HA-Dependent Tropism of H5N1 and H7N9 Influenza Viruses to Human Endothelial Cells Is Determined by Reduced Stability of the HA, Which Allows the Virus To Cope with Inefficient Endosomal Acidification and Constitutively Expressed IFITM3

Receptor specificity of the HA of IAVs is known to be a critical determinant of viral cell tropism. Here, we show that fusion properties of the HA may also play a key role in the tropism. Thus, we demonstrate that IAVs having a relatively low pH optimum of fusion cannot efficiently infect human endothelial cells owing to their relatively high endosomal pH and increased expression of fusion-inhibiting IFITM3 protein. These restrictions can be overcome by IAVs with elevated pH of fusion, such as zoonotic H5N1 and H7N9. Our results illustrate that the infectivity of IAVs depends on an interplay between HA conformational stability, endosomal acidification and IFITM3 expression in target cells, and the extracellular pH. Given significant variation of levels of HA stability among animal, human, and zoonotic IAVs, our findings prompt further studies on the fusion-dependent tropism of IAVs to different cell types in humans and its role in viral host range and pathogenicity.

ABSTRACT

Previous studies revealed that certain avian influenza A viruses (IAVs), including zoonotic H5N1 and H7N9 IAVs, infect cultured human lung microvascular endothelial cells (HULEC) more efficiently than other IAVs and that tropism to HULEC is determined by viral hemagglutinin (HA). To characterize mechanisms of HA-mediated endotheliotropism, we used 2:6 recombinant IAVs harboring HAs from distinctive avian and human viruses and found that efficient infection of HULEC correlated with low conformational stability of the HA. We next studied effects on viral infectivity of single-point amino acid substitutions in the HA of 2:6 recombinant virus A/Vietnam/1203/2004-PR8 (H5N1). Substitutions H8Q, H103Y, T315I, and K58₂I (K58I in the HA2 subunit), which increased stability of the HA, markedly reduced viral infectivity for HULEC, whereas substitutions K189N and K218Q, which altered typical H5N1 virus-like receptor specificity and reduced binding avidity of the HA, led to only marginal reduction of infectivity. None of these substitutions affected virus infection in MDCK cells. We confirmed the previous observation of elevated basal expression of IFITM3 protein in HULEC and found that endosomal acidification is less efficient in HULEC than in MDCK cells. In accord with these findings, counteraction of IFITM3-mediated restriction by amphotericin B and reduction of endosomal pH by moderate acidification of the extracellular medium enhanced infectivity of viruses with stable HA for HULEC without significant effect on infectivity for MDCK cells. Collectively, our results indicate that relatively high pH optimum of fusion of the HA of zoonotic H5N1 and H7N9 IAVs allows them to overcome antiviral effects of inefficient endosomal acidification and IFITM3 in human endothelial cells.

IMPORTANCE Receptor specificity of the HA of IAVs is known to be a critical determinant of viral cell tropism. Here, we show that fusion properties of the HA may also play a key role in the tropism. Thus, we demonstrate that IAVs having a relatively low pH optimum of fusion cannot efficiently infect human endothelial cells owing to their relatively high endosomal pH and increased expression of fusion-inhibiting IFITM3 protein. These restrictions can be overcome by IAVs with elevated pH of fusion, such as zoonotic H5N1 and H7N9. Our results illustrate that the infectivity of IAVs depends on an interplay between HA conformational stability, endosomal acidification and IFITM3 expression in target cells, and the extracellular pH. Given significant variation of levels of HA stability among animal, human, and zoonotic IAVs, our findings prompt further studies on the fusion-dependent tropism of IAVs to different cell types in humans and its role in viral host range and pathogenicity.

INTRODUCTION

Wild aquatic birds represent the major natural reservoir of influenza A viruses (IAVs), from which they can transmit and adapt to other animal species. IAVs of domestic birds and animals, such as gallinaceous poultry and pigs, occasionally infect humans. On very rare occasions, animal IAVs adapt for efficient transmission in humans, initiate influenza pandemics, and continue to circulate and evolve in the human population, causing seasonal influenza epidemics (for reviews, see references 1 and 2). Thus, humans can be infected by animal, pandemic, and seasonal IAVs that have distinguishable properties.

Seasonal IAVs typically replicate in the epithelial cells of the upper respiratory tract and cause self-limited infection with mild respiratory symptoms. In marked contrast, avian viruses of the H5N1/1997 and H7N9/2013 lineages, which are responsible for the majority of zoonotic IAV infections known to date, often replicate in the epithelial cells of terminal bronchioles and type II pneumocytes. This leads to destruction of the alveolar epithelium and epithelial-endothelial barrier, induction of high levels of proinflammatory cytokines, strong infiltration of innate immune cells, and acute lung injury (3, 4).

The alveolar tropism of the zoonotic H5N1 and H7N9 IAVs in humans is thought to be associated, at least in part, with the avian virus-like receptor-binding specificity of their envelope protein hemagglutinin (HA) (reviewed in references 5 and 6). The HA mediates virus attachment to target cells by binding to sialic acid-containing cell surface glycans. Avian and equine viruses typically bind to receptors containing terminal Neu5Ac2-3Gal residues, whereas swine and human viruses bind to receptors terminated with Neu5Ac2-6Gal moieties. In addition, IAVs from different host species may differ by their recognition of the penultimate sugar residues and more distant parts of the receptors. An interplay between fine-receptor specificity of the HA and spectra of sialoglycans in target tissues of different species represents the major determinant of host range and tissue tropism of IAVs (for reviews, see references 7, to ,9).

Following endocytosis, IAVs are carried through early, maturing, and late endosomes toward lysosomes. Exposure to gradually decreasing pH triggers conformational transition of the HA, fusion of the viral with the endosomal membrane, and penetration of the viral genome into the cytoplasm (reviewed in references 10, to ,12). The pH threshold of this transition depends on conformational stability of the HA and determines both the pH range in which viral-endosomal fusion takes place and the survival of the virus in the environment. There is increasing evidence that conformational stability of the HA, pH optimum of fusion, and environmental stability differ between IAVs from different host species and that these differences may affect viral host range and cell tropism (13–16). However, the role of HA stability in IAV tropism and pathogenicity in humans remains elusive.

In addition to alveolar epithelial cells, microvascular endothelial cells represent the major cell type in the lung. These cells are separated from lung epithelial cells by two thin basement membranes and can be exposed to IAVs released from infected pneumocytes. Although infected endothelial cells were only rarely found in lung autopsies of patients with severe influenza (17, 18), several groups demonstrated efficient in vitro infection of primary human lung microvascular endothelial cells (HULEC) by IAVs (19–21). The infection was accompanied by pronounced induction of adhesion molecules and cytokines and elevated cell death. Remarkably, zoonotic H5N1, H7N9, and some other poultry IAVs infected and activated HULEC more efficiently than human viruses and IAVs of aquatic birds (19–23), highlighting the potential clinical significance of the ability of some IAV to infect HULEC.

