Embryogenesis of flowering plants establishes a basic body plan with apical–basal, radial and bilateral patterns from the single-celled zygote. Arabidopsis embryogenesis exhibits a nearly invariant cell division pattern and therefore is an ideal system for studies of early plant development. However, plant embryos are difficult to access for experimental manipulation, as they develop deeply inside maternal tissues. Here we present a method for the culture of zygotic Arabidopsis embryos in vitro. The technique omits excision of the embryo by culturing the entire ovule, thus greatly facilitating the time and effort involved. It enables external manipulation of embryo development and culture from the earliest developmental stages up to maturity. Administration of various chemical treatments as well as the use of different molecular markers is demonstrated together with standard techniques for visualizing gene expression and protein localization in in vitro cultivated embryos. The presented set of techniques allows for so far unavailable molecular physiology approaches in the study of early plant development.
Flowering plants vary remarkably in their overall appearance. However, mature embryos look much the same, because only basic body parts are formed in the course of embryogenesis. A mature embryo, which develops from the single-celled zygote, is composed of a shoot and a root meristem placed at opposite poles of the apical–basal axis. The shoot meristem is located at the top, flanked by the cotyledons (embryonic leaves). It is connected with the root meristem at the bottom by the hypocotyl (embryonic stem) and the root itself. The apical–basal axis, radial and lateral symmetry, as well as patterns of different cell fates along these symmetries, are established during embryogenesis, laying the foundations for all post-embryonic development (Jürgens, 2001).
In the model plant Arabidopsis thaliana, the embryo develops through a highly uniform series of cell divisions and cell fate changes (reflected also by molecular markers), leading to morphologically distinct intermediate stages (e.g. Friml et al., 2003; Vroemen et al., 1996). The initial zygotic division is asymmetric, giving rise to a smaller apical and a larger basal cell. The basal cell and its lineage divide anticlinally and form the suspensor, an extraembryonic file of cells which connects the embryo to the maternal tissue. After the 32-cell stage, the uppermost suspensor cell becomes incorporated into the embryo as the hypophysis cell, from which most of the future root meristem is derived. All other embryonic parts are derived from the apical cell, which divides peri- and anticlinally to first form an eight-cell stage proembryo, then establishes the protoderm (at the 16-cell stage) and finally gives rise to the round globular embryo. At this stage, bilateral symmetry is established, with the positions of the cotyledon primordia becoming apparent at the triangular stage. At the heart stage, the patterning is complete and during later development this pattern is merely elaborated upon (Jürgens and Mayer, 1994).
In both plant and animal systems, initial patterning events, and therefore embryogenesis, are topics of great interest to developmental biology. However, morphological characterization alone is not sufficient to elucidate the underlying mechanisms. In order to scrutinize embryogenesis thoroughly, it is necessary to employ additional techniques such as the analysis of mutants and transgenic lines, microsurgery or external manipulation of embryo development. The latter, in particular, requires the possibility of in vitro embryo cultivation.
In some animal systems such as Drosophila or mice this is possible (for a review see New, 1990). In contrast, such experimental approaches are not very well established in plant systems, and extremely limited in Arabidopsis.
One possible way to culture plant embryos is associated with somatic embryogenesis. Under certain culture conditions plants will form somatic embryos that undergo an embryogenesis-like development in vitro and eventually become mature plants (e.g. De Vries et al., 1988; Huang and Yeoman, 1984; Luo and Koop, 1997; Mordhorst et al., 1998). There are two main limitations to this approach for studying embryogenesis. The first is that the culture media usually contain high levels of phytohormones, or that specific Arabidopsis mutants, such as the primordia timing (pt) are used, both of which affect normal zygotic embryogenesis (Friml et al., 2003; Liu et al., 1993a). The second limitation lies in the question of whether somatic embryogenesis follows the same developmental program as zygotic embryogenesis. Although somatic embryogenesis requires genes also known to be required for zygotic embryogenesis (Mordhorst et al., 2002), all experiments based on somatic embryogenesis suffer from the absence of well defined early stages. Somatic Arabidopsis embryos (derived from zygotic embryos) up to the four-cell stage have been described (Luo and Koop, 1997), however, the subsequent cell division pattern becomes very irregular. Moreover, somatic embryos derived from leaf protoplasts exhibit some resemblance to normal zygotic embryos, but they are less regular and naturally lack a suspensor. In summary, somatic embryo culture appears to be a limited substitute for the study of the natural course of early zygotic embryogenesis.
