Conservation of class C function of floral organ development during 300 million years of evolution from gymnosperms to angiosperms


For correspondence (fax +65 6779 5671; e-mail


Flower development in angiosperms is regulated by the family of MADS-box transcription factors. MADS-box genes have also been reported from gymnosperms, another major group of seed plants. AGAMOUS (AG) is the class C MADS-box floral organ identity gene controlling the stamen and carpel development in Arabidopsis. We report the characterization of an ortholog of the AG gene, named Cycas AGAMOUS (CyAG), from the primitive gymnosperm Cycas edentata. The expression pattern of CyAG in Cycas parallels that of AG in Arabidopsis. Additionally, the gene structure, including the number and location of the introns, is conserved in CyAG and other AG orthologs known. Most importantly, functional analysis shows that CyAG driven by the AG promoter can rescue the loss-of-function ag mutant of Arabidopsis. However, the ectopic expression of CyAG in ag mutant Arabidopsis cannot produce the carpeloid and stamenoid organs in the first and second whorls, although the stamen and carpel are rescued in the third and fourth whorls of the transformants. These observations show that the molecular mechanism of class C function controlling reproductive organ identity (stamen and carpel of angiosperms or microsporophyll and megasporophyll of gymnosperms) arose before the divergence of angiosperms and gymnosperms, and has been conserved during 300 million years of evolution thereafter.


In the ‘ABC’ model of flower development, three classes of transcriptional regulators act in combination to specify sepals (A only), petals (A and B), stamens (B and C), or carpels (C only) of Arabidopsis and Antirrhinum flowers (Coen and Meyerowitz, 1991). Molecular cloning revealed the original genes of classes A–C. In Arabidopsis, class A genes are AP1 and AP2, but no corresponding genes were found in Antirrhinum. However, the classes B and C genes have identifiable orthologs (or functional equivalents) in the two species. DEFICIENS (DEF)/GLOBOSA (GLO) and APETALA (AP3)/PISTILLATA (PI) are class B genes, and PLENA (PLE) and AGAMOUS (AG) are class C genes in Antirrhinum and Arabidopsis, respectively (Weigel and Meyerowitz, 1994). Recently, the SEPALLATA genes were identified as cofactors of the ABC genes involved in specifying petals, stamens, and carpels in Arabidopsis (Lohmann and Weigel, 2002; Pelaz et al., 2000). Based on these data, the ‘quartet model’ was proposed to explain the molecular interaction of floral organ identity genes (Theissen and Saedler, 2001).

Molecular studies showed that all these floral organ identity genes in Arabidopsis and Antirrhinum, except AP2, are MADS-box genes (Lohmann and Weigel, 2002). The transcription factors encoded by MADS genes share a highly conserved MADS-box region (approximately 60 amino acids) for DNA binding, dimerization and accessory-factor interactions (Shore and Sharrocks, 1995). Plant MADS-box genes have an additional conserved region, called K-box, characterized by regular spacing of hydrophobic residues suitable for protein–protein interactions (Ma, 1994). The region between the MADS- and K-boxes is weakly conserved (approximately 30 amino acids) linker domain (I or L domain). Downstream of the K-box, the C terminal of the protein is the most variable region in sequence and length, and the function of this region is unknown so far.

Although the molecular mechanisms of flower development in higher eudicots, such as Arabidopsis and Antirrhinum, have been illustrated extensively, the origin and evolution of the flowering plants remains an ‘abominable mystery’ as it was in the time of Charles Darwin, over 100 years ago. Based on the recent molecular data, it seems likely that the origin and evolution of flowers of angiosperms (or reproductive organs of gymnosperms) may be reflected by the origin and evolution of MADS-box floral organ identity genes. The observations from maize and rice strongly support the conservation of the ABC model (at least the B and C functions) in monocots (Ambrose et al., 2000; Ma and dePamphilis, 2000). In some basal eudicots, however, orthologs of the B class genes do not show a uniform expression pattern in the petals as predicted by the ABC model, which suggests a diversity of the B function in petal identity (Kramer and Irish, 1999). Orthologs of class B and C genes have also been reported from gymnosperms, another major group of seed plants (Rutledge et al., 1998; Tandre et al., 1995, 1998; Winter et al., 1999). However, class B and C genes have not been found in ferns and other more basal plants after extensive search despite the identification of general MADS-box genes in such species (Theissen et al., 2000). It is not entirely unlikely that the orthologs of class B and C genes had not evolved before the seed plants emerged about 300 million years ago. These data also imply that the establishment of molecular mechanisms (including B and C functions) for floral (or reproductive) organ development could be a prerequisite for the origin of the flower (or reproductive organs). These mechanisms could have occurred after the common ancestors of the present-day seed plants (gymnosperms and angiosperms) separated from the other land plants, but before the separation of gymnosperms and angiosperms.