Experiments with reassortant IAVs suggest that viral HA is the major determinant of the HULEC tropism (19, 23), but the underlying molecular mechanisms remained obscure. It was found that HULEC express both 3-linked and 6-linked sialic acids and that binding preference of IAVs for either Neu5Ac2-3Gal (avian type) or Neu5Ac2-6Gal (human type) terminal receptor moieties could not alone account for the observed differences in levels of viral infectivity (19–21). As enhanced infectivity for HULEC correlated with the virus origin from gallinaceous poultry, it was hypothesized that this tropism may be determined by distinctive receptor-binding properties shared by H5, H7, and H9 poultry viruses (19, 24). Another mechanism was suggested in a recent report by Sun and colleagues (23). They found that human IAVs were internalized by HULEC and initiated hemifusion but failed to release their genome into the cytoplasm. The restriction was mediated, at least in part, by relatively high constitutive expression in HULEC of the interferon-inducible transmembrane protein 3 (IFITM3), which is expressed at different levels in vertebrate cells and strongly upregulated by interferon. IFITM3 resides in late endosomes and lysosomes and inhibits fusion of various enveloped and some nonenveloped viruses, presumably, by blocking formation of fusion pores at the posthemifusion stage (for reviews, see references 25 and 26). Sensitivity of IAVs to IFITM3 depends on the viral pH optimum of fusion, with viruses that fuse at a higher pH being less sensitive to IFITM-mediated restriction (27). However, there was no clear correlation between the pH optimum of fusion and ability of IAVs to infect HULEC (23), suggesting that the IFITM3-mediated restriction cannot fully account for the differences in viral tropism to these cells.

In this study, we aimed to validate previous reports about the critical role of the HA in the ability of IAVs to infect HULEC and to determine whether receptor-binding specificity, conformational stability, or both characteristics of the HA determine viral endotheliotropism. We also wanted to specify characteristics of HULEC which determine differences in their susceptibility to IAVs. Our results show that reduced conformational stability of the HA is primarily responsible for enhanced endotheliotropism of zoonotic avian IAVs and that low infectivity for HULEC of IAVs with pH-stable HA is determined by the combined effects of elevated constitutive expression of IFITM3 and relatively high endosomal pH.

RESULTS

Tropism of IAVs to HULEC depends on the HA and correlates with HA conformational stability.

Previous studies on IAV tropism to HULEC revealed significant differences in the ability of different viruses to infect these cells and suggested that infectivity depends on the origin of the viral HA. As the majority of these studies were performed with wild-type IAVs differing by all eight gene segments, we decided to validate previous findings using a panel of 2:6 recombinant viruses that shared six gene segments of laboratory strain A/Puerto Rico/8/1934 (PR8) and carried HA and neuraminidase (NA) segments from distinctive avian and human IAVs. The homologous NAs were included together with the HAs to preserve the functional HA-NA balance. The viruses were prepared, and their receptor-binding and membrane fusion properties have been previously characterized (27). They represented typical properties of IAVs from aquatic birds, gallinaceous poultry, and humans (Table 1). Thus, avian-origin A/mallard/Alberta/119/1998 (H1N1) (mal-H1N1), A/mallard/New York/6750/1978 (H2N2) (mal-H2N2), A/Vietnam/1203/2004 (H5N1) (VN-H5N1), and A/Shanghai/2/2013 (H7N9) (Sh-H7N9) preferentially bound to terminal Neu5Ac2-3Gal moieties whereas human A/Memphis/14/1996 (H1N1) (Mem-H1N1) and A/Hong Kong/1/1968 (H3N2) (HK-H3N2) preferred Neu5Ac2-6Gal moieties. The HAs from duck, pandemic, and seasonal human viruses underwent acid-induced conformational transition in the pH range from 4.7 to 5.1, whereas HAs of two zoonotic poultry viruses had a higher pH range of the transition (pH 5.6 to 5.7). A single-point HA mutant of HK-H3N2 (17R) that differed from the parent virus by a reduced conformational stability and elevated pH optimum of fusion was also included in the study.

TABLE 1.

The 2:6 recombinant PR8-based viruses with wild-type HA and NA used in this study^a

Designation	Origin of HA and NA	Host species (IAV type)	HA preference for Neu5Ac-Gal linkageb	Low-pH conformational transition of HA (pH50)c
mal-H1N1	A/mallard/Alberta/119/1998 (H1N1)	Aquatic bird	α2-3	4.9
mal-H2N2	A/mallard/New York/6750/1978 (H2N2)	Aquatic bird	α2-3	5.1
Mem-H1N1	A/Memphis/14/1996 (H1N1)	Human (seasonal)	α2-6	4.7
HK-H3N2	A/Hong Kong/1/1968 (H3N2)	Human (pandemic)	α2-6 > α2-3	4.9
17R	HK-H3N2 with H17R mutation in HA	Human (pandemic)	α2-6 > α2-3	5.3
VN-H5N1	A/Vietnam/1203/2004 (H5N1)	Poultry (zoonotic)	α2-3	5.7
Sh-H7N9	A/Shanghai/2/2013 (H7N9)	Poultry (zoonotic)	α2-3 > α2-6	5.6

^aThe viruses were generated and characterized previously (27).

^bVirus ability to bind either Neu5Acα2-6Gal-terminated (α2-6) receptors, Neu5Acα2-3Gal-terminated (α2-3) receptors, or both receptor types. The symbol > designates stronger binding to the receptor type indicated.

^cpH values at which 50% of the HA undergoes acid-induced conformational transition as judged by altered viral sensitivity to protease digestion; correlate of the viral pH optimum of fusion.

HULEC cultures employed in the previous experiments on IAV infection were prepared using commercial primary cells and medium; the cells were cultivated either on plastic (19, 21, 23) or on Transwell membrane supports (20, 21). We used the second cultivation variant as a more representative model of endothelium in vivo. Our initial experiments (data not shown) confirmed previous findings which indicated that replication of IAVs with a monobasic cleavage site in HULEC is limited to one cycle and that IAVs do not infect polarized HULEC cultures from the basolateral side of the Transwell membrane (19, 21). To compare levels of viral infectivity for HULEC, we inoculated apical sides of polarized cultures with identical virus doses based on infectious titers determined in MDCK cells. We incubated cultures for 18 h and detected infected cells by immunostaining for viral nucleoprotein (NP). Absolute numbers of infected cells varied between cultures prepared from cells of different donors and between experiments performed on different days; however, the relative infectivity of the viruses illustrated in Fig. 1 was reproducible. Namely, VN-H5N1, Sh-H7N9, and 17R always infected HULEC significantly more efficiently than the other four viruses. These results confirmed that the origin of the HA represents an important determinant of IAV tropism to endothelial cells. The infectivity of recombinant IAVs correlated with the conformational stability of the HA; viruses with less stable HAs were more infectious than their relatively stable counterparts. It should be noted, however, that even the most efficient IAVs were significantly less infectious for HULEC than for MDCK cells; for example, fewer than 1.5 × 10⁴ cells were infected in HULEC inoculated with 10⁵ focus-forming units (FFU) of VN-H5N1 (Fig. 1b). In contrast to HA stability, binding preference of the HA for the type of Neu5Ac-Gal linkage did not correlate with infectivity. For example, viruses that bound to Neu5Ac2-3Gal were both highly infectious (VN-H5N1 and Sh-H7N9) and poorly infectious (mal-H1N1 and mal-H2N2), as were the viruses that bound to Neu5Ac2-6Gal (17R versus HK-H3N2 and Mem-H1N1). The dominant role of the HA conformational stability in the tropism was particularly appealing in the case of HK-H3N2 and its single-point fusion mutant 17R, which have identical receptor-binding properties and differed solely by the pH of HA conformational transition.

FIG 1.

Infection of HULEC with 2:6 recombinant IAVs described in Table 1. The cells were infected with 10⁵ FFU of virus per culture (based on titers determined in MDCK cells). Infected cultures were incubated for 18 h, fixed, and immunostained for NP. (a) Microscopic images. Red, viral NP; blue, counterstaining with hematoxylin. Objective, ×2. (b) Numbers of infected cells per culture (mean values and SD from the experiment with cells of three different donors each tested in duplicate; n = 6). Filled circles, open circles, and triangles depict donors 1, 2, and 3, respectively. Asterisks indicate viruses with statistically significant differences in results with respect to those with HK-H3N2 (***, P < 0.001).

