• Open Access

Tubers from potato lines expressing a tomato Kunitz protease inhibitor are substantially equivalent to parental and transgenic controls

Authors


*(fax +1 418 656 7856; e-mail dominique.michaud@fsaa.ulaval.ca)

Summary

Recombinant protease inhibitors represent useful tools for the development of insect-resistant transgenic crops, but questions have been raised in recent years about the impact of these proteins on endogenous proteases and chemical composition of derived food products. In this study, we performed a detailed compositional analysis of tubers from potato lines expressing the broad-spectrum inhibitor of Ser and Asp proteases, tomato cathepsin D inhibitor (SlCDI), to detect possible unintended effects on tuber composition. A compositional analysis of key nutrients and toxic chemicals was carried out with tubers of SlCDI-expressing and control (comparator) lines, followed by a two-dimensional gel electrophoresis (2-DE) proteomic profiling of total and allergenic proteins to detect eventual effects at the proteome level. No significant differences were observed among control and SlCDI-expressing lines for most chemicals assayed, in line with the very low abundance of SlCDI in tubers. Likewise, proteins detected after 2-DE showed no quantitative variation among the lines, except for a few proteins in some control and test lines, independent of slcdi transgene expression. Components of the patatin storage protein complex and Kunitz protease inhibitors immunodetected after 2-DE showed unaltered deposition patterns in SlCDI-expressing lines, clearly suggesting a null impact of slcdi on the intrinsic allergenic potential of potato tubers. These data suggest, overall, a null impact of slcdi expression on tuber composition and substantial equivalence between comparator and SlCDI-expressing tubers despite reported effects on leaf protein catabolism. They also illustrate the usefulness of proteomics as a tool to assess the authenticity of foods derived from novel-generation transgenic plants.

Introduction

Numerous studies reported the potential of recombinant protease inhibitors as useful antinutritive compounds to protect crop plants from herbivory or pathogenic infection (Michaud, 2000; Haq et al., 2004). Ser protease inhibitors, in particular, have been readily identified as potential candidates for the production of insect-resistant transgenic crops (Hilder et al., 1987; Johnson et al., 1989; Duan et al., 1996; Xu et al., 1996), and their usefulness to reduce insecticide loads in the field has been documented in recent years (Huang et al., 2005; Qiu, 2008). Most protease inhibitors in plants are proteinaceous competitive inhibitors acting as pseudo-substrates to enter the active site of target proteases. Following inhibition, the proteases can no longer cleave peptide bonds, thereby resulting in a detrimental inhibition of extracellular protein digestive functions in herbivores leading to significant growth and development delays (Arai et al., 2002; Birk, 2003; Haq et al., 2004).

Despite promising developments, the expression of recombinant protease inhibitors in plants raises a number of questions about the occurrence of unintended metabolic interference (or pleiotropic) effects on endogenous proteolysis eventually altering growth, development or chemical composition of the host-plant tissues. Proteolytic enzymes are ubiquitous effectors in living cells involved in the regulation of numerous metabolic processes, ranging from housekeeping functions such as protein turnover and the elimination of misfolded polypeptides to the processing of polypeptide pre- and pro-regions on maturing protein backbones (Estelle, 2001; Schaller, 2004; Smalle and Vierstra, 2004; Faye et al., 2005; van der Hoorn, 2008). Whereas studies reported negligible phenotypic effects for protease inhibitors expressed in transgenic plants based on the assessment of macroscopic indicators such as growth rate, stem diameter or leaf number (e.g. Masoud et al., 1993; Brunelle et al., 2004; Rivard et al., 2006; Badri et al., 2009), recent reports suggest the onset of more subtle effects at the metabolic level. For instance, several studies documented the pesticidal effects of recombinant cystatins expressed in transgenic plants (e.g. Lepléet al., 1995; Atkinson et al., 2004; Outchkourov et al., 2004), but the constitutive accumulation of these proteins in planta could also be associated with an alteration of flower development (Gutiérrez-Campos et al., 2001), an inhibition of the hypersensitive response (Belenghi et al., 2003), a modified physiological behaviour of the host plant under low temperature regimes (Van der Vyver et al., 2003), or an altered total protein content in leaves (Van der Vyver et al., 2003; Prins et al., 2008). In a similar way, no visible phenotypic effects were observed for the Ser protease inhibitors bovine aprotinin and tomato Kunitz cathepsin D inhibitor (SlCDI) expressed in transgenic potato lines (Brunelle et al., 2004; Rivard et al., 2006), but metabolic interference significantly altering leaf protein levels were observed for both inhibitors (Badri et al., 2009; Goulet et al., 2009).

To further document the pleiotropic effects of recombinant protease inhibitors in plants and to assess their possible impact on the composition and authenticity of food products, we examined the chemical composition and proteome profile of potato tubers harvested from a collection of SlCDI-expressing, transgenic potato lines (Brunelle et al., 2004; Rivard et al., 2006). A classical scheme for the safety assessment of foods derived from transgenic crops consists of comparing their chemical composition with the composition of a conventional counterpart referred to as the comparator line, preferentially derived from a sister line or the parental line used for genetic transformation (OECD, 1993; George et al., 2004). Based on this, a transgenic product is considered as likely safe for human use if it is found to be ‘substantially equivalent’ to a conventional comparator with a documented history of safe use, provided that the encoded recombinant protein is not allergenic, antinutritive or toxic to eventual consumers (OECD, 2002; El Sanhoty et al., 2004). For potato, about 30 chemicals have been identified as key nutritional or toxic indicators for the assessement of substantial equivalence among tubers, including proximate components such as total solids, total sugars, total proteins, total lipids, ash and fibres; and specific nutrients or toxic compounds such as sucrose, glucose, starch, vitamin C, amino acids, potassium, magnesium, copper and glycoalkaloids (AOAC, 2000a; OECD, 2002). Detailed studies have been published in recent years comparing the chemical composition of transgenic potato tubers with the composition of a conventional comparator (Rogan et al., 2000; El Sanhoty et al., 2004; Shepherd et al., 2006). In this study, we combined this approach with a detailed proteomic profiling of tuber proteins (Lehesranta et al., 2005; Bauw et al., 2006) to compare in an integrated way the composition of tubers harvested from different SlCDI-expressing and comparator potato lines. We recently observed a significant impact of SlCDI on the leaf proteome of potato lines constitutively expressing this protein (Goulet et al., 2009), which raised questions about the chemical composition of tuber tissues in these plants and suggested their possible usefulness as models for the validation of safety assessment strategies adapted to food products derived from novel-generation transgenic plants.

