To begin to decipher the intracellular targeting information within p36, we took advantage of tobacco (Nicotiana tabacum) Bright Yellow-2 (BY-2) suspension-cultured cells as a well-characterized in vivo import system 192021. Specifically, BY-2 cells were transiently transformed (via biolistic bombardment) with plasmid DNA encoding p36 fused at either its N or C terminus to the myc-epitope recognition motif (-EQKLISEEDL-; 22) and, then, following a 4 h incubation period to allow for gene expression and protein sorting, cells were processed for indirect immunofluorescence confocal-laser scanning microscopy (CLSM). As shown in Figure 1, both transiently-expressed myc-p36 and p36-myc localized exclusively to globular structures that were found mostly within the perinuclear region of the cell and contained endogenous E1β, a protein subunit of the pyruvate dehydrogenase complex located in the mitochondrial matrix 23. Interestingly, close inspection of these p36-containing globular structures at higher magnification revealed that they actually consisted of numerous torus or donut-shaped structures that contained myc-p36 or p36-myc and that enclosed spherical fluorescent structures containing matrix-localized E1β (refer to arrowheads in Figure 1). These toroidal structures also contained the endogenous mitochondrial outer membrane protein porin 24, as evidenced by the colocalization of p36-myc and porin (Figure 1). Taken together, these results indicate that p36 sorts to mitochondrial outer membranes in BY-2 cells and that the expression of this viral protein causes mitochondria to coalesce in perinuclear regions, in contrast to mitochondria in non-transformed cells that are distributed throughout the cytosol; refer to images of the immunostaining of endogenous mitochondrial porin and E1β in the non-transformed cells shown in the bottom row of Figure 1.

While it remains to be determined whether the aggregated mitochondria in p36-transformed cells were modified also in terms of their ultrastructure, this possibility is likely since expression of p36 (e.g., p36-GFP) in various other cell types (e.g., leaf mesophyll protoplasts and yeast cells) led to similar aggregations of mitochondria, as well as a proliferation of their outer membranes 161718. Indeed, high-magnified views of myc-p36- or p36-myc-transformed BY-2 cells in this study revealed that the immunostaining pattern attributable to endogenous porin in these cells was more diffuse than that observed in non-transformed cells (Figure 1), suggesting that the outer mitochondrial membranes in the former cells were altered in terms of their morphology. The nature of these changes or the mechanism(s) by which p36 participates in this process were not, however, examined further.

The localization of myc-p36 and p36-myc to mitochondria indicates also that appending the myc epitope to the N or C terminus of p36 did not disrupt the protein's normal targeting behaviour, since non-tagged (wild-type) p36 colocalized also with endogenous porin in the outer membranes of aggregated mitochondria (refer to Additional file 1A, top row). p36 was immunodetected in these cells using polyclonal antibodies raised against a synthetic peptide corresponding to an amino acid sequence in the C-terminal half of TBSV p33 and p92 replicase proteins (residues 184–203; 25) and conserved in CIRV p36 and p95 (residues 218–237). However, due to limited availability of this antibody reagent, only myc-tagged versions of p36, primarily p36-myc, were employed in the remainder of the experiments described in this manuscript.

The localization of epitope-tagged and wild-type p36 to mitochondrial outer membranes, and their subsequent effects on mitochondrial morphology and distribution, were similar also to when p36 was co-expressed with its allied replication protein, p95, either together alone (Rep) or together in the context of full-length CIRV, i.e., a CIRV cDNA positioned within a plant expression plasmid (Additional file 1A). Confirmation of the infectivity of this CIRV cDNA was that Chenopodium quinoa leaves 7 to 10 days after rub-inoculation displayed local lesions (Additional file 1B) that resembled the disease symptoms reported for leaves infected with native CIRV RNA 4. Moreover, electron microscopic analyses of these CIRV cDNA-inoculated leaf samples revealed the presence of mitochondrial-derived MVBs that were not observed in cells from mock-transformed leaves (Additional file 1C). The results for the mitochondrial localization of p36 expressed in the context of full-length CIRV (or Rep) are important because they indicate that, compared to the expression of p36 alone (either wild-type or epitope tagged), the viral protein behaves in a similar manner in terms of its intracellular localization. Thus, we deemed it appropriate to study the mitochondrial targeting information of p36 when expressed on its own.

The topological orientation of p36 in mitochondrial outer membranes was assessed using a differential detergent permeabilization assay 26. Specifically, BY-2 cells were transformed transiently with N- or C-terminal myc-tagged versions of p36 and then fixed and incubated either with triton X-100, which perforates all cellular membranes, or with digitonin, which perforates only the plasma membrane.

As shown in Figure 2 (top row), control experiments with myc-p36-transformed cells verified that permeabilization with triton X-100 allowed for the immunodetection (via epifluorescence microscopy) of endogenous proteins within the cytosol (i.e., α-tubulin) and within subcellular compartments (i.e., E1β within the mitochondrial matrix), whereas permeabilization with digitonin allowed for the immunodetection of only cytosolic proteins (i.e., α-tubulin, but not E1β). Figure 2 shows also that when myc-p36- or p36-myc-transformed cells were permeabilized with either triton X-100 or digitonin both expressed proteins were immunodetected, whereas endogenous E1β was only immunodetected in the corresponding triton-X-100-permeabilized cells. Likewise, myc-p36 was immunodetected in both triton X-100- and digitonin-permeabilized cells that were incubated with antibodies specific for myc or the p36 C-terminal peptide mentioned above, i.e., amino acid residues 218–237 located downstream of the second (of two) predicted TMDs in p36.

Overall, these data are in agreement with the previously proposed model for the membrane topology of p36 based on the hydropathy profile of the protein's primary amino acid sequence 15 and protease resistance analyses of p36 in either CIRV-infected plants or isolated mitochondria 18. In this model, p36 is predicted to be an integral membrane protein that contains two TMDs (residues 102–119 and 166–188), an intervening loop region (residues 120–165) orientated towards the mitochondrial intermembrane space, and N- and C terminal portions of the protein (residues 1–101 and 189–296) that face the cytosol. Unfortunately, our efforts to confirm the inward orientation of the intervening loop sequence in p36 were unsuccessful, because addition of the myc-epitope sequence to this portion of the protein caused the resulting mutant to be mislocalized from mitochondria to the cytosol in transformed cells (data not shown), a result that is likely due to the fact that, as discussed below, the loop sequence of p36 contains essential mitochondrial targeting information that was disrupted by the addition of the myc-epitope motif.

