An Arabidopsis thaliana F1-ATPase pFAγ subunit mitochondrial targeting peptide (MTP; pFAγ) that has previously been functionally validated in Arabidopsis protoplasts (Lee et al., 2012) was tested for its ability to traffic transgenic proteins to the matrix of intact plant leaf cells using the N. benthamiana transient leaf assay system. The entire pFAγ (77 AA) as previously described was translationally fused to the N terminus of GFP (pFAγ::GFP), where the first 42 AA (~4.6 kDa) of pre-sequence is predicted to be removed via the matrix-located peptidase (Figure 1A). As a control to discriminate matrix-processed GFP, several alanine amino acid substitutions were introduced in regions of the MTP required for its mitochondrial recognition and processing by the mitochondrial processing peptidase (MPP; Lee et al., 2012), thus this version (mFAγ::GFP) should be unable to be correctly processed and rather produce a full length fusion protein ~4.6 kDa larger than pFAγ::GFP. Five-week-old N. benthamiana leaves were infiltrated with either pFAγ::GFP or mFAγ::GFP. SDS-polyacrylamide gel electrophoresis (PAGE) and Western blots were carried out on crude protein extracted 4 days post infiltration (4 dpi) using a GFP antibody. For pFAγ::GFP, a band was observed corresponding to the expected size (~30 kDa) for matrix-processed GFP polypeptide, whereas for mFAγ::GFP, a higher MW band was observed (~35 kDa) of the expected size for unprocessed GFP fusion polypeptide (Figure 1B). A fainter band at ~28 kDa was observed for both pFAγ::GFP and mFAγ::GFP. This was not observed in negative controls and therefore may represent a degradation product, or possibility a product arising from alternative transcription or translation. To determine the subcellular localization of the GFP fusion proteins, protoplasts were isolated 3 dpi from leaf tissues containing either pFAγ::GFP, mFAγ::GFP, or no vector, and were examined by confocal microscopy after staining their mitochondria with MitoTracker. In protoplasts expressing pFAγ::GFP, the GFP signal fully co-localized with the mitochondrial marker (Figures 1G–J, white arrowheads) and was absent in control protoplasts (Figures 1C–F). In contrast, in protoplasts expressing mFAγ::GFP the GFP signal only partially localized to mitochondria (Figures 1K–N, white arrowheads), and a large fraction of the GFP signal was found to be targeted to other subcellular areas (Figures 1K–N, open arrowheads). Taken together these analyses indicate that the pFAγ MTP was capable of translocating the GFP fusion polypeptide to the matrix in N. benthamiana leaf cells and cleavage by the MPP, while the use of mFAγ MTP resulted in largely mis-targeted fusion proteins.

We next wanted to test if the pFAγ MTP was capable of directing bacterial nitrogenase proteins to the matrix. For this, two Nif proteins were chosen (NifF and NifZ) because their relatively small sizes would enable clear discrimination of the processed size by Western blotting. Protein-coding regions for the K. pneumoniae NifF and NifZ polypeptides were codon optimized for eukaryotic expression and fused to either HA or FLAG epitopes as C-terminal fusions (Figures 2A,B and Supplementary Table 1). The pFAγ MTP was fused to the N terminus of these polypeptides as for GFP above. To generate unprocessed versions of these pFAγ:: Nif proteins the same constructs were expressed in Escherichia coli by T7 RNA polymerase. Given that the MTP cannot be processed in bacteria which have neither mitochondria nor MPP, the difference in size between plant and bacterially- expressed proteins would enable processing to be validated by SDS-PAGE and Western blot. The Western blot revealed that the size of the polypeptides extracted from the N. benthamiana leaves was smaller in each case than the corresponding polypeptide produced in E. coli (Figures 2A,B). For each of pFAγ::NifF::HA and pFAγ::NifZ::FLAG, the polypeptides extracted from the plant cells corresponded to the sizes predicted for cleavage of the fusion polypeptides in their MTPs, whereas the polypeptides detected in E. coli extracts were of the expected sizes for unprocessed pFAγ::Nif fusion polypeptides. From these data we could conclude that the pFAγ MTP was capable of both directing Nif-fused polypeptides into the matrix and cleavage of the MTP by the MPP in plant cells.

