Ebola Virus Glycoprotein Needs an Additional Trigger, beyond Proteolytic Priming for Membrane Fusion

Background Ebolavirus belongs to the family filoviridae and causes severe hemorrhagic fever in humans with 50–90% lethality. Detailed understanding of how the viruses attach to and enter new host cells is critical to development of medical interventions. The virus displays a trimeric glycoprotein (GP1,2) on its surface that is solely responsible for membrane attachment, virus internalization and fusion. GP1,2 is expressed as a single peptide and is cleaved by furin in the host cells to yield two disulphide-linked fragments termed GP1 and GP2 that remain associated in a GP1,2 trimeric, viral surface spike. After entry into host endosomes, GP1,2 is enzymatically cleaved by endosomal cathepsins B and L, a necessary step in infection. However, the functional effects of the cleavage on the glycoprotein are unknown. Principal Findings We demonstrate by antibody binding and Hydrogen-Deuterium Exchange Mass Spectrometry (DXMS) of glycoproteins from two different ebolaviruses that although enzymatic priming of GP1,2 is required for fusion, the priming itself does not initiate the required conformational changes in the ectodomain of GP1,2. Further, ELISA binding data of primed GP1,2 to conformational antibody KZ52 suggests that the low pH inside the endosomes also does not trigger dissociation of GP1 from GP2 to effect membrane fusion. Significance The results reveal that the ebolavirus GP1,2 ectodomain remains in the prefusion conformation upon enzymatic cleavage in low pH and removal of the glycan cap. The results also suggest that an additional endosomal trigger is necessary to induce the conformational changes in GP1,2 and effect fusion. Identification of this trigger will provide further mechanistic insights into ebolavirus infection.


Introduction
Ebolaviruses cause severe hemorrhagic fever in humans and nonhuman primates with 50-90% lethality. No specific vaccines or treatments for ebolavirus infection have yet approved for human use [1][2][3][4][5]. Among the five different members of the ebolavirus genus, Zaire ebolavirus (ZEBOV) and Sudan ebolavirus (SEBOV) are the most lethal and are the most commonly associated with outbreaks among humans [6]. The virus displays a trimeric glycoprotein (a class I fusion protein) on its surface, termed GP 1,2 , which is solely responsible for attachment and internalization of the virus [7,8]. The glycoprotein is initially expressed as a single polypeptide that is then cleaved by furin in the producer cell to yield two disulphide-linked subunits termed GP1 and GP2 [9]. Of these, GP1 attaches to target cells while GP2 drives the fusion of viral and host cell membranes for the delivery of viral RNA into the host cells. The GP 1,2 trimer is extended by heavily glycosylated mucin-like and ''glycan cap'' regions that are attached to the top of GP1 by a single polypeptide and reach upwards and outwards toward the target cell. The extensive glycocalyx provided by these domains and other glycans on GP1 and GP2 may shield the complex from immune surveillance and/or play an additional role in the natural host reservoir [10].
Numerous studies have revealed that the 450 kDa trimeric GP 1,2 is further proteolytically cleaved, after entry into target cells, by the endosomal cathepsins L and B [11][12][13][14]. This trimming operates on the loop formed by residues 190-213 in GP1 and yields a ,39 kDa fragment containing the N-terminal portion of GP1 (prior to the cleavage site) and all of GP2. Cleavage is thought to expose the receptor-binding region (RBR) on the remaining GP1 core and enhance fusion of viral and cellular membranes [11,13,15]. Cathepsins L and B cleave at slightly different sites. Cathepsin L cleaves at residue 201 [15,16] and is sufficient to remove the mucinlike domain and glycan cap. Cathepsin B deletes additional residues N-terminal to the site of cathepsin L cleavage, removing an additional ,1 kDa of mass from GP1 [11,15].
Cleavage by the combination of cathepsins L and B can be functionally mimicked by thermolysin [12,15]. Thermolysin . * cleaves GP around residue 190, leaving residues 33-190 of GP1 and all of GP2 [15] that together assemble a ,39 kDa GP 1,2 core. Although it is known that these cleavage events are usually required for infection, the structural manifestation of enzymatic cleavage is as yet unclear.
Here we demonstrate by antibody binding and peptide amide hydrogen-deuterium exchange mass spectrometry (DXMS) that priming of the GP 1,2 ectodomain and endosomal pH themselves are insufficient for triggering the conformational changes necessary for fusion, and that an additional trigger must be required in the infected cell.

