The most relevant structural and functional characteristics of representative natural NVPNs are summarized in Table 1. Vault nanocages (Figure 1A) have extending caps at their ends, a hinged waist region, 8 small pores (2 nm in diameter) and a significant internal volume. Each vault nanocage is composed of several protein and non-protein elements, with the major vault protein (MJV) representing 70% of the overall mass. An interesting property is the dynamic ability of the vault nanocages, since they can open and close transiently, allowing the incorporation of small molecules and macromolecules within the inner core (Kickhoefer et al., 1996; Kickhoefer et al., 2005; Kong et al., 1999; Poderycki et al., 2006; Anderson et al., 2007; Ryu et al., 2008; Goldsmith et al., 2009; Tanaka et al., 2009; Yu et al., 2017).

PepA nanocages are formed by the self-assembly of aminopeptidase (PepA), a zinc-dependent metallopeptidase (Figure 1B). The resulting nanoparticles have an inner cavity with negative charge and 8 pores on the edges and faces with diameters of 1 and 3 nm. The channels are useful for the traffic of molecules. PepA nanocages were used as templates for the size-controlled synthesis of ultrasmall platinum nanoparticles, with the formation of a multifunctional biohybrid catalyst (Kim et al., 2010; San et al., 2011).

DNA-binding protein from starved cells (Dps) belongs to the ferritin subfamily. Dps nanocages (Figure 1C) have small pores (0.8 nm) through which small molecules can diffuse. Dps nanocages were used as templates for the formation of metallic-protein nanoparticles to improve the endogenous catalytic activity of ferritins (Ilari et al., 2000; Kang et al., 2008; Haikarainen and Papageorgiou, 2010).

Ferritin nanocages (Figure 1D) have an internal cavity that matches the size of the iron core. In this core, iron is stored in an insoluble non-toxic state. The cages have 8 hydrophilic pores (4 nm) that facilitate the movement of iron atoms and other small molecules. Ferritin nanocages are quite stable at high temperature and in a wide range of pH values. Self-assembly, disassembly and reassembly can be controlled by metal ions. The nanocages were applied in biomineralization and as nanocarrier of biological and non-biological molecules (Lawson et al., 1991; Hempstead et al., 1997; Zhang and Orner, 2011; He and Marles-Wright, 2015; Kuruppu et al., 2015; Chen et al., 2016; Wang et al., 2017; Palombarini et al., 2020; Zhang et al., 2020).

Small heat shock protein (sHSP) nanocages (Figure 1E) have 8 pores (3 nm and 17 nm) that allow molecular trafficking. Advantages of the cages include high stability at high temperatures (up to 70°C) and in a broad range of pH values (5–11). sHSP nanocages were used as nanoreactors, for biomineralization, for drug delivery and for bioimaging. The functionalization reported includes the introduction of modifications with organic molecules outside and inside of the nanocages (Kim et al., 1998a; Kim et al., 1998b; Kim R. et al., 1998; Bova et al., 2002; Flenniken et al., 2003; Kim et al., 2003; Varpness et al., 2005; Flenniken et al., 2006; Abedin et al., 2009).

Dihydrolipoyl acetyltransferase (E2) nanocages from Bacillus stearothermophilus (Figure 1F) have 12 pores (5 nm) and high stability at extreme temperatures due to the thermophilic nature of the native organism. E2 proteins can be modified simultaneously at the outer and inner surfaces to allow loading of drugs inside and to display functional epitopes outside (Izard et al., 1999; Domingo et al., 2001; Jung et al., 2002; Milne et al., 2006; Dalmau et al., 2008).

Encapsulin nanocages (Figure 1G) have a characteristic large central cavity (20 nm) that makes them interesting as a cargo delivery nanoplatform and as a nanoreactor (Sutter et al., 2008; Rahmanpour and Bugg, 2013; Moon et al., 2014; Tamura et al., 2015; Cassidy-Amstutz et al., 2016; Giessen, 2016). Lumazine synthase nanocages (Figure 1H), which have a 7.8 nm inner cavity with negative charge, were used as molecular carriers through encapsulation. Other applications include the biomineralization of iron. A key characteristic of these nanocages is the stability at high temperatures (up to 95°C) (Schott et al., 1990; Ritsert et al., 1995; Shenton et al., 2001; Zhang et al., 2001; Seebeck et al., 2006; Ladenstein et al., 2013; Ra et al., 2014).

