We previously developed a uniform rice-derived oral cholera vaccine (MucoRice-CTB) using an overexpression system of CTB (N4Q), a modified cholera toxin B subunit, and RNAi to inhibit the production of key rice endogenous storage proteins such as a prolamin and glutelin (Yuki et al., 2013). To establish MucoRice-CTB for human use, we developed MucoRice-CTB without the hygromycin phosphotransferase (HPT) selection marker by co-transforming two Agrobacterium tumefaciens, each with different T-DNA. ( Supplementary Figure 1 ) and had produced six selection maker–free MucoRice-CTB lines. Of the 6 lines advanced to the T4 generation (Mejima et al., 2015), we used MucoRice-CTB lines 19A, 9B, 43B, and 50A in this paper ( Supplementary Table 1 ). The T6 generation of MucoRice-CTB line 19A was produced as a master seed bank in the MucoRice-CTB production facility at the Institute of Medical Science, University of Tokyo, on September 28, 2016.

Leaf genomic DNA was isolated with a Nucleon PhytoPure kit (GE Healthcare, WI, USA). PCR was performed by using the EmeraldAmp MAX PCR Master Mix (PR320A, Takara, Tokyo, Japan)) on a Veriti Thermal Cycler (Thermo Fisher Scientific) as follows: 35 cycles of 30 s at 94°C, 30 s at 60°C, and 1 min at 72°C. The reaction products were separated by electrophoresis on a 1.5% (w/v) agarose gel. To determine full transgene sequences, PCR was performed with KOD FX polymerase (KFX-101, Toyobo, Osaka, Japan) on the same thermal cycler as follows: 2 min at 94°C and 35 cycles of 10 s at 98°C, 5 min at 68°C. The reaction products were separated by electrophoresis on a 0.7% (w/v) agarose gel. All primers used are listed in Supplementary Table 2 . The pZAAMP-CTB-10Li45GB3A and pZH2B plasmids were used as positive controls, and WT genomic DNA was used as a negative control.

Seeds of non-transgenic Oryza sativa L cv. Nipponbare (WT) and MucoRice-CTB lines were harvested 14 days after flowering. Total RNA was extracted with an RNeasy Plant Mini Kit (Qiagen, Hilden, Germany) and treated with DNase (Takara Bio Inc.). Total RNA (0.5 µg) was used to synthesize cDNA with a Prime Script Reagent Kit with gDNA Eraser (Takara Bio Inc.). cDNA (20 µl) was diluted 50 times with distilled water and used as a template. PCR was performed in 20 µL; the Fast SYBR Green Master Mix (Applied Biosystems) and a StepOne Plus real-time PCR system (Applied Biosystems) were used. The primer sets are shown in Supplementary Table 2 . Triplicate reactions were performed as follows: 20 s at 95°C; 40 cycles of 3 s at 95°C, 30 s at 60°C, 15 s at 95°C; and 1 min at 60°C. Transcript levels were normalized to that of 17S RNA. The specificity was assessed by analyzing the product melting curves.

Leaf genomic DNA was extracted from young leaves (2 g) as above. Standard procedures (Sambrook et al., 1989) were used. Genomic DNA (~20 µg) was digested with EcoRI or SacI, followed by fractionation on a 0.7% (w/v) agarose gel and blotting onto a Hybond N+ nylon membrane (GE Healthcare). The CTB DNA probe was amplified from pZAAMP-CTB-10Li45GB3A by PCR with the CTB-F and -R primers ( Supplementary Table 2 ). An AlkPhos Direct Labelling Module and Detection System with CDP-Star Detection Reagent (GE Healthcare) were used. Bands were detected with an ImageQuant LAS 4000 mini camera system (GE Healthcare). pZAAMP-CTB-10Li45GB3A was used as a positive control, and WT rice genomic DNA was used as a negative control.

