Introduction

Insoluble enzyme preparations are useful catalysts for industrial biotransformations. Being more easily separated and recycled than their soluble counterparts, they are conducive to continuous process development and are therefore often the preferred choice of biocatalyst for manufacturing scale process operation (Buchholz et al. 2005; Bommarius and Riebel 2005). The originally soluble enzyme is usually made insoluble by its attachment to the surface of an insoluble carrier, a process called immobilization (Cao et al. 2003; Cao 2005; Sheldon 2007a). There are innumerable variants of the general procedure, attributable to the large number of different carriers available and the various exploitable strategies for fixing the enzyme to them (Cao et al. 2003; Cao 2005). Encapsulation in organic (Buchholz et al. 2005; Cao 2005) or inorganic gels (Betancor and Luckarift 2008) constitutes a special mode of carrier-mediated enzyme immobilization. An often made observation for carrier-bound insoluble enzymes is that in comparison to the base preparation, they have reduced specific activity but have acquired extra shelf-lifetime and stability under the operational conditions. Despite notable recent efforts (Hanefeld et al. 2009), the molecular basis for gain and loss of activity and stability in immobilized enzymes is not well understood. Unfortunately, design of enzyme immobilization for optimum catalyst performance is still largely empirical (Sheldon 2007a).

There are two other main drawbacks of carrier-bound insoluble enzymes. A large amount of non-catalytic mass (≥90% of the total mass of biocatalyst) is introduced to a bioreactor operating with carrier-bound enzymes. This can be a serious problem when enzymes are employed that have a low specific activity. Furthermore, the carriers used are often considerably more expensive than the enzymes attached to them. Therefore, if one managed to precipitate the target enzyme into an insoluble particle that shows good retention of the specific activity of the soluble protein and at the same time has useful resistance to mechanical stresses, a powerful carrier-free, yet heterogeneous biocatalyst might be obtained and pitfalls with the use of solid carriers could be partly eliminated. This review summarizes recent advances in methodology for carrier-free enzyme immobilization. Reported approaches are distinguished according to the amount of biological recognition and specificity involved in the process of protein aggregate formation.

Cross-linked enzyme aggregates

A solid preparation of an enzyme can principally be obtained by precipitation or crystallisation. Cross-linking with bifunctional reagents such as glutardialdehyde is usually required to stabilize the insoluble form in suspension during further application. Cross-linked enzyme crystals and aggregates have been introduced, with variable success, to the field of biocatalysis.

The method of cross-linked enzyme crystals (CLECs), discovered in the 1960s (Quiocho and Richards 1964), has since been applied to a limited number of enzymes including ribonuclease A, subtilisin, caprboxypeptidase B, alcohol dehydrogenase, and some lipases (Cao et al. 2003). The CLECs prepared often displayed useful resistance against denaturation by heat, organic solvents and proteolysis (Sheldon 2007b). Crystal size and retention of specific activity could be tuned in some cases by variation of the conditions for crystallization and cross-linking (Margolin 1996). As shown for Burkholderia cepacia lipase, coating of CLECs with surfactants or β-cyclodextrin could contribute to further enhancement of shelf-life and operational stability (Rajan and Abraham 2008). The scope of the CLEC technology is currently limited by the requirement for a successful crystallisation of the (purified) target enzyme.

Insoluble catalysts were also obtained by cross-linking enzymes (CLEs) directly in solution. However, the resulting enzyme preparations were difficult to handle and mechanically unstable. The precipitation required delicate balancing of a number of process variables (protein and cross-linker concentration, pH, temperature, ionic strength, mixing and others) and was therefore generally not reproducible (Cao et al. 2003).

