Human Monoclonal Antibodies That Neutralize Pandemic GII.4 Noroviruses

Abstract

Background & Aims:

Human noroviruses are responsible for approximately 200,000 deaths worldwide each year. In 2012, the GII.4 Sydney strain emerged and became the major circulating norovirus strain associated with human disease. Our understanding of the human norovirus-specific antibody response is limited because few human monoclonal antibodies (mAbs) to noroviruses have been described and there are no functional assays to measure virus neutralization. We studied the antibody-mediated response to the GII.4 strain by isolating mAbs to GII.4 from infected patients and developing virus neutralization assays.

Methods:

We used a robust human hybridoma technique to isolate mAbs from patients previously infected with norovirus and identified mAbs that blocked virus binding to cell receptors, using virus-like particles to test blockade ability. We tested the ability of select mAbs to neutralize live human noroviruses using stem cell-derived human enteroids.

Results:

We isolated a panel of 25 IgG or IgA human mAbs that recognized norovirus GII.4 Sydney 2012 and determined their potential to block virus binding to cell receptors. In competition binding studies, most antibodies recognized 3 major antigenic sites on the GII.4 Sydney 2012 protruding (P) domain.

Conclusions:

We isolated and characterized human mAbs that neutralize live human norovirus GII.4 Sydney 2012—the predominant strain responsibl e for recent outbreaks. Analyses of these antibodies identified neutralizing epitopes; further studies will provide insight into the human immune response to this deadly virus.

Introduction

Since the licensure and use of rotavirus vaccines, human noroviruses (HuNoV) have become the major etiologic agent of epidemic and sporadic acute gastroenteritis¹. The persistence of HuNoVs is attributed to many factors, such as a low infectious dose, extreme environmental viral stability, high levels of shedding, and prolonged shedding even after symptoms have resolved². According to the Centers for Disease Control and Prevention, HuNoVs cause on average 19 to 21 million cases of infection and between 570 to 800 deaths in children under the age of five each year in the United States. HuNoVs infect people of all ages and, even though infection is characteristically acute and self-limiting, disease can become life threatening in children, the elderly, and the immunocompromised³. The correlates of HuNoV immunity in humans are poorly understood.

One of the challenges for developing antibodies or vaccines to prevent HuNoV-associated disease is the extreme antigenic diversity of field strains. HuNoVs currently are classified phylogenetically into 7 different genogroups (GI-GVII) and at least 41 different genotypes⁴. Viruses from genogroup I (GI) and the rapidly evolving genogroup II (GII) account for nearly all human infections. The HuNoV genome contains 3 open reading frames (ORF1, ORF2, and ORF3). ORF1 encodes nonstructural proteins, while ORF2 and ORF3 encode the major and minor capsid proteins, respectively. In the past, HuNoVs could not be cultivated in cell culture, but the VP1 and VP2 protein sequences could be expressed using a baculovirus expression system to produce HuNoV virus-like particles (VLPs)⁵. These VLP reagents have facilitated the study of HuNoV evolution, antigenicity and the emergence of new virus strains^6,7.

Since the mid-1990s, viruses from genogroup II genotype 4 (GII.4) have caused the majority of outbreaks, with new strains emerging every 2–3 years⁴. In 2012, the GII.4 Sydney strain emerged and since then has continued to predominate among circulating strains. The molecular basis for antibody-mediated recognition of these strains and their mechanisms of action are not well characterized. In this study, we describe the isolation and characterization of a panel of human monoclonal antibodies (mAbs) that bind to GII.4 Sydney 2012 VLPs. The majority of these antibodies also block receptor binding, as inferred by their ability to inhibit hemagglutination of human O+ red blood cells (RBCs) or the interaction between GII.4 Sydney 2012 VLPs and porcine gastric mucin (PGM). Both of these assays are surrogate systems for testing HuNoV neutralization^8,9. For over 40 years there have been numerous documented attempts to cultivate HuNoVs in vitro, but previously none of them resulted in the establishment of a robust reproducible system of viral growth^10–12. Recent breakthroughs in the development of an in vitro replication system using human intestinal organoid technology have now made it possible to cultivate HuNoV and to test inhibition of growth, or neutralization, using antibodies^13,14. Here, we used a human jejunal monolayer culture system to identify antibodies that neutralize live GII.4 Sydney 2012 HuNoV. We identified the first neutralizing human mAbs against norovirus, as well as a panel of human anti-GII.4 Sydney 2012 VLP binding IgGs and the first anti-GII.4 human IgA molecules. Almost 70% of the mAbs that we isolated exhibited a high level of potency, inhibiting GII.4 Sydney 2012 VLPs from binding to PGM at half-maximal effective concentrations (EC₅₀) below 24 µg/mL. We also used this panel of mAbs to identify major antigenic sites on the GII.4 Sydney 2012 major capsid protein. These studies contribute new insights into natural human humoral immunity to HuNoVs and provide mAbs that have the potential to be used for diagnostic and therapeutic purposes.