Results

SlCDI-expressing and control potato tubers are substantially equivalent

Seven SlCDI-expressing potato lines exhibiting variable transgene expression and SlCDI protein levels in leaves were selected for the compositional analyses (‘CD’ lines in Rivard et al., 2006; Goulet et al., 2009), along with three comparator lines consisting of the parental line K and two transgenic control lines (‘SPCD’ lines) bearing the slcdi transgene with no promoter for heterologous expression. PCR assays were first performed to confirm the presence of the nptII (selection marker) and slcdi transgenes in tubers of the nine transgenic lines, and to confirm the absence of slcdi mRNA transcripts in control lines (Figure 1). As expected, the nptII transgene was amplified from total DNA extracts of all CD and SPCD lines, unlike nontransgenic line K showing no positive signal (Figure 1a). Likewise, the slcdi transgene was present in all CD and SPCD lines but absent from the parental line (Figure 1c), in contrast with (an) endogenous CDI-encoding gene(s) amplified in all lines using potato-specific oligonucleotide primers (Figure 1b). In agreement with the absence of a promoter sequence in gene constructs, no slcdi transcript signal was amplified from the SPCD lines by reverse transcriptase (RT)-PCR, similar to parental line K (Figure 1d,e). By comparison, slcdi transcripts were easily detected in the seven CD lines, albeit at variable levels from one line to another. Unlike leaf protein extracts, tuber protein extracts from the SlCDI-expressing lines had no measurable trypsin inhibitory activity in vitro against bovine trypsin used as a target (Figure 1f), which suggests either a very low abundance of SlCDI in tuber tissues and/or a negligible functional impact of this protein in the presence of highly abundant endogenous trypsin inhibitors.

Figure 1.

 Integration and expression of the tomato cathepsin D inhibitor (slcdi) transgene in tubers of transgenic potato lines. (a–c) Polymerase chain reaction (PCR) detection of the nptII marker transgene (a), endogenous SlCDI-like gene(s) (b) and slcdi transgene (c) in control (K, SPCD) and SlCDI-expressing (CD) lines. (d, e) Reverse transcriptase (RT)-PCR detection of slcdi mRNA transcripts in control and CD lines. (f) Trypsin inhibitory activity in leaf and tuber protein extracts of control and CD lines. DNA amplicons following PCR and RT-PCR were submitted to agarose gel electrophoresis and visualized by staining with ethidium bromide. Amplicons produced from DNA templates were used as positive controls for the tests [CTRL(+)]. Data in (e) are expressed as relative amounts of slcdi transcripts compared with the amount in leaves of clone CD21A (arbitrary value of 1.0). Data in (f) are expressed as relative activities compared with the activity in leaf or tuber protein extracts of control line K. Each bar on panels (e) and (f) is the mean of three independent (biological) replicates ±SE.

A comparative compositional analysis was performed with the ten lines selected to address this question, using as source material 5 cm-long tubers from 4 month-old plants acclimated and grown in greenhouse. In brief, no significant difference was observed among the lines for a number of proximate components including total solids, total proteins, total lipids, total sugars, ash and crude fibres (Table 1). As well, no difference was observed for key nutritional indicators of potato tubers such as glucose, sucrose, starch, soluble proteins, vitamin C, amino acids and selected minerals (Tables 2 and 3), except for magnesium levels showing slight but significant variations in one control (SPCD4) and two SlCDI-expressing (CD26A and CD21A) lines (anova: < 0.05). Overall, these data showing a similar chemical composition for the different tubers, together with the absence of glycoalkaloids in both control and SlCDI-expressing lines (Table 2), indicate that tubers harvested from SlCDI-expressing and comparator lines were substantially equivalent despite detectable expression of the slcdi transgene in tuber RNA extracts (Figure 1d,e), measurable SlCDI inhibitory activity in leaves (Figure 1f), and significant interference of SlCDI on the overall proteome of leaf tissues (Goulet et al., 2009).

Table 1.   Proximate chemical composition of tubers from control and tomato cathepsin D inhibitor (SlCDI)-expressing potato lines
ComponentConcentration (% dry weight)*
K (ctrl)Transgenic controlsTransgenic test lines
SPCD4SPCD7CD26ACD32ACD18ACD33ACD34ACD3ACD21A
  1. *Data are presented as mean contents. No significant difference was observed among the ten lines for all components assessed (anova; > 0.05).

Total solids20.520.221.021.020.820.721.420.620.522.2
Total proteins9.929.639.469.449.329.619.319.739.929.11
Total lipids1.041.140.951.051.101.111.030.970.981.04
Total sugars77.978.378.878.878.078.479.178.177.879.7
Ash4.484.524.374.344.424.354.264.434.494.15
Crude fibre1.221.291.281.291.251.211.221.261.271.22
Table 2.   Selected nutrients and antinutritional compounds in tubers of control and tomato cathepsin D inhibitor (SlCDI)-expressing potato lines
ComponentConcentration (% fresh weight)*
K (ctrl)Transgenic controlsTransgenic test lines
SPCD4SPCD7CD26ACD32ACD18ACD33ACD34ACD3ACD21A
  1. *Data are presented as mean contents. No significant difference was observed among the ten lines for all components assessed (anova; > 0.05).

  2. †Expressed in mg/100 g fresh weight.

  3. ‡ND, not detected (limit of detection: 0.1 mg/100 g fresh weight).

