• Open Access

Tomato cystatin SlCYS8 as a stabilizing fusion partner for human serpin expression in plants

Authors


Summary

Studies have reported the usefulness of fusion proteins to bolster recombinant protein yields in plants. Here, we assess the potential of tomato SlCYS8, a Cys protease inhibitor of the cystatin protein superfamily, as a stabilizing fusion partner for human alpha-1-antichymotrypsin (α1ACT) targeted to the plant cell secretory pathway. Using the model expression platform Nicotiana benthamiana, we show that the cystatin imparts a strong stabilizing effect when expressed as a translational fusion with α1ACT, allowing impressive accumulation yields of over 2 mg/g of fresh weight tissue for the human serpin, a 25-fold improvement on the yield of α1ACT expressed alone. Natural and synthetic peptide linkers inserted between SlCYS8 and α1ACT have differential effects on protease inhibitory potency of the two protein partners in vitro. They also have a differential impact on the yield of α1ACT, dependent on the extent to which the hybrid protein may remain intact in the plant cell environment. The stabilizing effect of SlCYS8 does not involve Cys protease inhibition and can be partly reproduced in the cytosol, where peptide linkers are less susceptible to degradation. The effect of SlCYS8 on α1ACT yields could be explained by: (i) an improved translation of the human protein coding sequence; and/or (ii) an overall stabilization of its tertiary structure preventing proteolytic degradation and/or polymerization. These findings suggest the potential of plant cystatins as stabilizing fusion partners for recombinant proteins in plant systems. They also underline the need for an empirical assessment of peptide linker functions in plant cell environments.

Introduction

A number of recent successes in the industrial-scale production of pharmaceutical proteins in plants (D'Aoust et al., 2010; Lai and Chen, 2012; Pogue et al., 2010) can be attributed to the development of high-yielding expression systems for the transient transformation of Nicotiana spp. leaf tissue (Marillonnet et al., 2005; Pogue et al., 2002; Sainsbury and Lomonossoff, 2008). An additional factor determining final yields is the overall stability of the recombinant polypeptide in a given cellular context. Proteolytic degradation is often cited as a yield-limiting problem in plant-based expression systems (Benchabane et al., 2008), and a number of studies have shown that co-expression of protease inhibitors can improve protein yields by reducing the impact of endogenous proteolysis (Goulet et al., 2012; Komarnytsky et al., 2006; Pillay et al., 2012; Rivard et al., 2006). While plant cells are able to process, fold and assemble complex mammalian proteins, it is also likely that these processes have a considerable influence over expression levels. For example, the introduction or modification of post-translational modification sites can influence the stability and accumulation of secreted recombinant proteins in plants and plant cells (Lombardi et al., 2012; Meyers et al., 2008; Xu and Kieliszewski, 2011).

Similarly to prokaryotic expression systems, strategies to increase the stability of recombinant proteins in plants include the use of fusion proteins or peptides. Indeed, the first attempt to express a clinically relevant protein was as a hybrid transcript (Barta et al., 1986), and since then the accumulation of recombinant proteins in plants has been improved by fusion to a diverse range of proteins including ubiquitin (Hondred et al., 1999), cholera toxin B subunit (Kim et al., 2004) and human immunoglobulin α-chains (Obregon et al., 2006). Recent examples involving hydrophobin (Joensuu et al., 2010) or elastin-like polypeptides (Conley et al., 2009; Floss et al., 2008; Patel et al., 2007) also assist purification of the protein by stimulating the assembly of aggregates or inclusion bodies in leaf tissue (Conley et al., 2011). A key consideration in the design of fusion proteins is the choice of a proper peptide linker sequence. In general, adding a peptide linker to separate two functional domains is thought to have a positive impact on the function of each domain, but empirical assessment of a variety of linkers within a given expression context always remains ideal (Sainsbury et al., 2012a; Wriggers et al., 2005). The choice and number of amino acids used are critical, as the linker itself may have a substantial impact on the performance of hybrid partners (Lu and Feng, 2008; Zhao et al., 2008); the yield of hybrid protein (Sauer et al., 2001); and the stability of the fusion, in terms of susceptibility to proteolysis (Kavoosi et al., 2007; Wu et al., 2009; Zhao et al., 2008), thermal stability (Kavoosi et al., 2007; Sauer et al., 2001) and tendency to aggregate (Zhao et al., 2008).

Our goal in this study was to assess the potential of fusion proteins and peptide linkers for the production of human alpha-1-antichymotrypsin (α1ACT) in plants. α1ACT is a member of the serpin protein superfamily, which includes Ser protease inhibitors involved in a wide range of biological processes (Silverman et al., 2010). Because deficiencies in certain serpins, including α1ACT, may result in human disease (Law et al., 2006), recombinant serpins are attractive candidates for replacement therapy. Like most serpins, α1ACT is biologically active in a stressed metastable state, which converts to a relaxed state upon interaction with target proteases. This interaction results in cleavage of the reactive centre loop, which in turn results in a significant conformational change (Baumann et al., 1991) and covalent linkage, making the target protease and the inhibitor inactive (Law et al., 2006). The intrinsic conformational mobility of serpins could partly explain their relative instability in plants (Benchabane et al., 2009a; Huang et al., 2009). It also suggests that recombinant serpin production in plants and plant cells might be improved by the use of a stabilizing fusion partner, which could help stabilize their structure and eventually prevent their degradation by host plant endogenous proteases along the cell secretory pathway (Benchabane et al., 2009a; Huang et al., 2009). Here, we assessed the potential of tomato cystatin SlCYS8 (Girard et al., 2007), a Cys protease inhibitor of the cystatin protein superfamily (Benchabane et al., 2010), as a stabilizing fusion partner for the protection and increased accumulation of α1ACT in leaves of Nicotiana benthamiana. We show a strong stabilizing impact of SlCYS8 on α1ACT, influenced by the peptide linker inserted between the two protein partners.