A better approach is the culture of zygotic embryos. In Brassica juncea, successful methods for in vitro embryo culture and treatment with substances have been shown (Hadfi et al., 1998; Liu et al., 1993a,b). A similar method has also been described in Arabidopsis (Kost et al., 1992). However, these works incorporated removal of the embryos from their ovules, which especially in the case of Arabidopsis, makes it almost impossible to analyze stages younger than the globular stage. Nevertheless, the use of Arabidopsis is highly favorable, due to the obvious advantages of a fully sequenced genome, easy transformation techniques as well as readily available mutant, T-DNA insertion and marker lines.
To overcome the manual problems of embryo excision, we looked for alternative approaches. Methods based on the culture of whole ovules are widespread in the plant field. However, most published procedures do not focus on embryogenesis itself, but were developed in order to rescue specific hybrids or to investigate fiber development (e.g. for grapes: Cain et al., 1983; for cotton: Beasley, 1971; for wheat: Kumlehn et al., 1997). Their greatest advantage is the relative ease of ovule preparation in contrast to embryo excision, thus allowing access to early stages and the use of large sample sizes. Additionally, these protocols do not require phytohormones in the medium, nor are they limited to specific mutants.
Here we present a method whereby Arabidopsis embryos can be cultivated in their ovules, in vitro, for prolonged periods of time from the very first developmental stages onward. We provide examples of its application for embryological studies incorporating exogenous treatments with substances, reporter gene-based markers as well as immunocytochemical methods.
Results and discussion
Procedure of in vitro culture
To establish the procedure, we had to optimize ovule isolation under semi-sterile conditions, culture conditions such as light and temperature, and especially, medium composition. For the culture, we first selected siliques of the appropriate stages and sterilized them by a short dip in 70% EtOH. The siliques were allowed to dry and placed onto double adhesive tape that was likewise sterilized. Under a dissecting scope, the siliques were cut open along the replum with needles and the ovules were carefully transferred onto plates containing in vitro culture medium (ICM). Simple shielding of the dissecting scope from air draughts decreased fungal contamination problems considerably, with typically about 5–8% of the plates contaminated. This low contamination rate, given the relative ease and speed of preparation, does not justify more elaborate sterilization or ovule preparation protocols. Microscopic examination of a few ovules is sufficient to determine the developmental stage of the embryos, as embryos within one silique do not differ much in their developmental stage (Bowman, 1994). The plates were then sealed with parafilm and turned upside down, to prevent condensating water from dripping onto the ovules. The plates were kept in the dark. It did not matter if the ovules were placed on the medium singly or in small clusters. Furthermore, it did not significantly affect survival if the funiculus or even more maternal tissue was still attached to the ovule (not shown). However, submerging the ovules in the medium seemed to have adverse effects on survival. Of particular importance was the composition of the culture medium. In previous experiments, we used a lower sucrose content of only 2%, but the survival rates were much lower, probably due to osmotic problems as the ovules looked bloated (not shown). Using a higher (10%) sucrose concentration improved survival rates. However, the high sucrose content negatively affects the later steps of embryogenesis and plant development (see below). Thus for longer culture durations (typically exceeding 5 days), we found it best to transfer the ovules from ICM and darkness to normal Arabidopsis medium (AM) and dim light under sterile conditions.
