A large number of nuclear-encoded proteins are targeted to the organelles of endosymbiotic origin, namely mitochondria and plastids. To determine the targeting specificity of these proteins, fluorescent protein tagging is a popular approach. However, ectopic expression of fluorescent protein fusions commonly results in considerable background signals and often suffers from the large size and robust folding of the reporter protein, which may perturb membrane transport. Among the alternative approaches that have been developed in recent years, the self-assembling split-fluorescent protein (sasplit-FP) technology appears particularly promising to analyze protein targeting specificity in vivo. Here, we improved the sensitivity of this technology and systematically evaluated its utilization to determine protein targeting to plastids and mitochondria. Furthermore, to facilitate high-throughput screening of candidate proteins we developed a Golden Gate-based vector toolkit (PlaMinGo). As a result of these improvements, dual targeting could be detected for a number of proteins that had earlier been characterized as being targeted to a single organelle only. These results were independently confirmed with a plant phenotype complementation approach based on the immutans mutant.
Many nuclear-encoded proteins are targeted across cellular membranes to reach their final destination in the cell, i.e. a sub-cellular compartment or cell organelle. To determine the specificity of such targeting, fluorescent protein tagging – the fusion of the candidate protein with a fluorescent reporter (e.g. green fluorescent protein, GFP) and subsequent in vivo imaging by fluorescence microscopy – is a popular approach. However, this widely utilized method carries some inherent pitfalls (reviewed by Moore and Murphy, 2009). Firstly, the large size and intrinsic folding properties of the reporter protein might perturb membrane transport of the candidate protein (Marques et al., 2004). Secondly, considerable background fluorescence signals occur as a result of protein overexpression, and hence saturation of the organelle transport machinery is sometimes observed (Sharma et al., 2018a). And finally, proteins targeted to an organelle in low amounts are difficult to visualize using this approach owing to weak fluorescence signals originating from a limited number of protein molecules.
These constraints become particularly prominent when it comes to the analysis of proteins targeted to mitochondria and/or plastids (Tanz et al., 2013). These organelle-targeted proteins usually carry an N-terminal transport signal (or transit peptide) to facilitate transport of the passenger protein across the membranes via organelle-specific translocation machineries. Amongst a repertoire of proteins encoded in the plant cell nucleus, more than 3000 are transported into mitochondria and/or plastids (Van Wijk and Baginsky, 2011; Rao et al., 2017). While most of these proteins are targeted to only one type of organelle, a group of proteins exists with dual-targeting specificity, i.e., they carry an ‘ambiguous’ transit peptide capable of translocating the passenger protein into both mitochondria and plastids (reviewed in Sharma et al., 2018b). Despite having dual-targeting properties, some of these proteins are still preferentially targeted to one of the two organelles. In this case, determining targeting specificity using a conventional fluorescent protein-tagging approach is particularly tedious and error-prone, as highly intensive fluorescence signals coming from one organelle can mask signals of low-intensity coming from the other (Duchêne et al., 2005; Sharma et al., 2018a).
One possibility to circumvent these problems is to detect the fluorescence signals coming from each organelle in separate cells. This has become technically feasible with the invention of self-assembling split green fluorescent protein (sasplit-GFP) technology (Cabantous et al., 2005). Spontaneous self-assembly of split-GFP relies on two highly engineered fragments derived from the ‘superfolder’ GFP variant (sfGFP). The large fragment, GFP1-10 OPT (hereafter referred to as GFP1-10), comprises ten N-terminal antiparallel β-sheets of GFP. The smaller fragment, GFP11 M3 (hereafter referred to as GFP11), comprises only 16 amino acids and represents the C-terminal eleventh β-sheet of GFP. When brought into close proximity, the two fragments can assemble spontaneously to reconstitute the functional fluorophore without the need of an additional interacting partner (Fig. 1A) (Cabantous et al., 2005). For subcellular protein localization studies, GFP1-10 is fused to a transport signal of known organelle specificity and analyzed together with a chimeric protein comprising the candidate protein fused to GFP11. Fluorescence complementation is achieved specifically and exclusively if the transport signal of the candidate mediates transport of GFP11 into the organelle housing GFP1-10. The fluorescence signal remains limited to the compartment containing GFP1-10, irrespective of whether the GFP11-fused candidate protein is targeted to further subcellular compartments or not (Fig. 1B). Thus, sasplit-GFP technology allows selective in vivo imaging of a protein of interest in a respective compartment with enhanced signal-to-noise ratio. Furthermore, the small-sized GFP11 tag could be hypothesized to have a lower propensity for interfering with membrane transport in comparison to the full-length fluorescent proteins used for direct fluorescent protein tagging.
