The Aquilegia FRUITFULL-like genes play key roles in leaf morphogenesis and inflorescence development

Authors


For correspondence (e-mail alitt@nybg.org).

Summary

The APETALA1/FRUITFULL (AP1/FUL) MADS box transcription factors are best known for the role of AP1 in Arabidopsis sepal and petal identity, the canonical A function of the ABC model of flower development. However, this gene lineage underwent multiple duplication events during angiosperm evolution, providing different taxa with unique gene complements. One such duplication correlates with the origin of the core eudicots, and produced the euAP1 and euFUL clades. Together, euAP1 and euFUL genes function in proper floral meristem identity and repression of axillary meristem growth. Independently, euAP1 genes function in floral meristem and sepal identity, whereas euFUL genes control phase transition, cauline leaf growth and fruit development. To investigate the impact of the core eudicot duplication on the functional diversification of this gene lineage, we studied the role of pre-duplication FUL-like genes in columbine (Aquilegia coerulea). Our results show that AqcFL1 genes are broadly expressed in vegetative and reproductive meristems, leaves and flowers. Virus-induced gene silencing of the loci results in plants with increased branching, shorter inflorescences with fewer flowers, and dramatic changes in leaf shape and complexity. However, aqcfl1 plants have normal flowers and fruits. Our results show that, in contrast to characterized AP1/FUL genes, the AqcFL1 loci are either genetically redundant or have been decoupled from the floral genetic program, and play a major role in leaf morphogenesis. We analyze the results in the context of the core eudicot duplication, and discuss the implications of our findings in terms of the genetic regulation of leaf morphogenesis in Aquilegia and other flowering plants.

Introduction

The APETALA1/FRUITFULL (AP1/FUL) MADS box transcription factors have been implicated in multiple developmental processes, from phase transition to floral organ identity to fruit development; however, it remains unclear how these functions were acquired over evolutionary time. It has been shown that the AP1/FUL genes are angiosperm-specific and have undergone several duplications (Litt and Irish, 2003; Shan et al., 2007). The most significant of these duplications, the duplication that produced the euAP1 and euFUL clades, is correlated with the origin of the core eudicots, a large clade that comprises approximately 75% of extant species of flowering plants (Angiosperm Phylogeny Group, 2009). Sequence analyses have shown that proteins in the two clades have divergent motifs at the C–terminus (Cho et al., 1999; Yalovsky et al., 2000; Litt and Irish, 2003; Shan et al., 2007), and functional studies have demonstrated some overlapping and some unique roles. Together, euAP1 and euFUL genes are implicated in proper floral meristem identity and axillary meristem repression (Ferrandiz et al., 2000). Independently, euAP1 genes play a role in sepal identity (and, in Arabidopsis, petal identity) (Irish and Sussex, 1990; Huijser et al., 1992; Berbel et al., 2001; Vrebalov et al., 2002; Benlloch et al., 2006), whereas euFUL genes control the transition to the reproductive meristem, the transition to first-order inflorescence meristems, cauline leaf growth, determinacy and fruit development (Gu et al., 1998; Immink et al., 1999; Müller et al., 2001; Melzer et al., 2008; Jaakola et al., 2010; Berbel et al., 2012; Torti et al., 2012).

