Laser capture microdissection for the analysis of gene expression during embryogenesis of Arabidopsis


(fax +0191 334 1201; e-mail


It is during embryogenesis that the body plan of the developing plant is established. Analysis of gene expression during embryogenesis has been limited due to the technical difficulty of accessing the developing embryo. Here we demonstrate that laser capture microdissection can be applied to the analysis of embryogenesis. We show how this technique can be used in concert with DNA microarray for the large-scale analysis of gene expression in apical and basal domains of the globular-stage and heart-stage embryo, respectively, when critical events of polarity, symmetry and biochemical differentiation are established. This high resolution spatial analysis shows that up to approximately 65% of the genome is expressed in the developing embryo, and that differential expression of a number of gene classes can be detected. We discuss the validity of this approach for the functional analysis of both published and previously uncharacterized essential genes.


Higher plant embryogenesis is a critical phase of the life cycle, and has been the object of intense study by geneticists, developmental biologists and biochemists for decades (Franzmann et al., 1995; Laux et al., 2004; Mayer et al., 1991; Scheres et al., 1994). It is during this part of the life cycle that the basic body plan of the seedling is established. An understanding of the control mechanisms of embryogenesis can provide insight into the fundamentals of developmental regulation, including pattern formation, morphogenesis and cell differentiation; of metabolic regulation, such as the developmentally regulated synthesis and accumulation of seed storage oils and proteins; and the integration of these processes by inter- and intra-cellular signalling systems. Arabidopsis thaliana has proved to be a very powerful model organism in which to dissect the regulatory mechanisms of embryo development (Laux et al., 2004; Souter and Lindsey, 2000).

The establishment of polarity is an important feature of many developmental systems and differentiation processes, including Arabidopsis embryogenesis. The first division of the Arabidopsis zygote is asymmetric, producing two daughter cells that have quite distinct fates, forming the embryo-proper (which contributes to the seedling) and the multicellular suspensor, which (with the exception of a single, upper cell) does not, and undergoes programmed cell death. As development proceeds, polarity is reinforced with the establishment of meristematic poles at either end of the embryo-proper, forming the apical-basal axis. Superimposed is a radial axis, defined by the formation of concentric layers of cells that differentiate into the epidermis, ground tissue and vascular cylinder. Several studies have demonstrated the requirement for tightly controlled gene expression to allow correct embryogenesis to occur (Friml et al., 2003; Helariutta et al., 2000; Mahonen et al., 2000; Schoof et al., 2000).

Previous research has exploited a variety of techniques to identify genes expressed in the developing embryo, including the isolation of RNA for cDNA library construction and screening (Goldberg et al., 1989), promoter/enhancer trapping (Topping et al., 1994) and, most powerfully in terms of identifying essential genes, mutational screens (Mayer et al., 1991; Meinke and Sussex, 1979). It has been estimated that there are approximately 40 genes required for pattern formation in Arabidopsis embryogenesis (Mayer et al., 1991), but a larger number of genes are essential, based on the frequency of embryonic-lethal mutations (Franzmann et al., 1995; Tzafrir et al., 2004). However the total number of genes expressed has been estimated in the thousands (Girke et al., 2000; Goldberg et al., 1989; White et al., 2000). Sequencing of ESTs from Arabidopsis seed cDNA indicated about 3000 genes are expressed, though this number would be biased downwards as the lowest abundance transcripts would be under-represented (White et al., 2000). A limitation to the cDNA approach, which has the potential to reveal large numbers of expressed genes, is the small amounts of RNA that can be isolated from the youngest stages of embryogenesis. The use of DNA microarray technology potentially takes this expression analysis further, and has been used to study transcriptional changes in seed development in Arabidopsis (Girke et al., 2000). To date this has involved the use of DNA chips with limited genome coverage (<30%), with RNA isolated from whole seeds.

Analysis of gene expression by in situ hybridization or promoter-driven reporter genes shows that even from the earliest stages of development, different cell types and regions of developing embryos differ not only in their morphology and position but also in their transcriptomes (Friml et al., 2003; Haecker et al., 2004; Mansfield and Briarty, 1991; Mayer and Jürgens, 1998). Therefore, being able to access the transcriptome of different regions of developing embryos would provide crucial information on what, if any, qualitative or quantitative differences exist and how these may govern the fates of these regions. The small size of Arabidopsis embryos and their inaccessibility means that the ability to identify and isolate small groups of specific cells is technically demanding.

Recently, the development of laser capture microdissection (LCM) has enabled the isolation of a target group of cells from a complex tissue (Bonner et al., 1997; Emmert-Buck et al., 1996). LCM was originally developed to isolate pure populations of cells from histological sections of complex tissue in animal systems (Emmert-Buck et al., 1996). Reports on the application of LCM to plant systems (Asano et al., 2002; Kerk et al., 2003; Nakazono et al., 2003) illustrate its applicability for cDNA library construction or the RT-PCR analysis of specific plant genes. In this study we show the generation of target RNA from very small tissue samples that represent specific domains of the developing embryo and discuss its value for high resolution transcriptional profiling of embryogenesis.

Results and discussion

Sample preparation, LCM and RNA amplification

One of the most important factors for performing LCM successfully is achieving good histological tissue sections. In the case of plant tissue the best histological sections are achieved using either paraffin or plastic embedding. Unfortunately in the case of these techniques, the fixatives used and other aspects of tissue processing negatively affect the quality of the RNA that is isolated (Bustin, 2002; Gillespie et al., 2002). Fresh frozen tissue and cryosectioning give the best yield and quality of RNA for LCM (Gillespie et al., 2002). A concern with plant tissue is that ice crystals can form in vacuoles and the large air spaces found between cells in mature tissues. However, we found that the histology of small, cytoplasmically dense embryonic cells was not adversely affected by freezing. Good quality sections were achieved by dissecting embryo sacs from siliques (Col-0 ecotype), embedding these in inert OCT medium in moulds, freezing in liquid nitrogen-cooled isopentane and then sectioning blocks on a cryostat at −22°C (see Experimental procedures). For mature tissue types we have found that a brief fixation with ethanol:acetic acid and infiltration with 10–15% sucrose prior to freezing in isopentane is required to generate sections suitable for LCM (data not shown).

