Norwalk virus (NV) is an important agent of epidemic gastroenteritis, and an oral subunit vaccine shows potential for protection. Recombinant Norwalk virus (rNV) capsid protein expressed in plants assembles virus-like particles (VLPs) that are orally immunogenic in mice and humans. In this article we examine rNV expression in tomato and potato using a plant-optimized gene, and test the immunogenicity of dried tomato fruit and potato tuber fed to mice. The synthetic gene increased rNV expression fourfold in tomato and potato plants, which assembled VLP. Four doses of 0.4 g freeze-dried tomato fruit containing 64 µg rNV (40 µg VLPs) induced NV-specific serum IgG and mucosal IgA in ≥ 80% of mice, while doses of 0.8 g elicited systemic and mucosal antibody responses in all mice. Feedings of 1 g freeze-dried potato tuber containing 120 µg rNV (90 µg VLPs) were required to produce 100% responsiveness. Oxidation of phenolic compounds upon rehydration of dried tuber caused significant VLP instability, thus decreasing immunogenicity. Air-dried tomato fruit stimulated stronger immune responses than freeze-dried fruit of the same mass, perhaps by limiting the destruction of plant cell matrix and membrane systems that occurs with freeze-drying. Thus, rNV in dried transgenic tomato fruit was a more potent immunogen than that in dried potato tubers, based on the total VLPs ingested. These findings support the use of stabilized, dried tomato fruit for oral delivery of subunit vaccines.
Norwalk virus (NV) is a Norovirus of the Caliciviridae family and is a major cause of acute gastroenteritis in humans (Jiang et al., 1990; Clarke and Lambden, 2000). Recent estimates indicate that NV and NV-like agents are responsible for greater than 96% of outbreaks of acute nonbacterial gastroenteritis in developed and developing countries (Vinje et al., 1997; Fankhauser et al., 1998). The Centers for Disease Control and Prevention in Atlanta attributes about 23 million cases of gastrointestinal illness in the USA each year to Norovirus infections (Mead et al., 1999). Epidemic outbreaks of this disease frequently occur in cruise ships, schools, day care centres, nursing homes, and hospitals, and can be especially problematic in military settings (Hyams et al., 1993). The increasing clinical significance of these infections suggests the pressing need of an efficacious vaccine against NV (Estes et al., 2000).
The NV genome contains a positive-sense, single-stranded, non-enveloped RNA. Its second open reading frame encodes a single NV capsid protein (NVCP) that self-assembles into empty virus-like particles (VLPs), lacking viral RNA, when expressed in the baculovirus/insect cell expression system (Jiang et al., 1992, 1993) and plants (Mason et al., 1996). X-ray crystallography of recombinant NV VLPs (rNV VLPs) showed that these VLPs are composed of 90 dimers of the NVCP that form a T = 3 icosahedral structure with a diameter of 38 nm (Prasad et al., 1999). The rNV VLPs are stable at low pH, when lyophilized, and when stored long-term at 4 °C (Estes et al., 2000). The insect cell-derived VLPs are immunogenic in experimental animals (Mason et al., 1996; Ball et al., 1998) and in human volunteers following oral administration (Ball et al., 1998, 1999), and in mice when administered parenterally (Jiang et al., 1992) and intranasally (Guerrero et al., 2001). These qualities make the rNV VLPs potentially useful as a vaccine against NV.
In this article, we evaluate tomato fruit as a vehicle for expression and oral delivery of rNV and show that freeze-dried or air-dried fruits are highly immunogenic in mice. Our previous studies in potato were limited by low levels of expression of NVCP, inefficient assembly of immunogenic VLPs and dependence on ingestion of raw tuber for delivery (Mason et al., 1996; Tacket et al., 2000). In the present work we used a plant-optimized NVCP to show that transgenic tomato fruit provides higher expression of rNV than potato tuber and is a more potent oral immunogen. Furthermore, we find that air-dried tomato fruit can induce more robust immune responses than freeze-dried powder, indicating that a very convenient stabilization process has utility for tomato-derived vaccine production. These studies serve as a model for oral immunization with drying-stabilized plant-made vaccines.
Optimized sNVCP gene enhanced mRNA and protein accumulation
The sNVCP gene was designed by altering 108 of 1584 nucleotides in the native NVCP gene to substitute dicotyledonous plant-favoured codons, and to remove 28 ‘CG’ dinucleotides, 4 ‘CNG’ potential methylation sites and 4 potential polyadenylation motifs. The percentage of GC changed from 47% in the native gene to 45% in the designed gene, and a total of 86 codons were changed. pNV110 and psNV110 (Figure 1) were used to transform tomato and potato plants. For tomato, 74 and 24 independent rNV-expressing plantlets were obtained for pNV110 and psNV110, respectively. For potato, 57 and 36 positive lines were obtained for pNV110 and psNV110, respectively. The five highest-expressing tomato lines (T1 generation) and 10 best potato lines (T0 generation) for each construct were grown in soil in the greenhouse to allow comparison of their expression performance in fruits and tubers. The sNVCP gene allowed approximately fourfold higher mRNA and rNV accumulation in both tomato (P = 0.0002, t-test) (Figure 2A,C) and potato plants (P = 0.0001, t-test) (Figure 2B,D). The slopes of the linear regression lines (rNV levels vs. mRNA abundance) were similar for NV110 and sNV110 constructs (data not shown), indicating that more stable transcript accumulation in sNV110 lines was a key factor for the higher accumulation of rNV protein in both plants. The maximum rNV level in sNV110 lines were up to 8% of the total soluble protein (TSP) in tomato fruit and 0.4% in potato tubers. Fruits in the development stages from mature green to turning/pink consistently showed higher levels of rNV than did fully red ripened fruit. The selected tomato line sNV110-2 contained one and potato line sNV110-40 contained two copies of sNV gene, as determined by Southern blot (data not shown). Potato line sNV110-40 was chosen for further studies because it produced higher tuber yields than some higher-expressing lines.
