Systematic analysis of protein subcellular localization and interaction using high-throughput transient transformation of Arabidopsis seedlings


*(fax +33 1 30833099; e-mail


The functional genomics approach requires systematic analysis of protein subcellular distribution and interaction networks, preferably by optimizing experimental simplicity and physiological significance. Here, we present an efficient in planta transient transformation system that allows single or multiple expression of constructs containing various fluorescent protein tags in Arabidopsis cotyledons. The optimized protocol is based on vacuum infiltration of agrobacteria directly into young Arabidopsis seedlings. We demonstrate that Arabidopsis epidermal cells show a subcellular distribution of reference markers similar to that in tobacco epidermal cells, and can be used for co-localization or bi-molecular fluorescent complementation studies. We then used this new system to investigate the subcellular distribution of enzymes involved in sphingolipid metabolism. In contrast to transformation systems using tobacco epidermal cells or cultured Arabidopsis cells, our system provides the opportunity to take advantage of the extensive collections of mutant and transgenic lines available in Arabidopsis. The fact that this assay uses conventional binary vectors and a conventional Agrobacterium strain, and is compatible with a large variety of fluorescent tags, makes it a versatile tool for construct screening and characterization before stable transformation. Transient expression in Arabidopsis seedlings is thus a fast and simple method that requires minimum handling and potentially allows medium- to high-throughput analyses of fusion proteins harboring fluorescent tags in a whole-plant cellular context.


Systems biology aims to integrate large-scale data from the transcriptome, the proteome and the metabolome to provide a comprehensive view of the functional structure of a cell or organism. Such approaches require identification and integration of protein networks that rely on subcellular protein partitioning and interaction. To address the question of protein localization and interaction on a large scale, it is essential to express tagged proteins in cells in an efficient, rapid, versatile and non-destructive way. It is also important that the experiment is performed under physiological conditions, in particular by preserving tissue organization and cellular integrity and minimizing transformation and culture stresses as much as possible.

Several methods for transient transformation have been described in plants, particularly Arabidopsis. For example, Yoo et al. (2007) described an efficient and versatile transient assay involving transformation of mesophyll protoplasts. However, protoplasts are individual cells that have lost their identity and positional information, and which very quickly undergo profound de-differentiation processes accompanied by extensive chromatin remodeling (Tessadori et al., 2007). Alternatively, transient transformation of cultured Arabidopsis cells with Agrobacterium allows systematic subcellular protein localization or analysis of transcription factor–promoter interactions (Berger et al., 2007; Koroleva et al., 2005). However, like protoplasts, cultured Arabidopsis cells are physiologically distinct from native plant cells. They have lost their identity and cell-to-cell connections, and have often been sub-cultured for many years in synthetic medium which may cause extensive genetic and phenotypic changes. Moreover, cultured cells have limited genetic origins, and the considerable genetic diversity existing in Arabidopsis cannot be exploited using this method.

To avoid these drawbacks, transient plant transformation has been achieved through particle bombardment or Agrobacterium infiltration. Particle bombardment allows direct transformation of plant cells that are resistant to Agrobacterium infection, but shows relatively poor efficiency and induces significant mechanical stress. The most versatile and efficient method of transient plant transformation still relies on Agrobacterium-based infiltration. Agrobacterium can transfer foreign DNA into plant cells with intact cell walls and without excessive mechanical stress. Several reports have described protocols for transient transformation of Arabidopsis and tobacco leaves, stems and flowers by Agrobacterium, which mainly rely on the expression of GUS-based constructs (Kapila et al., 1997; Lee and Yang, 2006; Wroblewski et al., 2005; Yang et al., 2000). Assays for GUS activity are destructive and do not allow visualization of protein subcellular localization and interactions. Conversely, fluorescent protein (FP)-based constructs have been widely used to monitor protein subcellular localization and protein–protein interaction via fluorescence resonance energy transfer or bimolecular fluorescent complementation (BiFC). Indeed, tobacco epidermal cells have been used to efficiently express FP-based constructs with minimal stress responses. This allows in-depth analysis of endomembrane dynamics, mainly of the ER and Golgi apparatus (Batoko et al., 2000; Boevink et al., 1998; Nebenfuhr et al., 1999; Runions et al., 2006; daSilva et al., 2004), but also monitoring of protein–protein interactions (Bracha et al., 2002; Bracha-Drori et al., 2004; Citovsky et al., 2006; Walter et al., 2004; Zamyatnin et al., 2006). However, tobacco epidermal cells represent an heterologous system for expression of Arabidopsis proteins, and, like cell culture, cannot be used to test expression in an extensive collection of mutants and transgenic marker lines. Therefore, an efficient transient expression system in Arabidopsis seedlings for FP-based constructs is essential for functional analysis of protein networks in various genetic backgrounds. Manual leaf infiltration of FP constructs developed for tobacco has been scarcely used in Arabidopsis (Lavy et al., 2002), as the method induces mechanical and biotic stresses often associated with extensive non-specific autofluorescence that can impair FP detection. Direct transient transformation of seedlings with FP constructs has been documented but was not thoroughly tested (Lagrange et al., 2003). Recently, a new procedure for epidermal root cell transformation by Agrobacterium rhizogenes has been described, allowing dynamic analysis of endosomal compartments in root hair cells (Campanoni et al., 2007). Here, we describe the optimization of an infiltration protocol of Arabidopsis seedlings with Agrobacterium tumefaciens that allows rapid and efficient expression of fluorescent proteins (GFP, YFP, mRFP1, mCherry) in cotyledons and leaves. It is possible to label various cell types and all the subcellular compartments. Co-expression of GFP- and mRFP1-tagged proteins in the same cell allows direct co-localization studies, as illustrated by systematic analyses of the subcellular localization of enzymes involved in the sphingolipid metabolic pathway. Finally, we show that transient co-expression of proteins in Arabidopsis can be used for monitoring of protein–protein interactions by BiFC.


