Serologic Correlate of Protection against Norovirus-Induced Gastroenteritis

Abstract

Background

Norovirus infection is the leading cause of acute non-bacterial gastroenteritis. Histoblood group antigens (HBGA) are host susceptibility determinants for Norwalk virus (NV) infection. We hypothesized that antibodies that block NV-HBGA binding are associated with protection from clinical illness following NV exposure.

Methods

We developed an HBGA blocking assay to examine the ability of human serum to block the interaction of NV virus-like particles with H type 1 and H type 3 glycans. Sera from persons experimentally challenged with NV were evaluated.

Results

There was a high correlation between the H type 1 and H type 3 synthetic glycan assays(r=0.977, p<0.0001); the H type 1 assay had higher quantitative sensitivity (p<0.0001). Among 18 infected secretor-positive individuals, blocking titers peaked by day 28 post-challenge and were higher for individuals who did not develop gastroenteritis than for those who did at days 0,14,28, and 180 (p<0.05 for each). Additionally, 6/6 without gastroenteritis had measurable blocking titers (>25)compared to 2/12 with gastroenteritis (p=0.0015).

Conclusions

Blocking antibodies correlate with protection against clinical NV gastroenteritis. This knowledge will help guide the evaluation of new vaccine strategies, and elucidation of the nature of immunity to the virus.

Introduction

Human noroviruses are the leading cause of epidemic non-bacterial gastroenteritis worldwide [1], and account for an estimated 23 million cases in the United States annually [2]. No small animal model exists, and human noroviruses have not successfully been grown in cell culture. Instead, researchers have relied on data from natural outbreaks and experimental human challenge studies to examine host immune responses to infection. Human susceptibility to infection by the prototype genogroup I.1 (GI.1) virus, Norwalk (NV), has been linked in recent studies to the genetically determined expression of histoblood group antigens (HBGAs). Approximately 20% of Northern Europeans and Caucasian Americans lack a functional fucosyltransferase 2 (FUT2) enzyme encoded by the FUT2 gene, and do not express H type 1 or Lewis b (Le^b) antigens on their mucosae or in secretions; this phenotype is termed non-secretor or secretor-negative [3–5]. Blood type B or AB individuals also are less susceptible to NV infection [4]. The development of recombinant expression systems, in which recombinant expression of viral capsid proteins leads to spontaneous self-assembly into virus-like particles (VLPs) [6, 7], has provided reagents to enable study of virus-cell interactions and host immune responses to infection [8, 9]. These studies have shown that norovirus VLPs interact with a variety of HBGAs, including A, B, H types1, 2, and 3, Le^y, Le^x, and Le^b [10–14].

Previous volunteer studies have yielded conflicting evidence of a protective immune response to NV infection. Although studies showed short-term protection to experimental infection [15, 16], long-term immunity has been difficult to elucidate. When a group of volunteers were experimentally re-challenged with NV 27–42 months after the initial challenge, they were not protected against illness despite having developed serum antibodies against NV [17]. Another report showed that high levels of pre-existing serum NV antibody did not protect against infection [18]. These early studies were limited by the use of ELISA to measure NV-specific binding antibody levels, as methods were not available to measure NV neutralizing antibodies. More recent studies have used blocking assays as a surrogate for NV virus-serum neutralization [19–22]in which serum antibodies block the binding of NV VLPs to HBGAs. These studies reported that while all the subjects tested had pre-existing anti-NV antibody by ELISA, only 20–30% had pre-existing blocking antibody titers. Following NV challenge 90–100% of the subjects developed blocking titers [19, 20]. Although these studies evaluated HBGA blocking titers in samples from experimental and natural infections, none tested for a correlation between blocking titers and clinical outcomes. The purposes of this study were to optimize an assay to measure antibodies that block the binding of VLPs to HBGAs and to determine whether the presence of antibodies correlates with protection against disease.

Materials and Methods

Volunteer study

Norwalk virus challenge studies were conducted from September 2004 to March 2008, as previously described [23]. The clinical protocol was reviewed and approved by the Institutional Review Board at Baylor College of Medicine. After providing informed consent, healthy adults (18–50 years) received an oral inoculation of either live NV at a range of doses from 4.8 to 4800 RT-PCR units, or placebo. Serum samples were collected before inoculation (day 0) and at 2, 7, 14, 28, and 180 days after inoculation. All stool samples were collected for the 21 days following challenge. Clinical signs and symptoms were evaluated every 4 hours after inoculation up to 96 hours. NV infection was defined as excretion of virus in stool (detection of antigen or virus by ELISA or RT-PCR, respectively) or a ≥ 4-fold increase in serum antibody titer by ELISA (pre-inoculation to 28 days post-inoculation). Viral gastroenteritis was defined as either 1 episode of vomiting plus 1 other symptom (abdominal cramps/pain, nausea, bloating, loose feces, fever ≥ 37.6 °C, myalgia, or headache), or moderate diarrhea alone (>200 g watery feces)for any continuous 24-hour period.

