (sent for review August 14, 2008) The HA of influenza virus is a receptor-binding and fusion protein that specimens outside the respiratory tract, including blood, cere- is required to initiate infection. The HA receptor-binding domain brospinal fluid, intestine, feces, placenta, and a fetus (16–20). determines the species of sialyl receptors recognized by influenza The mechanism of systemic spread of H5N1 viruses in mammals viruses. Here, we demonstrate that changes in the HA receptor- has not been established, although detection of virus in blood binding domain alter the ability of the H5N1 virus to spread system- (18, 21) or nerves (22) suggests possible routes. ically in mice. The A/Vietnam/1203/04 (VN1203) and A/Hong Kong/ We and others have found that A/Vietnam/1203/04 (VN1203) 213/03 (HK213) viruses are consistently lethal to domestic chickens virus, whose HA recognizes synthetic ␣2,3-linked sialyl receptor, is but differ in their pathogenicity to mammals. Insertion of the VN1203 highly pathogenic in the ferret and mouse models, whereas A/Hong HA and neuraminidase (NA) genes into recombinant HK213 virus Kong/213/03 (HK213) virus, whose HA recognizes both ␣2,3- expanded its tissue tropism and increased its lethality in mice; con- linked and ␣2,6-linked synthetic sialyl receptors, is less lethal (8, 9, versely, insertion of HK213 HA and NA genes into recombinant 23, 24). Surface glycoproteins have also been reported to play a role VN1203 virus decreased its systemic spread and lethality. The VN1203 in the pathogenicity of H5N1 viruses isolated in Hong Kong in 1997 and HK213 HAs differ by 10 aa, and HK213 HA has shown greater (11, 25), H7N7 virus isolated in the Netherlands (26), and the H1N1 binding affinity for synthetic ␣2,6-linked sialyl receptor. Introduction 1918 pandemic virus (27). We hypothesized that changes in the HA of an S227N change and removal of N-linked glycosylation at residue receptor binding domain might modulate the virulence of H5N1 158 increased the ␣2,6-binding affinity of VN1203 HA. Recombinant viruses in mammals by affecting the spread of virus from the VN1203 virus carrying the S227N change alone or with the residue- respiratory tract to other tissues. 158 glycosylation site removed showed reduced lethality and systemic spread in mice but not in domestic chickens. Wild-type VN1203 Results virus exhibited the greatest efficiency in systemic spread after intra- VN1203 HA and Neuraminidase Increase the Virulence of HK213 in muscular inoculation and in infection of mouse bone marrow-derived Mice. The VN1203 virus spreads systemically in inoculated mice and dendritic cells and conventional pulmonary dendritic cells. These ferrets whereas the HK213 virus is detected largely in their respi- results show that VN1203 HA glycoprotein confers pathogenicity by ratory tracts (8, 9, 23, 24). To determine whether the HA and facilitating systemic spread in mice; they also suggest that a minor neuraminidase (NA) of VN1203 contribute to its systemic spread, change in receptor binding domain may modulate the virulence of we first investigated the tissue tropism of HK213 virus carrying H5N1 viruses. VN1203 HA or VN1203 HA and NA. HK213, HK213xVN1203HA, and HK213xVN1203HANA recombinant viruses grew to compara- dendritic cell ͉ H5N1 ͉ pathogenesis ͉ mice bletitersinvitro(Table S1), and one 50% mouse lethal dose (MLD50) in BALB/cJ mice was 4,295, 413.3, and 41.8 pfu, respec- he HA glycoprotein of influenza A virus is a type I transmem- tively. Therefore, the lethality of HK213 virus was increased one log Tbrane protein that binds to sialic acid residues on target-cell by inserting VN1203 virus HA and two logs by inserting VN1203 glycoproteins or glycolipids (1). The receptor-binding domain is HA and NA. Furthermore, Ϸ40% of mice inoculated with located at the HA membrane distal end, where the 130-loop, HK213xVN1203HA or HK213xVN1203HANA virus, but none of the 220-loop, 190-helix, and other conserved residues (Y98, W153, mice inoculated with HK213, developed neurological signs (hind H183, Y195) form the receptor-binding site (1, 2). HA is a limb paralysis or convulsion). The lung titers of HK213 and host-range determinant, because human and avian influenza virus HK213xVN1203HANA virusesdifferedonday1(P ϭ 0.0174, HAs preferentially recognize ␣2,6- and ␣2,3-linked sialyl receptors, Tukey’s test) but were comparable on days 3, 5, and 7 after respectively (3, 4). In addition to the terminal sialic-acid linkages, inoculation (P ϭ 0.0802, 2-way ANOVA) (Fig. 1A). internal linkages, as well as fucosylation, sulfation, and sialylation at HK213xVN1203HA and HK213xVN1203HANA viruses were de- the inner oligosaccharides, also determine receptor-binding affinity tected in brain or blood, suggesting that infection was not contained and specificity (2, 5). Despite their preferential recognition of in the mouse lungs (Fig. 1A). The results suggest that the HA ␣2,3-linked sialyl receptors, highly pathogenic H5N1 avian influ- glycoproteins of VN1203 contribute to viral spread from the enza viruses are able to infect humans and cause severe disease. respiratory tract. Glycan-array profiling has shown differences between seasonal human influenza virus and H5N1 viruses (2, 6), but current techniques are at an early stage to map the glycans in tissues, and Author contributions: H.-L.Y., A.C.M.B., R.J.W., J.E.R., and R.G.W. designed research; H.- it is not known whether differences in glycan recognition affect L.Y., J.R.A., A.C.M.B., N.A.I., H.M., J.F., D.A.B., D.B., A.S.L., and J.E.R. performed research; R.S. and D.J.H.