Differences in function among paralogs may be the result of the divergent C-terminal sequences, differences in expression domains, and/or differences in interacting partners, as MADS box proteins function in complexes (Riechmann et al., 1996; Theissen and Saedler, 2001; Smaczniak et al., 2011). For instance, in Arabidopsis, whereas AP1 is expressed in the floral meristem and later in sepals and petals (Bowman et al., 1993; Blázquez et al., 2006), FUL is expressed in the inflorescence meristem, cauline leaves and carpels (Mandel and Yanofsky, 1995; Gu et al., 1998; Ferrandiz et al., 2000). Both AP1 and FUL can interact with a common set of other MADS box transcription factors involved in the flowering transition, inflorescence meristem identity and floral organ identity (Pelaz et al., 2001; DeFolter et al., 2005; Gregis et al., 2006; Kaufmann et al., 2010); however, AP1 has also been shown to form higher-order complexes with the B function proteins APETALA3 (AP3) and PISTILLATA (PI), whereas FUL has been shown to interact with the C function protein AGAMOUS (AG) (Riechmann et al., 1996; DeFolter et al., 2005). Other euAP1 genes are expressed similarly to AP1 (e.g. Hardenack et al., 1994; Ferrandiz et al., 2000; Berbel et al., 2001; Benlloch et al., 2006), whereas euFUL genes are broadly expressed in vegetative parts, floral organs and fruits (e.g. Gu et al., 1998; Fornara et al., 2004; Shchennikova et al., 2004; Preston and Kellogg, 2007; Danilevskaya et al., 2008; Sather and Golenberg, 2009; Pabón-Mora et al., 2012). Protein interaction data from core eudicots (Huijser et al., 1992; Leseberg et al., 2008) have shown that other euAP1 proteins have interaction capabilities similar to AP1, whereas the interactions of euFUL proteins vary greatly, and occasionally have expanded to include orthologs of AP3 and PI, which are typically euAP1 partners (Immink et al., 2003; Shchennikova et al., 2004; Leseberg et al., 2008). Collectively, the expression, function and protein interactions in core eudicots suggest a more conserved euAP1 functional program in floral meristem determination and sepal identity, and a more flexible recruitment of euFUL proteins into multiple developmental processes. However, determining the origin of these divergent core eudicot functions and whether this is correlated with the duplication event requires examination of the ancestral role of the gene lineage prior to the duplication.

The pre-duplication ancestors in the AP1/FUL gene lineage are the FUL-like genes, which are present in all flowering plants outside of the core eudicots (Litt and Irish, 2003; Preston and Kellogg, 2006; Shan et al., 2007). In terms of sequence, FUL-like proteins are more similar to euFUL than to euAP1 proteins, and data from rice (Oryza sativa) and rye (Secale cereale) suggest that they interact with a similar suite of proteins to euFUL proteins (Fornara et al., 2004; Ciannamea et al., 2006). Grass FUL-like genes have been implicated in control of the transition from vegetative to reproductive meristems in response to vernalization, based on the inability of wap1 (wheat apetala1-like) mutants and quadruple rice mutants [defective in three FUL-like and one SEPALLATA (SEP) gene] to flower, as well as on the accumulation of FUL-like genes transcripts during vernalization in temperate grasses (Murai et al., 2003; Trevaskis et al., 2003, 2007; Preston and Kellogg, 2008; Kobayashi et al., 2012). In contrast, FUL-like genes in poppies (Papaver somniferum, basal eudicots) play pleiotropic roles in flowering time, cauline leaf development, and proper floral meristem and perianth identity, as well as late roles in fruit development (Pabón-Mora et al., 2012). The poppy data suggest that most of the functions attributed to core eudicot paralogs were present before the duplication and were subsequently divided between euAP1 and euFUL genes. However, the available functional information is limited, preventing us from generalizing the role of FUL-like genes, and from understanding the effect of the core eudicot gene duplication on the functional evolution of AP1/FUL genes.

The main objective of this study is to expand the characterization of FUL-like genes in basal eudicots, to explore whether functions of FUL-like genes in poppies may be extrapolated to other members of the order Ranunculales, and to develop a broader and firmer basis on which to evaluate the functional divergence of the core eudicot AP1/FUL homologs. We chose to work with Aquilegia coerulea ‘Origami’ (Ranunculaceae), an emerging model system with a complete genome sequence (http://www.phytozome.net), an extensive EST database, optimized methods for analysis of gene function, and a growing body of data on gene function and evolution (Kramer et al., 2007; Voelckel et al., 2010; Ballerini and Kramer, 2011; Sharma et al., 2011). The present study explores the expression and function of the Aquilegia AP1/FUL homologs, and shows that Aquilegia coerulea FUL-like1 loci (AqcFL1) share some functions with poppy counterparts, but also play a novel role in leaf morphogenesis. In contrast to all other AP1/FUL homologs characterized to date, AqcFL1 may not function in flower or fruit development. We compare our findings to the known roles of AP1/FUL genes in core and non-core eudicots, and highlight their implications for our current understanding of leaf development gene networks in angiosperms.