Glucose0.120.130.120.120.130.130.110.120.120.12
Sucrose0.200.200.210.210.180.210.210.210.200.21
Starch14.714.414.314.814.714.514.614.614.814.6
Soluble proteins0.610.710.660.630.630.710.730.710.720.79
Vitamin C†79.179.079.278.979.379.879.079.579.277.9
Glycoalkaloids‡NDNDNDNDNDNDNDNDNDND
Table 3.   Selected minerals and amino acids in tubers of control and tomato cathepsin D inhibitor (SlCDI)-expressing potato lines
ComponentConcentration (mg/100 g fresh weight)*
K (ctrl)Transgenic controlsTransgenic test lines
SPCD4SPCD7CD26ACD32ACD18ACD33ACD34ACD3ACD21A
  1. *Data are presented as mean concentrations. No significant difference was observed among the ten lines for all components assessed (anova; > 0.05), except for magnesium showing a significant difference in some transgenic control or test lines (*< 0.05) compared with untransformed mother line K.

Minerals
 Potassium326340348385303329381342388388
 Magnesium17.622.6*20.721.5*20.319.820.620.020.523.5*
 Copper0.140.180.220.220.230.180.190.240.170.18
Amino acids
 Aspartate30.434.034.035.030.727.832.733.030.634.7
 Methionine14.214.220.318.820.314.221.316.221.316.2
 Glutamate7.858.119.639.939.607.097.199.478.418.71
 Asparagine11.010.811.710.210.710.49.5310.910.910.4
 Glutamine535538622598553469551631572664
 Cysteine33.431.430.436.536.034.535.535.533.434.0
 Glycine3.553.504.052.522.533.543.042.502.512.03
 Histidine244255258205209211247230238242
 Phenylalanine8.6110.110.69.127.099.609.637.608.116.59
 Serine3.554.566.086.086.086.597.605.078.117.60
 Threonine7.098.608.598.598.1110.66.597.6010.19.12
 Tryptophan713648694727746702735710683695
 Tyrosine72.070.474.069.958.873.068.968.469.973.0
 Arginine11699.390.710610511511598.886.194.8

Slcdi transgene expression has no significant impact on the tuber proteome

A proteomic analysis was carried out to determine whether the null impact of SlCDI on tuber composition could be associated with a negligible amount of this protein in planta, and to look for eventual minor pleiotropic effects detectable at the proteome level. To this end we compared the two-dimensional gel electrophoresis (2-DE) protein profiles of tubers harvested from two SPCD comparator lines, lines SPCD4 and SPCD7, with the 2-DE tuber protein profiles of two SlCDI-expressing lines, lines CD3A and CD21A (Figure 2), exhibiting SlCDI levels in leaves estimated at ∼0.1% of total soluble proteins (Goulet et al., 2009), comparable with expression levels reported earlier for insect-resistant transgenic tobacco lines expressing Ser-type inhibitors (Hilder et al., 1987; Johnson et al., 1989). Because most proteins in tuber extracts exhibited a neutral or mildly-acidic pI value (Figure 2a), a 47 pH gradient was used for isoelectric focusing (IEF) to minimize physical interference during migration or masking effects upon staining due to the presence of highly abundant proteins (Figure 2b).

Figure 2.

 Representative two-dimensional gel electrophoresis protein profiles for tubers of transgenic control line SPCD4. Isoelectric focusing for the 1st dimension was performed along 3–10 (a) or 4–7 (b) pH gradients. About 550–600 protein spots were observed on gels after colloidal blue staining. Numbered spots on panel (b) refer to proteins showing altered levels in at least one line (see Figure 3 for quantitative details). Square boxes on panels (a) and (b) point to multiple components of the patatin storage protein complex (see Bauw et al., 2006). Numbers and protein scales on the left correspond to commercial molecular weight markers (kDa).

In brief, most of the 550–600 proteins detected in 2-D gels after colloidal blue staining were found at similar levels in the four lines tested, including multiple components of the major potato storage protein, patatin. By contrast, 21 proteins showed a significantly altered level in at least one line (Figure 2b). Protein spots #3 and 12, for instance, were less abundant in lines SPCD7 and CD3A than in lines SPCD4 and CD21A, compared with protein spots #1, 7 and 18 being less abundant in line SPCD4 than in the three other lines (Figure 3). Despite significant differences, none of the proteins showing variation did correspond to abundant (e.g. storage) proteins, and most alterations fell within a less than twofold variation range. Most importantly, no variation measured in the 2-D gels could be correlated with the presence or absence of the slcdi transgene. Several proteins, for instance, were found at different levels in the two SPCD lines or followed diverging variation patterns in the CD lines (Figure 3), which might indicate somaclonal variation or false-positive alterations following anova, within the error range of 5% adopted for statistical significance. Possibly accounting for the negligible impact of slcdi transgene expression, no protein spot was detected in 2-D gels for the two CD lines at the 22.0 kDa/pI 8.9 intersect corresponding to the expected migration coordinates of recombinant, nonglycosylated SlCDI. In agreement with compositional data indicating substantial equivalence among tubers of test and control lines (see above), these observations suggest, overall, a negligible effect of slcdi at the proteome level associated with a very low abundance of SlCDI in tuber cells despite the presence of mRNA transcripts in tuber RNA preparations (Figure 1e) and the detection of active SlCDI in leaves (Figure 1f).

Figure 3.

 Relative levels of protein spots exhibiting altered levels among control and test lines (see Figure 2b for spot numbering). Data are expressed as relative variations in lines SPCD7, CD3A and CD21A compared with line SPCD4 (arbitrary value of 1.0). Positive (negative) values indicate higher (lower) relative levels compared with line SPCD4. Asterisks indicate significant (*) or highly significant (**) variations among the four lines assessed (anova; < 0.05 or < 0.01).

