Detection of Murine Norovirus on Fresh Produce Through a CRISPR/Cas13a RNase-Based Capsid Integrity Assay

Abstract

Standard food detection methods do not distinguish between infectious and non-infectious human norovirus leading to uncertainty in the management of a norovirus positive food sample. These methods also require expensive RT-qPCR-based equipment and reagents. In contrast, CRISPR-based, compared to RT-qPCR-based, detection methods are generally less expensive and yield similar sensitivity and specificity. Our goal was to detect norovirus with an intact capsid, a proxy for infectivity, through a CRISPR–Cas13a-based detection method together with an RNase-capsid integrity assay. We termed this assay: Foodborne RNA-virus Enzymatic Sensing for High-throughput on fresh produce (CRISPR FRESH) reflecting its potential to detect infectious or potentially infectious virus particles. Our CRISPR FRESH method detected murine norovirus (MNV-1), with an intact capsid, at a limit of detection of 2.59 log10 gc/25 g (5 gc/rx). This method did not cross-react with other targets (synthetic DNA targets for hepatitis A virus; human norovirus GI, GII; rotavirus). Compared with RT-qPCR, CRISPR FRESH showed an increased sensitivity when detecting low copy numbers of RNase-pre-treated MNV-1 in lettuce and blueberries samples. Viral detection with the RT-qPCR assay is quantifiable while the CRISPR assay is present/absent. This report describes a CRISPR-based detection of potentially infectious viruses in food samples.

Supplementary Information

The online version contains supplementary material available at 10.1007/s12560-025-09651-5.

Introduction

Human norovirus is the most common cause of acute gastroenteritis worldwide, with an estimated 685 million cases annually (Kirk et al., 2015). Noroviruses are non-enveloped, single-stranded RNA viruses that infect a wide range of mammalian host species, including humans (Chhabra et al., 2019; De Graaf et al., 2016). Based on phylogenetic clustering of the complete VP1 amino acid sequence, noroviruses are classified into ten genogroups, GI-GX (GI, GII, and GIV cause human infections), and 49 genotypes (Chhabra et al., 2019).

Human norovirus is considered an important foodborne pathogen in outbreaks associated with produce including fruits (e.g., berries) and vegetables (e.g., leafy greens), which are often consumed raw or minimally processed (Bosch et al., 2018; Yeargin & Gibson, 2019). Contamination with human norovirus, due to poor hygiene, may occur at any point of the “farm to fork” supply chain, including handling during harvest, postharvest and food preparation by handlers. Furthermore, the ID50 of only 18 infectious viral particles requires sensitive detection methods of low viral copies in food samples (El-Senousy et al., 2013; León-Félix et al., 2010; Teunis et al., 2008). Based on viral RNA or surrogate viruses, human norovirus is considered to be stable under different conditions used in the food industry such as high and low temperatures, extreme pH, pressure conditions, and chemical disinfectants; this stability also contributes to its transmission (Ausar et al., 2006; Butot et al., 2008; Park et al., 2010a; Shoemaker et al., 2010). There is a need for laboratory methods that can detect infectious viruses in food, which can cause infection and illness, rather than inactive viral fragments that persist in the environment, which do not cause infection and illness (Knight et al., 2013; Sobolik et al., 2021). These methods will help in accurately evaluating the risk of foodborne norovirus infections.

Among viral detection methods for food samples, RT-qPCR is the gold standard. However, because RT-qPCR detects only genetic material, it does not discriminate between norovirus genetic material in an intact capsid, a non-intact capsid, or from free environmentally persistent RNA (Diez-Valcarce et al., 2011; Lowther et al., 2019). Human norovirus cell culture systems can detect infectious norovirus but lack replication efficiency and cannot be applied to all norovirus strains (Cheng et al., 2024). Thus, several groups have developed indirect detection methods of infectivity, including molecular approaches to detect intact viral capsids, a suggested proxy of infectivity (Moore et al., 2017). Some of these approaches include ligand binding-based assays using synthetic human norovirus histo-blood group antigen (HBGA) or porcine gastric mucin, and capsid and genome integrity assays such as RT-qPCR preceded by RNase and/or proteinase and the use of photo activatable intercalating agents like ethidium monoazide and propidium monoazide (Liu & Moore, 2020; Moore et al., 2017; Nuanualsuwan & Cliver, 2002; Rudi et al., 2005).

In addition to RT-qPCR, other detection methods of viral genetic material include isothermal amplification methods such isothermal Recombinase Polymerase Amplification (RPA), LAMP and NASBA, whole-genome or amplicon-based sequencing methods, and droplet digital PCR (ddPCR) (Fukuda et al., 2006; Ma et al., 2018a; Moore & Jaykus, 2017; Rexin et al., 2024). Although these technologies offer high specificity and sensitivity, like RT-qPCR, they cannot discriminate between infectious and non-infectious viruses and some of them are expensive (e.g., sophisticated equipment, trained personnel) and time consuming.

