Use of Zebrafish Embryos To Reproduce Human Norovirus and To Evaluate Human Norovirus Infectivity Decay after UV Treatment

Zebrafish embryos were more efficient and robust than zebrafish larvae in supporting HNoV replication and enabled continuous passaging.

The clinical (stool) sample of HNoV GII.4[P16] (8 log genome copies/g stool) was tested by injecting 3 nL of the virus suspension in the yolk of each larva at 3 days postfertilization (dpf) (same procedure as described previously [18, 19]) and in the yolk of each embryo at 6 h postfertilization (hpf). Using phenol red to simulate the virus suspension, it was shown that the manipulation of embryo yolk injection was much easier with less resistance and thus more time-efficient than the larva yolk injection as shown in the supplemental videos (see Movies S1 and S2 in the supplemental material).

The HNoV loads in the zebrafish embryos/larvae were followed up after injection. At each time point (0 day postinfection [dpi] was taken as immediately after injection), as shown in Fig. 1, 10 injected embryos or larvae were pooled as one sample and subjected to homogenization, RNA extraction, and reverse transcription-quantitative PCR (RT-qPCR) for HNoV GII detection. Consistent with our previous reports (18, 19), when the viruses were injected into the zebrafish larvae (Fig. 1a), low levels of HNoV below 2 log genome copies/10 larvae were recovered at 0 dpi, and a significant increase of virus genome copies (4.4 ± 0.8 log genome copies/10 larvae) was detected at 2 dpi (P < 0.05). The viral loads as detected from the zebrafish larvae started to decrease gradually from 3 dpi. At 5 dpi, the viruses already dropped to levels with no significant difference from those at 0 dpi (P > 0.05).

In comparison, when the viruses were injected into the zebrafish embryos (Fig. 1b), higher levels of HNoV above 3 log genome copies/10 embryos were recovered at 0 dpi. This was possibly due to less virus loss during the injection of embryos than of larvae, which had more rigid structures to prick (Movies S1 and S2). Significant virus replication was already noticed from 1 dpi and lasted to 6 dpi, which was the last day as tested (P < 0.05). The peak levels of virus detection were reached at 2 dpi and lasted for 4 days (no significance from 2 dpi to 5 dpi [P > 0.05]). The virus levels detected at 3 dpi had the smallest variation (7.7 ± 0.2 log genome copies/10 embryos), and thus, 3 dpi (4.2 ± 0.4 log genome copies/10 embryos; increase in comparison with 0 dpi) was selected as the time point to evaluate the virus infectivity in the following studies.

HNoVs are genetically diverse RNA viruses with 10 genogroups (GI to GX) and 49 genotypes (20). Next to GII.4[P16], we have also tested two other strains, GII.2[P16] and GII.17[P31], in this study. All three strains belong to the globally prevalent genotypes and/or P-types in recent years (21). When the same protocol was applied to HNoV GII.2[P16] and GII.17[P31] (both with ~8 log genome copies/g stool), results comparable to those of HNoV GII.4[P16] were obtained at 3 dpi (P > 0.05) (Fig. 2). In addition, the high levels of virus replication enabled continuous passaging for all three tested strains up to four passages as tested (Fig. 2). Previously, the zebrafish larvae allowed successful passaging only up to the second passage (16). In this study, with the use of zebrafish embryos, significant increases of virus titers were obtained for HNoV GII.4[P16] (from passage 4, P < 0.05) and GII.17[P31] (from passage 3, P < 0.05). For HNoV GII.2[P16], although no significance was obtained (P > 0.05), a trend of increase in virus titers along with the passaging was also observed (Fig. 2). In fact, the passaging was still able to continue (we have continued with HNoV GII.4[P16] until passage 8 [data not shown]). Therefore, in the future, we anticipate this tool could be used to study the adaptive mutation of HNoVs along with virus passaging, which has been reported over multiple viruses (22–25).

For both zebrafish larva and embryo models, no visual pathological change was observed after virus injection and during virus replication within the tested periods as shown in Fig. 1. Since the expression of innate immune response-related genes, including ifn, mx, and rsad2/viperin, was observed to be significantly increased in the HNoV-infected zebrafish larvae (16, 18), the higher replication levels of HNoVs in zebrafish embryos than in larvae might be due to the less efficient antiviral innate immune response in the zebrafish embryos than in the larvae. We also hypothesize that the incomplete organ formation and less development of tight biological barriers between organs and tissues might have allowed enhanced dissemination and/or replication of the viruses, which remains to be investigated in the future. On the other hand, since not every organ is fully functional and not every biological barrier is fully in place by 3 dpf, there might be limitations using the zebrafish embryos/larvae to study the HNoV tropism, pathogenesis, and host-pathogen interactions directly as an in vivo animal model.

