A Dual Motif in the Hemagglutinin of H5N1 Goose/Guangdong-Like Highly Pathogenic Avian Influenza Virus Strains Is Conserved from Their Early Evolution and Increases both Membrane Fusion pH and Virulence

To reveal differences in the HA proteins of H5 LPAIV and HPAIV, we surveyed the fluDB (38) and GISAID (https://www.gisaid.org/) public databases. For analysis, we included 4,188 aligned sequences to generate the consensus HA sequences from LPAIV (803 entries) and HPAIV (3,385 entries). They differ by 35 amino acid exchanges in HA1 and by 8 exchanges in HA2 (Fig. 1). As parental viruses for the generation of HA mutants by reverse genetics (39), we used the recombinant H5N1 HPAIV A/Swan/Germany/R65/06 (R65) (17) and the polybasic HA cleavage site mutant (TG05-HA_poly) of H5N1 LPAIV A/Teal/Germany/Wv632/2005 (TG05) (18). Their HA sequences and the consensus sequences from LPAIV and HPAIV share 22 amino acid exchanges in HA1 and 5 exchanges in HA2 (Fig. 1), narrowing the range of relevant mutations in the HA.

The substitution T124I reduces acid stability accompanied by differentially increased virulence in chicken (37) (corresponding to position 108 in that study). Strikingly, within the HPAIV HA consensus sequence, including the R65 HA (Fig. 1), the preceding amino acid residue S123 was altered to R. In contrast to LPAIV, in which positions 123 and 124 contain S and T in 93.8% of sequences, in HPAIV, both residues are replaced by R and I in 97.5% of sequences (Fig. 2A). Furthermore, R65 carries an incomplete potential N-glycosylation motif (40) due to R at position 156 (Fig. 1). In most H5 HPAIV (80.4%) (38) (https://www.gisaid.org/), this motif is incomplete (Fig. 2B). Whereas its absence is crucial for high virulence and contact transmission in mammalian models (40, 41), its phenotype in chicken is unknown.

Taken together, the sequence survey indicates that position 156 and in particular the dual motif at positions 123 and 124 are candidate virulence determinants in H5 HPAIV.

To investigate the impact of all amino acid exchanges in the entire HA1 and HA2 regions beside the dual motif 123R/124I, we first constructed HA chimeric viruses by replacing the HA1 or HA2 part in R65 with that of TG05-HA_poly, resulting in the HA chimeras R65-HA1_TG05poly/HA2_R65 and R65-HA1_R65/HA2_TG05, respectively (Fig. 2C).

According to our surveys using sequence databases, we introduced three TG05 amino acid residues into R65 HA1, resulting in the mutants R65-HA_R156N, R65-HA_R123S, and R65-HA_I124T (Fig. 2C). Remarkably, we could not rescue a double-virus mutant carrying both R123S and I124T in the authentic R65 HA. We speculate that this result is due to severely impaired fitness.

We also introduced the same HA1 point mutations into the HA2 chimera R65-HA1_R65/HA2_TG05, resulting in R65-HA_R156NHA2_TG05, R65-HA_R123SHA2_TG05, R65-HA_I124THA2_TG05, and R65-HA_R123S+I124T/HA2_TG05 (Fig. 2C).

After a first passage of the rescued viruses on MDCK cells, we also found the acquired HA mutations M405I in R65-HA1_TG05poly/HA2_R65, G217W in R65-HA_R123S, and S123I in R65-HA_R123S+I124THA2_TG05, which we therefore renamed R65-HA_R123I+I124THA2_TG05. This finding of a pseudoreversion in the latter HA chimera corresponds to our failure to rescue the R65 nonchimeric position 123/124 double mutant. Taken together, these results indicate that the two amino acid residues R123 and I124 within HA1 form a coupled motif, which is crucial for the fitness of an H5N1 HPAIV like R65.