The Clustered Regularly Interspaced Short Palindromic Repeats (CRISPR) system, when coupled with Cas enzymes, offers an alternative to traditional RT-qPCR methods in detecting pathogens (Freije & Sabeti, 2021). Further, CRISPR is traditionally coupled with isothermal pre-amplification methods, like RPA (as used in this study) to increase the amount of genetic material for detection (Gootenberg et al., 2017). These CRISPR Cas systems are not only comparable in sensitivity and specificity to RT-qPCR but also boast advantages such as reduced assay times, lowered costs, and the option to replace costly RT-qPCR equipment with equipment less expensive than a traditional thermocycler (Palaz et al., 2021; Sashital, 2018; Wang et al., 2020; Williams et al., 2021). For example, CRISPR-based assays require equipment, less expensive than a thermocycler, to control temperature (e.g., thermoblocks) for assay reactions and to read assay results (e.g., fluorometers, microplate readers). Moreover, they can be integrated into laboratories currently performing RT-qPCR assays. The application of CRISPR with the Cas13a enzyme has demonstrated the capability to identify as few as 5 genomic copies per reaction on 50 µl of norovirus GII.4 from stool samples within 2 h at a temperature of 39 °C (Duan et al., 2022). On foods, there have been CRISPR-Cas-based assays to detect E. coli O157:H7, Listeria monocytogenes, Salmonella Typhimurium, Staphylococcus aureus, and Vibrio parahaemolyticus (Hadi et al., 2023; Lu et al., 2022), and recently, for human norovirus and rotavirus on lettuce and strawberries (Le et al., 2025). Despite these advances, the detection of viral RNA by CRISPR does not directly indicate viral infectivity. There is a need for methods that are as sensitive and specific as RT-qPCR, are less expensive and laborious, and can distinguish infectious from non-infectious norovirus (Liu & Moore, 2020; Raymond et al., 2021). Such methods would enhance the food industry’s access to viral detection tools. Given this need, the aim of this study was to detect norovirus with an intact capsid, a suggested proxy for infectivity, through a CRISPR–Cas13a-based detection method in conjunction with an RNase-capsid integrity assay. We termed this assay: Foodborne RNA-virus Enzymatic Sensing for High-throughput on fresh produce (CRISPR FRESH) designed to indicate potentially infectious viruses on fresh produce. We evaluated this system on murine norovirus (MNV-1), as a human norovirus surrogate, on fresh produce, including blueberries and lettuce.

Materials and Methods

Viral Stock

Murine norovirus CW1 strain (MNV-1, ATCC^® VR-1937™) was used as a human norovirus surrogate. MNV-1 was propagated in RAW 264.7 cells (ATCC^® TIB-71™) maintained in Dulbecco's Modified Eagle's Medium” (DMEM) (Sigma-Aldrich^® St. Louis, MO) supplemented with 10% FBS, penicillin (100 U/ml), and streptomycin (100 µg/ml) (Sigma-Aldrich^®) at 37 °C with 5% CO₂. Viral stocks were obtained after four consecutive rounds of virus infection in confluent cells within one passage (75-cm² culture flasks). Cells were scraped and the suspension was subjected to three freeze–thaw cycles followed by clarification using low-speed centrifugation of 1460×g for 15 min. Further, cells were adjusted and infected at a multiplicity of infection (MOI) of 0.1. The viral aliquots were stored at − 80 °C. For titration, viral RNA was extracted and quantified by RT-qPCR as described below.

Viral RNA Extraction, Reverse Transcription, and RT-qPCR

All viral RNA extractions were conducted on 140 µl of sample, post elution and RNase/no-RNase treatment, using the QIAamp viral RNA mini kit^® (Qiagen^©, CA, USA), following the methodology described by the supplier. Viral RNA was eluted in 60 µl of Qiagen^© AVE buffer. The two-step RT-qPCR was used for viral detection. Reverse transcription (cDNA) was carried out following supplier instructions from the High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems™, CA, USA) at a final volume of 20 µl. The cDNA synthesis conditions were 10 min at 25 °C, 120 min at 37 °C and 5 min at 85 °C using a Veriti™ 96-Well thermal Cycler (Applied Biosystems™). Then the RT-qPCR was performed using the SensiFAST™ Probe No-ROX Kit (Meridian Bioscience^®, OH, USA) with a final volume of 25 µl and concentration as recommended by the manufacturer. Cycling conditions at a Thermo Scientific™ PikoReal™ Real-Time PCR System (Thermo Scientific™, MA, US) were set at 3 min at 95 °C and 45 cycles of 10 s at 95 °C and 30 s at 60 °C; 3 µl of extracted viral RNA and 0.8 µl of forward primer (10 μM) and 0.8 µl of reverse primer (10 μM). Primer sequences described in Table 1 were used in the reaction.

Table 1.

Sequences used in this study

Target/name	Sequence (5′–3′)	Positiona
MNV-1 RT-qPCR Forward	AACGCTCAGCAGTCTTTGT	5043–5061
MNV-1 RT-qPCR Reverse	GGCGGTAGGAACAAGAT	5122–5138
MNV-1 RT-qPCR Probe	AATGAGGATGAGTGATGGCGCAGCG	5063–5087
MNV-1 RPA T7 promoter/ Forward	TAATACGACTCACTATAGGGAGACGCTTTGGAACAATGGATGCTGAGACCCCGCAGG	4986–5042
MNV-1 RPA Reverse	CCAGCCACGGGYTGAATGGGGACGGCCTGT	5144–5173
MNV-1 crRNA	GAUUUAGACUACCCCAAAAACGAAGGGGACUAAAACAUCCUCAUUCACAAAGACUGCUG	5049–5072

^aPosition sequences based on Murine_norovirus_1_ATCC_VR_1937 from ATCC bank. Target region: ORF1-ORF2 junction (RNA polymerase-capsid protein VP1)

For absolute quantification by RT-qPCR, a standard curve (Supplementary Fig. 3) was generated using a ten-point, 10-fold serial dilution of an MNV-1 amplicon (5.06 × 10⁹ to 5.06 × 15 genomic copies per reaction [gc/rx]), with each dilution tested in triplicate. The amplicon was purified from agarose gel using the Wizard^® SV Gel and PCR Clean-Up System (Promega, WI, USA), and DNA concentration was determined via spectrophotometry using a NanoDrop™ 2000 (Thermo Scientific™). Copy number was calculated using the formula: copy number = (DNA concentration [ng/µl] × 10⁹)/(Molecular weight of DNA [g/mol] × Avogadro's number). To determine the limit of quantification (LOQ) for the RT-qPCR assay targeting murine norovirus (MNV-1), the ten-point, 10-fold serial dilution standard curve of quantified RNA standard (5.06 × 10⁹ to 15 genomic copies per reaction) was analyzed in three independent experiments, each with three technical replicates (n = 9 per dilution). LOQ was defined based on both precision and accuracy, following established guidelines (Forootan et al., 2017; Taylor, 1987). Precision was assessed using the coefficient of variation (CV): CV (%) = (Standard deviation of copy number / Mean copy number) × 100. A CV threshold of ≤ 35% was used (Klymus et al., 2020). Accuracy was evaluated using the absolute relative error (ARE): ARE (%) = |(Measured copy number − Input copy number)/Input copy number| × 100. In the absence of standardized qPCR-specific criteria, an ARE threshold of ≤ 20% was adopted based on U.S. FDA bioanalytical method validation guidance (FDA, 2018).