Table 2 summarizes structural characteristics of artificial NVPNs reported in the literature. The formation of symmetric and homogenous 16-nm protein nanocages (Figure 2A) results from the self-assembly of monomers obtained by a fusion process. Each monomer is a geometrically controlled fusion of two natural protein oligomers, which are connected by an α-helical linker. One of those oligomers is the trimeric bromoperoxidase and the other is a dimeric M1 virus matrix protein (Lai et al., 2012a).

His6-HuHF and His6-SF nanocages (Figure 2B) with regulatable self-assembly were created based on two recombinant ferritins (rHuHF and rSF). Histidine motifs were incorporated in one of their subunit interfaces. Two different switches (metal- and pH-based) were developed to control the assembly-disassembly of the nanocages. This may lead to more efficient encapsulation of molecules within the nanocages compared to the traditional methods reported (Gu et al., 2020).

The trp RNA-binding attenuation protein (TRAP) nanocages (Figure 2C) result from the assembly of 24 ring-shaped proteins derived from the natural TRAP from B. stearothermophilus. Each constituent TRAP ring is formed by 11 monomers that were engineered to include a cysteine residue. Contrary to the usual situation, the complete nanocage is formed not through a network of protein-protein interactions but through the bridging of opposing thiols of the cysteine residues between TRAP rings via single gold (I) ions. The fully assembled TRAP nanocages present six square apertures (4 nm). TRAP nanocages are stable up to 95°C and at high concentration of denaturing agents (e.g., 7 M urea). Although they are susceptible to reducing agents, this could be a promising characteristic for delivery to targets that contain this type of agents. The fact that the assembly relies on a metal-induced process is very useful since it provides a more rigorous control of assembly-disassembly and, eventually, a programmable mechanism. The nanocages can also be labelled with a dye in each ring-shaped monomer, which could be useful for bioimaging applications (Heddle et al., 2006; Heddle et al., 2007; Malay et al., 2019; Naskalska et al., 2021; Majsterkiewicz et al., 2022; Stupka et al., 2022).

I3-01 nanocages (Figure 2D) are hollow architectures that result from the self-assembly of multiple monomers corresponding to a trimeric 2-keto-3-deoxy-6-phosphogluconate (KDPG) aldolase from Thermotoga maritima. These aldolase monomers were engineered to contain complementary hydrophobic interfaces. The nanocages have several large pores (9 nm). I3-01 nanocages are stable up to 80°C and in the presence of high concentration of denaturing agents (e.g., 6.7 M guanidine hydrochloride). Applications include synthetic biology, targeted drug delivery, and vaccine development (Hsia et al., 2016; Votteler et al., 2016; Bruun et al., 2018; McCarthy and Gonen, 2022).

Prototype oxygen-impermeable protein nanocages (OIPNC) (Figure 2E) are derived from the pentameric β-carboxysome from Thermosynechococcus elongatus BP-1 (CcmL). Self-assembly of the proteins into nanocages occurs in the presence of quantum dots as templates by protein-quantum dots interfacial engineering. An advantage of these nanocages is the permeability to O₂ in a switchable process controlled by molecular patches. Future interesting applications include their use as nanocarriers and nanoreactors (Gao et al., 2022).

TIP60 nanocages (Figure 2F) are hollow spheres with 20 triangular pores that were created using a fusion protein design approach. Each monomer is a genetic fusion of two proteins (one pentameric- LSm and one dimeric- MyoX-coil) with a three-residue linker. A reversible assembly and disassembly mechanism based on metal ions and chelators was developed for TIP60 nanocages (Kawakami et al., 2018; Obata et al., 2021; Ohara et al., 2023).

The upstream processing comprises the selection of the host cells, the preparation and optimization of the expression vector and its subsequent transformation/transfection into the host. All these aspects are crucial for obtaining a quality bioproduct and a high productivity (Puetz and Wurm, 2019; Tripathi and Shrivastava, 2019). While the natural host can be selected if the desired NVPN is expressed naturally (Schott et al., 1990; Sutter et al., 2008), nanocage production uses mainly expression systems/host dedicated to recombinant protein production (Diaz et al., 2018). Here, a variety of expression hosts are available that include bacteria, yeasts and multicellular fungi, insect, mammalian and plant cells (Diaz et al., 2018; Puetz and Wurm, 2019; Tripathi and Shrivastava, 2019). NVPNs are mostly produced recombinantly in bacterial cells, and in particular in Escherichia coli, a very-well studied host that grows fast, is easy to cultivate and propagate at low cost, and displays high productivity. However, E. coli is also associated with disadvantages such as the lack of proper post-translational modifications (PTMs), the formation of inclusion bodies and the propensity to generate endotoxin contamination due to its Gram-negative nature (Owczarek et al., 2019; Puetz and Wurm, 2019; Tripathi and Shrivastava, 2019). These can be circumvented using yeasts such as Saccharomyces cerevisiae and Pichia pastoris, which grow fast, are easy to manipulate genetically and can perform PTMs. Mammalian cells are more suitable for production of larger and more complex nanocages that might also require PTMs. The most common and used cell lines are Chinese hamster ovary (CHO), Sp2/0 and NS0. Insect cells like Sf9 can also be used for nanocage production through a baculovirus expression vector system (Owczarek et al., 2019; Puetz and Wurm, 2019; Tripathi and Shrivastava, 2019).