To assess the efficiency of our culture system, we cultured ovules for 5 days on ICM. The initial embryonic stage was determined as mentioned above. The ‘young’ group contained mostly embryos between four- and 16-cell stage, the ‘old’ group 32 cell and globular stage embryos at the start of culture. Using a stereoscope, the vitality of the ovules was easily evaluated. The viable ones were bigger, translucent or pale white and round compared with necrotic ovules of brownish color or flat appearance (Figure 2b). The survival rate was influenced by the starting stage, older ovules being less sensitive than younger ones, probably due to their greater mechanical stability (Figure 1a). Overall, the vitality of the embryo, determined by microscopic analysis, correlated with the vitality of the ovule (Figure 1a). We confirmed this correlation by staining with propidium iodide (PI), which stains dead cells and fluorescein diacetate (FDA), which stains living cells (Rotman and Papermaster, 1966). As expected, embryos from necrotic ovules were strongly stained by PI (Figure 2c), whereas embryos from viable ovules were not stained by PI but by FDA (Figure 2d). Thus the fast analysis with a stereomicroscope is a reasonable way to non-invasively determine the vitality of the cultured embryos. Keeping the plants on ICM medium for periods longer than 5 days affected the survival rate negatively (Figure 1b). For even longer culture durations (>10 days), ICM abolished the development of real leaves and a root system, the plants rather formed many leaf primordia and produced increased amounts of anthocyans (Figure 2e). The transfer to AM after 5 days improved the general survival rate (Figure 1b). When ovules were cultured on AM medium for the rest of embryogenesis, the embryos were able to germinate, developed rosette leaves and a root system, finally bolted and flowered (Figure 2a). The duration until they started to bolt varied in the range of 40–50 days after start of culture. Based on these analyses, approximately 20% of the plants will reach maturity using the method with the transfer from ICM to AM after 5 days of culture.
For the very early stages up to 24 h after pollination, when the zygotes or embryos are still not detectable, our method is less efficient. About 50% of the ovules were dead (small or flat) already after 1 day (n > 50), which did not change after longer culture duration.
In summary, this analysis shows that complete embryogenesis starting from a zygote and finally yielding a mature plant can be readily reproduced in our culture system.
Next we addressed the speed of development in our culture system. We cultivated ovules from young siliques of approximately the same age. At the start, and after 12, 24, 48, 72 and 96 h of culture, we took a random sample and determined the youngest embryo stages in each sample. At the start the youngest stage was one cell, after 12 h, embryos had reached two or four-cell stage, after 24 h eight to 16-cell stage, after 48 h early to mid globular stage, after 72 h triangular stage and after 96 h young to mid heart stage (Figure 2a).
As a complementary approach, we cultured ovules from siliques roughly divided into a ‘young’ and an ‘old’ group. We determined the developmental stages of randomly selected embryos at the start and after 5 days of culture, and calculated the distribution of developmental stages (Figure 1c). The later stages (torpedo and beyond) are less well defined and include embryos with different cell numbers but similar shape, consequently the graph is compressed toward the later stages. At the start, the most represented stages were eight- and 16-cell stages in the ‘young’ group. After 5 days of culture, most embryos were at late heart stage. For the ‘old’ group, the most abundant starting stage was the globular stage, after 5 days most embryos had reached walking stick to bent cotyledon stage. Compared with uncultured embryos, this seems to be slightly slower. It may be that already after 5 days adverse effects of the high sucrose concentration in ICM become apparent, as keeping the ovules on ICM for prolonged periods of time leads to aberrant and delayed development (see above). The presence or absence of dim light (culture plates covered with a sheet of plain paper or aluminum foil, respectively) did not influence the speed of development, but the embryos cultured in light developed chlorophyll at later stages and the ovules became green (not shown).
To verify the somewhat reduced speed of development, we compared the timing of development in planta with our culture system. To synchronize embryo development, we used ovules from emasculated and manually pollinated siliques. Twenty-four hours after pollination, we started culturing the ovules. We then compared the cultured embryos with embryos from a control group whose siliques remained on the plant. After 4 days, 46% (n = 28) of the control embryos reached the late globular stage and 50% reached the triangular stage. Of the surviving cultured embryos, 42% were at the 32-cell stage, 28% at the globular and 21% at the triangular stage. This confirms that the speed of development in culture is reduced compared with development in planta.