Based on the defined criteria (CV ≤ 35%, ARE ≤ 20%), the LOQ for the MNV-1 RT-qPCR assay was determined to be 5 gc/rx (CV = 19.7%, ARE = 3.3%). For assays performed on lettuce and blueberry matrices, LOQ could not be established due to excessive ARE values (lettuce > 50%, blueberry > 97%), likely resulting from low viral recovery efficiency (see Tables 2, 3). Consequently, for these matrices, the limit of detection (LOD) was calculated instead, as described below.

Table 2.

RT-qPCR detection of MNV-1 in inoculated lettuce samples

Total inoculuma	Lettuce		Lettuce RNase ONE™
	log10 gc/ 25 g  ± SD (Cq)	% Recovery  ± SDb	log10 gc/ 25 g  ± SD (Cq)	% Reduction  ± SDc
8.59	6.54 ± 0.28 (26.81)	1.00 ± 0.50	5.91 ± 0.25 (28.72)	76.72 ± 4.06
7.59	5.90 ± 0.29 (28.98)	2.33 ± 1.21	4.99 ± 0.16 (31.85)	88.64 ± 7.21
6.59	4.93 ± 0.31 (32.21)	2.52 ± 1.39	4.23 ± 0.18 (34.47)	81.92 ± 7.21
5.59	4.32 ± 0.37 (34.46)	6.72 ± 4.89	3.45 ± 0.21 (37.83)	88.25 ± 8.61
4.59	3.72 ± 0.29 (35.68)	15.46 ± 9.44	2.92 ± 0.09 (39.44)	85.87 ± 11.49
3.59	3.25 ± 0.36 (36.81)	58.62 ± 50.56	Not detected	Not detected
2.59	2.99 ± 0.12 (38.68)	Not detected	Not detected	Not detected

^aTotal inoculum was calculated as MNV-1 genomic copies and expressed as log10 gc/ in 25 g

^bSee methods. % recovery compared the MNV-1 gc quantified after elution from the produce sample to the MNV-1 gc used to inoculate the produce sample

^cSee methods. % reduction compared the MNV-1 gc quantified after RNase treatment to the MNV-1 gc quantified without RNase treatment

Table 3.

RT-qPCR detection of MNV-1 in inoculated blueberries samples

Total inoculuma	Blueberries		Blueberries RNase ONE™
	log10 gc/ 25 g  ± SD (Cq)	% Recovery  ± SDb	log10 gc/ 25 g  ± SD (Cq)	% Reduction  ± SDc
8.59	5.95 ± 0.12 (29.45)	0.23 ± 0.06	5.34 ± 0.08 (31.07)	75.41 ± 6.25
7.59	5.31 ± 0.08 (31.65)	0.52 ± 0.09	4.54 ± 0.05 (33.83)	83.08 ± 3.55
6.59	4.32 ± 0.04 (34.68)	0.53 ± 0.04	3.65 ± 0.04 (36.86)	78.56 ± 3.53
5.59	3.58 ± 0.15 (36.62)	1.00 ± 0.33	2.89 ± 0.06 (39.35)	80.30 ± 11.46
4.59	3.05 ± 0.23 (38.20)	3.13 ± 1.42	Not detected	Not detected
3.59	2.65 ± 0.02 (40.05)	11.69 ± 0.66	Not detected	Not detected
2.59	Not detected	Not detected	Not detected	Not detected

^aTotal inoculum was calculated as MNV-1 genomic copies and expressed as log10 gc/ in 25 g

Sample Collection and Inoculation

Fresh iceberg type lettuce (Lactuca sativa var. capitata) and the common North American blueberry (Vaccinium corymbosum) were bought at local food stores in Northern Mexico. These food matrices were selected as representative fruits (e.g., berries) and vegetables (e.g., leafy greens) associated with norovirus outbreaks (Cook et al., 2019; Sobolik et al., 2021; Torok et al., 2019; Yang & Scharff, 2024). Both food matrices were independently subdivided in 25 g samples and placed in Whirl–Pak^® bags (VWR™, WI, USA). Any adhered soil was removed by gentle scrubbing. Seven ten-fold diluted doses (3.9 × 10⁸ to 3.9 × 10² gc/25 g of food in 300 μl of PBS) of stock murine norovirus inoculum were spot inoculated (6 spots of 50 µl) on replicate 25 g produce samples and allowed to dry air for 1 h. For both food matrices, a no-virus inoculum control of 270 μl PBS with was included as negative control. For each food matrix and each dilution, three independent experiments with three replicates were performed.

