Identification of genes expressed during earlyArabidopsiscarpel development by mRNA differential display: characterisation ofATCEL2, a novel endo-1,4-β-D-glucanase gene


*For correspondence (fax +64 9373 7416; e-mail


The floral homeotic geneAGAMOUS (AG) imparts carpel identity on the fourth whorl of floral organs in wild-typeArabidopsisflowers. Less is known about the genes that regulate carpel patterning and differentiation. To identify candidate regulators, we screened for genes expressed in developing carpels. SinceArabidopsiscarpels are difficult to isolate, we used whole inflorescence apices of two floral homeotic mutants (pi and pi ag) and mRNA differential display, to identify carpel transcripts. Two of the resulting cDNA clones were shown to be expressed predominantly in flowers. They encodedAGL11,a MADS box transcription factor known to be expressed in the carpel and ovules, and a novelArabidopsisendo-1,4-β-D-glucanase (ATCEL2).In situhybridisation localised theATCEL2transcript to the developing septum and ovule primordia of young carpels.


Carpels are complex floral organs essential for seed and fruit development in angiosperms. In Arabidopsis thaliana L. Heynh (Arabidopsis), two congenitally fused carpels differentiate into a cylindrical ovary, enclosing the ovules, topped by a short style and stigma (reviewed by Gasser & Robinson-Beers 1993). Collectively these tissues are known as the gynoecium or pistil. The carpels occupy the centre of the Arabidopsis flower and comprise the fourth whorl of floral organs ( Fig. 1).

Figure 1.

Photographs of flowers of Arabidopsis wild-type (left), pi (middle) and pi ag (right).

The diagrams below show the domains of gene activity in the four floral whorls predicted by the ABC model of flower development ( Coen & Meyerowitz 1991). In the model, PI is a B class gene and AG is a C class gene. Se: sepals; Pe: petals; St: stamens; Ca: carpels.

The best characterised regulator of carpel development is the floral homeotic gene AGAMOUS (AG) which specifies carpel identity in the fourth whorl ( Fig. 1;Coen & Meyerowitz 1991). AG and other floral homeotic genes encode MADS box transcriptional regulators expressed in the primordia of the floral whorl whose identity they control (reviewed by Yanofsky 1995).

Much less is known about the genes that translate the positional information encoded by AG into the specialised tissues of the Arabidopsis carpel. However, genetic studies have identified a few genes, including CRABS CLAW and SPATULA, that are needed for aspects of Arabidopsis carpel development ( Alvarez & Smyth 1994a; 1994b; 1997). Several other genes, such as ETTIN, have more extensive pleiotropic effects, influencing both carpel and floral development ( Sessions & Zambryski 1995;Sessions et al. 1997 ).

To identify more regulators of carpel patterning and differentiation, we screened for genes expressed predominantly in developing Arabidopsis carpels. A previous differential cDNA library screen in tomato identified seven genes expressed in young pistils. Most of them encoded enzymes likely to be involved in pistil growth, maintenance or defence ( Gasser et al. 1989 ;Milligan & Gasser 1995). The screen used isolated tomato pistils as a source of mRNA. Similar screens are difficult in Arabidopsis as the developing carpels are small and inaccessible.

To facilitate identification of Arabidopsis carpel transcripts, we used whole inflorescence apices of two floral homeotic mutants and mRNA differential display ( Liang & Pardee 1992). pistillata (pi) flowers have only two kinds of floral organs, carpels and sepals ( Fig. 1;Bowman et al. 1989 ;Hill & Lord 1989). Flowers of the double mutant pi ag consist only of repeated whorls of sepals ( Fig. 1;Bowman et al. 1991 ). Hence, we expected that transcripts present in pi, but missing in pi ag inflorescence apices, would be those expressed in the developing carpel.

