The cell wall of Gram-positive bacteria (Figure 1A) is characterized by its main entities, namely, an inner cytoplasmic membrane and an outer peptidoglycan layer facing the external environment. In contrast, Gram-negative bacteria have a significantly thinner peptidoglycan layer that is surrounded by a second outer phospholipid membrane, which is decorated with poly-saccharides facing the extracellular environment (Figure 1B) (Rohde, 2019). The peptidoglycan layer is a water-containing cross-linked system of di-saccharides formed by N-acetylglucosamine (NAG) and N-acetylmuramic acid (NAM), which are cross-linked by flexible pentapeptide amino acid chains forming a mesh-like system (Vollmer et al., 2008), and has embedded proteins and teichoic acid. The main component of teichoic acid is a co-polymer (poly-di-ester) of glycerol and phosphate groups, which is negatively charged at neutral pH. Therefore, the surface of Gram-positive bacteria is considered to be hydrophilic and negatively charged due to the nature of the peptidoglycan layer and the teichoic acid content. However, as discussed in the later section, the natural variation of the cell wall composition also manifests itself by varying its surface charge (from almost neutral to negatively charged) and hydrophobicity. The joint electrostatic effect of the charged groups on the bacterial surface are routinely characterized using zeta-potential measurements, either performed at neutral (physiological) pH or as function of pH to reveal the effect of the protonation and deprotonation of bacterial surface groups (Jiang et al., 2023a).

Revealing the structure and properties of bacterial surfaces are non-trivial both experimentally and theoretically. Small- and wide-angle X-ray scattering measurements have been widely used to determine the supramolecular self-assembly of purified lipopolysaccharides (LPSs) in suspension experimentally (Brandenburg et al., 1992; Garidel et al., 2005). Specular neutron and X-ray reflectivity at the solid/water and air/water interface have been used to further investigate the molecular-level structures in the z-direction perpendicular to the membrane xy-plane. The monolayer of LPSs at the air/water interface can be used to study the structural integrity of the outermost LPSs (Oliveira et al., 2010; Thoma et al., 2020), whereas asymmetric bilayers formed at the solid/aqueous interface can be used as a more realistic model of the outer membrane, demonstrating the ion-specific collapse of the saccharide head groups (Schneck et al., 2009b; Clifton et al., 2013). The neutron-based techniques provide structural information with Å resolution by optimizing the contrast in the scattering length densities, by selective deuteration and using appropriate mixtures of H₂O and D₂O. However, these methods are less sensitive to specific ions/elements. Grazing-incidence X-ray fluorescence (GIXF) was proposed as a promising strategy to analyze the electrostatics of LPS surfaces covered by charged saccharide head groups because the X-ray fluorescence signals measured near the critical angle of total reflection can be used to calculate the density profiles of ions in the head group of LPS (Schneck et al., 2010; Abuillan et al., 2013). It is notable that the charge density near the LPS surface is neither uniform nor static, and charged LPS moieties can be crosslinked using multivalent ions. These considerations render the classical Derjaguin-Landau-Verwey-Overbeek theory inadequate. Moreover, atomistic molecular dynamics simulations are not suited to handle the typical length scale (>10³ nm³) and characteristic time window of diffusion-limited reactions (∼10^–3 s). Pink et al. employed a coarse-grained, minimal model of LPS membranes and calculated the ion density profiles and conformational changes within the framework of a linearized Poisson-Boltzmann theorem derived by Netz (Netz, 1999). They demonstrated that the divalent cations bound to core saccharides form electrostatic potential barrier (Pink et al., 2003; Oliveira et al., 2009), which explains the data obtained by GIXF and reflectivity experiments.

Microbial adhesion to hexadecane (MATH) or microbial adhesion to solvents (MATS) are microbiological standard techniques that provide indirect information about the bacterial surface properties (Rosenberg et al., 1980; Geertsema-Doornbusch et al., 1993). Upon vigorous shaking of a two-phase system, a fraction of the bacteria is trapped in the organic phase and a MATH value is estimated from the optical density (OD) measurements. A larger MATH value means that the surface is predominantly hydrophobic. It has been observed that a coarse and creaming emulsion that retains a fraction of bacteria through a Pickering mechanism is formed (Rosenberg, 1984; Rosenberg, 2006). Unlike real distribution coefficients, no simple equilibrium thermodynamic interpretation can me made using MATH values.

The contact angle at the three-phase contact is related to interfacial tensions of wetting liquid, between the liquid and the bacterial film, and the bacterial film and air through the Young’s equation, that is, in principle the thermodynamic quantities that can be extracted from contact angle measurements. It should be noted that the contact angle of the “dry” bacterial layer in contact to air does not represent the hydrated surfaces of bacteria under water; these measurements represent the contact angles with the “collapsed and dry constituents” of the bacterial cell wall (Van Oss et al., 1988; Jiang et al., 2023a). The Lewis base character (A ⁻ ) is defined as the difference between the MATS values for chloroform and hexadecane and the Lewis acid character (A ⁺ ) is derived from the differences of the solvent pair, acetylacetate, and decane. Using contact angle measurements in combination with reference solvents, e.g. water, formamide, and 1-bromo-naphtalene, and the van Oss−Good model, the Lewis acid and base components of the bacterial interfacial tensions can be separated and the bacterial-water interfacial tensions can be estimated. In this approach, the bacterial cohesive free energy density, ∆ G B W B = − 2 γ B W , is the underlying driving free energy change for bacterial aggregation.

Atomic force microscopy (AFM) can be used to characterize the bacterial cell wall and isolated cell wall components such as the peptidoglycan layers. For a review, see (Dufrêne et al., 2021). AFM can be used to detect the surface morphology of the bacterial cell wall, revealing that Gram-positive bacteria exhibit a regular architecture of a peptidoglycan layer and adhesins (Dupres et al., 2005). Moreover, the force-distance curves can be obtained with the µm-scale spatial resolution. A small deformation can be used to detect the elasticity of the cell wall, whereas a larger deformation provides information on the cell stiffness, reflecting the turgor pressure of the cell.