Results and discussion

Co-expression of α1ACT and SlCYS8 in N. benthamiana leaves

Given that α1ACT is unstable when expressed in plant tissues (Benchabane et al., 2009a), we attempted to increase the level of expression through two means. First, we employed the highly efficient transient expression system CPMV-HT (Sainsbury and Lomonossoff, 2008); second, we co-expressed a ‘companion’ protease inhibitor along with α1ACT, with the aim to protect it from degradation by Cys proteases while in transit along the cell secretory pathway. For this, we chose the eighth domain of tomato multicystatin, SlCYS8, which has been shown to be an effective competitive inhibitor against plant papain-like Cys proteases of the Solanaceae (Goulet et al., 2008). Cys protease activity is high in both whole cell and apoplast extracts of N. benthamiana, and we have recently shown that transient expression of cystatins can reduce these activities and even result in increased yields of co-expressed recombinant proteins (Goulet et al., 2012; Robert et al., 2013). Expression constructs for α1ACT and SlCYS8 targeted to the secretory pathway were assembled into pEAQ expression vectors (Figure 1) where coding sequences are inserted between CPMV-HT enhancers found between the CaMV 35S promoter and nopaline synthase terminator (Sainsbury et al., 2009). These constructs were co-infiltrated in N. benthamiana leaves along with a protein P19-encoding construct to reduce transgene silencing (Voinnet et al., 2003). Infiltration with either construct resulted in detectable expression of recombinant protein (Figure 2), estimated in both cases at approximately 1% of total extractable proteins (TEP) by densitometric analysis of immunoblots using standard amounts of each protein. SlCYS8 (10.7 kDa) is about one quarter the size of α1ACT (45 kDa) and was, therefore, more abundant than the human protein on a molar basis. In contrast to the single band detected for the nonglycosylated SlCYS8, α1ACT was detected as multiple bands (Figure 2), presumably due to both proteolytic trimming and variable glycan-processing in the ER and the Golgi (Benchabane et al., 2009a).

Figure 1.

Constructs for SlCYS8, human α1ACT and SlCYS8-α1ACT hybrid inhibitors. The hybrids were designed so that the regions important for interaction with proteases, the N-terminus of SlCYS8 and the C-terminus of α1ACT, were at opposite ends of the molecule. (a) Diagram of individual inhibitors and representative structure of a hybrid protease inhibitor for expression with pEAQselectK-based constructs for expression in Nicotiana benthamiana, or with pGEX-3X-based constructs for expression in Escherichia coli as glutathione-S-transferase (GST) fusions. Black box, factor Xa cleavage site for release from GST; black line, CPMV-HT UTRs; PDI, signal peptide from Medicago sativa protein disulphide isomerase; L, linker peptide. (b) Diagram of hybrid constructs for E. coli and N. benthamiana expression, with linker sequences and corresponding properties given were appropriate. t-α1ACT, N-terminally truncated α1ACT (see text for details), with native signal peptide removed. full-α1ACT, full-length (wild-type) α1ACT, with native signal peptide removed.

Figure 2.

Human α1ACT expression in the secretory pathway of Nicotiana benthamiana leaves. pEAQ-based constructs were agroinfiltrated into N. benthamiana leaves alongside pEAQexpress for protein P19. Soluble protein extracts were submitted to 12% (w/v) SDS-PAGE and then electrotransferred to nitrocellulose for immunoblotting. The membranes were probed with anti-SlCYS8 or anti-α1ACT primary antibodies, followed by incubation with alkaline phosphatase-conjugated anti-rabbit secondary antibodies. The position of molecular size markers is given on the left. EV stands for empty vector with pEAQexpress, α1ACT + SlCYS8 for co-expression of free α1ACT and SlCYS8 from a single plasmid. The arrowhead points to the position of intact hybrids, the star to free α1ACT and the circle to free SlCYS8.

SlCYS8 was co-expressed with α1ACT to prevent its degradation by leaf Cys proteases, but this approach surprisingly resulted in a marked decrease in α1ACT (Figure 2). Considering that both proteins were expressed from a single T-DNA with the same regulatory sequences, this result suggested the SlCYS8 construct to be favoured over α1ACT during pre- and/or post-translational processes. In an attempt to overcome this apparent competitive effect, we fused the two proteins to create a hybrid protein, Hybrid-A (Figure 1). Because the C-terminus of α1ACT and the N-terminus of SlCYS8 are important for the activity of each protein, SlCYS8 was positioned in N-terminal position, upstream of α1ACT. Fusing the two proteins resulted in a considerable restoration of α1ACT accumulation, clearly overcoming the detrimental effect of SlCYS8 co-expression as a separate protein (Figure 2). The SlCYS8-α1ACT fusion was, however, nearly completely cleaved to release the two domains. This is consistent with the previously reported proteolytic susceptibility of the α1ACT N-terminal region in plant cells (Benchabane et al., 2009a).

