Pepper Mild Mottle Virus, a Plant Virus Associated with Specific Immune Responses, Fever, Abdominal Pains, and Pruritus in Humans

Background Recently, metagenomic studies have identified viable Pepper mild mottle virus (PMMoV), a plant virus, in the stool of healthy subjects. However, its source and role as pathogen have not been determined. Methods and Findings 21 commercialized food products containing peppers, 357 stool samples from 304 adults and 208 stool samples from 137 children were tested for PMMoV using real-time PCR, sequencing, and electron microscopy. Anti-PMMoV IgM antibody testing was concurrently performed. A case-control study tested the association of biological and clinical symptoms with the presence of PMMoV in the stool. Twelve (57%) food products were positive for PMMoV RNA sequencing. Stool samples from twenty-two (7.2%) adults and one child (0.7%) were positive for PMMoV by real-time PCR. Positive cases were significantly more likely to have been sampled in Dermatology Units (p<10−6), to be seropositive for anti-PMMoV IgM antibodies (p = 0.026) and to be patients who exhibited fever, abdominal pains, and pruritus (p = 0.045, 0.038 and 0.046, respectively). Conclusions Our study identified a local source of PMMoV and linked the presence of PMMoV RNA in stool with a specific immune response and clinical symptoms. Although clinical symptoms may be imputable to another cofactor, including spicy food, our data suggest the possibility of a direct or indirect pathogenic role of plant viruses in humans.


Introduction
Little is known regarding the viral flora present in the human gastrointestinal tract [1][2][3][4][5]. Recently, viral communities have been examined in non-diarrheic human stool in a systematic and unbiased fashion using metagenomics [6][7]. Interestingly, the vast majority of the recovered viral sequences corresponded to uncultured pathogenic plant RNA viruses [7]. The most abundant viral sequences isolated were from the Pepper mild mottle virus (PMMoV), with up to 10 9 virions per gram of dry weight fecal matter. PMMoV is a non-enveloped, rod-shaped, singlestranded positive sense RNA virus classified in the genus Tobamovirus, which includes viruses extremely resistant to physical and chemical agents [8][9]. It is one of the major pathogens of Capsicum spp (chili peppers). Complementary data from Zhang et al.'s study have shown that PMMoV could be detected in non-diarrheic stool from 12 out of 18 individuals living in San Diego, USA or in Singapore, suggesting it might be geographically widespread, and in 3 out of 22 fresh and processed pepper samples. Moreover, the fecal PMMoV was viable and could infect host plants.
In the present study, we determined if PMMoV could be detected in commercial food products in our geographical area or the stool of patients and if these viruses were associated with biological and clinical symptoms.

Food products and human stool collection
Twenty-eight commercial food products were purchased in various grocery stores and supermarkets in Marseille (France) and stored at room temperature until processing (Table 1). They included four fresh peppers, thirteen manufactured products containing peppers, four products containing spice and seven products not known to contain pepper. For solid food products, 1 ml of the product was diluted in 1 ml of 16 PBS. Additionally, 565 stool samples were obtained between March and September 2008 for microbial diagnosis independent of the present study and approximately 1 ml collected randomly from each stool was stored in 1 ml of 16 PBS at +4uC. Of these samples, 357 were from adults (.18 years of age) and 208 were from children (,18 years of age). This study has been approved by our institutional ethics committee (Comité d'Ethique de l'IFR48, Université de la Méditerranée; review no. 09-001), that waived the need for consent.

PMMoV RNA extraction
Viral RNA was extracted from 200 ml of the fecal supernatant or the solid food product suspensions following centrifugation at 10,000 g for 10 min or directly from the liquid food products by using the MagNA Pure LC RNA Isolation Kit (Roche Diagnostics, Meylan, France). The prepared RNA was directly analyzed or stored at 280uC until processing.