The selection of host cells is accompanied by the design and construction of an expression vector/system that can drive a highly efficient nanocage expression (Puetz and Wurm, 2019). The expression vectors will comprise the gene that codes for the NVPN but also any additional genetic elements required to improve stability or enable functionalization. These can be point mutations to change or remove standard reactive residues or to introduce unnatural amino acids, or sequences of functional peptides or proteins that are fused to the N- or C-terminus or internal loops of the monomers (Diaz et al., 2018).

Most types of NVPNs reported in the literature (PepA, Dps, E2 nanocages, encapsulin, ferritin, sHSP, lumazine synthase, I3-01 nanocages, Tet8-M nanocages, TIP60 nanocages and TRAP nanocages) were produced in E. coli (Table 3) using pET-based expression vectors. Table 3 lists the most used key strains of E. coli as well as alternatives in terms of producer host organisms. Contrary to most NVPNs, vault nanocages (Figure 1A) cannot be produced in E. coli due to their eukaryotic origin, which implies complex post-translational modifications and protein folding (Diaz et al., 2018). In this specific case, yeast, insect, or mammalian cells must be used as producer hosts (Table 3). In yeast, the expression vector comprises a glyceraldehyde 3-phosphate dehydrogenase promoter (Wang et al., 2018) and in insect cells a Bac-to-Bac method based on infection with a recombinant baculovirus is necessary (Stephen et al., 2001; Kickhoefer et al., 2005; Poderycki et al., 2006; Champion et al., 2009; Kar et al., 2011; Kar et al., 2012; Wang et al., 2015).

Once recombinant host cells with the expression vector are established, host cell banks are prepared that must be properly characterized and stored. This allows the maintenance of the reproducibility and consistency of the process, since each batch of protein nanocages will be manufactured using the same cell source (Puetz and Wurm, 2019; João et al., 2021).

Following bank preparation, host cells are cultivated and the NVPN is expressed. Initially, a screening should be performed using small-scale cultures to identify and evaluate the impact of cultivation and operation parameters on protein expression levels. These parameters include media composition, temperature, agitation, aeration, cell density, pH, inducer concentration and induction time, among others (Tripathi, 2016; Tripathi and Shrivastava, 2019). After the process is defined at a small scale, scale-up and pilot studies are performed in highly controlled bioreactors as a steppingstone to industrial scale implementation. Bioreactors can be operated in batch, semi-batch, and continuous/perfusion modes, depending on the method used to supply nutrients, circulate the culture medium or recover the target NVPN. Large scale manufacturing needs to be run using optimal operation conditions that maximize expression and yield of NVPNs. Specific parameters of the bioreactor (aeration, dissolved oxygen, CO_2, and hydrodynamic shear) also need to be tested and optimized to guarantee high specific and volumetric productivities. A Design of Experiments (DOE) approach may be used as a more efficient strategy to optimize these parameters and understand the interaction effects between them (Tripathi, 2016; Tripathi and Shrivastava, 2019; João et al., 2021).

NVPNs studied in the literature are produced mostly in shake flask cultures with volumes up to 4 L (Zou et al., 2016a; 2016b), mainly using Luria-Bertani (LB) (Allen et al., 2002; Bellini et al., 2014; Zou et al., 2016b) or 2× YT growth media (Murata et al., 2015; Guan et al., 2018; Kawano et al., 2018; Cristie-David and Marsh, 2019). When under the control of a T7 promoter, expression is induced with the addition of 0.1 mM (Cristie-David and Marsh, 2019; Gu et al., 2020; Ohara et al., 2023) to 1 mM (Santambrogio et al., 2000; Dalmau et al., 2008; 2009; Jeon et al., 2013; Kawano et al., 2014; Zou et al., 2016a; 2016b) of isopropyl β-D-1-thiogalactopyranoside (IPTG) to the culture during the exponential growth phase. Other parameters that may vary are the induction time, between 2 h (Kim R. et al., 1998; Kim et al., 1998a; Santambrogio et al., 2000; Allen et al., 2002; Choi et al., 2011) and 30 h (Jeon et al., 2013; Zou et al., 2016a; Zou et al., 2016b; Cristie-David and Marsh, 2019; Palombarini et al., 2019), and the temperature at and after induction, which is frequently 20°C–25°C (Jeon et al., 2013; Zou et al., 2016a; Zou et al., 2016b; Palombarini et al., 2019) or 37°C (Zou et al., 2016a; Zou et al., 2016b). Overall, data regarding selection and optimization of process and operating conditions, as well as potential scale-up approaches, is very limited.

