The pH of Activation of the Hemagglutinin Protein Regulates H5N1 Influenza Virus Replication and Pathogenesis in Mice

Abstract

After receptor binding and internalization during influenza virus entry, the hemagglutinin (HA) protein is triggered by low pH to undergo irreversible conformational changes that mediate membrane fusion. To investigate how mutations that alter the activation pH of the HA protein influence the fitness of an avian H5N1 influenza virus in a mammalian model, we infected C57BL/6J or DBA/2J mice and compared the replication and virulence of recombinant A/chicken/Vietnam/C58/04 (H5N1) HA-Y23₁H mutant, wild-type, and HA-H24₁Q and HA-K58₂I mutant viruses that have HA activation pH values of 6.3, 5.9, 5.6, and 5.4, respectively. The HA-Y23₁H mutant virus was highly susceptible to acid inactivation in vitro and was attenuated for growth and virulence in mice, suggesting that an H5N1 HA protein triggered at pH 6.3 is too unstable for the virus to remain fit. Wild-type and HA-H24₁Q viruses were similar in pathogenicity and grew to similar levels in mice, ducks, and cell cultures derived from both avian and mammalian tissues, suggesting that H5N1 HA proteins triggered at pH values in the range of 5.9 to 5.6 broadly support replication. The HA-K58₂I mutant virus had greater growth and virulence in DBA/2J mice than the wild type did, although the mutant virus was highly attenuated in ducks. The data suggest that adaptation of avian H5N1 influenza virus for infection in mammals is supported by a decrease in the HA activation pH to 5.4. Identification of the HA activation pH as a host-specific infectivity factor is expected to aid in the surveillance and risk assessment of currently circulating H5N1 influenza viruses.

INTRODUCTION

Highly pathogenic avian influenza (HPAI) H5N1 viruses were first detected in geese in 1996 in Guangdong Province, China. In 1997, Hong Kong reported the first human outbreak of H5N1 influenza, which caused six deaths (1). Since 2003, H5N1 influenza viruses have spread across Asia and into Europe and Africa (2), causing 360 deaths in 610 reported human cases as of 17 December 2012 (http://www.who.int/influenza/human_animal_interface/en/). H5N1 has become endemic in domestic poultry in Indonesia and Egypt, causing large economic losses (3, 4). Surveillance studies suggest that currently circulating H5N1 viruses may lack the ability to be transmitted efficiently between humans (4, 5). Nevertheless, H5N1 remains a pandemic threat, as H5N1 viruses continue to circulate in domestic poultry, frequently infecting humans. Recently, H5 influenza viruses have been shown to be capable of acquiring airborne transmissibility in ferrets (6–8), highlighting the potential threat of circulating H5 viruses. For surveillance, risk assessment, and preventive control measures directed toward HPAI viruses of the H5N1 subtype, there is an urgent need to understand the molecular properties required for replication and pathogenesis in mammalian hosts.

The replication efficiency, pathogenicity, and transmissibility of influenza viruses depend on multiple viral genetic and host factors (9). The present study focused on the hemagglutinin (HA) protein, which binds receptors and mediates viral-cellular membrane fusion during viral entry and is the major antigenic target during infection (10, 11). The HA protein is a trimeric class I membrane fusion protein (11, 12) that contains in its ectodomain a membrane-proximal, metastable stalk domain capped by a membrane-distal receptor-binding domain (RBD) (13, 14). The HA protein is primed for membrane fusion activity, and consequently infectivity, by posttranslational cleavage of the HA0 precursor into the fusion-capable HA1-HA2 complex (11). Intracellular furin-like proteases can cleave the polybasic cleavage sites of some H5 and H7 HA proteins, enabling systemic virus spread and enhancing the virulence of these highly pathogenic avian influenza (HPAI) viruses (15–17). Infection by influenza viruses is initiated when the HA surface glycoprotein binds sialic acid-containing receptors on the surface of the host cell. The receptor-binding specificity of the HA protein has been shown to be a major determinant of the host range, tissue tropism, pathogenicity, and transmissibility of influenza viruses (18). Currently circulating H5N1 influenza viruses have HA proteins that tend to bind preferentially to α(2,3)-linked sialosides and thus are poorly adapted for growth in the upper respiratory tracts of humans (19). Alternatively, human-adapted influenza viruses tend to bind preferentially to α(2,6)-linked sialosides that are predominant in the human upper respiratory tract (16, 20). A switch from α(2,3) receptor binding specificity to α(2,6) receptor binding specificity is generally thought to be a necessary, but not necessarily sufficient, step in the adaptation of avian influenza viruses for efficient growth in the upper respiratory tracts of mammals and airborne transmissibility (6, 7, 21).