In addition to AFM, defined model systems can be used to study the mechanical properties of bacterial surfaces more quantitatively. Schneck et al. fabricated planar stacks of purified LPS membranes with various strains and preformed off-specular neutron scattering at defined disjoining pressures. Although the model consisted of symmetric bilayer stacks of LPSs, this approach enabled quantitative insights into how saccharide chain modifications modulate the mechanical parameters of outer membranes, such as the bending rigidity and vertical compression modulus (Schneck et al., 2009a). Measuring the response function under the application of a shear stress, which is called interfacial stress rheology, is another technique that is employed to measure the complex viscoelastic modulus of model membranes (Brooks et al., 1999; Schneider et al., 2002). Compared to the commonly used rotating disk rheometry, a small probe-membrane contact enables sensitivity for determining the loss and storage moduli of “soft” membranes, such as the membranes of LPSs of Gram-negative bacteria and lipids mimicking lipoteic acid of Gram-positive bacteria (Tanaka et al., 2004; Herrmann et al., 2015). Experiments using model systems, as compared to experiments using living (or fixed) bacteria, enable the understanding of the contributions of the outermost surfaces, such as the viscoelasticity of the peptidoglycans and inner membranes, as well as the turgor pressure pushing the cell wall from the inside.

Yet, despite the major progress made over the past decades as described above, there is a clear demand for experimental tools that bridge the well-defined model systems and more complex, real bacterial surfaces. On one hand, well-defined two-dimensional model systems, such as LPS membranes, enable the precise measurement of detailed structural features, ion density profiles, and interfacial viscoelasticity. On the other hand, most of the model systems cannot address the effect of the physico-chemical properties of food on biological functions. Moreover, from the theoretical viewpoint, the development of multiscale modeling that connects atomistic simulations and coarse-scale models is required to cover several orders of magnitude in space and time.

Different approaches have recently been used to change the native surface properties of bacteria, with a particular focus on increasing hydrophobicity and thereby the adhesion behavior of Gram-positive bacteria to promote the ability of bacteria to be active in building structure and texture of food by for example be active Pickering stabilizers at surfaces and interphases. Three approaches can be implemented, namely i) adsorption of macro molecules, ii) chemical grafting of molecules to the bacterial wall components, and iii) selective removal of chemical constituents from the bacterial cell wall using enzymes. The first category contains studies by means of the deposition of cationic and anionic polymers (Priya et al., 2011; Yucel Falco et al., 2017b). This procedure changes the sign of the surface potential of the bacterial surface, depending on the effective charge of the outer polyelectrolyte layer. As the polymers used in these studies were mostly hydrophilic, the bacterial surface remained hydrophilic.

Probiotic lactic acid bacteria can be incorporated into food structures and have the dual role of being a functional part of the food structure as well as delivering the probiotic function of the bacteria. Several studies have focused on Pickering emulsions stabilized by bacteria adhering to the oil/water interface using natural UM bacteria (Firoozmand and Rousseau, 2016; Muhammed et al., 2023) and chemically modified bacteria (Jiang et al., 2019; Jiang et al., 2021). The emulsions exhibited large emulsions droplets of 10—100 μm, which was one to two orders of magnitude larger than the size of bacteria (∼1 µm). For food emulsions, the size limitation mainy comes from the size of bacteria that are used as the stabilizers. Mayonnaise-like emulsions were produced using food grade, Streptococcus thermophilus with a high internal oil phase load exceeding 40% (Firoozmand and Rousseau, 2016). MCT oil in water emulsions were produced using a selection of 31 naturally hydrophobic lactobacillus bacteria (Muhammed et al., 2023), as well as enzymatically treated LGG and LBD bacteria (Jiang et al., 2021). The highly hydrophobic LGG bacteria modified by laurylchloride exhibited the surprising ability to promote the formation of W/O/W double emulsions (Figures 3A–C using a simple one-step procedure (Jiang et al., 2021). The outer surface of the oil droplets (diameter ≈60 µm) was stabilized by a densely packed layer of adsorbed bacteria, whereas the inner water droplets (diameter ≈5 µm) were jointly stabilized by the bacteria and unknown low molecular weight surfactants (Figure 3D). Foams can also be made using natural and modified lactic acid bacteria as Pickering stabilizers. Falco et al. used β-casein- and sodium caseinate-modified La-5 to produce dry foams (foams showing fast drainage of the serum phase) that exhibited stability for more than 180 min (Yucel Falco et al., 2017a). Jiang et al. modified the same bacteria using OSA to produce dry foams that were stable for more than 180 min with superior stability, as compared to those using UM bacteria (Jiang et al., 2019). Because the air bubbles in food foam are typically significantly larger than the typical oil droplets of emulsions and due to the micron size of bacteria, bacterial stabilization may be more suited for foams than for emulsions. The new field of bacteria as Pickering stabilizers for emulsions and foams lacks systematic studies of the interrelationships between bacterial surface properties of natural or modified bacteria and the properties of emulsions.

A more complex structure was fabricated using bacteria as colloidal building blocks (Jiang et al., 2022; Jiang et al., 2023b). Jiang et al. noted that fat globules and lactic acid have the same approximate size and share functionality with respect to aggregation and the ability to adhere at the water/air interphase, and proposed that the fat globules of a whippeable emulsion could be substituted with lactic acid bacteria (Jiang et al., 2023b). A whipping cream was formulated using bacteria (35% (V/V), either the hydrophobic LBC or more hydrophilic LBD, as well as sodium caseinate, and a thickening agent, both in low concentrations. The naturally hydrophobic LBC imparted the foam structure with bacteria densely covering the air phase with a network of bacterial aggregates, producing foam with a high modulus and good drainage stability (Figures 3E–I). In contrast, the LBD-based foam was softer with poorer drainage resistance and exhibited a structure where the internal air/water interphase was depleted of bacteria. The bacteria culturability for both formulations was intact upon whipping, although the aggregation of LBC lowered the apparent CFU quantification.

