The diversity of drugs, particularly biologics, with different mechanisms of action presents unique challenges for drug delivery and has led to the development of new delivery systems that facilitate transport across biological barriers, delay degradation, or alter the distribution of a drug to increase its efficacy and/or reduce its toxicity (Boyd, 2008). For example, incorporation of drugs into liposomes, nanoparticles, microparticles, or polymer-based carriers may protect the cargo from environmental effects and enable controlled and targeted delivery of the active ingredients (Schwestka and Stoger, 2021). Particle-based delivery systems also offer opportunities for increasing the efficacy of vaccines by enhancing their stability and ability to interact with target cells in vivo (Scotti and Rybicki, 2013; Marsian and Lomonossoff, 2016; Schwestka and Stoger, 2021; Ward et al., 2021). Particulate structures also possess inherent immunostimulatory properties, promoting more efficient uptake by antigen presenting cells than corresponding soluble antigens (Zhu et al., 2014; Snapper, 2018). Particulate protein assemblies display multiple copies of the antigen, typically with regular spacing, which enhances the immune response, and the particle itself may also confer adjuvant properties (Zhu et al., 2014; Jiao et al., 2014; Snapper, 2018; Torres et al., 2019).

Encapsulation is particularly useful for the mucosal delivery of vaccines, which induces not only systemic immunity but also protects mucosal barriers such as the intestine and respiratory tract from invading pathogens (Zhu and Berzofsky, 2013; Kim and Jang, 2017). Such non-invasive drug delivery strategies are particularly suitable for the administration of veterinary vaccines to control diseases that pose an economic risk for farmers and/or the risk of zoonotic transmission to humans. Mucosal immunization, such as the delivery of vaccines via drinking water or feed enables the vaccination of large numbers of animals over a short period of time without the effort and stress caused by injections. Most currently licensed veterinary mucosal vaccines are based on live-attenuated viruses and bacteria, but due to the risks associated with those vaccines, there is interest for new vaccine technologies involving synthetic particle carriers to combat emerging zoonotic diseases (Vacher et al., 2013; Topp et al., 2016; Mohan et al., 2017; Schwestka et al., 2020).

One way to achieve encapsulation of pharmaceutical compounds is based on in vitro techniques such as coacervation and spray drying (Perry and McClements, 2020). In this context, natural plant-derived polymers such as storage proteins, cell-wall carbohydrates and starch are often extracted and reformulated to encapsulate drugs in vitro. For example, certain subfractions of zeins, the large family of maize storage proteins, have been studied due to their unique physicochemical and biological properties. They form edible films that are tough, hydrophobic and resistant to microbial degradation (Shukla and Cheryan, 2001), allowing the use of zein nanoparticles for the in vitro encapsulation of pharmaceutical and nutraceutical compounds (Patel et al., 2010; Luo et al., 2011; Karthikeyan et al., 2014; Penalva et al., 2017). Given the ability of plants to assemble protein storage organelles, which are particulate structures formed in the endoplasmic reticulum (ER) or vacuoles, mainly consisting of seed storage proteins such as zeins, recombinant proteins can also be directly bioencapsulated during the manufacturing process when plants are used as production hosts. In addition, this strategy can leverage many other advantages that plants offer in providing a safe, scalable, and cost-effective platform for the rapid and sustainable production of vaccines and drugs (Lico et al., 2020; Capell et al., 2020; Margolin et al., 2020; Rosales-Mendoza, 2020). Natural plant storage organelles are mostly restricted to seeds and tubers, and these organs have been used to achieve the in vivo encapsulation of proteins by directing them to accumulate in storage organelles holding promise for the development of new drug delivery vehicles (Stoger et al., 2005; Takaiwa et al., 2015; Endo et al., 2021). For example, transgenic rice seeds have been used for the production and delivery of T-cell epitopes representing allergenic proteins (Takagi et al., 2010; Takaiwa et al., 2015; Ishida et al., 2021). By targeting cedar pollen allergens to protein storage organelles such as ER-derived protein bodies (PBs) and protein storage vacuoles, the encapsulated allergens were protected against proteolytic digestion and long-term oral administration of transgenic rice improved medication scores long-term administration (Takaiwa et al., 2019; Endo et al., 2021). While this is a successful approach and cereal seeds are valuable bioreactors for the production and storage of high-value compounds (Zhu et al., 2022), the generation of transgenic cereals is a lengthy process, rendering seed-based systems most suitable for products that are required in large quantities and over long periods of time (Tschofen et al., 2016; Vamvaka et al., 2016). Vaccines are often needed on a seasonal basis and have to be adapted to novel pathogens favoring transient expression systems involving the short-term expression of proteins in plants infiltrated with recombinant bacteria. Such systems are widely used, and several large-scale production facilities have been established in the last 10 years based on the tobacco relative Nicotiana benthamiana (Lomonossoff and D’Aoust, 2016; Fischer and Buyel, 2020; Kurokawa et al., 2021). Although the process of PB biogenesis is not fully understood, some aspects have been characterized in sufficient detail to induce these protein storage organelles ectopically in non-storage tissues such as leaves of Nicotiana benthamiana, and this offers a faster, more versatile and controllable method to encapsulate proteins in the protective environment of PBs. Usually this has been achieved by means of fusion proteins joining the therapeutic candidate to a partial prolamin sequence. For example, fusion to the N-terminal part of 27-kDa-γ-zein (also called Zera) can induce the formation of ectopic PBs in heterologous expression systems (Mainieri et al., 2004; De Virgilio et al., 2008; Torrent et al., 2009b). The resulting PBs are spherical particles with unique properties (~1 µm diameter, ~1.20 g/cm³ density) that allow separation during downstream processing, which can be either based on gradient ultracentrifugation (Whitehead et al., 2014; Van Zyl et al., 2017) or serial tangential flow filtration steps (Schwestka et al., 2020).