After binding to cellular receptors, influenza viruses are internalized by endocytosis. As the pH is progressively decreased, a threshold is eventually reached at which the metastable HA surface protein is triggered to undergo irreversible structural changes that facilitate fusion of the viral envelope with the endosomal membrane (22, 23). The HA proteins from different strains and subtypes can vary in their activation pH values, which range from approximately 4.6 to 6.0 (24). The HA proteins from HPAI viruses tend to have activation pH values near the higher end of the range, toward 6.0, whereas those from human seasonal viruses tend to have lower pH values, nearer to 5.0. For a limited sampling of H5N1 influenza virus isolates, the HA activation pH has been measured to range from 5.3 to 5.9 (7, 25–27). For H1, H3, and H7 influenza viruses, mutations that alter the HA activation pH have been associated with changes in virulence in mice (28–31). For experimental infection of H5N1 influenza viruses in ducks and chickens, the highest levels of replication and pathogenesis appear to correlate with HA activation pH values that range between 5.6 and 6.0, while HA activation pH values lower than 5.6 have been found to attenuate replication and pathogenesis (25, 26, 32). In contrast, the replication of attenuated or reassorted H5 viruses in the upper respiratory tracts of mice and ferrets was enhanced by mutations that lowered the activation pH of the H5 HA protein to 5.6 or lower (7, 34).

Knowing the factors and molecular signatures that govern the efficient growth of a virus in one host species, tissue, or cell culture versus another is of fundamental importance in viral infectious disease. Such an understanding is an essential requirement to effectively conduct surveillance, perform risk assessments of viruses, make decisions to cull animals or quarantine humans, develop therapeutics that alleviate pathogenesis, identify and validate suitable drug targets, decide which virus seed stocks to prepare, efficiently and rapidly produce vaccines, and even decide which avenues of research are worthy of pursuit. The rationale for this and related studies is to understand how one such fundamental molecular property, the HA activation pH, governs the growth of H5N1 influenza virus in various species and cell types so as to benefit public health and agriculture in the aforementioned ways.

Here we investigated how the pH of activation of the HA protein regulates the replication and virulence of H5N1 influenza virus in mice. The wild-type (WT) virus selected for the present study was A/chicken/Vietnam/C58/04 (H5N1), a clade 1 influenza virus that has avian-virus-like α(2,3) receptor binding specificity and a polymerase poorly suited for replication in mammals (33) but was not engineered to be attenuated, reassorted, or mammal adapted, as had been done in previous studies (6, 7, 34). As a result of these molecular properties, the WT C58 H5N1 influenza virus does not cause weight loss, death, transmission (either contact or airborne), systemic spread, or robust nasal shedding in ferrets (33), further mitigating the risks involved in the use of the moderately pathogenic C58 strain for H5N1 research.

In the present study, mice were infected either with the WT C58 virus (HA activation pH of 5.9) or with a C58 virus containing a single point mutation in the HA1 subunit, HA-Y23₁H (HA activation pH of 6.3) or HA-H24₁Q (HA activation pH of 5.6), or in the HA2 subunit, HA-K58₂I (HA activation pH of 5.4) (26). These mutations in the HA stalk domain have been previously shown to alter the pH of activation of the C58 HA protein without altering HA protein expression, cleavage, or receptor-binding affinity (27), and viruses containing these mutations have replication rates in MDCK cells similar to that of the WT C58 virus (26). Another advantage of using the C58 viruses to study avian H5N1 infection in mice is that these same viruses were previously used to investigate replication, pathogenesis, and transmission in mallards (26). Therefore, the present results for infection in a mammalian model can be compared to those obtained for infection in an avian model. The results from the present study show that the C58 HA-Y23₁H mutant virus, which has an HA activation pH higher than that of the WT, is attenuated for replication and pathogenesis in mice, just as it was in ducks. In contrast, the C58 HA-K58₂I mutant virus, which has a decreased HA activation pH, promoted high levels of replication in the lungs and pathogenesis in mice despite being severely attenuated in ducks. The C58 HA-K58₂I mutant virus also replicated better in the murine nasal cavity than did the C58 WT virus, albeit to maximal levels that were relatively low, most likely because of an avian-like polymerase complex. Overall, the data from both the present study and a previous study (26) on the C58 viruses support the notion that a decrease in the activation pH of the HA protein that is detrimental to H5N1 replication in avian species may be necessary, but not sufficient, for adaptation to a mammalian host. Thus, this work provides evidence that the HA activation pH is an important molecular factor involved in the interspecies adaptation of highly pathogenic H5N1 influenza virus.

MATERIALS AND METHODS

Viruses.

Recombinant influenza viruses of the A/chicken/Vietnam/C58/04 (H5N1) strain (33) were generated by reverse genetics and characterized previously (26, 27). These viruses were C58 WT, C58 HA-Y23₁H (Y23H mutation in the HA1 subunit, H5 numbering), C58 HA-H24₁Q (H24Q mutation in the HA1 subunit, H5 numbering), and C58 HA-K58₂I (K58I mutation in the HA2 subunit, H3 and H5 numbering). All viruses were grown in eggs and plaque titrated in eggs and MDCK cells. All experiments with HPAI H5N1 viruses were conducted before the moratorium on avian influenza virus transmission research (35).

Biosafety and biosecurity.

