The B–class of MADS box genes has been studied in a wide range of plant species, but has remained largely uncharacterized in legumes. Here we investigate the evolutionary fate of the duplicated AP3-like genes of a legume species. To obtain insight into the extent to which B-class MADS box gene functions are conserved or have diversified in legumes, we isolated and characterized the two members of the AP3 lineage in Medicago truncatula: MtNMH7 and MtTM6 (euAP3 and paleoAP3 genes, respectively). A non-overlapping and complementary expression pattern of both genes was observed in petals and stamens. MtTM6 was expressed predominantly in the outer cell layers of both floral organs, and MtNMH7 in the inner cell layers of petals and stamens. Functional analyses by reverse genetics approaches (RNAi and Tnt1 mutagenesis) showed that the contribution of MtNMH7 to petal identity is more important than that of MtTM6, whereas MtTM6 plays a more important role in stamen identity than its paralog MtNMH7. Our results suggest that the M. truncatula AP3-like genes have undergone a functional specialization process associated with complete partitioning of gene expression patterns of the ancestral gene lineage. We provide information regarding the similarities and differences in petal and stamen development among core eudicots.
Our understanding of the molecular mechanisms controlling the genetic regulation of flower development rests largely on genetic analyses performed in the core eudicot species Arabidopsis thaliana and Antirrhinum majus. These studies led to the ABC model (Bowman et al., 1989; Coen and Meyerowitz, 1991). However, additional information is emerging regarding the identification of candidate genes that control floral organ identity in other angiosperms, including legumes (Baum et al., 2002; Hecht et al., 2005; Irish, 2006; Theissen and Melzer, 2007; Soltis et al., 2009).
According to the ABC model, a combination of B- and A-function genes specifies petal identity in the second floral whorl, whereas a combination of B- and C-function genes controls stamen identity in the third whorl. A pair of MADS box genes, APETALA3 (AP3) and PISTILLATA (PI), encode the B-function activity in A. thaliana, and DEFICIENS (DEF) and GLOBOSA (GLO) encode the B-function activity in A. majus. Mutations in either one of these genes produce transformations of petals into sepals and stamens into carpels (Bowman et al., 1989; Sommer et al., 1990; Schwarz-Sommer et al., 1992; Tröbner et al., 1992; Goto and Meyerowitz, 1994; Jack et al., 1994).
Phylogenetic analysis of B-class MADS box genes from representatives of all major lineages of angiosperms shows a high frequency of gene duplications within both AP3 and PI lineages (Kramer et al., 1998, 2003; Kramer and Irish, 2000; Becker et al., 2002; Aoki et al., 2004; Kim et al., 2004; Stellari et al., 2004). The ancestral gene of the AP3/PI lineages is thought to have undergone a duplication event yielding the AP3 and PI lineages before diversification of the angiosperms (Aoki et al., 2004; Kim et al., 2004; Stellari et al., 2004). Subsequently, the AP3 lineage underwent another major duplication at the base of the core eudicots, giving rise to two AP3-like lineages: one termed euAP3 that contains AP3 itself, and another lineage named TM6, which lacks a representative in A. thaliana, although it is present in many other eudicot taxa (Kramer et al., 1998, 2006).
Duplicated genes are generally considered to adopt one of three possible fates: non-functionalization, in which one copy is silenced, neo-functionalization, in which one copy acquires an entirely new function whereas the other copy maintains the original function (Ohno, 1970; Force et al., 1999), and sub-functionalization, in which each descendant copy adopts part of the function of the ancestral gene (Hughes, 1994; Force et al., 1999; Lynch and Force, 2000; Lynch et al., 2001).
Most studies proposing a cause–effect relationship between gene duplication and functional divergence and evolution rely on phylogenetic analysis and expression patterns of duplicated genes. In many of these cases, functional studies are lacking. However, studies performed using various members of the Solanaceae showed sub-functionalization of duplicated genes in the AP3 lineages, with euAP3-type genes evolving to play a primary role in petal and stamen development, whereas TM6-type genes have a partially overlapping function in stamen development (Liu et al., 2004; Vandenbussche et al., 2004; de Martino et al., 2006; Geuten and Irish, 2010). Recently, AP3-like genes have been functionally characterized in Gerbera hybrida, and were shown to play similar roles as their orthologous genes in Petunia hybrida, Solanum lycopersicum and Nicotiana benthamiana (Broholm et al., 2010).
Flowers of model dicot species, such as A. thaliana or A. majus, form the primordia of petals and stamens independently, but in some legumes, including Medicago truncatula, these organs derive from special structures called common primordia that probably represent an evolutionary specialization. These structures comprise four ephemeral meristems that develop between sepal and carpel primordia, and, upon division, each follows a characteristic pattern to produce the petal and stamen primordia (Ferrándiz et al., 1999; Benlloch et al., 2003; Tucker, 2003). B-class genes have been suggested to contribute to control of the identity and determinacy of common primordia in legumes (Ferrándiz et al., 1999; Taylor et al., 2001; Berbel et al., 2005), but experimental data are required to reinforce this hypothesis.