Most in vitro ovule culture systems report developmental aberrations in varying frequencies (Kumlehn et al., 1997). In these systems, this is usually not of great importance, as the purpose is the regeneration of adult plants from otherwise non-viable embryos. However, for studying embryogenesis itself, it is desirable to have a system which keeps aberrations to a minimum. Using our method, it was possible to culture embryos from the one-cell stage on for up to 3 days without inducing aberrations (when only embryos from macroscopically viable ovules were analyzed). This is sufficient for most studies, as major patterning events like apical–basal axis initiation as well as radial and bilateral symmetry establishment are completed during these 3 days. After 3.5 days, about 10% of the embryos exhibited defects in the root pole, with extra and irregular cell divisions (Figure 2f). The frequency of these aberrations increased with culture duration to about 30% after 4.5 days; other defects such as apically broadened heart stages (Figure 2g) were also observed. It seemed that embryos cultured from younger stages (one to four cell) were more sensitive than those of later stages (eight to 16 cell). This might be due to an already preset patterning in the older embryos or a negative effect of the medium to such young embryos. In most cases, severe aberrations, which were visible after 5 days, were linked to a morphologically clearly discernible defect of the ovule (Figure 2h), so it was easy to exclude these embryos from later analysis. Nonetheless, 2.4% (n = 380) strong and 12.1% slight embryo defects were not associated with obvious ovule defects after 5 days of culturing. The slight defects were mainly in the root pole or a ‘folded’ hypocotyl phenotype (Figure 2i), where it seems that the ovule did not expand as fast as the embryo. Functionally, these aberrations were of little relevance, because these embryos still had all necessary organs to reach maturity. These data show that our in vitro culture method induces aberrations only after a prolonged period of culturing, these aberrations are predictable and their frequencies are acceptably low. Special caution should be exerted when investigating subtle phenotypes in embryos cultured from very young embryonic stages onward. However, these problems can be overcome by larger sample numbers, which are quite easy to realize.
Use of treatments
The most important advantage of the in vitro culture system is the possibility to chemically interfere with various cellular processes and study the consequences for embryo development. Therefore, we supplemented the culture medium with different drugs and examined their effects on embryo development. Such studies using the presented system were instrumental in establishing the role of the plant hormone auxin and its spatial distribution in apical–basal axis formation (Friml et al., 2003) and bilateral symmetry establishment (Benkováet al., 2003).
To further demonstrate the applicability of this approach, we administered drugs known to interfere with vesicle transport and cytoskeletal structure. First, we compared the influence of brefeldin A (BFA) and monensin, both drugs known to affect intracellular vesicle trafficking. BFA specifically inhibits ARF-GEFs, including GNOM, which mediates trafficking of the PIN1 auxin transport regulator (Geldner et al., 2003). By this means BFA should indirectly affect auxin transport. Indeed, after treatment with BFA (20 μm), ball-shaped embryos with no defined apical–basal axis developed (Figure 3a), which strikingly resembled gnom or multiple pin mutant embryos (Friml et al., 2003). As expected, these BFA effects were not observed (Figure 3b compared with 3a), when we cultured ovules of GNOMM696L-myc transgenic plants, in which GNOM is engineered to be resistant to BFA inhibition (Geldner et al., 2003). These results suggest that BFA exhibited auxin transport-related effects on embryo development mainly through GNOM. On the contrary, monensin – another established vesicle trafficking inhibitor (Mollenhauer et al., 1990) – had very different effects on embryogenesis. Monensin treatment (20 μm), in contrast to BFA, did not lead to any specific patterning defects but rather to arrested development (Figure 3c compared with 3b) or frequent embryo and ovule lethality (more than 80%).
As another example of drug treatment, we interfered with the cytoskeleton. The herbicide oryzalin is known to depolymerize tubulin filaments and thus inhibits cell division (Bajer and Mole-Bajer, 1986). We found that after 2 days of culturing in the presence of oryzalin (5–10 μm), embryos remained at earlier developmental stages and displayed big, bloated and multinucleate cells (Figure 3d, compared with control 3e after same culture duration and at same scale). These defects are strikingly similar to those observed in tubulin-deficient mutants (Mayer and Jürgens, 2002). In contrast, latrunculin B, which interferes with the actin cytoskeleton, had different, more variable effects on embryo development. Most ovules did not survive the treatment. In those that survived, the morphology of embryonic cells changed and irregular cell divisions were found (not shown).
