Expression of floral MADS-box genes in basal angiosperms: implications for the evolution of floral regulators

Authors

  • Sangtae Kim,

    Corresponding author
    1. Department of Botany, University of Florida, Gainesville, FL 32611, USA,
      (fax +1 352 846 2154; e-mail sangtae@botany.ufl.edu; dsoltis@botany.ufl.edu).
    Search for more papers by this author
  • Jin Koh,

    1. Department of Botany, University of Florida, Gainesville, FL 32611, USA,
    Search for more papers by this author
  • Mi-Jeong Yoo,

    1. Department of Botany, University of Florida, Gainesville, FL 32611, USA,
    Search for more papers by this author
  • Hongzhi Kong,

    1. Department of Biology, the Institute of Molecular Evolutionary Genetics, and the Huck Institutes of the Life Sciences, the Pennsylvania State University, University Park, PA 16802, USA,
    2. Laboratory of Systematic and Evolutionary Botany, Institute of Botany, the Chinese Academy of Sciences, Beijing 100093, China, and
    Search for more papers by this author
  • Yi Hu,

    1. Department of Biology, the Institute of Molecular Evolutionary Genetics, and the Huck Institutes of the Life Sciences, the Pennsylvania State University, University Park, PA 16802, USA,
    Search for more papers by this author
  • Hong Ma,

    1. Department of Biology, the Institute of Molecular Evolutionary Genetics, and the Huck Institutes of the Life Sciences, the Pennsylvania State University, University Park, PA 16802, USA,
    Search for more papers by this author
  • Pamela S. Soltis,

    1. Florida Museum of Natural History, University of Florida, Gainesville, FL 32611, USA
    Search for more papers by this author
  • Douglas E. Soltis

    Corresponding author
    1. Department of Botany, University of Florida, Gainesville, FL 32611, USA,
      (fax +1 352 846 2154; e-mail sangtae@botany.ufl.edu; dsoltis@botany.ufl.edu).
    Search for more papers by this author

(fax +1 352 846 2154; e-mail sangtae@botany.ufl.edu; dsoltis@botany.ufl.edu).

Summary

The ABC model of floral organ identity is based on studies of Arabidopsis and Antirrhinum, both of which are highly derived eudicots. Most of the genes required for the ABC functions in Arabidopsis and Antirrhinum are members of the MADS-box gene family, and their orthologs are present in all major angiosperm lineages. Although the eudicots comprise 75% of all angiosperms, most of the diversity in arrangement and number of floral parts is actually found among basal angiosperm lineages, for which little is known about the genes that control floral development. To investigate the conservation and divergence of expression patterns of floral MADS-box genes in basal angiosperms relative to eudicot model systems, we isolated several floral MADS-box genes and examined their expression patterns in representative species, including Amborella (Amborellaceae), Nuphar (Nymphaeaceae) and Illicium (Austrobaileyales), the successive sister groups to all other extant angiosperms, plus Magnolia and Asimina, members of the large magnoliid clade. Our results from multiple methods (relative-quantitative RT-PCR, real-time PCR and RNA in situ hybridization) revealed that expression patterns of floral MADS-box genes in basal angiosperms are broader than those of their counterparts in eudicots and monocots. In particular, (i) AP1 homologs are generally expressed in all floral organs and leaves, (ii) AP3/PI homologs are generally expressed in all floral organs and (iii) AG homologs are expressed in stamens and carpels of most basal angiosperms, in agreement with the expectations of the ABC model; however, an AG homolog is also expressed in the tepals of Illicium. The broader range of strong expression of AP3/PI homologs is inferred to be the ancestral pattern for all angiosperms and is also consistent with the gradual morphological intergradations often observed between adjacent floral organs in basal angiosperms.

Introduction

One of the most important developments in our understanding of floral development was the formulation of the ABC model for controlling floral organ identity (Coen and Meyerowitz, 1991). This model is based on genetic studies in Arabidopsis (Brassicaceae) and Antirrhinum (Plantaginaceae, formerly placed in Scrophulariaceae; see APGII, 2003) and posits that the specification of floral organ identity is controlled by three genetically separate functions (Figure 1a). The A function specifies sepal identity (whorl 1). The A and B functions together direct petal identity (whorl 2). The combination of both B and C functions determines stamen identity (whorl 3). Finally, the C function alone controls the identity of carpels (whorl 4). A-function genes include the Arabidopsis apetala1 (ap1) and apetala2 (ap2) genes (Bowman et al., 1993; Mandel et al., 1992). The B function requires the deficiens (def) and globosa (glo) genes in Antirrhinum, and their respective orthologs in Arabidopsis, apetala3 (ap3) and pistillata (pi) (Goto and Meyerowitz, 1994; Jack et al., 1992, 1994; Schwarz-Sommer et al., 1992; Sommer et al., 1990; Tröbner et al., 1992). The Arabidopsis agamous (ag) and Antirrhinum PLENA (PLE) genes are required for the C function (Bradley et al., 1993; Yanofsky et al., 1990). With the exception of ap2, all of these organ identity genes are members of the MADS-box family (reviewed in Ma and dePamphilis, 2000) and are collectively referred to as floral MADS-box genes.

Figure 1.

(a) The classic ABC model (Coen and Meyerowitz, 1991) for floral organ identity in Arabidopsis is shown as gray boxes. Based on recent additions to the ABC model (Colombo et al., 1995; Ditta et al., 2004; Pelaz et al., 2000; Theissen, 2001), D- and E-class genes are shown as white boxes. SEP4 (AGL3), a recently recognized E-function gene, is included in the box of E-class genes (Ditta et al., 2004).
(b) Modified ABC model (van Tunen et al., 1993): the boundary of B-class gene function is extended to the first whorl to explain the petaloid perianth (‘shifting boundary’ of Bowman, 1997, and ‘sliding boundary’ of Kramer et al., 2003).

The identification of floral MADS-box genes as essential regulators of early flower development has led to a huge effort at isolation and molecular analysis of additional members of the MADS-box gene family, including the Arabidopsis AGL genes (Ma et al., 1991; Mandel and Yanofsky, 1998; Rounsley et al., 1995). Since the original ABC model was presented, MADS-box genes that specify ovule identity were proposed to define the D function (Colombo et al., 1995) (Figure 1a). More recently, the Arabidopsis MADS-box genes AGL2, AGL4, AGL9 and AGL3 were found to have redundant function in specifying the identity of petals, stamens and carpels and were renamed sepallata1, sepallata2, sepallata3 and SEPALLATA4 (SEP1, −2, −3 and −4); the SEP genes were proposed to define the E function (Ditta et al., 2004; Pelaz et al., 2000; Theissen, 2001) (Figure 1a). In addition to model plants such as Arabidopsis and Antirrhinum, MADS-box genes have also been isolated from other core eudicots (sensuAPGII, 2003), including Petunia (Solanaceae), tobacco (Nicotiana, Solanaceae) and Gerbera (Asteraceae) (Angenent et al., 1993; Davies et al., 1996; Kater et al., 1998; Kempin et al., 1993; van der Krol et al., 1993; Yu et al., 1999). Floral MADS-box genes have also been characterized from several basal eudicot families, including Papaveraceae and Ranunculaceae (Kramer and Irish, 2000; Kramer et al., 1998, 2003, 2004), as well as from a diverse array of monocots (Ambrose et al., 2000; Kang et al., 1995; Kyozuka et al., 2000; Mena et al., 1995; Nagasawa et al., 2003; Schmidt et al., 1993) and basal angiosperms (Kim et al., 2004, 2005; Kramer and Irish, 1999, 2000; Kramer et al., 2003, 2004; Litt and Irish, 2003; Zahn et al., 2005a).

