Results and Discussion

The premise behind this study of sGP structure, folding, and stability is that structural elucidation might clarify the poorly understood function of this viral protein. The use of the authentic form (i.e., derived from live EBOV infection) of sGP eliminates any issues related to the fidelity of glycoprotein processing and/or collateral influences by EBOV proteins expressed concomitantly with sGP. Herein, we present the results of extensive physicochemical characterizations of the molecule.

Production and purification of sGP from culture fluid from EBOV-infected Vero E6 cells. sGP was purified to homogeneity by lectin and GFC from culture fluids of EBOV-infected Vero E6 cells incubated in serum-free medium. The advantages of purifying sGP from EBOV-infected Vero E6 (or comparable) cultures is that it offers a simple, rapid, and reproducible means of producing relatively large quantities of authentic protein. GFC, SDS-PAGE, and mass spectrometry analysis (figure 1) indicate that the purified protein is highly (>95%) pure and is the disulfide-linked homodimer sGP in which the subunits are joined in a parallel orientation via 2 intermolecular (C53–C53′ and C306–C306′) cystine bridges.

Primary sequence and posttranslational modifications (PTMs). Confirmation of the encoded amino acid sequence and identification of PTMs requires peptide mapping with complete sequence coverage. Our previous analysis (with sequence coverage of 97%) of tryptic peptides mass-to-charge ratio (m/z) provided convincing evidence for the complete assignment of all disulfide linkages and the detection that all 6 sites bearing the consensus sequence for N-glycosylation (NX- T/S) are fully occupied [8]. In addition to these PTMs, we found that 1 hexose unit is added to peptides T26 and T26T27 (see peptide notation in figure 2). The 2 strong peaks at m/z 2298.19 and 2426.29 observed in figure 1C were matched to T26 and T26T27, respectively, by use of the FindMod tool (http://ca.expasy.org/tools/findmod/) to predict a potential PTM. This search indicated that these peptides bear the C-mannosylation recognition signal WXXW, which potentially explains the increment of 162 atomic mass units (amu), compared with the peptide masses of T26 and T26T27.

Enzymatic cleavage of sGP and [RCM]sGP with trypsin and GluC followed by RP-HPLC helped to further define primary structure and PTMs. Peaks were collected and analyzed by MALDI-TOF MS, and ESI-MS/MS and N-terminal sequencing by Edman degradation were used as indicated. The primary sequence and all previously found PTMs were reconfirmed in this extensive analysis (PTMs highlighted in figures 2 and 3). No evidence for unoccupied N-glycosylation sites, amino acid mutations, or other PTMs, including tryptophan oxidation and O-linked glycans, was found by use of our methodology; however, the question remains open as to whether other techniques or methodologies might reveal otherwise.

C-mannosylation of sGP at position W288. Edman degradation and ESI-MS/MS of RP-HPLC-purified peptides T26 were performed to gain information on the position of the hexose unit. As is shown in figure 4A, the amino acid sequence, as revealed by Edman degradation, matches that predicted for peptide T26, except that no phenylthiohydantoin derivative of tryptophan (PTH-W) eluted at cycle 12, which suggests that W288 has been modified. ESI-MS/MS analysis corroborated this conclusion (figure 4B). T26 was identified as doubly charged ion at m/z 1149.47. MS/MS analysis of this species revealed the sequences GE(C2-Man)WAFWETK (y9*), AFWETK (y6), FWETK (y5), and WETK (y4). The singly charged y9 fragment was also accompanied by a mass loss of 120 amu (y9*-120), known as the fingerprint mass for α-C2-mannosyltryptophan ([C2-Man]W) because of the stability of the CC bond in aromatic C-glycosides [19]. Therefore, we conclude that a mannosyl residue is attached to W288 and that W291 is unmodified. The structure of the modified tryptophan, (C2-Man)W, is shown in figure 3B. sGP is the first viral protein reported to bear this modification, and the sequence motif WAFW (also present in GP) is conserved among all 4 strains of EBOV.