Efficient transient expression of FP constructs in Arabidopsis seedlings

We developed a transient expression method based on batch infiltration of young Columbia (Col-0) seedlings by Agrobacterium that allows rapid and robust monitoring of protein localization and interaction in Arabidopsis seedlings. Briefly, seedlings were grown in six-well plates for 3–4 days on regular Arabidopsis medium. Seedlings were then submerged in an Agrobacterium tumefaciens culture and vacuum-infiltrated twice for 1 min. Agrobacterium solution was then discarded, and the seedlings were grown for a further 3 days before observation (Figure 1a). Vacuum infiltration tended to force young seedlings down on the medium, which was systematically associated with a low transformation rate and high autofluorescence. In contrast, a high number of transformed cells were observed on seedlings that were still erect after vacuum infiltration. A sterile polyvinyl grid (0.5 mm grid) was thus placed on the medium prior to sowing to provide a holding frame for the young seedlings and to prevent their collapse after infiltration (Figure S1). In the absence of a grid, we usually obtained few transformed seedlings (<20%), but its presence tripled the transformation yield (see below). Compared to vacuum- or syringe-based infiltration of rosette leaves, this method has the advantages of being fast (less than a week for plant growth), requiring minimum handling, being completed in vitro, and generating a high number of transformed cells and seedlings in a limited space.

Figure 1.

 Transient expression of fluorescent fusion proteins in Arabidopsis seedlings.
(a) Expression of a 35S:GFP construct in a 7-day-old seedling (top). Transformation mostly targets the cotyledon (bottom). Samples were visualized under the stereo microscope (left), and total (middle) or GFP-specific (right) fluorescence was recorded. Scale bar = 1 mm.
(b–e) Transient transformation of various cell types in Arabidopsis seedlings. GFP expression was observed in epidermal cells (b), stomatal companion cells (c), guard cells (d) and mesophyl cells (e). Scale bar = 10 μm.
(f–o) Localization of subcellular markers in epidermal cells. CD3-955–GFP labels the ER (f), PMA4–GFP labels the plasmamembrane (g), ERD2–GFP (h) and QUA2–GFP (i) label the Golgi, FBR–GFP labels the nucleolus (j), γ-TIP–GFP labels the tonoplast (k), SEC22–YFP labels the Golgi (l), CD3-989–YFP labels the mitochondria (m), citrate synthase–GFP labels the peroxisome (n) and CD3-997–YFP labels the plastids (o). Merged images are shown with GFP or YFP (green channel), propidium iodine (red channel) and chloroplast autofluorescence (blue channel). Scale bar = 10 μm.

Transient transformation was monitored by using the Agrobacterium tumefaciens C58C1 strain carrying the binary pGWB2 vector which allows the expression of a 35S:GFP construct (Koncz and Schell, 1986; Nakagawa et al., 2007). GFP fluorescence was recorded using either a stereo or confocal microscope (Figure 1a). Most of the GFP fluorescence was observed in the cotyledons, scattered in individual cells or within cell patches (Figure 1a and Figure S2). Occasional GFP fluorescence was also observed in young leaves and petioles (Figure S2). Various cell types were successfully transiently transformed. Epidermal and mesophyll cells were routinely transformed, but stomata guard and companion cells, as well as epidermal cells at the cotyledon edges, could also be transformed (Figure 1b–e). Roots showed a high level of autofluorescence, which prevented easy observation of GFP fluorescence, but were not transformed (Figure 1a). Failure to transform root cells was not due to the GFP reporter, as infiltration with 35S:GUS did not lead to any detectable GUS activity in roots (Figure S3). However, GUS expression in cotyledons showed a pattern similar to that observed for GFP constructs, demonstrating that the seedling infiltration method could be used successfully for various types of constructs (Figure 1a and Figure S3). Arabidopsis seedlings with GFP-expressing cells were used to estimate the efficiency of cotyledon cell transformation.