The zebrafish embryo tool enabled direct profiling of HNoV binding to human HBGAs.

The entry of HNoV in host cells has been demonstrated to be closely correlated with histo-blood group antigens (HBGAs) (26, 27). HBGAs are a group of complex terminal carbohydrates that can be found as soluble antigens in saliva and are expressed on the mucosal epithelium of the digestive tract (28). Therefore, human clinical samples (stool samples) are typically not used to study the binding between the viruses and HBGAs, largely due to the presence of various types and abundances of HBGAs in the stool samples. Preparation of baculovirus-expressed virus-like particles (VLPs), on the other hand, has high cost and technical issues, including the nonuniformity in size during assembly and the particle dissociation during storage that may happen as reported previously (29). In addition, gene cloning and baculovirus expression might introduce certain discrepancies between genuine virus particles and VLPs, such as the inconsistency regarding the involvement of the minor capsid protein (VP2) (30). Here, as the HNoVs produced by zebrafish embryos have both high titers and clean background, the binding patterns of the three tested HNoV strains were characterized by a saliva-binding RT-qPCR to compare the stool samples and zebrafish embryo-amplified samples (passage 4).

Four saliva samples from different individuals with representative HBGA phenotypes were used, including two secretors (S1 and S2) and two nonsecretors (NS1 and NS2) based on the detection of Fucα1-2Gal-R. According to the ABO(H) typing, S1 was type A and S2 was type B. Both S1 and S2 were positive for Lewis b and y antigens and negative for Lewis a and x antigens. For nonsecretors’ saliva, both NS1 and NS2 were negative for all of the ABO(H) antigens. NS1 was positive for all four Lewis antigens, and NS2 was positive for only Lewis a antigen (Table 1).

The binding levels of the HNoVs and HBGAs in saliva were characterized by a “binding indicator” defined as the logarithms of the bound virus ratios (binding indicator = log [genome copies of HNoVs bound on saliva-coated plates/genome copies of HNoVs in the samples added to the plates]) as recorded in Fig. 3. As expected, the HNoVs amplified by zebrafish embryos indeed showed clearer binding patterns than the viruses in stool samples. For all of the three strains tested in this study, HNoVs GII.4[P16], GII.2[P16], and GII.17[P31], viruses from zebrafish embryos showed significantly higher levels of binding than viruses from stool samples to the secretors’ saliva S1 and S2 (P < 0.05), while no such significance was observed in the binding to the nonsecretors’ saliva NS1 and NS2 (P > 0.05), except GII.4[P16] to NS1 (P < 0.05). In the meantime, the levels of binding of all three strains from zebrafish embryos to the secretors’ saliva S1 and S2 were also significantly higher than to the nonsecretors’ saliva NS1 and NS2 (P < 0.05). These results were in high accordance with the clinical results reported by researchers from multiple countries and regions in the world that the nonsecretor individuals, who express only Lewis antigens, are less susceptible to HNoV infection (26). Therefore, this HNoV replication tool with the use of zebrafish embryos supplied an efficient and straightforward way to investigate the patterns of binding of the HNoV strains to HBGA without the need of preparing VLPs.

The zebrafish embryo tool enabled evaluation of HNoV inactivation, identifying up to a 2-log reduction.

For viruses well adapted to cell culture, the virus infectivity titer is usually determined by plaque assay or 50% cell culture infectious dose (TCID₅₀) endpoint dilution assay, both of which remain challenging for HNoVs. The current reports evaluating HNoV inactivation efficacy with the use of HIEs were mainly for (qualitative) validation purposes, confirming that a certain treatment could inactivate the viruses to noninfectious levels (31–33). Previously, with the use of zebrafish larvae, we managed to determine the titer of HNoV as TCID₅₀ units, but only an 0.8-log virus reduction by heat treatment was able to be identified (19). On one hand, this was due to the limited virus titer in the stool samples. On the other hand, stool samples had high complexity and diversity in the matrix and thus needed to be diluted to minimize the matrix effect. Previously, a protective effect on the viruses against inactivation treatment was reported in the presence of organic materials such as proteins and lipids (34) as well as HBGA-like components, which have specific binding affinities with HNoVs (35). In comparison, as the zebrafish embryos generated viruses with higher titers and consistent background by continued passaging, this new tool enabled a larger scale of log reductions to be identified.