The intent of the study was to enroll secretor positive individuals based on previous observations that nonsecretors were universally resistant to experimental infection with NV [3, 5]. The presence of HBGAs in saliva was determined by detection of A, B, Lewis a (Le^a) or b (Le^b) glycans in saliva by ELISA using monoclonal antibodies against A (Immucor, Houston, TX), B (Immucor), Le^a (Immucor), and Le^b (Immucor). Persons who had A, B or Le^b antigens in their saliva were identified as secretor positive, while persons who had no antigens detected or only Le^a antigens were excluded as secretor negative or secretor status unknown (no reliable antibody against H type 1 was then available). After all subjects were enrolled and all study procedures had been completed, the antibody against Le^b was found to cross-react to Le^a, falsely identifying secretor negative (expressing only Le^a) persons as secretor positive. A second anti-Le^b monoclonal antibody (Covance, Princeton, NJ) was Le^b-specific using saliva samples from persons with known FUT2 genotypes, and UEA-1 lectin (Sigma-Aldrich, St Louis, MO) was used to identify the presence of H type 1 glycans. These reagents were used to reanalyze saliva from enrolled subjects, and several were found to be secretor negative (only Le^a antigens were present in saliva).

HBGA blocking assays

Blocking assays to measure the ability of serum antibodies to inhibit NV VLP binding to H type 1 or H type 3 synthetic carbohydrates were developed and optimized. All reagents and test samples were diluted in 0.1M sodium phosphate (pH 6.4) buffer with 0.25% fatty-acid free Bovine Serum Albumin (BSA) (Sigma-Aldrich, St Louis, MO), and all assay volumes were 100 μL. NV VLPs (0.32 μg/mL) for the blocking assays were produced using a baculovirus expression system as described previously [6], were incubated with an equal volume of serum that had been serially two-fold diluted from the starting dilution(1:25) , for 1 h at 4°C in LoBind 1.5 mL tubes (Eppendorf, Hauppage, NY). Neutravidin-coated, 96-well microtiter plates (Pierce Thermofisher Scientific, Rockford, IL) were washed and then coated with 2.5 μ/mL of either synthetic polyvalent Le^d (H type 1)-PAA-biotin or polyvalent (H type 3)-PAA-biotin (Glycotech Corp., Gaithersburg, MD) for 1 h at 22°C. Plates were washed 6 times between each incubation step with 0.1M sodium phosphate(pH 6.4) buffer. The sera-VLP solutions were added and incubated at 4 °C for 2 h. Plates were washed, then NV-specific rabbit polyclonal sera (1:5000) was incubated for 1 h at 4°C, washed, and followed by horseradish peroxidase-conjugated, goat anti-rabbit IgG (1:5000) (Sigma) for 1 h at 4°C. The color was then developed by adding tetramethylbenzidine peroxidase liquid substrate (Kirkegaard & Perry Laboratory, Gaithersburg, MD) and stopped after 10 minutes incubation at 22°C by adding 1M phosphoric acid. Optical density was measured at 450 nm using the SpectraMax M5 (Molecular Devices, Sunnyvale, CA) plate reader. Blank wells were incubated with buffer instead of serum/VLP and VLP binding to carbohydrate in the absence of a serum sample was used as a positive control. Results were rejected if OD values for the positive control were outside the range of 0.7 to 1.3. Fifty-percent blocking titers (BT50), defined as the titer at which OD readings (after subtraction of the blank) were 50% of the positive control, were determined for each sample. A value of 12.5 was assigned to samples with a BT50 less than 25. A blocking control serum sample was used as another control, with plates rejected if the BT50 for the blocking control was greater than one dilution above or below the known BT50.

An assay to confirm the specificity of the blocking was performed using the same protocol for the blocking assay with the following exceptions: after coating with carbohydrate, sera were incubated directly on the plate without first pre-incubating with VLP. After washing, VLPs were then incubated on the plate and detected as for the blocking assay.

ELISA to detect NV-specific antibodies

Norwalk virus specific serum antibody was detected by ELISA using NV VLPs, as previously described [24].

qRT-PCR for NV quantitation

The concentration of NV genome was determined for each stool sample by qRT-PCR as described previously [23]. Total virus shedding was calculated by adding the product of fecal virus concentration and stool weight for each individual.

