• Open Access

A protease activity–depleted environment for heterologous proteins migrating towards the leaf cell apoplast


(fax +1 418 656 7856; email dominique.michaud@fsaa.ulaval.ca)


Recombinant proteins face major constraints along the plant cell secretory pathway, including proteolytic processing compromising their structural integrity. Here, we demonstrate the potential of protease inhibitors as in situ stabilizing agents for recombinant proteins migrating towards the leaf apoplast. Genomic data for Arabidopsis, rice and Nicotiana spp. were assessed to determine the relative incidence of protease families in the cell secretory pathway. Transient expression assays with the model platform Nicotiana benthamiana were then performed to test the efficiency of protease inhibitors in stabilizing proteins targeted to the apoplast. Current genomic data suggest the occurrence of proteases from several families along the secretory pathway, including A1 and A22 Asp proteases; C1A and C13 Cys proteases; and S1, S8 and S10 Ser proteases. In vitro protease assays confirmed the presence of various proteases in N. benthamiana leaves, notably pointing to the deposition of A1- and S1-type activities preferentially in the apoplast. Accordingly, transient expression and secretion of the A1/S1 protease inhibitor, tomato cathepsin D inhibitor (SlCDI), negatively altered A1 and S1 protease activities in this cell compartment, while increasing the leaf apoplast protein content by ∼45% and improving the accumulation of a murine diagnostic antibody, C5-1, co-secreted in the apoplast. SlCYS9, an inhibitor of C1A and C13 Cys proteases, had no impact on the apoplast proteases and protein content, but stabilized C5-1 in planta, presumably upstream in the secretory pathway. These data confirm, overall, the potential of protease inhibitors for the in situ protection of recombinant proteins along the plant cell secretory pathway.


The growing demand for recombinant therapeutic and diagnostic proteins worldwide (Leader et al., 2008), and the structure of these proteins often requiring complex post-translational maturation (Walsh and Jefferis, 2006; Gomord et al., 2010) strongly contribute to the rising popularity of plant protein expression platforms (Faye and Gomord, 2010). Plants offer obvious advantages as protein biofactories in terms of cost efficiency, product safety, flexible scalability and competence to perform post-translational modifications typical of mammalian proteins (Karg and Kallio, 2009). Many recombinant proteins of medical interest have been successfully expressed in plant systems over the last 20 years, including a variety of blood proteins, hormones, growth regulators, enzyme inhibitors, vaccine antigens and mammalian antibodies (Ma et al., 2003; Daniell et al., 2009; De Muynck et al., 2010; Rybicki, 2010). Major progress has been achieved over the years towards the optimization of transgene transcription and translation in plant systems, the development of highly efficient expression platforms and the understanding of protein post-translational modifications specific to the plant cell machinery (Potenza et al., 2004; Faye et al., 2005; Gleba et al., 2007; Streatfield, 2007; Daniell et al., 2009; Sainsbury et al., 2009; Gomord et al., 2010).

A key challenge, at this stage, is to protect the integrity and overall quality of the recombinant protein products. Unlike several less complex pharmaceuticals, proteins present a natural tendency to structure heterogeneity, often giving a complex mixture of protein variants differing in their primary or tertiary structure (Faye et al., 2005). One factor strongly influencing the accumulation of recombinant proteins is their inherent instability in heterologous environments (Doran, 2006; Benchabane et al., 2008). Proteolytic enzymes are involved in numerous vital processes in plants (Schaller, 2004; van der Hoorn, 2008), but their abundance in plant tissues often represents a burden to the effective heterologous production of proteins. Whereas a number of recombinant proteins are accumulated under a stable form in plant systems, several others undergo partial or extensive hydrolysis negatively impacting their final yield and quality (Benchabane et al., 2008).

In particular, studies reported the unintended processing of recombinant proteins along the cell secretory pathway by resident proteases of the ER, the Golgi or the apoplast (e.g. Stevens et al., 2000; Sharp and Doran, 2001; Outchkourov et al., 2003; Schiermeyer et al., 2005; Badri et al., 2009a; Benchabane et al., 2009; De Muynck et al., 2009). A well-known example of this is the systematic processing of mammalian antibodies in plant systems (Gomord et al., 2004; De Muynck et al., 2010). Another example is the anti-inflammatory protein bovine aprotinin targeted to the secretory pathway of transgenic potato leaf cells, shown to undergo partial trimming at the C and N termini when retained in the ER using a C-terminal KDEL signal (Badri et al., 2009a). In a similar way, human α1-antichymotrypsin targeted to the secretory pathway of BY-2 tobacco cells could be detected in the culture medium but was cleaved in the C-terminal region by intracellular and apoplastic resident proteases, within the protease inhibitory site providing biological activity against chymotrypsin (Benchabane et al., 2009).