The sasplit-GFP system has already been applied to a wide range of organisms and adapted to elucidate a variety of cellular functions or processes including, for example, in vivo protein solubility, sub-cellular localization of pathogen effectors, endogenous protein labeling for in vivo imaging, protein–protein interaction studies and membrane protein topology determination (Cabantous and Waldo, 2006; Van Engelenburg and Palmer, 2010; Machettira et al., 2011; Cabantous et al., 2013; Kamiyama et al., 2016; Henry et al., 2017). The principal suitability of the sasplit-GFP system for in vivo imaging of protein targeting in plant cells has also been demonstrated recently (Park et al., 2017). In this case, transgenic Arabidopsis thaliana lines expressing the organelle-targeted GFP1-10 receptor were transiently transformed with constructs expressing a candidate protein fused to the GFP11 tag. However, the requirement of transgenic ‘receptor’ plant lines and low transient transformation efficiency hampers the utilization of this approach for high throughput protein targeting studies. Along with a limited yield of transformed cells, the relatively low brightness of sasplit-GFP prevented the visualization of proteins targeted in low amounts into the organelles.
Therefore, we systematically evaluated and optimized the sasplit-GFP system for analysis of protein targeting specificity in plant cells and assessed the effect of multimerization of the GFP11 tag on the intensity of the fluorescence signals inside mitochondria and plastids. A Golden Gate-based vector toolkit named PlaMiNGo (analysis of plastid and/or mitochondrial targeted proteins N-terminally fused to GFP11 tags via Golden Gate cloning) was developed to facilitate high-throughput analyses of candidate proteins. With this approach, dual targeting to mitochondria and plastids could be detected for several proteins that were previously characterized as being targeted to a single organelle only. Importantly, plastid targeting of two of these proteins was independently confirmed using a phenotype complementation-based approach in stable transgenic Arabidopsis plants, which proves the in vivo relevance of results obtained with the PlaMiNGo system.
The sasplit-GFP system to determine protein targeting specificity
In order to study protein targeting into plastids, the transit peptide (FNR1-55) of a chloroplast protein, ferredoxin-NADP+ oxidoreductase (PETH) of spinach (Zhang et al., 2001), was fused to the GFP1-10 receptor (the large fragment of the sasplit-GFP system) to facilitate its localization into plastids. For mitochondrial localization of GFP1-10, the N-terminal 100 amino acid residues (mtRi1-100) comprising the pre-sequence of the mitochondrial Rieske iron-sulfur protein (FES1) of potato (Emmermann et al., 1994) were used. Both transport signals had previously been characterized for their targeting specificity to a single organelle with in vivo and in vitro approaches (Rödiger et al., 2011). In the initial experiments, evaluating the suitability of the sasplit-GFP system for our purposes, each transport signal was likewise combined with a GFP11 tag (the small fragment of sasplit-GFP system). A sevenfold repeat of this GFP11 tag (GFP11×7; separated via a five-amino-acid linker) was used, as such multiple GFP11 tags had been reported to intensify the fluorescence signals in mammalian cells (Kamiyama et al., 2016). When these FNR1-55/GFP11×7 and mtRi1-100/GFP11×7 constructs were co-expressed with the FNR1-55/GFP1-10 and mtRi1-100/GFP1-10 gene constructs, fluorescence signals were exclusively obtained in those instances in which the same transport signal was present in both chimeras, i.e. in plastids after co-expression of FNR1-55/GFP1-10 and FNR1-55/GFP11×7 and in mitochondria after co-expression of mtRi1-100/GFP1-10 and mtRi1-100/GFP11×7 (Fig. 1C,D). When infiltrated alone, neither of these constructs generated any detectable fluorescence signal. The chimeric GFP1-10 accumulated in the organelles and could be detected in protein extracts from the infiltrated leaf tissue (Figs S1, S2). This demonstrates the suitability and specificity of the sasplit-GFP system for the analysis of protein targeting.