Previous studies showed that the heterologous ectopic expression of deficiens-agamous-like 2 (DAL2) and spruce-agamous 1b (SAG1b) (AG orthologs from conifers) in wild-type Arabidopsis produced a phenotype similar to that produced by the ectopic expression of AG gene (Rutledge et al., 1998; Tandre et al., 1998). However, the possibility that these phenotype changes could be because of the activation of the endogenous AG gene by DAL2 or SAG1b, rather than the direct effects of DAL2 or SAG1b in the heterologous plants, cannot be ruled out (Winter et al., 2002). Complementation experiments of gnetum gnemon MADS2 (GGM2) (class B genes pistillata (PI)/AP3 ortholog from gnetophyte) indicated that it could only partially rescue the class B function in class B mutant Arabidopsis (Winter et al., 2002). When GGM2 was expressed ectopically in the class B mutant, GGM2 could complement the stamen identity function of class B genes, but not the petal identity function. Therefore, whether the B and C functions are conserved between gymnosperms and angiosperms still remains inconclusive.

Molecular evidence suggests that gymnosperms may be monophyletic. Compared with other gymnosperms, the most characteristic and unique feature of Cycas is the apparently primitive nature of the reproductive organs (Figure 1a–c). The megasporophylls of Cycas do not form a determinate female cone, although the megasporophylls are somewhat tightly clustered together when they first emerge (Figure 1b). Female Cycas plants produce successive and repeated zones of vegetative leaves, cataphylls, and megasporophylls. Each megasporophyll has a petiole, expanded at the distal end into a terminal blade. The blade is entire serrated or sometimes deeply lobed. The serrations or lobes are suggested to be the equivalent of the leaflets of foliage leaves. The ovules are attached to both sides of the petiole. The short and thickened microsporophylls of Cycas (male) are organized spirally in compact cones (or strobilus) as in other cycads (Figure 1c). The microsporophylls are much less leaflike than the megasporophylls. Even so, they have a resemblance to the megasporophylls in having a middle fertile portion and a terminal ‘blade’ region. It is generally accepted that in its female reproductive structures, Cycas has essentially retained the seed fern (extinct) condition where the ovules were borne on the foliage (Mamay, 1969). If this interpretation is correct, the genus is a valuable ‘missing link’, showing an intermediate stage in the evolution of female cones.

Figure 1.

Morphology of C. edentata.

(a) Two C. edentata plants.

(b) A female pseudocone. The inset shows close-up of a megasporophyll with ovules along the margins.

(c) A male cone. The inset shows a microsporophyll with numerous microsporangia.

Scale bar: 0.5 m for (a) and 10 cm for (b,c). Scale bar for insets: 2 cm for megasporophyll and 1 cm for microsporophyll.

To investigate the molecular mechanisms controlling the reproductive organ development in Cycas edentata, we cloned CycasAGAMOUS (CyAG), an ortholog of AG. The intron–exon organizations are also conserved between CyAG and other AG group members. CyAG is expressed only in the male and female reproductive organs, and not in the vegetative tissues of Cycas, suggesting the involvement of CyAG in reproductive organ development. Functional analysis of CyAG by introducing it into ag loss-of-function mutant of Arabidopsis demonstrated that CyAG could complement the AG function in Arabidopsis. Our study reveals that at least part of the molecular and genetic mechanisms controlling flower and cone development existed before the divergence of angiosperms and gymnosperms, and has been conserved during the nearly 300 million years of evolution. Based on our data and results from other plant species, a model for the evolution of the classes A–C functions of flower development is proposed.


Molecular cloning of CyAG gene from C. edentata

To clone the MADS-box genes expressed in the ovules of C. edentata, we performed PCR with a degenerate primer (MADS1). The degenerate primer MADS1 was designed based on the most conserved 10 amino acids (KRRNGLLKKA) within the MADS-box region of all known plant MADS-box proteins. Reverse-transcribed first-strand cDNA from ovules with oligo(dT) primer was used as a template for PCR amplification. As the predicted length of the MADS-box proteins is about 200 amino acids, the PCR products above 600 bp were cloned and examined. Sequence analysis showed that an 800-bp PCR product was the 3′ portion of a MADS-box gene. A homology search showed that its closest homologs were DAL2 and SAG1 from conifers and GGM3 from gnetophyte, which were known homologs of AG from Arabidopsis (data not shown). The new gene was named CyAG.