All work with highly pathogenic H5N1 influenza virus was performed in an enhanced animal biosafety level 3 (ABSL-3+) laboratory that is select agent approved and routinely inspected by both institutional biosafety and USDA officials. The ABSL-3+ facility has entry and exit access control with both a card scanner and a biometric fingerprint reader. Personnel enter through a shower area and then take off all items and wear a scrub suit, a Tyvek suit, a disposable outer gown, gloves, and powered air-purifying respirators that HEPA filter the breathing air. All rooms are under negative air pressure, and there is a double-door autoclave and a double-HEPA-filtered air exhaust, and security cameras are placed throughout the laboratory. All in vitro work is performed in class II biosafety cabinets, and animal work is performed in negatively pressurized flexible-film isolators. All personnel are required to shower upon exit and comply with a quarantine policy to prevent outside contact with birds or immunocompromised hosts. Only personnel who receive training with H5N1 HPAI virus and who receive select agent security clearance can access the facility. ABSL-3+ personnel also receive annual refresher training to ensure adherence to regulations. Emergency plans are in place, and annual drills are performed to minimize biological risks and ensure personnel safety. The virus inventory is secured in locked freezers and is under constant security monitoring. The lab manager controls access to the virus inventory, and a logbook and database of all inventory are kept up to date. The ABSL-3+ laboratory is inspected biannually by the USDA, is in compliance with all USDA regulations, and meets or exceeds all standards outlined in Biosafety in Microbiological and Biomedical Laboratories, 5th edition (http://www.cdc.gov/biosafety/publications/bmbl5/BMBL.pdf).

Virus growth kinetics.

Multiple-step growth kinetics of the WT and mutant viruses were determined in the following cell lines: MDCK (Madin-Darby canine kidney), A549 (CCL-185, human lung carcinoma), NHBE (normal human bronchial epithelium), DF1 (CRL-12203, chicken embryo fibroblast), and CCL-141 (duck embryo fibroblast). Confluent monolayers of cells were infected with a multiplicity of infection (MOI) of approximately 0.01 PFU/cell (the PFU titer was determined in MDCK cells). After 1 h of incubation at 37°C, cells were washed twice with phosphate-buffered saline (PBS) plus calcium and magnesium (PBS+) to remove nonbound virus particles and reincubated at 37°C. Culture supernatants were collected at indicated time points and stored at −80°C. Samples were titrated in MDCK cells by using a 50% tissue culture infective dose (TCID₅₀) assay, and virus titers were calculated by using the Reed and Muench method (36).

Animal experiments.

Seven-week-old female DBA/2J or C57BL/6J mice were obtained from The Jackson Laboratory (Bar Harbor, ME). Mice were inoculated intranasally under isoflurane anesthesia. DBA/2J mice were inoculated with 2.8 × 10⁴ 50% egg infective doses (EID₅₀; equivalent to ∼1 50% minimum lethal dose [MLD₅₀] determined by a pilot experiment) contained in 50 μl of PBS. In the case of C57BL/6J mice, we used 1.6 × 10⁶ EID₅₀ in 50 μl, which is the highest concentration we could attain with our virus stock. Mice were then observed daily for survival and weight loss for 17 days. Animals having signs of severe illness (e.g., paralysis) or more than 25% weight loss were euthanized for humane reasons. Virus titers in tissues were determined only for DBA/2J mice. For determination of tissue titers, groups of mice were euthanized at 2, 4, and 7 days postinfection. Tissues were collected and homogenized in PBS, and aliquots were stored at −80°C until further use. Samples were titrated in 10-day-old embryonated chicken eggs, and titers were expressed as EID₅₀/ml calculated by the Reed and Muench method (36).

Acid stability.

To measure the effect of acid exposure on the retention of infectivity in vitro, virus stocks were diluted in PBS+, adjusted to the desired pH by using 0.1 M citric acid, and incubated at 37°C for 1 h. The infectivities of these viruses were then determined by measuring TCID₅₀s.

Statistical analysis.

All statistical analyses were performed with GraphPad Prism5 software. One-way analysis of variance (ANOVA), two-way ANOVA, or a log-rank chi-square test was used to test differences between different groups. P values of less than 0.05 were considered statistically significant.

RESULTS

In vitro replication kinetics of H5N1 influenza viruses containing HA mutations.

In a previous study, we found that a C58 virus containing activation pH-altering mutation HA-Y23₁H (activation pH, 6.3) or HA-K58₂I (activation pH, 5.4) had single- and multistep replication kinetics in MDCK cells similar to those of the WT C58 virus (activation pH, 5.9), despite the two mutant viruses having reduced growth and earlier clearance in the trachea and cloaca of mallards (26). To examine how the pH of activation of the HA protein contributes to the growth kinetics of the C58 viruses in cultured cells derived from various species, we determined multistep growth curves in (i) mammalian MDCK and A549 cell monolayers, (ii) avian DF1 and CCL-141 monolayers, and (iii) differentiated NHBE cells (Fig. 1).

Fig 1.

In vitro replication kinetics of reverse genetic C58 WT and mutant viruses. MDCK (A), CCL-141 (B), DF-1 (C), A549 (D), or NHBE (E) cells were infected with reverse genetic C58 WT or mutant virus at an MOI of 0.01 PFU/cell. Supernatants were collected at the indicated time points, and virus infectious titers in MDCK cells were quantified by performing TCID₅₀ assays. Error bars represent the standard deviations of triplicate samples. Graphs are representative of two independent experiments. Statistical analysis was performed by two-way ANOVA. Asterisks indicate P values of <0.05.