Detection of PMMoV RNA by real-time PCR
PMMoV RNA was detected using a real-time reverse transcription (RT)-PCR assay with SuperScript III Platinum One-Step Quantitative RT-PCR System (Invitrogen Life Technologies, Carlsbad, Calif., USA) on an Mx3000P thermocycler (Stratagene, La Jolla, CA 92037 USA). Each 25 ml reaction contained 10 ml of the extracted viral RNA, 12.5 ml 26 reaction mix, 0.5 ml of SuperScript III/RT Platinum Taq mix, and 0.75 ml (10 pmol/ml) of each primer and 0.5 ml (10 pmol/ml) of probe described by Zhang et al [7]. Positive samples were further tested by RT-PCR amplification and sequencing, as well as electron microscopy.

Infection of host plants with PMMoV RNA-positive food products
The viability and infectivity of PMMoV recovered from two PMMoV RNA-positive powder or dried food-products (no. 8,17) and from a PMMoV RNA-positive Tabasco sample were tested by mechanical inoculation to host plants. Powder or dried foodproducts were crushed then mixed with an inoculation solution containing Na 2 HPO 4 0.03M and 0.2% (wt/vol) Na-diethyldithio-

PMMoV RNA sequencing and phylogenetic analysis
PMMoV sequences corresponding to two different regions of the PMMoV genome were obtained by RT-PCR amplification using the SuperScript One-Step RT-PCR System (Invitrogen Life Technologies) and primers that had been designed with the SVARAP program [11]. The first targeted region is located near the 59 end of the genome (primers used are PMMoV5pFwd: 59-ATATCTGATGATGCAAGTTC-39, 348 (the primer location was defined in reference to sequence GenBank accession nb NC_003630); PMMoV5p_Rev2: 59-TAAACCTCTCTATTA-GAGGC-39, 928), and the second region corresponds to a fragment of the capsid gene (primers are PMMoVcaps_Fwd: 59-CGTTAGGyAATCAGTTTCAA-39, 5776; PMMoVcaps_Rev2: 59-CGAACTAACTCATTCATGA-39, 6070). For RT-PCR amplification, five ml of extracted viral RNA were added to 20 ml reaction solution containing 0.5 ml (10 pmol/ml) of primers, 12.5 ml of 26 reaction buffer, and 0.5 ml of RT/Platinum Taq mix. RT-PCR reaction was conducted under the following conditions: 45uC for 30 min for RT, 94uC for 2 min for initial denaturation, followed by 40 cycles of denaturation at 94uC for 30 s, annealing at 56uC for 45 s, elongation at 72uC for 2 min, then final elongation step at 72uC for 5 min. PCR products were purified with Sephadex G-50 Superfine on MAHVN 4550 plates (Millipore, Molsheim, France) then sequenced by using the amplification PCR primers and the Big Dye Terminator cycle sequencing kit version 1.1 on the ABI Prism 3130 genetic analyzer (Applied Biosystems, Branchburg, NJ, USA). The phylogenetic analysis was performed using Mega v.4.1 software (http://www. megasoftware.net/).

PMMoV purification from Tabasco sauce
Tabasco sauce (50 ml) was centrifuged at low speed (2,000 rpm) for 20 min. The PMMoV-containing supernatant (30 ml) was then treated with 1 M NaCl (30 ml) at 4uC overnight and filtered using a 5 mm Millex-SV filter (Millipore). The filtrate was loaded onto a caesium chloride (CsCl) density gradient consisting of 1 ml layers of 1.7, 1.5 and 1.30 g/ml CsCl in 16 PBS. After ultracentrifugation (22,000 g for 2 h at 4uC), the 1.30-1.50 g/ml fraction containing the purified PMMoV (r = 1.31) was collected. This fraction was washed on a Microcon 100 centrifugal filter (Millipore) with 16 PBS.