Protein expression and purification
The design of a construct amenable to high-level expression of Zaire ebolavirus GP 1,2 (ZEBOV-GP 1,2 ) and KZ52 is described previously [17,18]. Briefly, ZEBOV-GP 1,2 has an N-terminal HA tag and comprises residues 33-632 with the mucin and transmembrane regions (residues 313-465 and 633-676) deleted, and a T230V mutation that significantly improves expression yields. The protein was transiently expressed in HEK293T cells and purified using an anti-HA column followed by size exclusion chromatography. ZEBOV-GP 1,2 is cleaved by thermolysin (75 mg/ml) overnight and the cleaved protein (ZEBOV-GP 1,2CL ) is separated from released fragments by size exclusion chromatography using a Superdex-200 10/300 GL column equilibrated in PBS buffer. To obtain a complex between ZEBOV-GP 1,2CL and KZ52, the proteins were mixed in a ratio of 1:1.5 by weight respectively and incubated on ice for 2 h. The complex was separated from excess KZ52 by size exclusion chromatography (Suppl. Figure S1A).
A construct for Sudan ebolavirus GP 1,2 (SEBOV-GP 1,2 ; strain Gulu) was cloned into the pDISPLAY vector and contains a Nterminal HA tag for purification. The construct comprises residues 33-637 with the mucin and transmembrane regions (residues 314-472 and 638-676) deleted. SEBOV-GP 1,2 is more sensitive to cathepsin and thermolysin cleavage and it is difficult to obtain sufficient yields of cleaved, wild-type SEBOV-GP 1,2 for DXMS studies. Hence, for production of homogeneous SEBOV-GP 1,2CL , a furin cleavage site was introduced with the mutation 205 RKKR 208 into the 190-213 loop upon which cathepsins and thermolysin act (gift of Paul Bates, University of Pennsylvania). This protein is thus ''born'' cleaved twice by furin in the producer cell: one furin-cleavage site separates GP1 from GP2, while the second removes the glycan cap from GP1. This ''born-cleaved'' protein was transiently expressed in HEK293T cells supplemented with a plasmid encoding excess furin (in a glycoprotein-tofurin plasmid ratio of 4:1 by weight). The expressed protein was purified using an anti-HA column. Further size exclusion chromatography illustrates it is the same size as cleaved wildtype GP 1,2 .
Effect of pH on KZ52 binding to GP 1,2CL Binding of KZ52 to ZEBOV-GP 1,2CL under various pH conditions was determined by ELISA. Briefly, aliquots of ZEBOV-GP 1,2CL were buffer exchanged to a range of pH (pH 4.2, 4.6, 5.0, 5.4, 6.0, 6.6 and 7.0, each obtained from a 1 M stock solution pH screen (Hampton) using Millipore Ultramax 10 K cutoff centrifugal concentrators [17]), and incubated in 10 mM buffer and 150 mM NaCl at room temperature for 2 h. The proteins were subsequently neutralized in 100 mM Tris-HCl buffer pH 7.5 (note that any conformational changes in GP2 leading to fusion are irreversible). The changes in pH were monitored by litmus paper.
0.25 mg of the protein per well was adsorbed (,50 ml/well) overnight at 4uC in a 96-well microtiter plate. The wells were washed with PBS containing 0.05% Tween (PBS-Tween) and blocked with 100 ml/well of 3% bovine serum albumin (BSA) in PBS for 1 h at room temperature. The wells were washed (with PBS-Tween) and incubated with 50 ml of the antibody between 0.1 and 1 mg/ml for 1 h, washed again, and incubated with HRP conjugate (1:2000 dilution in 1% BSA and PBS). The wells were washed thoroughly with PBS-Tween, and bound HRP was detected by incubation with 50 ml/well of 1:1 solution of TMB substrate (Pierce). The reaction was terminated by addition of 50 ml/well of 2 N H 2 SO 4 . The resulting solutions were analyzed by a UV/Vis spectrophotometer microtiter plate reader at 450 nm.