In MDCK cells, a prototypic mammal-derived cell type for in vitro assays of influenza virus growth kinetics, the HA-Y23₁H, HA-H24₁Q, and HA-K58₂I mutant C58 strain viruses generally had replication rates similar to that of the WT virus (Fig. 1A). In CCL-141 duck and DF1 chicken embryo fibroblast cells (Fig. 1B and C), the mutant viruses containing the activation pH-lowering mutations HA-H24₁Q and HA-K58₂I had replication rates similar to that of the WT virus. The mutant virus containing the activation pH-increasing mutation HA-Y23₁H had significantly reduced replication rates in the two avian-derived cell lines (P values of <0.01 for both cell lines, two-way ANOVA), consistent with the previously reported attenuation of this virus in mallards (26). In A549 human lung carcinoma cells (Fig. 1D), the WT and HA-Y23₁H and HA-H24₁Q mutant viruses had replication kinetics that were not statistically significantly different (P values of >0.05, two-way ANOVA), except for the HA-H24₁Q mutant at the 12-h time point. In contrast, the virus with the HA-K58₂I mutation and the lowest activation pH had significantly lower titers between 12 and 36 h after infection (P values of <0.01, two-way ANOVA) before reaching a maximum after 72 h of infection that was less than 1 log₁₀ lower than that of the WT virus. In NHBE cells (Fig. 1E), all four viruses grew at relatively similar rates.

In summary, the HA-Y23₁H mutation, which raised the activation pH from 5.9 to 6.3, contributed to an H5N1 virus with reduced replication in avian-derived cells (Fig. 1B and C) and in mallards (26). The HA-H24₁Q mutation, which lowered the activation pH to 5.6, did not produce attenuation in any of the cell lines tested (Fig. 1), just as it did not produce attenuation in mallards (26). Finally, the HA-K58₂I mutation, which lowered the activation pH to 5.4, contributed to attenuation in A549 cells (Fig. 1D) and in mallards (26) but, unexpectedly, not in avian CCL-141 or DF1 cells (Fig. 1B and C).

HA-Y23₁H, with an increased activation pH, reduces the virulence of the C58 H5N1 strain in C57BL/6J mice.

We have previously discovered that efficient replication, virulence, and transmission of C58 strain H5N1 viruses in mallards are promoted by HA proteins that have activation pH values of 5.9 (WT) and 5.6 (HA-H24₁Q) but not by an HA protein that has an activation pH value of 6.3 (HA-Y23₁H) or 5.4 (HA-K58₂I) (26). In the present study, we investigated how the pH-altering mutations might alter the virulence of C58 viruses in mice, first by using the C57BL/6J strain, which is more resistant to H5N1 influenza viruses than the DBA/2J strain is (37). We inoculated groups of C57BL/6J mice intranasally with 50 μl of PBS containing 1.6 × 10⁶ EID₅₀ of WT or mutant C58 virus and then monitored the mice for weight loss and survival for 17 days (Fig. 2). All (100%) of the C57BL/6J mice inoculated with this relatively high dose of WT C58 virus survived the infection and had an average maximum weight loss of less than 15% of their starting weight 7 days after inoculation (Fig. 2A). The low virulence of the WT virus in C57BL/6J mice is consistent with the C58 virus being highly attenuated in mammalian species because of its avian-like polymerase complex (33). All of the mice infected with the HA-H24₁Q mutant virus also survived and suffered an average maximum weight loss of ∼10% 7 days after inoculation. The virulence of the HA-K58₂I mutant virus in C57BL/6J mice was largely similar to that of the WT, as the mutant virus caused a slightly lesser extent (∼12%) and delay (by ∼1 day) of weight loss and recovery yet, on the other hand, increased the mortality rate by 15%, a difference not significant by log rank chi-square test. All of the C57BL/6J mice inoculated with the HA-Y23₁H mutant survived, and the animals did not suffer substantial weight loss compared to a group mock inoculated with PBS. Thus, the C58 virus containing an HA-Y23₁H mutation that raises the activation pH from 5.9 to 6.3 was avirulent in C57BL/6J mice (Fig. 2A), just as it was in mallards (26). In contrast, the HA-K58₂I mutation, which lowered the activation pH to 5.4 and eliminated virulence in mallards, was not found here to have attenuated virulence in C57BL/6J mice compared to that of the WT C58 virus.

Fig 2.

Virulence of reverse genetic C58 WT and mutant viruses in C57BL/6J mice. Shown are the mean percent weight change (A) and survival (B) of C57BL/6J mice (n = 7) after intranasal inoculation with 1.6 × 10⁶ EID₅₀ of reverse genetic C58 WT or mutant virus. Error bars represent the standard deviations. Statistical analysis was performed by two-way ANOVA for weight loss and log-rank chi-square test for survival curves. Asterisks indicate P values of <0.05.

HA-K58₂I, with a decreased activation pH, contributes to the increased virulence of the C58 H5N1 strain in DBA/2J mice.