-P. contributed new reagents/analytic tools; H.-L.Y. and J.R.A. analyzed data; influenza virus pathogenicity. and H.-L.Y., J.R.A., and R.G.W. wrote the paper. The highly pathogenic H5N1 viruses are consistently lethal to The authors declare no conflict of interest. domestic chickens but differ in their pathogenicity to small 1To whom correspondence should be addressed at: Department of Infectious Diseases, St. mammals, including mice and ferrets (7–9). Although H5N1 Jude Children’s Research Hospital, 332 N. Lauderdale, Memphis, TN 38105. E-mail: virulence in mammals is known to be a polygenic trait (10–15), [email protected]. H5N1 viruses that are highly lethal to mice or ferrets commonly This article contains supporting information online at www.pnas.org/cgi/content/full/ spread systemically after intranasal inoculation (8, 9). Further- 0811052106/DCSupplemental. more, H5N1 virus or RNA has been isolated from human © 2008 by The National Academy of Sciences of the USA

286–291 ͉ PNAS ͉ January 6, 2009 ͉ vol. 106 ͉ no. 1 www.pnas.org͞cgi͞doi͞10.1073͞pnas.0811052106 Downloaded by guest on September 28, 2021 Table 1. Afﬁnity of the recombinant H5N1 inﬂuenza viruses to sialyl substrates Recombinant virus Fetuin p6ЈSL p3ЈSL p3ЈSLN

HK213 0.18 Ϯ 0.03 5.78 Ϯ 0.82* 1.05 Ϯ 0.21* ND VN1203 0.14 Ϯ 0.06 Ͼ100 0.23 Ϯ 0.04 1.81 Ϯ 0.22 VN1203 HAS227N 0.26 Ϯ 0.03* Ͼ100 0.13 Ϯ 0.05 6.35 Ϯ 0.34* VN1203 HAS227Nϩ158Gly⌬ 0.11 Ϯ 0.04 19.49 Ϯ 1.21* 0.17 Ϯ 0.03 6.81 Ϯ 0.29*

diss ϭ mean Ϯ SE ϫ t␣,n-1 (␮M sialic acid), where t␣ is Student’s coefﬁcient with probability ␣ ϭ 0.95, from four independent experiments. *P Ͻ 0.05.

alone is not sufficient to confer preferential binding to p6ЈSL. Our results suggest that the increased binding affinity of HK213 for p6ЈSL was mediated mainly by the S227N change and the absence of N-linked glycosylation at residue 158.

Introduced HA Mutations Do Not Affect Viral Growth In Vitro or Membrane Fusion pH. We assessed whether the introduced HA changes affected growth of recombinant VN1203 viruses in vitro. VN1203xHK213HANA virus was generated for comparison. The recombinant viruses grew comparably in Madin-Darby canine kidney (MDCK) cells but formed different plaque sizes (see Table S1). At a multiplicity of infection (MOI) ϭ 0.0001 pfu/cell, multistep growth curves showed comparable titers of VN1203, VN1203- Fig. 1. VN1203 HA contributes to systemic spread in mice. Virus titers in HAS227N, and VN1203-HAS227Nϩ158Gly⌬ viruses in MDCK cells lungs, brains (including olfactory bulbs), and blood of BALB/cJ mice inoculated (Fig. S1). We also evaluated whether the introduced HA mutations intranasally with 100 pfu/50 ␮L of recombinant of HK213 (A) and VN1203 (B) Ϯ would affect membrane fusion activity of the VN1203 recombinant viruses. Each data point represents the mean virus titer SD from 3 mice. *, viruses by determining the pH of syncytium formation in Vero cells. P Ͻ 0.05; , P Ͻ 0.01. ** Vero cells infected with VN1203, VN1203-HAS227N, and VN1203- HAS227Nϩ158Gly⌬ viruses formed syncytia at or below pH 5.8, 5.9, Determinants of HA Affinity for ␣2,6-Linked Sialyl Receptor. We and 5.8, respectively. Syncytium formation was observed at or below assess possible functional differences between the HK213 and pH 5.7 for the VN1203xHK213HANA virus. The results suggest that VN1203 HA glycoproteins. The two HA glycoproteins were the introduced HA mutations did not significantly impair viral shown to differ in their affinity for synthetic ␣2,3-linked 3Ј- fitness in vitro. sialyllactose (p3ЈSL) and ␣2,6-linked 6Ј-sialyllactose (p6ЈSL) receptors (24), although a recent glycan array study found no Changes in the HA Receptor-Binding Motif Affect Virulence in Mice but significant difference between their recognized glycan species Not in Chickens. Despite their comparable growth in vitro, the four (28). Ten amino acid differences between the VN1203 and recombinant viruses differed in their lethality to BALB/cJ mice. ϭ HK213 HA glycoproteins were identified (Table S2); differences VN1203 virus was most lethal (1 MLD50 0.6 pfu), followed by ϩ ⌬ within or near the receptor binding site included one change VN1203-HAS227N 158Gly (8.6 pfu), VN1203xHK213HANA (74.8 among the conserved residues in the 220-loop (S227N in pfu), and VN1203-HAS227N (85.4 pfu) viruses. The Ϸ100-fold HK213), one difference in the 190-helix (R193 in HK213 and reduction in the lethality of VN1203 virus carrying HK213 HA and K193 in VN1203), and two differences (residues 158–160: NST NA confirmed that VN1203 HA and NA surface glycoproteins are for VN1203, NNA for HK213) that resulted in a potential virulence factors. N-linked glycosylation site at residue 158 of VN1203 HA. We observed neurological signs in Ϸ40% of VN1203-inoculated ϩ ⌬ Another human H5N1 isolate (A/Turkey/65–596/06), which ex- and 20% of VN1203-HAS227N 158Gly -inoculated mice, but none in hibits dual binding affinity for p3ЈSL and p6ЈSL (24), also shows VN1203-HAS227N or VN1203xHK213HANA-inoculated mice. Viral the S227N change and de-glycosylation at residue 158. To replication in mouse lungs was comparable at days 1, 3, and 5 (Fig. investigate the effect of these differences on receptor binding 1B), but the day-7 lung titer of VN1203xHK213HANA was signifi- MICROBIOLOGY ϩ ⌬ affinity and specificity, we generated recombinant VN1203, cantly lower than those of the VN1203 or VN1203-HAS227N 158Gly VN1203-HAS227N, and VN1203-HAS227Nϩ158Gly⌬ viruses (de- viruses (P ϭ 0.0043, Tukey’s test) (P ϭ 0.0113, 2-way ANOVA). glycosylation by introducing S159N and T160A mutations). Although all viruses were detectable in the brain (including olfac- The binding affinity of the recombinant viruses to synthetic sialyl tory bulbs) at day 7 after inoculation, only VN1203 and VN1203- receptor substrates was measured (Table 1). The S227N change HAS227Nϩ158Gly⌬ replicated to high titers (Fig. 1B). Detection of alone did not significantly increase binding affinity for p6ЈSL, VN1203xHK213HANA virus in mouse brain on day 7 in the absence consistent with previous report (2). VN1203-HAS227N also exhib- of neurological signs may be attributable to virus replication in the ited reduced binding to fetuin (containing ␣2,3- and ␣2,6-linked olfactory bulbs after intranasal inoculation. Alternatively, the effi- sialyl receptors) and to 3Ј-sialyl (N-acetyllactosamine) (p3ЈSLN). cient transcription and replication activity of VN1203 polymerase The S227N change and de-glycosylation at residue 158 significantly complex (13) may facilitate the spread of VN1203xHK213HANA increased the binding affinity of VN1203 HA for p6ЈSL (see Table viruses by a route other than blood (viremia was detected only in 1) without reducing its binding affinity for p3ЈSL. The importance mice inoculated with VN1203 virus) (Fig. 1B). of glycosylation at residue 158 was recently demonstrated (28). The To test whether VN1203 virus possessed the greatest ability to majority (99%) of the clade 2.2 H5N1 viruses lack the glycosylation spread systemically after respiratory tract infection, we monitored site at residue 158 (28) and retain binding specificity for ␣2,3-linked virus replication kinetics after intramuscular (i.m.) inoculation with sialyl receptors. Thus, the absence of glycosylation at residue 158 100 pfu. All viruses were detectable at the inoculation site (thigh

Yen et al. PNAS ͉ January 6, 2009 ͉ vol. 106 ͉ no. 1 ͉ 287 Downloaded by guest on September 28, 2021 systemically in mice but remained universally lethal to domestic chickens.

Target Cells of the H5N1 Viruses. Systemic spread requires both target cell binding and systemic trafficking. We first compared the cell types bound by the surface glycoproteins of HK213 and VN1203 viruses by immunohistochemistry (IHC) with FITC- labeled viruses and normal mouse tissues (29). To perform IHC in a biosafety level 2 laboratory, we used recombinant A/Puerto Rico/8/34 (PR8) viruses carrying the HA (with the basic amino acid motif removed) and NA glycoproteins of HK213 or VN1203 (designated PR8xHK213HANA and PR8xVN1203HANA). The overall patterns of attachment of the PR8xHK213HANA and PR8xVN1203HANA viruses in the mouse respiratory tract (moder- ately positive in trachea, bronchi, bronchioles, and alveoli) (see Table S3) were similar to those of A/Vietnam/1194/04 (H5N1) virus (29). Viral attachment was observed outside the respiratory tract in endothelial cells of the brain, biliary epithelial cells of the liver, and ciliated ependymal cells lining the brain ventricles (Table S3 and Fig. S2). The similar attachment patterns of the PR8xHK213HANA and PR8xVN1203HANA viruses suggest that the HK213 and VN1203 viruses have similar affinity for these target cells; therefore, the different systemic spread of these viruses is likely to reflect their ability to reach these target cells.

Infection of Mouse Monocytes In Vitro and In Vivo. Because systemic spread depends on systemic trafficking and target-cell binding, we next investigated the ability of the H5N1 viruses to infect Fig. 2. Viral replication kinetics in BALB/cJ mice i.m. inoculated with 100 mouse dendritic cells (DCs). DCs appeared to be good candi- pfu/100 ␮L of H5N1 viruses. (A–E) Virus titers at the inoculation site (thigh and dates for systemic trafficking vehicles because they are profes- popliteal lymph node) (A) and in inguinal lymph node (B), blood (C), lungs (D), sional antigen-presenting cells that are widely distributed in and brain (including olfactory bulbs) (E) after inoculation with the indicated tissues and can migrate to draining lymph nodes for antigen viruses. Each data point represents the mean titer Ϯ SD from 3 mice. (F) presentation. The detection of VN1203 virus in the popliteal Survival of mice i.m. inoculated with recombinant VN1203 viruses. lymph node and blood after i.m. inoculation supports the role of DCs in facilitating systemic spread. Furthermore, recent studies have shown that DCs are susceptible to H5N1 infection and can muscle and popliteal lymph node) until day 3, and VN1203 virus potentially disseminate the virus in vivo (30, 31). remained detectable until day 7 (Fig. 2A). In addition, virus was We first infected differentiated mouse bone marrow-derived detected in the inguinal lymph nodes of mice inoculated with S227Nϩ158Gly⌬ DCs (BMDCs) in vitro with recombinant VN1203, VN1203- VN1203 or VN1203-HA virus (Fig. 2B). Virus was HAS227N, VN1203-HAS227Nϩ158Gly⌬, and VN1203xHK213HANA vi- detected in the blood only in mice inoculated with VN1203 (Fig. ruses and with PR8 (H1N1) virus at MOIs of 0.2, 2, or 10 pfu/cell. 2C). VN1203 virus was detected in the lungs as early as day 3 and After 15 h, cells were analyzed by flow cytometry for CD11b in the brain as early as day 5 after inoculation (Fig. 2 D and E); (monocyte marker), CD11c (DC marker), and influenza nucleo- S227Nϩ158Gly⌬ VN1203-HA virus was detected in lungs but not in protein (NP). At an MOI of 0.2 or 2.0, the percentage of S227N HANA brain, and the