To prevent separation of the two domains, we constructed a second SlCYS8-α1ACT hybrid, HybridΔ23, lacking 23-amino acids at the N-terminus of mature α1ACT (Figure 1). This region of the protein, not detected by X-ray crystallography and presumably flexible, is considered not essential for α1ACT activity (Baumann et al., 1991). Expression of HybridΔ23 also resulted in restored, even increased levels of α1ACT as detected by immunoblotting (Figure 2), and as for α1ACT expressed alone the hybrid was detected as multiple bands due to variable glycosylation and proteolytic trimming. Rather than protection from proteolysis, SlCYS8 co-expression thus resulted in what appeared to be a significant competition effect. As both the cleaved (Hybrid-A) and stable (HybridΔ23) hybrids restored α1ACT accumulation, we hypothesized that there may have been improved translation of the hybrids imparted by the more efficiently expressed SlCYS8. However, the slightly better α1ACT increase with HybridΔ23 compared with the nearly totally cleaved Hybrid-A suggests that there was also an impact of the cystatin on the overall stability of α1ACT.

Activities of α1ACT, SlCYS8 and SlCYS8-α1ACT hybrids

In general, separation of proteins by a linker peptide improves the activity of the hybrid partners. To find a stable linker peptide and to further investigate the possible impact of a general stabilizing effect while preserving α1ACT activity, we assembled a number of SlCYS8-α1ACT hybrids with the inhibitor domains separated by various linker peptides (Figure 1) such that the linkers replaced the presumably flexible α1ACT N-terminus that was seen to be susceptible to cleavage in planta (Figure 2). Due to the impact that a given linker sequence may have in certain contexts, empirical assessment of various linker peptides is usually preferable (Sainsbury et al., 2012a). In this study, we chose to separate SlCYS8 and α1ACT with classic synthetic flexible (Wriggers et al., 2005) and rigid (Arai et al., 2001) linkers or with a naturally occurring peptide of the multifunctional Dengue virus (DENV) NS3 protein. We refer to this linker as twistable, as it is able to rotate the protease domain of NS3 by approximately 161° with respect to the helicase domain (Luo et al., 2010).

Because serpin activity depends on a significant conformational change, we first quantified the impact of fusing α1ACT and SlCYS8 on the activity of each domain, using free and hybrid inhibitors produced in Escherichia coli. Individual inhibitors, the hybrid with no linker (HybridΔ23) and hybrids with domains separated by the α1ACT N-terminus (Hybrid-A) linker, the flexible linker (Hybrid-F), the rigid linker (Hybrid-R) or the twistable linker (Hybrid-T) were all successfully purified from E. coli as glutathione-S-transferase (GST) fusions (Figure 1 and Figure S1). Following removal of the GST moiety and confirmation of the presence of constituent domains (Figure S1), SlCYS8 and α1ACT activities were measured by protease inhibition fluorimetric assays. Anti-chymotrypsin activity of α1ACT expressed alone was similar to the corresponding activity measured previously for the bacterially expressed protein (Benchabane et al., 2009b), in contrast with SlCYS8 showing no such activity. As expected, adding purified SlCYS8 to the reaction mixture had no impact on α1ACT antichymotrypsin activity thereby permitting to clearly assess the impact of SlCYS8 fused to α1ACT (Figure 3a). As the hybrid proteins were composed of a truncated version of α1ACT (Figure 1), we also confirmed that the truncated version expressed alone possessed the same level of antichymotrypsin activity as full-length α1ACT (not shown). By contrast, the peptide linkers were shown to have a considerable impact on the function of both SlCYS8 and α1ACT (Figure 3). Direct grafting of SlCYS8 to α1ACT with no synthetic or natural linker, such as in HybridΔ23, had a substantial negative impact on α1ACT activity. By comparison, Hybrid-A displayed no difference in activity compared with α1ACT alone, suggesting a utility for the α1ACT N-terminus as a peptide linker in vitro. The twistable linker of Hybrid-T also allowed a similar level of antichymotrypsin activity, unlike the flexible and rigid synthetic linkers, which both carried a penalty for the fusion protein (Figure 3a). Differences were statistically significant between hybrids with Hybrid-A and Hybrid-T more potent than both HybridΔ23 and Hybrid-F (P < 0.05) in their antichymotrypsin activity.

Figure 3.

In vitro protease inhibitory activities of SlCYS8, human α1ACT and SlCYS8-α1ACT hybrid inhibitors. Purified individual and hybrid inhibitors were used to inhibit synthetic fluorigenic substrates hydrolysis by papaya latex papain or bovine pancreatic chymotrypsin. (a) Chymotrypsin inhibition. (b) Papain inhibition. Values are the mean of three biological replicates ± SE. The gel insert in (b) shows papain interacting with SlCYS8 or α1ACT, as assayed by gelatin/SDS-PAGE. 1, papain alone (noninhibited); 2, papain + SlCYS8; 3, papain + α1ACT.