In this study, a dilution series of the three HNoV strains GII.4[P16], GII.2[P16], and GII.17[P31] produced by zebrafish embryos (passage 4, adjusted to ~8 log genome copies/mL) was first tested in parallel. Since each data point was from a pool of 10 zebrafish embryos, when the viruses were 10 and 100 times diluted (~30 and 3 genome copies injected in each embryo, respectively), the virus replication showed large variations between of the biological replicates. All three tested virus strains became noninfectious consistently when they were 1,000 times diluted (<1 genome copies injected in each embryo) (Fig. 4a). Therefore, when the same samples (produced by zebrafish embryos, passage 4, adjusted to ~8 log genome copies/mL) were inactivated and showed consistently no infectivity after the treatment, we were able to make a conservative estimation that the treatment was able to inactivate the HNoVs by at least 2 logs. In comparison with our previous study (19), which was very laborious and in which each data point was generated from 240 zebrafish larvae, this system enabled a more effective evaluation of HNoV inactivation.

In order to demonstrate this concept, in the following experiments, we measured the inactivation of the three HNoV strains with a series of UV irradiation doses at 254 nm (UV₂₅₄, from 1.5 mJ/cm² to 7.5 mJ/cm²) and showed that a dose of 6 mJ/cm² UV₂₅₄ was able to induce a >2-log HNoV infectivity reduction for all three strains tested (Fig. 4b). UV₂₅₄ is a commonly used disinfection process in water treatment as well as for environmental surfaces including those in food processing facilities. Researchers have been dedicating numerous efforts in evaluating the efficacy of UV on enteric viruses until very recent years (36–38). In 2020, Rockey et al. (38) reported a genome-wide PCR-based approach evaluating UV disinfection of HNoVs and constructed a linear inactivation curve with an inactivation constant of 0.27 cm²/mJ (thus, a 1.62-log reduction after 6 mJ/cm² UV₂₅₄ treatment). As discussed by the authors, this result was a significant departure from previous reports of 0- to 1.5-log inactivation after UV₂₅₄ doses as high as 300 mJ/cm², which relied on short regions of the genome (39–41). As a further improvement, in this study, by measuring the virus infectivity directly, we revealed that HNoVs were not as UV resistant as previously predicted by PCR-based methods (>2-log reduction after 6-mJ/cm² UV₂₅₄ treatment for HNoV GII.4[P16], GII.2[P16], and GII.17[P31]). This result also suggests that HNoVs are generally more UV susceptible than multiple enteric viruses and commonly used HNoV surrogates as tested before (0.84-log reduction for bacteriophage MS2, 1.86-log reduction for echovirus 12 [EV12], 1.92-log reduction for murine norovirus [MNV] [38], 1.08-log reduction for Tulane virus [TV], and 0.50-log reduction for rotavirus [RV] [36] after 6-mJ/cm² UV₂₅₄ treatment, calculated based on the linear inactivation curves as reported).

Conclusions.

In summary, the zebrafish embryo tool as reported in this study represents an essential improvement in the HNoV replication method. With high efficiency and robustness, this tool was able to strengthen HNoV-related research in a versatile way. In a pathogenesis study, the zebrafish embryo-generated viruses had clear patterns of binding to HBGAs in human saliva. In a disinfection study, this tool was used to show that a dose of 6 mJ/cm² UV₂₅₄ was able induce a >2-log HNoV infectivity reduction for all three strains tested, suggesting that HNoVs were more UV susceptible than multiple enteric viruses and commonly used HNoV surrogates as tested before. In the future, it is expected that this tool will not only benefit epidemiological research on HNoV but also be used to generate HNoV inactivation parameters which are highly needed by the water treatment and food industries.

HNoVs from stool samples.

HNoVs in stool samples were kindly provided by the Molecular Laboratory, Department of Molecular Pathology, Singapore General Hospital, and genotyped in our laboratory at National University of Singapore over the B-C genome region (42). In order to prepare the virus suspension for injection, an aliquot of 100 mg of the stool sample was suspended in 1 mL of sterile phosphate-buffered saline (PBS), thoroughly vortexed, and centrifuged for 5 min at 9,000 × g; the supernatant was harvested and stored at −80°C until further use.

