INTRODUCTION
Highly pathogenic avian influenza viruses (HPAIV) cause enormous losses in poultry worldwide and have raised concerns about a novel pandemic due to repeated zoonotic transmissions to humans (
1). HPAIV emerge from low-pathogenic precursors after circulation in poultry (
2–5) by an extension of the common monobasic hemagglutinin (HA) cleavage site (HACS) to a polybasic motif (
6–8). To mediate membrane fusion between the virion and endosome, the HA has to be activated by proteolytic cleavage (
9–11). The precursor protein HA0 is cleaved into the HA1 (head) and HA2 (stem) fragments by several host proteases whose recruitment depends on the HACS motif. Whereas the monobasic HACS prevalent in low-pathogenic avian influenza viruses (LPAIV) and seasonal human strains is susceptible to proteases like TMPRSS2 and HAT (
12), HPAIV carry a polybasic HACS that is susceptible to ubiquitous furin (
13), permitting systemic viral spread. Except for one H4 LPAIV (
14), so far, polybasic HACS are naturally confined to HA subtype H5 or H7. Because experimental reversion to a monobasic motif abolishes virulence (
15,
16), the polybasic HACS is considered the prime virulence determinant in HPAIV (
6,
8,
16,
17). In contrast, conversion of the monobasic HACS of an LPAIV to a polybasic motif does not result in a highly pathogenic phenotype in most cases (
15,
18–21). Those two findings reveal the existence of additional virulence determinants (
15,
18,
20). An increase in virulence was already attributed to the PB2, PB1, NP, neuraminidase (NA) (in particular the stalk deletion), M, and NS genes (
22–28). In this study, we focused on novel virulence determinants within the HA. Following receptor binding, depending on the cellular receptor specificity (
29–32), HA mediates the fusion of the virion envelope with the endosomal membrane (
9,
33,
34). The activation of fusion competence requires a conformational change of HA that is triggered at a particular pH (
35). Remarkably, the pH values required for this essential HA rearrangement differ considerably among different influenza virus strains. In human strains, the HA-activating pH values appear to decrease with adaptation to the human host (
36). In contrast, among avian strains, LPAIV generally display a lower activating pH, whereas contemporary H5 HPAIV exhibit a higher pH (
37). These different activating pH are likely involved in virulence and host adaptation.
DISCUSSION
HPAIV arise from low-pathogenic precursor strains upon the acquisition of the polybasic HACS, the prime virulence factor (
16). Its artificial introduction into the HA of low-pathogenic avian strains, however, does not necessarily lead to a highly pathogenic phenotype (
15,
18–21), corresponding to the existence of further virulence determinants in other gene segments, such as in PB2, PB1, NP, M, or NA (
22–28). Beyond that, the required coadaptation of the HA itself along with the acquisition of a polybasic HACS during early HPAIV evolution has remained uncertain. To this end, we generated a panel of HA mutants of clade 2.2.2 H5N1 HPAIV R65 (
53) carrying HA1 or HA2 parts from H5N1 LPAIV TG05 (
18) and/or several TG05 point mutations. We found that the R65 HA1 part and, in particular, the two HA amino acid residues 123R and 124I as a dual motif are required for extreme virulence in both chickens and mice.
After receptor-mediated endocytosis of the incoming virion along the endosomal pathway, the cleaved HA is subjected to a continuous pH decrease until an irreversible conformational change takes place that enables fusion of the membranes from the virion and endosome (
9,
33,
34). A mild acidic fusion activation pH is common to HPAIV, in contrast to LPAIV (
54), and regulates high virulence in chicken (
37). In our study, we found that in R65, each of these two HA residues, 123R and 124I, is crucial for an elevated HA fusion activation pH.