Tomato fruit produces both 23 nm and 38 nm VLPs
Sucrose gradient sedimentation of i-rNV showed two peaks: a more abundant faster-sedimenting VLP (38 nm) and a slower-sedimenting VLP (23 nm) (White et al., 1997). We estimated the proportion of plant-derived NVCP that was assembled as 23 nm or 38 nm VLP by sucrose gradient sedimentation (Figure 3). The percentage of NVCP sedimenting as 23 nm or 38 nm VLPs averaged about 41% and 22%, respectively, in fresh tomato fruit and about 24% and 40%, respectively, in fresh potato tubers (Figure 3A, Table 1). Thus, tomato and potato contained similar total amounts of VLP, but differed in the relative proportions of 23 nm and 38 nm VLPs. To determine if rNV accumulation levels affected VLP assembly efficiency, we analysed crude protein extracts from tomato fruit having different rNV expression levels and found no significant differences among lines (data not shown). We also compared VLP proportions in fresh, freeze-dried, and air-dried materials and found freeze-drying and air-drying processing did not affect VLP content in either tomato or potato plants (data not shown).
Table 1. The feeding and ingested dosages in mouse experiments. *, 23 nm rNV particles from sucrose fractions 3–5, which may also contain oligomers. **, the majority of rNV in fractions 3–5 were likely soluble subunits and oligomers, since potato browning disassembled VLPs
Further analysis of VLPs by negative staining and transmission electron microscopy (TEM) revealed that rNV VLPs from tomato fruit and potato tubers, which cosedimented with faster-sedimenting insect cell-derived VLPs were about 38 nm in diameter and resembled i-rNV in appearance (Figure 3B-a and c). The slower-sedimenting peak fractions 3–5 showed smaller particles (∼23 nm) (Figure 3B-b). Further analysis with immuno-electron microscopy (immuno-EM) demonstrated that both 23 nm and 38 nm VLPs from tomato bound to rNV-specific antibody (data not shown), consistent with similar observations of i-rNV (Hardy et al., 1996). EM examination of extracts of nontransgenic materials did not reveal any such particles (data not shown).
Oral immunogenicity of transgenic potato expressing synthetic NVCP gene
Potato powder (1.5, 2.5 or 5 g) was fed to each mouse on days 1, 4, 17 and 20. The actual doses ingested by each mouse were, on average, 1, 1.2 and 2 g per feeding, containing 120, 144 and 240 µg rNV/dose, respectively (Table 1). Potato powder in two groups (2.5 and 5 g) was reconstituted with 0.6 volumes of sterile water to facilitate ingestion by mice. Nontransgenic potato Desiree powder (5 g per mouse per feed) was given to a negative control group, and 100 µg purified i-rNV VLPs were orally administered by gavage to a positive control group. Serum and fecal NVCP-specific antibodies were measured by ELISA, and geometric mean titres (GMTs) were calculated for each group of mice. All pre-immune serum and fecal samples were negative (titre < 20), as were those of the control group that received nontransgenic potato. The time courses of rNV-specific serum IgG and intestinal IgA responses are shown in Figure 4(A,B). No primary serum IgG was detected after two doses in the three groups receiving transgenic potato, but fecal IgA GMT at 11 days postingestion (dpi) were of 49, 15 and 29 for the groups that ingested 1, 1.2 or 2 g of potato powder, respectively. After the third and fourth doses (days 17 and 20), serum antibody responses were detected in all mice in the group that ingested 1 g of powder, with IgG GMT peaking at 201. Mucosal responses increased quickly and reached a peak in the 1-g group with IgA GMT = 1293 at 27 dpi. The serum IgG responses to potato rNV were lower than those of mice that were gavaged with 100 µg i-rNV VLPs, but fecal IgA responses were similar. In the other two groups that received more potato materials, only 20%−40% mice gave measurable antibody responses with IgG GMTs less than 36 and IgA GMTs less than 117.
We checked the stability of VLPs in potato materials, in order to determine why higher doses performed poorly. The two higher-dose potato powders had been mixed with water, which resulted in potato browning due to oxidation, and the 38 nm VLP proportion was decreased to 15 µg/g dry weight (P = 0.0004, t-test) (Figure 4C,D). Thus, the actual doses for these groups ingesting 1.2 and 2 g were ∼18 and ∼30 µg of 38 nm VLPs/dose, while the 1 g dose group received ∼63 µg of 38 nm VLPs/dose.