Slcdi transgene expression has no impact on the deposition pattern of protein allergens

Immunodetection assays were carried out to formally confirm the negligible abundance of SlCDI in tuber protein extracts and to assess the overall deposition pattern of endogenous Kunitz protease inhibitors (Pouvreau et al., 2001; Heibges et al., 2003), identified earlier as potential protein allergens in potato tubers along with the multifunctional storage protein patatin (Seppäläet al., 2001). We observed above a conserved deposition pattern for components of the patatin complex in tubers of control and SlCDI-expressing lines (see Figure 2). As a complement, we immunodetected and assayed Kunitz inhibitors in tuber extracts, after designing and producing chicken polyclonal antibodies for the probing of a 15-aa loop protruding at the surface of SlCDI and potato Kunitz inhibitors (see Figure 4a). Four bands c. 17–23 kDa in size were detected on immunoblots for all lines tested, including control parental line K (Figure 4b). No extra signal was detected for SlCDI in tubers of the CD lines, in contrast with (healthy) leaves of line CD3A exhibiting a ∼20-kDa protein undetected in line K, immunologically related to bacterially expressed SlCDI (Figure 4c). The detection of a unique, multiprotein pattern in test and control lines, along with the absence of an extra signal for SlCDI in the CD lines, again suggested the very low abundance of SlCDI in transgenic tubers (see above) and a null impact for this protein on the deposition of endogenous Kunitz proteins.

Figure 4.

 Immunodetection of tomato cathepsin D inhibitor (SlCDI) and endogenous Kunitz protease inhibitors in leaf and tuber protein extracts of control and SlCDI-expressing potato lines. (a) Tertiary structure model for SlCDI showing the position and spatial orientation of a 15-aa hydrophilic surface loop targeted for the production of anti-SlCDI polyclonal antibodies (in red) (see Experimental procedures for details). A sequence alignment generated with CLUSTAL W (Higgins et al., 1994) compares the loop sequence of SlCDI with the corresponding loop of selected potato Kunitz proteins including S9C11, a distant homologue down-regulated in leaves of SlCDI-expressing potato lines (see Goulet et al., 2009). Alphanumeric codes refer to NCBI GenBank accession numbers (http://www.ncbi.nlm.nih.gov). The ‘---’ signs indicate identity with corresponding residues in SlCDI. (b) Immunodetection of SlCDI and potato Kunitz inhibitors in tuber protein extracts using the anti-SlCDI loop polyclonal antibodies. (c) Immunodetection of SlCDI in leaf protein extracts of lines K and CD3A using the same antibodies. rSlCDI, nonglycosylated SlCDI expressed in Escherichia coli (Brunelle et al., 2005). Numbers on the left of panels (b) and (c) refer to molecular weight markers (kDa).

A 2-D immunoblot study was conducted to confirm in more detail the null impact of slcdi transgene expression on Kunitz proteins (Figure 5). The tuber proteins of lines SPCD4, SPCD7, CD3A and CD21A were first resolved in 2-D polyacrylamide slab gels as described above (Figure 5a), and then transferred onto nitrocellulose sheets for immunodetection with the anti-SlCDI loop polyclonal antibodies. About 25 proteins were immunodetected for the different lines (Figure 5b), in accordance with the occurrence of multi-member Kunitz protein families in potato (Heibges et al., 2003). Most of these proteins showed a neutral or basic pI and a molecular weight in the c. 1625-kDa range (Figure 5b) typical of Kunitz protease inhibitors (Suh et al., 1990, 1991; Kreft et al., 1997; Pouvreau et al., 2001; Kang et al., 2002; Heibges et al., 2003). Protein spots matching the immunodetected proteins in Coomassie blue-stained template gels (Figure 5a) were found at comparable levels in all lines tested (anova; > 0.05), thus confirming the null impact of the slcdi transgene on the overall pattern of proteins immunologically related to SlCDI. Several of these proteins were successfully identified by mass spectrometry (MS) (Figure 5a,b, spots A–M) and shown to belong to the Kunitz family of protease inhibitors, as expected (Tables 4 and S1). Except for abundant isoforms of the structurally distinct proteinase inhibitor II (spots A–D), all identified proteins included in their primary sequence a hydrophilic amino acid sequence highly similar to the SlCDI loop targeted by the chicken antibodies (Table 4). In agreement with total protein 2-DE analyses (see above), no specific signal could be detected for SlCDI in SlCDI-expressing potatoes, in sharp contrast with five endogenous CDI homologues easily detected in both control and CD lines (Figure 5 and Table 4, spots G–K). These observations confirming the absence of detectable SlCDI in tuber extracts of lines CD3A and CD21A, along with the identical protein profiles observed for patatin and Kunitz inhibitors in CD and SPCD lines, unequivocally indicated a negligible impact of slcdi transgene expression on the deposition pattern of potential protein allergens in tubers of SlCDI-expressing potatoes.

Figure 5.

 Electrophoretic separation and immunodetection of Kunitz protease inhibitors in a tuber protein extract of control line SPCD4. (a) Colloidal blue-stained proteins resolved by two-dimensional gel electrophoresis (2-DE). Protein bands and numbers on the left correspond to molecular weight markers (kDa). (b) Immunodetection of tomato cathepsin D inhibitor (SlCDI) and immunologically related (Kunitz-type) tuber proteins following 2-DE (left panel) or sodium dodecyl sulphate–polyacralyamide gel electrophoresis (right panel). Polyclonal antibodies raised against the 15-aa hydrophilic surface loop of SlCDI were used as primary antibodies (see Figure 4, above). Letters A to M refer to proteins identified by LC-MS/MS (see Tables 4 and S1). A 3–10 pH gradient was used for isoelectric focusing. Numbers on the left indicate the position of commercial molecular weight markers (kDa).

Table 4.   Proteins immunodetected in potato tuber protein extracts using chicken polyclonal antibodies directed against the 15-aa surface loop of tomato cathepsin D inhibitor (SlCDI) and potato Kunitz inhibitors (these proteins were detected at statistically similar levels in control and SlCDI-expressing lines*)
Putative proteinAccession no.Spot no.†Mr estim./calc.pI estim./calc.Sequence coverage (%)Hydrophilic loop motif
  1. *Protein sequences retrieved from the NCBI GenBank database (http://www.ncbi.nlm.nih.gov). Sequence details for LC-MS/MS identifications are given online in Table S1.

