A deletion and mutagenesis study was performed on the mitochondrial presequence of the β-subunit of the F1-ATP synthase from Nicotiana plumbaginifolia linked to the green fluorescent protein (GFP). The various constructs were tested in vivo by transient expression in tobacco protoplasts. GFP distribution in transformed cells was analysed in situ by confocal microscopy, and in vitro in subcellular fractions by Western blotting. Despite its being highly conserved in different species, deletion of the C-terminal region (residues 48–54) of the presequence did not affect mitochondrial import. Deletion of the conserved residues 40–47 and the less conserved intermediate region (residues 18–39) resulted in 60% reduction in GFP import, whereas mutation of conserved residues within these regions had little effect. Further shortening of the presequence progressively reduced import, with the construct retaining the predicted N-terminal amphiphilic α-helix (residues 1–12) being unable to mediate mitochondrial import. However, point mutation showed that this last region plays an important role through its basic residues and amphiphilicity, but also through its hydrophobic residues. Replacing Arg4 and Arg5 by alanine residues and shifting the Arg5 and Leu6 (in order to disturb amphiphilicity) resulted in reduction of the presequence import efficiency. The most dramatic effects were seen with single or double mutations of the four Leu residues (positions 5, 6, 10 and 11), which resulted in marked reduction or abolition of GFP import, respectively. We conclude that the N-terminal helical structure of the presequence is necessary but not sufficient for efficient mitochondrial import, and that its hydrophobic residues play an essential role in in vivo mitochondrial targeting.
Although mitochondria possess their own genome, most mitochondrial proteins are encoded by the nucleus, synthesized as precursors in the cytosol, then imported into the mitochondria. Precursors contain an N-terminal extension, the presequence, involved in targeting. Precursors pass through the mitochondrial double membrane via the translocases of the outer and inner membranes (TOM and TIM, respectively), and their presequence is cleaved off during import, or once inside the mitochondria, by a specific mitochondrial processing peptidase (MPP) (for review see Glaser et al., 1998; Neupert, 1997). Presequences do not share a common primary structure, but have a common secondary structure and are enriched in basic and hydrophobic residues. Arginine, alanine, leucine and serine are present at a higher frequency in the presequence, and the N-terminal region is predicted to fold into an amphiphilic α-helix (Schneider et al., 1998; von Heijne, 1986).
The potential of the N-terminal part of the presequence to fold into an amphiphilic α-helix in a phospholipid environment has been confirmed experimentally (Roise et al., 1988), and the importance of this helical structure has been assessed in vivo (Lemire et al., 1989). Moreover, the fact that it is required for mitochondrial import (Wang and Weiner, 1993) but is not the only structural property of a mitochondrial presequence has been demonstrated in vitro (Tanudji et al., 1999). The roles of residues in the presequence have been assessed by mutagenesis studies, which show that the positively charged residues are important for import (Vassarotti et al., 1987). Amphiphilicity and positive charges are the two main features reported for presequences. Substitution of all arginine residues in the N-terminal segment in rat liver aldehyde dehydrogenase drastically reduced in vitro import competence, while mitochondrial import was maintained in the presence of a more stable α-helix (Hammen et al., 1996; Heard and Weiner, 1998).
However, a recent NMR structural analysis of Tom20 complexed with a presequence peptide has revealed the importance of hydrophobic residues. The presequence was shown to bind in a helical conformation to an apolar groove in Tom20, with binding depending mainly on hydrophobic, rather than hydrophilic, interactions (Abe et al., 2000). However, the importance of hydrophobic residues has still to be assessed at the transport level.
The import mechanism has been intensively studied in recent years in yeast and Neurospora crassa, leading to the description of the TOM and TIM complexes (for review see Koehler, 2000; Neupert, 1997). Although the import mechanism appears to be well conserved in different organisms, the plant kingdom shows several particularities in terms of the proteins involved in the TIM and TOM complexes, the MPP, and the primary and secondary structures of the presequences (for review see Glaser et al., 1998). Plant mitochondrial presequences are longer and have a higher serine content than those in yeast (Schneider et al., 1998), but there is little functional information about the structural elements involved in import. This is especially true in the case of plants in which the presence of chloroplasts is suggested to have led to the establishment, during evolution, of a more discriminating system (Macasev et al., 2000). For instance, chloroplast transit peptides also contain a basic residue-rich region, and some can target reporter proteins in yeast mitochondria. A consequence of this uncertainty is that no good software program exists to predict plant mitochondrial presequences with high accuracy. For instance, only 349 genes of the genomic sequence of Arabidopsis thaliana were predicted with accuracy to encode mitochondrial proteins (Arabidopsis Genome Initiative, 2000). Tanudji et al. (1999) carried out an in vitro mutagenesis analysis of soybean alternative oxidase, and showed that an amphiphilic structure was not sufficient for import of this precursor and that positive residues along the entire presequence play an important role in import, but that none was essential for import. However, like most studies on the import of modified presequences, this study was carried out in vitro, and it has been shown that in vitro import into isolated mitochondria leads, in some cases, to artefactual results (de Castro Silva-Filho et al., 1997; Ni et al., 1999).