“Colloidosomes” are hollow, water-filled structures analogue to liposomes, but formed by colloidal particles rather than lipid molecules (Thompson et al., 2015). Simple stable colloidosomes were generated by spontaneous water ingression into micro-bubbles Pickering stabilized by OSA-modified LBC bacteria (Jiang et al., 2022). The fact that the structure does not disintegrate after removal of the air/water interphase indicate that a stabilizing principle beyond Pickering phenomena is in action, that is, inter-bacterial attraction of the modified hydrophobic bacteria. The basic anionic bacterial colloidosome was modified using the layer-by-layer method involving cationic and anionic polymers or hybrid combinations of polymers and a second layer of the strongly anionic bacterial strain Pediococcus pentosaceus (PCP), thus forming a bacterial double layer of two strains (Figure 3J). The permeability of the anionic fluorescein permeant could be rationalized by electrostatic repulsions (capsule zeta potential) for anionic colloidosomes and porosity for cationic capsules (Jiang et al., 2022).

Finally is should be mentioned that the health effect of ingestion of modified bacteria is currently understudied. Physical modification based on food components such as proteins will not require approval and is unlikely to induce any toxic effects. Chemical modification by amide and ester bonded fatty acids will most like require approval from authorities, however it can be argued that upon hydrolysis natural fatty acids will be released and a toxic response is unlikely. An interesting future research topic is that the relationship between bacterial surface properties, their binding to the intestinal interphase and the ability of bacteria to colonize the intestine.

Polar lipids are key components of cell membranes that on their own form liquid crystals that possess the long-range orientational order of crystals and the high molecular mobility of liquids. The interplay between the hydrocarbon chain and polar head group interactions controls the assembly in aqueous solvents. This can be understood by the packing parameter of the amphiphilic molecule (Israelachvili et al., 1976; Mitchell and Ninham, 1981). The packing parameter (v/al) is defined as the ratio between the volume of the hydrophobic chain (v) and the product of the head group area (a) and the chain length (l). It is important to note that the packing parameter is not a fixed molecular parameter but depends on the environment. Therefore, different lipid liquid crystalline (LLC) structures illustrated in Figure 4 can be formed by both changing the lipid composition and environmental factors, such as temperature, pH, and ionic strength (Larsson, 1989; Larsson et al., 2006; Barriga et al., 2019). The so-called inverse LLC, where the lipid aqueous interface is curved towards the aqueous solvent, forming aqueous cavities, is of particular interest. This is because the size of the cavity often matches or can be formulated to match the size of functional protein or an enzyme. Varying the lipid composition results in an LLC with different structures and curvatures of the aqueous lipid interface, including inverse bicontinuous cubic (Q₂), sponge phase (L₃), or inverse hexagonal (H₂) structure (Larsson, 1989; Seddon and Templer, 1995; Mulet et al., 2013; Gilbert et al., 2019). The phase behavior of polar lipids with decreasing water content and increasing temperature often follows the sequence (left to right in Figure 4): hexagonal phase (H_I) → lamellar phase (L_α) for water soluble lipids and lamellar phase (Lα) → reversed hexagonal phase (H_II) → inverse micellar phase for lipids (L₂) for lipids with low water solubility. Different types of phases with cubic symmetry (micellar cubic I₁ and I₂, bicontinuous cubic V₁ and V₂) often occur in between these phases. When it comes to bicontinuous cubic phases the two systems of water channels allow the substrates to be transported in and out of the structure. The self-assembled character of the structure also offers advantages when it comes to formulation. The flexibility and softness of the structure makes it easier to include enzyme that are as large or even larger than the water channel dimensions.

The LLC phases can be dispersed into colloidally stable nanoparticles, such as cubosomes, spongosomes/sponge-like nanoparticles, and hexosomes, which are more convenient for many scientific and technological applications that require continuous processing, such as dispensing or injection (Larsson, 2000; Mulet et al., 2013; Barriga et al., 2019; Valldeperas et al., 2019b). Inverse bicontinuous phases (V₂) are nanostructures where the lipid bilayer curves to form water channels in a three-dimensional network (Larsson, 1989). The curvature of such a bilayer can be described as a minimal surface, which is as concave as it is convex, that is, it has a zero mean curvature. The bicontinuous lipid sponge phase can be regarded as a “melted” bicontinuous cubic phase, and as such, it does not present the long-range order of cubic phases (Anderson et al., 1989). The scattering pattern only features two broad diffraction peaks, indicating the mobility of the system (Porcar et al., 2003). It has a more flexible bilayer structure than that of the V₂ phases, and consequently, it is more macroscopically fluid and swollen than the V₂ phases. A dispersing agent is usually added to form LLC nanoparticles in a large excess of water (>90%), but the inner structure is closely related to the corresponding fully hydrated bulk phase (Valldeperas et al., 2016).