Parameters involved in the efficiency of transient transformation of Arabidopsis seedlings

First, we investigated the influence of the Agrobacterium infiltration medium on transformation efficiency, as it has been reported that its composition can affect the rate of Arabidopsis transformation (Wroblewski et al., 2005). The infiltration medium used for floral dip stable transformation, based on 5% sucrose and 0.01% Silweet, did not lead to any detectable transgene expression and produced a high seedling mortality rate. We tested other infiltration media containing 10 mm MgCl2 and 200 μm acetosyringone and buffered or not with 5 mm MES at pH 5.6. These media resulted in more than 50% of seedlings being transformed (68 and 58%, respectively) (Figure 2a). We then replaced the magnesium salts by sucrose, which resulted in a further increase in transformation efficiency (approximatively 75% GFP-positive seedlings) (Figure 2a). Vacuum infiltration (10 mmHg pressure) also appeared to be essential, as the seedling transformation rate was reduced by half when vacuum was not applied (Figure S4). The best results were observed with two successive vacuum infiltrations.

Figure 2.

 Transient expression of 35S:GFP fluorescent constructs in Arabidopsis seedlings.
(a) Effect of the infiltration medium on transient transformation. Transformation efficiency is expressed as the ratio of GFP-positive seedlings to the total number of seedlings (n = 80).
(b) Effect of Agrobacterium density in the culture and infiltration media on seedling transformation. Three concentrations of bacterial culture were used [OD600 = 0.5 (white bars), 1 (grey bars) or 5 (black bars)] and diluted to the indicated infiltration density. The transformation efficiency is expressed as in (a) (= 80).
(c) Effect of Agrobacterium density in the culture and infiltration media on epidermal cell transformation. Three concentrations of bacteria culture were used [OD600 = 0.5 (white bars), 1 (grey bars) or 5 (black bars)]. Transformation efficiency is expressed as the ratio of GFP-positive cells to the total number of cotyledon cells (= 1500).
(d) Effect of Arabidopsis accession on transient transformation. Transformation efficiency is expressed as the ratio of GFP-positive seedlings to the total number of seedlings (= 40).
Values are means, with standard errors indicated by bars.

The effect of Agrobacterium concentration on transient transformation of seedlings was determined by testing various bacterial densities in the culture and infiltration media. We did not test the efficiency of various Agrobacterium strains as it was previously reported that the laboratory strain C58C1 was the most competent for transforming Arabidopsis (Koncz and Schell, 1986). Bacteria were harvested at three growth stages, i.e. the exponential phase [attenuance at 600 nm (OD600) = 0.5], and the early (OD600 = 1) and late (OD600 = 5) stationary phases. For each culture phase, five bacteria concentrations were used for infiltration. Transient transformation efficiency was directly correlated with Agrobacterium concentration in both the culture and the infiltration medium (Figure 2b). The highest efficiency (92% of transformed seedlings) was obtained with a saturated culture (OD600 = 5) and the most concentrated infiltration solution (OD600 = 2) (Figure 2b). Bacteria from saturated cultures could still achieve a reasonably good transformation yield (>50%) even at low infiltration densities (≥0.01), while bacteria from low-density cultures required higher infiltration densities (>0.5) to achieve similar yields. We also determined the cellular transformation efficiency for each transformed plant by counting the number of GFP-expressing cells per cotyledon. As expected, the number of transformed cells per cotyledon was directly correlated with Agrobacterium density in the culture and infiltration medium (Figure 2c). However, the bacterial density of the culture seemed to have a greater influence on the efficiency of cell transformation, as use of fully saturated cultures doubled the number of transformed cells compared to early stationary phase cultures. Increasing the Agrobacterium density in the infiltration medium improved the transformation yield by only 20–35%, except for OD600 = 0.5 (Figure 2c). In conclusion, a saturated Agrobacterium culture associated with a high infiltration density led to >90% transformed seedlings, with, on average, 25% of epidermal cotyledon cells expressing the GFP marker.