Here we report an in vivo mutagenesis analysis of a plant presequence using the presequence (54 residues) of the β-subunit of the F1-ATPsynthase (F1β) from Nicotiana plumbaginifolia as a model. A previous semiquantitative analysis showed that C-terminal-shortened presequences are still able to import a reporter protein (Chaumont et al., 1994). However, no deletion of the predicted amphiphilic α-helix and no site mutagenesis were carried out. In the present study, we show that the predicted N-terminal amphiphilic α-helix between residues 1 and 12 does not, by itself, allow mitochondrial import of a reporter protein. The importance of this region, notably of its leucine residues, is demonstrated.
Presequence of the F1-ATPase β-subunit
Based on the Garnier–Robson (Garnier et al., 1996), Geourjon (Geourjon and Deléage, 1995) and Chou–Fasman (Chou and Fasman, 1978) methods of secondary structure prediction, and the Kyte–Doolittle (Kyte and Doolittle, 1982) and Eisenberg (Eisenberg et al., 1984) methods of hydropathic moment calculation, the N. plumbaginifolia F1β presequence was predicted to fold into an amphiphilic α-helix between residues 1 and 12, with a helical structure extending to residue 15. This helix is followed by a short β-sheet structure between residues 21 and 24, and a second helical amphiphilic structure between residues 46 and 57 (Figure 1a).
To detect conserved and potentially important regions, sequence comparison was performed on several plant F1β presequences (Figure 1b). Two conserved regions corresponding to the two predicted helices could be seen, the first between residues 1 and 17 and the second between residues 40 and 61, with the intervening region being poorly conserved.
The F1β presequence imports the green fluorescent protein into protoplast mitochondria
The N. plumbaginifolia F1β presequence (54 residues) and 12 residues of the mature protein were placed in front of the gfp gene under the control of a strong plant transcription promoter, giving the construct β-GFP. A GFP construct without the presequence was used as a control. Both plasmids were introduced into Nicotiana tabacum protoplasts (Figure 2a), and GFP and chloroplast autofluorescence were examined by confocal microscopy.
With the GFP control, fluorescence was localized in the nucleoplasm and the cytosol surrounding the large central vacuole (Figure 2b,c). The presence of GFP in the nucleoplasm has been observed previously, and was explained by its small size allowing passive diffusion into the nucleus (Grebenok et al., 1997).
When GFP was associated with the presequence, the fluorescence was localized in small particles of 1–2 µm throughout the cytoplasm, while the nucleoplasm and cytosol showed no fluorescence (Figure 2d,e). When these protoplasts were subjected to immunolocalization of mitochondrial lipoamide dehydrogenase, the images corresponding to GFP (Figure 2f) and lipoamide dehydrogenase (Figure 2g) merged (Figure 2h) to show co-localization of GFP and the mitochondrial marker, indicating mitochondrial import of GFP.