Materials and Methods

Virus-like particles.

GII.4 Sydney 2012 virus-like particles (VLPs), based on strain AFV08795.1, were produced and purified, as previously described¹⁵. Briefly, VP1 and VP2 capsid protein sequences were cloned into the transfer vector pVL1392 (Epoch Life Sciences, Inc.). The vector was co-transfected with a bacmid vector into Sf9 insect cells. Recombinant virus then was used to inoculate Sf9 cells. VLPs were purified from the culture supernatant using a cesium chloride cushion gradient. GII.4 VLP assembly was verified visually using electron microscopy, and antigenicity was tested by western blot.

VLP binding assay.

Antibody reactivity to GII.4 VLPs was tested using an indirect enzyme-linked immunosorbent assay (ELISA). Microtiter plates were coated with 1 µg/mL of GII.4 VLPs in PBS at 4°C ove rnight. Wells then were blocked with 5% nonfat dry milk in PBS with 0.05% Tween-20 for 1 hour at room temperature. Purified antibodies were diluted serially in PBS and added to VLP-coated plates for 1 hour at room temperature. Microtiter plates were washed 3 times with PBS-0.05% Tween-20 in between each step. Antibodies that bound to VLPs were detected using horseradish peroxidase tagged anti-к or -λ chain secondary antibodies (Southern Biotech) for 1 hour at room temperature. Plates were developed using the ultra-TMB reagent (Pierce ThermoFisher) and stopped using sulfuric acid. Absorbance was measured at 450 nm using a BioTek Synergy HT Microplate Reader.

VLP-carbohydrate binding antibody blockade assay.

Microtiter plates were coated with 10 µg/mL of pig gastric mucin (PGM) Type III (Sigma) in PBS for 4 hours at room temperature, and then were blocked overnight at 4°C in 5% nonfat dry milk in PBS with 0.05% Tween-20. GII.4 Sydney 2012 VLPs (0.5 µg/mL) were pretreated with each mAb applied in serial 3-fold dilutions with decreasing concentrations. Complexes then were applied to PGM-coated plates for 1 hour at room temperature. Microtiter plates were washed 3 times with PBS-0.05% Tween-20 in between each step. Bound VLPs were detected using guinea pig serum containing anti-GII.4 Sydney 2012 polyclonal antibodies, followed by an alkaline phosphatase-conjugated anti-guinea pig IgG. Optical density was measured at 405 nm using a Synergy HT Microplate Reader (BioTek).

Hemagglutination inhibition assay.

Human type O+ red blood cells were purchased from Rockland Immunochemicals, Inc. Cells were pelleted at 500 × g for 10 minutes at 4°C and washed twice with PBS wit hout Ca²⁺ or Mg²⁺. GII.4 Sydney 2012 VLPs (3.5 µg/mL) were pretreated with decreasing concentrations of each mAb, from 15 to 0.007 µg/mL, in PBS pH 5.5 and incubated at room temperature for 30 minutes. VLP-mAb complexes were added to an equal volume of 0.5% washed red blood cells in PBS pH 6.2 and incubated for 2 hours at 4°C in a 96-well V-bottom microtiter plate. The HAI titer was determined as the lowest concentration of antibody that completely inhibited hemagglutination.

Human subjects.

We studied otherwise healthy adult subjects with a history of acute gastroenteritis contracted during a HuNoV outbreak in North Carolina between February 27 and March 1, 2013. The cause of the outbreak was determined by the Orange County, NC health department to be a GII.4 Sydney 2012 norovirus strain. Subjects were recruited after recovery to donate a one-time peripheral blood sample. The research study was approved by the Vanderbilt University Medical Center Institutional Review Board; all subjects provided written informed consent prior to participation.

Peripheral blood mononuclear cell (PBMC) isolation and hybridoma generation.