Extensive molecular phylogenetic analyses indicate that floral MADS-box genes from model organisms and their homologs form several well-supported major clades, which can be recognized as separate subfamilies (Becker and Theissen, 2003; Nam et al., 2003; Parenicova et al., 2003; Theissen et al., 1996). In fact, functionally similar genes in Arabidopsis and Antirrhinum are generally homologs that belong to the same subfamily. For example, the C-function genes AG and PLE are functional homologs (although not orthologs) in the AG subfamily (named after the first described member of the subfamily, as proposed by Becker and Theissen, 2003). Similarly, the B-function genes, DEF and GLO from Antirrhinum and AP3 and PI from Arabidopsis, are members of the DEF/GLO (or AP3/PI) subfamily, which can be further divided into the DEF (or AP3) and GLO (or PI) lineages. The Arabidopsis A-function gene AP1 is a putative ortholog of the Antirrhinum SQUA gene; SQUA plays a less prominent role in the A function than AP1 (Huijser et al., 1992). Phylogenetic analyses of the SQUA subfamily (A-class) identified two gene clades within the core eudicots, the euAP1 clade (which includes Arabidopsis AP1 and Antirrhinum SQUA) and the euFUL clade [which includes Arabidopsis FRUITFULL (FUL)] (Litt and Irish, 2003). The FUL gene is important for normal fruit development (Gu et al., 1998) and has a redundant role with AP1 and CAULIFLOWER in regulating meristem identity (Ferrandiz et al., 2000). In angiosperms outside the core eudicots, SQUA subfamily members are more similar to FUL than to AP1 and are often referred to as FUL-like (Litt and Irish, 2003). The SEP genes and their close relatives also form a separate subfamily, whereas the MADS-box genes required for the D function are members of the AG subfamily (Becker and Theissen, 2003; Kramer et al., 2004; Zahn et al., 2005a). Other MADS-box genes, such as the Arabidopsis AGL6 gene (Ma and dePamphilis, 2000; Ma et al., 1991), which defines another subfamily closely related to the AGL2 (SEP) subfamily (Becker and Theissen, 2003; Zahn et al., 2005a), are expressed in the flower and may play a role in flower development.

Functional studies in several eudicots and grasses indicate that homologs of AP3/PI and AG often exhibit conserved B and C functions, respectively (Ma and dePamphilis, 2000). In addition, relatively recent gene duplications in some groups of species have resulted in sets of paralogs that carry out a subset of the functions performed by their homologs in Arabidopsis and Antirrhinum (reviewed in Baum, 1998; Ma and dePamphilis, 2000; Soltis et al., 2002). For example, two AG homologs in maize have undergone subfunctionalization (Mena et al., 1995, 1996). Similarly, multiple paralogs of AP3 and PI are present in petunia, and they seem to fulfill collectively the functions of AP3 and PI (van der Krol and Chun, 1993; van der Krol et al., 1993; Tsuchimoto et al., 2000; Vandenbussche et al., 2004; Zahn et al., 2005b). These gene duplication and putative subfunctionalization events suggest that, while the functions of individual members of the DEF/GLO and AG subfamilies may vary from species to species, members of each subfamily collectively have a conserved function in regulating floral organ identity as proposed in the ABC model (Zahn et al., 2005b).

Further evidence for conservation and diversification of MADS-box gene expression (and by inference, function) has been found in non-grass monocots. In the monocot Asparagus (Asparagaceae), which exhibits only slight morphological differentiation between the outer and inner perianth whorls, AP3/PI homolog expression follows the classic ABC model (Park et al., 2003, 2004). On the other hand, in the monocot Tulipa (Liliaceae), Kanno et al. (2003) demonstrated that the organs of both floral whorls 1 and 2, which are morphologically similar, express both A- and B-class genes. These two whorls therefore have the same ‘petaloid’ identity. To explain the morphology of the lily flower, van Tunen et al. (1993) proposed a modified ABC model (Figure 1b) in which the expression of B-class genes was extended to the first floral whorl. The expression of AP3/PI homologs in both the first and second whorls in Tulipa (Kanno et al., 2003) supported this modified ABC model and is consistent with B-class gene expression in the petaloid perianth of Ranunculus (Ranunculaceae; Kramer et al., 2003). The ‘shifting boundary’ (Bowman, 1997) and ‘sliding boundary’ (Kramer et al., 2003) models allow the boundary of the B function to ‘slide’ from that observed for Arabidopsis and Antirrhinum to include the outer perianth whorl (outer tepals) of Ranunculus, Tulipa and other species with an entirely petaloid perianth.

Molecular genetic studies in Arabidopsis and Antirrhinum, as well as several other species, indicate that the function of floral MADS-box genes is very well correlated with the expression patterns of these genes, particularly when expression levels are high (Ma and dePamphilis, 2000). In Arabidopsis, AP1 and AG are expressed in the perianth and reproductive regions of the floral meristem, respectively, corresponding to the A and C functions. Although the DEF/GLO and AP3/PI genes are initially expressed somewhat broadly, they become restricted to the second and third whorls, as predicted by the B function (Goto and Meyerowitz, 1994; Jack et al., 1992; Schwarz-Sommer et al., 1992; Tröbner et al., 1992). Also, the petunia DEF/GLO homologs show differential expression between petals and stamens in a manner consistent with their functions (Angenent et al., 1992; Immink et al., 2003). Conversely, members of the same subfamily that have diverged in function also show distinct expression patterns. For example, the SHP1 and SHP2 genes are members of the AG subfamily, but have more specialized functions in carpel and ovule development (Liljegren et al., 1999; Pinyopich et al., 2003). They are expressed in developing carpels and ovules, but not stamens (Flanagan et al., 1996; Savidge et al., 1995). Similarly, the PLE paralog FAR in Antirrhinum is strongly expressed in the anther and is required for male reproductive development (Davies et al., 1999). Therefore, within a given subfamily of floral MADS-box genes, expression patterns can be very good predictors of gene function.

Although eudicots comprise approximately 75% of all angiosperm species, the organization of the flower is fairly constant throughout this clade, with floral organs typically arranged in distinct whorls and floral parts in fours, fives or multiples thereof. In contrast, non-monocot basal angiosperms represent only 3% of angiosperm species diversity (Drinnan et al., 1994) but display enormous floral diversity, with some taxa exhibiting an undifferentiated perianth of spirally arranged tepals [e.g. Amborella (Amborellaceae); Figure 2a], others having a well-differentiated perianth of distinct sepals and petals [e.g. Asimina (Annonaceae) and Saruma (Aristolochiaceae); Figure 2d], and still others that apparently lack a perianth altogether [e.g. Eupomatia (Eupomatiaceae); Figure 2f]. Reconstructions of perianth evolution indicate, in fact, that a differentiated perianth of sepals and petals evolved independently in several basal angiosperm lineages, as well as in eudicots (Albert et al., 1998; Ronse De Craene et al., 2003; Zanis et al., 2003).

Figure 2.

Photographs of flowers of basal angiosperms investigated in this study.
(a) Amborella trichopoda.
(b) Nuphar advena (photo credit: V. Remay).
(c) Illicium floridanum.
(d) Asimina longifolia.
(e) Magnolia grandiflora (photo credit: D. Callaway).
(f) Eupomatia bennettii (photo credit: H. Teppner).
Arrows indicate sepals of A. longifolia.

Despite the diversity of floral form and structure in basal angiosperms, information on the expression of floral MADS-box genes in these plants is limited. For example, AP3/PI homologs are expressed throughout the perianth in a species of Magnolia (Magnoliaceae) and in Calycanthus (Calycanthaceae) (Kramer and Irish, 2000), both members of the large magnoliid clade of basal angiosperms (Figure 3d; APGII, 2003; D. Soltis et al., 2000). Likewise, AP1 and AP3/PI homologs are expressed more broadly than expected from studies of Arabidopsis in both Magnolia and Eupomatia (Eupomatiaceae), a close relative of Magnolia that lacks a perianth (Figure 2; Kim et al., 2005). An AGL2 (SEP1) homolog from Nuphar (Nymphaeaceae, water lilies) is expressed in all floral organs (Zahn et al., 2005a).

Figure 3.