Results

Aquilegia coerulea is a perennial rosette-forming herb that requires vernalization to flower (Kramer, 2009). Wild-type A. coerulea leaves consist of three parts: (i) a leaf base with a large pair of stipules that enclose the younger leaves and the shoot apical meristem (SAM), (ii) a petiole, and (iii) a peltately palmate compound lamina with three leaflets that are deeply segmented into secondary leaflets and have lobed margins (Figure S1). Upon the transition to flowering, the shoot apical meristem is converted to an inflorescence meristem that elongates and forms two lateral cauline leaves, each subtending axillary meristems, before producing a terminal flower (Figure S1). Axillary meristems initially have inflorescence meristem identity and repeat the same developmental pattern. A. coerulea ‘Origami’ produces one (or occasionally two) main inflorescence axes (Figure S1). Cauline leaves have short petioles and petiolules, and are consistently smaller than rosette leaves; they vary in size and shape, ranging from highly dissected lower leaves, always with three leaflets, to upper leaves with as few as one leaflet (Figure S1). A. coerulea flowers are polysymmetric, with five free petaloid sepals, five free spurred petals, numerous stamens in 4–6 whorls and five staminodes that surround five free carpels (Tucker and Hodges, 2005; Ronse de Craene, 2010).

Identification of Aquilegia FUL-like genes

Aquilegia coerulea has two FUL-like genes in its genome, the result of a tandem duplication (http://www.phytozome.net). The two copies, AqcFL1A and AqcFL1B, share 97% nucleotide identity and 96% amino acid identity in their coding sequences. Sequence analysis predicts that both copies have the typical FUL-like C-terminal amino acid motif (Figure S2, Table S2) and are expressed (http://compbio.dfci.harvard.edu/cgi-bin/tgi/gimain.pl?gudb=aquilegia). Because of the high sequence similarity, the constructs used in this study for in situ hybridization, virus-induced gene silencing and yeast two-hybrid analyses did not distinguish between the copies; hence, for these experiments, we refer to the two copies together as AqcFL1.

AqcFL1 genes are broadly expressed during vegetative and reproductive growth

We evaluated AqcFL1A and AqcFL1B gene expression in a series of dissected organs from A. coerulea using quantitative RT-PCR (Figure 1a). The results show that both are expressed in very young shoot apical meristems and unfolding leaves before and after vernalization, but expression is stronger after vernalization. In addition, both genes are expressed in all floral organs before anthesis, and in the developing fruit (Figure 1a). However, AqcFL1A is more strongly expressed in floral organs and fruits, whereas AqcFL1B is more strongly expressed in leaves after vernalization.

Figure 1.