Finally, we validated that the predicted MTP processing site was cleaved using mass spectrometer (MS) analysis. For this we chose NifH-HA, due to its importance as a core component of the nitrogenase enzyme complex. pFAγ::NifH::HA was expressed in plants, immunoprecipitated on HA-antibody agarose, further enriched via denaturing SDS electrophoresis (Figure 2C), and subjected to in-gel digestion with trypsin followed by tandem MS analysis of the resultant peptides. This analysis found 5 fully tryptic peptides identical to regions within NifH and a sixth semi-tryptic peptide consistent with exact cleavage of the MTP between residues 42 and 43 (Figure 2D). The tryptic peptide SISTQVVR that would been obtained from an unprocessed MTP was not observed. Instead, the most N-terminal peptide that was detected was the semi-tryptic ISTQVVR, confirmed by a complete series of y-ions in its MS/MS spectrum. These data conclusively demonstrated that the pFAγ::NifH::HA polypeptide had been cleaved at the preselected site in the MTP within the N-terminal extension by the matrix processing protease. The data implied that the pFAγ MTP contained all of signals necessary for translocation and processing in the matrix.

Given the success of expression, detection, and processing of GFP, NifF, NifZ, and NifH polypeptides in N. benthamiana mitochondria, we attempted to determine if all remaining Nif proteins required for nitrogenase biosynthesis and function could be expressed in N. benthamiana mitochondria. In the model diazatroph K. pneumoniae, 16 Nif proteins are involved in nitrogenase biosynthesis or catalysis, while four others are of unknown function or involved in transcriptional regulation (Oldroyd and Dixon, 2014). We therefore chemically synthesized eukaryotic expression codon-optimized versions of all these essential biosynthetic and catalytic nif genes of K. pneumoniae, each fused to the previously validated pFAγ MTP (Figure 3A, Supplementary Table 1). Each genetic construct encoded a fusion polypeptide having an N-terminal pFAγ MTP, then the Nif sequence, followed by a C-terminal extension comprising either an HA or FLAG epitope for detection by the appropriate antibody. For plant expression, each construct included the 35S promoter and nopaline synthase (nos) 3′ transcription terminator regions flanking the protein coding region.

A. tumefaciens cells containing each of the 16 genetic constructs were separately infiltrated into N. benthamiana leaves and, 4 days later, protein extracts prepared and analyzed by SDS-PAGE and Western blotting as before. For each of the constructs encoding HA-tagged pFAγ-Nif polypeptides, bands were detected on the Western blots which were approximately the size as predicted for the matrix-processed polypeptide (Figure 3B). Protein abundance varied among the HA-tagged polypeptides, with the easiest to detect being the NifB, NifH, NifK, NifS, and NifY fusion polypeptides. The NifF, NifE, and NifM polypeptides were present at lower levels, whereas detection of the NifQ polypeptide required a longer exposure of the blot to be visible (Supplementary Figure 1). Interestingly, additional, higher-molecular weight bands were detected for the infiltrations with the NifB, NifS, NifH, and NifY constructs that were size-specific to each individual construct. These additional bands were approximately twice the molecular weight of the primary band, suggesting that these polypeptides were dimerising despite the denaturing conditions during gel electrophoresis. It has been reported the NifB, NifS, and NifH proteins function as homo-dimers in bacteria (Yuvaniyama et al., 2000; Rubio and Ludden, 2008).

