Plant biology is currently experiencing a growing demand for easy and reliable mRNA and protein localisation techniques. Here, we present novel whole mount in situ hybridisation and immunolocalisation protocols, suitable to localise mRNAs and proteins in Arabidopsis seedlings. We demonstrate that these methods can be used in different organs of Arabidopsis seedlings as well as in other plant species. In order to achieve better reproducibility and higher throughput, we modified these protocols for automation to be performed by a liquid handling robot. In addition, we show that other procedures such as reporter enzyme assays and tissue clearing can be similarly automated. We present examples of application of our protocols including mRNA localisation and proteins and epitope tag (co)localisations which demonstrate that these methods provide reliable and versatile tools for expression, localisation and anatomical studies in plants.
With several plant genomes sequenced or close to completion (The Arabidopsis Genome Initiative, 2000; Yu et al., 2002) and with diverse ‘high-throughput’ technologies applicable for various problems in molecular genetics and molecular biology, work in plant biology has shifted towards the precise elucidation of the roles of individual genes and gene families in plant growth and development. The important and time-consuming step in these studies is the correlation of the gene expression and protein localisation patterns with the corresponding mutant phenotype. Also, with a rapidly growing list of characterised plant proteins, more and more of them can be used as markers of cell identity or polarity, and eventually as markers for intracellular organelle identity. Thus, antibodies and gene-specific probes become important tools for plant developmental or cell biology. All these developments cause a growing demand for rapid and reproducible techniques allowing specific transcript and protein localisation in planta.
The most popular localisation methods are in situ hybridisation for mRNA and immunocytochemistry for protein. These detection techniques were originally developed for application to sections from animal tissue (Angerer and Angerer, 1981; Deinhardt and Dedmon, 1965); however, they have also been successfully adapted for plants (Meyerowitz, 1987). These protocols have been demonstrated to provide results with high sensitivity and resolution. However, they are laborious and time consuming because they include the embedding of the tissue in various matrices and the sectioning of the embedded material. Therefore, the processing of multiple samples is difficult. In addition, the embedding procedure often changes the antigenic properties of proteins, thus rendering useless some antibodies that are fully functional in other experiments. To circumvent these problems, whole mount methods have been developed, which allow mRNA or protein localisation in the whole organism. Such methods are much faster than the use of tissue sections because they omit embedding and sectioning steps. Additionally, they preserve the tissue as well as antigenic properties of studied proteins better.
Originally, the whole mount in situ method was introduced and optimised for localisation of transcripts in Drosophila embryos (Tautz and Pfeifle, 1989). In animal systems, the whole mount methods work effectively. However, their adaptation to plants is hindered by several problems resulting from poor permeability of reagents through cell walls. Based on the described procedures for whole mount in situ hybridisation (de Almeida Engler et al., 1998; Gajewski et al., 1996; Tautz and Pfeifle, 1989) and whole mount immunodetection (Goodbody and Lloyd, 1994; Webb and Gunning, 1990), we have developed modified protocols enabling rapid and reproducible in situ localisation of mRNA and proteins in plant tissues. Moreover, to achieve higher reproducibility and throughput, we have adapted these methods to automation with a programmable, liquid handling robot. This robot has previously been used for automated whole mount in situ detection in animal species such as Drosophila, hydra and mouse (Plickert et al., 1997). It automates all the incubation and washing steps of the procedures such as in situ hybridisation or immunostaining. The specimens are processed in flow-through columns which are arranged in a thermostated rack. Within a liquid exchange circle, the buffer is expelled by gentle air pressure delivered through the outer needle of the dual probe. The expelled liquid is replaced immediately by fresh solvent delivered via the central channel of the probe. The robot is able to process up to 96 samples with 96 different probes in parallel.
Here we show the use of the automated protocols for localisation of various mRNAs in Arabidopsis seedlings and several proteins in Arabidopsis, parsley and tobacco roots. We also demonstrate the use of these methods in several ‘display of differential pattern’ studies, including antibody-based cell type identification in developmental mutants and comparison of certain treated and untreated groups. Moreover, we show examples of similar serial incubation procedures, such as the substrate reactions in enzyme-based reporter gene assays or tissue clearing procedures, which have also been automated.
