The M2 proteins of bat influenza A viruses reveal atypical features compared to conventional M2 proteins

ABSTRACT The influenza A virus (IAV) M2 protein has proton channel activity, which plays a role in virus uncoating and may help to preserve the metastable conformation of the IAV hemagglutinin (HA). In contrast to the highly conserved M2 proteins of conventional IAV, the primary sequences of bat IAV H17N10 and H18N11 M2 proteins show remarkable divergence, suggesting that these proteins may differ in their biological function. We, therefore, assessed the proton channel activity of bat IAV M2 proteins and investigated its role in virus replication. Here, we show that the M2 proteins of bat IAV did not fully protect acid-sensitive HA of classical IAV from low pH-induced conformational change, indicating low proton channel activity. Interestingly, the N31S substitution not only rendered bat IAV M2 proteins sensitive to inhibition by amantadine but also preserved the metastable conformation of acid-sensitive HA to a greater extent. In contrast, the acid-stable HA of H18N11 did not rely on such support by M2 protein. When mutant M2(N31S) protein was expressed in the context of chimeric H18N11/H5N1(6:2) encoding HA and NA of avian IAV H5N1, amantadine significantly inhibited virus entry, suggesting that ion channel activity supported virus uncoating. Finally, the cytoplasmic domain of the H18N11 M2 protein mediated rapid internalization of the protein from the plasma membrane leading to low-level expression at the cell surface. However, cell surface levels of H18N11 M2 protein were significantly enhanced in cells infected with the chimeric H18N11/H5N1(6:2) virus. The potential role of the N1 sialidase in arresting M2 internalization is discussed. IMPORTANCE Bat IAV M2 proteins not only differ from the homologous proteins of classical IAV by their divergent primary sequence but are also unable to preserve the metastable conformation of acid-sensitive HA, indicating low proton channel activity. This unusual feature may help to avoid M2-mediated cytotoxic effects and inflammation in bats infected with H17N10 or H18N11. Unlike classical M2 proteins, bat IAV M2 proteins with the N31S substitution mediated increased protection of HA from acid-induced conformational change. This remarkable gain of function may help to understand how single point mutations can modulate proton channel activity. In addition, the cytoplasmic domain was found to be responsible for the low cell surface expression level of bat IAV M2 proteins. Given that the M2 cytoplasmic domain of conventional IAV is well known to participate in virus assembly at the plasma membrane, this atypical feature might have consequences for bat IAV budding and egress.

The ion channel activity of M2 protein is crucial for the IAV uncoating process (22). Following internalization of IAV by receptor-mediated endocytosis, a vacuolar-type H + -ATPase leads to a progressive increase of the proton concentration in the endoso mal lumen. The acidic milieu in the endosome activates the M2 ion channel protein embedded in the viral envelope (23). The inward proton flux into the virion promotes loosening of M1 interactions with itself and other viral components such as HA and viral ribonucleoprotein complexes (vRNP), thereby decreasing rigidity of the virus particle (23)(24)(25). When the pH value drops beyond a critical pH threshold level, the viral hemagglutinin (HA) undergoes a conformational change and triggers the fusion of the viral envelope with the endosomal membrane (26). As the M1 protein has already separated from the RNPs, the nuclear localization signal of the nucleoprotein (NP) can be recognized which allows the active transport of the RNPs into the nucleus, where replication and transcription of the viral genome takes place. The uncoating step is inhibited by the antiviral compounds AMT and rimantadine, which prevent protons from passing the channel (21).
For some IAV, an additional function has been assigned to the M2 protein. The ion channel protein raises the pH in the secretory pathway and thereby preserves the native conformation of HA during its transport through acidic compartments such as the Golgi apparatus (27). If the ion channel is blocked by AMT, HA is at risk of undergo ing a premature conformational change, which is irreversible. As a consequence, HA is transported to the plasma membrane and integrated into the viral envelope in a fusion-incompetent form. M2 ion channel activity also affects cellular ion homeostasis and contributes to IAV pathogenesis by dysregulating protein transport and maturation in the secretory pathway (28,29), by altering cellular ion channel functions in the airways (30)(31)(32), and by triggering the inflammasome (33).
In 2012 and 2013, the genomic sequences of novel influenza A-like viruses were identified in Guatemala in specimens of the little yellow-shouldered bat Sturnira lilium (34) and in Peru in feces of the fruit bat Artibeus planirostris (35). The phylogenetic analysis of the genomic RNA segments encoding the putative envelope glycoproteins HA and NA led to the classification of bat IAV into the new subtypes H17N10 and H18N11. Bat IAV differ from conventional IAVs in several fundamental aspects. The surface glycoproteins lack the canonical receptor-binding and -destroying activities of classical HA and NA, respectively (35)(36)(37). The bat IAV HA glycoproteins H17 and H18 do not use sialic acid residues for attachment to the host cell but rather take advantage of the MHC class II protein complex for entry into host cells of different species (38). The ectodomain of the NA-like protein seemed to be dispensable for virus replication in cell culture and in mice (39), but turned out to be important for efficient transmission in the bat natural host (40). These findings indicate that the NA-like protein has an impact on tissue tropism and transmission, although the exact biological function of this envelope glycoprotein is not known yet.
Another unusual feature of bat IAV is the incompatibility of the packaging signals of most of their RNA segments with those of conventional IAV (41,42). Only when the packaging signals of bat IAV RNA segments four and six were used, chimeric bat IAV harboring the envelope glycoproteins of the mouse-adapted A/SC35M (H7N7) could be generated by reverse genetics (41). The chimeric virus turned out to be resistant to amantadine, in line with the presence of an asparagine residue at position 31. Interest ingly, when the chimeric virus was passaged on chicken DF-1 cells or embryonated chicken eggs, the asparagine at position 31 rapidly mutated to serine, which rendered the virus sensitive to amantadine (41). The functional relevance of this adaptive mutation remains unclear since the ion channel activity of the bat IAV M2 proteins has not been studied so far.
In the present study, we analyzed the bat IAV M2 proteins with respect to subcellular localization, proton channel activity (PCA), and functional importance for virus entry and HA metastable conformation in the secretory pathway. Our findings suggest that bat IAV M2 proteins differ fundamentally in several aspects from the M2 proteins of classical IAV.

