Different outcomes for the MYB floral symmetry genes DIVARICATA and RADIALIS during the evolution of derived actinomorphy in Plantago

Authors


Author for correspondence:

Jacqueline M. Nugent

Tel: +353 1 708 3857

Email: Jackie.Nugent@nuim.ie

Summary

  • The gene network that specifies flower shape in Antirrhinum majus (bilateral floral symmetry or zygomorphy) includes two MYB-class genes – RADIALIS (RAD) and DIVARICATA (DIV). RAD is involved in establishing the dorsal identity program and its role is to regulate the domain of activity of DIV (the ventral identity program) by restricting it to ventral regions of the flower.
  • Plantago is in the same family as Antirrhinum but has small, radially symmetrical (actinomorphic) flowers derived from a zygomorphic ancestral state. Here we investigate the MYB-class floral symmetry genes and the role they have played in the evolution of derived actinomorphy in Plantago lanceolata.
  • A DIV ortholog (PlDIV) but no RAD ortholog was identified in P. lanceolata. PlDIV is expressed across all petals and stamens later in flower development, which is consistent with the loss of RAD gene function. PlDIV expression in anther sporogenous tissue also suggests that PlDIV was co-opted to regulate cell proliferation during the early stages of pollen development.
  • These results indicate that evolution of derived actinomorphy in Plantago involved complete loss of dorsal gene function, resulting in expansion of the domain of expression of the ventral class of floral symmetry genes.

Introduction

The floral symmetry gene network is best understood in the model plant Antirrhinum majus, where bilateral symmetry (zygomorphy) is exhibited in the distinct shape of dorsal, lateral and ventral petals and by dorsal stamen abortion (Coen, 1996; Luo et al., 1996). CYCLOIDEA (CYC) and DICHOTOMA (DICH) are key genes that determine the shape of dorsal floral organs (in particular the petals) in Antirrhinum (Luo et al., 1996, 1999; Almeida et al., 1997; Galego & Almeida, 2002). CYC and DICH are paralogous ECE-CYC2-class genes of the TCP (TEOSINTE BRANCHED 1 (TB1), CYCLOIDEA (CYC) and PROLIFERATING CELL NUCLEAR ANTIGEN FACTOR (PCF)) transcription factor family and are thought to have a role in regulating cell proliferation and expansion (Cubas et al., 1999a; Howarth & Donoghue, 2006; Martin-Trillo & Cubas, 2009). In Antirrhinum, both CYC and DICH (CYC2 genes) are expressed in the dorsal region of very early flower meristems and expression of both genes is maintained throughout flower development in dorsal floral organs (Luo et al., 1999). In cyc/dich double mutants, Antirrhinum flowers lose their distinct bilateral symmetry and become radially symmetrical (actinomorphic) and all petals and stamens resemble the ventral petals and stamens of wild-type flowers (Luo et al., 1999; Corley et al., 2005). Two additional genes, DIVARICATA (DIV) and RADIALIS (RAD), act with the CYC2 genes to specify flower shape in Antirrhinum. Both DIV and RAD proteins belong to the MYB protein superfamily (Galego & Almeida, 2002; Corley et al., 2005). CYC2 genes positively regulate RAD expression in the dorsal regions of the flower and rad mutant flowers resemble cyc/dich mutant flowers in their phenotype (Corley et al., 2005). DIV is required to specify organ shape in ventral regions of the flower and in div mutant flowers the ventral petals adopt lateral petal shape (Almeida et al., 1997; Galego & Almeida, 2002). DIV is transcribed throughout the floral meristem; however, DIV activity is repressed in dorsal regions, that is, where the domains DIV, CYC2 and RAD gene expression overlap, by RAD binding to a co-regulator of DIV transcriptional activity called DIV- and RAD-interacting factor (DRIF) and sequestering it in the cytoplasm (Raimundo et al., 2013). When dorsal and ventral identity programs are eliminated in a cyc/dich/div triple mutant, the default lateral floral organ identity program is revealed in A. majus (Almeida et al., 1997; Corley et al., 2005).