Electron microscopy study of PMMoV
A 40 ml aliquot of Tabasco sauce or fecal supernatant was deposited on a nickel 400 mesh Formvar carbon grid (EMS, Fort Washington, PA). After 15 min at 37uC, staining was performed with 1% ammonium molybdate for 10 sec. The stained grids were examined with a transmission electron microscope (Philips Morgagni 268D at 80Kv) after they were dried for 30 min at room temperature. The sensitivity of detection is estimated to be 10 6 -10 8 virus particles/ml [12]. In addition, immunogold staining was performed, prior to electron microscopy, on PMMoV purified from PMMoV RNA-positive Tabasco sauce and stool, as follows: the electron microscopic grids were fixed with PBS and 1% glutaraldehyde for 2 min. After rinsing with PBS for 2 min, the grids were incubated twice for 5 min with 50 mM NH 4 Cl in PBS, preincubated twice for 2 min with Solution 1 (1% bovine serum albumine (BSA), 1% normal goat serum, 0.2% tween 20 in PBS) and incubated for 90 min with a polyclonal anti-PMMoV mouse antibody at 1:1 dilution in Solution 1. These antibodies had been produced in two 6-week-old female Balb/C mice inoculated intraperitoneally with 1.3 mg of purified PMMoV using 400 mg aluminum hydroxide and 10 mg non-methylated oligonucleotides containing CpG motifs as an adjuvant, as previously described [13]. Blood collection was performed at day 38 following two booster doses administered at day 14 and day 28. After rinsing the stained grids with 0.1% BSA in PBS for 2 min, the grids were preincubated in Solution 2 (0.01% fish gelatin in PBS), transferred to a drop of gold-conjugated goat anti-mouse antibodies (diluted 1:40 in Solution 2) for 60 min, washed six times with Solution 2 and twice with PBS for 10 min each. The grids were fixed in 2% glutaraldehyde for 15 min, then washed with PBS four times for 10 min and stained with 1% ammonium molybdate for 10 sec.
Two negative controls were concurrently tested using antibodies derived from a non-immunized mouse or rotavirus instead of PMMoV.

Anti-PMMoV ELISA
ELISA Nunc Maxisorp plates were coated with 100 ml PMMoV solution per well (0.6 mg/ml purified from a PMMoV RNApositive Tabasco sauce) overnight at 4uC. All wells were machinewashed thrice with 300ml 0.1% Tween in PBS and blocked with 200 ml of PBS containing 5% dried skimmed milk for 30 min at room temperature with gentle agitation. The wells were washed with 0.1% Tween in PBS, and 100 ml of serum diluted in 3% milk / 0.1% Tween in PBS was added. The plates were incubated for 1 h at room temperature and washed as described above. After washing, 100 ml of horseradish peroxidase-conjugated goat antihuman total Ig or IgM (Beckman Coulter, Marseille, France) diluted 1:5000 in 3% milk /0.1%Tween in PBS was added. The plates were then incubated for 1 hr at room temperature and washed as described above. The plates were developed by adding 200 ml of ortho-phenylenediamine per well and the absorbance was read at 492 nm.

Analysis of risk factors and biological and clinical signs associated with the presence of PMMoV RNA in stool or anti-PMMoV antibodies in serum
A case-control study was performed to assess the potential risk factors and the biological and clinical signs associated with the presence of PMMoV in the stool. Eighteen case-patients were compared to thirty-one control-patients with PMMoV RNAnegative stool. When possible, two control-patients from the same clinical units as the cases were selected for each case. The caseand control-patients were matched by age and sex. The variables analyzed included demographic characteristics, underlying diseases and clinical and biological signs. A Microsoft Excel spread sheet was used to enter the data. To identify risk factors for PMMoV detection, univariate analysis was performed with the Epi Info software version 3.3.2 (CDC, Atlanta, GA, USA). Proportions were compared by using the Yates Chi-square corrected test or the Fisher's exact test while continuous variables were compared by using either the ANOVA test or the Mann-Whitney/Wilcoxon two-sample test when the data were not normally distributed. For multivariate analysis, only variables with a p value #0.1 in univariate analysis were kept in the final model. Binary logistic regression analysis was performed using the SPSS software version 10 (Chicago, IL, USA).