We report here the existence of two duplicated AP3-like genes in M. truncatula (an euAP3 gens named MtNMH7 and a paleoAP3 gene named MtTM6). To shed light on the specific contribution of these paralogous genes to the control of petal and stamen identity, we investigated their detailed expression pattern and used reverse genetics approaches to uncover the function of these genes in legumes. We show that the duplicated AP3-like genes in this legume species control petal and stamen identity. Our results suggest that the MtNMH7 and MtTM6 paralogous genes have undergone a post-duplication sub-functionalization process, associated with complete partitioning of gene expression patterns of the ancestral gene lineage.
Two Medicago truncatula AP3-like genes
We previously isolated and functionally analyzed two M. truncatula B-class MADS box genes belonging to the PI/GLO sub-family (Benlloch et al., 2009). With the aim of isolating the other B-class MADS box genes involved in control of petal and stamen identity in the M. truncatula flower, we screened a floral cDNA library using the MADS box fragment of the A. majus DEFICIENS gene as a probe. Eight of the isolated clones showed significant similarity to the AP3 gene from A. thaliana (56% amino acid identity) and the other seven clones were similar to the TM6 gene from tomato (59% amino acid identity). The obtained sequences were named M. truncatula TM6 (MtTM6, GenBank accession number JN412097) based on its similarity to TOMATO MADS BOX GENE6 (TM6; Pnueli et al., 1991) and M. truncatula NMH7 (MtNMH7, GenBank accession number JN412096) based on its homology with the AP3 ortholog in Medicago sativa (NMH7).
The longest MtTM6 cDNA clone was 1135 bp long and contained a 696 bp open reading- rame encoding a 232 amino acid protein. The longest MtNMH7 cDNA clone was 984 bp long and contained a 687 bp open reading frame encoding a 229 amino acid protein. The MtTM6 genomic sequence indicates that the MtTM6 gene is organized in seven exons of 188, 70, 62, 109, 42, 45 and 183 bp, and six introns (Figure S1a), and that MtNMH7 is organized similarly, with seven exons of 188, 67, 63, 99, 42, 46 and 185 bp, and six introns. These sequences were used as BLASTN (Altschul et al., 1997) queries, and displayed using the Chromosome Visualization Tool (CViT; http://www.medicagohapmap.org/tools/blastform). The MtTM6 gene is located on chromosome 5 in M. truncatula bacterial artificial chromosome AC136451, clone mth2-17d19 (Figure S1b), and the MtNMH7 gene is located on chromosome 3 in M. truncatula bacterial artificial chromosome AC151483, clone mth2-157f20 (Figure S1c). Both genes are present as a single copy as confirmed by Southern blot analysis (Figure S3).
Alignments of the inferred amino acid sequences with selected AP3 orthologs showed that MtTM6 and MtNMH7 show all the common domains of a MADS box MIKC-type protein and the PI-derived motif (consensus FxFRLQPSQPNLH; Kramer et al., 1998; Lamb and Irish, 2003). Moreover, MtTM6 shows the characteristic ancestral paleoAP3 motif in the C-terminal region, while MtNMH7 shows the euAP3 motif. Each of these motifs is highly conserved within lineages (Figure S2a,b).
Phylogenetic analysis reveals an ancient duplication in the ancestral gene of the M. truncatula AP3 lineage
To analyze the duplication event of the M. truncatula AP3-like genes in the context of time, we inferred the phylogeny using a nucleotide dataset containing AP3 homologs from a wide variety of species (Table S1). We included ‘B gene’ (GGM2-like) sequences isolated from diverse gymnosperms, and AP3-like genes from basal angiosperms, magnoliids, monocots, early-divergent eudicots and core eudicots. In agreement with previous analyses (Kramer et al., 1998; Kim et al., 2004), we observed two major core eudicot lineages (euAP3 and TM6) derived from the ancestral paleoAP3 lineage. The topology of the phylogenetic tree confirms that the M. truncatula paralogs isolated here fall within the clade of AP3-like genes, and that both MtNMH7 and MtTM6 are specifically placed within their respective lineages, euAP3 and TM6. Interpretation of this phylogeny suggests that the duplication resulting in the paralogs MtNMH7 and MtTM6 occurred early, coinciding with the base of the core eudicot radiation (Figure 1).
Complementary expression patterns between AP3-like genes in M. truncatula
Northern blot and quantitative RT-PCR analyses revealed that both genes were expressed exclusively in floral tissues but with different expression levels (Figure 2). We did not detect any expression in root nodules, in contrast to what has been described previously for other AP3-like genes from legume species such as Medicago sativa and Glycine max (Heard and Dunn, 1995; Wu et al., 2006; Páez-Valencia et al., 2008).