Zebrafish maintenance and husbandry.

Wild-type adult zebrafish (Danio rerio) were maintained in the aquatic facility of the National University of Singapore with water temperature at 28°C and a 14-/10-h light/dark cycle. Fertilized eggs were collected from adults placed in mating cages and kept in petri dishes containing embryo medium E3 (5 mM NaCl, 0.17 mM KCl, 0.33 mM CaCl₂, 0.33 mM MgSO₄). All zebrafish experiments were performed in compliance with the Institutional Animal Care and Use Committee guidelines, National University of Singapore.

Microinjection of zebrafish larvae and embryos in the yolk.

Seventy-two-hour-postfertilization larvae were first anaesthetized for 2 to 3 min in embryo medium E3 containing 0.2 mg/mL Tricaine (Sigma-Aldrich, USA), after which, they were aligned on the agarose surface with their yolk facing the microinjection needle. Three nanoliters of the viral suspension was injected into the yolk of each larva. For embryo injection, fertilized eggs were collected and washed with embryo medium E3 three times. Prior to injection, embryos were inspected under an inverted microscope to ensure that the embryos had not developed past the spherical stage (~4 hpf). Thereafter, the embryos were transferred to a petri dish with grooves of a mold imprint (6 rows of V-shaped grooves) in 1.5% agarose. Three nanoliters of the viral suspension was injected into the yolk of each embryo with a pulled borosilicate glass capillary needle. After the injection, the zebrafish (either the embryos or larvae) were transferred to a petri dish containing E3 and further maintained at 29.5°C. Every day postinfection, the embryos or larvae were observed to ensure that they were properly developing; dead embryos or larvae were removed, and the medium was changed daily. Videos demonstrating the injection procedures were filmed using 0.05% phenol red (Sigma-Aldrich, USA) solution to simulate the virus suspensions with a Nikon motorized fluorescence stereomicroscope (SMZ25-FL) equipped with a Nikon digital camera (DS-10).

Harvesting of zebrafish larvae/embryos for analysis.

The zebrafish larvae/embryos were harvested with 10 larvae/embryos pooled as one sample. After being euthanized on ice, the larvae/embryos were mixed with 100 μL PBS in 1.5-mL tubes and homogenized with 3 cycles of 15 s at 6,500 rpm with rest intervals of 60 s by a FastPrep 24-5G tissue-cell homogenizer (MP Biomedicals, USA). The homogenates were clarified by centrifugation at 9,000 × g for 5 min, and the supernatants were used as the virus suspension for RT-qPCR analysis or further investigation (passaging of viruses by being injected into new zebrafish embryos, profiling of capabilities of binding to HBGAs, and UV inactivation study).

RNA extraction and RT-qPCR for HNoV detection.

RNA was extracted using an RNeasy minikit (Qiagen, Germany) following the manufacturer’s protocol.

RT-qPCR analyses of NoV GII were carried out using the GoTaq Probe 1-Step RT-qPCR system (Promega, USA). Primers and the 6-carboxyfluorescein (FAM)–6-carboxytetramethylrhodamine (TAMRA)-labeled probe of HNoV GII were constructed according to ISO 15216-1:2017. Sequences were as follows: forward primer QNIF2, 5′-ATGTTCAGRTGGATGAGRTTCTCWGA-3′; reverse primer G2SKR, 5′-TCGACGCCATCTTCATTCACA-3′; probe QNIFs, 5′-FAM-AGCACGTGGGAGGGCGATCG-3′ TAMRA. Cycling conditions were 45°C for 15 min and then 95°C for 10 min, followed by 40 cycles with 95°C for 15 s and 60°C for 30 s in each cycle. Cycle threshold (C_T) values were determined during RT-qPCR analysis using a StepOne system (Applied Biosystems, USA).

Double-stranded DNA (dsDNA) containing the specific primer-probe binding sites was synthesized for NoV GII and cloned into the pGEM-T vector (Promega), resulting in the NoV-GII plasmids. The plasmids with inserts were purified by using a Plasmid Midi kit (Qiagen). Plasmid concentration was determined by photospectroscopy at 260 nm using the BioDrop Duo spectrophotometer (BioDrop, United Kingdom). Tenfold serial dilutions ranging from 5 × 10⁶ to 5 copies of all positive-control plasmids were used to prepare a standard curve and enumerate the NoV GII.