  2. †See arrows on Figure 5.

  3. ‡No SlCDI was detected in the potato lines analyzed.

  4. §(−) indicate identity with the corresponding residue in SlCDI.

SlCDI–‡GDVYLGKSPRSSAPC§
Serine proteinase inhibitor I (PSPI-21)P58514E18.4/16.38.5/5.018- - - - - - - - -N-D- - -
F21.3/16.38.4/5.018
Cathepsin D inhibitorS10721G21.3/20.88.0/7.856- - - - - - - - -N-D- - -
P58521H21.7/18.99.8/9.256
Aspartic proteinase inhibitorCAA45723I20.6/24.16.8/7.520- - - - - - - - -N-D- - -
AAB23206J21.5/24.66.5/6.343
P58518K22.3/20.59.8/8.660
Proteinase inhibitorBAA04152L20.7/23.57.6/6.812- - - - - - - - -N-D- - -
M21.0/23.59.8/6.751
Proteinase inhibitor II (Pin II)CAA27730A14.4/17.58.1/6.859–CN–YSANGAFICEG
B14.2/17.57.8/6.860
C14.3/17.57.6/6.850
D14.5/17.57.4/6.855

Discussion

High-throughput profiling technologies such as transcriptomics, proteomics and metabolomics present a number of advantages over classical compositional studies for the characterization of transgenic plant lines and foods derived from transgenic crops (Kuiper et al., 2003; Rischer and Oksman-Caldentey, 2006; Kok et al., 2008; Davies, 2009; Miki et al., 2009), including a dramatically improved resolution power allowing for the simultaneous analysis of hundreds or thousands of chemical components, or ‘characters’. From a regulatory or safety monitoring viewpoint, these technologies make it possible not only to assess a defined number of well-characterized nutritional indicators selected a priori, but also to detect uncharacterized (and even unknown) factors eventually affected in planta by the insertion or expression of the transgene. As importantly, they allow for the a posteriori identification and characterization of any unintended effect using appropriate gene sequencing, chromatographic or mass spectrometric methods, thereby simplifying data interpretation in terms of biological effects and significance. Proteomic approaches, in particular, appear of obvious interest to detect and monitor the unintended or pleiotropic effects of a transgene or recombinant protein in planta (Corpillo et al., 2004; Lehesranta et al., 2005; Ruebelt et al., 2006a,b,c; Lovegrove et al., 2009), given the central structural and functional roles assigned to proteins in plant cells and tissues. In the present study we followed an integrated, dual evaluation scheme to assess the impact of SlCDI on the chemical composition of tubers harvested from SlCDI-expressing transgenic potato lines, first involving a compositional analysis of key nutrients, and then the monitoring of tuber proteome profiles.

SlCDI constitutively expressed in transgenic lines of potato was shown recently to significantly change the overall proteome profile in leaves via an unelucidated proteome-wide alteration of the protein synthesis/turnover balance (Goulet et al., 2009). In sharp contrast with such dramatic effects on leaf protein catabolism, we observed here no impact for SlCDI on both the nutritional composition and general proteomic profile of SlCDI-expressing tubers. Small significant variations were noted for magnesium among SPCD and CD tuber extracts, but mean levels fell, in all lines, within the concentration range naturally observed for this element in tubers of commercial potato varieties (Leszczynski, 1989). In a similar way, a number of significant changes were detected at the proteome level, but no formal link could be made with the presence or expression of the tomato transgene in SlCDI-expressing lines. These observations suggest, overall, a negligible impact of SlCDI expression on tuber composition, and the onset of interline somaclonal or epigenetic variations among test and control lines possibly induced during genetic transformation or plantlet regeneration (Jongedijk et al., 1992; Kawchuk et al., 1997).

Complementary 2-DE analyses were carried out to visualize the deposition pattern of endogenous Kunitz protease inhibitors, abundant in tuber tissues (Pouvreau et al., 2001; Heibges et al., 2003) and identified earlier as potential allergens in potato tubers, along with multiple components of the patatin protein complex (Seppäläet al., 2001). Empirical analysis of the host-plant allergenic potential still represents a key step in the overall safety evaluation of transgenic food products (Kok et al., 2008), as there is yet no logical scheme to predict in silico, with a high degree of confidence, the impact of transgene integration or recombinant protein expression on the deposition pattern of endogenous allergens (FAO/WHO, 2001; Codex Alimentarus Commission, 2003; Kleter and Peijnenburg, 2004; Sanchez-Monge and Salcedo, 2005). Such an analysis was of particular relevance in the present case given the high structural homology between SlCDI and potato Kunitz inhibitors, readily recognized as belonging to the same protein family (Werner et al., 1993). In summary, our data indicated no significant difference among SlCDI-expressing and control tubers for the deposition pattern of Kunitz-type proteins in tubers, similar to patatin subunits showing unaltered levels in 2-D gels following colloidal Coomassie blue staining. Along with data obtained for the general proteome profile, these observations supported, overall, the conclusions of our compositional data showing substantial equivalence between comparator and SlCDI-expressing tubers. They also unequivocally confirmed the very low abundance of SlCDI in tuber tissues, which in turn likely accounted for the negligible effects of slcdi transgene expression on tuber composition.