A key aspect to consider when using prokaryotic expression systems is whether protein overexpression leads to the accumulation of NVPNs as intracellular inclusion bodies. To minimize this, conditions should be optimized to increase nanocage solubility. For example, Zou and co-workers evaluated the effect of temperature (20°C, 15 h and 37°C, 5 h) on the expression of the heavy and light chains of human ferritin and on the co-expression of molecular chaperones. The authors concluded that the amount of soluble protein nanocages increased with the lower temperature and with the presence of chaperones to help in the folding process (Zou et al., 2016a; Zou et al., 2016b). If formation of inclusion bodies cannot be avoided altogether, solubilization and refolding steps must be considered in the downstream processing (Tripathi, 2016; Tripathi and Shrivastava, 2019). A study by Palombarini et al. (2019) on the expression of ferritin monomers optimized the concentration of IPTG (0.1, 0.5, and 1 mM), the induction time (4, 8, and 16 h) and the induction temperature (25°C and 37°C). The addition of 0.5 mM IPTG at 25°C for 16 h was identified as optimal and subsequently implemented at a large-scale production of ferritin nanocages by the biotechnology company GeneScript (Palombarini et al., 2019).

Martín and co-workers developed a strategy for the production of vault nanocages in mammalian cells as a faster and more efficient alternative to the traditional expression in yeast or insect cells. An engineered vault nanocage (His-tagged major vault protein) was successfully produced in the human embryonic kidney 293F cell line through transient gene expression (Martín et al., 2022).

Cell-free protein synthesis has been widely explored in recent years as an alternative to cell-based expression systems but reports of its use in NVPNs production are scarce. In one example, Mrazek described the obtention of engineered vaults using a cell-free wheat germ expression system and either a DNA vector or an mRNA encoding the major vault protein. Notably, the author was able to simultaneously package passenger molecules in the internal cavity of the formed vaults by adding them to the synthesis mixture (Mrazek, 2016). However, due to its current limitations and challenges, cell-free protein synthesis does not seem to be the most suitable strategy for consistent production of NVPNs at large-scale (Kwon and Jewett, 2015; Des Soye et al., 2018; Failmezger et al., 2018; Tran et al., 2018; Gregorio et al., 2019; Colant et al., 2021).

The downstream processing encompasses the extraction, isolation, and purification of the target NVPNs from the broth culture obtained in the upstream stage. The final end-product must conform to a predetermined set of specifications that are established with the final intended use in consideration. Ideally, the overall downstream process will comprise a small number of high-yield unit operations so as to minimize complexity, residence time and processing costs (João et al., 2021). Typically, unit operations are selected and implemented that explore different physical-chemical properties of the specific protein nanocages and of the associated impurities (genomic DNA, RNA, host proteins, and endotoxins). In general, the downstream processing of NVPNs will include primary isolation and recovery steps such as cell harvesting, cell lysis and clarification, and then a number of purification steps (Tripathi, 2016; Tripathi and Shrivastava, 2019; João et al., 2021). An overall description of this downstream processing is illustrated in Figure 4.

In the first step of the downstream processing, NVPN producer cells are harvested from the culture broth, for example, by centrifugation at moderate speeds (e.g., 5,000 × g) (Kim et al., 1998b; Zou et al., 2016a; 2016b; Martín et al., 2022). Next, cells are resuspended in an appropriate buffer, and the intracellular nanocages are released via a specific cell lysis step. The most common options at lab scale are ultrasonication (Allen et al., 2002; Suci et al., 2010; Choi et al., 2011; Jeon et al., 2013; Zhen et al., 2013; Bellini et al., 2014; Cassidy-Amstutz et al., 2016; Martín et al., 2022), French press (Jung et al., 2002; Suci et al., 2010; Peng and Lim, 2011) and Dounce homogenization cell lysis (Stephen et al., 2001; Poderycki et al., 2006). In some instances, lysozyme is added to the lysis buffer to weaken/break down bacterial cell walls and hence improve lysis (Allen et al., 2002; Suci et al., 2010; Bellini et al., 2014). Nucleases (e.g., DNase, RNase) are also added often to lysis buffers to promote degradation of nucleic acids as they are released (Allen et al., 2002; Suci et al., 2010; Bellini et al., 2014). Protease inhibitors are also an option in some cases (Stephen et al., 2001; Poderycki et al., 2006; Bruun et al., 2018; Martín et al., 2022). For large-scale operation, high-pressure homogenizers and bead milling will certainly be more suitable than sonication, French press or Dounce homogenization cell lysis (Tripathi and Shrivastava, 2019).