As infection of the relatively resistant C57BL/6J strain of mice with the C58 strain viruses resulted in relatively low levels of weight loss even at a high dose, we next compared the viruses for pathogenicity in the relatively susceptible DBA/2J strain of mice (37). We inoculated groups of DBA/2J mice intranasally with 50 μl of PBS containing 2.8 × 10⁴ EID₅₀ of WT or mutant C58 virus and then monitored the mice for weight loss and survival for 17 days (Fig. 3). DBA/2J mice inoculated with either WT or HA-H24₁Q virus had similar average maximum weight losses of ∼12% of their starting weight and had mortality rates of 25% and 17%, respectively. Just as in the resistant C57BL/6 mice (Fig. 2A), in the susceptible DBA/2J mice, the HA-Y23₁H mutant virus was attenuated compared to the WT virus, contributing to a maximum weight loss of ∼5% compared to PBS-inoculated mice (Fig. 3A). DBA/2J mice inoculated with the HA-K58₂I mutant virus had an average maximum weight loss of ∼17% (∼5% greater weight loss than the WT group) and a mortality rate of 73% (∼50% higher mortality rate than the WT group, P value of <0.05 by log-rank chi-square test). In summary, the rank order of pathogenicity of the C58 viruses in DBA/2J mice was HA-Y23₁H (activation pH, 6.3) ≪ WT (activation pH, 5.9) ≈ HA-H24₁Q (activation pH, 5.6) < HA-K58₂I (activation pH, 5.4).

Fig 3.

Virulence of reverse genetic C58 WT and mutant viruses in DBA/2J mice. Shown are the mean percent weight change (A) and survival (B) of DBA/2J mice (n = 12) after intranasal inoculation with 28,000 EID₅₀ of reverse genetic C58 WT and mutant viruses. Error bars represent the standard deviations. Statistical analysis was performed by two-way ANOVA for weight loss and log-rank chi-square test for survival curves. Asterisks indicate P values of <0.05.

The pH of HA activation influences the growth of the C58 H5N1 strain in the murine respiratory tract.

To investigate how the activation pH of the HA protein may influence tissue-specific replication of the avian C58 H5N1 influenza viruses, we intranasally inoculated groups of DBA/2J mice with 50 μl of PBS containing 2.8 × 10⁴ EID₅₀ (∼1 MLD₅₀) of virus and collected tissues 2, 4, and 7 days later so that tissue virus titers could be measured in the nasal cavities, tracheas, lungs, brains, and kidneys (Fig. 4). While all four viruses disseminated to the brain and kidneys, none of the viruses grew to high levels (>10³ EID₅₀/ml), perhaps because of their inefficient polymerase complex activity in mice (33). Just as the WT and HA-H24₁Q mutant viruses induced similar weight losses and death rates in DBA/2J mice (Fig. 3), these two viruses also grew to similar levels in the lungs, with an average peak of ∼10⁵ EID₅₀/ml 4 days after inoculation (Fig. 4C). The WT and HA-H24₁Q mutant viruses grew to similarly low levels (<10² EID₅₀/ml) in the trachea and nasal cavities (Fig. 4A and B), consistent with these viruses having avian-like polymerase activity (26, 27, 33).

Fig 4.

Viral titers of reverse genetic C58 WT and mutant viruses in tissues of infected DBA/2J mice. Shown are the titers of reverse genetic C58 WT and mutant viruses in the nasal cavities (A), tracheas (B), lungs (C), brains (D), and kidneys (E) of infected DBA/2J mice at days 2, 4, and 7 postinfection. The ratio above each bar represents the number of mice with detectable virus titers divided by the total number of infected mice. The dotted horizontal line indicates the assay detection limit (TCID₅₀/ml = 1 log₁₀). Data from two independent experiments were combined, and the error bars represent the standard deviations. Statistical analysis was performed by one-way ANOVA. Asterisks indicate P values of <0.05.

In contrast to the HA-H24₁Q mutant, notable differences in virus growth in respiratory tract tissues were observed between the WT virus and the HA-Y23₁H and HA-K58₂I mutants. Just as the HA-Y23₁H mutant virus induced less weight loss in DBA/2J mice than did the WT virus (Fig. 3A), the HA-Y23₁H mutant virus also grew to significantly lower levels in the lungs 4 days after inoculation (nearly 2 log₁₀ titer reduction; P value of <0.01; one-way ANOVA) (Fig. 4C). The average virus titer and percentage of virus positivity after 2 and 7 days of infection were also substantially lower in mice inoculated with the HA-Y23₁H mutant than in mice inoculated with the WT. The opposite trend was observed for the HA-K58₂I mutant virus. In the lungs, the HA-K58₂I mutant continued to grow to a titer approximately 10-fold higher than that of the WT 7 days after inoculation (Fig. 4C), consistent with the HA-K58₂I mutant virus inducing greater weight loss and death in DBA/2J mice (Fig. 3). Moreover, the average virus titers and percentages of virus positivity after 7 days of infection in the trachea and nasal cavities were also substantially greater in mice inoculated with the HA-K58₂I mutant than in those inoculated with the WT. Notably, the HA-K58₂I mutant virus grew in the nasal cavities to an average peak titer of 10³ EID₅₀/ml, 100-fold higher than the peak titer of the WT virus in the nasal cavities. In summary, the rank order of C58 virus growth in the respiratory tracts of DBA/2J mice was HA-Y23₁H (activation pH, 6.3) ≪ WT (activation pH, 5.9) ≈ HA-H24₁Q (activation pH, 5.6) < HA-K58₂I (activation pH, 5.4).

An increase in the pH of HA activation coincides with increased sensitivity to acid inactivation.