It is not clear, at this point, why SlCDI could not be detected in tuber extracts of lines CD3A and CD21A despite the presence of slcdi transcripts at levels roughly comparable to those measured in leaves and the detection of biologically active SlCDI in leaf tissues of both two lines. One hypothesis could involve the variable (in)stability and turnover rate of SlCDI in cellular environments harbouring distinct proteolytic machineries, as discussed earlier for a number of recombinant proteins in plant systems (Benchabane et al., 2008). An alternative hypothesis would be the natural sink effect of highly abundant storage proteins on amino acid resources, inherently unfavourable to the biosynthesis and accumulation of transgene-encoded proteins (Schmidt and Herman, 2008). We recently established a link between SlCDI biosynthesis in potato leaves and the down-regulation of its endogenous homologue, S9C11 (Goulet et al., 2009). By contrast, SlCDI accumulation could have been compromised in tubers by highly expressed SlCDI homologues strongly attracting amino acids at the mRNA translation stage. Although protein accumulation in storage organs is influenced by mineral nutrition and some other environmental factors (Tabe et al., 2002), the relative distribution of proteins in storage tissues exhibits a very limited plasticity and is, for the most part, genetically determined (Schmidt and Herman, 2008). Studies reported the potential of low storage protein mutants (Tada et al., 2003) or transgenic host plants deficient in major storage proteins (Goossens et al., 1999; Kim et al., 2008; Schmidt and Herman, 2008) as efficient production platforms for foreign proteins, thereby suggesting both the natural intrinsic inertia of the storage organ proteome and the practical relevance of redirecting the host-plant endogenous protein biosynthesis machinery to the expression of foreign proteins for applications such as molecular farming or nutritional engineering involving recombinant protein deposition. This limited plasticity of the proteome in storage organs could represent, on the other hand, an advantage from a regulatory or safety viewpoint for biotechnological applications targeted to nonstorage organs, especially for those proteins such as SlCDI exhibiting high homology with endogenous (and potentially allergenic) storage proteins. Work is underway to test this idea, using a number of recombinant proteins and promoter sequences expressed in different organs of potato.

Experimental procedures

Potato lines and tuber processing

Seven transgenic lines of potato (Solanum tuberosum L., cv. Kennebec) expressing SlCDI under the control of the cauliflower mosaic virus 35S constitutive promoter were used as source material for the experiments (CD lines), along with two transgenic sister lines bearing the slcdi transgene with no promoter (SPCD lines) and in vitro clones of control parental line K used for genetic transformation (Brunelle et al., 2004). These 10 lines exhibit different levels of slcdi mRNA transcripts in leaves, ranging from null or barely detectable in untransformed and transgenic controls to moderate or high in transgenic test lines (see Figure 1a). At least three tubers of each line were sown and grown for four months in greenhouse at the Centre de recherche en horticulture of Laval University (Québec, QC, Canada). Three to five tubers of similar size (ca. 5 cm long) were harvested in each pot and used for chemical and molecular analyses. The tubers were washed, peeled, sliced and ground to a fine powder in liquid nitrogen. The powder was lyophilized to dryness and stored at −80 °C until use.

Insertion of the nptII and slcdi transgenes

Stable insertion of the nptII gene selection marker (see Brunelle et al., 2004) and slcdi transgenes was confirmed by PCR amplification from genomic DNA samples using appropriate oligonucleotide primers. Genomic DNA was extracted essentially as described by Hwang and Kim (2000). In brief, 250 mg of tuber powder was mixed with 300 μL of modified CTAB buffer (mCTAB) consisting of 2% (w/v) CTAB (Sigma-Aldrich, Oakville, ON, Canada), 1.4 M NaCl, 100 mM Tris-HCl (pH 8.0), 50 mM ethylenediamine tetraacetic acid (EDTA), 1% (w/v) PVP40 (Sigma-Aldrich) 0.5% (w/v) sodium bisulfide and 1% (v/v) 2-mercaptoethanol, and incubated with gentle agitation at 65 °C for 20 min. Three hundred microlitres of chloroform was added to each sample, the tubes were vortexed for 10 s at maximal speed, and centrifuged at 12 000 g for 5 min at room temperature. The upper phase (containing DNA) was transferred in a sterile tube, and DNA was precipitated with an equal volume of isopropanol for 5 min at room temperature. After centrifugation for 5 min at 12 000 g, the pellet was Speedvac-dried and resuspended in 50 μL of sterile TE buffer containing RNAse A (50 μg/mL), before measuring A260/A280 ratios using an UV spectrophotometer to monitor the quality of the DNA samples. The nptII and slcdi transgenes were detected using the following primers: 5′-ACTGA AGCGG GAAGG GACTG GCTGC TATTG and 3′-GATAC CGTAA AGCAC GAGGA AGCGG TCAG for nptII; and 5′-AAGGA TCCGT GCACA AAAGA TGGCT GCTTC TAAAC CTAAT CCAGT AC and 3′-AACCC GGGAA GCCGA GACTT TCTTG AAGTA GACCC CCAAG for slcdi. Potato endogenous CDI-encoding gene sequences were monitored as a control, using the following degenerated primers: 5′-AGGAT CGGTG MWTCT CCTMW ACCTA AKCCR GTAC and 3′-AGAAT TCTAG CTAGA CTTCC TTGAA GTTGA C, where M = A or C, W = A or T, K = G or T and R = G or A (Brunelle et al., 2005). After PCR, the amplicons were visualized by ethidium bromide staining following 1% (w/v) agarose gel electrophoresis.

Expression of the slcdi transgene

Expression of the slcdi transgene was monitored by RT-PCR, using tuber total RNA samples and oligonucleotide primers for slcdi amplification (see above). Total RNA was extracted from tuber powder using the Concert Plant RNA Reagent from Invitrogen Life Technologies (Burlington ON, Canada), following the supplier’s instructions. After precipitation in isopropanol, total RNA was washed in 75% (v/v) ethanol, dissolved in RNAse-free water, and proofchecked for quality by the monitoring of A260/A280 and A260/A230 absorbance ratios. cDNA populations were synthesized using the Superscript kit for cDNA synthesis (Invitrogen), and used as templates for PCR. The PCR amplicons were resolved by 1% (w/v) agarose gel electrophoresis and visualized by ethidium bromide staining. For quantitation, the gels were scanned using an Amersham Image Scanner digitalizer (GE Healthcare, Baie d’Urfé, QC, Canada) prior to computer processing and image analysis. Image analysis was carried out as described recently (Badri et al., 2009), using the Phoretix 2D Expression software, v. 2005 (NonLinear USA, Durham, NC, USA). Three biological (tuber) replicates were used for each line to allow for statistical assessments (anova; α = 0.05).