Different strategies have been proposed to minimize unintended proteolysis in planta (Doran, 2006; Benchabane et al., 2008), typically involving targeting to specific cellular locations with peptide sorting signals (e.g. Wandelt et al., 1992; Schouten et al., 1996; Gomord et al., 1997; Conley et al., 2009), or the addition of a stabilizing fusion partner to the recombinant polypeptide backbone (e.g. Obregon et al., 2006; Scheller et al., 2006; Patel et al., 2007; Floss et al., 2009). Here, we explored the potential of plant protease inhibitors to protect secreted proteins in situ, during their migration through the cell secretory pathway. Experimental support for the potential of this avenue was first provided by Komarnytsky et al. (2006), who improved the accumulation of a recombinant antibody in the hydroponic culture medium of transgenic tobacco plants by the secretion of a Bowman-Birk inhibitor in root exudates. Similarly, Kim et al. (2007) succeeded in reducing Ser protease activities in the growth medium of rice cell suspension cultures engineered to secrete a hybrid inhibitor with trypsin and chymotrypsin inhibitory activity. After an assessment of protease-related genomic data for Arabidopsis, rice and Solanaceae suggesting the general occurrence of proteases from different functional families in the plant cell secretory pathway, we demonstrate the potential of two broad-spectrum protease inhibitors from tomato as effective companion proteins for the in situ protection of secreted proteins transiently expressed in leaves of the model expression platform Nicotiana benthamiana.


A wide variety of secreted protease sequences in the genomes of Arabidopsis, rice and Solanaceae species

Hundreds of genes code for proteins involved in proteolytic processes in plants (Beers et al., 2004), with an estimate of 1900 genes in Arabidopsis directly or indirectly involved in the hydrolysis of peptide bonds (Schaller, 2004). Here, we performed systematic searches in major genomic and protease databases to identify protease families possibly represented in the plant cell secretory pathway. Protease sequence data for Arabidopsis, rice and different Solanaceae species were retrieved by sequence searches in the MEROPS peptidase database (Rawlings et al., 2008) or the SOL Genomics Network (SGN) database (Mueller et al., 2005), and then assessed for the presence of an N-terminal signal peptide indicating ER-mediated inclusion in the secretory pathway (Bendtsen et al., 2004). The abundance and relative distribution of secreted proteases among protease families in the whole genome were similar, overall, in Arabidopsis and rice (Table 1). More than 550 protease sequences were retrieved for both species, distributed in more than 20 functional families among the four major protease functional classes. Proteases presumably targeted to the secretory pathway were, for the most part, pepsin-like (A1 family) and, to a lesser extent, impas peptidase-like (A22 family) Asp proteases; papain-like (C1A family) and, to a lesser extent, legumain-like (C13 family) Cys proteases; trypsin/chymotrypsin-like (S1 family), subtilisin-like (S8 family) and serine carboxypeptidase-like (S10 family) Ser proteases; and matrix metallopeptidase-like (M10 family) and glutamate carboxypeptidase-like (M20 family) metalloproteases. This general pattern for the genome complement of secreted proteases could also be drawn for Solanaceae species, despite a limited amount of sequence data for this plant family in the MEROPS database (Table 1).

Table 1.   Relative incidence of secreted proteases among major protease families in Arabidopsis (Arabidopsis thaliana), rice (Oryza sativa) and Solanaceae species* Thumbnail image of

Complementary analyses were performed with ∼110 000 EST-derived unigenes available in the SGN database for Nicotiana tabacum, Nicotiana sylvestris and N. benthamiana. Putative protease sequences were first retrieved based on homology alignments and automatic annotations of the unigene sequences (Mueller et al., 2005), then submitted to the MEROPS batch blast tool to formally assign them to a specific protease family (Rawlings and Morton, 2008), and finally assessed for the presence of an N-terminal signal peptide indicative of inclusion in the cell secretory pathway (Bendtsen et al., 2004) (Table 2). As for Arabidopsis and rice, hundreds of secreted protease sequences were retrieved for Nicotiana species and assigned for the most part to the A1, A22, C1, C13, S1, S8, S10, M10 and M20 protease families (Table S1). Because unigene sequences are often made of incomplete gene sequences, the proportion of proteases with an N-terminal peptide signal for each Nicotiana species within each protease family was lower than the corresponding proportions calculated for the Arabidopsis and rice proteases (Tables 1 and 2). Despite this technical limitation underestimating the number of secreted proteases in Nicotiana species, the overall distribution of protease family members was found to be comparable among the plant species assessed, with A1, C1, S1, S8 and S10 proteases appearing, as a whole, as the most represented protease families in the plant cell secretory pathway (Figure 1).

Table 2.   Relative incidence of secreted proteases among major protease families in Nicotiana tabacum, Nicotiana sylvestris and N. benthamiana*
Protease familyN. tabacumN. sylvestrisN. benthamiana
No.% SP§No.% SPNo.% SP
  1. *Protease sequences were retrieved from unigenes databases of the SOL Genomics Network database and assigned to specific protease families using the MEROPS database batch blast tool (see text for details). The presence of an N-terminal signal peptide for cellular secretion was inferred using the SignalP algorithm (Bendtsen et al., 2004).

  2. After the classification of the MEROPS protease database.

  3. Number of protease sequences retrieved in the unigenes database.

  4. §Number of sequences with a predicted signal peptide (SP), relative to the total number of sequences retrieved (%).

A22520.00 0
Figure 1.