Next, we wanted to evaluate the performance of this system by determining the protein-targeting behavior of several dual-targeted proteins. For this purpose, we selected three previously characterized nuclear-encoded organelle proteins from Arabidopsis thaliana with proven dual-targeting characteristics, namely TyrRS (tyrosyl-tRNA synthetase; At3g02660), GrpE (co-chaperone GrpE1; At5g55200), and PDF (peptide deformylase 1B; At5g14660) (Berglund et al., 2009; Baudisch et al., 2014). The experiments employing eYFP (enhanced yellow fluorescent protein) fusions indicated that all three proteins are targeted to both endosymbiotic organelles as seen previously (Sharma et al., 2018a), either in comparable amounts (TyrRS1-91/eYFP) or preferentially to either mitochondria (GrpE1-100/eYFP) or chloroplasts (PDF1-100/eYFP) (Fig. 2). The respective N-terminal amino acid sequences carrying the organelle transport signals of the candidate proteins were fused to GFP11×7 tags and analyzed in our system. All three candidates (TyrRS1-91/GFP11×7, GrpE1-100/GFP11×7 and PDF1-100/GFP11×7) showed targeting to both plastids and mitochondria when co-transformed with the respective organelle-targeted receptors, FNR1-55/GFP1-10 and mtRi1-100/GFP1-10 (Fig. 2). Even plastid targeting of GrpE1-100/GFP11×7 was clearly visible, which is remarkable considering the vague fluorescence signals obtained in this organelle with the classical fluorescent reporter fusion (Fig. 2B). Thus, separation of the fluorescence signals for the two organelles into different cells proved to be advantageous to determine the low plastid-targeting properties of GrpE.
Multimerization of the GFP11 tag leads to fluorescence signal enhancement in plastids but not in mitochondria
In the initial experiments described above, we used GFP11×7, the sevenfold repeat of the GFP11 tag. However, the requirement or benefits of such multiple GFP11 tags to enhance fluorescence signals in mitochondria and plastids of plant cells were not systematically assessed. Hence, we compared the fluorescence signal intensity obtained in the two organelles with either seven (GFP11×7), three (GFP11×3) or a single (GFP11×1) repeat of the GFP11 tag when fused to the dual-targeted protein transport signal of TyrRS. It turned out that the use of a single GFP11 tag yields only very faint signals in plastids while the GFP11×3 and GFP11×7 tags significantly enhance the fluorescence signals (Fig. 3A). In contrast, for mitochondria, no correlation between the number of GFP11 tag repeats and fluorescence signal intensity was found. Instead, the signal intensities were largely similar for all three constructs (Fig. 3B). However, since the fluorescence signals obtained with the single GFP11 tag in mitochondria were brighter than in plastids, they are usually sufficient for proper visualization of this organelle.
Construction of the PlaMiNGo toolkit
To facilitate easy combination of sasplit-GFP technology and fluorescence signal enhancement for high-throughput screening of protein targeting specificity, we developed a set of Golden Gate-based vectors. Designed to analyze the targeting specificity of candidate proteins to plastids and mitochondria, these vectors enable rapid cloning of the candidate genes upstream of single or multiple GFP11 tags. In this PlaMiNGo toolkit, we used highly efficient gene regulatory elements, namely a ‘long’ 35S promoter (Engler et al., 2014) and ocs or rbcs E9 transcriptional terminators (De Greve et al., 1982; Coruzzi et al., 1984), to control expression of the chimeric genes (Fig. 4A; Fig. S3). Moreover, to avoid the requirement for performing co-transformation using two plasmids, we used a single transfer (T)-DNA expression system (Grefen and Blatt, 2012; Hecker et al., 2015) comprising two gene expression cassettes. This has the advantage that each transformed cell expresses both chimeras simultaneously. Consequently, the final T-DNA vectors contain two gene expression cassettes between left and right border elements. One of the expression cassettes encodes GFP1-10 gene chimeras to be targeted to either plastid or mitochondria by fusion with mtRi1-100 or FNR1-55, respectively, while the other expression cassette encodes either one, three or seven repeats of GFP11 tags. The latter expression cassette additionally carries a ccdB cassette upstream of the respective GFP11 tag, which can be replaced by the candidate gene in a single Golden Gate reaction step, allowing for background-free selection of positive clones. As a result, six vectors, three destined to analyze potential plastid-targeting and three for mitochondria-targeting analysis, were generated (Fig. 4A; Fig. S3).
To evaluate the functionality of these vectors, the dual-targeting transport signal of TyrRS (N terminal 1–91 amino acids) was cloned upstream of the GFP11 tags of all six vectors and analyzed via Agrobacterium-mediated transient transformation of Nicotiana benthamiana leaves. In comparison to the previous experiments involving co-infiltration of two vectors, the signal intensities were significantly improved ∼ninefold in plastids and ∼threefold in mitochondria with the use of GFP11×7 and GFP11×1 tags, respectively (Fig. 4B; Fig. S4). Now, even in plastids a single copy of the GFP11 tag was sufficient for reliable detection of the reconstituted GFP fluorescence, although considerable signal enhancement could still be observed with GFP11×3 and GFP11×7 tags. This suggests that the GFP11×7 tag, when co-expressed with the plastid-targeted GFP1-10 receptor, should allow for detection even of minute amounts of plastid-localized proteins. Likewise, upon imaging of mitochondria the fluorescence signals obtained with the single GFP11 tag were stronger than the signals obtained in the previous experiments using separate vectors. Still, as observed earlier, no further improvement of signal intensities by multimerization of the GFP11 tag could be obtained in mitochondria (Fig. S4). Instead, artificial protein aggregates were observed in some cells expressing constructs with multiple GFP11 tags (GFP11×3 and GFP11×7) (Movie 1).