Actinomorphy is considered the ancestral floral shape condition in flowering plants (Doyle & Endress, 2000; Endress & Doyle, 2009). However, phylogeny mapping suggests that zygomorphy has evolved independently at least 70 times during angiosperm evolution: once in basal angiosperms, at least 23 times independently in monocots and at least 46 times independently in eudicots (Citerne et al., 2010). In particular, zygomorphy is more prevalent within some of the more recently derived, species-diverse lineages of flowering plants (Ronse De Craene et al., 2003; Sargent, 2004). A surprising aspect of zygomorphy is that, although it evolved independently many times, on several occasions it did so by co-opting the same CYC2-class genes into the floral symmetry gene network. Arabidopsis thaliana produces radially symmetric flowers and, although an A. thaliana CYC2 gene is expressed in dorsal regions of the early flower meristems, expression is not maintained during later stages of flower development (Cubas et al., 2001). Maintenance of CYC2 gene expression in dorsal floral organs throughout flower development correlates very strongly with zygomorphy in a range of diverse angiosperms including Lamiales, Fabales, Brassicales, Asterales, and Malpighiales (Hileman et al., 2003; Feng et al., 2006; Busch & Zachgo, 2007; Broholm et al., 2008; Kim et al., 2008; Wang et al., 2008; Zhou et al., 2008; Zhang et al., 2010; Chapman et al., 2012). CYC2 genes, therefore, are a conserved feature of floral symmetry gene networks across diverse plant taxa as a result of parallel evolutionary processes.

Molecular phylogenetics places Plantago and Antirrhinum within a highly supported clade within the Lamiales, sensu lato called the Veronicaceae (Olmstead et al., 2001) or the Plantaginaceae (Rønsted et al., 2002). Despite their relatively close phylogenetic relatedness, these genera produce strikingly different flowers and have adopted very different reproductive strategies (Reeves & Olmstead, 1998; Reardon et al., 2009). In contrast to the large zygomorphic, insect-pollinated flowers of Antirrhinum, Plantago flowers are small and actinomorphic and are wind-pollinated (Reardon et al., 2009). Plantago species are also gynodioecious, that is, both female and hermaphrodite plants coexist within a population: females are effectively male-sterile, whereas hermaphrodites have functional female and male reproductive organs (van Damme & van Delden, 1982; De Haan et al., 1997). The genetics of sex determination in plants involves both nuclear and cytoplasmic gene interactions – mitochondrial genes cause male sterility (female flowers) and nuclear genes restore male fertility (hermaphrodite flowers) (Delph et al., 2006). Thus, mitochondrial-to-nuclear retrograde signaling has a profound influence on flower morphology (Carlsson et al., 2008). Two female types (male sterile 1 (MS1) and male sterile 2 (MS2)) have been described for Plantago lanceolata (van Damme & van Delden, 1982; De Haan et al., 1997).

Derived actinomorphy in a range of flowering plant genera including Cadia (Citerne et al., 2006), Bournea (Zhou et al., 2008), Linaria (Cubas et al., 1999b) and several genera within the Malpighiaceae (Zhang et al., 2013) has been shown to correlate with changes in the domain of expression of CYC2 genes. Gene loss events have also been implicated as facilitators of flower shape change. Preston et al. (2011) have proposed that derived actinomorphy in Plantago is the result of loss of the entire floral symmetry gene network (CYC2, DIV and RAD). They showed that a CYC2 gene loss event (A-clade CYC) correlates with derived actinomorphy in Plantago and is associated with expansion of the domain of expression, and possible neofunctionalization, of a second CYC2 gene in Plantago major (B-clade PmCYC). In a previous study we reported a single CYC2 gene in P. lanceolata (B-clade PlCYC) that has no dorsal-specific expression during flower development and has a much more restricted domain of expression compared with PmCYC (Reardon et al., 2009). The two studies are consistent with loss of A-clade CYC correlating with the evolution of derived actinomorphy in Plantago but suggest that B-clade CYC gene regulation followed different trajectories in different Plantago lineages: some losing B-clade CYC expression during early stages of flower development (PlCYC; Reardon et al., 2009) and others maintaining a broad pattern of B-clade CYC expression during flower development (PmCYC; Preston et al., 2011). The aim of this study was to investigate the fate of the MYB-class floral symmetry genes DIV and RAD and the role they may have played in the evolution of derived actinomorphy in P. lanceolata.

