Coronaviruses are large positive-strand RNA viruses with a broad host range (
47,
74). Like other enveloped viruses, CoVs enter target cells by fusion between the viral and cellular membranes, a process mediated by the viral spike (S) protein (
25). The CoV S protein, as characterized to date, consists of two noncovalently associated subunits, S1 and S2. The S1 subunit of the S glycoprotein mediates receptor binding (
12,
82), while the S2 subunit is responsible for driving viral and target cell membrane fusion (
83). The S2 subunit is a prototypical class I viral fusion protein containing common structural features (
11,
26-
28,
92) such as (i) a hydrophobic fusion peptide (
50,
51), (ii) a pair of extended α-helices, specifically 4,3-hydrophobic heptad repeats (HR) (
7,
97), and (iii) a cluster of aromatic amino acids proximal to (iv) a hydrophobic transmembrane anchoring domain.
Although the SARS-CoV S protein shares only 20 to 27% amino acid (aa) homology with the S proteins of other CoVs (
69), recent studies have confirmed that the putative SARS-CoV S2 subunit is also a class I viral fusion protein. Using computational analysis, Gallaher and Garry (
29) first proposed that the portion of the SARS-CoV S protein corresponding to the S2 subunit fit the prototypical model of a class I viral fusion protein based on the presence of two predicted HR regions at the N and C termini of S2 and an aromatic-aa-rich region just prior to the transmembrane anchor domain. Using synthetic peptides analogous to the two HR regions of S2, several groups have demonstrated that SARS-CoV HR1 and HR2 interact with one another to assume a coiled coil conformation (
6,
49,
84,
99). Most recently, Xu et al. showed by crystal structure analysis that the SARS-CoV S protein fusion core forms a typical six-helix coiled coil bundle (
98), as seen with the murine hepatitis virus (MHV) S protein (
97). Furthermore, we have shown that the aromatic-aa-rich region of the SARS-CoV S2 subunit has similar functionality to the aromatic regions of both the human immunodeficiency virus (HIV) transmembrane (TM) glycoprotein (GP) (
80) and Ebola virus (EboV) GP2 (
71), in that peptides analogous to this aromatic region can induce the permeabilization of lipid vesicles (
73). Although the putative fusion peptide of the SARS-CoV S2 subunit has yet to be identified, it has been predicted that the SARS-CoV S fusion peptide lies within the N-terminal region of the S2 portion (residues 851 to 882), preceding HR1 (
84).
Like other enveloped viruses encoding class I viral fusion proteins (
27,
28), it is presumed that SARS-CoV uses membrane fusion mechanisms for viral entry (
98,
99). After binding of the SARS-CoV S1 subunit to the mammalian receptor angiotensin-converting enzyme 2 (ACE2) (
48,
85,
95) and/or CD209L (L-SIGN) (
44), a conformational change in the S protein results in the exposure of an unidentified hydrophobic fusion peptide within S2. As with other class I viral fusion proteins (
27,
28), the fusion peptide is believed to penetrate the target cell membrane, initiating the virion-cell membrane fusion event. Numerous mutagenesis studies of other enveloped viruses encoding class I viral fusion proteins (
8,
23,
32,
38,
43,
51), as well as synthetic peptide studies (
1,
14,
15,
17,
21,
34,
46,
58,
59,
62,
63,
66,
70), have provided substantial evidence of the role of the fusion peptide in initiating membrane fusion. Following insertion of the fusion peptide into the target cell membrane, HR interactions between residues 916 to 950 of HR1 and residues 1151 to 1185 of HR2 (
84) mediate the formation of a six-helix coiled coil bundle (
98,
99). The formation of this structure, also known as the trimer of hairpins, is believed to facilitate the apposition of both the viral and target cell membranes, resulting in fusion and subsequent entry of the viral core into the target cell.