Effects of amino acid substitutions in the HA on infectivity of VN-H5N1 for HULEC.

To further study the roles of membrane fusion and receptor-binding properties of the HA in the endotheliotropism of IAVs, we generated seven variants of VN-H5N1 with point mutations in the HA (Fig. 2 and Table 2). This virus strain was chosen as a model because zoonotic H5N1 IAVs displayed particularly strong tropism to HULEC (19–21, 23) and because the structure-function relationship of the HA of A/Vietnam/1203/2004 (H5N1) and closely related virus strains has been extensively characterized. Three variants were generated with substitutions in the HA1 subunit (H8Q, H103Y, and T315I) and one variant with a K58I substitution in the HA2 (K58₂I); these substitutions were shown to increase HA stability and decrease the pH optimum of fusion of the H5N1 viruses (28–32). We previously hypothesized (19, 24) that a stronger endotheliotropism of H5 and H7 IAVs than of other avian IAVs can be associated with distinctive receptor-binding properties, namely, high-affinity binding of H5 and H7 IAVs to 3-linked sialyloligosaccharide moieties 6-Su-3′SLN and 6-Su-SLe^x (where 6′-Su-3′SLN is 6′-sulfated 3′-sialyl-N-acetyllactosamine and 6-Su-SLe^x is 6′-sulfated sialyl Lewis X) containing fucose and/or sulfo group at the subterminal GlcNAc residue (Table 3 gives the structures and designations of the sialyloligosaccharides). To test this hypothesis, we prepared three variants of VN-H5N1 containing single substitutions K189N and K218Q and their combination. Based on the crystal structure of H5 HA complex with 6-Su-SLe^x (33) (Fig. 2b), we assumed that the first substitution would destroy the ionic bond between the HA and the sulfo group of the receptor, whereas the second substitution would affect HA interactions with the fucose moiety. Analysis of the published H5 HA sequences indicated that positively charged amino acids (either K or R) were conserved in HA positions 189 and 218 of H5N1 viruses isolated before 2008; however, substitutions with 189N and 218Q were commonly observed after 2009 (data not shown). We therefore assumed that introduction of these substitutions into VN-H5N1 will be tolerated.

FIG 2.

Amino acid substitutions introduced into the HA of VN-H5N1 in this study shown on the X-ray structure of the H5 HA complex with 6-Su-SLe^x (Protein Data Bank accession number 3ZNL) (33). (a) Two HA monomers are shown in gray, and the third monomer is shown in green (HA1) and cyan (HA2). Positions of substitutions are shown on this monomer as yellow space-filled models; H5 numbering and amino acid changes are displayed on the right. The 6-Su-SLe^x molecule in the receptor-binding site of the third monomer is shown as a stick model; the fusion peptide is shown in red. (b) Close-up view of the receptor-binding pocket illustrating the ionic bond between the sulfate group of the 6-O-sulfo-GlcNAc residue and the side chain of the amino group of lysine in position 189 and the close proximity of the fucose residue to the side chain of lysine in position 218.

TABLE 2.

HA mutants of 2:6 recombinant virus VN-H5N1

HA type and/or substitution(s)a	Plaque diam in MDCK cells (mm)b
	Mean	SD (n)
Wild type
VN-H5N1	3.3	1.0 (27)
Fusion mutants
H8Q (18)	3.3	0.74 (18)
H103Y (110)	3.5	1.1 (20)
T315I (318)	3.4	0.90 (20)
K582I (58)	3.1	0.95 (23)
Receptor-binding mutants
K189N (193)	3.6	0.78 (27)
K218Q (222)	3.5	0.80 (25)
KK/NQ (193, 222)	3.2	0.81 (19)

^aAll substitutions other than K58₂I (K58I in HA2) are in the HA1 subunit. The H5 numbering system is used to define the position of substitution; numbers in parentheses refer to H3 numbering.

^bThe mean diameters and SD of plaques formed by the viruses in MDCK cells under semisolid overlay medium in the experiment performed on the same day. All P values for the differences with respect to VN-H5N1 are >0.05. n, number of plaques.

TABLE 3.

Structure and designation of sialyloligosaccharide moieties of SGPs

Structure	Designation
Neu5Acα2-3Galβ1-4GlcNAcβ	3′SLN
Neu5Acα2-3Galβ1-4(6-HSO3)GlcNAcβ	6-Su-3′SLN
Neu5Acα2-3Galβ1-4(Fucα1-3)GlcNAcβ	SLex
Neu5Acα2-3Galβ1-4(Fucα1-3)-(6-HSO3)GlcNAcβ	6-Su-SLex
Neu5Acα2-3Galβ1-3GlcNAcβ	SLec

All seven variants and VN-H5N1 formed plaques of similar sizes in MDCK cells (Table 2), indicating that introduced substitutions had no major effect on virus replication in these cells. To verify the effects of the substitutions on the fusion properties of the HA, we determined the pH of the conformational transition of the HA and inhibition of the viral infection by ammonium chloride (Fig. 3a and b). As expected, all four fusion mutants (H8Q, H103Y, T315I, and K58₂I) underwent conformational transition at significantly lower pH than the parental VN-H5N1. Three mutants were also significantly more sensitive than VN-H5N1 to neutralization by NH₄Cl, indicating that they require a lower endosomal pH for fusion and penetration of the genome into the cytoplasm. Among the four fusion-affecting substitutions, H103Y and K58₂I induced the largest alteration of the HA conformational stability. The receptor-binding mutants K189N, K218Q, and KN/KQ displayed either marginal alteration or no alteration of the acid stability of the HA (Fig. 3a) and showed no statistically significant differences in infection inhibition by NH₄Cl (Fig. 3b).

FIG 3.

Phenotypic analyses of VN-H5N1 and its single-point HA mutants. (a) Acid-induced conformational transition of HA. Viruses adsorbed in the wells of microtiter plate were incubated in buffers with pHs in the range from 4.8 to 7.2, followed by treatment with proteinase K. Binding of peroxidase-labeled fetuin to viruses was assayed, and values of pH that corresponded to 50% inactivation of HA binding activity (pH₅₀) were determined from binding-versus-pH curves. Two to four experiments were performed on different days with two replicates for each virus (n = 4 to 8). (b) Inhibition of viral infection by ammonium chloride. MDCK cells were infected in the presence of various concentrations of NH₄Cl, incubated overnight, fixed, and immunostained for NP. Concentrations of NH₄Cl that reduced numbers of infected cells by 50% (IC₅₀) were determined from dose-response curves. Two to three experiments were performed on different days with 1 to 4 replicates per virus (n = 6 to 7). (c) Inhibition of infection by Vibrio cholerae sialidase. MDCK cells were incubated with serial dilutions of sialidase, followed by addition of viruses, overnight incubation, fixation, and immunostaining for NP. Concentrations of sialidase that reduced numbers of infected cells by 50% were determined from dose-response curves. Three to four experiments were performed on different days with 3 replicates per virus (n = 9 to 12). All panels show mean values and SD. Asterisks indicate P values for differences in results with respect to those with VN-H5N1 (*, P < 0.05; **, P < 0.01; ***, P < 0.001).

To determine the effect of the substitutions on viral receptor-binding properties, we first compared the ability of IAVs to infect MDCK cells that were treated with Vibrio cholerae sialidase and as a result had reduced levels of receptors on the cell surface (Fig. 3c). A higher tolerance of the virus infection to desialylation (a higher value of sialidase 50% inhibitory concentration [IC₅₀]) was interpreted as an indication of a higher binding avidity (34). Among the fusion variants, H8Q and T315I did not differ from VN-H5N1, whereas the two other variants (H103Y and K58₂I) showed marginal alteration of binding avidity. In contrast, the three receptor-binding variants displayed significant reduction of binding avidity for MDCK cells.