While previous studies using chimeras of CIRV p36 and CymRSV p33, expressed either in the context of the full-length virus or alone, revealed that the N terminal half of each protein is responsible for its sorting to mitochondria and peroxisomes, respectively 1415, results from subsequent mutational analyses of p36 expressed alone provided only limited insight to the precise nature the protein's mitochondrial targeting information 18. For instance, based on the targeting results of various p36 deletion mutants, Weber-Lotfi et al. 18 concluded that the mitochondrial sorting of p36 is mediated by multiple determinants located within a relatively large portion of the protein's N-terminal half, i.e., residues 84–196 which includes both of the protein's TMDs and their flanking regions, and that together these determinants might function cooperatively as a so-called signal loop-anchor type mitochondrial targeting sequence. These authors also speculated that this putative signal loop-anchor targeting sequence for p36 was unique because, rather than being comprised of non-continuous structural elements and/or three-dimensional folding features, such as those that typically constitute the targeting determinants in host-cell multi-spanning mitochondrial outer membrane proteins with β-barrel structures, the mitochondrial targeting determinants in p36 (a non-β-barrel protein) were part of a linear stretch of the protein. The physicochemical properties of these potentially novel targeting determinants in p36 were not, however, defined. Thus, to better understand the mitochondrial targeting information in p36 we carried out a comprehensive mutational analysis of the protein, employing initially both gain-of-function targeting assays with different portions of p36 fused to a reporter protein (Figure 3) and domain-swapping assays with p36 and p33 (Figure 4).

As shown in Figure 3A, gain-of-function targeting experiments were carried out using a series of fusion proteins that consisted of p36, or portions thereof, appended to the bacterial passenger protein chloramphenicol acetyltransferase (CAT) and that were expressed transiently in BY-2 cells. The representative micrographs presented in Figure 3B show that while CAT expressed alone accumulated throughout the cytosol, p36-CAT, consisting of full-length p36 fused to the N terminus of CAT, localized exclusively to aggregated mitochondria. These latter data are consistent with the localization of myc-tagged and wild-type p36 to (aggregated) mitochondria in BY-2 cells (refer to Figure 1 and Additional file 1), as well as p36 fused to GFP and expressed transiently (via agrobacterium infiltration) in tobacco leaf mesophyll cells 1718, and indicate that the CAT moiety used here as a reporter passenger protein does not alter the sorting behaviour of p36.

Similar to p36-CAT, p36 1-190-CAT, containing residues 1–190 of p36 including the protein's N-terminal hydrophilic domain, TMDs 1 and 2, and the intervening loop sequence appended to CAT, localized to (aggregated) mitochondria (Figure 3B), confirming also the results presented previously by Weber-Lotfi et al. 18 that the N-terminal half of the protein contains its mitochondrial targeting information. By contrast, p36 1-120-CAT, which includes only the N-terminal hydrophilic domain and TMD1 of p36, accumulated in the cytosol. Notably, the mitochondria in these p36 1-120-CAT-transformed cells, similar to mitochondria in cells transformed with CAT alone, were not altered in terms of their morphology and/or distribution (compare endogenous E1β immunostaining in p36 1-120-CAT-and CAT-transformed cells; Figure 3B), and reinforcing the notion that this fusion protein was not targeted to mitochondria.

Other p36-CAT fusion proteins comprised of smaller portions of the p36 N-terminal half also did not target to mitochondria. That is, both p36 90-164-CAT and p36 120-164-CAT, consisting of TMD1 and the intervening loop sequence and the intervening loop sequence alone, respectively, localized to the cytosol (Figure 3B). On the other hand, both p36 120-190-CAT and p36 90-190-CAT localized to (aggregated) mitochondria, albeit the former fusion protein only did so in a relatively inefficient manner since it accumulated also in the cytosol (Figure 3B). Based on the results presented here (Figure 3), p36 90-190-CAT contains the minimally sufficient portion of the protein capable of efficiently targeting CAT to mitochondria and, similar to wild-type p36, is localized to the outer membranes of mitochondria in an N_out-C_out topology (refer to Additional files 2A and 2B for high-magnification CLSM and differential permabilization assays with p36 90-190-CAT).

Characterization of the mitochondrial targeting information in p36 was also carried out using domain-swapping assays with TBSV p33, an ortholog of CymRSV that has been recently well characterized in terms of its peroxisomal membrane targeting signal 25. Alignment of the deduced amino acid sequences for p36 and p33 revealed that while these proteins are nearly identical in their C-terminal halves, they are significantly divergent with respect to their N termini (Figure 4A). For instance, although both proteins are predicted to contain two TMDs located at relatively similar positions within their N-terminal halves (underlined, Figure 4A), p36 also possesses a number of amino acid residues in this region of the protein that are not found in p33. The most conspicuous of these being stretches of unique residues located near the extreme N terminus and within the intervening loop sequence of p36 (Figure 4A). Alignment of the p36 and p33 sequences revealed also that the multiple targeting signal motifs responsible for sorting nascent p33 initially to peroxisomes (i.e., -K₁₁K₁₂-, K₇₆RRQR₈₀- and -R₁₂₄PSVPKK₁₃₀-) (shaded in grey in Figure 4A) and then from peroxisomes to a subdomain of the endoplasmic reticulum (ER) referred to as peroxisomal ER (pER) (i.e., -K₅R₆-) (bolded in Figure 4A) 25 are not conserved in p36, consistent with the apparently distinct intracellular sorting pathways employed by these two proteins.