Multicolor imaging of dual-protein targeting to two organelles
A further objective of the study was to establish simultaneous multicolor imaging with the sasplit-FP system. For this purpose, we modified the GFP1-10 receptor to generate a yellow-shifted variant (YFP1-10) using a single amino acid substitution (T203Y) as reported earlier (Kamiyama et al., 2016). To test if this variant can assemble with GFP11 to generate a functional fluorophore inside the organelles, the plastid (FNR1-55) or mitochondrial (mtRi1-100) transport signals were separately fused to the N-terminus of YFP1-10. These fusions were co-infiltrated with a gene construct encoding the dual-targeting transport signal TyrRS1-91 fused to either GFP11×1 or GFP11×7. When imaged with an YFP-specific filter set, fluorescence signals were solely obtained in mitochondria (with mtRi1-100/YFP1-10 and TyrRS1-91/GFP11×1) or in plastids (with FNR1-55/YFP1-10 and TyrRS1-91/GFP11×7), demonstrating that the YFP1-10 fragment can indeed assemble with GFP11 in both organelles (Fig. S5A,B).
Next, we tested if multicolor imaging, i.e. the simultaneous labeling of plastids and mitochondria with YFP1-10 and GFP1-10, respectively, within the same cell is possible. For this purpose, the FNR1-55/YFP1-10 fusion was co-infiltrated with a pMiNGo11-1:TyrRS vector comprising mtRi1-100/GFP1-10 and TyrRS1-91/GFP11×1. This should result in transformed cells co-expressing the dual-targeted GFP11 (via fusion with TyrRS1-91) and two different receptors targeted to two different organelles, namely YFP1-10 to plastids (FNR1-55/YFP1-10) and GFP1-10 to mitochondria (mtRi1-100/GFP1-10). Indeed, the resulting transformed cells emitted fluorescence signals of different spectra in the two organelles owing to reassembly of the GFP11 tag with both the GFP1-10 and YFP1-10 receptors within mitochondria and plastids, respectively (Fig. S5C). However, in all transformed cells a certain degree of bleed-through of signals, i.e., the appearance of YFP fluorescence signals in GFP channel and vice versa, could be detected. The adjustment of the filter sets to avoid such bleed-through inevitably led to significant reduction of the fluorescence signal intensity. In summary, the YFP1-10 derivative of the sasplit-GFP system is not yet perfect for multicolor imaging but represents a promising basis for development of such tools.
Analysis of protein targeting specificity with PlaMiNGo
Finally, we examined the suitability of the PlaMiNGo toolkit using eight candidates with presumed targeting specificity either to plastids, mitochondria, or to both organelles (Table 1). Three of these proteins, namely Gtred (Monothiol glutaredoxin-S15, also known as GRXS15), GCS (Glycine cleavage system H protein 1, also known as GDH1) and GAPDH (Glyceraldehyde-3-phosphate dehydrogenase B, also known as GAPB) had earlier been reported to be dual-targeted (Baudisch et al., 2014). Two other candidates, namely FNR (also known as PETH) and RbcS (small subunit of Rubisco from pea), are well-characterized plastid proteins (Highfield and Ellis, 1978; Zhang et al., 2001), while the residual three candidates, namely mtRi (also known as FES1), ATPS (ATP synthase subunit beta-3; At5g08680) and CoxIV (Cytochrome c oxidase subunit IV) of yeast, are known for their mitochondrial targeting specificity (Maarse et al., 1984; Emmermann et al., 1994; Baudisch et al., 2014). The protein fragments comprising the transport signals of these proteins were cloned as fusions with GFP11 tags into the PlaMiNGo vectors, additionally comprising the organelle-targeted receptor fusions FNR1-55/GFP1-10 or mtRi1-100/GFP1-10.