After screening an ovule-derived cDNA library of Cycas with the 3′-specific sequence of incomplete CyAG cDNA, three clones were obtained. DNA sequence analysis showed that these clones encoded the same open-reading frame (ORF), 5′ and 3′ untranslated regions (UTRs). However, the lengths of the 5′ UTR of these clones varied. To clone the full-length 5′ UTR of CyAG cDNA, 5′ Rapid Amplification of cDNA Ends (RACE) was performed with internal gene-specific primers. The overlapping regions of the 5′ UTR of the longest 5′ RACE product and that of the three cDNA library-derived CyAG clones matched exactly (data not shown). The full-length CyAG cDNA was 1164 bp in length and contained a 675-bp coding region from position 279 to position 953 including the TGA stop codon. The 5′ and 3′ UTRs of CyAG cDNA were 278 and 212 bp in length, respectively. The 3′ UTR possesses a putative polyadenylation signal (TATAA), 49 nucleotides upstream of the polyA tail (Supplementary Material).

Alignment of CyAG with DAL2 and AG revealed three highly conserved regions (MADS, I, and K domains), which are involved in DNA–protein and protein–protein interactions, and divergent N- and C-terminals in the protein (Figure 2a). A phylogenetic tree based on the amino acid sequences of the MIK domains of various MADS-box proteins revealed that the AG group forms a distinct clade (Figure 2b; Supplementary Material Figure S1). In the AG clade, the gymnosperm orthologs, namely, CyAG, GBM5, DAL2, SAG1b, and GGM3, separated from their angiosperm counterparts into a distinct clade. In this phylogenetic analysis, DAL3 and GGM1 were used as the outgroups.

Figure 2.

Characterization of CyAG gene.

(a) Alignment of the predicted amino acid sequences of CyAG with DAL2 and AG. Identical or similar residues are shaded black or gray. Dashes indicate gaps. Groups of amino acids considered to be similar were L, I, V, M (hydrophobic), D, E (acidic), N, Q (amide), F, Y, W (aromatic), H, K, R (hydrophilic, basic), and P, A, G, S, T (small, neutral, or weakly hydrophobic).

(b) Phylogeny of plant MADS-box proteins (MIK region) in the AG, DEF, and SQUA subfamilies. A simplified phylogenetic tree is shown here, with SQUA, DEF and AG representing gene clades, not individual genes. The complete tree with details of the sequences used for these groups is shown in the Figure S2.

Genomic DNA blot analysis with a CyAG-specific probe indicated that only one copy of the CyAG gene exists in the Cycas genome (Figure 3a). Based on the known gene structures of MADS-box genes, the possible intron positions of AG group MADS-box genes were determined. According to the CyAG cDNA sequence, primers were designed to clone the introns flanking the possible intron positions (Supplementary Material Figure S2). As predicted by the cDNA sequence, the first intron of CyAG gene (326 bp) was obtained by PCR with primers AG1F and AG1R. PCR with AG3F and AG3R produced introns 3 (737 bp) and 4 (103 bp). Finally, the genomic DNA fragment including introns 5 (161 bp), 6 (170 bp), 7 (190 bp), and 8 (122 bp) of CyAG gene were amplified by AG4F and AG4R. The overlapping regions of these amplified fragments matched exactly, and the exon sequences of these amplified fragments also matched the cDNA sequence of CyAG (100%), indicating that they were truly the genomic sequence of CyAG. However, attempts to clone the intron 2 of CyAG gene by genomic PCR with AG2F and AG2R failed, probably because the intron 2 might be very long (approximately 3–5 kbp in other AG orthologs). Therefore, genome walking was performed. With primer AG2F or AG2R, partial sequence of intron 2 was obtained. Although the full-length intron 2 was not obtained so far, the sequence that we know has confirmed the existence of intron 2 and its position in CyAG gene. CyAG gene contains eight introns, all of which conform to the AG–GT rule governing the intron–exon junctions. The positions of these introns in relation to the deduced amino acid sequence of CyAG and other known AG orthologs are shown in Figure 3(b). When we compare AG orthologs from angiosperms, the first intron of AG is in the predicted translation region, while all the others are located in the 5′ UTR. This is because the N-terminal domain of AG is longer than that in the other orthologs (Riechmann et al., 1999; Yanofsky et al., 1990). In contrast to the AG orthologs from angiosperms, gymnosperm AG orthologs, SAG1b, GBM5, and CyAG do not encode any additional amino acids upstream of the MADS-box (N-terminal region), but the positions of all the introns are essentially identical in all the seven genes compared (Figure 3b). However, if we compare the gene organizations of AG orthologs with those of other MADS-box genes, such as AP3/PI and AP1 group genes, the similarity is less (Goto and Meyerowitz, 1994; Jack et al., 1992; Mandel et al., 1992). This observation further supports the phylogenetic association of CyAG with the other class C genes from gymnosperms and angiosperms.

Figure 3.

CyAG gene organization.

(a) Genomic DNA blot analysis of CyAG. Genomic DNA was digested by HindIII, EcoRI, PstI, or XbaI, as indicated above the lanes. The size of the DNA markers is also indicated in kbp.

(b) Comparison of the gene structures of AG group genes available in the database. N/M/I/K/C refer to the N-terminal, MADS, I, K, and C-terminal domains, respectively, described in the text. The positions of the introns are indicated by arrows, and the intron sizes in bp are given below. *The full length of the 2nd intron of CyAG gene has not yet been determined.