Expression of AqcFL1 genes. (a) Quantitative RT-PCR results showing expression of AqcFL1A and AqcFL1B in shoot apices (shoot apical meristem and unexpanded leaves) and leaves before and after vernalization, in dissected floral organs prior to anthesis, and in fruits. SBV, shoot apical meristem before vernalization; LBV, leaves before vernalization; SAV, shoot apical meristem after vernalization; LAV, leaves after vernalization; Sep, sepal; Pet, petals; Sta, stamens; Std, staminodia; Car, carpels; Fr, flower. (b–l) In situ mRNA hybridization of AqcFL1. cl, cauline leaf; fm, floral meristem; l, leaf primordia; la, leaf lamina; lb, leaf base; lf, lateral flower; p, petal; s, sepal; st, stamens; tf, terminal flower; arrowhead indicates carpel primordia; arrows indicate staminodia; asterisks indicate axillary floral buds; dashed line indicates the leaf petiole. Scale bars = 50 μm. (b–d) Shoot apical meristem before the transition to flowering, showing young leaf primordium (b) and older leaves (c, d). (e) Close-up of a dissected lamina of a rosette leaf. This section shows the folded leaf; segments correspond to overlapping tips of the leaflets. (f) Young inflorescence with a terminal flower and two lateral flowers. (g) Flower meristem and two axillary meristems. (h) Elongating floral bud with young sepals. (i) Floral bud with sepals and incipient petal and stamen primordia. (j) Floral bud with mature sepals and differentiated petal, stamen, staminode and carpel primordia. (k) Mature floral bud with young carpel primordia. (l) Mature floral bud with fully differentiated carpels.

A more detailed examination of expression based on in situ mRNA hybridization shows that the AqcFL1 genes are expressed in the shoot apex before vernalization as well as in the vasculature and young rosette leaf primordia (Figure 1b–d). As the leaves enlarge, AqcFL1 expression becomes restricted first to the lamina (Figure 1b–d) and later to the growing tips (Figure 1e). After the reproductive transition, AqcFL1 genes are expressed in inflorescence and floral meristems (Figure 1f), in young floral buds during sepal initiation (Figure 1g,h), and in petal and stamen primordia (Figure 1h–i). When the carpel primordia initiate, AqcFL1 expression is reduced in sepals, but remains in the tips of the petal, stamen, staminode and carpel primordia (Figure 1j–l). Hybridization with the sense AqcFL1 probe showed no signal (data not shown). These results show that AqcFL1 genes are expressed broadly in time and space during vegetative and reproductive growth, similar to what was shown previously for Aquilegia formosa (Ballerini and Kramer, 2011).

AqcFL1 genes regulate leaf morphology, axillary meristem growth, and inflorescence height

To investigate the function of FUL-like genes in Aquilegia, we used the bipartite tobacco rattle virus (TRV) to perform virus-induced gene silencing (VIGS) as previously described (Liu et al., 2002; Gould and Kramer, 2007; Pabón-Mora et al., 2012). One of the two viral particles, TRV2, includes a multiple cloning site into which a fragment of the target gene may be inserted to induce silencing. We used a TRV2 construct that includes a region of the Aquilegia gene AqANS encoding anthocyanidin synthase (Gould and Kramer, 2007); down-regulation of AqANS levels results in a reduction of floral pigmentation, facilitating the screening of plants (Kramer et al., 2007; Sharma et al., 2011). We cloned a fragment from a conserved region of the AqcFL1A and B genes (Figure S2) into TRV2-AqANS to produce TRV2-AqcFL1-AqANS. A total of 150 plants with four or five leaves were infiltrated with TRV1 and TRV2-AqcFL1-AqANS. In parallel, 100 control plants at the same developmental stage were treated with TRV1 and TRV2-AqANS. Because the viability of TRV is poor at low temperature, infiltration was performed after vernalization. At this stage, plants had already become competent to flower; as a result, the function of AqcFL1 genes in the transition from the vegetative to the reproductive state could not be evaluated.