Mitochondrial import by F1β presequence deletions
Progressive C-terminal deletion of the F1β presequence was carried out to identify domains essential for mitochondrial import and to determine whether the predicted amphiphilic α-helix between residues 1 and 12 was sufficient for mitochondrial import. GFP constructs containing the N-terminal 11, 14, 17, 22, 30, 42, 47, 53 or 60 residues of the F1β precursor (Figure 3a) were electroporated into protoplasts, then GFP import into mitochondria was monitored by confocal microscopy in at least 50 GFP-expressing protoplasts, and quantified as the ratio of the number of protoplasts in which all GFP was imported into mitochondria relative to the total number of GFP-expressing protoplasts. To rule out a confusion between mitochondria and possible GFP aggregates in the cytosol, co-localization of mitochondria labelled with Mitotracker and GFP was performed for some constructs (Figure 3c). GFP import was also quantified by Western blot analysis of crude mitochondrial and cytosolic fractions from transformed protoplasts. As some mitochondria were fragmented during homogenization and thus released GFP into the cytosolic fraction, immunodetection of lipoamide dehydrogenase was used as a mitochondrial marker (de Castro Silva-Filho et al., 1996). With the full β-GFP construct, the mitochondrial/cytosolic GFP ratio was similar to that seen for lipoamide dehydrogenase, while with the control GFP construct GFP was found only in the cytosolic fraction (Figure 4a). With constructs βΔ53-GFP and βΔ47-GFP, mitochondrial import of GFP was similar to that seen with the full β-GFP (Figure 4c). In situ observation (Figure 3b) confirmed that, with these three constructs, all the GFP was imported into mitochondria in more than 80% of the transformed protoplasts. With construct βΔ42-GFP, import was reduced to 89% (Figure 4c) and, although this level of import was not statistically different from that seen with β-GFP (Student’s test, α = 0.05), a lower import performance was clearly observed in situ as, in 44% of the transformed cells, GFP was found in the cytosol and the nucleus as well as in the mitochondria (Figure 3b). With βΔ30-GFP import decreased to 72% compared to that seen with β-GFP. Shortening the presequence to 22 and 17 residues further decreased import to 45 and 41%, respectively. This was confirmed by in situ analysis which showed that 55 and 100% of transformed cells, respectively, showed accumulation of GFP in the cytosol as well as in the mitochondria. This is clearly shown by the observation that GFP fluorescence superimposes on Mitotracker labelled mitochondria but also extends to the cytosol (shown for βΔ22-GFP, Figure 3c). Constructs with only the first 14 or 11 residues led to all the GFP accumulating in the cytosol in all transformed protoplasts, as seen in the control GFP (Figure 3b). With βΔ14, import efficiency fell below 20% (Figure 4c) indicating that, although it contained the predicted amphiphilic α-helix, this construct failed to direct GFP import into mitochondria.
Western blotting using anti-GFP antibodies was performed to quantify import, but also made it possible to follow cleavage of the GFP precursor (Figure 4b). With β-GFP, a major band was seen at the expected size for the mature protein, while an additional band corresponded to the precursor. For the shorter constructs which lacked the processing site, the main band was seen at the expected size of the precursor. However, with constructs βΔ30-GFP to βΔ53-GFP an additional band was seen, suggesting the presence of a cryptic cleavage site within the presequence. Cleavage probably occurred in vivo rather than in vitro after homogenization, as the same band was seen by Western blotting of a protein extract obtained after phenol treatment of the plant material before homogenization (data not shown).
In summary, transient expression analysis leads to the conclusion that the predicted N-terminal amphiphilic α-helix by itself cannot efficiently address GFP into mitochondria.
Expression of F1β presequence deletions in stable transformants
We noticed (for example with the β-GFP construct) that, while the majority of transformed cells contained GFP only in the mitochondria, some also had GFP in the cytosol and nucleoplasm (1–12% depending on the experiment). To rule out a possible saturation effect due to the transient expression system and, consequently, an underestimation of the import capacity of the various constructs, the constructs GFP, βΔ14-GFP, βΔ30-GFP, βΔ47-GFP and β-GFP were expressed stably in N. tabacum. Transgenic plants were produced using an Agrobacterium tumefaciens T-DNA-derived vector. With the GFP construct, GFP fluorescence was seen in the cytosol and nucleoplasm, whereas with the β-GFP construct, fluorescence was restricted to mitochondria. With the other three constructs (βΔ47-GFP, βΔ30-GFP and βΔ14-GFP), the fluorescence in the cytosol and nucleoplasm increased as the presequence was progressively shortened (data not shown), as seen with transient expression. The large quantity of material available using transgenic plants allowed us to further purify mitochondria by Percoll gradient centrifugation, and the subsequent quantification by Western blot analysis of subcellular fractions confirmed that import efficiency in stable transformants was similar to that seen using transient expression (Figure 4d).