Bat IAV M2 primary sequences reveal remarkable differences compared to conventional IAV M2 proteins
The open reading frames of the bat IAV M2 proteins are composed of 96 aa, whereas the M2 proteins of conventional IAV always comprise 97 aa (Fig. 1A). For prediction of putative TM domains, we ran a hydrophobicity plot for the M2 proteins of A/ little yellow-shouldered bat/Guatemala/164/2009 (H17N10) (denoted as M2 G ) and A/ flat-faced bat/Peru/033/2010 (H18N11) (denoted as M2 P ) and compared them with the M2 protein of the conventional avian IAV A/FPV/chicken/Rostock/1934 (H7N1) (denoted as M2 R ) (Fig. 1B). For all three M2 proteins the hydropathy graph followed a very similar trend except for the region between aa 44 and 62, which corresponds to the AH of conventional IAV (Fig. 1B). The AH domain of M2 R protein revealed a higher hydrophi licity than the bat IAV M2 proteins. Nevertheless, a hydrophobic region between aa 26 and 43, which correspond to the TM domain of conventional IAV M2 proteins (7), was also detected in the bat IAV M2 protein. When a plot for charged aa residues was performed, the predicted TM domains of all three M2 proteins were flanked N-terminally by negatively charged aa and C-terminally by positively charged ones (Fig. 1C), in line with the topology of type III membrane proteins (43).
The M2 proteins of conventional IAV reveal an extremely high sequence conservation throughout the whole molecule (44). When the primary sequences of the H17N10 and H18N11 M2 proteins were aligned with an M2 consensus sequence representing 6,413 different isolates of conventional IAV (Fig. 1A), similarities but also remarkable differences were detected. While the M2 ED of conventional IAV normally contains two cysteine residues (Cys17 and Cys19), which stabilize the M2 tetramer by forming intermolecular disulfide bonds (45), only Cys19 was present in the bat IAV M2 proteins. Similar to conventional IAV, the TM domain of bat IAV M2 proteins harbors the His37-x-x-x-Trp41 motif, suggesting that the bat IAV M2 proteins may have PCA. In conventional IAV M2 proteins, a serine is normally found at amino acid position 31, while in AMT-resistant strains this residue is frequently mutated to asparagine (46,47). Surprisingly, the TM domains of bat IAV M2 proteins contain Asn31 which was found to confer resistance of bat IAV to this drug (41). Using the HeliQuest algorithm (48), amphipathic helices were predicted for M2 G (H17N10) and M2 P (H18N11) with hydrophobic moments (<µH>) similar to that of M2 R (Fig. 1D). The AH domain of conventional IAV contains a conserved Cys residue at position 50 which is subject to post-translational palmitoylation (49,50). This cysteine residue is absent in the CT domains of bat IAV M2 proteins. Finally, the CT domains of the bat IAV M2 proteins, including the M1 interaction site (aa 71-76) (51, 52), did not show significant homology with the corresponding region of conventional IAV. The bat IAV M2 proteins also lack the LIR motif that was shown to mediate recruitment of the autophagy protein LC3 to the plasma membrane (53).

The H18N11 M2 protein assembles into homo-tetramers
To analyze the expression and subcellular localization of bat and conventional IAV M2 proteins, a plasmid-based system was employed. An [HA] peptide tag (YPYDVPDYA) was genetically fused to the M2 N-termini to allow detection of the different proteins with the same antibody. BHK-21 cells were transfected with the [HA] M2 expression plasmids and lysed 24 hours post-transfection (p.t.). The cell lysates were separated by SDS-PAGE under non-reducing (− DTT) or reducing conditions (+ DTT) and analyzed by Western blot using an [HA]-specific monoclonal antibody (mAb). Following separation under non-reducing conditions, proteins bands migrating according to 17, 32, and 55 kDa were detected in lysates of [HA] M2(H7N1)-transfected cells which likely represent the mono meric, dimeric, and tetrameric forms of the protein, respectively ( Fig. 2A). Only the 17 kDa band was detected when the gel was run under reducing conditions, indicating that the oligomeric forms of mature [HA] M2 R are normally stabilized by intermolecular disulfide bonds. The mutation S31N did not affect [HA] M2 R oligomerization but reduced band intensity, most likely because this mutation affected M2 protein stability. The [HA] M2 P protein of bat IAV H18N11 appeared as multiple protein species under non-reducing conditions with the most prominent one corresponding to approximately 24 kDa ( Fig.  2A). Under reducing conditions, only the 12 kDa form was detected suggesting that this band represents the monomeric [HA] M2 P and the 24 kDa band the dimeric form of the protein.
The primary sequence of [HA] M2 P and [HA] M2 R have similar predicted molecular masses of 12.6 and 12.4 kDa, respectively. Nevertheless, the apparent molecular masses of the two proteins differed significantly. While the [HA] M2 P monomer appeared as a 12 kDa band, which corresponded to its predicted molecular mass, the M2 R monomer migrated as a 17 kDa band ( Fig. 2A, + DTT). To figure out which domain would be responsible for this discrepancy, expression plasmids were generated encoding chimeric M2 proteins in which the ED, TM, AH, and CT domains of M2 P (each domain labeled by P) and M2 R (each domain denoted by R) were shuffled (Fig. 2B). Western blot analysis of transfected cell lysates showed that the chimeric M2(RRR-P) protein containing the CT domain of M2 P migrated significantly faster than M2 R and this shift was even more pronounced with the chimeric M2(RR-PP) protein (Fig. 2B). In contrast, the TM and ED domains had no significant effect on migration of the M2 proteins [compare M2(RR-PP) with M2(R-PPP), M2(PPPP) with M2(R-PPP), and M2(RRRR) with M2(R-P-RR)]. Vice versa, the M2(PPP-R) protein migrated slower than M2(PPPP) but faster than M2(PP-RR), while M2(P-RRR) showed the same apparent molecular mass than M2(PP-RR). The M2(PPP-R), M2(PP-RR), M2(P-RRR) proteins and in particular the M2(P-R-PP) protein showed lower band intensities compared to the parental M2(PPPP), suggesting that these chimeric proteins might be less stable. Collectively, these data suggest that determinants in the long cytoplasmic domain were responsible for the differential migration of the M2 proteins.

The M2 P (H18N11) protein is expressed at the cell surface at low levels
The IAV M2 protein plays an active role in virus budding and release (44). Furthermore, M2 is incorporated into the viral envelope and participates in virus uncoating (22,23,25,54). For all these functions it is necessary that the M2 protein is transported to the cell surface where the budding process takes place. To   although not as pronounced as the [ (Fig. S1), suggesting that [HA] M2 P was less stable. By swapping the AH and CT domains with those from [HA] M2 R , the total expression levels increased, suggesting that these domains had a stabilizing effect on the protein. When the AH and CT domains of [HA] M2 R were replaced by those from [HA] M2 P , the overall expression levels stayed the same, but the percentage of [HA] M2(RR-PP) at the cell surface declined. Compared to the parental [HA] M2 R protein, [HA] M2(R-P-RR) protein revealed a slightly reduced overall expression level, while the relative cell surface expression level was even higher. In contrast, the [HA] M2(P-R-PP) protein showed a reduced relative cell surface expression compared to the parental [HA] M2 P protein. Collectively, these findings suggest that the CT and AH domains have an impact on [HA] M2 protein stability and/or subcellular localization.
To analyze subcellular localization of M2 proteins, BHK-21 cells were transfected with expression plasmids encoding [HA]-tagged M2 proteins. The cells were fixed with formalin at 24 hours p.t., permeabilized with Triton-X100, and analyzed by indirect immunofluorescence. Cells transfected with expression plasmid encoding [HA] M2 P protein showed a punctate signal for the [HA] epitope in the cytosol, while the plasma mem brane was not labeled with the antibody (Fig. 3B, upper panel). In contrast, a strong labeling of the plasma membrane was observed for [HA] M2 R (Fig. 3B, lower panel), indicating that this M2 protein accumulates at the cell surface. The transfected cells were also labeled with a rhodamine conjugate of wheat germ agglutinin (WGA), a lectin that binds to sialic acid and N-acetylglucosaminyl residues and labels cellular compartments in which glycoconjugates are highly abundant such as the Golgi apparatus (55). Interest ingly, in cells expressing [HA] M2 R the Golgi apparatus appeared dispersed (Fig. 3B, lower panel), a phenomenon that has been linked to high M2 proton channel activity in the Golgi membrane (56). In contrast, the Golgi apparatus seemed to be intact in cells expressing [HA] M2 P (Fig. 3B, upper panel). We hypothesized that [HA] M2 P is transported to the plasma membrane but does not stay there because it is rapidly internalized. To test this hypothesis, we incubated the [HA] M2 P transfected cells with the [HA]-specific antibody for 1 hour at 37°C prior to fixation with formalin. When the non-permeabilized cells were subsequently incubated with the Alexa Fluor 488-labeled secondary antibody, none or very faint specific labeling of the cells was observed (Fig. 3C, right hand panel). However, when the fixed cells were permeabilized with Triton-X-100 and then incubated with the secondary antibody, several punctate vesicle-like structures were labeled, indicating that the primary anti-[HA]tag antibody was internalized when incubated with the cells at 37°C. In contrast, [HA] M2 R was labeled with the [HA]tag antibody at the cell surface, irrespective of the permeabilization of the fixed cells (Fig. 3C, left hand panel).