Site-specific N-glycosylation and glycan structures. We extended our analysis to a comprehensive characterization of all N-linked glycans of sGP. We used ZIC-HILIC (a hydrophobic stationary-phase liquid chromatography) in batch mode, a robust method for glycopeptide enrichment [20], to isolate the glycopeptides from the tryptic digests and to determine the specific carbohydrate structures, as well as their sites of attachment to the protein. The resulting glycopeptide-containing eluate was separated by reverse-phase HPLC (RP-HPLC; figure 5A. The HPLC fractions were analyzed by MALDI-TOF MS. After deglycosylation with PGNase, the carbohydrate-free peptides were identified by their mass (or Edman sequence, as in the case of peptide T24) (figure 5A and table 1). Peptide T21 (comprising masses of 6500–8500) contains 2 sites for N-glycosylation (N228 and N238). When GluC was used to digest the peptide, 2 peptides also resulted, each with a single site (referred to in table 1 as ΔT21_220–231 and ΔT21_236–245, respectively); each was purified and analyzed. Thus, all 6 potential sites for N-glycosylation were shown to be occupied.

Multiple forms of glycosyl structure were found at each site, reflecting microheterogeneity in glycosylation (table 1). The sugar composition of the oligosaccharide attached at each glycosylation site was deduced from the difference between the observed mass and the theoretical mass of a corresponding nonglycosylated peptide through the use of the Glycomod Tool provided through the EXPASY Web site. From their estimated carbohydrate compositions, the glycan structures were predicted using GlycoSuite. Resistance of the majority of the oligosaccharides to endoglycosidase H hydrolysis reveals a predominant presence of complex-type glycans, with only few structures that appear to be sensitive to hydrolysis (hybridtype) (the classification is shown in table 1). When necessary, multiple structural possibilities were narrowed down by sequentially digesting the glycopeptides with multiple glycosidases. A representative spectrum is illustrated for the first eluting chromatographic peak (retention time, 22.2 min) (figure 5B and 5C). Sialylated glycans were detected in the positive ion spectra recorded in linear mode (figure 5B). To simplify the analysis, glycans were desialylated (by sialidase treatment) before sequential digestion with other exoglycosidases (figure 5C).

Planar representations of all of the proposed structures are summarized in figure 3B. The majority of the glycoforms of sGP belong to the complex type, with core fucosylated structures with or without bisecting GlcNAc. Sialylated glycans were detected at all 6 sites. Only N257 bears mixed-type oligosaccharides. It is important to note that the proposed structures, although likely, are considered to be hypothetical because MS does not provide information on the stereochemical properties of the linkages, and limited information on linkages was obtained by using specific enzymes.

Biophysical characterization of sGP. The native conformation of sGP was assessed by CD, fluorescence spectroscopy, and MALLS. The far-UV CD spectra of sGP recorded at 20°C have 2 bands, which we interpret as evidence that the protein is folded and contains a mixture of elements of secondary structure (figure 6A). The negative bands at 210 and 220 nm are typical of β-sheet and α-helix content, respectively. The βsheet content is predominant, whereas the a-helical content is estimated to be 15%.

Steady-state tryptophan fluorescence of native sGP, recorded at 20°C with excitation at 280 nm, shows a blue-shifted emission maximum (λ_max) at 335 nm, reflecting asymmetry in the environment of the indole ring(s) (proof that they exist in the context of a tertiary structure) (figure 6B). Unfolding of sGP induced by 6N GdnHCl resulted in a decrease in the fluorescence quantum yield and a red shift of the λ_max from the native value of 335 nm to 347 nm (figure 6B). The position of the λ_max in the unfolded state reflects a highly polar (i.e., solvent-exposed) microenvironment.

sGP was also characterized using GFC monitored by online MALLS, UV absorbance, and RI detectors. With this setup, molar mass distribution, absolute molar masses, and number (M_n), weight (M_w), and z (M_z)—average molar mass moment can readily be computed. Figure 6C shows a plot of sGP molar mass versus elution time. The computed molar mass of the purified sGP is 102,933 ± 3017 Da. This value is in agreement with the calculated molar mass from the sequence (65.3 kDa) plus an estimated carbohydrate content (∼35–41 kDa) of dimeric sGP; this figure is construed as a more accurate measurement than that estimated by GFC or SDS-PAGE (figure 1A and 1B). Moreover, the linear nature of the absolute molar mass distribution across the peak in figure 6C revealed that the sample is monodisperse, with M_w/M_n and M_z/M_n values of 1.006±0.019 and 1.013±0.034, respectively. The range of molar mass is likely due to microheterogeneity associated with variations in N-linked sugar composition.