Expression of F1β presequence deletions in yeast
The observation that sequences C-terminal to the predicted N-terminal amphiphilic α-helix were essential for efficient mitochondrial targeting in plants contradicted the view that this structure is generally sufficient in yeast. This prompted us to test some of the GFP constructs in the yeast Saccharomyces cerevisiae after adapting the expression system. Yeast GFP expression was analysed by confocal microscopy of transformed cells and Western blotting of subcellular fractions (Figure 4e and 5d), using the mitochondrial F1-ATPase α- and β-subunits as the mitochondrial marker.
With the construct Y-β-GFP, GFP and the mitochondrial marker were clearly co-localized (Figure 5a–c). The presence of the 47-residue presequence resulted in complete GFP import in yeast (Figure 4e and 5d), as in tobacco. Shortening the presequence to the first 30 residues reduced GFP mitochondrial import by 50%, while shortening it to only 14 residues resulted in a further decrease, but not to the level seen in tobacco protoplasts, indicating that this short presequence is about twice as efficient in yeast (Figure 4e); this result was confirmed in situ (Figure 5d) as GFP was still detected in yeast mitochondria, despite the cytosolic fluorescence. We therefore conclude that the predicted N-terminal amphiphilic α-helix kept some capability to address the reporter protein into yeast mitochondria, while this was less the case for plant mitochondria.
Site-directed mutagenesis analysis of the presequence
A site-directed mutagenesis approach was then used to evaluate separately the requirement for the N-terminal region and the role of the internal region. We produced 48 point mutants of the presequence involving 27 different residues, either alone or in combinations of two or three. For each mutant, the mitochondrial import efficiency of the mutated presequences was calculated as the fraction of the total number of GFP-expressing cells (at least 50 cells) that only showed mitochondrial labelling. Four categories, I (>80%), II (>50%), III (>20%), and IV (≤20%), were defined (Table 1).
Table 1. Import efficiency of mutant F1β-GFP constructs
Constructs expressing the mutated presequence linked to GFP were distributed into four categories according to the fraction of total number of GFP-expressing cells that showed only mitochondrial labelling (according to in situ analysis): I, >80%; II, >50%; III, >20%; IV, ≤20% (see text for details).
Alignment of several F1β presequences (Figure 1b) showed that the internal region (residues 18–47) is less conserved (residues 40–47 excepted), but contains a few residues which are more conserved, these being Pro31, Arg36, Arg40, Pro43 and Leu47. When various mutations of these conserved residues were produced using degenerated primers, reduced import was seen with L47C, but not with L47F, L47R or L47D, or any other of these mutations (Table 1); even two or three mutations did not modify targeting efficiency, with the exception of the A37V-L47H-V52I mutant which modified the hydrophobic moment of the presequence, moderately reducing import. The rest of the mutations were localized in the N-terminal region.
The possibility for interrupting the predicted α-helix without also modifying charged or hydrophobic residues was limited. For instance, construct L6P had an effect on import, but this mutation also reduced the hydrophobicity (Figure 6). However, the S9G mutant and the double mutant A8G-S9G, both predicted to reduce the length of the helix, had no effect on import (Table 1).
When the amphiphilicity of the presequence was reduced by inverting Arg5 and Leu6 or by replacing Arg4 with Val, microscopy showed that both mutations moderately reduced import (Figure 6).
As regards the positively charged residues, mutation of an arginine residue resulted in reduced import in the case of R4V and R17G, but not R12A, while replacement of Arg17 by His in the R17H-T56S mutant had no effect on the targeting efficiency of the presequence. Introduction of a negative charge (R4E) and the double or triple removal of an arginine (positions 4, 5 and 12) did not further reduce GFP import (Figure 6), indicating that, although the basic residues in the N-terminal part of the presequence were important for targeting, their substitution still allowed complete import in more than 50% of the transformed cells.
A dramatic effect was seen when the hydrophobic residue leucine was replaced by the hydrophilic residue glutamine. Constructs L6Q, L7Q, L10Q and L11Q all showed reduced import, and the double mutants L6Q-L7Q and L10Q-L11Q showed no GFP import into mitochondria, indicating that these leucine residues are essential for targeting (Figure 6).
The characterization of deletions or mutations of a mitochondrial presequence in an in vitro import system can lead to artefactual results (de Castro Silva-Filho et al., 1997; Ni et al., 1999). We therefore dropped this approach for in vivo analysis. As the production of many stable transformants is time consuming, transient expression in electroporated protoplasts appeared to offer a very convenient alternative. We observed that, even with the full presequence, some protoplasts (usually less than 5%) showed reduced import. As this did not necessarily occur in those cells that had the highest level of GFP, it does not seem to be related to gene overload. Alternatively, it could result from damage to the import machinery caused in some cells by electroporation. Moreover, this transient expression system was validated by the similar data obtained for five constructs expressed in stable transgenic plants.