To study the impact of the HA mutations on replication efficiency, we determined growth curves on DF1 cells. The HA2_TG05 chimera R65-HA1_R65/HA2_TG05 exhibited a replication kinetic that was decreased by 2 orders of magnitude compared to the parental R65, the TG05 HA reassortant R65-HA_TG05poly, and the HA1_TG05 chimera R65-HA1_TG05poly/HA2_R65 (Fig. 3A). All viruses with the HA1 region modified, such as R65-HA1_TG05poly/HA2_R65 or the single point mutant R65-HA_R156N, R65-HA_R123S, or R65-HA_I124T, and wild-type R65 displayed titers within 1 order of magnitude at 48 h postinfection (p.i.). However, R65-HA_R156N came close to its maximum already at 24 h p.i. (Fig. 3B). In contrast to R65-HA1_R65/HA2_TG05, which replicated approximately 100-fold less than R65, all the other HA2_TG05 chimeras carrying an additional TG05 point mutation within HA1 displayed a replication efficiency similar to that of R65: R65-HA_R156N/HA2_TG05, R65-HA_R123S/HA2_TG05, R65-HA_I124T/HA2_TG05, or R65-HA_R123I+I124THA2_TG05 (Fig. 3C). Taken together, HA2 replacement, as with R65-HA1_R65/HA2_TG05, led to a titer reduction, which, however, is offset by any of the HA1 point mutations from TG05. Furthermore, we observed no apparent growth disadvantage of the R65 mutants in DF1 cells in the case of HA1 replacement or one of the R123S and I124T substitutions.

To investigate the virulence and transmission of R65, R65-HA1_R65/HA2_TG05, R65-HA1_TG05poly/HA2_R65, R65-HA_R123S, and R65-HA_I124T in chicken, we infected six birds per group oculonasally with 10⁵ PFU and added four contact chickens at day 1 p.i. All birds infected with R65 or R65-HA1_R65/HA2_TG05, including the contact animals, succumbed to death, whereas among the R65-HA1_TG05poly/HA2_R65-infected chickens, two primarily infected birds and one contact bird survived (Fig. 4). Those two primarily infected birds displayed clinical symptoms from day 3 p.i., one recovered until day 9 p.i., and both animals seroconverted (data not shown). The surviving contact animal developed mild symptoms on day 9 p.i. and was still seronegative on day 10 p.i. In contrast, virulence and transmissibility were reduced drastically for R65-HA_R123S and R65-HA_I124T; all animals remained seronegative. Infection with the latter resulted in death of three primarily infected animals and one contact bird. Another contact bird developed mild symptoms on day 10 p.i. and was seronegative at that time. R65-HA_R123S did not cause any clinical symptoms or seroconversion in primarily infected or contact animals (Fig. 4).

For investigation of the point mutants and HA2 chimeras R65-HA_R156N, R65-HA_R156N/HA2_TG05, R65-HA_R123S/HA2_TG05, R65-HA_I124T/HA2_TG05, and R65-HA_R123I+I124T/HA2_TG05, we infected three chickens per group oculonasally. All animals died, with total clinical scores ranging from 2.0 to 2.2, close to that of R65, with a score of 2.2. Only R65-HA_R156N displayed a slightly increased value of 2.6 (Fig. 4 and Table 1). Therefore, the replacement of R65 HA2 by that of TG05 can restore the virulent phenotype even in the presence of the point mutations R123S and I124T. Overall, the strong attenuation of R65 in the presence of the two HA mutations R123S and I124T in chicken demonstrates their crucial role in the virulence of HPAIV.

To address the impact of the most attenuating mutation, HA R123S, in a mammalian model, we intranasally inoculated five mice per group with phosphate-buffered saline (PBS) or 10⁴ PFU of R65 or R65-HA_R123S. Whereas all R65-infected animals succumbed to death, the mock- and R65-HA_R123S-infected animals remained unaffected (Fig. 5A). To determine the viral titers in organs by a plaque assay, we infected mice and sacrificed them on day 3. In R65-infected mice (n = 3), we found high viral loads in lungs and hearts and a notable titer in the brain of one animal. However, in the R65-HA_R123S-infected mice (n = 4), we detected no lung titers in two animals or titers considerably reduced by 4 to 2 orders of magnitude in the other two animals, a noticeable titer in the heart of one mouse, and no titers in the brains (Fig. 5B). Taken together, the HA R123S mutation in the clade 2.2.2 virus R65 (H5N1) abolishes virulence in both avian and mammalian hosts.