VN1203-HA and VN1203xHK213 viruses CD11cϩNPϩ cells was greater (59.4% and 83.6%, respectively) in were not detected in lungs or brain. All mice inoculated with BMDCs infected with VN1203 than in BMDCs infected ϭ VN1203 (n 17) died 7–8 days after inoculation; 4 of 17 mice with VN1203-HAS227N (9.8% and 63.8%, respectively), S227Nϩ158Gly⌬ inoculated with VN1203-HA and 1 of 7 mice inocu- VN1203-HAS227Nϩ158Gly⌬ (11.9% and 68%, respectively), HANA lated with VN1203xHK213 were euthanized because of hind- VN1203xHK213HANA (12.1% and 64.8%, respectively), or PR8 limb paralysis, but mice inoculated with VN1203-HAS227N (n ϭ 17) (4.4% and 18.1%, respectively) viruses (Fig. 3A). At MOI ϭ 10, the showed no clinical signs (Fig. 2F). At an i.m. dose of 105 pfu, all percentage of CD11cϩNPϩ cells was similar in all cells infected with VN1203-inoculated mice (n ϭ 8) died 7 to 8 days after inoculation; the H5N1 recombinants (69.2%–78.7%) but was lower in those ϩ ⌬ 7 of 8 VN1203-HAS227N 158Gly -inoculated mice, 3 of 8 infected with PR8 virus (45.1%). VN1203xHK213HANA-inoculated mice, and 1 of 8 VN1203- To differentiate active viral replication from antigen uptake, we HAS227N-inoculated mice were euthanized because of hind-limb analyzed BDMCs incubated for1hor16hwithliveorheat- paralysis. All mice that survived the 105 pfu i.m. inoculation were inactivated (60°C for 15 min) viruses at MOI ϭ 0.2 or 2 pfu/ cell. protected from homologous intranasal rechallenge with 103 pfu. After 1-h incubation at MOI ϭ 0.2, the percentage of CD11cϩNPϩ The HA and NA genes of virus isolated from two mouse lungs 7 cells was similar in BMDCs incubated with heat-inactivated (0.4– ϩ ⌬ days after i.m. inoculation with 100 pfu of VN1203-HAS227N 158Gly 0.8%) and live (0.3–0.7%) viruses. However, after 16-h incubation, virus were analyzed; the introduced HA mutations were retained the percentage of CD11cϩNPϩ cells was significantly different in and no additional NA mutations were identified. BMDCs incubated with heat-inactivated (0.3–1.2%) and live (19.3– Because VN1203-HAS227N virus showed significantly reduced 53.2%) viruses. The percentage of CD11cϩNPϩ cells was highest in lethality in mice, we also tested its lethality in domestic chickens. BMDCs infected with VN1203 (53.2%), followed by The i.v. pathogenicity indexes of the VN1203, VN1203-HAS227N, VN1203xHK213HANA (34%), VN1203-HAS227Nϩ158Gly⌬ (30.7%), and VN1203-HAS227Nϩ158Gly⌬ viruses were 3.00, 2.95, and 2.98, VN1203-HAS227N (20.8%), and PR8 (19.3%). Similarly, after 1-h respectively. All chickens died within 48 h after inoculation. Overall, incubation at MOI ϭ 2.0, the percentage of CD11cϩNPϩ cells was our results demonstrated that the VN1203 recombinant viruses comparable in BMDCs incubated with heat-inactivated (0.4–3.1%) with different receptor recognition differed in their ability to spread and live (0.5–6.6%) viruses. The percentage of live CD11cϩNPϩ

288 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0811052106 Yen et al. Downloaded by guest on September 28, 2021 overall Ͼ99% of the conventional DC population (CD11chiCD11bint) was NPϩ. However, the percentage of conventional DCs in total lung homogenate was lowest in VN1203- inoculated (4.5 Ϯ 1.2%) and VN1203-HAS227N-inoculated (4.5 Ϯ 0.8%) mice, followed by VN1203-HAS227Nϩ158Gly⌬-inoculated (7.7 Ϯ 2.2%) and VN1203xHK213HANA-inoculated (6.1 Ϯ 1.5%) mice (P ϭ 0.0108, 1-way ANOVA). In Tukey’s test, these percent- ages differed significantly between mice inoculated with VN1203 or VN1203-HAS227N vs. VN1203xHK213HANA virus (P Ͻ 0.05). In a repeated experiment, the number of conventional DCs in the mouse lung 5 days after inoculation was calculated. The mean number of conventional DCs was lowest with VN1203-inoculated mice (84,337.6), followed by VN1203-HAS227N-inoculated (110,697.2), VN1203-HAS227Nϩ158Gly⌬-inoculated (149,171.2), and VN1203xHK213HANA-inoculated (169,913.3) mice. We further differentiated the conventional DC population into NPint and NPhi based on the level of cellular NP expression (Fig. 3B). As DCs mature, antigen internalization is down-regulated (32). Therefore, a high level of NP expression (NPhi) is indicative of cells in which virus is replicating, whereas an intermediate level of NP expression (NPint) is indicative of antigen internalization. The percentage of conventional DCs expressing NPhi was higher in VN1203-inoculated mice than in mice inoculated with VN1203- HAS227N, VN1203-HAS227Nϩ158Gly⌬, or VN1203xHK213HANA (P ϭ 0.001, 1-way ANOVA) (Fig. 3C); the percentage of conventional DCs expressing NPint was lowest in VN1203-inoculated mice (P ϭ 0.001, 1-way ANOVA) (Fig. 3D). Overall, the total lung homogenate of VN1203-inoculated mice showed the lowest percentage of conventional DCs but the highest percentage of conventional DCs expressing NPhi (indicating active viral replication). The low pro- portion of DCs may therefore reflect the death of VN1203-infected Fig. 3. Infection of mouse monocytes in vitro and in vivo. Dendritic cells derived cells. These results support the potential mediation of systemic from C57BL/6 mouse bone marrow were incubated for 12 h with the indicated spread by VN1203-infected DCs. viruses (MOI ϭ 0.2 or 2.0) and infection was analyzed by ﬂow cytometry (A). Representative FACS plot showing conventional DC population Effect of Surface Glycoproteins on Cytokine/Chemokine Production. (CD11chiCD11bint) in mouse lung homogenate, further gated for NP signal inten- sity (NPint,NPhi). A higher percentage of VN1203-inoculated mouse