Tomato expressing sNVCP induced robust systemic and mucosal antibody responses
Fruits from turning to pink ripening stages were used for oral immunogenicity studies. Freeze-dried fruit powder of tomato line sNV110-2 (1.5, 2.5 or 5 g) was fed to mice on days 1, 4, 17 and 20. The actual doses ingested by each mouse averaged 1.2, 1.5 and 2.2 g per feeding, containing 192, 240 and 352 µg rNV/dose, respectively, or 60, 75 and 110 µg VLPs/dose (Table 1). Nontransgenic tomato TA234 powder (5 g per mouse per feed) were given to a negative control group, and 100 µg purified i-rNV VLPs were orally administered by gavage to a positive control group. All pre-immune serum and fecal samples were negative for rNV-specific antibodies, as were control group (titres < 20; the lowest dilution tested). The time courses of NVCP-specific serum IgG and intestinal IgA responses are shown in Figure 5. All mice that ingested 1.2 g or more of tomato sNV110-2 powder consistently showed rapid responses for both serum and intestinal antibodies after the first two feedings (day 11) with IgG GMTs ranging from 64 to 192, and IgA GMTs from 1718 to 2465, respectively. No significant differences (P = 0.05) in day 11 GMTs were observed among groups fed different doses of transgenic tomato. In contrast, for the i-rNV gavage group, only one of the five mice showed a serum antibody response on day 11, and two of the five had primary intestinal antibody responses. After the third (day 17) and the fourth (day 20) feeding of transgenic tomato powder, IgG titres in all groups substantially increased, then levelled off on day 55 with GMTs ranging from 6775 to 8221. For the group that ingested 1.2 g sNV110-2 powder, serum antibody responses remained high (IgG GMT = 2000) for at least 6 months (see booster experiment below). Intestinal antibody responses reached their peaks at 27 dpi with IgA GMTs ranging from 10 681 to 19 127. Linear regression analysis revealed that increases in IgG and IgA GMTs were highly correlated with dosage. However, only the group that consumed 2.3 g of sNV110-2 tomato powder consistently had significantly higher IgA GMTs than the one consuming 1.2 g at all time points (P < 0.025–0.0367), indicating that 1.2 g of tomato powder was sufficient for inducing robust serum antibody responses in mice. In contrast, the control i-rNV gavage group had their maximum levels of IgG GMT = 826 and IgA GMT = 1748, which were 8.2- and 6.2-fold less than the group that ingested 1.2 g of sNV110-2 powder (P < 0.0183); and their long-term serum antibody responses declined to baseline in 3 months (IgG GMT = 58 at 87dpi).
Minimum dosages of transgenic tomato required for eliciting antibody responses
In order to determine the minimum dosage required for inducing immune responses, we tested doses of 0.1, 0.4 and 0.8 g of either sNV110-2 freeze-dried tomato powder (mixed with 0.6 volumes water) or air-dried fruit (Figure 6). Immunization schedules were similar to the previous studies, except that for some groups only two (days 1 and 20) or three (days 1, 4 and 20) feedings of 0.8 g of transgenic materials were given. One week after the first two feedings, serum antibody responses were detected in 20% of mice receiving two doses of 0.8 g of either powder or air-dried fruit. Intestinal antibody responses were stronger than serum responses, with 20% of mice showing IgA increases with two doses of 0.4 g of sNV110-2 powder or air-dried fruit, and 40% and 100% of mice showing intestinal antibody responses with 0.8 g of materials, with IgA GMTs = 45 and 564, respectively. Further, 60% of mice that ate 0.8 g air-dried fruit showed IgA responses after one dose (GMT = 53), while none of the freeze-dried powder group had detectable fecal antibody after one feeding of 0.8 g of powder having the same amount of rNV or VLPs. t-Test indicated a significant difference (P < 0.036–0.047) between these groups, indicating that at 0.8 g dose, air-dried fruit was a more potent immunogen.
On day 27, 1 week following two additional feedings of 0.1, 0.4 and 0.8 g of sNV110-2 powder (on days 17 and 20), 20%, 80% and 100% mice showed significantly increased IgG responses with IgG GMTs of 30, 117 and 350, respectively (Figure 6A). There were significant differences between the 0.1-g group and the latter two (P = 0.0457 and 0.0216, Mann–Whitey test). Similar trends were observed in three groups that ingested air-dried materials, with IgG GMTs of 36, 120 and 4487, respectively. Serum antibody responses in the group that received 0.8 g air-dried fruit were significantly greater than those of groups receiving 0.4 g of air-dried fruit (P < 0.0122, Mann–Whitey test) and 0.8 g of powder (P < 0.035, Mann–Whitey test). In groups that ate a total of two or three feedings of 0.8 g of sNV110-2 powders, only 20%−40% of mice showed serum antibody responses with very low IgG GMTs (ranging 27–54). However, 100% of the group receiving only two doses of 0.8 g of air-dried fruit had IgG responses (GMT = 314), which was significantly higher than the counterpart group receiving powder (P < 0.0122).
Intestinal IgA responses on day 27 showed similar responses (Figure 6B). All mice receiving 0.4 g or more powder showed IgA responses with IgA GMTs = 397 for the 0.4-g group and = 730 for the 0.8-g group. The counterparts consuming air-dried materials had IgA GMTs of 710 and 8828, respectively. For the doses of 0.1 g tomato powder or dried fruit, 20% and 40% of mice had IgA responses with IgA GMTs of 25 and 50, respectively. These titres were similar to those of the groups that received two or three doses of 0.8 g powder. For the two-dose group with 0.8 g air-dried fruit, IgA responses, as with serum IgG, were detected in all mice, with GMT of 702.
The overall profiles of antibody responses among the groups at 41 dpi were similar to those at 27 dpi. Two exceptions were observed: (i) IgG and IgA GMTs continued to increase for the group receiving a total two doses of 0.8 g powder; and (ii) IgA GMTs for all other groups began to decline.
Booster responses show immune memory is established
In order to determine if immune memory was established, we boosted groups that ingested two doses of tomato powder (0.1 and 1.2 g) or potato powder (1.0 g), by gavage with 10 µg i-rNV VLPs at the times indicated (Figure 7). For the group that was primed with 0.1 g tomato powder, only one of the five mice showed a primary antibody response with a peak titre of 184 (group GMT = 30) (Figure 7A). At 55 dpi, the IgG GMT returned to a baseline. One week later, all mice received a booster of 10 µg VLPs. An immediate and strong secondary antibody response was induced in all five mice, including the mice whose primary antibody response was below detectable levels. A peak IgG response with GMT of 2487 was measured 2 weeks after boosting. Mice fed control tomato powder showed no immediate response and 3 weeks after boosting showed IgG GMT < 40.