Five of the eight candidate proteins showed the same targeting behavior in our assay system as reported in the literature (Table 1): mtRi and FNR showed exclusive transport into either mitochondria or plastids, respectively (Fig. 5A,B) (Rödiger et al., 2011), and the dual-targeting candidates GCS, Gtred and GAPDH showed transport into both organelles (Baudisch et al., 2014) (Fig. 5C–E). It should be noted though that in the case of GAPDH, mitochondrial targeting was rather weak in our assays and could be observed only in a few transformed cells. However, it was rather unexpected that the remaining three monospecific candidates, namely ATPS1-100, RbcS1-79 (SSU1-79) and CoxIV1-29, showed dual targeting in our experiments (Fig. 6). In the literature, mitochondrial targeting of RbcS has already been described once (Rudhe et al., 2002) but these results were solely based on in vitro assays. In addition, dual targeting of the yeast mitochondrial pre-sequence, CoxIV1-29, could be assumed considering the high degree of freedom of non-plant mitochondrial transport signals (Staiger et al., 2009). However, this dual targeting was entirely unexpected for the plant mitochondrial protein ATPS1-100, which has never previously been shown to target to plastids when analyzed with in vivo fluorescent protein tagging and in vitro protein transport experiments (Baudisch et al., 2014) (Fig. S6A).
Phenotype complementation confirms the plastid targeting properties of ATPS
These unexpected results demanded independent confirmation. Thus, to re-evaluate the plastid-targeting properties of the transport signal of mitochondrial ATPS, we applied a phenotype complementation approach. In this approach, we made use of the immutans mutant of Arabidopsis thaliana that shows a white-green sectored (variegated) leaf phenotype and stunted growth when grown under daylight conditions (Fig. 7A,B). The mutant phenotype is caused by the absence of a functional nuclear-encoded plastid protein, namely plastid terminal oxidase (PTOX, also known as AOX4) (Carol et al., 1999; Aluru et al., 2001). Complementation of the variegated phenotype of immutans requires a functional PTOX protein in plastids (Fu et al., 2005) (Fig. 7). To adapt PTOX for our analysis, we replaced the authentic transit peptide of PTOX with the N-terminal 100 amino acid residues of two candidate proteins, namely ATPS and GCS (a validated dual-targeted protein), generating ATPS1-100/mPTOX and GCS1-100/mPTOX, respectively. Two further constructs encoding either the authentic PTOX precursor or the transit peptide-free mature PTOX protein (mPTOX) were generated for comparison. These gene chimeras were expressed under the control of the CaMV35S promoter in the immutans mutant background. As expected, both the authentic PTOX precursor and GCS1-100/mPTOX were able to complement the variegated phenotype (Fig. 7C,E). In contrast, the expression of mPTOX cannot complement the mutant phenotype (Fig. 7D). These results confirm that targeting of PTOX into plastids is essential for complementation of the immutans variegated phenotype. Remarkably, the expression of ATPS1-100/mPTOX also resulted in mutant phenotype complementation (Fig. 7F). However, such complementation was observed in only four out of six transgenic lines analyzed, suggesting that plastid targeting of ATPS1-100/mPTOX is, in principle, possible but apparently less efficient than with typical plastid-targeted transit peptides. These results clearly underline the basic plastid-targeting property of mitochondrial protein ATPS and thus re-confirm the results obtained with the sasplit-GFP technology established here.
The goal of this study was to assess the suitability of the signal-enhanced sasplit-GFP system to determine the targeting specificity of nuclear-encoded organelle proteins and to develop tools for rapid cloning and subsequent analysis of their targeting behavior. The characterization of the targeting specificity of organelle proteins is crucial (i) to elucidate the functional properties of organelle transport machineries, (ii) to study evolutionary aspects of protein targeting, and (iii) for biotechnological applications employing organelles. The application of the sasplit-GFP system, as demonstrated in this study, provides a novel toolbox to quickly determine the targeting properties of candidate proteins with high sensitivity.
Selective imaging and fluorescence signal enhancement with sasplit-GFP technology
The selective imaging of organelles is one of the major advantages of the sasplit-GFP system in comparison with classical fluorescent protein tagging approaches. The requirement for the presence of the non-fluorescing GFP1-10 receptor in a specific subcellular location is the key for selective imaging (Kaddoum et al., 2010). In this study, two transport signals, namely FNR1-55 and mtRi1-100, were selected for localization of the receptor specifically within two sub-cellular locations, the plastid stroma and the mitochondrial matrix, respectively. As a result, fluorescence signals will appear only if the GFP11-tagged protein is completely imported into the same sub-cellular location and not if a protein is merely binding to the organelle surface. This was otherwise difficult to distinguish with fluorescent protein-based approaches, specifically for mitochondria because of their small size.