Finally, we investigated the influence of the origin of Arabidopsis accession lines on the transient transformation efficiency (Figure 2d). We tested the widely used accessions Columbia 0 (Col-0), Landsberg erecta (Ler), Wassilewskija (WS), C24 or Cape Verde Islands (CVi), and also accessions used to generate recombinant inbred lines, such as Bay-0 or Shadarah (Sha). The accessions showed significant variations in transformation yield, with Ler and Cvi being the most (85%) and the least (30%) susceptible accessions, respectively. We observed differences among the Arabidopsis accessions not only with regard to the yield of seedling transformation but also the efficiency of transient expression at the cell level. For instance, Ler showed a much higher number of GFP-expressing cells per cotyledon than Col-0 (73 ± 23 and 8 ± 3%, respectively, with an infiltration OD600 of 1).

Transient co-expression of FP-based constructs in Arabidopsis epidermal cells

Having obtained a high transformation rate of seedlings, we then tried to co-express into cotyledon cells two constructs carried in two independent Agrobacterium strains. For this purpose, we co-transformed seedlings with 35S:GFP and 35S:mRFP1-AtYPC1 constructs that encode fluorescent proteins localized in the cytosol and nucleus (GFP) and in the ER (AtYPC1). A large number of cells were able to co-express both fluorescent proteins (Figure S5). We then monitored the effect of Agrobacterium culture and infiltration density on the number of plants showing GFP and mRFP1 co-expressing cells (Figure 3a). Similar to the single transformation experiments, an increase of Agrobacterium concentration in the infiltration medium up to an OD600 of 2 had a positive effect on co-transformation yield, with almost 60% of seedlings showing co-transformation events. Above an OD600 of 2, the high infiltration density reduced co-transformation efficiency, especially when bacteria were grown at low density (Figure 3a). We then evaluated the number of GFP and mRFP cells that co-expressed both markers (Figure 3b and Table S1). In contrast to seedlings, the yield of co-transformed cells was not greatly affected by Agrobacterium concentration in the culture or the infiltration medium. Around 17–19% of transformed cells expressed both the GFP and mRFP1 markers, whatever the concentration of Agrobacterium. However, there were two exceptions: first, the highest culture concentration (OD600 = 5) and infiltration concentration (OD600 = 4) significantly reduced the number of co-transformed cells (12%), and second, an OD600 of 1 and 2 for the culture and infiltration medium, respectively, led to the highest yield of co-transformed cells (26%). Finally, to further improve the co-transfomation rate, we also determined the influence of seedling age on the ability to co-express two markers. During initial tests on transient expression, we had noticed that seedlings growing for less than 3 days or more than 4 days showed much lower transformation efficiency than 3–4-day-old seedlings. Indeed, after 2 days of growth, only 15% of seedlings were co-transformed, compared to 65–70% for 3- and 4-day-old seedlings (Figure 3c). Similarly, 6- and 8-day-old seedlings showed reductions in co-transformation efficiency to 45 and 27%, respectively. The co-transformation yields for 3- and 4-day-old seedlings were then compared at various Agrobacterium concentrations in the infiltration medium (Figure 3c). Three-day-old seedlings showed a higher co-transformation rate compared to four-day-old seedlings, mainly at low Agrobacterium concentration (OD600 = 0.1 and 0.5). At higher concentration (OD600 = 2), transformation competency was similar between the two ages of seedlings.

Figure 3.

 Transient co-expression of GFP and mRFP1 fluorescent constructs in Arabidopsis seedlings.
(a) Effect of Agrobacterium density in the culture and infiltration media on seedling transient co-transformation. Equal quantities of bacteria for each construct were mixed to reach the final OD600 used for infiltration. For each construct, two densities of bacteria culture were used [OD600 = 1 (grey bars) or 5 (black bars)] for preparing the infiltration solution. Co-transformation efficiency is expressed as the ratio of seedlings with doubly transformed cells (GFP- and mRFP1-positive) in their cotyledons to the total number of seedlings (= 40).
(b) Effect of Agrobacterium density in the culture and infiltration media on transient co-expression. The Agrobacterium solutions used for co-transformation were prepared as in (a). Two concentrations of bacteria culture were used [OD600 = 1 (grey bars) or 5 (black bars)]. Co-expression efficiency is expressed as the ratio of doubly transformed cotyledon cells (GFP- and mRFP1-positive) to the total number of transformed cells (GFP- or mRFP1-positive) (= 300).
(c) Effect of seedling age on transient seedling co-transformation. Seedlings were grown for 3 days (grey bars) or 4 days (black bars) after imbibition. The Agrobacterium solutions used for co-transformation were prepared as in (a). Co-transformation efficiency is expressed as the ratio of seedlings with doubly transformed cells (GFP- and mRFP1-positive) in their cotyledons to the total number of seedlings (= 40).