SlCDI inhibitory activity

SlCDI protease inhibitory activity in leaf and tuber protein extracts was assayed in vitro using the colorimetric protein substrate azocasein and bovine trypsin as a target protease (Michaud and Vrain, 1998). Soluble proteins were extracted from tuber or leaf powder in 50 mm Tris–HCl, pH 7.5, containing 1 mm EDTA (see Rivard et al., 2006). Trypsin activity was monitored in 50 mm Tris–HCl, pH 8.0, after adding 100 μL (∼100 μg protein) of tuber or leaf extract from either SlCDI-expressing or control plants (see Michaud and Vrain, 1998). An arbitrary value of 100% residual activity was assigned to trypsin activities measured with leaf or tuber extracts of control line K.

Immunodetection of SlCDI

SlCDI in tuber and leaf extracts was probed by immunodetection with IgY primary antibodies generated in chicken by AgriSera (Vännäs, Sweden), against the 15-aa peptide sequence ‘GDVYLGKSPRSSAPC’ forming a hydrophilic loop at the surface of the inhibitor (see below for SlCDI structure modelling). The proteins were resolved by 12% (w/v) or 15% (w/v) sodium dodecyl sulphate–polyacralyamide gel electrophoresis and electrotransferred onto nitrocellulose membranes for immunodetection. The membranes were saturated with 5% (w/v) skim milk in Tris-buffer saline containing 0.05% (v/v) Tween-20, washed three times for 10 min in Tris-buffer saline, and then incubated for 1 h at 22 °C with the anti-SlCDI IgY chicken antibodies. After washings in Tris buffer, the membranes were incubated for 1 h at 22 °C with goat anti-chicken secondary antibodies labelled with peroxidase, washed three times, and revealed by chemiluminescence using the Western Lightning immunodetection kit (Perkin Elmer Life Sciences, Whaltham, MA, USA).

Tertiary structure modelling

A tertiary structure model for SlCDI was inferred based on the spatial coordinates of soybean Kunitz trypsin inhibitor [Protein Data Bank (http://www.rcsb.org/pdb), Accession no. 1AVU) (Song and Suh, 1998). The model was constructed using the software Modeller, v. 6.2 (Sanchez and Sali, 2000), and tested for energy distribution and stereochemical quality using the in-built Energy command of Modeller and the Procheck program (Laskowski et al., 1993), respectively. SlCDI was visualized with the RasMol software (Sayle and Milner-White, 1995).

Compositional analysis

The chemical composition of potato tubers was assessed by a two-step approach first involving a proximate analysis to compare relative levels of major constituents such as total solids, total proteins, total lipids, total carbohydrates, ash and crude fibres; followed by the targeted analysis of specific nutritive and toxic compounds typically assessed for potato tubers (AOAC, 2007a; OECD, 2002). All measures were repeated three times with samples from distinct tubers. Data for each compound were analyzed using F statistics obtained from a one-way anova, with plant line as the main effect. An error rate of 5% was used to interpret the results.

Proximate analysis

Total solids were determined by the estimation of tuber powder weight loss upon drying in an oven at 100 °C (AOAC, 2007a). Total proteins were determined from total nitrogen quantitation by the Kjeldahl method (AOAC, 2007b), using the multiplication factor ‘N X 6.25’ for crude protein content estimation. Fat content was determined by the Soxhlet extraction method (AOAC, 2007c) after tuber sample extraction in petroleum ether and extract evaporation, drying and weighing. Ash content was estimated by ignition to a constant weight in an electric furnace at 550 °C, and subsequent quantification of the residues (AOAC, 2007d). Total carbohydrates were estimated by difference using the following equation: carbohydrates (%) = 100% − (% protein + % fat + % ash + % moisture) (USDA, 1973). The content in crude fibres was determined following sample digestion in 1.25% (v/v) sulfuric acid and 1.25% (v/v) sodium hydroxide (AOAC, 2007e). Crude fibre content was estimated as residue weight loss after drying, weighting and ignition at 550 °C.

Ascorbic acid (vitamin C)

Ascorbic acid was extracted from lyophilized tuber samples and quantitated by high performance liquid chromatography (HPLC) under a UV detector as described by Spares et al. (1990). l-Ascorbic acid (Sigma-Aldrich) was used as a reference standard.

Starch and soluble sugars

Soluble sugars in tuber powder samples were extracted for 30 min at 80 °C in 80% (v/v) ethanol. After centrifugation for 5 min at 3000 g, the supernatant was recovered for glucose, fructose and sucrose enzymatic quantitation in the presence of ATP and NAD (Stitt et al., 1989). The pellet was resuspended in 0.2 m KOH for starch quantitation. After a 30 min incubation at 100 °C, the pH was adjusted to 5.5 with 1 m acetic acid. The starch fraction was hydrolyzed with amylase (EC 3.2.1.1) and amyloglucosidase (EC 3.2.1.3), and assayed by enzymatic determination of released glucose (Stitt et al., 1989).

Soluble proteins

Soluble proteins were assayed according to Bradford (1976) with chicken egg white albumin as a standard, after protein extraction from lyophilized tuber powder in 50 mm Tris–HCl, pH 7.5, containing 1 mm EDTA and 1 mm phenylmethylsulfonyl fluoride (PMSF) (see Rivard et al., 2006).

Minerals

Potassium, magnesium and copper were analyzed by emission spectrometry as described (AOAC, 2007f). The tuber samples were dried, precherred and ashed overnight in a muffle furnace at 550 °C. Ashed samples were treated with hydrochloric acid, taken to dryness, and dissolved in 5% (v/v) hydrochloric acid. The amount of each element was determined at the appropriate wavelength using inductively coupled plasma, by comparing emission of the unknown sample with the emission of standard solutions.

Amino acids

Amino acids were extracted in methanol : chloroform : H2O 12 : 5 : 2, derivatized and analyzed by gas chromatography–MS according to Pearson and Nasholm (2001). Internal standards (Sigma-Aldrich) were used for quantification.