Results and discussion
Whole mount in situ hybridisation procedure
The previously described whole mount in situ hybridisation techniques for plants (de Almeida Engler et al., 1998; Bennett et al., 1996; Ludevid et al., 1992) did not yield entirely satisfactory results for us, especially for less abundant mRNAs and mRNAs with spatially less restricted localisation patterns. In order to improve the signal/background ratio, which was a limiting factor for these methods, we developed a new procedure based on these protocols as well as on protocols used for the mRNA localisation in Drosophila embryos (Tautz and Pfeifle, 1989) and hydras (Gajewski et al., 1996). The typical in situ hybridisation experiment consists of several steps: (i) tissue fixation and permeation, (ii) hybridisation and (iii) signal detection. Most in situ hybridisation protocols start with a fixation step using cross-linking by aldehyde. This is followed by proteinase K treatment for degrading the protein bound to the RNA. This standard step can be fully replaced by a brief heat denaturation as has been described for animal species (Plickert et al., 1997). However, because of the specific properties of plant material, namely the presence of cell walls and a waxy layer at the surface of most of the plant body, additional permeation steps have to be included. These include heptane treatment during material fixation, alcohol treatment, which also removes chlorophyll and other pigments, and finally xylene treatment. In case of persisting penetration problems, the duration of these treatments was simply prolonged as required by the particular plant material.
For hybridisation, different types of probes such as radioactive or digoxigenin-labelled antisense RNAs can be used. We used the latter, and obtained the best results with short probes (150–220 base pairs) produced by alkaline hydrolysis of longer transcripts. For most of the probes, the optimal hybridisation temperature was 50°C, which ensured sufficient hybridisation specificity. However, with less abundant transcripts or less specific probes, the temperature was decreased to 45°C or increased to 55°C, respectively. Detection of the bound probe was performed by using commercially available anti-DIG antibody linked to alkaline phosphatase.
The crucial question of how to ensure that observed signals specifically reflect the localisation pattern of RNA of interest in every in situ hybridisation protocol remains. This is particularly important for whole mount techniques because, with most of the probes, after a sufficiently long staining time, signals can almost always be obtained, especially in the meristem regions. This is also often true for typical sense probe controls. Therefore, the best way to validate results obtained by in situ detection is with a parallel experiment using the corresponding knock-out mutant. If this is not available, simultaneous use of a probe with a known staining pattern as a positive control and a non-plant RNA probe as a negative control is helpful in ensuring that the observed experimental staining is not artifactual. In that case, an additional independent confirmation of staining pattern, e.g. by promotor::GUS analysis, is highly recommended. We used both manual and automated protocols for the detection of various mRNA transcripts in Arabidopsis seedlings with fully satisfactory results. For example, the localisation of root-specific putative auxin efflux regulator AtPIN2 mRNA showed the strongest signals in cortex cells, accompanied by epidermal and lateral root cap staining (Figure 1b). This localisation pattern correlated well with the pattern of AtPIN2 promotor activity monitored in AtPIN2::GUS transgenic plants (Figure 1a) and was similar to previously reported results (Chen et al., 1998; Müller et al., 1998). However, we achieved better signal resolution with our protocol. Additionally, we examined the specific expression pattern of another auxin-related gene, PINOID (PID), encoding the Ser/Thr kinase (Christensen et al., 2000) in root tissue. The PID mRNA was found in the differentiating vasculature of the root (Figure 2a) in accordance with the previous reports (Benjamins et al., 2001). The specificity of this signal was further indicated by the lack of staining in a parallel experiment with pid knock-out mutant (Figure 2b). The method presented here can be applied not only to the root tips, but also to all other organs of the Arabidopsis seedlings. To demonstrate this, we investigated the expression of another auxin efflux regulator, AtPIN1, in aerial seedling tissue. In addition to confirming the reported expression in root (Friml et al., 2002a) and stem (Gälweiler et al., 1998), our results show that AtPIN1 is also expressed in the cotyledon vasculature (Figure 2c). As another example, we analysed the aerial tissue expression of the small ribosomal protein RPS5A (Weijers et al., 2001). We detected RPS5A mRNA localisation in shoot apical meristem and developing leaf primordia (Figure 2d). Both aerial expression patterns corresponded with promotor activity studies (data not shown; Weijers et al., 2001). The protocol described here has also been successfully used by several other laboratories to localise various transcripts in Arabidopsis (Benjamins et al., 2001; Friml et al., 2002a,b; Weijers et al., 2001; Bennett, Hejátko, Scheres, personal communication), thus demonstrating reliability of this method.