Here, we describe the selection of pi-expressed cDNA clones by differential display and report the molecular characterisation and spatial localisation of one carpel-predominant clone, ATCEL2, in more detail. ATCEL2 encodes a novel Arabidopsis endo-1,4-β- d-glucanase which is predominantly expressed in the developing septum and ovule primordia of young carpels.


pi cDNA clones identified by mRNA differential display

Development of the Arabidopsis flower up to anthesis has been divided into 12 stages ( Smyth et al. 1990 ). The Arabidopsis gynoecium develops as a slotted tube, comprised of two congenitally fused carpels, from the centre of stage 7 flowers. A false septum forms dividing the cylinder into two locules in stage 9. Four rows of ovule primordia develop from placental tissue at the intersection of the septum and carpel valves during stage 9. The top of the cylinder starts to close and ovule primordia elongate in stage 10. During stages 11–12 stigmatic papillae appear, the style becomes distinct from the ovary, ovule tissues differentiate, and megasporogenesis occurs ( Hill & Lord 1989;Smyth et al. 1990 ).

To focus on the earliest stages of carpel development, we extracted RNA from inflorescence apices carrying young flowers (≤ stage 9) of the floral homeotic mutants, pi and pi ag. Transcripts in the two RNA populations were compared by mRNA differential display. Bands present at higher levels in the pi lanes were cloned and sequenced. Thirty-five different clones were obtained. Database searches indicated that half of the pi clones (18 clones) were identical or similar to genes encoding known proteins. Four of these clones were predicted to encode regulatory proteins and included the Arabidopsis MADS box gene, AGL11, known to be expressed in Arabidopsis ovules and seeds ( Rounsley et al. 1995 ). AGL11 transcript was detected from the earliest visible stage of Arabidopsis ovule primordia development (stage 9) until late in seed development by in situ hybridisation analysis ( Rounsley et al. 1995 ). Most of the other known proteins were enzymes, including three detected previously in pistils of other plants (endo-1,4-β- d-glucanase, Brummell et al. 1997 ;Milligan & Gasser 1995; chitinase, Lotan et al. 1989 ;Budelier et al. 1990 ; lipoxygenase,Rodriguez & Beltran 1995).

The expression of the pi clones was re-tested in mutant tissues by Northern or RT-PCR analysis. Of 15 pi clones detected, five were differentially expressed between pi and pi ag (shown for ATCEL2, Fig. 2). Subsequently, the expression of these clones in different wild-type tissues was analysed. Two of the clones, AGL11 and ATCEL2, were shown to be expressed predominantly in flowers. ATCEL2 (Arabidopsis thaliana CELLULASE 2) encoded part of a novel Arabidopsis endo-1,4-β- d-glucanase gene named after cel1, the first reported Arabidopsis endo-1,4-β- d-glucanase gene ( Shani et al. 1997 ). We studied ATCEL2 further as AGL11 had already been analysed in previous work.

Figure 2.

Northern hybridisation analysis of pi clone ATCEL2 to RNA from pi and pi ag inflorescences and flowers.

Lanes contain approximately 20 μg total RNA from flowers at the stages shown. Whole inflorescences carrying young flowers were the source of ≤ stage 9 RNA. Flower stages are approximations based on wild-type flowers ( Smyth et al. 1990 ). Hybridisation with an rRNA probe to the same blot is shown in the panel below.

Localisation of ATCEL2 expression in flowers by in situ hybridisation

We analysed the temporal and spatial pattern of ATCEL2 expression in flowers in more detail by in situ hybridisation ( Fig. 3). In wild type, ATCEL2 transcript was first detected along the inner carpel walls in late stage 7 flowers ( Fig. 3a). ATCEL2 signal was localised to the medial ridge which gives rise to the septum and placental tissues ( Sessions 1997). ATCEL2 signals concentrated in the developing septum and ovule primordia during stages 8–9 ( Fig. 3b,c). By stage 10, the signal reduced in intensity in the septum and ovules ( Fig. 3d,e). Distinct ATCEL2 signals were not seen above background levels beyond about stage 10.

Figure 3.

In situ hybridisation of ATCEL2 to wild-type and pi inflorescences.

Wild-type tissue (a–e);pi tissue (f–h). (a) Longitudinal section of wild-type inflorescence apex with late stage 7 flower. (b) Cross-section of stage 8 wild-type flower. (c) Cross-section of stage 9 wild-type flower. (d) Cross-section of stage 10 wild-type flower. (e) Longitudinal section of stage 9 wild-type flower. (f) Cross-section of approximately stage 9 pi flower. (g) Longitudinal section of pi inflorescence apex with flower bud at ~ stage 9. (h) Longitudinal section of pi inflorescence with arrowhead at pedicel/stem junction. Se: sepal; Pe: petal, St: stamen, Ca: carpel, Ov: ovule, Sep: septum; Ped: pedicel.