After a first cell culture passage of our recombinant viruses, secondary HA mutations were acquired in R65-HA1
TG05poly/HA2
R65 (M405I), R65-HA
R123S (G217W), and R65-HA
R123S+I124T/HA2
TG05 (S123I). The latter pseudoreversion of S123 to the larger side chain of an isoleucine clearly indicates the reduced fitness of the R123S/I124T double mutant on the R65 HA1 background. Moreover, the other two secondary mutations occurred at critical positions in the protein structure: M405 marks the N-terminal end of the B loop that connects it to α helix A (
Fig. 8A). This region likely regulates the nucleation of α helix B during the conformational switch that triggers membrane fusion. G217, on the other hand, is located directly at the trimer interface in the receptor-binding domain of the HA1 subunit (
Fig. 8A). Dissociation of HA1 is an essential step during the structural rearrangements promoting fusion. Both of these secondary mutations likely decrease the overall stability of the prefusion conformation and thus may partially compensate for the artificial introduction of our stabilizing mutations.
Furthermore, in viruses reisolated from chicken oral swabs and mouse heart homogenates, we found several mutations, such as the reversion HA S123R or HA2 mutations, in the TG05 HA2 chimeric viruses that were identical to or at the same positions as those in R65, indicating strong selection pressure. Moreover, we observed mutations in the M2 ion channel exit region (
42,
44,
45,
55), D44N (
43), R45L, and R45H, which were also selected in chickens infected with highly pathogenic reassortants carrying a non-H5/non-H7 HA (
21). These mutations are prevalent in a few native HPAIV and LPAIV (
21) and were demonstrated to compensate for the lower acid sensitivity of the HA (
56).
The additional N-glycosylation motif formed by the R156N exchange in the HA of R65 resulted in somewhat increased growth properties, fusion activation pH, and virus inactivation pH and slightly increased virulence in chicken. Therefore, the very common loss of this N-glycosylation site in H5 HPAIV (
Fig. 1 and
2B) is not detrimental in chickens. In contrast, its absence enhances virulence in mice, facilitates contact transmission in guinea pigs (
40,
41), and is involved in airborne transmissibility in ferrets (
57). Therefore, the absence of an N-glycosylation site at positions 156 to 158 might have supported the transit of the clade 2.2.2 viruses through some intermediate mammalian host(s) like, perhaps, pikas (
58), as previously proposed for the mammalian marker mutation PB2 627K (
59).
Some HA virulence determinants could already be revealed by the determination of different 50% lethal dose (LD
50) values of closely related H5N1 HPAIV in chicken mutated at HA1 positions 97, 108, 126, 138, 212, and 217 (
60) (corresponding to positions 113, 124, 142, 154, 228, and 233, respectively, in the full-length R65 HA translated sequence). All these residues are identical in the R65 HA and H5 HPAIV consensus sequence (determined from 3,385 sequences) but are different in the corresponding positions of the TG05 HA and LPAIV consensus sequence (from 803 sequences), except for the rather conserved residue 97 (residue 113) (
Fig. 1). Among these mutations, HA 124I was attributed to increased acid sensitivity and virulence in chickens (
37). Furthermore, sequence database searches revealed that both 123R and 124I are predominant as a dual motif in the overwhelming majority of 97.5% of H5 HPAIV, in contrast to H5 LPAIV. Remarkably, this dual motif is retained in later and contemporary H5 HPAIV, including novel H5NX reassortants, like the H5N8 or the H5N6 strains, which can cause experimental infections in dogs and zoonotic infections in humans (
61–65).
However, a few HPAIV (2.0%) carry 123S and 124T or only a single exchange (up to 0.5%) (
Fig. 2A). Among them, A/Duck/Guangxi/07/1999 (
66) carries neither 123R, 124I, nor any of the candidate HA1 mutations established previously (
60) (
Table 3). Interestingly, this strain killed only two of seven oculonasally infected chickens without systemic replication and was avirulent in mice (
66). This observation supports the notion that the HA 123R/124I dual motif is crucial for high virulence in avian and mammalian hosts.
