Rescuing Recombinant Proteins by Sequestration Into the P22 VLP

Abstract

Here we report the use of a self-assembling protein cage to sequester and solubilize recombinant proteins which are usually trafficked to insoluble inclusion bodies. Our results suggest that protein cages can be used as novel vehicles to rescue and produce soluble proteins that are otherwise difficult to obtain using conventional methods.

Currently, E. coli is the most widely used host system for the heterologous expression of recombinant proteins. A large variety of E. coli strains, expression vectors, and other supporting materials are commercially available and enable expression of everything from eukaryotic codon-optimized genes to the addition of affinity tags to any protein for rapid and easy purification. Unfortunately, recombinant proteins are sometimes expressed as insoluble aggregates or inclusion bodies. Recovering proteins from inclusion bodies often requires time consuming, low yielding methodology that might not produce the active protein/enzyme desired. Preventing proteins from forming insoluble aggregates would be beneficial since many proteins are not studied due to the difficulty of recovering them. Here we describe a strategy for preventing proteins from forming inclusion bodies, and protecting them, through a genetic co-expression of the protein of interest with a protein cage that can sequester the recombinant protein from the cellular milieu shortly after expression (Figure 1).

Figure 1.

Schematic for insoluble protein formation during expression and a proposed mechanism for rescuing these proteins inside a VLP

The strategy of sequestering proteins inside protein cages to solubilize, stabilize, and protect them is intrinsic to nature which, for example, uses chaperones to encapsulate and refold proteins. In our strategy we look to utilize engineered virus like particles (VLPs), a class of protein cages, which have been shown to encapsulate proteins in vivo.^1–10 For the study described here, we use the inherent characteristics of the T = 7 icosahedral bacteriophage P22 capsid. The P22 VLP is assembled from 420 copies of a 46.6 kDa coat protein (CP) as well as 130–300 copies of 33.6kDa scaffold protein (SP) that aids in the assembly process of a 58 nm procapsid.^11–13

We have shown that we can encapsulate a number of soluble proteins by heterologous co-expression of protein cargo-truncated SP fusion proteins with CP, including GFP and enzymes that retain their catalytic activity.^4–8 Rescuing proteins in vivo before they form inclusion bodies, by co-expression of a fusion protein of the insoluble protein of interest with the truncated SP of P22 and P22 CP, provides a novel method to rescue proteins and allow their further study. We have examined the encapsulation of two different proteins, a novel α-galactosidase (GalA) from the hyperthermophilic archaeon, Pyrococcus furiosus, and the hemagglutinin head (HA_head) from influenza, which both intractably form inclusion bodies when recombinantly expressed in E. coli. GalA is a highly thermostable enzyme that catalyzes the hydrolysis of α-1,6-linked galactose saccharides and can also be used to create complex sugars from monosaccharide building blocks, of interest for applications in biofuels production and utilization.¹⁴ However, exploration of this enzyme has been limited because GalA is produced as an inclusion body from which only about 10–15% of total activity can be extracted into a soluble form.¹⁴ In addition to this enzyme, we investigate an antigenic protein from influenza, the hemagglutinin head (HA_head), which is of broad interest for the development of flu vaccines. HA_head has been expressed in high yield in E. coli where it forms an inclusion body and has been successfully solubilized and refolded using denaturant.¹⁵ A means to rapidly produce HA_head would be beneficial for gaining materials for studying these antigens as agents for producing influenza vaccines.

The genes encoding GalA and HA_head were individually subcloned into the previously described pETDuet assembler vector⁵ for co-expression of either GalA-SP or HA_head-SP fusion proteins and CP, and subsequently expressed in E. coli (BL21(DE3)) and purified. Initial (unoptimized) yields of approximately 100 mg of the different Cargo-P22 VLPs (GalA-P22 or HA_head-P22, utilized to denote non-covalent association of the GalA-SP or HA_head-SP to CP to produce the P22 VLP complex as shown in Figure 1) per liter of media were observed. Further purification by size exclusion chromatography (SEC) was performed on all samples and it was observed that GalA-P22 and HA_head-P22 eluted with the same retention time as wtP22 VLPs. SDS-PAGE confirmed the co-purification of the GalA-SP (57.6 kDa) and HA_head-SP (44.2 kDa) fusion proteins together with CP (46.7 kDa) (Figure 2) and when imaged by TEM the VLP particles (54.2nm ± 1.97nm and 55.4nm ± 2.35 nm, respectively) appeared to be densely packed with the same morphology as wt P22 VLP. The purification route for our P22 encapsulated proteins is particularly unique since it requires only cell lysis, low speed centrifugation to remove insoluble cellular components, and lastly ultracentrifugation of the supernatant through a sucrose cushion to obtain the target protein within the P22 VLP in high purity. An additional SEC step provides samples that are nearly devoid of contaminating proteins and aberrant VLP assemblies..

Figure 2.

Characterization of the GalA-P22 and HA_head-P22 particles by SDS-PAGE (A), TEM with 200 nm scale bars (B), HPLC-MALS (C), and kinetics of the encapsulated GalA enzyme (D).