Following sectioning, tissues were dehydrated through an ethanol gradient and finally in xylene before proceeding to LCM (see Experimental procedures). Uncoated slides were the most reliable in terms of tissue adherence for LCM and a 5–10 μm tissue section gave the best transfers. Due to the limitations of the PixCell IITM LCM system used, which has a minimum laser beam diameter of 7.5 μm, we found it almost impossible to isolate specific, single embryonic cells, which were often smaller than this. However, it was possible reproducibly to capture small (approximately five to 10) cell clusters from globular stage embryos onwards. As transfer was not always complete for such small clusters, a number of replicates were performed, with approximately 15 cell clusters (typically 100–200 cells in total) captured ready for RNA isolation, and for transcript profiling using the Arabidopsis ATH1 GeneChip.

There are several histologically distinct embryonic stages in Arabidopsis (Jürgens, 1996). The heart-stage embryo displays both apical-basal polarity and radial symmetry, and is the first stage at which the cells that will form the meristems of the seedling can be distinguished. These features are not evident in globular-stage embryos and therefore the transition between globular- and heart-stage is a key developmental stage. For this reason embryos of these stages were chosen for analysis. Both the apical and basal domains of globular-stage embryos (Figure 1a–g), and also the root pole and cotyledonary tissue of heart-stage embryos (Figure 1h–n) were captured for gene expression analysis. Transcriptome analysis of these cell clusters was expected to provide insights into the transcriptional changes occurring during embryo development, and also specifically with regard to the formation of the root meristem.

Figure 1.

Laser capture microdissection of cryosections of globular- and heart-stage Arabidopsis embryos.
(a–g) LCM of globular-stage embryo. (a) Globular-stage embryo. (b) Basal region after targeting with laser, (c) after removal of the cap and (d) basal cells captured on cap. (e) Targeting of the apical region, (f) after removal of the cap, and (g) apical cells on the cap.
(h–n) LCM of heart-stage embryo. (h) Heart-stage embryo. (i) Root pole after targeting with laser, (j) after removal of the cap, (k) root pole cells on cap. (l) Targeting of cotyledon pole cells, (m) after removal of the cap, and (n) cotyledon pole cells on the cap.

Zimmerman and Goldberg (1977) estimated that a single tobacco leaf cell contains approximately 38 pg of polysomal RNA. As LCM does not necessarily capture a complete cell, it was expected that the yield of RNA from captured cells would be insufficient both to quantify accurately and for GeneChip analysis. Therefore, whilst RT-PCR could be performed successfully on RNA extracted from LCM embryonic cells (see below), RNA amplification steps were required to generate sufficient RNA for microarray analysis.

One problem with amplifying such small quantities of mRNA is maintaining the population distribution of the transcripts. Standard PCR methods are generally considered to increase bias in a heterologous population by preferentially amplifying smaller or more abundant targets. More specialized PCR methods have been developed in which product size is limited such that only the 3′ end of cDNA is amplified (Iscove et al., 2002), but these are more suited to cDNA and not to Affymetrix microarray formats. We therefore used linear T7 RNA polymerase-mediated amplification (Van Gelder et al., 1990). This approach has previously been used for amplification of animal cell RNA prior to DNA microarray analysis (Luo et al., 1999; Salunga et al., 1999). Following DNase treatment of the RNA isolated from LCM cells, three rounds of amplification generated approximately 5–10 μg of amplified RNA (aRNA). Gel analysis indicated a size distribution of between 100 and 500 nt (Figure 2a,b).

Figure 2.

Gel and RT-PCR analysis of aRNA.
(a) aRNA from globular embryos after three rounds of amplification. AP, globular apical; Ba, globular basal.
(b) aRNA from heart embryos after three rounds of amplification. CO, heart cotyledonary pole; RT, heart root pole.
(c) RT-PCR analysis of aRNA (three rounds of amplification) from heart-stage embryos. Rt, root pole; Co, cotyledonary pole.
(d) RT-PCR analysis of aRNA from the apical pole of globular-stage embryos.
(e) RT-PCR analysis of aRNA from the basal pole of globular-stage embryos.

Prior to microarray analysis, RT-PCR experiments were performed to determine whether the aRNA was a suitable template and also whether differential gene expression could be detected. For example, in situ hybridization experiments show localization of PINOID (PID) to the cotyledon poles of heart-stage embryos (Christensen et al., 2000), AINTEGUMENTA (ANT) is localized to the apical region of globular- and the cotyledon poles of heart-stage embryos (Long and Barton, 1998), and AtPIN4 protein is predominantly localized at the root pole of globular-stage embryos (Friml et al., 2002). RT-PCR experiments on the aRNA supported the in situ data, besides detecting a number of other key embryonic genes (Figure 2c–e). All the RT-PCR products were mRNA specific, because oligonucleotide primers were designed around introns.