 Relative distribution of secreted protease-encoding genes among protease families in Arabidopsis (Arabidopsis thaliana), rice (Oryza sativa) and Nicotiana species. Percentages indicate the number of secreted protease sequences for each family, relative to the number of sequences retrieved for all families. Protease families were classified as in the MEROPS database (Rawlings et al., 2008). A, Asp proteases; C, Cys proteases; S, Ser proteases; M, metalloproteases.

In situ inhibition of protease activities in the leaf apoplast of N. benthamiana

In vitro protease assays with family-specific fluorigenic peptide substrates were conducted to empirically identify (endo)protease activities possibly damaging secreted proteins in the apoplast of N. benthamiana leaves (Figure 2). A malate dehydrogenase assay with whole-leaf and leaf apoplast protein extracts first showed that <0.2% of total activity detected in leaves for the intracellular enzyme was found in the apoplast extracts, thereby confirming the quality and validity of these preparations for subsequent analyses (Table 3). As expected considering the in silico inferences above and previous work showing the occurrence of Asp, Cys and Ser peptidases in the apoplast of N. tabacum leaves (Delannoy et al., 2008), a diversity of endoprotease activities were detected in the apoplast extracts, including cathepsin D/E-like (A1), papain-like (C1A), legumain-like (C13), chymotrypsin-like (S1), trypsin-like (S1) and subtilisin-like (S8) activities (Figure 2). A1 and S1 protease activities were easily detected in the apoplast protein extracts but barely detectable in whole-leaf protein extracts (Student’s t-test; < 0.05), which suggests an apoplastic destination for most of these enzymes and the potential of Asp and/or Ser protease inhibitors as modulators of protease activities in the apoplast milieu of N. benthamiana leaves. By contrast, S8, C1 and C13 protease activities were found at significantly lower levels in the apoplast preparations compared to whole-leaf extracts (Student’s t-test; < 0.05), thereby suggesting an intracellular location for several of these enzymes and a possible protein-stabilizing effect of Cys protease inhibitors within the secretory system, before secretion in the apoplast.

Figure 2.

 Major endoproteolytic activities in whole-leaf and apoplastic protein extracts of Nicotiana benthamiana leaves. Protease names refer to protease subfamilies expected to cleave the corresponding test substrate. Each bar is the mean of three independent (biological replicate) values ±SE.

Table 3.   Malate dehydrogenase activity in the apoplast of agroinfiltrated Nicotiana benthamiana leaves transfected with different pcambia vectors
TreatmentMDH activity (%)*
  1. *Data are expressed as the percentage of activity in leaf apoplast extracts, compared to the activity measured in whole-leaf protein extracts (value of 1.00). Absolute activities were measured on a fresh weight basis (nkat/g). Data are the mean of three independent (biological replicate) values ±SD.

No treatment (control)0.15 ± 0.02
Original (‘empty’) pcambia 2300 vector0.17 ± 0.04
SlCDI pcambia vector0.18 ± 0.04
SlCYS9 pcambia vector0.16 ± 0.02

Transient expression assays in agroinfected N. benthamiana leaves were conducted with protease inhibitor–encoding sequences to test these hypotheses (Figure 3). Two broad-spectrum inhibitors were selected for the assays: (i) an apoplast-targeted version of tomato cathepsin D inhibitor (SlCDI) (Werner et al., 1993), reported to inhibit protease members of the A1 and S1 families (Cater et al., 2002; Brunelle et al., 2005; Lison et al., 2006), and (ii) an apoplast-targeted version of the tomato cystatin SlCYS9 (Girard et al., 2007), expected to inhibit C1 and C13 proteases (Martinez et al., 2007; Benchabane et al., 2010). Unlike SlCYS9 having no significant effect, SlCDI transiently expressed in leaves (Figure 3a) caused a measurable decrease in A1 and S1 protease activities in the apoplast compared to the empty vector treatment, concomitant with a significant, ∼45% increase in total protein content as measured on a fresh weight basis (Figure 3b,c) (ANOVA; < 0.05).

Figure 3.

 Proteolytic activities and total protein content in the apoplast of Nicotiana benthamiana leaves transiently expressing SlCDI or SlCYS9, 6 days after agroinfection with the corresponding inhibitor-encoding sequences. Plants transfected with Agrobacterium cells harbouring the original pcambia 2300 vector were used as negative controls for comparative purposes. (a) Immunodetection of SlCDI in plants transfected with the SlCDI-encoding DNA sequence. Numbers on the left refer to commercial molecular weight markers (kDa). SlCYS9 expression in slcys9-transfected plants was also confirmed by immunodetection (not shown). (b) Protein content in the apoplastic extracts, assayed according to Bradford (1976). Each bar is the mean of three independent (biological replicate) values ±SE; 100 ng protein/μL of apoplast extract corresponded to ∼15 ng protein/mg leaf fresh weight under the conditions used. (c) Endoprotease activities in the apoplast extracts, as detected using protease family–specific fluorigenic peptide substrates. Protease names refer to protease subfamilies expected to cleave the corresponding test substrate. Each bar is the mean of three independent (biological replicate) values ±SE.