Whole mount immunolocalisation procedure
The whole mount immunocytochemistry protocol, mainly intended for in situ localisation of proteins in Arabidopsis seedlings, was developed based on the protocol for cytoskeletal immunofluorescence visualisation (Goodbody and Lloyd, 1994; Webb and Gunning, 1990). The typical immunolocalisation procedure consists of several steps: (i) tissue fixation, (ii) permeation, (iii) blocking and primary antibody incubation, (iv) secondary antibody incubation and (v) signal detection. The fixation step is generally critical and requires individual optimisation. However, we found that with our whole mount procedure, simple fixation by 4% formaldehyde was sufficient for most of the proteins. After fixative removal, permeation steps followed. First, the cell walls were partially digested by driselase; then membranes were permeabilised by dimethylsulfoxide and detergent treatment. The driselase step seemed to be the most critical because insufficient digestion resulted in drastic reduction of the signal intensity, while overdigestion strongly affected the tissue integrity and could result in the loss of some signals (e.g. in case of microtubule detection). Moreover, driselase showed significant batch-to-batch variation with respect to its lytic activity. Therefore, adjustment of incubation time was recommended for each batch. Another major factor influencing the results of immunolocalisation experiments was the quality of the primary antibody. We obtained the best results in using polyclonal or monoclonal antibodies raised against heterologously expressed proteins. Surprisingly, the use of shorter peptides for the immunisation often yielded antibodies which were not suitable for immunolocalisation protocols. Moreover, we did not find any correlation between suitability of the antibody for Western blot analysis and for immunolocalisation experiments (data not shown). Most of the sera were used directly, but for some antibodies, subsequent affinity purification of the sera was unavoidable because the crude serum gave a high non-specific background. The suitable working concentration had to be adjusted (the effective concentration usually varied between 1 : 100 and 1 : 2000 for crude serum). To verify the data obtained by whole mount immunolocalisation, the only convenient control is the use of the knock-out mutant versus wild type in parallel immunolocalisation experiments. If the mutant is not available, the application of an independent method such as promotor::GUS fusion or in situ hybridisation to verify at least the localisation pattern at the tissue or cellular level is useful. In our experience, antigen competition or pre-immune serum experiments as negative controls are insufficient if not combined with the above-mentioned controls.
Comparing detection methods, we obtained the best results with fluorescence-labelled secondary antibodies (FITC, CY3, DATF), which allow the use of confocal scanning microscopy. The entire protocol was designed, similar to that used for in situ hybridisation, to be performed either manually in eppendorf tubes or automated using the liquid handling robot. Our data showed that the method was particularly suitable for localisation of proteins in the root tips and root hairs. To achieve signals at a reasonable intensity for other tissues, such as developing lateral roots, hypocotyls or cotyledons, the whole mount procedure should be carried out on slides with eventual gentle sample squashing and an additional freezing step in liquid nitrogen. Immunolocalisation, according to our protocol, was performed manually and automated to localise a variety of plant proteins. Using anti-AtPIN2 antibodies (Müller et al., 1998), we localised the AtPIN2 protein in outer cell layers of the root (Figure 1c). This pattern corresponds to AtPIN2 promotor activity (Figure 1a) and AtPIN2 mRNA localisation (Figure 1b). In addition to previous reports demonstrating a presence of AtPIN2 in the epidermis and cortex cells (Müller et al., 1998), we could clearly detect the AtPIN2 also at upper ends of the lateral root cap cells (Figure 1c, inset). Besides AtPIN2, other AtPIN proteins, such as AtPIN3 (Friml et al., 2002b), AtPIN1 and AtPIN4 (Friml et al., 2002a), have been successfully localised in Arabidopsis roots. All the aforementioned proteins are located in the plasma membrane. To validate this method also for other subcellular localisations, we performed detection experiments for KNOLLE syntaxin (Lauber et al., 1997; Lukowitz et al., 1996), known to be localised in cell plate and Golgi apparatus (Figure 3a) or tubulin, visualising microtubule cytoskeleton (Figure 3b). As mentioned before, the method can be routinely used for root tips as well as for root hairs. This was demonstrated by visualisation of actin cytoskeleton (Figure 3c) in root hairs, and Rop GTPases (Molendijk et al., 2001) at their growing tip (Figure 3d). The entire methodology has been developed primarily for Arabidopsis; however, its application is transferable to other plant species as well, as shown by visualisation of tubulin cytoskeleton in tobacco (Figure 3e) and actin cytoskeleton in parsley (Figure 3f).