Entrapping proteins, including enzymes, with retained bioactivity is challenging due to their large size, complexity, and limited stability outside their natural biological environment. A lipid self-assembly with large water pores is required for the encapsulation of such large biomolecules. Inverse bicontinuous cubic (V₂) and sponge (L₃) phases have been successfully applied for this purpose (Razumas et al., 1994; Angelova et al., 2011; Sun et al., 2015; Valldeperas et al., 2019c; Gilbert et al., 2019). A recent summary shows that cubosomes can be loaded with a wide range of compounds from small drug molecules (<1 kDa) to large proteins with a molecular weight up to 174 kDa (Barriga et al., 2019). However, the activity of entrapped enzymes depends on the size of the water channels in relation to size of the enzyme molecules (Razumas et al., 1994; Sun et al., 2015). It has been shown that the long-term stability from days to a month in terms of retained activity of biosensors based on electrodes coated with monoolein and entrapped enzyme increase in the urease (molecular weight, M_w ≈ 590 kDa) < creatinine deiminase (M_w ≈ 200 kDa) ≈ glucose oxidase (M_w ≈ 160 kDa) < lactate oxidase (M_w ≈ 80 kDa) (Razumas et al., 1994). The enzyme stability can be improved when the water channels are larger than the protein size (Sun et al., 2015). It should be noted that the bicontinuous cubic phase used here are liquid crystalline and can adopt to size of the water channel to the enzyme size. This is even easier for the more flexible sponge (L₃) phase. Here the inclusion of enzymes was found to increase the bending rigidity of the lipid bilayer of the sponge phase as determined by neutron spin echo measurements (Gilbert et al., 2022). The size of the water channels in the LLC structure could also be tuned by changing the environmental conditions, which also opens up the possibility to control the encapsulation and release of enzymes using external stimuli (Fong et al., 2016). Oral delivery of lipid-based carriers, such as liquid crystalline nanoparticles, vesicles, or nano-emulsions, also need to survive the harsh environment through the gastrointestinal tract, where lipolysis is a particular challenge as pointed out by Dully et al. (2020). The explored a strategy using lipase inhibitors, pointed out that there are other strategies like use of non-digestible lipids or use of capsules. If the lipid particle survives the harsh conditions in the stomach all the way down the small intestine the cargo need to traverse the mucus layer prior to absorption of the drug. In some cases, such as in inflammatory bowel disease, local delivery to the intestine is desirable. In this case, absorption is not needed, but the drug or nanoparticle still must penetrate the mucus (Date et al., 2018). In other cases, mucus adhesion is needed, where the nanoparticle adheres to the mucus for a prolonged time, releasing the drug and resulting in a prolonged effect (Gamboa and Leong, 2013). Therefore, the colloidal aspect is clearly crucial for the optimization of the formulation. This is a very wide field of research beyond the scope of this review. A comprehensive review of the topic and associated challenges has been presented by Zhai et al. (2019).

Amorphous forms of silica and silicates are recognized by the US Food and Drug Administration (FDA) and the European Food Safety Authority (EFSA) as the safe ingredients for oral delivery up to the dose of 1.5 g per day (Diab et al., 2017). In biological systems, silica can be found in a variety of nano- and micro-structures, such as silica diatoms (Sumper and Brunner, 2006) and spicules of sponges (Uriz et al., 2003), which inspired the synthesis of various hybrid materials (Nassif and Livage, 2011). Mesoporous silica nanoparticles are extensively used for the encapsulation of drugs with small molecular weights (Manzano and Vallet-Regí, 2020) and for the imaging of biological specimens (Tasciotti et al., 2008). The sol-gel process is a promising and effective strategy to synthesize mesoporous silica for food applications because of the compatibility and non-cytotoxicity to bio-macromolecules, bioactive enzymes, and probiotic bacteria (Diab et al., 2017).

Mesoscopic pores can be used as cavities that facilitate the stability of the encapsulated bioactive enzymes (or proteins in general) under broader pH and temperature conditions (Hudson et al., 2008; Carlsson et al., 2014). As the pore size, typically in the order of several nm to 10 nm, is large enough to allow large molecules to enter and diffuse inside the pores. Therefore, enzymes or their aggregates bind not only to the outer surface of particles, but many of them can also bind to the significantly larger, inner pore surfaces either by physisorption or by chemisorption. It should be noted that the immobilization of the bioactive enzymes into small pores or into cavities with only one opening can block the diffusion of substrates and products, which could reduce the catalytic activity of the confined enzymes (Lee et al., 2009). In contrast, the use of “macropores” facilitates the diffusion of substrates and improves the enzymatic activity, but the enzymes can also exit confinement and diffuse away from the formulation. Therefore, the chemical coupling of enzymes to the inner pore surface or the use of crosslinked or aggregated enzymes is desirable to keep the enzymes inside the mesopores (Wang et al., 2010).

A variety of formulations have been developed and examined for the delivery of probiotics, such as spray-drying, freeze-drying, extrusion, and emulsification (Gouin, 2004; Champagne and Fustier, 2007; Gasperini et al., 2014). Among these methods, emulsification with the aid of surfactants is a gentle technique that causes minimal damage to microorganisms (Krasaekoopt et al., 2003). When an aqueous polymer solution is mixed with oil under stirring, the presence of surfactants (such as polyglycerol polyricinoleate (PGPR)) further contributes to stabilizing the water droplets by reducing the surface tension at the oil/water interface. If the polymers in aqueous droplets can be crosslinked, the microparticles of crosslinked hydrogels can easily be extracted from the oil-in-water emulsions using centrifugation.

The use of crosslinked hydrogels based on biopolymers approved by the US FDA is a straightforward strategy to deliver such formulations to the gastrointestinal tract (Figure 5A). For this purpose, probiotic bacteria can be encapsulated in particles of hydrogels based on anionic polysaccharides, such as alginate, pectin, and xanthan gum, to form hydrogels crosslinked by divalent cations, such as Ca²⁺ (Figure 5B) (Kwiecień and Kwiecień, 2018). Polyalginate (PA⁻), which is crosslinked by divalent cations such as Mg²⁺ and Ca²⁺ by the formation of “egg-box” junctions hosting divalent cations between the stacks of guluronic acid blocks (Katchalsky et al., 1961), is one of the most common encapsulation matrices for probiotic bacteria and enzymes (Orive et al., 2002). PA⁻ is an FDA-approved food ingredient and is a linear polysaccharide extracted from the cell wall of marine brown algae, consisting of disaccharide units based on L-guluronic and D-mannuronic acids (Lee and Mooney, 2012). It is expected that formulations based on anionic polysaccharides, such as PA⁻, are protonated and hence are less soluble at low pH. The formulations are meant to protect probiotic bacteria under gastric conditions, whereas the bacteria are released in the intestines because the de-protonated polyelectrolytes are more soluble at higher pHs.