MucoRice-CTB 19A was produced from master seed bank seeds in a closed plant production system at Chiba University. The production process flow from seed to rice flour (or active pharmaceutical ingredient) is shown in Figure 1A . Cultivation was performed by hydroponics using recirculating nutrient solution and polystyrene foam trays. Seeds of MucoRice-CTB 19A were sterilized by immersion in 70% ethanol followed by 1.25% sodium hypochlorite solution. These seeds were sown individually on a net attached to a square wooden frame at approximately 1.5 cm intervals and floated on water adjusted to pH 5.0 to 5.5. Liquid fertilizers (OAT House Fertilizers 1, 2, and 5; OAT Agrio Co., Ltd., Tokyo, Japan) were added five days after sowing. Seedlings were grown under automated control of temperature, humidity, and CO₂ concentration and illuminated by multiple 40 W white LED lamps for 21 days. Seedlings were selected based on plant height and leaf number and transplanted at a density of approximately 200 plants/m². Transplanted rice plants were grown under an automated controlled environment for approximately three months. Supplementary Table 3 provides details on the settings for the seedling and cultivation environments, and Figure 1B shows the schematic diagram of the cultivation. Fertilizers were added to the nutrient solution according to the water supply, and the pH was controlled to be below 6.0 by the automated addition of pH adjusting reagent (OAT Agrio Co., Ltd.). The panicles were harvested at least 40 days after heading, dried in a well-ventilated room until the moisture content of the grains was sufficiently reduced, and threshed. Hulls were removed manually using a husker. Extracted brown rice was polished in a rice polisher (Kett Electric Laboratory Co. Ltd., Tokyo, Japan) to achieve a polishing rate of 95% and ground into a coarse powder in a mill (Iwatani Corporation, Osaka, Japan) to produce rice flour. To prevent the spread of genetically modified plants, exhaust air from the cultivation room was passed through HEPA filters to collect particulate matter. Waste nutrient solution was collected in a waste water tank. Condensed water from the air conditioner was either reused for humidification or discharged to the effluent tank. Water in the holding tank was automatically autoclaved and disposed of.

Genomic DNA was extracted and sequenced in a HiSeq2000 platform (Illumina) in 2016 as described (Mejima et al., 2015). Paired-end read sequences of 100 bp per read (Sanger FASTQ format) from both sides of each fragment were obtained using Casava software (ver. 1.13.48; Illumina).

The flanking regions of each integrated sequence were determined as described (Mejima et al., 2015). Sequences LB-1 to -3 located near the left border (LB) and RB-1 to -4 near the right border (RB) ( Supplementary Table 4 ) in the T-DNA vector were used to identify reads containing insert-specific sequences. The matched reads were mapped onto the insert construct by using the CLC Workbench 6.0 program (CLC bio, Aarhus, Denmark), and reads with insert sequences only were excluded. The remaining read sequences were subjected to National Center for Biotechnology Information (NCBI) BLAST (Oryza Blast) to estimate the insert position in the genome. Based on this information, primers ( Supplementary Table 2 ) were designed for amplification, isolation, and cloning of the DNA, and the sequence of the genome flanking DNA and the entire sequence of the inserted DNA were determined.

Sequences of the cloned PCR products, including the genomic flanking region of chromosome 1 and the entire insertion region, were determined with Genetic Analyzer 3500 and compared to the WT Nipponbare rice genome using Genetyx software Oryza Blast (Genetyx, Tokyo, Japan).

Salt-soluble proteins from mature brown seed of MucoRice-CTB 19A and WT (Nipponbare) were extracted by using 1 M NaCl. Shotgun proteomic analysis of peptide mixtures was performed as described in Kurokawa et al. (2013) using a nanoflow LC system (Dina-2A, KYA Technologies, Tokyo, Japan) in combination with a linear ion trap Orbitrap Mass spectrometer (LTQ-Orbitrap Velos, Thermo Fisher Scientific). Protein identification was performed by searching MS and MS/MS data against the NCBI nonredundant rice protein database using Mascot (Matrix Science). Decoy database searches were also performed using Mascot and filters were applied to meet a false positive rate of less than 1%.

Mature seeds of MucoRice-CTB 19A were harvested, and total seed protein was extracted and subjected to SDS-PAGE followed by Western blot analysis as previously described (Mejima et al., 2015). CTB levels in MucoRice-CTB were determined by SDS-PAGE densitometric analysis (Yuki et al., 2013). Briefly, SDS-PAGE gels were stained with Simply Blue Safe Stain kit (Invitrogen Life Technologies) for 12 h at room temperature, and the stained spot intensity of CTB on SDS-PAGE was determined on a GS-900 calibrated densitometer (Bio-Rad Laboratories, Tokyo, Japan) using a Bio-Rad Quantity One software. Because the CTB sequence of MucoRice CTB was authentic except for the amino acid change to Gln(Q) and the addition of the N-terminal S(Ser)-R(Arg) peptide, recombinant SR-CTB (N4Q) serial dilutions (0.2, 0.4, 0.8, and 1.6 μg) were used as standards for densitometric analysis (Yuki et al., 2013).