In situ hybridization analysis during early stages of floral development showed that the M. truncatula AP3-like genes have non-overlapping expression patterns in the second and third floral whorls. MtTM6 transcripts were initially detected at the centre of the floral meristem at stage 2 of floral development (Benlloch et al., 2003), whereas MtNMH7 was not detected at this stage. At stage 4, expression of both genes was observed in the common primordia that differentiate into petal and stamen primordia. Remarkably, we observed a complementary expression pattern at the cellular level at this early developmental stage: MtTM6 transcripts were strongly detected in the outer cell layers that surround the common primordia, whereas MtNMH7 transcripts were located in the inner cell layers of the common primordia (Figure 3a,e). This differential expression pattern persisted throughout the development of petals and stamens until complete flower development (Figure 3b–d,f,g). At late stages, MtNMH7 expression was also detected in the ovules (Figure 3d), while MtTM6 continues to be expressed in the outer cell layers of both floral organs (Figure 3h).
Protein–protein interactions of the M. truncatula B-class MADS box proteins
The maintenance of AP3 and PI expression depends upon the interaction of both gene products (heterodimers) as part of larger MADS box protein complexes (Riechmann et al., 1996).
We previously reported the expression pattern and functional analysis of the two PI-like genes of M. truncatula (Benlloch et al., 2009). MtPI expression was detected at high levels in flowers, whereas MtNGL9 expression was detected at very low levels in flowers and in all tissues analyzed, being slightly higher in nodules and leaves. We re-tested the expression pattern of the two M. truncatula PI paralogs in inflorescence apices, and found that MtNGL9 is weakly expressed in the floral meristem in the cells that give rise to common petal/stamen primordia and later on during the development of petals and stamens. Moreover, MtNGL9 mRNA was also detected in ovules (E. Roque, M. Fares, L. Yenush, J.P. Beltrán, L.A. Cañas, IBMCP, Valencia, unpublished data).
To test the ability of the gene products to interact in vitro, we studied protein–protein interactions among the four M. truncatula B-class MADS box products in pairwise combinations using the yeast two-hybrid system (Figure 4). We found that both MtNMH7 and MtTM6 were able to interact with MtPI. In contrast to MtNMH7, which shows heterodimerization capability with both PI proteins, MtTM6 was able to interact with MtPI. Interestingly, we observed homodimerization only in the case of MtPI. Further studies are in progress to compare these protein–protein interactions quantitatively and confirm them in planta.
Loss-of-function analyses of MtNMH7 and MtTM6
To assess the functional roles of AP3-like genes in M. truncatula, we used two reverse genetic approaches: RNAi gene silencing in stably transformed plants and Tnt1 retrotransposon insertion in the MtTM6 gene (Tadege et al., 2008; Cheng et al., 2011).
For RNAi silencing of MtNMH7, we generated a hairpin RNAi construct using the C-terminal region of the MtNMH7 cDNA (positions 326-734 from the ATG codon), and transformed M. truncatula cv. 2HA plants with this construct. Flowers of the 35S::RNAi-MtNMH7 plants with strong phenotypes (lines Tr13 and Tr14) displayed apparent morphological alterations (Figure 5b). The yellow petals in the second whorl (W2) showed patches of green tissue. The mid-veins of the standard or vexillum (V) became broader and greener, especially toward the edge of the corolla, and the wings (W) and keel (K) petals also showed green patches. In some cases, the green tissue included almost the whole petal, and, in others, the green tissue was present only at the border of petals (Figure 5b, arrows). In the third whorl (W3) the staminal tube with anthers and filaments was formed. However, the anthers fail to produce pollen grains. Moreover, the filaments of the two stamens flanking the staminal tube were green and contained trichomes, and the anthers had lost their characteristic morphology, suggesting that reduction of MtNMH7 expression causes slight stamen carpelloidy (Figure 5b, arrows).
These transformations were more evident at the cellular level, with adaxial second-whorl cells showing sepaloid characteristics. The typical conical epidermal cells present in the wild-type petals (Figure 6a) were replaced by stomata-containing sepal-like epidermal cells, indicating a shift of petal toward sepal identity in the green petal patches (Figure 6b,c). Cells of the 35S::RNAi-MtNMH7 carpelloid stamens flanking the staminal tube (Figure 6g,o) had lost the narrow and longitudinally elongated morphology that is typical of a wild-type anther filament (Figure 6e), and now resembled the cells present in the middle wild-type carpel, with the presence of trichomes (Figure 6f). Moreover, the characteristic irregular epidermal cells found in wild-type anthers (Figure 6i) were replaced by cells that were quite similar to those present at the wild-type ovule surface in the anthers of 35S::RNAi-MtNMH7 flowers (Figure 6j). This change is shown in Figure 6(k).
In summary, reduction in MtNMH7 expression resulted in a partial homeotic transformation of petals into sepal-like mosaic organs, and a slight homeotic conversion of stamens into carpelloid organs. These phenotypic changes indicate that MtNMH7 is involved in petal and stamen development, but its reduced expression is not sufficient to completely abolish the B function.