Assessment of SlCYS8 antipapain activity was complicated by the fact that α1ACT appeared to inhibit both chymotrypsin and papain (Figure 3). The apparent interaction between α1ACT and papain, first observed during fluorimetric assays, was confirmed by a gelatin-PAGE assay, where pre-incubation of papain with α1ACT before electrophoresis prevented migration of the test protease under mildly denaturing conditions (Figure 3b). The formation of a (covalent) complex with papain was also detected by immunoblotting following SDS-polyacrylamide gel electrophoresis (SDS-PAGE) under reducing conditions (not shown). Cross-class inhibition of Cys proteases has been demonstrated earlier for a number of serpins (Schick et al., 1998). The inhibitory process depends on the reactive centre loop, suggesting that Cys protease–serpin interactions are analogous to the Ser protease–serpin interaction (Irving et al., 2002). Given the possibility of two modes of papain inhibition in our reaction mixtures, we have estimated antipapain activity as a percentage of inhibition. In brief, free α1ACT and SlCYS8 had an additive effect against papain in vitro, giving inhibition percentages comparable overall to the SlCYS8-α1ACT hybrids. Interestingly, the linkers also had a differential impact on total papain inhibition, indicating that they imposed different constraints on the SlCYS8 and α1ACT hybrid domains. One hybrid, Hybrid-A, even outperformed the additive effect of the two purified inhibitors and was significantly more active than HybridΔ23 and Hybrid-R (< 0.05).

Taken together, these results demonstrate a significant impact of the SlCYS8 fusion partner on α1ACT inhibitory activity. The negative impact of directly fusing the functional domain of α1ACT could not be overcome by the insertion of either a flexible or a rigid synthetic linker, emphasizing the need for an empirical assessment of linker peptides. In the present case, using the native α1ACT N-terminus as a potentially flexible linker or the noncognate DENV NS3 linker peptide restored α1ACT activity. This good performance of the twistable linker, about the same length as the other linkers tested (Figure 1), provides a first demonstration of its potential in creating hybrid proteins. Its unique properties may provide a means of retaining activity of fusion partners that require significant conformational changes for activity, by allowing a re-orientation of the functional domains.

Hybrids expression in N. benthamiana

As the different linkers had a differential impact on α1ACT activity, we expressed all hybrids in N. benthamiana to determine whether they also had an impact on their accumulation in planta. The hybrids were assembled into the pEAQ expression vector and again targeted to the secretory pathway. Coomassie blue staining of leaf soluble proteins following SDS-PAGE showed the hybrids to accumulate extremely well in infiltrated leaves and contribute a significant portion of the extracted proteins (Figure 4). Immunoblotting confirmed that both α1ACT and SlCYS8 were expressed from all hybrid constructs, but showed considerable differences between hybrids in the extent to which the two domains were separated due to cleavage of linkers during their migration through the cell secretory pathway (Figure 4). As expected, Hybrid-A was almost completely cleaved (see Figure 2). Hybrid-R and Hybrid-T also underwent considerable cleavage, as shown by the detection of free SlCYS8 and the detection of α1ACT mostly at 45 kDa. The stability of linker peptides in heterologous environments is known to be context-dependent, as illustrated, for example, by a variant of the rigid linker used here, which is susceptible to cleavage in yeast (Zhao et al., 2008) but not in E. coli (Lu and Feng, 2008), and susceptible to auto-proteolytic cleavage under certain pH conditions (Wu et al., 2009). Flexible Gly-rich linkers have been shown to perform well in plant biotechnology with respect to their stability in planta (Peschen et al., 2004; Werner et al., 2006). Accordingly, the Gly-containing linker remained intact in the present study. Although the twistable linker allowed for a good amount of α1ACT activity in vitro, it appeared to be unsuitable for the in planta generation of a stable fusion protein.

Figure 4.

Hybrid inhibitor expression in the secretory pathway of Nicotiana benthamiana leaves. pEAQ-based constructs were agroinfiltrated into N. benthamiana leaves alongside pEAQexpress to provide protein P19. Leaf soluble proteins resolved by 12% (w/v) SDS-PAGE were stained with Commassie blue or electrotransferred to nitrocellulose sheets for immunoblotting. The membranes were probed as described in Figure 2. The position of molecular size markers is given on the left. EV stands for empty vector with pEAQexpress, α1ACT + SlCYS8 indicates co-expression from a single plasmid. The arrowhead points to the position of intact hybrids, the star to free α1ACT, and the circle to free SlCYS8.

An ELISA was performed to quantify the expression of α1ACT alone, co-expressed with SlCYS8 or expressed in fusion with the plant cystatin (Figure 5). α1ACT accumulated at 115 mg/kg of fresh weight tissue (FWT; i.e. approximately 1% TEP) when expressed alone, compared with a reduced amount of 24 mg/kg when co-expressed with SlCYS8. Fusion to SlCYS8 increased the amount of α1ACT to a remarkable 2354 mg/kg (or approximately 20% TEP) for HybridΔ23, a 20-fold increase over the expression of α1ACT alone. A comparable yield was achieved with Hybrid-F, also visualized as a stable product following SDS-PAGE and immunoblotting (Figure 4). By comparison, hybrids undergoing proteolytic processing showed lower α1ACT rates, estimated at 285 mg/kg FWT for Hybrid-A, 539 mg/kg for Hybrid-R and 977 mg/kg for Hybrid-T. These data are in contrast to the immunoblots indicating roughly equal accumulation of α1ACT between the hybrids and suggest that differences in α1ACT accumulation expressed as hybrids reflect the extent to which the domain partners are separated due to linker cleavage along the cell secretory system. Polymerization of some serpins can be induced by low pH, resulting from a loss of secondary structure (Devlin et al., 2002). Therefore, perhaps polymerized α1ACT is lost during ELISA plate binding under native conditions, in contrast to the strongly denaturing conditions of SDS-PAGE. This would indicate that in the low-pH environment of the Golgi or the apoplast, the positive influence of SlCYS8 fusion on α1ACT accumulation might result from the prevention of polymerization of the serpin. The expression levels achieved with HybridΔ23 and Hybrid-F, which are similar to the highest levels achieved for stable proteins such as green fluorescent protein (GFP; Lindbo, 2007; Marillonnet et al., 2005; Sainsbury et al., 2009), represent an astonishing improvement for a recombinant protein previously considered as poorly expressed in the plant cell secretory pathway (Benchabane et al., 2009a).