FLAG-tagged pFAγ::Nif::FLAG fusion polypeptides were observed in the Western blots for each of the constructs including the NifJ, NifN, NifV, NifU, NifX, and NifZ sequences (Figure 3B). The FLAG antibody yielded more background bands from the N. benthamiana extracts than the HA antibody. Nevertheless, the results were similar to those for the HA-tagged proteins. Considerable variation was seen in signal intensity for the different pFAγ::Nif::FLAG fusion polypeptides. An additional, higher-molecular-weight band was observed that was specific for the NifU construct (Figure 3B). The pFAγ::NifX::FLAG construct also yielded an additional, smaller band of greater intensity than expected for the predicted, processed molecular weight. Despite numerous replications of the infiltrations, we were unable to detect any specific bands for the pFAγ::NifD::FLAG construct in these plant assays, even though expression of the same genetic construct in E. coli readily yielded a visible band of the expected molecular weight (Figure 4A). The result from the bacterial extracts confirmed that the genetic construct containing pFAγ::NifD::FLAG was translationally competent.

Given the key role of NifD in the catalytic activity of nitrogenase (Eady et al., 1972), we tried to identify the reason for the lack of NifD fusion polypeptide production and tested several approaches to improve its abundance in plant assays. Firstly, we tested whether the lack of MTP-NifD fusion polypeptide production/accumulation could be attributed to low transgene transcription or mRNA instability, by measuring the mRNA expression level. To do this, the level of mRNA in the infiltrated N. benthamiana cells from pFAγ::NifD::FLAG was first measured by qRT-PCR and compared to the level of mRNA transcribed from the construct encoding pFAγ::NifU::FLAG in N. benthamiana cells (Figure 4B). This second construct was used as a control since it provided high levels of polypeptide production in the plant cells, as described above (Figure 4C). In order to remove any bias in amplification efficiency, oligonucleotide primers were used which annealed within the pFAγ MTP region shared by both nif fusion genes (illustrated in Figure 4B, Supplementary Table 2). The results from the qRT-PCR assay showed that the level of pFAγ::nifD::FLAG mRNA was actually slightly higher, although not significantly (P = 0.0683), than the level of pFAγ::nifU::FLAG mRNA. Secondly, cDNA was synthesized from the plant-produced mRNA transcribed from pFAγ::nifD::FLAG, cloned and sequenced, and its nucleotide sequence proved to be base perfect. Therefore, the lack of accumulation of the pFAγ::NifD::FLAG fusion polypeptide was not due to low mRNA expression or instability. These experiments also showed that the T-DNA containing and encoding pFAγ::NifD::FLAG was fully functional and that the 35S promoter in that construct was likewise functional. Therefore, the inability to detect the MTP-NifD fusion polypeptide in N. benthamiana cells was not due to any lesions in expression of the mRNA.

As transcription of the gene encoding pFAγ::NifD::FLAG and mRNA accumulation was clearly not limiting MTP-NifD production, several modifications were made to the genetic construct in attempts to improve protein accumulation (Figure 5A). Firstly, the possibility was tested that the presence of the FLAG epitope in the C-terminal extension was causing either lack of production or instability of the NifD fusion polypeptide. For this, a construct was designed in which the FLAG epitope was substituted with an HA epitope, designated pFAγ::NifD::HA. The HA epitope had allowed for accumulation and detection of 9 other MTP-Nif proteins (Figure 3B). Secondly, the codon usage of the NifD::HA open reading frame was modified in an attempt to determine whether a different mRNA sequence might improve translation efficiency. For this, an A. thaliana codon optimization algorithm was utilized (Graf et al., 2009), whereas the previous construct used a H. sapien algorithm (Graf et al., 2009). This construct was designated pFAγ::NifD_At::HA, and encoded an identical polypetide to pFAγ::NifD::HA. Additionally, a genetic construct was made encoding a version of the NifD fusion polypeptide with the mutated mFAγ N-terminal extension rather than pFAγ (mFAγ::NifD::HA) as described earlier (Figure 1A). This construct was made in order to test whether mitochondrial targeting and/or processing, if they occurred, were at least partially responsible for the lack of NifD fusion polypeptide production.

