Biochemical characterization of Ebola virus GP

Transcription

1 University of Iowa Iowa Research Online Theses and Dissertations Spring 2014 Biochemical characterization of Ebola virus GP Nicholas Joseph Lennemann University of Iowa Copyright 2014 Nicholas Joseph Lennemann This dissertation is available at Iowa Research Online: Recommended Citation Lennemann, Nicholas Joseph. "Biochemical characterization of Ebola virus GP." PhD (Doctor of Philosophy) thesis, University of Iowa, Follow this and additional works at: Part of the Microbiology Commons

2 BIOCHEMICAL CHARACTERIZATION OF EBOLA VIRUS GP by Nicholas Joseph Lennemann A thesis submitted in partial fulfillment of the requirements for the Doctor of Philosophy degree in Microbiology in the Graduate College of The University of Iowa May 2014 Thesis Supervisor: Professor Wendy J. Maury

3 Graduate College The University of Iowa Iowa City, Iowa CERTIFICATE OF APPROVAL PH.D. THESIS This is to certify that the Ph.D. thesis of Nicholas Joseph Lennemann has been approved by the Examining Committee for the thesis requirement for the Doctor of Philosophy degree in Microbiology at the May 2014 graduation. Thesis Committee: Wendy J. Maury, Thesis Supervisor Paloma H. Giangrande Aloysius J. Klingelhutz Kevin L. Legge Richard J. Roller

4 ACKNOWLEDGMENTS I would like to thank my mentor, Dr. Wendy Maury for providing an optimal training environment. Throughout my training she has always kept an open door, which allowed for a great balance between independent and guided learning. I would also like to thank Dr. Paloma Giangrande, Dr. Aloysius Klingelhutz, Dr. Kevin Legge, and Dr. Richard Roller for their time and advice. I would like to thank the present members of the Maury lab, Abigail Ashton, Rachel Brouillette, Sven-Moller Tank, Bethany Rhein, and Madeline Walkner. I would like to thank Dr. Andrew Kondratowicz for providing advice and guidance. I would like to thank Bethany Rhein for isolating and characterizing murine macrophages. I would like to thank Madeline Walkner for helping with cloning and characterization of mutants. I would also like to thank previous members Abigail Berkebile and Neil Patel for making several mutants. I would also like to thank Ashley Cooney for helping with mouse studies, making mutants, and characterization of mutants. I would like to thank our collaborators for their contributions to my projects. Robert Davey provided multiple expression vectors and antibodies. Kartik Chandran provided an expression vector. Xiangguo Qiu provided several monoclonal antibodies. Gene Olinger and John Dye provided hours of work in the USAMRIID Biosafetly Level 4 facility. Lastly, I would like to thank my friends and family. Especially my parents, who worked hard to make sure my brother, sister, and I grew up in a supportive environment, which has instilled in me the importance of hard work and perseverance, and has allowed me to pursue all of my interests. ii

5 ABSTRACT Filoviruses cause sporadic outbreaks of highly lethal hemorrhagic fever throughout central Africa. Virus entry is mediated by the sole viral glycoprotein, GP. Furthermore, GP is the main target for neutralizing antibodies. Thus, a better understanding of GP and its functions is critical for the development of antivirals and vaccines. GP contains a high number of N- and O-linked glycans, which shield the majority of the protein. These glycans are required for cell surface interactions with C-type lectins that mediate internalization of the virus. We found that GP1, but not GP2, N-linked glycans were required for efficient entry into cells expressing the C-type lectins: L-SIGN, DC-SIGN, and LSECtin expressing cells, but O-linked glycans were sufficient for ASGPRI- and hmgl-dependent entry. However, filoviruses also utilize phosphatidylserine (PS) receptors, which bind PS in the viral membrane, to mediate entry into host cells. We found that all N-linked glycosylation sites in GP1 could be mutated without significant impact on expression. Furthermore, removal of all N-linked glycans increased entry into a PS receptor-dependent cell line and primary murine macrophages. These results correlated with an increase in sensitivity to proteolysis, which is required within the late endosome/lysosome to expose the receptor-binding domain. Surprisingly, removal of N-linked glycans that directly shield the receptor-binding domain did not allow for binding to the intracellular receptor, NPC1. Thus, proteolytic removal of heavily glycosylated domains within the late endosome/lysosome exposes critical receptor-binding residues that are masked by polypeptides and not N-linked glycans. Furthermore, removal of the conserved N-linked glycan on the heptad repeat 1 region in iii

6 GP2 led to an increase in entry. Conversely, removal of the conserved N-linked glycan on the heptad repeat 2 region decreased entry. Removal of either glycan resulted in a decrease in entry mediated by protease-treated GP. Together, these results suggest N- linked glycans on GP2 are involved in controlling fusion. Interestingly, removal of N- linked glycans masking conserved regions of GP led to a significant increase in convalescent antibody-mediated neutralization. Overall, these results indicate that there is an evolutionary trade-off that results in a decrease in entry efficiency in order to protect virus from the immune system. Analysis of entry mediated by multiple species of ebolavirus indicated that the residue occupying position 95 is a critical determinant of entry. For Ebola virus (EBOV) GP, Sudan virus (SUDV) GP, and Bundibugyo (BDBV) GP, a lysine at position 95 imparts dependence on the cysteine protease cathepsin B. However, a glutamine at this position alleviates this dependence and is found in some early isolates of SUDV. Furthermore, cathepsin B dependence inversely correlated with an increase in sensitivity to protease-mediated degradation of GP. Mutation of K95 to a glutamine in EBOV GP and BDBV GP led to decreased sensitivity to NPC1 and voltage-operated calcium channel inhibitors. Conversely, mutation of the Q95 to a lysine in SUDV GP decreased sensitivity to NPC1 inhibitors and had no impact on voltage-operated calcium channel inhibitors. However, all proteins regardless of the residue at position 95 required NPC1 for entry. Together these results indicate that a single amino acid polymorphism in GP of ebolaviruses has dramatic impacts on entry factor dependence, suggesting potential differences in entry pathways. iv

7 TABLE OF CONTENTS LIST OF TABLES... viii LIST OF FIGURES... ix LIST OF ABBREVIATIONS... xi CHAPTER I INTRODUCTION...1 Filovirus hemorrhagic fever outbreaks...1 Filovirus discovery...1 The viral reservoir and transmission...2 Characterization of filoviral hemorrhagic fever disease and pathogenesis...5 Characterization of filoviruses...6 Classification and phylogenetics...6 Genome organization...7 Ebola virus glycoproteins...7 sgp...7 Structure of GP...8 Glycosylation of GP...10 Host factors involved in filovirus entry...11 Phosphatidylserine receptors...11 TAM family receptor kinases...12 TIM family receptors...13 C-type lectins...14 DC-SIGN and L-SIGN...15 LSECtin...16 ASGPRI...17 hmgl...17 Cysteine proteases...18 Niemann-Pick C Rationale and objectives for current studies...20 List of specific aims...20 Knowledge gained by current studies...21 CHAPTER II COMPREHENSIVE FUNCTIONAL ANALYSIS OF N-LINKED GLYCANS ON EBOLA VIRUS GP Introduction...31 Materials and Methods...33 Cells lines and plasmids...33 N-linked glycan modeling...34 Site-directed mutagenesis...34 Transfection...34 Pseudovirion production...35 Immunoblot of pseudovirions...35 Pseudovirion matrix and glycoprotein quantitation...35 Transduction assay...36 Cell binding assay...36 CatB inhibition assay...37 v

8 Thermolysin sensitivity assay...37 Purification of soluble NPC1 C Loop...37 Soluble NPC1 C Loop GP binding assay...38 Isolation and transduction of murine peritoneal cells...38 Antibody/antisera neutralization assay...39 Results...40 Loss of the N-linked glycan shield does not affect EBOV GP expression, but enhances virion transduction...40 GP1 N-glycans do not impact virion attachment to Vero cells...41 Removal of GP1 N-glycans imparts CatB independence and increases protease sensitivity...41 Removal of glycans shielding the RBD does not allow for NPC1 binding...43 Domain specific N-linked deglycosylation altered C-type lectin utilization...44 N-linked glycans in GP1 are not required for entry into macrophages...46 Removal of N-linked glycans enhances antisera sensitivity...47 Discussion...48 CHAPTER III THE ROLE OF EBOLA VIRUS GP2 N-LINKED GLYCANS...67 Introduction...67 Materials and Methods...69 Cell lines and plasmids...69 Modeling of GP N-linked glycans...69 Site-directed mutagenesis...70 Transfections...70 Production of VSVΔG-GFP pseudovirions...70 Pseudovirion EBOV GP and VSV-matrix quantification...71 Transduction assays...71 Immunoblots...71 Results...72 Mutation of GP2 N-linked glycosylation sites decreases protein expression...72 Removal of GP2 N-linked glycans impacts GP-mediated transduction...73 Removal of N-linked glycans decreases entry mediated by primed GP...74 A single N-linked glycan on GP is sufficient for expression and function...75 GP2 N-linked glycans can mediate C-type lectin-dependent entry in absence of MLD...75 Identification of N-linked glycan species on GP The N-linked glycan at N618 is not required for efficient C-type lectin-dependent entry mediated by GP...77 Discussion...78 CHAPTER IV A SINGLE RESIDUE IN THE GLYCOPROTEIN OF EBOLAVIRUSES ALTERS ENDOSOMAL EVENTS REQUIRED FOR ENTRY...88 Introduction...88 Materials and Methods...90 vi

9 Cells lines and plasmids...90 Site-directed mutagenesis...91 Transfections...91 Pseudovirion production...91 Transduction assays...92 Thermolysin digest reactions...92 Immunoblots...92 Entry factor inhibition assays...93 Soluble NPC1 C loop GP binding assay...93 Generation of CT43-NPC1-myc stable cell line...94 Results...95 SUDV-Bon GP mediates significantly lower transduction than EBOV GP...95 Sequence and structural comparison of GP...96 SUDV GP Q95K mutation increases pseudovirion transduction...97 The residue at position 95 is a determinant of CatB-dependence...97 A glutamine at position 95 increases protease sensitivity of GP...98 EBOV/BDBV K95Q GP mediated entry is less sensitive to inhibitors of NPC Mutation of residue 95 in EBOV GP and BDBV GP decreases sensitivity to inhibition of voltage-operated calcium channels Mutation of residue 95 does not impact entry kinetics Discussion CHAPTER V DISCUSSION AND FUTURE DIRECTIONS The role of Ebola virus GP1 N-linked glycans during infection The role of GP1 N-linked glycans in Ebola virus pathogenesis The role of N-linked glycans on Marburg virus GP and Lloviu virus GP Determine residues masked by glycan cap/mucin-like domain polypeptide that bind NPC Mechanism of enhanced neutralization Re-evolution of GP in vitro and in vivo The role of N-linked glycans on EBOV GP2 during entry Role of GP2 N-linked glycans in triggering of fusion Role of GP2 N-linked glycans in antibody evasion Determine minimum glycosylation requirements for expression and entry Effects of residue occupancy at position 95 in GP Role of residue at position 95 during fusion Further characterize the entry pathway of filoviruses APPENDIX USE OF N-LINKED GLYCAN DEFICIENT EBOV GP1 AS AN IMMUNOGEN REFERENCES vii

10 LIST OF TABLES Table 1-1 Chronological order of filovirus outbreaks in human populations Filovirus nomenclature Nomenclature of N-glycan site mutations in EBOV GP1 and GP1Δmuc...55 viii

11 LIST OF FIGURES Figure 1-1 Filovirus genome organization Linear model comparison of GP and sgp Models of EBOV GP Model of filovirus entry Schematic diagrams of Ebola virus GP Immunoblot analysis of GP1 NGS mutant pseudovirions from HEK293T supernatants Summary of relative expression and Vero cell transduction efficiency of GP1 N-glycan site mutants Expression and entry efficiency of selected EBOV GP N-glycan mutants (see Figure 2-3 in the for a full analysis) Vero cell binding assay of WT or mutant GPs depleted of all N-linked glycans in the core and glycan cap domains (7G) or throughout GP1 (7Gm8G) Effects of N-glycan removal on entry processes Immunoblot of lysates from HEK293T cells transfected with the indicated myc-tagged CLEC CLEC utilization of N-glycan site mutants Comparison of Vero cell transduction efficiency (black) and expression (gray) of GP and 5GΔmuc N-glycan site mutant-mediated entry into murine macrophages Enhanced neutralization of glycan cap N-glycan mutants by cynomolgus macaque anti-ebov IgG Convalescent mouse antisera neutralization of VSV pseudotyped with the indicated NGS mutants Model of N-glycans at conserved sites in EBOV GP Expression and entry of EBOV GP2 NGS mutants Removal of GP2 N-linked glycans decreases entry of thermolysin treated GP The glycan at N563 is sufficient for GP-mediated entry and expression...84 ix

12 3-5 GP2 contains ligands for CLECs Identification of the N-linked glycan species on EBOV GP Characterization of receptor utilization by GP2 NGS mutants Comparison of EBOV GP and SUDV GP mediated entry Sequence alignment and structural mapping of potential interactions of residues at position 95 in GP Mutation of Q95K in SUDV GP rescues poor titer phenotype The presence of lysine vs. glutamine at position 95 alters endosomal protease requirements Mutation of the residue at position 95 alters GP protease sensitivity Cysteine protease dependence of VSVΔG pseudovirions treated with THL Mutation of residue 95 in GP decreases sensitivity to the NPC1 inhibition, but does impart NPC1-independent entry Mutation of K95 in EBOV and BDBV decreases sensitivity to voltage-operated calcium channel inhibition Kinetics of entry Linear model comparison of EBOV GP and MARV GP Addition of a N-linked glycan at a conserved Asn in EBOV GP Purified EBOV sgp-his and trimeric EBOV GPΔmuc-His A-1 Vaccination of Balb/c mice with VSV pseudovirions protects from lethal MA- EBOV challenge x

13 LIST OF ABBREVIATIONS Ab APC ASGPR ASGPRI BDBV CatB CatL CLEC CMV CRD Antibody Antigen-presenting cell Asialoglycoprotein receptor Asialogylcoprotein receptor H1 subunit Bundibugyo virus Cathepsin B Cathepsin L C-type lectin Cytomegalovirus Carbohydrate recognition domain DC-SIGN Dendritic cell-specific intercellular molecule-3 grabbing nonintegrin DMEM DRC EBOV egfp ELISA Endo H ER Dulbecco's modified Eagle's medium Democratic Republic of the Congo Ebola virus Enhanced green fluorescent protein Enzyme-linked immunosorbent assay Endoglycosidase H Endoplasmic reticulum Gas6 Growth-arrest-specific 6 GP GP 0 Glycoprotein Glycoprotein precursor xi

14 GP1 Glycoprotein subunit 1 GP1Δmuc Mucin-like domain deleted glycoprotein subunit 1 GP2 Glycoprotein subunit 2 GP THL HA HEK HIV hmgl Thermolysin treated glycoprotein Hemagglutinin Human embryonic kidney Human immunodeficiency virus Human macrophage lectin specific for galactose/nacetylgalactosamine HOPS HR IC 50 IFL IFNW IFN IFNAR -/- Ig L L-SIGN Homotypic fusion and vacuole protein sorting Heptad repeat 50% inhibitory concentration Internal fusion loop Interferon omega Interferon Interferon-α/β receptor-deficient Immunoglobulin RNA-dependent RNA polymerase Liver/lymph node-specific intercellular molecule-3 grabbing nonintegrin LASV LLOV LSECtin Lassa virus Lloviu virus Liver/lymph node sinusoidal endothelial cell C-type lectin xii

15 M M-CSF mab MARV MHC MLD MWCO NGS NP NPC1 NPC2 ORF PBS PEI Matrix Macrophage colony-stimulating factor Monoclonal antibody Marburg virus Major histocompatability complex Mucin-like domain Molecular weight cut-off N-linked glycosylation site Nucleoprotein Niemann-Pick C1 protein Niemann-Pick C2 protein Open reading frame Phosphate buffered saline Polyethylenimine PNGase F Peptide-N-glycosidase F PRR PS RAVV RBD RESTV SARS SD SEM Pattern-recognition receptor Phosphatidylserine Ravn virus Receptor-binding domain Reston virus Severe acute respiratory syndrome Standard deviation Standard error of the mean xiii

16 sgp SP SUDV TAFV TAM TCID 50 THL TIM TM TU VP VSV WT Soluble glycoprotein Signal peptide Sudan virus Taï Forest virus Tyro3/Axl/Mer 50% tissue culture infectious dose Thermolysin T-cell immunoglobulin and mucin domain Transmembrane domain Transducing unit Viral protein Vesicular stomatitis virus Wild-type xiv

17 1 CHAPTER I INTRODUCTION Filovirus hemorrhagic fever outbreaks Filovirus discovery Outbreaks of Ebola and Marburg hemorrhagic fever occur sporadically, and are mostly isolated to central Africa (Table 1-1). In 1967, laboratory workers in Marburg and Frankfurt, Germany and in Belgrade, Yugoslavia developed severe hemorrhagic fever after being in close contact to African green monkeys or materials obtained from these animals that were exported from Uganda (1). While these individuals were hospitalized, the disease spread to healthcare personnel. Altogether 31 individuals developed disease, resulting in seven deaths. These small outbreaks led to the discovery of the first filovirus, Marburg virus. Eight subsequent outbreaks have occurred, caused by two viruses from the same species as the 1967 outbreaks, Marburg marburgvirus, Marburg virus and Ravn virus (MARV and RAVV, respectively). In 1976, two large outbreaks of viral hemorrhagic fever occurred in northern Zaire (today the Democratic Republic of the Congo, DRC; 318 reported cases, 88% fatality rate) and southern Sudan (284 reported cases, 53% fatality rate) (1). It was discovered that the viruses responsible were antigenically distinct from Marburg virus and were classified as members of a new genus, ebolavirus, named after the Ebola River in the DRC. The two outbreaks were determined to be caused by related, but antigenically distinct viruses, Ebola virus (EBOV) and Sudan virus (SUDV), respectively. Since 1976, 27 separate ebolavirus outbreaks have occurred and are attributed to five different ebolavirus species: Zaire ebolavirus, Sudan ebolavirus, Reston

18 2 ebolavirus, Taï Forest ebolavirus, and Bundibugyo ebolavirus (filovirus nomenclature is described in Table 1-2). All species of ebolavirus cause disease in humans and nonhuman primates, except for Reston virus, which only causes disease in non-human primates. In 2011 a novel filovirus of a new genus, cuevavirus, was detected in bat carcasses from two geographically distinct caves in northern Spain (2). Massive bat dieoffs led to the discovery of pathological signs of viral pneumonia in lung tissues from dead Schreiber s bats (Miniopterus schreibersii). Lloviu virus (LLOV) RNA was found in tissues from dead M. schreibersii samples but was absent in samples from healthy M. schreibersii samples and dead greater mouse-eared bats (Myotis myotis) from the same locations. Sequence analysis indicated that LLOV is more closely related to ebolaviruses (~48% nucleotide identity) than marburgviruses (~43% nucleotide identity). Although there is no direct evidence to link LLOV infection and mortality in M. schreibersii populations, this would represent the first documented case of filovirus pathogenesis in bats. The viral reservoir and transmission Several studies have suggested that insectivorous bats are the natural reservoir for ebolaviruses and marburgviruses (3-9). While filovirus-specific antibodies and/or RNA have been found in several species of bat, only the Egyptian fruit bat (Rousettus aegyptiacus) and the Hammer-headed bat (Hypsignathus monstrosus) have been shown to harbor both ebolaviruses and marburgviruses (4, 5, 8, 9). Analysis of an Egyptian fruit bat colony in Kitaka Cave, Uganda, the site of a 2007 MARV outbreak, revealed that 5% of bats had evidence of MARV and RAVV infection (8). Another study traced the source

19 3 of the 2007 EBOV outbreak in the DRC to the first victim coming in contact with a freshly killed fruit bat that was to be eaten (3). Furthermore, RESTV-specific antibodies were found in Geoffroy s Rousette (R. amplexicaudatus), in the Philippines at sites near areas where RESTV infection of cynomolgus macaques and pigs had been identified (10). These studies provide indirect evidence that common fruit bats can serve as the source of filovirus outbreaks. There is little information available regarding the course of filovirus infections in bats. However, two studies provided evidence that experimental inoculation of bats with EBOV and MARV resulted in viremia from 1-12 days post infection, but virus was absent in samples taken from later time points (11, 12). Furthermore, virus was detected in feces of bats 21 days after inoculation with EBOV (12). Interestingly, despite high titers of virus being present in multiple organs, including the liver and spleen, inoculated bats did not develop clinical disease. Over 100 viruses have been detected in bats, and only three have been shown to cause disease in bat populations: two lyssaviruses, rabies virus and Australian bat lyssavirus, and the arenavirus Tacaribe virus (13). Currently, there are limited molecular and immunological mechanism studies that have been performed to explain how bats harbor a number of viruses that are pathogenic in other species without developing disease. One proposed hypothesis is innate antiviral mechanisms limit viral replication to allow time for adaptive immune responses to develop. An interesting aspect of the bat immune system is the presence of up to 12 interferon-ω (IFNω) genes, while humans and mice have only one (13). Furthermore, type III IFNs are present in the bat genome and unlike humans and mice the type III IFN receptor has a broad tissue distribution, suggesting a potentially significant role in

20 4 controlling virus replication throughout the body (13). Adaptive immune responses (humoral and cell-mediated) of several species of bat have been described to be slightly delayed compared to similar infections in mice (13). Recent evidence suggests that bats have a subclass of IgM that is not present in humans (14). Since IgM is the first antibody class produced in response to infection, diversification of this class may enhance the initial antibody response thereby providing an advantage to the host. Most fruit bats (Rousettus aegyptiacus) that were experimentally inoculated with MARV developed an IgG response by 21 days post infection, which correlated with a decrease in viremia (11). Interestingly, antisera collected on day 21 post infection had relatively poor neutralization titers (1:4 to 1:8), leaving the possibility of a strong cell-mediated immune response (11). While the role of natural reservoirs in the spread of filoviruses remains controversial, the routes of entry into humans and primates are well established. Most filovirus infections occur through direct contact with virus from infected or dead people or animals (15). The virus enters the body via mucosal surfaces, skin abrasions, or the use of contaminated needles (16). Non-human primate studies have shown that large amounts of virus are shed in saliva, feces, and urine (1). Additionally, examination of lung tissue from infected monkeys and humans indicated that the virus is present in the aveoli and bronchi (1). Spread of the virus through ingestion or direct contact with the conjunctiva was lethal in monkeys (17). Aerosolized transmission was observed in a biocontainment laboratory that housed both infected and uninfected rhesus macaques (18). Furthermore, EBOV infected pigs were able to transmit the virus to uninfected cynomolgus macaques without direct contact (19).