Microarray analysis

The amount of information to be gained from RT-PCR experiments is limited and we therefore investigated the use of the Affymetrix ATH1 GeneChip for a more detailed transcriptional analysis. Three biological replicates were performed for each tissue (globular apical, globular basal, heart cotyledon, heart root) with each replicate consisting of approximately 10–20 cell clusters. In contrast to unamplified mRNA, aRNA exhibits a certain degree of 3′ bias, in that there is an enrichment of 3′ sequence information compared with 5′. This occurs during cDNA synthesis, where random priming is required for first or second strand synthesis. Multiple rounds of cDNA synthesis and RNA amplification increase the amount of 3′-5′ bias. This was taken account of during the GeneChip analysis by reducing the number of probe pairs used and restricting them to those designed towards the 3′ end of transcripts.

In a recent study of the root transcriptome, the application of a minimal Affymetrix signal of 75 indicated that 10 465 genes were expressed, corresponding to approximately 46% of the genes represented on the ATH1 GeneChip (Birnbaum et al., 2003). Using the same cut-off value of 75 for the mean value of the three replicates analysed, our data indicate that between 8027 and 10 591 genes (36–47%) are expressed in the apical and basal domains of the globular- and heart-stage embryos, respectively, with the number dropping to between approximately 18 and 22% at a cut-off threshold of 150 (Table S1). This is therefore a similar number of genes as found to be expressed in the Arabidopsis root. There is some variance in the data between replicates, which may be due either to biological variability between samples or to technical variation. This variation in signal call (positive versus negative) is most apparent in the case of genes with very low expression levels, around the cut-off threshold.

If a lower threshold mean signal value of 40 is used, the data indicate that up to 65% of the genes are expressed. An argument for using the lower mean signal value of 40 is that genes that have found to be expressed at low levels by other techniques such as RT-PCR or in situ hybridization analysis (e.g. AINTEGUMENTA in the apical region of globular-stage embryos or MONOPTEROS in the basal region of the heart-stage root, Table 1) would be included as positives at that signal value, but as negatives at a mean signal cut-off value of 75.

Table 1.  The expression of embryonic genes based on published in situ hybridization or promoter::GUS analysis compared with their expression determined by LCM and GeneChip analysis. The mean signal of the three replicates for each tissue is shown
GeneExpression in globular (G) and heart (H) stageReferenceGlobular apicalGlobular basalHeart cotyledonHeart root
AINTEGUMENTAG, presumptive cotyledons; H, cotyledonsElliott et al. (1996), Long and Barton (1998)49.129.11247.516.0
ARGONAUTE1G + H, throughout embryoLynn et al. (1999)200.686.21198.92057.1
AtPIN1G, central cells; H, vascular precursorsSteinmann et al. (1999)196.6189.6188.5148.8
AtPIN2Not expressedFriml et al. (2003)25.651.645.126.7
AtPIN3H, root poleFriml et al. (2003)123.8108.480.277.3
AtPIN4G, lens-shaped cell + provascular initial; H, root poleFriml et al. (2002)76.8134.6113.8499.8
BODENLOSG + H, provascular cellsHamann et al. (2002)65.923844.3258.9
FUS3G, throughout embryo; H, protodermTsuchiya et al. (2004)140.4180.0186.2356.7
LEC1G + H, outer protoderm and ground tissueLotan et al. (1998)169.6201.5228140.3
MONOPTEROSG + H, provascular cellHamann et al. (2002)92.618384.552.7
PINOIDG, apical region; H, outer layers of cotyledon and hypocotylChristensen et al. (2000)135.699.392143.5
WOX2G, apical region; H, epidermal cells of root-hypocotyl junctionHaecker et al. (2004)496.3170.1130348.8
ZWILLEG + H, provascular cellsMoussian et al. (1998)366.5171.1747.61223.6

Taken together, our gene coverage data show that there are differences in the numbers of genes expressed in apical versus basal domains in globular- and heart-stage embryos, and, depending on the cut-off threshold, up to between approximately 40 and 65% of the genes represented on the ATH1 GeneChip are expressed. Our ATH1 GeneChip data can be fully accessed through the Nottingham Arabidopsis Stock Centre (Craigon et al., 2004;

Comparisons with published expression patterns of embryonic genes

One approach we used to validate the data was to compare the expression data obtained following LCM with those reported in the literature for genes for which embryonic expression is known (Table 1). For example, the ANT gene encodes a protein with APETALA2-like DNA binding domains (Klucher et al., 1996). The ant mutant displays a number of floral and ovule defects. In situ hybridization determined that the ANT gene is expressed throughout the floral organs, as well as in the embryo (Elliott et al., 1996; Long and Barton, 1998). ANT expression is first seen in late globular-/early heart-stage embryos in the cotyledon primordia and is then seen throughout the cotyledons in older heart- and torpedo-stage embryos, with no expression elsewhere in the embryo. When we compared our Affymetrix GeneChip data for heart-stage cotyledon versus heart-stage root we found that ANT expression was absent in all three root replicates (mean signal, 16) and highly expressed in the cotyledon replicates (mean signal, 1247.5). According to our data, ANT is 78 times more abundant in cotyledon tissue than root. By using a signal threshold value of 40, ANT expression is detectable in the apical but not basal cells of globular embryos, a result which is in agreement with our RT-PCR analysis in which ANT mRNA was found specifically in the apical region (Figure 2d,e), and also with in situ hybridization data (Long and Barton, 1998).

The BODENLOS/IAA12 gene encodes a member of the short-lived IAA proteins (Hamann et al., 2002), which are required for mediating auxin responses. Defects in bdl mutants are observed during embryogenesis, as early as the two-cell stage, with bdl seedlings defective in primary root development (Hamann et al., 1999). BDL transcript was detected by in situ analysis in the provascular cells from the globular stage onwards, whereas a pBDL:GUS construct indicates that expression is strongest in the root pole of heart-stage embryos (Hamann et al., 2002). According to our Affymetrix GeneChip data, BDL mRNA is more abundant in the basal half of the globular embryo (mean signal, 238) compared with the apical cells (mean signal, 65.9). In heart-stage embryos, BDL is considered to be expressed in cotyledon tissue (mean signal, 44.3) at a signal threshold value of 40, but is expressed at higher levels in root tissue (mean signal, 258.9).