Expression pattern of CyAG in different tissues of C. edentata

RNA blot analysis revealed that the 1.2-kbp CyAG transcripts are detectable specifically in the entire reproductive organs of male and female plants, including the megasporophyll, ovule, central axis of the male cone, microsporophyll, and microsporangium of Cycas (Figure 4). Its expression was not detectable in the vegetative tissues, such as root, stem, and leaf, suggesting that CyAG is involved only in the reproductive organ development of Cycas.

Figure 4.

RNA blot analysis of CyAG of C. edentata.

Total RNA samples were prepared from different tissues shown above each lane.

The spatial expression patterns of the CyAG gene in the different tissues of microsporophyll and ovule were determined by in situ hybridization. Both vegetative plant parts, such as the roots and leaves, as well as reproductive tissues, were used for in situ hybridization. Similar to the RNA blot analysis, the CyAG transcripts were detectable in the whole ovule and microsporophylls of C. edentata (Figure 5). In the ovule, its expression level was relatively focused in the inner layer of integuments and nucellus, rather than the outer layer of integuments (Figure 5a,b). In microsporophylls, CyAG expression was focused in the pollen grains and the tapetal layer, while the signals were weakly detectable in the sterile part of microsporophylls (Figure 5d,e). Control sections, which were probed with a sense CyAG probe, showed no signal, indicating that the post-hybridization washes were sufficiently stringent to eradicate non-specific signals (Figure 5c,f). However, the CyAG transcripts were undetectable in the roots and leaves of C. edentata (data not shown). As all of the tissue sections were placed on the same slide, the result ensured that the reproductive organ-specific expression of CyAG gene in C. edentata is a true reflection of the in vivo situation and not because of inconsistent processing conditions.

Figure 5.

In situ localization of CyAG transcripts in reproductive organs of C. edentata.

Longitudinal sections of ovule (a–c) and cross-sections of microsporophylls (d-f) and microsporangia were hybridized with the antisense probe (a,b,d,e) or sense probe (c,f).

(a, b) CyAG expression detected in the ovule.

(c) No signal is detectable with the sense probe.

(d, e) CyAG expression detected in microsporophylls and microsporangia.

(f) No signal is obtained with the sense probe. Arrows indicate regions of hybridization signals.

in, integuments; nu, nucellus; fg, female gametophyte; mp, microsporophyll; mg, microsporangia. Scale bar, 0.3 mm for (e,f); 0.7 mm for (b,c); 1 mm for (c,d).

Functional analysis of CyAG in ag mutant of Arabidopsis

To investigate whether the structural similarity between CyAG and angiosperm C class genes extends to functional conservation, we attempted to rescue the loss-of-function ag mutant (ag-2) of Arabidopsis by CyAG. In one construct, CyAG cDNA was under the control of the constitutive cauliflower mosaic virus 35S promoter (CaMV35S::CyAG) to achieve the ectopic expression of CyAG in transgenic lines. In the other construct, CyAG was placed downstream of the intragenic regulatory region of AG from Arabidopsis (Busch et al., 1999; Deyholos and Sieburth, 2000; Sieburth and Meyerowitz, 1997) fused with the minimal 35S promoter (AGenhancer::Δ35S::CyAG). Earlier studies had confirmed that the intragenic regulatory region of AG and minimal 35S promoter drives the expression of the downstream gene as the normal AG expression pattern. Therefore, the expression pattern of CyAG in transgenic Arabidopsis plants recovered should be the same as the AG expression in wild-type Arabidopsis.

Heterozygotes of ag-2 mutants of Arabidopsis that carried a large T-DNA insertion (Yanofsky et al., 1990) were identified by genomic PCR and used for the floral dip transformation. The genotypes of the T1 generation transgenic plants harboring the CyAG gene with homozygous ag-2 background were determined again by genomic PCR.

A total of 15 transformants (ag-2 homozygotes) were generated, of which seven lines were with the CaMV35S::CyAG and eight were with AGenhancer::Δ35S::CyAG. Compared with the wild-type (Figure 6a) and ag-2 mutant (Figure 6b), six of the eight transgenic lines containing AGenhancer::Δ35S::CyAG produced floral phenotypes nearly identical to the wild-type Arabidopsis (Figure 6c,d). The petals and sepals in the third and fourth whorls of the transgenic flowers were converted to normal stamens and carpel, like in a wild-type flower. The plants could set seed by selfing, and the number of floral organs was comparable to that of the wild type, except that the stamens in the complemented flowers ranged from four (Figure 6c) to six (Figure 6d). Five of the seven transgenic lines harboring CaMV35S::CyAG showed phenotypes almost identical to those in wild-type Arabidopsis (Figure 6e), and one showed a very weak overexpression of AG phenotype, where the flowers have stamens and carpels in the third and fourth whorls, but slightly shorter and narrower petals than those in the wild type (Figure 6f). The existence of CyAG gene in the genome of all transgenic lines was confirmed by genomic DNA blot analysis, but CyAG expression was detectable only in the lines showing the rescued phenotype by RT-PCR (data not shown). As ag-2 is a null mutant of AG that carries a 35-kbp T-DNA insertion within the gene (Yanofsky et al., 1990), our results indicate that the complementation of AG function is by the expression of CyAG in the transgenic Arabidopsis.