21 5 Characterization of filoviral hemorrhagic fever disease and pathogenesis Individuals infected with filoviruses experience initial flu-like symptoms such as fever, chills, muscle aches, malaise, and headache after a 4 to 10 day incubation period (15). Symptoms become more severe as the disease progresses with patients experiencing abdominal pain, vomiting, and non-bloody diarrhea. Around five days after the onset of symptoms patients develop a characteristic maculopapular/rubeola-like rash that is the most reliable diagnostic sign (1). Severe hemorrhaging occurs during the terminal stage of the disease, usually at 6-12 days, including bleeding from the gums, nose, and gastrointestinal tract due to breakdown of endothelial barrier function and disseminated intravascular coagulation (15). Death usually occurs at 7-16 days after infection as the result of multiple organ failure and shock. Currently, there are no specific antivirals or vaccines to combat filovirus outbreaks. Symptoms are treated with oral and intravenous nutritional support and rehydration. Animal studies have indicated that monocytes, macrophages, and dendritic cells are early sites of virus replication (15). Infection of these early targets leads to release of proinflammatory cytokines, which results in tissue damage and promotes the movement of more monocytes and macrophages to the site of viral replication. Infection of these cells results in the dissemination of virus to endothelial and epithelial cells in multiple target tissues, including the liver, spleen, and lung. Virus replication in hepatocytes leads to impairment of liver function, which may account for coagulation disorders during fatal infection. Increased vascular permeability is likely due to the release of inflammatory cytokines and chemokines by infected monocytes and macrophages, and viral replication

22 6 within endothelial cells. In fatal cases up to viral RNA copies/ml of serum can be detected, whereas nonfatal patients have 10 7 copies/ml (20). Survival appears to be highly dependent on the early, transient, innate immune response to control virus replication and activate the adaptive immune response. The presence of antibodies at the onset of symptoms and prominent activation of CD8 + T-cells are associated with nonlethal infections (15). Characterization of filoviruses Classification and phylogenetics Filoviruses are enveloped and contain a negative sense, non-segmented, singlestranded genome of ~19 kb and thus are members of the order mononegavirales. Individual virions range in length from 800-1,000 nm and have a uniform ~90 nm width, which produces long, flexible filaments with a helical nucleocapsid (21). In addition to the formation of individual virions, multiple genome copies can be packaged into continuous or linked particles that can be greater than 7,000 nm in length (21). Phylogenetic analysis of filoviruses suggests that these viruses originated over 10,000 years ago (22). Marburgviruses diverged from a common ancestor around 10,000 years ago, while ebolaviruses and cuevaviruses diverged around 7,000 years ago. Ebolaviruses, marburgviruses, and cuevaviruses differ by greater than 50% at the nucleotide level (23). New species within the three different genera are characterized by having greater than 50% nucleotide identity to the type species (EBOV, MARV, or LLOV). Genome sequence analysis of virus isolates from different outbreaks indicates that there is little evolution occurring in human populations (24, 25). However, isolation of RNA and virus from bats in the Kitaka Cave, Uganda after the 2007 outbreak of

23 7 Marburg hemorrhagic fever indicated that the bat colony contained a diverse population of MARV and RAVV (8). Therefore, bat colonies appear to be a potential source of molecular evolution of these viruses. Genome organization The filovirus genome contains seven ORFs: (3 to 5 ) NP, VP35, VP40, GP, VP30, VP24, and L (Figure 1-1). Transcription of mrnas are initiated and terminated at conserved start and stop sites, respectively. The sequence of the LLOV genome indicates that there are six mrna transcripts with the last ORF containing a bicistronic mrna encoding the VP24 and L proteins (2). The NP gene encodes for the nucleoprotein that encapsidates the RNA genome. VP35, VP30, and VP24 function as bridging molecules between the nucleocapsid and VP40 matrix protein; and together with the RNAdependent RNA polymerase they form the nucleocapsid (21). VP35 and VP24 also function to inhibit the interferon response in infected cells and aid in replication and transcription (26, 27). In marburgviruses the GP ORF codes for the surface glycoprotein (GP); however, the primary transcript (~80%) from the ebolavirus and cuevavirus GP ORF encodes a soluble glycoprotein (sgp). Due to transcriptional editing by the L protein, there is a frame-shift about 20% of the time that results in the production of a transcript that encodes for surface GP (28). Ebola virus glycoproteins sgp The most abundant protein produced from the ebolavirus GP ORF is the 364 amino acid secreted glycoprotein (sgp) (29). The first 295 amino acids are completely

24 8 conserved between GP and sgp, which contains the highly conserved receptor-binding domain (RBD) that is on average 87% identical between species (Figure 1-2). A cysteine residue in the non-homologous C-terminus results in the formation of parallel homodimeric structure (30). Since sgp is overwhelmingly more abundant, it skews the immune response to produce antibodies against epitopes that are absent from or masked in GP, or shared between both which then allows sgp to absorb anti-gp antibodies (31). This phenomenon is predicted to result in poor neutralization of virus during infection and, indeed, poor neutralization is observed in antisera from convalescent guinea pigs and monkeys (32). Modest levels of antisera neutralization is an important consideration for the development of vaccines. Structure of GP Early biochemical studies identified the filovirus GP to be a trimer of GP1/GP2 heterodimers, indicative of a class I glycoprotein (33). The protein is expressed as a single precursor polypeptide (GP 0 ) that is cleaved into the two subunits by furin in the Golgi prior to trafficking to the cell surface (34). Deletion mutagenesis of EBOV GP and MARV GP determined the receptor-binding domain to reside within GP1 from amino acids and , respectively (35). Extensive mutagenesis of this region defined residues that were required for efficient viral entry, notably EBOV GP residues K114, K115, and F88 (36-40). GP1 contains two heavily glycosylated domains, the glycan cap and mucin-like domain (MLD). The glycan cap of EBOV GP is predicted to contain six N-linked glycans, defined by an N-X-S/T sequon (X P), and mutagenesis of individual sites had little impact on expression or GP-mediated entry (41). The EBOV GP MLD contains

25 9 eight N-linked glycan sites and potentially 80 O-linked glycans. Occupancy of individual serine and threonine residues by O-linked glycans is difficult to predict since O-linked glycosylation does not follow a defined motif. The GP1 core, lacking the MLD (GP1Δmuc), is sufficient to mediate entry (41). Removal of the MLD increases protein expression and dramatically increases GP-mediated entry (41), indicating that this domain is not required for function and may act to suppress the amount of protein present on virions. Structural data indicate that EBOV and SUDV GPΔmuc forms a chalice-like trimer of heterodimers (Figure 1-3) (42-44). In each heterodimer, GP1 is covalently linked to GP2 by a disulfide bond near the base of the chalice (43). The GP1 subunits are cradled by GP2 subunits that wrap around the base of GP1. Interestingly, the highly conserved GP1 RBD is located within the chalice, which is masked by N-glycans within the glycan cap, the glycan cap polypeptide and potentially the MLD. The GP2 subunit that anchors the trimer to the viral membrane is responsible for mediating fusion of viral and host cell membranes. GP2 contains an internal fusion loop that resides on the exterior of the structure, thus it more closely resembles that of class II glycoproteins of flavivirus than those of other class I glycoproteins (42). The cradle formed by GP2 consists of the segmented heptad repeat region 1 (HR1), which has an extended coil connecting two alpha helical structures. The post-fusion structure of filoviral GP2 indicates that HR1 and HR2 form an antiparallel six-helix bundle to drive fusion of viral and host cell membranes (45-47).

26 10 Glycosylation of GP Glycosylation is a common post-translational modification of proteins. Over 50% of all eukaryotic proteins are glycosylated (48). Glycans have a variety of functions, including facilitating protein folding and stability, protecting from proteases and antibody binding, and serving as ligands for lectins (49). During translation glycan precursors are attached to certain serine or threonine residues (O-linked glycosylation) within mucinlike proteins or asparagine residues of the N-X-S/T sequon (X P; N-linked glycosylation) within the endoplasmic reticulum (ER). A large number of glycosyltransferases, glycosidases, and glycan-processing enzymes modify glycans within the ER and Golgi to create a vast array of heterogeneous glycans (50). EBOV GP is one of the most heavily glycosylated viral glycoproteins. Similar to HIV gp120, approximately half of the mass of GP is attributed to glycosylation (41, 51). The extent of glycosylation on GP varies between ebolavirus species. GP of ebolaviruses contains between N-linked glycosylation sites (NGS), 9-16 in GP1 and two conserved sites in GP2 (52). There are close to 60 different species of high mannose, complex, and hybrid N-glycans on GP1 (53), indicating that there is heterogeneity between molecules produced from the same cell type. The O-linked glycans attached to the MLD of GP1 are predominantly small core 2 structures composed of 3-6 monosaccharides with varying amounts of sialic acid (53). Both types of glycans have been shown to bind to a variety of C-type lectins (CLECs) at the cell surface and have been proposed to play a role in evasion of neutralizing antibodies (52-57).

27 11 Host factors involved in filovirus entry The entry pathway for filoviruses has proven to be a complex process (Figure 1-4). Virus initially binds to the cell surface through non-specific interactions with host surface proteins followed by internalization into endosomes through a macropinocytosislike process (58). To date there have not been any GP-specific host factor interactions described at the cell surface. Inside the acidic environment of the endosome, host proteases are activated that result in the enzymatic removal of both the MLD and glycan cap from GP1 (Figure 1-4) (59-62). After this initial priming event, the receptor-binding domain is exposed and interacts with an intracellular receptor, Niemann-Pick type C1 (NPC1) (36, 63). The trigger for GP2-mediated fusion has not been characterized; however, it is hypothesized that GP must undergo further modifications prior to fusion (64). Much is known about these events for the ebolavirus type species, EBOV, however less is known regarding additional species and other genera. Furthermore, information on GP sequence determinants for interactions with specific host factors is limited. Phosphatidylserine receptors Several viruses have been shown to utilize phosphatidylserine (PS) receptors during entry in a process called apoptotic mimicry. PS is an anionic lipid that is normally sequestered to the inner leaflet of the plasma membrane. During apoptosis, PS flips to the outer leaflet and serves as signal for phagocytosis by macrophages and dendritic cells (65). Initial studies using vaccinia virus indicated virions contain PS on the outer leaflet of the viral envelope, which is required for internalization of virus; thus vaccinia virus mimics apoptotic bodies to gain entry into host cells (66). Additional work has indicated that entry of viruses from diverse families is dependent on exposed viral PS

28 12 interacting directly (TIM family) or indirectly with PS receptors (TAM family kinase receptors/tam ligands), including dengue virus (Flaviviridae; (67)), HIV (Retroviridae; (68), Pichinde virus (Arenaviridae; (69), Ross River virus (Alphaviridae; (70), and EBOV (Filoviridae; (70). Thus PS receptors represent a class of non-specific entry factors for a variety of viruses, however the impact of these molecules has yet to be determined in vivo. TAM family receptor kinases The TAM (Tyro3, Axl, and Mer) receptors are single pass transmembrane proteins with two terminal extracellular Ig-like domains followed by two fibronectin type III domains, a transmembrane domain, and a cytoplasmic tail that contains a functional protein tyrosine kinase domain (71). These receptors function in the clearance of apoptotic bodies and inhibition of type I interferon and cytokine-dependent signaling. Both of these roles require the binding of TAM receptors to one of the soluble PSbinding ligands, growth-arrest-specific 6 (Gas6) or protein S. Activation of TAM receptors occurs when the TAM ligands bridge TAM receptors to apoptotic cells through binding of PS in membranes (72). Ectopic expression of all three TAM receptors in poorly permissive Jurkat cells increased the entry of retroviruses pseudotyped with GP from a variety of ebolavirus species or MARV, and infectious EBOV (73). The enhancement of virus entry mediated by Axl required the terminal Ig-like domain and was independent of direct interaction with GP. These results suggested that interaction with TAM ligands is important for TAM-dependent virus entry (71, 74). Furthermore, activation of Axl in the highly permissive SNB-19 cell line resulted in actin polymerization and macropinocytosis (75).

29 13 Together, these results suggest that filoviruses utilize the apoptotic clearance function of the TAM ligands/receptors during entry. Additionally, engagement of TAM receptors during virus entry negatively impacts the type I interferon response, which allows productive viral replication (76). Interestingly, not all cell populations that endogenously express Axl require this protein for entry, such as the highly permissive VeroE6 cell line. These cells also do not express other TAM family members, Tyro 3 or Mer, nor do they express CLECs, (73). Therefore, there must be additional cell surface host proteins that mediate entry of filoviruses. TIM family receptors The T-cell immunoglobulin and mucin-domain (TIM) proteins are PS receptors involved in the regulation of immune responses, including clearance of apoptotic bodies (77). There are three TIM proteins in humans (TIM-1, -3, and -4) that are expressed on T cells (TIM-1 and TIM-3), epithelial cells (TIM-1), and antigen presenting cells (TIM-3 and TIM-4). These proteins consist of an extracellular variable Ig-like (IgV) domain followed by a mucin-domain of varying length and a single pass transmembrane domain and cytoplasmic tail. The cytoplasmic tails of both TIM-1 and TIM-3 contain tyrosine phosphorylation motifs, however the cytoplasmic tail of TIM-4 does not contain any tyrosine residues (78). The IgV domain of the TIM proteins binds directly to PS in a metal ion-dependent manner and engulfment of apoptotic bodies is dependent on this interaction. TIM-4 mediated clearance of apoptotic bodies is independent of the cytoplasmic tail, indicating that additional proteins are likely required for signaling (77). Additionally, TIM-1 and TIM-3 have not been reported to facilitate apoptotic body

30 14 clearance by T cells, but have been shown to mediate this process in kidney epithelial cells and antigen presenting cells (APCs), respectively (77). A comparative genetics analysis revealed that TIM-1 gene expression correlated with the transduction efficiency of VSV pseudotyped with EBOV GPΔmuc in a panel of human cell lines (79). TIM-1 is expressed at the surface of highly permissive Vero cells and ectopic expression in poorly permissive cells enhanced entry into these cells mediated by GP of EBOV and MARV. Targeting TIM-1 with an anti-igv domain antibody inhibited pseudovirus and infectious EBOV entry. TIM-1 is expressed on epithelial cells of airway, kidney, cornea and conjunctiva, which represent tissues that support EBOV replication. Additional work revealed that both TIM-1 and TIM-4 enhance entry of a wide range of viruses through binding PS in the viral membrane (67, 70, 80). Expression of TIM-1 enhances the binding and internalization of virions, including those lacking a viral glycoprotein. It is currently not known how TIM-1 and TIM-4 mediate internalization after binding to PS in viral membranes. Since TIM-4 has not been shown to mediate direct signaling, it is possible that there are additional surface proteins that form a complex with virions to facilitate the internalization of virus into endosomes. C-type lectins CLECs are a large family of cellular receptors that recognize various carbohydrate structures in a calcium-dependent manner. CLECs are a type of patternrecognition receptor (PRR) expressed on a variety of cell types that can recognize a wide array of pathogens, including fungi, parasites, bacteria, and viruses. Recognition of pathogens by CLECs has been shown to result in the development of protective immune

31 15 responses through upregulation of cytokines, induction of T cells, and internalization of pathogens resulting in antigen processing and presentation (81). In contrast, viruses utilize CLECs to gain entry into the host cell. CLECs have been shown to be receptors/attachment factors for a number of viruses, including HIV, dengue virus, West Nile virus, SARS-coronavirus, measles virus, and filoviruses (82). Due to differences in expression among cell types and ligand specificity, utilization of CLECs as entry factors may be a significant determinant of tissue tropism for viruses. Several C-type lectins have been shown to enhance entry mediated by GP of ebolaviruses and marburgviruses, including DC-SIGN, L-SIGN, ASGPR, LSECtin, and hmgl (53-57). All of these proteins are expressed on filovirus target cells, however the importance of these molecules in vivo has not been determined. DC-SIGN and L-SIGN Dendritic cell-specific intercellular molecule-3 grabbing non-integrin (DC-SIGN) is a type II transmembrane protein that is expressed on mature and immature dendritic cells in the dermis, mucosa, spleen, and lymph nodes, and specialized macrophages in the lung and placenta (83). DC-SIGN was originally cloned as a protein capable of binding to HIV gp120 in the absence of CD4 (84) and was found to allow DCs to facilitate infection to susceptible cells in trans (81). Liver/lymph node-specific intercellular molecule-3 grabbing non-integrin (L-SIGN) shares 77% amino acid identity with DC-SIGN and is present at the cell surface of endothelial cells of the lymph nodes, liver, placenta, lung, and intestine. The cytoplasmic tail of DC-SIGN contains a dileucine motif, which is conserved in L-SIGN, that is required for the internalization of bound ligands (85, 86). The ectodomain consists of a neck domain made up of repeat regions and the

32 16 carbohydrate recognition domain (CRD). Tetramerization of the neck domain has been shown to be important for high avidity binding of clustered ligands, which are present on the glycans attached to viral glycoproteins. The CRDs of DC-SIGN and L-SIGN (84% amino acid identity) bind with high affinity to high-mannose glycans. However, a valine at position 351 in the CRD allows DC-SIGN to bind fucose-containing glycans as well (87). Expression of DC-SIGN and L-SIGN on poorly permissive cells was found to enhance entry mediated by the Ebola virus GP using lentiviral pseudovirions (56). This enhancement could be neutralized with the addition of mannan, indicating that these CLECs were binding to high-mannose glycans present on GP. Additional studies have shown that GP-dependent entry of all ebolavirus species tested and MARV is enhanced upon expression of DC-SIGN and L-SIGN (56, 57, 88). However, SUDV GP-mediated entry was poorly enhanced compared to other species, which correlated with a lower degree of high-mannose glycans compared to EBOV GP (88). LSECtin Liver/lymph node sinusoidal endothelial cell C-type lectin (LSECtin) is closest in homology to DC-SIGN and L-SIGN and is co-expressed with L-SIGN on sinusoidal endothelial cells in the liver and lymph node. Accordingly, LSECtin has been shown to bind mannose and internalize upon ligand interaction (89, 90). However, the ability of LSECtin to enhance entry mediated by ebolavirus GP is independent of this interaction. The highest affinity ligands for LSECtin were determined to be truncated complex and hybrid glycans with terminal N-acetylglucsoamine-β-1-2-mannose residues, which are

33 17 present on EBOV GP (91). MARV GP-dependent entry was enhanced more robustly by LSECtin than other filoviruses that were tested (57). ASGPRI Asialoglycoprotein receptor (ASGPR) was the first mammalian lectin to be described (92) and the first cell surface receptor shown to mediate filovirus infection (55). ASGPR is expressed on hepatocytes as a hetero-oligomeric complex composed of two homologous subunits, H1 and H2, that interact with terminal galactose and N- acetylgalactosamine residues on glycans (93, 94). These sugars are present on both N- and O-linked glycans on EBOV GP (53). The H1 subunit of ASGPR (ASGPRI) contains a tyrosine-based internalization motif in the cytoplasmic tail; interaction with glycans results in internalization and subsequent trafficking to lysosomes (92, 95). Overexpression of ASGPRI in poorly permissive cells is sufficient to enhance the entry of filoviruses (88). SUDV GP-mediated ASGPRI-dependent entry was greater than that of EBOV GP, consistent with SUDV GP having a lower degree of high-mannose N- glycans and a higher degree of complex/hybrid N-glycans (88). hmgl Human macrophage lectin specific for galactose/n-acetylgalactosamine (hmgl) is a homo-oligomer that is expressed on dendritic cells and macrophages, but is absent on monocytes, plasmacytoid dendritic cells, and lymphocytes (95). Studies have shown that hmgl interacts with terminal galactose, N-acetylgalactosamine, and fucose; however, a subsequent study has shown that this protein has a high specificity for N- acetylgalactosamine residues (95). Mass-spectrometric analysis of glycans on the GP1 subunit of EBOV indicated that terminal N-acetylgalactosamine residues are present in

34 18 both N- and O-linked glycans (53). Similar to DC-SIGN and ASGPRI, hmgl contains a tyrosine-based internalization motif in the cytoplasmic tail and is able to mediate internalization upon ligand binding (96). Overexpression of hmgl in poorly-permissive cells enhanced entry mediated by filovirus GPs (54). Enhancement of entry by hmgl was highest for TAFV GP and both MARV GP and RESTV GP were poorly enhanced (54). Cysteine proteases Cysteine proteases are a large class of enzymes that are present in all living organisms and viruses (97). In mammals, most cell types constitutively express cysteine proteases, which localize to the late endosome and lysosome (98). There are several families of cysteine proteases, which include caspases, calpains, and papains. The cathepsins are well-studied members of the papain family that play an essential role in the digestion of proteins for presentation by major histocompatibility complex (MHC) class II molecules and protein turnover (98). Upon internalization filoviruses are trafficked from early endosomes to late endosomes/lysosomes. Within the acidic environment of the endosome/lysosome GP1 is processed by cellular cysteine proteases (61). All filoviruses require the activity of cysteine proteases for entry, though the specific protease requirements vary between viruses (99). Protease digestion of GP1 results in the removal of both the glycan cap and MLD, which exposes residues within the RBD (60). Treatment of EBOV GP with cathepsin L (CatL) results in the removal of both of these domains, however this enzyme is dispensible for entry (64). Instead, EBOV GP-mediated entry is dependent on cathepsin B (CatB), which removes an additional kilodalton peptide from GP1 (61, 64).

35 19 CatB-mediated removal of the glycan cap and MLD from EBOV GP results in a metastable structure that requires one or more additional events to trigger fusion (59). Inhibition of all cysteine proteases blocks entry mediated by CatB-treated EBOV GP, therefore an additional proteolysis event may be required to trigger fusion (59, 64). In vitro assays have indicated that, in an acidic environment, mild reduction or mild denaturing with urea is sufficient to trigger fusion after the glycan cap and MLD have been removed (100). However, the cellular trigger for fusion has not been identified. Niemann-Pick C1 Mutation of the NPC1 gene results in a fatal lysosomal lipid storage disease, called Niemann-Pick disease type C. Niemann-Pick C1 protein (NPC1) is a thirteen-pass transmembrane protein of the late endosome/lysosome involved in the trafficking of cholesterol in all cells (101). There are three large luminal loops that are required for cholesterol transport. The second loop binds to cholesterol-bound soluble NPC2, which positions NPC2 in close proximity to transfer cholesterol to the first NPC1 luminal loop (102). Studies have shown that the expression of NPC1 is required for filovirus infection (36, 63, 103). Endosomal proteolysis of GP1 potentiates interaction with the second luminal loop of NPC1 (36), thus this protein represents a novel intracellular viral receptor. However, binding of NPC1 is not sufficient to trigger fusion in an acidic environment (36) and the role of NPC1 during filovirus infection remains unknown. Heterozygous knockout of NPC1 in mice significantly increases survival compared to WT mice after challenge with mouse-adapted EBOV and MARV (103). Therefore, the amount of NPC1 expression appears to be an important determinant for filovirus

36 20 infection. NPC1 is an attractive target for the development of antivirals since it is the only entry factor, with non-redundant function, utilized by all filoviruses. Rationale and objectives for current studies Filoviruses cause sporadic, highly lethal hemorrhagic fever in human populations. These viruses express a single viral envelope glycoprotein, GP, which is responsible for mediating entry and is a primary target for antibodies. In order to develop antivirals and vaccines it is critical to determine how this protein efficiently provides the virus with (1) access to the cytoplasm for replication and (2) protection from immune responses, which may also be useful in understanding viral evolution in the context of the natural reservoir. The receptor-binding domain of EBOV GP has been characterized, however structural data indicate that this region is masked by a glycan shield. Furthermore, the RBD is highly conserved among ebolaviruses and to a lesser extent marburgviruses; however, there remain differences in entry factors required for infection between species of ebolavirus and between ebolaviruses and marburgviruses. These studies were designed to identify the impact of glycosylation and sequence differences of GP on expression, entry events, and antibody evasion. List of specific aims 1. Determine the role of GP N-linked glycans during Ebola virus entry and antibody evasion. 2. Characterize GP1-dependent differences in ebolavirus species-specific entry.

37 21 Knowledge gained by current studies 1. Removal of more than three N-linked glycans from EBOV GP1Δmuc dramatically decreases protein expression. 2. All 15 N-linked glycans can be removed from EBOV GP1 without significantly affecting protein expression, indicating that the presence of the MLD provides stability in the absence of N-linked glycans. 3. Removal of N-linked glycans from GP1 significantly increases entry in EBOV GP-mediated entry into Vero cells. 4. Removal of the N-linked glycan at residue N40 decreases the sensitivity to CatB inhibition, however further deglycosylation does not increase this effect. 5. N-linked glycans on GP1 impede proteolysis of GP and decreased protease sensitivity correlates with reduced GP-mediated entry. 6. Removal of the glycan shield surround the RBD is not sufficient for NPC1 binding, suggesting that the glycan cap polypeptide and/or MLD polypeptide/olinked glycans mask residues required for interaction. 7. Removal of N-linked glycans resulted in unique patterns of CLEC-mediated entry. a. DC-SIGN: ligands are present in both the GP1 core and MLD, however removal of all N-linked glycans from GP1 did not completely abolish entry. b. L-SIGN: ligands are present in both the GP1 core and MLD. c. LSECtin: ligands are present in both the GP1 core and MLD, however the MLD appears to contain more ligands than the GP1 core.

38 22 d. ASGPRI: ligands are present in the MLD, but not the GP1 core. e. hmgl: removal of all N-linked glycans from GP1 has a modest impact on entry, indicating that this CLEC interacts with O-linked glycans in the MLD or N-linked glycans on GP2. 8. Removal of the MLD from GP1 alters the species of N-linked glycans attached to the GP1 core, evidenced by the differences in CLEC utilization by deglycosylated mutants of GP1Δmuc and GP1. 9. Entry into primary murine peritoneal macrophages is independent of N-linked glycans and resembles entry into PS receptor-dependent Vero cells. 10. Removal of N-linked glycans from the GP1 core increases sensitivity to antibody neutralization. 11. Removal of the N563 glycan from GP2 abolishes KZ52 binding to GP. However, GP1 is detected by anti-gp1 mab 5E6, but at 50% of WT. 12. Removal of the N618 glycan from GP2 has a modest impact on KZ52 binding to GP. 13. Removal of the N563 glycan enhances transduction by 2-fold and removal of the N618 glycan decreases transduction by 2-fold. 14. GP2 N-linked glycans are involved in stability of protease treated GP. 15. The N-linked glycan at N563 is sufficient for protein expression and GP-mediated entry. 16. The complex GP2 N-linked glycan at N618 and high mannose N-linked glycan at N563 are sufficient for CLEC-dependent entry in the absence of the MLD. In the presence of the MLD, ASGPRI- and hmgl-dependent entry is independent of the

39 23 N618 complex glycan and the N563 high-mannose glycan is sufficient for DC- SIGN-dependent entry. 17. SUDV GP-mediated entry is over a log decreased compared to EBOV GP. Sequence analysis indicates that position 95 is occupied by a glutamine in SUDV- Boniface GP and a lysine in EBOV GP. Mutation of SUDV-Boniface GP Q95 to a lysine increases virus entry to levels mediated by EBOV GP. 18. A glutamine at position 95 in EBOV GP, SUDV GP, and BDBV GP increases protease sensitivity and GP1 degradation, whereas a lysine at this position stabilizes a protease cleavage intermediate. 19. The residue at position 95 of EBOV GP, SUDV GP, and BDBV GP is a determinant of sensitivity to CatB inhibition. GPs with a glutamine at position 95 are completely resistant to CatB inhibition. A lysine at position 95 results in complete dependence on CatB activity for EBOV GP- and BDBV GP-dependent entry. However, a lysine at position 95 in SUDV-Boniface GP is 20. The residue at position 95 of EBOV GP, SUDV GP, and BDBV GP is a determinant of sensitivity to NPC1 inhibition by U18666A. 21. Ebolavirus species have different sensitivities to inhibition of NPC1-GP interaction by compound 3.47 and mutation of the residue at position 95 decreases sensitivity to this compound. 22. Ebolavirus species are sensitive to verapamil-mediated inhibition of voltageoperated calcium channels. 23. Mutation of K95 in EBOV GP and BDBV GP decreases sensitivity to verapamil, but mutation of Q95 in SUDV GP does not impact sensitivity to this drug.

40 Vaccination of Balb/c mice with VSV pseudovirions bearing EBOV GP or 7G mutant protect against lethal challenge with mouse-adapted EBOV GP.