Localisation of ATCEL2 transcript to wild-type carpels is consistent with the high levels of expression detected by Northern analysis in pi inflorescence tissue and young flowers ( Fig. 2). However, it does not explain detection of ATCEL2 transcript in pi ag inflorescences ( Fig. 2). To address this question, we carried out in situ hybridisation analysis to localise ATCEL2 expression in inflorescences and flowers of pi, pi ag and ag. ATCEL2 transcript was detected in the developing septum and ovules in pi carpels ( Fig. 3f,g), but not in the floral organs of pi ag and ag (data not shown). ATCEL2 transcript was also detected at the junction of the floral pedicels and the inflorescence stem in all three of the mutants (shown for pi,Fig. 3h), but not in wild type (data not shown). This observation explains ATCEL2 detection in pi ag inflorescences by Northern hybridisation. However, it is not clear why ATCEL2 is expressed at the pedicel/stem junction in the three homeotic mutants, but not in wild type. One factor that the mutants have in common is that they are infertile and this may somehow be responsible for altering ATCEL2 expression patterns in this region (see Hensel et al. 1994 ).

Analysis of the sequence of a full length ATCEL2 cDNA

A full length ATCEL2 cDNA clone was isolated from an Arabidopsis cDNA flower library ( Weigel et al. 1992 ). The ATCEL2 cDNA sequence is 1748 bases in length with a predicted 1503 base open reading frame (ORF), 78 bases of 5′ untranslated region, and 167 bases of 3′ untranslated region. The predicted ORF encodes a 501 aa protein with a molecular mass of approximately 55 kDa ( Fig. 4). The most similar protein to ATCEL2 was the tomato endo-1, 4-β- d-glucanase, TPP18, which was 78% identical ( Fig. 4;Milligan & Gasser 1995). Both proteins encode motifs typical of the plant endo-1,4-β- d-glucanases ( Fig. 4).

Figure 4.

Alignment of the predicted protein sequences of ATCEL2 and TPP18.

Shaded boxes show identical residues. The proteins have two conserved domains (dashed lines) found in all plant endo-1,4-β- d-glucanases (reviewed by Brummell et al. 1994 ). They also have two endo-1,4-β- d-glucanase motifs (underlined) identified by PROSITE. These contain residues (white lettering) needed for the hydrolysis of β-1–4-glucan linkages in polysaccharides by bacterial endo-1,4-β- d-glucanases. Most plant endo-1,4-β- d-glucanases are secreted enzymes and putative signal peptides (lower case) were identified using SignalP V1.1 ( Nielsen et al. 1997 ).

TPP18 is one of seven characterised tomato endo-1,4-β- d-glucanase genes (reviewed by Rose et al. 1997 ). Our database searches identified at least 10 different Arabidopsis endo-1,4-β- d-glucanase genes (R. Schaffer and J. Putterill, unpublished results), two of which have been reported previously (CEL1, Shani et al. 1997 ;KORRIGAN, Nicol et al. 1998 ).


We isolated ATCEL2, a novel Arabidopsis endo-1, 4-β- d-glucanase, and the likely ortholog of TPP18, a gene expressed predominantly in young tomato pistils ( Brummell et al. 1997 ;Milligan & Gasser 1995). We extended the analysis of these genes in flowers by localising ATCEL2 transcript using in situ hybridisation to the developing septum and ovule primordia during the early stages of flower development (from stage 7). Expression decreased in the septum by the end of stage 9. By then, the two sides of the septum had fused, dividing the ovary into two locules. This postgenital fusion event occurs by merging of the epidermal layers of the opposing septal outgrowths (reviewed by Verbeke 1992) and may be one of the signals that down-regulates ATCEL2 expression.