The PIN family of auxin efflux carriers is a good example of genes with differing expression patterns. PIN1 expression is found in the central cells of the globular embryo and vascular precursors of the heart stage (Steinmann et al., 1999). This expression pattern overlaps with all of the tissues sampled by LCM and is reflected in the similar levels of expression. In agreement with previously published data (Friml et al., 2003), PIN2 expression is around a cut-off value of 40, indicating no or very low expression levels. PIN4 is expressed most strongly in the root pole of heart-stage embryos (Friml et al., 2002, 2003) and this is correctly predicted by the LCM data (root mean signal, 499.8 versus cotyledon mean signal, 113.8). PIN3 is also expressed in the root pole of heart-stage embryos as determined by promoter:GUS and in situ analysis (Friml et al., 2003). In contrast to this, our Affymetrix GeneChip data show low levels in each of the tissues sampled and this may reflect a technical difficulty in being able to determine differential expression of genes when using a lower threshold signal value.

WOX2, a member of the WUSCHEL-related homeobox gene family, provides another good example for the validation of our data, in that expression is initially strongest in the apical cells of globular embryos. It then becomes restricted to a small number of cells in the epidermis of the root-hypocotyl junction of heart stage embryos (Haecker et al., 2004); this correlates well with the Affymetrix GeneChip data. Other members of the WOX family have been shown to have spatially restricted expression patterns (Haecker et al., 2004), but were either absent from the ATH1 GeneChip or were expressed at very low levels. The reliable detection of genes with lower transcript levels is one potential limitation of this technique, and whilst the Affymetrix GeneChip data correlate well with known gene expression data any discrepancies observed were mainly restricted to those genes with low expression levels (mean signal <100).

Embryonic expression data for previously characterized genes are limited to a relatively small number of genes, and so we carried out further validation of the GeneChip expression data by comparing it with activities of a number of promoter-GUS fusions, which were constructed and introduced into transgenic plants. Six genes, expressed respectively either constitutively (At1g48630, At2g02760) or predominantly in the apical (At2g21320) or basal (At1g15750, At5g16780, At5g17430) regions of either the globular- or heart-stage embryo, were chosen for analysis (Table 2). In all cases, the promoter-GUS activities correlated with the embryo RNA analysis (Figure 3a–f). In seedlings, the GUS expression data also correspond with the GeneChip expression data produced for different root cell types (Birnbaum et al., 2003; Table S2).

Table 2.  Genes chosen for promoter:GUS analysis. The mean expression determined by LCM and GeneChip analysis is shown
GeneDescriptionGlobular apicalGlobular basalHeart cotyledonHeart root
  1. *Indicates an abnormally high mean signal due to a large signal variation in one replicate.

At1g48630Guanine nucleotide-binding family protein967.612672037.51230.2
At2g21320CONSTANS-like B-box zinc finger protein1757.322.5314.1106.8
At2g02760Putative ubiquitin-conjugating enzyme E2639.5596.2101.73878*
At1g15750WD-40 repeat family protein123.395.9185.4491.6
At5g16780SART-1 family protein32.713.541692.1
At5g17430AINTEGUMENTA-like protein81.4975.228.1359.1
Figure 3.

Promoter:GUS analysis of embryonic genes.
Genes chosen for analysis included:
Constitutively expressed genes. (a) At1g48630, globular embryo, heart embryo, 7-day-old seedling root and tip; (b) At2g02760, heart embryo, 7-day-old seedling apical and root meristems.
Apically expressed genes. (c) At2g21320, 16-cell embryo, cotyledonary embryo and 7-day-old seedling.
Basally expressed genes. (d) At1g15750, late heart-stage embryo, cotyledonary stage embryo and 7-day-old seedling root tip. (e) At5g16780, late torpedo-stage embryo, 7-day-old seedling root tip, mature root and shoot apical meristem. (f) At5g17430, heart-stage embryo, 7-day-old seedling root tip and emerging lateral root.

Expression patterns of known EMB genes

It has been estimated that there are approximately 500 genes in Arabidopsis that when mutated produce an embryo-defective phenotype (Franzmann et al., 1995; Tzafrir et al., 2004). Tzafrir et al. (2004) have characterized a collection of 220 EMB genes that give embryonic phenotypes when disrupted. However, for large numbers of these genes there are no data on their expression during embryogenesis. Applying the signal threshold of 75 we can determine that approximately 60% of these genes are expressed in at least one of the embryo domains sampled (Figure 4, Table S3). If we apply a signal threshold of 40, then the data indicate that approximately 75% of EMB genes are expressed in at least one embryonic domain analysed. As our data only deal with a limited number of tissue zones it is possible that the remaining EMB genes are expressed during embryogenesis at either a different stage or cellular domain to those analysed.

Figure 4.

Detection of EMB gene expression in globular and heart-stage embryos.
(a) The Venn diagram indicates the number of EMB genes (Tzafrir et al., 2004 and supporting information) expressed (mean signal value >40) in either the globular apical or basal pole. The intersection indicates the number of genes expressed in both poles whilst numbers in parentheses indicate the number of genes not detectably expressed in either the apical or basal poles.
(b) Venn diagram as in (a) indicating EMB genes expressed in heart-stage cotyledonary or root poles.