Furthermore, a survey for the HA 123R/124I motif among 1,003 H5 HPAIV sequences assigned to the H5 clades (
67) by the LABEL algorithm (
68) confirmed its absence among American and Eurasian non-Goose/Guangdong-like strains and its presence in all main H5 clades (
Tables 4 and
5).
Changes in the pH required for triggering HA-mediated fusion have been recognized as part of the process of adaptation of influenza viruses to novel hosts. Whereas human-adapted strains show a decrease in pH sensitivity (
36), HPAIV display an increased fusion pH accompanied by increased virulence in chickens (
37,
69) and enhanced replication in the upper respiratory tracts of mice and ferrets (
70,
71). An explanation for such divergent evolution could be that that both virus lineages undergo different selection pressures. In this regard, the demonstration that human strains primarily target nonciliated cells in the respiratory tract, whereas avian strains infect ciliated cells dependent on the virus receptor preference for α2-6- or α2-3-linked sialic acids (
72), might suggest that the different cell populations exert different evolutionary pressures. Alternatively, it is tempting to speculate that the acidification of the endosome is enhanced by upregulated V-ATPase activity due to the early activation of extracellular signal-regulated kinase (ERK) and phosphatidylinositol 3-kinase (PI3K) by the virus (
73). It remains to be elucidated whether differences in primary cellular tropism and/or virus-mediated regulation of the endosomal pH determine the different evolutionary pathways of avian and human virus lineages. Remarkably, the earliest H5N1 HPAIV strains from 1996/1997 (
74,
75), A/Goose/Guangdong/1/1996 (
76) and A/Goose/Guangdong/3/1997, carry only the HA 123R mutation but neither 124I nor any of the other identified HA1 residues relevant for virulence (
60). Therefore, 123R was crucial in early H5 HPAIV evolution, whereas 124I was acquired later in the clade 0 viruses from 1997, as seen in the first known chicken isolate, A/Chicken/Hong Kong/220/1997 (
77). Later on, other HA1 exchanges (
60) also appeared: 126E (142E) and 217S (233S) in 1997 and then 212K (228K) in 2000. Eventually, 138N (154N) complemented the whole set of all six key positions in the 2000 HPAIV (
Table 3).
Because several cell lines and tissues display different final (minimum) pH in their endosomes (
78,
79), they may not be permissible to influenza virus infection if the required HA fusion activation pH of a particular strain is lower than the endosomal endpoint pH, as is the case with LPAIV (
54,
78). Therefore, an increased acid sensitivity of the HA activating fusion competence enables the virus to infect cells with a milder acidic final pH in their endosomes, resulting in extended cell tropism (
78,
79). To this end, the RI motif at HA positions 123 and 124, in addition to the polybasic HACS, may serve as a door opener to the respiratory tract for H5 HPAIV (
70,
78,
79).
Taken together, the HA 123R/124I dual motif is crucial for high virulence in avian and mammalian hosts, indicating substantial relevance for the high zoonotic potential of H5 HPAIV. Accordingly, this virulence determinant is still predominant in Asian H5 HPAIV, including the novel H5NX reassortants. For HPAIV emergence, the acquisition of the polybasic HACS is the critical step but requires parallel coadaptation of the HA. The HA 123R/124I dual motif leads to an increased acid sensitivity, allowing broader viral tissue tropism for cells with an elevated (less-acidic) endpoint pH in their endosomes (
78,
79). This acquired gain of function might enable HPAIV to infect the respiratory tract of avian and mammalian hosts (
37,
70,
71,
79) and therefore facilitate zoonotic spillover infections. Whether such an adaptation would result in dead-end evolution compared to established human strains or airborne-transmissible H5N1 HPAIV mutants requiring sensitivity to lower pH (
36,
57,
80,
81) remains elusive.
Overall, we propose that, in addition to the acquisition of a polybasic HACS, the coadaptation of the HA to sensitivity to higher pH results in a broader viral organ tropism extended to the respiratory tract, enabling the emergence of novel HPAIV.