The transcript level of MtNMH7 in 35S::RNAi-MtNMH7 plants was considerably reduced (15% of wild-type level) but not completely eliminated (Figure S4), whereas the expression levels of MtTM6 and MtNGL9 were similar to the wild-type. MtPI expression was reduced by 30%. These results indicate that MtNMH7 may be required only to maintain full MtPI expression without affecting MtNGL9 or MtTM6 expression, or, alternatively, that 15% of the total MtNMH7 expression is sufficient to directly or indirectly maintain MtNGL9 and MtTM6 expression, but not to fully maintain MtPI expression levels.
To elucidate the function of MtTM6, we used both RNAi silencing and Tnt1 insertion approaches. For RNAi silencing of MtTM6, we generated a hairpin RNAi construct using the C-terminal region of MtTM6 cDNA (positions 444–716 from the ATG codon), and transformed M. truncatula cv. R108 plants. The mttm6-1 Tnt1 insertion allele was identified as described in Experimental procedures. The Tnt1 insertion lies at position 105 of the first exon of MtTM6.
The mttm6-1 flowers (Figure 5c) showed a stronger loss-of-B-function phenotype than the 35S::RNAi-MtTM6 flowers (Figure 5d). The petals of the mttm6-1 flowers had small patches of green tissue at the borders, and, in some cases, a carpel-like structure fused to the edge of the petal (Figure 5c, arrows). The 35S::RNAi-MtTM6 petals exhibited defects only in their shape. Reduction in MtTM6 expression resulted in some stamens differentiating into filaments and anthers that fail to produce pollen grains. Other stamens failed to develop anthers, producing stigma-like structures in their place (Figure 5d, arrow). The filaments were green with some trichomes (Figure 5d). However, total loss of MtTM6 expression caused an almost full conversion of all anthers into carpels. These were broader and greener than wild-type anthers and displayed typical carpel trichomes on their surface (Figure 5c).
Scanning electron microscopy of the border of the second-whorl petals in mttm6-1 plants revealed a conversion of the conical epidermal cells (Figure 6a) into stomata-containing sepal-like epidermal cells (Figure 6b), suggesting a shift of identity from petal towards sepal in this petal part (Figure 6d).
mttm6-1 flowers never developed a staminal tube. Instead, green carpelloid-like structures with trichomes at their surface were formed (Figures 5c and 6q). Closer inspection showed that, while the epidermis of wild-type filaments comprises narrow and longitudinally elongated cells (Figure 6e), the cell types in the mttm6-1 third-whorl organs resembled cells present in the low and middle parts of the wild-type carpel (Figure 6f), which develop trichomes (Figure 6h).
Although cells in the wild-type anthers displayed their typical irregular shape (Figure 6i), the mttm6-1 anthers contain cells similar to those present at the end of the wild-type carpel, in the region most proximal to the stigma (Figure 6l). This transformation is shown in Figure 6(h).
We analyzed the expression of the M. truncatula B-class MADS box genes in the 35S::RNAi-MtTM6 (line Tr1) and mttm6-1 plants by quantitative RT-PCR. The RNAi line showed reduced MtTM6 expression (20% of the wild-type level), and all other B-class genes showed reduced expression levels compared to the wild-type (Figure S4). Similar results were observed for the mttm6-1 line, in which MtTM6 expression disappeared completely as consequence of the Tnt1 insertion in the MtTM6 gene (Figure S4). These results suggest that expression of MtTM6 is required to maintain the expression of the other M. truncatula B-class genes.
B-class MADS box genes have been implicated in the regulation of floral organ identity in the second whorl (petals) and third whorl (stamens) in angiosperms. Recent functional analysis and expression pattern studies of the AP3 and PI lineages have demonstrated that, in various instances, these genes may have swapped roles, retained or lost their ancestral roles, or acquired novel functions. The multiple gene duplication events within both lineages have led to sub- and non-functionalization in some cases, processes that are often associated with changes in expression patterns (Vandenbussche et al., 2004; de Martino et al., 2006; Rijpkema et al., 2006; Drea et al., 2007).
Functional specialization of M. truncatula AP3-like genes
We have generated RNAi-induced MtTM6 and MtNMH7 loss-of-function transgenic plants and identified a complete MtTM6 loss-of-function insertional (Tnt1) mutation. Functional analysis of these plants uncovered interesting aspects of the regulatory control of petal and stamen development in this legume species.