We obtained PBMCs from heparinized blood by density gradient centrifugation using Ficoll-Histopaque from 7 donors who had recovered recently from natural infection with HuNoV. B cells were transformed with Epstein Barr virus substrain B95.8 in the presence of 2.5 µg/mL of CpG10103, 10 µg/mL of cyclosporine A, and 10 µM Chk2 inhibitor. Approximately 10⁷ PBMCs were plated into a 384-well plate in transformation medium, and a week later were expanded into four 96-well plates containing irradiated human PBMCs as a feeder layer. After an additional 7 days of culture, the supernatants were screened by indirect ELISA for the presence of antibodies that bound to GII.4 Sydney 2012 VLPs. Antibodies that bound to GII.4 Sydney 2012 VLPs were detected using horseradish peroxidase tagged anti-human IgA or IgG secondary antibodies (Southern Biotech). Wells containing transformed B cells secreting anti-GII.4 Sydney 2012 VLP antibodies were fused with HMMA2.5 myeloma cells using a CytoPulse Sciences Generator. After fusion, hybridomas were plated in selection medium containing 100 µM hypoxanthine, 0.4 µM aminopterin, 16 µM thymidine and 7 µg/mL ouabain. After two weeks, hybridomas were screened for production of human antibodies reacting with GII.4 Sydney 2012 VLPs and then cloned biologically using single cell sorting on a FACSAria III flow cytometer in the Vanderbilt Flow Cytometry Shared Resource.

Competition-binding assay.

To identify groups of antibodies binding to similar antigenic sites on norovirus GII.4 Sydney 2012, we performed biolayer interferometry using an Octet^® Red96 or Octet^® HTX biosensor system (FortéBio). The Octet^® HTX is a high-throughput biosensor system that was used to validate results obtained from the Octet^® Red96 system. With both biosensor systems, antibodies and antigen were diluted in 1X kinetic buffer (FortéBio 18–5032). Glutathione S-transferase (GST)-tagged GII.4 Sydney 2012 P domain dimers were immobilized onto anti-GST biosensor tips (FortéBio 18–5096). The P domain dimers were coated onto the biosensor tip by immersing the tip in a solution containing dimers at a concentration of 5 µg/mL. The biosensor tip with the bound P domains was washed and then submerged into a well containing 50 µg/mL of the first antibody and then dipped into another well containing 50 µg/mL of the second antibody. If binding of the first antibody still resulted in greater than 66% of binding of the second antibody, the result was interpreted to be no competition. If binding of the second antibody was between 34 and 66% in the presence of the first antibody, there is believed to be partial competition. If 33% or less binding of the second antibody was noted in the presence of the first, both antibodies are believed to be in competition with each other. Antibodies then were clustered based on their binding patterns.

Stool filtrates.

To prepare 10% stool filtrates, 4.5 mL of sterile PBS was added to 0.5 g of GII.4 Sydney 2012 positive stool sample. The stool suspension was sonicated using a cup horn sonicator and centrifuged at 1,500 × g for 10 minutes at 4°C. Supernatant was collected and transferred t o a new tube and centrifuged once again at 1,500 × g, for 10 minutes at 4°C. The resulting supernatant then was passed serially through 5 µm, 1.2 µm, 0.8 µm, 0.45 µm and 0.22 µm filters, and aliquoted and frozen at −80°C.

Expression and purification of GST-GII.4 Sydney 2012 P domain.

P1 and P2 domain sequences of GII.4 Sydney 2012, AFV08795.1, VP1 were cloned into the pGEX-4T-1 expression vector with a glutathione S-transferase (GST) tag and thrombin cleavage site. The P domain was expressed in Escherichia coli BL-21 cells and purified using standard column chromatography techniques with a prepacked Glutathione Sepharose Fast Flow column (GE Healthcare). GST-tagged proteins were eluted using 50 mM Tris-HCl, 10 mM reduced glutathione, pH 8.0 and stored at 4°C.

GII.4 Sydney 2012 virus neutralization assay.