DXMS methodology -Optimization of fragmentation conditions
Prior to the exchange experiments, disulfide reduction and protein denaturation conditions were identified that produced an optimal pepsin fragmentation pattern under exchange-quench conditions. It was found that incubation with quench buffer with TCEP and denaturant, followed by dilution of denaturant prior to exposure to pepsin gave the best peptide probe coverage map. For ZEBOV, 1 ml of ZEBOV-GP 1,2 or ZEBOV-GP 1,2CL stock solution (at 5.3 mg/ml and 5.5 mg/ml respectively) was diluted into 1 ml of non-deuterated buffer (8.3 mM Tris, pH 7.5, 150 mM NaCl, in H 2 O) at 0uC. Next, 3 ml of 1 M TCEP, 6.4 M GuHCl, pH 2.85 (quench buffer) was added and the sample was incubated for 5 min on ice. Then 15 ml of 0.8% formic acid (FA), 16.6% glycerol (quench diluent) was added to the denatured sample prior to protease digestion.

Author Summary
Ebolavirus causes often fatal hemorrhagic fever in humans and nonhuman primates. During infection, the virus is internalized into the low pH endosomes prior to the delivery of viral RNA to the infected cell. Cleavage by endosomal cathepsins of the heavily glycosylated mucinlike domain and glycan cap from the ebolavirus surface glycoprotein GP 1,2 is an essential step in infection. The effect of cleavage and the low pH of the endosomes on the conformation of GP 1,2 is as yet unknown. To investigate the effect of priming, we cleaved the mucinlike domain and glycan cap of Zaire ebolavirus (ZEBOV) GP 1,2 with thermolysin and engineered a mutant of Sudan ebolavirus (SEBOV) GP 1,2 that is cleaved with furin. We demonstrate by DXMS and antibody binding studies that cleavage of the mucin domain and glycan cap and incubation at low pH are insufficient to trigger the conformational changes of GP 1,2 that effect fusion. Unraveling the trigger that leads to the conformational change of GP 1,2 to its fusogenic form will enhance the understanding of ebolavirus infection and pinpoint key sites for therapeutic intervention.
Effect of Priming of Ebolavirus Glycoprotein www.plosntds.org 15 ml of quench diluent (0.8% FA, 16.6% glycerol) was added. The samples were then frozen on dry ice and stored at 280uC for further analysis.
In order to demonstrate the ability of our methods to detect differences in deuteration of these proteins, GP1 and GP2 of SEBOV-GP1,2 CL were separated from each other by adding 10 mM DTT and heating at 37uC for 30 min. The disulfidereduced sample was prepared and analyzed in a similar fashion as SEBOV-GP1,2 CL for DXMS experiments ( Figures S4A-S4C in supplementary information).
All samples were subsequently thawed at 4uC and passed over an AL-20-pepsin column (16 ml bed volume, 30 mg/ml porcine pepsin (Sigma)), at a flow rate of 20 ml/min. The resulting peptides were collected on a C18 trap (Michrom MAGIC C18AQ 0.262) and separated using a C18 reversed phase column (Michrom MAGIC C18AQ 0.2650 3 mm 200 Å ) running a linear gradient of 0.046% (v/v) trifluoroacetic acid, 6.4% (v/v) acetonitrile to 0.03% (v/v) trifluoroacetic acid, 38.4% (v/v) acetonitrile over 30 min with column effluent directed into an LCQ mass spectrometer (Thermo-Finnigan LCQ Classic). Data were acquired in both data-dependent MS1:MS2 mode and MS1 profile mode. SEQUEST software (Thermo Finnigan Inc.) was used to identify the sequence of the peptide ions. DXMS Explorer (Sierra Analytics Inc., Modesto, CA) was used for the analysis of the mass spectra as described previously [19].