Class I viral fusion proteins generally contain one fusion peptide, located (i) internally (
27,
34,
38), as seen for avian sarcoma virus (ASV) TM and EboV GP2, or (ii) either at or near the N terminus of the protein, as seen for HIV TM and influenza virus hemagglutinin (HA) (
28,
61,
66,
89). Variations in the number of aa and the position within the fusion protein are apparent between fusion peptides of different class I viral fusion proteins; however, distinct features are conserved. In general, fusion peptides are short (16 to 26 residues), hydrophobic sequences (
8,
24,
88) that are rich in alanine, glycine, and phenylalanine residues (
8,
26,
41). The presence of a canonical fusion tripeptide (YFG or FXG) is highly conserved among the fusion peptides of retroviruses, paramyxoviruses, influenza virus, and filoviruses (
21,
26,
64,
70). It is believed that the canonical fusion tripeptide contributes to the functional organization of the fusion peptide itself (
64). Lastly, the presence of a proline residue at or near the center of many fusion peptides has been implicated as critical for the interaction of the peptide with the target cell lipid membrane (
19,
37,
43,
70). Taken together, the presence of these conserved features and the inherent hydrophobicity of fusion peptide sequences allow for their preferential interaction with lipid membranes.
DISCUSSION
Dissecting the mechanisms by which the SARS-CoV S protein mediates fusion between the virion envelope and the cellular membrane could significantly contribute to our understanding of SARS-CoV pathobiology and to the design of antiviral drugs and vaccines. Recent studies have determined that the SARS-CoV S fusion protein is a prototypical class I viral fusion protein in that it contains two 4,3-hydrophobic HR regions responsible for the formation of a six-helix bundle, similar to the fusion proteins of EboV (
53,
86,
87) and the lentiviruses HIV and simian immunodeficiency virus (
10,
13,
53). The formation of this hairpin structure is believed to drive membrane fusion by mediating the juxtaposition of both the viral and cellular membranes. The formation of the six-helix bundle, however, is preceded by the insertion of a hydrophobic fusion peptide located within the N-terminal region of the fusion protein (
36). Insertion of the fusion peptide into the target cell membrane facilitates both target cell membrane disruption and the subsequent formation of the six-helix bundle. Although studies by Luo and Weiss have identified putative fusion peptides of MHV (
50,
51), no fusion peptide has been identified in the N-terminal region of the SARS-CoV S2 subunit.
Using the characteristics of known viral fusion peptides, including their hydrophobicity and aa composition and the presence of a canonical fusion tripeptide, we identified two putative fusion peptides with high interfacial hydrophobicities in the N-terminal region of the SARS-CoV S2 subunit (Fig.
1 and Table
1). SARS
WW-I is located at the extreme N-terminal end of S2, 9 aa downstream of a minimum furin cleavage site. In viral fusion proteins that undergo proteolytic cleavage (e.g., HIV TM and influenza virus HA), fusion peptides are situated at the N-terminal region. Although it has not been conclusively determined whether the SARS-CoV S protein is proteolytically cleaved during maturation, a minimum furin cleavage site is present within the S protein (758R-N-T-R761), and recent studies reported that the SARS-CoV S protein is proteolytically cleaved in vitro (
2,
96). Studies examining the conserved furin cleavage sites in other coronaviruses have shown conflicting results regarding whether cleavage is necessary for infectivity and/or cell-cell fusion (reviewed in reference
18). Most recently, de Haan et al. demonstrated that the S protein of MHV strain A59 is proteolytically cleaved; however, the requirements for cleavage during virion-cell and cell-cell fusion differ (
18).
The second possible SARS-CoV fusion peptide, SARS
WW-II, is similar to the fusion peptides of EboV GP2 (
27,
70) and ASV TM (
27) and the internal fusion peptides of class II viral fusion proteins (
31), as it is located distal to the furin cleavage site but still within the N-terminal region of S2. Despite their different locations within the S2 subunit, both SARS
WW-I and SARS
WW-II contain several features which are conserved among all known viral fusion peptides. For example, both demonstrated a high interfacial hydrophobicity when analyzed with the WWIH scale (Fig.
1C and Table
1), suggesting an inherent propensity to partition into lipid membranes (
94). Similar regions of high interfacial hydrophobicity are apparent in the fusion proteins of HIV and EboV (Fig.