Detection of Fucα1-2Gal and HBGA phenotyping in saliva samples.

Human saliva samples were collected from healthy adults under the ethical approval of National University of Singapore Institutional Review Board (NUS-IRB; reference code N-20-035) and characterized as described previously (43). Specifically, the saliva samples were boiled at 95°C for 10 min, followed by centrifugation at 10,000 × g for 5 min. The supernatant was collected and diluted at 1:100 with carbonate-bicarbonate buffer (pH 9.6; CBS) and was used to coat 96-well microtiter plate (100 μL/well) at 4°C overnight. The coated plate was washed three times with PBS containing 0.05% Tween 20 (PBS-T) and blocked with 5% nonfat dried milk (Blotto, 200 μL/well) at 37°C for 1 h. In order to detect Fucα1-2Gal-R, which is specifically present in secretor but not in nonsecretor saliva, horseradish peroxidase (HRP)-conjugated Ulex europaeus agglutinin (UEA-I; Sigma-Aldrich, USA) diluted at 1:3,200 (starting concentration, 1 mg/mL) was added and incubated for 1 h at 37°C. After three washes with PBS-T, the presence of Fucα1-2Gal-R was detected with a TMB (3,3′,5,5′-tetramethylbenzidine) kit (Sigma-Aldrich, USA), and the signal intensities (optical density at 450 nm [OD₄₅₀]) were read with a Multiskan Sky plate reader (Thermo Fisher Scientific, China). Similarly, for the determination of the HBGA phenotype of the saliva samples, to the blocked wells were added 1:200-diluted HBGA monoclonal antibody (MAb) anti-A (BG 2), anti-B (BG 3), anti-H type 1 (BG 4), anti-Lewis a (BG 5), anti-Lewis b (BG 6), anti-Lewis x (BG 7), and anti-Lewis y (BG 8) (Covance, Emeryville, CA, USA), and the mixtures were then incubated for 1 h at 37°C. After three washes with PBS-T, the HBGA phenotype was detected with horseradish peroxidase conjugated with goat anti-mouse IgG (1:1,500; Sigma-Aldrich, USA) for type A, H, and Lewis a and goat anti-mouse IgM (1:1,500; Sigma-Aldrich, USA) for type B and Lewis b, x, and y. Horseradish peroxidase activity was detected with a TMB kit, and the OD₄₅₀ was read with the Multiskan Sky plate reader. The cutoff value was taken as two times the absorbance value of the negative controls as previously performed by Nordgren and Svensson (26).

HNoV binding to saliva.

A microtiter plate was coated with saliva samples at 1:100 with CBS (100 μL/well) at 4°C overnight. The coated wells were washed three times with PBS-T and blocked with 5% nonfat dried milk at 37°C for 1 h. Thereafter, HNoV samples from human stool or zebrafish embryo-amplified virus suspensions prepared as described above were added at 10⁶ virus genome copies/well and incubated for 1 h at 37°C. After three washes with PBS-T, direct lysis was performed by adding lysis buffer from the RNeasy minikit (Qiagen) to individual wells. The subsequent RNA extraction procedures were conducted according to the manufacturer’s protocol. RT-qPCR analysis of the bound viruses was performed as described above.

UV₂₅₄ treatment of viruses.

The UV treatment was performed in a biosafety cabinet (ESCO, Singapore; LA2-4A1-E) with low-pressure mercury lamps producing mainly 254-nm UV light (UV-30A). The viral suspensions prepared as described above (zebrafish embryos amplified, passage 4) were spotted into a sterile petri dish (20 μL as one droplet) and exposed to the UV light until reaching a dose of 1.5, 3.0, 4.5, 6.0, and 7.5 mJ/cm², respectively. The UV irradiance values were measured by an optical radiometer (UVX radiometer; Analytik Jena, USA). After treatment, the samples were recovered by pipetting and homogenizing for several times and then used for injection into zebrafish embryos directly.

Statistical analysis.

Statistical analyses were performed using the software SPSS for Windows, version 22.0. A paired t test was used for data with two groups, and one-way analysis of variance was used for data with more than two groups. Differences were considered significant when the P value was <0.05.