Fluorescence signal enhancement
The self-assembling split-GFP molecules have been reported to produce fluorescence signals of lower intensity than classical fluorescent proteins (Köker et al., 2018). This problem can be circumvented with the use of multiple GFP11 tags. However, the two organelles respond differently to this modification. While signal enhancement with multiple GFP11 tags works well in plastids, fluorescence signal enhancement could not be observed in mitochondria (Fig. 3). One possible reason might be the size difference between these organelles. Plastids are comparatively larger in size and the proteins in the plastid stroma are more dispersed. Consequently, the chances for self-assembly of sasplit-GFP fragments in this organelle are lower than in mitochondria and thus fluorescence signal enhancement could be observed by increasing number of GFP11 tags in plastids. Furthermore, differences in the physicochemical properties of the two organelles, e.g. pH, might likewise contribute to this phenomenon.
The use of more efficient gene regulatory elements, i.e. promoter and terminator, in the PlaMiNGo toolkit also led to significant enhancement in fluorescence signal intensity. However, in combination with multiple GFP11 tags, such increased gene expression can lead to the formation of aggregations in transformed cells, particularly if these tags are combined with mitochondria-targeting transport signals. The intrinsic property of the sasplit-GFP fragments to form dimers and aggregates (Cabantous et al., 2005) and comparatively less efficient unfoldase activity of the mitochondrial protein translocation machinery (Agarraberes and Dice, 2001) could be one of the possible reasons for this phenomenon. Since protein unfolding prior to translocation is apparently more efficient in plastids, the multiple GFP11 tags can be efficiently imported into this organelle. By contrast, high expression of plastid-targeted GFP1-10 alone could result in the appearance of faint fluorescence signals, probably due to the formation of dimers. In most instances, these faint signals are clearly distinguishable from ‘actual’ fluorescence signals obtained via self-assembly of split-GFP, but experimental controls should still take these signals into account to avoid the misinterpretation of results (Fig. S7).
High sensitivity of sasplit-GFP system
Fluorescence signal enhancement in combination with selective imaging makes the sasplit-GFP system highly sensitive with respect to targeting specificity determination. Subsequently, dual targeting of several proteins was newly detected with the PlaMiNGo toolkit developed here, which had previously been missed owing to the inherent limitations of classical fluorescent protein tagging approaches. For example, GAPDH shows dual targeting with the sasplit-GFP system, in line with the results of in vitro import experiments (Baudisch et al., 2014). In contrast, when analyzed with the classical fluorescent protein tagging approach, GAPDH appeared to be solely transported into plastids (Baudisch et al., 2014). Similarly, the transit peptide of RbcS is able to translocate the GFP11 tag into mitochondria, but this property remained undetected with the fluorescent protein tagging approach. Remarkably, supporting the above observation, such dual targeting of the RbcS transit peptide was also witnessed in a recent study employing sulfadiazine-resistant plants (Tabatabaei et al., 2019). Conversely, the dual targeting of the yeast CoxIV transport signal had never been reported previously. The plastid targeting of the yeast mitochondrial transport signal might be a consequence of the fact that yeast does not contain plastids and thus the transport signal of yeast mitochondria has not evolved to be able to distinguish between the two endosymbiotic organelles (Huang et al., 1990; Staiger et al., 2009). However, such dual targeting was most unexpected for the transport signal of ATPS because neither in vitro nor in vivo approaches had previously given any hint of its plastid-targeting characteristic (Baudisch et al., 2014). Even transgenic plants expressing ATPS1-100/eYFP did not show any plastid localization (Fig. S6). However, with the use of the sasplit-GFP system the dual-targeting specificity of ATPS1-100 was clearly detectable and this result could be independently confirmed by a phenotype complementation approach using immutans mutants (Fig. 6A, Fig. 7F). These observations highlight the high sensitivity of the sasplit-GFP technology over classical fluorescent protein tagging approaches. Nonetheless, it cannot be completely excluded at this point that the observed high sensitivity of the sasplit-GFP system is due to the use of vector sets carrying highly efficient regulatory elements.
Modularity of PlaMiNGO toolkit
The PlaMiNGO vector toolkit is based on the principle of modular cloning (Weber et al., 2011). Hence, the components of the toolkit can easily be rearranged for in vivo imaging of proteins targeted to various other sub-cellular compartments. The vectors constructed in module I and II (Fig. S3) are binary vectors and can be utilized for plant cell transformation via Agrobacterium or via several other methods, e.g. protoplast transformation or particle bombardment. The gene of interest can be cloned upstream of one of the GFP11 tags of module I vectors with a single Golden Gate cloning reaction. Similarly, GFP1-10 can be targeted to the different sub-cellular or even sub-organellar compartments via cloning of the specific transport signal N-terminally in a Golden Gate ‘ready’ vector pTEI177. These vectors carry a ccdB negative selection cassette upstream of GFP1-10 or GFP11 tags for background-free cloning of the candidate gene. Consequently, the two vectors carrying GFP11 and GFP1-10 gene chimeras should be co-expressed in a single cell in order to determine protein targeting specificity to the organelle of interest.