Figure 5.

Quantitative analysis of α1ACT expressed alone or in fusion with SlCYS8 in the secretory pathway of Nicotiana benthamiana. pEAQ-based constructs were agroinfiltrated alongside pEAQexpress to provide protein P19. Soluble protein extracts were used on a volume basis and diluted for use in the ELISA such that each well was coated with protein from 0.333 mg of fresh weight tissue (FWT). A standard curve was generated from Escherichia coli-produced α1ACT diluted in control extract. Values are the mean of three biological replicates ± SD. α1ACT + SlCYS8 indicates co-expression from a single plasmid.

The stabilizing effect of SlCYS8 does not involve Cys protease inhibition

To confirm, as suggested by the co-expression of α1ACT and SlCYS8, that the stabilizing influence of SlCYS8 on α1ACT was not due to anti-Cys protease activity, we assembled a version of Hybrid-F, Hybrid-Fm, including a point mutation in the first inhibitory loop of SlCYS8 changing the residue Gln-47 (Q47) for a proline (P). This mutation disrupting the conserved motif QxVxG (where x is any amino acid) is known to make cystatins inactive against papain, without significantly affecting their tertiary structure (Arai et al., 1991; Sainsbury et al., 2012b). Accordingly, Q47P used at the IC50 of SlCYS8 inhibited papain-like Cys proteases from leaf extracts by only 5%. When provided in excess, the same mutant did not give more that 10% inhibition, in sharp contrast with SlCYS8 giving close to 90% inhibition (Figure 6a).

Figure 6.

The α1ACT-stabilizing effect of SlCYS8 is protease inhibition-independent. A variant of Hybrid-F, Hybrid-Fm, was created using a mutated, inactive variant of SlCYS8, Q47P, to rule out the possible impact of Cys protease inhibition in the stabilization of α1ACT by SlCYS8. (a) Inhibition curves for wild-type SlCYS8 (closed circles) and the Q47P mutant (open circles) against cathepsin L-like Cys proteases of Nicotiana benthamiana leaf tissue. Infiltration of pEAQ-based constructs for targeting to the secretory pathway in N. benthamiana leaves was followed by extraction of soluble proteins diluted for use in the ELISA (b). Samples and standards were prepared for ELISA as stated in Figure 5. (c) Immunoblotting with anti-α1ACT antibodies. The position of molecular size markers is given on the left. EV stands for empty vector with pEAQexpress; α1ACT + SlCYS8 indicates co-expression from a single plasmid.

Despite this drastically impaired efficiency of the mutant cystatin, infiltrating leaves with the Hybrid-Fm construct resulted in the same increase in α1ACT accumulation as with Hybrid-F (Figure 6b). This finding provides compelling evidence that the stabilizing effect of SlCYS8 was not due to its antiproteolytic activity. However, immunoblotting for α1ACT revealed some cleavage of the flexible linker with Hybrid-Fm, suggesting that wild-type SlCYS8 inhibited a linker-specific protease to keep Hybrid-F intact (Figure 6c). Interestingly, flexible linker cleavage in Hybrid-Fm was not accompanied by a reduced α1ACT ELISA yield as observed with the other linkers. We speculate that cleavage of the flexible linker occurred subsequent to the major stabilizing effect of SlCYS8, whereas the other linkers were cleaved while the presence of fused SlCYS8 was still critical in preventing α1ACT polymerization. This would mean that the apoplast is less acidic than late secretory organelles, which appears to be the case (Gao et al., 2004), although the baseline pH is variable due to low buffering capacity (Gjetting et al., 2012). This result also presents the intriguing possibility that nonfunctional SlCYS8 could be used to enhance stability in a therapeutic context where the inhibitory activity would otherwise preclude its use.

SlCYS8 imparts a generic stabilizing effect on α1ACT accumulation

We targeted α1ACT and the hybrid proteins to the cytosol of N. benthamiana leaf cells (Figure 1) to further investigate the stabilizing effect of SlCYS8 on α1ACT. Although α1ACT is a glycoprotein, it is fully functional when expressed in bacteria and is, therefore, presumably folded into its proper metastable active state without the help of ER-resident chaperones (Benchabane et al., 2009b). Following infiltration of N. benthamiana leaves with pEAQ-based hybrid constructs designed for cytosolic expression, Coomassie blue staining of leaf proteins following SDS-PAGE revealed high levels of recombinant protein (Figure 7a). Immunoblots showed that in contrast to their secreted counterparts, α1ACT and the hybrids accumulated mostly as single species around the expected size, in line with the absence of glycosylation and limited proteolytic trimming of α1ACT in the cytosol (Benchabane et al., 2009b). A protein fragment detected below α1ACT for all hybrids suggests however that there was a small amount of activity-related cleavage of the reactive centre loop. Hybrid-A, Hybrid-R and Hybrid-T were also cleaved within the linker, to separate SlCYS8 from α1ACT. This resulted in an additional α1ACT-containing fragment on immunoblots compared with HybridΔ23 and Hybrid-F (Figure 7b) and in the detection of free SlCYS8 (not shown), showing that the flexible linker was the most stable not only in the secretory pathway but also in the cytosol.