SlCDI impacts the leaf apoplast proteome of N. benthamiana

A comparative proteomic assessment of major protein families in the apoplast of agroinfected N. benthamiana leaves was conducted using digitalized proteome reference maps for this expression platform (Goulet et al., 2010a) to further characterize the impact of SlCDI on the apoplastic proteins (Figure 4). Protein families resolved by two-dimensional gel electrophoresis (2-DE) (Figure S1), including a variety of cell wall–modifying enzymes (e.g. peroxidases and copper-binding proteins) and Agrobacterium-inducible pathogenesis-related (PR-) proteins, were represented at normalized levels in SlCDI-transfected plants comparable to those observed in control and SlCYS9-transfected plants (ANOVA; > 0.05) (Figure 4a). By contrast, the net levels of these proteins after correction for protein content (see Figure 3b) were significantly increased in the presence of SlCDI, as notably illustrated by the significant, 54% increase in mean level of PR-3 proteins in SlCDI-transfected leaves compared to control plants transfected with the original pcambia vector (ANOVA; < 0.05) (Figure 4b).

Figure 4.

 Proteomic monitoring of major protein families in the apoplast of Nicotiana benthamiana leaves transiently expressing SlCDI or SlCYS9. The apoplastic proteins were resolved by 2-D gel electrophoresis as described earlier (Badri et al., 2009b) and identified by comparative spot matching based on digitalized reference maps for the leaf apoplast proteome of N. benthamiana (Goulet et al., 2010a). Plants transfected with Agrobacterium cells harbouring the original pcambia 2300 vector were used as negative controls for comparative purposes. (a) Relative abundance of major protein families in the apoplast, 6 days postagroinfection. Data are the sum of normalized spot volumes for each protein family member, i.e. the sum of protein member spot volumes in 2-D gels, divided by the total volume of all protein spots. Each bar is the mean of three independent (biological replicate gel) values ±SD. PR-1, PR-2, PR-3, PR-5 and PR-16: pathogenesis-related proteins 1, 2, 3, 5 and 16, respectively; Px, peroxidases; CBP, copper-binding proteins. (b) Net amounts of PR-3 proteins in the apoplast extracts. Data are the sum of normalized spot volumes for each PR-3 protein on 2-D gels (Panel a, above), corrected for protein content in the corresponding extracts (see Figure 3b). Each bar is the mean of three independent (biological replicate) values ±SE. See Table 4 for details on the relative and net levels of each PR-3 protein, and Supplementary Figure 1 for PR-3 protein spot numbering on 2-D gel reference maps.

A closer look at the data for the different PR-3 proteins suggested a protein-specific stabilizing effect for the tomato inhibitor (Table 4). For instance, PR-3 proteins 58, 87, 90 and 157 (see Goulet et al., 2010a and Figure S1 for protein numbering) showed net level increases of 50%–70% in the apoplast milieu when co-expressed with SlCDI, compared to lower (e.g. Protein 136), negligible (e.g. Protein 62), or by contrast more than twofold (e.g. Protein 125), increases for other proteins of the same family. These data do not formally exclude specific indirect effects at the gene level or upstream along the cell secretory pathway, but they point in sum to a limited impact of SlCDI on both the biosynthesis and maturation of secreted proteins, and the relative predominance of major protein families in the N. benthamiana leaf apoplast. They suggest, instead, that the up-regulating effect of this protein on the apoplastic protein content was the result of a direct inhibitory effect against apoplastic Asp (A1) and/or Ser (S1) proteases leading to an overall, albeit protein-specific, stabilization of resident proteins in situ.

Table 4.   Relative (normalized) and net amounts of PR-3 proteins in the apoplast of Nicotiana benthamiana leaves co-expressing C5-1 with SlCDI or SlCYS9*
PR-3 form*Normalized amounts (%)Net amounts (ng/μL)
Control+ SlCDI+ SlCYS9Control+ SlCDI+ SlCYS9
  1. *2-DE spatial coordinates for, and basic information on, N. benthamiana leaf PR-3 protein forms are given in the study by Goulet et al., 2010a and Figure S1. MS data for identity and function assignment are available in the Proteomics Identifications Database (PRIDE) (http://www.ebi.ac.uk/pride) (Vizcaino et al., 2009), under accession numbers 10538 to 10626.

  2. Values are presented as relative spot volumes for each individual PR-3 protein form, or as the sum of spot volumes for all forms of the PR-3 proteins functional group, compared to total protein volumes in 2-D gels (see Goulet et al., 2010a for the proteome reference map). Each datum is the mean of three independent (biological replicate) values ±SE. Ratio values correspond to normalized PR-3 protein amounts in protease inhibitor-expressing leaves, divided by the corresponding amount in control leaves.

  3. Values correspond to normalized values corrected for the absolute amount of proteins in the apoplast (see Figure 3b). Each datum is the mean of three independent (biological replicate) values ±SE. The asterisk indicates a significantly different value compared to control plants (ANOVA, < 0.01). Ratio values correspond to net PR-3 protein amounts in protease inhibitor–expressing leaves, divided by the corresponding amount in control leaves.

  4. §As reported in Figure 4a,b.