Significance of high dual targeting frequency
The results obtained in this study strongly suggest that the number of dual-targeted proteins is much higher than previously assumed. More than just increasing the list of such proteins, these findings also highlight the evolutionarily conserved nature of organelle translocation machineries in plant cells. Even if some of these proteins are mistargeted to the wrong organelle, for example due to protein overexpression, the technique reveals the fundamental targeting properties of the respective transport signal. The reasons for such widespread dual targeting of nuclear-encoded proteins cannot be clearly defined at this stage. On the one hand, it supports the hypothesis that dual targeting is an evolutionary remnant (Staiger et al., 2009). On the other hand, it highlights that the acquisition of dual-targeting properties might still be an ongoing process, which allows for the development of completely new biochemical pathways in an organelle (Martin, 2010; Xu et al., 2013).
MATERIALS AND METHODS
Generation of vectors for co-infiltration
The GFP1-10 and GFP11×7 fragments were amplified by PCR from plasmids pcDNA3.1-GFP (1-10) and pACUH-GFP11×7-mCherry-β-tubulin (deposited by the Bo Huang lab; Addgene #70218 and #70219) and cloned into pRT100mod-based vectors (Baudisch et al., 2014) either with digestion–ligation or with restriction-free cloning (Bond and Naus, 2012). Primers for gene amplification are summarized in Table S1. The gene sequence coding for the N-terminal 91 amino acids of TyrRS was amplified from a vector provided by Elzbieta Glaser (Stockholm University, Sweden) and cloned accordingly (Baudisch et al., 2014). The above constructs, comprising promoter and terminator regions (CaMV35S::Gene of interest:GFP1-10/GFP11×7::t35S), were later sub-cloned into a Golden Gate-compatible pLSU4GG binary vector using a BsaI restriction–ligation reaction (Erickson et al., 2017). Golden Gate cloning was performed in a 15 µl reaction with the following conditions: 2.5 units of T4 DNA ligase (Thermo Fisher Scientific), 5 units of BsaI (Thermo Fisher Scientific), 1×BSA, 20 cycles of incubation at 37°C for 2 min and 16°C for 5 min, final deactivation and denaturation at 50°C for 10 min and 80°C for 10 min, respectively. The construction of respective eYFP chimeras is described in Sharma et al. (2018a).
Construction of the PlaMiNGo toolkit
The modular cloning principle and DNA fragments of the Plant Parts I and II toolkits were used for vector construction (Weber et al., 2011; Engler et al., 2014; Gantner et al., 2018). The modules utilized for cloning of Golden Gate-based vectors are illustrated in Fig. S3. Golden Gate reactions were performed with 20 fmol of each DNA module with the following conditions: 2.5 units of T4 DNA ligase (Thermo Fisher Scientific), 5 units of BsaI or BpiI (New England Biolabs), 30 cycles of incubation at 37°C for 2 min and 16°C for 5 min, final denaturation at 80°C for 10 min. When required, the restriction–ligation reactions were subsequently supplemented with fresh ligation buffer and ligase for terminal ligation, and incubated for ≥3 h at 16°C. Ligation mixtures were transformed into Dh10b or ccdB survival II cells (Thermo Fisher Scientific) and grown on plates with appropriate selective medium. PCR primers for module 0 cloning and gene sequences are summarized in Tables S2 and S3. The YFP1-10 construct was generated by introducing a point mutation in GFP1-10. Two fragments were amplified via PCR from GFP1-10 template using primer set 1 (GFPf- tttgaagacataATGTCCAAAGGAGAAGAAC; YFPr- tttgaagacatGataTGAGAGGTAGTGATTATCAG) and primer set 2 (YFPf- tttgaagacattatCAAACAGTCCTGAGCAAAG; GFPr-tttgaagactaaagCTAACTTCCGCCGCCACCTG), and later ligated via BpiI Golden Gate reaction to generate YFP1-10.
Cloning of candidate proteins into PlaMiNGo vectors
The candidate targeting signals (TyrRS1-91, FNR1-55, mtRi1-100, ATPS1-100, GCS1-100, GAPDH1-100, Gtred1-100, RbcS1-79 and CoxIV1-29) were amplified from the corresponding cDNA templates (Nelson et al., 2007; Berglund et al., 2009; Baudisch et al., 2014) and cloned via standard Golden Gate reaction (see above) into PLaMiNGo vectors in exchange for a ccdB negative selection cassette, upstream of GFP11 tags. Overhangs of the fragment to be cloned were AATG at the 5′ end and TTCG at the 3′ end. Two additional nucleotides were inserted into some of the fusions to maintain the reading frame, resulting in an additional codon for an alanine residue.