The main advantages of our automated protocols compared to the manual methods are the possibility of investigating a large number of samples in parallel and the superior reproducibility of the experiments. The combination of these advantages allows experiments for display of differential expression and localisation patterns. These include, for example, screening libraries of pharmacological compounds for their effects on the localisation of subcellular markers or the characterisation of developmental mutants using antibody-based cell type identification. The data showing an effect of actin depolymerisation induced by Latrunculin B on localisation of the AtPIN1 protein provide an example of such an application. The original polar plasma membrane localisation of AtPIN1 was disrupted by the addition of Latrunculin B, and hence internalisation of the protein occurred (Figure 3g). This effect was similar to the reported disruption of AtPIN1 localisation by another actin depolymerising drug, Cytochalasin D (Geldner et al. 2001). Another example demonstrates the use of our method for analysing mutants in developmental biology. The splitting of the localisation domain of the columella cell type marker AtPIN3 (Friml et al., 2002b) (Figure 3h) demonstrated the presence of multiple root meristems in fackel mutant (Schrick et al., 2000). These examples and a number of other previously reported protein localisation patterns (Heese et al., 2001; Souter et al., 2002; Swarup et al., 2001) demonstrate the versatility and applicability of our immunolocalisation methods.
Using tags for localisation of proteins
Despite the demonstrated reliability of our automated immunolocalisation technique, the most critical aspect of a successful protein localisation method is the availability of suitable antibodies. Some of the most common problems are that an antibody does not recognise the desired protein or it recognises additional antigens and therefore yields unspecific staining. The most appropriate way to circumvent this problem is through the generation of transgenic plants expressing the protein of interest fused to a marker peptide. This marker should not interfere with the function of the protein it is fused to and its detection should be as easy as possible. A fusion with green fluorescent protein (GFP), which can be easily detected by its own fluorescence, is one of the most favoured options. We found that the GFP retained its fluorescent properties during our immunolocalisation protocol. This enabled visualisation of both immunolocalised proteins and GFP-tagged proteins in parallel. Figure 4(a) shows an example of GFP fluorescence in seedlings of endodermis/cortex marker line J0571 (http://www.plantsci.cam.ac.uk/Haseloff/IndexCatalogue.html), which were subjected to our immunolocalisation procedure. In addition, the use of commercially available anti-GFP antibodies enables the immunodetection of GFP (Figure 4b,c). This is especially useful if GFP cannot be detected by its own fluorescence, which may result if there is a low level of expression or if the fusion protein is targeted to the apoplast. An alternative to GFP fusion proteins is the use of uidA (GUS) enzyme for this purpose. GUS can be detected easily by a chromogenic reaction resulting in blue precipitates in tissues and cells where the fusion protein has been expressed. In addition, anti-GUS antibodies are commercially available and can be used for more precise localisation of fusion protein within the cell (data not shown). However, both GFP and GUS fusion proteins share the disadvantage of their size, which may alter the subcellular localisation and/or function of the protein of interest. Using smaller epitope tags for which antibodies are commercially available (for overview see Fritze and Anderson, 2000) may help to solve this problem. The most widely used small tags