1A and B). Moreover, both SARS
WW-I and SARS
WW-II are rich in alanine, glycine, and/or phenylalanine residues and contain a canonical fusion tripeptide and a proline residue (Table
1), making both peptides potential fusion peptide candidates. Classifying SARS
WW-I and SARS
WW-II as internal or N-terminal putative fusion peptides will depend on deciphering whether and where the SARS-CoV S protein is proteolytically cleaved. It is important that while cleavage of the S protein of CoV appears to enhance fusion (
35,
81), cleavage is not absolutely required for fusion (
4,
5,
39,
75,
76).
Although the transition of the fusion protein core to the six-helix bundle conformation has been shown to mediate fusion by repositioning the cell and viral lipid membranes, the fusion peptide has been implicated as the mediator of the fusion process through its ability to induce the fusion and permeabilization of lipid membranes. We employed biophysical assays to identify which of the two possible SARS-CoV fusion peptides could partition into lipid membranes as well as induce fusion and permeabilization of lipid vesicles. These criteria were chosen based on the experimental approaches used to identify the fusion peptides of several other viral fusion proteins. Only SARS
WW-I strongly partitioned into the lipid membranes of all LUV tested (Table
2). While SARS
WW-II appeared to weakly bind to LUV composed of POPC and PI (9:1) (Fig.
2B), the calculated partition coefficient was significantly lower than that of SARS
WW-I (Table
2). In a FRET fusion assay, we observed a rapid exponential increase in NBD fluorescence following the addition of SARS
WW-I, corresponding to 58% of the vesicles undergoing one fusion event at a P:L molar ratio of 1:10. In contrast, SARS
WW-II caused a marginal increase in NBD fluorescence, indicative of fewer vesicles (∼18%) undergoing membrane fusion events (Fig.
3 and Table
3). Even at a P:L ratio of 1:50, 38% of the vesicles underwent one fusion event in the presence of SARS
WW-I, whereas only 11% of vesicles underwent one fusion event in the presence of SARS
WW-II. The differences in fusion may correlate with the differing capacities of both peptides to partition into lipid membranes. While other studies have shown a correlation between a cation-mediated preaggregation of vesicles and peptide fusion activity (
59,
64,
79), we observed no enhancement of NBD fluorescence for either SARS
WW-I or SARS
WW-II in the presence of 5 mM Ca
2+. Therefore, the peptide-mediated fusion observed was not dependent on a cation-mediated preaggregation of vesicles (data not shown).
When tested in the Tb
3+/DPA leakage assay, SARS
WW-I induced measurable lipid vesicle permeabilization at all P:L ratios tested, whereas SARS
WW-II induced minimal to no observable leakage (Fig.
5). Although both SARS
WW-I and SARS
WW-II showed similar propensities to interact with lipids based on their high interfacial hydrophobicities (Table
1), not all small hydrophobic peptides, even those with high WWIH scores, are capable of membrane disruption (
94). Furthermore, the scrambled peptide of SARS
WW-I did not interact with membranes (Fig.
2C and Table
2) and was unable to induce fusion (Fig.
4) or leakage (Fig.
5) of lipid vesicles, indicating that we measured a sequence-specific membrane disruption mediated by SARS
WW-I. The SARS
WW-I peptide used throughout the aforementioned studies contained an aa substitution at residue 2 (Table
1), where the aromatic aa tyrosine was replaced with tryptophan. The rationale for this conservative aromatic aa replacement was that the intrinsic fluorescence of tryptophan is stronger and thus easier to quantitate spectrofluorometrically than that of tyrosine. To rule out the possibility that the capacity of SARS
WW-I to induce both fusion and leakage of lipid vesicles was a consequence of the aromatic aa substitution, we tested an unmodified peptide in both the FRET fusion and Tb
3+/DPA leakage assays. As expected, the unmodified peptide exhibited activity identical to that of SARS
WW-I in both assays (data not shown).