Human intestinal enteroids (HIEs) were generated and cultured as described previously¹³. Briefly, HIEs were grown as three-dimensional cultures in Matrigel (Corning) for 5 days and then plated as cell monolayer cultures in 96-well plates. Before plating, 96-well plates were pre-coated with collagen IV (Sigma) at 33 µg/mL in sterile cold water for 1.5 hours at 37°C. Three-dimensional HIEs were collecte d in 0.5 mM ethylenediaminetetraacetic acid diluted in ice-cold Dulbecco’s PBS, no calcium, no magnesium (Life Technologies, Cat# 14190–144) and spun down at 200 × g for 5 minutes at 4°C in a swinging bucket rotor. Th e pellet then was suspended in 0.05% trypsin/0.5 mL ethylenediaminetetraacetic acid and incubated at 37°C for 4 minutes. Trypsin then was inactivated with complete medium without growth factors [CMGF(−)] supplemented with 10% fetal bovine serum (FBS). The resulting pellet was suspended and passed through a 0.4 µm cell strainer and spun down at 400 × g at room temperature for 5 minutes. The pellet then was suspended in complete medium with growth factors [CMGF(+)] containing 10 µM Y-27632 (Sigma-Aldrich; Y0503) and seeded into a 96-well plate. After 24 hours, the culture medium was removed and replaced with differentiation medium. Cells were differentiated for 5 days. HuNoV GII.4 Sydney 2012 (TCH12–580)¹³ stool filtrate (2 × 10⁷ genome equivalents/µL) was used to test neutralization. Serial dilutions of the mAbs were prepared in CMGF(−) medium and each dilution was pre-incubated with 2.5 × 10⁵ genome equivalents of GII.4 Sydney 2012 at 37°C for 1 hour. Samples were diluted with equal volume of CMGF(−) medium supplemented with 1000 µM sodium glycochenodeoxycholate. Monolayers then were inoculated with pre-incubated samples. At 1 hour post-infection (HPI), monolayers were washed twice and incubated with differentiation medium supplemented with 500 µM glycochenodeoxycholate. After 1 and 24 HPI, cells and medium were collected and RNA was extracted using KingFisher Flex Purification System and MagMax Viral RNA Isolation kit. For RT-qPCR, a primer pair (COG2R /QNIF2d) and probe (QNIFS)^16,17 were used with qScript XLT One-Step RT-qPCR ToughMix reagent with ROX (Quanta Biosciences). Reactions were performed on an Applied Biosystem StepOne Plus thermocycler. A recombinant HuNoV RNA transcript was used to create a standard curve to quantitate viral genome equivalents in new RNA samples.

Results

Isolation of GII.4 VLP-reactive human mAbs.

The first step here was to isolate naturally occurring human mAbs to GII.4 Sydney 2012 virus capsid protein from human subjects with prior GII.4 Sydney 2012 virus infection. We used PBMCs collected from subjects with previous history of laboratory-confirmed GII.4 Sydney 2012 virus infection to generate human hybridoma cell lines secreting GII.4 VLP-reactive human mAbs. PBMCs were transformed with EBV and supernatants then were collected from lymphoblastoid cell lines and screened by ELISA for binding to GII.4 Sydney 2012 VLPs. Recombinant expression of norovirus genome ORF2 and ORF3 in a baculovirus expression system were used to generate VLPs that are antigenically and morphologically indistinguishable from native virions^5,18. Antibodies that bound to VLPs were detected using horseradish peroxidase conjugated anti-λ or -к light chain secondary antibodies. We used an anti-light chain secondary antibody for detection in order to isolate antibodies of varying Ig heavy chain isotypes. Previous studies have noted the presence of diverse isotypes in the human polyclonal antibody response to infection, including an increase in IgA, IgG, and IgM antibodies in serum^19,20, and we have shown previously that human IgAs can be more potent than IgGs in blocking GI.1 VLPs from binding to histo-blood group antigens (HBGAs)¹⁵. Transformed B cell lines corresponding to supernatants that contained antibodies that bound to GII.4 VLPs were fused with myeloma cells to create human mAb-secreting hybridoma cells. We isolated a panel of 25 hybridomas secreting VLP-reactive antibodies (21 IgGs and 4 IgAs) from 7 different donors (Table 1 and Supplementary Table 1).

Table 1.