Materials and Methods

Plant material

Fresh P. lanceolata L. and P. major L. leaf and inflorescence tissue used for gene cloning experiments was obtained from plants growing wild at the National University of Ireland, Maynooth, Ireland. In situ hybridization was carried out on inflorescence tissue obtained from P. lanceolata plants growing wild at the University of Edinburgh, Scotland and on inflorescence tissue obtained from segregating hermaphrodite and MS1 (female) plants generated by crossing an MS1 maternal parent with a hermaphrodite paternal parent (Cross B: NPZH 22-9 X FREQ 1-13; results in a 1 : 1 segregating population of hermaphrodite and MS1 plants). Fresh Digitalis purpurea and Veronica longifolia leaf tissue was obtained from plants grown from seed obtained from Kings Wholesale Seed Merchants and Growers, Essex, UK. Fresh Streptocarpus prolixus leaf tissue was obtained from plants growing in the National Botanic Gardens, Glasnevin, Dublin.

DIV and RAD gene isolation

Plantago lanceolata DIV-like PCR products were generated using MYB1F and DIV1R primers (Supporting Information Table S1) using inflorescence cDNA isolated from total RNA as a template. PCR cycling conditions were 94°C for 1 min; 40 cycles of 94°C for 45 s, 45°C for 90 s, and 72°C for 3 min; a final extension at 72°C for 10 min and a 4°C hold. Plantago major DIV-like PCR products were generated using PlDIVqPCRF and DIV1R primers (Table S1) using genomic DNA isolated from leaf tissue as the template DNA. PCR cycling conditions were 95°C for 1 min; 35 cycles of 95°C for 45 s, 50°C for 30 s, and 72°C for 45 s; followed by a final extension at 72°C for 10 min and a 4°C hold. PCR products were cloned into either the pCR2.1 or pCR4 TOPO cloning vector and sequenced. DNA sequences were assessed for similarity to A. majus DIV by BLAST analysis. Sequence alignments were generated using the profile alignment option in ClustalX (Thompson et al., 2002; GenBank references, Table S2). The final alignments were manually inspected to ensure no ambiguities existed. The optimum model of molecular evolution (HKY+G) was determined using modelgenerator v0.85 (Keane et al., 2006). One hundred bootstrap replicates were carried out with the appropriate model of evolution using the software program phyml (Guindon & Gascuel, 2003) and summarized using the majority-rule consensus method. Alternative tree topologies were assessed using the approximately unbiased test for statistical differences in alternative phylogenetic trees implemented in the software package consel (Shimodaira, 2002). A PCR strategy was also used to amplify RAD gene orthologs from P. lanceolata; D. pupurea, Vlongifolia, and S. prolixus template DNAs were used as positive controls. Initially PCR was carried out using various combinations of forward and reverse RAD gene-specific primers (Table S1) and using genomic DNA isolated from leaf tissue as the template DNA. All PCR reactions were repeated using P. lanceolata inflorescence cDNA as the template. The RAD-like gene family was more broadly assessed in Planceolata using the primer combinations designed by Boyden et al. (2012) (GB-RAD primers; Table S1). PCR products were cloned into the pCR4 TOPO cloning vector and 34 confirmed recombinants, with inserts of the expected size, were sequenced. The sequences were cropped to remove all primer sequences. Nucleotide and amino acid alignments were performed manually using only the unique sequences. A nucleotide alignment was used to construct a phylogeny in paup using the maximum likelihood (ML) method with the Generalised time-reversible (GTR) model of sequence evolution (GenBank references, Table S3). One hundred bootstrap replicates were also performed.

Quantitative RT-PCR

Total RNA was extracted from mature stamens (late-stage flowers: petals were enlarged and folded back, and stamens were exserted) and whole inflorescences from P. lanceolata hermaphrodite and MS1 plants using an RNeasy Plant Mini Kit (Qiagen). cDNA synthesis was carried out using 1 μg of total RNA using a Transcriptor High Fidelity cDNA Synthesis Kit (Roche). Tissue-specific qRT-PCR analyses were performed using the SYBR-Green methodology and gene-specific primers (Table S1). Three biological replicates (including three technical replicates for each) were included for each tissue type. The levels of PlCYC and PlDIV transcripts were normalized to the level of a CALMODULIN (CAM) control. CAM was chosen as the control gene after preliminary qRT-PCR experiments showed that the CAM primers gave very consistent cycle threshold (Ct) means when used across a dilution series of petal, stamen and inflorescence cDNAs. Relative gene expression levels were graphed using the iQ5 Optical System Software version 1.0 (Bio-Rad). All classes of qPCR products were cloned and sequenced to confirm the identity of the amplified gene products. PlCAM has GenBank accession number KF964144.