RNAi-induced loss of MtNMH7 function results in defects predominantly in petal development. The 35S::RNAi-MtNMH7 plants showed partial conversion of petals into sepaloid structures, producing petals with patches of green tissue and sepal-like cells, whereas the stamens were weakly transformed into carpel-like organs. Changes in the epidermal cell types of petal and stamens may be explained by protein movement from the inner cells, where the MtNMH7 gene is expressed, to external cell layers. Usually the mRNA and its MADS box protein co-localize in the same cells or tissues, but it is known from studies on the B-class genes DEF and GLOB from A. majus (Perbal et al., 1996) that some MADS domain transcription factors are able to transfer from one cell to another, where they may have a non-cell-autonomous function. Urbanus et al. (2009) determined the in planta protein localization patterns of four MADS domain proteins, and reported that the discrepancies between mRNA and protein patterns occasionally observed suggest a possible non-cell-autonomous action of these factors by intercellular transport via plasmodesmata or alternative control mechanisms. In MtNMH7 RNAi loss-of-function flowers, MtPI expression was reduced, suggesting that MtNMH7 positively regulates MtPI. Alternatively, a low level of MtNMH7 may be sufficient for auto-regulation and to directly or indirectly maintain MtNGL9 and MtTM6 expression, but may be insufficient to maintain the MtPI expression level. If the protein–protein interactions detected reflect those that occur in vivo, the MtTM6/MtPI heterodimeric complex is the most abundant B-class protein complex formed in the 35S::RNAi-MtNMH7 line. Taking into account the low level of MtNMH7 expression in this line, other complexes (MtNMH7/MtNGL9 or MtNMH7/MtPI) may be formed, but they would be insufficient to confer petal identity, indicating the importance of MtNMH7 in petal identity specification. Therefore, MtTM6/MtPI is sufficient to confer almost full stamen identity but is insufficient to confer complete petal identity.
In contrast, the mttm6-1 mutant and RNAi-induced loss-of-MtTM6 function plants produce flowers that display homeotic transformations predominantly in the third whorl. Down-regulation of MtTM6 activity in 35S::RNAi-MtTM6 plants did not result in homeotic defects in petal development. The only phenotypic effect observed was a change in petal shape. The lack of homeotic conversions in petals is the result of low-level residual expression of MtTM6 in these lines as corroborated by the comparison of the phenotypes of MtTM6-RNAi and mttm6-1 lines. Similarly, de Martino et al. (2006), after phenotypic analysis of tomato TM6 RNAi transgenic lines, were unable to exclude the possibility that lack of a homeotic phenotype in petals was the result of low-level residual expression of TM6 in these lines. mttm6-1 plants show almost full conversion of stamens into carpels, and a weak homeotic conversion of petals into sepal-like organs. In both 35S::RNAi-MtTM6 and mttm6-1 flowers, all the B-class genes showed reduced expression compared with the wild-type (Figure S4), suggesting that MtTM6 positively regulates MtPI, MtNGL9 and MtNMH7. In mttm6-1, in which the stamens are almost fully replaced by carpels, two heterodimers are hypothetically active: MtNMH7/MtPI and MtNMH7/MtNGL9. Both complexes are sufficient to confer almost full petal identity, but are not sufficient to confer stamen identity.
MtNMH7 and MtTM6 appear to regulate the other B-class genes at the transcriptional level. Their expression may be required directly or indirectly (perhaps via larger multimeric MADS box protein complexes) in the auto- and cross-regulatory loop, in which formation of a heterodimer between PI/GLO and AP3/DEF members is required to maintain expression of the M. truncatula B-class genes in petals and stamens (Riechmann et al., 1996).
Based on the protein interaction data, the expression patterns of B-class genes in RNAi or Tnt1 lines, as well as the phenotypical differences of individual MtNMH7 or MtTM6 loss-of-function plants, we conclude that both AP3-like genes are necessary for petal and stamen identity. MtNMH7 appears to play a major role in determining petal identity, whereas MtTM6 appears to do the same with respect to stamen identity. However, both genes are implicated in specification of the second and third whorls.
Differences in expression patterns of duplicated M. truncatula AP3-like genes may partially explain their functional specialization
Few studies have investigated the functional divergence process related to duplicated AP3-like genes in the core eudicot species. Data concerning functional analyses of the AP3-like genes are available for Petunia hybrida (PhDEF and PhTM6), Solanum lycopersicum (TAP3 and TM6), Nicotiana benthamiana (NbDEF and NbTM6) and Gerbera hybrida (GDEF2 and GDEF1) (Vandenbussche et al., 2004; de Martino et al., 2006; Rijpkema et al., 2006; Broholm et al., 2010; Geuten and Irish, 2010). euAP3-like genes from these species show similar expression patterns during petal and stamen development, and their loss-of-function phenotypes in tomato (TAP3) and Nicotiana (NbDEF) are quite similar, displaying marked homeotic conversions of petal into sepals and of stamens into carpelloid structures (Liu et al., 2004; de Martino et al., 2006; Geuten and Irish, 2010). In Gerbera, down-regulation of GDEF2 also causes changes in the second and third whorls (Broholm et al., 2010). Unlike euAP3-like genes, some TM6-like genes, such as PhTM6, NbTM6 and GDEF1, are initially expressed in the four floral organs (Vandenbussche et al., 2004; Broholm et al., 2010; Geuten and Irish, 2010), while TM6 is expressed throughout the three inner whorls (de Martino et al., 2006). Expression of TM6-like genes becomes predominant in stamens, and at later stages they are expressed strongly in the developing stamens and carpel (Vandenbussche et al., 2004; de Martino et al., 2006; Broholm et al., 2010; Geuten and Irish, 2010). The suppression of TM6 orthologs is qualitatively different. Loss-of-function of TM6 and GDEF1 produces flowers with weak conversion of stamens into carpels (de Martino et al., 2006; Broholm et al., 2010). Transgenic Nicotiana plants in which NbTM6 expression was suppressed reveal weakly homeotic transformations of stamens into carpelloid structures and also weak defects in petals and ovules (Geuten and Irish, 2010), while loss-of-function of PhTM6 has no obvious phenotype (Rijpkema et al., 2006), suggesting various degrees of functional redundancy in stamen identity of TM6-like genes with their respective euAP3 genes.