Silencing of AqANS in both the control and experimental groups was evident 4–5 weeks after treatment, when plants started flowering. Rosette and cauline leaves of plants in both groups (40 control and 45 experimental plants) that showed silencing of AqANS were screened using RT-PCR and quantitative RT-PCR to detect changes in the amount of AqcFL1 transcript. A reduction in transcript abundance was observed in the leaves of 34 (75%) of the 45 TRV2-AqcFL1-AqANS plants when compared to pooled samples from TRV2-AqANS plants (Figure 2a). In addition, to confirm reduction of the AqcFL1 transcript in floral tissue, a total of 15 each of TRV2-AqcFL1-AqANS sepal, petal and carpel samples were screened using quantitative RT-PCR. Most samples showed some degree of down-regulation when compared with pooled TRV2-AqANS sepals, petals and carpels (Figure 2b–d). Down-regulation was stronger in sepals and petals than in carpels (Figure 2b–d); carpels in flowers showing strong ANS silencing were left on the plant for evaluation of fruit phenotypes, and carpels from adjacent flowers, that may not have been as strongly silenced, were screened. Because AqANS silencing does not result in any morphological differences other than changes in flower color (Gould and Kramer, 2007; Kramer et al., 2007), the TRV2-AqANS-treated plants are hereinafter designated ‘wild-type’ and serve as the control. Plants that showed down-regulation of the target gene are hereinafter referred to as aqcfl1 plants.

Figure 2.

Down-regulation of AqcFL1 in VIGS-treated plants. (a) RT-PCR using cDNA prepared from leaves of VIGS-treated plants showing the fold change in AqcFL1 expression relative to pooled wild-type leaves in six aqcfl1 plants (numbered). (b–d) Quantitative RT-PCR using cDNA prepared from sepals (b), petals (c) and carpels (d) of VIGS-treated plants that showed down-regulation of AqcFL1 in leaves. Values are means ± SD for three technical replicates. AqIPP2 was used as the endogenous control.

In contrast to the range of phenotypes associated with other ap1/ful mutants (reviewed in Pabón-Mora et al., 2012), aqcfl1 plants showed no defects in floral meristem identity, floral organ identity or fruit development (Figure 3a–d; data not shown). However, differences in inflorescence size and axillary meristem growth were noted (Figure 3e–i). aqcfl1 plants consistently showed reduced inflorescence height (Figure 3e–g): whereas wild-type inflorescence axes often reached over 10 cm (mean 13.13 ± 3.09 cm), aqcfl1 inflorescences often arrested at 2–5 cm (mean 7.69 ± 4.6 cm; < 0.001), resulting in open flowers at the level of the rosette leaves (Figures 3e–g and 4b). To determine whether this stunted growth was associated with fewer flowers per inflorescence, we counted the number of flowers per inflorescence in wild-type and aqcfl1 plants and found that, whereas wild-type inflorescences often bear three to five flowers (mean 3.9 ± 0.80), aqcfl1 plants produce a minimum of one and a maximum of four (mean 2.7 ± 1.1; < 0.001). Outgrowth of buds in axils of the rosette leaves was also observed (Figure 3h,i and Table S1); whereas wild-type plants (= 30) produced one to two inflorescence stems (mean 1.2 ± 1.43), aqcfl1 plants (= 34) produced up to four (mean 1.78 ± 0.88; < 0.005). Thus, aqcfl1 plants exhibited shorter central inflorescences with fewer nodes and flowers, and an increase in the number of axillary inflorescences.

Figure 3.

Overview of flowers and inflorescences in wild-type and down-regulated aqcfl1 plants. (a, b) Wild-type flowers: front view (a) and lateral view (b). (c, d) aqcfl1 flowers: front view (c) and lateral view (d). (e) Wild-type inflorescence: lateral view. (f, g) aqcfl1 inflorescences: top view (f) and lateral view (g). (h, i) Close-up of the axillary meristems in rosette leaves at early (h) and late developmental stages (i). Arrows indicate additional elongating axillary meristems. Scale bars = 2 cm (a–g) and 1 cm (h, i).

Figure 4.