Isotype, Light Chain and ELISA Binding Characterization of GII.4 P VLP-Specific Human mAbs

Isotype	mAb clone, NORO-	Light chain	EC50 (μg/mL)
IgG	115	κ	0.1
	313.1	κ	0.1
	246A	λ	0.2
	250B	λ	0.2
	279A	λ	0.2
	329A	λ	0.2
	118	λ	0.3
	316	λ	0.3
	202A.1	λ	0.4
	312A	λ	0.4
	317	λ	0.4
	303	λ	0.5
	263	λ	0.6
	296A	λ	0.6
	327A	λ	0.6
	315B	λ	0.7
	251A	λ	0.9
	256A	λ	0.9
	278	λ	4.0
	123	λ	5.4
	310A	κ	6.2
IgA	318	κ	0.1
	320	κ	0.1
	273A	κ	0.2
	232A.2	κ	0.3

NOTE. Listed are the half-maximal concentrations (EC₅₀) at which half-maximal binding was obtained when tested by ELISA using VLPs as the antigen. MAbs are organized by isotype (IgG or IgA) and arranged in order from lowest to highest EC₅₀ value when binding to GII.4 VLPs.

MAb binding and blockade of GII.4 VLPs.

Next, we sought to determine how well the antibodies bound to GII.4 Sydney 2012 antigens, and to identify if any blocked attachment of VLPs to surrogate receptor molecules. Half-maximal effective concentrations (EC₅₀) were determined for the panel of GII.4 Sydney 2012 VLPs reactive mAbs using indirect ELISA. For the 21 isolated IgGs, EC₅₀ values ranged from 0.1 to 6.2 µg/mL (Table 1 and Supplementary Figure 1). For the 4 isolated IgAs, EC₅₀ values ranged from 0.1 to 0.4 µg/mL. We initially used a surrogate system to assess the neutralizing capacity of all 25 mAbs. The presence of antibodies that block VLPs from binding to HBGAs in vitro correlates with protection against clinical NoV gastroenteritis^21,22. To test blockade potential, serial dilutions of isolated mAbs were pre-incubated with GII.4 VLPs, and then VLP-antibody complexes were added to microtiter plates that had been coated previously with porcine gastric mucin type III (PGM)⁹. Previous studies have validated PGM as a reliable substrate to be used in VLP blockade assays²³. We then determined EC₅₀ values for the 4 IgAs and 13 IgGs, which ranged between 2.4 to 23.9 µg/mL (Figure 1A and B, and Supplementary Figure 2). Blockade activity was not detected for 8 of the IgGs when using antibody concentrations as high as 100 µg/mL. We determined the antibody variable gene heavy and light chain sequences for 17 of the 25 isolated mAbs (Supplementary Table 1). All 17 mAbs, both mAbs that did and did not inhibit GII.4 VLP binding to PGM, had distinct variable gene sequences, suggesting that blockade response is not restricted to a specific genetic sequence motif in antibody repertoires.

Figure 1. Neutralization of GII.4 Sydney 2012 using virus-like particles (VLPs) or live virus.

A) Half-maximal effective concentrations (EC₅₀) for all isolated IgGs and IgAs using a VLP blockade assay, and the antibody concentration at which hemagglutination was inhibited when using VLPs and O+ red blood cells. > symbols indicate blockade EC₅₀ values >100 µg/mL for IgGs or >113 µg/mL for IgAs, or HAI titers > 15 µg/mL.

B) Plotted are the absorbance values when plates were read at optical density (O.D.) 450nm for selected IgGs and IgAs when antibodies were diluted serially, combined with GII.4 Sydney 2012 VLPs and added to porcine gastric mucin (PGM).

C) Inhibition of replication of GII.4 Sydney 2012 virus using selected IgGs, IgAs and a non-specific human monoclonal antibody (mAb) were tested in a human intestinal enteroid system. An additional control for each experiment was virus incubated without a mAb. The half-maximal inhibitory concentration (IC₅₀) for each mAb is indicated in each individual graph. The data presented is an average of two independent experiments for NORO-263, −250B, −320, −273A, −318 and a dengue virus-specific control antibody 2D22. The number of genome equivalents for each concentration tested for each mAb including the no antibody control was the average of 6 replicates tested.

Hemagglutination inhibition (HAI) assay confirms mAb blockade activity.

We used a second functional assay to confirm the activity we observed in the blockade assay above. Previous studies have shown that an additional surrogate system to determine mAb neutralization is HAI, and that HAI serum antibody levels correlate with protection from symptomatic infection^8,23. We used serial dilutions of the isolated mAbs and pre-incubated them with GII.4 Sydney 2012 VLPs. VLP-antibody complexes were then added to human type O+ RBCs. HAI activity was assessed visually, and HAI titers were determined (Figure 1A). The majority of the mAbs, about 84%, had HAI activity similar to that of the measured blockade activity. Four mAbs differed in these measures, having either a greater than 2-fold difference in activities or exhibiting only HAI activity or only blockade activity.