Glycoalkaloids

Total glycoalkaloids were extracted in acetic acid according to Sotelo and Serrano (2000), and quantified by HPLC using the Waters system controller 600-MS, a Waters 431 pump and the UV photodiode array detector 996 (Waters, Mississauga ON, Canada). A C-18 column was used for chromatographic separation, with a water : acetonitrile (50 : 50 + NH4K2HPO4) mobile phase. The solvent flow rate and UV absorbance detector were adjusted at 1.0 mL/min and 280 nm, respectively.

Protein extraction for two-dimensional gel electrophoresis

Proteins for proteomic analyses were extracted according to Espen et al. (1999), with some modifications. In brief, 100 mg of lyophilized tuber powder was extracted in 750 μL of extraction buffer (50 mm Tris–HCl, pH 7.6, 0.25 m sucrose, 10 mm MgCl2, 1 mm EDTA, 10 mm phenylmethylsulfonyl fluoride) for 30 min at 4 °C with gentle stirring, in the presence of 1% (w/v) polyvinylpolypyrrolidone. After centrifugation at 20 000 g for 30 min at 4 °C, the supernatant was recovered for protein precipitation in methanol/chloroform according to Wessel and Flügge (1984). In brief, 150 μL of supernatant was mixed with 600 μL of methanol and 150 μL of chloroform. After a 2-h incubation at −20 °C, 450 μL of double-distilled water was added, and the whole mixture was mixed by inversion. The samples were then centrifuged at 14 000 g for 30 min at 4 °C, and the pellets were washed with 600 μL of methanol before being recentrifuged. The precipitates were Speedvac-dried and solubilized in 0.5% IPG buffer (pH 3–10 or pH 4–7) (GE Healthcare) containing 8 m urea, 2% (w/v) 3-[(3-cholamidopropyl)-dimethylammonio]-1 propanesulfonate (CHAPS) and 60 mm dithiothreitol. The mixtures were vortexed extensively and sonicated for 1 h. Protein content in the samples was determined according to Bradford (1976), with chicken egg white albumin as a protein standard.

Two-dimensional gel electrophoresis and image analysis

Two-dimensional gel electrophoresis first involved IEF (1st dimension), followed by standard 15% (w/v) sodium dodecyl sulphate (SDS)-polyacrylamide gel electrophoresis (PAGE) (2nd dimension). IEF was performed at 16 °C along a 4 to 7 (or 3 to 10) pH gradient in 13-cm Immobiline DryStrip gel strips (GE Healthcare), with 200 μg of leaf protein diluted in 250 μL of electrophoretic sample buffer (see above). Proteins were applied to the strips and resolved using an IPGphor apparatus (GE Healthcare). The program for IEF separation involved the following sequential steps: rehydration for 12 h at 30 V; for 1 h at 100 V; for 1 h at 500 V; for 1 h at 1000 V; for 1 h at 5000 V; and at 8000 V to reach 24 000 Vh. Following IEF, the strips were incubated for 25 min in a 50 mm Tris–HCl equilibration buffer, pH 8.8, containing 6 m urea, 30% (v/v) glycerol, 2% (w/v) SDS, 10 mg/mL dithiothreitol and traces of bromophenol blue, and used immediately for the second dimension. SDS-PAGE was performed in 1-mm thick polyacrylamide slab gels according to Laemmli (1970).

After migration, the gels were either fixed overnight in water containing 10% (v/v) acetic acid and 40% (v/v) methanol for subsequent protein staining; or electroblotted onto a nitrocellulose membrane without prior fixation and processed for immunodetection as described above. After fixation, the gels for staining were washed three times in water for 30 min, and the proteins were visualized by staining with the colloidal Coomassie blue GelCode reagent (Pierce, Rockford IL, USA). Gel images were digitalized and subjected to computer processing and image analysis, with three biological replicates for each plant line to allow for statistical assessments. Image analysis was carried out using the Phoretix 2D Expression software, v. 2005 (NonLinear USA). Spot detection and background subtraction were performed following the supplier’s recommendations. The gel containing the highest number of protein spots was identified, and used as a reference for protein spot matching. For each set of three gels (replicates), an average virtual gel was constructed that included protein spots found on at least two gels. Spot matching was performed with the average gels, and spots showing significantly variable intensities were identified based on a one-way anova using an error rate of 5%.

Tryptic digests and protein identification

Protein spots for identification were cut out from the gels manually, washed for 15 min in double-distilled water, and incubated for 15 min in 250 μL of water/acetonitrile (1 : 1 v/v). After additional washes in acetonitrile and 100 mm NH4-HCO3, the proteins were reduced with dithiothreitol, alkylated with iodoacetamide, and digested to peptides overnight at 37 °C with MS grade Trypsin Gold (Promega Corporation, Madison, WI, USA). The tryptic digests were sent to the Institute for Biomolecular Design (Edmonton, AB, Canada) or to the Génome Québec Innovation Centre at McGill University (Montréal, QC, Canada) for liquid chromatography–MS/MS with an electrospray ionization-ion trap mass spectrometer. Protein identification was performed using the mascot algorithm (http://www.matrixscience.com) with the following criteria for positive identification: a mascot score significant at < 0.05, at least two matched peptides, and a minimal sequence coverage of 10%. The following search parameters were used for database mining: Type of search = MS/MS ion search; Fixed modifications = carbamidomethyl (Cys); Variable modifications = oxidation (Met); Mass values = monoisotopic; Peptide mass tolerance ± 1.5 Da; Fragment mass tolerance = ±0.8 Da; Maximum missed cleavages = 1.

Acknowledgements

This work was supported by the Natural Science and Engineering Research Council (NSERC) of Canada. M. Khalf was the recipient of a Ph.D. scholarship from the Government of Egypt. C. Goulet was the recipient of an NSERC Ph.D. scholarship. J. Vorster was the recipient of a postdoctoral fellowship from the National Research Foundation of South Africa.

Ancillary