N. benthamiana leaves were infiltrated with A. tumefaciens containing these constructs and protein extracts prepared from the infiltrated tissues and analyzed. Encouragingly, the Western blot showed HA-containing bands of the molecular weights expected for both matrix-processed and unprocessed NifD fusion polypeptides when either pFAγ::NifD::HA or pFAγ::NifD_At::HA constructs were introduced (Figure 5B). Introduction of mFAγ::NifD::HA yielded only the larger (unprocessed) fusion polypeptide, showing that as for mFAγ::GFP (Figure 1A), this modification of the pFAγ MTP disrupts matrix processing. Matrix processing was further verified by comparing the position of the bands with that produced from mFAγ::NifD::HA and the bacterially produced polypeptides (Figure 5B). The observation of two bands, corresponding in size to processed and unprocessed forms of pFAγ::NifD::HA or pFAγ::NifD_At::HA indicated that processing of this MTP-NifD fusion polypeptide was not as efficient as for the other MTP-Nif fusion polypeptides as described above. The observed level of the pFAγ::NifD::HA or pFAγ::NifD_At::HA polypeptides were still much lower than the level of the pFAγ::NifK::HA fusion polypeptide used as a positive control, despite the same expression construct design and expression conditions. Furthermore, for each of the three modified NifD::HA constructs, additional bands of lower molecular weight were observed on the Western blots, some of which were specific to a particular pFAγ::NifD::HA version. For example, an intense band at about ~48 kDa was present for all of the modified NifD::HA constructs, but a different, intense band at about 40 kDa appeared unique when mFAγ::NifD::HA was introduced. As these bands were not present for either pFAγ::NifK::HA or GFP controls, they could represent NifD::HA degradation products or possibly the product of alternative transcription or translation initiation signals.

In diazatrophic bacteria the abundance of NifD and NifK are almost equal (Poza-Carrion et al., 2014) and both are found in crystal structures of nitrogenase in a 1:1 ratio as a heterodimer (Schmid et al., 2002). However, it was clear from the above analysis that NifK accumulation was significantly higher than NifD. Given that NifD and NifK form the catalytic unit of nitrogenase, functional activity in plants would likely benefit from a higher level of NifD and a stoichiometric 1:1 NifD:NifK ratio. We therefore designed a translational fusion between NifD and NifK to link the expression of these proteins, as similar recombinant fusion strategies have been shown to improve accumulation (Hondred et al., 1999). This approach required the design of an amino acid linker joining the NifD and NifK units whilst still allowing the proper protein folding of the entire fused D-K polypeptide, as follows. A homology model of the FeMoco protein (NifD and K) α₂β₂ hetero-tetramer from K. pneumoniae was constructed using the crystal structure of the FeMoco protein complex from Azotobacter vinelandii (PDB ID: 1FP4; Schmid et al., 2002) as a template. In the structural model for the K. pneumoniae D₂K₂ hetero-tetramer, the C-terminus of each NifD subunit was approximately 47 Å from the N-terminus of the NifK partner. Therefore, an unstructured and flexible amino acid linker of 30 residues was designed to span this 47 Å length (Figure 5A; see Section Materials and Methods for full details of the in silico design protocol). The predicted structure of the NifD-linker-NifK composite polypeptide, including its 30-amino acid linker, was examined, showing that addition of the linker allowed the full fusion polypeptide to fold into a structure that mimicked the native MoFe complex (Figure 5C). Finally a HA epitope tag was appended to the C-terminus of NifK (as previously for singly expressed NifK) to enable discrimination of the fusion protein with HA antibody, here termed pFAγ::NifD-linker-NifK::HA (Figure 5A). When this vector was introduced into N. benthamiana cells and protein extracts examined by Western blotting, two specific bands were detected in the size range predicted for the polypeptide at ~120 kDa. The closeness of the two bands suggested that both unprocessed and processed forms for the MTP presequence were present. The upper band was of greater intensity than the lower band, suggesting that processing occurred only partially. Nevertheless, the level of accumulation of the pFAγ::NifD-linker-NifK::HA polypeptide was greater than for pFAγ::NifD::HA, but much less than for pFAγ::NifK::HA.