Similar trends resulted in the group gavaged with 100 µg i-rNV VLPs and the group that ingested potato vaccine. For potato vaccine, a single boost after IgG titres returned to a baseline gave an immediate and strong increase in antibody titres, with a peak of 3688 at 2 weeks after boosting. For groups fed with high dosages of tomato, the induced antibody responses were much longer lived; for example, IgG GMT in the group consuming 1.2 g tomato powder was still as high as 2000 at 180 dpi. This group did not return to baseline before boosting, but a booster gavage of 10 µg i-rNV VLPs resulted in 10-fold increase in IgG GMT. Fecal IgA titres in this group (Figure 7B, tomato) declined greatly after a peak on day 25, but remained stable at GMT 100–200 from day 55 to day 180. The booster gavage of 10 µg i-rNV VLPs at that time provoked an immediate 50-fold increase in IgA GMT.
We previously reported production of rNV in tobacco and potato plants (Mason et al., 1996), and oral immunization with either tobacco-derived or raw potato tubers stimulated the production of humoral and mucosal antibody responses in mice. Furthermore, human clinical trials demonstrated that 95% of volunteers who ate uncooked potatoes containing rNV developed an immune response of some kind (Tacket et al., 2000). In both animal and clinical studies, however, the antibody titres were low or moderate. We realized that immunogenicity was limited by both the amount of rNV subunits produced and their assembly to make VLPs in plant materials, which affects the actual dose of immunogenic material that mice and human volunteers ingested.
In order to develop a more efficacious oral vaccine, we sought to enhance expression of rNV and/or assembly efficiency of rNV VLPs in plants. In other studies we showed that tomato fruit could be an effective vehicle for oral delivery of rNV: freeze-dried fruit expressing the native NVCP gene is orally immunogenic at approximately 80 µg per dose (3 g dry fruit powder) in mice (Huang et al., 2005). However, a larger dose of antigen (Tacket et al., 2003) in a smaller mass of vehicle tissue is required for human use. Thus, increased expression of NVCP in tomato was needed in order to make it a feasible system.
By using a plant-optimized sNVCP gene, we achieved a fourfold increase in rNV accumulation in tomato and potato plants compared to the native NVCP gene (Figure 2). Because mRNA levels were higher in the sNVCP plants, the optimized gene probably increased transcription rate, mRNA processing efficiency and/or stability. The removal of potential DNA methylation sites (28 ‘CG’ and 4 ‘CNG’) could have reduced methylation-mediated gene silencing. However, we cannot rule out the possibility that increased translational efficiency was a factor, since 86 of 528 codons were changed in the plant-optimized gene. Although this increased accumulation of NVCP did not enhance VLP assembly (proportion of NVCP in VLP fraction), the amount of VLPs was increased accordingly on the basis of dry weight.
We tested the potential of tomato fruit and potato tuber, with enhanced accumulation of rNV and VLPs, to provoke stronger immune responses in a mouse model. We also examined whether there was correlation between immune response and dosage (Tacket et al., 2003), which will inform the design human clinical trials using plant-derived rNV immunogen. We show here that four feedings of 0.4 g freeze-dried tomato powder, without any adjuvant, induced NV-specific serum IgG in 80% and mucosal IgA in 100% of mice, while 0.8 g tomato powder doses gave 100% responses for both serum and intestinal antibodies. Furthermore, two feedings of 1.2 g of transgenic tomato powder, containing ∼192 µg rNV/dose (60 µg 38 nm VLPs/dose), given 3 days apart, consistently stimulated primary serum and fecal antibody responses in all mice (IgG and IgA GMTs of 64 and 1718 at 11 dpi). In our hands, oral administration of two doses of 100 µg purified i-rNV VLPs by gavage gave primary responses in at best 40% of mice. Four feedings of the high-dose (192 µg) tomato vaccine resulted in peak IgG and IgA titres that were significantly higher than those produced by 100 µg i-rNV VLP gavage. Although 192 µg of tomato-derived rNV contained 60 µg 38 nm VLPs, the 23 nm particles contributed to the total VLP content, which we estimate was 120 µg per dose.
The enhanced potency of immunization by the tomato rNV has several possible explanations. First, the cell wall matrix and membrane materials of plants may serve as bioencapsulation elements, which could allow rNV to survive exposure to the enzymes in the gut. Oral immunization by gavage with purified i-rNV VLPs required delivery of relatively large doses to achieve immunization (Ball et al., 1998), even though they are stable at the pH of the stomach. This may be due to enzymatic degradation as the VLPs traverse the gastrointestinal tract. The natural bioencapsulation of the protein within plant cells may permit slow and longer-lasting release of VLPs that can be taken up by gut epithelial cells and delivered to the mucosal immune system. Alternatively, the VLPs may interact with a specific enterocyte receptor to facilitate uptake and presentation to immune cells (White et al., 1996; Ball et al., 1998; Guerrero et al., 2001). Since tomato-expressed rNV was a more effective immunogen than potato-expressed rNV, we hypothesize that plant cell encapsulation is more substantially retained in dried tomato fruits than in dried potato tuber.