Neutralization assay using stem-cell derived enteroids.

Inhibition of replication of GII.4 Sydney 2012 virus by mAbs NORO-250B, −263, −320, −273A, −318 and a non-GII.4 VLP binding control antibody, 2D22, were tested using an intestinal epithelial stem-cell derived in vitro cultivation system. We selected two IgGs, NORO-263 and −250B and three IgAs, NORO-320, −273A and −318. These five mAbs were chosen so that we could test neutralization by representative mAbs belonging to the three major antigenic sites we identified on the GII.4 Sydney 2012 P domain and to test at least two mAbs belonging to each isotype (Figure 2). Neutralization was measured by comparing the percent reduction of genome equivalents when compared to a no-antibody control within each assay using RT-qPCR (Figure 1C). A no-antibody control was used in each assay to normalize for any variability between experiments. Variability was noted due to the high sensitivity when using genome equivalents to measure replication. To account for these differences, we used six replicates for each mAb concentration tested within each assay and for the no-antibody control. We then averaged the genome equivalents from two separate assays for each antibody to obtain an IC₅₀ value. To verify that equal amounts of virus were added to each monolayer, each assay was performed in duplicate and RNA was collected from one assay at 1 HPI and from the other at 24 HPI. Four of the five antibodies tested, NORO-250B, −263, −273A and −318, had approximately 5 to 883-fold lower IC₅₀ values compared to blockade EC₅₀ values, or between 17 and 1,227-fold lower than HAI titers (Figure 1A and 1C). NORO-320 had a higher IC₅₀ in comparison to its blockade EC₅₀ or HAI titer. The dengue virus mAb 2D22 used as a similarly prepared negative control did not exhibit concentration-dependent inhibit of replication of GII.4 Sydney 2012 virus. Previous studies with polyclonal serum have shown that neutralization IC₅₀ values of GII.4 and GII.3 noroviruses are lower in comparison to blockade EC₅₀ values¹³. These studies as well as our data suggests that the HIE neutralization assay is likely more sensitive than the HBGA blockade or HAI assays.

Figure 2. Competition binding of GII.4 specific mAbs on GII.4 Sydney 2012.

P domain with the Octet^® Red96 system. Epitope binning was performed using biolayer interferometry. GST-tagged GII.4 Sydney 2012 P domain was loaded onto anti-GST tips, then the first antibody was loaded followed by loading of the second antibody. The numerical data indicate percent binding of the second antibody in the presence of the first antibody. Yellow, green, and magenta boxes indicate potential binding groups.

Binding studies using GII.4 Sydney 2012 protruding (P) domain dimers and shell (S) domain.

We next sought to determine if the antibodies were binding to the P or S domains of the major capsid protein VP1. Glutathione S-transferase (GST)-tagged recombinant GII.4 Sydney 2012 P domain dimers were expressed and purified using affinity column chromatography, as previously described²⁴. P domain dimers or recombinantly expressed S domains then were used as antigen for an indirect ELISA assay with serial dilutions of each mAb to determine if the 25 isolated mAbs were specific to the P or S domain of the GII.4 major capsid protein VP1. The VP1 region is divided into the S domain, which is not expressed on the recombinant GST-GII.4 P domain dimers, and the P1 and P2 subdomains that are expressed in the dimers. The P domain is the surface-exposed protruding region of the norovirus virion and also believed to include determinants for host cell attachment and antibody binding epitopes^25–27. The S domain is connected to the P domain by a flexible hinge region and forms the interior core of the viral capsid. The S domain has the highest degree of genetic sequence conservation of any protein domain in diverse norovirus strains²⁸. When binding was tested by ELISA, 20 of the 25 isolated mAbs bound to the GST-GII.4 P domain (Supplementary Figure 3). The five mAbs that did not bind to the P domain did bind to the S domain. NORO-329A, −312A, −296A and −232A.2, bound to both the P and S domain proteins, indicating that these antibodies likely bind to a quaternary epitope on the GII.4 Sydney 2012 major capsid protein. Isolated mAbs and GII.4 P domain dimers also were used for competition-binding studies. We used a real-time biolayer interferometry biosensor system to identify potential major antigenic sites recognized by the GII.4 Sydney 2012 P domain binding mAbs. Neutralizing mAbs and mAbs that did or did not block GII.4 Sydney 2012 VLPs from binding to PGM were classified into three major competition-binding groups, with some overlap between two groups (Figure 2). Despite multiple attempts, we were not able to detect binding using biolayer interferometry for 2 of the 20 mAbs that bound to the GII.4 Sydney 2012 P domain by ELISA. Competition-binding studies were performed using the Octet^® RED96 and Octet^® HTX systems, which are both instruments that can measure biomolecular interactions. The HTX system is a high-throughput system that has the ability to compete all 18 P domain binding mAbs within the same experiment. The RED96 system was only able to compete mAbs in groups of 8. Using data from the RED or HTX experiments, we noted three major competition-binding groups on the GST-GII.4 Sydney 2012 P domain (Figure 2 and Supplementary Figure 4). By using both instruments, in two independent laboratories, we were able to validate the reproducibility of the results.