As shown in Figures 4B and 4C, p36-myc and p33-myc (TBSV p33 appended to a C-terminal myc-epitope tag) sorted exclusively to aggregated mitochondria and peroxisomes, respectively, in transiently-transformed BY-2 cells. The aggregation of peroxisomes in p33-myc-transformed cells is apparent by comparing the immunostaining pattern attributable to the endogenous peroxisomal matrix enzyme catalase in these cells with that of p36-myc-transformed cells, wherein catalase-containing peroxisomes display a normal (punctate) morphology and distribution (Figure 4C). These data for the localization of p33-myc to altered peroxisomes was expected, since previously published results with wild-type (non-tagged) p33 expressed for the same length of time (i.e., 4 h following biolistic bombardment) revealed that it was localized exclusively to peroxisomes and caused the organelles to coalesce 25. Also similar to wild-type p33 25, p33-myc sorted from peroxisomes to pER at later time points following bombardment (i.e., 24 h) (data not shown), indicating that the myc epitope also did not disrupt the normal peroxisome-to-pER sorting of the protein. Nevertheless, in order to minimize any potential sorting complexities associated with the targeting of p33 from peroxisome to pER at later time points, all of p33/p36-myc hybrids employed in domain-swapping assays were assessed only at 4 h post-bombardment for peroxisomal versus mitochondrial targeting.

Replacement of the N-terminal half of p33 (residues 1–156) with the corresponding sequences from p36 (residues 1–190), including the N-terminal hydrophilic domain and both TMDs plus the intervening loop sequence, resulted in the hybrid protein (p36 1–190 p33-myc) being sorted to aggregated mitochondria, but not to peroxisomes (Figure 4C). Conversely, p33 1–156 p36-myc, consisting of the N-terminal half of p33, including all three of the protein's peroxisomal targeting signals (i.e., -K₁₁K₁₂-, K₇₆RRQR₈₀- and -R₁₂₄PSVPKK₁₃₀-; refer to Figure 4A) 25, fused to the C-terminal half of p36, sorted exclusively to peroxisomes in a manner similar to full-length p33-myc (Figure 4C).

Figures 4B and 4C show also that fusion of the N-terminal 120 amino acids of p36 (including the protein's N-terminal hydrophilic domain and first TMD) to the remaining C-terminal portion of p33-myc resulted in the hybrid protein (p36 1–120 p33-myc) being localized to the cytosol, a result similar to when this N-terminal region of p36 was fused to CAT (p36 1-120-CAT) (Figure 3). Localization to the cytosol was observed also when the intervening loop sequence from p36 was replaced with the loop sequence from p33 (p33 103–131 p36-myc), indicating that this region of p36 contains essential mitochondrial targeting information. In accordance with this premise, the p36 intervening loop sequence together with its second TMD was capable of redirecting, albeit inefficiently, p33-myc to mitochondria, i.e., p36 120–190 p33-myc was sorted to both mitochondria and peroxisomes (Figure 4C), and further showing that this hybrid protein contains targeting information for both organelles. By contrast, mitochondrial targeting of p36 was completely abolished when its intervening loop sequence and TMD2 was replaced with the corresponding sequences from p33 (residues 103–156), i.e., p33 103–156 p36-myc was not localized to mitochondria, but instead localized to both peroxisomes and the cytosol (Figure 4C). The apparent inefficient sorting of this hybrid protein to peroxisomes is likely due to the fact that it possessed only one of the three peroxisomal targeting signals within p33, i.e., -R₁₂₄PSVPKK₁₃₀- (refer to p33 sequence shown in Figure 4A).

Overall, the data presented in Figures 3 and 4 confirm and extend the observations of Weber-Lotfi et al. 18 that the minimally sufficient mitochondrial targeting information in p36 is located within protein's N-terminal half, including both TMDs and the intervening loop sequence. Additional flanking sequences immediately upstream and downstream of TMD1 and TMD2, respectively, are not essential, however, for targeting p36 to mitochondria. For instance, the N-terminal hydrophilic portion of p36 was insufficient in sorting CAT (Figure 3) or a p36/p33 hybrid protein containing the remaining C terminal portion of p33 (Figure 4) to mitochondria. Thus, our data contradict the previously proposed notion that the N-terminal hydrophilic domain of p36, specifically, the residues immediately upstream of TMD1, contain a distinct mitochondrial targeting determinant 18. It appears instead that p36 contains only one mitochondrial targeting signal: a signal that consists of a relatively large portion of the protein (residues 90–190) and includes several different structural elements, i.e., two TMDs and an intervening soluble loop sequence. The relative contribution of these elements likely mediates the various aspects of p36 biogenesis including maintaining solubility before membrane insertion, targeting to mitochondria, and/or ensuring proper assembly in the mitochondrial outer membrane.

The results presented above indicate that the sorting of p36 to mitochondria is, at least in part, dependent upon its TMDs. Thus, to test whether specific amino acid sequences or more global properties, e.g., overall hydrophobicity and/or length, within one or both of the TMDs contribute important mitochondrial targeting information, we generated several different mutant versions of p36-myc containing altered TMD sequences (Figure 5A). Included among these p36 TMD mutants were those that contained sequences derived from either TBSV p33 or the mitochondrial outer membrane-localized isoform of cytochrome b₅ (Cb5), a tail-anchored protein whose mitochondrial targeting signal consists of several unique physicochemical and sequence-specific characteristics within its single C-terminal TMD and hydrophilic tail domain 27. By replacing TMD1 or TMD2 in p36 with either the corresponding TMDs from p33 or the TMD from Cb5 (and vice versa, i.e., Cb5 with its TMD exchanged for TMD1 or TMD2 from p36), we sought to determine what role the p36 and p33 TMDs play in mitochondrial versus peroxisomal targeting and, in the case of p36 and Cb5, whether these two proteins possess functionally equivalent mitochondrial targeting information within their TMDs.

As shown in Figure 5B, when the sequences for TMD1 and TMD2 in p36 were exchanged the resulting mutant protein (p36-myc TMD1 ⇔ TMD2) localized exclusively to mitochondria. Interestingly, the morphology and distribution of the mitochondria in these cells, unlike in wild-type p36-myc-transformed cells (refer to Figure 1), did not appear to be altered (e.g., aggregated). These data suggest that the relative positions of the two TMDs within p36 influence the effects that this protein has on mitochondria morphology/distribution, but not its intracellular sorting.