Among positive PCR samples from oral swabs from infected chickens, we found the following HA changes. R65-HA1_R65/HA2_TG05 acquired the I394G exchange in the HA2 region; at this position, R65 carries V. Another HA2 mutation was found in shed R65-HA_R156N/HA2_TG05, G421R, being identical in HA2 of R65 (Table 2). Therefore, both chimeric viruses appear to adapt their heterologous HA2 part, originating from TG05, to R65 HA1. Moreover, R65-HA_R123S/HA2_TG05 reverted to S123R (Table 2). Furthermore, in R65-HA_I124T/HA2_TG05 and R65-HA_R123I+I124T/HA2_TG05, we found mutations in the exit region of the proton channel M2 (42–45), D44N and R45L (Table 2), indicating enforced coadaptation to the HA mutations at positions 123 and 124.

From the R65-HA_R123S-infected mouse displaying the highest virus titers (Fig. 5B), we reisolated virus from the heart and sequenced the HA and M genes. Whereas the introduced HA R123S exchange remained unaffected, M2 displayed the R45H exchange. Taken together, these observations support the notion that the HA1 R123S exchange plus the HA1 and HA2 parts from TG05 are incompatible with the R65 background and therefore entail compensatory mutations in HA or M2.

To determine the activation pH, enabling HA-mediated fusion, of each virus, we performed syncytium formation assays, whereas we assessed the acid stability in the environment by measurement of the virus inactivation pH (Fig. 6A). HPAIV R65 displays a high fusion activation pH of 5.9 and an inactivation pH of 5.7. However, TG05-HA_poly, which is an HACS mutant of LPAIV TG05 (18), has considerably lower pH values of 5.1 and 4.9, respectively. The introduction of the R65 HA into the TG05 background (TG05-HA_R65) resulted in increases of both pH values (5.8 and 5.4, respectively), but still below those of R65, suggesting the involvement of both HA and the other viral proteins. The reciprocal reassortant R65-HA_TG05poly also displayed a lower fusion pH of 5.3 but an extremely low virus inactivation pH of 4.1. Furthermore, the replacement of the HA1 or HA2 regions decreased the pH values to different extents (R65-HA1_R65/HA2_TG05 and R65-HA1_TG05/HA2_R65). Restoration of the N-glycosylation site at positions 156 to 158 (R65-HA_R156N and R65-HA_R156N/HA2_TG05) resulted in somewhat lower pH values, indicating enhanced stability. Remarkably, the HA R123S single mutation in R65 decreases the fusion activation pH from 5.9 to 4.9 and the virus inactivation pH from 5.7 to 4.4. Similarly, the I124T exchange leads to a decreased fusion activation pH of 5.6 and a virus inactivation pH of 4.7 (Fig. 6A). The introduction of TG05 HA2 changed the pH stability of all point mutants to various extents. Overall, we found notable decreases of the fusion activation pH and virus inactivation pH in the presence of substitutions R123S and I124T.

Because an increased temperature may lead to fusion competence due to the spring-loaded metastable conformation of the HA (46), we investigated the influence of the HA mutations on thermal stability. To this end, we incubated R65 and all mutants at different temperatures for 30 min prior to an HA test. In contrast to all other viruses, R65-HA_R123I/I124T-HA2_TG05 displayed a reduced HA titer of 1:8 after incubation at 42°C, compared to an HA titer of 1:64 following incubation at 37°C (Fig. 6B). This result suggests that the TG05 HA2 part might destabilize the prefusion conformation of the HA from R65-HA_R123I/I124T-HA2_TG05 to some extent.