Murine nasal epithelium is surrounded with glands similar to those surrounding human nasal epithelium, and both are slightly acidic (38). Acid secretions in the respiratory tract increase upon irritation or infection with influenza virus (39). Therefore, we hypothesized that a greater sensitivity to acid inactivation may contribute to the attenuation of the HA-Y23₁H mutant virus and that a greater resistance to acid inactivation may contribute to the enhanced fitness of the HA-K58₂I mutant virus in the murine respiratory tract. To test this, we incubated aliquots of each virus in pH-adjusted buffers ranging from pH 7.0 to pH 4.5 for 1 h and then after neutralization measured the titers of the viruses (Fig. 5). As might have been expected from its relatively high activation pH of 6.3, the HA-Y23₁H mutant virus's infectivity was reduced by >10-fold after exposure to pH 6.0 buffer, reduced by >100-fold after exposure to pH 5.5 buffer, and completely eliminated by pH 5.0 buffer. The WT virus was more resistant to acid inactivation than the HA-Y23₁H mutant was, as the WT did not lose infectivity after exposure to pH 6.0 buffer, lost <1 log₁₀ infectivity due to pH 5.5 buffer, and lost ∼2 log₁₀ infectivity due to pH 5.0 buffer. Similar to the WT virus, the HA-H24₁Q and HA-K58₂I mutant viruses did not lose infectivity after exposure to pH 6.0 and were completely inactivated after exposure to pH 4.5. After incubation at pHs 5.5 and 5.0, the HA-H24₁Q and HA-K58₂I mutant viruses retained slightly more infectivity (<1 log₁₀) than the WT virus, consistent with these two mutant viruses having a lower pH of HA activation. In summary, the rank order of C58 virus resistance to acid inactivation was HA-Y23₁H (activation pH, 6.3) ≪ WT (activation pH, 5.9) < HA-H24₁Q (activation pH, 5.6) ≈ HA-K58₂I (activation pH, 5.4).

Fig 5.

Acid stability of H5N1 influenza virus. Prestandardized virus stock was diluted in PBS buffer adjusted to the indicated pH and incubated for 1 h at 37°C. The remaining infectious virus titer was quantified by performing TCID₅₀ assays with MDCK cells. Statistical analysis was performed by two-way ANOVA. Asterisks indicate P values of <0.05.

DISCUSSION

The goal of this study was to investigate how mutations that alter the pH of activation of the HA protein influence avian H5N1 influenza viral infection in a mouse model. We infected groups of C57BL/6J and DBA/2J mice with either WT A/chicken/Vietnam/C58/04 (H5N1) virus or a recombinant virus containing a Y23₁H, H24₁Q, or K58₂I mutation in the HA protein. Compared to the WT C58 HA protein, which is activated to undergo irreversible conformational changes and cause membrane fusion at pH 5.9, the Y23₁H, H24₁Q, and K58₂I mutant HA proteins have been previously shown to be activated at pHs 6.3, 5.6, and 5.4, respectively (26), yet have expression, receptor binding, and cleavage phenotypes similar to those of the WT (27). Infection of mallards with these four C58 strain viruses has been investigated previously (26), thereby allowing one to compare and contrast the roles of HA activation pH in the fitness of an avian H5N1 influenza virus in avian and mammalian models. Overall, the data support the notion that a decrease in the pH of activation of the HA protein supports the adaptation of an avian H5N1 influenza virus to a mammalian host. However, the data also suggest that a mammalian-preferred HA activation pH is insufficient for robust growth in the mammalian upper respiratory tract in the absence of mammalian-adapted polymerase activity and α(2,6) receptor binding specificity, two well-established characteristics of mammalian-adapted influenza viruses.

Our present study and a previous (26) study have shown that the destabilizing HA-Y23₁H mutation, which increases the HA activation pH from 5.9 to 6.3, attenuates virus replication in avian-derived cell lines, the trachea and cloaca of ducks, and the respiratory tracts of mice. The overall lack of fitness of the C58 HA-Y23₁H mutant virus is most likely due to the HA protein of this virus being rather susceptible to inactivation, as the destabilizing mutation was also shown to increase susceptibility to acid inactivation in the present study. In general, an HA protein with an activation pH of 6.3 may be too unstable to support efficient replication in vivo, consistent with the activation pH values of HA proteins from diverse influenza virus subtypes ranging from 4.8 to 6.0 (24).

The C58 WT and HA-H24₁Q mutant viruses have HA activation pH values of 5.9 and 5.6, respectively, and these two viruses grow to similar levels in mice, ducks, and cell cultures derived from both avian and mammalian hosts. Both the WT and HA-H24₁Q mutant viruses were found to be highly pathogenic and transmissible in ducks (26) yet only moderately pathogenic in mice. Previous studies have shown that robust replication and high pathogenicity of H5N1 influenza viruses in chickens are supported by HA activation pH values of 5.7 and 6.0, while HA activation pH values lower than 5.5 are associated with decreased virulence (25, 26). Overall, these studies suggest that the preferred HA activation pH range for H5N1 influenza virus infection in avian species may be approximately 5.6 to 6.0, although additional studies with a broader array of viruses and avian hosts are needed to test this notion comprehensively.