5812.6 ± 1.2813.9 ± 2.3713.0 ± 1.1328.3 ± 1.5145.3 ± 5.4332.6 ± 5.78
590.66 ± 0.030.54 ± 0.100.66 ± 0.161.54 ± 0.251.73 ± 0.181.59 ± 0.29
620.66 ± 0.180.41 ± 0.350.43 ± 0.211.49 ± 0.371.34 ± 1.171.02 ± 0.45
870.60 ± 0.170.62 ± 0.120.70 ± 0.201.42 ± 0.532.10 ± 0.501.65 ± 0.36
900.97 ± 0.361.16 ± 0.211.17 ± 0.172.37 ± 1.123.97 ± 0.992.82 ± 0.24
1250.21 ± 0.070.30 ± 0.070.29 ± 0.060.45 ± 0.110.97 ± 0.170.74 ± 0.21
1360.33 ± 0.060.34 ± 0.090.34 ± 0.090.77 ± 0.211.08 ± 0.240.88 ± 0.27
1572.59 ± 0.702.84 ± 0.432.74 ± 0.396.25 ± 2.399.28 ± 1.056.88 ± 1.40
Total (PR-3 family)§18.6 ± 0.5620.1 ± 2.1719.3 ± 0.9442.6 ± 4.5065.8 ± 2.75*48.2 ± 6.65
Ratio (/ctrl) 1.081.04 1.541.13

SlCDI and SlCYS9 facilitate the accumulation of a diagnostic murine antibody along the leaf cell secretory pathway of N. benthamiana

Transient expression assays were carried out with DNA sequences for the light and heavy chains of a murine blood group–typing monoclonal antibody, C5-1 (Khoudi et al., 1999), to assess the protein-stabilizing effects of SlCDI and SlCYS9 co-secreted with a clinically useful recombinant protein (Figure 5). The C5-1 immunoglobulin has been used in several instances as a model to monitor the expression, maturation and processing of heterologous proteins in plant systems (e.g. Khoudi et al., 1999; Bardor et al., 2003; Sainsbury et al., 2008; Vézina et al., 2009; D’Aoust et al., 2009). Compared to the control pcambia vector treatment, transiently expressed SlCDI and SlCYS9 increased C5-1 light chain accumulation by 70%–80% in leaves when co-expressed with the recombinant antibody (ANOVA; < 0.05) (Figure 5a,b). SlCDI also increased the accumulation of the heavy chain by ∼85% (ANOVA; < 0.01), in contrast to SlCYS9 showing no effect (ANOVA; > 0.05).

Figure 5.

 Stabilization of C5-1 light and heavy chains in agroinfected Nicotiana benthamiana leaves transiently expressing SlCDI or SlCYS9. (a) Immunodetection of C5-1 light (LC) and heavy (HC) chains expressed for 6 days alone (Ctrl) or along with SlCDI or SlCYS9. The two polypeptides were detected on nitrocellulose sheets with peroxidase-conjugated goat anti-mouse IgG, after SDS–PAGE separation of the leaf proteins in reducing conditions. (b) Relative amounts of C5-1 light and heavy chains in leaves, expressed on a total soluble protein basis. Protein chain signals were measured by densitometric analysis of the immunoblots. Each bar is the mean of three independent (biological replicate) values ±SE. (c) Immunodetection of C5-1 multimers with goat anti-mouse IgG (immunoblotting) or human IgG1 (activity immunoblotting) conjugated with peroxidase. Numbers on both sides of the blots refer to commercial molecular weight markers (kDa). (d) Relative amounts of C5-1 multimers in leaves. Human IgG-binding multimers were quantified by densitometric analysis of the immunoblots. An arbitrary value of 1.0 was given to the first sample (first repetition) of control plants expressing C5-1 alone. Each bar is the mean of six independent (biological replicate) values ± SE.

Immunodetection assays following SDS–PAGE in nonreducing conditions were carried out to confirm the ability of the light and heavy chains to assemble and produce biologically active diagnostic antibodies (Figure 5c). Immunodetection with goat anti-mouse IgG first allowed to confirm the presence of a high molecular weight, multi-band C5-1 pattern in the leaf extracts, similar to those observed earlier for C5-1 expressed in the same expression platform (D’Aoust et al., 2009; Vézina et al., 2009). Immunodetection with human IgG1 also allowed for the specific detection of multi-band C5-1 patterns as observed earlier, thereby confirming the ability of most light and heavy chain combinations to effectively recognize human immunoglobulins. As inferred by comparative densitometry of the immunoblots, the four active, human IgG1-binding multimers were found at significantly higher levels in SlCDI- and SlCYS9-transfected plants (Student’s t-test; < 0.05) (Figure 5d), as observed initially for the light and heavy chains resolved in reducing conditions (see Figure 5a). These observations confirm, overall, the potential of companion protease inhibitors for the in situ protection of clinically useful recombinant antibodies travelling the plant cell secretory pathway. In a more specific perspective, they point to differential protein-stabilizing effects for recombinant protease inhibitors in N. benthamiana leaves, likely associated with distinct protease targets along the secretory pathway.