When a subset of the EMB genes was analysed it was found that approximately 36% have a terminal phenotype at the cotyledon stage suggesting that they may not be expressed until the later stages of embryogenesis (Tzafrir et al., 2004). By analysing the microarray data it was also possible to determine that a number of the EMB genes are expressed in a spatially restricted or temporal manner (Table 3). For example, in heart-stage embryos At2g34650 (PID) is approximately 21-fold more abundant in cotyledon than root tissue, whilst At5g10480 [PEPINO/PASTICCINO2 (PEP/PAS2)] is approximately 9- to 10-fold more abundant in the apical versus the basal tissues of globular- and heart-stage embryos. Mutation of each of these genes leads to defects in the apical regions of the embryo, seedling or mature plant (Bennett et al., 1995; Haberer et al., 2002). In the case of pepino mutants, defects are observed throughout development with the most pronounced abnormalities found in the shoot, which displays aberrant cell proliferation and an enlarged shoot apical meristem (SAM). Defects in the apical cells are evident in heart-stage embryos, where the SAM appears broader and there is reduced outgrowth of the cotyledons (Haberer et al., 2002).

Table 3. EMB genes determined to be spatially or temporally expressed during embryogenesis. Genes that show at least a fourfold difference in expression between embryo zones
GeneGene symbolFold difference
  1. G-Ap, globular apical pole; G-Ba, globular basal pole; H-Co, heart cotyledonary pole; H-Rt, heart root pole.

At5g10480PAS29.7 × G-Ap > G-Ba
At1g78630EMB 14735.3 × G-Ap > G-Ba
At3g52940FK4.6 × G-Ap > G-Ba
At1g80070SUS24 × G-Ap > G-Ba
At5g61410EMB 27287 × G-Ba > G-Ap
At5g07500PEI4.5 × G-Ba > G-Ap
At2g34650PID21 × H-Co > H-Rt
At2g39770CYT6.7 × H-Co > H-Rt
At3g46560EMB 247411 × H-Rt > H-Co
At3g16290EMB 20838.1 × H-Rt > H-Co
At2g41720EMB 26544.5 × H-Rt > H-Co
At2g01420PIN44.4 × H-Rt > H-Co
At3g25860PLE228 × H-Co > G-Ap
At1g10510EMB 200425 × H-Co > G-Ap
At4g04350EMB 23697.8 × H-Co > G-Ap
At3g16290EMB 208330 × H-Rt > G-Ba
At4g10760EMB 170626.5 × H-Rt > G-Ba
At3g46560EMB 247411.5 × G-Ap > H-Co
At4g11150EMB 244810.8 × G-Ap > H-Co
At3g54010PAS14.4 × G-Ap > H-Co
At5g06240EMB 27357.5 × G-Ba > H-Rt
At5g40600EMB 18754.1 × G-Ba > H-Rt

Not all genes function in a cell-autonomous manner (Nakajima et al., 2001). However, studies have shown that a correlation exists between the expression pattern of embryonic genes and the observed defects in gain- and loss-of-function mutants (Friml et al., 2002, 2003; Haecker et al., 2004; Hamann et al., 2002). Gain-of-function mutants in BDL show defects in the basal region of embryos (Hamann et al., 1999), pin4 mutants display abnormal cell divisions in the basal cells of globular- and heart-stage embryos (Friml et al., 2002), and wox2 mutants show defective divisions of cells in the apical regions of globular embryos (Haecker et al., 2004). As these genes and the examples of PID and PEP illustrate, expression data can potentially be useful either to predict or confirm the organs or regions most likely to be defective in gain- or loss-of-function mutants. We have therefore compared the expression data for each tissue and have been able to identify a large number of genes which can be classified as either spatially restricted within the embryo at a given stage or are temporally regulated.

Summarized in Table 4 is a comparison of the number of expressed genes (signal >40 in reference tissue) that differ by fourfold or greater in one tissue versus another. For example, of the 14 587 genes expressed in the globular-apical sample, 615 (4.2%) showed expression levels at least fourfold lower in the globular-basal sample. Table S4 gives examples of genes that are both spatially and temporally expressed. The data in Table 4 indicate that, of the genes expressed in restricted apical-basal patterns in the globular embryo, 5–8% are expressed at lower levels in the heart-stage embryo. Interestingly, of the genes expressed in apical and basal domains of the heart-stage embryo, a higher proportion (8–10%) were expressed at higher levels than at the globular stage. There is also a greater complexity of gene expression in the apical and basal domains of the heart-stage, versus globular-stage, embryo. Taken together, this may reflect the increased complexity at the cellular level as the embryo makes the transition from the globular to the heart stage.

Table 4.  The expression of genes (mean signal >40) in the reference tissue was compared with their expression in the other tissues sampled. Shown is the number of genes in the reference tissue which show at least a fourfold lower expression in the comparison tissue
Reference tissue (no. of genes expressed)Comparison tissue
  1. Numbers in parentheses represent the number of genes expressed as a percentage of the total in the reference. Abbreviations are as in Table 3.

Globular apical (14 587)615 (4.2)730 (5.0)993 (6.8)
Globular basal (15 165)752 (5.0)908 (6.0)1271 (8.2)
Heart cotyledon (15 461)1338 (8.7)1241 (8.0)1565 (10.1)
Heart root (12 870)1277 (9.9)1313 (10.2)1139 (8.9)