The pH of activation of the H5N1 HA protein appears to be a host-specific replication and pathogenicity factor. While an HA activation pH of less than 5.5 substantially attenuates H5N1 influenza virus replication and virulence in avian species (25, 26), the HA-K58₂I mutant virus, with an HA activation pH of 5.4, was shown here to have enhanced replication and virulence in mice compared to those of the WT C58 virus, whose HA is activated at pH 5.9. The attenuated replication, pathogenicity, and transmission of the HA-K58₂I mutant virus in ducks (26) are not consistent with the mutant virus being shown here to replicate with WT-like efficiency in duck and chicken embryo fibroblasts, although it is possible that pH gradients resident in the endocytic pathways of respiratory and enteric tissues of mallards differ from those of duck-derived cultured cells. On the other hand, the increased fitness of the HA-K58₂I mutant virus in mice may be due in part to a small, but perhaps biologically important, increase in its resistance to inactivation by exposure to mildly acidic environments in the respiratory tract. Airway epithelium is a primary line of innate defense against inhaled pathogens (40), and mice have a cellular and glandular composition similar to that of humans (41). Normal human airway epithelial tissue, especially in the nasal cavity, is acidic (pH 5.5 to 6.9) because of secretions by submucosal glands (38, 42). Moreover, acid secretions into the airway are increased upon irritation, inflammation, or infection with influenza viruses, decreasing the pH in nasal passages to 5.2 (39, 43). Thus, better growth of the HA-K58₂I mutant virus (activation pH, 5.4) than the WT (activation pH, 5.9) in the nasal cavity and lungs 7 days after inoculation may be due to increased resistance to extracellular acid inactivation. On the other hand, the attenuated growth of the HA-Y23₁H mutant virus (activation pH, 6.3) is most likely due to its greatly enhanced susceptibility to extracellular acid inactivation.

As the optimal HA activation pH for influenza virus growth differs in various hosts and tissues, it may be possible to optimize live attenuated influenza virus vaccines by introducing mutations that yield a suitable HA activation pH value for vaccine virus growth both in eggs or Vero cells and in the respiratory tract. Several recent reports are consistent with this notion. Introduction of the previously described HA-K58₂I mutation (26, 27) into a live attenuated (with NS1 deleted) H5N1 vaccine candidate was shown to lower the HA activation pH to 5.3, lower the 50% mouse infective dose by 25-fold, and induce greater systemic and mucosal antibody responses in mice (34). In another recent study (44), an HA-N117₂D stalk mutation in PR8 virus was found to increase the HA activation pH from 5.2 to 5.4 and, consequently, increase virus growth in Vero cells 10,000-fold, most likely because Vero cells have a relatively high endosomal pH (45). Introduction of the HA-N117₂D mutation into various 2009 pandemic H1N1, H3N2, and seasonal H1N1 viruses was also shown to increase virus growth 100- to 1,000-fold in Vero cells (44), suggesting that the production of live attenuated vaccine viruses in Vero cells with human-adapted influenza viruses may, in general, be enhanced via mutations that increase the HA activation pH. Of course, care should be taken not to increase the activation pH of a live attenuated vaccine too much, otherwise the infectivity, growth, and immunogenicity of the vaccine may be reduced. For example, the introduction of an HA-G75₂R mutation into an A/Vienna/28/06 (H3N2) virus with NS1 deleted raised the HA activation pH from 5.4 to 5.8 and simultaneously impaired the immunogenicity of this vaccine candidate in ferrets (46).

The activation pH of the HA protein may regulate the replication and virulence of a wide variety of high- and low-pathogenicity influenza viruses in mice. A G23₂C mutation in the fusion peptide of the HA protein from HPAI A/Netherlands/219/03 (H7N7) has been shown to decrease the HA activation pH from 5.4 to 4.4 and simultaneously reduce virulence and virus growth in mice (29). The adaptation of low-pathogenicity avian influenza virus A/Hong Kong/1/68 (H3N2) to the lungs of mice led to the discovery of several HA mutations that increase the pH of hemolysis (a surrogate for membrane fusion) from 5.2 to 5.6 while simultaneously increasing HK68 virulence and growth in the lungs (30). Similarly, the serial passage of A/PR/8/34 (H1N1) in Mx1-positive mice (47) led to the discovery that a combination of HA mutations P78₁L and H354₁Q increases the pH of hemolysis from 5.3 to 5.8 while simultaneously increasing PR8 virulence in mice (48).

The adaptation of H5 influenza viruses to support airborne transmission in ferrets has recently been associated with a decrease in the HA activation pH, along with changes in receptor-binding specificity and glycosylation (7). Three sequential mutations were required before airborne transmissibility was acquired by a reassortant influenza virus that contains seven genes from a 2009 H1N1 pandemic virus and the H5 HA-encoding gene from A/Vietnam/1203/04: (i) N224₁K/Q226₁L in the receptor-binding pocket to switch from α(2,3)- to α(2,6)-linked receptor binding specificity, (ii) N158₁D to remove a glycosylation site from the RBD head, and (iii) T318₁I in the stalk domain to decrease the HA activation pH from 5.8 to 5.6 (7). Three functionally similar mutations were also sequentially introduced into A/Indonesia/5/2005 (H5N1) before this virus acquired airborne transmissibility in ferrets: (i) Q222₁L/G224₁S in the receptor-binding pocket to switch from α(2,3)- to α(2,6)-linked receptor binding specificity, (ii) N182₁K to remove a glycosylation site from the RBD head, and (iii) H103₁Y at the interface of the HA1 RBD and the HA2 coiled-coil stalk adjacent to a residue shown to regulate the acid stability of the H5N1 HA protein (6, 25). Thus, in both cases, after receptor-binding specificity was switched from α(2,3) to α(2,6) and a glycosylation site was deleted, a final mutation required for airborne transmission in ferrets was one that has been directly shown to decrease the HA activation pH or was one that likely decreases the HA activation pH.