Bat IAV M2 proteins do not preserve the metastable conformation of acidsensitive HA
Previous work has demonstrated that glycoprotein (G)-deficient vesicular stomatitis viruses (VSV) expressing the envelope glycoproteins HA and NA of the highly pathogenic avian influenza viruses (HPAIV) A/chicken/Yamaguchi/7/2004 (H5N1) and A/turkey/Italy/ 4580/1999 (H7N1), respectively, were able to replicate in vitro in an autonomous manner (57). By generating viruses with a similar gene arrangement, we show that the chimeric virus VSV∆G(HA R :NA R :GFP), which encodes for the HA R and NA R envelope glycoproteins of A/FPV/chicken/Rostock/1934 (H7N1) (Fig. 4A), did not replicate autonomously on MDCK-II cells (Fig. 4B). In contrast, VSV∆G(HA R :NA R :M2 R ), which additionally encodes the M2 R protein (Fig. 4A), replicated efficiently on MDCK-II cells reaching maximum titers of almost 10 7 f.f.u./mL at 48 hours post-infection (p.i.) (Fig. 4B). To figure out whether PCA of M2 R conferred infectivity to VSVΔG(HA R :NA R :M2 R ), we infected MDCK-II cells using a multiplicity of infection (m.o.i.) of 1 f.f.u./cell and subsequently incubated the cells with AMT. At 24 hours p.i., the number of infectious viruses released into the cell culture supernatant was determined (Fig. 4C). Infectious titers of VSVΔG(HA R :NA R :M2 R ) dropped by 4.5 log 10 when the cells were treated with 10 µM of AMT and by 3.4 log 10 and 0.7 log 10 when the cells were incubated with 1 µM and 0.1 µM of AMT, respectively, demon strating that VSVΔG(HA R :NA R :M2 R ) was inhibited by AMT in a dose-dependent manner. In contrast, VSVΔG(HA R :NA R :M2 R [S31N]) expressing the S31N mutant M2 R protein, was highly resistant to inhibition by AMT (Fig. 4C). VSV∆G(HA R (mb):NA R :GFP), a chimeric virus encoding a modified HA R with a monobasic (mb) proteolytic cleavage site replicated independently of M2 (Fig. 4D), indicating that proteolytic activation in the secretory pathway primes HA R for pH-dependent conformational change. These findings are in line with previous reports which demonstrated that the HA of the highly pathogenic avian influenza virus A/FPV/chicken/Rostock/34 (H7N1) is particularly sensitive to low pH-triggered conformational change in the secretory pathway. In order to stay in its biologically active metastable conformation, the HA of this virus strain highly depends on M2 proton channel activity (27,(58)(59)(60)(61)(62). As the infectivity of VSVΔG(HA R :NA R :M2) relies on the incorporation of biologically active, fusion-competent HA into the viral envelope, we regard this virus as a suitable biological probe to indirectly measure M2 proton channel activity. When we refer to M2 PCA in the following text, the capacity of the M2 protein to preserve the metastable conformation of acid-sensitive HA is implied.
VSVΔG(HA P :NA P :GFP) encoding the HA P and NA P proteins of A/flat-faced bat/ Peru/033/2010 (H18N11) has recently been shown to replicate on MDCK-II cells in the presence of trypsin (63). We compared the replication kinetics of VSVΔG(HA P :NA P :GFP) with that of VSVΔG(HA P :NA P :M2 P ), which additionally encoded for the M2 P protein, but did not notice any enhanced virus replication in MDCK-II cells in the presence of M2 P (Fig.  4E). Since HA P (H18N11) contains a monobasic proteolytic cleavage site, it may not be primed for a pH-dependent conformational change and thus may not rely on M2 PCA. We, therefore, generated two additional chimeric viruses, VSVΔG(HA P (pb):NA P :GFP) and VSVΔG(HA P (pb):NA P :M2 P ), both encoding a modified HA P protein with a polybasic (bp) proteolytic cleavage site (63). Both viruses replicated independently of trypsin on MDCK-II cells but showed no significant difference in replication kinetics (Fig. 4F), suggesting that the HA P protein does not require help from the M2 P protein to maintain its metasta ble conformation.

Bat IAV M2 proteins imperfectly rescue the metastable conformation of acidsensitive HA
To assess the PCA of bat IAV M2 proteins, we replaced the M2 R gene in VSVΔG(HA R :NA R :M2 R ) with either M2 G (H17N10) or M2 P (H18N11) and performed multicycle replication kinetics on HEK 293T cells. Compared to VSVΔG(HA R :NA R :M2 R ), the chimeric viruses replicated only inefficiently, suggesting that the bat IAV M2 proteins have lower PCA than conventional IAV M2 proteins (Fig. 5A). Furthermore, VSVΔG(HA R :NA R :M2 G ) replicated less efficiently than VSVΔG(HA R :NA R :M2 P ), suggesting that the M2 P protein has higher PCA than the M2 G protein. The impact of M2 proteins on chimeric VSV replication was also analyzed by a plaque size assay on BHK-21 cells (Fig.  5B), which was found to correlate with virus titers (Fig. S2). While VSVΔG(HA R :NA R :M2 R ) produced large plaques in the cell monolayer, the infection with VSVΔG(HA R :NA R :GFP) remained localized to the primary infected cell. Plaques formed by VSVΔG(HA R :NA R :M2 G ) were not significantly larger than those formed by VSVΔG(HA R :NA R :GFP), indicating very low PCA. Infection with VSVΔG(HA R :NA R :M2 P ) resulted in plaques that were significantly larger than those formed by VSVΔG(HA R :NA R :M2 G ) but still smaller than those induced were either left non-permeabilized (-Tx) or were permeabilized with Triton X-100 (+Tx) prior to staining with anti-mouse AlexaFluor 488 and WGA-rhodamine.
Full-Length Text by VSVΔG(HA R :NA R :M2 R ). We also analyzed chimeric VSV which encoded the ion channel proteins of other viruses. Chimeric VSV encoding the BM2 protein of influenza B virus (B/Lee/40) or the M2 protein of A/Yamaguchi/7/2004 (H5N1) formed plaques that were as large as those formed by VSVΔG(HA R :NA R :M2 R ). However, the M2 protein of A/bat/ Egypt/381OP/2017 (H9N2) produced plaques that were smaller than those formed by VSVΔG(HA R :NA R :M2 R ) but still larger than those produced by VSVΔG(HA R :NA R :M2 G ). These findings suggest that M2 proteins derived from bat IAV may exhibit reduced proton channel activity compared to M2 proteins of conventional IAV.

N31S mutant bat IAV M2 proteins augment replication of chimeric VSV
Our previous work with recombinant IAV containing six genomic RNA segments of bat IAV H17N10 and 2 chimeric genomic segments encoding HA and NA of a classical IAV (H7N7) revealed that Asn31 of bat M2 protein rapidly mutates to a serine or histidine residue when the virus was grown in avian DF-1 cells or embryonated chicken eggs (41). We hypothesized that the H7N7 HA, which contains a multibasic cleavage site, might have required an M2 protein with high PCA to stabilize its metastable conformation in the secretory pathway. To test this hypothesis, we introduced the N31S mutation into the bat IAV M2 proteins and generated chimeric VSV. Both VSVΔG ( ) and found that both viruses were inhibited by the drug in a concentration-dependent manner (Fig. 5E). Collectively, these results indicate that the N31S mutation significantly increased the PCA of bat IAV M2 proteins, whereas the capacity of conventional M2 R protein to support replication of VSVΔG(HA R :NA R :M2 R ) stayed similarly high, irrespective of a serine or asparagine at position 31.