Virus Elution and Concentration

For elution and concentration from each food sample, the ISO 15216-1:2017 (ISO, 2017) was followed. Briefly, 40 ml of Tris (1 mol/l), glycine 0.05 M, and beef extract 1% (TGBE) buffer were added to each sample. Samples were incubated at room temperature with constant rocking for 20 min. The pH of samples was monitored every 10 min and, if necessary, adjusted to 9.5 ± 0.5 with NaOH 10 N. The supernatant was transferred to a 50 ml tube and clarified by centrifugation at 10 000g for 30 min at 5 °C. The supernatant was collected and adjusted to a pH of 7.0. A total of 0.25 volumes of 5 × PEG/NaCl solution (500 g/lPEG 8000 (Sigma-Aldrich, USA), 1.5 mol/l NaCl) were added to the recovered supernatant and homogenized by shaking for 60 s and incubated with constant rocking (approximately 60 oscillations min − 1) at 5 °C for 60 min. The samples were centrifuged at 10,000g for 30 min at 5 °C and the supernatant was discarded. The pellet was resuspended in 500 μl of PBS. Samples were subdivided for further RNase pre-treatment and limit of detection determination. The limit of detection (LOD) for the MNV-1 RT-qPCR assay on lettuce and blueberry was defined as the lowest dilution at which all replicates produced Cq values below 40.

RNase Pre-treatment

Prior to RNA extraction, the eluted virus (500 µl of PBS) was divided into a 100 µl sample that was discarded, a 200 µl no-RNase control sample, and a 200 µl RNase-treated sample. The 200 µl RNase-treated sample was incubated with 1 µl of 10 U RNase ONE™ Ribonuclease (Promega, WI, USA) and a 99 µl of 1X reaction buffer (total reaction volume of 300 µl) for 30 min at 37 °C on the thermocycler, used as a constant temperature block. The reaction was stopped with the addition of lysis buffer before the RNA extraction. A heat-treated control was also included where MNV-1 was heated to 80 °C for 5 min prior to RNase treatment. For each amplification, negative (water) and positive (11.59 log 10 gc of MNV-1/25 g of food) amplification controls were also included.

Sequence Design (Primers and CRISPR crRNA)

The RT-qPCR and RPA primers and crRNA are listed in Table 1. All primers and crRNA were designed within the same region. All available sequences of the CW1 strain of MNV-1 at GenBank were aligned using Clustal Omega (Sievers et al., 2011). The complete sequenced genome from ATCC^® VR-1937™ was used as the main CW1 MNV-1 strain, and a conserved region was determined through a multiple alignment (Supplementary Fig. 1). The polyprotein coding region of the RNA-dependent RNA polymerase (RdRp) and the capsid protein were chosen as the RT-qPCR (ORF1-ORF2) and CRISPR target (ORF-1ORF2). The RT-qPCR primers and probe were designed using NCBI-Primer-BLAST and IDT-PrimerQuest™ Tool, along with quality analysis tools (Integrated DNA Technologies IDT- OligoAnalyzer™, Primer3). Isothermal Recombinase Polymerase Amplification (RPA) primers were designed following the TwistAmp^® Assay Design Manual (TwistDx™, Cambridge, England). Additionally, a T7 promoter was added to the 5’ end of the forward RPA primer for further RNA transcription. The crRNA for MNV-1 was designed following the published guidelines from original Cas13a-based detection developed (Kellner et al., 2019) and obtained from Integrated DNA Technologies IDT (IA, USA). A full DNA map of the detection design is presented in Supplementary Fig. 2.

Cas13a-Based Assay (CRISPR FRESH)

CRISPR FRESH detection consisted of the combination of RPA amplification, T7 transcription, and Cas13a system into a single reaction, plus an RNase treatment prior to RNA extraction. Without the RNase treatment, the method is a CRISPR–Cas13a-based MNV-1 detection system. A reaction volume of 50 μl was used with the following reagents and concentrations: RPA mix volume containing 20 μl of 2 × Reaction Buffer, 5 μl of 10 × Basic E-mix (TwistAmp^® Liquid Basic), 2.4 μl forward primer (10 μM) and 2.4 μl reverse primer (10 μM, (RPA primer sequences in Table 1), 2 μl of total 1.8 mM dNTP´s (Thermo Scientific™, MA, US), 2.5 μl of 20 × Core Reaction Mix, 0.25 μl of Cas13a (10 μM) (SignalChem Biotech, Richmond, Canada), 0.125 μl crRNA (10 μM), 1.5 μl T7 RNA polymerase (50 U/μl; Thermo Scientific™, MA, US), 2 μl ribonucleoside triphosphate (rNTP) mix (50 mM; Thermo Scientific™, MA, US), 2 μl RNase inhibitor (40 U/μl; Promega© WI, USA), 0.5 μl MgCl₂ (500 mM; Promega© WI, USA), 1 μl of RNaseAlert™ Lab Test Kit v2 (1 μM; Invitrogen™, MA, US), and RNase-free H₂O. Then 3 μl of template and 2.5 μl of magnesium acetate (280 mM) were added to the tube cap and the tube cap closed. This mixture was vortexed and centrifuged to start the reaction. Samples were subdivided into 25 μl replicates (considering at least two reactions). The fluorescence intensity was measured every minute in the PikoReal™ Real-Time PCR System (Thermo Scientific™, MA, US). Fluorescence measurements were taken at 120 min to capture the linear phase of the reaction, prior to plateauing. At this point, the lowest concentration tested showed a fluorescence signal significantly higher than the no-template control confirming positive detection without the influence of signal saturation. A cost analysis between the reagent cost per reaction for CRISPR detection compared to RT-qPCR was also performed (Supplementary Table 3). Because the CRISPR, unlike the RT-qPCR, assay cannot provide an absolute quantification of samples, instead of a limit of quantification (LOQ), we calculated the limit of detection (LOD). For the limit of detection (LOD) for CRISPR, the MNV-1 product used for the RT-qPCR LOQ (described above) was used for the LOD. The LOD was set at the lowest dilution (using same dilutions as with RT-qPCR) where all replicates were higher than the positivity threshold for CRISPR (mean fluorescence of the negative controls plus three times their standard deviation).