In recognition of the limitations of CLECs and CLEs, Cao et al. (2000) introduced, and later commercialized (http://www.cleatechnologies.com), the technology of cross-linked enzyme aggregates. The original abbreviation CLEA is now also a registered trade name. As first demonstrated for penicillin G acylase, and subsequently applied to an impressive number of enzymes (e.g. lipases, proteases, amidases, esterases, phytase, oxynitrilase), CLEAs are made by a general procedure that involves salt- or solvent-induced (micro)aggregation of the enzyme or a desired combination of several enzymes, followed by chemical cross-linking to obtain a stable all-protein precipitate. For example, ammonium sulfate, tert-butyl alcohol or polyethylenglycol were useful to trigger the aggregation of penicillin G acylase. The cross-linking reagents can likewise be varied according to requirements of the particular enzyme used (Mateo et al. 2004; Kubáč et al. 2008). However, glutardialdehyde was most often applied. CLEAs were generally found suitable for biocatalysis in organic solvents (Cao et al. 2000). Two or more enzymes can in principle be joined in an aggregate, yielding a so-called Combi-CLEA (Mateo et al. 2006). Preparation of a Combi-CLEA containing pectinase, xylanase and cellulase activity was a notable success (Dalal et al. 2007). Combi-CLEAs might be interesting multi-purpose catalysts. They might be also useful for multi-step biocatalytic transformations (Vafiadi et al. 2008).

While CLEAs do eliminate the need to crystallize the enzyme of interest, there remains the inherent disadvantage that the exact protocol of aggregation and cross-linking must be established for each enzyme anew, implying that a substantial effort must be devoted to the optimization of conditions, as discussed below. In addition, some purification of the crude protein will in most cases be required prior to its precipitation.

Conditions used for aggregation and cross-linking

High-throughput experimentation facilitates optimization of the different interacting process variables (precipitant and the concentration thereof, cross-linker, etc.) that are involved in the preparation of CLEAs (Schoevaart et al. 2004). Maximum retention of specific activity of the soluble enzyme is a principal concern. However, structural properties like size, porosity and resistance to mechanical stress are also important (Yu et al. 2006). Choice of the best applicable precipitant is usually empirical but can draw on the long-standing experience in protein purification (Bommarius and Riebel 2005). Several studies underscore the importance of precipitant selection in the preparation of CLEAs (López-Serrano et al. 2002).

Despite its widespread use for cross-linking of proteins, the small and highly reactive glutardialdehyde, particularly when used in high concentrations, often causes large losses of activity in enzymes (Schoevaart et al. 2004), either due to chemical modification of groups essential for function or general denaturation induced by the derivatization. Addition of substrates or ligands during cross-linking can therefore provide useful protection. It is also possible to replace glutardialdehyde by dextran polyaldehyde, a commercial polymer of 100–200 kDa that is obtained by oxidation of dextran. Due to size exclusion, penetration of dextran polyaldehyde into the enzyme precipitate is restricted as compared to glutardialdehyde and activity retention may be improved significantly, as shown for penicillin G acylase (Mateo et al. 2004). Some enzymes that were inactivated by the use of glutardialdehyde in the CLEA process like alcohol dehydrogenase and some nitrilases and nitrile hydratases could be successfully cross-linked with dextran polyaldehyde (Kubáč et al. 2008).

Addition of bovine serum albumin (BSA) or polyethylenimine during the cross-linking step served the dual purpose of stabilizing the enzyme activity and providing additional amino groups for reaction with glutardialdehyde. CLEA preparations of a lipase, penicillin G acylase, and laccase were notably improved with respect to specific activity and resistance to elevated temperature or chemical denaturation using cross-linking in the presence of BSA (Shah et al. 2006; Cabana et al. 2007). Paucity of surface amino groups in glutaryl acylase has probably prohibited production of CLEAs using the “conventional” approach. Using polyethylenimine to provide extra sites for cross-linking, stable CLEAs of this enzyme were obtained (López-Gallego et al. 2005).