What role might ATCEL2 play in the developing Arabidopsis carpel? In plants, endo-1,4-β- d-glucanases have tissue-specific patterns of expression associated either with tissue expansion, fruit ripening or organ abscission (reviewed by Brummell et al. 1994 ). During tissue expansion, endo-1,4-β- d-glucanases are thought to hydrolyse the 1,4-β-glucan backbone of xyloglucans in the cell wall thereby facilitating wall extension ( Hayashi 1989; reviewed by Cosgrove 1997). Recent genetic evidence also suggests a role for endo-1,4-β- d-glucanases in cell elongation in Arabidopsis ( Nicol et al. 1998 ). The Arabidopsis KORRIGAN gene is predicted to encode a plasma membrane-bound endo-1,4-β- d-glucanase. korrigan mutants are dwarfed and have defects in cell elongation and wall assembly ( Nicol et al. 1998 ). The expression pattern of ATCEL2 in developing carpels suggests that ATCEL2 may be involved in wall extension during growth and division of cells in the septum and ovules. However, analysis of ATCEL2 mutant plants is needed to clarify the function of the gene. To date, we have screened the Feldmann T-DNA lines by PCR, but have not identified any plants with insertions in ATCEL2 ( Feldmann 1991;McKinney et al. 1995 ; R. Schaffer and J. Putterill, unpublished results).

Finally, how successful was our use of a ′molecular-genetic trick’, the floral homeotic mutants, combined with mRNA differential display to identify carpel regulators? Our method yielded at least four predicted regulators, while previous differential cDNA library hybridisation screens in tomato identified none. However, many clones we isolated did not turn out to be differentially expressed. In future, analysing microarrays of all Arabidopsis genes with RNAs from mutants such as pi and pi ag, promises to be a more direct way to identify most carpel-expressed genes ( Schena et al. 1996 ).

Experimental procedures

Plant material

Wild-type Arabidopsis (Landsberg erecta) was obtained from Lehle Seeds (Round Rock, Texas, USA). Segregating populations of pi-1(pi), and pi-1 ag-3 (pi ag) mutants in Landsberg erecta were a gift from John Bowman (University of California, Davis, California, USA). Plants were grown in the greenhouse under natural light supplemented in winter with mercury vapour lamps.

mRNA differential display, cloning and sequencing of cDNA fragments

Total RNA was extracted from pi and pi ag inflorescence apices ( Stiekema et al. 1988 ) after removing the older flowers by hand. Sectioning of the pi apices indicated that they carried floral buds ≤ stage 9 ( Smyth et al. 1990 ). Differential display was performed as described by Oh et al. (1995) except that [33P]dATP (ICN Pharmaceuticals Inc., Costa Mesa, California, USA) was used in the place of [32P]. Bands present at higher levels in the pi lanes than the pi ag lanes were excised, re-amplified by PCR with the appropriate primers, cloned into the pGEM-T vector (Promega Corporation, Madison, Wisconsin, USA) and sequenced. All nucleotide sequence was derived using fluorescently labelled primers using an Applied Biosystems Catalyst 800 robotic work station and a 373-A automated sequencer (Applied Biosystems/Perkin Elmer, Foster City, California, USA). Thirty-five different cDNA clones were obtained (pi clones). Their DNA sequence was compared to sequences in GenBank and dBEST using BLAST ( Altschul et al. 1990 ). Seventeen of the clones did not overlap with Arabidopsis DNA sequences in the databases and were submitted to dBEST (R. Schaffer and J Putterill, unpublished results).

ATCEL2 probe

The ATCEL2 probe used in Northern and in situ hybridisation analyses and in cDNA library screening is a cloned 219 bp cDNA fragment corresponding to bases 573–791 of the ATCEL2 nucleotide sequence. It was amplified in mRNA differential display in a PCR reaction containing T11 (GG) and OPAA 02 (GAGACCAGAC) primers.

In situ hybridisation analysis

In situ hybridisation analysis on 10 μm sections was performed using digoxigenin-labeled ATCEL2 RNA probes (Boehringer Mannheim Corporation, Indianapolis, Indiana, USA) as described by Coen et al. (1990) . Sense probes were made using T7 polymerase and plasmid template cleaved with NotI, and antisense probes were made using Sp6 polymerase and plasmid template cleaved with ApaI. The hybridisation signal was detected by NBT/BCIP (Boehringer Mannheim) and serial sections photographed using Nomarski differential interference contrast microscopy.


We thank Yuval Eshed and John Bowman for stimulating discussions and commenting on the manuscript, J.B. for giving us segregating populations of pi and pi ag plants, the Arabidopsis Biological Resource Centre for supplying the CD4–6 cDNA library, and Stuart Baum for photography of wild-type and mutant flowers. The research was supported by NZFRST (contract number CO6403).