Several recombinant pharmaceutical proteins have been produced in an active form by inducing ectopic PBs in N. benthamiana leaves, including calcitonin, human epidermal growth factor and growth hormones, as well as vaccine candidates (De Virgilio et al., 2008; Torrent et al., 2009a; Whitehead et al., 2014; Hofbauer et al., 2016). In addition to incorporating and protecting recombinant proteins zein–antigen microparticles were found to function as adjuvants, triggering a stronger immune response than the soluble antigen. This was shown in mice where the immune response to antigens incorporated into PBs could not be further enhanced by adding adjuvant (Hofbauer et al., 2016). Empty PBs administered together with the soluble antigen also enhance the immune response (Whitehead et al., 2014). Another important characteristic of γ-zein PBs is their ability to interact with cell membranes, likely reflecting the presence of proline-rich repeats in the 27-kDa-γ-zein polypeptide, which share some similarity with cell-penetrating peptides (Kogan et al., 2001; Fernandez-Carneado et al., 2004). We recently showed that γ-zein-EGFP (enhanced green fluorescent protein) PBs were taken up more efficiently than synthetic polystyrene particles of the same size when administered to intestinal epithelial cells and to antigen-presenting cells (Schwestka et al., 2020).

However, the induction of artificial PBs is not always successful when the Zera sequence is used as a fusion tag. For some recombinant viral antigens, including the HIV negative factor (Nef) and CAP256 gp140 envelope, the addition of Zera was not sufficient to lead to PB formation, although the fusion antigen accumulated in the ER (De Virgilio et al., 2008; Ceresoli et al., 2016; Ximba et al., 2020).

We therefore embarked on a novel synthetic approach aiming at the in vivo generation of designer PBs with a core-shell structure that may be applicable to a wider range of protein pharmaceuticals and allow the incorporation of several components and functions into the same PB. Based on existing knowledge on the mechanisms of storage organelle formation and on the interaction between protein components of naturally occurring multi-layered PBs found in maize seeds (Lending and Larkins, 1989) we explored selected members of the zein protein family in a combinatorial approach. Natural PBs of maize are composed of a shell containing γ- and β-zeins, and a core containing α-and δ-zeins. The γ- and β-zeins are expressed early in seed maturation, and are rich in cysteines and cross-linked by disulfide bonds. After PB initiation by 27-kDa-γ-zein, the remaining γ-zeins as well as 15-kDa-β-zein constitute the early PB, while α-and δ-zeins are only later incorporated by penetrating the existing matrix, which requires the rearrangement of existing disulfide bonds (Lending and Larkins, 1989; Guo et al., 2013). The 19-kDa-α-zein (B1) is the most abundant zein in the core of native zein PBs in maize, and δ-zeins are interesting candidates for the generation of layered PBs as well because they colocalize in the core of maize PBs alongside α-zein (Bagga et al., 1997; Guo et al., 2013; Kim and Krishnan, 2019).