The AH and CT domains do not affect PCA of chimeric M2 proteins
Our analysis of the expression levels of chimeric M2 proteins suggested that the CT and AH domains have a strong impact on stability and cellular targeting of the proteins (see Fig. 3). To study the role of these domains in PCA, we analyzed VSVΔG(HA R :NA R :X) encoding chimeric M2 genes by plaque size assay (Fig. 6A). Viruses encoding hybrid M2 proteins with the domain architecture PPP-R, PP-RR, or P-RRR formed smaller plaques compared to M2 P , indicating that these chimeric M2 proteins exhibited lower PCA than M2 P , although their stability was generally higher (PPP-R, PP-RR) or at least equally high compared to M2 P (see Fig. 3A; Fig. S1). Virus encoding hybrid M2 protein with the domain architecture RRR-P or RR-PP formed plaques of similar size compared to M2 R , indicating that the CT as well as the AH domain of M2 P did not affect PCA, although they affected M2 cell surface expression levels (see Fig. 3A). Finally, when the M2 domain architecture was changed to R-PPP, the chimeric M2 was associated with a mid-range plaque size which was in between the plaque size linked to the chimeric viruses encoding M2 R and M2 P . This finding suggests that the PCA of M2 P is enhanced if the ectodomain is replaced by the one of M2 R . In contrast to PPP-R and PP-RR, for which the N31S substitution caused a dramatic increase of plaque size, the N31S mutation did not affect plaque formation by chimeric VSVΔG(HA R :NA R :X) encoding the M2(R-PPP) protein (Fig. 6B). Collectively, the analysis of chimeric M2 PCA suggested that the CT and AH domains affect protein stability and subcellular targeting but have a minor effect on proton channel activity. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001).

Stabilization of the bat IAV M2 tetramer does not increase PCA
As specified in Fig. 1A, the M2 proteins of H17N10 and H18N11 lack Cys17 in the ectodomain, which may result in lower stability of the M2 tetramers as only one intramolecular disulfide bond can be formed (Fig. 2C) (45). To study the impact of a second cysteine residue in the M2 P ectodomain on the stability of the tetrameric form of the protein, we introduced the S17C substitution into the [HA] M2 P cDNA. Addition ally, the [HA] M2 P (N31S) and [HA] M2 P (S17C/N31S) mutants were generated as the N31S substitution was linked to enhanced PCA. In addition, we destroyed cysteine 17 of [HA] M2 R by introducing the C17S substitution, while S31N and C17S/S31N served as controls. Western blot analysis of transfected cell lysates revealed a lower abundance of stabilized tetramers M2 R (S31N) compared to M2 R (Fig. 7A). A reduced proportion of Full-Length Text stabilized tetramers and an increase in monomeric protein was observed for the mutant M2 R (C17S), while M2 R (C17S/S31N) appeared only as monomeric and dimeric but not as tetrameric form. The M2 P (N31S) substitution did not lead to the appearance of M2 P tetramers in the non-reducing SDS-PAGE (Fig. 7A). However, an increased formation of stabilized tetramers was observed for both M2 P (S17C) and M2 P (S17C/N31S).
To study the impact of a cysteine 17 on M2 PCA, we generated chimeric VSVΔG(HA R :NA R :M2) encoding the corresponding wild type and mutant M2 proteins. Analysis of plaque size formation by these viruses revealed that VSVΔG[HA R :NA R :M2 P (S17C)] induced smaller plaques than VSVΔG(HA R :NA R :M2 P ) and VSVΔG[HA R :NA R :M2 P (N31S)] (Fig. 7B), indicating that the new cysteine at position 17 did not increase but rather reduced plaque size. In contrast, infection with VSVΔG[HA R :NA R :M2 R (C17S)] led to reduced plaque size compared to VSVΔG(HA R :NA R :M2 R ), and this was even more reduced for VSVΔG[HA R :NA R :M2 R (C17S/ S31N)]. Thus, while M2 R requires Cys17 to fully preserve the metastable conformation of acid-sensitive HA, a cysteine at this position is detrimental for the corresponding activity of M2 P .

M2 P protein is rapidly internalized from the cell surface of H18N11-infected cells
Analysis of transfected BHK-21 cells suggested that M2 P is rapidly internalized from the cell surface while M2 R stays at the plasma membrane (see Fig. 3C). To rule out the possibility that M2 P associates with other viral proteins to stay at the plasma membrane, we infected MDCK-II cells with either H18N11, H5N1 or H18N11/H5N1(6:2), a chimeric virus containing six out of the eight RNA segments of H18N11 while two modified RNA segments encoded the HA and NA proteins of H5N1. At 24 hours p.i., the live cells were incubated for 1 hour at 37°C with the E10 antibody directed to the conserved N-terminal region of M2, and with antibodies specifically recognizing either the H5 or H18 hemag glutinin. Thereafter, the cells were washed, fixed with formalin, and either permeabilized or not permeabilized with Triton X-100 prior to incubation with a fluorescent secondary antibody (Fig. 8). Although several non-permeabilized cells were stained positive for the HA antigen of H18N11, only some of these cells also showed labeling of M2 P by the E10 antibody. In permeabilized cells, the E10 antibody bound to intracellular vesicles, indicating that the antibody/antigen complex had been internalized during the incuba tion at 37°C. MDCK-II cells that had been infected with H5N1 showed a completely different phenotype: All infected (HA-positive) cells were also labeled at the plasma membrane with the E10 antibody. In permeabilized cells, the E10 staining was still confined to the cell surface indicating that the M2 Y (H5N1) protein stayed at the plasma membrane and was not internalized at 37°C. When the MDCK-II cells were infected with the chimeric H18N11/H5N1(6:2) virus, a large fraction of the infected cells showed M2 P expression at the surface. In permeabilized cells, the E10 antibody was bound to intracellular vesicles, however, a considerable fraction was still found at the cell surface, most likely because one of the H5N1 glycoproteins interfered with M2 P internalization (see Discussion).

Detection of proton flux across the plasma membrane of infected cells
To address the question of whether cell surface M2 P would show PCA, we took advant age of MDCK-II cells stably expressing the pH-sensitive yellow fluorescent protein (YFP) (Fig. 9A). Mock-infected cells exhibited very little (5%) change in YFP fluorescence after 2 minutes incubation at pH 5.5 (Fig. 9B). However, MDCK-YFP cells showed a marked decline of fluorescence at 14 hours p.i. with rH5N1 if the cells were incubated with buffer adjusted to pH 5.5 (Fig. 9B). The influx of protons was not inhibited by amantadine if the cells were infected with rH5N1-M2(N31) encoding an AMT-resistant M2 protein (Fig. 9B, left hand graph). However, almost complete inhibition of YFP quenching was observed if the cells were infected with rH5N1-M2(S31) encoding the AMT-sensitive M2(N31S) mutant (Fig. 9B, right-hand panel). These findings indicate that the observed proton flux is mediated by the M2 protein and is not a consequence of virus-mediated damage of the plasma membrane. When MDCK-YFP cells were infected with rH18N11/H5N1(6:2) and incubated at pH 5.5, a progressive drop of fluorescence intensity was observed that reached 64% of the initial intensity after 2 minutes of incubation (Fig. 9C). Significant proton flux was also observed with cells that had been infected with rH18N11/H5N1(6:2)-M2(S31) and rH18N11/H5N1(6:2)-M2(N31). YFP quenching in cells infected with rH18N11/H5N1(6:2)-M2(S31) was sensitive to inhibition by AMT since this virus encoded the AMT-sensitive M2 P (N31S) mutant (Fig. 9C, cf. left and right-hand graphs). In contrast to rH5N1 and H18N11/H5N1(6:2), infection of MDCK-YFP cells with rH18N11 wild-type virus did not result in quenching of the YFP fluorescence signal (Fig. 9D), suggesting that proton flux across the plasma membrane of rH18N11-infected cells is much lower than that of rH18N11/H5N1(6:2) infected cells.