sGP exists preferably in the dimeric-folded state. At this juncture, we stress that the functionality of sGP remains undefined, and there is the possibility that the folded, dimeric conformation observed in solution may or may not be functionally relevant. There are well-known cases in which aggregation, conformational changes, or misfolding of the protein is closely connected with protein functionality. In this context, we searched for possible mechanisms that could lead to protein oligomerization, conformational change or misfolding.We used biophysical methodology similar to that described by Barrientos et al. [17, 18]. First, we studied GdnHCl-induced and thermally induced (un)folding of sGP. The effect of the chaotropic agent was tested at 3 different protein concentrations (0.05, 0.5, and 5 µg/µL) and 2 pH levels (7.4 and 6.0). CD and fluorescence analysis indicated that chaotropic agent—induced folding/unfolding was completely reversible (figure 6A and 6B, orange traces). MALLS/RI/UV-GFC further indicated that, at low protein concentrations (0.05 and 0.5 µg/µL), refolding to the folded, dimeric state of sGP was quantitatively reversible (figure 6C, orange trace). In addition, we did not detect formation of higher-order aggregates (figure 6D, top panel, black trace). Only at the highest protein concentration (5 µg/µL) were small amounts of higher-order species detected (<2%) (figure 6D, bottom panel). The effect of thermal denaturation was also tested in the same fashion. Natively folded protein was unfolded at 80°C, but, on cooling to 20°C, it refolded to its original state. Both fluorescence and MALLS-GFC analysis indicated reversible folding at the lowest protein concentration (0.05 µg/µL), with no aggregation or precipitation. At a higher protein concentration, however, protein aggregation was detected by both techniques; even so, the native dimer form predominated (∼70%).

We also tested whether a long incubation period at 37°C has an effect on conformational integrity and oligomerization of the dimer. Purified dimeric sGP at 3 different concentrations (0.05, 0.5, and 5 µg/µL) and 2 different pH levels (7.4 and 6.0) was incubated at 37°C for up to 72 h. Tryptophan fluorescence (excitation at 280 nm; emission spectra of 300–400 nm) was measured at 24-h intervals. After 72 h, each sample was subjected to MALLS/RI/UV-GFC analysis. Figure 6D (top panel, black and orange traces) and the insert in figure 6D (top panel) illustrate one example of the results that were obtained. In all cases, the fluorescence spectra of the native protein remained unchanged, and no precipitation or protein aggregation occurred. MALLS-GFC detected only dimeric sGP, and the RI signal indicates no loss of sample. These results indicate that sGP in the dimeric folded state is the predominant and a relatively stable form of the protein. Consistent with these experimental results, we observe that sGP stock solutions have long shelf storage at 4°C, from which we estimate that the dimeric folded form has a half-life of >100 years.

The effect of protein precipitants was also investigated. sGP (1 µg/mL) was precipitated by using a mixture of acidified methanol/acetone and then redissolved in buffer. A mixture of states was detected, including soluble and insoluble high-molecular- weight aggregates and soluble hexamers, tetramers, and dimers. Therefore, high-order aggregates were formed only under aggressive conditions that are unlikely to be encountered physiologically.

sGP appears to have a bizarre folding, free-energy landscape that allows it to refold smoothly and efficiently, under a wide range of conditions, to the preferential dimeric form. We have shown that the dimer is held together by 2 intermolecular disulfide bonds [2, 8], and it is possible that the propensity to revert to a soluble dimer is merely a consequence of this covalent fact. Given that the oligomerization constant is ∼10,000-fold higher than the concentration of sGP in tissue culture (0.5–1 ng/µL), we conclude that aggregation of sGP is not important in the context of EBOV infection, and we postulate that the folded dimeric conformation is the functional form of sGP (if it has any functionality).

Thermodynamic stability of sGP. The stability of dimeric sGP toward chemical and thermal denaturation was further probed by tryptophan fluorescence at the lowest protein concentration (0.05 µg/µL), in an attempt to observe equilibrium between folded and unfolded dimeric states. Figure 6E (top panel) displays the unfolding profile generated for sGP in the presence of increasing amounts of GdnHCl. As is discussed above, folded dimer and unfolded dimer are the only 2 species in equilibrium at the low-protein concentration (0.05 µg/µL). We consider the fractional extent of emission red-shift between its limits as a measure of fractional unfolding; that is, it quantifies a simple 2-state equilibrium between folded and unfolded dimer:

The midpoint of the GdnHCl denaturation curve is 0.9±0.1 mol/L. Analysis of the unfolding curve allows the extraction of thermodynamic parameters: the free energy of unfolding (ΔG) is 1.7±0.1 kcal/mol, with an m value of 1.86±0.1 kcal·mol⁻²·L. These results indicate that sGP is highly susceptible to physicochemical denaturation.