In situ hybridization and toluidine blue staining

RNA in situ hybridization was carried out as described previously (Reardon et al., 2009). PlDIV, PlCYC and P. lanceolata HISTONE4 (PlH4, a marker for cells in the DNA replication phase of mitosis) gene-specific regions were cloned into the pCR®II-TOPO® vector (Invitrogen). PlH4 has GenBank accession number KF964141. Antisense and sense DIG-labeled riboprobes were generated by transcription from the SP6 and T7 promoters using a DIG RNA Labeling Kit (SP6/T7; Roche). Wax-embedded inflorescence tissue from hermaphrodite and MS1 plants (similar to that used for in situ hybridization experiments) was sectioned (7-μM sections) and stained with 1% toluidine blue for 5 min before the sections were washed in dH20.

Results

DIV- and RAD-like genes in Plantago

Initially, two Planceolata DIV-like sequences were cloned and sequenced using PCR and primers specific to regions of the conserved MYBI and MYBII domains of DIV (Galego & Almeida, 2002). Phylogenetic analyses of these and related sequences in the databases showed one of these DIV-like sequences nested within a well-supported clade that includes the Antirrhinum DIV and DIV-like genes, suggesting that this sequence is from a P. lanceolata DIV ortholog (PlDIV; Fig. 1a). A P. major DIV ortholog was also identified using PCR and primers based on the PlDIV sequence (PmDIV; Fig. 1a). Broadly speaking, the inferred gene relationships within the DIV clade (Fig. 1a) are consistent with known species relationships. The main exception is the inferred sister relationship between Plantago/Aragoa and Digitalis, which has relatively low bootstrap support (Fig. 1a). Other more extensive gene phylogenies support a relationship between Plantago/Aragoa and Wulfenia/Veronica with Digitalis the outgroup to these genera (Preston et al., 2011). However, there is no statistical difference (< 0.05) between the log likelihood value of this tree (−4769.80, Fig. 1a) and that of an alternative tree with Plantago/Aragoa and Wulfenia/Veronica as sister groups to the exclusion of Digitalis (−4771.30); thus, either tree topology is equally likely. The second P. lanceolata DIV-like sequence identified in this study (PlDIV2; Fig. 1a) is similar to other more distantly related DIV paralogs previously identified in Plantago coronopus and P. lanceolata (Preston et al., 2011).

Figure 1.

DIVARICATA (DIV)-like and RADIALIS (RAD)-like genes in Plantago lanceolata and related species. (a) Maximum likelihood (ML) tree showing the relationships between representative DIV-like loci. Bootstrap values are displayed at selected branches. (b) PCR amplification of a RAD gene region in Digitalis purpurea, Veronica longifolia and Streptocarpus prolixus but not in P. lanceolata. (c) ML tree showing the relationships between RAD-like loci in Plantago and Antirrhinum majus (Veronica intercedens (Vi) and Digitalis pupurea (Dp) sequences are also included in the RAD2 clade). Bootstrap values are displayed at selected branches. The deduced amino acid sequence for each nucleotide sequence is shown at the relevant branch tip.