Uncoating of rH18N11 in MDCK-II cells is delayed
The PCA of classical IAV M2 proteins is known to play an important role in virus uncoat ing by allowing protons to enter the virus interior (24). To see whether bat IAV M2 proteins are also important for virus uncoating, we generated recombinant rH18N11 encoding either the AMT-resistant M2 P (N31S) or the M2 P (A30T/N31S) protein. Characteri zation of these viruses by multistep replication kinetics on MDCK-II cells revealed that the rH18N11-M2(N31S) replicated to ten-fold lower titers compared to wild-type rH18N11 (Fig. 10A), although this M2 P mutant was found to exhibit higher PCA than wild-type M2 P protein in the VSVΔG(HA R :NA R :M2) system (see Fig. 5D). However, the rH18N11-M2(A30T/ N31S) mutant, which was unable to preserve the metastable conformation of acidsensitive HA (see Fig. 5C), was attenuated by three log 10 compared to wild-type rH18N11 (Fig. 10A), suggesting that PCA is important for efficient replication of this virus.
The role of M2 PCA in virus entry was analyzed by a plaque reduction assay using AMT as specific inhibitor of M2 PCA. For rH5N1-M2(N31), the number of infectious foci was not affected by AMT in line with the resistance of this M2 mutant to AMT (Fig. 10C, left panel). However, rH5N1-M2(S31) encoding an AMT-sensitive M2 protein, was significantly inhibited by the drug. On the contrary, AMT had only a modest inhibitory effect on plaque formation by the AMT-sensitive rH18N11-M2(S31) with 10 µM of AMT leading to only a 50% reduction of the number of infectious foci (Fig. 10C, central panel). We hypothesized that this attenuation might be due to a lower incorporation rate of the mutant M2 P (N31S) protein into the viral envelope. However, mass spectrometry analysis of purified rH18N11-M2(S31) and rH18N11-M2(N31) particles showed that the mutant M2 P (N31S) protein was equally well incorporated into the viral envelope as the wild-type M2 P protein (Fig. S3). Finally, we analyzed the effect of AMT on entry of chimeric rH18N11/H5N1(6:2). AMT had a significant inhibitory effect on the entry of the AMTresistant rH18N11/H5N1(6:2)-M2(N31) (Fig. 10C, right panel), which might be due to a higher sensitivity of this virus to the lysosomotropic properties of AMT. However, the inhibitory effect of AMT on the rH18N11/H5N1(6:2)-M2(N31) encoding the AMT-sensitive i., the cells were exposed to pH 5.5, and quenching of YFP fluorescence was recorded.
It has previously been reported that AMT-mediated inhibition of M2 PCA results in delayed IAV uncoating and consequent later start of viral protein expression (64). We hypothesized that the low PCA of bat IAV M2 proteins would also lead to a slower virus entry process. To test this hypothesis, we infected MDCK-II with either rH5N1, rH18N11, and rH18N11/H5N1(6:2) and lysed the cells at 2, 4, 6, and 8 hours p.i.. Western blot analysis of the cell lysates showed that NP antigen could already be detected at 4 hours p.i. with rH5N1, while in cells infected with rH18N11 and rH18N11/H5N1(6:2) the NP antigen was not detected before 8 and 6 hours p.i., respectively (Fig. 10D). These findings support the idea that the low PCA of bat IAV M2 proteins may result in delayed virus uncoating.

DISCUSSION
In the present study, we provide evidence that the M2 proteins of bat IAV H17N10 and H18N11 differ from the M2 proteins of conventional IAV in several aspects: (i) The bat IAV M2 proteins have highly divergent primary sequences compared to the highly conserved M2 protein of conventional IAV. (ii) The M2 P (H18N11) protein demonstrated significantly reduced capacity to protect acid-sensitive HA from premature low pH-induced confor mational change and this effect was even more pronounced for M2 G (H17N10). (iii) The N31S substitution dramatically increased the capacity of bat IAV M2 proteins to rescue acid-sensitive HA. (iv) The M2 P (H18N11) protein is readily internalized from the plasma membrane. (v) The AH and CT domains of M2 were found to affect cell surface expression levels. All these special features of bat IAV M2 proteins may reflect functional differences between the bat and conventional IAV M2 proteins.
Using chimeric VSVΔG(HA R :NA R :M2), we observed that the M2 proteins of bat IAV H17N10 and H18N11 did not fully preserve the metastable conformation of the acidsensitive HA R in the secretory pathway (see Fig. 5A and B), indicating that the PCA of the bat IAV M2 proteins might be too low to sufficiently raise the pH in this compartment. In line with this finding, we previously observed that the N31S mutation was rapidly acquired when the chimeric bat IAV rH17N10/H7N7(6:2), which encoded for the HA and NA glycoproteins of A/seal/Massachusetts/1-SC35M (H7N7), was passaged in mammalian or avian cells (41). As the N31S substitution was shown in our study to dramatically enhance rescue of acid-sensitive HA by bat IAV M2 proteins (see Fig. 5C and D), this mutation was most likely selected in rH17N10/H7N7(6:2) to protect the HA (H7N7) from premature low pH-induced conformational change in the secretory pathway.
S31N is the prevalent mutation in conventional IAV M2 proteins, which confers resistance to AMT (65). The crystal structure of the drug-resistant M2(S31N) proton channel showed that the N31 side chains point directly to the center of the channel pore where they form a hydrogen-bonded network that disrupts the drug-binding site (10). Although it is unlikely that bat IAV have ever been exposed to AMT, the N31 residue is conserved in the M2 proteins of all bat IAV isolated so far and also confers resistance of bat IAV M2 proteins to AMT (41). We speculate that N31 has been maintained in bat IAV M2 proteins not because it confers resistance to AMT but rather because it keeps the PCA of these proteins low. It has recently been shown that IAV infection activates NLRP3 inflammasome through trans-Golgi network (TGN) dispersion and that high M2 PCA is responsible for this effect (56). In line with these previous findings, the Golgi apparatus stayed intact in M2 P -transfected cells, whereas it appeared dispersed in cells expressing