An examination of the TMD amino acid sequences within p36 (underlined, Figure 4A) did not reveal any noticeable conserved features. For instance, with the exception that both TMDs possess similar (moderate) average hydrophobicity indexes (according to the Kyte-Doolittle algorithm 28) of 1.58 (TMD1) and 2.05 (TMD2), they vary in their overall length (i.e., TMD1 is predicted to be 18 amino acid residues long, whereas TMD2 is 23 residues long). Moreover, they contain no obvious conserved amino acid sequence-specific motifs or an enrichment of particular amino acid residues. Therefore, the possibility was tested that the overall hydrophobic nature of the TMDs was the functional determinant in sorting p36 to mitochondria. Toward this end, either TMD1 or TMD2, or both TMDs together, in p36 were replaced with artificial (idealized) amino sequences composed of multiple -LALV-repeats 29 while still maintaining the length of the TMDs in p36, i.e., 18- and 23-long repeats of -LALV-corresponding to the length of TMD1 and TMD2 in p36, respectively. As shown in Figure 5B, all three of the resulting mutant versions of p36-myc (i.e., p36-myc TMD1ΔsynTMD, p36-myc TMD2ΔsynTMD and p36-myc TMD1/TMD2ΔsynTMD) mislocalized to cytosol and to some unknown punctate compartment(s) that did not colocalize with endogenous mitochondrial E1β (refer to open arrowheads in high magnified views shown in Figure 5B), nor peroxisomal catalase (data not shown). Similarly, replacement of TMD1 or TMD2 in p36 with the 18 amino-acid long single TMD from the mitochondrial isoform of Cb5 27 resulted in both mutant proteins (p36-myc TMD1ΔCb5TMD and p36-myc TMD2ΔCb5TMD) being mislocalized to the cytosol. By contrast, both p36-myc TMD1Δp33TMD1 and p36-myc TMD2Δp33TMD2 in which TMD1 or TMD2 of p36 were replaced with TMD1 or TMD2 from p33, respectively, sorted to (aggregated) mitochondria in a manner similar to wild-type p36-myc (Figure 5B).

As shown also in Figure 5B, N-terminal myc-tagged Cb5 (myc-Cb5) localized exclusively to mitochondria, as previously published 27. Replacement of the single TMD in myc-Cb5 with TMD1 or TMD2 from p36, however, resulted in the corresponding mutant proteins being mislocalized either to the cytosol (myc-Cb5Δp36TMD1) or to the ER (myc-Cb5Δp36TMD2) (Figure 5B); although in a small proportion (~15%) of the cells transformed with myc-Cb5Δp36TMD2 the mutant protein targeted to mitochondria in a manner similar to that of wild-type myc-Cb5 (Figure 5B).

Together, the data presented in Figure 5 imply that not all hydrophobic sequences are sufficient to replace TMD1 or TMD2 in p36 in terms of mediating its proper targeting to mitochondria. However, a pertinent question based on these data is why the TMDs from p33, a peroxisomal membrane-localized protein 15253031, but not an idealized hydrophobic TMD, nor the TMD from an authentic mitochondrial outer membrane protein Cb5 27, are capable of preserving the targeting of p36 to mitochondria? One possibility is that the relatively moderate overall hydrophobicity of the TMDs in p36 is an essential feature of the protein's mitochondrial targeting signal. Based on this premise, the mislocalization of p36 when one (or both) of its TMDs was replaced with either an artificial TMD(s) or the Cb5 TMD was due to the relatively high hydrophobic indexes of these introduced sequences, i.e., 3.33 and 3.36 for the 18 and 23 amino acid-long synthetic TMDs and 2.42 for the 18 amino acid-long Cb5 TMD, compared to those for p36 TMD1 (1.58) and TMD2 (2.05). By contrast, the mitochondrial targeting of p36 was not disrupted when its TMDs were replaced individually with the relatively moderate hydrophobic TMDs from p33, i.e., 2.01 and 1.98 for TMD1 and TMD2, respectively.

This scenario that the mitochondrial sorting of p36 is mediated, at least in part, by the moderate hydrophobic nature of its TMDs is in agreement with results from several other studies on the sorting of authentic mitochondrial outer membrane proteins. That is, depending on the protein, a moderate hydrophobic TMD(s), combined with positive charges at its flanking soluble regions, is essential for mitochondrial outer membrane targeting [reviewed in 3233]. For example, the 20-kD and 70-kD subunits of the TOM complex (TOM20 and TOM70) and TOM45 (a mitochondrial outer membrane protein of 45 kD) from yeast all contain moderately hydrophobic TMDs that are functionally interchangeable in terms of their targeting behaviour 34. The mammalian (human) TOM20 and yeast mitochondrial outer membrane protein fission 1 (FIS1) also appear to rely on the moderate hydrophobicity of their TMDs, since increasing their overall hydrophobic score via mutagenesis resulted in both modified proteins being mislocalized 3536. Conversely, reducing the overall hydrophobicity of an artificial TMD within a yeast ER-localized reporter membrane protein by introducing one or more hydrophilic residues caused it to relocalize from the ER to mitochondria 37.

A notable extension of this model is that, for at least some mitochondrial outer membrane proteins, they appear to rely also on subtle sequence-specific features within their TMDs. For instance, both the yeast 5-kD and 7-kD TOM subunits (TOM5 and TOM7) possess a conserved proline residue near the center of their TMDs that is required for their efficient targeting to mitochondria 3839. While the precise role of this proline in the mitochondrial targeting of these proteins has not been determined, it may be due to the residue's ability to destabilize the TMD's α-helical structure and, thus, dictate its overall configuration and subsequent interaction with the proper cognate receptor and/or integration/assembly factor(s) 32. Interestingly, targeting of mitochondrial plant Cb5 also appears to rely on a conserved proline residue that forms part of a short hydrophilic surface situated within the protein's TMD 27. Based on this, it is perhaps not surprising that replacement of the Cb5 TMD with either TMD1 or TMD2 from p36 resulted in both of the modified Cb5 proteins being either completely (myc-Cb5Δp36TMD1) or partially (myc-Cb5Δp36TMD2) mislocalized within cells (Figure 5B). That is, the nature of the targeting information within the TMDs of p36 and Cb5 appear to be sufficiently distinct in terms of their relative overall hydrophobicity and the presence (or the lack) of sequence-specific motifs, such that they are not functionally interchangeable. As discussed below, these apparent differences in the targeting elements within the TMDs of p36 and Cb5 also appear to reflect the relative contribution of unique positively-charged targeting elements at their flanking regions.