We also used CD spectroscopy to determine the propensity of the SARS
WW-I peptide to adopt a defined secondary structure (α-helix or β-sheet) upon interaction with lipid membranes. Studies examining the secondary structures of viral fusion peptides have been conflicting. While numerous studies suggest that a predominantly α-helical structure is the single fusion-active conformation adopted by viral fusion peptides (reviewed in reference
21), other studies argue that a β conformation is the active fusion state (
1,
17,
65,
72,
80). CD analysis of the SARS
WW-I peptide showed a propensity for the formation of a β-sheet structure in buffer and upon the addition of LUV (Fig.
6A). The formation of a β-sheet structure was independent of the presence of a cation, which has been shown to be necessary for the fusion peptides of EboV and HIV to assume a defined β-sheet secondary structure as well as to induce the aggregation and fusion of lipid vesicles (
59,
80). The ability of the SARS
WW-I peptide to adopt a secondary structure appears to be necessary for the peptide to interact with lipid membranes. Scrambling the peptide resulted in a loss of the β-sheet conformation (Fig.
6B) and an inability to bind to and interact with LUV. The secondary structure is a strongly driven thermodynamic consequence of membrane partitioning for hydrophobic peptides (
90) which can be influenced by the peptide charge, aa organization, peptide-membrane surface density, lipid composition, and monolayer lateral pressure (
79). Therefore, assumptions regarding the secondary structures of synthetic peptides are dependent on experimental conditions and are far from reflective of the true propensity of a viral fusion peptide to adopt a defined secondary structure in the context of the full glycoprotein. Nonetheless, CD analysis is a useful tool to study the behavior of these peptides in the context of model lipid membranes. The data presented herein for the SARS-CoV putative fusion peptide are consistent with the idea that the biologically relevant activity of viral fusion peptides is based on the propensity to adopt a defined secondary structure upon interaction with lipid membranes.
The biophysical studies presented above provide evidence that SARS
WW-I behaves similarly to the synthetic viral fusion peptides of HIV, EboV, and influenza virus (
14,
17,
21,
46,
58,
59,
62,
63,
65,
66,
70), in that the peptide has the capacity to partition into lipid membranes, adopt a well-defined secondary structure, and induce both fusion and permeabilization of lipid vesicles. The nominal P:L molar ratio was ∼1:25 for both fusion and leakage assays; however, peptide binding assays demonstrated that at this molar ratio, ∼40% of the peptide was actually bound to the lipid membrane under our experimental conditions. This suggests that less peptide is required to induce the observed levels of fusion and leakage. Observations that synthetic peptides corresponding to the fusion peptide domains of class I viral fusion proteins can induce measurable levels of fusion and leakage, separate from the context of the entire glycoprotein, provide support for models of viral fusion in which the fusion peptide and other domains actively facilitate lipid mixing (
3,
73). The transition to the six-helix-bundle state may not only facilitate membrane juxtaposition but also function to align the fusion peptide with the aromatic domain and the transmembrane domain. The alignment of these three domains may then form a continuous track of hydrophobic, membrane-interacting surfaces that provide a low-energy (low-barrier) path for lipid flow and subsequent membrane fusion during the transition to or formation of the six-helix-bundle configuration (
73).
Lastly, since the SARS
WW-II peptide was capable of inducing one-third the amount of fusion and permeabilization of lipid vesicles that the SARS
WW-I peptide induced, this less active sequence may work in conjunction with the SARS
WW-I peptide or may constitute a second fusion peptide. Although its location within the S2 subunit is consistent with the position of the known internal fusion peptides of ASV and EboV, its inability to partition into lipid membranes (Fig.
2 and Table
2) and to adopt a well-defined secondary structure (data not shown) argues against the latter possibility. Although further mutagenesis studies need to be conducted to further verify the role of the SARS
WW-I sequence in vitro, based on the biophysical assays presented above, we propose that residues 770 to 788 constitute the putative fusion peptide of the SARS-CoV S2 subunit. A working schematic of the SARS-CoV class I viral fusion protein, including the putative fusion peptide, the two well-characterized α-helices (HR1 and HR2), the transmembrane anchor, and other conserved structural domains present among many other class I viral fusion proteins, such as the aromatic domain and stem-loop region, is depicted in Fig.
7.