After cell lysis, a clarification step based on centrifugation or filtration is normally implemented to remove cell debris and particulate matter. If a situation exists where nanocage inclusion bodies are formed, these must be separated from cell debris and recovered, solubilized and then adequately refolded (Dalmau et al., 2008; Dalmau et al., 2009; Tripathi, 2016; Tripathi and Shrivastava, 2019). For example, Jeon and co-workers have shown that inclusion bodies of ferritin nanocages can be solubilized with 8 M urea and then refolded on a nickel ion chelate affinity column with a downward gradient of urea from 8 M to 0 M (Jeon et al., 2013).

Clarified lysates containing the NVPNs are sometimes subjected to a heat treatment prior to purification. This approach explores the fact that protein nanocages are in general thermostable at high temperatures. Thus, heating clarified E. coli cell lysates at 65°C–90°C will promote the denaturation and precipitation of host proteins, which are subsequently removed by centrifugation (e.g., 12,000 × g), without affecting the structure of the thermostable nanocages (Diaz et al., 2018). This strategy was applied successfully to lysates containing a range of different nanocages, including Dps (Allen et al., 2002; Suci et al., 2010), E2 nanocages (Dalmau et al., 2008; 2009), encapsulin (Moon et al., 2014; Cassidy-Amstutz et al., 2016), ferritin (Santambrogio et al., 2000; Sana et al., 2010; Falvo et al., 2013; Zhen et al., 2013; Bellini et al., 2014; He and Marles-Wright, 2015), sHSP (Flenniken et al., 2006) and I3-01 nanocages (Hsia et al., 2016).

Several studies also report the enzymatic treatment of nanocage-containing clarified E. coli cell lysates. In the majority of cases, a DNase or an RNase is used directly in the clarified lysates (Santambrogio et al., 2000; Murata et al., 2015) or previously added to the lysis buffer (Allen et al., 2002; Flenniken et al., 2003; Flenniken et al., 2006; Suci et al., 2010; Bellini et al., 2014; Ohara et al., 2023) to degrade host cell nucleic acid impurities such as genomic DNA and RNA that are co-released from cells during lysis. The resulting products of degradation (e.g., short oligonucleotides) are easier to remove in subsequent purification steps. Protease inhibitors may also be added at this stage to minimize enzymatic degradation of nanocages (Bruun et al., 2018).

The NVPNs-containing clarified lysates generated in the intermediate recovery stage and entering the final purification stage will often contain host derived impurities that must be removed to obtain a bulk product complying with final-specifications (Figure 4). Apart from genomic DNA, RNA, host proteins and other macromolecules (e.g., lipopolysaccharides in the case of E. coli), it is also important to remove misfolded nanocages or aggregates. Given this range of impurities, the purification stage is likely to comprise more than one step. Data on the downstream processing available in the literature primarily comes from laboratory protocols. This limited information makes it difficult to compare and weigh the merits of the different methodologies used.

To illustrate the diversity of options available in terms of producer hosts and unit operations, four different block diagrams of NVPNs manufacturing processes are schematized in Figure 5. These block diagrams were adapted from the literature and are representative of the current panorama in the downstream processing of protein nanocages.

The majority of the downstream processing studies available refer to NVPNs produced in E. coli, which is clearly the most common producer microbial host. Blocks diagrams A (Figure 5A) and B (Figure 5B) exemplify the downstream processing respectively of ferritin (Santambrogio et al., 2000) and sHSP (Guan et al., 2018) nanocages produced in E. coli and purified by chromatography. On the other hand, block diagram C (Figure 5C) illustrates the production of vault nanocages in Spodoptera frugiperda insect cells (Stephen et al., 2001). Finally, block diagram D (Figure 5D) shows a downstream process of ferritin nanocages produced in mammalian (Homo sapiens) cells (Kanekiyo et al., 2013).