Rosette and cauline leaf phenotypes in aqcfl1 plants. (a, b) Top view of (a) Wild-type and (b) aqcfl1 plants. Arrows indicate the abnormal leaves in aqcfl1 plants. (c, d) Close-up of selected wild-type (c) and aqcfl1 rosette leaves (d). ‘1’ indicates the central leaflet; ‘2’ indicates the lateral leaflets; ‘s’ indicates segments or secondary leaflets. (e, f) First cauline leaves of wild-type (e) and aqcfl1 (f). (g, h) Second cauline leaves of wild-type (g) and aqcfl1 (h). (i, j) Upper-most cauline leaves of wild-type (i) and aqcfl1 (j). (k, l) Scanning electron microscopy of the abaxial (k) and adaxial (l) wild-type rosette leaf surface. (m, n) Scanning electron microscopy of the abaxial (m) and adaxial (n) aqcfl1 rosette leaf surface. (o, p) Scanning electron microscopy of the abaxial (o) and adaxial (p) wild-type cauline leaf surface. (q, r) Scanning electron microscopy of the abaxial (q) and adaxial (r) aqcfl1 cauline leaf surface. (s–v) Clearings of wild-type (s) and aqcfl1 (t) rosette leaves, and wild-type (u) and aqcfl1 (v) cauline leaves; a single leaflet is shown in (s), (t) and (v). (w–z) Cross-sections of wild-type (w) and aqcfl1 (x) rosette leaves, and wild-type (y) and aqcfl1 (z) cauline leaves. Arrowheads indicate mid-vascular bundle. Scale bars = 2.5 cm (a, b), 1 cm (c–j, s–v), 30 μm (k–r) and 50 μm (w–z).

Unexpectedly, the most conspicuous abnormality in aqcfl1 plants was a change in leaf morphology (Figure 4a–j and Table S1). Rosette leaves in 13 aqcfl1 plants (38%) exhibited reduced lamina complexity, often showing two to three leaflets with smooth entire margins and little lobing (Figure 4d). These leaves showed no defects in the base, the petiole or the petiolules. Cauline leaf morphology (Figure 4e–j) was also abnormal in 29 aqcfl1 plants (85%), which exhibited: (i) increased length of the petiolules, particularly in the most basal but occasionally in the upper-most cauline leaves (Figure 4f,j), (ii) a reduction in leaf complexity, often to two leaflets with a reduced number of secondary leaflets and smooth entire margins (Figure 4f,h,j), and (iii) broadening of the lamina, especially in the upper-most cauline leaves (Figures 4j and S3).

aqcfl1 rosette and cauline leaves were compared with wild-type under the scanning electron microscope to determine whether there was any loss of leaf identity or epidermal abnormality associated with the different shape. aqcfl1 rosette leaves showed no epidermal abnormalities (Figure 4k–n). aqcfl1 cauline leaves (Figure 4o–r) possess a normal abaxial epidermis, but the smooth rounded adaxial epidermal cells of wild-type are replaced by papillate cells, which are more similar to the adaxial epidermal cells of the rosette leaf (Figure 4p,r). In addition, leaf clearings show that, although the number of main vascular traces entering each leaflet is the same in aqcfl1 and wild-type plants, their trajectory along the lamina differs. In wild-type leaves, vascular traces extend outwards to irrigate the central lobes of each leaflet, whereas in aqcfl1 leaves, the vascular traces converge at the tip of the simple leaf or leaflet (Figure 4s–v). Finally, cross-sections of the leaves show that the mid-vein of both the aqcfl1 rosette and cauline leaves is less developed and has fewer vascular cells than wild-type (Figure 4w–y). These results show that AqcFL1 plays a role in repression of axillary bud growth, inflorescence height and leaf morphogenesis.