41 25 Table 1-1 Chronological order of filovirus outbreaks in human populations. Year Country Genus Virus Reported no. cases Reported no. (%) deaths among cases 1967 Germany/Yugoslavia Marburgvirus Marburg 31 7 (23%) 1975 South Africa Marburgvirus Marburg 3 1 (33%) 1976 Zaire (DRC) Ebolavirus Ebola (88%) 1976 Sudan Ebolavirus Sudan (53%) 1976 England Ebolavirus Sudan 1 0 (0%) 1977 Zaire Ebolavirus Ebola 1 1 (100%) 1979 Sudan Ebolavirus Sudan (65%) 1980 Kenya Marburgvirus Marburg 2 1 (50%) 1987 Kenya Marburgvirus Marburg 1 1 (100%) DRC (Zaire) Marburgvirus Marburg (83%) 1989 USA Ebolavirus Reston 0 0 (0%) 1990 USA Ebolavirus Reston 4* 0 (0%) Philippines Ebolavirus Reston 3* 0 (0%) 1992 Italy Ebolavirus Reston 0 0 (0%) 1994 Gabon Ebolavirus Ebola (60%) 1994 Ivory Coast Ebolavirus Tai Forest 1 0 (0%) 1995 DRC (Zaire) Ebolavirus Ebola (81%) 1996 Gabon Ebolavirus Ebola (57%) Gabon Ebolavirus Ebola (74%) 1996 South Africa Ebolavirus Ebola 2 1 (50%) 1996 USA Ebolavirus Reston 0 0 (0%) 1996 Philippines Ebolavirus Reston 0 0 (0%) Uganda Ebolavirus Sudan (53%) Gabon Ebolavirus Ebola (82%) Rep. of Congo Ebolavirus Ebola (75%) Rep. of Congo Ebolavirus Ebola (89%) 2003 Rep. of Congo Ebolavirus Ebola (83%) 2004 Sudan Ebolavirus Sudan 17 7 (41%) Angola Marburgvirus Marburg (90%) 2007 DRC Ebolavirus Ebola (71%) 2007 Uganda Marburgvirus Marburg 4 1 (25%) Uganda Ebolavirus Bundibugyo (37%) 2008 Philippines Ebolavirus Reston 6* 0 (0%) 2008 USA Marburgvirus Marburg 1 0 (0%) 2008 Netherlands Marburgvirus Marburg 1 1 (100%) DRC Ebolavirus Ebola (47%) 2011 Uganda Ebolavirus Sudan 1 1 (100%) 2012 Uganda Ebolavirus Sudan (67%) 2012 DRC Ebolavirus Bundibugyo 10 6 (60%) * indicates asymptomatic infections, indicates patients had recently returned from Uganda Data obtained from CDC: and

42 26 Table 1-2 Filovirus nomenclature. Order Family Genus Species Virus Abbreviation Mononegavirales Filoviridae Marburgvirus Marburg marburgvirus Marburg virus MARV Ravn virus RAVV Mononegavirales Filoviridae Ebolavirus Zaire ebolavirus Ebola virus EBOV Sudan ebolavirus Sudan virus SUDV Reston ebolavirus Reston virus RESTV Taï Forest ebolavirus Taï Forest virus TAFV Bundibugyo ebolavirus Bundibugyo virus BDBV Mononegavirales Filoviridae Cuevavirus Lloviu cuevavirus Lloviu virus LLOV

43 Figure 1-1 Filovirus genome organization. Open reading frames for each of the three filovirus genera are shown, drawn to scale. The leader and trailer regions are marked at either end of the genome. Sites of gene overlap are marked with * above the transcription start site. The sgp ORF is shown above the GP ORF. A scale is shown at the bottom in kilobases. 27

44 28 Conserved region Figure 1-2 Linear model comparison of GP and sgp. SP, signal, peptide; RBD, receptorbinding domain; Cap, glycan cap; Mucin, mucin-like domain; TM, transmembrane domain. Disulfide bond between GP1 and GP2 is shown. Conserved region between the two proteins is noted.

45 29 A B C D E Figure 1-3 Models of EBOV GP. (A) Linear model of EBOV GP. Average amino acid sequence identity between EBOV GP and SUDV GP is shown for different regions. The regions included in the GP1 crystal structure are noted, GP1 core. Each N-linked glycosylation site is noted with a Y. SP, signal peptide; RBD, receptor-binding domain; MLD, mucin-like domain; IFL, internal fusion loop; HR1/2, heptad repeat 1/2; and TM, transmembrane domain. Furin cleavage site between GP1 and GP2 is noted with an arrow. (B-E) 3D structural representation of EBOV GP (GPΔmuc PBD ID: 3CSY and HR2 stalk PDB ID: 2EBO). Structures were glycosylated in silico with GlyPro server. (B) Top-down view and (C) side view of GP. (D) Top-down view and (E) side view of predicted CatB cleaved form of GP. MLD is shown for each monomer as a gray sphere in a predictive fashion, GP1 is shown in teal, RBD is shown in red, GP2 is shown in tan, and complex N-linked glycans are shown in orange.

46 30 1. Virion binding to cell surface TIM family TAM receptors / TAM ligands C-type lectins 2. GP cleavage Cysteine proteases!" #" $"!" #" $" NPC1 3. NPC1 binding 4. Fusion trigger: - cysteine protease? - reductase?!" #" $" Figure 1-4 Model of filovirus entry. (1) Virus binding to the cell surface occurs through non-specific interactions. TIM family cell surface receptors bind directly to PS in the viral membrane. TAM receptors interact with TAM ligands bound to PS in the viral membrane. C-type lectins bind to N- and O-linked glycans present on GP. Virus is internalized through a macropinocytosis-like process and trafficked to the late endosome/lysosome where (2) GP is cleaved by cysteine proteases, such as CatB. After removal of the glycan cap and MLD by cysteine protease cleavage, (3) primed GP interacts with NPC1. After NPC1 interaction, (4) fusion is triggered and the nucleocapsid is released into the cytoplasm. Current hypotheses for triggering fusion include further cleavage of GP by cysteine proteases and reduction of disulfide bonds by a reductase.

47 31 CHAPTER II COMPREHENSIVE FUNCTIONAL ANALYSIS OF N-LINKED GLYCANS ON EBOLA VIRUS GP1 Introduction Two genera compose the family Filoviridae: Ebolavirus and Marburgvirus. These viruses cause outbreaks of severe hemorrhagic fever, with associated fatality as high as 90% (104). There are five antigenically distinct species of ebolaviruses (105, 106), and the prototypic member of each species is referred to by the location of the initially described outbreak: Ebola virus (EBOV), Sudan virus (SUDV), Bundibugyo virus (BDBV), Taï Forest virus (TAFV), and Reston virus (RESTV) (23). All of these viruses have been reported to cause disease in humans, except for RESTV, which is pathogenic in nonhuman primates (104). Currently there are no licensed antivirals or vaccines and a better understanding of the structure and function of the viral glycoprotein may lead to novel approaches for the development of antivirals. The negative-sense, single stranded RNA genome of filoviruses encodes a single viral envelope glycoprotein (GP), which is highly glycosylated and forms a trimer of GP1/GP2 heterodimers on the surface of virions (Figure 2-1A). The GP1 subunit is required for receptor interactions and the transmembrane-associated GP2 subunit is required for membrane fusion. GP1 contains four different domains: base, receptorbinding domain (RBD), glycan cap and mucin-like domain (MLD) (Figure 2-1B). The first three domains compose the core of GP1 and are required for expression and function of the pre-fusion glycoprotein. EBOV GP with the MLD deleted (GP1Δmuc) has

48 32 increased protein expression and provides higher viral titers when pseudotyped onto retroviral (41) or vesicular stomatitis virus particles (personal observations). Although ebolaviruses are antigenically distinct, the RBD is highly conserved, with an average of 87% amino acid identity. This conservation suggests either an absence of selective pressures driving diversification of this region of the protein or high selective pressures maintaining sequence conservation. In contrast, the glycan cap and MLD have extensive sequence diversity between the different ebolavirus species. As the names suggest, both the glycan cap and the MLD are highly glycosylated, resulting in the majority of the protein being masked by glycans (Figure 2-1A) (42). The glycan cap contains only N-linked glycan sites (NGS) and, despite generally poor sequence conservation within the glycan cap, these sites are well conserved among ebolaviruses, suggesting functional significance (Figure 2-1C). In contrast to the glycan cap, the glycosylation events found in the MLD are highly variable between the different viruses and include both N- and O-linked glycans. Despite extensive glycosylation, the MLD remains highly targeted by neutralizing antibodies ( ). Several roles have been attributed to glycans attached to viral glycoproteins. Glycans can serve as ligands for C-type lectins (CLECs), facilitating viral attachment and internalization in a variety of cell types (82). Additionally, glycans promote protein folding/stability and virion incorporation of GP, as demonstrated in studies with Newcastle Disease virus and Lassa virus (110, 111). In the case of Nipah virus G/F proteins, not only does glycosylation help protein expression, it also decreases membrane fusion efficiency, thereby controlling premature fusion events (112, 113). Furthermore, the glycans on HIV gp120/gp41, Nipah virus G/F, hepatitis C virus E1/E2, and influenza

49 33 A virus HA protect virions from antibody-mediated neutralization (52). Despite the high degree of glycosylation found on filovirus GPs, the importance of the N-linked glycans on EBOV GP1 to the structure and function of the protein has not been well studied. Materials and Methods Cells lines and plasmids HEK 293T cells and Vero cells were maintained in high glucose DMEM (Gibco) supplemented with 10% FBS and 1% penicillin/streptomycin. EBOV studies were performed using the Mayinga strain of EBOV GP and GP1Δmuc (deletion of aa ) (41) expressed by the CMV promoter in pcdna3.1. The Marburg virus GP (Musoke) was expressed from the pcaggs vector (BEI Resources, NIAID, NIH, NR-19815). Plasmids containing the following C-type lectins were obtained through the Arizona State University Plasmid Repository: asialoglycoprotein receptor I (ASGPRI), liver and lymph node sinusoidal endothelial cell C-type lectin (LSECtin), human macrophage galactose/n-acetylgalactosamine-specific C-type lectin (hmgl) and soluble dendritic cell-specific ICAM3 grabbing non-integrin 1B type I isoform (sdc-sign1b). A plasmid containing the L-SIGN ORF was obtained through ATCC. All C-type lectins were PCR amplified and subcloned upstream of a Myc-epitope tag (EQKLISEEDL) in the pcdna3.1 vector. As the DC-SIGN cdna from the repository expressed a soluble version of the protein, a chimeric L-SIGN/DC-SIGN-myc construct was created by replacing the ectodomain of L-SIGN-myc with the homologous sequence of the sdc- SIGN1B to produce L DC-SIGN. In addition, we observed poor hmgl expression with the cdna construct, therefore a similar chimeric L-SIGN/hMGL-myc construct was

50 34 created to produce ( L hmgl). The plasmid expressing human T-cell immunoglobulin and mucin domain protein (TIM-1) has previously been described (79). N-linked glycan modeling A model of EBOV GP1,2 ΔTM corrected for the two N-glycan site mutations (T42V/T230V) required to produce the crystal structure, PDB ID: 3CSY (114), was generated using the protein fold prediction server PHYRE2 (115). This model was glycosylated in silico with the GlyProt server (glycosciences.de). Complex N-glycans were modeled at potential N-linked glycosylation sites with default torsion angles. Site-directed mutagenesis Indicated mutations were made in either an EBOV GP or GP1Δmuc expression vector using the QuikChange Site-Directed Mutagenesis kit (Stratagene, La Jolla CA) according to manufacturer s protocols. All mutations were confirmed by sequencing the length of the gene. Primer sequences will be made available upon request. Transfection HEK 293T cells were transfected by 25 kda-linear polyethylenimine (PEI) method, as previously described (116). Briefly, HEK 293T cells were plated in 6-well dish 24 h prior to transfection. On the day of transfection, 2 µg of DNA was diluted in 50 µl of 150 mm NaCl and combined with 6 µl of PEI (1 mg/ml) diluted in 50 µl of 150 mm NaCl. The DNA:PEI mixture was vortexed and incubated at room temperature for 10 min prior to addition to cells.

51 35 Pseudovirion production EBOV GP pseudotyped vesicular stomatitis virus (VSV) production has been previously described (54, 79). Briefly, HEK 293T cells were transfected with indicated glycoprotein expression vectors. At 24 h post transfection, cells were transduced with VSVΔG-eGFP pseudovirions for 2 h, then washed twice with PBS. Supernates were collected after 24 h, passed through a 0.45 µm filter and stored at -80 C. When indicated, virions were purified through a 20% sucrose cushion for 2 h at 83,000 x g. Immunoblot of pseudovirions Equal volumes of supernates containing pseudovirions were denatured and reduced in SDS-PAGE loading buffer + β-mercaptoethanol. Samples were then separated on 4-15% Tris-HCl polyacrylamide gels (BioRad). Protein was transferred to nitrocellulose and EBOV GP1 was detected with either rabbit anti-gp1 polyclonal Ab (gift from Anthony Sanchez) or a MLD-specific mab, 5E6 (117). VSV-matrix was detected with the 23H12 mab described above. Western blots were visualized as described above. Deglycosylation with peptide-n-glycosidase F (PNGaseF, New England Biolabs) was performed according to manufacturer s protocol. Pseudovirion matrix and glycoprotein quantitation Equal volumes of cell supernates containing pseudovirions from three independent stocks were lysed in native lysis buffer (PBS % NP-40) and then passed through a dot blot apparatus onto nitrocellulose (Whatman). Wells were washed 5x with PBS before blocking in PBS + 10% nonfat milk for 1 h. Dot blots were incubated with mouse α-vsv matrix mab, 23H12 (a gift from Douglas Lyles) and human α-ebov GP mab, KZ52 (a gift from Erica Saphire and Dennis Burton) diluted in PBS + 10%

52 36 nonfat milk % Tween-20 overnight at 4 C. Through utilization of separate IRDyeconjugated secondary antibodies (LI-COR) directed towards the primary antibodies, we were able to detect both VSV matrix and EBOV within a single well for each sample. Signals were visualized and quantified using an Odyssey Imaging Station and Image Studio software (LI-COR), which has been shown to be more sensitive and quantitative than enhanced chemiluminescence (118). Transduction assay Vero cells or HEK 293T cells, transfected with one of four different C-type lectins or TIM-1 expression plasmids 24 h prior, were plated in a 48-well dish. Vero cells were transduced with the indicated VSV pseudovirions normalized for VSV matrix (WT MOI of ~0.2) and HEK 293T cells were transduced at an MOI of ~0.01. At h post transduction, cells were lifted with Accumax or Accutase (Sigma) and GFP expression was quantified by flow cytometry. A different stock of pseudovirus was used for each transduction experiment. Cell binding assay Vero cells, plated in a 12-well dish, were cooled to 4 C and incubated with equal volumes of VSV pseudovirions in growth media, in the presence or absence of 2mM EGTA. Plates were centrifuged at 700 x g and 4 C for 1 h to facilitate virus attachment to cells. Unbound virus was removed and cells were washed 3 times with cold PBS with or without 2mM EGTA. Cells were gently scraped and lysed in SDS loading buffer, followed by brief sonication. Immunoblots were performed on lysates and input virus using anti-vsv matrix mab 23H12 and anti-beta actin mab AC-15 (Thermo Scientific).

53 37 CatB inhibition assay Inhibition of EBOV GP-dependent entry with CatB inhibitor, CA-074 (Sigma), has previously been described (61, 64). Briefly, Vero cells were incubated with 80 µm CA-074 or equal volume of DMSO for 2 h prior to transduction. Cells were transduced with the indicated VSV pseudovirion, normalized to the amount of VSV matrix, in the presence of inhibitor or DMSO. Entry was quantified by egfp expression 18 to 20 h later by flow cytometry. Cathepsin B activity was assayed in Vero cells treated for 2 h with 80 µm CA-074 or equal volume of DMSO using the Z-Arg-Arg-MCA (Peptides International), as previously reported (119). Thermolysin sensitivity assay VSV pseudovirions, normalized to expression of GP, were incubated with thermolysin (THL) at 200 µg/ml, 2-fold serial dilutions starting with 5 µg/ml, or a single concentration of 1.25 µg/ml at 37 C for 15 minutes. Reactions were immediately placed on ice and diluted 20-fold in growth media containing 50 µm phosphoramidon (Sigma), THL inhibitor. THL treated pseudovirions were then used to transduce Vero cells. GFP positive cells were analyzed by flow cytometry as described above. Purification of soluble NPC1 C Loop Construction of a soluble NPC1 C Loop expression plasmid has previously been described (102). Briefly, the protein was expressed from a construct containing the NPC1 C Loop flanked by sequences that express a stable anti-parallel coiled-coil with both His6x and FLAG tags and preprotrypsin signal sequence. 293T cells were transfected with this construct and maintained in Opti-MEM (Gibco) + 1% penicillin/streptomycin (Gibco). At 72 h post transfection supernatants were passed through a 0.2µm filter and

54 38 concentrated in a 30K MWCO protein concentrator (Sartorius Stedim). Concentrated protein was incubated with HisPur Ni-NTA resin (Thermo Scientific) for 4 h and purified by batch method. Protein was exchanged into imidazole-free elution buffer (50mM MES ph mM NaCl) with 7K MWCO Zeba spin desalting columns (Thermo Scientific). Protein concentration was determined with a NanoDrop 3300 by absorbance at 280 nm (A 280 ). Soluble NPC1 C Loop GP binding assay 96-well Maxisorp ELISA plates (Thermo Scientific Nunc) were coated with anti- EBOV GP mab KZ52 (2 µg/ml) overnight at 4 C. Plates were blocked with PBS + 3% BSA for 2 h at room temperature. VSV pseudovirions (5x10 4 TU of EBOV GP or GPequivalent) were bound to ELISA plates at 37 C for 1 h. Pseudovirions bearing EBOV GP were treated with THL (200 µg/ml) for 1 h prior to addition to ELISA plates. The indicated amounts of soluble NPC1 C Loop were added and plates were incubated overnight at 4 C. Wells were washed extensively and bound soluble NPC1 C Loop was detected with an anti-dykddddk antibody conjugated to horseradish peroxidase and Ultra-TMB (Thermo Scientific). Isolation and transduction of murine peritoneal cells BALB/c IFN-α/β receptor-deficient (IFNAR -/- ) mice, a gift from Joan Durbin, NYU, were crossed with BALB/c T cell immunoglobulin and mucin domain protein 1 deficient (TIM1 -/- ) mice (120), a gift from Paul B. Rothman, Johns Hopkins University, to produce IFNAR -/- TIM1 -/- double knockout mice. All mouse experiments were approved by the University of Iowa Animal Care and Use Committee. Mice were sacrificed and resident peritoneal cells were harvested by lavage using 10 ml of ice-cold

55 39 RPMI 1640 medium (Gibco) containing 10% FBS. The recovered cells were washed three times in cold media and ~3.0x10 5 cells were plated per well in a 6-well for characterization or 96-well plate for transductions in the presence of 50 ng/ml of murine macrophage colony stimulating factor (M-CSF) for 72 h (Affymetrix ebioscience, ). M-CSF differentiated adherent murine peritoneal cells were phenotyped by staining for cell surface markers with the following antibodies from ebioscience: CD11c- PerCP-Cyanine5.5 (N418), CD11b-PE-Cyanine7 (M1/70) and F4/80-PE (BM8). For transductions, non-adherent cells were removed from each well and adherent cells were washed twice with PBS prior to the addition of serial dilutions of VSV pseudovirions, performed in at least quadruplicate. After 24 h, three individuals scored transductions by determining the wells that contained at least one GFP positive cell. This data was used to calculate the TCID50 values for each pseudovirion. Antibody/antisera neutralization assay Serial dilutions of fractionated IgG pooled from 36 convalescent, vaccinated cynomolgus macaques challenged with EBOV (gift from John M. Dye, USAMRIID) or pooled convalescent sera from mouse adapted EBOV-challenged mice (gift from Gene Olinger, USAMRIID) were incubated at 37 C for 30 min with VSV pseudotyped with the indicated glycoprotein, normalized to the amount of matrix protein. Mixtures were diluted 5-fold in growth media and added to Vero cells plated in a 48-well format. GFP positive cells were enumerated with flow cytometry. The relative antisera sensitivity was calculated as the reciprocal of the IC 50 values determined with GraphPad Prism 5.

56 40 Results Loss of the N-linked glycan shield does not affect EBOV GP expression, but enhances virion transduction In order to determine the role of GP1 N-glycans in EBOV GP-dependent entry we created a library of over 40 individual and combinatorial mutations to disrupt all the N- linked glycosylation sites (NGS; N-X-S/T sequons) within the GP1 subunit in the presence or absence of the MLD (GP and GP1Δmuc, respectively; Table 2-1). NGS mutants were expressed in HEK 293T cells and pseudotyped onto VSVΔG-eGFP. The relative expression of wild-type (WT) or NGS mutant GP was determined through assessing the GP to VSV-matrix ratio present in cell supernates by dot blot analysis and entry (transduction efficiency) was assessed in Vero cells. Initial studies were performed to disrupt the seven NGS within the GP1 core of GP1Δmuc and GP. Loss of glycosylation on mutant GPs pseudotyped onto VSV pseudovirions was apparent in immunoblots due to faster migration of the mutant proteins (Figure 2-2A). Total expression of WT and mutant GPs in supernates and transduction of Vero cells mediated by these virions is summarized in Figure 2-3. Notably, removal of more than 3 glycans from the GP1 core significantly decreased expression in GP1Δmuc but all could be removed from GP. Furthermore these mutants were incorporated into purified virions to an equivalent degree as WT (Figure 2-2B). Surprisingly, we were able to disrupt all NGS within GP1 without affecting expression (7Gm8G; Figure 2-4A). Comparison of the 7Gm8G mutant and PNGaseF treated GP with immunoblot analysis confirmed that the 7Gm8G mutant lacked all N- linked glycans (Figure 2-2C). Furthermore, entry into Vero cell was significantly

57 41 enhanced in mutants that lacked the glycan at N40 (Figure 2-4B). Additionally, removal of all NGS within the GP1 core (7G), MLD (GPm8G), and GP1 subunit (7Gm8G) resulted in a significant increase in GP-mediated entry. These results indicated that N- linked glycans are not required for entry into Vero cells, but rather decrease the efficiency of entry. GP1 N-glycans do not impact virion attachment to Vero cells As extensive deglycosylation of EBOV GP led to greater transduction, we sought to identify which step(s) in EBOV entry were affected by the loss of N-linked sugars. Removal of GP1 N-glycans within the GP1 core (7G) or throughout GP1 (7Gm8G) did not impact binding of pseudovirions to the cell surface (Figure 2-5, top) and binding of all tested virions was decreased to similar levels (~50% reduction) in the presence of a calcium chelator, EGTA (Figure 2-5, middle). This finding was not unexpected since EBOV entry into Vero cells has been shown to be mediated by the phosphatidylserine receptor, TIM-1, which binds to virion-associated phosphatidylserine in a calciumdependent manner (80). Removal of GP1 N-glycans imparts CatB independence and increases protease sensitivity Endosomal processing of EBOV GP by cathepsin B (CatB) is an important step in EBOV entry that results in removal of both the MLD and glycan cap, exposing the RBD (59-62). Additionally it has been shown that disruption of the N40 NGS (T42A mutation) leads to CatB-independence in the absence of the MLD (64). Pseudovirion entry mediated by GP1Δmuc, GP, and the GP mutant 6G that contains a fully deglycosylated

58 42 glycan cap was abrogated by treatment of Vero cells with the CatB inhibitor CA-074, which corresponded with a dramatic decrease in CatB activity (Figure 2-6A), indicating that these GPs were dependent upon CatB processing for subsequent transduction steps (Figure 2-6B). In contrast, VSV pseudotyped with the native glycoprotein (G) was completely resistant to the drug. Consistent with the previous report, mutation of T42 in the absence of the MLD was significantly less sensitive to the CatB inhibitor (T42VΔmuc, Figure 2-6B), however we did not observe complete CatB-independence possibly due to a Val rather than Ala substitution at this site. Assessment of the T42V mutant in the context of GP also demonstrated partial CA-074 independence. The 7G mutant provided similar levels of independence of CatB cleavage as the GP T42V mutant, consistent with our findings with 6G and suggesting that the 6 glycans present in the glycan cap had little to no effect on CatB dependence of GP. Interestingly, removal of all N-linked glycans in GP1 (7Gm8G) decreased the sensitivity to CA-074 to the greatest extent for any of the EBOV GP mutants and to an equivalent extent as T42V GP1Δmuc. Yet, removal of all N-linked glycans from only the MLD (GPm8G) did not impart any degree of CatB-independence. Removal of the glycan at N40 has been hypothesized to increase access to the cathepsin cleavage site (64) and glycosylation is known to protect from proteases (121), therefore we determined if removal of glycans increased protease sensitivity of GP. Previous work has shown that treatment with 200 µg/ml thermolysin (THL) results in a ~20-fold increase in GP-mediated transduction associated with complete removal of the glycan cap and MLD (59, 64). To determine if deglycosylation of GP leads to increased protease sensitivity we compared the transduction of Vero cells by VSV pseudotyped

59 43 with GP or 7Gm8G treated with increasing concentrations of THL to transduction levels obtained at a high concentration of THL (200 µg/ml). 7Gm8G was dramatically more sensitive to low concentrations of THL than GP, providing evidence that the deglycosylated GP was more readily proteolytically processed (Figure 2-6C). Given these results a panel of NGS mutants was tested to determine the relationship between entry and protease sensitivity of deglycosylated GP (Figure 2-6D). As expected, there was a correlation between the number of mutations and the THL sensitivity. Results from 7G and GPm8G indicated that removal of glycans from the core has a greater effect on protease sensitivity, possibly due to large number of O-glycans still present in the MLD. Furthermore, the results from 7Gm8G showed there was an additive effect on THL sensitivity when the mutations of 7G and GPm8G were combined. These findings indicated that N-glycans in the glycan cap, and to a lesser extent in the MLD, control the efficiency of proteolytic processing, thereby regulating subsequent steps in entry such as binding to the endosomal receptor, NPC1. Removal of glycans shielding the RBD does not allow for NPC1 binding Since enzymatic removal of the glycan cap and MLD allows interaction with the endosomal receptor NPC1 (36, 63), we postulated that exposure of the RBD by disrupting NGS within the GP1 core and MLD might enhance binding of GP to NPC1. In an NPC1 binding assay, a soluble form of the second luminal domain (C loop) of NPC1, which directly interacts with residues in the RBD of EBOV GP (36), was used to bind VSV pseudovirions. Treatment of GP with THL, which is commonly used to mimic GP cleavage by CatB/L in removing the MLD and glycan cap (59, 64), resulted in binding of