are the synthetic peptides such as DYKDDDDK (FLAG), part of the haemagglutinin influenza protein (HA) and part of the human c-Myc protein (myc). Despite the much smaller size of these tags (8–10 amino acids), however, a test of the functionality of the fusion proteins, e.g. by complementing the corresponding mutant phenotype, is essential. An additional test by Western blot should be performed especially with GUS and GFP to make sure that the fusion protein is stable. The additional presence of the truncated protein with tag within the tissue of interest can cause false localisation results despite the functionality of the complete fusion protein. We tested several HA- and myc-tagged proteins in Arabidopsis and found that they worked equally well. For example, the immunolocalisation of C-terminal HA fusion to the putative auxin influx carrier AUX1 showed signals in columella, lateral root cap and stele cells of Arabidopsis root, which agrees with the results shown in previous reports for the N-terminal HA fusion to the AUX1 protein (Swarup et al., 2001). Another advantage of tag fusion proteins is that the anti-tag antibodies raised in different organisms are available and thus can be easily combined with polyclonal antibodies for multiple immunolocalisation studies (see double staining of AUX1-HA and AtPIN1 proteins in Figure 4e) and with many standard histological staining procedures (e.g. DAPI or YOYO-pro for nucleic acids). An example of such a staining is depicted in Figure 4(f), showing co-localisation of myc-tagged Arabidopsis t-SNARE AtSNAP33 (Heese et al. 2001) (in red) and the KNOLLE syntaxin (in green) in the cell plate and presumably GA and additional staining of nuclei (in blue) in the dividing root cell. Thus, the combination of different methods for protein immunolocalisation and histological staining gives us the desired versatility of the tools to localise various molecules and their spatial relationships within the plant cells.
Use of the liquid handling robot
In addition to other restrictions, manual in situ detection is mainly limited by throughput and reproducibility. This can be overcome by the use of an automated system that performs all the steps of the tedious manual protocol automatically. Most of the experimental data presented have been obtained by using a liquid handling instrument (Plickert et al., 1997). Whole mount in situ hybridisation and immunolocalisation techniques are the most common methods which can be performed with the robot. However, all the procedures with plant material encompassing frequent exchange of solvents and treatments at different temperatures can be automated on the instrument as well. We used it for special clearing protocols, which make plant tissues transparent and cellular structures more visible in Nomarski optics (Malamy and Benfey, 1997), e.g. for observing different stages of Arabidopsis embryogenesis (see two-cell stage embryo in Figure 5a) or lateral root development (see stage IV in Figure 5b), or for clearing roots prior to Lugol staining of starch in root columella (see Arabidopsis wild-type root meristem in Figure 5c). The automation of these methods greatly facilitates the larger scale experiments such as screening for mutants. The same clearing method gives very good tissue and signal preservation if performed after the regular GUS staining assay (see staining in AtPIN2::GUS transgenic plants in Figure 1a). In this case it is possible to perform the whole GUS staining assay with subsequent clearing in the robot. It enables screening of larger amounts of samples, e.g. a library of enhancers or genetrap lines for desired expression patterns. Thus, we were able to show that the robot can be successfully adapted to automate a variety of manual protocols to study mRNA and protein localisation, gene expression and tissue anatomy or morphology in plants.