Several primer combinations and a range of PCR conditions were used to try to amplify a RAD gene ortholog from P. lanceolata. All attempts to generate a convincing PCR product failed, this despite the fact that PCR products of the expected size were easily obtained from two other Veronicaceae species (D. purpurea and V. longifolia) and one non-Veronicaceae species (S. prolixus; Fig. 1b). The same primer combinations and range of PCR conditions also failed to generate a PCR product from P. lanceolata inflorescence cDNA. These data suggested the absence of a RAD gene ortholog in P. lanceolata. A more comprehensive investigation of RAD-like genes in P. lanceolata was carried out, using more universal, gene-family-specific PCR primers designed by Boyden et al. (2012). Three classes of P. lanceolata RAD-like sequences were identified using this approach (Fig. 1c): (1) sequences that fall within the RAD1-clade of RAD-like genes identified by Boyden et al. (2012); (2) sequences that fall within the RAD3-clade of RAD-like genes identified by Boyden et al. (2012); (3) sequences that are most similar to A. majus RAD-like5 (Baxter et al., 2007; Fig. 1c; designated as a RAD4 clade). None of the P. lanceolata sequences fall within the RAD2 clade that includes A. majus RAD and other RAD gene orthologs (Fig. 1c; Boyden et al., 2012). This despite the fact that, in the Dipsacales, the same PCR primers selectively amplified more RAD2-clade sequences than any other RAD-like sequences (Boyden et al., 2012). These data strongly support the conclusion that P. lanceolata does not have a RAD gene ortholog and confirm previous work that reported the absence of this gene in a range of Plantago species (Preston et al., 2011).

PlDIV is expressed in all petals and stamens during late-stage flower development in P. lanceolata

RNA in situ hybridization was carried out to establish the spatial and temporal pattern of expression of PlDIV during flower development in P. lanceolata (Fig. 2a). The early pattern of PlDIV expression was very similar to that reported for DIV in A. majus (AmDIV; Galego & Almeida, 2002). A low level of PlDIV expression was seen in the apical inflorescence meristem (Fig. 2b); PlDIV was expressed throughout early floral meristems before the initiation of floral organs and was transiently expressed in the bract primordia that subtend floral meristems (Fig. 2b,c); PlDIV was expressed in all early developing floral organs (Fig. 2d) before being down-regulated in sepals (Fig. 2e). However, the later pattern of PlDIV expression differed from that reported for AmDIV. PlDIV was initially expressed across all tissues of all four petals (Fig. 2e); as petal development proceeded, expression became restricted to the lateral edges of petals (Fig. 2g) until eventually expression ceased completely in these floral organs (Fig. 2h,i). The most distinguishing aspect of later PlDIV expression was the extremely high level of mRNA accumulation seen in anther tissue throughout stamen development; a low level of PlDIV expression was also maintained within the ovary of the carpel (Fig 2g–i). No hybridization signal was detected in tissue sections probed with the same amount of PlDIV sense control probe (Fig. 2f).

Figure 2.

Plantago lanceolata DIVARICATA (PlDIV) is expressed in the inflorescence meristem and in all lateral organs. (a) Hermaphrodite P. lanceolata inflorescence – the flowers are protogynous; young flowers with exserted stigmas are located toward the top of the inflorescence, and older flowers with exserted stamens are located toward the bottom of the inflorescence. (b) Longitudinal section through the main inflorescence apex hybridized with a PlDIV antisense probe. PlDIV is expressed in the inflorescence meristem (IM), throughout early flower meristems (FM) and transiently in bract primordia (B). (c–e) PlDIV is expressed in all developing floral organs (S, sepals; P, petals; St, stamens; C, carpels), before being down-regulated in the sepals (e). (f) Similar stage tissue sections probed with the PlDIV sense control probe. (g–i) Later in flower development PlDIV expression ceases in petals, although expression is maintained for longer at lateral petal edges (P). A high level of PlDIV expression is maintained in the anthers of developing stamens (A), and a low level of expression is also maintained throughout ovule development within the carpel (O).

In order to understand the role PlDIV might play in anther tissue, gene expression was assessed more closely at different stages of anther and pollen development. First, anther and early pollen development was examined in P. lanceolata by staining semi-thin tissue sections with toluidine blue. A developmental series showing representative pre-meiotic, meiotic and post-meiotic anther developmental stages is presented in Fig. 3 (a–f). Stage-matched anther sections probed with PlDIV are shown in Fig. 3 (g–l). PlDIV was highly expressed in the microspore mother cells during pre-meiotic stages of anther development (Fig. 3g,h). PlDIV expression decreased and eventually ceased in the microspore mother cells during meiosis but expression was maintained in the tapetal cell layer (Fig. 3i,j). A high level of PlDIV expression was maintained within the tapetal cells up until the time at which tetrads were formed (Fig. 3k), after which expression decreased in tapetal cells and ceased completely by the time the haploid microspores were released (Fig. 3l). This pattern of expression suggests that PlDIV is specifically expressed in cells that are proliferating and undergoing mitosis during anther development. This was confirmed by carrying out in situ hybridization using PlH4 as a marker for cell division. The congruence between the pattern of PlDIV and PlH4 expression during anther development is striking. Like PlDIV, PlH4 was expressed in the sporogenic tissue until the microspore mother cells entered meiosis and in the tapetum cells up until the tetrad stage (Fig. 3m–q). These data suggest that both AmDIV and PlDIV have a conserved role in regulating cell proliferation – albeit cell proliferation in petals in Antirrhinum (Galego & Almeida, 2002) and in stamens in P. lanceolata.