We next compared binding of VN-H5N1 and the three receptor variants to a panel of synthetic sialylglycopolymers (Table 3), all of which contained terminal Neu5Ac2-3Gal moieties but differed by the linkage to and substituents at the subterminal GlcNAc residue. VN-H5N1 bound 3′SLN and its sulfated analogues 6-Su-3′SLN and 6-Su-SLe^x more strongly than sialyl Lewis C (SLe^c) (Fig. 4). This binding profile is typical for the H5 and H7 viruses circulating in gallinaceous poultry (24, 35). In accord with our prediction, the substitution K189N markedly reduced virus binding to both sulfated receptors without affecting its binding to SLe^c. In addition, this substitution strongly decreased virus binding to nonsulfated 3′SLN and SLe^x. Of note, the binding pattern displayed by the K189N variant (preferential binding to SLe^c, no significant effect on binding of receptor sulfation, and negative effect of fucosylation) is typical for the IAVs with different HA subtypes circulating in ducks (24, 35). The substitution K218Q had a less pronounced effect than the substitution K189N on the virus binding profile. The K218Q mutant showed reduced binding to nonfucosylated analogues 3′SLN, SLe^c, and Su-3′SLN and increased binding to fucosylated receptors SLe^x and 6-Su-SLe^x. These observations confirm previous reports on the important role of the amino acid in position 218 for the recognition of fucose (24, 33, 35–38). The binding of the double mutant KN/KQ to all tested sialylglycopolymers (SGPs) was significantly lower than the binding of single mutants and below the detection limit of the assay. Neither VN-H5N1 nor its receptor-binding variants bound to receptor analogues containing terminal Neu5Ac2-6Gal moieties (data not shown).

FIG 4.

Binding of biotinylated sialylglycopolymers to VN-H5N1 and its HA mutants K189N, K218Q, and KN/KQ. (a) Primary binding data from one representative experiment (absorbency in the wells at 450 nm versus molar concentration of sialic acid moieties in solution). (b) Association constants of virus-SGP complexes calculated from three independent experiments performed on different days with one replicate per each virus-SGP pair (n = 3). The dotted line shows detection limit of the assay. Asterisks indicate P values for differences in results with respect to those with VN-H5N1 (*, P < 0.05; **, P < 0.01; ***, P < 0.001).

It has to be noted that positively charged amino acids were conserved in HA positions 189 and 218 of the H5N1 viruses isolated before 2008 and that substitutions K189N and K218Q accompanied emergence and global spread of the reassortant H5Nx viruses of the lineage 2.3.4.4 first identified in 2008 in China (for a review, see reference 39). These H5Nx viruses contained additional substitutions in the HA, among them S224R, which seemed to compensate for the negative effects on binding of substitutions in positions 189 and 218 (37). Thus, although the three receptor-binding mutants studied here were viable and replicated in MDCK cells, it must be considered that their reduced binding avidity may not fully reflect avidity of natural H5N1 isolates.

Figure 5 shows data on the infectivity of the HA mutants for HULEC. The fusion mutants displayed pronounced reduction of infectivity with respect to the parent VN-H5N1 (from 14-fold for H8Q to 70-fold for K58₂I), and the magnitude of the reduction closely correlated with the effect of these substitutions on the conformational stability of the HA. In contrast to the fusion mutants, the single-point receptor-binding mutants K189N and K218N were only 2-fold less infectious than VN-H5N1, and the double mutant KN/KQ showed approximately a 4-fold reduction of infectivity. Decreased infectivity of receptor-binding mutants could indicate that the poultry virus-like preference of the parent VN-H5N1 for 3′SLN and its sulfated analogues contributed to its high tropism to HULEC. However, a decreased infectivity of the receptor-binding variants could also be associated with their generally reduced binding avidity (Fig. 3c and 4).

FIG 5.

Infectivity of HA mutants of VN-H5N1 for HULEC. The cells were infected with 5 × 10⁴ FFU of each virus per culture, incubated for 18 h, fixed, and immunostained for NP. The numbers of infected cells per culture were counted and expressed in percentages with respect to cultures infected with the parent VN-H5N1. Data show mean values and SD from experiments with cultures prepared from cells of three different donors, each tested in triplicate (n = 9). All P values for the differences in results with respect to those with VN-H5N1 are lower than 0.0001. Values below the strain names show data on HA conformational stability (pH₅₀) and sensitivity of the virus infection to V. cholerae sialidase (IC₅₀ V. Chol.), determined in the experiments presented in Fig. 3a and c, respectively.

Collectively, our data on infectivity of VN-H5N1 point mutants (Fig. 5) together with the data on infectivity of recombinant viruses with wild-type HAs (Fig. 1) indicate that the conformational stability of HA represents the principal determinant of HA-dependent tropism of IAVs to HULEC and that receptor-binding properties of the viruses, including recognition of terminal Neu5Ac2-6Gal and Neu5Ac2-3Gal moieties and penultimate sulfated and fucosylated residues, play a less important role. We next aimed to understand which properties of HULEC determine differences in their susceptibilities to IAVs depending on HA stability.

IFITM3-mediated restriction of IAVs in HULEC depends on conformational stability of the HA.

Sun and colleagues found that infectivity of human IAVs is restricted in HULEC at the posthemifusion stage(s) of virus entry into cells (23). The restriction correlated with a relatively high constitutive expression of antiviral IFITM3 protein, and knockdown of IFITM3 by a small interfering RNA (siRNA) partially rescued susceptibility of cells to IAVs. We found more recently that the sensitivity of IAVs to the antiviral activity of IFITM2 and IFITM3 proteins depends on the conformational stability of the HA (27). Given these notions, we decided to test whether IFITM3-mediated restriction is responsible for the observed reduced infectivity for HULEC of IAVs with pH-stable HA compared to that of the pH-unstable counterparts (Fig. 1 and 5). To this end, we first compared basal expression of IFITM3 in polarized HULEC cultures and in two human cell lines of epithelial origin, A549 and Calu-3 (Fig. 6a). IFITM3 was readily detected by Western blotting in HULEC, whereas its expression was below the detection limit in the epithelial cells. Stimulation of A549 and Calu-3 cells with beta interferon (IFN-β) induced strong expression of IFITM3 that was significantly higher than the basal IFITM3 expression in HULEC. These results aligned with the previous report of elevated constitutive expression of IFITM3 in nonpolarized HULEC compared to that in epithelial cells (23).

FIG 6.

Basal expression of IFITM3 in HULEC and effects of AmpB and IFN pathway inhibitors on infectivity of IAVs. (a) Levels of IFITM3 in cell lysates were analyzed by immunoblotting using antibodies against IFITM3 and GAPDH (loading control) as described in Materials and Methods. Replicate cultures of A549 and Calu-3 cells were incubated for 24 h with 100 U/ml of IFN-β in the growth medium or without IFN-β. Cultures of HULEC from two different donors were used without IFN treatment. Pictures were made using ChemiDoc with Image Lab, version 5.2, software. (b) HULEC were preincubated with either ruxolitinib (Rux), TPCA-1 (TPCA), or pyridine 6 (P6) for 24 h or with AmpB for 1 h. Duplicate treated cultures and nontreated control cultures (Contr) were infected with 5 × 10³ FFU of VN-H5N1 and K189N and 5 × 10⁴ FFU of K58₂I, incubated for 18 h, fixed, and immunostained for NP. The numbers of infected cells per culture were counted and expressed with respect to infection in control nontreated cultures. Data shown are mean values and SD (n = 2). Asterisks indicate statistically significant differences between results for treated cultures and those of control cultures (*, P < 0.05; **, P < 0.01).