Expression patterns of putative regulatory genes

Amongst those that are predicted to be spatially restricted are genes encoding putative DNA or RNA binding proteins, kinases, cytoskeletal-associated proteins, potential hormone pathway proteins, protein–protein interactions, G-protein signalling components, transporter-like proteins, phosphatases and proteins involved in the ubiquitin-mediated degradation pathway. By way of illustration, the expression of approximately 1400 predicted transcription factors and receptor kinases (Shiu and Bleecker, 2003; were examined (Table S5). No particular family of transcription factor appears to be over-represented in the data, with genes from a number of different classes showing differential regulation (Table S6). For example, of the basic helix-loop-helix family, At4g16430 shows an approximately 228-fold higher expression in the apical versus basal region of the globular embryo, but by the heart stage has approximately fourfold higher expression level in the heart root region versus cotyledon. Of the BZIP class, At1g42990 is approximately 14-fold, and At4g02640 sixfold, more highly expressed in heart-stage cotyledon versus root. In the case of the AP2/EREBP family At5g17430 is 12-fold more abundant in heart-stage root versus cotyledon, whereas another family member, ANT (At4g37750) is 78-fold more abundant in the cotyledon versus root. In contrast, some classes of regulatory gene are not dramatically differentially regulated at the transcriptional level, either spatially or temporally between globular- and heart-stage. Included here are the ARR genes, which are generally expressed at low levels in early embryogenesis.

Auxin is an important regulator of embryogenesis and efflux carriers regulate its distribution such that by the globular stage onwards, auxin accumulates in the basal region of the embryo (Friml et al., 2003). However, analysis of the expression of the ARF and AUX/IAA family of transcription factors did not reflect this polar distribution of auxin. Of those genes that were expressed in the sampled tissues, the majority were expressed at a relatively low level in a non-polar manner (Table S5). ARF1 and BDL/IAA12 however were expressed more strongly in the basal regions of globular- and heart-stage embryos respectively (Table S6). In contrast, ARF3/ETTIN, which is required for correct flower and gynoecium development (Sessions et al., 1997) was expressed more strongly in the apical regions of globular- and heart-stage embryos. A possible explanation is that the expression of specific auxin-regulated gene family members is not necessarily restricted to the regions of auxin maximum in the embryo, and that individual family members may exhibit differential auxin sensitivity or respond to other signals.

Identification of novel mutants

To correlate gene expression patterns with functionality, we screened T-DNA insertion populations for putative knockouts of a number of these genes. For one gene (At5g16780), encoding a novel class of predicted transcription factor, the expression data show that mRNA abundance is below the level of detection in the globular-stage embryo, but is clearly detectable in the basal domain of the heart-stage embryo and at threshold levels in the cotyledonary region (Table 2, Figure 3). GUS analysis showed promoter activity of the gene in the embryonic root tip, and seedlings showed activity in the lateral root cap and meristem, and in the SAM, a tissue excluded in the laser capture procedure (Figure 3). Root cell type gene expression data (Birnbaum et al., 2003; Table S2) confirm At5g16780 expression in the youngest cells of the stele, endodermis-cortex, epidermis and lateral root cap, though no expression in older cells of those tissues. Plants heterozygous for a T-DNA insertional mutation in the eighth intron show segregation of an embryonic-defective phenotype in siliques, with pre-globular stage embryos exhibiting a defective pattern of cell division in the lower tier (Figure 5a). This indicates the requirement of very low levels of transcript of the gene at the globular stage, below the level detectable by this method. Heart-stage embryos are shorter and wider than wild type (Figure 5b), and seedlings develop very short roots with defective meristems, and have abnormal leaf development (Figure 5c–e), consistent with the observed embryonic and seedling gene expression. This example illustrates how the expression data can lead to the identification of previously uncharacterized genes with essential roles.

Figure 5.

Mutation of At5g16780 leads to embryo and seedling defects.
(a) A pre-globular stage embryo with abnormal cell division of the basal pole (arrow).
(b) A heart-stage embryo with flattened apical and basal axis.
(c) Ten-day-old seedlings. Wild-type segregant (left) and two mutant seedlings.
(d, e) Lugol staining of the root cap of 4-day-old seedlings. (d) Wild-type segregant with normal root cap, and (e) mutant with a defective root cap and root meristem.

Limitations to the technique and conclusions

Taken together, these data show that the developing embryo of Arabidopsis expresses a transcriptional profile that is regulated both spatially and temporally. However, there are clear limitations to the conclusions that can be drawn. Whilst LCM in combination with DNA microarray analysis allows the large-scale identification of strongly differentially expressed genes in specific tissue domains of early embryos, it is less useful for reliably characterizing at high resolution the expression of genes encoding very low abundance transcripts. The choice of the cut-off threshold for the Affymetrix signal is somewhat arbitrary, but by choosing a mean value of 40–75 we have related it to previously characterized gene expression patterns and to our own RT-PCR results. However, researchers should be cautious in the interpretation of such low levels of expression. This issue of defining basal levels of expression is not unique to the use of LCM or the study of embryogenesis, but is common to the application of DNA microarray analysis in other systems. Clearly, those interested in embryonic gene expression should rely on more than a single approach to construct a picture of molecular events in the developing embryo, and LCM represents a powerful tool to augment other techniques. It provides new opportunities to identify functionally important genes through a targeted reverse genetics approach and, in the longer term, provides a basis for the computational modelling of gene networks. These combined approaches will further illuminate genetic control mechanisms in the early development of higher plants.

Experimental procedures

The procedure for the transcriptome analysis of cell populations is divided into four phases: cryosectioning of embryos; LCM to isolate specific cell populations; RNA extraction and amplification; and cDNA microarray analysis.


Embryo sacs containing either globular- or heart-stage embryos were dissected from siliques and embedded in OCT embedding medium (RA Lamb, Eastbourne, UK) in a base mould (Merck Eurolab, Dorset, UK). Samples were then frozen in liquid nitrogen-cooled isopentane prior to sectioning. Sections of 6 μm were cut on a Leica CM3050S cryostat (Leica Microsystems, Nussloch, Germany) at −22°C. Sections were collected on RNAse-free uncoated glass slides and stored in 70% ethanol at −22°C ready for processing. Sections were processed through the following series: 30 sec in 70% ethanol, 15 sec in 0.01% phosphate-buffered saline pH 7.4, 30 sec in 70% ethanol, 30 sec in 95% ethanol, 30 sec in 100% ethanol, 2× 10 min in xylene. Slides were air dried and kept dry in boxes containing silica gel (Sigma, Poole, UK).