Since the intervening loop sequence in p36 (residues 120–165 between TMD1 and TMD2) is significantly divergent from the corresponding loop sequence in peroxisomal-targeted p33 (Figure 4A) and since replacement of this region in p36 with that from p33 resulted in the hybrid protein (p33 103–131 p36-myc) being mislocalized to the cytosol (Figure 4B), we analyzed this region in p36 to determine if it contained any distinctive features involved in mitochondrial targeting. Figure 6A shows that the deduced amino acid sequence of the p36 intervening loop, specifically the central portion of this sequence that is devoid in p33 (residues 131–157, shaded in Figure 6A; refer also to Figure 4A), is enriched in both positively-charged (bolded) and hydroxylated (i.e., Ser, Thr and Pro) amino acid residues. This central region of the p36 intervening loop also has the propensity (based on secondary structure prediction programs) to form an α-helix, one with amphipathic characteristics conveyed by a positively-charged face (refer to amino acid residues in red; Figure 6A).

Based on these observations and previous reports that the targeting of several other authentic mitochondrial outer membrane-destined proteins rely on positively-charged residues together with a flanking TMD(s) 3233, we examined further the positively-charged face of the amphipathic helix in p36 by mutational analyses. As shown in Figures 6B and 6C, deletion of amino acid residues 131–157 in p36-myc resulted in the mutant (i.e., p36-mycΔ131-157) being mislocalized to the cytosol. These data are consistent with the mislocalization of p36-myc when its intervening loop sequence was replaced with that of p33 (i.e., p33 103–131 p36-myc; Figure 4B).

Mitochondrial targeting of p36-myc was also abolished when glycine residues were exchanged for each of the four positively-charged residues predicted to be situated along the same face of the amphipathic α-helix within the protein's intervening loop sequence, i.e., p36-myc K₁₃₄K₁₃₇R₁₄₄K₁₅₁ΔG mislocalized to the cytosol (Figure 6C). By contrast, glycine substitutions of another cluster of positively-charged residues (i.e., K₉₃K₉₄R₉₈R₁₀₁) located immediately upstream of TMD1 did not disrupt the sorting of p36-myc to mitochondria (p36-myc K₉₃K₉₄R₉₈R₁₀₁ΔG; Figure 6C). These latter results, as well as those presented earlier for the mislocalization of p36 1-120-CAT (Figure 3) and p36 1–120 p33-myc (Figure 4) reinforce our conclusion that, in contrast to the findings reported by Weber-Lotfi et al. 18, the N-terminal hydrophilic region located upstream of TMD1 in p36 does not contain mitochondrial targeting information.

The results for p36-myc K₉₃K₉₄R₉₈R₁₀₁ΔG also indicate that not all mutations to the protein affected its mitochondrial targeting fidelity. An observation that underscores a substantive caveat of this study (and any analysis of membrane protein targeting) whereby mutations of p36 that resulted in the protein's mislocalization could have been due to aberrant protein folding, rather than a disruption of a specific targeting determinant(s). Although this possibility cannot be excluded, our combined use of loss-of-function and gain-of-function targeting assays strongly supports our definition of the bona fide mitochondrial targeting information within p36.

Collectively, our findings on the combined importance of positively-charged residues (Figure 6) and moderate hydrophobic TMDs (Figure 5) in the mitochondrial targeting of p36, as well as the N_out-C_out topological orientation of the protein in the mitochondrial outer membrane (Figure 2), suggests that the insertion and assembly of p36 involves a signal loop-anchor mechanism 40. That is, interaction of the positively-charged residues in the intervening loop sequence of nascent p36 with the mitochondrial outer surface might promote a conformational change that drives insertion of its moderate hydrophobic TMDs into the bilayer, and results in the proper N_out-C_out topology. This proposed signal loop-anchor mechanism is consistent, as mentioned above, with results from previous in vitro insertion experiments with p36 and isolated mitochondria 18. Moreover, since the topological orientation of p36 is identical to the topology of the authentic outer mitochondrial membrane proteins yeast fuzzy onion 1 (Fzo1) and its mammalian homologs, mitofusin (Mfn) 1 and 2, i.e., anchored in an N_out-C_out manner by two α-helical TMDs connected via a short loop sequence located in the intermembrane space 4142, it is possible that the signal loop-anchor insertion of p36 is mediated by the same mitochondrial machinery that is utilized by Fzo1 and Mfn1/2. Results of experiments aimed at addressing this possibility are described next.

Almost all mitochondrial proteins are encoded by nuclear genes, synthesized on free polyribosomes in the cytosol, and targeted post-translationally to the organelle. Thereafter, the recognition, translocation, and sorting of these proteins is usually mediated by the TOM complex and, depending on the protein's submitochondrial destination, one of several other translocase complexes located in the outer and inner mitochondrial membranes [reviewed in 434445]. For instance, based on studies carried out primarily with yeast and mammalian model systems, it is well established that nascent proteins destined for the mitochondrial matrix or inner membrane are recognized by the import receptors TOM20 and TOM70 and then delivered to one of two discrete translocase of the inner mitochondrial membrane (TIM) complexes. The majority of proteins destined for the outer mitochondrial membrane in yeast and mammals also rely on the TOM complex, but appear to do so in a number of different ways depending on their topology [reviewed in 33]. For example, mitochondrial outer membrane proteins with multiple α-helical TMDs, including Mfn1/2, utilize the receptor TOM70, but not other components of the TOM complex 46. β-barrel outer mitochondrial membrane proteins, such as porin and the channel forming subunit of the TOM complex, TOM40, as well as C-terminal tail-anchored subunits of the TOM complex (i.e., TOM22, TOM5, TOM6 and TOM7), are recognized mainly by TOM20 and then subsequently transferred to the SAM complex 47, a hetero-oligomeric complex located in the mitochondrial outer membrane where it functions in β-barrel protein insertion and assembly 3348. Interestingly, another subset of C-terminal, tail-anchored outer membrane proteins, including BAX and FIS1, do not require any known import (TOM) components, but instead appear to rely on the unique lipid composition of the mitochondrial outer membrane for their proper targeting and insertion 4950.