In general, for the virus variants studied, their virulence in chicken increased with their fusion activation pH (Fig. 7). In stark contrast to R65 and the N-glycosylation mutant R65-HA_R156N, displaying the highest clinical scores and pH values, the two single point mutants R65-HA_R123S and R65-HA_I124T have an avirulent or lower-pathogenicity phenotype accompanied by the lowest fusion activation pH. Therefore, an increased acid sensitivity of the HA is required for the high virulence of R65 mediated by the 123R/124I dual-amino-acid motif.

The fusion of virion and endosomal membranes results from the low-pH-triggered irreversible switch between the metastable “spring-loaded” prefusion conformation and the low-energy postfusion conformation of the HA. A key event during this process is the conversion of the B loop within the HA2 subunit into an α helix (Fig. 8A). Most available prefusion structures of H2, H5, and H6 hemagglutinins display similar conformations, in which the B loop is tightly attached to the central coiled coil of the HA2 stem (Fig. 8B). It has been suggested that early events in the conformational switch of these proteins involve a reversible detachment of the B loop from its prefusion position (47). The crystal structure of such an early fusion intermediate has been solved using a prefusion-stabilized mutant of the A/Japan/305/1957 H2 hemagglutinin at pH 5.3 (47) (Fig. 8C). Whereas the only three crystal structures of a canonical LPAIV H5 HA with the 123S/124T motif show the typical “attached” conformation of the B loop at pH 7.5 (48, 49) (Fig. 8D), some crystal structures of HPAIV H5 HA with the 123R/124I motif have the “detached” conformation of the B loop, even under neutral conditions, between pH 6.5 and pH 8.5 (37, 50–52). In contrast to A/Japan/305/1957 H2, where the initial detachment of the B loop appears to be low-pH dependent, the same process is therefore likely pH independent in HPAIV H5 hemagglutinins. For example, both conformations, the “attached” and the “detached” ones, were found in two different crystal forms of the same HPAIV H5 HA under identical crystallization conditions at pH 6.6 (37) (Fig. 8E and F). On the atomic level, the “detached” B loop within HA2 in the respective HPAIV H5 structures is held in suspension via multiple attractive interactions with the vestigial esterase domain of HA1. Of note, among the involved residues in the HA1 subunit is R123, which holds the benzene ring of F409 from the B loop via a cation-π interaction (Fig. 8F). In light of these observations, we propose that R123 contributes to the destabilization of the HPAIV H5 prefusion conformation via pH-independent priming of the HA2 subunit for its conformational switch. Although the supporting role of I124 is less apparent from the available structures, this residue likely affects the local structural dynamics of the vestigial esterase domain and thus influences the pH sensitivity of HPAIV H5 HA, as previously discussed (37).

On 24 July 2017, we downloaded all available H5 protein sequences from fluDB (38) and GISAID (https://www.gisaid.org/) (an acknowledgment spreadsheet of all laboratories contributing to GISAID is available upon request) and performed an alignment by using the MAFFT algorithm (82) implemented in the Geneious package (83). For further analysis, we removed all duplicated, truncated, or gapped sequences. Sequence positions refer to the R65 HA sequence, including the signal peptide (16 amino acids).

Human embryonic kidney (293T), Madin-Darby canine kidney (MDCK), and baby hamster kidney (BHK-21) cells were maintained in Dulbecco's modified Eagle's medium (DMEM) containing 10% fetal calf serum (FCS), whereas chicken fibroblast cells (DF1) were cultured in Iscove's DMEM plus 10% FCS.

The recombinant H5N1 HPAIV A/Swan/Germany/R65/06 (R65) (GenBank accession numbers DQ464354 to DQ464361) was grown in MDCK cells, and the existence of unwanted mutations was excluded by Sanger sequencing (18). TG05_poly, a polybasic HACS mutant from H5N1 LPAIV A/Teal/Germany/Wv632/2005 (18), was propagated in 11-day-old embryonated chicken eggs. Chimeric HA from R65 and TG05_poly were assembled by target-primed plasmid amplification (39); HA single point mutants were obtained by QuikChange site-directed mutagenesis (primer sequences are available upon request). The mutants were rescued as described previously (84) and then propagated in MDCK cells; virus gene compositions and full-length HA genes were verified by Sanger sequencing of reverse transcription-PCR (RT-PCR) amplicons. All viruses were handled under biosafety level 3 plus (BSL3⁺) conditions.