Second, α-tomatine, an alkaloid glycoside of tomato (Friedman, 2002), may serve as a natural adjuvant potentiating immune responses in mice. A molecular aggregate formulation, which was based on α-tomatine, cholesterol, and phosphatidylethanolamine, was reported to act as an effective adjuvant to augment either humoral or cellular immune responses (Rajananthanan et al., 1999a, 1999b; Heal et al., 2001). We did not measure α-tomatine levels in our fruit preparations, and thus we can only conjecture regarding its possible effect in this experiment. Tomatine is abundant in Lycopersicon pimpinellifolium, but substantially lower in cultivated tomato. Furthermore, the α-tomatine level is high in green fruit (Lycopersicon esculentum cv. ‘TA234’, 500 µg/g fresh weight), but decreases substantially by 10 days after flowering, and is only about 5 µg/g fresh weight in ripening and ripened fruit (Friedman, 2002). Nonetheless, potato tubers also contain limited amounts of α-tomatine (Kozukue et al., 1999), but did not immunize as efficiently as tomato rNV.
A third possibility is that smaller 23 nm VLP contributed greatly to immunogenicity in mice. The proportion of the total NVCP antigen represented by 23 nm particles was ∼25%−42% in tomato extracts (Figure 3). Similarly, the expression of rNV in an insect cell system resulted in the assembly of particles of two sizes, where the amount of smaller particles varied from one preparation to another, from 0 to 30% of the total VLP population (White et al., 1996; White et al., 1997). Particles of two sizes with similar morphology also occurred in stool samples from patients infected with a human calicivirus (Taniguchi et al., 1981). In fact, the self-assembly of viral capsids composed of a single structural protein into two structures has been found for a number of T = 3 icosahedral RNA viruses (Harrison et al., 1996). The mechanism that controls the assembly of capsid proteins into larger or smaller particles is still unknown, but it might be related to charge configuration in the putative nucleation complex (five-dimer nuclei), and thus might be regulated by pH, divalent cations, and the presence of additional charged domains (Erickson et al., 1985). Since several properties are conserved between the two particle types, including biochemistry, antigenicity and ability to bind to cultured human intestinal cells (White et al., 1996, 1997), it would be reasonable to believe that 23 nm VLPs in tomato materials are immunogenic in our experiment. Nonetheless, further study on the immunogenicity of 23 nm particles is needed.
We have shown that rNV in air-dried tomato fruit was a significantly more potent immunogen than that in freeze-dried tomato powder. A possible explanation is that more of the plant cell matrix and membrane system is conserved by air-drying, thus providing greater protection of rNV. This result has a particular importance for application of plant-derived vaccine technology in developing countries, because air-drying is a more economical method of fruit processing than freeze-drying. While we realize that vaccine production in any location must adhere to stringent protocols of good manufacturing practice, air-drying of fruit, when justified by comprehensive studies, could have a significant cost benefit.
We showed that freeze-dried potato tuber rNV provided less immunogenicity than tomato materials. One reason was that VLPs were disassembled in some higher-dose groups where potato powder was mixed with water. The disassembled subunits (soluble and oligomers) might be cross-linked due to oxidized polyphenolic compounds (Belitz and Grosch, 1987; deMan, 1999) that are common in potato tuber. Thus, only 20−40% of mice that ingested higher doses of rNV potato powder had detectable immune responses (Figure 4). These results confirmed the suggestion that soluble proteins or oligomers induce little or no immune response (Ball et al., 1998; Tacket et al., 2000). For the group that ate nonbrowned powder (not mixed with water prior to delivery), low titres of serum antibodies were detected, though all mice had positive responses. This result suggests that the potato variety ‘Desiree’ might contain some immuno-inhibiting substance. Similar results were seen in the studies with recombinant Escherichia coli heat-labile enterotoxin B (rLT-B) expressed in Desiree potato (Lauterslager et al., 2001), compared to potato variety ‘FL1607’ (Mason et al., 1998). Oral immunization of naive mice by feeding rLT-B tubers or by intragastric intubation of rLT-B extracts did not evoke detectable IgG and IgA antibody titres; however, subcutaneous immunization with preparation of rLT-B extracted from tubers yielded high antibody titres in serum (Lauterslager et al., 2001). In our studies, we found that, although lower IgG titres were detected in the group ingesting 1.0 g potato powder, fecal IgA response was similar to that obtained from the group receiving 100 µg i-rNV VLPs, indicating that VLPs in potato tubers were able to induce and develop immunization at local mucosal sites. This result was consistent with the finding by Lauterslager et al. (2001) that oral booster immunization with rLT-B Desiree tubers in primed mice resulted in anti-LT-B IgA responses in serum and faeces, but did not stimulate production of IgG. All of these observations suggest that ‘Desiree’ potato tubers contain higher levels of phenolic compounds that may contribute to tissue browning and affect rNV antigenicity upon oxidation than do ‘FL1607 tubers’. Further experiments directly comparing the same antigen expressed in ‘Desiree’ and ‘FL1607’ potatoes may resolve this question.
Oral immunogens, especially nonreplicating antigens, often induce tolerance after multiple exposures in small doses or even a single high dose (Sosroseno, 1995). Our results with both tomato and potato expressing NVCP showed that they were not tolerogenic for this antigen. After each oral feeding, the IgG and IgA titres increased or the number of responders increased. Furthermore, one booster gavage with a low dose of i-rNV VLPs (10 µg/mouse) 6 months after priming with plant rNV increased serum IgG and intestinal IgA responses (Figure 7). We used oral rather than parenteral (injection) delivery for the boosting experiment because we anticipate that a human vaccine for NV will be orally delivered, and because the published mouse studies have used mucosal (oral or intranasal) delivery routes (Mason et al., 1996; Ball et al., 1998; Guerrero et al., 2001). Although the gastrointestinal tract of mouse is quite different from that of human, several clinical studies have shown oral immunogenicity of rNV in humans similar to that in mice (Ball et al., 1999; Tacket et al., 2000, 2003), thus indicating that mouse is a good model.