Summary of angiosperm phylogeny.
Examples of some clades are indicated after clade names. Names in bold indicate genera included in this study. Nearly all analyses of basal angiosperms have identified Amborella as the sister to all other extant angiosperms (e.g. Borsch et al., 2003; Graham and Olmstead, 2000; Graham et al., 2000; Hilu et al., 2003; Magallón and Sanderson, 2001; Mathews and Donoghue, 1999, 2000; Nickerson and Drouin, 2004; Parkinson et al., 1999; Qiu et al., 1999; P. Soltis et al., 1999; D. Soltis et al., 2000; Zanis et al., 2002). In all of these studies Nymphaeaceae and Austrobaileyales followed Amborella as successive sisters to the remaining extant angiosperms, as shown here. An alternative topology in which Amborella and Nymphaeaceae are sister to each other, with this clade sister to all other extant angiosperms, has been found in some analyses (e.g. Barkman et al., 2000; Kim et al., 2004; Mathews and Donoghue, 2000; Parkinson et al., 1999; Qiu et al., 2000; P. Soltis et al., 2000).

The basal-most lineages of extant flowering plants, and successive sisters to all other angiosperms, are Amborellaceae, Nymphaeaceae (the water lilies in the broad sense, including Cabombaceae) and Austrobaileyales (e.g. Mathews and Donoghue, 1999; Qiu et al., 1999; Soltis et al., 1999; reviewed in Soltis and Soltis, 2004; see APGII, 2003) (Figure 3). This basal grade is followed by (i) Chloranthaceae, (ii) monocots, (iii) a large magnoliid clade, which comprises four orders (Magnoliales, Laurales, Canellales and Piperales) and includes a number of well-known basal angiosperms, such as Magnolia, Persea (avocado), Piper (black pepper) and Asimina (paw-paw) and (iv) eudicots (Figure 3). Although each of these four clades is well supported, relationships among them are unclear.

To investigate the conservation and divergence of expression patterns of floral MADS-box genes, we isolated genes from several basal angiosperms and analyzed their expression. The taxa included in this study are Amborella (Amborellaceae), Nuphar (Nymphaeaceae), Illicium (Illiciaceae; Austrobaileyales), Magnolia (Magnoliaceae; Magnoliales) (Figure 3) and Asimina (Annonaceae; Magnoliales). The latter is unusual among basal angiosperms in that species of this genus possess a well-differentiated perianth of sepals and petals (Figure 2d). We compare our results here with expression data we reported earlier for several MADS-box genes from Magnolia and Eupomatia (Eupomatiaceae; Magnoliales) (Kim et al., 2005) and Nuphar (Zahn et al., 2005a). We also compare all of the data now available for basal angiosperms, eudicots and monocots. Expression data from basal angiosperms, when coupled with data for model organisms, serve as important reference points for understanding the evolution of floral regulatory genes throughout angiosperms. Our gene expression data can stimulate additional analyses of floral gene expression and function in basal angiosperms, ultimately providing important information regarding the origin of the flower.

Results

Orthology of MADS-box genes from basal angiosperms

We report here the following homologs of MADS-box genes: Am.tr.AG (AY936231), Am.tr.AGL2 (AY936232), Am.tr.AGL6 (AY936234), Nu.ad.AP1 (AY936223), Nu.ad.AP3.1 (DQ004465), Nu.ad.AP3.2 (DQ004464), Nu.ad.AG (AY936230), Il.fl.PI (AY936224), Il.fl.AP3.1 (AY936225), Il.fl.AP3.2 (AY936226), Il.fl.AP3.3 (AY936227), Il.fl.AG (AY936229), Ma.gr.AG (AY936228) and Ma.gr.AGL6 (AY936233) (Table 1). We isolated two AP3 homologs from Nuphar and three from Illicium. In these instances the abbreviation for the taxon and gene is followed by a full stop and then a number to distinguish these multiple homologs. Blast searches in GenBank identified all of these new genes as putative members of the MADS-box gene family. Our phylogenetic analysis of these genes and other MADS-box genes from basal angiosperms showed bootstrap support greater than 70% for clades of A-, B- (DEF and GLO), C- (and D-) and E-class genes (Figure 4). Considering the new genes reported here, one Nuphar gene was identified as an ortholog of AP1 (A class), two Nuphar genes and three Illicium genes were identified as orthologs of AP3 (B class), one Illicium gene was identified as an ortholog of PI (B class), genes from Amborella, Nuphar, Illicium and Magnolia were identified as orthologs of AG (C class), one Amborella gene was identified as an ortholog of AGL2 (E class) and genes from Amborella and Magnolia were identified as orthologs of AGL6.

Table 1.  Summary of expression patterns of floral genes in basal angiosperm flowers just prior to anthesis
TaxaGene nameOuter perianthe 1stInner perianth 2ndInner stamens 3rdCarpels 4thLeaves
  1. Newly identified genes in this study are indicated in bold. NA indicates not applicable. Eupomatia lacks a perianth.

  2. aData from Kim et al. (2005).

  3. bData from Chanderbali et al. (in prep.).

  4. cData from Zahn et al. (2005a).

  5. dWe did not obtain information from the outermost perianth organs of Amborella because of the small size of the outer tepals.

  6. eSome taxa investigated do not have a whorled arrangement of floral parts (e.g. Amborella has all parts spirally arranged; Magnolia has whorled perianth parts, but spirally arranged stamens and carpels).

SQUA subfamily
 NupharNu.ad.AP1++++++++++
 EupomatiaEu.be.AP1aNANA++++++++
 MagnoliaMa.gr.AP1a+++++++++++
 PerseaPe.am.AP1b++++++++++++
GLO subfamily
 AmborellaAm.tr.PI?d++++++++++
 NupharNu.ad.PI+++++++++++
 IlliciumIl.fl.PI+++++++++
 AsiminaAs.lo.PI++++++
 EupomatiaEu.be.PIaNANA+++++
 MagnoliaMa.gr.PIa+++++++++
 PerseaPe.am.PI.1b++++++++++
 PerseaPe.am.PI.2b+++++++++
DEF subfamily
 AmborellaAm.tr.AP3?d+++++++++
 NupharNu.ad.AP3.1++++++++++++
 NupharNu.ad.AP3.2++++++++++++
 IlliciumIl.fl.AP3.1++++++
 IlliciumIl.fl.AP3.2++++++++
 IlliciumIl.fl.AP3.3+++
 AsiminaAs.lo.AP3+++++++
 EupomatiaEu.be.AP3aNANA++++++++
 MagnoliaMa.gr.AP3a+++++++++
 PerseaPe.am.AP3b++++++++
AG subfamily
 AmborellaAm.tr.AG?d+++++++
 NupharNu.ad.AG++++++
 IlliciumIl.fl.AG+++++++
 MagnoliaMa.gr.AG++++++
 PerseaPe.am.AGb++++++++++++++
AGL2 subfamily
 AGL2/3/4 lineage
  AmborellaAm.tr.AGL2?d+++++++++
  NupharNu.ad.AGL2c++++++++++++
  MagnoliaMa.gr.AGL2b++++++++++++++
 AGL9 lineage
  EupomatiaEu.be.AGL9bNANA+++++++
  MagnoliaMa.gr.AGL9b+++++++++++++
AGL6 subfamily
 AmborellaAm.tr.AGL6?d+++++++
 MagnoliaMa.gr.AGL6++++++
Figure 4.

Strict consensus of two shortest trees from a maximum parsimony analysis of MADS genes (7608 steps, Consistency Index (CI) = 0.40 and Retention Index (RI) = 0.58).
Selected representatives of each major clade of MIKCC-type MADS-box genes were analyzed together with genes newly identified in this study (bold) and genes used in this study (underlined). Each new gene is a member of a well-supported (>70% bootstrap value) major clade of the MADS-box family (thickened nodes).

Analysis of gene expression using relative-quantitative RT-PCR

Representative gel photographs (Figure 5) illustrate our relative-quantitative RT-PCR (RQ RT-PCR) results, and a summary of the data is provided in Table 1. No signal was detected in any of the negative controls (i.e. samples that did not contain a cDNA template). Results for the Antirrhinum DEF and GLO genes, which served as reference samples (Figure 5f), were almost identical to the patterns previously observed using RNA in situ and Northern blot hybridizations (e.g. Sommer et al., 1991; Tröbner et al., 1992): both genes exhibited strong expression in petals and stamens, and GLO showed very weak expression in carpels in our experiments. Although very weak expression of DEF in sepals and carpels was reported in Antirrhinum (Sommer et al., 1991), no signal was detected in these organs in our study.