Recombinant protease inhibitors have been mainly considered, until now, as antidigestive compounds useful to protect crops from insect herbivory or pathogenic infection (Ryan, 1990; Haq et al., 2004; Schlüter et al., 2010). Reports also described, in recent years, ‘pleiotropic effects’ for these proteins in planta (e.g. Gutiérrez-Campos et al., 2001; Belenghi et al., 2003; Van der Vyver et al., 2003; Huang et al., 2007; Prins et al., 2008; Shan et al., 2008; Zhang et al., 2008; Badri et al., 2009b; Srinivasan et al., 2009; Goulet et al., 2010b), which could be eventually harnessed to rationally control endogenous proteolytic processes and provide crops with new traits of agronomic value (Schlüter et al., 2010). From an industrial viewpoint, studies have shown the positive impact of ‘in-built’ heterologous inhibitors on the accumulation of recombinant proteins in the root growth medium of transgenic tobacco lines (Komarnytsky et al., 2006), or on the stability of these proteins in transgenic potato leaf crude protein extracts (Rivard et al., 2006). In a similar perspective, we recently reported the repressing effect of cytosol-targeted SlCDI on the general turnover of cytosolic proteins in potato leaf cells, correlated with a significant increase in endogenous and nonsecreted recombinant proteins in leaf tissues (Goulet et al., 2010b). Extending the use of heterologous protease inhibitors as protein-stabilizing agents in plant expression systems, we confirm here their effectiveness for the in situ protection of secreted proteins migrating towards the leaf cell apoplast.

A number of apoplast-targeted recombinant proteins have been successfully expressed in plant systems (e.g. Gaume et al., 2003; Streatfield et al., 2003; Sojikul et al., 2003; Komarnytsky et al., 2004; Wirth et al., 2004; Gils et al., 2005; Hellwig et al., 2004; Yang et al., 2005; Liénard et al., 2007; Loos et al., 2011), but the abundance and poor specificity of proteolytic enzymes in the apoplast often represents a significant hurdle, hardly compatible with effective recombinant protein production schemes (Hellwig et al., 2004; Schiermeyer et al., 2005; Doran, 2006; Benchabane et al., 2008; Delannoy et al., 2008; De Muynck et al., 2009). Bovine aprotinin, for instance, was detected as an intact form in the apoplast of transgenic potato leaves, but final yields in planta were significantly lower than those obtained with an ER-retained, KDEL-variant for comparable levels of mRNA transcripts (Badri et al., 2009b). In a similar way, the light and heavy chains of the blood typing C5-1 antibody were easily detected here in N. benthamiana leaves agroinfected with the corresponding gene sequences, but the accumulation of both polypeptides was clearly not optimal considering their higher accumulation rate in plants co-secreting SlCDI. Although work still needs to be carried out to properly characterize the structure and biological activity of the resulting protein product, our data represent a promising step in the successful stabilization of plant-made antibodies, expected to provide a range of clinically useful drugs and diagnostic tools during the next decade (Jefferis, 2009; De Muynck et al., 2010).

In a broader context, the targeted reconfiguration of host-resident protease activities with companion protease inhibitors might represent an advantageous alternative to current protein engineering and cell-targeting strategies for recombinant protein stabilization in vivo, which often lead to proteins structurally distinct from their native, original counterpart. Recombinant protein retention in the ER is a useful strategy to improve protein yields in planta (Stoger et al., 2002; Ma et al., 2003; Vitale and Pedrazzini, 2005; Torrent et al., 2009), but it requires the addition of non-native amino acids and compromises complete maturation of N-glycan chains along the secretory pathway, essential to sustain the stability or pharmacokinetics of certain proteins in a therapeutic context (Gomord and Faye, 2004; Sethuraman and Stadheim, 2006). From this perspective, the in situ modulation of protease activities might not only represent an alternative to protein engineering approaches for proteins such as antigens or antibodies that do not require extensive maturation or strict identity with the native protein, but also be a useful complement to currently developed glycoengineering strategies for the production of ‘humanized’ therapeutic proteins highly similar to their original mammalian homologue (Strasser et al., 2004, 2008; Cox et al., 2006; Saint-Jore-Dupas et al., 2006; Sourrouille et al., 2009; Frey et al., 2009; Vézina et al., 2009; Gomord et al., 2010).

In theory, proteolytic processes affecting recombinant protein accumulation could be contained in vivo using antisense or RNA silencing strategies implemented in transgenic host plants (Schiermeyer et al., 2005; Watson et al., 2005). Our genomic inferences indicating a wide array of possible protease forms for each functional class along the secretory pathway also suggest the relevance of strategies involving broad-spectrum protease inhibitions, such as those conducted here with SlCDI and SlCYS9 for the inhibition of endogenous A1, C1A, C13 and S1 proteases. Additional work is now required to further characterize the distribution of SlCDI and SlCYS9 target proteases along the secretory pathway of N. benthamiana leaf cells, to assess the possible impacts of these proteins on the host plant cell metabolism and to identify alternative protease targets, including membrane-bound proteases (Zhang et al., 2011), along the secretory pathway. The differential stabilization of C5-1 light and heavy chains by SlCDI and SlCYS9, together with the differential in vivo effects of these two inhibitors on leaf apoplast protease activities, clearly indicates protein- and inhibitor-specific outputs for the proposed approach, determined both by the cellular location of the target proteases and by the number and position of intrinsic cleavage sites on the protein to express. Whereas the apoplast protein- and C5-1 antibody-stabilizing effects of SlCDI were clearly associated with the apoplast, the stabilizing effect of SlCYS9 on the antibody light chain and multimers likely involved the inhibition of intracellular proteases and remains to be characterized.