Cloning for complementation of the immutans phenotype
The gene sequence coding for mature PTOX57-295 (mPTOX) was amplified from a cDNA clone provided by Steven Rodermel (Iowa State University, USA) and sub-cloned via restriction-free cloning into pRT100mod-based vectors downstream of the gene sequences coding for the transport signals of either GCS1-100 or ATPS1-100. Additionally, the PTOX full-length gene and coding sequence for mPTOX protein (lacking the transit peptide) were cloned into empty pRT100mod vectors, downstream of the CaMV35 promoter, via restriction digestion and ligation. The candidate gene constructs with promoter and terminator (CaMV35S::Candidate:mPTOX::t35S) were then cloned into a Golden Gate-compatible pLSU4GG binary vector with BsaI restriction–ligation reaction.
Agrobacterium infiltration of Nicotiana benthamiana
The constructs were transformed into electro-competent cells of Agrobacterium tumefaciens GV3101 (pMP90) (Koncz and Schell, 1986). Agrobacterium infiltrations of 6–8-week-old fully expanded leaves of Nicotiana benthamiana plants, grown in soil under long-day conditions (16 h:8 h light:dark cycle) in a greenhouse, at a day temperature of 21°C, night temperature of 19°C and 50–60% relative humidity, were performed as described by Sharma et al. (2018a). For co-infiltration, each bacterial strain was adjusted to OD600=0.8 and mixed in a 1:1 ratio prior to infiltration.
Microscopy and imaging
Confocal laser scanning microscopy was carried out as described by Sharma et al. (2018a). For GFP–YFP dual-channel imaging, the 493–518 nm (GFP) and 519–620 nm (YFP) filter ranges were used to collect the emitted fluorescence signals. When required, brightness and contrast of the images were equally adjusted for each image to avoid any discrepancy in visualization of signal intensities.
For quantification of signal intensities, infiltration of all relevant constructs was carried out on different spots of the same leaf. At least three images from each infiltration spot were used for quantification. Quantification of the signals was performed with raw images using the Fiji program (Schindelin et al., 2012). For the purpose of quantification, image acquisition was done with the 20× objective in 7–8 z-stacks covering the epidermal cell layer and later stacked to project the maximum intensities. The mean gray values of stacked images were calculated using the ‘Measure’ option of Fiji and further utilized for comparison of the fluorescence signal strength in arbitrary units (A.U.).
Growth and transformation of immutans plants
For the purpose of stable transformation, the Arabidopsis thaliana immutans seeds (provided by Steven Rodermel, Iowa State University, USA) were germinated at 5 μmol m−2 s−1 light in a 8 h:16 h light:dark cycle. After three weeks of germination, the seedlings were transferred to 50 μmol m−2 s−1 light. For induction of flowering, the plantlets were transferred to >150 μmol m−2 s−1 light and incubated in a 16 h:8 h light:dark cycle. Floral-dip transformation was performed as described by Davis et al. (2009). At least three independent T2 transgenic lines were selected in each case for the analysis of phenotype. After germination at >150 μmol m−2 s−1 light in a 8 h:16 h light:dark cycle, these plants were transferred to a 16 h:8 h light:dark cycle. Pictures presented in Fig. 7 are from 5-week-old plants grown in >150-μmol m−2 s−1 light.
We would like to thank Theresa Ilse for assistance in cloning, Elzbieta Glaser (Stockholm University, Sweden), Martin Schattat (Martin Luther University Halle-Wittenberg, Germany) and Bo Huang (University of California, San Francisco, USA; via Addgene) for kindly providing constructs and Steven Rodermel (Iowa State University, USA) for providing immutans mutants and the PTOX clone.
Conceptualization: M.S., R.B.K.; Methodology: M.S., C.K., C.L., J.S.; Validation: C.K., C.L.; Formal analysis: M.S.; Investigation: M.S.; Resources: R.B.K.; Data curation: M.S.; Writing - original draft: M.S.; Writing - review & editing: M.S., J.S., R.B.K.; Visualization: M.S., J.S., R.B.K.; Supervision: M.S., R.B.K.; Project administration: M.S., R.B.K.; Funding acquisition: J.S., R.B.K.
This work is supported by Martin Luther University Halle-Wittenberg, Halle (Saale), Germany and M.S. was supported by a fellowship funded by the Erasmus Mundus Action 2 program of the European Union (BRAVE project).
The vectors of PlaMiNGo toolkit have been deposited to Addgene, https://www.addgene.org/browse/article/28203315/.
The authors declare no competing or financial interests.