As PA⁻ microcapsules are not sufficiently stable to protect probiotic bacteria under gastric conditions, the deposition of additional protective layers is desired (Matricardi et al., 2008). For example, the layer-by-layer deposition of cationic and anionic polyelectrolytes (Lvov et al., 1993; Decher, 1997) has been used for the stable encapsulation of enzymes (Caruso et al., 2000), antigens (De Koker et al., 2009), and living cells (Diaspro et al., 2002). ¹H NMR data showed that polyelectrolyte multilayers act as a diffusion barrier by modulating the mobility of water molecules, especially when are stacked (Schwarz and Schönhoff, 2002). Notably, only one additional layer of chitosan on PA⁻ capsules increases the survival rate of Bifidobacterium and Lactobacillus in gastric fluids by two to three orders of magnitude (Figure 5C) (Krasaekoopt et al., 2004; Cook et al., 2011). However, it should also be noted that the size of the particles used in these studies was one order of magnitude larger than the sensory limit of the human tongue (Krasaekoopt et al., 2004), or were not even controlled (Cook et al., 2011). Consequently, several key factors remain unclear, such as (a) the minimum requirements for stable formulations, (b) the status of the probiotic bacteria release from the polymer capsules after reaching the target colon, and (c) whether the functionality of the probiotics is sustained after delivery. Systematic investigations addressing these points are still largely missing.

As examples of the delivery of bioactive molecules using lipid-based formulations, the encapsulation of two key types of enzymes of different sizes, namely aspartic protease (34 kDa or a diameter of about 5.7 nm) and β-galactosidase (460 kDa or a diameter of about 14 nm), are discussed (Valldeperas et al., 2019c; Gilbert et al., 2019). Here we note that the typical water channel diameter for fully swollen, bicontinuous cubic phase is about 5 nm while the corresponding value for a sponge phase is about 13 nm, bearing mind that both phases are flexible structures. Both enzymes are important for food processing, but the storage and delivery of the bioactivity is challenging. Aspartic protease from C. parasitica is a milk-clotting enzyme that is highly specific for κ-casein and promotes curd formation as well as cheese ripening. β-galactosidase from K. lactis (465 kDa, size = 14 nm) hydrolyses lactose to a mixture of glucose and galactose; therefore, it is used in food formulations intended for lactose-intolerant consumers. The ability to deliver iron is becoming increasingly important as the shift to more plant-based foods can lead to iron deficiency. The delivery of iron-carrying heme proteins has become important for food iron supply (Gilbert et al., 2023). Although the curvature of the lipid aqueous interfaces in the non-lamellar phases determines the size of the aqueous cavities, and hence the space given to the enzyme, the interaction between the enzyme and the lipid layer is an important factor that controls the efficiency of the encapsulation.

Mixtures of food-grade acylglycerides and acyldiglycerides can form highly swollen sponge phases (L3) with aqueous pores up to 13 nm in diameter, and can form well defined nanoparticles in excess water with the help of the dispersing agent polysorbate 80 (P80) (Figure 6) (Valldeperas et al., 2016). Aspartic protease (Mw = 34 kDa) assayed by milk clotting capability (78) and β-galactosidase (Mw = 476 kDa, tetramer) assayed by monitoring the Ortho-Nitrophenyl-β-galactoside - > O-nitrophenol (yellow) transition was found to be still active after inclusion into the sponge phase lipid matrix even after 3 months at 25°C (Gilbert et al., 2019). This is not possible if the enzyme is delivered from a solution where the stability at 25°C is a matter of days. It should here be noted that the lipid matrix used only consisted by approved food grade lipids. Interestingly, the two proteins affect the phase behavior of the lipid system differently (Figure 7). The smaller enzyme has a larger effect on the phase structure. The enzyme “likes” the lipid bilayer to be wrapped around itself, indicating that the protein is inside the water channels and that the change in curvature effect is induced by protein-lipid interaction.

The particles adsorb at the interface to form a lipid bilayer, as shown using quartz crystal microbalance with dissipation (QCM-D) and neutron reflectometry (Valldeperas et al., 2019a). Raman spectroscopy results for the sponge phases show large similarities in the lipid chain confirmation and head group interactions as in the lamellar and reverse bicontinuous cubic phase in the same lipid system (Talaikis et al., 2019). Size exclusion chromatography demonstrated efficient encapsulation of both enzymes, yet they retained their enzymatic activity over months, surpassing the storage stability of pure enzymes in solution (Valldeperas et al., 2019c; Gilbert et al., 2019). This may be owing to the penetration of the enzymes into the formed lipid bilayer, as shown using Raman spectroscopy (Valldeperas et al., 2019c) and neutron reflectometry (Valldeperas et al., 2019a). Neutron spin echo and molecular dynamics simulations have confirmed this (Figure 8) (Gilbert et al., 2022).