60 44 C loop binding over a wide range of concentrations (Figure 2-6E). However, exposure of the RBD through deglycosylation (7Gm8G) did not lead to binding of the NPC1 C loop above background, indicating that loss of N-linked glycans on GP1 does not abrogate the need for endosomal proteolysis that is required for receptor interactions. This result suggests that while N-glycans on GP may interfere with the ability of the RBD to interact with this lysosomal receptor, glycan interference is not the sole explanation for the inability of GP to bind to NPC1. The requisite proteolytic processing may also expose residues involved in C loop binding, such as F88 (36), that may only be available after proteolytic removal of the glycan cap and MLD. Domain specific N-linked deglycosylation altered C-type lectin utilization CLECs, that mediate GP-dependent entry, are expressed on a wide variety of EBOV target cell types including macrophages, dendritic cells, and hepatic cells and each binds to specific glycan moieties (83, 87, 91, 122, 123). To better understand the role of EBOV GP1 N-linked glycans in CLEC-dependent tropism, we determined the receptor utilization of mutants that lack N-linked glycans on the GP1 core (7G), the MLD (GPm8G), or both (7Gm8G). Constructs expressing Myc-tagged CLECs were transfected into poorly permissive HEK 293T cells in order to observe the impact on entry by each. All constructs expressed well in transfected cells (Figure 2-7). In these experiments, transfection of a TIM-1-expressing plasmid served as a transduction control since TIM-1-dependent entry into endosomes is GP-independent and should be independent of the GP glycosylation status (70, 80). Consistent with this, all

61 45 pseudovirions were enhanced by transfection of TIM-1 in a similar manner (Figure 2-8A). While all five CLECs enhanced transduction of WT GP between 6- and 13-fold, the impact of the loss of GP N-glycans on transduction varied depending on the CLEC examined (Figure 2-8B-F). Entry mediated by L-SIGN, L DC-SIGN, and LSECtin were highly dependent on N-glycans present throughout GP1, however LSECtin-dependent entry was less dependent on N-glycans present in the GP1 core (Figure 2-8B-D). ASGPRI-mediated entry was completely independent of N-glycans in the GP1 core, however this CLEC was able to mediate entry independent of GP1 N-glycans (Figure 2-8E). N-linked deglycosylation had minimal impact on GP-mediated entry into cells expressing L hmgl (Figure 2-8F). Results obtained with ASGPRI and L hmgl were consistent with previous reports indicating that ligands for these CLECs are present in O- linked glycans present in the MLD or potentially in N-linked glycans in GP2. CLEC-expressing HEK 293T cells were also used to evaluate the impact of N- linked glycans on our MLD-deleted GP, GP1Δmuc. In these studies, we compared transduction of GP1Δmuc to 5GΔmuc. While our 5GΔmuc mutant was poorly expressed and resulted in low levels of transduction compared to GP1Δmuc (Figure 2-3), levels of 5GΔmuc transduction were about 50% of that observed with GP and sufficient to allow transduction studies to be performed (Figure 2-9). All five CLECs enhanced GP1Δmuc transduction, increasing entry by 4.5- to 22-fold (Figure 2-8B-F). In all cases, loss of five of the N-linked glycans (5GΔmuc) on the glycan cap resulted in a dramatic decrease in transduction. However, the reduction was more modest for transduction mediated by L DC-SIGN, ASGPRI, and L hmgl. This enhancement in entry for 5GΔmuc cannot result

62 46 from interactions with MLD glycans and therefore must be due to either the two remaining intact N-glycans within the GP1 core at N40 and N204 or the two N-glycans in GP2. Additionally, the differences in 5GΔmuc-mediated transduction observed between L-SIGN and L DC-SIGN indicates that these CLECs have different ligand specificities, despite both binding to high mannose oligosaccharides. N-linked glycans in GP1 are not required for entry into macrophages Macrophages are major, early targets during filovirus infections and others have suggested CLECs may be important for entry into these cells (54, 56, 124). Previous work has shown that murine MGL and SIGNR1, the murine homolog of DC-SIGN, are expressed on peritoneal macrophages (125, 126). We used our most extensively deglycosylated mutants to determine the role of N-linked and O-linked glycans in entry into these cells. Since VSV is highly sensitive to the type I interferon response (127), resident peritoneal cells were isolated from IFNAR -/- mice for these studies. Additionally, these mice lacked TIM-1, which eliminated the possibility for entry in epithelial cells that may have contaminated the preparations. Cells were treated for 3 days with murine M- CSF and phenotyped prior to use. Eighty-five percent of adherent cells from the peritoneal cavity were CD11b + /F480 +, indicative of matured macrophages (data not shown). These cells were highly permissive for VSV pseudotyped with GP (Figure 2-10). Entry mediated by 7Gm8G was enhanced ~3-fold over WT, which is more consistent with our Vero cell data than CLEC utilization data. Furthermore, macrophages supported entry mediated by the 5GΔmuc mutation although this was dramatically decreased compared to WT. These findings demonstrated that CLEC/EBOV GP1 N-glycan

63 47 interactions are not necessary for entry into these cells and in the absence of the MLD a limited number of N-glycans are sufficient for entry. Removal of N-linked glycans enhances antisera sensitivity Our studies indicated that the most deglycosylated EBOV GP1 mutants were expressed to WT levels and provided the highest levels of transduction into Vero cells and peritoneal macrophages. Nonetheless, conservation of the glycan sites, particularly in the GP core sequences, across the species suggested that positive selection for these sites was occurring. Therefore, we evaluated our deglycosylated EBOV GPs for their sensitivity to antibody-mediated neutralization. Anti-EBOV IgG was purified from the pooled antisera of 36 convalescent cynomolgus macaques that were vaccinated with Venezuelan equine encephalitis virus replicon particles expressing GP prior to challenge with Ebola virus (a gift from John M. Dye, USAMRIID). VSV pseudovirions that were normalized for the amount of matrix protein were then incubated with various dilutions of this anti-ebov IgG. Neutralization efficiency of pseudovirions bearing WT or EBOV GP mutants were compared since deglycosylation may increase exposure of potential EBOV GP epitopes. Little to no enhancement of antibody sensitivity was observed with a mutant lacking three glycans in the GP core; however, removal of six glycans from the glycan cap increased neutralization sensitivity over 4-fold (Figure 2-11A and C). In contrast VSV pseudotyped with the Marburg virus GP was not neutralized by anti-ebov IgG. Complete deglycosylation of the GP1 core (7G) led to a further increase in antibody sensitivity (Figure 2-11B and C). Since the MLD is a major target of neutralizing antibodies ( ), we investigated the impact of removal of N-glycans from this domain had on anti-ebov IgG sensitivity. Surprisingly, GPm8G was no more sensitive

64 48 to neutralization than WT GP (Figure 2-11B), indicating that NGS in the MLD had little to no positive or negative effect on neutralization by this anti-ebov IgG prep. Consistent with this, combined removal N-linked glycans from the MLD with deglycosylation of the GP1 core (7Gm8G) did not further increase the sensitivity of 7G to antibody neutralization. We found similar, but less pronounced neutralization sensitivity with pooled sera collected from mice surviving challenge with mouse-adapted EBOV (Figure 2-12). Discussion EBOV GP1 is highly glycosylated, yet limited studies to date have investigated the role of these sugar chains on GP function. In this report, we sought to determine the effects of removing the N-linked glycans present on the EBOV GP1 subunit. We examined the impact on protein expression/stability, viral entry and antisera/antibody sensitivity. Surprisingly, elimination of all N-linked glycans on GP1 had no effect on expression levels of the protein, suggesting that, in the presence of the MLD, N-linked glycans attached to GP1 are not critical for GP folding. However, as our studies were performed with pseudovirions in tissue culture, it remains possible that these N-linked glycans are critical for GP stability in the context of infectious EBOV virions and/or in vivo. Loss of GP1 NGS significantly decreased utilization of four out of the five C-type lectins known to be used by EBOV GP for entry, while enhancing EBOV transduction into Vero cells and peritoneal macrophages. Enhanced levels of transduction were associated with enhanced sensitivity to thermolysin cleavage, but deglycosylation did not alleviate the need for proteolysis prior to NPC1 binding. Lastly, loss of highly conserved glycans on the core of GP1 increased virus sensitivity to antibody neutralization.

65 49 Therefore, we propose that the strong conservation of individual glycosylation sites in the GP1 core across the ebolavirus genus results, at least in part, from selective pressures to protect the virus against immune responses despite the fact that these sites collectively can have a negative impact on GP-dependent entry. Future studies testing our deglycosylated GPs in the context of recombinant filoviruses to verify our findings is warranted. Modeling N-linked glycans onto the GP1Δmuc structure suggests that the highly conserved RBD of ebolaviruses is effectively masked and protected by glycan cap glycans attached to well-conserved NGS, likely protecting this conserved region from selective immune pressures. The presence of N-linked glycans in the glycan cap of EBOV GP has been previously proposed to protect the virus from antibody-mediated neutralization (42, 52, 62, 108, 109, 128); however, we provide the first experimental evidence to support this claim. Our results suggest that there are neutralizing epitopes in the GP1 core that are masked by N-glycans that normally provides some protection against antibody binding. Surprisingly, removal of N-glycans from the MLD, which is highly targeted by neutralizing antibodies ( ), did not impact antibody sensitivity. Neutralizing epitopes within the MLD that are not blocked by N-linked glycans may be obstructed by the large number of O-glycans attached to the MLD. Since filovirus hemorrhagic fevers are acute, and often lethal infections in primates, it is likely that antibody-driven positive selection for NGS surrounding conserved regions occurs in nonprimate reservoirs, such as bats. Thus it is important to understand filovirus infection and persistence in bat populations.

66 50 A role for CLECs in EBOV GP-dependent entry has been firmly established by others (54-57). We have shown for the first time that removal of N-glycans from either the glycan cap (7G) or MLD (GPm8G) dramatically decreases utilization of DC-SIGN, L-SIGN, and LSECtin. Surprisingly, these same CLECs enhanced transduction mediated by GP1Δmuc, despite this form of the protein sharing the exact pattern of NGS as GPm8G. Additionally, hmgl utilization was largely unaffected by loss of GP1 N- glycans, providing indirect evidence that hmgl-dependent EBOV transduction is strongly dependent on O-linked glycans in GP. Despite the apparent utilization of O- linked glycans by hmgl, transduction mediated by GP1Δmuc was greatly enhanced by expression of this CLEC (54). In combination, these results suggest that deletion of the MLD alters the species of glycans present on the protein. Previous work has shown that there are differences in glycosylation between GP and the small, soluble form of GP (sgp), which has shared NGS with GP1Δmuc (53). A higher degree of glycan processing was found on sgp, indicated by the higher percentage of galactose residues and lower percentage of high mannose glycans. Since the types of glycans present on proteins are largely dependent on the environment in which the protein is produced (129), it is possible that the cytotoxicity associated with over-expression of GP (130, 131) or the large number of O-glycans present, changes the glycosylation machinery in the ER/Golgi. In fact, EBOV GP pseudovirions produced in monocyte-derived macrophages altered the CLEC utilization compared to virus produced in HEK 293T cells (88). Additionally, previous work has identified differences in glycan species and glycosylation site occupancy in HIV gp120 produced from different cell lines (132). Since all of our pseudovirions were produced in HEK 293T cells, glycosylation on our

67 51 viral particles may not represent the glycosylation diversity of GP found on virions that are produced from the broad spectrum of infected cell types during in vivo filovirus infection. Since CLECs are expressed on a wide variety of cells and bind to their respective sugar ligands in an EBOV GP domain-specific manner, our deglycosylated mutants can be used in future studies in the context of infectious EBOV to determine the role of CLECs in tissue tropism during in vivo infections. Previous work has shown that macrophages are important cellular targets during early days of filovirus infection (124). Effective transduction of murine peritoneal macrophages by N-glycan-deficient GP1-containing pseudovirions along with our findings that hmgl-dependent enhancement of transduction was relatively independent of the presence of GP1 N-glycans supports the idea proposed by others (54) that EBOV entry into macrophages may be mediated by this CLEC. However, additional mechanisms of EBOV entry into macrophages may also be important. We and others have identified that TIM family members mediate virus uptake into cells by interacting with phosphatidylserine on the surface of virions (70, 80). Further, we report here that TIM-dependent entry into cells was independent of the glycan status of the GP. Since TIM-4 is highly expressed on the surface of peritoneal macrophages (133), this lipiddependent mechanism may also be important for filovirus entry into APCs. Additionally, there may be cooperative mechanisms of virus attachment and internalization between cellular CLECs and phosphatidylserine receptors to efficiently mediate entry of viral particles. Future studies need to explore the role for phosphatidylserine receptors during early filovirus infection.

68 52 The ability of deglycosylated GP pseudovirions to enhance transduction of Vero cells and murine peritoneal macrophages provided evidence that N-glycans on GP1 can decrease the efficiency of the entry process. We found that removal of glycans masking the RBD enhanced proteolytic processing of GP, but did not result in the ability of unprocessed EBOV GP to bind the C loop of NPC1, consistent with an earlier study demonstrating that removal of the MLD did not unmask the RBD for NPC1 interaction (36). This result suggests that there are residues, critical for NPC1 binding, that are masked by the glycan cap/mld polypeptide rather than being concealed by the heavy glycan shield. This finding should help to identify other, previously undefined, residues that interact with the intracellular receptor. Thus, it is likely that increased sensitivity of our deglycosylated mutants to proteolytic processing results in more efficient transit through the endosomal compartments, leading to greater transduction efficiency. While it is possible that enhanced protease sensitivity may be deleterious to fusion events mediated by some viral glycoproteins, our results indicate that for filoviral GPs, which require proteolytic processing, access to protease-sensitive cleavage sites made available by deglycosylation leads to facilitation of the entry process. In these studies, we found that protease sensitivity and CatB dependence are independently controlled within GP1. For instance, GPm8G that retains the N40 glycosylation site was completely dependent on CatB, but had enhanced thermolysin sensitivity. In contrast, T42V was found to be partially CatB independent, but no more sensitive to thermolysin than WT GP. It was previously hypothesized that mutation of residue T42, that eliminates the N-glycan at N40, allows for other endosomal proteases, apart from CatB, to process GP within the endosome (64). Extrapolating from these

69 53 studies, we hypothesized that systematic deglycosylation would yield an EBOV GP that was progressively more CatB-independent. However, this was not the case; our results indicate that extensive deglycosylation beyond loss of the N40 glycan sequon did not further alleviate dependence on CatB, but the extensive deglycosylation found in 7G or 7Gm8G did increase thermolysin sensitivity. Additionally, mutation of N40 versus T42 was shown to display different protease sensitivity, suggesting that the glycan is not the only factor in determining CatB-dependence (64). Previous work has shown that cellular receptors influence the intracellular trafficking and endosomal fate of viruses (134). EBOV entry requires a specific intracellular compartment that requires CatB, HOPS complex, and NPC1 (61, 63, 103). However, similar to our CatB-independent EBOV mutants, not all filoviruses require CatB cleavage for entry (99). Therefore, deglycosylated mutants, with distinct protease dependence and sensitivity profiles, may be useful in the further characterization of the endosomal compartments that allow for the efficient entry of filoviruses.

70 54 A B C Figure 2-1 Schematic diagrams of Ebola virus GP. (A) A molecular model of EBOV GP1/2 shown in atop-down view. Complex N-glycans are shown in orange, GP is shown in light gray, RBD is shown in red, and MLD structure that has not been solved is represented as a gray sphere. PBD ID: 3CSY. (B) Linear model of EBOV GP. The disulfide bond between GP1 and GP2 is indicated, as well as the locations of N-linked glycans (marked with Ys ) in the GP1 and -2 domains, and the known protease cleavage sites are noted. SP, signal peptide; RBD, receptor-binding domain; MLD, mucin-like domain; IFL, internal fusion loop; HR1 and -2, heptad repeats 1 and 2; TM, transmembrane domain. (C) Alignment of predicted N-linked glycan sites within the GP1 core of the five Ebola virus species. N-X-S/T sequons are highlighted with a black background.

71 55 Table 2-1 Nomenclature of N-glycan site mutations in EBOV GP1 and GP1Δmuc. Name Mutations Mutation present in GP GP!muc T42V T42V / T206V / T230V / N238D / T259V / T270V / T298V + + T206V T42V / T206V / T230V / N238D / T259V / T270V / T298V + + T230V T42V / T206V / T230V / N238D / T259V / T270V / T298V - + N238D T42V / T206V / T230V / N238D / T259V / T270V / T298V + + T259V T42V / T206V / T230V / N238D / T259V / T270V / T298V - + T270V T42V / T206V / T230V / N238D / T259V / T270V / T298V - + T298V T42V / T206V / T230V / N238D / T259V / T270V / T298V - + T230V/N238D T42V / T206V / T230V / N238D / T259V / T270V / T298V + + N238D/T259V T42V / T206V / T230V / N238D / T259V / T270V / T298V - + T230V/T259V T42V / T206V / T230V / N238D / T259V / T270V / T298V - + N238D/T270V T42V / T206V / T230V / N238D / T259V / T270V / T298V + + T42V/T206V T42V / T206V / T230V / N238D / T259V / T270V / T298V + - T206V/T230V T42V / T206V / T230V / N238D / T259V / T270V / T298V + + T206V/T259V T42V / T206V / T230V / N238D / T259V / T270V / T298V + + T230V/N238D/T259V T42V / T206V / T230V / N238D / T259V / T270V / T298V + + T42V/T206V/T230V T42V / T206V / T230V / N238D / T259V / T270V / T298V + - T42V/T206V/T259V T42V / T206V / T230V / N238D / T259V / T270V / T298V + + T206V/T230V/T259V T42V / T206V / T230V / N238D / T259V / T270V / T298V + + 4G T42V / T206V / T230V / N238D / T259V / T270V / T298V + + 5G T42V / T206V / T230V / N238D / T259V / T270V / T298V + + 6G T42V / T206V / T230V / N238D / T259V / T270V / T298V + - 7G T42V / T206V / T230V / N238D / T259V / T270V / T298V + - Name Mucin-like domain N-linked glycan site mutations Mutation present in GP 7G m1g S319A / T335V / S348A / T388V / S415A / S438A / S456A / T464V - + m2g S319A / T335V / S348A / T388V / S415A / S438A / S456A / T464V - + m3g S319A / T335V / S348A / T388V / S415A / S438A / S456A / T464V - + m4g S319A / T335V / S348A / T388V / S415A / S438A / S456A / T464V - + m5g S319A / T335V / S348A / T388V / S415A / S438A / S456A / T464V - + m6g S319A / T335V / S348A / T388V / S415A / S438A / S456A / T464V - + m7g S319A / T335V / S348A / T388V / S415A / S438A / S456A / T464V - + m8g S319A / T335V / S348A / T388V / S415A / S438A / S456A / T464V + +

72 56 A B C Figure 2-2 Immunoblot analysis of GP1 NGS mutant pseudovirions from HEK293T supernatants. (A) Shown is the migration pattern of GP1 from various glycan cap and base domain N-linked glycosylation mutants. GP mutants are shown at the top, and GP1Δmuc mutants are shown at the bottom. GP1 was detected with a rabbit anti-ebov GP1 polyclonal Ab. (B) GP incorporation into VSV pseudovirions. VSV pseudotyped with the indicated NGS mutants were purified through a 20% sucrose gradient and analyzed by immunoblotting to determine the ratio of GP to VSV matrix (shown at the bottom, normalized to WT). (C) Immunoblot of mock or PNGase F-treated VSV pseudotyped with the indicated GP NGS mutant. (B and C) EBOV GP1 was detected with mab 5E6, and VSV matrix was detected with mab 23H12.

73 57 A Rel. Expression GP:Matrix Signal (GP1!muc = 100%) B Rel. Transduction (GP1!muc = 100%) C Rel. Expression GP:Matrix Signal (GP = 100%) D Rel. Transduction (GP = 100%) E Rel. Expression GP:M Signal (GP = 100%) GP GP T42V T42V T206V T206V N238D T42V/T206V N238D T42V/T206V T206V/T230V T206V/T230V T206V/T259V T42V/T206V/T230V T206V/T259V T42V/T206V/T259V T206V/T230V/T259V T42V/T206V/T230V T206V/T230V/T259V Figure 2-3 Summary of relative expression and Vero cell transduction efficiency of GP1 N-glycan site mutants. (A, C, and E) Expression of GP1Δmuc (A) and GP (C and E) NGS mutants. VSV pseudovirions in HEK293T cell supernatants were evaluated by dot blots for the amount of EBOV GP/GP1Δmuc and VSV matrix (M). The signal was quantified as described in Materials and Methods, and data are represented as ratios of the averages ± SD of the GP/M of three independent stocks, normalized to WT. (B, D, and F) Transduction efficiencies of GP1Δmuc (B) and GP (D and F) NGS GP mutants. VSV-eGFP pseudovirions (WT MOI, ~0.2) were normalized to the amount of VSV matrix protein and transduced into Vero cells. Transduction findings are represented as percentages of WT GP1Δmuc (B) or GP (D and F). Data are shown as the averages ± SEM of three stocks of virus assayed independently. Significance was calculated by 1- sample t test. *, P < 0.05; **, P < 0.005; ***, P < (A to D) Findings with N- glycan mutants in the base and glycan cap of GP1. (E, F) Findings for mutants that stepwise combine the 7G mutant with mucin-like domain N-linked glycan mutations. T230V/N238D T42V/T206V/T259V N238D/T270V 250 * GP 7Gm1G * * * * * 7Gm2G 7Gm3G 7Gm4G 7Gm5G 7Gm6G 7Gm7G * 7Gm8G GPm8G GP!muc T42V T206V T230V N238D T259V T270V T298V T206V/T230V T206V/T259V T230V/N238D N238D/T259V T230V/T259V N238D/T270V T42V/T206V/T259V T206V/T230V/T259V T230V/N238D/T259V 4G 5G * * * * T230V/N238D N238D/T270V * GP!muc T42V T206V T230V N238D T259V T270V T298V T206V/T230V T206V/T259V T230V/N238D N238D/T259V T230V/T259V N238D/T270V T42V/T206V/T259V T206V/T230V/T259V T230V/N238D/T259V 4G 5G F Rel. Transduction (GP = 100%) * T230V/T238V/T259V 4G 5G 6G 7G T230V/T238V/T259V 4G 5G 6G 7G * * * * * * * * GP 7Gm1G 7Gm2G 7Gm3G 7Gm4G 7Gm5G 7Gm6G 7Gm7G 7Gm8G GPm8G

74 58 A Expression GP:Matrix Signal (% GP control) B Transduction % GP Control P V * V * GP T42V/T206V/T259V G 7G * G * * G * P * * * * GPm8G 7Gm8G No GP Figure 2-4 Expression and entry efficiency of selected EBOV GP N-glycan mutants (see Figure 2-3 in the for a full analysis). (A) Relative expression of GPs. Dot blots were used to determine the ratio of EBOV GP to VSV matrix in supernatants. No GP represents pseudovirus produced in the absence of an envelope glycoprotein. (B) Transduction of Vero cells by VSV pseudotyped with the indicated NGS mutant (normalized to the amount of VSV matrix protein), presented as the percentage of the GP control value. (A and B) Data represent the averages ± SEM of three independent stocks of virus. Significance was calculated by 1-sample t test. *, P < 0.05; **, P < 0.005; ***, P <

75 59 Pseudovirion Figure 2-5 Vero cell binding assay of WT or mutant GPs depleted of all N-linked glycans in the core and glycan cap domains (7G) or throughout GP1 (7Gm8G). Cells were bound by the indicated pseudovirions without (top) and with (middle) 2 mm EGTA, washed, and lysed. Immunoblots of cell lysates were probed for VSV matrix and for cellular actin that serves as a loading control. Equivalent amounts of VSV matrix in the input viruses are shown at the bottom. Representative immunoblots for two independent experiments are shown.

76 Figure 2-6 Effects of N-glycan removal on entry processes. (A) CatB activity assay. Vero cells were pretreated with 80 µm CA-074 or an equal volume of DMSO for 2 h prior to cell lysis. Cell lysates were assayed for CatB activity with a fluorogenic substrate. CatB activity is presented as the average percentage of activity observed with DMSO treatment ± SD (n = 2). (B) CatB inhibition assay. Vero cells were transduced with pseudovirions bearing the indicated GPs in the presence of 80 µm CA-074. Data are presented as percentage of cells transduced compared to results for DMSO control. Shown are averages ± SEM of three independent experiments. (C and D) THL sensitivity assays. Pseudovirions, normalized for GP expression, were incubated with (C) the indicated concentration of THL or (D) 1.25 µg/ml of THL for 15 min at 37 C prior to transduction of Vero cells. Data are presented as percentage of transduction observed at 200 µg/ml THL. Relative transduction efficiency after treatment at each THL concentration is presented as the percentage of transduction at 200 µg/ml THL treatment. The data represent the averages ± SEM of three independent experiments, performed with three independent stocks of pseudovirus. (E) NPC1 C loop binding assay. VSV pseudovirions bearing the indicated GP were bound to an enzyme-linked immunosorbent assay (ELISA) plate and incubated with soluble NPC1 C loop. The amount of bound protein was quantified by ELISA. As a positive control, VSV pseudotyped with GP was treated with THL (GP THL ) (200 µg/ml) for 1 h at 37 C prior to incubation with C loop. Graph is representative of two independent experiments. 60

77 Figure 2-7 Immunoblot of lysates from HEK293T cells transfected with the indicated myc-tagged CLEC. Endogenous c-myc is marked with an asterisk and serves as a loading control. Locations of molecular mass markers are shown on the left. 61

78 62 Fold change transduction (Empty HEK293T cells) A B C D E F Figure 2-8 CLEC utilization of N-glycan site mutants. (A to F) HEK293T cells transfected with the indicated entry factor were transduced at an MOI of 0.01 with VSV pseudotyped with the NGS mutant noted. Enhancement of transduction is presented as the fold change in GFP-positive cells compared to the results for empty vectortransfected cells, which are indicated with dashed gray lines. Data represent the averages ± SEM between three independent experiments. Significance was determined by Student s t test. *, P < 0.05; **, P < 0.005; ***, P < Note that the scale of the y axis varies between panels.