Figure 6.

Phenotypes of CyAG expression in Arabidopsis.

(a) Wild-type Arabidopsis flower.

(b) ag-2 mutant flower.

(c) Flowers of ag-2 harboring AGenhancer::Δ35S::CyAG showing that the stamens and carpel are rescued (four stamens).

(d) Flowers of ag-2 harboring AGenhancer::Δ35S::CyAG showing that the mutant is fully rescued (six stamens).

(e) Flowers of ag-2 harboring CaMV35S::CyAG showing phenotype similar to the wild type.

(f) Flower of ag-2 harboring CaMV35S::CyAG showing phenotype similar to weak overexpression of AG.

se, sepal; pe, petal; st, stamen; ca, carpel. Scale bar, 1 mm for (a–f).


CyAG is the ortholog of AG in the primitive gymnosperm C. edentata

The sequence and phylogenetic analyses revealed that CyAG is a member of AG group MADS-box genes. In other words, CyAG is the ortholog of AG in C. edentata. The amino acid sequence of CyAG shared high similarity to other AG orthologs. CyAG and AG had nearly 100% identity within the MADS-box. In the I and K domains, they still had 57 and 62% identity, respectively. Among the conserved residues in the I domain are those that constitute the TIERYKK motif found in class C proteins (Tandre et al., 1998). Furthermore, the K domain of AG shows significantly higher similarity to that of CyAG (62% identity) than to the K domains of other classes of MADS-box proteins in Arabidopsis (e.g. 24% identity between AG and AP3). Four AG group MADS-box genes from gymnosperms shared the highest sequence similarity. In the MADS-box, they had 100% identity; in the I and K domains, they had 80% identity; and even in the C domain, they had 60% identity. These results suggested that AG group MADS-box genes are highly conserved in sequence despite the evolutionary distance between gymnosperms and angiosperms. Although many MIKC-type MADS-box genes have been isolated from basal plants, such as ferns and mosses, no homolog of the ABC function genes has been found in non-seed plants so far (Hasebe et al., 1998; Munster et al., 1997; Theissen et al., 2000). As there is no evidence that orthologs of floral homeotic genes exist in these non-seed plants, the presence of floral homeotic genes in seed plants might be correlated to the emergence of seed or flowering plant-specific structures, such as ovules, carpels, stamens, or floral perianth organs. Our results show that the AG group MADS-box genes might be one of the earliest ABC class genes emerging at least 300 million years ago and that they would have regulated the reproductive organ development in the common ancestor of gymnosperms and angiosperms. Phylogenetic analysis gave us a clearer picture of the relationship of AG group MADS-box genes from gymnosperms to angiosperms. The phylogenetic tree also supports the view that gymnosperms are monophyletic, and hence gymnosperm lineage should have had a common ancestor after separation from the common ancestor of angiosperms and gymnosperms.

Conserved gene organizations of AG orthologs from angiosperms and gymnosperms

An important reference for identifying the evolutionary relationships of the homologous genes is the intron and exon structures of genes. The conservation of intron structure between the homologous genes provides further evidence of their orthology (Doyle, 1994). The genomic DNA blot hybridization showed that CyAG is a single-copy gene in the C. edentata genome. This result made it possible to clone the CyAG gene by genomic PCR and genome walking. In our study, all the introns of CyAG gene were found by genomic PCR and genome walking method. Gene structures of AG group genes have been reported for seven other members, which are AG (Yanofsky et al., 1990), plena (PLE) (Bradley et al., 1993), farinelli (FAR) (Davies et al., 1999), populus trichocarpa AGAMOUS (PTAG)1/PTAG2 (Brunner et al., 2000), Ginkgo biloba MADS5 (GBM5) (Jager et al., 2003), and SAG1b (Rutledge et al., 1998). Except GBM5 (where only six introns were cloned), all of these AG orthologs have eight introns, and the intron positions are conserved (Figure 3b). However, when we compared the gene structure of AG group genes with those of other MADS-box genes, significant differences in the intron number and position were observed. These data strongly support the orthology of CyAG and AG group genes from other plant species.