To assess the effect of basally expressed IFITM3 on susceptibility of HULEC to IAVs, we determined infectivity of VN-H5N1 and its variants K189N and K58₂I in the presence of amphotericin B (AmpB), an antifungal compound that counteracts IFITM3-mediated restriction of IAVs (40–43). AmpB enhanced infectivity of the viruses (Fig. 6b). The magnitude of the effect correlated with the HA conformational stability, the enhancement being much higher in the case of the pH-stable variant K58₂I than of the less stable VN-H5N1 and K189N. This result demonstrated that differential tropism of IAVs to HULEC depends, at least in part, on a stronger IFITM3-mediated restriction of IAVs with a pH-stable HA.

Elevated basal expression of IFITM3 in HULEC may be IFN independent. However, it may also result from either constitutive expression of IFN by HULEC or induction of IFN by components of the growth medium. To test the potential role of IFN in enhanced IFITM3 expression in HULEC, we treated the cultures for 24 h prior to infection with the inhibitors TPCA-1, ruxolitinib, and pyridine 6 targeting components of IFN induction and signaling pathways (44, 45). The inhibitors had minor if any effect on susceptibility of HULEC to viral infection (Fig. 6b), suggesting that elevated basal expression of IFITM3 is independent of IFN.

Relatively weak endosomal acidification, expression of IFITM3, and elevated extracellular pH limit infectivity of pH-stable IAVs in HULEC.

Growing experimental evidence suggests that infectivity of IAVs with pH-stable HA may be reduced in cells having a relatively high endosomal pH (31, 46–49). We therefore decided to test whether insufficient endosomal acidification, in addition to IFITM3, contributed to reduced susceptibility of HULEC to some IAVs. To address this question, we first characterized kinetics of endosomal acidification in HULEC. MDCK cells were used as a comparison as these cells are highly susceptible to a variety of IAVs and do not seem to discriminate between viruses differing by the pH optimum of fusion. The endosomes of HULEC and MDCK cells were pulse-loaded with a ratiometric mixture of fluorescein isothiocyanate (FITC)- and tetramethylrhodamine isothiocyanate (TMR)-labeled dextrans, the ratio of pH-sensitive green fluorescence and pH-insensitive red fluorescence (G/R) was determined using confocal microscopy, and the endosomal pH at different time points was calculated based on G/R versus pH calibration curves. The pH in endosomes of MDCK cells gradually decreased from 6.4 to 5.1 over the observation period from 5 min to 50 min postloading (Fig. 7a), thus covering the pH range required for acid-induced initiation of membrane fusion by most IAVs from different host species (for review, see reference 16). The endosomal pH in HULEC was significantly less acidic than that in MDCK cells and reached 5.5 only at the end of the observation period. This level of acidification in HULEC should be sufficient to induce HA conformational transition and fusion of IAVs with relatively unstable HAs, such as VN-H5N1 and its receptor-binding mutants (Fig. 3), but may be suboptimal for the fusion of more stable viruses. To corroborate results of direct endosomal pH measurements in HULEC and MDCK cells, we compared inhibition of viral infection in these cells by ammonium chloride (Fig. 7b). The infectivity of VN-H5N1 was significantly more sensitive to inhibition by NH₄Cl in HULEC than in MDCK cells, indicating that endosomal acidification is low in HULEC.

FIG 7.

Endosomal pH dynamics and inhibition of viral infection by ammonium chloride in HULEC and MDCK cells. (a) The cells grown in six-chamber microscopic slides were allowed to take up a mixture of pH-sensitive FITC-dextran and pH-insensitive TMR-dextran, and the ratio of FITC-to-TMR fluorescence intensity (G/R) of endosomes at 37°C was monitored over time using confocal laser microscopy, as described in Materials and Methods. The pH values were calculated from the G/R-versus-pH calibration curve generated using dextran-loaded cells incubated in pH-adjusted buffers supplemented with ionophores. Data shown are mean values and 95% confidence intervals from the experiment with HULEC from three different donors, with 8 to 10 microscopic fields per donor (n = 27). Seven microscopic fields were analyzed in two replicate MDCK cultures (n = 7). The P values for the differences between results for HULEC and those for MDCK cells at time points from 5 to 35 min are below 0.05. (b) Replicate HULEC and MDCK cell cultures were infected with 4 × 10⁴ FFU and 500 FFU per culture, respectively, of VN-H5N1 in the EGM-2MV-BSA medium in the presence of various concentrations of NH₄Cl. The cells were incubated for 8 h, fixed, and immunostained for NP. Numbers of infected cells per culture were counted and expressed in percentages with respect to cultures infected without NH₄Cl. Data show mean values and SD from one experiment with three replicate HULEC and 5 to 6 replicate MDCK cell cultures per condition. All P values for the differences between results in HULEC and those in MDCK cells in the presence of NH₄Cl are below 0.003.

To determine whether the relatively high endosomal pH of HULEC restricts IAV infection, we decided to facilitate acidification of endosomes by reducing the pH of the extracellular medium (Fig. 8a). The commercial basal medium EGM-2MV used for cultivation and infection of HULEC had a pH of 7.7. The infectivity of VN-H5N1 increased about 3.5-fold at pH 7.0 compared to that at pH 7.7 and then gradually decreased at pH 6.5, 6, and 5.5. In marked contrast to the infectivity of VN-H5N1, the infectivity of K58₂I was very low at pH 7.7, and it was enhanced by 2 orders of magnitude when the pH of the infection medium was lowered to 7.0. Furthermore, whereas the infectivity of VN-H5N1 reached its maximum at pH 7.0, the peak of infectivity of K58₂I was shifted to pH 6.0 to 6.5. Importantly, K58₂I and VN-H5N1 displayed comparable levels of infection at their distinctive pH optima of the extracellular medium, highlighting the critical role of pH in IAV tropism to HULEC. In contrast, reduction of extracellular pH had no effect on the infectivity of K58₂I and did not enhance the infectivity of VN-H5N1 in MDCK cells (Fig. 8c).

FIG 8.

Effects of extracellular pH and AmpB on infection in HULEC and MDCK cells. (a) HULEC cultures were infected with 5 × 10⁴ FFU of VN-H5N1 and K58₂I in standard EGM-2MV-BSA medium (pH 7.7) and in the same medium adjusted to different pH values. The inoculum was removed after 1 h, the cultures were incubated in standard EGM-2MV-BSA for 7 h, fixed, and immunostained for NP. Data shown are mean values and SD from one experiment with 2 to 3 replicate cultures. Asterisks indicate differences between results for the viruses. (b) HULEC cultures were infected with VN-H5N1 (10⁴ FFU), K189N (10⁴ FFU), and K58₂I (10⁵ FFU) using either standard or pH-adjusted EGM-2MV-BSA and processed as described above. AmpB (1 μM) was added to some of the cultures at 1 h prior to infection and maintained in the medium during the whole experiment. Data show the ratio of infected cells in low pH- and/or AmpB-treated cultures and in cultures infected at pH 7.7 without AmpB. Asterisks indicate differences in results with respect to those of the cultures infected at pH 7.7 without AmpB. (c and d) MDCK cells were infected, processed, and analyzed as described for panels a and b, using 300 FFU of each virus per culture and pH-adjusted DMEM-BSA instead of EGM-2MV-BSA. Data shown are mean values and SD from one experiment with 4 replicate cultures per condition (n = 4). Asterisks in panel d indicate differences in results with respect to results with infection at pH 8 without AmpB (*, P < 0.05; **, P < 0.01; ***, P < 0.001).