Figure 3.

Plantago lanceolata DIVARICATA (PlDIV) is expressed during the early stages of pollen development in hermaphrodite flowers. (a–f) P. lanceolata anther sections stained with 1% toluidine blue showing representative pre-meiotic (a, b), meiotic (c, d) and post-meiotic (e, f) stages of pollen development. PlDIV is expressed in primary sporogenous tissue and in the tapetum and microspore mother cells within developing locules (g, h). Expression ceases in the microspore mother cells during meiosis (i, j). Expression is maintained within the tapetal cells until tetrads are formed (k) and ceases by the time the haploid microspores are released (l). The pattern of PlH4 expression closely resembles that of PlDIV (m–r). Ms, microspore mother cells; T, tapetum; Tds, tetrads; Msp, microspores.

Premature down-regulation of PlDIV expression in MS1 (female) flowers

As the expression data suggested that PlDIV may have a role in regulating cell proliferation during stamen development, we examined PlDIV expression in P. lanceolata MS1 plants. Flower development is normal in MS1 plants except that stamen filaments are shorter than in hermaphrodites and the anthers do not produce pollen (Fig. 4a–c; van Damme & van Delden, 1982). Real-time qPCR analysis showed no significant difference in the level of transcription of PlDIV in mature stamens of hermaphrodite and MS1 plants. However, PlDIV expression levels were on average 50% reduced in MS1 inflorescences compared with hermaphrodites (Fig. 4d). PlCYC transcript levels were also assessed as an internal control and were similar in the two inflorescence types (Fig. 4d). Because PlDIV expression levels appeared reduced in MS1 inflorescences, in situ hybridization was carried out to see how this might be reflected in the tissue-specific pattern of expression of the gene. The pattern of PlDIV expression in MS1 inflorescence tissue was similar to that seen in hermaphrodites up to the point when microspore mother cells entered meiosis (Fig. 5a,b). After the commencement of meiosis in MS1 anthers, PlDIV expression decreased significantly in the sporogenic tissue and also in the tapetal cells (Fig. 5c,d). Normal pollen development does not proceed beyond this point in MS1 anthers: anther locules became more flaccid than in hermaphrodites, the sporogenous tissue within the locules begins to disintegrate and tetrads and microspores are not produced (Fig. 5e). Again, the pattern of PlH4 expression mirrored that of PlDIV in MS1 anthers; PlH4 was expressed in the sporogenic tissue and in the tapetal cells until the microspore mother cells entered meiosis (Fig. 5f,g) after which point PlH4 expression ceased in both these tissues (Fig. 5h,i). These data demonstrate premature down-regulation of PlDIV expression in tapetal cells in P. lanceolata MS1 flowers and that this down-regulation correlates with the failure of microspore mother cells to complete meiosis. In P. lanceolata MS1 flowers, PlCYC was expressed in anther connective tissue in a pattern similar to that reported previously for P. lanceolata (Fig. 5j–l; Reardon et al., 2009).

Figure 4.

Plantago lanceolata DIVARICATA (PlDIV) is prematurely down-regulated during the early stages of pollen development in male sterile 1 (MS1) flowers. (a) Mature Plantago lanceolata hermaphrodite flower – the petals are folded back and the mature stamens are exserted. (b, c) Mature P. lanceolata MS1 flowers: (b, c): the bracts and sepals have been removed; (c) a petal has been removed to reveal the short MS1 stamen filaments. S, stigma; A, anther; F, filament; P, petal. (d) Quantitative RT-PCR analyses showing relative PlCYC and PlDIV expression in P. lanceolata hermaphrodite (green) and MS1 (orange) inflorescence tissue. Error bars indicate ± 1 SD.