Expression pattern of CyAG provides further support to the functional equivalency of CyAG and AG

Although phylogenetic analysis and gene structure indicated that CyAG is the ortholog of AG, it did not necessarily prove their functional equivalency because their functions might change during speciation (Theissen, 2002). However, the specific expression of CyAG in the male cone and female pseudocone not only showed that CyAG was involved in the reproductive organ development of Cycas, but also supported that CyAG had similar function with other AG orthologs. Despite the extensive differences in reproductive organs between Cycas and angiosperms, RNA blot and in situ hybridization revealed that CyAG expression had parallels with that of AG orthologs. For example, in Arabidopsis, the AG gene is expressed in the third and fourth whorls (stamens and carpels) of the hermaphroditic flowers (Yanofsky et al., 1990). Stamens represent the male reproductive organ (equivalent of microsporophylls), and carpels represent the female reproductive organ (equivalent of megasporophylls) in Arabidopsis. AG is not expressed in the petals and sepals of Arabidopsis– organs corresponding to which are absent in gymnosperms. CyAG transcripts were also found to be restricted to megasporophylls and ovules of the female and the central axis, microsporophylls, and microsporangia of male cone. So far, four other AG group MADS-box genes are reported from gymnosperms: DAL2 and SAG1b from the conifers Picea abies and P. mariana, respectively, GGM3 from Gnetum, and GBM5 from Ginkgo. The expression of DAL2 was detected in ovuliferous scales, but not in the bracts, the cone axis, or the apical meristem (Tandre et al., 1998). The expression of SAG1b also was found to be very similar in female cones (Rutledge et al., 1998). In the male cones of conifers, SAG1b expression was detected at a low level in the tapetal layer, but not in either the developing pollen or central vascular region. The structural differences between conifer cones and Cycas cones make it difficult to fully compare these expression data directly. Nonetheless, all the AG orthologs reported from gymnosperms, with the exception of GBM5 (which is also expressed in the vegetative tissues), were expressed only in reproductive organs of male and female plants, and not in vegetative tissues. These data suggest that all AG orthologs of gymnosperms are involved in reproductive structure formation and that their function might be conserved during the evolution from gymnosperms to angiosperms.

Full complementation of Arabidopsis class C null mutants by CyAG

In our study, the most convincing evidence of functional conservation between CyAG and AG genes came from the complementation analysis, which showed the complete rescue of AG function in the loss-of-function ag mutant by the CyAG gene. Unfortunately, gymnosperms are not amenable to the standard techniques of functional genomics because they are woody plants with large, uncharacterized genomes and lengthy generation periods. Hence, we transformed an Arabidopsis ag mutant (ag-2) either with CyAG under the control of the constitutive CaMV35S promoter (CaMV35S::CyAG) or with the intragenic regulatory region of AG from Arabidopsis fused to the minimal 35S promoter (AGenhancer::Δ35S::CyAG). Our data showed that CyAG gene was able to rescue the AG function of stamen and carpel identities in Arabidopsis. However, ectopic expression of CyAG in the first and second whorls was not sufficient to convert the petals and sepals to stamens and carpels, respectively. This result is not consistent with the phenotype of ectopic expression of AG in Arabidopsis (Mizukami and Ma, 1992), and the phenotype of ectopic expression of DAL2 and SAG1b (conifer AG orthologs) in wild-type Arabidopsis. The phenotypic differences of CyAG and AG ectopic expression lines might be because of the possibility of evolution of novel downstream mechanisms independent of AG context in angiosperms that are required for the development of petals and sepals. However, heterologous ectopic expression of DAL2 and SAG1b (AG orthologs from conifers) in wild-type Arabidopsis produced a phenotype similar to the ectopic expression of AG gene (Rutledge et al., 1998; Tandre et al., 1998). It has been pointed out that data from complementation of null mutants should be more reliable than data from ectopic expression in wild-type background (Theissen et al., 2000). Therefore, in the earlier studies the possibility that ectopic expression phenotypes could arise because of the activation of the endogenous AG gene by DAL2 or SAG1b, rather than the direct effects of DAL2 or SAG1b in wild-type Arabidopsis cannot be ruled out. Our results are generally consistent with the partial complementation of GGM2 (PI/AP3 ortholog from gnetophyte) in class B mutant Arabidopsis, which showed that GGM2 could only rescue stamen identity, but not the petal identity of the class B function (Winter et al., 2002).