Deuterium exchange experiments
Functional deuteration of ZEBOV glycoproteins were performed by diluting 1 ml of ZEBOV-GP 1,2CL or ZEBOV-GP 1,2 stock solution into 3 ml of D 2 O buffer (8.3 mM Tris, 150 mM NaCl, in D 2 O, pD READ 7.2) at 0uC. At 10 sec and 1000 sec, 6 ml of ice-cold quench buffer (6.4 M GuHCl, 1 M TCEP, pH 2.85) was added, samples incubated on ice for 5 min, 10 ml quench diluent (0.8% FA, 16.6% glycerol) was added, and samples frozen on dry ice. The equilibrium-deuterated back-exchange control samples were prepared by diluting 1 ml of ZEBOV-GP 1,2CL or ZEBOV-GP 1,2 into 3 ml of 1% formic acid in 99.9% D 2 O with incubation overnight at room temperature, cooled to 0uC and mixed with 6 ml quench buffer, incubated for 5 min, supplemented with 10 ml quench diluent, and frozen and further processed as above.
Functional deuteration of SEBOV-GP 1,2 was performed by first diluting 2 ml of SEBOV-GP 1,2 with 2 ml of non-deuterated buffer The centroids of the isotopic envelopes of nondeuterated, functionally deuterated, and fully deuterated peptides were measured using DXMS Explorer and then converted to corresponding deuteration levels with corrections for backexchange [20].

Figure preparation
All figures in the manuscript were generated using Pymol and Adobe photoshop.

Structure and Neutralizing antibodies of ZEBOV and SEBOV
The crystal structure of Zaire ebolavirus GP 1,2 (ZEBOV-GP 1,2 ) has been previously determined in its prefusion form, in complex with an antibody termed KZ52 that was derived from a human survivor of Zaire ebolavirus infection [17]. The structure reveals that antibody KZ52 binds a conformational epitope on GP 1,2 where the GP1 and GP2 subunits meet, and illustrates the specificity of KZ52 for the prefusion state of ZEBOV-GP 1,2 . In addition, we have also determined the crystal structure of the prefusion form of Sudan ebolavirus GP 1,2 (SEBOV-GP 1,2 ) bound to a novel neutralizing antibody termed 16F6 [21]. 16F6, like KZ52, recognizes a conformational epitope on the prefusion form of SEBOV-GP 1,2 that bridges the two subunits together. The loop of GP 1,2 upon which both endosomal cathepsins and thermolysin act (residues 190-213) is mobile and disordered in both structures, and is located at the outer (''side'') surface of the trimeric GP 1,2 .
In these prefusion GP 1,2 structures, each GP2 in the trimer wraps around its corresponding GP1 in a metastable state. The GP1s form a ''clamp'' on the metastable prefusion conformation of GP2 and prevent it from springing irreversibly into its fusionactive and more stable six-helix bundle conformation, which has also been crystallized [22,23]. In order for fusion to occur, GP2 must unwrap from each GP1 in the prefusion complex and rearrange to yield the six-helix bundle conformation of GP2 alone.