Comparison of F1β presequences from different plant species led to their rough division into three regions, consisting of two highly conserved N- and C-terminal regions separated by a less-conserved region between residues 18 and 39 in N. plumbaginifolia. Progressive deletion and site-directed mutation of the presequence both led to the conclusion that no targeting information was located within the C-terminal region, which contains the cleavage site.
The fact that processing of presequences of matrix proteins is not conditional for their import (Zwizinski and Neupert, 1983) was confirmed, as the mutation containing the first 47 residues, but lacking the cleavage site located between residues 54 and 55, allowed all the GFP to be imported into mitochondria. However, a cryptic cleavage site was detected N-terminal to the normal cleavage site and, on the basis of Western blot analysis, was tentatively located between residues 22 and 30.
Deletions within the internal region (residue 18–47) indicated that it contributes significantly to the effectiveness of import. However, single or combined mutations of conserved residues had no marked effect on import. It is well established that the positively charged residues are essential and that their gradual elimination or mutation reduces the mitochondrial import efficacy of the presequence (Hammen et al., 1996; Ni et al., 1999; Tanudji et al., 1999). The reduced import seen with gradual C-terminal deletion of the F1β presequence supports this relationship, as most deletions resulted in the elimination of one or two positively charged residues. The regular distribution of positive charges prevented a marked change in the net charge balance, and might explain why single basic residue mutations did not significantly affect mitochondrial import. This shows the usefulness of combining deletion and single-residue mutation analysis.
The F1β presequence, like most mitochondrial presequences, is predicted to fold into an N-terminal amphiphilic α-helix between residues 1 and 12, with the helix extending to residue 15. The role of the α-helical structure is controversial, as it is absent in some presequences (Allison and Schatz, 1986; Roise et al., 1988) whereas in other cases there is evidence that it is required (Abe et al., 2000; Heard and Weiner, 1998; Wang and Weiner, 1993). Its role may be to stabilize the interaction between the positive and hydrophobic residues and the import machinery. In this work we tested three deletions, βΔ17, βΔ14 and βΔ11, in this N-terminal part of the presequence, and showed that import was dramatically reduced by 59, 81 and 94%, respectively, indicating that the residues between 11 and 17 play an important role in GFP import and that the N-terminal part of the presequence containing the putative amphiphilic α-helix (residues 1–12) is not sufficient to direct the reporter protein to mitochondria efficiently. Although not sufficient, the region encompassing the predicted amphiphilic α-helix could be necessary, and this was tested by modifying its amphiphilicity, putative helical structure, and basic and hydrophobic residues. Attempts to disrupt the α-helix (residues 1–15) by introducing glycine and proline residues (S9G, A8G-S9G and L6P) showed that only the proline mutation resulted in modified import efficiency; however, in this mutation the leucine residue was subsequently shown to be essential. Thus the helical structure of the presequence may not be essential for import, but proteins interacting in vivo with the presequence may stabilize and facilitate its helical conformation, even when mutated.
In this study the two mutants (R4V and R5L-L6R) that were generated in order to disturb the amphiphilicity of the N-terminal part of the presequence both caused reduced mitochondrial import of GFP. This supports the suggestion that an amphiphilic structure in a mitochondrial presequence plays an important role, as previously indicated (e.g. Lemire et al., 1989).
Mutation of the arginine residues in this region showed that some were important for efficient import, with import being reduced with the R17G and R4A-R5A mutants, but not with R17H-T56S; moreover, the R12A mutant did not show reduced import and its addition to the R4A-R5A mutant did not cause any additional decrease. This suggests that not all basic residues are equally important, and that some are more essential because they are possibly involved in a recognition or transmembrane translocation step.