MtNMH7 and MtTM6 are expressed in the regions of the floral meristem that give rise to petal and stamens (common primordia), and their expression in these floral organs persists until later stages of development. At early developmental stages, MtTM6 expression was also observed in the central region of the floral meristem, but we did not observe expression of MtTM6 outside the second and third whorls at later developmental stages, in contrast to the expression of their TM6-like orthologs. Moreover, MtNMH7 is expressed in ovules at later stages of development, an expression pattern that may be characteristic of legumes. Expression of the euAP3 gene GmNMH7 in soybean (Glycine max) was also observed in fully differentiated carpel and ovules (Wu et al., 2006). MtNMH7 and MtTM6 genes appear to be expressed at different levels and with a mutually antagonistic expression pattern in different sub-populations of cells within petals and stamens. The MtTM6 transcript was strongly expressed in the outer cell layers that surround the second and third whorls, whereas MtNMH7 was expressed inside them. Our results suggest that coordinated expression of B-class genes in the inner cell layers of the organ primordia plays an important role in specifying petal identity, while their expression in the outer cell layers may be mainly required to specify stamen identity. In agreement with this complementary expression pattern, MtTM6 and MtNMH7 do not appear to have any degree of functional redundancy as described for their orthologs in Petunia, tomato, Nicotiana or Gerbera. These observations suggest that regulatory changes played prominent roles in diversification of their functions. MtNMH7 and MtTM6 may have experienced a sub-functionalization process, concomitant with complete partitioning of the expression pattern of the ancestral gene lineage.
Our results provide information on the similarities and differences in petal and stamen development between core eudicots, and we also provide evidence regarding the evolutionary fate of the duplicated AP3-like genes of a legume species, in which B-function MADS box genes had remained largely uncharacterized.
Medicago truncatula cv. Jemalong lines A17 and R108, Medicago sativa cv. RSY-27 and Pisum sativum cv. Alaska plants were used in this study. Plants were grown in the greenhouse at 22°C (day) and 18°C (night) with a 16 h light/8 h dark photoperiod, in a mixture of soil/sand (3:1) irrigated with Hoagland no. 1 solution (Hewitt, 1966).
Identification of Tnt1 insertion sites in MtTM6 and co-segregation test
Medicago. truncatula R108 was transformed with a construct containing the complete Tnt1 retroelement of tobacco as described previously (d'Erfurth et al., 2003). Tnt1 tagging was performed using one of these transgenic lines (Tnk7-7) to activate Tnt1 transposition. Mutagenized plants that contain multiple independent Tnt1 inserts were regenerated from leaf explants of the starter line as described previously (Tadege et al., 2005, 2008; Cheng et al., 2011). The mttm6-1 allele was identified by screening of a segregating population of approximately 9000 independent R1 lines. For pooled DNA screening, we used primers based on the MtTM6 sequence (MtTM6-F and MtTM6-F1; Table S2) in combination with primers for both sides of the Tnt1 long-terminal repeats (Tnt1-F, Tnt1-F1, Tnt1-R and Tnt1-R1; Table S2). PCR was performed using genomic DNA pools from 9000 Tnt1-carrying transgenic plants (Cheng et al., 2011). A PCR product of 750 bp was obtained from one of these pools using the primers MtTM6-F and Tnt1-F. One part of the PCR product corresponds to the first 105 bp of the coding sequence of the MtTM6 gene. The rest of the amplified sequence corresponds to the border of the long-terminal repeats of the Tnt1. We confirmed that the amplified fragment represents an insertion of the retrotransposon into the M. truncatula MtTM6 gene (Figure S1a). Fifteen T2 plants from the NF1297 line were grown in the greenhouse, and their phenotype was analyzed. Three plants of this population exhibited the mttm6-1 mutant phenotype. PCR analysis was performed in plants showing the wild-type phenotype using primer combinations MtTM6-F/Tnt1-F and MtTM6-F/MtTM6-505 (Table S2) to distinguish heterozygous and wild-type plants. Heterozygous lines were self-pollinated, and one line (NF1297-4) was selected for further analysis. Approximately a quarter of the resultant progeny (5/22 plants) exhibited the floral phenotype described above and co-segregated with the Tnt1 insertion.