Figure 5.

Figure 5.

Relative quantitative RT-PCR results of floral MADS-box genes in basal angiosperms. Standard deviations are indicated for each value.
Flower buds collected just before anthesis were used. In the case of Magnolia, floral buds of 15 and 30 mm in diameter (just before anthesis) were examined. The open arrowhead indicates a longer band than expected (see text). For Nu.ad.AP3.2, the expression signal of expected bands and the sum of signals of longer and expected bands were calculated separately in carpels. TE, tepals; TE(M), tepals in male flowers; TE(F) tepals in female flowers; OTE, outer tepals; ITE, inner tepals; SE, sepals; OPE, outer petals; IPE inner petals; SN, stamens; SD, staminodes; CA, carpels; LE, leaves; BR, spathaceous bracts; TE1, three outer tepals; TE2, three middle tepals; TE3, three inner tepals; SE1, immature seeds of 1 mm in diameter; SE2, immature seeds of 2 mm in diameter.

Figure 5.

Figure 5.

Relative quantitative RT-PCR results of floral MADS-box genes in basal angiosperms. Standard deviations are indicated for each value.
Flower buds collected just before anthesis were used. In the case of Magnolia, floral buds of 15 and 30 mm in diameter (just before anthesis) were examined. The open arrowhead indicates a longer band than expected (see text). For Nu.ad.AP3.2, the expression signal of expected bands and the sum of signals of longer and expected bands were calculated separately in carpels. TE, tepals; TE(M), tepals in male flowers; TE(F) tepals in female flowers; OTE, outer tepals; ITE, inner tepals; SE, sepals; OPE, outer petals; IPE inner petals; SN, stamens; SD, staminodes; CA, carpels; LE, leaves; BR, spathaceous bracts; TE1, three outer tepals; TE2, three middle tepals; TE3, three inner tepals; SE1, immature seeds of 1 mm in diameter; SE2, immature seeds of 2 mm in diameter.

In Amborella, Am.tr.AP3 and Am.tr.PI were expressed in the tepals and stamens, consistent with the expression of AP3 and PI in Arabidopsis and DEF and GLO in Antirrhinum. Expression of both Am.tr.AP3 and Am.tr.PI was also detected in carpels, and weak expression of Am.tr.PI was found in leaves. Am.tr.AG was expressed in reproductive organs, and Am.tr.AGL2 in tepals, stamens and carpels, both similar to the expression reported for their orthologs in eudicots. In addition, Am.tr.AGL6 was strongly expressed in tepals, with intermediate levels of expression detected in stamens and carpels (Figure 5a).

The outer and inner tepals of Nuphar are morphologically similar, but the outer tepals are green, whereas the inner tepals are yellow (Figure 2b). Because of this color difference, the outer and inner tepals are often considered to be sepals and petals, respectively (e.g. Cronquist, 1988; Judd et al., 2002). However, outer and inner tepals of Nuphar exhibited very similar expression levels for all of the genes we investigated (Figure 5b; Table 1). Nu.ad.AP1 was expressed in all floral organs and leaves, with the strongest expression observed in carpels and leaves (Figure 5b; Table 1). Expression of Nu.ad.PI, Nu.ad.AP3.1 and Nu.ad.AP3.2 (Figure 5b) was detected in both outer and inner tepals, as well as in stamens and staminodes. Furthermore, the level of expression of Nu.ad.PI in the carpels was dependent on the stage of the floral bud (data not shown). In the carpels from floral buds 10–13 mm in diameter (near the time of male meiosis) relatively high expression was observed. However, expression of Nu.ad.PI was lower in carpels from flowers that were 30 mm in diameter (just before anthesis) and was not detected in carpels from open flowers (40 mm in diameter) (see also the results for real-time PCR). Hence, the data indicate a gradual decrease in expression of this PI homolog in carpels as the flower matures.

Although the two AP3 homologs of Nuphar had very similar expression patterns, the expression of Nu.ad.AP3.2 was relatively weak compared with Nu.ad.AP3.1: the signal of Nu.ad.AP3.2 was only detectable after 29 cycles (all other RQ RT-PCR experiments were performed with 26 or 27 cycles; see Experimental procedures). For Nu.ad.AP3.2 an additional band was detected only in carpels (open arrowhead in Figure 5b, fourth panel). The genomic sequence of the corresponding region of this additional band (DQ070749) indicates that this band represents an unspliced precursor RNA. The primer pairs used for RQ RT-PCR of Nu.ad.AP3.2 correspond to putative exon3 and exon7 of Nu.ad.AP3.2. The sequence of the additional band contained intron3, intron4, intron5 and intron6. Alternatively spliced or partially spliced RNA fragments, which are restricted to certain floral organs, have been reported for several MADS genes (Kim et al., 2005; Stellari et al., 2004). Nu.ad.AG was expressed in stamens (and staminodes) and carpels.

The Illicium PI homolog (Il.fl.PI) was strongly expressed in outer and inner tepals and stamens. The three Illicium AP3 homologs (Il.fl.AP3.1, Il.fl.AP3.2 and Il.fl.AP3.3) exhibited different expression levels among floral organs. Il.fl.AP3.1 was expressed strongly in stamens and inner tepals. Il.fl.AP3.2 was expressed at a high level in inner tepals and stamens and a medium level in outer tepals. Il.fl.AP3.3 was strongly expressed in inner tepals. Strong expression of the AG homolog (Il.fl.AG) was observed in inner tepals and stamens, and weak expression was observed in carpels.

Expression of the homologs of AP1, AP3, PI and AGL2/9 from Magnolia grandiflora was described previously (Kim et al., 2005) and is summarized in Table 1. Here, we present expression data for an AG homolog (Ma.gr.AG) and an AGL6 homolog (Ma.gr.AGL6) from M. grandiflora. The large size of Magnolia flowers also permits a comparison of expression levels between floral organs from floral buds of 15 and 30 mm in diameter. Expression of Ma.gr.AG was observed in both stamens and carpels (Figure 5d). For Ma.gr.AGL6, strong expression was observed only in tepals, a result similar to that observed for the Amborella AGL6 homolog. The levels of expression at two different stages of floral development in M. grandiflora (flowers of 15 and 30 mm in diameter) were very similar for all four genes investigated (Figure 5d), but with minor differences. For example, a weak signal of Ma.gr.AGL6 was detected in the spathaceous bract of the15 mm floral buds but not in the spathaceous bract of older buds (30 mm; just before anthesis). Also, weak expression of Ma.gr.AP3 was detected in seeds 2 mm in diameter, but not in seeds 1 mm in diameter.

Homologs of AP3 and PI were previously isolated (As.lo.AP3 and As.lo.PI; Kim et al., 2004) from Asimina longifolia. Both As.lo.AP3 and As.lo.PI were expressed in petals and stamens (Figure 5e), but were either not expressed or only weakly expressed in sepals; the lack of detected expression in sepals differs from the results reported here for AP3 and PI homologs in the outer perianth of other basal angiosperms.

Determining expression levels using real-time PCR

The real-time PCR results for Nu.ad.PI and Nu.ad.AP3.1 generally agree with those obtained using RQ RT-PCR (Figure 5b). Both genes showed strong expression in all floral parts in relatively young floral buds (10–13 mm in diameter). However, Nu.ad.PI expression was not detected in carpels from open flowers (40 mm in diameter) using real-time PCR (Figure 6a) whereas expression was detected using RQ RT-PCR. Importantly, identical results were obtained for Nu.ad.PI using both RQ RT-PCR and real-time PCR when the same floral samples were used (Figure 6b). In addition, Nu.ad.AP3.1, but not Nu.ad.PI, was also expressed in primary roots and the immature seed (1 mm in diameter) (Figure 6a,b).

Figure 6.

(a) The real-time PCR results for AP3 and PI homologs of Nuphar (Nu.ad.PI and Nu.ad.AP3.1). The Ct value of the 18S rRNA gene control was divided by those of Nu.ad.PI and Nu.ad.AP3.1.
(b) The relative quantitative (RQ) RT-PCR result for Nu.ad.PI using the same samples that were used in the real-time PCR experiment. Abbreviations are as in Figure 5. PRO, primary roots; SRO, secondary roots.