Establishment of a memory response is one of the key factors for vaccine design and implementation (Kong et al., 2001). Our studies show that feeding 0.1 g tomato powder or 1 g of potato powder stimulated a primary immune response and that a single booster of low dose provoked a rapid and robust recall response after the specific antibodies had declined to baseline. No booster responses were obtained in mice that were fed nontransgenic materials, showing that the boosting response elicited in mice was indeed the result of priming and establishment of immune memory to rNV VLPs presented in the gut. Interestingly, the mice that were immunized with potato or low-dose tomato (0.1 g) mounted a rapid antibody boosting response even after titres had fallen to near baseline levels at 120 dpi. The higher-dose tomato group showed much more robust antibody production and less decrease over time than the low-dose groups, yet they still were strongly boosted. Thus, repeated exposures of mice to low doses of tomato-derived rNV do not cause immune tolerance, and immune memory cells were established in these mice. This suggests a practical immunization strategy: multiple low-dose delivery of an oral subunit vaccine as a useful approach to large-scale immunization in developing countries, especially for infants and young children, who are unable to consume large doses.
Taken together, our data show that oral immunization of rNV expressed in transgenic tomato and potato can elicit systemic and mucosal antibody responses. Furthermore, rNV in tomato fruit, especially in air-dried material, is a more potent immunogen than potato. The robust immunogenicity of the tomato-derived vaccine likely resulted from natural bioencapsulation by the plant cell matrix and membrane systems, larger amounts of smaller 23 nm particles and natural adjuvants like tomatine. These preclinical data will be used to design a phase I/II clinical trial with tomato-derived rNV.
Construction of plant expression vectors
The plasmid pNVCP3, containing the plant-optimized NVCP gene (GenBank accession AY360474) in an XbaI/SacI fragment, was obtained from Axis Genetics PLC (UK). The binary vector plasmid pPS1 (Huang and Mason, 2004) contains the nptII gene, dual-enhancer 35S promoter, tobacco etch virus (TEV) 5′ UTR and vspB 3′ sequence. The plasmids were digested with XbaI and SacI and the two fragments ligated to create psNV110. The native NVCP gene from pNV102 (Mason et al., 1996) was ligated with pPS1 to make pNV110.
pNV110 and psNV110 were electroporated into Agrobacterium tumefaciens LBA4404, which were then used to transform tomato (L. esculentum cv. TA234) and potato (Solanum tuberosum L. cv. Desiree) by modified stem or cotyledon cocultivation methods (Frary and Hamilton, 2001). Kanamycin-resistant transformants were screened by PCR, RNA blots and ELISA (below). Transgenic tomato and potato lines with high expression levels were clonally propagated and transferred to soil in pots to produce tomato fruit and potato tubers by standard greenhouse culture.
Protein extraction and NVCP ELISA
Plant materials, except tomato fruits, were homogenized in extraction buffer (10 mm sodium phosphate, pH 6.6; 100 mm NaCl; 1 mm EDTA; 1 mm phenylmethylsulphonyl fluoride; 0.1% Triton X-100; 50 mm sodium ascorbate). For tomato fruit, the buffer was modified by excluding sodium ascorbate and adjusting pH to 8.0. NVCP in plant extracts was quantified by ELISA as described previously (Mason et al., 1996), with slight modifications. Rabbit anti-i-rNV serum diluted 1 : 10 000 in 10 mm phosphate-buffered saline (5.6 mm Na2HPO4, 3 mm KH2PO4, 100 mm NaCl. PBS, 50 µL per well) was bound to 96-well polyvinylchloride microtiter plates (Fisher) for at least 4 h at room temperature, or overnight at 4 °C, and the plates were blocked with 5% nonfat dry milk in PBS (5% BLOTTO) for at least 1 h at 37 °C. After washing the wells four times with PBS containing 0.05% Tween 20 (PBST), samples (50 µL each well) diluted in PBS were added and incubated 16 h at 4 °C, or 2 h at 37 °C. The wells were washed and incubated in succession with guinea pig anti-i-rNV serum, diluted 1 : 40 000 in 1% BLOTTO, and goat anti-guinea pig IgG-horseradish peroxidase conjugate, diluted 1 : 5000 in 1% BLOTTO, for 2 h at 37 °C. The plate was developed with 3, 3′, 5, 5′-tetramethylbenzidine (TMB) substrate (Kierkegaard & Perry Laboratories, Gaithersburg, MD) for 5–10 min at room temperature, and the reaction was stopped by adding an equal volume of 0.5 N H2SO4. For a standard curve, i-rNV (Mason et al., 1996) was diluted with PBS to concentrations between 0.2 and 25 ng/mL and processed as above. TSP was measured by the Bradford dye-binding procedure (Bio-Rad, Hercules, CA), using bovine serum albumin (BSA) as a standard.