Systematic subcellular localization in Arabidopsis epidermal cells and application to study of long-chain-base metabolic enzymes

We then assessed whether our protocol for transient expression in Arabidopsis epidermal cells could be used to systematically investigate protein subcellular distributions. We first chose to express at least one representative marker of the best characterized subcellular compartments (Table S2). Transient expression in cotyledon epidermal cells labeled various membrane compartments such as ER, plasma membrane, Golgi, nucleolus and tonoplast, and also organelles such as mitochondria, peroxisomes and plastids (Figure 1f–o). As shown previously, the cytosolic and nuclear compartments could also be labeled (Figure 1b–e).

We then used our transient expression system to investigate the subcellular distribution of proteins by systematic co-localization with known markers. To demonstrate the potential of our method, we focused our study on the enzymes involved in the metabolism of a specific class of lipids, the long-chain-base (LCB) precursors of sphingolipids. Lipids in general, and sphingolipids in particular, are molecules for which subcellular localization is difficult to determine. One way to partially address the problem is to localize the enzymes that synthesize and modify these molecules. The LCB sphinganine is synthesized from palmitoyl CoA and serine, and is used to produce the sphingolipid ceramide. Alternatively, LCBs can be phosphorylated by LCB kinases and probably act as LCB-1P lipid messengers, which are either dephosphorylated by a LCB phosphatase or degraded by an LCB-1P lyase. In yeast, most of this metabolic pathway is localized in the ER (Natter et al., 2005). Altogether, these enzymatic reactions represent six major enzymes that are potentially encoded by a total of ten genes in Arabidopsis (Dunn et al., 2004). To better characterize this metabolic pathway, the corresponding cDNAs were systematically cloned as N- and C-terminal fusions in GFP and mRFP1 Gateway expression vectors (Karimi et al., 2005) (Table S3). Of the 36 available FP constructs, ten were not functional in tobacco (mainly GFP–X and mRFP1–X), among which three were still not functional in Arabidopsis (Table 1). In seven cases, the FP constructs gave a fluorescence signal in Arabidopsis but not in tobacco, demonstrating that homologous expression systems are preferable to investigate protein expression. To address the subcellular distributions of the various enzymes, FP constructs were co-transformed with known subcellular markers into Arabidopsis seedlings and also into tobacco leaves for comparison (Figure 4a–i and Table 1). The lyase AtDPL1 was the only protein among our dataset that had been previously characterized as associated with the ER (Tsegaye et al., 2007). As expected, AtDPL1 was observed in the ER in Arabidopsis seedlings for all the FP configurations tested (Table 1 and Figure 4g). The subcellular distribution was often similar in both expression systems. However, we observed two cases (LOH3 and LOH1) for which ambiguous localizations occurred in the heterologous tobacco system. The position of the FP tag led to a clear difference in protein subcellular distribution. The most striking examples were observed with the putative LCB hydroxylases (AtSUR2A and AtSUR2B), ceramidase (AtYPC1) and ceramide synthases (LOH1, LOH2 and LOH3). For these six proteins, all the FP–X constructs showed ER localization, but the X–FP constructs accumulated in the Golgi apparatus (Figure S6). Comprehensive databases of information on the subcellular distribution of Arabidopsis proteins, such as SUBA (Heazlewood et al., 2007;, did not provide a clear conclusion (Table S4), stressing the need for experimental validation of subcellular localizations. The fact that all these enzymes are associated with the ER in yeast and mammals strongly suggests that the FP–X construct indicates the true subcellular localization (Lahiri and Futerman, 2007; Natter et al., 2005). The presence of an ER-retrieving such a KxxK or KxK in the C-terminus of three of these proteins (Table S4) suggests that C-terminal addition of an FP tag exerts a masking effect (Huh et al., 2003; Natter et al., 2005). The two putative LCB kinases AtSPHK1 and AtSPHK2 were found to be associated with the tonoplast (AtSPHK2) and the mitochondria (AtSPHK1) (Figure 4h,i). The mitochondrial localization of SPHK1 was confirmed by most of the predictions in the SUBA database (Table S4). Only one protein of the ten (the putative phosphatase AtLCB3) showed an ambiguous subcellular distribution that could not be resolved using the four construct combinations in the two expression systems. Indeed, the phosphatase AtLCB3 FP constructs did not give any signal in tobacco, and only one construct was functional in Arabidopsis but was associated with variable subcellular distribution (Table 1).