The most dramatic result was the drastic reduction in import seen when hydrophobic residues were replaced by hydrophilic residues (L6Q, L7Q, L10Q, L11Q, L6Q-L7Q and L10Q-L11Q). Each of these mutations had a strong effect (<50% import efficiency), while the two double mutants completely abolished import, showing that these hydrophobic residues are essential for import competence of the precursor. Recently, Abe et al. (2000) produced an NMR structure for the rat import receptor Tom20, complexed with a presequence peptide, and showed that presequence binding to Tom20 is mediated by hydrophobic interactions and the presence of a helical structure. They also demonstrated that the ionic interactions (mainly involving arginines) are not essential for binding, and are therefore probably involved in other steps of import, such as the movement of the presequence through the inner mitochondrial membrane along the potential gradient. Our data therefore provide a clear in vivo demonstration of the importance of the four hydrophobic residues located in the N-terminal part of the F1β precursor. In the N. plumbaginifolia presequence these are all leucine residues, but in other plant species some of these leucines are replaced by other hydrophobic residues (Figure 1b), suggesting that it is the presence of hydrophobic residues rather than leucine residues per se that is important. The sequence of the plant Tom20 is poorly conserved compared to its animal and fungal counterparts, and is predicted to be anchored in the membrane by a C-terminal, rather than an N-terminal, membrane span (Heins and Schmitz, 1996). Nevertheless, our data showing the importance of hydrophobic residues suggest that a similar mechanism exists in all these organisms. This is also supported by the efficacy of the N. plumbaginifolia F1β presequence in GFP import into yeast mitochondria, the only significant difference compared with import into plant mitochondria being that, with the construct βΔ14-GFP, the reporter protein was more efficiently imported in yeast. This observation might further support the hypothesis that the plant import machinery has to be more discriminating because of the presence of chloroplasts (Macasev et al., 2000), and thus requires longer presequences.
Plasmids and DNA constructs
The plant transient expression vectors were derived from pTZ-19U (Stratagene, La Jolla, CA, USA). The β-GFP plasmid contained the plant enhanced transcription promoter of the plasma membrane H+-ATPase isoform 4 gene (pma4) (Zhao et al., 1999) and the nopaline synthase terminator controlling the expression of the F1-ATPase (subunit presequence from Nicotiana plumbaginifolia (54 residues of presequence and 12 residues of mature protein) linked to the green fluorescent protein (gfp) gene. The GFP plasmid was identical to the β-GFP plasmid, except that it did not contain the presequence. To prepare deletions within the presequence, a primer located around the BglII restriction site (upstream of the initiating ATG codon) and downstream primers provided with a KpnI site were used to obtain, by PCR, fragments corresponding to the first 11, 14, 17, 22, 30, 42, 47, 53 or 60 residues of the F1β precursor. The amplified fragments were digested with KpnI and BglII and cloned into β-GFP by replacement of the wild-type presequence, giving the βΔ11-GFP to βΔ60-GFP plasmids. Site-directed mutagenesis of the presequence was performed by triple PCR, using a 5′ primer located around the BglII site, a 3′ primer at the beginning of the gfp sequence, and pairs of internal primers containing the modified nucleotides; in some cases degenerated primers were used to increase the number of mutations. Amplified fragments were introduced as above. All constructs were checked by sequencing.
For stable plant transformation, the genes for gfp or gfp linked to the first 14, 30, 47 or 66 codons of the presequence were released from their vector by HindIII/SacI digestion and transferred into pBI-101 (Clontech, Palo Alto, CA, USA), an Agrobacterium tumefaciens T-DNA-derived vector.
The yeast plasmids were constructed from the cp(PMA1)pma2 plasmid (de Kerchove d'Exaerde et al., 1995) by removing pma2 cDNA using BamHI–HindIII (blunt-ended) and replacing it with DNA fragments coding for gfp or gfp linked to the first 14, 30, 47 or 66 amino acids of the F1β precursor, obtained by SacI (blunt-ended) and BglII digestion of the above plant transient expression plasmids. The resulting plasmids (Y-GFP, Y-βΔ14-GFP, Y-βΔ30-GFP, Y-βΔ47-GFP and Y-β-GFP) were introduced by heat shock into the W303-1B Saccharomyces cerevisiae strain (Thomas and Rothstein, 1989).
Transient expression and plant transformation
Nicotiana tabacum leaf protoplasts were purified and transformed as described by Lukaszewicz et al. (1998). Stable transgenic N. tabacum plants were produced using A. tumefaciens as described by Horsch et al. (1986). Transformants were grown on kanamycin medium (100 mg l−1), and plants expressing GFP were selected by epifluorescence microscopy.