Purified samples of GalA-P22 and HA_head-P22 were further characterized by HPLC-SEC monitored by multiangle light scattering (MALS), quasi elastic light scattering and refractive index detectors. Results from SEC-MALS yielded average molar masses of 30.13 ± 0.29 and 32.38 ± 0.31 MDa for GalA-P22 and HA_head-P22, respectively, corresponding to 183 ± 5.0 GalA and 286 ± 7.0 HA_head proteins per capsid, after subtracting the mass of the capsid (19.7 MDa) from the observed masses. Radius of hydration (R_h) values for each construct were consistent with values expected for the procapsid P22 (29 nm), with observed values of 28.01 ± 0.55 and 28.67 ± 0.38 for GalA-P22 and HA_head-P22, respectively. The radius of gyration (R_g) for each construct, which is dependent on the center of mass of the particle, showed values of 24.2 ± 0.55 and 23.23 ± 0.06 nm for GalA-P22 and HA_head-P22, respectively. The ratio of R_g/R_h for all cases was ~0.86, indicative of densely packed spherical particles, as opposed to the thin shelled hollow sphere value of 1 observed for the empty P22. We also examined the ability of the P22 VLP constructs to undergo the transition to alternate P22 morphologies, expanded shell ¹⁶ and wiffleball (WB), by heating. All constructs were found to undergo such changes as observed by agarose gel shift assays and SEC-MALS/RI, with expected radii and mass changes observed by SEC-MALS (Figure S1 and S2). We have encapsulated protein cargo-SP fusions of up to ~180 kDa in size (unpublished data), allowing rescue of much larger proteins than described here. In addition, native proteins without SP can be obtained by engineering a protease cleavage site between the two components, shown previously for the separation of GFP-SP inside P22 by thrombin.⁴

Having determined that proteins can be successfully rescued by encapsulation inside the P22 VLP, we examined the ability to maintain proper folding. For this we focused on examining the activity of the GalA-P22 by performing kinetic assays on both the PC and WB (to assay effects of substrate accessibility and higher temperature on kinetics) morphologies of the construct to see if the packaged enzyme was active, which would imply proper folding and structure of the protein. Assays were performed at 60°C and 75°C (only WB at 75°C to prevent expansion) in a PBS buffer pH 7.0 using the substrate 4-nitrophenyl-α-D-galactopyranoside (pNp-αGal). Previous studies using P22 to encapsulate enzymes have shown that diffusion of substrates is not inhibited by the protein cage wall.^{5, 6}

Experimental kinetic parameters for GalA-P22 constructs are summarized in Table 1 and show that the encapsulation produces highly active GalA enzyme (GalA-SP is discussed as GalA). Previously, a turnover rate (k_cat) of 610.1sec⁻¹ for GalA at 90 °C was reported¹⁴, 15 °C higher than the highest temperature assessed here. Since the P22 capsid is not stable at 90°C we could not make a direct comparison but our observed rate for GalA was nearly one third greater than observed previously at the higher temperature. These findings strongly suggest that encapsulation either enhances the rate of GalA or, more likely, more correctly folded and active enzymes are produced by encapsulation. Interestingly, the tightly confined and crowded environment within P22, with an internal GalA concentration of 300 g/L, closely mimics that of native cellular conditions (500 g/L) and may also effect the overall activity. In comparison with other thermophilic enzymes, the reduction in turnover due to the lower temperature is in good agreement with previous observations.^{14, 17–20} Little change was observed between binding affinities (K_M) of GalA in different P22 morphological states at different temperatures, consistent with values found in literature and following the trend observed for other glycosidases.^{14, 17–20}

Table 1.

Kinetic parameters determined for P22-GalA in different morphological states at different temperatures. Gal-A from inclusion body data is provided from reference 14 for comparison.

Construct	Temperature (°C)	Kcat(sec−1)	KM(mM)
PC GalA-P22	60	361.9±16	0.22±0.04
WB GalA-P22	60	341.4±18	0.22±0.05
WB GalA-P22	75	860.1±32	0.26±0.07
GalA (from inclusion body)	90	610.1	0.25

Although the P22 VLP is not a barrier for substrate access to enzymes on the interior and P22 encapsulating antigenic proteins have been demonstrated to be extremely effective at producing a stimulative immune response⁷, the ability to release the cargo proteins from the protein cage to obtain “free” protein would be ideal for more extensive applications. We examined the ability to disassemble P22 containing GalA and HA_head by lowering the solution pH to 3 and 4, which has been shown to disassemble VLPs.²¹ Lowering the pH to 3 caused massive precipitation of all protein and we were unable to obtain any protein from the soluble fraction. At pH 4, there was no precipitation, and SEC showed promising separation of peaks consistent with assembled capsid and disassembled CP with GalA protein subunits. However, these fractions showed no GalA activity. These results are not surprising, since harsh conditions are needed to disrupt the cage structure that functions naturally to protect viral genome from the harsh environments it encounters. Future investigations will focus on developing a milder programmed P22 VLP, allowing internal protein cargoes to be easily removed.

Taking inspiration from the complex hierarchical assemblies found in biology we have designed and constructed a self-assembling protein based recovery vessel for in vivo sequestration of inclusion body forming proteins. The P22 VLP system was successfully exploited to encapsulate both enzyme and antigenic proteins that normally are expressed as inclusion bodies. Future work in improving the ability to remove and recover internalized proteins are needed and could provide VLPs capable of stimuli responsive disassembly, which has significant implications for not only obtaining difficult protein targets from E. coli, but also for cellular delivery of therapeutics.