Table 1.   Subcellular distribution of LCB metabolic enzymes by transient expression in Nicotiana benthamiana and Arabidopsis thaliana epidermal cells
Accession no.NameFunctionN. benthamianaA. thaliana
  1. ER, endoplasmic reticulum; G, Golgi; M, mitochondria; PM, plasma membrane; N, nucleus; Ch, chloroplast; Cy, cytosol; T, tonoplast; ND, not detected; NA, not available. Proteins for which the localization is given as ‘ambiguous’ showed different localizations in different cells.

At1g13580LOH3Ceramide synthaseER, GER, GER, GER, GERERERER
At3g25540LOH1Ceramide synthaseGAmbiguous (G, N, Cy)GAmbiguous (G, N, Cy)GERGER
At3g19260LOH2Ceramide synthaseGNDGNDGERGER
At3g58490AtLCB3LCB phosphataseNDNDNDNDAmbiguous (T, ER, G)NDNDND
At4g21540AtSPHK2LCB kinaseTTTTTTTT
Figure 4.

 Subcellular distribution and interaction of FP proteins in Arabidospsis seedlings.
(a–i) Co-localization of enzymes involved in sphingolipid metabolism in epidermal cells. Sphingolipid enzymes fused with GFP were co-expressed with various subcellular mCherry makers in Arabidopsis epidermal cells.
(a–g) GFP (left), the ER CD3-959–mCherry marker (middle) and merged (right) signals are shown. (a) GFP–SUR2A; (b) GFP–SUR2B; (c) GFP–YPC1; (d) GFP–LOH3; (e) GFP–LOH1; (f) GFP–LOH2; (g) GFP–DPL1.
(h) SPHK1–GFP (left), the mitochondrial CD3-989–mCherry marker (middle) and merged image (right).
(i) SPHK2–GFP.
(j, k) Protein–protein interaction by BiFC of Deficiens (Def) and Globosa (Glob). Co-transformation of N-terYFP–Def and C-terYFP–Glob leads to YFP-positive Arabidopsis epidermal cells (k), but co-transformation of N-terYFP–AKINβγ and C-terYFP–Glob did not show any YFP signal (j).
Red indicates chloroplast autofluorescence. Scale bar = 10 μm.

Transient analysis of protein–protein interactions by BiFC

Finally, we investigated whether transient co-transformation of Arabidopsis seedlings could also be used for detection of protein–protein interactions by BiFC. We evaluated BiFC in Arabidopsis seedlings using the Deficiens (Def) and Globosa (Glob) proteins from Arabidopsis for which an interaction has previously been reported in the nucleus and cytosol (Davies et al., 1996; Zachgo et al., 1995). Co-transformation with the N-terYFP–Def and C-terYFP–Glob fusion proteins in Arabidopsis epidermal cells showed clear YFP fluorescence signals in the nucleus and cytosol, as expected (Figure 4k). To demonstrate the specificity of the Def–Glob BiFC interaction, we used the cytosolic protein AKINβγ as a negative control (Gissot et al., 2006). Co-transformation of N-terYFP–Def and C-terYFP–Glob with C-terYFP–AKINβγ and N-terYFP–AKINβγ fusion proteins, respectively, did not give any detectable YFP signal (Figure 4j and data not shown).


Transient transformation of Arabidopsis seedlings is an improved in planta expression system

We developed a method for transient expression in Arabidopsis seedlings that presents several advantages compared to previously published protocols. We first demonstrated that a high yield of transient expression of one or two FP constructs in epidermal cells can be achieved in young Arabidopsis seedlings. In contrast with expression systems in tobacco, the Arabidopsis assay is rapid, requiring <10 days from sowing to analysis, requires minimal handling, can be performed in various genetic backgrounds (mutants, reporter lines, etc.), and can be performed in vitro on a large scale. This assay can also be used for the analysis of GUS expression, thus allowing quantitative assays as required for experiments such as promoter analysis. We found that Agrobacterium density and Arabidopsis genetic background are important parameters for optimum transient expression. In particular, high Agrobacterium density in the culture phase seems to enhance bacterial virulence. Similar findings have been reported for A. rhizogenes-mediated transient transformation of root epidermal cells (Campanoni et al., 2007). Surprisingly, the Ler accession appears to be the most competent accession for transient transformation, but has been previously described as having a low efficiency of stable transformation by floral dip procedures (Clough and Bent, 1998). However, we cannot exclude the possibility that slight germination delays, for example in dormant Cvi, could result in the varied transient transformation efficiencies observed in the Arabidopsis accessions.