Plant nodulation tests were performed in the greenhouse as described previously (Coronado et al., 1995; Benlloch et al., 2009).
Isolation and sequence analysis of cDNA and genomic sequences of MtNMH7 and MtTM6 cDNAs
cDNAs were isolated from a library of M. truncatula A17 inflorescence apices (Benlloch et al., 2006) using the MADS box fragment of the A. majus DEFICIENS gene as a probe. Sequence alignments and similarity comparisons of the inferred proteins were performed using MacVector 9.5 software (MacVector Inc., http://www.macvector.com/). The deduced amino acid sequence was aligned using the CLUSTAL W tool in MacVector 9.5. The genomic sequences of MtNMH7 and MtTM6 were obtained by PCR using genomic DNA as a template and primers MtNMH7-ATG/MtNMH7-766 and MtTM6-ATG/MtTM6-725 respectively (Table S2).
Isolation and sequence analysis of MsTM6, PsNMH7 and PsTM6 cDNAs
MsTM6 (JN412100), PsNMH7 (JN412099) and PsTM6 (JN412098) were obtained by RT-PCR and 3′ RACE from M. sativa and P. sativum floral cDNA samples. TM6-like genes from M. sativa and P. sativum were obtained using primers designed against the mRNA of the M. truncatula MtTM6 gene (MtTM6-ATG and MtTM6-622; Table S2). The 3′ UTR and C-terminal region were obtained from the M. sativa and P. sativum floral cDNAs using the 3′ RACE system for rapid amplification of cDNA ends (Invitrogen, http://www.invitrogen.com/) and nested gene-specific primers MsTM6-362 and PsTM6-362, respectively (Table S2). The euAP3-like gene from P. sativum was obtained using primers designed against the mRNA for the M. truncatula MtNMH7 gene (MtNMH7-ATG and MtNMH7-671; Table S2). The 3′ UTR and C-terminal region were obtained from the P. sativum floral cDNA using the 3′ RACE system for rapid amplification of cDNA ends (Invitrogen) and the nested gene-specific primer PsNMH7-384 (Table S2). Sequence alignment and similarity comparisons of the inferred proteins were performed using MacVector 9.5 software (MacVector Inc.). The deduced amino acid sequence was aligned using the CLUSTAL W tool in MacVector 9.5, and further refined by hand.
The phylogenetic tree was inferred by the neighbor-joining method using Poisson-corrected amino acid distances. A total of 1000 bootstrap pseudo-replicates were used to stimate reliability of internal nodes. The tree was rooted using five PI-like sequences: AtPI, AmGLO, NtGLO, PhGLO1 and PhGLO2. This choice was based on the fact that the AP3 and PI are known to be paralogous lineages and are therefore each other's natural outgroups. Tree inference was performed using MEGA version 5 (Tamura et al., 2007). The dataset comprised 74 previously reported AP3-like genes obtained from GenBank, and five AP3-like sequences that are not found in GenBank. All sequences used in this analysis, with their GenBank accession numbers and respective species, are listed in Table S1.
Real-time RT-PCR analysis
Total RNA was isolated from roots, nodules and inflorescences of wild-type M. truncatula A17 plants and inflorescences of 35S::RNAi-MtTM6, 35S::RNAi-MtNMH7 and mttm6-1 plants using an RNeasy plant mini kit (Qiagen, http://www.qiagen.com/) according to the manufacturer's instructions. RNA samples were measured and equalized, and their integrity was analyzed by gel electrophoresis. Total RNA was treated using rDNaseI of the DNase treatment and removal kit (Ambion, http://www.invitrogen.com/site/us/en/home/brands/ambion.html?CID=fl-ambion). Primers were designed for all genes from the 3′ end of the gene using Primer Express version 2.0 (Applied Biosystems, http://www.appliedbiosystems.com/) with default parameters. For first-strand synthesis, total RNA (1 μg) was reverse-transcribed in a 20 μl reaction mixture using the PrimerScript 1st strand cDNA synthesis kit (Takara, http://www.takara-bio.com/). A 1 μl aliquot of the reverse transcription reaction was used for real-time RT-PCR analysis with 300 nM of each primer mixed with Power SYBR® Green PCR master mix (Applied Biosystems) according to the manufacturer's instructions. The reaction was performed in 96-well optical reaction plates using an ABI PRISM 7500 sequence detection system and appropriate software (Applied Biosystems). The relative levels were determined by the method. To normalize the variance among samples, ‘Secret Agent’ (O-linked N-acetylglucosamine transferase, TC77416; Hartweck et al., 2002) was used as an endogenous control. All reactions were performed in triplicate, with biological replicates for each sample. The primers used were MtNGL9-qRTDIR/MtNGL9-qRTREV, MtPI-qRTDIR/MtPI-qRTREV, MtNMH7-qRTDIR/MtNMH7-qRTREV, MtTM6-qRTDIR/MtTM6-qRTREV, and Sec.Ag-qRTDIR/Sec.Ag-qRTREV (Table S2).