To confirm these observations and to dissect the effects of pH- and IFITM3-mediated restrictions, we studied infection in HULEC using the standard infection medium (pH 7.7) and medium adjusted to pH 6.5, each with and without the addition of AmpB. Figure 8b shows enhancement of viral infectivity under different conditions with respect to infectivity in cultures infected at pH 7.7 without AmpB. In agreement with the previous experiment, the infectivity of viruses with pH-unstable HA, VN-H5N1 and K189N, was comparable at pH 7.7 and pH 6.5, whereas K58₂I infected cells much more efficiently at pH 6.5 than at pH 7.7. The addition of AmpB increased the infectivity of VN-H5N1 and K189N from 4- to 5-fold at either pH, indicating that reduction of extracellular pH does not affect the inhibition of these IAVs by IFITM3. In contrast, AmpB increased the infectivity of K58₂I more strongly at pH 7.7 than at pH 6.5. A combination of mild acidification of the infection medium and AmpB treatment resulted in the highest levels of HULEC infection by this pH-stable virus. In MDCK cells, the pH of the medium, the addition of AmpB, and a combination of these conditions had only a minor, if any, effect on the infectivity of each virus.

From these results we concluded that a combination of cell-type-specific factors, including relatively weak endosomal acidification, elevated endogenous expression of IFITM3, and slightly alkaline extracellular cultivation medium, is responsible for the differences in HA-dependent tropism of IAVs to HULEC.

DISCUSSION

Previous studies revealed that the infectivity of zoonotic H5N1 and H7N9 IAVs for HULEC was an order of magnitude higher than the infectivity of other human and avian IAVs tested and that the differences in infectivity correlated with the viral HA (19, 23). Our study was initiated to understand which characteristics both of the HA and of the cells determine these differences. Here, we found that the conformational stability of the HA is the major determinant of IAV tropism to HULEC and that IAVs with pH-stable HA are strongly restricted in HULEC owing to a combined effect of endogenous IFITM3, relatively inefficient endosomal acidification, and elevated pH of cultivation medium.

Our data on relatively high basal expression of IFITM3 and its contribution to IAV restriction in HULEC agree with the observations made by Sun and colleagues (23). In our experiments, counteraction of IFITM3 by AmpB enhanced infectivity of pH-stable K58₂I virus to a significantly higher extent than that of pH-unstable variants VN-H5N1 and K189N (Fig. 6b and 8b). The exact mechanisms by which IFITMs inhibit cell entry by IAVs and other sensitive viruses are not well defined (for reviews, see references 25 and 26). IFITM2 and IFITM3 mainly localize to late endosomes and lysosomes, but neither inhibits viral entry into these compartments or their acidification. One plausible model suggests that accumulation of IFITM3 at the sites of IAV fusion in the late endosomes is a prerequisite for its antiviral activity and that IFITM3 traps viral fusion at a hemifusion stage by reducing fluidity and changing the curvature of the membrane (43, 50). Based on this model, it is assumed that AmpB counteracts IFITM3 by affecting membrane fluidity (43). In addition to directly inhibiting viral fusion, IFITM2 and IFITM3 can reduce efficiency of virus penetration by promoting trafficking of late-endosomal cargo to lysosomes for degradation (51). Either mechanism of IFITM-mediated restriction would be consistent with a lower sensitivity to IFITM3 of IAVs with pH-unstable HA observed here and in our previous study (27). Indeed, a higher pH optimum of HA conformational transition could facilitate fusion and penetration of IAVs from less acidic IFITM3-deficient early and maturing endosomes prior to their merging with IFITM3-competent vesicles.

We found that in addition to a high basal level of IFITM3, the endosomes are less acidic in HULEC than in MDCK cells. A link between differential susceptibility of some continuous cell lines to IAVs and levels of endosomal pH in these cells has been described in previous reports (31, 47, 49). Thus, Vero, A549, and Raw264.7 cells were found to be less susceptible to IAVs with pH-stable HA than to pH-unstable IAVs whereas no apparent differences in the infectivities of all tested viruses were observed in MDCK cells. This pattern correlated with a higher endosomal pH in Vero, A549, and Raw264.7 cells than in MDCK cells. A similar interplay between endosomal pH, pH stability of the HA, and infectivity of IAVs was observed by Daidoji and colleagues in their study on infectivity of a broad panel of IAVs for different clones of human airway epithelial cells transformed with the simian virus 40 large T-antigen (48). Collectively, these reports suggested that infectivity of IAVs with pH-stable HAs, such as human-adapted pandemic and seasonal IAVs, is reduced in cells with relatively inefficient endosomal acidification, such as Vero, A549, and Raw264.7, presumably owing to the reduced competence of these cells to trigger HA-mediated membrane fusion. It should be mentioned that none of these studies determined expression and antiviral activity of cellular IFITMs. Based on our results, we speculate that basal IFITMs could contribute in some of the cell lines tested to the observed HA stability-dependent differences in IAV infectivity. Furthermore, we predict that a relatively high endosomal pH potentiates the antiviral effect of IFITM3 by delaying virus fusion and penetration and increasing the chance of virus encounter with IFITM3 later in the endocytic pathway. Thus, in our experiments on the combined effects of AmpB and extracellular pH, counteraction of IFITM3 by AmpB enhanced infectivity of K58₂I much more strongly at pH 7.7 than at pH 6.5 (Fig. 8b). This result indicates that elevation of the initial pH in endosomes of HULEC enhances the antiviral effect of IFITM3.

Romanova and colleagues investigated the effects of HA stability on replication of IAVs in cell culture and on immunogenicity and quality of influenza vaccines. They noticed that pH-stable primary human isolates of H1N1 and H3N2 IAVs displayed a higher infectivity for Vero cells and primary human airway epithelial cells at pH 5.6 to 5.8 than at pH 7.5 (29, 46, 52). The mechanisms of this phenomenon remained unexplored. More recently, Desai et al. studied cell-type-specific factors controlling virus-endosome fusion of avian sarcoma and leukosis retrovirus (53). In particular, they found that endosomal acidification is less robust in A549 cells than in CV-1 cells and that a modest elevation of extracellular pH increased the pH in early endosomes in a cell-type-dependent manner. This delayed acid-induced fusion of endocytosed virus in A549 cells but not in CV-1 cells. Together, these reports corroborate and explain our observation of a remarkably strong effect of extracellular pH on tropism of IAVs to HULEC but not to MDCK cells (Fig. 8a and c). For example, HULEC were almost completely resistant to the virus with pH-stable HA at mildly alkaline pH (7.7), while having at least 1,000-fold higher susceptibility to the pH-unstable counterpart. The difference in infectivities between these two IAVs was smaller but still substantial at pH 7. We conclude that, owing to insufficiently robust endosomal acidification in HULEC, these cells cannot promptly acidify neutral or mildly alkaline extracellular medium in endosomal vesicles to the levels required to trigger conformational transition of pH-stable HA. This effect reduces virus penetration from early/maturing endosomes and increases the number of virions inhibited by IFITM3 in late endosomes and/or degraded in lysosomes. Furthermore, this effect can be alleviated and the infectivity of IAVs can be increased by either lowering extracellular pH or decreasing the conformational stability of HA (Fig. 8a). Together with other reports (29, 46, 52, 53), our results emphasize that the pH and the buffering capacity of the extracellular medium may affect the infectivity of IAVs and that the magnitude of the effect depends on the parameters of the cellular endocytic system as well as on the conformational stability of the viral HA. This notion must be considered in studies on mechanisms of virus replication in vitro and for optimization of virus propagation in cell culture. It is tempting to speculate that the extracellular pH also has an effect on IAV tropism and pathogenicity in vivo. For example, whereas VN-H5N1 was more infectious for HULEC than the K58₂I variant at a pH above 7, the opposite pattern was observed after acidification of the extracellular medium to pH 6.5 owing to a simultaneous drop of infectivity of VN-H5N1 and an increase of infectivity of K58₂I (Fig. 8a). In humans, the average extracellular pH appears to be lower in the nasal and tracheo-bronchial epithelium than in the lung (6.3 and 7.0 versus 7.5, respectively) (reviewed in reference 16). These differences in extracellular pH correlate with preferential replication of pH-stable human IAV viruses in the upper respiratory tract and with the lung tropism of pH-unstable zoonotic H5N1 and H7N9.