Altogether, our protocol showed 85% efficiency for single-cell transformation in Ler seedlings, matching the most effective transient transformation protocols in Arabidopsis (Koroleva et al., 2005). High-throughput transient expression of FP-based constructs has been described previously for Arabidopsis but involved cultured cells and not seedlings (Koroleva et al., 2005). However, cultured cells have specific constraints, such as their unknown genetic background and their de-differentiated state, that might interfere with some protein subcellular localizations or interactions. Cultured cells are also not easily produced from plant material and thus present a very limited choice in terms of genetic diversity compared with seedlings derived from Arabidopsis collections. Our method is also amenable to high throughput as the analysis could be performed on a minimal number of very young seedlings that could be efficiently grown and transformed in multi-well plates. The limiting step in the procedure is the analysis of fluorescence rather than the transformation step. The availability of sensitive fluorescence detection methods should further improve the throughput rate.

Transient expression in epidermal cells from cotyledons of FP-fused proteins allows determination of proper subcellular targeting and protein–protein interactions

Transient seedling transformation targets mostly cotyledons and only occasionally leaves. Our data demonstrate that epidermal cells from Arabidopsis cotyledons could be used to co-localize FP-tagged proteins, as previously established for tobacco leaf epidermal cells. The subcellular distribution of protein fusions is often investigated in tobacco epidermal cells (mostly Nicotiana benthamiana) as transient expression is efficient and the large epidermal cells facilitate imaging of cellular compartments. However, this represents an orthologous system for Arabidopsis proteins, and we showed that several FP constructs that were functional in Arabidopsis were not expressed in tobacco or did not show similar subcellular distribution. Moreover, tobacco requires a large greenhouse space to grow the plants, and manual leaf infiltration is too labor-intensive for a large number of experiments. The development of a homologous system for Arabidopsis will be an important tool to address protein function, as protein subcellular distributions and interactions may be regulated by endogenous scaffold or partner proteins (Smyczynski et al., 2006).

Transient protein co-localization of sphingolipid metabolic enzymes in Arabidopsis

As a proof of concept, we investigated the precise subcellular distribution of putative enzymes of the sphingolipid metabolic pathway. This pathway has been thoroughly characterized in yeast but is still poorly understood in plants (Dunn et al., 2004). Few enzymes in the pathway have been functionally characterized (Chen et al., 2006; Ryan et al., 2007; Tsegaye et al., 2007). Our analysis provides an extensive comparison of FP constructs with two different tags in both the N- and C-terminal positions and expressed in two biological systems. During the course of this study, we were able to compare more than 80 FP co-localizations, a number that was made possible by the efficiency of our Arabidopsis seedling transformation technique. Of the ten proteins analyzed in our transient assay in Arabidopsis seedlings, seven are clearly associated with the secretory pathway. The position of FP tag was found to change the subcellular distribution from the ER to the Golgi apparatus in several cases. Clear ER retention motifs were present in the C-terminus of several of these proteins, and were masked by the presence of the FP tag. The influence of the FP tag on protein subcellular distribution has been reported previously (Huh et al., 2003; Natter et al., 2005). The LCB kinase AtSPHK2 was not found in the ER but was clearly associated with the tonoplast. Interestingly, the yeast homolog LCB4p was localized at the plasma membrane in actively growing cells but was targeted to the vacuole during the stationary phase (Iwaki et al., 2005, 2007). The vacuolar distribution of AtSPHK2 was also associated with non-dividing, resting cells. Surprisingly, AtSPHK1 was not found to be associated with endomembrane compartments but with mitochondria. Our results stress the need to evaluate several combinations of FP constructs before drawing conclusions on protein localization. Secondly, testing is better performed in a homologous system. Finally, it is clearly necessary to confirm the localization observed in a transient expression system in stable transgenic lines, and also by other methods such as cell fractionation or immunolocalization.

In conclusion, our transient seedling transformation method provides a simple and rapid tool to study Arabidopsis protein localization and function in live cells and a whole-tissue context. It allows complete control of the genetic background as well as the culture conditions, and does not require a large greenhouse space. This protocol could be used for routine analysis of constructs before stable transformation, and also for large-scale, high-throughput co-localization of protein fusions and studies of protein–protein interactions in Arabidopsis in the context of functional genomics programs. Moreover, this method could potentially be used for other applications that rely on DNA transfer into plant cells by Agrobacterium, such as virus-induced gene silencing, microRNA interference or biochemical complementation.

Experimental procedures

Plant culture and infiltration

Sterile Arabidopsis thaliana seeds were sown on a sterile filter (Saatitech,, reference PA500/38, pore size 500 μm) placed on Arabidopsis agar medium (Duchefa, in six-well culture plates (TPP,, reference 92406) and stratified for 2 days at 4°C. Seedlings were grown for 4 days (18°C temperature, 60% humidity, 16/8 h day/night cycle) prior to infiltration.