79 63 Rel. Transduction/Expression (GP = 100%) 150 Transduction Expression GP 5G!muc GP 5G!muc Figure 2-9 Comparison of Vero cell transduction efficiency (black) and expression (gray) of GP and 5GΔmuc. Vero cells were transduced with VSV pseudovirions normalized to matrix protein. Expression was determined by comparison of GP/M supernatant ratios. Data are represented as the percentages of GP values. Shown are the averages ± SEM of three independent experiments. Each experiment was performed with independent stocks of pseudovirus.

80 TCID 50 /ml GP 7Gm8G 5G!muc Figure 2-10 N-glycan site mutant-mediated entry into murine macrophages. Peritoneal macrophages from BALB/c IFNAR / TIM1 / mice were plated in 96-well plates and treated for 72 h with M-CSF. Following M-CSF treatment, cells were transduced with VSV pseudotyped with the indicated GP. TCID 50 values were calculated from two independent experiments performed at least in quadruplicate and are presented as the averages ± SD.

81 65 A B C Transduction (No IgG = 100%) Log IgG Dilution GP T42V/ T206V/ T259V 6G MARV Transduction (No IgG = 100%) 125 GP GPm8G 100 7G 7Gm8G 75 MARV Log IgG Dilution Antibody sensitivity (1/IC 50 ) * ** GP 42/206/259 6G7G GPm8G 7Gm8G * Figure 2-11 Enhanced neutralization of glycan cap N-glycan mutants by cynomolgus macaque anti-ebov IgG. (A and B) Neutralization of pseudovirus, normalized to the amount of VSV matrix protein, by fractionated IgG from convalescent vaccinated and challenged cynomolgus macaques. For ease of viewing the findings, results were separated to show neutralization of the mutants with 3 or 6 glycans removed within the GP1 core (A) or all 7 core glycans removed compared to all 15 in GP1 (B). Data points for neutralization assays are presented as the percentages of the no-igg control and represent the averages ± SD (n = 3). MARV, Marburg virus GP pseudotyped onto VSV. (C) The relative antibody sensitivity values, reciprocals of the IC 50 s, were determined for each independent experiment. Data are presented as the averages ± SD (n = 3). Significant differences between WT GP and the other constructs were determined by Student s t test. *, P < 0.05; **, P <

82 66 A B C Transduction (No Serum = 100%) Log Serum Dilution GP T42V/ T206V/ T259V 6G MARV Transduction (No Serum = 100%) 125 GP GPm8G 100 7G 7Gm8G 75 MARV Log Serum Dilution Antisera sensitivity (1/IC 50 ) GP 42/206/259 6G7G * ** GPm8G 7Gm8G Figure 2-12 Convalescent mouse antisera neutralization of VSV pseudotyped with the indicated NGS mutants. (A and B) Neutralization of pseudovirus, normalized to the amount of VSV matrix, by pooled convalescent murine antiserum. For clarity, results were separated to show neutralization of the mutants with 3 or 6 glycans removed within the GP1 core (A) or all 7 core glycans removed compared to mutants that included removal of N-linked glycan sites within the MLD (B). Data points for neutralization assays are presented as the percentages of the no-serum control and represent the averages ± SD (n = 3). (C) Relative antiserum sensitivity values, representing the reciprocals of the IC 50 s, were determined for each independent experiment. Data are presented as the averages ± SD (n = 3). Significant differences between WT GP and the other constructs were determined by Student s t test. *, P < 0.05; **, P <

83 67 CHAPTER III THE ROLE OF EBOLA VIRUS GP2 N-LINKED GLYCANS Introduction Ebolaviruses and marburgviruses are members of the family Filoviridae that cause highly-lethal, sporadic outbreaks of viral hemorrhagic fever throughout central Africa. Recently, a new virus, Lloviu virus (LLOV) of a novel genus within the family, cuevavirus, was identified circulating in bat populations in northern Spain (2). There is only one species of marburgvirus, Marburg marburgvirus (type virus Marburg virus, MARV). However there are five ebolavirus species with the type viruses named after geographical regions where they were first discovered: Ebola virus (EBOV), Sudan virus (SUDV), Taï Forest virus (TAFV), Bundibugyo virus (BDBV), and Reston virus (RESTV) (23). Currently there are no vaccines or antivirals available for prevention/treatment of filoviral hemorrhagic fevers. Filovirus entry is mediated by the sole viral glycoprotein (GP), which is composed of a trimer of GP1/GP2 heterodimers. To date there have not been any GPspecific cell surface receptors described. Instead, these viruses utilize phosphatidylserine (PS) receptors, that recognize PS lipids in the viral membrane, or C-type lectins, that recognize a broad range of glycan species on GP, to facilitate cellular binding and internalization (54-57, 70, 79, 80). Virus is trafficked to the late endosome/lysosome where cellular cysteine proteases remove heavily glycosylated regions of GP1 (glycan cap and mucin-like domain, MLD) to expose the highly conserved receptor-binding domain (RBD) (59-62). The RBD then interacts with an intracellular receptor, Niemann- Pick C1 (NPC1) (36, 63). The role of NPC1 has yet to be discovered, however GP

84 68 interaction with this protein is required during entry prior to GP2-mediated fusion of the virus-host membranes (36, 63, 103). The EBOV GP is extensively glycosylated, with over half of the molecular weight attributed to carbohydrates (41). The GP1 subunit contains 15 N-linked glycosylation sites (NGS) and potentially 80 O-linked glycosylation sites (52). We have previously shown that removal of all N-linked glycan sites from this subunit had minimal impact on expression and resulted in an increase in GP-mediated entry, which correlated with an increase in protease sensitivity (135). Removal of GP1 N-linked glycans decreased CLEC-dependent entry, but enhanced entry into primary murine macrophages. Furthermore, removal of N-linked glycans surrounding the highly conserved RBD resulted in enhanced antibody-mediated neutralization of virus entry. The GP2 subunit of filoviruses contains two completely conserved NGS, N563 and N618, which are located in the heptad-repeat 1 (HR1) and HR2 regions of GP2, respectively (Figure 3-1). Previous work has shown that both of these sites are occupied by glycan modifications (136). The conservation of these sites within the family suggests functional significance. We performed site-directed mutagenesis on both GP2 NGS to determine the role of these glycans during entry. We found that disruption of either site impacted the ability of a conformation-specific antibody to recognize the protein. However, both mutants were able to support entry. Mutation of the N563 NGS resulted in a 2-fold increase in entry, while mutation of the N618 NGS resulted in a 2-fold decrease. We found that these glycans are sufficient to support entry into DC-SIGN, ASGPRI, and hmgl expressing cells in the absence of the GP1 MLD, however in the context of the full-length protein

85 69 removal of these sites had no impact on CLEC-dependent entry. Overall we have provided evidence that these sites are involved at a post-internalization step during GPmediated entry, potentially through coordinating fusion. Materials and Methods Cell lines and plasmids Vero cells and HEK293T cells were maintained in Dulbecco s Modified Eagle Medium (DMEM; Gibco) + 10% fetal bovine serum (FBS) + 1% penicillin/streptomycin. The pcdna3.1 expression plasmids for EBOV GP, EBOV GP1Δmuc, and the N- linked glycan-deficient GP1 mutant, 7Gm8G, have been previously described (135). Construction of C-terminal myc-tagged CLEC expression plasmids was previously described (135). The full EBOV GP ORF was PCR amplified from the pcdna3.1 expression vector inserted upstream of a HA-tag (amino acid sequence YPYDVPDYA) followed by a stop codon. Modeling of GP N-linked glycans The pre-fusion EBOV GP 1,2 ΔTM structure (PDB ID: 3CSY) lacks the C-terminus of the protein (amino acids ). Therefore, the heptad repeat 2 from the post-fusion EBOV GP2 structure (PBD ID: 2EBO) was placed at the base of the ectodomain, using PyMol software, to serve as a predictive model of the stalk region. Additionally, four NGS in the ectodomain were lacking due to disordered structures (N204 and N296) or were mutated in order to promote crystallization (N40 and N228). In order to model predicted NGS the EBOV GP sequence was submitted to the PHYRE2 protein fold recognition server (115), resulting in a model that contained NGS at N40 and N228. This

86 70 structure was then submitted to in silico glycosylation using the GlyProt server, which produced a model containing complex N-linked glycans at all NGS, except N204 and N296 which are part of disordered regions (42). Complex glycans were then modeled onto the glycosylated structure in a predictive fashion with PyMol. Site-directed mutagenesis Primers were designed to mutate the asparagine residues of NGS and introduced into the EBOV GP or mutant GP expression vectors using the QuickChange Site-Direct Mutagenesis Kit (Stratagene, La Jolla, CA) according to the manufacturer s protocol. The 7GΔmuc mutant was made by addition of T42V and T206V to the 5GΔmuc mutant, which has been previously described (135). All mutations were confirmed by sequencing the full length of the EBOV GP ORF by Sanger sequencing. Transfections All transfections were performed in HEK293T cells seeded in a 6-well plate by polyethylenimine (PEI) method, as previously described (116, 135). Production of VSVΔG-GFP pseudovirions Pseudovirions were produced in HEK293T cells as previously described (135). Briefly, HEK293T cells were transfected with the various EBOV GP constructs and transduced by VSVΔG-GFP pseudotyped with LASV GPC the following day. After 24 h, cell supernatants were collected and filtered through 0.45 µm syringe-filters followed by storage at -80 C. When indicated, pseudovirions were pelleted through a 20% sucrose cushion for 2 h at 83,000 x g and resuspended in PBS.

87 71 Pseudovirion EBOV GP and VSV-matrix quantification Assessment of EBOV GP to VSV-matrix (M) ratios of pseudovirion preparations was performed as previously described (135). Briefly, pseudovirion-containing supernatants were passed through a dot blot apparatus onto nitrocellulose. EBOV GP was detected with anti-ebov GP human monoclonal antibody (mab) KZ52 (137) or mouse anti-ebov GP1 mab 5E6 (117) and VSV-M was detected with mouse anti-vsv-m mab 23H12 (138). Signals were quantified using the Odyssey Imaging Station and Image Studio software (LI-COR). Transduction assays Vero cells or HEK293T cells were seeded in 48-well plates 24 h prior to transduction. Pseudovirions were normalized to WT matrix expression prior to addition to Vero cell monolayers (WT MOI of ~0.1). When indicated pseudovirions were treated with 200 µg/ml thermolysin (THL; Sigma) for 1 h prior to addition to Vero cell monolayers. HEK293T cells were transduced at an MOI of ~0.01, in order to observe receptor-dependent enhancement of entry, 48 h post transfection. Transduction was determined by quantification of GFP expressing cells by flow cytometry h post addition of pseudovirions. Immunoblots Pelleted pseudovirions bearing HA-tagged GP or mutant GPs were treated with peptide-n-glycosidase F (PNGaseF) or Endo H (endoglycosidase H) to remove all N- linked glycans or only high mannose N-linked glycans, respectively, for 3 h. Reactions were separated on a 4-20% Mini-PROTEAN TGX gel (Bio-Rad) and transferred to

88 72 nitrocellulose. GP2-HA was detected with a polyclonal anti-ha antibody (Sigma) and signals were visualized with the Odyssey Imaging Station. HEK293T cells transfected with CLEC-myc expression vectors were harvest 48 h post transfection and lysed in 1% SDS, followed by brief sonication to promote lysis. Cell lysates were separated on a 4-20% Mini-PROTEAN TGX gel (Bio-Rad) and transferred to nitrocellulose. CLEC-myc fusion proteins and cellular myc (loading control) were detected with the 9E10 anti-myc mab. The 9E10 anti-myc mab, developed by J.M. Bishop was obtained from the Developmental Studies Hybridoma Bank, created by the NICHD of the NIH and maintained at The University of Iowa, Department of Biology, Iowa City, IA Results Mutation of GP2 N-linked glycosylation sites decreases protein expression To investigate the role of GP2 N-linked glycans during entry the asparagine residue, to which the glycan is added, was mutated to an aspartate to disrupt both NGS within GP2. These GP mutants were transfected into 293T cells and pseudotyped onto VSVΔG-GFP. A previous study found that both of these sites were glycosylated (136). To evaluate protein expression pseudovirion containing supernatants were passed through a dot blot apparatus and the ratio of VSV matrix to GP was determined within a single well using a mouse anti-vsv matrix monoclonal antibody (23H12) and a conformationspecific human anti-ebov GP monoclonal antibody (KZ52). Detection of mab binding was assessed using two different IR dyes conjugated to species-specific secondary antibodies (135).

89 73 The N563D mutation resulted in a significant decrease in KZ52 antibody binding compared to wild-type (Figure 3-2A). Disruption of the N618 site also resulted in decreased antibody binding, however this was more modest compared to the N563D mutation. Antibody binding to GP containing both mutations was indistinguishable from background (No GP; Figure 3-2A). Since part of the KZ52 epitope lies within GP2, these mutations may impact the ability of this antibody to bind (42). Therefore, we utilized a GP1-specific monoclonal antibody that targets the MLD, 5E6 (117), to detect protein levels. In contrast to the data obtained with KZ52, the N563D mutant GP was clearly recognized by mab 5E6, however this was significantly decreased compared to WT (Figure 3-2B). Interestingly, the N618D mutation had no effect on expression determined by mab 5E6 binding. Additionally, the N563D/N618D double mutant could also be detected by 5E6, but was significantly decreased compared to WT. Overall these results indicate that both GP2 N-linked glycosylation sites are important for promoting native conformation of this subunit, or are a part of the KZ52 epitope, and the N563 site is important for overall GP expression. Removal of GP2 N-linked glycans impacts GP-mediated transduction Since all the mutants generated were detected with the 5E6 antibody we assessed the ability for these proteins to mediate transduction of Vero cells. We have previously shown that entry into this cell type is dependent on the phosphatidylserine (PS) receptor TIM-1 (79), which enhances internalization of virus by a GP-independent mechanism through binding (PS) in the viral envelope (70, 80). Furthermore, we have found that removal of N-linked glycans from GP1 increased entry into Vero cells, which correlated

90 74 with an increase in protease sensitivity (135). Transduction mediated by the N563D mutant pseudotyped onto VSVΔG-GFP was significantly enhanced by 2-fold (Figure 3-2C), despite the decrease in expression and altered conformation as determined by binding of 5E6 and KZ52 antibodies, respectively (Figure 3-2A and 2B). This suggests that the N-linked glycan at N563 impedes one or more entry steps. Transduction mediated by VSVΔG-GFP pseudotyped with the N618D mutant was decreased by almost 2-fold, however this was not significant. Despite being detected by the 5E6 antibody the N563D/N618D mutant was unable to mediate transduction above background. Overall these data indicate that GP2 N-linked glycans have distinct roles during GP-mediated entry; the glycan at N563 decreases entry efficiency and the glycan at N618 is required for efficient entry. Removal of N-linked glycans decreases entry mediated by primed GP A critical step in entry of filoviruses is the removal of the glycan cap and MLD by endosomal cysteine proteases prior to binding the intracellular receptor, NPC1 (36, 59, 61, 63, 103). Interestingly, the bacterial metalloproteinase, thermolysin (THL), is able to mediate this priming event (36, 59, 64). THL treatment of virions pseudotyped with GP results in ~20-fold increase in transduction (59, 139). This allows for the assessment of GP2 N-linked glycans during entry in the absence of the glycan cap and MLD. We found that THL treatment of pseudovirions bearing the N563D mutant GP decreased transduction by ~75% compared to WT (Figure 3-3A). Transduction was still enhanced compared to mock treated virions, but only by 3-fold compared to the 21-fold observed with WT (Figure 3-3B). This result indicates that this glycans may stabilize the primed

91 75 form of GP or may be important for controlling coordinated post-internalization entry events. Transduction mediated by THL treated virions pseudotyped with the N618D mutant GP were ~50% decreased compared to WT, consistent with reduction in entry observed with untreated virions (Figure 3-2C). Furthermore, THL treatment enhanced entry mediated by the N618D mutation similar to WT (Figure 3-3B), indicating that this glycan is involved in a post-endosomal proteolysis event, possibly fusion. A single N-linked glycan on GP is sufficient for expression and function We have previously shown that removal of all N-linked glycans from the GP1 subunit does not affect protein expression and results in enhanced entry (135). Since the N618D mutation had minimal impact on expression we introduced this mutation into our GP1 N-linked glycan deficient construct, 7Gm8G, and assessed expression and entry. Introduction of the N563D mutation into constructs that have only 7 mutations in GP1 abolished expression and entry (data not shown). Addition of the N618D mutation to 7Gm8G led to a significant decrease in protein expression in virion-containing supernatants (Figure 3-4). However, virions pseudotyped with this mutant GP exhibited significantly enhanced entry compared to WT. Therefore, 16 out of 17 N-linked glycans can be removed from EBOV GP without sacrificing sufficient expression and efficient entry. GP2 N-linked glycans can mediate C-type lectindependent entry in absence of MLD Removal of the MLD from GP (GP1Δmuc) enhances protein expression and entry (41). We previously found that extensive deglycosylation of GP1Δmuc led to decreased

92 76 protein expression and entry compared to GP1Δmuc, but entry was comparable to GP (135). Removal of all N-linked glycans from GP1Δmuc (7GΔmuc) resulted in entry 50% decreased compared to GP (Figure 3-5A); thus, this protein lacks all N- and O-linked glycans on GP1 and can be used to investigate the role of GP2 N-linked glycans on CLEC-dependent entry. We fused a myc-tag to all CLECs that are known to enhance EBOV entry and overexpressed them in poorly permissive HEK293T cells (Figure 3-5B). The ectodomains of DC-SIGN and hmgl were fused to the transmembrane domain and cytoplasmic tail of L-SIGN in order to enhance cell surface expression ( L DC-SIGN and L hmgl, respectively). GP1Δmuc-mediated entry was enhanced ~10-25-fold in cells expressing CLECs (Figure 3-5C). This enhancement was abolished in cells expressing L- SIGN and LSECtin upon removal of all N-linked glycans from GP1Δmuc. However, cells expressing L DC-SIGN, ASGPRI, and L hmgl were still able to support entry mediated by 7GΔmuc above background, indicating that ligands for these proteins are present in N-linked glycans on GP2. Identification of N-linked glycan species on GP2 We utilized two glycosidases to determine the glycan species present on GP2. PNGaseF removes all N-linked glycans, while Endo H is only able to remove high mannose species. An HA tag was fused to the C-terminus of GP2 and GP2 mutants for detection via immunoblot. The migration patterns of PNGaseF and Endo H treated GP2 indicated that this subunit contains both high mannose and complex/hybrid N-linked glycans (Figure 3-6). Treatment of N563D with Endo H did not affect protein migration, indicating that the glycan attached to N618 is either a complex or hybrid species. Consistent with these observations and a previous study (136) treatment of N618D with

93 77 Endo H resulted in protein migration comparable to PNGaseF treated WT, confirming that there is a high mannose glycan at N563. The N-linked glycan at N618 is not required for efficient C-type lectin-dependent entry mediated by GP Our results with the 7GΔmuc contruct indicated that GP2 N-linked glycans are sufficient to enhance DC-SIGN-, ASGPRI-, and hmgl-dependent transduction. We therefore, performed similar CLEC-dependent enhancement of entry studies with our GP2 NGS mutants. Elimination of individual NGS within GP2 had no effect on CLECdependent entry, which is not surprising given the number of N-linked glycans attached to GP1 (Figure 3-7A-E). Consistent with our previous findings, removal of all N-linked glycans from GP1 decreased the enhancement of entry mediated by CLEC expression (135). Entry mediated by 7Gm8G was similar to background in L-SIGN and LSECtin expressing cells (Figure 3-7B and C), however entry into DC-SIGN expressing cells was still enhanced by ~2-fold (Figure 3-7A). As expected, addition of the N618D mutation to 7Gm8G did not further decrease DC-SIGN-dependent enhancement of entry since this mutant GP would still retain the high mannose glycan at N563 (Figure 3-6 and Figure 3-7A). Entry into ASGPRI and hmgl cells was only decreased ~50% upon removal of all N-glycans from GP1, suggesting that either O-linked glycans in the MLD or the complex/hybrid N-linked glycan at N618 is serving as a ligand for these receptors. Interestingly, removal of the N618 glycan from 7Gm8G had no impact on ASGPRI- and hmgl-dependent entry. These results indicate that the high mannose glycan at N563 is sufficient to mediate a low degree of DC-SIGN-dependent entry and the complex/hybrid glycan at N618 is not required for ASGPRI- and hmgl-dependent entry.

94 78 Discussion N-linked glycans on viral glycoproteins are known to serve a variety of functions, including CLEC binding, antibody evasion, and promotion of expression and stability (82). We have previously shown that relatively conserved N-linked glycans on EBOV GP1 serve several roles, including CLEC binding, limiting proteolysis, and antibody evasion (135). The GP2 subunit of filoviruses contains two completely conserved N- linked glycosylation sites that, to date, have not been investigated for their role in entry. We find that the glycan at N563 impedes entry while the glycan at N618 is important for efficient entry. Furthermore, 16 out of 17 N-linked glycans could be removed from GP without abolishing expression or entry. Next we found that the glycan at N563 is able to support minimal entry into DC-SIGN expressing cells, however the glycan at N618 had no detectable role in mediating entry into CLEC-expressing cells. Thus, there is likely minimal evolutionary selection to maintain these conserved glycans in GP2 as ligands for CLEC-dependent entry. Further work needs to be done to characterize the role of GP2 N- linked glycans on endosomal entry events and potentially antibody evasion. N-linked glycans in GP2 are attached to the HR1 and HR2 regions, which are involved in the formation of the six-helix bundle during fusion. Interestingly, removal of these glycans had opposite effects on entry. Mutation of the N563D NGS in HR1 enhanced entry by 2-fold, while the N618D mutation in HR2 decreased entry by 2-fold. Work with paramyxovirus fusion protein has indicated that glycans attached to HR regions are involved in controlling fusion (110, 113). Thus, it is possible that removal of the N563 glycan enhances fusogenticity of GP and removal of the N618 glycan decreases this event. The trigger for GP fusion has yet to be determined, which has limited studies

95 79 on this critical step during entry. Two hypotheses exist for the triggering of fusion, 1: GP requires additional proteolysis after initial protease-mediated removal of the glycan cap/mld (64) and 2: the endosomal/lysosomal reduction of GP disulfide bonds, potentially by a reductase, destabilizes the structure (100), resulting in insertion of the fusion loop into the host membrane. Our results showing that proteolysis partially inactivates the N563D and N618D mutants, suggests that these glycans are important for limiting initial proteolysis of GP to protect from triggering fusion until the virus is in close proximity to the lysosomal membrane to allow for insertion of the fusion loop. Alternatively, N-linked glycans within GP1 and the N563 glycan in GP2 mask a conserved disulfide bond between the internal fusion loop and HR1 (42). Thus, the glycan at N563 might impede reduction of this bond after proteolysis, thereby limiting fusogenicity. Given the location of N618 in HR2, a role for the attached glycan could be to coordinate proper formation of the six-helix bundle, which could explain why disruption of this NGS results in a decrease in entry efficiency. Future studies should focus on the potential roles of these glycans in mediating efficient fusion virus-host membranes. We have previously shown that N-linked glycans on GP1 are important for efficient utilization of CLECs to gain entry into the cell (135). Upon removal of the MLD from GP1, N-linked glycans on GP2 were able to support entry into cells expressing several CLECs utilized by EBOV for entry. High mannose N-linked glycans are known to interact with DC-SIGN and L-SIGN (87). Thus, it was surprising that removal of all N-linked glycans in GP1 abolished L-SIGN- but not DC-SIGN-dependent entry. The ability of DC-SIGN to enhance entry, albeit poorly, but not L-SIGN could be due to

96 80 differences in binding to both the high mannose structure in GP2 and fucose-containing O-linked glycans present in the MLD, thereby increasing avidity. In the presence of the MLD, the complex/hybrid glycan at N618 did not impact entry into ASGPRI and hmgl expressing cells. Since these receptors bind to terminal galactose-containing structures, which are present in O-linked glycans on the MLD, it is not surprising that removal of the glycan at N618 did not impact ASGPRI and hmgl-dependent entry. Overall, these studies indicate that the primary ligands that allow for CLEC-dependent entry are present in the GP1 subunit. Structural studies of EBOV GP and SUDV GP have highlighted a hot-spot for neutralizing antibody binding located at the base of the protein where GP1 and GP2 interact (42, 43). Interestingly, removal of the glycan at N563 completely abolished binding of the in vitro neutralizing antibody, KZ52. However, this mutation also affected overall protein expression, as determined by mab 5E6 binding to the GP1 MLD. Despite the observed decrease in expression the N563D mutation enhanced transduction in Vero cells. Therefore, this site is important for optimal expression but is not absolutely required. Incorporation of this glycan may have occurred due to selective pressures by antibody neutralization in the natural reservoir, which is likely to be bat populations (8, 140). Furthermore, glycosylation of the stem domain of the Nipah virus fusion protein has been shown to protect from antibody neutralization (113). Therefore, it is possible that the N618 glycan may also have a role in antibody evasion. Further work needs to be performed to clarify a role for these glycans in evasion from the host immune response.

97 81 A B N563 N618 EBOV RQLANETTQAL DWTKNITDKID SUDV RQLANETTQAL DWTKNITDKIN TAFV RQLANETTQAL DWTKNITDKID BDBV RQLANETTQAL DWTKNITDKID RESTV RQLANETTQAL DWTKNITDEIN MARV RRLANQTAKSL DLSKNITEQID LLOV RELANTTTKAL DWSANITAEIN Figure 3-1 Model of N-glycans at conserved sites in EBOV GP2. (A) Structural model of N-linked glycans on EBOV GP, shown from the side. The heptad repeat 2 of EBOV GP2 (PDB ID: 2EBO) was modeled at the base of the EBOV GP 1,2 Δmuc structure (PDB ID: 3CSY), and a short trimer of alpha helices was used to represent the transmembrane domain (shown in gray within the viral membrane). Complex N-linked glycans were added to each N-linked glycosylation site in silico with the GlyProt server. The mucinlike domain for each monomer is shown as a gray sphere in a predictive manner. GP1 is shown in teal, GP2 is shown in tan, GP1 N-linked glycans are shown in orange, and GP2 N-linked glycans are shown in purple and are labeled according to the asparagine residue position to which they are attached. (B) Alignment of the conserved N-linked glycosylation sites (N-X-S/T, where X is any amino acid except proline), highlighted in black, within GP2 of representative viruses in the filovirus family.