The formulation of bioactive enzymes in porous silica particles is a highly promising approach (Lee et al., 2009). To address the trade-off between “mesopores” vs. “macropores” (Section 3.2), the use of materials displaying both types of pores has been proposed. Yang et al. reported the “fish-in-net” encapsulation of enzymes by the condensation of pre-formed mesoporous silica particles around the enzyme aggregates in the presence of glycerol (Yang et al., 2006). Bernal et al. synthesized meso-macroporous silica materials (mesopore diameter: 10–40 nm, macropore diameter: 0.1–20 µm) by the polycondensation of sodium silicate using the micelles of cationic surfactants (cetyltriammoniumbromide (CTAB)) in the presence of ethylacetate as the template and encapsulated bioactive enzymes, such as β-galactosidase and lipase (Bernal et al., 2011; Bernal et al., 2014). Blin et al. employed double emulsions based on water/oil/water (W/O/W) or water/solid lipid nanoparticles/water (W/SLPs/W) as templates and synthesized meso-macroporous silica particles, resulting in hexagonally aligned mesopores and scalable macropores (Blin et al., 2014). Pavel et al. applied this formulation (mesopore diameter: 9 nm, macropore diameter: 0.2 µm) to encapsulate β-galactosidase (Pavel et al., 2017). The adsorption of enzymes to meso- or macropores show a dependence on concentration and thus the aggregation state of enzymes. Notably, the oligomers of β-galactosidase adsorbed on the surface of macropores exhibit a high catalytic activity.

The encapsulation of probiotic bacteria requires another strategy because the pore size is too small to accommodate µm-sized cells. For this purpose, a protective silicate layer can be mineralized on polymer-based microparticles using electro-spraying or emulsification. Callone et al. (Callone et al., 2008) reported the encapsulation of Saccharomyces cerevisiae and Oenococcus oeni in PA⁻ particles by injecting cell/PA⁻ droplets into an M²⁺-loaded bath, resulting in the formation of 400–500 µm-sized particles. Subsequently, the surface was coated with amorphous silica synthesized using three methods, namely (i) dipping the particles in silica sol, (ii) consolidating the particles with tetraethoxysilane in an apolar solvent, and (iii) coating the particle surface with methyltriethoxysilane in the gas phase. The particles prepared using (ii) exhibited cytotoxicity, but those prepared using (i) and (iii) exhibited an increase in the function of the cells, such as the fermentation of glucose by Saccharomyces and maleic acid by Oenococcus. Haffner et al. (Haffner et al., 2016) implemented a similar strategy and fabricated Lactobacillus/PA⁻ particles by electrospraying into a bath containing Ca²⁺ ions, yielding 200 µm-sized hydrogel particles encapsulating Lactobacillus. The protective amorphous silica layer was deposited by the electrostatic adsorption of aminotrimethoxysilane, followed by the condensation of tetramethyl orthosilicate. The bacterial viability was assessed by counting the colony forming units (CFUs) before and after incubation in solutions mimicking gastric and intestinal fluids (Minekus et al., 2014). The Lactobacillus encapsulated in silica-PA⁻ particles exhibited only very minor leakages and maintained the viability of the bacteria, whereas all the naked bacteria with no protection died in both fluids. The metabolic function of bacteria can be further evaluated by using an in vitro gastrointestinal simulator (see Section 5.1 for more details) (Haffner et al., 2017). The coating of organic particles with “porous” silica layers further allows for the diffusion of molecules, which helps the bacteria maintain their functions and stabilizes the encapsulated enzymes (Blin et al., 2014; Haffner et al., 2016).

As summarized in Section 3.3, probiotic bacteria can be encapsulated in formulations based on biopolymer-based hydrogel droplets using emulsification (Krasaekoopt et al., 2003). Maintaining the size of the formulation particles to smaller than the sensory limit of the human tongue (80 μm, according to (Tyle, 1993)) is an important requirement for food applications, otherwise consumers would feel the formulations as “additional, foreign ingredients”. For this purpose, it is necessary to optimize the emulsification process. For example, the combination of shear-thinning agents and surfactants enables the formation of small and stable droplets even at a significantly lower emulsification speed. However, this has been either difficult to control or even overlooked in previous studies (Krasaekoopt et al., 2004; Cook et al., 2011). Moreover, hydrogel-based microcapsules prepared by emulsification, such as those based on PA⁻ hydrogels crosslinked by Ca²⁺, cannot maintain their structural integrity in gastrointestinal environments or sustain the functions of the encapsulated probiotic bacteria (Eshrati et al., 2018). The deposition of additional polymer layers is a possible strategy to increase the stability of PA⁻ microparticles and hence the protection of the probiotic bacteria (Matricardi et al., 2008). For example, the layer-by-layer deposition of cationic and anionic polyelectrolytes (Lvov et al., 1993; Decher, 1997) has been used for the stable encapsulation of enzymes (Caruso et al., 2000), antigens (De Koker et al., 2009), and living cells (Diaspro et al., 2002). The deposition of one chitosan layer increases the survival rate of Bifidobacterium and Lactobacillus encapsulated in PA particles by a factor of 100–1,000 under gastric conditions, suggesting that even one additional polymer layer could increase the stability of the formulations (Krasaekoopt et al., 2004; Cook et al., 2011).

Eshrati et al. demonstrated the encapsulation of LGG into small, highly monodisperse PA⁻ particles by the systematic optimization of the preparatory parameters. To avoid mechanical damage to probiotic bacteria during emulsification, it is necessary to stir the oil/water mixture at a low speed (such as 300 rpm) in the presence of surfactants (such as PGPR) and shear-thinning agents (such as xanthan gum). Although LGG encapsulated in crosslinked PA^–particles exhibits an increase in viability, which was assessed by the number of CFUs, by four orders of magnitude, as compared to “naked” LGG, the fraction of viable LGG after 2 h in gastric fluid (3 g/L pepsin and 85 mM NaCl, pH 2.0) is still 0.1%. The fraction of viable LGG in crosslinked PA^–particles is 0.01% after incubating LGG/PA⁻ particles in intestinal fluid (1 g/L pancreatin, 1 g/L bile salt, and 85 mM NaCl, pH 6.5) for 3 h. These results clearly indicate that PA^–particle cores alone do not have a sufficiently high protective function. In contrast, the layer-by-layer deposition of positively charged chitosan (Ch⁺) and negatively charged PA^– (Figure 9A) significantly increases the viability of LGG (Caruso et al., 2000; Cook et al., 2011). Remarkably, the deposition of only three polyelectrolyte layers increases the CFU values by a factor of more than 10⁶ in stomach fluid and 10⁵ in small intestine fluid (Figure 9B). It is notable that a further increase in the layer number did not improve the viability, suggesting that the deposition of three polyelectrolyte layers (chitosan + PA + chitosan), corresponding to a total thickness of only 20 nm, is the minimum requirement to achieve the highest protective function (Eshrati et al., 2018).