Agrobacterium tumefaciens cells of strain C58C1 GV3101 (Koncz and Schell, 1986) with pMP90 helper plasmid and transformed with the various constructs were grown overnight in 5 ml pre-culture (2YT liquid medium: 16 g l−1 bacto-tryptone, 10 g l−1 yeast extract, 5 g l−1 NaCl, pH 7), and used to inoculate a 30 ml culture (2YT liquid medium).

After overnight growth at 28°C, A. tumefaciens cells were collected and resuspended at the appropriate OD600 in 2 ml of 5% sucrose, 200 μm acetosyringone. Three other infiltration media were tested in this study: (i) 10 mm MgCl2, (ii) 10 mm MgCl2 and 5 mm MES pH 5.6, and (iii) 5% sucrose and 0.01% Silweet ( In the case of co-infiltration, each culture was equally diluted in 4 ml to achieve the final infiltration concentration as measured by the OD600. Infiltration was performed by covering the seedlings with the Agrobacterium solution and by applying vacuum (10 mmHg) twice for 1 min. Excess infiltration medium was subsequently removed and the plates were transferred to a culture room for 3 days (same growth conditions as above). Nicotiana benthamiana leaves were agro-infiltrated as previously described (Gissot et al., 2006).

GFP-based constructs

Gateway cloning (Invitrogen, was performed by amplifying full-length cDNA with and without stop codons with respectively, the primer pairs Gene1-F and Gene1-R1 or Gene1-F and Gene1-R2 (containing half attB1 and attB2 sequences). A second round of PCR was performed using full-length attB1 and attB2 primers (94°C for 3 min, then 25 cycles of 94°C for 30 sec, 55°C for 30 sec and 72°C for 2 min), and PCR products were purified using a Nucleospin extract kit (Macherey-Nagel, PCR products were introduced into the pDONR207 vector (Invitrogen), via recombination with a BP reaction (Gateway, Invitrogen). An LR reaction was used to transfer cDNAs from the resulting entry vectors to the appropriate destination vectors pB7WGF2, pB7FWG2, pK7WGR2 or pK7RWG2 (Karimi et al., 2005) according to the manufacturer’s instructions (Invitrogen).

Cytology and imaging

GUS staining was carried out as described previously (Harrar et al., 2003). For confocal microscopy, 7-day-old seedlings were incubated for 30 min in 15 μm propidium iodide, and cotyledons were imaged using an inverted TCS-SP2-AOBS spectral confocal laser scanning microscope (Leica, GFP and YFP samples were excited with 488 or 514 nm argon laser lines, respectively, with an emission band of 495–540 nm for GFP detection, 520–560 nm for YFP detection, 605–660 nm for propidium iodide, and 675–765 nm for chlorophyll autofluorescence. Co-localization experiments using GFP and mRFP1 constructs were performed simultaneously using the same settings as above for GFP detection and the 594 nm HeNe laser and an emission band of 600–645 nm for mRFP1. The mRFP1 settings were used to localize mCherry constructs as described above. The number of transformed cells was determined using Optimas software (Imasys, Images of whole seedlings and cotyledons were taken using a Nikon SMZ1500 stereomicroscope with GFP band-pass and long-pass filters (Nikon, All quantitative data on the number of seedlings or cells expressing fluorescent constructs were obtained from at least two independent experiments. At least two Arabidopsis infiltrations were performed for each experiment.


J.M. and this work were supported by the 6th European Integrated Project AGRON-OMICS (grant number LSHG-CT-2006-037704), and L.B. is funded by Cancéropole (région Ile de France). We are grateful to François Parcy (Commissariat à l’Énergie Atomique Grenoble, France) for the gift of N-terYFP–Def and C-terYFP–Glob BiFC destination vectors, to Patrick Moreau (Centre National de la Recherche Scientifique, Bordeaux, France) for the gifts of PMA4–GFP, BETII–YFP and sec22–YFP, to Nemo Peeters for the gift of FBR–GFP, Citsynthase–GFP and γ-TIP–GFP, to Oumaya Bouchakbé-Coussa (Laboratoire de Bioloigie Cellulaire, INRA, Versailles, France) for the 35S:GFP construct, and to Sylvie Dinant for the 35S:GUS construct. We would like to thank Olivier Loudet (Station de Génétique et Amélioration des Plantes, Versailles, France) for providing the various accessions, and Olivier Grandjean (Laboratoire Commun de Cytologie, Versailles, France) for help with the Optimas software. We thank also Elizabeth Crowell (Laboratoire de Bioligie Cellulaire, INRA, Versailles, France) for critical reading of the manuscript. We are grateful to the Région Ile-de-France for its financial support.