DCs exhibited The pathogenicity of H5N1 viruses is reportedly associated with NPhi than that of VN1203xHK213HANA inoculated mice (B). Percentage of high levels of proinflammatory cytokines (14, 18). There has been CD11chiCD11bintNPhi cells (C) and CD11chiCD11bintNPint cells (D) in BALB/cJ lung evidence suggesting that viral glycoproteins may interact with homogenate 5 days after intranasal inoculation of 100 pfu of VN1203, VN1203- pattern recognition receptors to induce inflammatory cytokine HAS227N, VN1203-HAS227N ϩ 158Gly⌬, or VN1203xHK213HANA virus. production (33). To determine whether changes in surface glycoproteins affect the level of proinflammatory cytokine production, we assayed IFN-␣, IL-6, TNF-␣, MCP-1 (CCL2), MIP-1␣ (CCL3), cells remained low at 0.2 to 0.3% in BMDCs incubated with and IFN-␥ in mouse lung homogenate on days 3 and 6 after heat-inactivated viruses for 16 h; however, the percentage of inoculation with 1,000 pfu of recombinant HK213, HK213x ϩ ϩ CD11c NP cells was 68.4%, 64.2%, 58.6%, 58.5%, and 49.8% in VN1203HANA, VN1203, or VN1203xHK213HANA virus (Fig. S3). ϩ ⌬ BMDCs incubated with live VN1203, VN1203-HAS227N 158Gly , The pulmonary levels of MCP-1 and TNF-␣ differed significantly VN1203-HAS227N, VN1203xHK213HANA, and PR8 viruses, respec- on days 3 and 6 between mice inoculated with HK213 vs. tively. Overall, these results suggest that all 5 viruses tested were HK213xVN1203HANA viruses and between mice inoculated with able to replicate (produce NP protein) in the infected BMDCs, and VN1203 vs. VN1203xHK213HANA viruses). Therefore, VN1203 that VN1203 virus infects and replicates most efficiently in these HA and NA glycoproteins specifically induce greater production of cells. these cytokines. Virus containing the VN1203 internal genes We next analyzed the mononuclear infiltrate in mouse lungs 5 (VN1203 and VN1203xHK213HANA) induced higher levels of MICROBIOLOGY days after intranasal inoculation of 100 pfu of the recombinant IFN-␣ on day 3 and higher levels of TNF-␣ and MCP-1 on days 3 H5N1 viruses (n ϭ 5 per group). All 4 viruses had been shown to and 6 than did HK213 and HK213xVN1203HANA viruses. On day replicate to comparable titers in mouse lungs on day 5 (see Fig. 1D). 6, VN1203-inoculated mice had a significantly higher pulmonary ϩ In total lung homogenate, the mean percentage of NP cells was level of IL-6 and lower level of MIP-1␣ and IFN-␥ than did other higher in mice inoculated with VN1203 [43.8%Ϯ10.0% (SD)] than mice. The low MIP-1␣ level on day 6 suggested the possibility of in mice inoculated with VN1203-HAS227N (37.3 Ϯ 6.6%), VN1203- reduced trafficking of CD8ϩ T cells into the lungs. Poor IFN-␥ HAS227Nϩ158Gly⌬ (27.0 Ϯ 5.7%), or VN1203xHK213HANA (27.7 Ϯ induction indicates an inefficient cell-mediated immune response, 4.4%) (P ϭ 0.0039, 1-way ANOVA). Tukey’s test showed a and is consistent with the comparatively low percentage of con- significantly higher percentage of NPϩ cells after VN1203 inocu- ventional DCs observed in the lungs of VN1203-inoculated mice. lation than after VN1203-HAS227Nϩ158Gly⌬ (P Ͻ 0.01) or VN1203xHK213HANA (P Ͻ 0.05) inoculation. The mean percent- Discussion age of CD11bϩNPϩ cells was also higher, although not significantly Influenza virus surface glycoproteins are known to contribute to so, after inoculation with VN1203 (54.9 Ϯ 14.0%) vs. VN1203- pathogenicity (11, 25–27). Here, we showed that changes in the HA HAS227N (46.3 Ϯ 8.9%), VN1203-HAS227Nϩ158Gly⌬ (39.2 Ϯ 10.0%), receptor-binding domain affect the ability of highly pathogenic or VN1203xHK213HANA (39.5 Ϯ 7.7%) virus (P ϭ 0.0959, 1-way H5N1 virus to spread systemically, thereby modifying its pathoge- ANOVA). The percentage of CD11cϩNPϩ cells did not differ and nicity in a mouse model. The differing ability of VN1203, VN1203-

Yen et al. PNAS ͉ January 6, 2009 ͉ vol. 106 ͉ no. 1 ͉ 289 Downloaded by guest on September 28, 2021 HAS227N, VN1203-HAS227Nϩ158Gly, and VN1203xHK213HANA vi- affect viral replication efficiency in mouse lungs and further mod- ruses to infect BMDC in vitro and conventional DCs in mouse lungs ulate the cytokine response. in vivo suggests that the efficient systemic spread of VN1203 may Influenza pathogenicity is a polygenic trait (10–15). Although we be mediated by the interaction of its HA glycoprotein with specific show here that the HA glycoprotein can facilitate systemic spread, cellular receptors in the monocyte population. Detection of the role of NA glycoprotein should not be overlooked. Further VN1203 virus in the inguinal lymph node and in blood from the studies are also needed to elucidate whether VN1203 HA glycop- retro-orbital plexus after i.m. inoculation further supports the rotein facilitates systemic spread through efficient interaction with possibility of viral spread via infected monocytes. specific receptors on monocytes. Recent studies have identified that A change from ␣2,3-linked to ␣2,6-linked sialyl receptor recog- DC-SIGN, a member of the C-type lectin family, may interact with nition has been observed repeatedly during the evolution of pan- the HA of H5N1 viruses by facilitating capture and attachment demic influenza viruses (i.e., the 1918, 1957, and 1968 pandemics) of the virus particles to DCs (31, 39). DC-SIGN is known to mediate (5, 34). As a consequence, human influenza viruses recognize the HIV-1 infection of T cells (40); it also reportedly interacts with predominantly ␣2,6-linked sialyl receptors (6). Whereas this change the glycoproteins of Ebola virus, hepatitis