Unfortunately, chitosan is not an ingredient approved by the FDA or EFSA, and it is claimed that chitosan is antimicrobial (No et al., 2002; Tsai and Hwang, 2004). The use of food-derived milk proteins is a promising strategy for the stabilization of PA^–particles by the layer-by-layer deposition of charged macromolecules because milk proteins possess both positive and negative net charges near neutral pH, such as lactoferrin (+), caseinate (−), and whey protein isolates (−). Falco et al. used several milk proteins for the layer-by-layer coating of 150 µm-diameter PA⁻ particles encapsulating LGG (Yucel Falco et al., 2019). The layer-by-layer deposition was monitored using the change in the zeta potential and QCM-D. The coating with only milk proteins did not lead to sufficiently high stability. However, the deposition of Ch⁺ as the first layer improved the stability. The treatment of PA^–particles coated with Ch⁺ and additional milk protein layers in gastric fluid for 30 min resulted in a decrease in the CFU number by three orders of magnitude. Unfortunately, a longer exposure to gastric fluids resulted in the death of all bacteria, indicating insufficient stability. The use of materials approved by the FDA and EFSA is a necessary step to apply such (bio)polymer-based formulations as food ingredients.

Moreover, layer-by-layer coating with polyelectrolytes can be applied on the surfaces of the negatively charged cell wall of bacteria, enabling the use of a single bacteria as the smallest delivery vehicle (Priya et al., 2011; Yucel Falco et al., 2017b). Falco et al. coated individual bacteria with sulfated β-glucan and Ch⁺, because β-glucan can resist the human enzymatic system but not the bacterial glycanases in the small intestine (Yucel Falco et al., 2017b). The layer-by-layer deposition was confirmed on planar model surfaces using QCM-D and spectroscopic ellipsometry, whereas the deposition on the bacteria was confirmed using zeta-potential measurements. Although the formulations could resist the acidic gastric conditions, they were partially degraded during exposure to a combination of simulated intestinal fluid containing glucanase, indicating the release mechanism of bacteria in the intestine.

Cellulose derivatives, such as carboxymethyl cellulose (CMC) and hydroxyethyl cellulose (HEC), have been used for the development of probiotic delivery systems based on crosslinked edible films (Singh et al., 2019), water in water (W/W) emulsions with gelatin (Singh et al., 2018a), and coacervates with Ch⁺ with genepin (Singh et al., 2018b) or without (Singh et al., 2017). Food grade cellulose derivatives are water soluble, typically due to the presence of carboxylic acid groups, and serve as thickeners in the food industry. Therefore, these components require additional gelling agents or crosslinkers to form stable particles or films that do not readily dissolve or disintegrate. CMC- and HEC-based edible films were cast and crosslinked under mild conditions by drying at 50°C after the addition of citric acid (Singh et al., 2019). The lactic acid bacteria in LGG could be absorbed into the films and the pH dependency of film swelling could be tuned to provide a low degree of swelling at gastric pH and high degree of swelling at intestinal pH by adjusting the CMC/HEC ratio and citric acid content. Overall, the films could be loaded with a high quantity of bacteria and provided good protection.

Water in water emulsion is a phenomenon driven by thermodynamic incompatibility between pairs of water-soluble polymers. The interfacial tensions are extremely small, the interphase has a large width, and both inhibit the stabilization with small (molecular) emulsifiers; therefore, Pickering stabilization using nano- or micron-scaled particles must rather be employed. Singh et al. (2018a) used CMS and gelatin at temperatures above 35°C to avoid the gelation of gelatin and to introduce LGG bacteria in dispersed CMC droplets. The bacteria exhibited a specific affinity for the CMC phase and was intensively present at the interphase of the particles, which was attributed to Pickering stabilization phenomena and the high desorption energy of the large size of the bacteria. The formulation strategy resulted in the enhanced survival of entrapped LGG when exposed to simulated gastric and intestinal conditions. Sub-millimeter CMC particles were produced following a polymer coacervation route by the dropwise addition of a CMC solution into a chitosan solution (Singh et al., 2018b) in combination with a mild crosslinker, namely genipin. The addition of sodium caseinate prevented particle aggregation due to electrostatic repulsion of the otherwise nearly neutral particles. The formulation significantly improved the survival rate of the bacteria when exposed to simulated gastric and intestinal conditions, as compared with the unformulated LGG bacteria. Since probiotics can be administered through food, the long-term culturability was tested over 30 days, showing a significantly improved stability, as compared with the unformulated samples. However still up to five log units of loss was detected even for the improved CMC-chitosan formulations which still limits the practical applications.