At least for some CLEAs, size clearly matters and structural features of the resulting preparation are relevant for process optimization. Yu et al. (2006) showed optimum activity for CLEAs of Candida rugosa lipase in particles having a diameter of about 40–50 μm. Control over CLEA size is anything but easy because the (possible) influence from essentially all process variables has to be considered. An interesting development was the preparation of dendritic CLEAs of trypsin using CO2-expanded reverse micelles, obtained from bis(2-ethylhexyl)sulfosuccinate and isooctane (Chen et al. 2006). Under raised CO2 pressure, trypsin precipitates in the reversed micelles and could subsequently be cross-linked with glutardialdehyde. The size of the resulting CLEAs varied between 7 and 38 nm and was adjustable by varying the water-to-surfactant ratio and the concentration of the enzyme in the reverse micelle. The specific activity of dendritic CLEAs exceeded in some cases the specific activity of conventionally prepared CLEAs.

Operational stability

Published work on CLEAs has focused mostly on biocatalytic applications in organic synthesis (Sheldon 2007b). A detailed biochemical engineering analysis of the use of CLEAs in (continuous) processes is to our knowledge lacking. Several authors have recognized the problem of weak resistance of CLEAs to mechanical stresses such as those applied by stirring (Khare et al. 1991; Kim et al. 2007). CLEAs of penicillin G acylase were encapsulated in a polymeric matrix prepared from polyvinylalcohol (Wilson et al. 2004), the so-called LentiKats (Jekel et al. 1998). The resulting enzyme preparation was remarkably stable (>50 days) under shaken conditions in buffer and also resisted incubation in aqueous dioxane (70% v/v) for several hundreds of hours. Encapsulation resulted in a decrease in activity to 60% of the free CLEAs (Wilson et al. 2004). Entrapment into hierarchically-ordered mesoporous silica is another approach for CLEA stabilization against mechanical stress. Stability of α-chymotrypsin and lipase under rigorous shaking conditions was increased more then tenfold (Kim et al. 2007).

In summary, CLEAs have become widely known in the field of biocatalysis and have proven their usefulness for preparative organic synthesis on small scale of operation. Given suitable optimization, the technology has broad scope and is in principle applicable to any available enzyme. However, it is currently not clear if conventionally prepared CLEAs are robust enough to compete with established carrier-bound insoluble enzyme preparations used in industrial scale biotransformations, the conversion of β-lactam antibiotics for example. However, CLEAs are certainly compatible with existing technologies of encapsulation that could be used to make them more resistant to operational stress factors. Application of CLEAs in fixed or fluidized bed reactors used in continuously operated processes should be more often demonstrated in the future (Hara et al. 2007; Aytar and Bakir 2008; Vafiadi et al. 2008). CLEAs of a thermostable γ-lactamase were recently employed in a flow microreactor built from a capillary column (Hickey et al. 2009). The highly miniaturized reaction system was applied for substrate screening and enzyme kinetic characterization, and the CLEA-microreactor showed useful stability under conditions of continuous processing (Hickey et al. 2009). The microfluidic reactor containing immobilized enzymes, perhaps in the form of CLEAs, could be a generally useful tool for biocatalytic process development (Hickey et al. 2007; Thomsen and Nidetzky 2009; Schwarz et al. 2009).

Selective pull-down domains

A significant improvement of CLEA production could come from a molecular design of the aggregation process such that it now features elements of biological recognition and specificity. Not only would optimization of conditions for enzyme precipitation become unnecessary, but it should also be much easier to achieve control over retention of activity and final particle size during formation of the insoluble protein.

Fusion to a highly soluble protein such as the maltose-binding protein (Kapust and Waugh 1999) is a well known approach to improve the solubility of target proteins that are otherwise prone to aggregation (inclusion body formation) under conditions of recombinant protein production. The authors have suggested exploitation of the reverse approach for production of enzyme aggregates. Conceptually, a chimeric protein is generated that harbors the enzyme of interest fused to a very poorly soluble protein such as the cellulose-binding module of Clostridium cellulovorans (CcCBM) (Nahálka and Nidetzky 2007). Protein production in a heterologous host such as Escherichia coli is carried out under conditions that induce selective pull-down of the folded chimeric protein via intermolecular self-aggregation of the CcCBM. The idea that enzymes trapped in inclusion bodies can be catalytically active and therefore are not completely useless is certainly not new. In their significant studies on the dynamics of protein deposition in inclusion bodies, Villaverde and Carrió (2003) showed that retention of biological function in such aggregated proteins is actually quite common.