Detection of PMMoV RNA by real-time PCR
Food products. PMMoV RNA sequences were recovered from twelve (57%) of the twenty-one pepper-or spice-containing food products (Table 1). Tabasco sauce contained the highest viral load, estimated to be nearly 10 7 PMMoV RNA copies/ml based on a cycle threshold of 22 by real-time PCR or the presence of more than one viral particle/field by electron microscopy (Figure 1a).
Stool from children. PMMoV RNA was detected in the stool from only 1 child out of 137 (0.7%) using real-time PCR ( Table 2). The presence of PMMoV RNA was confirmed by sequencing and viral particles compatible with PMMoV were observed using electron microscopy. The patient was a 5-year-old boy living in Marseille who was admitted to our institution because of abdominal pains, fever, asthenia and hyperventilation.
Stool from adults. The stool from 22 of 304 (7.2%) adults was positive for PMMoV by real-time PCR (Table 2). This proportion was significantly higher than the incidence in children (p = 0.0043). PMMoV RNA sequence could be recovered from all but two patients, whereas PMMoV-like viral particles were observed, by electron microscopy, in the stool from all but one patient, who was found positive for PMMoV RNA sequencing (Figure 1d) . The proportion of PMMoV RNA-positive patients was significantly higher in Dermatology Units than in other clinical units (7/20 (35.0%) versus 15/284 (5.3%); p = 7610 27 ) ( Table 2). Clinical and biological data were obtained from 18 of the 22 PMMoV-positive patients and compared to the characteristics of 31 patients with PMMoV-negative stool. By univariate analysis, abdominal pains, diverticulosis or diverticulitis, and fever were significantly associated with PMMoV RNApositive stool (Table 3). No biological sign was significantly associated with the presence of PMMoV RNA. By multivariate analysis, using a model including age, gender and the variables significantly associated with the case-patients in the univariate analysis, fever (p = 0.0447 (relative risk (RR), 5.290 (95% confidence limits for RR (CI95), 1.040-26.911)), abdominal pains (p = 0.0376 (7.069 (1.119-44.660)), and pruritus (p = 0.0460 (6.131 (1.033-36.402)) were independently associated with the presence of PMMoV RNA. In addition, an interaction was observed between abdominal pains and diverticulitis. Overall, the model fitted the data well (Hosmer-Lemeshow test equal to 0.94). Three of four patients with concurrent abdominal pains and fever and the two patients who had abdominal pains and pruritus were PMMoV RNA-positive.

Infection of host plants with PMMoV RNA-positive food products
All N. tabacum cultivar Xanthi NN plants inoculated with each of the three PMMoV RNA-positive food products developed local  (Figures 2a-h). These lesions were observed on the three inoculated leaves for each plant, whereas control plants remained devoid of symptoms during an observation period of two weeks post-inoculation. PMMoV RNA was detected using RT-PCR from local lesions, PMMoV-like viral particles were observed by electron microscopy, and systemic mosaic symptoms typical of PMMoV were observed within 2-3 weeks post-inoculation on C. annuum cultivar Yolo Wonder or N. benthamiana.

PMMoV RNA sequencing and phylogenetic analysis
PMMoV sequences corresponding to the 59 region of the PMMoV genome were obtained from four Tabasco sauce samples and thirteen patients, whereas sequences corresponding to a fragment of the capsid protein gene were obtained from eleven food products, including three Tabasco sauce samples, and eighteen patients (Figures 3a and 3b, respectively; GenBank accession numbers GQ427021-GQ427066). The BLAST analysis of the nucleotide sequences from the 59 region produced only ten relevant hits, after which the sequences no longer corresponded to PMMoV (http://blast.ncbi.nlm.nih.gov/Blast.cgi). The sequences recovered from Tabasco sauce tended to cluster separately from those recovered from patients (Figure 3a) with Tropical soda apple mosaic virus sequence. One cluster (cluster II-1) contained four sequences from the present study (Figure 3b), and 100% nucleotide identity was observed between the sequences from two patients that received care in a Dermatology Unit and from a spicy powder.