In situ hybridization studies of AP3 and PI homologs in Amborella and Nuphar

Our RQ RT-PCR experiments suggest that AP3 and PI homologs in basal angiosperms tend to have broader expression than that reported for their eudicot counterparts. To obtain information on the spatial expression pattern of AP3 and PI homologs, particularly at relatively early stages of flower development before the stages analyzed by the RT-PCR experiments, we performed RNA in situ hybridization experiments using gene-specific probes for AP3 and PI homologs of Amborella and Nuphar. For Amborella, only male flowers were included in this study because of the very limited amount of material available for this taxon, which is restricted in nature to New Caledonia and is cultivated only rarely in botanical gardens.

A recent study of floral development in Amborella (Buzgo et al., 2004) suggested the following developmental stages for male flowers: stage 1, flower initiation; stage 2, initiation of the transverse receptacular bracts and outer tepals; stage 3, initiation of inner perianth organs; stage 4, initiation of stamens; stage 6, development of sporophylls and of microsporangia; and stage 8, male meiosis (stages 5 and 7 occur in female flowers). For Am.tr.PI a strong signal was detected in the initiating tepals and in the primordia of other floral organs during stages 1–3 (Figure 7a). At stage 4, a strong signal was detected in the outer and inner tepals and also in initiating stamens (Figure 7b). At stages 6 and 8 (which follow the development of anthers), signals detected in tepals and in the connective tissue of the stamens were particularly high. A very strong signal was also detected in the anthers (Figure 7c). No expression was detected in bracteoles or receptacular bracts in any stage investigated (Figure 7a–c). The expression pattern of Am.tr.AP3 was similar to that of Am.tr.PI. In stages 1–3, a strong signal was detected in the initiating tepals and in the primordia of stamens (Figure 7d). In stages 6 and 8, a relatively weak signal was detected in the tepals and filaments of the stamens, and a strong signal was detected in anthers (Figure 7f). Weak expression was also detected in the vascular bundles (Figure 7f).

Figure 7.

In situ hybridization using Am.tr.PI and Am.tr.AP3 gene probes to longitudinal sections of developing Amborella flowers. Developmental stages follow Buzgo et al. (2004): a–c, Am.tr.PI; d–f, Am.tr.AP3 (a and d stages 1–3, b and e stage 4, c and f stages 6, 8). Solid arrows indicate bracteoles, and open arrows indicate receptacular bracts. T, tepal; S, stamen; C, connective; A, anther. In each pair of images, the left image is bright field and the right image is dark field. All scale bars are 0.5 mm.

The floral materials of Nuphar advena used in this study were limited; floral buds must be collected from natural populations and are only available seasonally. In addition, Nuphar floral buds develop individually in the axils of leaf primordia. Therefore, the total number of very young floral buds available for analysis was small. Nonetheless, our experiments revealed that both PI and AP3 homologs were expressed strongly in stamen and staminode primordia (Figure 8). In addition, expression of the Nu.ad.AP3.2 and Nu.ad.PI genes was also clearly detectable in the inner tepals, and Nu.ad.AP3.1 is also possibly expressed in the inner tepals (Figure 8, arrows). Very weak expression of these genes may have been present in the tips of the outer tepals (Figure 8, arrowheads). This pattern of expression was observed for a number of sections, suggesting that it is representative of the floral material at this stage of development. There was no detectable expression in the fused carpels at the center of the flower. The bright spots in the region below the gynoecium are not a true signal because they are not from the silver grains of the photosensitive emulsion, but rather from the tissue. Therefore, these spots represent non-specific background which can also be seen to some extent in the sense control (Figure S1).

Figure 8.

In situ hybridization using Nu.ad.PI, Nu.ad.AP3.1 and Nu.ad.AP3.2 gene probes to longitudinal sections of developing Nuphar flowers: (a) Nu.ad.AP3.1, (b) Nu.ad.AP3.2, (c) Nu.ad.PI. In each pair of images, the left image is bright field and the right image is dark field. All scale bars are 1.0 mm.

We tested the specificity of our in situ probes using Southern blots. For the various MADS genes we tested, each probe was specifically hybridized to the gene from which it was derived under the same hybridization temperature (42°C) and washing stringency (salt conditions) as those of the in situ experiments (Figure S2).

Evolutionary reconstruction of gene expression patterns

Our analyses of evolutionary transformations of floral MADS-box gene expression revealed that the AP3 and PI homologs were ‘strongly expressed’ in all floral organs in the common ancestor of all extant angiosperms (Figure 9). However, after the first two branches of extant angiosperms (represented here by Amborella and Nuphar), the ancestral state of both AP3 and PI homologs was ‘not expressed/weakly expressed’ in the carpels, and AP3 was ‘equivocal’ in the outer perianth. The ancestral state for PI and AP3 homologs in the eudicots was reconstructed in each as ‘not expressed/weakly expressed’ in the outermost floral organs.

Figure 9.

Evolution of gene expression patterns of floral MADS-box genes in angiosperms. Character states are indicated for each floral part: colored floral parts represent ‘strongly expressed’ genes (+++ and ++ in Table 1); open floral parts indicate that the genes are ‘not expressed/weakly expressed’ (− and + in Table 1). Dashed organs in the nodes indicate ‘equivocal’ status, and those in the terminals indicate ‘uncertain (contain both character states)’ status. Symplesiomorphic character states are indicated at each node. Each color represents a MADS gene subfamily/lineage: brown, SQUA (excluding the euAP1 lineage); yellow, the euAP1 lineage; red, GLO; blue, DEF; green, AG. Where expression data are not available or the homologous organ is not present, we used black.

The ancestral pattern of expression of AG homologs in angiosperms was restricted to the reproductive organs (Figure 9). However, in Illicium the AG homolog was expressed in the inner tepals and stamens, with ‘no/weak’ expression in carpels. In Persea, AG homologs were expressed in all floral organs. These expression patterns in Illicium and Persea appear to represent derived states for these taxa, based on our reconstructions, but more data are needed. For AP1 homologs we reconstructed expression patterns for euFUL genes and euAP1 genes separately. AP1 homologs from basal angiosperms were compared with euFUL genes. The ancestral state of the euFUL lineage for all angiosperms was ‘strongly expressed’ in reproductive organs and ‘equivocal’ in perianth (Figure 9). Because euAP1 genes were found only in core eudicots (Litt and Irish, 2003), we reconstructed the ancestral state of euAP1 gene expression in core eudicots separately. Strong expression was restricted to the perianth in the ancestor of core eudicots (Figure 9). When we compare euAP1 genes and FUL-like genes of basal angiosperms, the perianth-specific expression reported for core eudicots (Hardenack et al., 1994; Mandel et al., 1992; Taylor et al., 2002) is derived in our reconstructions.

Discussion

Conservation and divergence of expression patterns of floral MADS-box genes in angiosperms

To gain insights into the evolution of expression and function of floral MADS-box genes we examined the expression of homologs of known floral MADS-box genes from several basal angiosperms. Included among the taxa analyzed here are Amborella and Nuphar, phylogenetically pivotal taxa that are the successive sister groups to all other extant flowering plants. We obtained expression data using RQ RT-PCR, real-time PCR and RNA in situ hybridization experiments. The results obtained from these different approaches agreed closely when the same stage of floral development was assayed, supporting the reliability of the results. The minor differences in expression patterns among methods are likely to be due to sampling and technical differences. For example, the RQ RT-PCR data are generally for later stages of floral development, but in situ hybridization data are presented for early stages of floral development. It will be very useful in the future to obtain additional in situ hybridization data for earlier stages of floral development for additional genes in more basal angiosperm taxa to complement the RQ RT-PCR results reported here.