Laser capture microdissection

Laser capture microdissection was performed using a PixCell IITM system using CapSureTMHS LCM caps (Arcturus, CA, USA). Briefly, a slide was placed into position on the LCM microscope stage. An HS LCM cap was positioned over the section using the placement arm and the laser beam was set to 7.5 μm and focused. Cells of interest were then located and LCM performed with the laser beam set to either 7.5 or 15 μm. The power and duration of the laser beam was variable and dependent on beam size, but was typically 80 mW and 70 μsec for a 7.5-μm beam. Captured cells were then removed by rapidly lifting the LCM cap using the placement arm. Typically, between 10 and 15 sections were processed per LCM cap and non-specific material was removed from the surface of an LCM cap using a Post-It note. An ExtracSureTM Sample Extraction Device (Arcturus) was then attached to the LCM cap in preparation for RNA extraction.

RNA extraction and RNA amplification

RNA from LCM cells was extracted using the Absolutely RNATM Nanoprep kit (Stratagene, CA, USA) according to the manufacturer's instructions with slight modifications. Briefly, 100 μl of lysis buffer was applied onto the cap via the ExtracSureTM Sample Extraction Device that was then connected to a 0.5-ml microcentrifuge tube. After vortexing, the sample was incubated at 60°C for 5 min. The lysis buffer was then collected and mixed with an equal volume (approximately 100 μl) of 70% ethanol before applying to the RNA-binding column. Following DNase treatment and washes, RNA was eluted in 2 × 20 μl of DEPC-treated water, and was vacuum concentrated to a volume of 11 μl. cDNA synthesis and RNA amplification were performed using the MessageAmpTM aRNA kit [Ambion (Europe) Ltd, Huntingdon, UK] according to the manufacturer's instructions. Briefly, first round cDNA synthesis was primed with an T7 Oligo(dT) primer. One microlitre T7 oligo(dT) primer was added to 11 μl RNA, heated to 70°C for 10 min and to this, at 42°C, were added 2 μl 10x first strand buffer, 1 μl ribonuclease inhibitor, 4 μl dNTP mix and 1 μl reverse transcriptase. After a 2-h incubation, second strand synthesis was performed by the addition of 63 μl nuclease-free water, 10 μl 10x second strand buffer, 4 μl dNTP mix, 2 μl DNA polymerase and 1 μl RNase H. Reactions were incubated for 2 h at 16°C. cDNA was purified according to the manufacturer's instructions and vacuum concentrated to 16 μl. For in vitro transcription, 8 μl of the cDNA was mixed with 8 μl NTP mix, 2 μl 10x reaction buffer and 2 μl T7 enzyme mix. Reactions were incubated at 37°C for 16–24 h and then treated to DNase I digestion. aRNA was purified and vacuum concentrated to 10 μl.

Subsequent rounds of cDNA synthesis were performed as follows. To 10 μl of aRNA was added 2 μl random primers and the reaction was incubated at 70°C for 10 min. At 42°C the following were added: 2 μl 10x first strand buffer, 1 μl ribonuclease inhibitor, 4 μl dNTP mix and 1 μl reverse transcriptase and the reaction incubated for 2 h. One microlitre of RNase H was added and incubated for 30 min at 37°C. Second strand synthesis was primed with 5 μl T7 Oligo(dT) primer. Following a 10-min incubation at 70°C the remaining components were added at room temperature: of 58 μl nuclease-free water, 10 μl 10x second strand buffer, 4 μl dNTP mix, 2 μl DNA polymerase. The reaction was incubated at 16°C for 2 h. cDNA was then purified and in vitro transcription performed as described. Typically three rounds of RNA amplification were required to produce microgram quantities of aRNA.


RT-PCR was performed using the OneStep RT-PCR kit (Qiagen Ltd, Surrey, UK) according to the manufacturer's instructions. Typically, 10–50 ng of aRNA was used per reaction with conditions of 50°C for 30 min, 95°C for 15 min, and 40 cycles of 94°C for 30 sec, 55°C for 30 sec and 72°C for 1 min. The primer pairs used for RT-PCR were as follows: ACT1 forward GATCCTAACCGAGCGTGGTTAC, reverse GACCTGACTCGTCATACT-CTGC; AtPIN4 forward 5′-CATTGCTTGTG-GGAACTCTGTC-3′, reverse 5′-CTATTCCTTGAGGCAACGCA-GC-3′; ANT forward 5′-CAAGCACGGATTGGTAGAGTCG-3′, reverse 5′-CATTAGGGTTTGA-TGTCCAAGG-3′; CUC1 forward 5′-CATGCATGAGTATCGCCTTGACG-3′, reverse 5′-GGAGGAGGAGGAAGAACCGTGG-3′; MP forward 5′-GATGTCCAG-TCGCAGATCACATC-3′, reverse 5′-CTCATCTGCTGGACCTCAGTTGG-3′; SCR forward 5′-CGTGCGACTCACGGGACTTGG-3′, reverse 5′-CAGTAGAGTCGCTTGTGTAGC-3′; PID forward 5′-CTCTCTCCGTCATAGACAACC-3′, reverse 5′-GCATTACCATGTGATCCACCTG-3′; SHR forward 5′-CCTAGAGACGCTGTGA-TATCGAG-3′; WOL forward 5′-GTGCAGAGAGTGGTCAAGTTGC-3′, reverse 5′-CAGCAGAATCGGTCCACTCTTCC-3′.

cDNA microarray analysis

The Affymetrix GeneChip Arabidopsis ATH1 Genome Array (approximately 24 000 genes) was used for cDNA microarray analysis. The Arabidopsis Microarray and Bioinformatics service at GARNET performed probe labelling, hybridization and analysis (Craigon et al., 2004). One microgram of cDNA generated from aRNA after three rounds of amplification was provided as the template for probe preparation. For each tissue (globular apical and basal, heart root and cotyledon pole), three replicates were performed. Data analysis was performed using TIGR MultiExperiment Viewer version 2.2 (Saeed et al., 2003).