Leaflets and lobes form very early in Aquilegia leaf development

To determine whether leaflet initiation and secondary leaflet formation as well as lobing of the lamina overlap with AqcFL1 expression, we investigated in detail the stages of wild-type Aquilegia leaf development during which leaflets and lobes initiate. Formation of the three main leaflet primordia takes place in plastochron 1 (P1) (a plastochron being the time between successive leaf initiation events) (Figure 5a). Leaflets are initiated basipetally, and the central leaflet is slightly larger than the lateral ones (Figure 5a). Each leaflet primordium further differentiates into three main secondary leaflets by P2–3 (Figure 5b). These developmental stages (P1–3) overlap with the onset of AqcFL1 expression throughout leaf and leaflet primordia (Figure 1b). The lamina elongates without further segmentation in P4–5 (Figure 5c,d), until marginal meristematic activity resumes at P6–7. At this time, additional lobes are formed on the secondary leaflets (Figure 5e). During P4–7, AqcFL1 genes are expressed in the lamina and the incipient lobes, but not in the leaf base or petiole (Figure 1c,d). All leaflets, secondary leaflets and lobes are formed before the petiole and petiolules elongate (Figure 5f,g). The lower-most cauline leaves follow the same pattern as rosette leaves; in contrast, upper-most cauline leaves do not form lobes (Figure 5h,i). During cauline leaf development, AqcFL1 transcripts accumulate in the growing tips of all cauline leaves (Figure 1f,g,i). These results show that AqcFL1 expression in the leaf primordium, the margin and the growing tips of the lamina during P1–P7 and during cauline leaf initiation and differentiation (Figure 1) largely overlaps with the stages at which leaflets, secondary leaflets and lobes are initiated.

Figure 5.

Rosette and cauline leaf development in Aquilegia coerulea. (a) Shoot apical meristem flanked by two rosette leaves. (b–g) Leaf development series showing leaflet inception and early lobing of each leaflet, concurrent with differentiation of the leaf base (b–d) and subsequent marginal growth and increase in leaf dissection (e–g). (h) Lateral view of the inflorescence showing the terminal flower and the lateral floral primordia protected by the most basal cauline leaf. (i) Elongating inflorescence with three cauline leaves protecting the flowers. c, cauline leaf; lb, leaf base; s, segments in each leaflet; tf, terminal flower; ‘1’ indicates the central leaflet; ‘2’ indicates the lateral leaflets; asterisks indicate the leaflet primordia; arrows indicate petiolules; dashed line indicates the leaf petiole. Scale bars = 100 μm (a–f, h, i) and 1 mm (g).

Pairwise yeast two-hybrid assays of AqcFL1A show strong interactions with flowering regulators and floral meristem identity proteins

To compare interaction partners between the pre-duplication Aquilegia FUL-like protein and Arabidopsis AP1 and FUL, we tested two-way interactions between AqcFL1A and the Aquilegia MADS box homologs of the AP1 and FUL protein partners. Given that the two paralogous copies of AqcFL1 differ by eight amino acids, of which only one difference (195S→P) is radically non-conservative, testing both proteins was considered unnecessary. Furthermore, AqcFL1A shows higher expression in floral tissues and all of the known interactors are likewise expressed in reproductive tissues, so we focused this analysis on AqcFL1A. Common partners of Arabidopsis AP1 and FUL identified through yeast two-hybrid (Y2H) assays include flowering regulators, such as SUPPRESSOR OF OVEREXPRESSION OF CONSTANS1 (SOC1) and AGAMOUS-LIKE24 (AGL24), and the floral meristem and floral organ identity proteins AGAMOUS-LIKE6 (AGL6) and SEP; in addition, FUL interacts with AG. Aquilegia possesses two AGL24 copies (Ballerini and Kramer, 2011), three SEP copies, four SOC1 and four AP3 copies, one PI copy, and two AG copies (http://www.phytozome.net; Kramer et al., 2007; Sharma et al., 2011). With the exception of AqcSOC1.3 and AqcSOC1.4, whose expression was not detected in any of the tissues tested for expression of AqcFL1 (Figure S4), all genes are co-expressed with AqcFL1 (Figures 1a and S4) (Kramer et al., 2007; Ballerini and Kramer, 2011). We used a Y2H system to test protein interactions based on genes that are co-expressed with AqcFL1, with the exception of AqcAP3-3b, which is expressed at very low levels, and the recently identified AqcAG2 (E. Kramer and B. Sharma, unpublished data). Our results show strong growth for AqcFL1A homodimers, as well as combinations with AqcAGL24.1, AqcAGL6, AqcSEP1 and AqcSEP3. In addition, weak growth is seen with AqcSOC1.1, AqcSOC1.2 and AqcSEP2. No growth was recovered for AqcFL1A with AqcAGL24.2, AqcPI, AqcAG1 or the AqcAP3 homologs (Figures 6, S5 and S6). Thus, pairwise Y2H tests suggest that AqcFL1A may have a similar repertoire of interacting partners to AP1 in Arabidopsis.