98 82 A B C Figure 3-2 Expression and entry of EBOV GP2 NGS mutants. (A and B) EBOV GP to VSV matrix ratios of the indicated pseudovirion preparations, determined by dot blot with anti-vsv matrix mab, 23H12 and (A) conformation-specific anti-ebov GP mab, KZ52, or (B) anti-ebov GP1 mab, 5E6. Data represent the average ± SD GP:M ratio of three independent stocks of virus, presented as a percentage of WT. (C) The indicated pseudovirions, normalized to VSV matrix expression, were used to transduce Vero cells. GFP positive cells were quantified the following day by flow cytometry. Data represent the average ± SEM of transduction observed with three independent stocks of virus, performed in triplicate, and is presented as a percentage of WT. No GP, represents pseudovirions produced with a viral glycoprotein. Significance was determined by onesample t test, * p < 0.05, * p < 0.005, and *** p <

99 83 A Transduction (GP = 100%) B Titer (TU/mL) Mock THL GP N563D N618D 10 4 GP N563D N618D THL Figure 3-3 Removal of GP2 N-linked glycans decreases entry of thermolysin treated GP. The indicated pseudovirions, normalized to VSV matrix expression, were treated with 200 µg/ml THL for 1 h at 37 C prior to addition to Vero cell monolayers. GFP positive cells were quantified the following day by flow cytometry. (A) Data represent the average ± SEM of transduction observed with two independent stocks of pseudovirions, performed in triplicate, presented as a percentage of WT. (B) The same data are shown as the transducing units/ml with the fold enhancement observed with THL treatment labeled for each virion.

100 Figure 3-4 The glycan at N563 is sufficient for GP-mediated entry and expression. Relative expression (white) was determined by calculating the ratio of EBOV GP, detected with mab KZ52, to VSV matrix, detected with mab 23H12, by quantitative dot blot of virion containing supernatants. Data represent the average ± SD of GP:M ratios for three independent stocks of pseudovirions. Vero cell transduction (black) was performed with pseudovirions normalized to matrix expression. The following day GFP was quantified by flow cytometry. Data represent the average ± SEM of transduction observed with three independent stocks of pseudovirions. All data are presented as a percentage of WT. Significance was determined by one-sample t test,* p < 0.05 and *** p <

101 85 A B C Figure 3-5 GP2 contains ligands for CLECs. (A) Immunoblot of transfected HEK293T cell lysates. CLEC-myc proteins and cellular myc (*; loading control) were detected with the 9E10 anti-myc mab. (B) Enhancement of entry observed with CLEC-myc expressing HEK293T cells. The indicated pseudovirions were used to transduce empty plasmid or CLEC-myc transfected HEK293T cells expressing myc-tagged. Data are presented as the fold increase in entry over empty plasmid transfected cells (gray dashed line) and represents the average ± SEM of three independent experiments. Significance between GP1Δmuc and 7GΔmuc for each CLEC was determined by Student s t test, * p < 0.05.

102 Figure 3-6 Identification of the N-linked glycan species on EBOV GP2. (A) Immunoblot of glycosidase treated GP2 NGS mutants. Pelleted pseudovirions bearing the indicated GP, with a C-terminal HA tag, were treated for 3 h with Endo H or PNGaseF, E and P respectively. GP2-HA was detected with a polyclonal anti-ha antibody and VSV-M was detected with a monoclonal anti-vsv M antibody, 23H12. 86

103 87 A D Fold change in entry (over Empty HEK293T B E C Figure 3-7 Characterization of receptor utilization by GP2 NGS mutants. (A-E) HEK293T cells were transfected with the myc-tagged CLEC constructs and transduced with the indicated pseudovirions. Data are presented as the fold increase in transduction over empty plasmid transfected cells (gray dashed line) and represents the average ± SD of one of three experiments, each performed in duplicate. Note differences in y-axis between panels.

104 88 CHAPTER IV A SINGLE RESIDUE IN THE GLYCOPROTEIN OF EBOLAVIRUSES ALTERS ENDOSOMAL EVENTS REQUIRED FOR ENTRY Introduction Recent advances in the filovirus field have uncovered a complex entry pathway. A macropinocytosis-like internalization process is preceded by initial cell surface interactions with phosphatidylserine (PtdSer) receptors (TIM family proteins and Gas6/TAM family kinases), which interact with PtdSer in the outer leaflet of the virion membrane (70, 73, 79, 80, 141) and/or C-type lectins (54-57), which bind N- and O- linked glycans attached to the viral envelope glycoprotein (GP). In the late endosome/lysosome, GP is primed for fusion through proteolysis mediated by cysteine proteases (59, 61). This event is characterized by the removal of two heavily glycosylated regions, the glycan cap and mucin-like domain (MLD), resulting in the exposure of the receptor-binding domain (RBD) (42, 59-62). Following endosomal proteolysis, the RBD of GP1 interacts with the second luminal loop (C loop) of Niemann-Pick C1 (NPC1), a multipass transmembrane protein of the late endosome/lysosome involved in cholesterol transport (36, 63). We recently showed that removal of N-linked glycosylation from the glycan cap and MLD of GP1 is insufficient for RBD binding to NPC1 and that these two domains must be proteolytically removed (135). The role of NPC1 in facilitating entry of filoviruses has not been defined, however previous work has suggested that further proteolysis or mild reduction are required post NPC1-binding to destabilize GP1/GP2 interactions leading to virus-host membrane fusion mediated by the GP2 subunit (59, 64,

105 89 100). Thus, a potential role for NPC1 is to traffic virus to a pro-fusion compartment. Alternatively, NPC1 binding of the viral GP may be directly involved in viral/cell membrane fusion. Ebolaviruses and marburgviruses make up the family Filoviridae. All of the marburgviruses belong to the Marburg marburgvirus species. There are five species of ebolaviruses; the viruses within each species are named after the geographical location of the initial outbreak: Ebola virus (EBOV), Sudan virus (SUDV), Bundibugyo virus (BDBV), Taï Forest virus (TAFV), and Reston virus (RESTV) (23). All filoviruses require endosomal processing by cysteine proteases for infection. Initial reports identified that cathepsins L and B (CatL and CatB, respectively) mediate cleavage of EBOV GP1 to 18 kda and 17 kda forms, respectively, with CatB being essential for entry (59, 61, 64). However, recently there have been conflicting results published regarding the role of CatB in entry mediated by the different ebolavirus species (99, 142, 143). Misasi et al. showed that loss of CatB activity resulted in a log decrease in VSV pseudovirus entry mediated by GP of the Boniface strain of SUDV (SUDV GP) (99). However, Marzi et al. utilized infectious SUDV and found that entry was completely independent of CatB activity (142). We used sequence comparisons to identify GP residues that might be responsible for potential differences in the previous results obtained with EBOV GP and SUDV GP. We found that some SUDV isolates encode a glutamine at residue position 95, whereas others encode a lysine. Further, a lysine is found at position 95 in all other ebolavirus species. A previous study has shown that mutation of K95 in EBOV GP resulted in decreased entry efficiency (38). Furthermore, a structural study indicated that the residue

106 90 at this position might be involved in promoting protein stability or triggering of fusion (44). Mutagenesis of residue Q95 in SUDV GP to a lysine dramatically increased SUDV GP-mediated entry and decreased sensitivity to proteolysis, whereas mutation of K95 in EBOV/BDBV GP to a glutamine imparted enhanced protease sensitivity. Furthermore, mutations at position 95 within EBOV GP, SUDV GP, and BDBV GP significantly altered endosomal protease requirements. Mutation of this residue in EBOV GP and BDBV GP resulted in a decrease in sensitivity to the NPC1 inhibitors, U18666A and 3.47, but did not result in entry independent of NPC1. Interestingly, mutations at position 95 in EBOV GP and BDBV GP, but not SUDV GP, also decreased sensitivity to calcium channel inhibition. Overall, we have identified a single amino acid within multiple ebolavirus species that affects endosomal processes and sensitivity to inhibition of entry factors. Materials and Methods Cells lines and plasmids HEK 293T cells and Vero cells were maintained in DMEM + 10% FBS + 1% penicillin/streptomycin. NPC1 null CHO cells (CT43 cells) were maintained in DMEM/F-12 (50:50 mixture) + 10% FBS + 1% penicillin/streptomycin (144). The EBOV GP (Accession: NP_066246) and EBOV GPΔmuc (deletion of amino acids composing the MLD) in pcdna3.1 (Invitrogen, Carlsbad, CA) have been previously described (39, 41). A codon optimized SUDV-Boniface GP (Accession: Q66814) gene was cloned into pcdna3.1 and amino acids were deleted to create SUDV GPΔmuc. Both EBOV GPΔmuc and SUDV GPΔmuc were also cloned upstream of a HA-tag (amino acids: YPYDVPDYA) within pcdna3.1. The BDBV GP (Accession:

107 91 YP_ ) gene was amplified from TRIzol-extracted viral RNA and PCR amplified and cloned into pcdna3.1. The pcaggs vector was used to express a codon optimized LASV-Josiah GPC (Accession: NP_694870). Site-directed mutagenesis Mutations were introduced using the QuikChange site-directed mutagenesis kit (Stratagene, La Jolla, CA) and verified by sequencing the length of the GP gene. Primer sequences will be made available upon request. Transfections HEK 293T cells were transfected by a modified polyethylenimine (PEI; MW ~25,000, Polysciences Inc., Warrington, PA) protocol (116, 135). Briefly, 2 µg of plasmid DNA was incubated with 6 µl of PEI in a total volume of 100 µl of 150mM NaCl for 10 min prior to the addition to adherent cells in a 6-well dish (80% confluent). Pseudovirion production Production of VSVΔG-eGFP pseudovirions was performed as previously described (145). Briefly, HEK 293T cells were transfected with glycoprotein expression vectors. At 24 h post transfection, cells were transduced at an MOI of 3 with VSVΔG pseudotyped with LASV GPC for 1 h followed by extensive washing. Cell supernates containing newly pseudotyped VSVΔG were collected 24 h post transduction, filtered through 0.45 µm syringe filters, and stored at -80 C. Where indicated, virions were purified by pelleting through a 20% sucrose cushion at 83,000 x g and 4 C for 2 h, resuspended in PBS, and stored at -80 C.

108 92 Transduction assays Pseudovirions were added to monolayers of Vero cells and CT43 cells +/- NPC1- myc in 48-well dishes. After 16 h, cells were lifted with Accutase (Sigma-Aldrich, St. Louis, MO) and the percentage of GFP positive cells was detected by flow cytometric analysis. Titers (transducing units/ml) were determined by calculating the amount of virus that resulted in 5 to 25% GFP positivity which fell in the linear range of transduction. Background levels of transduction were determined by the GFP positivity obtained with VSVΔG produced in cells not expressing a glycoprotein. Thermolysin digest reactions Pseudovirion stocks were incubated with indicated concentrations of thermolysin (THL; Sigma-Aldrich) at 37 C for 15 min or 1 h followed by incubation on ice and the addition of phosphoramidon (50 µm; Peptides International, Louisville, KY), to stop the reaction. Immunoblots Purified, concentrated pseudovirion preparations were separated on Mini- PROTEAN TGX gels (4-20%; Bio-Rad, Hercules, CA) and transferred to nitrocellulose. GP1 of the various ebolavirus species was detected with a rabbit polyclonal Ab developed against EBOV GP1 amino acids (TKRWGFRSGVPPKV). In some experiments, EBOV GP1 and SUDV GP1 were detected by the MLD specific monoclonal antibodies 5E6 (117) and 17F6 (a gift from John R. Dye), respectively. The GP2-HA was detected with rabbit anti-ha polyclonal Ab (Sigma-Aldrich), and the 23H12 mab was used to detect VSV-matrix (VSV-M). Secondary antibodies conjugated

109 93 to IRDyes (LI-COR) were used to visualize and quantify the band intensities with an Odyssey Imaging Station and Image Studio software (LI-COR). Entry factor inhibition assays Vero cells were treated for 4 h with the pan cysteine protease inhibitor E-64 (300 µm; Peptides International), 2 h with the CatB-specific inhibitor, CA-074 (80 µm; Peptides International), or equal volumes of DMSO prior to addition of pseudovirus. At h post transduction, cells were lifted and GFP positive cells were detected by flow cytometric analysis. CatB activity was determined by incubation of cell lysates from DMSO or 80 µm CA-074-treated Vero cells with Z-Arg-Arg-MCA (Peptides international) for 30 min at room temperature, as previously described (119). Fluorescence was measured by excitation at 380 nm and fluorescence emission at 460 nm. Inhibition of NPC1 in Vero cells was carried out by addition of U18666A (20 µm; Sigma Aldrich) added at the time of transduction or by a 90 min pre-incubation of the cells with compound 3.47 (2 µm). Voltage-operated calcium channels were inhibited by pretreatment of Vero cells for 1 h with several dilutions of verapamil (Sigma Aldrich). Entry in the presence of inhibitors was compared to entry in cells treated with an equal volume of DMSO (NPC1 inhibitors) or H 2 O (verapamil). Ammonium chloride (NH 4 Cl) was added to pseudovirus-containing culture media at a final concentration of 20mM to block endosomal acidification every 2 h during transduction. Soluble NPC1 C loop GP binding assay The soluble NPC1 C loop construct has been previously described (36, 102). Briefly, the construct contained a preprotrypsin signal sequence, a FLAG tag

110 94 (DYKDDDDK), and a 6xHis tag upstream of sequences that form an antiparallel coiled coil flanking the NPC1 C loop (amino acids ). Purification of soluble NPC1 C loop was performed as previously described (135). Briefly, soluble NPC1 C loop from transfected HEK 293T cells was collected from the supernatant and purified with HisPur Ni-NTA resin (Thermo Scientific, Rockford, IL). Elutions were exchanged into PBS with Zeba Spin desalting columns (Thermo Scientific). Detection of soluble NPC1 C loop binding to GP was performed as previously described (36, 135). Mock or thermolysin (THL)-treated EBOV GP and SUDV GP pseudovirions were captured on ELISA plates by anti-ebov GP2 mab 2G4 (a gift from Qui Xianggou) and anti-sudv GP mab 16F6 (a gift from John R. Dye), respectively. After blocking the wells in PBS + 3% BSA, serial dilutions of soluble C loop were incubated overnight. After extensive washing, GP-bound C loop was detected with an anti-dykddddk antibody conjugated to HRP (GenScript, Piscataway, NJ) and developed with UltraTMB substrate (Thermo Scientific). Generation of CT43-NPC1-myc stable cell line The NPC1 ORF was PCR amplified from a plasmid obtained from OriGene (Rockville, MD) and inserted upstream of a c-myc tag (amino acids: EQKLISEEDL) in pcdna3.1 (Invitrogen). This sequence was subcloned into the pbabe-puro retroviral vector. The resulting plasmid was packaged into MuLV and pseudotyped with VSV G. MuLV pseudovirions were used to transduce CT43 cells, as previously described (36). Transduced cells were selected for with puromycin (10 µg/ml). Resistant cells were used for transduction assays, as described above.

111 95 Results SUDV-Bon GP mediates significantly lower transduction than EBOV GP To assess if transduction efficiency of Vero cells by EBOV GP and SUDV GP differed, we pseudotyped these ebolavirus glycoproteins onto non-infectious vesicular stomatitis virus that expressed egfp in place of the native G glycoprotein (VSVΔGeGFP). We consistently observed a log lower titer in virions bearing the SUDV GP (Figure 4-1A). Since the MLD (amino acids ) contains the greatest sequence diversity, with only 17.8% identity, we deleted this domain to determine if differences in this region were responsible for the difference in titer (Figure 4-1B). Consistent with previous work with EBOV (41), removal of the MLD resulted in a 20 to 30-fold increase in entry mediated by both EBOV GPΔmuc and SUDV GPΔmuc (Figure 4-1A). However, SUDV GPΔmuc titers remained a log lower than EBOV GPΔmuc, indicating that sequence variation in the MLD was not responsible for the decreased titer. To evaluate if titer differences were due to differences in GP expression, we analyzed both the expression and virion incorporation of EBOV GPΔmuc and SUDV GPΔmuc. A C-terminal HA-tag was added to EBOV GPΔmuc and SUDV GPΔmuc to allow for quantitative assessment of the same epitope. These plasmids were transfected into HEK 293T cells and transduced 24 h later with VSVΔG-eGFP to produce pseudovirions bearing HA tagged GP. Pelleted virions from the cell supernatants were assessed for the ratio of GP to VSV matrix expression by immunoblot. Analysis of pelleted virions indicated that incorporation of SUDV GPΔmuc into pseudovirions was

112 96 similar to that of EBOV GPΔmuc (Figure 4-1C). These results indicated that GP expression was not a significant factor for the differences in the observed titers. Sequence and structural comparison of GP Alignment of the RBD of filovirus isolates indicated that there was a polymorphism at position 95 (Figure 4-2A), which has previously been shown to be important for entry (38, 60). Two of the first identified SUDV isolates, including the Boniface sequence that we were studying, contained a glutamine at residue 95, whereas all other filoviruses contain a lysine at that position. Structures have been solved for GP from a single strain of EBOV GP and two strains of SUDV, including GP from a SUDV-Boniface isolate that contains a lysine at position 95 (42-44). These structures reveal that K95 interacts with the carboxyl group of the T576 main-chain in the unstructured hinge region in heptad repeat 1 (HR1) via hydrogen bond (44). This indicated that a lysine at position 95 might stabilize the interactions between GP1 and an unstructured region of GP2 that is involved in formation of the six-helix bundle during fusion. However, in silico substitution of K95 with a glutamine indicated that the hydrogen bond switches from T576 to the main-chain carboxyl group of L573, which resides in the structured alpha helix region of HR1 (Figure 4-2B). The loss of the hydrogen bond to the unstructured region may result in destabilization of HR1 in GP2, which may account for the decreased titer observed with the SUDV GP containing a glutamine at position 95.

113 97 SUDV GP Q95K mutation increases pseudovirion transduction To test the hypothesis that the residue at position 95 in SUDV GP was responsible for the poor viral titer a Q95K substitution was made. Titers observed with VSVΔG pseudotyped with SUDV Q95K GP increased by an order of magnitude over SUDV GP and were indistinguishable from those obtained with EBOV GP (Figure 4-3A). Interestingly, a K95Q substitution in EBOV GP or BDBV GP did not result in a decrease in titer (Figure 4-3B and C). However, pseudovirions bearing BDBV GP had a ~log decrease in titer compared to EBOV GP and the BDBV GP K95Q mutation enhanced entry 2-fold compared to BDBV GP (Figure 4-3C). Immunoblot analysis with the EBOV GP MLD-specific monoclonal antibody, 5E6 (a gift from Qui Xiangguo), and the SUDV GP MLD-specific monoclonal antibody, 17F6 (a gift from John R. Dye), indicated that expression of the respective mutants were similar to (EBOV) or decreased (SUDV) compared to WT, therefore differences in expression could not explain the differences in pseudovirion titer (Figure 4-3D). These results suggest that a lysine at position 95 in SUDV GP increases the efficiency of one or more steps in the entry pathway, but a glutamine at position 95 in EBOV GP or BDBV GP resulted in no impact on entry or a modest increase in entry, respectively. The residue at position 95 is a determinant of CatBdependence Previous work indicated that transduction mediated by EBOV GP and SUDV GP had different CatB activity requirements (99). To compare the CatB dependence of our mutants to WT GPs, we utilized the CatB specific inhibitor, CA-074, that is known to

114 98 block entry by several species of ebolavirus (59, 61, 64, 99, 142, 143). Pretreatment of cells with CA-074 for 2 h resulted in a dramatic decrease in CatB activity (Figure 4-4A). Consistent with previous studies (59, 61, 64, 99, 142, 143), entry mediated by EBOV GP and BDBV GP was dependent on CatB activity (Figure 4-4B and C). Consistent with studies performed by Marzi at al., our results indicate that SUDV GP-mediated entry was not blocked by inhibition of CatB (142). The SUDV Q95K GP mutant had significantly increased sensitivity to the CatB inhibitor compared to WT, although this effect was not complete, whereas changing residue K95 of EBOV GP and BDBV GP to glutamine abrogated the need for CatB activity (Figure 4-4B and C). Nonetheless, entry mediated by all pseudovirions tested was blocked by the pan cysteine protease inhibitor, E-64, as others have shown (Figure 4-4B and C) (59, 61, 99). Together these results indicated that the amino acid that occupies position 95 significantly alters CatB-dependence of GPmediated transduction for multiple species of ebolavirus. A glutamine at position 95 increases protease sensitivity of GP During filovirus entry, both the glycan cap and mucin-like domain are digested by endosomal proteases that result in exposure of the RBD and subsequent binding to NPC1 (36, 63). Since the residue at position 95 lies at the base of the RBD, we determined whether lysine versus glutamine at this position would impact entry mediated by the processed form of GP. Previous work has established that treating EBOV GP with thermolysin (THL), results in a processed form of the protein (GP THL ) that mimics the product of CatB digestion (59, 61, 64). An earlier study suggested that, in some instances, loss of CatB dependence of a filovirus GP resulted in greater sensitivity to protease

115 99 treatment (64). Thus, we sought to determine if the GPs that had a glutamine at position 95 and had reduced CatB-dependence were more sensitive to THL proteolysis. Consistent with previous reports (59), we observed ~30-fold increase in entry when pseudovirions bearing the EBOV GP were treated with 200 µg/ml of THL for 1 h (Figure 4-5A). Surprisingly, when VSVΔG-SUDV GP virions were treated with THL, titers decreased to near background levels. However, upon mutation of Q95 to match the residue within EBOV GP, we observed a rescue in the ability of THL treated pseudovirions to mediate entry to similar levels as EBOV GP THL. Extrapolating from our SUDV GP findings, we predicted that THL treatment of EBOV GP K95Q pseudovirions would reduce transduction efficiency. Surprisingly, these protease treated pseudovirions had only 2.5- fold lower titers than untreated virus (Figure 4-5B). However, the titer of VSVΔG-EBOV K95Q GP THL was 2.5 orders of magnitude higher than that of SUDV GP THL, but was decreased ~40-fold compared to the titer obtained with EBOV GP THL. Treatment of pseudovirions bearing BDBV GP with THL enhanced titers by ~30-fold, similar to the effect observed with EBOV GP (Figure 4-5C). However, unlike results obtained with EBOV K95Q GP and more similar to SUDV GP, occupation of position 95 by a glutamine in BDBV GP decreased entry to near background levels when virions were treated with THL. Given the possibility that the residue at position 95 is involved in maintaining GP stability during proteolysis we tested whether a gentler proteolysis would allow for enhancement of transduction for proteins containing a glutamine at position 95. Pseudovirions bearing the various WT and mutant GPs were treated with either 20 µg/ml or 200 µg/ml THL for 15 min at 37 C, rather than 1 h. Incubation with either

116 100 concentration of THL resulted in the typical ~20 fold increase in EBOV GP-mediated entry (Figure 4-5D). Enhancement of entry could be observed for all GPs containing a glutamine at position 95 when treated with 20 µg/ml THL, ranging from ~10-30-fold over mock treated; however, incubation of virions, bearing GP with a glutamine at position 95, with 200 µg/ml THL resulted in significant decreases in enhancement of entry. Surprisingly, the fold enhancement of entry for pseudovirions bearing BDBV GP, with either concentration of THL, was greater than that observed with EBOV GP. The 200 µg/ml THL treatment enhanced entry almost 70-fold, indicating that BDBV GP is sensitive to overdigestion with THL. In order to detect the THL-processed form of GP we utilized a polyclonal Ab raised against amino acids of EBOV GP1. This antibody was able to detect GP in all ebolaviruses due to the high conservation of this epitope (Figure 4-3C and D). However, a glutamine at residue 95 reduced the signal observed on immunoblots. Despite the clear enhancement of entry observed with brief THL treatments, we were unable to detect cleavage products of GPs that contained a glutamine at position 95, but these products were apparent in GPs with a lysine at this position (GP THL Figure 4-5E to G). Due to a high degree of background signals it was difficult to determine if intermediate cleavage products were present for Q95 containing GPs, however there were intermediate forms clearly present for K95 GPs with 20 µg/ml THL treatment. Overall these results indicate that the residue that occupies position 95 has a role for stabilizing the post-catb cleavage conformation and/or alters GP conformation resulting in hypersensitivity to proteases.

117 101 Previous work has shown that cysteine proteases are required for EBOV entry post CatB-cleavage (59, 61, 64). Since we have shown that entry mediated by EBOV K95Q GP is independent of CatB activity and provides detectable levels of entry post THL-cleavage, despite no detectable cleavage product, we hypothesized that this mutation would be less sensitive to cysteine proteases after GP priming. Pretreatment of Vero cells with E-64 decreased entry mediated by both EBOV GP THL and SUDV Q95K GP THL by over 2-logs and as expected had no effect on LASV GPC dependent entry (Figure 4-6). However, EBOV K95Q GP THL mediated entry was significantly less sensitive to E-64 (>1-log) compared to WT. EBOV/BDBV K95Q GP mediated entry is less sensitive to inhibitors of NPC1 Since we were unable to detect the primed EBOV K95Q GP, we used a binding assay to determine if there was a cleavage product that could bind NPC1. Since EBOV GP interacts with C loop of NPC1, we utilized a soluble C loop binding assay to detect GP THL /C loop interactions (36). As expected, EBOV GP was unable to bind soluble C loop without proteolytic removal of the glycan cap and MLD (Figure 4-7A). Consistent with transduction and immunoblot data, SUDV Q95K GP THL bound soluble C loop in a similar manner to EBOV GP THL. However, both SUDV GP and EBOV K95Q GP were unable to interact with soluble C loop, despite being present at five times the amount of EBOV GP and SUDV Q95K GP. Furthermore, we were unable to detect C loop binding to GPs, except SUDV Q95K GP, treated with 20 µg/ml THL for 15 min (data not shown), consistent with immunoblot data that showed poor or no detection of GP cleavage products.