Without characterizing the role of vesicle traffic or the cytoskeleton in embryos in detail, our results demonstrate that embryo development can be effectively manipulated by exogenous chemical treatments in our culture system. In most cases, we had to use higher substance concentrations than those proved effective in root treatments. This might be due to the diffusion barrier of the ovule or lower stability of some substances under culture conditions. Therefore, the optimal concentration for any new substance to be used has to be determined in advance.
Use of markers
One of the main advantages of Arabidopsis as a model system is the availability of various mutants and transgenic marker lines as well as antibodies against different proteins. To test the use of markers in the in vitro system, we used various GFP-, GUS- and antibody-based markers. As examples of GFP based markers, we selected and characterized several embryonically expressed enhancer trap lines from available collections (M0148, M0164, M0171; Haseloff, http://www.plantsci.cam.ac.uk/Haseloff/catalogues/Mlines/index.html). First, we tested whether culturing interferes with marker expression. Cultured embryos showed the same pattern of marker expression as in planta (Haseloff, http://www.plantsci.cam.ac.uk/Haseloff/catalogues/Mlines/index.html). In the M0148 line, we found the GFP fluorescence in a broad band in the transition zone between hypocotyl and cotyledons (Figure 4a); in the M0164 line, the signal was restricted to the cotyledon provasculature (Figure 4c,e); and for M0171, expression was detected between the cotyledons and at the root pole (Figure 4g). Treatment with the synthetic auxin analogue 2,4-D or the auxin transport inhibitor NPA, both of which are known to interfere with embryo patterning (Friml et al., 2003) also interfered with marker expression. The expression of the cotyledon-based marker M0148 was lost (Figure 4b), correlating with the morphologically visible failure of cotyledon establishment. The cotyledon provasculature marker M0164 shifted higher up into the cotyledons and the pattern in the vascular system was disrupted (Figure 4d,f). This indicated interference with the early stages of vascular differentiation, similar to reported post-embryonic NPA-induced vascular defects (Mattsson et al., 1999). Expression of M0171 in the upper parts was no longer detectable after the treatments, but in the lower parts, the signal was enhanced and expanded, with additional signals in the epidermal layer (Figure 4h). These results confirm a role of auxin and its distribution in specification of different embryo parts. They also demonstrate the usability of GFP-based marker lines, as we did not encounter any problems with changed expression patterns, weaker signal strength or high background fluorescence after in vitro culture.
Next, we tested the applicability of GUS-based markers using a CUC3pr::GUS reporter line, which is specifically expressed at organ borders (Vroemen et al., 2003). We performed histochemical detection of GUS activity (Weijers et al., 2001) on NPA treated and control embryos. In controls (Figure 4i), GUS activity was localized at the border between the cotyledons, whereas upon NPA treatment, the cotyledons had not separated and GUS activity was mislocalized (Figure 4j). This result shows that the GUS assay can be successfully used on in vitro cultured embryos. We also tested whether immunolocalization procedures can be used for cultured embryos. We adapted the previously described protocol (Lauber et al., 1997), to be performed semi-automated by a blot-processing robot. Given that about half of the embryos are lost during culture and another part during the immunolocalization procedure, it is necessary to culture large numbers of ovules, which is easy to achieve with our system. To test both the in vitro culture as well as the immunolocalization method, we used two well-characterized antibodies – anti-PIN1 and anti-PM-ATPase. After culturing, the pattern of PIN1 was localized to the basal and apical sides of embryo provascular and epidermal cells, respectively (Figure 4k). This pattern matches the one previously reported for uncultured embryos (Steinmann et al., 1999). Similarly, PM-ATPase was found at the borders of all embryo cells (Figure 4l) again showing the same localization pattern as reported for embryos developing in planta (Rober-Kleber et al., 2003). Based on these results, we conclude that our in vitro culture protocol does not interfere with the subsequent use of immunocytochemical methods.