As nitrogenase activity requires the concerted action of many Nif proteins, it is anticipated that functional nitrogenase reconstitution in plants will require many of the Nif proteins utilized by diazotrophic bacteria for biosynthesis and function. We therefore wanted to determine if multiple Nif proteins could be expressed in N. benthamiana using the pFAγ MTP. To test this concept, genetic constructs encoding four Nif fusion polypeptides of different sizes were chosen that would enable the resultant polypeptides to be identified by Western blot analysis, namely the constructs encoding pFAγ::NifB::HA, pFAγ::NifS::HA, pFAγ::NifH::HA, and pFAγ::NifY::HA. The four A. tumefaciens cultures transformed with these constructs were mixed in equal amounts and the mixture was then infiltrated into N. benthamiana leaves. The accumulated polypeptide levels in this four construct combination were compared to infiltrations with single constructs. It was observed that each polypeptide was more abundant when expressed from a single construct than from the four gene combination (Figure 6). Nevertheless, all four Nif fusion polypeptides were readily detected in the protein extracts from the gene combination, and the molecular weight observed for each polypeptide was identical for the individual and combination infiltrations. This showed that combinations of Nif fusion polypeptides could be produced in the plant cells with the desired targeting and processing of each Nif fusion polypeptide to the mitochondria.

Vector pCW441 was designed as dual purpose bacterial and plant expression construct. A region of the pET14b vector (Novogene) was used to design a region of DNA encompassing the T7 promoter, a simple multiple cloning site (NotI and AscI) and the T7 terminator, and this region was chemically synthesized (Geneart) and ligated into a T-DNA-based plant expression vector, pORE1 (Coutu et al., 2007) in between the 35S promoter and the NOS terminator regions, generating pCW441. This vector is stable in E. coli, and can produce proteins via the standard T7 polymerase promoter system (Studier and Moffatt, 1986). The 240 bp sequence incorporating the 77 amino acid MTP of the F1-ATPase γ subunit (Lee et al., 2012) was chemically synthesized with a NotI and AscI flanking region permitting ligation into pCW441, generating pRA1. Klebsiella oxytcoa Nif genes were codon optimized for eukaryotic expression and commercially synthesized (Geneart). Nif gene sequences were flanked by AscI sites for subcloning and contained either HA or FLAG tags as C terminal fusions and designed to allow a translational fusion with the N-terminal MTP. GFP sequence was derived from S65T GFP sequence (Ormö et al., 1996). The full compendium of constructs used in this study are found in Supplementary Table 1.

Plant growth and Agrobacterium infiltrations were carried out as described by Naim et al. (2012).

Infiltrated N. benthamina leaf proteins were extracted by grinding an ~2 × 2 cm leaf disc in liquid nitrogen then transferring the powder to 300 μL of buffer comprised of 125 mM Tris-HCL pH 6.8, 4% SDS, 20% glycerol, 60 mM DTT. Samples were heated at 95°C for 3 min before centrifugation at 12,000 g for 2 min. Protein samples (20 μL) were separated by SDS-PAGE (NuPAGE Bis Tris 4–12%, Thermofisher.com) at 200 V for 1 h. Proteins were transferred to PVDF membranes using an iblot system (Thermofisher). After blotting gels were Coomassie stained for 1 h then rinsed in water for visualization of remaining protein. Membranes were blocked overnight in TBST + 5% skim milk powder at 4°C. Anti-HA, and anti-FLAG were purchased from Sigma, anti-GFP was a gift from Leila Blackman (ANU). Antibodies were added at 1:5,000 in TBST with 5% skim milk powder and incubated for 2 h. Membranes were washed for 3 × 20 min with TBST and the secondary antibody [Immun-Star Goat Anti-Mouse (GAM)-HRP conjugate] (Biorad) was added at 1:5,000 in TBST + 5% skim milk for 1 h, followed by 3 × 15 min TBST washes. For secondary antibody detection Amersham ECL reagent was used and membranes were developed either with an X-ray developer or on an Amersham imager (Amerhsam).

Protoplasts from infiltrated and untransformed N. benthamiana leaves were prepared as described before (Rolland et al., 2016). Mitochondria were stained for 10–20 min using a 100 nM solution of MitoTracker Red CMXRos (ThermoFisher Scientific). Protoplasts were then imaged using an upright Leica SP8 laser-scanning confocal microscope with a 40x water immersion objective (NA = [1.1]). GFP was excited at 488 nm and emission was recorded at 495–520 nm. In the same track, MitoTracker Red CMXRos was excited at 580 nm and emission was recorded at 595–620 nm. In a separate track, chloroplasts were excited at 633 nm and emission was recorded at 650–690 nm.