Figure 5.

Plantago lanceolata DIVARICATA (PlDIV) expression is prematurely down-regulated in Plantago lanceolata male sterile 1 (MS1) anthers. (a–e) Expression pattern of PlDIV in MS1 anthers. Developmentally matched sections are shown for two different inflorescences (a1–e1 and a2–e2). After the commencement of meiosis, PlDIV expression decreases in the sporogenic tissue and in the tapetal cells (c, d) and normal pollen development does not proceed beyond this point (e). (f–i) Expression pattern of PlH4 in MS1 anthers: PlH4 is expressed in the sporogenic tissue and in the tapetal cells until the microspore mother cells enter meiosis (f–g), then PlH4 expression ceases in both tissues (h–i). PlCYC is expressed in MS1 anther connective tissue (j–l). Ms, microspore mother cells; T, tapetum; CT, connective tissue.

Discussion

This study shows that Plantago retains just one of the MYB genes (DIV) that make up the flower shape gene regulatory network found in A. majus. AmDIV has a role in regulating cell division in the ventral petals and adjacent regions of lateral petals, giving these petals their distinctive shape (Galego & Almeida, 2002; Corley et al., 2005). While the early pattern of PlDIV expression in P. lanceolata is similar to that reported for AmDIV (Galego & Almeida, 2002), the two genes differ in their later pattern of expression. Later PlDIV expression in hermaphrodite and MS1 flowers suggests that PlDIV has been redeployed to regulate cell proliferation during stamen development rather than petal development in P. lanceolata. Although the MYB floral symmetry genes, RAD and DIV, have not been studied as extensively as the CYC2 genes in species other than A. majus, their role appears to be conserved at least across the Lamiales (Zhou et al., 2008). For example, Bournea leiophyll flowers have a single plane of symmetry (zygomorphy) during early floral organ initiation but transition to polysymmetry (actinomorphy) by the time the floral organs are mature. This switch from bilateral to radial symmetry at anthesis is achieved by the down-regulation of BlRAD in dorsal floral regions later in flower development and is associated with expression of BlDIV across all petals and stamens (Zhou et al., 2008). The later pattern of PlDIV expression shares similarities with the expression pattern reported for BlDIV. In B. leiophyll, BlDIV is expressed later in flower development at the lateral edges of all petals and in all stamens before expression is down-regulated in both organ types (Zhou et al., 2008). B. leiophyll flowers have five near-equal size stamens and it has been suggested that BlDIV expression in all stamen primordia may be involved in synchronizing stamen growth (Zhou et al., 2008). Plantago lanceolata has four stamens of equal size (Fig. 4a) and continuous expression of PlDIV in all stamen primordia may also be synchronizing growth in these organs. Uniquely in P. lanceolata, though, a very high level of PlDIV expression is maintained within stamen sporogenic cells until the stage when microspore mother cells enter meiosis and in tapetal cells until the microspores are released; this is not seen for either the AmDIV or BlDIV genes. The pattern of PlH4 expression very closely mirrors that of PlDIV (Fig. 3g–r). In several plant species, cytoplasmic male sterility (CMS) is associated with tapetal programmed cell death (PCD) occurring earlier in anther development than in wild type (Horner & Rogers, 1974; Bino, 1985; Grant et al., 1986; Balk & Leaver, 2001; Wilson et al., 2001; Kapoor et al., 2002; Ku et al., 2003). Tapetal PCD normally occurs late in pollen development and is required to provide the precursors for the biosynthesis of the pollen outer wall (Parish & Li, 2010). It has been proposed that cell proliferative and degenerative forces act simultaneously during tapetum development and when the gene, or genes, required for cell proliferation are down-regulated, the degenerative cell death mechanism proceeds unchecked, causing tapetum degeneration (Kapoor et al., 2002). Vacuolation and enlargement of tapetal cells commence at microspore mother cell meiosis in many plant species (Parish & Li, 2010). In hermaphrodite P. lanceolata, the tapetal cells also become enlarged and very clearly distinct at this stage (Fig. 3c,d). However, no distinct tapetal cell layer is seen in MS1 plants and this correlates with a premature down-regulation of PlDIV expression in tapetal cells (Fig. 5c,d). Although the tapetal layer is not as obvious in MS1 plants as it is in the hermaphrodites, it is clear that a tapetal cell layer is initiated, as evidenced by the pattern of PlH4 expression (Fig. 5g). Eventually PlDIV and PlH4 expression ceases in all cells within the MS1 anthers (Fig. 5e). At this stage it is not possible to say if cessation of PlDIV and PlH4 expression in MS1 tapetum cells is attributable to gene-specific down-regulation in these cells and/or PCD. The cytology in the MS1 probed sections suggests tapetal cell shrinkage, which is an indicator of PCD (Parish & Li, 2010). It is possible then that the P. lanceolata MS1 phenotype is a result of premature tapetal PCD brought about by the loss of a counter-balancing PlDIV-mediated cell proliferative force. However, rather than initiating a tapetum cell death program, the down-regulation of PlDIV expression in MS1 anthers could equally well be a consequence of the cell death program itself, initiated by entirely different factors. Additional work is required to investigate the precise role of PlDIV in tapetum and pollen development.