For plaque assays, we infected MDCK cells, as described previously (85), with 400 μl virus inoculum in the absence of trypsin. For growth curves, we inoculated DF1 cells at a multiplicity of infection (MOI) of 0.001. From two independent replicates, we harvested supernatants 0, 24, 48, and 72 h after inoculation for titration by plaque assays (85).

The animal experiments were evaluated by the responsible ethics committee of the State Office for Agriculture, Food Safety, and Fishery in Mecklenburg-Western Pomerania (LALFF M-V) and gained governmental approval (registration numbers LALLF M-V/TSD/7221.3-1.1-018/07 and LALLF M-V/TSD/7221.3-1.1-023/13). We infected 2-week-old White Leghorn specific-pathogen-free chickens (three or six per group) oculonasally at a dosage of 10⁵ PFU. For transmission studies, we added four contact animals at 1 day p.i. Each bird was observed daily for 10 days for clinical signs and classified according to OIE criteria; moribund birds were killed and scored as dead on the next observation day (86). Sera of surviving birds were collected at day 10 and tested for influenza A virus nucleoprotein-specific antibodies by a competitive enzyme-linked immunosorbent assay (ELISA) (ID-Vet), as recommended by the manufacturer.

Four-week-old female BALB/c mice (Charles River, Sulzfeld, Germany) were housed in ISOcages (Tecniplast, Buguggiate, Italy). After anesthesia by isoflurane inhalation, we intranasally inoculated the animals with 30 μl phosphate-buffered saline (PBS) (mock) or PBS-diluted virus at a dosage of 10⁵ PFU. Over a period of 14 days, we observed each mouse daily regarding body weight changes, other clinical symptoms, or death. By a plaque assay, we titrated entire organ homogenates of lungs, hearts, and brains from inoculated mice, sacrificed at 3 days p.i.

In two independent experiments, we incubated the viruses in a thermoblock for 30 min at 4°C, 37°C, 42°C, or 50°C prior to hemagglutination assay at room temperature using 1% chicken red blood cells.

BHK-21 monolayers in six-well plates were infected at an MOI of 2 and incubated at 37°C for 30 min. We then washed the cells with PBS and overlaid them with DMEM containing 0.2% bovine serum albumin (BSA) and antibiotics. After 6 h, the cells were incubated at 37°C for 5 min in PBS in parallel at different pH adjusted in 0.1-unit increments from 4.8 to 6.0. We then washed the cells, incubated them in DMEM containing 10% FCS at 37°C 2 h, and stained them eventually with a Hema-3 Stat Pack staining kit (Fisher, USA). We noted the fusion activation pH as the highest pH sufficient for the formation of any syncytia.

First, we incubated DF1 cells for 30 min at 37°C in parallel with virus at an MOI of 2 diluted in PBS at different pH adjusted in 0.1-unit increments from 4.0 to 6.0. After the removal of the inoculum, we kept the cells in Iscove's DMEM containing BSA (0.2%) and antibiotics at 37°C for 5 h. Cells were then fixed with 3.7% formaldehyde (Roth, Germany), permeabilized with 0.5% Triton X-100 (Merck, Germany), and blocked for 30 min with 1% skim milk in PBS. Infected cells were stained with a polyclonal antibody to the NP protein from A/WSN/1933 (H1N1) (1:1,000; incubated for 1 h at room temperature) (GeneTex, USA) and washed with PBS three times prior to detection with goat anti-rabbit Alexa Fluor 488 antibody (Invitrogen, Germany) by fluorescence microscopy. All experiments were done in duplicates. The highest pH at which no virus-infected cells were found was designated the virus inactivation pH.