Statistical analyses

Pearson correlation was used to compare the H type 1 and H type 3 assays. Fifty percent blocking titers (BT50s) were converted to log10 values to calculate geometric mean titers (GMTs), and Wilcoxon signed-rank test was used for comparison of non-parametric data. Mann-Whitney test was used to compare BT50s between groups. Fisher’s exact test was used to compare the presence of measurable blocking titers per group. Fecal virus concentrations for those with and without blocking titers were compared using student’s t-test. All p values reported are 2-sided.

Results

Development of H type 1 blocking assay

Noroviruses and recombinant NV VLPs bind to several HBGAs that are expressed on the surface of human cells including the epithelium of the gastrointestinal tract. While the HBGA H type 3 is expressed within enterocyte mucosa, it is not expressed on the surface of enterocytes; in contrast, H type 1 is expressed on the surface of enterocytes [25]. Also, H type 1 shows a stronger level of binding to NV VLPs than H type 3 [22]. Due to its specific expression in the intestinal tract, its strong affinity for NV VLPs and the observation that binding of the NV VLPs to the gastroduodenal junction correlates with the presence of H type 1 antigen and not with that of H type 3 [25, 26], we chose to focus on development of an H type 1 blocking assay.

The synthetic H type 1 assay performance was affected by several factors including pH, buffer ionic strength, incubation temperature, NV VLP protein concentration, and HBGA concentration. We found that detection of VLP binding to H type 1 was optimal using a 0.1M sodium phosphate buffer, at a pH of exactly 6.4, as the signal:noise ratio was lower for other pH levels. At pH <6.4 background binding was high, and at pH >6.4 detection of binding was poor. The VLP-HBGA interaction was most stable in 0.25% BSA in 0.1M sodium phosphate buffer and less stable in PBS. Signals were higher when incubation steps were carried out at 4 C compared to room temperature or 37 C. The optimal concentration of NV VLPs was determined, and the concentration with the highest quantitative sensitivity for blocking that yielded an OD signal of approximately 1.0 was determined to be 0.16 μg/mL protein. The optimal concentration of H type 1 used for this amount of VLP was determined to be 2.5 μg/mL, with lower concentrations yielding unacceptable inter-well variability. Conditions for the H type 3 assay were not as temperature or pH dependent, but both assays were carried out in the same conditions for ease of comparison.

Human serum can contain anti-glycan antibodies. Huflejt et al. found measurable antibodies recognizing H type 1 and H type 3 carbohydrates in approximately 50% of 103 serum samples from healthy persons [27]. To determine whether observed HBGA blocking activity was mediated by direct binding of serum antibodies to glycans, we developed a control assay to measure blocking activity generated by preincubation of the capture glycan with serum prior to addition of NV VLPs. None of the samples tested showed anti-glycan blocking activity. Thus, any measurable blocking effect was the result of antibody binding to the VLP and preventing it from interacting with its carbohydrate ligand.

Comparison of the H type 1 and H type 3 blocking assays

Once optimized, we compared the performance of the H type 3 blocking assay to the H type 1 assay. There was a high correlation between the results from 52 serum samples tested by both assays under the same conditions(r=0.977, p<0.0001, Figure 1a). However, in most cases (50/52, 96%) the H type 1 assay was more quantitatively sensitive, achieving higher blocking titers than the H type 3 assay for matched serum samples (p<0.0001, Wilcoxon signed-rank test, Figure 1b). In two instances (4%) the H type 1 assay detected blocking activity while the H type 3 assay did not.

Figure 1. Comparison of H type 1 and H type 3 blocking assays.

Fifty-two samples representing different time-points were tested using both HBGA H types. (A) H type 1 BT50s are correlated with H type 3 BT50s (r = 0.977, Spearman correlation, p<0.0001). (B) BT50s for each sample are compared for both H types. Blocking antibody levels using the H type 1 assay were higher than those determined using the H type 3 assay, with 2 exceptions (4%) (p<0.0001, Wilcoxon signed-rank test).

Validation of the H type 1 assay

To validate the assay we tested pre-challenge (day 0) and post-challenge (day 28) sera from human subjects and compared fold-rises in blocking titers to total NV-specific antibody titers. All of the subjects (18/18) who met the definition of infection and had ≥ 4 fold rise in total anti-NV specific ELISA, also had a ≥ 4-fold rise in BT50 (assay sensitivity =100%). Conversely, for those who were not infected (including non-secretors, blood type B or AB, or challenged with placebo), 0/16 had a ≥ 4-fold rise in BT50 (assay specificity = 100%). To further validate the assay, we established several internal controls. First, we imposed an acceptable range of OD values for our positive control (VLP in the absence of sera) of 0.7–1.3. As a positive blocking control, we included a pooled lab serum sample with a known BT50 on each plate and rejected plates in which the BT50 for the blocking control was greater than one dilution above or below the known BT50. None of plates were rejected based on these criteria.