Full-Length Text
Journal of Virology the conventional M2 R protein (see Fig. 3B). Thus, a low PCA may be regarded as a strategy to circumvent cytotoxic effects and inflammation that are normally associated with high proton-conducting activity (28)(29)(30)(31)(32)(33), and this may favor a persistent infection of bats without causing disease (40). In contrast to bat IAV M2 proteins where the N31S substitution had a dramatic protective effect on the conformation of acid-sensitive HA, the reciprocal S31N substitution did not affect the corresponding activity of M2 R (H7N1) (see Fig. 5C and D) and similar observations were made with the M2 proteins of A/Udorn/72 (H3N2) (9) and rH5N1 (see Fig. 9C). We speculate that the bat IAV M2 proteins adopt a completely different conformation compared to conventional M2 proteins: the channel is probably tightly closed and only partially opens upon exposure to low pH. The N31S mutation may change the conformation of the channel protein in such a way that the channel pore can more easily open upon exposure to low pH. However, structural analysis of the wild type and N31S mutant of the bat IAV M2 proteins is needed to prove this hypoth esis. Surprisingly, when rH18N11/H5N1(6:2)-infected cells were analyzed for PCA, the corresponding activity of M2(N31)-and M2(S31)-encoding viruses was similarly high (see Fig. 9C). The reason for this phenomenon is not known yet, but co-clustering of M2 with HA at the plasma membrane might have had an impact on M2 protein conformation as well (66).
A well-established function of IAV M2 proteins is to preserve the metastable conformation of acid-sensitive HA in the acidic milieu of the secretory pathway. There are two features that may render an HA protein highly sensitive to low pH, and this is exemplified by the HA R of A/FPV/chicken/Rostock/1934 (H7N1). First, the irreversible conformational change of HA R is triggered at a relatively high pH threshold (pH ~5.9) (67). Second, HA R contains a multibasic cleavage site that is cleaved by the TGN resident prohormone convertase furin (68). If this cleavage site was changed into a monobasic sequence motif, HA R was not cleaved by furin, but required exogenously added trypsin for activation. The chimeric VSVΔG(HA R :NA R :GFP) encoding this HA R cleavage mutant replicated as efficiently as VSVΔG(HA R :NA R :M2 R ) (see Fig. 4D), indicating that it did not rely on M2 PCA. Thus, only HA which is proteolytically processed in the secretory pathway may be at risk for a low-pH induced premature conformational change. The HA proteins of bat IAV H17N10 and H18N11 possess a monobasic cleavage site, which is recognized by the plasma membrane-resident transmembrane serine protease TMPRSS2 (69). It is not known whether the bat IAV HA proteins would also be cleaved by a bat homolog of the human airway protease (HAT) that resides in the Golgi and was shown to activate the HA of pandemic 2009 IAV (H1N1) (70). The situation of HA P (H18N11) proteolytic cleavage in the Golgi can be mimicked by introducing a polybasic cleavage site into the bat IAV HA proteins (63). However, even in this context, M2 did not improve replication of chimeric VSVΔG(HA P (pb):NA P :M2 P ) when compared to the M2 P -deficient VSVΔG(HA P (pb):NA P :GFP) (see Fig. 4F). The modified HA G (H17N10) and HA P (H18N11) proteins containing multibasic cleavage motifs were shown to induce membrane fusion when exposed to pH 5.4 and 5.6, respectively (63), suggesting that the bat IAV HA proteins are relatively acid-stable and are probably not at risk of a premature low-pH induced conformational change in the Golgi. A similar observation has been made with the equine IAV A/Cornell/74 (H7N7) strain. The acid-stable HA of this virus is also cleaved in the TGN but does not rely on M2 PCA to keep its metastable conformation in the secretory pathway (71). Based on these findings we conclude that bat IAV do not depend Full-Length Text on M2 PCA to protect their HA proteins from a premature conformational change during transport through the acidic compartments of the secretory pathway.
A major function of the IAV M2 ion channel protein is to assist virus uncoating by allowing protons to enter the interior of the virion, which in turn facilitates the dissocia tion of the M1 protein from the inner leaflet of the viral envelope and from the ribonu cleoprotein complexes (22,54,72). To study the impact of M2 PCA for virus uncoating, we generated recombinant IAV encoding either AMT-sensitive M2(S31) or AMT-resistant M2(N31). A plaque reduction assay showed that entry of the AMT-sensitive rH18N11/ H5N1(6:2)-M2(N31S) was significantly inhibited by low concentrations of AMT (1 µM), indicating that M2 P PCA supported virus uncoating. Surprisingly, rH18N11-M2(S31) entry was only marginally affected by AMT, although this virus encoded for the same AMTsensitive M2 P (S31). We noticed that rH18N11-M2(S31) replicated to significantly lower titers compared to wild-type rH18N11-M2(N31). This attenuated phenotype was not due to impaired incorporation of M2 P (N31S) into the viral envelope (see Fig. S3), but might be due to an overall more compromised budding efficacy of the mutant virus. Interest ingly, the N31S substitution affected the fitness of rH18N11-M2(S31) but not that of rH18N11/H5N1(6:2)-M2(S31), suggesting that the defect caused by the N31S substitution could be somehow compensated by the envelope glycoproteins of H5N1 (see discussion below). When the N31S substitution was combined with the mutation A30T, M2 P lost its capacity to rescue acid-sensitive HA (see Fig. 5C), which resulted in severely compro mised replication of rH18N11-M2(A30T/N31S) (see Fig. 10A), suggesting that M2 PCA is important for efficient uncoating. Nevertheless, rH18N11-M2(A30T/N31S) still replicated at low levels, in line with the previous observation that IAV can undergo multiple cycles of replication in the absence of detectable PCA (73). All these experiments point to low PCA of wild-type M2 P protein which might explain the delayed entry of rH18N11 and rH18N11/H5N1(6:2) (64).
We surprisingly observed that in contrast to the conventional M2 R (H7N1) protein, the M2 P (H18N11) protein predominantly localized to intracellular compartments in cells transfected with the corresponding cDNA or infected with rH18N11. Only a small fraction of M2 P was detected at the plasma membrane at steady state. Using chimeric M2 proteins, we found that the AH and CT domains were responsible for this low cell surface expression level and it is tempting to speculate that these domains contain signals that mediate rapid internalization of the M2 protein from the plasma membrane. In fact, the M2 proteins of both H17N10 and H18N11 contain one or two potential internalization signals that match the sequence motif Y-X-X-Φ. However, these motifs are located in the membrane-proximal part of the cytoplasmic domain and it is currently not known whether they can actually interact with the AP-2 adaptor protein that participates in clathrin-mediated endocytosis from the plasma membrane (74). In addition, the bat IAV M2 proteins contain several lysine residues in the distal part of their CT domains (see Fig. 1A) that might be targeted by the E3 ubiquitin ligase MARCH8 for ubiquitinationdependent endocytosis and degradation (75).
Bat IAV (H18N11) replicated to only low titers on MDCK-II cells (see Fig. 10A). As IAV M2 proteins are major players in virus assembly and budding (51,52,(76)(77)(78), the low-level cell surface expression of the M2 P protein might be at least partially responsible for the low numbers of infectious particles produced. We found that the cell surface expression levels of M2 P were significantly enhanced in cells infected with rH18N11/H5N1(6:2), suggesting that the H5N1 envelope glycoproteins were able to prevent internalization and degradation of M2 P (Fig. 8). Interestingly, treatment of cells with sialidase has been shown to selectively inhibit caveolar endocytosis (79), while the cellular membrane-bound sialidase NEU3 has been shown to inhibit clathrin-mediated endocytosis (80). Thus, it might be that the sialidase activity of the H5N1 NA protein is responsible for the M2 P internalization block. M2 P protein that accumulates at the plasma membrane might assist in virus budding and release more efficiently than M2 protein that is internalized. If this idea holds true, it would explain why chimeric rH18N11/H5N1(6:2) replicated to higher titers than rH18N11 which is devoid of sialidase activity (35,81). It would also explain why infection of MDCK-II cells with H18N11 in the presence of exogenous sialidase resulted in higher virus titers (39).
In summary, our experimental work using acid-sensitive HA and YFP in combination with the specific M2 proton channel inhibitor amantadine suggest that the M2 proteins of H17N10 and H18N11 bat IAV have unusually low PCA, which might represent a strategy by which toxic effects that are associated with high PCA such us triggering the inflammasome or disturbing cellular ion homeostasis are avoided (31)(32)(33). Furthermore, the rapid internalization of M2 from the plasma membrane and the lack of sialidase activity which could interfere with M2 internalization might represent a viral strategy to evade the host immune responses by keeping the levels of viral particles and the amount of viral antigens that are presented to the immune system low. All this may favor chronic or persistent infections in bats.