Priming GP 1,2
To investigate the effects of enzymatic cleavage of GP 1,2 , we digested ZEBOV-GP 1,2 with thermolysin and purified the resulting protein (ZEBOV-GP 1,2CL ) by size exclusion chromatography. Note that the ZEBOV-GP 1,2 used for the studies has a deletion of the mucin-like domain to enhance protein expression. Thermolysin functionally mimics cleavage by endosomal cathepsins, but operates at physiological pH, while cathepsins require low pH (,5.5) for cleavage. Use of thermolysin allows us to decouple the effects of enzymatic cleavage from the effects of low pH. The elution profile of thermolysin-cleaved ZEBOV-GP 1,2CL suggests that ZEBOV-GP 1,2 remains trimeric after cleavage (Suppl. Figure  S1A). SEBOV-GP 1,2 is quite sensitive to proteolysis and cleavage by either cathepsins L/B or thermolysin results in significant degradation, making it difficult to purify sufficient material for DXMS studies. In order to analyze the effects of enzymatic cleavage of SEBOV-GP 1,2 , we engineered a version of SEBOV-GP 1,2 with an additional furin site at residues 206-210 in the loop of GP1 that is operated on by cathepsins and thermolysin (the wild-type furin-cleavage site that separates GP1 and GP2 also remains in this construct). The furin-engineered SEBOV-GP 1,2CL , is stable, is released from the expression host already deleted of its glycan cap and mucin-like domain and structurally mimics enzymatically cleaved SEBOV-GP 1,2 (Suppl. Figure S1B). Like ZEBOV-GP 1,2CL , SEBOV-GP 1,2CL also exists as a trimer.