All efforts to identify a P. lanceolata RAD ortholog failed, suggesting that this gene is absent in P. lanceolata. Several lines of evidence strongly support this conclusion: (1) RAD orthologs were easily identified in other Veronicaceae species (e.g. V. longifolia and D. purpurea) and in Sprolixus, a non-Veronicaceae species that also belongs to the Lamiales sensu lato, suggesting that the gene-specific primers and PCR methodologies used were appropriate for gene amplification; (2) several other P. lanceolata genes were easily identified using PCR in this study (including PlCYC, PlCAL, PlDIV, PlDIV2 and PlH4), suggesting that an accelerated nucleotide substitution rate in P. lanceolata nuclear DNA cannot explain the lack of a RAD gene PCR product; (3) analysis of the extended RAD-like gene family in P. lanceolata also revealed no RAD gene ortholog even though all other known classes of RAD-like genes were identified using this approach, and despite the fact that the same PCR primers selectively amplified RAD gene orthologs from a range of species within the Dipsacales (Boyden et al., 2012); (4) late PlDIV expression seen in all P. lanceolata petals is consistent with a RAD gene loss event. These data also confirm previous work that showed RAD gene loss in the lineage leading to Plantago (Preston et al., 2011).

The role of RAD in A. majus is to negatively regulate DIV in regions of the flower where CYC2 genes and DIV gene expression spatially overlap. RAD does this by sequestering co-regulators of DIV transcriptional activity in the cytoplasm, thus preventing or reducing DIV activity in the nucleus (Raimundo et al., 2013). In Antirrhinum, RAD is not expressed in cyc/dich double mutant flowers and late expression of DIV spreads to all five petals, resulting in radially symmetric (actinomorphic) flowers with all petals having ventral petal identity and shape (Luo et al., 1996, 1999; Galego & Almeida, 2002). Preston et al. (2011) previously proposed that the evolution of flower shape in Plantago involved loss of the entire floral symmetry gene network (A-clade CYC, RAD and DIV) which resulted in actinomorphic flowers with the default lateral floral organ identity program in place across the whole flower. However, this study demonstrates that a ventral identity program is maintained in Plantago. DIV orthologs were identified in both P. lanceolata and P. major, representative species from the two main Plantago subgenera (Psyllium and Plantago; Rønsted et al., 2002), which suggests that the program is probably retained throughout the genus. Thus, the route to actinomorphy in Plantago involved loss of only the dorsal identity program, leaving the ventral identity program in place and operational across the whole flower, and in addition acquiring a novel role during stamen and pollen development.

Acknowledgements

We thank Gwyneth Ingram and Andrew Hudson for fixing and embedding tissue; Mike Scanlon and Scanlon lab members for access to facilities for qPCR, fixing and embedding tissue and in situ hybridizations; Louise MacElwain for help with cloning PmDIV; Cian Lemass for help with identifying RAD-like genes in Plantago; Frank Wellmer and Emmanuelle Graciet for help with photography and Andrew Hudson and Liam Dolan for comments on the manuscript. We acknowledge the SFI/HEA Irish Centre for High-End Computing (ICHEC) for the provision of computational facilities. This work was part funded by a Basic Science Research Grant from Enterprise Ireland (J.M.N.) and an SFI UREKA Summer School grant (2008–2011).

Ancillary