Figure 7.

Cytosolic expression of human α1ACT alone or as a fusion with SlCYS8. Expression constructs for expression of human α1ACT, SlCYS8 and SlCYS8-α1ACT hybrids were generated for expression in the cell cytosol. Leaf soluble proteins resolved by 12% (w/v) SDS-PAGE were stained with Commassie blue (a) or transferred to nitrocellulose for immunoblotting with anti-α1ACT antibodies (b). The position of molecular size markers is given on the left. EV stands for empty vector with pEAQexpress, and α1ACT + SlCYS8 indicates co-expression from separate plasmids. Samples and standards for ELISA (c) were prepared as stated in Figure 5.

α1ACT expressed alone accumulated better in the cytosol than when secreted to reach 303 mg/kg FWT (Figure 7c), in contrast with other transiently expressed secreted proteins of human origin, such as epidermal growth factor (Wirth et al., 2004) and growth hormone (Gils et al., 2005). There was good agreement between the denaturing SDS-PAGE results and the ELISA, which could be due to a lower tendency for α1ACT to polymerize in the cytosol where the pH is slightly alkaline. This subcellular compartment, however, is not ideal for the expression of therapeutic serpins, as unglycosylated forms are rapidly cleared from the blood (Kwon et al., 1995) and exhibit a low thermal stability (Kwon and Yu, 1997). The impact of fusing α1ACT to SlCYS8 appeared to be much smaller in the cytosol than in the secretory pathway (Figure 7c). Again, the more stable Hybrid-F resulted in the best improvement of α1ACT accumulation, with 717 mg/kg FWT, indicating that post-translational structural integrity of the fusion was important as well in the cytosol. The fact that improved yields in α1ACT were measured for hybrids expressed in the cytosol and that co-expression with free SlCYS8 resulted in a 53% decrease in α1ACT accumulation, which was again overcome by its fusion to the cystatin, again suggests a possible positive influence on mRNA translation.

To determine whether the increase in accumulation was specific to SlCYS8, we fused α1ACT to the GFP, a marker protein known to accumulate very well in the plant cell cytosol. A GFP-α1ACT hybrid was assembled with the flexible linker, and following infiltration into N. benthamiana leaves, both GFP and α1ACT were detectable at the expected size on immunoblots (Figure 8a,b). The GFP hybrid led to a similar increase in α1ACT accumulation as Hybrid-F in the cytosol (Figure 8c), although a small proportion was apparently cleaved. A secreted version of the GFP hybrid was also constructed. This hybrid only led to a modest increase in α1ACT accumulation, showing the well documented instability of secreted GFP to result in reduced stabilization of α1ACT during transit through the cell secretory pathway. Indeed, only free α1ACT could be detected on the immunoblots, in contrast to fully intact Hybrid-F (Figure 8b). Co-expression of GFP alongside α1ACT resulted in a slight decrease in accumulation of the serpin in both the cytosol and the apoplast, further suggesting an impact on competition for protein translational resources.

Figure 8.

Expression of α1ACT as a fusion to green fluorescent protein (GFP). pEAQ-based constructs for GFP and GFP-hybrid expression in the cytosol or targeted to the cell secretory pathway were agroinfiltrated into Nicotiana benthamiana leaves alongside pEAQexpress to provide protein P19. Leaf soluble proteins resolved by 12% (w/v) SDS-PAGE were transferred to nitrocellulose for immunoblotting with anti-GFP (a) or anti- α1ACT antibodies (b). The position of molecular size markers is given on the left. EV stands for empty vector with pEAQexpress, and α1ACT + SlCYS8 indicates co-expression from separate plasmids. Samples and standards for ELISA (c) were prepared as stated in Figure 5.

Conclusions

Here, we assessed the potential of tomato SlCYS8 as a stabilizing fusion partner for human α1ACT targeted to the plant cell secretory pathway. Contrasting with recent studies reporting the potential of protease inhibitors as ‘companion’ proteins to improve the expression of recombinant proteins in Solanaceae (Goulet et al., 2012), SlCYS8 co-expressed with α1ACT had a significant repressive effect on α1ACT accumulation in leaves of N. benthamiana. On the other hand, the cystatin showed a strong stabilizing effect when expressed as a translational fusion with the human protein. Yields could be improved by up to 20-fold and reached levels similar to the highest levels reported up to now for recombinant proteins in plants. A number of peptide linkers were tested and shown to influence protease inhibitory potency of the two domains as well as fusion protein accumulation in plant tissue. The latter was dependent on the extent to which the hybrid protein could remain intact during its passage through the secretory pathway. While these results provide a valuable resource for the use of peptide linkers in N. benthamiana, they also highlight the need for an empirical assessment of linker properties. In conclusion, fusion to SlCYS8 confers a significant overall stability to α1ACT that may prevent proteolysis and/or polymerization and allows for a massive yield increase. Interestingly, SlCYS8 alone is expressed to a much higher level in the cytosol than when secreted (unpublished results); yet, the fusion leads to a considerably bigger effect when secreted resulting in a much higher accumulation. While the positive effect of SlCYS8 on α1ACT accumulation could be repeated in the cytosol by fusion to GFP, instability at low pH meant that this fusion partner was less useful in the plant secretory system. Work is underway to determine to what extent the stabilizing effect of SlCYS8 can be applied to the accumulation of other clinically useful recombinant proteins including alpha-1-antitrypsin (Karnaukhova et al., 2006), a structural homologue of α1ACT expressed earlier in stably transformed plant or plant cell lines (Agarwal et al., 2008; Huang et al., 2001, 2010; Jha et al., 2012; McDonald et al., 2005; Nadai et al., 2009; Zhang et al., 2012).