Evaluation of the Sensitivity and Specificity of CRISPR Detection

To evaluate the sensitivity of detection of our integrated method, six ten-fold serial dilutions ranging from 5.06 × 10⁵ to 5.06 × 10⁰ gc/ rx of MNV-1 were used to compare the limit of detection between CRISPR and RT-qPCR. To evaluate the specificity of our method, the amplification of MNV-1 RNA was compared to the amplification from other non MNV-1 RNA targets including hepatitis A virus (HAV), human norovirus genogroups GI and II, and rotavirus. As described in the FDA BAM Chapter 26 (Williams-Woods et al., 2022), the HAV target was the 5′-UTR region for all HAV strains (448–537 nt, 89 bp, GenBank accession M14707), the human norovirus target was the RNA polymerase (ORF1) region for GI (5287–5371 nt, 84 bp, GenBank accession KF039728) and GII (5,03–5100, 97 bp, GenBank accession EF684915). Both HAV and norovirus targets were generated using gBlocks gene fragment constructs (IDT, USA). The rotavirus target was the rotavirus region VP6 (24–1217 nt, 1,193 pb, based on GenBank accession NC_011509.2) from rotavirus strain A using the pT7-VP6SA11 plasmid, a gift from Takeshi Kobayashi (Addgene plasmid # 89166; http://n2t.net/addgene:89166; RRID:Addgene_89166).

Efficiency of Recovery

For the RT-qPCR method, the recovery efficiency of the elution and concentration steps was estimated using the equation: Virus recovery yields (%) = 10^(ΔCq/m) × 100% where ΔCq is the Cq value of extracted viral RNA from the food sample minus the Cq value of viral RNA extracted from the MNV inoculum dilution dose, and m is the slope of the virus RNA transcript standard curve (ISO, 2017). For the CRISPR method, as there is no published recovery efficiency method to date, we adapted a recovery efficiency method from viral food detection assays (Park et al., 2010b). The recovery efficiency of the elution and concentration steps was estimated using the equation: Virus recovery yields (%) = (mean fluorescence value of extracted viral RNA from the food sample/mean fluorescence value of viral RNA extracted from the MNV inoculum dilution dose) × 100%. Mean fluorescence data used, for the three replicates at each dose, were at the endpoint of the CRISPR reaction (120 min). Due to inherent differences in measurement principles and units between RT-qPCR and CRISPR detection methods, recovery data are presented descriptively and not directly compared.

Statistical Analysis

All assays were performed three separate times (repetitions) with three replicates each. The positivity threshold for CRISPR was defined as the mean fluorescence of the negative controls plus three times their standard deviation (mean + 3 × SD). Statistical differences between negative controls and samples were determined using a one-way ANOVA followed by Dunnett’s multiple comparisons test. Significant differences in assay values between heat-treated to non-heat-treated or RNase-treated to untreated samples were analyzed through a linear regression model using the value of the assay (RT-qPCR, log10 transformed viral titer; CRISPR, background-subtracted fluorescence) as the outcome and dose (log10 transformed), treatment (exposure of interest), and replicates and independent experiments (adjustment) as the exposures. Normality of data distribution was confirmed prior to analysis. Statistical analyses were performed using statistical packages in R programming language (R studio).

Results

RT-qPCR of RNase Pre-treated Lettuce and Blueberries Samples

To establish the limit of detection (LOD, see Materials & methods) of the RT-qPCR RNase pre-treatment MNV-1 detection assay, a ten-fold serial dilution of known concentrations were inoculated on lettuce (Table 2) and blueberry (Table 3) samples. The dilution range was 8.59 and 2.59 log10 genomic copies (gc) on 25 g of produce sample, equivalent to 5.06 × 10⁶ to 5.06 × 10⁰ gc/rx. The LOD for the RT-qPCR MNV-1 detection assay, without RNase ONE™, for lettuce samples, was 2.59 log10 gc/25 g sample (5.06 gc/rx) inoculated MNV-1 (Table 2). At this dose, the average eluted MNV-1 was 2.99 ± 0.12 log10 gc/25 g. For blueberries samples, the LOD was ten-fold higher at 3.59 log10 gc/25 g sample (5.06 × 10¹ gc/rx) inoculated MNV-1 (Table 3). At this dose, the average eluted MNV-1 was 2.65 ± 0.02 log10 gc/ 25 g. At each dose, RNase-treated samples, compared to untreated, had lower detectable RNA (Tables 2, 3). The RNase-treated samples, compared to untreated samples, ranged in % reduction, across doses, between 76.72 ± 4.06 to 88.64 ± 7.21% (lettuce) and 75.41 ± 6.25 to 83.08 ± 3.55 (blueberries). Heat-treated MNV-1 at 80 °C (Supplementary Table 1 and 2) was included as a positive RNase control. RNase-treated, compared to untreated, MNV-1 at 80˚C exhibited a significant reduction in titer across doses (lettuce 93.5–95.3%; blueberries 96.2–96.9%). MNV-1 viral titer that was heat-treated at 80 °C (Supplementary Table 1 and 2), compared to non-heat-treated (Tables 2, 3), was not significantly different for either lettuce or blueberries (data not shown).