The results we provide here, together with expression data reported from other studies (summarized in Table 1), indicate that most of the homologs of known floral genes isolated from basal angiosperms are expressed in those floral organs that are functionally and/or morphologically similar to the organs in genetic model organisms (e.g. Arabidopsis and Antirrhinum) that require these genes. For example, in basal angiosperms AP3 and PI homologs are expressed in petal-like perianth organs (e.g. tepals) and stamens, and AG homologs are expressed in stamens and carpels. Similarly, SEP homologs are expressed in all floral organs. At the same time, many of these genes also exhibit expression in organs other than those expected on the basis of the known function of their homologs in Arabidopsis or Antirrhinum. For example, expression of some AP3/PI homologs was detected in carpel tissue of basal angiosperms, and expression of AG homologs was observed in the perianth of some basal taxa. Therefore, homologs of AP3, PI, AG and SEP from basal angiosperms generally have expression patterns that include a conserved component, as well as a broader component, when compared with their counterparts in well-characterized genetic systems (Goto and Meyerowitz, 1994; Jack et al., 1992; Ma et al., 1991; Mandel and Yanofsky, 1998; Mandel et al., 1992). In contrast, homologs of the AP1 gene from basal angiosperms show a pattern of expression that is very different from that reported in model systems: homologs of AP1 from basal angiosperms show strong expression in leaves, perianth and reproductive organs, unlike AP1 which is expressed only in the perianth in Arabidopsis and other eudicots (Mandel et al., 1992).

In Amborella, Nuphar and the other basal angiosperms having an undifferentiated perianth (e.g. Magnolia, Illicium, Persea), homologs of the B-class genes PI and AP3 are generally expressed in both the outer and inner perianth organs. Hence, it may be that expression of PI and AP3 homologs in the outer whorl typifies most basal angiosperms. There is also evidence for the expression of PI and AP3 homologs in the outer perianth in some basal eudicots (e.g. Ranunculaceae). Even in the core eudicots, expression of PI and AP3 has been detected at very early stages in the outer perianth (Jack et al., 1992; van der Krol et al., 1993; Schwarz-Sommer et al., 1992; Tsuchimoto et al., 2000).

Considering the patterns of expression of homologs of AG, AP3/PI and AP1, our reconstructions indicate that, within angiosperms, the ancestral expression pattern of AG homologs is most similar to that observed in core eudicots (Bradley et al., 1993; Davies et al., 1999; Kater et al., 1998; Kempin et al., 1993; Yanofsky et al., 1990). In contrast, the AP3 and PI homologs have ancestral states suggesting broader expression patterns in early angiosperms than those observed in eudicots (Angenent et al., 1993; Davies et al., 1996; Goto and Meyerowitz, 1994; Hardenack et al., 1994; Jack et al., 1994; van der Krol et al., 1993; Schwarz-Sommer et al., 1992; Tröbner et al., 1992; Yu et al., 1999). Finally, the ancestral state for the expression of AP1 homologs differs dramatically from the expression pattern found in euAP1 genes of core eudicots (Hardenack et al., 1994; Mandel et al., 1992; Taylor et al., 2002), suggesting that the perianth-specific expression reported for the core eudicots (Hardenack et al., 1994; Mandel et al., 1992; Taylor et al., 2002) is derived and associated with a gene duplication (see below).

Functional implications for MADS-box genes in basal angiosperms

Genetic and molecular studies in core eudicots and monocots support a strong correlation between the pattern of expression and function of floral MADS-box genes, although sometimes transient and/or weak expression does not correspond to a known genetic function (Goto and Meyerowitz, 1994; Jack et al., 1992; van der Krol et al., 1993; Schwarz-Sommer et al., 1992; Tröbner et al., 1992; Tsuchimoto et al., 2000). Therefore, the expression patterns reported here strongly suggest that AG homologs in basal angiosperms probably promote the development of reproductive organs, as AG and PLE do in Arabidopsis and Antirrhinum, respectively. The detection of AG homologs in the perianth of both Persea and Illicium suggests that expression of AG in all floral organs in these taxa is independently derived. These two genera are not closely related among basal angiosperms, with Illicium being a member of Austrobaileyales and Persea a member of Laurales in the magnoliid clade (Figure 1). Our results indicate that patterns of expression of the AP3/PI homologs in basal angiosperms are generally broader than the patterns of expression of these genes in model core eudicots, such as Arabidopsis. For example, in both Amborella and Nuphar, AP3 and PI homologs are expressed in the perianth and carpels, as well as in stamens. Because the expression in the inner perianth and stamens is either stronger or detected at both early and late stages it is likely that these genes function to control the identity of these organs, similar to their respective homologs in core eudicots (Angenent et al., 1993; Davies et al., 1996; Goto and Meyerowitz, 1994; Hardenack et al., 1994; Jack et al., 1994; van der Krol and Chun, 1993; Schwarz-Sommer et al., 1992; Tröbner et al., 1992; Yu et al., 1999; Zahn et al., 2005b). The expression of these AP3/PI homologs in the outer perianth and carpels could represent function, because the outer perianth of many basal angiosperms has petal-like characteristics. Furthermore, a series of morphological transitions from bracts through tepals and stamens to carpels observed in Amborella (Buzgo et al., 2004) also supports possible functioning of these genes in outer perianth and carpels. Alternatively, such expression could be a ‘molecular fossil’ left from an ancestral state that is no longer needed today.

Functional studies of AP1 homologs are very limited, and members of this subfamily are functionally divergent (reviewed in Litt and Irish, 2003). To date, AP1 is the only gene in the SQUA subfamily that has been shown to confer the A function (Bowman et al., 1993; Litt and Irish, 2003). In Antirrhinum, the expression pattern of the AP1 homolog SQUA is the same as that of AP1; however, mutant analysis did not demonstrate the A function of SQUA (Huijser et al., 1992; Taylor et al., 2002). A gene duplication generated the euAP1 and euFUL lineages during the evolution of the SQUA subfamily, perhaps in the common ancestor of the eudicots (Litt and Irish, 2003). Expression of AP1 homologs is broader in basal angiosperms than in core eudicots (i.e. euAP1 genes), suggesting that AP1 homologs may have different functions in basal taxa. In Magnolia and Nuphar expression of the AP1 homolog was higher in leaves and carpels than in the perianth and stamens, resembling the expression of members of the euFUL lineage rather than that of members of the euAP1 lineage (Irish, 2003; Litt and Irish, 2003).

Phylogenetic analyses support a close relationship among the SQUA, AGL2 (SEP) and AGL6 subfamilies of the MADS-box gene family (Becker and Theissen, 2003). SEP homologs from basal angiosperms are expressed in all floral organs, similar to some SEP genes in Arabidopsis, suggesting that these genes may have similar functions in basal angiosperms to those reported in Arabidopsis.

When the ABC model was proposed to explain the genetic control of the identity of floral organs on the basis of similar homeotic mutants in Arabidopsis and Antirrhinum, most of the corresponding floral homeotic genes had not yet been cloned (Coen and Meyerowitz, 1991). The subsequent molecular analysis of these genes revealed that most of them encode highly similar MADS-box genes, raising the question of whether their homologs in other angiosperms also have similarly conserved functions. Our data support the idea that AG homologs have a conserved function in angiosperms that is required for the C function of the ABC model. Similarly, AP3 and PI homologs probably also have conserved functions necessary for the B function for petal (inner perianth) and stamen identities. In addition, AP3 and PI homologs in basal angiosperms may have a broader function, extending to the outer perianth, particularly if the outer perianth resembles petals as in the case of Amborella, Nuphar, Magnolia and many other basal angiosperms. Furthermore, this broader pattern of B function may represent the ancestral condition for angiosperms. On the other hand, AP1 homologs in basal angiosperms do not seem to share conserved functions that specifically control the identity of the outer perianth, as in Arabidopsis (Bowman et al., 1993; Mandel et al., 1992). Recent studies indicate that four SEP genes (previously AGL2, −3, −4 and −9) redundantly control sepal identity and contribute to the specification of other floral organs (Ditta et al., 2004; Pelaz et al., 2000; Theissen, 2001). Because the SQUA and AGL2 subfamilies are phylogenetically closely related it is possible that, in basal angiosperms, both SQUA and AGL2 subfamily members together control the identity of the outer perianth and also contribute to the identity of other floral organs. The expression of AP1 homologs in basal angiosperms is also similar to that of AGL3 (SEP4) in Arabidopsis (Huang et al., 1995). In addition, the AGL6 homolog of Magnolia was strongly expressed only in tepals (Figure 5d). In the case of Amborella, although Am.tr.AGL6 is expressed in all floral organs, the strongest expression was found in tepals (Figure 5a). Therefore, other candidates for A-function genes in basal angiosperms could be AGL6 homologs.