In the present study, the K58₂I stalk mutation that decreased the HA activation pH from 5.9 to 5.4 was associated with an increase in C58 H5N1 virus growth and virulence in DBA/2J mice. Describing increased virulence and stability of highly pathogenic H5N1 influenza virus qualifies as dual-use research (DUR), but we do not consider knowledge of this work or the C58 HA-K58₂I mutant virus itself to constitute DUR of concern (DURC). The C58 HA-K58₂I mutant virus is not a threat to agriculture because the mutant virus is attenuated and loses transmissibility in avian species compared to that of the WT C58 virus (26). We believe that the C58 HA-K58₂I mutant virus does not increase the risk of H5N1 influenza virus to human health for several reasons. First, it lacks the ability to bind to human α(2,6) receptors and has an avian-like polymerase deficient for growth in mammalian hosts; therefore, the virus does not have the capacity to be transmitted in humans. Second, the virulence of the C58 HA-K58₂I mutant virus in mice is orders of magnitude weaker than naturally occurring H5N1 viruses and is on a par with the virulence of currently circulating human H1N1 viruses that are considered to be clinically mild (33, 49). Third, the C58 HA-K58₂I mutant virus is susceptible to oseltamivir and is antigenically matched to an A/Vietnam/1203/04 (H5N1) experimental vaccine. Fourth, the mutant viruses were not actively adapted during animal experiments and tissues were destroyed after titers were measured.

We also do not believe that a knowledge of the biological importance of the HA activation pH could be directly misapplied to pose a significant threat to public health and is therefore not DURC. Some might reasonably question whether the introduction of an HA-K58₂I mutation into a ferret-transmissible virus (7) would be expected to yield a human-transmissible H5N1 virus with enhanced pathogenicity. The final mutation required for the acquisition of airborne transmissibility in ferrets in a study by Imai et al. (7) was an HA-T318₁I mutation that lowered the HA activation pH from 5.8 to 5.6. In the context of the C58 strain, we find that the HA-K58₂I and HA-H24₂Q mutations decrease the HA activation pH by 0.5 and 0.3 unit, respectively (26, 27), and in combination, the effect is additive, with the two mutations decreasing the HA activation pH of C58 WT by 0.8 unit, from 5.9 to 5.1 (unpublished data). The additive nature of activation pH-altering mutations in other strains of H5N1 influenza viruses has also been described previously (25). Therefore, one would expect the introduction of an HA-K58₂I mutation into the ferret-transmissible virus in the study by Imai et al. (7) to reduce the HA activation pH to ∼5.1, which is most likely too low for efficient HA activation during entry. H3N2 and H7N7 viruses with HA activation pH values of less than 5.3 have been shown to have lower replication and virulence in mice than related viruses that have HA activation pH values ranging from 5.4 to 5.6 (29, 30).

As H5N1 influenza viruses are currently endemic in Egypt and Indonesia, continuing to spread among domestic poultry and often infecting humans (3, 4), H5N1 constitutes an ever-present threat to both agriculture and human health. The key finding in this paper is that a decrease in the HA activation pH (from 5.9 to 5.4) supports H5 influenza virus growth in a mammalian model while it has a deleterious effect on H5 growth in avian species (25, 26). Thus, the data show that the HA activation pH is a novel interspecies adaptation marker, helping us understand the properties necessary for influenza viruses to cross the species barrier.

This work may benefit public health in several ways. First, this work assists surveillance by identifying individual mutations and specific HA activation pH values that promote adaptation to mammals. Second, risk assessment will be enhanced through the realization that avian H5N1 influenza viruses with low pathogenicity in avian species because of a relatively low HA activation pH (such as the C58 HA-K58₂I mutant) may constitute a greater risk to mammals. Third, the knowledge of molecular markers for increased adaptation to mammals should assist scientists and public health authorities in making decisions to cull animals, quarantine humans, select prepandemic vaccine seed stocks, rapidly produce immunogenic vaccines, and identify viable drug targets such as the HA stalk (a region of the protein known to regulate its HA activation pH and a target of experimental small-molecule drugs and universal antiviral antibodies). Finally, this work also has implications for viral infectious diseases in general. Many enveloped viruses invade cells after their fusion glycoprotein is triggered by a low pH, including hepatitis C virus, Epstein-Barr virus, vesicular stomatitis virus, avian leukemia virus, human rhinovirus, dengue virus, and severe acute respiratory syndrome coronavirus (50). Thus, the tropism and host range of other important human and agricultural pathogens may also be influenced by the pH of activation of their fusion protein.