Evolution of the ABC model of flower development in plants

Functional complementation of ag-2 by CyAG in Arabidopsis allows us to derive two important conclusions. First, it demonstrates that the C class floral organ identity gene emerged before the divergence of gymnosperms and angiosperms (at least 300 million years ago) and that its function has been conserved in all seed plants. The regulatory contexts of AG action for stamen and carpel identity, including protein–DNA and protein–protein interactions are conserved in gymnosperms and angiosperms, and probably the target gene recognition corresponding to this important transcription factor might also be conserved in the divergent species separated by over 300 million years. Second, our data, together with results from conifers, gnetophyte, and basal eudicots, also suggest that in the common ancestor of angiosperms and gymnosperms, only B and C functions existed. In the basal angiosperms, the A function was added first, while the other functions remained unchanged. In the lineage leading to monocots and dicots, the A–C functions co-evolved further. Meanwhile, novel interactions of class B proteins with class A proteins and novel specificities in DNA binding and in the regulation of target genes developed, leading to the production of two types of flower perianth organs, viz., petals and sepals. But these novel mechanisms in sepals and petals might not be identified or affected by the primitive class B and C genes, such as CyAG and GGM2, which explains the lack of strong ectopic expression phenotypes in the heterologous Arabidopsis system as reported.

Experimental procedures

Plant materials and growth conditions

Reproductive and vegetative tissues were collected from female and male C. edentata D.J. de Laubenfels (=C. rumphii Miq.) plants growing in the Singapore Botanic Gardens, Garden of the Department of Biological Sciences, National University of Singapore, Singapore and from the natural populations along the seashore of Desaru, Johor, West Malaysia. The selfed heterozygous seeds of ag-2 mutants and transgenic plants of Arabidopsis were grown under greenhouse conditions (22°C, 14 h of light/10 h of dark).

Molecular cloning and characterization of CyAG gene

Total RNA was isolated from microsporophylls and ovules of C. edentata by the modified method of Murray and Thompson (1980), and mRNA was isolated using Oligotex mRNA Mini kit (Qiagen, Germany). Reverse transcription, followed by PCR amplification, was performed using oligo(dT) linker primer 5′-GAGAGAGAGAGAGAGAGAGAACTAGTCTCGAG(T)18-3′ and MADS-box degenerate primer MADS1–5′-AARMGIMGIAAYGGIYTIYTIAARAARGC-3′. In the primer, R represents A/G, M represents C/A, I represents inosine, which can bind A/T/C/G, and Y represents T/C. PCR was carried out with the following steps: 93°C for 3 min for enzyme activation, 30 cycles of 15-sec denaturing at 95°C, 30-sec annealing at 42°C, and 1-min extension at 72°C, with a final 10-min extension at 72°C. PCR products above 600 bp were cloned, and the inserts were sequenced.

Partial gene-specific sequence of CyAG gene (approximately 580 bp) was used as the probe to screen a Cycas young female megasporophyll and ovule-derived cDNA library in λZAPII vector (Stratagene, USA). From the unamplified cDNA library, approximately 50000 plaques were plated and transferred to replica nylon membranes and then screened. A nearly full-length (28 nucleotides missing from the 5′ end) cDNA of CyAG was isolated. The missing 5′ UTR was amplified by 5′ RACE with 5′/3′ RACE Kit (Roche Diagnostics, Germany). Internal CyAG-specific primers for 5′ RACE were 5′-ACTCGGCTGAGGCCTCTTTC-3′ and 5′-GCTTGGTCCTGCTGCGTGTA-3′).

DNA and RNA gel blot analyses and in situ hybridization

Genomic DNA sample from C. edentata plants was extracted as described previously by Dellaporta et al. (1983), with some modifications. Total RNA samples from root, stem, leaf, male cone, and female pseudocone were isolated as described above. To each lane was loaded 20 μg of total RNA or 40 µg of digested total genomic DNA. RNA and genomic DNA blot analysis was performed following Sambrook et al. (1989). The 383-bp-long 3′ fragment (of which 69 bp is from 3′ UTR) served as the probe. In situ hybridization was carried out as described by Sakai et al. (1995) and Zhang et al. (2002) using DIG-labeled antisense and sense (control) RNA probes. The probes were transcribed from the same fragment used in blot studies under the T7 or SP6 promoter in vector pGEM-T Easy (SP6/T7 transcription kit; Roche Diagnostics, Germany).

Phylogenetic analyses

Multiple sequence alignment was performed using the clustalw and boxshade program. Phylogenetic trees were constructed based on a set of MADS-box protein sequences comprising the published AG, SQUAMOSA (SQUA), and DEF subclade sequences from gymnosperms and angiosperms and our CyAG sequence. Trees were generated from ‘170-amino-acid domain’ MIK domain sequences, which contained the 60 aa of MADS-box and the subsequent 110 aa of I and K domains. Phylogenetic relationships among these genes were inferred by parsimony analysis. Parsimony analysis was performed using the paup* 4.0b10 software. Consensus trees and estimates of statistical confidence were inferred from 100 bootstrapped (randomly sampled with replacement) data sets. Previously published plant MADS-box gene sequences were retrieved from the GenBank database: AP1, S27109; CAULIFLOWER, AAC67506; SQUA, CAA45228; ZAP1, T03410; DAL1, T09603; PrMADS3, T09603; GGM9, CAB44455; PrMADS1, T09569; PLE, P23706; NTDEF, CAA65288; AP3, AAD51890; LRDEF, AAM27456; PMADS2, Q07474; SLM2, CAA56656; NTGLO, CAA48142; PRGLO, AAF28863; PTAG1, ACC06237; PTAG2, ACC06238; CUM1, ACC08528; NAG1, Q43585; pMADS3, Q40855; TAG1, Q40168; FAR, CAB42988; FBP6, S60307; PLE, CAB42988; SLM1, X80408; BAG1, AAA32985; AG, P17839; OsMADS3, S59480; ZAG1, JQ2289; ZMM1, CAA57073; ZAG2, X80206; AGL1, P29381; AGL5, P29385; SAG1b, ACC97157; DAL2, T14847; GGM3, CAB44449; GBM5, AAM76208; GGM1, CAB44447; DAL3, T14848.