Experimental procedures

Plasmid constructs

Individual and hybrid protease inhibitors were assembled for bacterial and plant-based expression using a modified version of Golden Gate cloning (Engler et al., 2008). Briefly, PCR fragments encoding SlCYS8 (GenBank accession no. AF198390; Girard et al., 2007) and human α1ACT (GenBank accession no. J05176; Rubin et al., 1990) were generated with terminal extensions containing unique recombination sites inside of inverted BsaI recognition sites, oriented to cut inside the fragment leaving the recombination sites overhanging. Fragments were blunt-end ligated into pUC18strep (Sainsbury et al., 2012b) in the presence of SmaI to maintain the specificity of the ligation. Complementary oligonucleotides encoding various linker sequences (Figure 1) with terminal extensions for Golden Gate recombination were annealed and similarly ligated into pUC18strep. For assembly into expression vectors for prokaryotic and plant-based expression, recombination reactions between expression plasmids and donor clones were driven by the simultaneous application of BsaI and T4-DNA ligase to produce GST fusion constructs. Nonrecombined pUC18strep and expression plasmid clones were killed by the presence of ampicillin or the ccdB gene, respectively.

For plant-based expression of the inhibitors, pEAQselectK (Sainsbury et al., 2009) was also modified to contain the ccdB lethal gene flanked by the appropriate recombination sites outside of inverted BsaI recognition sites oriented to cut outside of the ccdB gene and leave the recombination sites overhanging (pEAQ-GG). Recombination reactions between pEAQ-GG and donor clones resulted in CPMV-HT expression constructs (Sainsbury and Lomonossoff, 2008). To achieve this, the ccdB gene was first inserted into the CPMV-HT expression cassette of pM81-FSC5 and pM81-FSC6, allowing recombination into these CPMV-HT cloning plasmids (Sainsbury et al., 2012c). The ccdB-containing cassette from pM81-FSC6 was then transferred to pEAQselectK via SbfI and AscI restriction sites allowing recombination directly into the pEAQ expression vector. To enable transit through the plant secretory pathway, the signal peptide of Medicago sativa protein disulphide isomerase (PDI; Shorrosh and Dixon, 1991) was also amplified with appropriate extensions and inserted into pUC18strep. To co-express SlCYS8 and α1ACT from a single plasmid, a PDI:ACT cassette was assembled in pM81-FSC5 and transferred to pEAQ-PDI:CYS8 via PacI and SbfI restriction sites. The hybrid inhibitors were generated using a donor clone containing the PDI:SlCYS8 fusion sequence amplified from the recombined expression clone, pEAQ-PDI:CYS8. A Q47P mutation changing the residue Gln-47 for a proline in SlCYS8 was created by site-directed mutagenesis of the donor clone by the Quickchange method. For expression in E. coli, the expression vector pGEX-3X (GE Healthcare, Baie d'Urfé, QC, Canada) was modified to contain the ccdB lethal gene flanked by the appropriate recombination sites outside of inverted BsaI recognition sites, to give pGEX-3X-GG (Sainsbury et al., 2012b).

Plant-based expression

pEAQ vectors for expression in planta were maintained in electroporated cells of Agrobacterium tumefaciens, strain AGL1 (Lazo et al., 1991). The bacterial cultures were grown to stable phase in Luria–Bertani medium supplemented with the appropriate antibiotics and pelleted by gentle centrifugation. Following resuspension in infiltration medium [10 mm MES (2-[N-morpholino] ethanesulphonic acid), pH 5.6, containing 10 mm MgCl2] to an OD600 of 0.5, and incubation for 2–4 h at room temperature, suspensions were either pressure infiltrated into N. benthamiana leaves using a needleless syringe. Expression constructs were co-infiltrated with pEAQexpress (Sainsbury et al., 2009) at the same optical density to provide the silencing suppressor protein P19, and leaf tissue was harvested 7 days postinfiltration (dpi), except where indicated otherwise.

Protein extraction and gel electrophoresis

To extract soluble proteins from leaf material, infiltrated leaf tissue was harvested as leaf discs representing 160 mg of control-infiltrated tissue to account for possible treatment effects on leaf mass, and homogenized in three volumes of phosphate-buffered saline (PBS), pH 7.3, containing 5 mm EDTA, 0.05% (v/v) Triton X-100, and complete protease inhibitor cocktail (Roche Diagnostics, Laval, QC, Canada). Leaf lysates were clarified by centrifugation, and protein concentrations were determined using the Bradford assay. Approximately 15 μg of protein extract for Coomassie blue staining, or 8 μg for immunological detection, were separated by SDS-PAGE. Gelatin-PAGE was performed as previously described (Michaud et al., 1996), following incubation of papain with recombinant inhibitors, at the same stoichiometric ratio used in fluorimetric assays (see below).