We therefore included in our study 27-kDa-γ-zein, Zera, 16-kDa-γ-zein, 15-kDa-β-zein, 22-kDa-α-zein, 19-kDa-α-zein (B1) and 10-kDa-δ-zein. Each of these zeins was fused to at least two different fluorescent protein tags to allow for different color combinations upon co-expression in N. benthamiana. We selected the best combination for obtaining multi-layered PBs and tested it for PB formation with an antigen peptide derived from the outer surface protein BB0172 of B. burgdorferi, that was earlier identified as suitable candidate for the development of a veterinary vaccine for protection against Lyme disease (Small et al., 2014).

Overall, our study demonstrates that it is possible to produce multicomponent ectopic PBs by transient co-expression of selected zeins in Nicotiana benthamiana, highlighting that the interactions between the storage protein components are both necessary and sufficient to achieve an orderly composition resembling the model of naturally occurring zein PBs in maize. This in turn emphasizes that native PBs are supramolecular structures formed by a complex interplay of storage proteins that self-assemble into spherical particles in the lumen of the ER through reversible interactions (Lending and Larkins, 1989; Pedrazzini et al., 2016). Unlike aggregation, which describes a process of irreversible binding of misfolded proteins, this process rather resembles coalescence or similar phenomena, which sometimes serve to protect parts of the proteome from degradation (Saarikangas and Barral, 2016). The prerequisite for the synthetic production of multi-layered PBs for bioencapsulation is therefore the understanding of the interaction of individual zeins as their building blocks (Guo et al., 2013). These building blocks comprise α-, β-, γ-, and δ-types of zein, of which the β-, γ-, and δ-types are usually encoded by single-copy genes and include one 15-kDa β-type zein (Pedersen et al., 1986), two δ-type zeins of 10- and 18-kDa (Kirihara et al., 1988; Chui and Falco, 1995), and three γ-type zeins of 16-, 27-, and 50-kDa (Prat et al., 1987; Woo et al., 2001; Li and Song, 2020). The α-zein subfamily can be further divided into 19- and 22-kDa subclasses which are encoded by multiple gene families (Wilson and Larkins, 1984). The functions of individual zeins in PB biogenesis have been studied in their native environment, using either natural mutants or knockdown lines (Segal et al., 2003; Wu and Messing, 2010; Guo et al., 2013). Although these studies provided important information on the localization and behavior of zeins, they could not analyze the role of individual zeins independently of other endogenous factors. A yeast two-hybrid study provided further important insights into direct zein interactions (Kim et al., 2002), but could not account for the effects of their spatial arrangement in the PB matrix. Other studies investigated if zeins accumulate into combined particles upon ectopic expression, but did not determine the localization of individual zeins (Bagga et al., 1997; Coleman et al., 2004; Kim and Krishnan, 2019). The additive approach in the present study aimed to reveal the individual positioning of different zeins upon ectopic co-expression. While this provides valuable insight into the mechanisms of storage organelle formation in seeds, it is important to note that the environment is not necessarily representative of the native state in all respects. For example, native PBs contain numerous non-zein proteins (NZPs) in addition to zeins, including FLOURY1 (FL1) and OPAQUE10 (O10) (Holding et al., 2007; Wang et al., 2016). Moreover, in maize seeds, other mechanisms in addition to protein-protein interactions between zeins have been identified as an important factor for the structural organization of PBs. For example, altered mRNA targeting can disrupt the localization pattern of zeins in maize PBs (Washida et al., 2004; Washida et al., 2009). The intricate temporal regulation of zein expression and the resulting relative abundance and stoichiometric ratio of zeins also contribute to the mechanism of endogenous PB formation (Guo et al., 2013; Li and Song, 2020). Nevertheless, the study of interactions between PB resident proteins in a heterologous system is a valuable complementary approach, and this was recently pursued to show that the accumulation of NZP1 in induced PB depends on its interaction with 22-kDa-α-zein (Feng et al., 2022).