A first case for application of the pull-down domain was provided using the dimeric flavoenzyme d-amino acid oxidase from Trigonopsis variabilis (TvDAO) (Nahálka and Nidetzky 2007). Upon fusion of CcCBM to the N-terminus of the enzyme, TvDAO synthesized in E. coli BL21 (DE3) was targeted almost completely into intracellular protein particles. Enzyme-lacking CcCBM was almost completely soluble under otherwise identical conditions for recombinant protein production, indicating that the pull-down domain fulfilled the expected solubility switch function efficiently. Aggregates of CcCBM-TvDAO displayed about 40% of the specific activity of the purified soluble oxidase. The high activity level in oxidase aggregates was remarkable for two reasons. First of all, it indicates that CcCBM shows little interference with folding and function of its fusion partner, despite the structural complexity of TvDAO. Secondly, it suggests that the overall catalytic activity deposited in the aggregates was not strongly compromised by mass transfer effects or in other words, the insoluble enzyme was readily accessible to its substrates, d-amino acid and O2. Using SDS–PAGE, it was shown that the CcCBM-TvDAO preparation as isolated contained mainly the protein of interest. CcCBM-TvDAO displayed weak bind affinity to microcrystalline cellulose and was substantially more stable than the soluble oxidase under conditions of d-methionine conversion in a bubble-aerated bioreactor. Encapsulation of CcCBM-TvDAO contributed additional stability (Nahálka and Nidetzky 2007).

Nahálka and colleagues have demonstrated the scope of CcCBM as a versatile pull-down domain. They recently obtained catalytically active aggregates of a maltodextrin phosphorylase (Pyrococcus furiosus) (Nahálka 2008), sialic acid aldolase (SAA; E. coli K-12) (Nahálka et al. 2008) and a polyphosphate kinase (PPK; Silicibacter pomeroyi) (Nahálka and Pätoprstý 2009) with useful retention (≥83%) of the specific activity of the respective soluble enzyme. CcCBM-SAA particles were further entrapped in alginate hydrogels stabilized by cross-linking with glutardialdehyde (Nahálka et al. 2008). The insoluble SAA catalyst was still reasonably active (51%) and displayed high operational stability as well as useful recyclability in a repeated batch process for neuraminic acid synthesis (20 rounds of conversion). It could be freeze-dried and re-hydrated with nearly full retention of activity. Work with CcCBM-PPK demonstrates yet another important advantage of selective pull-down of the target enzyme as in this case, contaminating phosphatase activity from the E. coli background is efficiently removed by simple washes of the insoluble protein fraction (Nahálka 2008).

In a recent study unrelated to biocatalysis, CcCBM was applied to prepare functional aggregates of different lectins (Nahálka et al. 2009). A microtiter plate assay was developed in which the insoluble lectins should be used for glycoprotein recognition in a high-throughput format. CcCBM fusions of two sialic acid-binding lectins were shown to “agglutinate” (precipitate) sialylated proteins under the assay conditions, leading to a positive read-out when multivalent binding by the lectin occurred (Nahálka et al. 2009).

In summary, fusion to CcCBM as pull-down domain appears to be a generally applicable method of producing insoluble, yet highly active aggregates of different enzymes. Noncovalent attachment of the CcCBM-enzyme to cellulose will facilitate separation of the protein particles from bulk liquid (Nahálka and Nidetzky 2007). However, additional processing steps (encapsulation, cross-linking) may be needed to finally obtain an operationally robust biocatalyst. Because aggregation under control of a pull-down domain is unlikely to enhance the intrinsic stability of the enzyme used, there is a clear potential in applying (thermo)stable catalysts (Littlechild et al. 2007), provided that extreme process conditions (like high temperature or high salt, for example) do not cause unwanted disaggregation.