Production of immune sera
Immune serum directed to the HA antigen of A/chicken/Yamaguchi/7/2004 (H5N1) has been produced in a previous study by immunization of chickens with VSV*ΔG(HA H5 ) vector vaccine (83). For production of immune serum directed against the H18N11 HA antigen, 7-week-old New Zealand white rabbits (Charles River, Lyon, France) were immunized via the intramuscular route with 10 8 f.f.u. of VSV*ΔG-H18pb, a recombinant VSV vector encoding the HA antigen of A/flat-faced bat/Peru/033/2010 (H18N11) (63). Four weeks after the primary immunization, the animals were immunized a second time using the same vector vaccine, route, and dosage. Three weeks after the second immunization, the rabbits were bled under anesthesia. Sera were prepared by centrifu gation of the coagulated blood and stored in aliquots at −20°C. Animal immunization experiments were performed in compliance with the Swiss animal protection law and approved by the animal welfare committee of the Canton of Bern (authorization number BE-128/19).
For generation of chimeric VSV (serotype Indiana, GenBank acc. no. J02428) encoding for all three IAV envelope glycoproteins, the plasmid pVSVΔG(HA:GFP) encoding an engineered VSV genome with six transcription units (85)  To study the role of proteolytic activation in priming of HA R for low-pH induced conformational change, the polybasic proteolytic cleavage site EPSKKRKKR↓GLF (furin consensus sequence in bold letters) was changed into PEIPKGR↓GLF resulting in the HA R (mb) gene which was then used to replace the HA R gene in pVSVΔG(HA R :NA R :GFP) and pVSVΔG(HA R :NA R :M2 R ). To study the role of the M2 P (H18N11) protein for preserva tion of the corresponding HA P protein, the recombinant plasmids pVSVΔG(HA P :NA P :GFP) and pVSVΔG(HA P :NA P :M2 P ) were generated by replacing the corresponding genes in pVSVΔG(HA R :NA R :GFP) by those from A/flat-faced bat/Peru/033/2010. In addition, the plasmids pVSVΔG(HA P (pb):NA P :GFP) and pVSVΔG(HA P (pb):NA P :M2 P ) were constructed by changing the HA P proteolytic cleavage site NPIKETR↓GLF into NPQRRRKKR↓GLF which matches the polybasic consensus sequence recognized by the cellular protease furin.

Generation of recombinant VSV
The generation of G-deficient recombinant VSV has been performed in principle as described in a previous report (85,89)

Generation of recombinant IAV
Recombinant IAV were generated by transfection of co-cultured HEK-293T and MDCK-II cells with 8 pHW2000 plasmids encoding all eight viral genomic segments and Lipofectamin 2000 (Life Technologies) as transfection reagent (91). At 24 hours p.t., the cells were washed with PBS and maintained in FBS-free medium supplemented with 0.2% (wt/vol) of bovine serum albumin, 1% (vol/vol) of penicillin/streptomycin solution (Life Technologies, cat. no. 15140122), and 1 µg/mL of acetylated trypsin (Merck KGaA, cat. no. T6763). At 6 days p.t., the supernatant of the transfected cells was supplemented with 5% FBS and cleared by low-speed centrifugation (1,200 × g, 10 minutes, 4°C). Infectious virus that had been released into the transfected cell culture supernatant, was passaged twice on MDCK-II or RIE-1495 cells in the presence of acetylated trypsin and 100 mU of C. perfringens sialidase (Merck KGaA, cat. no. N2876). At 2 days p.i., the viruses were harvested, supplemented with 5% FBS, and stored in aliquots at −70°C.
For titration of IAV, MDCK-II or RIE1495 cells were seeded into a 96-well cell culture plate (10′000 cells/well). Serial virus dilutions (10 −1 to 10 −6 ) were prepared in FBS-free MEM medium, and 40 µL of each dilution was added to the wells in duplicates. Fol lowing an incubation for 90 minutes at 37°C, 160 µL of MEM containing 1% (wt/vol) methylcellulose, 2% FBS, and 1% (vol/vol) penicillin/streptomycin solution were added to each well. At 24 hours p.i., the overlay medium was aspirated and the cells fixed for 15 minutes with 3.7% formalin in PBS. Excess formalin was quenched by washing the cells twice with PBS containing 0.1 M glycine and once with PBS. Virus-infected cells were detected by indirect immunofluorescence. For rH18N11-infected cells chicken anti-HA(H18) immune serum (1:400 in PBS) and goat anti-chicken IgG conjugated to Alexa Fluor 488 (1:400; Life Technologies, cat. no. A11039) were used. For cells infected with rH5N1 or rH18N11/H5N1(6:2), rabbit anti-HA(H5) immune serum (1:400 in PBS) and goat anti-rabbit IgG conjugated to Alexa Fluor 488 (Life Technologies, cat. no. A11034) were used for immunostaining. The number of infected cell foci was determined with an Observer.Z1 fluorescence microscope (Zeiss, Feldbach, Switzerland) using a 10× objective. Virus titers were calculated and expressed as f.f.u./mL.
To verify that any mutation introduced into the viral genome was still present after virus passaging, total RNA was either extracted from infected cell lysate (NucleoSpin RNA Mini Kit, Macherey-Nagel, Oensingen, Switzerland, cat. no. 740955) or from infectious virus stocks (NucleoSpin Virus, Macherey-Nagel). Reverse transcription (Invitrogen SuperScript III first-strand synthesis system, Life Technologies, cat. no. 10308632) was performed using the Uni-12 oligonucleotide (5′-AGCAAAAGCAGG-3′) for priming which is complementary to the 3′ end of the conserved region of the viral cRNA (92). Subse quently, viral genomic segments were amplified with segment-specific primers (92) using a high-fidelity DNA polymerase (Phusion high-fidelity DNA polymerase, Life Technolo gies, cat. no. F-530). Sanger DNA sequencing of the amplicons was performed using the BigDye Terminator v3.1 cycle sequencing kit (Life Technologies, cat. no. 4337458) and analyzed using a SeqStudio Genetic Analyzer System (Thermo Fisher Scientific, Bremen, Germany).