The thermal melting curve (figure 6E, bottom panel) for sGP exhibited 2 well-resolved transitions: 1 at 37°C and 1 at the melting temperature (T_m, 60°C). A transition just at 37°C suggests a physiological role of this state change, perhaps some regional melting or a conformational change, although our low-resolution methods do not shed light on the precise nature of the change.

Probing the structure of sGP with limited trypsin proteolysis. Vulnerability to limited proteolysis by endoproteases is a useful probe of protein structure. To gain insight into the protein tertiary fold of the native sGP, we exposed it to mild trypsin digestion at neutral pH (7.2) and 20°C and then evaluated the effects by use of RP-HPLC and mass spectrometry. Preferential cleavage sites were classified as primary, secondary, or tertiary on the basis of qualitative kinetic evaluation (figure 3A and 3B; shown in orange [primary], pink [secondary], and fuchsia [tertiary]). The primary and secondary sites are located in the N-terminal regions (residues 65–95) and C-terminal regions (residues 276–295), respectively, indicating that these regions are particularly flexible and exposed. The tertiary site (comprising both K190 and K191) was located between the “F2-like region” [8] and the glycan-rich region. Collectively, these sites presumably lie at flexible surface loops or links between subdomains. It is worth noting that the F2-like region (comprising residues V96-K190/K191) is highly resistant to proteolysis. In contrast, the C-mannosylated site is within a protease-sensitive segment. Perhaps this PTM at W288 confers functionally significant hydrophilicity on an otherwise highly hydrophobic region comprising residues W288-A28-F300-W301.

Summary. sGP is the most abundant viral protein in the culture fluid of EBOV-infected cells, a situation that facilitated the isolation and purification of biochemical quantities of pure protein for both structural and functional studies. The characterization of this preparation should prove to be a very useful guide to accelerate other studies of sGP produced from recombinant systems. The work presented here enriches our understanding of the biophysical and biochemical properties of the protein and guides the way to the determination of a highresolution, 3-dimensional structure and the grail of sGP functionality. A high-resolution model is required to inspire a better functional understanding of this unique protein. In the present study, we separated the tasks of characterizing the posttranslational modification of sGP into site-specific glycosylation, glycan structures, and residues most sensitive to protease cleavage. The composite of these tasks, illustrated in figure 3, provides sufficient prerequisite information to guide a domain-by-domain approach to the determination of high-resolution structures of putative 3-dimensional domains in sGP. Our ability to produce and purify milligrams of authentic sGP, reinforced by the wealth of biophysical and biochemical information from this study, offers hope regarding the feasibility of pursuing crystallization and X-ray diffraction of the intact, fully glycosylated sGP molecule (i.e., to conquer).

We also demonstrated that sGP exhibited all of the characteristics of a well-folded protein, with a free-energy landscape that facilitates reversible folding and a strong propensity for disulfide-linked dimeric quaternary structure, even at high concentrations (>5 µg/µL). The low apparent free energy of conformation transition of sGP (ΔG, 1.7±0.1 kcal/mol) suggests that the molecule is well suited as a thermodynamically facile switch, which would allow it to report on relatively subtle changes in milieu. In addition, a conformational transition at 37°C was obviously detected in the thermal denaturing experiment. On the basis of biophysical and biochemical considerations alone, we propose that the property of being a thermodynamically facile switch will be an important clue to reveal sGP functionality.

Ebola hemorrhagic fever in humans is an “accidental event” that presumably played no role in the past and continuing evolution of EBOV in nature. Therefore, the biological function of sGP, to the extent that it has one, is manifested only in its infection of the maintenance host(s), in which sGP arose before the divergence of the 4 EBOV species, and in its retention by natural selection as a result of some contribution it makes to the efficient transmission of virus from one member of the host species to the next. Two EBOV mutants in which sGP is not produced [21] or is produced at low levels as a secondary gene product [22] are significantly more cytopathic than their wildtype counterpart. Therefore, the production of sGP might be an exquisite evolutionary feature for viral survival. Possible roles of sGP in the reservoir host could include interaction with macrophages or dendritic cells that prolong the infection and ameliorate illness. The facile conversion between folded and unfolded states, thanks to the low energetic barrier separating these states, might be involved in switching to a conformational form, which, lacking binding affinity, does not inhibit viral entry into cells in the context of a rapidly progressing disease in human and nonhuman primates.