CRISPR–Cas13a-Based Detection Optimization and LOD (Sensitivity)

The Cas13a-based detection is intended to be a one-pot detection that uses RPA isothermal pre-amplification in combination with Cas13a. Before testing Cas13a-based detection in produce samples, we optimized the detection of MNV-1 along with crRNA concentration (e.g., 10 µM exhibited greatest signal yields, Supplementary Fig. 4), time of detection, limits of detection (LOD, see Materials & methods) and specificity with our target. We confirmed that MNV-1 was detected with all required reagents but not when any of the reagents were missing (Fig. 1A). Because the Cas13a-based detection, unlike RT-qPCR, cannot provide an absolute quantification of viral amounts, instead of LOQ, as a measure of assay sensitivity, we calculated the limit of detection (LOD). We selected 120 min as the cut-off time for detection because the fluorescence signal reached a plateau at that time. To establish the LOD of the Cas13a-based detection, ten-fold serially diluted MNV-1 (log10 8.59 to 2.59 gc/25 g) were tested (Fig. 1B). The LOD was at least 5.06 × 10⁰ gc/rx (total of log10 2.59 gc/25 g). Thus, a signal of 144.05 ± 107.26 fluorescence arbitrary units (a.u.) was the minimal positive signal of our method; samples below 144.05 fluorescence a.u. were considered negative. The specificity of the assay for MNV-1 was confirmed when the assay detected MNV-1 target but not synthetic DNA targets of GI and GII Norovirus, Hepatitis A virus and a plasmid rotavirus pT7-VP6SA11 (Fig. 1C).

Fig. 1.

Cas13a-based detection optimization. A The background-subtracted fluorescence signal increased over time and reached a plateau approximately at 120 min but only in a reaction mix with all required reagents. B The limit of detection was 5.06 gc/rxn for Cas 13a detection of the MNV-1 target.; ***p < 0.001. C Cas 13a-based detection detected MNV-1 but not HAV, GI, and GII norovirus synthetic targets and rotavirus VP6 containing plasmid (see Methods for details). Error bars represent standard deviation. Statistical analysis was performed using one-way ANOVA with Dunnett’s post hoc test

CRISPR FRESH on Lettuce and Blueberries Samples

The CRISPR FRESH method, consisting of the integration of RNase pre-treatment, RPA pre-amplification, and Cas13a detection, was evaluated on inoculated lettuce (Fig. 2A) and blueberry (Fig. 2B) samples. Consistent with the prior RT-qPCR without RNase pre-treatment, our CRISPR–Cas13a method, without RNase pre-treatment, detected, for lettuce, 2.59 log10 gc/25 g sample (5.06 gc/rx) and, for blueberries, 3.59 log10 gc/25 g sample (5.06 × 10¹ gc/rx) (Table 4). However, when the samples were pre-treated with RNase ONE™ (CRISPR FRESH), we observed a positive signal corresponding to 3.59 log10 gc/25 g sample (5.06 × 10¹ gc/rx) for both lettuce and blueberries. The CRISPR-based assay detected a lower RNase-treated dose than the RT-qPCR assay (Table 2 lettuce 4.59 log10 gc/25 g sample [5.06 × 10² gc/ rx], Table 3 blueberry 5.59 log10 gc/25 g sample [5.06 × 10³ gc/ rx]) (Table 4). RNase-treated MNV-1 at 80˚C (positive RNase control), compared to untreated, exhibited a significant reduction in fluorescence across doses (lettuce 93.5–95.3%; blueberries 96.1–96.8%). Heat-treated MNV-1 fluorescence at 80 °C (Supplementary Fig. 5; A and B), compared to non-heat-treated (Fig. 2A, B) MNV-1 was not significantly different for either lettuce or blueberries (data not shown).

Fig. 2.

Cas13a-based detection of spiked MNV-1. (log 8.59 to 2.59 gc/ 25 g) A Cas13a detection of RNase-treated and non-treated lettuce samples. B Cas13a detection of RNase-treated and non-treated blueberries samples. Background-subtracted fluorescence signals (arbitrary units, a.u.) are shown for serial dilutions of the positive control (Pst Ctrl, 11.59 Log10 MNV-1/25 g of food) and negative control (Negative Ctrl) in untreated (black bars) and RNase ONE™-treated (gray bars) samples. The Y axis of each graph represents the concentration of gc of spiked virus per 25 g of food. A signal of 144.05 ± 107.26 a.u. was the minimal positive signal of our CRISPR method. Error bars represent standard deviation. Statistical analysis was performed using one-way ANOVA with Dunnett’s post hoc test. ***p < 0.001

Table 4.

Comparison between the MNV-1 detection of RT-qPCR and CRISPR methods between lettuce and blueberries at each inoculated concentration

Inoculated concentrationb	RT-qPCRa				CRISPRa
	Lettuce		Blueberries		Lettuce		Blueberries
	No RNase ONE™	RNase ONE™	No RNase ONE™	RNase ONE™	No RNase ONE™	RNase ONE™	No RNase ONE™	RNase ONE™
8.59	+	+	+	+	+	+	+	+
7.59	+	+	+	+	+	+	+	+
6.59	+	+	+	+	+	+	+	+
5.59	+	+	+	+	+	+	+	+
4.59	+	+	+	−	+	+	+	+
3.59	+	−	+	−	+	+	+	+
2.59	+	−	−	−	+	−	−	−

^aMNV-1 detected (“+”; Cq ≤ 40) and undetected (“−”; Cq > 40) criteria for positivity are the same for both RT-qPCR and CRISPR-based assays

^bTotal inoculum was calculated as MNV-1 genomic copies and expressed as log10 gc/ in 25 g

Recovery Efficiency

We compared the recovery efficiencies of lettuce and blueberries using RT-qPCR and separately, using CRISPR. The results, and thus recovery efficiencies, from these assays cannot be directly compared due to different measurement units (RT-qPCR: gc/25 g food; CRISPR: fluorescence/25 g food). Using the RT-qPCR assay, the recovery yield range, across doses, of MNV-1 eluted and detected was 1.00–58.62% in lettuce samples and 0.23–11.69% in blueberry samples (Tables 2, 3). By RT-qPCR, across doses, lettuce recovery yields were significantly higher (p < 0.05) than blueberry recovery yields.

Using the CRISPR assay, the recovery yield range, across doses, of MNV-1 eluted and detected by CRISPR was 50.6–94.4% in lettuce samples and 60.8–91.8% in blueberry samples (Fig. 2, data not shown). By CRISPR, across doses, there was no significant difference between lettuce and blueberry recovery yields. Given that RT-qPCR and CRISPR detection are based on different measurement principles and units, recovery yields are described independently for each method without implying direct quantitative comparison.