RNA was extracted from infiltrated N. benthamina leaves by grinding an ~2 × 2 cm leaf disc in liquid nitrogen then transferring the powder to 500 μL of Trizol (Life Technologies) buffer for RNA purification. RNA expression analysis was as described previously (Allen et al., 2010) and N. bentahamiana GADPH was used to normalize gene expression. Primers are described in Supplementary Table 2.

A homology model of the NifD-K α₂β₂ heterotetramer from K. pneumoniae was constructed using the crystal structure of the NifD-K complex from A. vinelandii (PDB ID: 1FP4) as a template. The C-terminus of NifD was ~47 Å from the N-terminus of its NifK partner so a linker of appropriate length was designed to connect the two units. The linker was 30 residues in length consisting of an 11-residue section from a known unstructured linker region from Hypocrea jecorina cellobiohydrolase II (Accession no. AAG39980.1) with the final arginine replaced by an alanine, followed by an 8-residue FLAG-tag and finally by another copy of the 11-residue unstructured linker sequence with the arginine replaced by an alanine. A geometry optimization and equilibration at constant pressure was carried out on the tetramer using an octahedral TIP3P water box with minimum boundary distance from the solute of 10.0 Å (without the inclusion of any of the metal centers). This calculation was carried out using Amber 12 (Case et al., 2012) employing the ff99SB force field at 298 K over 5 ns in total using 2 fs time steps and SHAKE constraints. The final NifD-linker-NifK sequence is found in Supplementary Table 1.

Infiltrated N. benthamiana tissue was ground in liquid N₂ and processed using a Retsch tissuelyser in 2 mL Eppendorf tubes in 50 mM Tris-HCL pH7.5, 1 mM EDTA, 150 mM NaCl, 0.2% (w/v) SDS, 10% (v/v) glycerol, 5 mM DTT, 0.5 mM PMSF, and 1% (v/v) protease inhibitor cocktail for plants (Sigma). Protein extracts were cleared by centrifugation and used as input for overnight incubation with monoclonal anti-HA conjugated agarose beads (Sigma). Unbound proteins were removed by a series of washes with 150 mM Tris-HCL pH 7.5, 5 mM EDTA, 150 mM NaCl, 0.1% (w/v) Triton X-100, 5% (v/v) glycerol, 5 mM DTT, 0.5 mM PMSF, and 1% protease inhibitor cocktail for plants (Sigma). Bound proteins were eluted by incubating beads in Laemmli buffer at 95°C for 10 min. Input and IP protein samples were separated by SDS-PAGE and the area of gel determined to contain MTP-NifH-HA was determined by Western analysis upon a duplicated sample. From the replicate gel, the corresponding region was excised for in-gel tryptic digestion and tandem mass spectrometer analysis as previously described using an Agilent Chip Cube system coupled to an Agilent Q-TOF 6550 mass spectrometer (Campbell et al., 2014). Mass spectra derived from tryptic peptides from common contaminants such as the added trypsin and human keratin were identified before the remaining mass spectral data were used to search against a database containing all protein sequences from Nicotiana species from the NCBInr database (10/3/2015) plus the HA sequence using Spectrum Mill software (Agilent Rev B.04.01.141 SP1) with a precursor mass tolerance of 15 ppm, product mass tolerance of 50 ppm, default Q-TOF scoring, and stringent default “autovalidation” settings. Modification of cysteine residues by acrylamide was a required modification and oxidation of methionine was allowed as a variable modification. Initially, tryptic cleavage was required and up to two missed cleavages were allowed. After validating peptide matches, the search was repeated with the remaining unmatched spectra allowing non-tryptic cleavage.

This work is subject to a patent by CSIRO in which RA, KT, AW and CW are inventors. The other authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.