Western blotting

Following separation by SDS-PAGE, proteins were electroblotted onto nitrocellulose using transfer buffer [20 mm Tris–HCI, 152 mm glycine, 20% (v/v) methanol]. Blocking of nonspecific binding sites was achieved with blocking solution [5% (w/v) skim milk powder in PBS, containing 0.025% (v/v) Tween-20], which also served as antibody dilution buffer. SlCYS8, α1ACT and the inhibitor hybrids were detected with rabbit anti-SlCYS8 polyclonal antibodies (Agrisera, Vännäs, Sweden) or with commercial anti-human α1ACT polyclonal antibodies (US Biological, Swampscott, MA) as primary antibodies, followed by incubation with alkaline phosphatase-conjugated secondary antibodies raised in goat (Sigma-Aldrich, Oakville, ON, Canada). Colorimetric signals were developed by application of the alkaline phosphatase substrate 5-bromo-4-chloro-3-indolyl phosphate/nitro blue tetrazolium (Life Technologies, Burlington, ON, Canada).

Quantitative ELISA

For α1ACT quantitation, Immulon 2HB ELISA plates (ImmunoChemistry Technologies, Bloomington, MN) were coated with duplicate samples of soluble protein extracts diluted 1 : 50 to 27–30 μg/mL in PBS, pH 7.3, for 1 h at 37 °C. Following two washes with PBS, nonspecific binding sites were blocked with 1% (w/v) casein in PBS, for 1 h at 37 °C. The plates were then washed once with PBS and incubated for 1 h at 37 °C with anti-human α1ACT diluted in PBS with 0.25% (w/v) casein. After three washes with PBS, an anti-rabbit antibody conjugated to horseradish peroxidase also diluted in PBS with 0.25% (w/v) casein was applied for 1 h at 37 °C. Washes in PBS were repeated, and the plates were incubated with 3,3′,5,5′-tetramethylbenzidine and SureBlue peroxidase substrate (KPL, Gaithersburg, MD). The reaction was stopped with 1 N HCl, and the absorbance was read at 450 nm. A standard curve was generated for each plate with E. coli-purified α1ACT diluted in control extracts infiltrated with empty pEAQselect and pEAQexpress, to account for possible matrix effects. No difference was noted in the adsorption of α1ACT and hybrids under the neutral conditions used (not shown).

Bacterial expression and purification of recombinant proteins

Expression and purification of SlCYS8 and human α1ACT were carried out as reported (Benchabane et al., 2009a; Goulet et al., 2008), except that E. coli BL21 cells were lysed by 4 freeze-thaw cycles and that a 4-h, rather than a 16-h, induction was used for α1ACT. Purification of the hybrids was as for α1ACT. Protein concentrations were assayed by densitometry of Coomassie blue-stained gels after SDS-PAGE, using the PHORETIX 2-D image analysis software (NonLinear USA, Durham, NC) and bovine serum albumin (Sigma-Aldrich) as a protein standard.

Protease inhibitory activities

Rate constants for the association of individual and hybrid inhibitors with TLCK-treated bovine pancreas chymotrypsin (Sigma-Aldrich) were determined under pseudo-first-order conditions by the monitoring of substrate hydrolysis progress curves. Inhibitors (22 nm) were incubated in protease assay buffer (50 mm Tris–HCl, pH 8.0) in the presence of chymotrypsin (34 nm) and GLT-Ala-Ala-Phe-methylcoumarin (MCA) fluorigenic substrate (15 μm; Peptides International, Louisville, KY).

Apparent inhibition constants (kobs) were estimated by nonlinear regression fitting of each progress curve to the following model (Schechter and Plotnick, 2004):

display math

where [P] is the concentration of product at time t; [P]0, the concentration of product at time 0; v0, the (initial) rate of substrate hydrolysis; and kobs, the apparent first-order rate inhibition constant.

Second-order inhibition rate constants (kinh) were determined after empirical assessment of a KM value for the chymotrypsin–substrate interaction, using this equation:

display math

where [S]0 and [I]0 refer to the concentrations of substrate and inhibitor, respectively, at time 0.

Substrate hydrolysis progress curves were determined for the interaction of inhibitors with papain (from papaya latex; Sigma-Aldrich). Papain activity was measured in 50 mm Tris–HCl buffer, pH 8.0, using the synthetic fluorigenic substrate Z-Phe-Arg-MCA for cathepsin L-like enzymes (Peptides International). Hydrolysis was allowed to proceed in reducing conditions (10 mm l-Cys). Percentage inhibition values were determined by comparison to the protease alone. Chymotrypsin and papain were assayed with a Fluostar Galaxy microplate fluorimeter (BMG Labtech, Offenburg, Germany). Data were collected at 25 °C over 300 s, using excitation and emission filters of 360 and 450 nm, respectively. Three independent measurements were made, with three technical repetitions each, for each enzyme–inhibitor interaction assessed. Interactions were compared by one-way ANOVA followed by Tukey's multiple comparison test. Curve fitting and statistical analysis was performed with GraphPad Prism (GraphPad Software, La Jolla, CA).

Inhibition of plant cathepsin L-like proteases

Plant protease activities were assayed at 25 °C by the monitoring of Z-Phe-Arg-MCA hydrolysis progress curves, with leaf protein extracts (36 ng protein/μL) diluted in 50 mm MES, pH 5.8, containing 10 mm l-Cys. Protease activities were detected by spectrofluorimetry as described above for papain.

Acknowledgements

This work was supported by the Natural Science and Engineering Research Council of Canada, and by the Fonds de Recherche Québec/Nature et Technologies.

Ancillary