The possible impacts of in vivo protease inhibition on the host plant metabolism also remain to be characterized. Comparative proteome assessments with digitalized maps for the leaf apoplast proteome of N. benthamiana suggested a limited qualitative impact of SlCDI and SlCYS9 in planta, but eventual pleiotropic effects cannot be excluded with some inhibitors, notably for those inhibition strategies involving target proteases in the early compartments of the secretory pathway (Badri et al., 2009b). Work is underway to address these questions, using a collection of engineered inhibitors targeted to different locations along the cell secretory pathway. Work is also underway to determine the potential of companion inhibitors for the stabilization of secreted proteins in non-plant eukaryotic systems.

Experimental procedures

Genomic assessments

The overall diversity of secreted proteases in plant cells was assessed with Arabidopsis (A. thaliana), rice (Oryza sativa) and Solanaceae species (including Nicotiana spp.) as models. Amino acid sequences from the major protease families were retrieved from different gene and protein databases and assessed for the presence of an N-terminal signal peptide indicating their targeting to the secretory pathway. Protease sequences of Arabidopsis, rice and Solanaceae were retrieved from the MEROPS peptidase database (http://www.merops.sanger.ac.uk) (Rawlings et al., 2008), after discarding nonpeptidase homologues. Nicotiana spp. protease sequences were retrieved from the SGN unigene databases (solgenomics.net) for N. tabacum, N. sylvestris and N. benthamiana (Mueller et al., 2005). The proteases were identified using (i) the NCBI BLAST tool, v. 2.2.15, against known proteases of A. thaliana and O. sativa, and (ii) systematic searches among automatic annotations of the SGN unigene sequences. A further screening of the sequences was performed with the MEROPS database batch BLAST function (http://www.merops.sanger.ac.uk) to discard unclassified proteases (Rawlings and Morton, 2008). The presence or absence of an N-terminal signal peptide for cellular secretion was inferred for each protease sequence retrieved using the SignalP tool, v. 3.0 (http://www.cbs.dtu.dk/services/ SignalP) (Bendtsen et al., 2004).

Leaf total proteins

Seven-week-old greenhouse-grown N. benthamiana L. plants were used for the experiments. The plants were kept under a 14:10 day/night photoperiod, watered daily as needed and fertilized twice a week with a 400 ppm solution of equilibrated commercial fertilizer. Total proteins were extracted from leaves of the same physiological age ground to a fine powder in liquid nitrogen. For each extraction, 3 mL of 10 mm MES buffer, pH 5.8, was added to 1 g of leaf powder, and the mixture was incubated on ice for 15 min. The protein extracts were clarified by centrifugation at 10 000 g for 10 min at 4 °C and kept at −80 °C until use. Total proteins were assayed according to Bradford (1976), with bovine serum albumin as a protein standard (Sigma-Aldrich, Oakville, ON, Canada).

Leaf apoplast proteins

Apoplast proteins were recovered as described by Dani et al. (2005), with some modifications. Freshly harvested leaves of the same age were washed in chilled double-distilled water and submerged in chilled vacuum infiltration buffer (10 mm MES, pH 5.8). The leaves were vacuum-infiltrated twice for 20 s with infiltration buffer, dried off to remove excess buffer, carefully rolled in a home-made swiss-roll cylinder with a collector tube at one extremity and centrifuged at 4 °C for 10 min at 1000 g to collect the vacuum infiltrate. The protein solutions were centrifuged at 6000 g for 5 min at 4 °C to discard A. tumefaciens cells, and the protein content was assayed according to Bradford (1976) with bovine serum albumin as a standard. The extracts for proteomic analyses were dialysed overnight at 4 °C against chilled double-distilled water, lyophilized and kept at −80 °C until further analysis.

Malate dehydrogenase assay

Contamination of the apoplast preparations with intracellular proteins was assessed using a colorimetric activity assay for the intracellular enzyme malate dehydrogenase (Dani et al., 2005). Dehydrogenase activity was assayed after mixing protein extracts from whole leaves (intracellular controls; see Leaf total proteins, above) or apoplastic preparations (see Leaf apoplast proteins, above) with 50 mm Tris–HCl, pH 7.5, containing 0.2 mm NADH and 0.4 mm oxaloacetate (Sigma-Aldrich), and comparing absorbance decreases at 340 nm using a Multiskan Ascent microplate spectrophotometer (Thermo Fischer Scientific).

Protease assays

Protease activities were assayed by monitoring substrate hydrolysis progress curves (Goulet et al., 2008) using the following synthetic fluorigenic substrates: Z-Phe-Arg-methylcoumarin (MCA) for C1A papain-like proteases, Z-Arg-MCA for S1 trypsin-like proteases, Glt-Ala-Ala-Phe-MCA for S1 chymotrypsin-like proteases, Ala-Ala-Phe-MCA for S8 subtilisin-like proteases, Z-Ala-Ala-Asn-MCA for C13 legumain-like proteases and MOCAc-Gly-Lys-Pro-Ile-Leu-Phe-Phe-Arg-Leu-Lys(Dnp)-d-Arg-NH2 for A1 cathepsin D/E-like proteases (Pepnet, Louisville, KY). Substrate hydrolysis reactions with whole-leaf and apoplast protein extracts (3.6 μg protein in 100 μL) were allowed to proceed at 25 °C in 50 mm MES, pH 5.8, containing 10 mm l-Cys for Cys protease assays. Protease activity levels were monitored using a Fluostar Galaxy microplate fluorimeter (BMG, Offenburg, Germany), with excitation and emission filters of 360 and 450 nm, respectively, for the MCA substrates, or of 340 and 400 nm, respectively, for the MOCAc substrate. Protein samples from three biological (plant) replicates were used for each treatment to allow for statistical analyses.