118 102 Since EBOV K95Q GP THL was able to mediate entry despite any detectable cleavage product, we determined if this mutation could lead to entry independent of NPC1. Previous work has shown that treatment of cells with the NPC1 inhibitor, U18666A, blocks EBOV GP-dependent entry (36, 63, 103). While treatment of Vero cells with U18666A reduced entry mediated by all WT ebolavirus glycoproteins, EBOV K95Q GP- and BDBV K95Q GP-mediated entry was significantly less sensitive to inhibition of NPC1 by U18666A compared to WT (Figure 4-7B and C). In contrast to these results, WT SUDV GP-mediated entry was abolished by treatment with U18666A and mutation of the glutamine at position 95 in this protein to a lysine significantly decreased sensitivity to this drug, however this effect was not to the degree observed with EBOV GP and BDBV GP mutants (Figure 4-7B). Since the mechanism of U18666A inhibition of NPC1 is unknown (146), we utilized an inhibitor that has been shown to block EBOV GP/NPC1 interaction, compound 3.47 (63). Surprisingly, the effect of this drug on entry was species specific. Entry mediated by EBOV GP and BDBV GP was more sensitive to inhibition compared to SUDV GP, however BDBV GP-dependent entry was modestly less sensitive than that of EBOV GP (Figure 4-7D and E). Consistent with the results obtained with U18666A, mutations at position 95 of WT GPs decreased the inhibition of entry (Figure 4-7D and E). Entry mediated by EBOV K95Q GP and BDBV K95Q GP was inhibited to a similar extent as SUDV GP; however, contradictory to the EBOV GP results, SUDV Q95K GP was completely resistant to compound 3.47 inhibition (Figure 4-7D). Interestingly, compound 3.47 was unable to inhibit NPC1 C loop binding to both EBOV GP and SUDV

119 103 Q95K GP (Figure 4-7F). Overall, the residue at position 95 is a species-specific determinant for NPC1 inhibition of entry (Figure 4-7E). Given the results obtained with NPC1 inhibitors we sought to determine if mutant GPs could mediate entry into NPC1 null CHO cells (CT43 cells) (144). These cells were unable to support entry of all ebolavirus GP pseudovirions tested and transduction could be rescued by the expression of NPC1-myc (Figure 4-7G). These results indicated that NPC1 is essential for entry mediated by GP of ebolaviruses, but mutations in WT GPs at position 95, regardless of a glutamine or lysine occupying this position, are less sensitive to U18666A- and compound 3.47-inhibition pathways that are normally required for entry. Overall these drug inhibition studies indicate that NPC1 has multiple functions during entry, as previously suggested (147), and mutations at position 95 eliminates the requirement for one or more of these functions; however, entry remains dependent on NPC1 expression. Mutation of residue 95 in EBOV GP and BDBV GP decreases sensitivity to inhibition of voltage-operated calcium channels Ca 2+ is known to be an important signal for cellular trafficking (148). Inhibition of calcium efflux has been shown to block infection of several viruses, including coxsackievirus B3 and Junín virus ( ). Recently, it has been shown that verapamil, which inhibits voltage-operated calcium channels, is a potent inhibitor of EBOV entry (152). Given that mutation of residue 95 in GP of multiple species of ebolavirus alters sensitivity to inhibition of NPC1, which may be involved in intracellular trafficking of virus, we tested whether these mutations altered inhibition of entry by verapamil. We

120 104 found that entry of pseudovirions bearing all WT, mutant GPs, and LASV GPC was inhibited in a dose-dependent manner by pretreatment of Vero cells with verapamil (Figure 4-8). The highest concentration of drug (100 µm) completely abolished entry mediated by ebolavirus GPs but not LASV GPC. However, reducing the concentration of verapamil resulted in significant decreases in sensitivity for EBOV GP and BDBV GP mutants, with the lowest concentration tested (25 µm) having minimal impact of mutant EBOV and BDBV GP-dependent entry. In contrast, SUDV GP and SUDV Q95K GP had similar dose-response curves to verapamil-dependent inhibition. These results support our previous findings with other filovirus entry factor inhibitors, indicating that there are differences in entry processes between ebolavirus species and that the residue occupying position 95 is a critical factor in determining sensitivity to inhibition of these processes. Mutation of residue 95 does not impact entry kinetics Given our results showing that mutation of the residue at position 95 alters protease sensitivity of GP and sensitivity of GP-mediated transduction to entry factor inhibition, we wanted to determine if these mutations altered overall entry kinetics. We performed a time of addition assay with NH 4 Cl, which inhibits filovirus GP mediated entry by blocking endosomal acidification (153). We observed similar entry kinetics over 8 h for all pseudovirions tested (Figure 4-9). These results indicate that sensitivity to proteolysis and resistance to inhibition of entry factors does not increase the rate at which virions move through the entry pathway, but likely increases the efficiency of entry. Discussion Filovirus entry is a complex process that is initiated by virion-cell surface interactions, leading to virus internalization into endosomes. Once in the endosomal compartment, GP1 of the viral glycoprotein is processed by proteases. This results in a

121 105 small molecular weight form of GP that binds to its intracellular receptor, NPC1, prior to virus-host membrane fusion. We found that the residue that occupies position 95 in GP1 has a dramatic effect on entry through altering endosomal events. Our results highlight the diversity of the entry process between ebolavirus species and give insights into the complicated pathway required for efficient entry by these viruses. Furthermore, these results highlight the importance of noting the accession numbers for filovirus GP clones used in assays, since single amino acid differences have distinct differences in sensitivity to inhibition of entry factors. Previous reports have shown that the late endosomal/lysosomal protein NPC1 is critical for filovirus entry by binding to the RBD of endosomally processed GP, however the exact role(s) of this protein during entry has yet to be determined (36, 63, 103). We have shown that the K95Q mutation in both EBOV GP and BDBV GP can bypass a U18666A sensitive-npc1 step required for WT entry. Despite containing a glutamine at position 95, SUDV GP-mediated entry is completely sensitive to this drug and, contradictory to WT EBOV GP and BDBV GP results, mutation to a lysine significantly decreased inhibition, however this was not as dramatic as the resistance observed with EBOV GP and BDBV GP mutants. The exact mechanism of NPC1-inhibition by U18666A is unknown (146). It has been shown that this drug results in cholesterol accumulation in the lysosome by blocking NPC1-mediated cholesterol trafficking out of this compartment (154). However, ebolavirus entry has been shown to proceed independent of cholesterol trafficking (36), thus the accumulation of cholesterol does not appear to impact EBOV GP-mediated entry. Additionally, U18666A does not block interactions between processed GP and NPC1 (63). Thus, one possibility is that U18666A

122 106 inhibits NPC1-dependent trafficking of virus to a pro-fusion compartment; and for EBOV GP and BDBV GP, but not SUDV GP, a glutamine at position 95 alleviates the dependence on this step for entry but still requires binding to NPC1, possibly to induce a conformational change required to help in triggering of fusion. This suggests that there are other differences in the entry pathway between these three species that are independent of the residue occupying position 95. Future studies with these mutations with pseudovirions and recombinant filoviruses would help to understand the mechanism of U18666A inhibition of NPC1 and the role of NPC1 during filovirus entry. Binding of processed EBOV GP and NPC1 is blocked by compound 3.47, which results in decreased entry (63). Surprisingly, we found that this drug was less efficient at inhibiting SUDV GP and, to a lesser extent, BDBV GP-mediated entry. Mutation of EBOV GP or BDBV GP at position 95, to match the residue in SUDV GP, resulted in decreased sensitivity that was comparable to results with SUDV GP. Contrary to our hypothesis, mutation of SUDV GP, to match the residue at position 95 in EBOV/BDBV GP, resulted in complete resistance to this drug. Furthermore, binding of NPC1 C loop to EBOV GP and SUDV K95Q GP was not inhibited by compound The previous study utilized full length NPC1 for binding assays (63), therefore it is possible that this drug may bind to a region of NPC1, distinct from the C loop, that impacts the structure of the protein resulting in decreased capacity to bind processed EBOV GP. A previous study showed that treatment of cells expressing non-functional NPC1 with verapamil enhanced cholesterol accumulation (155). We found that verapamil was a potent inhibitor of entry mediated by EBOV/SUDV/BDBV GP. Interestingly, mutations at position 95 in EBOV GP and BDBV GP, but not SUDV GP, resulted in significant

123 107 resistance to this drug. Thus, treatment with verapamil may affect a similar NPC1- dependent process as U18666A and compound 3.47, however these effects on entry are distinct from those observed with the NPC1 inhibitors as evidenced by SUDV GP/ SUDV Q95K GP-mediated entry. One possibility is that there are Ca 2+ -dependent proteases, which may be inhibited by verapamil, that are required for destabilizing GP resulting in fusion. Given our results indicating that Q95 containing EBOV/BDBV GP are more sensitive to proteolysis, it is would make sense that these proteins are less sensitive to verapamil. However, results with both WT and mutant SUDV GPs, would suggest that the overall structure of these proteins is less permissive for Ca 2+ -dependent proteolysis. Another possibility is that there are additional, non-protease Ca 2+ -dependent proteins associated with NPC1 that are required for GP-mediated entry and that U18666A and compound 3.47 inhibit these interactions. Mutations in GP, potentially at position 95, may alleviate the need for these additional host factor proteins in order to mediate entry. One role for these unknown factors could be to mediate trafficking of virus to a profusion compartment, which may not be required for viruses that have mutations at position 95. In addition to providing insights into the potential role of NPC1 during entry, residue occupancy of position 95 had a significant impact on endosomal protease dependence. In all species tested, a glutamine at position 95 in GP resulted in complete independence of CatB activity for entry. However, when this position is occupied by a lysine, only EBOV GP and BDBV GP were completely dependent on CatB activity for entry. Interestingly, MARV GP mediated entry has been shown to be independent of CatB activity (99), despite containing a lysine at position 79 (corresponds to 95 in EBOV

124 108 GP; Figure 4-2A). The Q95K mutation in SUDV GP resulted in a significant increase on CatB activity compared to WT, however it was not complete. This suggests that there are other amino acid determinants in SUDV GP that impact protease dependence. We also observed a log decrease in sensitivity to the pan cysteine protease inhibitor, E-64, during entry mediated by the THL cleaved EBOV K95Q GP; however, entry was still inhibited by over a log. This result suggests that, in the context of EBOV GP, a glutamine at position 95 may increase the sensitivity to cysteine proteases required for further proteolysis to allow for fusion. Alternatively, overdigestion with THL may lead to a decrease in GP substrate, thus requiring a lower degree of cysteine protease activity to allow for efficient entry. Together these observations further highlight innate differences between the glycoproteins of the different filovirus species. Future work could utilize these mutations to determine other proteases that are required during entry for proteins that do not require CatB activity. Our results regarding the entry of THL cleaved GP support the previous hypothesis that the residue at position 95 plays a role in stabilizing the pre-fusion complex and/or triggering fusion (44). We found that entry by THL cleaved GP was significantly impacted for SUDV GP and BDBV GP when position 95 was occupied by a glutamine. However, the K95Q mutation in THL cleaved EBOV GP displayed an intermediate phenotype that resulted in titers more similar to uncleaved EBOV K95Q GP. These results suggest that the amino acid sequence in the RBD of EBOV GP provides greater stability than that of both SUDV GP and BDBV GP. Because our results indicated that SUDV GP entry is completely independent of CatB activity it is possible that less extensive proteolysis primes this protein for fusion, thus it is not surprising that

125 109 overdigestion with THL (which mimics CatB cleavage) resulted in GP inactivation. Interestingly, EBOV K95Q GP was also completely insensitive to CatB inactivation but remained capable of mediating entry, albeit ~40-fold less than EBOV GP THL. Treatment of pseudovirions bearing ebolavirus GPs with lower concentrations of THL resulted in enhanced entry, however cleavage products were not detectable. These results suggest that a minimal amount of primed GP or the presence of cleavage intermediates are able to enhance entry compared to untreated GP. Thus, the residue at position 95, which residues at GP1/GP2 interface, is important for stability of the processed form of GP, but not unprocessed GP. Additional mutagenesis will reveal which residues support this intermediate phenotype. Further work needs to be performed to implicate the role of the residue at this position in mediating fusion. The SUDV GP clone that we utilized in these studies was from an isolate of the Boniface strain that was responsible for the 1976 outbreak. Other isolates of this same strain contain a lysine at position 95, however a glutamine at position 95 was also present in the Maleo strain, which caused the 1979 outbreak. Thus, it is possible that these early outbreaks contained isolates that were representative of viruses circulating in the natural bat reservoirs, which may have selective pressures leading to CatB-independent virus populations. Interestingly, virions pseudotyped with SUDV GP provided titers that were over a log lower than those obtained with EBOV GP, independent of protein expression. This decrease was abolished by mutation of position 95 to match the residue present in EBOV GP. Together with our CatB-inhibition studies, these results suggest that CatBdependence can provide a fitness advantage in the case of SUDV GP-mediated entry. Surprisingly, BDBV GP-mediated entry was also a log lower than EBOV GP, however

126 110 this was independent of CatB-dependence. Furthermore, titers obtained through overdigestion of BDBV GP with THL maintained this defect. However, a less extensive THL digestion decreased the titer differences between BDBV GP THL and EBOV GP THL to ~5-fold. These results suggest BDBV GP may be more sensitive to protease inactivation, which could account for the differences observed during entry. Filoviruses have evolved to utilize non-specific cell-surface receptors with redundant functions (CLECs and PS receptors), but utilize specific endosomal host factors for entry. Our results show that a single nucleotide change in the GP ORF results in resistance to drugs that target specific, non-redundant, intracellular entry factors. This introduces a difficult hurdle in the development of anti-filoviral drugs, since the virus is likely to rapidly adapt to inhibition of drugs that target both viral proteins and host factors.

127 111 A Titer (TU/mL) B ** * GP GP!muc EBOV SUDV C Relative Expression (EBOV GP!muc = 100%) EBOV SUDV Figure 4-1 Comparison of EBOV GP and SUDV GP mediated entry. (A) Titer of VSVΔG pseudotyped with EBOV/SUDV GP and GPΔmuc in Vero cells. Data are represents the average ± SD transducing units (TU)/mL of three independent experiments. Significance was determined by Student s t test, * p < 0.05 and ** p < (B) Linear diagram of EBOV GP. Percent amino acid identity, by region, between EBOV GP and SUDV-Boniface GP is shown in parentheses. The cathepsin B/L cleavage site is marked by an arrow and the disulfide bond between GP1 and GP2 is shown. SP: signal peptide, RBD: receptor-binding domain, IFL: internal fusion loop, HR1/2: heptad repeat 1 and 2, respectively, and TM: transmembrane domain. (C) Incorporation of GPΔmuc-HA into VSVΔG pseudovirions. Left, immunoblot shows C-terminally HA tagged EBOV and SUDV GPΔmuc (EΔ and SΔ, respectively) from pelleted pseudovirus pelleted pseudovirus stocks. Right, ratio of GP2 (HA signal) to VSV-M, determined by the signal intensity of each band with LICOR Image Studio. Data are presented as the average ± SD of three independent virus stocks, shown as a percentage of EBOV GPΔmuc.

128 112 A # EBOV-May* (NP_066246) EGNGVATDVP SATKRWGFRS GVPPKVVNYE AGEWAENCYN LEIKKPDGSE SUDV-Mal (Q66798) EGSGVSTDIP SATKRWGFRS GVPPQVVSYE AGEWAENCYN LEIKKPDGSE SUDV-Bon* (Q66814) EGSGVSTDIP SATKRWGFRS GVPPQVVSYE AGEWAENCYN LEIKKPDGSE SUDV-Bon (ACR33190) EGSGVSTDIP SATKRWGFRS GVPPKVFSYE AGEWAENCYN LEIKKPDGSE TAFV-CI (Q66810) EGNGVATDVP TATKRWGFRA GVPPKVVNYE AGEWAENCYN LAIKKVDGSE BDBV* (YP_ ) EGNGVATDVP TATKRWGFRA GVPPKVVNYE AGEWAENCYN LDIKKADGSE RESTV-Pen (NP_690583) EGNGIATDVP SATKRWGFRS GVPPKVVSYE AGEWAENCYN LEIKKSDGSE MARV (ADM72998) SGQKVADSPL EASKRWAFRT GVPPKNVEYT EGEEAKTCYN ISVTDPSGKS LLOV EGLGEHADLP TATKRWGFRS DVIPKIVGYT AGEWVENCYN LEITKKDGHP B Figure 4-2 Sequence alignment and structural mapping of potential interactions of residues at position 95 in GP. (A) Alignment of residues within the RBD of GP from the indicated filovirus species. Residues shown in bold differ from EBOV GP and NCBI accession numbers are noted in parentheses. Asterisks indicate the sequences of the constructs used in this study and # marks position 95. (B) Crystal structures were produced from the SUDV-Bon GP structure; PDB ID: 3VE0. (B, left) Top down view of trimeric GP. (B, middle) Side view of monomeric GP1/GP2 heterodimer. Receptorbinding residues K114 and K115 are labeled to indicate where NPC1 binding occurs. (B, right) Zoom-in of area surrounding position 95. Potential hydrogen-bond interactions and distances are labeled.

129 113 A B C D Figure 4-3 Mutation of Q95K in SUDV GP rescues poor titer phenotype. (A and B) Titer of VSVΔG pseudotyped with the indicated species of ebolavirus GP or mutant GP in Vero cells. Data represent the average ± SD TU/mL of three independent experiments. Significance was determined by Student s t test, * p < 0.05, ** p < 0.005, and *** p < (C) Immunoblot of the indicated pseudovirions. EBOV GP and K95Q mutant were detected with the MLD-specific mab 5E6. SUDV GP Q95K mutant were detected with the MLD-specific mab 17F6. No GP represents pseudovirions produced without an envelope glycoprotein.

130 114 A B C Cathepsin B Activity (DMSO = 100%) DMSO CA-074 Transduction (DMSO = 100%) Residue 95: ** ** K Q K Q CA-074 n.s. K Q K Q E-64 n.s. Eb Su Eb Su Residue 95: Transduction (DMSO = 100%) * K Q K Q Eb Su Bu Eb Su Bu CA-074 E-64 n.s. K Q K Q Figure 4-4 The presence of lysine vs. glutamine at position 95 alters endosomal protease requirements. (A) CatB activity in Vero cells pretreated with 80 µm CA-074 or DMSO for 2 h. Cell lysates were incubated with the Z-Arg-Arg-MCA CatB substrate, after incubation at room temperature for 30 min fluorescence was measured. Data are presented as the percent CatB activity in cells treated with inhibitor compared to activity in DMSO treated cells and represents as the average ± SD, n=2. (B and C) Inhibition of VSVΔG pseudotyped with EBOV (Eb), SUDV (Su), and BDBV (Bu), GP or mutant GP entry by CA-074 and E-64. Vero cells pretreated for 2 h with 80 µm CA-074 or 4 h 300 µm E-64 prior to addition of pseudovirions. Significance was determined by Student s t test, * p < 0.05, ** p < 0.005, n.s. indicates values are not significant.

131 115 A Titer (TU/mL) 10 2 Residue 95: D THL Sensitivity (Fold inc. in entry) GP 10 8 *** GP THL * 10 5 ** Residue 95: K Q Q EBOV SUDV K Q K Q K Q Eb *** ** Su *** ** Bu *** B Titer (TU/mL) GP 10 8 * ** GP THL THL conc. 20 µg/ml 200 µg/ml K K Q EBOV E SUDV ** C Titer (TU/mL) 10 8 * * K Q K Q EBOV SUDV BDBV * *** F G FIG 5 Mutation of the residue at position 95 alters GP protease sensitivity. VSV G pseudotyped Figure 4-5 Mutation of the residue at position 95 alters GP protease sensitivity. VSVΔG pseudotyped with the indicated EBOV (Eb), SUDV (Su), and BDBV (Bu) GP or mutant GP was mock or THL treated prior to transduction/immunoblot. (A-D) Titer of THL treated pseudovirions in Vero cells. Data represent the average ± SD TU/mL of three independent experiments. (A-C) Pseudovirions treated with 200 µg/ml of THL for 1 h at 37 C prior to addition to cells. Background is indicated by a dashed line, determined by fluorescence of cells exposed to pseudovirions lacking a viral glycoprotein. (D) Pseudovirions were treated with the indicated concentration of THL for 15 min prior to addition to cells. Data are presented as the average ± SD fold increase in transduction compared to mock treated pseudovirus. (E-G) Immunoblot of pelleted pseudovirions mock or THL treated, for the indicated amount of time. GP1 was detected with anti- EBOV GP1 polyclonal Ab and VSV-M was detected with the 23H12 mab. Significance was determined by Student s t test, * p < 0.05, ** p < 0.005, and *** p <

132 116 Transduction (DMSO = 100%) * E-64 ** 0.1 Residue 95: K Q K - Eb Su LV GP THL Figure 4-6 Cysteine protease dependence of VSVΔG pseudovirions treated with THL. The indicated virions were treated with 200 µg/ml THL for 1 h prior to addition to Vero cells treated with 300 µm E-64. EBOV (Eb), SUDV (Su), and Lassa virus GPC (LV). Data are presented on a log scale. Significance was determined by Student s t test, * p < 0.05 and ** p <

133 117 A Absorbance (450nm) C Loop (µg/ml) - THL + EBOV EBOV K95Q (5x) SUDV (5x) SUDV Q95K B *** ** 10 0 Residue 95: K Q K Q - Transduction (DMSO = 100%) D Transduction (DMSO = 100%) F U18666A Eb Su LV 0 Residue 95: K Q K Q - * Eb 3.47 *** Su LV C ** Residue 95: K Q K Q - Transduction (DMSO = 100%) E Transduction (DMSO = 100%) G U18666A Eb Su Bu LV 3.47 ** 0 Residue 95: K Q K Q - Eb Su Bu LV 10 6 null NPC1 Titer (TU/mL) Residue 95: K Q K Q - Eb Su LV Figure 4-7 Mutation of residue 95 in GP decreases sensitivity to the NPC1 inhibition, but does impart NPC1-independent entry. (A) NPC1 C loop GP binding assay. VSVΔG virions pseudotyped with the indicated EBOVGP or SUDV GP were captured on ELISA plates with mabs 2G4 and 16F6, respectively. Five-fold greater amounts of EBOV K95Q GP and SUDV GP were added to the wells, noted with 5x in parentheses. Serial dilutions of soluble C loop were added to the plate overnight and bound protein was detected via HRP conjugated anti-dykddddk polyclonal Ab. Results are representative of a single assay performed in duplicate. Similar results were obtained in additional experiments. (B- E) Inhibition of NPC1 during entry. Vero cells were treated with (B and C) 20 µm U18666A during transduction or (D and E) 2 µm compound 3.47 for 90 minutes prior to and during transduction with the indicated pseudovirions. Transduction data in the presence of drug are shown as the average ± SEM percentage of transduction in DMSO treated cells. Significance was determined by Student s t test, * p < 0.05, ** p < 0.005, and *** p < (F) NPC1 C loop GP binding assay in presence of the indicated concentration of compound Data are representative of the average ± SD of two

134 Figure 4-7 continued, experiments performed in triplicate. Dashed line represents average absorbance from wells without virus added. (G) Titer of the indicated pseudovirions in NPC1 null CT43 cells and CT43 cells stably expressing NPC1-myc. EBOV (Eb), SUDV (Su), BDBV (Bu), and Lassa virus GPC (LV). 118

135 119 Transduction (H 2 O = 100%) Residue 95: *** *** ** *** K Q K Q K Q - Eb Su Bu LV Verapamil (µm) Figure 4-8 Mutation of K95 in EBOV and BDBV decreases sensitivity to voltageoperated calcium channel inhibition. Vero cells were treated for 1 h with the indicated concentration of Verapamil prior to the addition of pseudovirions bearing GP or mutant GP of EBOV (Eb), SUDV (Su), BDBV (Bu), and Lassa virus GPC (LV). Data are shown as the average ± SD percentage of transduction in H 2 O (vehicle) treated cells. Significance was determined by two-way ANOVA with a Bonferroni post test, ** p < and *** p <

136 Figure 4-9 Kinetics of entry. Vero cells were transduced with the indicated pseudovirions. At the indicated times, NH 4 Cl was added to the culture supernatant at a final concentration of 20 mm. Transduction at each time point is shown as a percentage of transduction at 8 hours post addition of virus. Data represent the average ± SD of three independent experiments. 120

137 121 CHAPTER V DISCUSSION AND FUTURE DIRECTIONS The role of Ebola virus GP1 N-linked glycans during infection N-linked glycans on viral glycoproteins are known to provide a variety of roles, including expression, binding C-type lectins, and immune evasion (82). Previous studies indicated that N-linked glycans on EBOV GP1 interact with a number of C-type lectins at the cell surface, however the impact of these carbohydrates on endosomal entry processes and immune evasion had not been investigated. The studies carried out in Chapter II indicate that N-linked glycans on EBOV GP1 are not required for expression and actually decrease entry efficiency through limiting proteolytic processing, a previously unappreciated impact of N-linked glycans during virus entry. Additionally, disruption of the glycan shield, directly surrounding the conserved receptor-binding domain, resulted in an increase in sensitivity to antibody-mediated neutralization. These results highlight an evolutionary trade-off where N-linked glycan sites have been incorporated into GP to aid in immune evasion at the expense of optimal entry efficiency. Furthermore, N-linked deglycosylation altered the utilization of C-type lectins, which may provide a significant tool in understanding the importance of different filoviruses receptors, including phosphatidylserine receptors, in tissue tropism and pathogenesis. The role of GP1 N-linked glycans in Ebola virus pathogenesis Filoviruses infect a large number of cell types in various tissues. Early target cells include macrophages, monocytes, and dendritic cells, which are known to express C-type