The emergence of new GII.4 strains has been associated with the evolution of the GII.4 major capsid protein and antigenic variation²⁹. To measure the reactivity of our panel of mAbs for another GII.4 strain, we tested binding reactivity and blockade activity to a GII.4 Houston 2002 (ABY27560.1) variant. About 93% of the amino acid sequence of the major capsid protein is conserved between GII.4 Houston 2002 and Sydney 2012, but there are remarkable differences among predicted GII.4 blockade epitopes (Supplementary Figure 5). All 25 mAbs exhibited binding reactivity to GII.4 Houston 2002 antigen (Supplementary Figure 6). Only 3 mAbs, NORO-115, −329A, and −318 had a greater than 10-fold higher binding EC₅₀ value when compared to GII.4 Sydney 2012 VLPs (Table 1). The 18 mAbs that either blocked with EC₅₀ values <100 µg/mL or had HAI titers < 15 µg/mL all blocked GII.4 Houston 2002 VLPs from binding to PGM. GII.4 Houston 2002 and Sydney 2012 had different amino acid sequences in four of the five predicted blockade epitopes, these results suggest the potential existence of additional blockade epitopes or the use of Epitope B, which was conserved among both strains (Supplementary Figure 5).

Discussion

Here we report the first instance of neutralization of HuNoV by mAbs and describe a large panel of human mAbs that neutralize the pandemic GII.4 Sydney 2012 strain. Previously it was not possible to test norovirus neutralization directly because of the lack of a reliable in vitro culture system for norovirus replication. A surrogate system to predict neutralization was devised and used to study inhibition of the interaction between VLPs and HBGAs. The presence of blocking antibodies in serum correlates with protection from clinical gastroenteritis induced by HuNoV infections, and therefore the HBGA blocking assay has been considered a surrogate system for HuNoV neutralization⁹. Recently, we developed an in vitro system using human enteroids to replicate multiple HuNoV strains¹³. Here, we used that system to identify the first norovirus human mAbs with demonstrated virus neutralizing activity. Of the 25 human mAbs isolated in this study, 17 of them blocked GII.4 Sydney 2012 VLPs from binding to PGM at concentrations as low as 2.4 µg/mL. The 18 mAbs that blocked GII.4 Houston 2002 from binding to PGM at concentrations as low as 2 µg/mL, also either blocked GII.4 Sydney 2012 VLPs from binding to PGM or inhibited hemagglutination at the concentrations tested. Interestingly, 13 of the 14 IgGs that blocked GII.4 Houston 2002 VLPs from binding to PGM did so at a lower EC₅₀ value in comparison to blockade EC₅₀ values for GII.4 Sydney 2012. This finding could indicate that our donors had prior exposure to an earlier norovirus variant similar to GII.4 Houston 2002. This panel also contains the first reported human IgA mAbs that bind to GII.4 Sydney 2012 VLPs and also inhibit receptor binding. We tested neutralization of live GII.4 Sydney 2012 HuNoV using the mAbs NORO-263, −320, −250B, −273A and −318. These antibodies were selected for testing so that we could investigate differences in neutralizing activity between mAbs of differing isotypes, those which belong to different competition-binding groups, and those with different binding and blockade EC₅₀ values. NORO-250B, −263, −273A and −318, had lower neutralization IC₅₀ values in comparison to blockade EC₅₀ values and HAI titers. Surprisingly, NORO-320 had an IC₅₀ value about 2-fold higher than its HAI titer and about 3-fold higher than its blockade EC₅₀ value. Previous studies have noted differences in blockade potency of GI.1 VLPs among human IgG and IgA mAbs with blockade potency being enhanced for IgAs¹⁵. To draw a similar conclusion for GII.4 Sydney 2012 neutralizing human antibodies, it would be essential to test mAbs binding to different epitopes with identical variable domain sequences and distinct isotypes. Such studies could determine if isotype plays a critical role in neutralization of GII.4 Sydney 2012 by human mAbs.