Mature seeds of MucoRice-CTB 19A were cut into semi-thin sections (0.5 μm) on an ultramicrotome (EM UC6, Leica, Wetzlar, Germany) as previously described (Kurokawa et al., 2014). For double staining of prolamin and glutelin, sections were incubated with rabbit anti-13 kDa prolamin antibody (1:1000) or mouse anti-glutelin A monoclonal antibody (5 μg/mL) for 1 h after blocking with 1% BSA-PBS. Then, sections were incubated with Alexa647-conjugated anti-rabbit IgG antibody (1:200) or DyLight488-conjugated anti-mouse IgG antibody (1:200) in 1% BSA-PBS for 1 h. For double staining of CTB and glutelin, sections were incubated with rabbit anti-CTB antibody (10 μg/mL) or mouse anti-glutelin A monoclonal antibody (5 μg/mL) for 1 h after blocking with 1% BSA-PBS. Then, sections were incubated with Alexa647-conjugated anti-rabbit IgG antibody (1:200) or DyLight488-conjugated anti-mouse IgG antibody (1:200) in 1% BSA-PBS for 1 h. Images were captured by a confocal laser scanning microscope (LSM 800 Axio Observer, Carl Zeiss, Oberkochen, Germany).

Six- to eight-week-old female BALB/c mice (CLEA, Tokyo, Japan) were used for immunization with MucoRice-CTB 19A. Mice were maintained at 22°C under a standard 12-h light/12-h dark cycle. Food and water were provided ad libitum. All experiments were performed in accordance with the “Guidelines for the Use and Care of Laboratory Animals” and approved by the Animal Committee of the Chiba University Institute of Medical Science. WT or MucoRice-CTB line 19A powder (150 mg containing 740 μg of CTB) suspended in 750 μL PBS was used for oral immunization by gastric intubation four times at 2-week intervals. Serum and fecal extracts were collected 1 week after the last immunization.

The CTB-specific ELISA was performed as described previously (Nochi et al., 2007).

Briefly, a 2-fold serial dilution of serum and nasal wash solution was added to the well of a 96-well microplate coated with 100 μL of 5 μg/mL CTB and incubated for 2 h at room temperature. After washing, HRP-conjugated goat anti-mouse IgG or HRP-conjugated goat anti-mouse IgA (Southern Biotechnology Associates, Birmingham, AL, USA) diluted 1:4000 was added and the plates were incubated for 1.5 h at room temperature. After developing by a TMB Peroxidase Substrate kit (XPL, Caithersburg, MD, USA), the endpoint titer was expressed as the reciprocal of the log2 of the last dilution that gave an OD₄₅₀ 0.1 greater than the negative control.

In vivo challenge experiments with CT were performed as described previously (Tokuhara et al., 2010). Briefly, vaccinated mice (10 mice per group) were orally administered 20 μg CT (List Biological Laboratories, Campbell, CA, USA); 12 h later, mice were euthanized and their small intestine and colon were removed for clinical evaluation of diarrhea and collection of intestinal contents. Samples were centrifuged and intestinal water content was measured.

Group differences in data were evaluated using the two-tailed Student t test; a P value of less than 0.01 was considered statistically significant.

Seeds of four marker-free MucoRice-CTB lines MucoRice-CTB 9B, 19A, 43B, and 50B, and WT Nipponbare were germinated and seedlings were planted in soil under sunlight in a greenhouse. Average fertility and number of seeds per plant in these lines relative to WT Nipponbare are shown in Supplementary Figure 2 ; the values of both parameters were higher in 19A than in the other lines.

In Southern blot analysis, hybridization with the CTB probe detected one EcoRI fragment (ca. 9 kb) and one SacI fragment (ca. 10 kb) in line 19A ( Supplementary Figure 3 ), indicating a single copy of the CTB gene in its genome. We detected three copies in line 9B, one in line 43B, and two in 50A. We selected line 19A as a new MucoRice-CTB line because of its higher fertility and number of seeds per plant under the conditions tested, and its genetic simplicity (single copy of the transgene), which we expected to be a stable producing line.

To evaluate production stability, three consecutive cultivation experiments were conducted using the master seed bank. The results of the germination rate and pre-transplant growth assessment demonstrated the viability of the seeds and the uniformity of the environment for seedling growth (data not shown). The rice flour or active pharmaceutical ingredient used for analysis was obtained according to the process flowchart shown in Figure 1A . The brown rice yield from each of the three cultivation rounds was calculated based on six randomly selected plants, respectively. The yields were either equal to or greater than 588.5 g/m² ( Figure 1C ).

The presence of the CTB gene and absence of the HPT gene were confirmed by genomic PCR ( Supplementary Figure 4 ), and the presence of CTB proteins was confirmed by western blot analysis ( Figure 2 ). The endosperm proteins glutelin A and 13K prolamin were downregulated ( Figure 2 ). Downregulation of mRNA of glutelins A and B and 13K prolamin by RNAi was confirmed by quantitative real-time PCR ( Supplementary Figure 5 ). SDS-PAGE densitometric analysis (n=3 lot) showed that the CTB protein level was 4.94 ± 0.29 μg/mg in line 19A and 6.52 ± 0.22 μg/mg in line 51A.