138 122 lectins that have been associated with filoviral entry. It is known that incorporation of N- linked glycans at NGS and the species of glycans can vary between cell types (129). Consistent with these results, GP produced from monocyte-derived macrophages resulted in a decrease in interaction with DC-SIGN, L-SIGN, and ASGPR compared to GP produced in HEK 293T cells (88). This suggests that during infection the cellular source of virus may impact tropism. To evaluate the differences in cellular tropism of EBOV, virus harvested from different tissues of infected mice, via multiple routes, could be used to infect various target cells or poorly permissive cells expressing the various CLECs utilized by EBOV. For these studies it would be important to use GP, rather than the commonly used laboratory-derived GPΔmuc form of the protein, since O-linked glycans in the MLD contain ligands for different CLECs (53). Initial studies should utilize infectious, recombinant VSV expressing the EBOV GP (or GP from other filoviruses) or N-linked glycosylation mutants in a BSL2 setting to provide preliminary data prior to moving into infectious filoviruses in the BSL4. VSV is highly sensitive to the murine IFN response, therefore IFNAR -/- mice would be a more appropriate model system to allow for sufficient viral replication. Entry into poorly permissive cells ectopically expressing TIM-1/4, which enhance entry by a GP-independent mechanism, would provide a means to normalize entry results in the same cell type expressing different CLECs. The ability of serum-derived virus to utilize different CLECs would provide information on tissue tropism of virus produced early during a blood-contact initiated infection. The tropism of virus from various organs, such as the liver and lungs, would provide information on the target tissues late during infection. These tissues would contain viruses from various cell

139 123 types since there are likely to be immune cells present at the site of infection, especially the liver, which is highly vascularized. However, previous results from serum-derived virus could be used for interpretation of potential differences in receptor-utilization by these viruses. Overall these studies would provide interesting details on the role of glycosylation on virus tropism and the importance of specific cellular receptors/attachment factors. Removal of N-linked glycans from EBOV GP1 abrogates 2 out of 3 liver CLECs to mediate entry, therefore these studies should elucidate the role of specific liver receptors during filovirus infection. Given the previous results that virus from monocyte-derived macrophages poorly utilize several C-type lectins, there could be a significant role for phosphatidylserine receptors during infection. TIM-1 is expressed on the apical surface of human airway epithelia and is required for EBOV GP-mediated entry into these cells (79), however alveolar macrophages, which express DC-SIGN, are also present in the airway (156). Therefore, infection via the intranasal route with deglycosylated GP mutants should provide interesting details regarding the role of phosphatidylserine during pathogenesis. Further, since complete removal of N-linked glycans from EBOV GP1 had minimal impact on hmgl-mediated enhancement of entry, utilization of MGL knockout mice would provide an interesting model to investigate the tropism of virus in the absence of this receptor. Altogether these results should provide information on receptors/attachment factors that are critical for filovirus entry, which would aid in focusing the development of antivirals. Additionally, given the decrease in a number of cell surface entry factors, it is likely that N-linked glycan-deficient EBOV GP1 would be attenuated and may aid in development of filovirus vaccines. Furthermore, investigation

140 124 of other filoviruses may provide insights into important differences in virus tropism between species and genera. The role of N-linked glycans on Marburg virus GP and Lloviu virus GP Analysis of N-linked glycosylation sites in MARV GP indicates that this protein is predicted to be more heavily glycosylated than other filoviruses, with ~22 total predicted sites. The LLOV GP has 16 predicted N-linked glycosylation sites, however only 3 of these sites are within the GP1 core. The strategy that was used in Chapter II to systematically remove N-linked glycans from EBOV GP could be employed to determine the role of glycans on MARV GP and LLOV GP. It would be interesting to determine if there is a conserved stability of deglycosylated GP of filoviruses and if entry is enhanced upon removal of glycans from MARV/LLOV GP, similar to EBOV GP. Additionally, limited work has been done on the binding of CLECs by LLOV GP, only DC-SIGN and hmgl have been investigated (157). Therefore a similar strategy to that described above could be used to investigate LLOV GP-mediated tropism. Results from these studies could suggest a potential explanation for the absence of LLOV disease in primates, especially if LLOV GP-mediated entry is poorly enhanced by liver-specific CLECs. There are eight sites within the MARV GP1 core, most of which are at relatively well-conserved locations compared to EBOV GP (Figure 5-1). Interestingly, there is an additional site in the RBD at N94 (N110 in EBOV GP), which would be predicted to mask potential receptor-binding residues after endosomal processing. The location of this glycan may give insight into the interaction of primed GP of filoviruses with NPC1, since the glycan would be predicted to obstruct part of the RBD. The Asn residue is conserved

141 125 between EBOV GP and MARV GP, therefore E112 in EBOV GP could be mutated to a Thr or Ser residue to introduce an NGS to determine if this site is glycosylated and the impact on entry. A similar strategy was used to obscure the sialic acid binding site in HA of influenza A virus (158). Previous work has shown that MARV GP does not form a stable intermediate upon thermolysin (THL) proteolysis (36), therefore introduction of this NGS into EBOV GP would allow for a system to assess the impact of this glycan on receptor-binding. Preliminary results obtained through the introduction of an N-X-T/S motif in EBOV GP at this conserved Asn results in an increase in molecular weight of THL cleaved GP, indicative of the addition of a glycan (Figure 5-2). This suggests that the complimentary NGS in MARV GP1 is also glycosylated. Furthermore, introduction of this NGS in EBOV GP resulted in a dramatic decrease in titer and a decrease in C loop binding, despite equivalent levels of expression compared to WT (Figure 5-2). Additionally, chimeric proteins could be produced that substitute the RBD of MARV GP into EBOV GP in order to look at MARV GP interaction with NPC1. Since there is relatively poor homology (less than 50% identity) between the RBDs of MARV GP and EBOV GP, this would provide a system to determine if there are differences in critical residues for NPC1 binding between these proteins and if a N-linked glycan in the RBD effects EBOV GP and MARV GP in different ways. Determine residues masked by glycan cap/mucin-like domain polypeptide that bind NPC1 The RBD of EBOV GP has been extensively characterized through mutagenesis studies in order to determine residues involved in receptor-binding (38-40). At the time these studies were performed it was hypothesized that there was a GP-specific cell

142 126 surface receptor that interacts with the RBD. However, upon the determination of the prefusion GP structure, modeling of N-linked glycans on GP1 indicates that receptorbinding residues are efficiently masked by a glycan shield (42, 60, 62). To date there has not been a GP-specific cell surface receptor identified, however the RBD has been shown to bind to an intracellular receptor, NPC1, after proteolytic removal of the glycan cap and MLD. Interestingly, removal of the glycan shield through mutagenesis of the N-linked glycan sites was not sufficient to allow for binding to the intracellular receptor, NPC1 (Figure 2-6). These results indicate that there are additional residues masked by the glycan cap and/or MLD polypeptides. Systematic mutagenesis could be performed on residues that are exposed only after proteolytic removal of these domains to determine residues critical for NPC1 binding. Crystal structure models could be used to identify the surface exposed residues of primed GP and these residues could be cross-referenced with previous mutagenesis studies that identified mutations that decrease EBOV GP-mediated entry (38-40). The NPC1 C loop binding assay employed in Chapter II could be used to determine which residues are required for receptor interaction. Mechanism of enhanced neutralization Removal of N-linked glycans surrounding the conserved RBD led to enhanced neutralization by antibodies from convalescent cynomolgus macaques and mice. Interestingly, removal of N-linked glycans from the highly variable MLD did not have an impact on sensitivity to antibody-mediated neutralization, despite being highly targeted by neutralizing antibodies ( ). Therefore, it would be interesting to determine the mechanism of enhanced neutralization upon removal of glycans from the GP1 core.

143 127 During a natural infection ebolaviruses produce a soluble GP homodimer (sgp) from the same ORF as the surface GP. This protein is produced in far excess compared to GP, with 80% of transcripts from this ORF coding for sgp. Furthermore, the first 295 amino acids of this protein are completely identical to that of GP, which includes the conserved RBD and the less conserved glycan cap. Most antibodies from human or primate survivors are directed towards sgp (52) and in the absence of sgp these antibodies have been shown to neutralize GP-mediated entry (31). Thus, one potential explanation for the enhanced neutralization is that upon removal of N-linked glycans from the sgp/gp conserved region anti-sgp antibodies present in convalescent serum bind to GP1 core deglycosylated mutants and not WT GP, or less efficiently. This would also explain why we observed no impact on neutralization upon removal of N-linked glycans from the MLD, since anti-sgp antibodies would not target this region. One way to determine if convalescent antibodies bind deglycosylated mutants more efficiently is to perform ELISAs. Additionally, purified sgp could be introduced into these assays to indicate whether anti-sgp antibodies are responsible for the increased binding. Adding purified sgp into neutralization assays in order to alleviate the enhanced antibody sensitivity observed with deglycosylated GP1 core mutants could support these studies. Additionally, antisera specifically raised against GP or sgp could be used for ELISAs and neutralization assays to determine if there are differences in antibody binding or neutralization of deglycosylated mutants. These studies could utilize 6xHis-tagged sgp and 6xHis-tagged trimeric GP expression plasmids that I have constructed, and I have demonstrated that protein can be efficiently purified from supernatants of cells expressing the tagged proteins (Figure 5-3).

144 128 Another explanation for enhanced neutralization is that upon antibody binding the protein structure is more readily perturbed, leading to GP inactivation. We have shown in Chapter II that removal of N-linked glycans does not impact the ability of a conformation-specific antibody to recognize GP, however a common role for N-linked glycosylation is to promote protein stability (82). In order to test this, GP mutants could be absorbed by convalescent sera on an ELISA plate and then detected with a number of conformation-specific monoclonal antibodies. This would indicate if binding of GP by convalescent antibodies alters the native conformation of the protein or locks the protein in an alternative conformation, which could explain the inability of GP to mediate entry. Re-evolution of GP in vitro and in vivo Removal of N-linked glycans from EBOV GP1 had minimal impact on expression and resulted in an increase in GP-mediated entry into Vero cells, which do not require GP for virus internalization. This forced evolution indicated that N-linked glycans decrease entry of virions pseudotyped with EBOV GP. However, it is possible that there are undetermined consequences of GP1 deglycosylation in vitro. Infectious, recombinant VSV expressing EBOV GP, which lacks all N-linked glycan sites in GP1, could be serially passaged in Vero cells in the presence of neutralizing antisera/antibodies or overexpression of endosomal proteases, followed by sequencing to determine if the virus reincorporates glycosylation sites into GP1 due to these introduced pressures. Previous work has shown that N-linked glycosylation site gain is more likely to occur through the generation of a Ser or Thr downstream of an Asn residue (159). Since all but one site (N238) were disrupted through mutation of the Ser or Thr during the systematic removal of N-glycosylation sites (see Chapter II) it is more likely that these sites can be

145 129 regenerated. Reintroduction of glycans into GP1 would suggest that there was a loss of fitness upon complete removal of N-linked glycosylation sites through mutagenesis. The frequency of generated sites could be calculated to determine if the sites are conserved or random. Generation of conserved sites could be pursued to determine the impact they have on expression and entry. Over 50% of the Asn residues in GP1 are part of N-linked glycosylation sites. Therefore, it is possible that novel sites could be generated, which could lead to interesting insights into GP1 structure and function. A similar strategy could be performed in vivo to solidify the role of N-linked glycans in immune evasion. IFNAR -/- mice would be the best model system to use in order to allow for sufficient viral replication to occur and provide the opportunity for mutations to arise. These mice are highly susceptible to infectious VSV, therefore they would need to be infected with a low dose of virus. Antisera from infected mice should be passaged with virus prior to introduction into the next animal to ensure that the selective pressures of the antibody system are transferred. Analysis of generated sites at different stages of passaging would indicate which regions of GP are highly targeted by the immune system. The generation of N-linked glycosylation sites in vivo could be compared to those that were generated in vitro in order to discern a potential role for individual sites in immune evasion, entry, or expression. The role of N-linked glycans on EBOV GP2 during entry In Chapter III we showed two N-linked glycans on EBOV GP2 are completely conserved among filoviruses, suggesting functional significance. These N-linked glycosylation sites (N563 and N618) are located on heptad repeat regions 1 and 2

146 130 (HR1/2), respectively. Disruption of the N563 site resulted in a significant decrease in expression, but a significant increase in GP-mediated entry. Conversely, disruption of the N618 site had minimal impact on expression, but decreased GP-mediated entry by 2-fold. These results indicated that N-linked glycans at these sites have different roles during entry, likely during fusion. Role of GP2 N-linked glycans in triggering of fusion The filovirus GP is a type I fusion glycoprotein, indicating that fusion is driven by the formation of a six-helix bundle. The filovirus GP2 subunit contains HR1 and HR2 that form this structure during fusion; however, the cellular trigger for this event remains unknown. A modified version of a previously established in vitro fusion assay could be used to determine the effects of GP2 N-linked glycans during fusion (100). This assay could be performed by mixing biotinylated liposomes with pseudovirions bearing HAtagged GP2 N-linked glycan mutants followed by treatment with THL, at neutral/low ph, heat, low concentrations of urea, or a reducing agent. Streptavidin coated beads could then be used to pull down liposomes, followed by immunoblot detection of GP2-HA. It is possible that only overdigestion with THL of the N563D mutant would be required to trigger fusion since entry mediated by this protein was increased 2-fold and THL treatment decreased entry by 75%. Alternatively, fusion may be triggered at more mild secondary treatments after THL digestion. Utilization of varying conditions could indicate that the N563D mutation is more efficient at mediating fusion compared to WT. Conversely, the N618D mutation would be predicted to mediate fusion less efficiently than WT, since this mutation resulted in a decrease in entry with and without THL treatment.

147 131 Role of GP2 N-linked glycans in antibody evasion Previous work has shown that removal of N-linked glycans from heptad repeat regions of Nipah virus fusion protein resulted in enhanced antibody-mediated neutralization (113). Neutralization assays could be performed with convalescent antibodies, as described in Chapter II, with GP2 N-linked glycosylation site mutants. These studies could be followed up with assays to determine the mechanism of enhanced neutralization, as described above for GP1 mutants. Determine minimum glycosylation requirements for expression and entry As shown in Chapter III, mutation of both NGS within GP2 abolished expression and entry. Recently a cell line has been developed that shortens the N-linked glycosylation pathway in the Golgi, which results in small 2-3 residue sugars attached at NGS (160). These shortened sugars have been shown to be sufficient for protein folding and expression. Furthermore, the sugars can be digested with sialidase and β- galactosidase, leaving a single N-acetylglucosamine residue attached to the Asn. A single N-acetylglucosamine residue has been shown to be sufficient for maintaining influenza A virus HA protein structure and function (161). This cell line can be used to produce EBOV GP, followed by glycosidase-mediated removal of the shortened sugars to essentially eliminate all glycosylation from the protein. Expression and entry studies can then be performed without the need for mutations, which can have glycan-independent consequences. These experiments would provide insight into the role of glycosylation on GP folding versus stability. Furthermore, transduction, protease sensitivity, and neutralization studies can be performed with virions produced in these cells to support our findings in Chapters II and III, which would indicate that the removal of glycans, not the mutations made to disrupt the NGS, determine the phenotype of NGS mutants.

148 132 Effects of residue occupancy at position 95 in GP Filovirus entry is a unique process that requires endosomal proteolysis prior to binding the intracellular receptor, NPC1. All filoviruses require cysteine protease activity during entry; however, the specific proteases are not conserved between viruses. The studies presented in Chapter IV indicate that a single amino acid substitution in GP at position 95 of three viruses from the ebolavirus genus dramatically altered the stability of protease-mediated GP-intermediates, with a lysine at this position stabilizing and a glutamine destabilizing the protein. Furthermore, mutation of this residue altered sensitivity to inhibition of cathepsin B and NPC1. Resistance to inhibition of cathepsin B activity correlated with a decrease in stability of the protease-primed form of GP. However, all species tested remained highly sensitive to inhibition of all cysteine proteases regardless of the residue occupying position 95, which indicates the requirement for further proteolysis during entry. All filoviruses have been shown to require NPC1 for entry (63). Consistent with this, all ebolavirus species tested in Chapter IV were sensitive to NPC1 inhibition by U18666A (Figure 4-7). Mutation of residue 95, in all instances, decreased sensitivity to this inhibitor; however, this effect was not consistent with the residue occupying position 95. For example, entry mediated by wild-type GP from all tested species is highly sensitive to the U18666A inhibition, but EBOV GP and BDBV GP have lysine residues and SUDV GP has a glutamine residue at this position. Similarly, compound dependent inhibition of entry mediated by wild-type GP of the different species was decreased upon mutation of this residue. A glutamine at position 95 of GP in the species tested resulted in modest inhibition (~50%). Interestingly, a lysine at this position in

149 133 EBOV GP and BDBV GP led to a more dramatic inhibition ( 75%), whereas a lysine at position 95 in SUDV resulted in complete resistance to this drug. Furthermore, inhibition of voltage-operate calcium channels, with verapamil, decreased entry mediated by GP of all species tested. Upon mutation of residue 95, in only EBOV GP and BDBV GP, the sensitivity to this drug was decreased. The role of calcium channels during entry is unknown, however these results indicate that mutation of position 95 in some species alters the dependence on this step during entry. These mutations may prove useful in characterizing the post-internalization steps in filovirus entry since protease requirements for several species, the role of NPC1 during filovirus entry, and the fusion trigger have not been defined. Role of residue at position 95 during fusion Structural data indicate that the amino acid at position 95 in GP interacts with HR1 in GP2 (Figure 4-2). A lysine at this position makes contact with an unstructured loop in the middle of HR1, whereas a glutamine interacts with an alpha-helix of HR1. Therefore, after proteolysis a lysine may be required to stabilize the pre-fusion structure. The same approach described above to assay fusion for GP2 N-linked glycosylation site mutants could be used to determine the conditions required for fusion to occur when a glutamine occupies position 95. Given the entry and immunoblot results obtained with THL treated pseudovirions, it is hypothesized that a glutamine at position 95 will only require THL to mediate fusion. Alternatively, it is possible that overdigestion with THL leads to proteolysis of GP2 due to altered protein conformation when a glutamine is present at this position. HA-tagged GP could be used to detect if GP2 is being degraded upon THL treatment by immunoblot analysis.

150 134 Further characterize the entry pathway of filoviruses Results from Chapter IV indicate that there are differences in entry factors and roles of conserved entry factors during entry of ebolaviruses. Furthermore, mutation of the residue at position 95 altered endosomal events required for entry. To further characterize the potential differences in entry between WT and residue 95 mutants, lipophilic dyes could be used to track single virus particles during entry. Utilization of distinct dyes could allow for the tracking of two or more different viruses within the same cell. Information from these studies would indicate whether there are differences in pathways between species or WT and mutants. Furthermore, fusion events could be observed through tracking loss of the fluorescent signal upon lipid mixing between the virus and host membranes. The entry factor inhibitors used in Chapter IV could be used to observe different impacts these drugs have on virus trafficking. Several specific entry factors are known to be required for filovirus entry. However, specific proteases for MARV GP and SUDV GP-mediated entry and the trigger for fusion remain unknown. Incorporation of photoreactive amino acids (leucine or methionine) into pseudovirus particles could be used to cross-link entry factors to GP. Protein complexes could be immunoprecipitated, potentially by using a HA-tagged GP, and subjected to mass spectrometrical analysis to identify the proteins present. Alternatively, the genome-wide haploid genetic screen used to identify the HOPS complex and NPC1 as critical factors during filovirus entry could be performed with WT and residue 95 mutants to identify differences in factors required for entry (103).

151 135 EBOV GP MARV GP GP1 core Figure 5-1 Linear model comparison of EBOV GP and MARV GP. SP, signal peptide; RBD, receptor-binding domain; MLD, mucin-like domain; IFL, internal fusion loop; HR1/2, heptad repeat 1/2; and TM, transmembrane domain. Location of furin cleavage site is shown with an arrow. Predicted N-linked glycosylation sites are shown with a Y.

152 136 A B C Transduction (GP = 100%) EBOV GP E112T Absorbance (450nm) C Loop (µg/ml) GP THL E112T THL Figure 5-2 Addition of a N-linked glycan at a conserved Asn in EBOV GP. (A) Transduction of Vero cells by the indicated VSV pseudovirion. Data are presented as a percentage of EBOV GP transduction, n = 1. (B) Immunoblot of the indicated pseudovirions. Top panel shows unprocessed GP1. Bottom panel shows THL processed GP, the increase in molecular weight for the mutant is indicative of the presence of an additional N-linked glycan. (C) NPC1 C loop binding assay. The indicated pseudovirions were treated with THL and used in a NPC1 C loop binding assay, n = 1. The dashed line represents background absorbance in wells without virus.

153 137 A B GP!muc-His C Figure 5-3 Purified EBOV sgp-his and trimeric EBOV GPΔmuc-His. (A) Coomassie stain of purified sgp-his. Gel was loaded with 17 µg purified protein per lane with or without β-mercaptoethanol (βme). (B) Silver stain of purified GPΔmuc-His. (C) Blue- Native PAGE of purified GPΔmuc-His with or without the detergent, digitonin, to reveal the trimeric structure.

154 138 APPENDIX USE OF N-LINKED GLYCAN DEFICIENT EBOV GP1 AS AN IMMUNOGEN Currently there are no approved vaccines for filoviruses. Several platforms have been shown to be effective for vaccinating against homotypic filovirus infections. Adenovirus-based vectors and recombinant vesicular stomatitis virus vectors expressing filoviral GPs (rvsv-gp) have been shown to completely protect against challenge (162); however, both raise safety concerns due to pre-existing immunity and use of infectious virus, respectively. Furthermore, univalent cross-protection is poor, which is an important consideration due to the unpredictable nature of filovirus outbreaks (163). The receptor-binding domain (RBD) of ebolaviruses is highly conserved (~83% amino acid identity), however this region is masked by N-linked glycans (42). In Chapter II, we found that removal of N-linked glycans surrounding this conserved region (7G mutant), but not the highly variable mucin-like domain, resulted in enhanced neutralization by antibodies from convalescent cynomolgus macaques and mice. Additionally, previous work has shown that vaccination with deglycosylated influenza HA resulted in enhanced survival upon lethal challenge (161). Therefore, we hypothesized that the 7G mutant or 8G mutant (7G + N618D mutation in GP2) could serve as efficient immunogens to protect against heterotypic infection. In order to address the safety concerns associated with adenovirus vectors and rvsv, we utilized noninfectious VSV pseudovirions. We utilized a prime/boost regimen for 45 Balb/c mice (20 female and 25 male, 4-8 weeks old) that were age and sex matched. Mice were immunized intramuscularly at

155 139 day 0 and day 21 with 2.5 x 10 7 WT transducing units, normalized to WT VSV matrix expression (see Chapter II Materials and Methods) of VSV pseudovirions bearing EBOV GP, EBOV 7G, EBOV 8G, SUDV GP, or PBS alone, in 100 µl. At day 49, mice were shipped to United States Army Medical Research Institute of Infectious Diseases (USAMRIID). After being quarantined for 2 weeks, vaccinated mice were challenged with 1000 plaque forming units of the mouse-adapted strain of Ebola virus (MA-EBOV) and monitored over 28 days for morbidity and mortality. This strain was created through serial passaging of EBOV in murine populations to allow for adaptations to the murine interferon response (164). However, analysis of the MA-EBOV GP reveals there were no mutations to the N-linked glycosylation sites (165). As expected, mice vaccinated with PBS all succumbed to disease by day 7 (Figure A-1A). Vaccination with the heterotypic GP of SUDV lead to a significant decrease in survival compared to WT EBOV GP vaccination. Vaccination with pseudovirions bearing WT EBOV GP resulted in 8 out of 9 mice surviving challenge, while the 7G mutant provided 100% protection. Addition of the N618D GP2 mutant to 7G (8G) protected 6 of 9 mice from lethal challenge, potentially due to decreased expression of this mutant compared to WT and 7G (data not shown). The average weight of all groups dropped on day 6, however the average weight rebounded by day 7 of surviving mice within each group (Figure A-1B). The drop in weight correlated with an increase in sickness score for all groups (Figure A-1C). Furthermore, the average sickness of each group correlated with the protection, with PBS vaccinated mice having the highest sickness score and 7G vaccinate mice having the lowest. By day 10 all surviving mice had recovered from disease.

156 140 This study provides, for the first time, evidence that non-infectious VSV pseudovirions can serve as efficient vaccines against filoviruses in a rodent model. Since there was no precedent, we were unable to gauge the optimal dose to provide information about whether removal of N-linked glycans increases vaccine efficiency. Therefore, future studies will aim to determine the dose of WT to give intermediate protection. This information can then be used in combination with the 7G, or potentially 7Gm8G (see Chapter II), mutant to determine if removal of N-linked glycans provides enhanced protection. Additionally, vaccination with deglycosylated GP will be performed to determine if they provide protection against challenge by heterologous species. Future studies will also be performed to determine the immune correlates for protection. Preliminary evidence suggests that the prime/boost regimen we used in the current study provides relatively poor neutralizing antibodies (~80% neutralization at a 1:4 dilution of serum collected at day 35). Thus, it is possible that pseudovirion vaccinations stimulate a protective T cell response or antibody-dependent cell-mediated cytotoxicity. One possibility is that removal of glycans may enhance protease processing of antigen for MHC class II presentation to CD4+ T cells, which are important for activating an adaptive immune response. Overall, we have provided evidence that the VSV pseudovirion vaccine platform can serve as a safe alternative to rvsv and adenovirusvector vaccines. Future studies can be performed in guinea pigs and primates to determine if this platform is efficient in more relevant model systems.

157 141 A B C Figure A-1 Vaccination of Balb/c mice with VSV pseudovirions protects from lethal MA-EBOV challenge. (A) Survival curve of mice vaccinated with indicated VSV pseudovirions. Mice were challenged with 1000 PFU of MA-EBOV and monitored for 28 days. (B) Average ± SD weight of surviving mice during the 28 day period post challenge. (C) Average ± SD sickness score for vaccination groups. 1: animals are moving normally but have ruffled fur, 2: Ruffled fur, behavior subdued until stimulated, 3: Behavior is subdued regardless of stimulation, mice tend to be in hunched position, fur can be extremely ruffled, 4: Minimal movement, unresponsive when stimulated, possible ocular/nasal discharge. Mice with sickness score of 4 were euthanized.