In summary, we find that all tested types of markers based on GFP or GUS reporters or antibodies, which are used to analyze embryo development in planta, can also be applied to the cultured embryos without any limitations.
Here we present a method for in vitro culture of Arabidopsis embryos within their ovules. This method has several advantages over previously published protocols, which involve embryo excision or which are based on somatic embryogenesis. Somatic embryogenesis involves phytohormone treatment and does not entirely resemble normal zygotic embryogenesis, whereas excision is impossible for the youngest stages and manipulation is tedious. The major advantages of the presented system are accessibility of even the youngest embryos, a hormone-free medium and the mechanical ease of preparation, allowing large sample numbers. It allows culturing from the zygote stage up to the viable grown-up plant. Longer culture periods cause predictable developmental irregularities at low rates, nonetheless, for shorter durations, in vitro development mirrors that in planta. We demonstrate that using this system, embryogenesis can be chemically interfered with and available markers and protocols established for in planta embryos can be used. In summary, the presented method provides a platform useful for physiological approaches in analyzing Arabidopsis embryogenesis. This, in conjunction with well established genetic, molecular biology and emerging cell biological tools in Arabidopsis will be instrumental in the future to gain new insights into the mechanisms controlling early events in plant development.
The medium for in vitro culture of ovules (ICM) contained 10% sucrose, Murashige and Skoog (1962) salts (×0.5) and 0.3% phytagel (Sigma, St Louis, MO, USA), pH was adjusted with KOH to 5.9. After brief autoclaving (10 min), the medium was allowed to cool down to 50°C, then 25 ml l−1 of a 16 g l−1 aqueous stock solution of glutamine (final concentration 400 mg l−1) was sterile-filtered into the medium. Other substances such as phytohormones were added before pouring into small (50 mm) Petri dishes. The medium should not be remelted. AM contained 1% sucrose, 0.5 × Murashige and Skoog (1962) salts, 0.05% MES and 0.8% Agar (Serva, Heidelberg, Germany). pH was adjusted to 5.8 with KOH. Unless otherwise noted, we used 50 μm of 2,4-D and NPA, 20 μm of BFA and monensin and 10 μm of oryzalin and latrunculin B and cultured for 3 days under dim light.
Immunolocalization was modified from Lauber et al. (1997). As buffer, PBS (pH 7,4) instead of MTSB was used. Prior to fixation we included a short washing step in water to remove any traces of medium from the ovules. Cell wall digestion with 2% driselase (Sigma) was carried out for 40 min, for membrane permeabilization, 3% NP40 10% DMSO in PBS was used (1h). All subsequent washing and incubation steps were automated using the BioLaneTM HTI blot processor (Hölle & Hüttner, Tübingen, Germany). Application of 2% BSA blocking solution (2h) and antibodies also in 2% BSA was carried out manually. Primary anti-PIN1 (Gälweiler et al., 1998; 1:400), anti-PM-ATPase (1:500; kindly provided by W. Michalke) and secondary anti-rabbit (1:400) as well as anti-mouse-CY3-conjugated (1:600) antibodies (Dianova, Hamburg, Germany) were used.
Histological staining and microscopy
The enzymatic reaction for determining GUS activity was performed as described by Weijers et al. (2001), with a previous washing step in water to remove traces of medium from the ovules. The material was observed with a Zeiss Axiophot (Carl Zeiss, Jena, Germany) with differential interference contrast optics or epifluorescence using FITC optimized filters. Images were taken with an Axiocam HR. Confocal laser scanning microscopy was carried out on a Leica TCS SP with the LCS Software Package. Photoshop 7.0 (Adobe Systems Inc.) was used for image processing.
We are very grateful to M. Griffith for substantial help with the establishment of the in vitro procedure. We thank S. Hiller, P. Hrušáková, and J. Mravec for technical assistance; E. Benková, K. Steinborn, A. Vieten, and D. Weijers for generous supply of plant material; W. Michalke for providing the anti-PM-ATPase antibodies and D. Weijers and P. Brewer for critical reading of the manuscript. This work was supported by the Volkswagenstiftung.