While it is generally accepted that the various mitochondrial import pathways and associated machinery are conserved among evolutionarily diverse organisms including yeast, animals and plants, considerable differences exist among some of their core protein subunit compositions and functional specificities, particularly in plants 5152. Thus, p36 may or may not use the same mitochondrial insertion machinery in plants as that employed by multi-spanning mitochondrial outer membrane proteins in yeasts and mammals, such as the TOM70-dependent targeting of Mfn1/2 46. The targeting and insertion of p36 may be also more similar to that of yeast FIS1, which is mediated primarily by the distinct lipid composition of the mitochondrial outer membrane 50, a possibility that appears to be consistent with previous data indicating that p36 does not to rely on any mitochondrial-surface proteinaceous receptors 18. This latter conclusion, however, was based solely on the observation that pre-treatment of isolated mitochondria with trypsin to remove any exposed parts of surface receptors did not prevent membrane insertion of p36. Indeed, this apparent lack of sensitivity to trypsin is not evidence a priori that p36 import does not utilize a proteinaceous receptor(s), since mitochondrial surface receptors can display differences in terms of their sensitivity/resistance to an applied protease 5354 and can even be bypassed in vitro such that a chloroplast protein can be imported into protease-pretreated mitochondria with the same efficiency as a mitochondrial protein that normally relies on a surface receptor 55.

To determine the insertion pathway of p36 into the outer mitochondrial membrane we tested whether it could interact directly with a variety of protein components of the plant TOM and/or SAM complexes. Specifically, we employed bimolecular fluorescence complementation (BiFC) assays to determine whether in vivo physiological interactions occur between p36 and four Arabidopsis TOM protein subunits including: i) the TOM20 import receptor 565758; ii) TOM22, which may function in a manner similar to yeast and mammalian TOM22 by providing some receptor function, but mostly as a link between the primary receptors TOM20 and TOM70 59; iii) TOM40, the central pore-forming subunit of the TOM complex 56; and iv) mtOM64 (outer mitochondrial membrane protein of 64-kD), which functions as an import receptor and, based on structural similarities, may be a paralog of yeast and mammalian TOM70 5260. We also tested whether p36 physically interacts with Arabidopsis METAXIN, a diverged form of the SAM component SAM37 and METAXIN1 in yeasts and mammals, respectively 5261.

BY-2 cells were transiently transformed with pairs of plasmids encoding p36 fused at its C terminus to the C-terminal half of the modified yellow fluorescent protein Venus 62 (cVenus) and one of the TOM or SAM proteins fused at either their N or C termini to the N-terminal half of Venus (nVenus) (e.g., nVenus-TOM20); refer to illustrations of various proteins in Figure 7B showing their predicted topological orientations in the mitochondrial outer membrane and the relative location of their appended Venus fragment. Because each half of Venus is not intrinsically fluorescent, Venus fluorescence is observed only when intermolecular interactions occur between nVenus- and cVenus-tagged proteins 636465, highlighting an advantage of BiFC over other protein-protein interaction assays (e.g., yeast two-hybrid and immunoprecipitations) in that the subcellular localization of the interacting proteins can directly be observed in living cells via microscopy. All cells were also co-bombarded with a third plasmid encoding the red fluorescent protein (RFP), which served as an internal standard for transformation efficiency and aided in identifying transformed cells.

Alternatively, BY-2 cells were co-bombarded with a plasmid coding for an individual nVenus- or cVenus-tagged protein along with a plasmid encoding βATPase-GFP (consisting of the N-terminal mitochondrial matrix targeting presequence of βATPase [residues 1–60] fused to the N terminus of GFP), which served to identify transformed cells and, acting as a well-characterized mitochondrial marker protein 6667. As presented in Additional file 3A, results from these latter experiments provided confirmation of the expression of each of the individual Venus half fusion proteins (via immunomicroscopy of an appended myc or hemaglutinin (HA)-epitope tag; see Methods and Materials for details on the construction of BiFC plasmids), as well as their mitochondrial localization. Additional results confirming the topological orientation of several of these Venus half fusion proteins in BY-2 cells using the differential detergent permeabilization assay are presented in Additional file 3B.

The results presented in Figure 7A show that p36 (i.e., p36-cVenus) interacted in vivo with all three of the nVenus-tagged versions of the putative import receptors, TOM20, TOM22 and mtOM64. Notably, the interaction of p36 with TOM20 was unaffected when the receptor protein's N-terminal tetratricopeptide motif-based domain responsible for the recognition and interaction with mitochondrial precursor proteins (residues 1–142; 58) was deleted (TOM20mut), suggesting that the interaction of p36 with the TOM20 is not mediated by this substrate-receptor-binding motif. p36 did not interact, however, with either TOM40, the SAM component METAXIN, or itself (i.e., p36-nVenus) (Figure 7A). While these latter data appear to contradict those published elsewhere for the homo-oligmerization of the p36-related protein, TBSV p33 [68 and references therein] and, p36-p36 and other p36-protein interactions likely occur within the context of the functional CIRV replication complex and, thus, mediated by the allied replication protein, p95, as well viral RNA template 69.

In a series of control experiments we also evaluated whether or not the βATPase matrix targeting presequence and/or the full-length outer mitochondrial β-barrel membrane protein, porin, interacted with the TOM and SAM complex protein subunits in a manner similar to p36. As shown in Figure 7A, the data obtained for βATPase and porin are mostly in accordance with the current working models for the mitochondrial import pathways used by these two proteins 43444551. For instance, both βATPase and porin (βATPase-cVenus and porin-cVenus) interacted with the nVenus-tagged versions of TOM20, TOM22 and mtOM64, reflecting the apparent functional redundancy of these import receptors in plants 52. Likewise, porin interacted with the N-terminal mutant version of TOM20 (TOM20mut), but βATPase did not, supporting the notion that the import of matrix-destined proteins is dependent primarily on TOM20 via the receptor's N-terminal tetratricopeptide motif domain 5658. The relative importance of the mtOM64 receptor in βATPase and porin (and p36) import using a similar mutagenesis-BiFC approach was not, however, assessed, since deletion of the predicted substrate-receptor binding motif in mtOM64 6070 disrupted its sorting to mitochondria in BY-2 cells (data not shown).