Anti-PMMoV Total Ig and IgM ELISA
The level of serum anti-PMMoV total and IgM antibodies was examined by ELISA in a subset of 29 adult patients, 13 of which had PMMoV-positive stool, as detected by PCR (Figures 4a and  4b). For the analysis of the IgM anti-PMMoV level, the optical density (OD) for the substrate and antibody controls was 0.079 and 0.099, respectively. The mean OD was significantly higher in patients with PMMoV RNA than in the controls (0.472 versus 0.325, p = 0.013 (ANOVA); Figure 4b). Using the receiver operating characteristic (ROC) curve test with a sensitivity of 100% and a specificity of 62%, a value of 0.320 was considered positive for the detection of PMMoV-specific IgM (AUC = 0.808; 95% confidence interval: 0.645-0.971; p = 0.005). The proportion of serum samples with an OD greater than 0.320 was significantly higher for patients with PMMoV RNA-positive stool than for the controls (p = 0.0004). With this threshold, all 13 serum from patients with PCR PMMoV-positive stool were seropositive, whereas 10/16 (63%) sera from patients with PCR PMMoVnegative stool were seronegative. Thus, the positive and negative predictive values for having PCR-positive stool were 68% and 100%, respectively. For the analysis of the total anti-PMMoV level, the optical density (OD) for the substrate and antibody   controls was 0.051 and 0.071, respectively. The mean OD was significantly higher in patients with PMMoV RNA than in the controls (0.663 versus 0.444, p = 0.028; Figure 4a). A cut-off value of 0.565 (estimated using the ROC curve test with a sensitivity of 69% and a specificity of 81%) was used (AUC = 0.724; 95% confidence interval: 0.529-0.918; p = 0.041). The proportion of serum samples for which the optical density (OD) was greater than 0.565 was significantly higher for patients with PMMoV-positive stool than for those with PMMoV-negative stool (p = 0.0070). Using this threshold, 9 (69%) of the 13 sera from patients with PCR PMMoV-positive stool were seropositive, whereas 13 (82%) of the 16 sera from patients with PCR PMMoV-negative stool were seronegative. Additionally, serological patterns indicating past infection (i.e., a positive total antibody result, but a negative IgM result) and negative total antibody and IgM testing were observed in two and eight PMMoV-negative patients, respectively, but in none of the PMMoV-positive patients (p = 0.0004). Finally, in the case-control study (Table 3), positive testing for anti-PMMoV IgM antibody was significantly more frequent in casepatients than in control-patients.