Evaluation of Blocking Titers in Response to NV Challenge

We tested samples from 34 subjects at days 0, 14, 28, and 180 days post-inoculation of virus at a range of doses from 4.8 to 4800 RT-PCR units. Of these, 18 secretor-positive individuals became infected and also had ≥ 4-fold rise in BT50 by day 28 compared to day 0, and the majority (14/18) still had ≥ 4-fold rise in BT50 at day 180 (Table, Figure 2). Blocking antibody titers peaked at day 14 for 6 subjects (33%) and at day 28 for 12 subjects (67%). All of the infected subjects still had detectable blocking titers at day 180. Twelve of the infected subjects met the definition of gastroenteritis; 6 did not. Blocking titers were higher at all time-points for those without gastroenteritis than for those with gastroenteritis, including day 0 (p=0.0024, Mann-Whitney test), day 14 (p=0.0076), day 28 (p=0.0168), and day 180 (p=0.0444).

Table.

HBGA BT50 ANTIBODY RISES (≥ 4-FOLD) AT DIFFERENT TIMES FOLLOWING CHALLENGE

Days after challenge	Infected (Number/Total)			Not Infected (Number/Total)
	No GE	GE	All	Placebo	Non-secretor or type B or type AB	Susceptible, Not infected	All
Day 14	6/6	11/12	17/18	0/5	0/8	0/3	0/16
Day 28	6/6	12/12	18/18	0/5	0/8	0/3	0/16
Day 180	3/6	11/12	14/18	0/5	0/8	0/3	0/16

Figure 2. GMTs for BT50s by H type 1 blocking assay.

Among persons infected with Norwalk virus, BT50 GMTs were significantly higher (asterisk, p<0.05, Mann-Whitney) at each time-point for those with no GE (n=6) compared with those with GE (n=12). Error bars represent 95% confidence intervals. GE = met clinical definition of gastroenteritis, No GE = did not meet definition of gastroenteritis.

Association of pre-existing blocking titers with outcome

Analysis of the day 0 samples revealed that the absence of measurable blocking titers (<25) prior to inoculation was associated with the development of clinical gastroenteritis (Figure 3a). All infected secretor-positive subjects challenged with virus without gastroenteritis (6/6) had measurable blocking titers pre-inoculation, while blocking titers were detected in only 2/12 subjects with gastroenteritis (p=0.0015, Fisher’s exact test). In contrast, there were no significant differences in ELISA antibody titers between subjects with and without gastroenteritis(Figure 3b). These findings were confirmed with the H type 3 assay, with the same level of significance between the gastroenteritis group and the no gastroenteritis group.

Figure 3. Effect of pre-challenge serum BT50 level (H type 1 assay) on development of gastroenteritis.

Pre-challenge serum samples were tested for the 18 infected subjects; 12 met clinical definition of gastroenteritis (GE, triangles), 6 did not meet definition of gastroenteritis (No GE, squares). (A) BT50s, (B) total anti-NV ELISA titers. The two empty triangles in panel A represent detectable BT50s in the GE group, and their respective anti-NV titers are shown in panel B.

Peak fecal viral shedding was higher in persons with no pre-existing serum blocking antibody compared to those with pre-existing blocking antibody (11.2 vs. 10.1 log10 genomic copies/gram, p=0.027, Student’s t-test). Similarly, the total amount of fecal virus shedding post-challenge was higher in persons with no pre-existing serum blocking antibody compared to those with pre-existing blocking antibody (13.4 vs. 12.3 log10 genomic copies, p=0.01, Student’s t-test).

Discussion

This study reports for the first time a correlation between the presence of pre-challenge HBGA blocking antibody titers and protection against clinical gastroenteritis induced by NV infection. Our findings support the hypothesis that HBGA blocking antibodies may be a surrogate method for measuring serum virus-neutralizing antibodies [19]. The blocking assay was used to screen serum samples from human volunteers who were experimentally challenged with NV. We also measured the ability of the serum antibodies to block recombinant NV VLP binding to either H type 1 or H type 3 synthetic glycans, and observed a high correlation between the two assays. We found that blocking titers peaked by day 28 post-challenge and were still elevated at day 180 relative to the pre-challenge levels.