Southern blot hybridization
Plant genomic DNA was extracted from leaves of M. truncatula A17 as described by Dellaporta et al. (1983), and 10 μg of DNA were digested with EcoRI, BamHI and HindIII, separated on 0.7% agarose gels in Tris acetate EDTA buffer pH 8 (40 mm Tris-acetate, 2 mm sodium EDTA) overnight at 1 V/cm, and transferred to a nylon membrane. Southern blot hybridization was performed by the standard method at 52 and 65ºC. cDNA probes were generated by PCR using primers MtNMH7-289/MtNMH7-734 and MtTM6-362/MtTM6-698 (Table S2).
The B-class MADS box transcription factors were sub-cloned into the GAL4-based two-hybrid vectors pGBD-C2 (James et al., 1996) and pACT-2 (Clontech, http://www.clontech.com/) and transformed into the PJ69-4A yeast strain (James et al., 1996). In accordance with the genotype of the PJ69-4A strain, positive interactions are indicated by growth in SD selective medium lacking histidine and adenine (SD-H-A). Interactions between the transcription factors were tested by spotting a 1:200 dilution of each strain onto minimal medium (SD) plates and recording growth after 3–5 days. Identical results were observed for three independent transformants.
RNA in situ hybridization
RNA in situ hybridization with digoxigenin-labeled probes was performed on 8 μM longitudinal paraffin sections of M. truncatula inflorescences as described previously (Ferrándiz et al., 2000). The RNA antisense and sense probes were generated using the T7 and SP6 polymerases, respectively, using a 337 bp fragment of MtTM6 (positions 362–698 from the ATG codon) and a 445 bp fragment of MtNMH7 (positions 289–734 from the ATG codon) cloned into the pGEM-T Easy vector (Promega, http://www.promega.com/).
Generation of transgenic plants
Transformation of M. truncatula R108 was performed as described previously (d'Erfurth et al., 2003). The 35S::RNAi-MtNMH7 construct was prepared using the 326 bp fragment of MtNMH7 (positions 326–734 from the ATG codon), amplified using primers MtNMH7-DIR-RNAi and MtNMH7-REV-RNAi (Table S2) that incorporate two restriction sites that are used for cloning into the pHANNIBAL vector (Wesley et al., 2001). The 35S::RNAi-MTM6 construct was prepared using the 278 bp fragment of MtTM6 (positions 444-716 from the ATG codon), amplified using primers MtTM6-DIR-RNAi and MtTM6-REV-RNAi (Table S2).
Northern blot analysis
Total RNA (15 μg) was isolated from frozen leaves, roots, root nodules, stems and flower buds by phenol/chlorophorm extraction and precipitated using 3 M lithium chloride. RNA electrophoresis was performed in formaldehyde/agarose gels, transferred to Hybond N+ membranes (Amersham Biosciences, http://www.gelifesciences.com), and hybridized with 32P-labeled probes under standard conditions. The probes used were a 337 bp fragment of MtTM6 (positions 362–698 from the ATG codon), amplified using primers MtTM6-362 and MtTM6-698 (Table S2), and a 445 bp fragment of MtNMH7 (positions 289–734 from the ATG codon), amplified using primers MtNMH7-289 and MtNMH7-734 (Table S2).
Light microscopy and cryo-SEM
Images of wild-type, 35S::RNAi-MtTM6; 35S::RNAi-MtNMH7 and MtTM6-Tnt1 (mttm6-1) flowers were obtained using a Leica MZ28 stereomicroscope (http://www.leica.com/). For cryo-SEM, samples were frozen in slush nitrogen and attached to the specimen holder of a CT-1000C cryo-transfer system (Oxford Instruments, http://www.oxford-instruments.com/) interfaced with a JEOL JSM-5410 scanning electron microscope (http://www.jeol.com/). The samples were then transferred from the cryostage to the microscope sample stage, where the condensed surface water was sublimed by controlled warming to –85ºC. Afterwards, the sample was transferred back to the cryostage for gold coating by sputtering. Finally, the sample was returned to the microscope sample stage and viewed at an accelerating voltage of 15 keV.
This work was funded by grants BIO2006-09374 and BIO2009-08134 from the Spanish Ministry of Science and Innovation. We are gratefully to Mario A. Fares and Santiago F. Elena (Instituto de Biología Molecular y Celular de Plantas, Valencia, Spain) for helpful comments and bioinformatics support. The collaboration and assistance of Rafael Martínez-Pardo in the greenhouse is gratefully acknowledged.
The GenBank accession numbers for the MtTM6 sequence and the MtNMH7 sequence are JN412097 and JN412096.