C virus, dengue virus, may facilitate efficient human-to-human transmission, reduced measles viruses, and SARS coronavirus (37). In our preliminary affinity for a spectrum of ␣2,3-linked sialyl receptors may affect experiment, recombinant human DC-SIGN, DC-SIGNR, and Lan- pathogenicity. H5N1 viruses continue to pose a serious pandemic gerin did not neutralize VN1203 infection of MDCK cells (data not threat, but it is not clear whether they would follow the evolutionary shown), suggesting that these C-type lectins may not served as the path of previous pandemic influenza viruses by gradually acquiring major receptors for entry (39). In addition to C-type lectins, changes in their receptor recognition pattern. In the present study, interaction between viral glycoproteins and Toll-like receptors 2 the differences we observed between VN1203, VN1203-HAS227N, (TLR2) and TLR4 have been reported (41), and interaction ϩ VN1203-HAS227N 158Gly, and VN1203xHK213HANA in systemic between UV-inactivated H5N1 virus and TLR4 has been spread cannot be simply explained by the differential usage of ␣2,3- recently shown to induce acute lung injury in mice (42). vs. ␣2,6-linked sialyl receptors. The introduced HA mutations did Experiments are being planned to further characterize the not affect the pH for syncytium formation or reduce viral lethality interaction between HA and DCs. Taken together, our results in domestic chickens, but other mechanisms that may explain our show that changes in the HA receptor binding domain can alter differing observations in vivo cannot be excluded. For example, the the ability of H5N1 virus to spread systemically in a mouse introduced HA mutations may affect HA susceptibility to protease model. Our results also suggest that H5N1 infected DCs may cleavage in mammalian cells or influence viral transport in the serve as vehicles for systemic spread. endocytic pathways. However, the location of these HA mutations and the fact that they affect binding affinity for p3ЈSL and p6ЈSL Materials and Methods leads us to speculate that they affected binding affinity for or Viruses. H5N1 isolates were obtained through the World Health Organization recognition of fine glycan structures that cannot be fully charac- Global Influenza Surveillance Network. The 8 gene segments of HK213 virus were terized by current methods. The glycan array profiles of VN1203 amplified by RT-PCR and cloned into plasmid vector pHW2000 (43). The plasmids and VN1203-HAS227N viruses (2) indicate that both viruses bind to encoding the 8 genes of VN1203 virus have been described (13). Amino acid the carbohydrate determinant sialyl lewisx [Neu5Ac␣2–3Gal␤1– changes were introduced by using the QuikChange site-directed mutagenesis kit 4(Fuc␣1–3)GlcNAc], but only VN1203 binds to sialyl lewisa (Stratagene). Recombinant HK213 and VN1203 viruses were generated by co- [Neu5Ac␣2–3Gal␤1–3(Fuc␣1–4)GlcNAc] (2). Both sialyl lewisx transfecting 293T cells (TransIT-LT1, Mirus) with the appropriate 8 plasmids (43), and sialyl lewisa are important in leukocyte rolling and cancer cell and virus stocks were prepared in embryonated chicken eggs. The full genome sequences of the recombinant viruses were verified. Experiments with H5N1 metastasis (35, 36). viruses were conducted in a United States Department of Agriculture-approved Hypercytokinemia and high viral load are associated with fatal biosafety level 3ϩ containment facility. human H5N1 infection (18). We observed that the most lethal virus studied, VN1203, induced an elevated proinflammatory Binding of Virus to Sialic Acid-Containing Substrates. Virus binding affinity to cytokine response in mouse lungs (TNF-␣, MCP-1, and MIP-1 HRP-conjugated fetuin was tested in a direct solid-phase assay with the immo- ␣ on day 3 after inoculation; IL-6 and MCP-1 on day 6). Infection bilized virus (44). Virus binding affinity to the synthetic sialyl receptors p3ЈSL and of monocytes with VN1203 virus may lead to apoptosis of these p6ЈSL (provided by Dr. Nicolai V. Bovin, Shemyakin and Ovchinnikov Institute of cells and increased level of proinflammatory cytokine response Bioorganic Chemistry, Moscow) was tested by competitive assay based on inhi- in mouse lungs. The observation that VN1203 surface glycop- bition of binding to HRP-labeled fetuin (45). The dissociation constants (Kdiss)of roteins specifically induced higher level of TNF-␣ and MCP-1 virus-receptor analog complexes were calculated as the means of 4 individual suggest that HA and NA may affect viral infectivity in mono- experiments. cytes. Furthermore, infection of the conventional pulmonary DC Animal Experiments. The MLD was determined by intranasally inoculating population would impede DC antigen presentation to T cells and 50 groups of 6-week-old female BALB/cJ mice (Jackson Laboratories) with 10-fold might convey inhibitory signals to T cells (37). Lymphopenia has serial dilutions of virus in 50 ␮l of PBS under isoflurane anesthesia. Mice were been reported in humans infected with H5N1 virus (38) and in inoculated i.m. with virus diluted in 100 ␮l of PBS with 26-gauge needles. Mouse mice inoculated with highly pathogenic H5N1 virus (21). blood and organs were collected and stored at Ϫ70°C before titration in MDCK The internal genes of VN1203 induced higher levels of IFN-␣ on cells. Clinical signs were monitored daily. Mice that developed neurological signs day 3 and of TNF-␣ and MCP-1 on days 3 and 6 after inoculation. or lost Ͼ25% of their original weight were euthanized. The