Self-assembling systems

Well-known for lipids and non-natural polymers (Graff et al. 2004), the self-assembly into insoluble ordered structures is a feature recognized for a growing number of proteins (Bayley et al. 2004; Hampp and Oesterhelt 2004; Sleytr et al. 2004). Using chimeric proteins in which the enzyme of interest is fused to a module capable of self-assembly it could be possible to obtain insoluble enzyme preparations that have a well defined, perhaps crystalline-like association state. Exploitation of self-assembly properties may allow one to achieve relatively precise control over size and shape of the resulting enzyme particles. Although of interest, encapsulation through self-assembly of another protein or peptide entity (disjoined from the target enzyme) is not considered here.

Fusion to a cell surface layer protein

Cell surface layers (S-layers) are common “protein membrane” structures of the cell envelope of archaea and bacteria (Schuster et al. 2008). They have a lattice-like appearance and are formed through a self-assembly process. Isolated protein units of natural S-layers are capable of recrystallization into precisely defined monolayers in suspension and at different interfaces. These properties of S-layer proteins have promoted a wide range of (nano)biotechnological applications (Sleytr et al. 2007). Although fusion to an S-layer protein has been explored for protein array development (Tang et al. 2008) and other analytical purposes (Sleytr et al. 2007), there are surprisingly few examples showing application of S-layer technologies to enzyme immobilization and biocatalyst development. However, a notable common feature of S-layer fusions was retention of the original function of the individual components of the chimeric protein, that is, the self-assembly properties and the respective biological activity (Sleytr et al. 2007).

A thermophilic laminarinase (LamA from Pyrococcus furiosus) fused to the S-layer protein SbpA from Bacillus sphericus was produced as an insoluble aggregate in E. coli that was unfortunately not characterized. However, after solubilization by chemical unfolding and renaturation, the SbpA-LamA fusion was successfully immobilized on different solid supports (Tschiggerl et al. 2008). Liposomes featuring a self-assembled fusion of an enzyme and an S-layer protein on the surface constitute an interesting approach towards potentially useful biocatalysts. The easily tuneable size of liposomes could provide flexibility in design of the immobilized enzyme preparation. Using glucose-1-phosphate thymidyltransferase fused to the C-terminus of the S-layer protein SgsE from Geobacillus stearothermophilus, Schäffer et al. (2007) also demonstrated the possibility for recycling of the insoluble, liposome-supported catalyst.

It is worth noting that production of S-layer fusion proteins in E. coli and other hosts (Saccharomyces cerevisiae or HeLa cells) appears to generally yield insoluble products resulting from intracellular self-assembly or aggregation (Blecha et al. 2005). These “inclusion bodies” might already be useful as carrier-free insoluble enzyme preparations.

Fusion to other scaffolding proteins

Proteins forming high-molecular weight assemblies have been used for display of catalytic modules on the surface of the resulting insoluble supramolecular structures. Relevant examples include the stress-responsive 145.8 kDa protein SP1 from aspen plant (Populus tremula) which forms a ring-shaped homododecamer (Wang et al. 2002) and is highly resistant to extreme conditions such as low and high pH, high temperature (T m of 107°C), organic solvents and the presence of various proteases (Wang et al. 2006). A construct in which SP1 was fused to glucose oxidase (Gox) from Aspergillus niger was produced as an insoluble aggregate in E. coli. Thus precipitated oxidase showed substantially improved stability as compared to the native enzyme. The non-catalytic mass of the Gox-SP1 aggregate was only 10% (w/w) of the total particle mass. Gox-SP1 was assembled into nanotube structures containing hundreds of enzyme molecules per tube (Heyman et al. 2007).

In another, however, rather preliminary example, Candida antarctica lipase B was designed for protein particle formation by joining the enzyme to a capside protein of the potato virus (Carette et al. 2006). Insoluble virus protein obtained through co-expression of the fusion protein and the native capside protein in the plant Nicotiana benthamiana displayed esterase activity towards p-nitrophenyl caproate as substrate.