RNA extraction and blot
Total RNA from leaf was extracted with Trizol Reagent (Gibco BRL, Carlsbad, CA). Total RNA from potato tubers was isolated by a modified phenol/chloroform method. Approximately 1.6 g frozen tuber tissue was ground in a prechilled mortar with liquid N2 and then transferred to a 13-mL polypropylene tube with the screw cap. Four millilitres of extraction buffer (100 mm LiCl; 100 mm Tris, pH 8.5; 10 mm EDTA; 1% SDS; 15 mm DTT) and 4 mL phenol (pH 4.3) were added to the frozen tuber powder and heated to 65 °C for 5 min and vortexed vigorously for 1 min. Four millilitres ofchlorophorm : isoamyl alcohol (24 : 1) were added and vortexed for 30 s. The homogenates were centrifuged 15 min at 7500 g at 4 °C, and the aqueous phase was transferred to a fresh 13 mL polypropylene tube, and further clarified by addition of 4 mL chlorophorm : isoamyl alcohol (24 : 1), vortex and centrifugation as above. The cleared aqueous phase was transferred to sterile, RNase-free tubes and mixed with an equal volume of 4 m LiCl and kept on ice overnight. The RNA was pelleted by centrifugation for 25 min at 16 000 g at 4 °C.
Total RNA was purified from tomato fruit as previously described previously (Lonneborg and Jensen, 2000). In an oakridge tube, 10 mL extraction buffer [4%p-aminosalicylic acid; 1% 1,5-naphthalenedisulphonic acid, disodium salt hydrate; 3 mL CTAB buffer (10% cetyltrimethylammonium bromide; 1.6 m NaCl); 10 mL buffer-saturated phenol (pH 4.3)] were mixed and heated to 70 °C in water bath for 10 min. Approximately 3.5 g fruit powders, ground previously in liquid N2 and kept −70 °C, were added to the tube and vortexed vigorously for 30 s. After mixing with 10 mL chloroform : isoamyl alcohol (24 : 1), the homogenates were centrifuged 20 min at 16 000 g at 4 °C. The aqueous phase was mixed with 2 volumes of ethanol and precipitated 15 min at room temperature. After centrifugation for 15 min at 16 000 g at 4 °C, the resulting nucleic acid pellet was resuspended in 2 mL diethylpyrocarbonate (DEPC)-treated water, mixed with an equal volume of 4 m LiCl, and then precipitated on ice overnight. The resulting RNA pellet was centrifuged for 20 min in a microcentrifuge (16 000 g) at 4 °C.
Five micrograms of total RNA was denatured and fractionated on 1% formaldehyde gels (Mason et al., 1996), capillary blotted to Zeta probe membrane (Bio-Rad, Hercules, CA), and hybridized with a combination of DIG-labelled NVCP and sNVCP probes, following the protocols provided by the manufacture (Boehringer Mannheim, Indianapolis, IN). The probes were generated by PCR using primers NV-for 5′-CCCCCGGTGATGTTTTGTTTGA-3′ and NV-rev 5′-GAGGGGCACGTGAGTTAGA-3′ for NVCP and sNV-for 5′-CTTGTTTTTAGTCCCTCCTACAGT-3′ and sNV-rev 5′-GCTAGCAAGATGTGAAATGTACTC-3′ for sNVCP. The signals were detected by Image station 440 CF and quantified with Kodak 1D IMAGE analysis software (Eastman Kodak Company, New Haven, CT).
Assay for rNV VLP assembly
NVCP assembled as VLP in freeze-dried tomato fruit and potato tuber was assayed by sucrose gradient sedimentation as described previously (Mason et al., 1996; Tacket et al., 2000). The gradients were constructed by layering 2.2 mL each of 50%, 40%, 30%, 20% and 10% sucrose/PBS into a Beckman SW41 Ti tube (Beckman Coulter, Fullerton, CA) and incubated for at least 2 h at 4 °C to allow diffusion into a continuous gradient. Extracts (1 mL) in protein extraction buffer were layered on the gradients, which were centrifuged in a Beckman SW41 Ti rotor at either 53 000 g for 14 h or 90 000 g for 3 h at 4 °C. Gradients were fractionated by bottom puncture and aliquots of each fraction were analysed by ELISA as described above or Western blotting (below). VLP standard (i-rNV, produced in insect cells and purified by sedimentation through 30% sucrose cushion followed by isopycnic centrifugation in CsCl) was sedimented in a similar gradient and was analysed by ELISA to identify the VLP fractions. The proportion of the total ELISA- or immunoblotting-positive NVCP material in the sucrose gradient that cosedimented with the VLP standard was calculated to obtain the percentage of VLPs in plant materials. Peak fractions (23 nm VLP in fractions 3–5 and 38 nm VLP in fractions 7–10) were pooled separately, and the particles were concentrated by pelleting for 3 h at 210 000 g. Pellets suspended in sterile MilliQ water were used for negative staining and EM.
Sodium dodecylsulphate–polyacrylamide gel electrophoresis and Western blotting
Sucrose gradient fractions were electrophoresed on 4%−20% gradient polyacrylamide gels, and proteins were electrophoretically transferred to polyvinylidene difluoride (PVDF) membrane (Amersham Pharmacia, Pittsburgh, PA) and probed in succession with rabbit anti-rNV serum at dilution of 1 : 3000 in 1% BLOTTO and goat anti-rabbit IgG-horseradish peroxidase conjugate (Sigma, St Louis, MO) diluted in 1 : 10 000 in 1% BLOTTO. The membranes were developed by chemiluminescence using the ECL plus detection reagent (Amersham Pharmacia) following the manufacture's protocol, and the signals were detected with a Storm system (Molecular Dynamics) and quantified with IMAGEQUANT software.
A 10 µL drop of the sample was applied to a parafilm sheet, and a formvar/carbon-coated grid (300 mesh) was set on the drop for 15 s. Excess liquid was absorbed from the side of the grid with a filter paper. Then the grid was placed on a drop of 2% aqueous uranyl acetate for 15 s, and excess fluid was blotted as before. After air-drying, the grid was examined on a Tecnai 12 Biotwin transmission electron microscope (FEI Company, Eindhoven, The Netherlands).