Analysis of conformational changes by DXMS
DXMS is able to measure the ability of deuterium in solvent water to exchange with hydrogens that are covalently bound to peptide amide nitrogen atoms in a protein [24]. To determine if cleavage of GP 1,2 results in conformational change, we analyzed uncleaved ZEBOV/SEBOV-GP 1,2 and cleaved ZEBOV/ SEBOV-GP 1,2CL proteins by DXMS, obtaining 49%/68% and 68%/83% coverage, respectively. Peptide regions throughout the glycoproteins were analyzed. Peptides covalently linked to glycans Effect of Priming of Ebolavirus Glycoprotein www.plosntds.org of heterogeneous mass are not accessible for measurement with currently employed DXMS methods, and therefore result in small gaps in the sequence coverage. Note that enzymatic cleavage removes nearly all N-linked glycans from GP1 and hence greater sequence coverage is obtained for cleaved GP (68%/83%) than uncleaved GP (49%/68%).
For ZEBOV-GP 1,2 the peptide fragmentation analysis confirms that thermolysin cleaves after the aromatic residues in the disordered loop 190 KKDFFSS 196 and deletes the glycan cap (as observed by Dube et. al. [15]). The deuteration levels of ZEBOV-GP 1,2 and SEBOV-GP 1,2 (measured over a time of 10-1000 sec) are consistent with the respective crystal structures (PDB code 3CSY and 3S88) in that peptide fragments that are buried or that have amide hydrogen atoms involved in hydrogen bonding show low levels of deuteration.
Comparison of uncleaved GP 1,2 with cleaved GP 1,2CL , for both ZEBOV and SEBOV, reveals that no significant changes in deuteration occur upon cleavage for any measured peptide spanning the whole of GP 1,2 , observed over a 10-1000 sec time scale (Figure 1 and S2A-S3C in supplementary information). The deuteration pattern of the residues in the fusion loop (residues 520-540 of GP2) is also identical between the cleaved and uncleaved proteins (Figure 2). The unwinding of the fusion loop is a required early step in the springing of GP 1,2CL to the post-fusion form. Alteration in the deuteration pattern of the fusion loop could indicate the beginning of unwinding of GP1,2 CL , but is not observed after enzymatic cleavage.
In ZEBOV, residues R64, F88, K95, K114, K115, and K140 are critical for binding and are thought to comprise a part of the receptor-binding region [15,25]. Peptides containing the residues R64, F88, K95, K114, and K115 are equally accessible by solvent before and after cleavage and removal of the glycan cap (Figure 3), indicating that no major conformational changes occur at these sites upon cleavage. Note that peptides containing residue K140   DXMS thus suggests that no residues in ZEBOV/SEBOV-GP 1,2CL dramatically change conformation when the mucin-like domain and glycan cap are released. These results also suggest that the glycan cap, itself, does not significantly occlude access by solvent to this site. The glycan cap, however, could block steric access by a protein molecule and receptor access would likely be enhanced by enzymatic removal of the glycan cap.
The epitope of the human neutralizing antibody KZ52 bridges GP1 and GP2 and KZ52 only binds when the subunits are assembled in their prefusion conformation. Conformational changes in GP2, such as those required for fusion, would likely abrogate KZ52 binding. Hence, binding studies of KZ52 to ZEBOV-GP 1,2CL provide additional insights into the structure of ZEBOV-GP 1,2CL . By both size exclusion chromatography and ELISA, we find that KZ52 binds well to cleaved, trimeric ZEBOV-GP 1,2CL (See Figure 4 and Suppl. Figure S1A). An equivalent, GP1/GP2-bridging, prefusion-specific antibody, termed 16F6 [21], that recognizes SEBOV-GP 1,2 also forms a stable complex with SEBOV-GP 1,2CL . The binding of the prefusionspecific conformational antibodies KZ52 and 16F6 indicates that the GP 1,2CL ectodomain from both viruses remains in its pre-fusion state upon thermolysin or furin cleavage of the 190-213 loop. In addition, successful binding of KZ52 to cathepsin L-cleaved ZEBOV-GP 1,2CL with a K D of 1.5 nM has been recently reported by Hood et al. using surface plasmon resonance [16]. By contrast, Shedlock et al. [26] find that KZ52 does not neutralize cathepsin L-cleaved ZEBOV-GP 1,2 that has been pseudotyped onto a viral surface (binding not directly measured in these studies). Hence, it seems that some as-yet-undetermined differences exist between ectodomain GP and viral-surface GP upon cathepsin L cleavage that do not occur upon thermolysin cleavage.
Effect of pH on conformation of primed GP 1,2 The endosomal compartments have an acidic pH ranging from ,5.9-6.0 in the early endosome to ,5.0-5.5 in the late endosome and the role of this low pH in triggering irreversible conformational changes leading to fusion has been speculated. To investigate the effect of pH on the conformation of ZEBOV-GP 1,2CL , we monitored binding of KZ52 to acid pH-treated ZEBOV-GP 1,2CL by ELISA (Figure 4). Bovine serum albumin (BSA), and reduced and denatured GP 1,2 were used as negative controls. Indeed, KZ52 binding is unaffected by incubation of ZEBOV-GP 1,2CL in endosomal pH, suggesting that pH alone does not cause rearrangement of ZEBOV-GP 1,2CL from the pre-fusion state.

Conclusions
The DXMS and ELISA binding studies together suggest that the priming of GP 1,2 of ebolaviruses and the low pH in which priming occurs, are themselves, insufficient for triggering the  Effect of Priming of Ebolavirus Glycoprotein www.plosntds.org conformational changes required for fusion. An additional trigger such as binding of the receptor to cleaved GP or the action of another cellular factor thus appear to be essential for fusion ( Figure 5).
The requirement of enzymatic cleavage for ebolavirus GPs may instead serve different, non-exclusive purposes. Cleavage might simply expose the receptor-binding site for binding to an endosomal receptor. Further, removal of the glycan cap and the heavily glycosylated mucin domain (,75 kDa of protein and carbohydrate) could facilitate membrane fusion by reducing steric barriers to GP2 rearrangement and membrane association. Alternatively, cleavage of the residue 190-213 loop that covers the outside of the fusion loop may remove a flexible tether that anchors the fusion loop in place on the outside of the prefusion trimer. The enzymatic cleavage step of ebolavirus GP 1,2 could indeed be required for one or all of these reasons, but the specific trigger of ebolavirus fusion remains to be identified.