Discussion

The goal of this study was to detect MNV-1 with an intact capsid, a proxy for infectivity, through a CRISPR–Cas13a-based detection method in conjunction with an RNase-capsid integrity assay. We termed this assay: CRISPR Foodborne RNA-virus Enzymatic Sensing for High-throughput on fresh produce (CRISPR FRESH). We found that this CRISPR-based assay was both sensitive in detecting MNV-1 on lettuce and blueberries and did not cross-react with other genetic material. We also noted that this CRISPR-based, compared to the RT-qPCR-based, assay could detect the same inoculation doses of MNV-1 on either lettuce (2.59 log10 gc/25 g) or blueberry (3.59 log10 gc/25 g) samples (Table 4) but had assay-specific differences by produce type. Lastly, CRISPR-, compared to the RT-qPCR-based, assay after RNase treatment, detected MNV-1 RNA at a lower MNV-1 inoculation dose on lettuce and blueberry samples.

Other norovirus reports suggested higher or comparable CRISPR-based assay LOD (sensitivity) results compared to this report (2.59 log10 gc/25 g, 5.06 gc/reaction within 120 min). These norovirus reports employed either Cas13a or Cas12a detection and used fluorescence readers or lateral flow strips (LFS) for readout (Duan et al., 2022; Qian et al., 2021). For instance, Jia et al., (2020) reported a sensitivity of 50 norovirus GII gc/reaction in human stool samples within 20 min using RPA and a lateral flow test. Han et al. (2020) reported a sensitivity of 166 copies/μl GII norovirus in shellfish, water, and feces within 20 min using RT-RPA. Ma et al., (2018a, 2018b) reported a sensitivity of 100 gc of MNV-1 in mice fecal, fecal, and gastric tissue specimens within 16 min using RPA. Duan et al., (2022) reported a sensitivity of 5 gc/reaction of GII.4 in stool samples within 40 min using RPA-Cas13a, a fluorescence reader, and LFS. Additionally, Li et al., (2024) reported a sensitivity of 2.5 gc/reaction of norovirus GII.4 and GII.17 in stool samples within 40 min using RPA-Cas13a and a portable blue light transilluminator. The lower detection time range (16–40 min) of these studies compared to our study’s detection time (120 min) could be attributed to the variation of fluorescence signal peaks or the designed crRNA, which affects cleavage efficiency and fluorescence intensity (Gootenberg et al., 2017; Ke et al., 2021; Leski et al., 2023; Tambe et al., 2018; Yang et al., 2023). The lack of cross-reaction of our assay with RNA from other enteric viruses was consistent with other CRISPR-based norovirus detection assays (Cas13a or Cas12a) for stool specimens, which also did not cross-react with other virus species such as rotavirus, enterovirus, and even other norovirus variants within the same GII genogroup (Duan et al., 2022; Qian et al., 2021).

This CRISPR-based and the RT-qPCR-based assay could detect the same inoculation doses of MNV-1 on either lettuce (2.59 log10 gc/25 g) or blueberry (3.59 log10/25 g) samples but had assay-specific differences by produce type (Tables 2, 3, 4). The similarity in detection efficiencies between CRISPR- and RT-qPCR-based assays, as we have observed, has been documented by other groups (Duan et al., 2022; Gootenberg et al., 2017). Both the CRISPR- and RT-qPCR-based assays exhibited a higher MNV-1 LOD on lettuce (Table 4), in comparison with blueberry. We hypothesized that the MNV-1 detection was similar across assays, but that the MNV-1 viral elution was superior in lettuce, in comparison with blueberry samples. In support of this hypothesis, when viral titer could be quantified (by RT-qPCR-based, but not CRISPR-based assays), we observed significantly higher overall MNV-1-specific percent recovery in lettuce, in comparison with blueberry samples (Tables 2, 3). These findings suggest that the food type matrix characteristics may influence viral recovery. Similar to our results, one other group also observed superior extraction efficiencies of MNV-1, MS2, and Tulane Virus in lettuce in comparison with blueberries (Tang et al., 2023).

CRISPR, compared to the RT-qPCR, assay could detect lower MNV-1 inoculated gc of MNV-1 RNA from lettuce or blueberry samples that were eluted and treated with RNase. It is unlikely that CRISPR-based, in comparison with RT-qPCR-based assays, are more sensitive at detecting, after RNase treatment, intact RNA in intact MNV-1 capsids. The reason for this is that we have already shown and asserted, based on other’s evidence (Duan et al., 2022; Gootenberg et al., 2017), that both CRISPR-based and RT-qPCR-based assays have the same sensitivity at detecting MNV-1 RNA. Instead, we hypothesize that the addition of RNase resulted in incomplete reduction of free MNV-1 RNA and generation of MNV-1 RNA fragments. We define free RNA as RNase-accessible RNA outside a capsid or inside a damaged “permeable” capsid. In support of this hypothesis, we and others have shown that RNase treatment of inoculated MNV-1 on produce does not result in complete (100%) reduction of MNV-1 RNA (Li et al., 2012; Marti et al., 2017; Monteiro & Santos, 2018). Our heat inactivation (80 °C) control, when comparing RNase-treated to untreated samples, exhibited a 93.5–96.8% (1.19–1.86 log10) reduction of MNV-1 RNA across MNV-1 doses on lettuce and blueberry samples. Marti et al. reported, in lettuce eluted samples using MNV-1, a 99.99% (5.7 log10) reduction by RT-qPCR and using RNase A and UV light (Marti et al., 2017). Other studies in non-food samples have also reported variation of 0.03–0.45 log10 in the reduction of heat-treated noroviruses by using RNase pre-treatment (Li et al., 2012; Monteiro & Santos, 2018). These results suggest RNase did not destroy all accessible RNA or that heat treatment (80 °C) does not fully provide access of all RNA to RNase. An additional possibility is that the MNV-1 stock used in our experiments contained a larger proportion of defective particles, potentially due to multiple passages in cell culture, which could affect the efficiency of heat inactivation and RNase treatment. If the hypothesis that RNase does not destroy all accessible RNA which is shown to be valid, then both CRISPR- and RT-qPCR-based assays, combined with RNase treatment to detect intact capsids (i.e., potentially infectious virus), may lead to false positive results: detection of incompletely degraded RNA fragments instead of protected RNA in intact capsids. To explain our results, we propose that CRISPR-based, in comparison with RT-qPCR-based, assays may be more effective at detecting RNase-generated MNV-1 fragments and thus more likely to generate false positive results (i.e., potentially infectious viruses). To address these hypotheses, our future research will focus on developing methodologies to adjust RNase treatment results based on the “effectiveness of free RNA reduction (e.g., using RNase-treated controls) and on improving RNase treatment conditions to ensure the elimination, and confirmation of the elimination, of free RNA. We also aim to further validate these CRISPR-based results with MNV-1 infectivity results.