Figure 7A shows also that porin, but not βATPase, interacted with METAXIN, consistent with the proposed role of this SAM component in the insertion and assembly of β-barrel proteins in yeasts, mammals and plants 474852. However, Lister et al. 52 recently showed using pull-down and yeast two-hybrid assays that METAXIN interacted with a number of plant matrix-destined precursor proteins as well as β-barrel porin and TOM40, suggesting that this protein receptor, unlike its counterparts in yeast and mammals, performs a variety of roles in plants in addition to the insertion and assembly of β-barrel membrane proteins. Other than the use of different experimental approaches in the latter study 52 compared to that employed here, a reason(s) for the apparent lack of interaction between βATPase and METAXIN is not known. Regardless, these results for βATPase (and p36) and METAXIN, combined with the negative interaction data for certain other protein pairs discussed below (e.g., p36 and Cb5) provides convincing evidence that the Venus fluorescence observed in (positive) BiFC assays (e.g., porin and METAXIN) resulted from specific protein-protein interactions and was not simply a consequence, for example, of the overexpression of protein pairs localized to the same intracellular local (i.e., mitochondrial outer membrane) 65.

At least one unexpected finding from the BiFC experiments presented in Figure 7A was that both βATPase and porin (and p36) did not interact with TOM40, the putative primary subunit of the TOM complex import channel 56. These results contradict those obtained primarily from studies with yeasts and mammals, wherein TOM40 participates in the import of most types of mitochondrial proteins including those destined for the matrix or outer membrane 434445. One possible reason for this apparent lack of interaction between TOM40 and βATPase and porin (and p36) is that this putative channel protein, while being efficiently targeted to mitochondria when expressed individually in BY-2 cells (see Additional file 3A), was not properly inserted/assembled into functional TOM complexes and, thus, could not participate in mitochondrial protein import. Another possible reason may be that the nVenus fragment appended to TOM40 (refer to illustration in Figure 7B) is sterically inaccessible, such that even if the fusion protein was properly assembled into the TOM complex it could not interact with cVenus appended to either co-expressed βATPase, porin or p36. Unfortunately, since another TOM40 fusion protein with nVenus appended to its C terminus (TOM40-nVenus) was not sorted to mitochondria in BY-2 cells (data not shown) and because no other experimental data has been published for the structure and/or function of a plant TOM40, resolution of this issue remains to be addressed.

In additional control experiments TBSV p33 (p33-cVenus) was not observed to interact with any of the TOM complex subunit proteins tested nor with p36 (Figure 7A). These results were entirely expected since p33 sorts to peroxisomes and not to mitochondria (25; see Additional Figure 3A for the localization of p33-cVenus in BY-2 cells). Interactions were also not observed between p36 and porin or p36 and Cb5 (Figure 7A), indicating that, similar to the negative results discussed above for βATPase (and p36) and METAXIN, the positive Venus signals detected for certain other protein pairs was due to their specific protein-protein interaction. However, related to these control experiments with p36, porin and Cb5, it is important to note that, whereas Cb5 and p36 are orientated in outer mitochondrial membranes in an N_out-C_in 27 and N_out-C_out manner (Figure 2; and 1518), respectively, porin contains two putative TMDs and orientated in an N_in-C_in manner (Figure 7B; refer also to differential permeabilization results for cVenus-porin in Additional file 3B). Thus, while cVenus appended to porin (porin-cVenus) is predicted to be located in the intermembrane space, nVenus appended to p36 (p36-nVenus) faces the cytosol (Figure 7B) and, thus, the lack of a BiFC signal for these two proteins when co-expressed in BY-2 cells was expected given the physical separation of their corresponding Venus halves 65.

Taken together, the BiFC results presented in Figure 7A indicate that p36 interacts with a variety of import receptors in the TOM complex including TOM20, TOM22, and mtOM64, consistent with recent findings that, compared to yeasts and mammals, the mitochondrial import apparatus in plants is highly flexible with overlapping specificity 52. Our BiFC results indicate also that whereas the porin interacts with the putative SAM component METAXIN, p36 (and βATPase) does not. Hence, METAXIN appears to participate specifically in the assembly of outer mitochondrial membrane proteins containing a β-barrel structure, a conclusion that extends the findings presented previously for the role of this mitochondrial import component in plants 52.

Standard recombinant DNA procedures were performed as described by Sambrook et al. 84. Molecular biology reagents were purchased either from New England BioLabs (Beverly, MA, USA), Promega (Madison, WI, USA), Perkin-Elmer (Perkin-Elmer Biosystems, Mississauga, Canada), Stratagene (La Jolla, CA, USA), or Invitrogen (Carlsbad, CA, USA). Synthetic oligonucleotides were synthesized by either Invitrogen (Frederick, MD) or University of Guelph Laboratory Services (Guelph, Canada); a complete list of the sequences of the oligonucleotides primer used in plasmid constructions described below are provided in Additional file 4. Plasmid DNA was isolated using Qiagen reagents (Qiagen, Mississauga, Canada), and dye terminated cycle sequencing was used with an Applied Biosystems Model 377 or 3730 automated sequencer (Perkin-Elmer Biosystems) to verify all DNA constructs. Mutagenesis was carried out using appropriate complementary forward and reverse mutagenic primers and procedures according to the QuickChange site-directed mutagenesis kit (Stratagene).

Tobacco (Nicotiana tabacum cv BY-2) suspension-cultured cells were maintained and prepared for biolistic bombardment as described previously 19. Briefly, all transient transformations, with the exception of those for BiFC (see below), were performed using tungsten particles coated with 10 μg of plasmid DNA (or 5 μg of each plasmid for co-transformations) and a biolistic particle delivery system-1000/HE (Bio-Rad Laboratories, Mississauga, Canada). Bombarded cells were incubated for 4 h to allow for expression and sorting of the introduced gene product(s), then fixed in formaldehyde, incubated with 0.01% (w/v) pectolyase Y-23 (Kyowa Chemical Products, Osaka, Japan), and permeabilized with either 0.3% (v/v) triton X-100 or 25 μg mL^-1 digitonin (Sigma-Aldrich Ltd.) 26. Cells were evaluated after 4 h to ensure that any potential negative effects due to (membrane) protein over-expression were avoided.