i.v. pathogenicity The polymerase complex of VN1203 is reported to confer virulence index was determined in groups of 10 8-week-old chickens by i.v. injection with in mammals by enhancing replication and transcription efficiency 0.1 ml of 10-fold dilutions of virus-containing allantoic fluid. Deaths and disease (13). In our studies, VN1203 polymerase complex showed signifi- signs were recorded daily for 10 days. All animal studies were conducted under cantly greater transcription and replication efficiency than HK213 applicable laws and guidelines and were approved by the St. Jude Children’s virus in human lung epithelial A549 cells but not in quail fibroblast Research Hospital Animal Care and Use Committee. QT6 cells (Yen et al., unpublished data). In this study, the VN1203 In Vitro Infection of Bone Marrow-Derived DCs. BMDC were generated as recombinant viruses replicated more efficiently in mouse lungs described (47, 48) by culturing bone marrow cells harvested from the femurs and (range, 4.8–6.5 log10TCID50/ gram tissue) than did HK213 recom- tibias of C57BL/6 mice in the presence of GM-CSF (200 units/ml) (Peprotech). Half binant viruses (range, 3.5–4.8 log10TCID50/ gram tissue) on day 1 of the medium was replaced every 2 days, and nonadherent cells were collected after inoculation with 100 pfu of virus, regardless of the surface on day 8. Harvested cells were resuspended (2.5 ϫ 106 cells per reaction) in glycoproteins they carried. Therefore, polymerase activity may RPMI1640 containing live or heat-inactivated influenza viruses at different MOIs

290 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0811052106 Yen et al. Downloaded by guest on September 28, 2021 and incubated at 37°C for 1 h. Cells were washed twice with HBSS, plated in 6-well cytometry data were collected on a BD FACSCalibur (BD) and analyzed by plates, and incubated at 37°C before flow-cytometry analysis. FlowJo software (Tree Star, Inc.). For additional information, see SI Text. Flow Cytometry. Mouse lung monocyte populations were analyzed after intranasal inoculation of 100 pfu of recombinant VN1203 viruses. Single ACKNOWLEDGMENTS. We thank the World Health Organization Global cell suspensions were prepared by passing lung tissue through a 100-␮m Influenza Surveillance Network for providing the H5N1 viruses; Nicolai V. nylon mesh (BD). For surface staining, isolated cells (1 ϫ 106) were stained Bovin for the gift of sialic polymer substrates; Mark L Reed, Nicholas Negov- etich, Elena Govorkova, and Paul Thomas for advice; Scott Krauss, Alexey with mAbs to CD11b (monocyte marker) (M1/70; BD) and CD11c (DC Khalenkov, Kelly Jones, Patrick Seiler, Heather Forrest, Jennifer McClaren, marker) (N418; eBiosciences). For intracellular staining, cells were fixed and David Carey, and Cedric Proctor for excellent technical assistance; and Sharon permeabilized (Cytofix/Cytoperm kit, BD), then stained with a FITC-labeled Naron for editorial assistance. This work was supported by Contract mAb specific for influenza NP (Argene). The experiments were repeated HHSN266200700005C with the National Institute of Allergy and Infectious independently and the representative result is reported. All flow- Diseases and by the American Lebanese Syrian Associated Charities.

1. Skehel JJ, Wiley DC (2000) Receptor binding and membrane fusion in virus entry: The 26. Munster VJ, et al. (2007) The molecular basis of the pathogenicity of the Dutch highly influenza hemagglutinin. Annu Rev Biochem 69:531–569. pathogenic human influenza A H7N7 viruses. J Infect Dis 196:258–265. 2. Stevens J, et al. (2006) Structure and receptor specificity of the hemagglutinin from an 27. Kobasa D, et al. (2004) Enhanced virulence of influenza A viruses with the haemag- H5N1 influenza virus. Science 312:404–410. glutinin of the 1918 pandemic virus. Nature 431:703–707. 3. Rogers GN, et al. (1983) Single amino acid substitutions in influenza haemagglutinin 28. Stevens J, et al. (2008) Recent avian H5N1 viruses exhibit increased propensity for change receptor binding specificity. Nature 304:76–78. acquiring human receptor specificity. J Mol Biol 381:1382–1394. 4. Rogers GN, Paulson JC (1983) Receptor determinants of human and animal influenza 29. van Riel D, et al. (2007) Human and avian influenza viruses target different cells in the virus isolates: Differences in receptor specificity of the H3 hemagglutinin based on lower respiratory tract of humans and other mammals. Am J Pathol 171:1215–1223. species of origin. Virology 127:361–373. 30. Perrone LA, Plowden JK, Garcia-Sastre A, Katz JM, Tumpey TM (2008) H5N1 and 1918 5. Stevens J, et al. (2006) Glycan microarray analysis of the hemagglutinins from modern pandemic influenza virus infection results in early and excessive infiltration of macro- and pandemic influenza viruses reveals different receptor specificities. J Mol Biol 355:1143–1155. phages and neutrophils in the lungs of mice. PLoS Pathog 4:e1000115. 6. Kumari K, et al. (2007) Receptor binding specificity of recent human H3N2 influenza 31. Thitithanyanont A, et al. (2007) High susceptibility of human dendritic cells to avian viruses. Virol J 4:42. influenza H5N1 virus infection and protection by IFN-alpha and TLR ligands. J Immunol 7. Gao P, et al. (1999) Biological heterogeneity, including systemic replication in mice, of 179:5220–5227. H5N1 influenza A virus isolates from humans in Hong Kong. J Virol 73:3184–3189. 