Digoxigenin-labelled probes were prepared by in vitro transcription according to the manufacturer's instruction (Roche). Three-day-old seedlings were fixed in a 1 : 1 mixture of fixation solution (5% formaldehyde, 15% DMSO, 0.08 m EGTA, 0.1% Tween in PBS pH 7.4) and heptane for 30 min. Incubations in MeOH (2 × 5 min), EtOH (3 × 5 min), EtOH/xylene mixture (1 : 1, 30 min); washings with EtOH (2 × 5 min) and rehydration in EtOH/PBS series (75, 50 and 25%, 10 min each) followed. Post-fixation in fixation solution (20 min) and washing by PBS (2 × 10 min) were performed. Protein was removed from RNA by proteinase K digestion (20 µg ml−1 in PBS) for 15 min at 37°C and the reaction was stopped with glycine (2 mg ml−1) in PBS for 5 min. Seedlings were washed in PBS/Tween (2 × 10 min). The samples were pre-hybridised in a hybridisation mix (50% formamide, 5× SSC, 1 mg ml−1 denatured salmon sperm DNA, 0.1 mg ml−1 heparin, 0.1% Tween 20) for 1 h at 50°C and hybridised in a hybridisation mix with 20–100 ng ml−1 probe (16 h, 50°C). Three washing steps were performed at 50°C: 50% formamide, 2× SSC, 0.1% Tween for 10, 60 and 20 min; 2× SSC, 0.1% Tween for 20 min and 0.2× SSC, 0.1% Tween for 20 min. Another washing at room temperature with PBS/0.1% Tween (3 × 10 min) followed. The samples were further pre-incubated in 1% BSA in PBS/0.1% Tween (90 min) and incubated with an antibody for 4 h (anti-DIG ALP conjugated antibody 1 : 2000 in 1% BSA in PBS/0.1% Tween). Finally, the samples were washed in PBS/0.1% Tween (8 × 20 min) and incubated at 37°C in ALP buffer (0.1 m Tris pH 9.5, 0.1 m NaCl, 50 mm MgCl2, 0.1% Tween) supplemented with 2 mm Levamisole (Sigma Chemical Co.) for 10 min. A substrate reaction was carried out according to the manufacturer's instruction (Roche) in ALP buffer with Levamisole. The programme script for this method using the InsituPro robot is available from Intavis AG (http://www.intavis.com).
Whole mount immunolocalisation
The 3–6-day-old Arabidopsis seedlings were fixed in 4% paraformaldehyde in MTSB (50 mm PIPES, 5 mm EGTA, 5 mm MgSO4 (pH 7) adjusted with KOH) for 1 h. Samples were washed with MTSB/0.1% Triton (5 × 10 min) and with deionised water (5 × 10 min). Cell walls were digested with 2% driselase in MTSB for 30–45 min, and samples were washed with MTSB/0.1% Triton (5 × 10 min). Incubation with 10% DMSO/3% NP-40 in MTSB for 1 h followed. After another washing in MTSB/0.1% Triton (5 × 10 min), seedlings were pre-incubated in 2% BSA/MTSB (1 h, 37°C) and incubated with the primary antibody in 3% BSA/MTSB (5 h, 37°C). After extensive washing with MTSB/0.1% Triton (8 × 10 min), the seedlings were incubated with a secondary antibody in 3% BSA/MTSB for another 3 h (37°C). Finally, the samples were washed with MTSB/0.1% Triton (5 × 10 min) and deionised water (5 × 10 min) and transferred into Slowfade Antifade mounting medium. The program script for this method using the InsituPro robot is available from Intavis AG (http://www.intavis.com).
Histological staining and microscopy
The enzymatic reaction for determining GUS activity was performed as described by Weijers et al. (2001), and samples were subsequently cleared (Malamy and Benfey, 1997). Starch granules in the Arabidopsis root cap were visualised by Lugol staining (Friml et al. 2002a) after clearing. Fox histological analysis of embryos, dissected ovules were fixed and cleared as described (Weijers et al., 2001). DNA was stained for 15 min with 1 µg ml−1 4,-6-diamidino-2-phenylindole (DAPI; Sigma Chemical Co.) before the last washing step of the immunolocalisation protocol. Both embryos and roots were viewed with Nomarski optics using a Leica DMRB microscope equipped with a video camera (Hitachi, HV-C20A). Fluorescent samples were inspected by a confocal laser scanning microscope (Leica) and the leica tcs-nt software. Images were processed using Adobe Photoshop 4.0 (Adobe Systems Inc.), and the final composition of figures was performed with Adobe Illustrator 9.0.
We extend our thanks to Rene Benjamins, Niko Geldner, Changhui Guan, Maren Heese, Arthur Molendijk, Andreas Müller, Mario Schelhaas, Ingeborg Schülz, Martin Souter and Dolf Weijers for providing material and technical help. We are very grateful to Günter Plickert and Martin Gajewski for the help with InsituPro robot. We acknowledge Leo Gälweiler, Jim Haseloff, Ranjan Swarup and Christian Luschnig for providing the material. We acknowledge support by the Volkswagen Stiftung.