Discussion
We detected PMMoV RNA by real-time PCR in stool from 22 of 304 (7.2%) adult patients and 1 of 208 children who received care in public hospitals. The present study therefore confirms the recent finding by Zhang et al. who identified PMMoV as the major RNA virus in human stool in a metagenomic study [7]. However, we found a much lower prevalence of PMMoV RNA-positive stool by real-time PCR among adult patients than Zhang et al., who detected the viral genome in 12 of 18 fecal samples (66.7%) using RT-PCR. This difference might be explained either by a lower sensitivity for real-time PCR or the lower exposure to PMMoV in our patients. We recovered PMMoV RNA sequences from 57% of food products containing pepper or spice and purchased from grocery stores in Marseille (southeastern France), indicating that these products represent a common source of viruses through ingestion. In addition, in food products and patient's stool, we could visualize by electron microscopy complete viruses whose shape and size are consistent with PMMoV and that reacted with PMMoV-specific antibodies in the immunogold staining experiments. Moreover, 46 new PMMoV sequences corresponding to two regions of the viral genome were recovered from food products or patients' stool. Four phylogenetic clusters were identified, and one of them (II-1) included sequences from both patients and a putative food source. Furthermore, three PMMoV RNA-positive food products were found to contain virus particles infectious to host plants.
We also identified statistically significant differences in the occurrence of fever, abdominal pains, and pruritus and the detection of specific immune responses to PMMoV in the casecontrol study. We, therefore, believe that we provide the first evidence that plant viruses may cause disease in humans. Fever and abdominal pains, which were significantly more common in PMMoV-positive patients (39% for both clinical features) than in patients with PMMoV-negative stool (13% and 7%, respectively), were also present in the only PMMoV RNA-positive child. These symptoms might not be caused by PMMoV but by consumption of spicy food. Otherwise, they may be fortuitously related to an associated unknown infectious agent or to an allergic or toxic cofactor. Nevertheless, the presence of a specific immune response to PMMoV reinforces the hypothesis that this plant virus was the cause of the symptoms in these patients. Indeed, anti-PMMoV IgM antibodies were detected in all PMMoV-positive patients, significantly more frequently than in the controls. The concurrent detection of PMMoV RNA and antibodies to PMMoV indicates that this virus is not a neutral component of the human gut flora. On the contrary, our findings suggest that PMMoV may infect humans and cause clinical symptoms. We believe that these preliminary data may prompt further studies on the pathogenic role of PMMoV in humans. Finally, the higher prevalence of PMMoV RNA among patients who received care in Dermatology Units deserves further attention, as it was high (35%).
To date, plant viruses and vertebrate viruses are believed to exist in two distinct and non overlapping biological niches, and they are not known to share common hosts in the plant or animal kingdoms [14][15]. Therefore, to date, plant viruses have not been described as pathogens for vertebrates or humans or known even to infect them [16]. This paradigm is underscored by the growing body of data about the use of plant virus-based vaccines, which are considered as safe tools that might be useful in treating chronic viral infections and cancer in the future [17][18][19][20]. However, the boundaries between plants and animals might not be so hermetical for plant viruses, and, to our knowledge, no experimental model has been described that would allow us to rule out a pathogenic role for plant viruses in vertebrates. The recent work of Zhang et al. challenged this dogma, as it was neutral, technology-driven research that was not trying to confirm a former hypothesis [7,21]. That work identified plant viruses as the most common viral inhabitants of the human gut flora, PMMoV being the most prevalent. In addition, a recent study demonstrated that PMMoV   is widespread in wastewater from the United States of America (USA), at concentrations ranging between 8.0610 5 and 2.2610 7 copies/ml [22]. Beyond these findings, our results, together with previous data, suggest that the relationship between plant viruses and animals may deserve further re-evaluation using modern microbiologic tools (Table 4). The oral administration of Cowpea severe mosaic virus, another plant virus, has been previously shown to induce a durable and systemic immune response in mice without requiring the co-administration of an adjuvant [23], and similar results have been observed in mice using Alfalfa mosaic virus and chimeric plant virus particles [24][25]. In addition, some plant viruses can replicate in insect tissues, as illustrated by Tomato spotted wilt virus and Maize rayado fino virus [26][27][28]. Furthermore, a reduction in the lifespan and fecundity of insects (Bemisia tabaci) during infection with Tomato yellow leaf curl virus has been previously observed, indicating that this plant virus had some features of an insect pathogen [29]. It has been also suggested, based on viral sequence analysis, that during evolution, plant-infecting viruses (previously known as plant circoviruses and recently reclassified in the genus Nanovirus) switched hosts to infect a vertebrate and then recombined with a vertebrate-infecting virus [16]. Moreover, it has been shown recently that Tomato spotted wilt virus could replicate in two human cell lines, HeLa and diploid fibroblasts [30]. This virus is one of the most important plant pathogens worlwide, and it causes mild infection on its main insect vector, Frankliniella occidentalis, in which it induces a strong immune response [26]. Finally, the possible involvement of plant viruses in human diseases was questioned several decades ago. Indeed, culturable Tobacco mosaic virus could be recovered from sputum and thoracentesis fluids obtained from cigarette smokers with a history of pulmonary disease, including lung cancer [31][32][33]. Taken together, these experimental data, coupled with the previous results reported by Zhang et al. and our findings, favor the hypothesis of a pathogenic role for PMMoV. It should be highlighted that PMMoV, like other tobamoviruses, are known to be extraordinarily resistant to heat and desiccation [34][35]. Moreover, although this deserves to be investigated in further studies, the viability and infectivity to host plants were evidenced for PMMoV recovered from food products in the present study and from stools in Zhang et al.'s study [7]. Additionally, we visualized complete viral particles in stool by electron microscopy.
In summary, we confirmed that humans may carry a high PMMoV load, likely acquired from food products, and we documented that PMMoV might not only be a common inhabitant of the human gut but may also interact with the human immune system and cause clinical symptoms. These results should prompt further studies to re-evaluate whether or not plant viruses, including PMMoV, may have a pathogenic role in humans.