HuNoV-specific antibodies have been described previously, but these were antibody fragments derived from phage display libraries³⁰, murine mAbs from infected mice³¹, nanobodies from alpacas immunized with VLPs³², or mAbs isolated from patient PBMCs with an unknown norovirus history of exposure²³. Such antibodies do not provide direct information about the physiologic human humoral immune response to HuNoV infection. The therapeutic potential of mouse mAbs is limited, since they have been shown to induce human anti-mouse antibody responses. There has been little progress in understanding individual HuNoV-specific antibodies in the past because of the difficulty in generating human mAbs with functional activity. Here we used a hybridoma technology³³ and circulating B cells from convalescent patients to produce human mAbs. This approach generates hybridoma cell lines from circulating B cells that express naturally occurring and matched heavy and light chain genes. An additional benefit of using this approach is that it does not involve the use of any laboratory animals to produce antibodies. Using this method, we isolated 25 GII.4 Sydney 2012 VLP-reactive mAbs. The majority of these antibodies were neutralizing when tested in a surrogate system for neutralization and all five of the mAbs tested also inhibited replication of live GII.4 Sydney 2012 virus by direct neutralization in vitro. Neutralizing human mAbs have potential for use in prophylactic, therapeutic or diagnostic applications. We currently do not have any drugs available to treat or prevent HuNoV infection, so our panel of neutralizing mAbs now have the high potential to impact the design of improved diagnostic and therapeutic measures for HuNoVs.

Since the mid-1990s, new antigenically diverse GII.4 pandemic viral strains have emerged continuously every 2 to 5 years, and today these strains continue to be the predominant cause of norovirus outbreaks. In 2012, the epidemic GII.4 Sydney variant emerged in Australia and began spreading globally. Even though blockade epitopes among some contemporary GII.4 strains have been predicted or identified, we have limited information about the neutralization determinants on GII.4 Sydney 2012 viruses ^34,23. Here, we determined that there are at least three major antigenic and neutralizing sites on the P domain of GII.4 Sydney 2012 viruses. In the future, defining neutralization epitopes in high resolution with neutralizing antibodies could contribute valuable insights for rational structure-based vaccine design efforts.

HuNoV is one of the leading causes of severe acute gastroenteritis, therefore the global burden of norovirus infection is extremely high in both developed and developing countries. Unfortunately, there is currently no licensed vaccine to revent norovirus infection. Efforts to design a vaccine have been hindered by the lack of a small animal model or tissue culture model to test neutralization or infection, the antigenic heterogeneity among noroviruses, and uncertainty about the durability of protective immunity ³⁵. Vaccine efforts have focused on the use of monovalent GI.1 or bivalent GI.1/GII.4 virus-like particles or P particle subunits^36–38. Clinical trials have shown that norovirus VLP vaccines are immunogenic and without frequent serious adverse events^39,40. We now have developed a reliable in vitro system to test the replication or inhibition of replication of live noroviruses. Mapping the neutralization or blockade epitopes using the panel of mAbs we isolated against this circulating pandemic strain of norovirus will provide critical information that can be used for the design of future VLP vaccines that can elicit a protective immune response.

Abbreviations:

CMGF

complete medium and growth factors

EC₅₀

half-maximal effective concentration

ELISA

enzyme-linked immunosorbent assay

G

genogroup

GST

glutathione S-transferase

HAI

hemagglutination inhibition

HIE

human intestinal enteroids

HPI

hours post-infection

HuNoV

Human Norovirus

mAb

monoclonal antibody

IC₅₀

half-maximal inhibitory concentration

ORF

open reading frame

PBMC

peripheral blood mononuclear cells

P domain

protruding domain

PBS

phosphate buffered saline

PGM

porcine gastric mucin

RBC

red blood cells

S domain

shell domain

VLP

virus-like particle