Our RNase reduction in MNV-1 inoculated lettuce and blueberries (75.41–88.64%, 0.61–0.91 log10 reduction, Tables 2, 3), measured by RT-qPCR, was higher than other published work. Monteiro and Santos (2018), using norovirus GII isolated from stools of infected patients with gastroenteritis and Hepatitis A virus (HAV) HM175/18f (VR-1402) from cell culture, achieved a mean of 0.03 log reduction in GII norovirus and 0.95 log reduction in HAV in non-food samples. Li et al. (2012), using MNV-1 from tissue culture, reported a log reduction for approximately 0.04 using MNV-1 suspension and RNase One RT-PCR in vitro. We hypothesize two possible explanations for our higher RNase-induced reduction of MNV-1 RNA compared to previous studies. First, it is possible that our RNase reaction conditions (e.g., enzyme concentration, incubation time, temperature, or buffer composition) were more effective, compared to previous studies’ reaction condition, at degrading free RNA, leading to greater RNA reduction. Second, it is possible that either our multiple MNV-1 cell culture passage process or the lettuce, and blueberry elution process resulted in a higher proportion of defective particles and RNase-accessible RNA, either due to partial damage of viral capsids during handling, or intrinsic matrix effects that make viral  RNA more susceptibe to RNase. Further optimization and validation studies will be necessary to distinguish between these two possibilities.

This study had strengths and limitations. The first strength of this study was that CRISPR FRESH leverages the precision of Cas13a to achieve qualitative, sensitive, and specific detection of viral RNA, providing a reliable indicator of the presence of a target. The second strength of this study was the direct relationship between the concentration of target RNA and the detection capability of the enzyme (Fig. 1), laying the foundation for a dependable quantitative assay. The third strength was the evaluation of viral recovery and detection on different produce types across multiple doses. One limitation of this study was that we could not directly compare RT-qPCR results to CRISPR FRESH results (e.g., viral recovery) because it is currently qualitative and not yet capable of precise quantification without further optimization. A second limitation are the difficulties in developing a quantitative CRISPR-based assay because of the need to calibrate each standard at each step of the assay (e.g., pre-amplification, T7 transcription, detection). A third limitation may be the methodological inability to achieve 100% reduction of free RNA across sample matrices (e.g., different produce types), thus needing new methodologies to adjust for this confounder.

There are several future research areas to apply this assay to human norovirus on foods. One area is to enable quantitative analysis of pathogens in food samples by optimizing primer concentrations and reaction conditions to avoid saturation effects (e.g., in the pre-amplification step) and estimating the ratios of active Cas13 to the target RNA. A second area is to develop a calibration workflow that provides reliable parallel analyses of varying target concentrations. A third area is to validate the method across different food matrices to ensure its accuracy and reliability, including the quantification of viral elution efficiency for each food matrix. A long-term goal would be integrated CRISPR FRESH into the existing detection toolkit for food safety professionals in industry and government. This would include a cost–benefit analysis of its use versus traditional viral detection methods.

In conclusion, CRISPR FRESH demonstrates a sensitive and specific detection of noroviruses on fresh produce. The ability of CRISPR FRESH to detect viruses is comparable to RT-qPCR. Further, CRISPR FRESH offers the option to replace expensive RT-qPCR equipment (e.g., thermocycler) with less expensive equipment for temperature-controlled reaction conditions (e.g., thermoblocks) and readout (e.g., fluorometers). Further, CRISPR FRESH reagents are less expensive than RT-qPCR reagents per reaction (Supplementary Table 3). We also demonstrate its potential of detecting RNA in intact capsids, a proxy for infectivity. This report provides a foundation for additional cost-efficient, rapid, and in-field detection assays for human foodborne viruses on foods to control the spread of foodborne outbreaks.

Supplementary Information

Below is the link to the electronic supplementary material.

Author Contributions

All authors contributed to the study conception and design. Material preparation, data collection, and analysis were performed by A.O., A. M.-M., J. S. L., N. H., and S. G.. The first draft of the manuscript was written by A. O. and all authors commented on the previous versions of the manuscript. All authors read and approved the final manuscript.

Funding

This work is supported in part by the Universidad Autónoma de Nuevo Leon, PROVERICYT program and by the Agriculture and Food Research Initiative, project awards no. 2019-67017-29642 and 2024-67017-42441, from the U.S. Department of Agriculture’s National Institute of Food and Agriculture.

Data Availability

No datasets were generated or analyzed during the current study.

Declarations

Conflict of interest

The authors have no competing interests to declare that are relevant to the content of this article.