32. Banchereau J, Steinman RM (1998) Dendritic cells and the control of immunity. Nature 8. Govorkova EA, et al. (2005) Lethality to ferrets of H5N1 influenza viruses isolated from 392:245–252. humans and poultry in 2004. J Virol 79:2191–2198. 33. Akira S, Uematsu S, Takeuchi O (2006) Pathogen recognition and innate immunity. Cell 9. Maines TR, et al. (2005) Avian influenza (H5N1) viruses isolated from humans in Asia in 124:783–801. 2004 exhibit increased virulence in mammals. J Virol 79:11788–11800. 34. Connor RJ, Kawaoka Y, Webster RG, Paulson JC (1994) Receptor specificity in human, 10. Rott R, Reinacher M, Orlich M, Klenk HD (1980) Cleavability of hemagglutinin deter- avian, and equine H2 and H3 influenza virus isolates. Virology 205:17–23. mines spread of avian influenza viruses in the chorioallantoic membrane of chicken 35. Sperandio M (2006) Selectins and glycosyltransferases in leukocyte rolling in vivo. FEBS embryo. Arch Virol 65:123–133. J 273:4377–4389. 11. Hatta M, Gao P, Halfmann P, Kawaoka Y (2001) Molecular basis for high virulence of 36. Ugorski M, Laskowska A (2002) Sialyl Lewis(a): A tumor-associated carbohydrate Hong Kong H5N1 influenza A viruses. Science 293:1840–1842. antigen involved in adhesion and metastatic potential of cancer cells. Acta Biochim Pol 12. Shinya K, et al. (2004) PB2 amino acid at position 627 affects replicative efficiency, but 49:303–311. not cell tropism, of Hong Kong H5N1 influenza A viruses in mice. Virology 320:258–266. 37. Pohl C, Shishkova J, Schneider-Schaulies S (2007) Viruses and dendritic cells: Enemy 13. Salomon R, et al. (2006) The polymerase complex genes contribute to the high mine. Cell Microbiol 9:279–289. virulence of the human H5N1 influenza virus isolate A/Vietnam/1203/04. J Exp Med 38. bdel-Ghafar AN, et al. (2008) Update on avian influenza A (H5N1) virus infection in 203:689–697. humans. N Engl J Med 358:261–273. 14. Cheung CY, et al. (2002) Induction of proinflammatory cytokines in human macro- phages by influenza A (H5N1) viruses: A mechanism for the unusual severity of human 39. Wang SF, et al. (2008) DC-SIGN mediates avian H5N1 influenza virus infection in cis and disease? Lancet 360:1831–1837. in trans. Biochem Biophys Res Commun 5:561–566. 15. Seo SH, Hoffmann E, Webster RG (2002) Lethal H5N1 influenza viruses escape host 40. Geijtenbeek TB, et al. (2000) DC-SIGN: A dendritic cell-specific HIV-1-binding protein anti-viral cytokine responses. Nat Med 8:950–954. that enhances trans-infection of T cells. Cell 100:587–597. 16. Buchy P, et al. (2007) Influenza A/H5N1 virus infection in humans in Cambodia. J Clin 41. Thompson AJ, Locarnini SA (2007) Toll-like receptors, RIG-I-like RNA helicases and the Virol 39:164–168. antiviral innate immune response. Immunol Cell Biol 85:435–445. 17. de Jong MD, et al. (2005) Fatal avian influenza A (H5N1) in a child presenting with 42. Imai Y, et al. (2008) Identification of oxidative stress and Toll-like receptor 4 signaling diarrhea followed by coma. N Engl J Med 352:686–691. as a key pathway of acute lung injury. Cell 133:235–249. 18. de Jong MD, et al. (2006) Fatal outcome of human influenza A (H5N1) is associated with 43. Hoffmann E, Neumann G, Kawaoka Y, Hobom G, Webster RG (2000) A DNA transfec- high viral load and hypercytokinemia. Nat Med 12:1203–1207. tion system for generation of influenza A virus from eight plasmids. Proc Natl Acad Sci 19. Gu J, et al. (2007) H5N1 infection of the respiratory tract and beyond: A molecular USA 97:6108–6113. pathology study. Lancet 370:1137–1145. 44. Gambaryan AS, Matrosovich MN (1992) A solid-phase enzyme-linked assay for influ- 20. Uiprasertkul M, et al. (2005) Influenza A H5N1 replication sites in humans. Emerg Infect enza virus receptor-binding activity. J Virol Methods 39:111–123. Dis 11:1036–1041. 45. Matrosovich MN, et al. (1993) Probing of the receptor-binding sites of the H1 and H3 21. Tumpey TM, Lu X, Morken T, Zaki SR, Katz JM (2000) Depletion of lymphocytes and influenza A and influenza B virus hemagglutinins by synthetic and natural sialosides. diminished cytokine production in mice infected with a highly virulent influenza A Virology 196:111–121. (H5N1) virus isolated from humans. J Virol 74:6105–6116. 46. Hoffmann E, Lipatov AS, Webby RJ, Govorkova EA, Webster RG (2005) Role of specific 22. Park CH, et al. (2002) The invasion routes of neurovirulent A/Hong Kong/483/97 (H5N1) hemagglutinin amino acids in the immunogenicity and protection of H5N1 influenza influenza virus into the central nervous system after respiratory infection in mice. Arch Virol 147:1425–1436. virus vaccines. Proc Natl Acad Sci USA 102:12915–12920. 23. Shinya K, et al. (2005) Characterization of a human H5N1 influenza A virus isolated in 47. Inaba K, et al. (1992) Generation of large numbers of dendritic cells from mouse bone 2003. J Virol 79:9926–9932. marrow cultures supplemented with granulocyte/macrophage colony-stimulating fac- 24. Yen HL, et al. (2007) Inefficient transmission of H5N1 influenza viruses in a ferret tor. J Exp Med 176:1693–1702. 48. Mahnke K, Bhardwaj RS, Luger TA, Schwarz T, Grabbe S (1996) Interaction of murine

contact model. J Virol 81:6890–6898. MICROBIOLOGY 25. Chen H, et al. (2007) Polygenic virulence factors involved in pathogenesis of 1997 Hong dendritic cells with collagen up-regulates allostimulatory capacity, surface expression Kong H5N1 inﬂuenza viruses in mice. Virus Res 128:159–163. of heat stable antigen, and release of cytokines. J Leukoc Biol 60:465–472.

Yen et al. PNAS ͉ January 6, 2009 ͉ vol. 106 ͉ no. 1 ͉ 291 Downloaded by guest on September 28, 2021