Our data confirm that among the zeins, the N-terminal part of 27-kDa-γ-zein (commercially developed as Zera) is necessary a for the initiation of PBs when fused to other polypeptides (Geli et al., 1994; Torrent et al., 2009a; Wu and Messing, 2010). The 27-kDa-γ-zein is the evolutionary oldest zein, and while its C-terminal part seems to be derived from 2-S-albumin, the N-terminal part with amphipathic repeats of the amino acid sequence PPPVHL promotes both membrane interaction and spontaneous self-assembly (Kogan et al., 2001) which are crucial for ER-retention. Additionally, seven cysteine residues drive polymerization and stabilize the resulting supramolecular structure via disulfide bridges (Geli et al., 1994; Torrent et al., 2009b; Mainieri et al., 2014). It has been previously shown that the formation of ectopic PBs can be triggered by Zera (Mainieri et al., 2004; Torrent et al., 2009b), but the potential for organization into stratified PBs in a heterologous expression system has not been studied in detail.

During native PB biogenesis in maize endosperm, 27-kDa-γ-zein interacts strongly with 15-kDa-β-zein and 16-kDa-γ-zein, which are present in the periphery of newly formed PBs and locate closer to the center at later stages of native PB biogenesis (Lending and Larkins, 1989). In the heterologous system, both 15-kDa-β-zein and 16-kDa-γ-zein co-localized very well with Zera-induced PBs, ranging from patches at the periphery of the PB, to a homogeneous distribution or even a separation into a core-shell structure within the PB. This diversity of co-localization behavior reflects to some extent their differential distribution during native PB maturation. Without co-expression of Zera the EGFP fusions of 15-kDa-β-zein or 16-kDa-γ-zein did not form regular, spherical PBs, but resulted in patchy fluorescence patterns indicating small protein aggregates. This is well in agreement with earlier observations in Arabidopsis leaves where 16-kDa-γ-zein formed dispersed electron-dense threads enlarging the ER lumen without assembling into PBs, and only when co-expressed with 27-kDa-γ-zein it co-assembles with the latter into insoluble polymers (Mainieri et al., 2018). In transgenic Arabidopsis seeds 16-kDa-γ-zein formed more compact aggregates, but remained largely insoluble in reducing conditions, also suggesting that 16-kDa-γ-zein is unable to form well-ordered polymers on its own but relies on the ability of 27-kDa-γ-zein to drive self-assembly (Arcalis et al., 2021). Although 16-kDa-γ-zein most likely evolved through gene duplication of 27-kDa-γ-zein (Xu and Messing, 2009), this different behavior may be explained by the loss of the repeat region as well as some cysteine residues reducing the strength of its intermolecular interactions (Mainieri et al., 2018). A similar explanation may apply to 15-kDa-β-zein, which lacks the amphipathic repeats, while still showing a sequence similarity of 85% to the N-terminal region of other proteins of the γ-zein family (Kim et al., 2002). However, when co-expressed with Zera-EGFP, 15-kDa-β-zein-mCherry efficiently contributed to two-component PBs, and the combination of these two zeins resulted in the desired core-shell structure of the PBs.

In contrast, the fluorescently labeled 10-kDa-δ-zein exhibited only rare and small inclusions in PBs, when co-expressed with Zera. In native PBs 10-kDa-δ-zein is found together with 19-kDa-α-zein in the PB core, but neither appears to directly interact with γ-zeins in the shell (Kim et al., 2002). The observation of occasional co-localization is consistent with a TEM analysis that indicated the formation of mixed PBs upon co-expression of δ-zeins with 27-kDa-γ-zein in soybean, as inferred from electron density (Kim and Krishnan, 2019).

We observed that the co-expressed α-zeins (19-kDa-α-zein and 22-kDa-α-zein) that are located in the core and interface of native PBs were clearly localized separately from the Zera-induced PBs when expressed individually, with 19-kDa-α-zein being incorporated only in very rare cases. This is consistent with data from a yeast two-hybrid study showing no direct interaction with 27-kDa-γ-zein (Kim et al., 2002). Consequently, it was speculated that other zeins and non-zein proteins may be required for their incorporation, and it has been shown, for example, that 22-kDa-α-zein remains outside of PBs in the absence of the membrane protein floury1 (Holding et al., 2007).