Transient expression assays

Different gene vectors were used, alone or in combination, for transient expression assays, including (i) an engineered pcambia vector for the light and heavy chains of C5-1 (Khoudi et al., 1999); (ii) an engineered pcambia vector for SlCDI (Werner et al., 1993; Brunelle et al., 2004); (iii) an engineered pcambia vector for SlCYS9 (Girard et al., 2007); and (iv) an original (‘empty’) version of the plant binary expression vector, pcambia 2300 (CAMBIA, Canberra, Australia), as a control for agroinfection. The pcambia 2300 was used as provided (CAMBIA). SlCDI- and SlCYS9-encoding vectors were devised as recently described (Goulet et al., 2010a). The C5-1 antibody sequences were cloned in the pcambia 2300 vector after PCR amplification of the two chain sequences using the following primers: 5′ ATGGTTTTCACACCTCAGATACTTGG and 3′ ATATGAGCTGCGATGCCTAACACTCATTCCTGTTGAAGC for the light chain, and 5′ ATGGC-TTGGGTGTGGACCTTGC and 3′ ATAAGAGCTCGCATGCTCATTTACCAGGAGAGTGGG for the heavy chain. The two genes were integrated in a gene expression cassette including a double cauliflower mosaic virus 35S promoter, the tobacco etch virus enhancer sequence in 5′ position and the Nos terminator sequence in 3′ position. The gene cassettes were inserted in a single expression vector to allow for co-expression of C5-1 light and heavy chains in transfected leaves. All gene vectors for expression assays were electroporated in A. tumefaciens, strain LBA4404, prior to plant infection. Leaf infiltration was performed as described (D’Aoust et al., 2009; Goulet et al., 2010a), after blending equal volumes of agrobacterial cultures, as needed. Infiltrated tissues for protein extraction and analysis were collected 6 days postinfection. Three (or six) biological (plant) replicates were used for each treatment to allow for statistical analyses.

Proteomic assessments

2-DE was performed as described by Goulet et al. (2010a), with 240 μg of leaf apoplast protein per gel. The proteins were first resolved by isoelectric focusing along a nonlinear 3–10 pI gradient, and then submitted to 15% (w/v) SDS–PAGE for separation based on polypeptide molecular weights. The 2-D gels were stained with Coomassie Blue, digitalized with an Amersham Image Scanner (GE Healthcare, Baie d’Urfé, QC, Canada) and analysed using the Phoretix 2-D Expression software, v. 2005 (NonLinear USA, Durham NC) (Badri et al., 2009b), with three biological (plant) replicates per treatment to allow for statistical analyses. The proteins were identified by protein spot matching using a digitalized reference map for the apoplast proteome of agroinfected N. benthamiana leaves (Goulet et al., 2010a), deposited in the Proteomics Identifications Database (http://www.ebi.ac.uk/pride) under the accession numbers 10538 to 10626.


SlCDI, SlCYS9 and C5-1 were immunodetected by a routine procedure (Towbin et al., 1979), following 10% (w/v) SDS–PAGE (Laemmli, 1970) and electrotransfer onto nitrocellulose membranes. Leaf total proteins were extracted in 50 mm Tris–HCl, pH 7.4, containing 150 mm NaCl, 0.1% (v/v) Triton X-100 and 1 mm phenylmethylsulfonyl fluoride, as described previously for leaf total proteins. SlCDI was detected with polyclonal IgY antibodies raised in chicken against a hydrophilic surface loop of the protein (Khalf et al., 2010) and peroxidase-conjugated rabbit anti-IgY as secondary antibodies (Sigma-Aldrich). SlCYS9 was detected with polyclonal IgG antibodies raised in rabbit against potato multicystatin (Girard et al., 2007) and peroxidase-conjugated goat anti-rabbit IgG as secondary antibodies (Sigma-Aldrich). The light and heavy chains of C5-1 were detected with peroxidase-conjugated goat anti-mouse IgG (Sigma-Aldrich) for immunodetection, or with peroxidase-conjugated human IgG1 (Sigma-Aldrich) for activity blotting. Antibody chains and complexes were quantified by densitometric analysis of the immunoblots, using the Phoretix 2-D Expression software (Nonlinear USA) and purified mouse IgG (Sigma-Aldrich) to generate a standard curve (Goulet et al., 2010b).


We thank Professor Kamal Bouarab (Université de Sherbrooke, Sherbrooke QC, Canada) for providing N. benthamiana seeds and Héma-Québec (Québec QC, Canada) for providing DNA material for C5-1 antibody cloning. This work was supported by the Natural Science and Engineering Research Council (NSERC) of Canada. C. Goulet was the recipient of an NSERC doctoral scholarship.