Fixed and permeabilized cells were processed for immunofluorescence microscopy as described by Trelease et al. 91. Primary and dye-conjugated secondary antibodies and sources were as follows: mouse anti-myc antibodies in hybridoma medium (clone 9E10; Princeton University, Monoclonal Antibody Facility, Princeton, NJ, USA); mouse anti-CAT antibodies in hybridoma medium (provided by S. Subramani, University of California, San Diego); rabbit anti-cottonseed catalase IgGs 92; mouse anti-maize porin 24; rabbit anti-pea E1β 93; mouse anti-α-tubulin (clone DM 1A) (Sigma-Aldrich Ltd.); rabbit anti-myc and mouse anti-HA (Bethyl Laboratories, Montgomery, TX, USA); rabbit anti-p36 antibodies were raised against a synthetic peptide corresponding to an identical amino acid sequence in both p36 and TBSV p33 (-VEPARELKGKDGEDLLTGSR-) (residues 218–237 in p36 and 184–203 in p33) (refer to the p33 and p36 sequence alignment shown in Figure 4A) 25; goat anti-mouse Alexa Fluor 488 IgGs (Cedar Lane Laboratories, Ontario, Canada); goat anti-rabbit rhodamine red-X IgGs (Jackson ImmunoResearch Laboratories, West Grove, PA, USA). Concanavalin A conjugated to Alexa 594 (Molecular probes, Eugene, OR) was added to BY-2 cells at a final concentration of 5 μg mL^-1 during the final 20 min of incubation with secondary antibodies. All antibodies raised in rabbits were IgG-affinity purified using protein A-Sepharose columns.

Epifluorescent images of BY-2 cells were acquired using a Zeiss Axioskop 2 MOT epifluorescence microscope (Carl Zeiss, Toronto, Canada) with a Zeiss 63× Plan Apochromat oil-immersion objective. Image captures were performed using a Retiga 1300 charge-coupled device camera (Qimaging, Burnaby, Canada) and Northern Eclipse 5.0 (Empix Imaging, Mississauga, Canada) or OpenLab 5.0 (Improvision Inc., Lexington, MA, USA) software. CLSM images of BY-2 cells were acquired using a Leica DM RBE microscope with a Leica 63× Plan Apochromat oil-immersion objective, a Leica TCS SP2 scanning head, and the Leica TCS NT software package (Version 2.61) (Leica, Heidelberg, Germany). Fluorophore emissions were collected sequentially in double- and triple-labeling experiments; single-labeling experiments showed no detectable crossover at the settings used for data collection. Confocal images were acquired as a z-series of representative cells and single optical sections were saved as 512 × 512-pixel digital images. All fluorescence images of cells shown in individual figures are representative of > 50 independent (transient) transformations from at least two independent transformation experiments. Figure compositions were generated using Adobe Photoshop CS (Adobe Systems, San Jose, CA).

Biolistic bombardment of cells for BiFC experiments was carried out essentially as described above with the exception that tungsten particles were coated with 200 ng each of plasmids encoding the appropriate combinations of proteins. Tungsten particles were coated also with 200 ng of pRTL2/RFP serving as an internal reference marker to help identify cells that were transformed and, thus, could be examined for the BiFC signal. Alternatively, particles were coated also with 200 ng of pUC18/βATPase-GFP or pRTL2/RFP-MFP serving as a mitochondrial or peroxisomal marker protein, respectively, and for assessing (mitochondrial or peroxisomal) localization of co-expressed Venus half fusion proteins.

Approximately 16 h after bombardment, cells were formaldehyde fixed and either examined (via fluorescence microscopy) for BiFC (and RFP) or processed for immunofluorescence microscopy as described above. The 16 h time point and the amounts of plasmid DNA employed in BiFC assays were chosen based on preliminary optimization experiments aimed at eliminating the possibility of any non-specific (false positive) reconstitutions that may occur due to protein (over)expression [reviewed in 9495]. Overall, the conditions used for BiFC were based on those in which nVenus- or nVenus-tagged versions of the various TOM proteins, p36, p33, porin and the βATPase presequence co-expressed with the corresponding "empty" vector containing the nVenus or cVenus alone yield no significant BiFC (Venus) fluorescence (data not shown). Also, transformations with RFP (or βATPase-GFP or RFP-MFP) alone, confirmed that there was no cross-bleed of the red (or green) fluorescence signal into the yellow channel.

Topological orientations of Venus half fusion proteins were determined using digitonin (differential) permeabilization as described above, with the exception that biolistic bombardments were performed with gold particles (Sigma-Aldrich Ltd.) coated with 5 μg of plasmid DNA, and cells were formaldehyde fixed ~4 h after bombardment.

Chenopodium quinoa plants were grown in chambers at 21°C with a 12 h light-12 h dark cycle and five days prior to rub inoculation plants were transferred to a laboratory bench top with lower light conditions to facilitate the infection process 86. Rub inoculations (including mock inoculations) were carried out with 6-to-8-leaf stage plants and 10 μg of CIRV infectious cDNA diluted in 30 μl of inoculation buffer 86. Approximately 7 to 10 days after inoculations, ~1 cm² sections of the rub-innoculated leaf, including several necrotic lesions, were removed and processed for electron microscopy.

CIRV cDNA rub-inoculated C. quinoa leaf pieces were processed for transmission electron microscopy (TEM) by fixing in 4% (v/v) gluteraldehyde (Fisher Scientific, Markam, Canada) and 1% (v/v) acrolein (Sigma-Aldrich Ltd.) in 0.025 M NaKPO₄ buffer (pH 7.2) for 2 h in a vacuum chamber and then at 4°C overnight. Samples were then washed three times in 0.025 M NaKPO₄ buffer at 2 h intervals and 4°C, then post-fixed overnight at 4°C in 1% (w/v) osmium tetroxide (Fisher Scientific), dehydrated in a graded series of ethanol (50 to 100% [v/v]), and embedded in Spurr's medium-grade epoxy 96. Thin sections were cut with a glass or diamond knife, mounted on copper grids and post-stained with 2% (w/v) uranyl acetate and Reynold's lead citrate 97 for 10 and 3 min, respectively. All images were acquired at 80 kV with a Phillips (Mahwah, NJ, USA) CM10 TEM.

National Center for Biotechnology Information (NCBI) and/or Arabidopsis Genome Initiative (AGI) database accession numbers for proteins described in this study are as follows: βATPase (P17614), Cb5D (AAT84461), CIRV p36 (CAA59477), CIRV p95 (NP_612580), METAXIN (NP_565446, At2g19080), MFP (AAL35606), mtOM64 (NP_196504; At5g09420), porin (Q9SRH5; At3g01280), TBSV p33 (NP_062898), TOM20 (NP_189344; At3g27080), TOM22 (NP_563699; At1g04070), TOM40 (NP_188634; At3g2000).