To investigate the location of the transgenes, we resequenced the whole genome of MucoRice-CTB line 19A; the data are available in the DDBJ and NCBI Sequenced Read Archive under the accession number SAMD00491847 and DRX362635, respectively. We determined the position of the insertion by identifying reads containing sequences near the RB or LB sequences of the T-DNA ( Figure 3A ) and by a BLASTN search of these sequences against the Nipponbare genome. The position of the insertion was 29,910,245 bp on chromosome 1. We also amplified the flanking regions in MucoRice-CTB line 19A with the primer pairs 19Ach1-R1 and 131out2 for LB, and 10TLend5 and 19Ach1-F for RB ( Figures 3A, B ; Supplementary Table 2 ). Sequencing of the PCR products showed a 57-bp deletion corresponding to 29,910,246–29,910,302 bp on chromosome 1 of the WT rice genome ( Figure 4C ).

To clarify structure of the transgene, we amplified the inserted sequence in MucoRice-CTB line 19A chromosome 1 with the primer pairs 19Ach1 R3 and RAPint3-R, and RAPint4-F and 19Ach1-F ( Figure 4A ; Supplementary Table 2 ), and cloned and sequenced it the PCR products. We demonstrated that a single copy of the transgene (6933 bp) was inserted into chromosome 1 in the reverse orientation ( Figure 4C ; Supplementary Figure 6 ). To avoid PCR errors, we carried out PCR and cloning three times ( Figure 4B ). In comparison with the T-DNA binary vector, the inserted sequence showed (1) a deletion of the entire LB 25 bp and its downstream region (163 bp), (2) a 17-bp deletion from the 3′-end of the RB and a C-to-A point mutation in one of remaining 8 bp of RB ( Figure 4C ). One copy each of the complete CTB gene cassette and RNAi cassette were consistent with the results of the Southern blot analysis ( Supplementary Figure 3 ).

In MucoRice-CTB line 19A, the CTB signal was most abundant in the subaleurone layer and was absent in the aleurone layer (data not shown). The intensity of the CTB signal decreased toward the interior of the seed as the protein granules decreased and starch grains dominated (data not shown). CTB was prominent in the cell wall, cytoplasm, in protein bodies (PB)-I and -II, although the total amount of CTB was small ( Figure 5 ). The number and size of PB-I (stained with 13-kDa prolamin antibody) and PB-II (stained with glutelin A antibody) appeared to be lower in MucoRice-CTB than in WT ( Figure 5 ), as expected from downregulation of these storage proteins by RNAi.

The CTB (N4Q) protein was present in MucoRice-CTB 19A, with a sequence coverage of 85% ( Supplementary Figure 7 ), but not in WT. The number of identified salt-soluble proteins was 452 in MucoRice-CTB line 19A and 405 in WT rice, with an overlap of 254 proteins ( Supplementary Table 5 ). The total number of matching peptides in the MS/MS spectra was then calculated for each protein to obtain the peptide spectral match (PSM). Since PSM is proportional to the amount of a protein (Liu et al., 2004), the relative amount of MucoRice-CTB line 19A proteins overlapping with WT proteins was calculated as the PSM ratio, 19A/WT. The major identified rice allergens and their PSM ratios are listed in Table 1 . The levels of 63-kDa globulins, glyoxalase I, and α-amylase/trypsin-like protein family such as RAG2, RA5R and RA16, were lower in line 19A than in WT seeds, whereas the levels of other allergens, 19-kDa and 52-kDa globulins, did not differ. These results suggest that line 19A raises fewer safety concerns than WT rice.

In order to determine whether MucoRice-CTB line19A could induce antigen-specific mucosal immune responses, mice were orally immunized with unpurified polished flour of line 19A or WT rice. MucoRice-CTB line 19A, but not WT rice, induced serum CTB-specific IgG and IgA and fecal CTB-specific IgA immune responses ( Figures 6A–C ). These mice were challenged orally with CT. Mice immunized with MucoRice-CTB were protected from CT-induced diarrhea, since mice immunized with the same amount of WT rice flour developed severe diarrhea, as evidenced by a significantly higher intestinal fluid volume ( Figure 6D ). To evaluate the protective efficacy, we measured the intestinal fluid volume in immunized mice without CT challenge in a control experiment ( Supplementary Figure 8 ). We found a significant difference (P = 0.0469) between the control mice (267 ± 74 μL) and CT-treated mice (411 ± 201 μL). These results indicate that MucoRice-CTB strain 19A has high oral immunogenicity and is capable of inducing protective immunity.