MADS-box gene expression and perianth differentiation

Most basal angiosperms do not exhibit a well-differentiated perianth of sepals and petals (e.g. Endress, 2001). Rather, the perianth typically consists of tepals. The broad expression of AP3/PI homologs observed throughout the undifferentiated perianth of many basal angiosperms (e.g. Amborella, Nuphar and Magnolia) complements the lack of a clear morphological distinction between sepals and petals in these taxa. In Nuphar, the outer perianth organs are green and more ‘sepal-like’; the inner perianth organs are brightly colored and more ‘petal-like’ (Figure 2b). However, this distinction is not clear-cut, and the perianth is often described as consisting of tepals.

In Asimina, a morphologically well-differentiated perianth of sepals and petals (Figure 2d) corresponds very well with the observation that the expression pattern of both AP3 and PI homologs in Asimina is the same as that in Arabidopsis: expression of AP3 and PI homologs was not observed in sepals, but was detected in petals and stamens. Asimina clearly represents an independent derivation of a differentiated perianth from that observed in eudicots (e.g. Albert et al., 1998; Ronse De Craene et al., 2003). It is noteworthy, therefore, that the pattern of expression of the AP3/PI homolog is similar in these phylogenetically well-separated taxa.

Experimental procedures

Plant materials

We collected samples from the following sources: Amborella trichopoda, plants cultivated at the National Tropical Botanical Garden, HI, USA (DL8346, TF6481, DHL8350, HAW); N. advena, plants collected in Black Moshannon State Park, PN, USA (S. Kim 1140, FLAS); Illicium floridanum, plants cultivated on the campus of the University of Florida, Gainesville, FL, USA (S. Kim 1139, FLAS); A. longifolia, a plant collected in Morningside Park, Gainesville, FL, USA (S. Kim 1129, FLAS); M. grandiflora, a plant cultivated on the campus of the University of Florida, Gainesville, FL, USA (S. Kim 1138, FLAS). Entire flowers at varying stages of early development up to anthesis were removed and dropped immediately into liquid nitrogen and stored at −80°C.

We used two general methods to isolate and characterize genes for floral organ identity. The first method used expressed sequence tags (ESTs) obtained as part of the Floral Genome Project (Albert et al., 2005; Soltis et al., 2002; http://fgp.bio.psu.edu/cgi-bin/fgpmine/index.cgi). We subsequently obtained finished sequences of these ESTs using M13 and M13 reverse universal primers. The second method involved isolating RNA from floral buds and developing flowers, making cDNA, and using RT-PCR following the general methods of Kim et al. (2004) (described below).

RNA extraction, RT-PCR and cDNA sequence determination

RNA extractions for all taxa were performed following a modified method of the RNeasy Plant Mini Kit (Qiagen, Stanford, CA, USA). The modification includes a portion of the CTAB DNA extraction protocol (Doyle and Doyle, 1987) and subsequent use of the RNeasy Plant Mini Kit. This method was originally developed for the successful extraction of RNA from basal angiosperms such as Amborella and Nuphar (Kim et al., 2004). Reverse transcription was performed following the manufacturer's directions using SuperScript II RNase H-reverse transcriptase (Invitrogen, Carlsbad, CA, USA) and polyT primer (5′-CCG GAT CCT CTA GAG CGG CCG C(T)17-3′). PCR reactions were performed using a MADS gene-specific degenerate primer (5′-GGG GTA CCA AYM GIC ARG TIA CIT AYT CIA AGM GIM G-3′) and the polyT primer used in reverse transcription (Kramer et al., 1998). PCR conditions were those employed by Kramer et al. (1998). PCR bands over 800 bp in size were excised from the agarose gel and purified using the Geneclean II Kit (Q·Bio Gene, Carlsbad, CA, USA). Purified DNAs were cloned using the TOPO TA Cloning Kit (Invitrogen). Plasmid DNAs were purified from cloned cells using the FastPlasmid Mini Kit (Eppendorf, Westbury, NY, USA). Cycle sequencing reactions were performed using the CEQ DTCS-Quick Start Kit (Beckman Coulter, Fullerton, CA, USA), and cDNA sequences were determined using a CEQ 8000 sequencing system (Beckman Coulter).

Characterization and identification of genes

We determined gene identity using a blast approach followed by a phylogenetic analysis. To verify the subfamily identities of newly isolated genes from the taxa under investigation, and to address their orthology to previously reported genes, we added our MADS-box sequences to a large data set of 82 sequences representing all subfamilies of MIKCC-type MADS genes (Becker and Theissen, 2003).

For each gene identified as a putative member of the MADS-box gene family, we used the following naming system (see also Kim et al., 2004). Each gene was named using the first two letters of the genus name, followed by a full stop, and the first two letters of the specific epithet; this was, in turn, followed by the abbreviated name of its phylogenetically closest homolog in Arabidopsis. For example, a putative AGAMOUS homolog isolated and identified from A. trichopoda was abbreviated as: Am.tr.AG.

Amino acid alignment was conducted using clustal x (ver. 1.83; Thompson et al., 1997) with manual adjustment. Maximum parsimony (MP) analysis was performed on the amino acid data set using paup* 4.0b10 (Swofford, 2001). The search strategy involved 100 random addition replicates with tree-bisection–reconnection (TBR) branch swapping, saving all optimal trees. To assess support for each node, bootstrap analysis (Felsenstein, 1985) was performed using 100 bootstrap replicates, each with 10 random addition replicates and TBR branch swapping, saving all optimal trees.

Expression studies

For the examination and quantification of gene expression, we used RQ RT-PCR. We also present comparable data for Nuphar B-class homologs based on real-time PCR. In situ hybridization studies were employed for B-class homologs in both Amborella and Nuphar.

For the RQ RT-PCR and real-time PCR analyses we used samples collected just before anthesis for all taxa used in this study. Flowers were dissected while completely frozen. Separated piles of floral parts from almost-opened flowers were made while carefully working to ensure that all parts remained frozen: tepals, stamens, staminodes (for Nuphar) and carpels. Young leaves were also analyzed. For Amborella, both male and female flowers were included. Although tepals of Amborella showed gradual morphological changes from the outermost to the innermost positions, only inner tepals were sampled separately because the outer tepals were too small to collect. Tepals from female flowers were collected together with staminodes. Tepals of Illicium exhibit similar morphological transitions to those of Amborella. Samples of tepals from both the outermost and innermost whorls of Illicium were prepared and analyzed for expression of floral genes. As a reference, we also examined the expression patterns of the Antirrhinum DEF and GLO genes using RQ RT-PCR.

Total RNAs were extracted from each sample using the RNeasy Plant Mini Kit (Qiagen). After RNA extraction we treated samples with DNase to avoid potential contamination by genomic DNA (DNase-free kit from Ambion, Austin, TX, USA). Reverse transcription using RNA from each floral part was performed following the manufacturer's directions using SuperScriptTM II RNase H-reverse transcriptase (Invitrogen). We used random-hexamer oligonucleotides for the reverse transcription instead of polyT primer because the 18S rRNA gene was used as an internal control of quantification in both RQ RT-PCR and real-time PCR.