Genomic PCR and genome walking

Using genomic DNA as template, genomic PCR was performed to investigate the gene organization of CyAG gene. Gene-specific primers were designed based on the CyAG cDNA sequence (see Supplementary Material). Among these primers, AG1F and AG1R were used to clone the possible introns in the 5′ UTR of CyAG gene, AG2F and AG2R were for the putative second intron of CyAG gene, and AG3F and AG3R were for the possible introns in the I domain and K domain. AG4F and AG4R were used for cloning the possible introns in the 3′ end of CyAG gene. The partial sequence of the second intron of CyAG gene was obtained by genome walking using Universal Genome Walker kit (Clontech, USA).

Construction of the binary vector and plant transformation

All cloning techniques were standard (Sambrook et al., 1989). The full-length ORF of CyAG was generated by PCR using a 5′ primer on the ATG containing a BamHI site and a reverse primer designed downstream of the stop codon with a SacI site. The primer sequences were 5′-GA(GGATCC)GCGGATAAAGTTC-3′ and 5′-GA(GAGCTC)TTTACTGGCATG-3′. This was cloned into pCAMBIA1302 vector (Clontech, USA) downstream of the CaMV35S promoter (CaMV35S::CyAG) or into pCAMBIA1300 downstream of the 3-kbp AG intragenic regulatory region (AGenhancer) and minimal 35S promoter (Δ35S) yielding AGenhancer::Δ35S::CyAG replacing the GUS gene. The AGenhancer region was placed into the HindIII site of pCAMBIA1300. Both constructs were sequenced to confirm the absence of PCR errors and mobilized into Agrobacterium tumefaciens GV3101 and used to transform Arabidopsis. The heterozygotes of ag-2 mutant plants were used for transformation by floral dip method (Clough and Bent, 1998). Heterozygotic seedlings of ag null mutant ag-2 of Arabidopsis were identified by genomic PCR with the primers: TLFP (T-DNA-specific) 5′-GATGCACTCGAAATCAGCCAATTTTAGAC-3′, AGX1 (AG-specific) 5′-TTCTCATTTGGTCAATACCC-3′ and AGX2 (AG-specific) 5′-TACCTCTCAATAGTCCCTTT-3′. Clorox-sterilized seeds were plated on MS medium with 50 µg ml−1 kanamycin, 25 µg ml−1 hygromycin, and 250 µg ml−1 carbenicillin (selection medium). The segregated ag-2 homozygotes carrying CyAG (transgenic lines) were also identified by PCR with primers described above.


We thank Drs Elliot Meyerowitz for providing ag mutant seeds (ag-2), Detlef Weigel for the clone containing the regulatory region of AG gene, and Prakash Lakshmanan and Hao Yu for critically reading the manuscript. We also thank the Singapore Botanic Gardens for making available some of the C. edentata materials. This work was supported by the research grants RP154-000-096-112 and RP154-000-073-112 and a PhD scholarship (to P.Z.) from the National University of Singapore.

Supplementary Material

The following material is available from

Figure S1. The full-length cDNA (upper row) and amino acid (lower row) sequences of CyAG gene.

For the amino acid sequence of CyAG, the MADS-box is shown in bold letters, I domain is shown by italic letters, K domain is boxed, and the rest is the C-terminal. For cDNA sequence, the translation stop codon is indicated by an asterisk, and the polyA signal (TATAA) is underlined. The primers (AG1F/AG1R, AG2F/AG2R, AG3F/AG3R, and AG4F/AG4R) designed for cloning the introns of CyAG gene are underlined with arrows indicating their directions. The primers AG4F/AG4R were used to synthesize the 3′CyAG-specific probe for genomic DNA and RNA blot analyses.

Figure S2. Phylogenetic tree of plant MADS-box proteins (MIK region) in the AG, DEF, and SQUA subfamilies.

A single most parsimonious tree was obtained. Bootstrap values for 100 replicates that are above 50% are shown at the branches.

Accession numbers of genomic sequences reported in this paper: AF492455 (CyAG cDNA sequence), AY295079 (genomic sequence of CyAG).