Confocal and epifluorescence microscopic analysis
Confocal microscopy was performed using a Bio-Rad (Hercules, CA, USA) MRC-1024 laser scanning confocal imaging system. For GFP detection, excitation was at 488 nm and detection between 506 and 538 nm. For Texas Red detection, excitation was at 568 nm and detection between 589 and 621 nm. Chloroplast autofluorescence was detected between 664 and 696 nm with an excitation at 488 nm. Mitotracker Red CM-H2XRos (Molecular Probes, Eugene, OR, USA) was detected between 589 and 621 nm with an excitation at 568 nm.
Epifluorescence microscopy was performed using a Leica DMR microscope and a Leica DC200 camera. GFP and fluorescein isothiocyanate filter sets were used to detect GFP and chloroplast autofluorescence, respectively.
In situ immunodetection and labelling of mitochondria
Nicotiana tabacum and S. cerevisiae protoplasts were fixed for 2 h in 0.1 m KH2PO4, 0.5 mm MgCl2, 4% paraformaldehyde pH 6.5; washed in 0.1 m KH2PO4, 1.2 m sorbitol pH 6.5; and permeabilized at −20°C in methanol for 6 min and in acetone for 30 sec. They were then incubated for 1 h with antibodies (dilution: 1/100) against mitochondrial proteins (lipoamide dehydrogenase (de Castro Silva-Filho et al., 1996) for plant protoplasts and ATPase α- and β-subunits (Boutry and Goffeau, 1982) for yeast protoplasts and for 1 h with Texas Red-X goat anti-rabbit IgG (H + L) conjugate (Molecular Probes) (dilution: 1/100).
Mitochondria of N. tabacum protoplasts were labelled with 400 nm of Mitotracker Red CM-H2XRos (Molecular Probes) using a labelling period of 40 min in culture medium and transferred in fresh medium after centrifuging at 600 g for 6 min.
Tobacco and yeast mitochondrial extracts
To prepare crude mitochondrial extracts from tobacco, after transformation and incubation the electroporated protoplasts were pelleted (600 g for 6 min) and suspended in 100 µl homogenization medium (50 mm Tris, 1 mm EGTA, 0.4 m sucrose, 10 mm KH2PO4, 5 mmβ-mercaptoethanol, 0.1% polyvinylpyrolidone pH 7.6) in an Eppendorf tube. After addition of 40 mg glass beads (1 mm diameter), the tubes were vortexed for 4 sec and centrifuged for 30 sec at 610 g. The supernatants were then centrifuged for 10 min at 20 000 g and the pellet (crude mitochondrial extract) resuspended in 40 µl suspension medium (0.4 m mannitol, 0.1% bovine serum albumin, 10 mm KH2PO4, 1 mm phenylmethanesulfonyl fluoride pH 7.2), while the proteins in the supernatant (crude cytosolic fraction) were precipitated using chloroform/methanol method and resuspended in 40 µl suspension medium.
Yeast mitochondrial extracts were prepared from yeast spheroplasts (Foury, 1989), and subcellular extracts (including purified mitochondria) of transgenic plants were prepared as described by Chaumont et al. (1994).
Antibodies directed against green fluorescent protein
The gfp gene was introduced in the pQE-16 expression vector (Qiagen, Hilden, Germany) and expressed in E. coli cells. After rabbit immunization against GFP, the antibodies were purified by affinity chromatography.
Protein analysis and quantification of import
Proteins of different extracts (same volumes) were electrophoresed (SDS–PAGE) and transferred to a nitrocellulose membrane, which was successively incubated with antibodies and 125I-labelled protein A. The signals were quantified using a phosphor-imager (Bio-Rad GS-525 Molecular Imager System). For each construction, import was defined as the ratio between GFP detected in the crude (transient expression) or purified (stable expression) mitochondrial fraction and the total GFP detected in the mitochondrial (crude or purified) and cytosolic fractions together. This ratio was then expressed as percentage of the ratio obtained for the β-GFP control. To take account of the variability in the fragmentation of mitochondria during homogenization, values were corrected by the ratio obtained for the mitochondrial markers lipoamide dehydrogenase (for plant) and F1-ATPase (for yeast).
G.D. is the recipient of a fellowship from the Fonds pour la Formation à la Recherche dans l'Industrie et dans l'Agriculture. This work was supported by grants from the Belgian National Fund for Scientific Research and the Interuniversity ‘Poles of Attraction’ program of the Belgian Government Office for Scientific, Technical and Cultural Affairs.