As the physical access to the digestive product after each compartment in the gastrointestinal tract in vivo is not realistic, several types of in vitro systems have been developed to simulate the entire gastrointestinal tract (Van de Wiele et al., 2015). In vitro gastrointestinal simulators are powerful tools to not only assess the protective capacity of formulations but also to quantify the metabolic function of LGG during its passage through the entire gastrointestinal tract in the presence of nutrients (starch and other carbohydrates) and other bacteria. This assay is distinct from the previous CFU counting assay because only the probiotic bacteria that are released from the formulation can be detected using flow cytometry. Namely, the bacteria that are encapsulated and protected by the formulation are not detected. Once the LGG/PA⁻ particles were coated with three polyelectrolyte layers, no released bacteria could be detected throughout the entire gastric process (1 h at pH 4.8, 0.5 h at pH 3.0, and 0.5 h at pH 2.0). Notably, the residence in the small intestine chamber showed an increase in the released LGG by a factor of five, as compared with the initial level, suggesting the release of LGG that proliferated in the formulations by the deprotonation of chitosan (pK_b = 6.5) (Guarino et al., 2015). Most significantly, the amount of lactate (metabolic product) produced increased in the ileum, small intestine, and colon chambers by five orders of magnitude (10⁵ ×), as compared with the free LGG (Eshrati et al., 2018), demonstrating that the 40 µm composite polymer formulations significantly improved the delivery of functional probiotic bacteria without being detected by the tongue of the consumer (Figure 9C).

After their release from the formulations, probiotic bacteria adhere on the surface of the colon epithelium and create colonies (Sonnenburg et al., 2004). The surface of the gastrointestinal tract is covered with a mucosal layer consisting of mucin, which protects against the invasion of pathological bacteria (Pelaseyed et al., 2014). An increasing number of studies have demonstrated that the protective function of mucus is supported by the presence of phospholipids, especially, phosphatidylcholine and lysophosphatidylcholine (Ehehalt et al., 2004).

AFM is a commonly used method to reveal the mechanism of the initial step, which is the adhesion of LGG on the mucosal surface of intestinal epithelial cells. Earlier studies have shown that bacteria can be immobilized on a cantilever surface by polycations, such as poly (lysine) and poly (ethyleneimine), and the force-distance curves can be measured on various substrates (Cao et al., 2006; Boks et al., 2008; Rowan-Nash et al., 2019). Dague et al. coated plastic dish surfaces with porcine gastric mucin using physisorption as a model of the gastrointestinal tract surfaces, and measured the force-distance curves with a Lactococcus lactis-coated cantilever (Dague et al., 2010). The force-distance curves measured using bacteria-coated cantilevers are distinct from those measured using intact Si₃N₄ cantilevers. As the surface density of mucin is hardly controllable by physisorption, Amadei et al. reported the immobilization of biotin-functionalized mucin onto planar, supported lipid membranes incorporating biotinylated lipids via neutravidin crosslinkers (Amadei et al., 2018). A high-affinity binding of biotin and neutravidin (K _D ∼ 10^–15 M) enables the grafting of biopolymers by controlling the self-assembly of biotinylated lipids (Kaindl et al., 2012; Rieger et al., 2015). Instead of measuring interfacial forces using AFM, the microinterferometry technique has been used to calculate the mean force potential from the height fluctuation of cell-sized beads (Higaki et al., 2017). Although mucin is highly glycosylated, carrying negative net charges, the mucin-functionalized surface exhibits a markedly stronger interaction with silica particles coated with phosphatidylcholine than those coated with positively charged lipids, suggesting that the mucin-phosphatidylcholine interaction is not driven by electrostatic interactions, as suggested by a previous isothermal calorimetry study (Korytowski et al., 2017).

As a more realistic model for the interaction of the released probiotic bacteria with the intestinal tract surfaces, Eshrati et al. fabricated a polarized monolayer of enterocyte-like cells and goblet cell-like cells differentiated from CaCo2 and HT29 MTX cells (Figure 10A). After the establishment of a polarized epithelial monolayer with tight junctions, the sample was transferred into a parallel plate flow chamber. The adhesion of LGG on the surface was monitored under the shear stress that covers the bacteria in vivo, 0.1–0.5 Pa (Eshrati et al., 2019). It should be noted that the cells need to be cultured in Transwell^® culture until they establish the apico-basal polarity because the commonly used cell culture on plastic or glass substrates is not able to establish the polarized epithelial cell monolayer connected with tight junctions to express the layer of mucin proteins. Therefore, the differentiated cell monolayer with a sufficiently high transepithelial electric resistance (≥400 Ω cm²) was transferred into a microfluidic chamber (Fröhlich et al., 2021). In addition to the establishment of tight junctions (Figure 10B), the expression of mucin 2, which is a major constituent of mucus layers that build up a protective barrier together with phospholipids on the epithelial surfaces against pathogens (Pelaseyed et al., 2014; Korytowski et al., 2017), was confirmed by immunofluorescence imaging (Figure 10B). Subsequently, the Young’s modulus of the mucin layer was measured using AFM nano-indentation, which indicated that the goblet cell-like cells derived from HT29 MTX cells produced thicker and softer mucin layers, E _HT29 = 0.2 kPa and E _CaCo2 = 10 kPa. Interestingly, the surface density of adhered LGG (Figure 10C) exhibited a clear maximum at approximately 0.3 Pa, suggesting that the adhesion of LGG to the mucosal surface in the colon is shear-dependent. Moreover, the persistence of LGG adhesion was further examined by a stepwise increase in the shear stress exerted on the adherent LGG, demonstrating that some LGG cells can withstand high shear stress even beyond the physiological levels. The presence of such “strong binders” might contribute to the maintenance of the chronic homeostasis of microbiomes in the gastrointestinal tract.

Finally, adhered bacteria start forming new microbiomes by forming colonies. To characterize the interactions between bacteria, the timescale of bacterial auto-aggregation is commonly used in microbiology. The assay involves monitoring the time evolution of the loss of OD due to sedimentation of the aggregated bacteria (Thomas et al., 2018). Freely suspended single bacteria remain in suspension for a significantly longer time because the sedimentation velocity (v) of a particle with a radius (r) increases with v ∝ r − 2 . The adhesion force between bacteria can also be measured both on the cellular level and on the single molecular level using optical tweezers and single molecule force spectroscopy, respectively (Verbelen et al., 2007; S Bezryadina et al., 2016).