Among the zeins, 16-kDa-γ-zein and 15-kDa-β-zein have been suggested to play a mediating role between other PB components because they interact with most other zeins (Kim et al., 2002). Suppression studies also indicated that 16-kDa-γ-zein or 15-kDa-β-zein might play a role in integrating other zeins (Wu and Messing, 2010; Guo et al., 2013). Indeed, a strong effect on the integration of other zeins into PBs was observed when 16-kDa-γ-zein was co-expressed, making it a key element in the design of synthetic PBs. Together with 15-kDa-β-zein, 16-kDa-γ-zein formed combined homogeneous inclusions within Zera PBs that were larger and more abundant than those of 15-kDa-β-zein alone, which was also confirmed by flow cytometry analysis of isolated PBs. Co-expression of 16-kDa-γ-zein even allowed integration of 19-kDa-α-zein, 22-kDa-α-zein, and 10-kDa-δ-zein into Zera PBs, albeit to a lesser extent. Moreover, 16-kDa-γ-zein was also observed at the interface between the PB core and the shell formed by Zera, consistent with its role as a linker.

It should be noted that co-infiltration of individual agrobacteria lines each carrying a single zein construct was used in this study to facilitate the screening of a large number of zein combinations. This approach however is prone to causing some variation in expression levels within the leaf (Bashandy et al., 2015). For future large-scale production the construction of a single vector, which contains all the required genes or their stable integration into the plant genome would therefore be favorable to increase the uniformity of PBs within a single cell and throughout the leaf tissue. For example, according to our results a platform could be envisaged, that stably expresses both Zera and 16-kDa-γ-zein at an optimized stoichiometric ratio, providing the chassis for bioencapsulation. A therapeutic protein could then be incorporated by fusing it to 15-kDa-β-zein or one of the other zein candidates identified and adding it to the expression host via agroinfiltration. This would allow the on-demand production of, for example, seasonal or emerging vaccine antigens.

The fact, that the multicomponent PBs can be isolated and retain their internal architecture further suggests, that various proteins of interest can be stably incorporated and accumulated by this approach. To confirm that multicomponent PBs can also be achieved with a pharmaceutically relevant model antigen, we used a peptide derived from B. burgdorferi outer surface protein BB0172, which was previously identified as a promising veterinary vaccine candidate for protection against Lyme disease (Small et al., 2014). This antigen did not accumulate to detectable levels when conventionally fused to Zera, but in fusion with 15-kDa-β-zein it was clearly detectable. As expected, fusion with 15-kDa-β-zein rendered the antigen insoluble, and no free or soluble antigen was present. After co-expression of Zera-EGFP, which resulted in incorporation of the 15-kDa-β-zein-fused antigen into Zera-induced PBs, the antigen accumulated to higher levels, revealing a positive effect of encapsulation on antigen yield. Additional co-expression of 16-kDa-γ-zein further doubled the antigen content. This confirms a positive role of 16-kDa-γ-zein in protecting the 15-kDa-β-zein-fusion protein within Zera-induced PBs. Our immunolocalization results confirmed the incorporation of antigen fused with 15-kDa-β-zein into multicomponent PBs. Previous studies have shown that the BB0172 derived peptide antigen is immunogenic but requires potent adjuvants to be effective (Small et al., 2014). Since the delivery of subunit antigens on particulate structures is an established approach to enhance their immunogenicity (Rybicki, 2020; Ximba et al., 2022), we anticipate that through encapsulation in PBs, the vaccine candidate will benefit not only from higher yield but also from the adjuvant effect of PBs. Further studies will be needed to confirm this.

In summary, the in planta encapsulation of antigens within newly designed multi-layered protein storage organelles with a core-shell structure is a novel approach that can be integrated in the upstream production process to enhance the accumulation and to achieve a particulate formulation. In future, this strategy may allow to further maximise the benefits of particulate plant-based drug delivery vehicles by functionalizing the individual PB components in different ways, for example combining the antigen, a receptor-binding ligand and an adjuvant in one particle.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

JS, MT, LZ and ES designed and carried out experiments, analyzed data, and wrote the manuscript. EA and FS designed and carried out experiments and analyzed data. ME-G, EP and AV discussed the data, and contributed to the manuscript. JS, LZ and ES designed the study, analyzed data, and wrote the manuscript. All authors contributed to the article and approved the submitted version.