YFP-based assay to measure M2 proton channel activity
The cDNA encoding the yellow fluorescent protein was amplified from the pcDNA6.2/ C-YFP-DEST Vector (Life Technologies, cat. no. V35720) by PCR using gene-specific forward and reverse primers harboring KpnI and HindIII endonuclease restriction sites at the 5′ prime ends, respectively. The amplicon was treated with KpnI and HindIII restriction endonucleases and ligated into the respective cloning sites of the pCEP4 plasmid (Life Technologies, cat. no. V04450). RIE-1495 cells were transfected with the recombinant pCEP4-YFP plasmid and selected for 14 days with 250 µg/mL of hygromycin B (Invivogen, Toulouse, France; cat. no. ant-hg-1) A cell clone expressing high levels of YFP was selected and used for subsequent ion channel studies.
RIE1495-YFP cells were seeded into 96-well cell culture plates (2 × 10 4 cells/well, six replicates for each experiment). The cells were cultured for 24 hours at 37°C and infected with IAV using an m.o.i. of 2 f.f.u./cell. At 14 hours p.i., the cells were washed once with 200 µL/well of MES buffer, pH 7.4 (50 mM MES, 100 mM NaCl, 50 mM KCl, and 1 mM CaCl 2 ), and incubated with this buffer (50 µL/well) for 5 minutes at 21°C. YFP fluorescence was recorded for 2 minutes on a GloMax Discover microplate reader (Promega, Dübendorf, Switzerland) with an excitation line at 475 nm and an emission filter at 500-550 nm. If the fluorescence signal did not change during this time by more than 5%, the plasma membrane was regarded as intact and the cells were incubated with 50 L/well of MES buffer adjusted to pH 5.5. As soon as the buffer was added, the plate was returned to the microplate reader, and YFP fluorescence was recorded for 2 minutes at 5 seconds intervals. For inhibition of M2 proton channel activity, the cells were treated for 30 minutes with 20 µM of amantadine in MEM medium which was also present in the subsequent incubations with MES buffers adjusted to pH 7.4 and pH 5.5, respectively. The relative light units (RLU) measured were normalized to the initial RLU that were recorded when MES buffer pH 5.5 was added. The mean values and standard deviations (SDs) of the normalized RLUs were calculated for the six replicate measurements.

Flow cytometric analysis of M2 cell surface expression
BHK-21 cells were seeded at 3 × 10 5 cells per well into 6-well plates and maintained at 37°C for 24 hours. The cells were transfected with 3 µg/well of pCDNA6-M2 plasmid using 6 µL/well of Lipofectamine 2000 reagent (Life Technologies). At 4 hours p.t., the cell culture supernatant was aspirated and replaced with fresh GMEM medium supplemen ted with 5% FBS. At 20 hours p.t., cells were washed with Ca 2+ /Mg 2+ -deficient PBS (PBS -/-) and suspended in PBS -/after placing μthem for 10 minutes at 37°C. Cells were stained with the LIVE/DEAD Fixable Violet Dead Cell Stain Kit (Life Technologies, cat. No. L34964) for 30 minutes at ambient temperature, washed with PBS, and then divided into two fractions. The first cell fraction was fixed/permeabilized using the BD Cytofix/Cytoperm Kit (Becton Dickinson, Basel, Switzerland, cat. no. 554714) and stained with a mouse monoclonal antibody directed to HA epitope (Merck KGaA, Darmstadt, Germany; cat. no. SAB272196, 1:3000). The second fraction of live, non-fixed cells was incubated for 20 minutes on ice with the HA epitope-specific mAb (1:100), washed once with PBS, and incubated for 15 minutes with goat anti-mouse IgG conjugated with AlexaFluor-647 (Life Technologies, cat. no. A-21235). The cells were washed, fixed as above, and analyzed with a FACSCanto II flow cytometer (Becton Dickinson). Flowjo v9.1 software (Treestars, Inc., Ashland, OR, USA) was used for analysis of intracellular and cell surface expression of M2 protein.

Indirect immunofluorescence analysis
BHK-21 cells (10 5 cells/well) were seeded into 24-well cell culture plates that had been coated with collagen (Merck KGaA, Darmstadt, Germany, cat no. 125-50). The following day, the cell culture medium was replaced with 500 µL of fresh GMEM with 5% FBS and the cells were transfected with 1 µg of pCDNA6- [HA] M2 plasmid and 2 µL of Lipofectamine 2000 transfection reagent according to the manufacturer's protocol. At 16-20 hours p.t., the cells were incubated for 1 hour either at 4°C or at 37°C with a mAb directed to the [HA] epitope (1:100; Merck KGaA, cat. no. SAB272196). The cells were washed 3 times with 1 mL/well of PBS and fixed for 15 minutes at 21°C with 3.6% formalin in PBS. Subsequently, the cells were permeabilized for 5 minutes with 0.25% (vol/vol) of Triton X-100 in PBS, or were left non-permeabilized. The primary antibody bound to the [HA] M2 antigen was detected by incubating the cells for 1 hour with goat anti-mouse IgG conjugated to Alexa Fluor-546 (1:500; Life Technologies, cat. no. A11030). Finally, the cells were stained for 5 minutes at 37°C with 4′,6-diamidino-2-phenylindole (DAPI, 0.1 µg/mL; Merck KGaA, cat. no. 9542) and mounted in ProLong Gold Antifade Mountant (Thermo Fisher, cat. no. P10144). Image acquisition was performed on the Nikon A1R confocal microscope (Nikon Europe B.V., Amsterdam, The Netherlands) using the 60×/1.4 NA oil objective.
Indirect immunofluorescence analysis was also performed with MDCK-II cells that had been infected with IAV (m.o.i. of 1 f.f.u./mL). For detection of cell surface M2 protein at 24 hours p.i., a mAb (clone E10) directed to the M1/M2 N-terminus was used (1:100; Merck, cat. no. MABF2165).

Mass spectrometry
MDCK-II cells were seeded into six T-150 flasks (8 × 10 6 cells/flask) and cultured for 24 hours at 37°C. The cells were infected with rH18N11 or rH18N11-M2(N31S) using an m.o.i. of 0.3 f.f.u./cell. At 48 hours p.i., IAV released into the cell culture supernatant was collected and purified by ultracentrifugation through a density gradient according to the protocol from Hutchinson et al. (94). Briefly, infectious cell culture supernatant was clarified by a first centrifugation step at 2,000 × g for 30 minutes at 4°C, and then by a second centrifugation step at 18,000 × g for 10 minutes and 4°C). The clarified supernatant was placed onto a 10% (vol/vol) OptiPrep density gradient medium (Merck KGaA, cat. no. D1556) cushion in NTC buffer (100 mM NaCl, 20 mM Tris-HCl, pH 7.4, 5 mM CaCl 2 ) and pelleted (112,000 × g, 90 minutes, 4°C) using an SW-32 Ti Rotor (Beckman Coulter, Nyon, Switzerland). After resuspending the pelleted virus, it was placed onto a 10%-40% OptiPrep (wt/vol) gradient and centrifuged at 209′000 × g for 150 minutes at 4°C using an SW-41 Ti rotor (Beckman Coulter). The virus migrated in the gradient as a visible band which was carefully aspirated with a needle, resuspended in NTC buffer, and pelleted by ultracentrifugation. The pelleted virus was finally suspended in 150 µL of NTC buffer. Fifty microliters of purified virus were denatured in lithium dodecyl sulfate (LDS) sample buffer by heating for 10 minutes at 95°C prior to loading on a 10% polyacrylamide gel. Polyacrylamide gel electrophoresis was performed for 5 to 10 minutes until the whole sample had entered the gel. The gel was fixed for 15 minutes with 20% ethanol/ 10% acetic acid and stained overnight with QC Colloidal Coomassie (Life Technologies, cat. no. LC6025). After watering the gel for 1 hour, protein bands were excised, added to a tube with 20% ethanol, and further processed by the mass spectrometry core facility of the University of Bern. The proteins were subjected to in-gel digestion with trypsin according to the method described by Gunasekera et al. (95). Peptides were run on the LTQ-orbitrap XL Mass Spectrometer (Thermo Fisher Scientific) and data interpretation was performed with MaxQuant v2.3.1.0 using the published protein sequences of A/flat-faced bat/Peru/033/2010 (H18N11) as reference (GenBank accession nos.: CY125942-CY125949).

Statistical analysis
Data were represented as the mean ± SD. Statistical analysis was performed using the GraphPad Prism program package v8. The statistical tests used to evaluate the data are indicated in the respective figure legends.