Relative-quantitative-RT PCR.  Methods followed those used previously (Kim et al., 2005). We performed multiplex PCR using a gene-specific primer pair (Figure S3), the 18S rRNA gene primer pair (internal control), and a competitive primer pair to the 18S rRNA gene primers (competimers) following the protocol of QuantumRNA (Ambion). The 18S rRNA gene was used for the internal control in each reaction. Because the PCR signal of the 18S rRNA gene is higher than that of the specific genes we are studying, the ‘competimers’ of the 18S primer pair included in the QuantumRNA kit were used to reduce the 18S rRNA PCR signal. The optimal ratio of the 18S primer pair to competimers was tested for each gene to obtain a similar level of PCR signal between the 18S rRNA and that of the mRNA of each gene. The optimal ratio ranged from 3:7 to 6:4 for the genes that we examined. PCR reactions for all genes were performed with 26 or 27 cycles (except Nu.ad.AP3.2 which needed 29 cycles) at 95°C (30 sec), 55°C (30 sec) and 72°C (30 sec) using an Eppendorf Mastercycler (Brinkmann, Westbury, NY, USA). A range of 1–256 ng of total RNA (back-calculated from the amount of cDNA used in the PCR reaction after the RT-PCR) was tested, and 16–64 ng of total RNA were found to generate accumulation of unsaturated PCR product for each gene through 26 or 27 cycles of PCR. We used 25 ng of total RNA for the RQ RT-PCR. For all PCRs, we used a negative control that did not contain cDNA template. Twenty microliters from each PCR reaction were fractionated in a 2% (w/v) agarose gel containing 10−4 (w/v) ethidium bromide in Tris-acetate EDTA buffer. Gel images were analyzed using kodark 1D Image Analysis Software (Kodak, Rochester, NY, USA). Three to 15 replicates of RQ RT-PCR were performed for each gene using cDNAs from more than two independent RNA samples extracted from different individuals. The relative PCR signal of the specific gene to the 18S rRNA gene and its standard deviation were calculated for each floral organ. The gene specificity of each PCR product was confirmed by sequencing all PCR products. To compare relative expression levels among taxa we (i) converted the highest expression value in each gene to 1, (ii) made a histogram showing the expression of each gene in each organ using relative values based on the conversion in (i), and (iii) evaluated the relative amount of expression in each histogram as follows: −, not expressed; +, <0.1; ++, 0.1–0.4; +++, 0.4–1. We analyzed two different developmental stages using floral materials of M. grandiflora for Ma.gr.AP3, Ma.gr.PI, Ma.gr.AG and Ma.gr.AGL6. Samples of floral organs were obtained from the dissection of floral buds 15 and 30 mm in diameter, respectively.

Real-time PCR.  We also used real-time PCR to investigate expression of the AP3 and PI homologs of N. advena (Nu.ad.AP3.1 and Nu.ad.PI). This method includes a third primer as a probe and a fluorescent dye-labeled system and can provide a relative quantification of expression (Chiang et al., 1996; Leutenegger et al., 1999). We analyzed samples of floral organs from young floral buds (1.0–1.5 mm in diameter) and open flowers (40 mm in diameter) and samples from vegetative organs. The probe and primers for real-time PCR were designed using Primer Express 2.0 (Applied Biosystems, Foster City, CA, USA). The designations of primers and probes are indicated in Figure S3. We performed PCR reactions with TaqMan® Universal PCR Master Mix (Applied Biosystems) using a GeneAmp® 5700 Sequence Detection System (Applied Biosystems) following the recommendations of the manufacturer. A TaqMan® Ribosomal RNA Control Reagent (Applied Biosystems) was used for the internal control of each sample. Five independent reactions were performed for each sample. The relative ratio of threshold cycle (Ct) values between the 18S rRNA gene and the specific gene and their standard deviations were calculated for each sample.

RNA in situ hybridization.  For the in situ hybridization study we employed B-class homologs of A. trichopoda and N. advena. Fresh Amborella and Nuphar floral bud samples were immediately fixed in formaldehyde–acetic acid–ethanol (FAA) for 4–6 h. For Amborella, gene-specific primers were designed in the C-domain and a part of the K-domain to make RNA probes (Figure S3). PCR products of these regions were purified and cloned using the pGEM® T-vector system (Promega, Madison, WI, USA). 35S-dUTP-labeled RNA probes were synthesized by in vitro transcription: antisense and sense (negative control) transcripts were generated by using either T7 or T3 RNA polymerase (Promega). For Nuphar, an antisense RNA probe for Nu.ad.PI was transcribed from the plasmid 30M-A20 from the Nad03 floral cDNA library linearized with BamHI with T7 RNA polymerase in the presence of 35S-UTP (K and C domains). Antisense RNA probes for Nu.ad.AP3.1 and Nu.ad.AP3.2 were synthesized using T7 RNA polymerase with EcoRI-linearized plasmid 19M-E24 and ApoI-linearized plasmid 37M-E01, respectively, from the Nad03 floral cDNA library. A sense RNA probe was generated using T3 RNA polymerase with XhoI-linearized 30MS2-A10. Information on clones used in this experiment is available at http://fgp.bio.psu.edu/cgi-bin/fgpmine/index.cgi. The transcripts were partially hydrolyzed by incubation at 60°C in 0.1 m Na2CO3·NaHCO3 buffer, pH 10.2, for 45 min. The sample embedding, hybridization, washing and autoradiography were performed as described previously (Drews et al., 1991; Flanagan and Ma, 1994).

Character-state reconstruction

To investigate the diversification of expression patterns of A-, B-, C- and E-class homologs across angiosperms we conducted a character-state reconstruction using MacClade (ver. 3.04; Maddison and Maddison, 1992) and a phylogenetic framework for angiosperms inferred from recent multigene analyses (e.g. Qiu et al., 1999; P. Soltis et al., 1999; D. Soltis et al., 2000; Zanis et al., 2002; reviewed in Soltis and Soltis, 2004). We used the ‘all most parsimonious states’ optimization in MacClade because the accelerated transformation (acctran) and delayed transformation (deltran) optimizations cannot be applied when a polytomy is present.

In addition to our expression data for A-, B-, C- and E-class homologs in basal angiosperms, we added expression data from major lineages of angiosperms (see Figure S4). We selected Silene, Arabidopsis, Antirrhinum, Petunia and Gerbera as well-studied representatives of core eudicots (summarized in De Bodt et al., 2003; Irish, 2003); Ranunculus was used as a representative of basal eudicots (Kramer et al., 2003), and Zea, Oryza, Tulipa, Asparagus and Sagittaria were added as representatives of monocots (summarized in De Bodt et al., 2003; Irish, 2003; Kanno et al., 2003; Park et al., 2003, 2004). For basal angiosperms, Nuphar (Zahn et al., 2005a), Asarum (Piperales; Kramer and Irish, 2000), Persea (Laurales; A. Chanderbali, SK, M. Buzgo, Z. Zheng, PS and DS, unpublished data) and Magnolia (Magnoliales; Kim et al., 2005) were added to the data presented here. Because some of the MADS-box gene subfamilies contain major duplications, only genes representing one of the duplicates were included. In our analyses of B-class genes, TM6 genes (in the DEF subfamily) were excluded because they do not show B function (Kim et al., 2004). D-class genes in the AG subfamily (Kramer et al., 2004) were excluded. The expression of AP1 homologs in basal angiosperms was more similar to that reported for euFUL genes than for euAP1 genes (Litt and Irish, 2003). Therefore, we compared AP1 homologs of basal angiosperms with euFUL genes. For euAP1 genes, we only reconstructed character evolution in core eudicots. Because our purpose was to address the evolution of ‘strong’ expression (which is more likely than ‘weak’ expression to be associated with function) and because we used binary character coding, weak expression (i.e. ‘+’ in Table 1) was treated as ‘0’. If patterns of expression of multiple genes within a species differed (e.g. the expression of Il.fl.AP3.1, Il.fl.AP3.2 and Il.fl.AP3.3 in outer perianth and stamens of I. floridanum), we considered the expression for that gene in that species as ‘equivocal’ (Figure S4). For Amborella, with the exception of PI and AP3 (which were examined via in situ hybridization), we did not address expression in the outermost tepals because of the small size of the flowers. Therefore, the expression of C-class homologs in the outer perianth of Amborella was considered ‘uncertain’ (Figure 9 and Figure S4).

Acknowledgements

We thank Doug Yul Sung and Ilho Kang for technical advice regarding RQ RT-PCR experiments and Laura Zahn and Donglan Tian for help with Nuphar in situ hybridization experiments. We also thank Andre Chanderbali, Matyas Buzgo, Günter Theissen, Samuel Brockington, Laura Zahn and Jim Leebens-Mack for helpful comments and discussion. This study was supported by NSF grant PGR-0115684.

Ancillary