Oral immunization and antibody measurement
Tomato fruit of line sNV110-2 was harvested at the pink stage of ripening, the seeds were removed and the remaining flesh was cut into pieces. Fruit tissue was either kept at −20 °C, to be freeze-dried at −40 °C (LABCONCO), or left in a standard chemical hood for air-drying for 2 weeks. The air-dried fruit was put in an oven at 37 °C for 24 h before feeding mice to allow more evaporation of moisture. Potato tubers of line sNV110-40 were cut into 1–2 cm3 cubes and left in 0.1% sodium ascorbate for 10 min before freeze-drying.
Mice were fasted for 4 h before feeding with plant-derived vaccine materials. We initially tested oral immunogenicity of large doses (1.5, 2.5 or 5 g) of freezer-dried tomato fruit or potato tuber given to each BALB/c mouse on days 1, 4, 17 and 20. After the first feeding on day 1, in order to increase ingestion by mice, the powder was reconstituted with 0.6 volumes of sterile water except for one group (1.5 g potato powder). Five gram of nontransgenic tomato and potato materials were fed to two negative control groups. The i-rNV VLPs (100 µL, 1 µg/µL in 0.9% sodium chloride) were orally delivered by gastric intubation (gavage) to a positive control group that was not fasted before immunization. Each mouse was housed separately for the feedings and the amount of diet weighed before and after feeding to monitor the amount ingested. The plant vaccine material was left in the cage until all of it was consumed or for 16 h, whichever was earlier. Other than during vaccine dosing, animals were fed regular mouse chow during the entire experiment.
In a second experiment, we tested the oral immunogenicity of 0.1, 0.4 or 0.8 g of tomato powder (reconstituted with water as above) or air-dried tomato fruit samples fed to feed mice on days 1, 4, 17 and 20. Alternative immunization schedules were tested in three groups that received two or three doses of 0.8 g powder or air-dried tomato fruit on days 1 and/or 4, and day 20. Finally, boosting experiments were performed on some groups at the times indicated in the legend to Figure 7. In these cases, 10 µg of i-rNV VLPs (100 µL, 100 ng/µL in 0.9% sodium chloride) were orally administered to each mouse by gavage. Individual serum and fecal samples were collected before the first immunization and on days 11, 27, 41, 55 and 69. For groups in boosting experiments, the long-term kinetics of the immune responses were examined in serum and fecal samples obtained every 2 or 3 weeks beyond day 69 until boosting, and weekly after boosting.
Immune responses were determined by antibody ELISAs as described previously (Ball et al., 1998; Guerrero et al., 2001). Polystyrene 96-well high binding plates (Fisher) were coated with rNV antigen by adding 50 µL of i-rNV standard particles per well (1 ng/µL in PBS) and incubating for 4 h at 37 °C. Nonspecific protein binding was blocked overnight at 4 °C with 5% BLOTTO (serum IgG assay) and 10% BLOTTO (fecal IgA assay). Test sera were prepared in 5% BLOTTO, serially diluted twofold, and incubated for 2 h at 37 °C to permit antibody binding. Initial dilution of each serum varied according to OD450 reading in a preliminary screen of IgG levels. Once the optimal dilution of each sample was determined, at least four dilutions of the test samples were assayed. Each serum sample was also analysed in a well-lacking rNV antigen to determine background binding. The plates were washed six times with PBST and treated for 2 h at 37 °C with HRP-conjugated goat anti-mouse IgG (Sigma) diluted in 1 : 5000 in 2.5% BLOTTO. The reaction was developed with 50 µL per well of TMB peroxidase liquid substrate for 5 min, and colour development was stopped by the addition of an equal volume of 1 N H2SO4. Absorbance was measured at 450 nm using a microplate reader (DYNEX Technologies MRX). Endpoint titre was reported as the reciprocal of the highest dilution that had an absorbance value ≥ 0.1 above the background (absorbance of the well-lacking antigen).
Fecal samples were extracted by making 1 : 10 (w/v) suspension in PBS containing 0.1% Tween 20 and 0.1 ng/µL leupeptin (Sigma). Suspensions were kept for 30 min at 4 °C and thoroughly homogenized with a FastPrep machine (Bio101) for 1 min. After incubation for another 30 min at 4 °C, suspensions were centrifuged for 10 min at 12 000 g at 4 °C. The supernatant was collected and clarified again by 1 min at 12 000 g at 4 °C. The twice-clarified supernatant was stored at −20 °C until assay. NV-specific IgA titres in fecal extracts were determined as for assay of serum IgG but using HRP-conjugated goat anti-mouse IgA (Sigma).
GMTs of IgG in serum and IgA in fecal pellets were determined for every group of mice. A mouse was considered to be a responder if its titre was fourfold greater than pre-immunization titre, which was assigned a value of 10. Nonresponders were included in the computation of the GMTs, unless otherwise noted (Figure 6, parentheses). The lowest dilution tested (1 : 20) for BALB/c was divided by 2 and used as the titre for the negative samples (i.e. negative samples were assigned a titre of 10). Standard errors were calculated on the log-transformed titres. Antibody titres between groups were compared by using the Kruskal–Wallis test followed by the Mann–Whitney U rank sum test (Ball et al., 1998).
We thank William Hamilton, Koen Hellendoorn and Timothy Jones for design of the plant-optimized NVCP gene, Shannon Caldwell and the Cornell Integrated Microscopy Center for technical assistance with the TEM study, and Yanning Liu and Min Zhu for their assistance with statistical analysis and helpful discussion on immunogenicity experiments. This work was supported by NSF grant #BES-0109936 to HSM and a fund from the Park Foundation.