Replication in endothelial cells is a hallmark of the fatal systemic disease of gallinaceous poultry caused by highly pathogenic (HP) avian IAVs with HA subtypes H5 and H7 (54, 55). Infection of endothelial cells also plays an important role in the pathogenesis of the HP H5N1 IAVs in cats, mice, and ferrets (56, 57). Proteolytic activation of the HA by ubiquitous cellular proteases is essential for productive replication of HP IAVs in the endothelium of these avian and mammalian species. The results of our study suggest that reduced conformational stability of the HA typical for HP avian IAVs viruses (16, 58) may represent another essential factor of their endotheliotropism and high pathogenicity in birds and mammals. It is currently believed that lung endothelial cells in humans become activated during infection of epithelial cells and contribute to influenza virus pathogenesis by secreting cytokines and chemokines, promoting the recruitment of neutrophils and macrophages and causing adhesion of platelets (reviewed in references 56 and 59). Postmortem analyses of autopsy samples from influenza patients are limited, and findings of infected endothelial cells in these samples are rare (17, 18). Thus, the extent of HULEC infection in vivo by zoonotic H5N1 and H7N9 IAVs and its role in disease severity remain unexplored.

Receptor-binding specificity of the HA has long been known to play a pivotal role in cell tropism, replication, and pathogenicity of IAVs in humans (7–9). The role of the conformational stability and of the membrane fusion properties of HA in human IAV infections is less well understood. Our study provides the first example of stability-dependent differential tropism of IAVs to primary human cells. As HA stability varies significantly among animal, human, and zoonotic IAVs, further studies are warranted on the stability-dependent tropism of IAVs to different types of primary human cells in vitro and in vivo.

MATERIALS AND METHODS

Cells.

All cells were propagated at 37°C in 5% CO₂. MDCK, 293T, A549, and Calu-3 cells were grown in Dulbecco’s modified Eagle medium (DMEM; Gibco) supplemented with 10% fetal calf serum (FCS; Gibco), 100 IU ml⁻¹ penicillin and 100 μg ml⁻¹ streptomycin (Pen-Strep), and 2 mM glutamine.

Cryopreserved primary human lung microvascular endothelial cells from three different donors were purchased from Lonza. The cells were passaged once in plastic flasks in the EGM-2MV growth medium (Lonza) composed of endothelial cell basal medium EBM-2 and an EGM-2MV SingleQuots kit (growth factors and fetal calf serum). The passage 1 cells were stored in aliquots in liquid nitrogen. Thawed passage 1 cells were propagated and polarized on semipermeable membrane supports as described by Chan et al. (20). In brief, the cells were seeded in the apical compartment of 12-mm Transwell-Clear supports (pore size, 0.4 mm; Corning) and maintained in EGM-2MV medium, which was changed in both apical and basolateral compartments every other day. Ten- to 15-day-old HULEC cultures were used for the experiments.

Plasmids and recombinant viruses.

A set of pHW2000 plasmids containing gene segments of A/PR/8/34 (H1N1) (PR8) and two pHW2000 plasmids containing HA and NA gene segments of A/Vietnam/1203/2004 (H5N1) were kindly provided by Richard Webby and Robert Webster, St. Jude Children’s Research Hospital, Memphis, TN, USA. The HA of the H5N1 virus contained a deletion of the polybasic cleavage site introduced by mutagenesis. Mutations introduced in the HA plasmid of A/Vietnam/1203/2004 in this study were made using a site-directed mutagenesis kit (QuikChange, Stratagene).

The 2:6 recombinant viruses containing HA and NA gene segments of A/Vietnam/1203/2004 and 6 other genes segments of PR8 were generated using eight-plasmid reverse genetics (60). In brief, a mixture of eight pHW2000 plasmids was transfected into 293T cells using Lipofectamine 2000 (Invitrogen). After 2 days, rescued viruses were amplified in MDCK cells, clarified by low-speed centrifugation, and stored in aliquots at –80°C. The identities of the HA and NA proteins of the viruses were confirmed by sequencing.

The 2:6 recombinant viruses that shared six gene segments of PR8 and contained HAs and NAs of the wild-type avian, human, and zoonotic viruses were prepared and characterized previously (27). They are listed in the Table 1.

Virus titration in MDCK cells.

Viral titers were determined as described previously (61). In brief, MDCK cells in 96-well plates were infected with 0.1 ml of serial 10-fold dilutions of viruses in DMEM containing 0.1% bovine serum albumin (BSA; PAA Laboratories), Pen-Strep, and 2 mM glutamine (DMEM-BSA). No trypsin was added to the medium in order to limit the infection to one replication cycle. The cultures were incubated overnight, fixed with 4% paraformaldehyde (PFA), and immunostained for viral NP. Numbers of infected cells per well were counted under the microscope for the virus dilution that produced from 30 to 300 infected cells per well and recalculated into numbers of focus-forming units (FFU) per milliliter of the original undiluted virus suspensions.

Plaque assay.

Virus titration in MDCK cells under semisolid overlay medium in the presence of trypsin and analysis of viral plaque size were performed as described previously (27).

Infection in endothelial cells.

The HULEC cultures on Transwell filters were washed twice with EGM-2MV medium, in which fetal serum was replaced by 0.1% BSA (EGM-2MV-BSA). The apical compartments of the replicate cultures were inoculated with 5 × 10⁴ FFU of the viruses (based on titers determined in MDCK cells) in 0.2 ml of EGM-2MV-BSA. The same medium without the virus was added to the basolateral compartments (0.5 ml of per culture). The cultures were incubated for 18 h, fixed with 4% PFA, immunostained for viral NP, and analyzed under the microscope as described previously (62). If fewer than 300 cells per culture became infected, all infected cells in each replicate culture were counted. In the case of higher infection levels, the number of infected cells was counted in the epithelial segments that included all consecutive microscopic fields along the diameter of the culture (segment surface area, 3 mm²). Three to four such segments per each replicate culture were counted; the results were averaged and used to calculate the total number of infected cells in this culture.

Effects of AmpB and inhibitors of IFN response.

Antimycotic amphotericin B (AmpB; Sigma) and the inhibitors of IFN induction and signaling pathways TPCA-1, ruxolitinib (both from Selleckchem), and pyridine 6 (Calbiochem) were prepared as 10 mM stock solutions in dimethyl sulfoxide (DMSO). HULEC were incubated with inhibitors added to the growth medium in both compartments to final concentrations of 1 μM (TPCA-1, ruxolitinib, and AmpB) and 2.5 μM (pyridine 6). The IFN inhibitors and AmpB were added 24 h and 1 h before the infection, respectively. Growth medium in both compartments was replaced before the infection by EGM-2MV-BSA medium containing the same concentrations of the inhibitors that were used for preincubation. The cultures were infected by adding 5 × 10⁴ FFU of viruses to the apical compartments. The cells were fixed at 18 h postinfection, and the numbers of infected cells in each replicate culture were determined as described above.

Effects of extracellular pH.