Materials and growth conditions

Arabidopsis thaliana ecotype Col-O was used exclusively for all experiments. For in vitro growth studies, seeds were vernalized and surface-sterilized (Clarke et al., 1992) and plated on growth medium (half-strength Murashige and Skoog medium (Sigma), 1% sucrose, and 2.5% phytagel (Sigma) at 22 ± 2°C at a photon flux density of approximately 150 μmol m−2 sec−1. Embryo sacs were obtained from siliques of 3–4 week-old plants grown in compost. T-DNA insertion lines were obtained from the Nottingham Arabidopsis Stock Centre (Nottingham, UK; NASC on-line catalogue,; Scholl et al., 2000).

Gene constructs and plant transformation

Promoter fragments were amplified using the Expand High Fidelity PCR system (Roche, Herts., UK). The promoters for At1g48630 and At2g21320 were cloned using the GATEWAY system (Invitrogen Ltd, Paisley, UK) by first cloning into pDONR207, followed by cloning into the binary vector pBI101-G (a gift from P. Hussey, Durham University). The At1g15750 promoter was transferred as a SalI–BamHI fragment into the binary vector pΔGUS-CIRCE. All other promoter fragments were amplified and ligated into T-tailed pΔGUS-CIRCE. All constructs were validated by sequencing. The primer pairs used for promoter amplification are as follows: At1g48630 forward 5′-GGGGACAAGTTTGTACAAAAAAGCAGG-CTGACGCCTGCTTCAGTCTGCGTTTCC-3′, reverse 5′-GGGGACCACTTTGTA-CAAGAAAGCTGGGTCGAGGGCAAAATGGAGGTAACGAG-3′; At2g21320 forward 5′-GGGGACAAGTTTGTACAAAAAAGCAGGCTCAGGATAATGGACA-CAGGGTGCTC-3′, reverse 5′-GGGGACCACTTTGTACAAGAAAGCTGGGTGG-TTGAAGCGGAAGATGGAGAGAC-3′; At1g15750 forward 5′-TCGTCGACCCAC-CATCTATCCCTCAATGCTC, reverse 5′-TCGGATCCCTAACCAAACTATCACC-AGATCTGC-3′; At2g02760 forward 5′-GAATCACCTGCTTTACGAGACC-3′, reverse 5′-CAAACCATTTGAATCTGCCAAGC-3′; At5g16780 forward 5′-CGAG-AAACCAGATCCTTCTTC-3′, reverse 5′-GAACTTTAGTCAAGACCTAACCAC-3′; At5g17430 forward 5′-CAATGACTCTGGCCGTGGCTG-3′, reverse 5′-CTA-CTCCTTGTGATAGATGAGAG-3′.

Plant transformation was performed by the floral dip method (Clough and Bent, 1998) using Agrobacterium tumefaciens strain C58C1 (Koncz and Schell, 1986). Tissue localization of β-glucuronidase (GUS) enzyme activity was performed as described (Topping and Lindsey, 1997).


Tissues were cleared and mounted for light microscopy in chloral hydrate (Topping and Lindsey, 1997) or 20% glycerol. For determining the prescence of starch seedlings were stained in Lugol's solution (Sigma) for 5 min. Photographs were taken using a CoolSNAP cf digital camera (Photometrics, Roper Scientific, Tucson, AZ, USA) with Openlab 3.1.1 software (Improvision Ltd, Coventry, UK) on Leica MZ125 [Leica Microsystems (UK) Ltd, Milton Keynes, UK], Olympus SZH10 (Olympus UK Ltd, Southall, UK), or Zeiss Axioskop (Carl Zeiss Ltd, Welwyn Garden City, UK) microscopes. Images were processed in Adobe Photoshop 5.0 (Mountain View, CA, USA).


This work was supported by funding to KL from the Biotechnology and Biological Sciences Research Council, which is gratefully acknowledged. MS was in receipt of a BBSRC CASE studentship in collaboration with Syngenta.

Supplementary Material

The following material is available from

Table S1 The number of genes expressed per replicate in the sampled embryo tissues above the signal thresholds indicated

Table S2 Expression data for selected genes identified as being expressed in the Arabidopsis root by Birnbaum et al. (2003)

Table S3 The expression of EMB genes (Tzafrir et al., 2004, and their supporting information) in embryonic tissue. For each gene, the expression in each tissue replicate and the signal mean is given as well as Affymetrix GeneChip Spot ID and Probe Names

Table S4 Examples of genes that show either spatial or temporal expression during globular and heart-stage embryogenesis. Mean signal values are given for the three tissue replicates. Each gene shows at least fourfold greater expression in the tissue indicated than the other embryo tissues

Table S5 Expression of putative transcription factors and receptor kinases. Transcription factors were taken from AtTFDB (, whilst a list of receptor kinases was published by Shiu and Bleecker (2003)

Table S6 Transcription factors and receptor kinases that show spatial or temporal differences in expression between the tissues sampled. Genes from Table S4 were analysed for differential expression in the embryonic tissues shown. Abbreviations are as in Table 3