The lack of a reliable cell culture system for the prototype NV has hindered study of specific immunological parameters such as detection of serum virus neutralization antibodies that are proven correlates of protection for many other virus systems [28]. Previous volunteer studies have yielded conflicting observations of protective immune responses to experimental NV infection [15–18]. The ability of serum antibodies to block binding of recombinant NV VLPs to their putative HBGA receptors in vitro has been hypothesized to be a surrogate measure for NV-neutralizing antibodies [19]. Although recent studies have evaluated HBGA blocking titers in serum samples from experimental and natural infections [19–22], none associated blocking titers with clinical outcomes. Our data indicate an association between the presence of blocking titers prior to inoculation and protection against disease (Figure 3). There also is an association between lower virus shedding (peak and total) and presence of blocking activity. We previously reported the association of lower virus shedding with absence of symptoms of gastroenteritis [23].

Many of the subjects tested in our study had measurable pre-existing blocking titers, in contrast to results recently reported by Lindesmith et al. in which few of the pre-inoculation samples tested were able to block 50% binding at the minimum dilution tested (1:400) [20]. Most likely these different observations are due to technical differences that resulted in increased quantitative sensitivity in our assay which enabled the detection of blocking titers starting at a lower serum dilution (1:25). Since there was no mention of either the selection criteria for inclusion of volunteers, or the clinical data for those selected in the Lindesmith et al. study, the reason for their finding a lack of correlation is not clear; however, our results predict that the lack of detectable blocking titers would correlate with a lack of protection from illness. We found that assay pH and temperature were critical factors in order to obtain a sufficient OD value for HBGA-VLP binding. Inclusion of a lab standard blocking control to monitor assay variability was also necessary to accurately interpret and validate the results. Additionally, it is known that some individuals possess anti-HBGA serum antibodies [27, 29] which would cause false positive results in our assay. This concern led us to develop an assay to test for anti-carbohydrate antibodies and allowed us to confidently interpret blocking titers as none of the volunteer samples tested had anti-HBGA antibodies.

There was a strong correlation for NV VLP-HBGA blocking activity of serum antibody between H type 1 and H type 3 glycans as assessed by BT50 levels (Figure 1a). This is expected because NV VLPs bind to terminal fucose and galactose moieties present on both H type 1 and H type 3 [30]. Previous studies have suggested that the H type 1 glycan has a greater affinity for NV VLPs than the H type 3 glycan based on greater in vitro VLP binding to the H type1 glycan [12] and greater inhibition of VLP binding to gastroduodenal tissue sections compared to H type 3 [25]. Our results are consistent with these data and may explain why the H type 1 blocking assay was more quantitatively sensitive in detecting blocking titers than H type 3 (Figure 1b). H type 1 is also the most biologically relevant HBGA, given its specific expression on the surface of the gut mucosa [25]. Despite the advantages of the H type 1 glycan, H type 3 is currently more easily obtained; therefore until H type 1 becomes readily commercially available, the use of H type 3 for blocking assays is a reasonable alternative.

The results of this study highlight a number of issues to be addressed in future studies. First, a larger number of test samples is needed to confirm an association between blocking titers and protection from illness. Second, what is the minimum threshold blocking titer that protects from illness? In this study we evaluated the presence or absence of a blocking titer based on a minimum serum dilution of 1:25. Third, what is the duration of immunity associated with blocking antibodies? We tested samples up to180 days post-challenge, and infection-induced blocking activity was still measurable 180 days after challenge for all infected subjects (Figure 2). Fourth, the interaction of inoculum dose and blocking activity needs to be assessed, as the protective effect of even high blocking titers may be overcome if individuals are exposed to a sufficiently high dose of virus. Fifth, do candidate vaccines against NV induce serum blocking activity, and does this activity correlate with protection from illness? Finally, are blocking antibodies cross-reactive with VLPs of other Norwalk-like virus genotypes?

The immune response to norovirus infection is complex, with some reports indicating that high levels of pre-existing antibodies measured by ELISA do not protect against infection, and long-term immunity is not consistently observed [11, 18, 31–33]. However, NV antibody titers measured by ELISA only detect antibody that binds to VLPs and not functional antibodies therefore it is not surprising that ELISA antibodies do not predict protection against future infection and symptomatic illness. In contrast, we have been able to link NV blocking titers and the development of clinical symptoms following experimental challenge, and thus identify a potential correlate of protection. This knowledge will help guide the development and evaluation of new vaccine strategies, and will help to elucidate the nature of host immunity to the virus.