The apical compartments of washed HULEC cultures on Transwell filters were inoculated with viruses in the EGM-2MV-BSA medium adjusted to different pH values. Medium with the same pH but without the virus was added to the basolateral compartment. After 1 h, the inoculum was removed, and the medium in both compartments was replaced by standard EGM-2MV-BSA (pH ∼7.7 to 7.8). The cultures were incubated for 7 h, fixed, immunostained, and analyzed for numbers of infected cells as described above. Similar experiments were performed in MDCK cells in 96-well plates using standard and pH-adjusted DMEM-BSA. In some of the experiments in both HULEC and MDCK cells, 1 μM AmpB was added to the growth medium 1 h before the infection and maintained in the infection medium (IM) during the whole experiment.

Low-pH-induced conformational transition of HA.

Alteration of the HA sensitivity to protease digestion that accompanies acid-induced conformational transition was quantified as described previously (34). In brief, viruses adsorbed in the wells of microtiter plates were incubated in either phosphate-buffered saline (PBS) or 0.1 M morpholineethanesulfonic acid (MES) low-pH buffers for 10 min at 37°C, followed by washing with PBS and incubation with 0.05 mg/ml of proteinase K in PBS for 1 h at 37°C. After the washing step, binding to the viruses of peroxidase-labeled fetuin was determined and expressed in percentages of low-pH-exposed virus with respect to the PBS-exposed control. Binding-versus-pH curves were plotted, and pH values that corresponded to HA inactivation by 50% (pH₅₀) were determined by linear interpolation.

Inhibition of viral infection by ammonium chloride.

The viral infection inhibition assay was described previously (63). MDCK cells in 96-well plates were infected in the presence of various concentrations of NH₄Cl, incubated overnight, fixed, and immunostained for viral NP. Concentrations of NH₄Cl that reduced numbers of infected cells by 50% (50% inhibitory concentration [IC₅₀]) were determined from dose-response curves by linear interpolation.

In the experiments on infection inhibition in HULEC (Fig. 7b), EGM-2MV-BSA medium was used instead of DMEM-BSA and the cultures were fixed at 8 h postinfection.

Inhibition of infection by Vibrio cholerae sialidase.

Viruses were compared for their sensitivities to gradual desialylation of receptors on cells as described previously (34). In brief, MDCK cells in 96-well plates were incubated with 0.05 ml per well of serial dilutions of sialidase in DMEM-BSA for 30 min at 37°C. A total of 200 FFU of the viruses in 0.05 ml of IM was added per well without removing sialidase. The cultures were incubated overnight, fixed, and immunostained for viral NP. Concentrations of sialidase that reduced numbers of infected cells by 50% (IC₅₀) were determined from dose-response curves by linear interpolation.

Detection of IFITM3 by Western blotting.

Six to 12 HULEC cultures grown on Transwell-Clear supports for 10 to 15 days were washed with ice-cold PBS and lysed in 20 mM Tris-buffered saline containing 5 mM EDTA and 2% Triton X-100 for 30 min on ice. Confluent A549 cells and Calu-3 cells grown in 12-well plates were incubated for 24 h with or without addition of 100 U/ml of recombinant IFN-β (Betaferon; Schering) to the growth medium. IFN-treated and control nontreated cells were lysed as described above for HULEC cultures. All cell lysates were clarified by centrifugation, supplemented with sample dye buffer containing β-mercaptoethanol, heated for 5 min at 95°C, and separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (12% gel) under reducing conditions. Proteins were transferred to a polyvinylidene difluoride membrane (GE Healthcare) by semidry electroblotting. IFITM3 and glyceraldehyde 3-phosphate dehydrogenase (GAPDH; loading control) were detected using rabbit anti-IFITM3 (Abcam) and mouse monoclonal anti-GAPDH antibodies (Santa Cruz Biotechnology), peroxidase-labeled secondary antibodies (Dianova), and enhanced chemiluminescence (ECL) peroxidase substrate (Pierce). Immunostained blots were analyzed using a ChemiDoc reader with Image Lab, version 5.2, software.

Receptor-binding assay.

Receptor-binding specificity of the viruses was characterized using soluble synthetic sialylglycopolymers (SGPs) (GlycoNZ, Auckland, New Zealand), kindly provided by Nicolai Bovin and Alexander Tuzikov, Shemyakin and Ovchinnikov Institute of Bioorganic Chemistry, Moscow, Russia. The SGPs contained 20 mol% specific sialyloligosaccharide (Table 3) and 5 mol% biotin attached to the 30-kDa poly-N-(2-hydroxyethyl)acrylamide backbone. The binding of the viruses to SGPs was determined in a direct solid-phase assay as described in detail elsewhere (64). In brief, viruses adsorbed in the wells of 96-well enzyme immunoassay plates were allowed to interact with serial dilutions of SGPs, followed by incubation with peroxidase-labeled streptavidin. Peroxidase activity was assayed using tetramethylbenzidine substrate solution. The data were transferred to a personal computer and processed using Microsoft Excel software. The apparent association constants of virus complexes with SGPs were determined from the slopes of the Scatchard plots (A₄₅₀/C versus A₄₅₀), where C is the concentration of the sialic acid in solution and A₄₅₀ is the absorbency in the corresponding well.

Endosomal pH measurement.

The endosomal pH in HULEC and MDCK cells was determined by confocal microscopy using pH-sensitive ratiometric fluorescent probes as described by Verkman and colleagues (65, 66) with modifications. In brief, the cells grown in six-chamber plastic slides (μ-Slide VI; Ibidi GmbH) were placed in temperature-controlled holder set on the stage of a Leica SP8 confocal laser scanning microscope with a 20× objective (numeric aperture, 0.75; zoom, ×3.5). The cells were washed with PBS containing 0.5 mM CaCl₂ and 0.9 mM MgCl₂ (PBS+) and overlaid with solution containing 2.5 mg/ml of fluorescein isothiocyanate (FITC)-dextran (40 kDa; Merck) and 7.5 mg/ml of tetramethylrhodamine isothiocyanate (TMR)-dextran (40 kDa; Merck) in PBS+. The cells were incubated for 5 to 7 min at 37°C, washed five times with ice-cold PBS+ containing 2% BSA, and overlaid with warm PBS+ to raise the temperature. Images of 5 to 10 microscopic fields for each cell type were acquired over a time period from 5 to 50 min with 5-min intervals using a 488-nm laser and 552-nm laser for green (FITC) and red (TMR) fluorescence, respectively, with corresponding filters. To prepare in situ calibration curves (FITC/TMR fluorescence ratio versus pH), the cells were loaded with dextrans and overlaid with buffers containing 120 mM KCl, 20 mM NaCl, 1 mM CaCl₂, 1 mM MgCl₂, 10 mM HEPES, 10 mM MES, nigericin, valinomycin, and monensin (20 μM each) with pH values adjusted to 4.5, 5.6, 6.1, 6.5, and 7.6. After incubation for 40 min to equilibrate extracellular and intracellular pH, the images of 5 to 10 fields for each pH value were acquired at a single time point.

The image analysis was performed using Imaris version 8.3.1 software (Bitplane/Oxford Instruments, Switzerland). The spot module of the software first identified endosomes in the red channel as spherical objects by calculating the mean intensity of high-value pixels (9 pixels together are defined as an endosome). Next, the mean intensity of each spot (same x-y pixels) in the green channel were chosen. The mean intensity of the background was determined for each channel in the cell area without endosomes using the surface module of Imaris version 8.3.1. After background subtraction, green-to-red intensity (G/R) ratios were calculated for each endosome and used to calculate mean G/R for each replicate microscopic field. The G/R-versus-pH calibration curves were built using a linear regression model.

Statistics.

Data are shown as mean values and standard deviations (SD) of biological replicates (the details are explained in the table footnotes and figure legends). The statistical significance of the differences between groups was estimated with a two-tailed unpaired Student's t test, and P values are indicated in the figure legends.