Fusion to an elastin-like polypeptide

Elastin-like polypeptides (ELPs) of variable length containing the pentapeptide repeat Val-Pro-Gly-Xxx-Gly where Xxx is any amino acid (Meyer and Chilkoti 1999) precipitate reversibly upon heating above the transition temperature. Numerous ELP fusions were developed with the aim of facilitating downstream processing recombinant proteins through their selective precipitation from crude cell extracts of the host organism (Meyer et al. 2001). The technology holds considerable promise for biocatalysis but has not been widely applied in the field (Fujita et al. 2009). One could imagine a general procedure in which an aggregate of ELP-enzyme is formed first and stabilized by an ensuing cross-linking step. Methods for cross-linking of elastin-like polypetides are proposed in the literature (Lim et al. 2007).

Fusion to proteins binding to polyhydroxyalkanoate granules

Polyhydroxyalkanoates (PHAs) are naturally occurring polyesters, found in various bacteria that stockpile them for energy storage (Madison and Huisman 1999). A wide range of bioplastic materials have been prepared from PHAs (Lu et al. 2009). An interesting and exploitable feature of PHAs is that the biosynthetic enzyme (PHA synthase) remains covalently attached to the emerging polyester granule. PHA synthase can be expressed in a wide range of host organisms (Rehm 2003). Fusion proteins harboring an active PHA synthase should therefore produce PHAs during their recombinant production and at the same time attach themselves to the polyester particles. Proof of concept (for biocatalysis) was obtained using a β-galactosidase-PHA synthase fusion (Peters and Rehm 2006). A sandwich-type binding of β-galactosidase to a polyester particle displaying a PHA synthase fusion with anti-β-galactosidase scFv (single chain variable fragment of an antibody) was demonstrated (Grage and Rehm 2008). While potentially useful for analytical purpose, it would seem that simple procedures of attaching the enzyme of interest to PHAs are preferred in biocatalysis.

An alternative way of immobilizing enzymes on PHAs might be through fusion with a phasin. The phasins are small (14–30 kDa) amphiphilic proteins which are the main constituents of the PHA-granule associated proteome (Moldes et al. 2004; Neumann et al. 2008). Co-expression of genes encoding the relevant phasin fusion and PHA biosynthetic enzymes might provide the desired enzyme activity immobilized on PHA granules.

(Auto)encapsulation using biomimetic silica

The silaffin polypeptide from the diatom Cylindrotheca fusiformis induces precipitation of silica from monosilicic acid solution (Kröger et al. 2002). The repeat unit of the silaffin polypeptide, the R5 peptide (H2N-SSKKSGSYSGSKGSKRRIL-COOH), was utilized to immobilize butyrylcholinesterase. Highly active and stable biosilica spheres were obtained (Luckarift et al. 2004). The addition of R5 peptide to a solution containing enzyme and silicic acid promoted bio-silification at room temperature and under mild chemical reaction conditions (Naik et al. 2004). Using R5 fusions, phosphodiesterase and organophosphate hydrolase peptide were recently produced as immobilized enzyme preparations in E. coli. Each of the fusion proteins initiated silica polycondensation, and enzymatic activity (69–97%) was retained in the resulting silica spheres (Marner et al. 2009). An additional metal ion complexing peptide (His)6-tag enabled direct recovery of the silified insoluble catalyst from crude cell extract, as demonstrated for d-amino acid oxidase (Chien and Lee 2007).

Conclusions

Table 1 compares, on a semi-quantitative basis using different criteria, the methods of carrier-free immobilization discussed in this paper. The pros and cons related to biocatalyst preparation and process operation are summarized. Among the methods available, CLEA technology is clearly the one most advanced and widely used. However, fusion proteins designed for aggregation could improve the current protocols for production of CLEAs. Encapsulation may in all cases be needed to eventually obtain mechanically stable biocatalysts.

Table 1 Semi-quantitative comparison of methods used for preparation of carrier-free insoluble enzymes
Full size table