Figure 6.

Summary of protein interactions of AqcFL1A with other MADS box proteins in contrast with reported Arabidopsis AP1 and FUL interactions. Solid lines show interactions from Y2H studies reported by Honma and Goto (2001), Pelaz et al. (2001) and DeFolter et al. (2005). Dashed lines show interactions based on immunoprecipitation studies reported by Riechmann et al. (1996). The asterisk indicates that the pairwise interaction between AqcFL1A and AqcAG2 has not been tested yet.

Discussion

The broad expression domain of AqcFL1 is similar to that of other core and basal eudicot euFUL and FUL-like genes

AqcFL1 genes are broadly expressed in the vegetative meristem, and in the young leaf primordium (P1–2) and developing lamina (P3–7). After the reproductive transition, AqcFL1 is expressed in the inflorescence and floral meristem, cauline leaves, all floral parts and fruits. This expression pattern is typical of euFUL and other FUL-like genes (Fornara et al., 2004; Kim et al., 2005; Preston and Kellogg, 2007; Danilevskaya et al., 2008; Berbel et al., 2012; Pabón-Mora et al., 2012), in contrast to the more restricted expression in floral meristems and perianth organs characteristic of euAP1 genes (Hardenack et al., 1994; Ferrandiz et al., 2000; Berbel et al., 2001; Sather and Golenberg, 2009). The AqcFL1 expression data support our previous observations in poppies, and support the hypothesis that pre-duplication FUL-like genes are broadly expressed spatially and temporally, and that, whereas euFUL genes maintain this broad expression, euAP1 expression is lost from leaves, stamens, carpels and fruits. These broad expression patterns may have facilitated the redeployment of euFUL and FUL-like gene into diverse new roles in different taxa.

AqcFL1 functions in inflorescence development, but not in flower or fruit development

Based on the roles of Papaveraceae FUL-like genes in branching, flowering time, perianth identity and fruit development, we previously concluded that the functional repertoire of the pre-duplication FUL-like lineage encompassed all the roles of the core eudicot paralogs euFUL and euAP1 (Pabón-Mora et al., 2012). euAP1 genes typically function in floral meristem and sepal identity (Irish and Sussex, 1990; Huijser et al., 1992; Ferrandiz et al., 2000; Berbel et al., 2001; Vrebalov et al., 2002; Benlloch et al., 2006), whereas euFUL genes redundantly control floral meristem identity but also regulate the transition to reproductive meristems, meristem determinacy and fruit development (Gu et al., 1998; Müller et al., 2001; Melzer et al., 2008; Jaakola et al., 2010; Pabón-Mora et al., 2012). This broad functional repertoire of poppy FUL-like genes suggests that sub-functionalization occurred after the core eudicot gene duplication (Pabón-Mora et al., 2012). In contrast, our current data suggest that AqcFL1 does not play a significant role in floral meristem or organ identity, or in fruit development. The lack of defects observed in flower and fruit development suggests three alternatives: (i) that low levels of residual expression of AqcFL1 after down-regulation are enough to maintain proper floral meristem identity, floral organ identity and fruit development, (ii) that AqcFL1 is completely redundant with other transcription factors with regard to these functions, or (iii) that AqcFL1 has been decoupled from the floral development program. Regardless of which of these possible explanations is correct, our functional data suggest that the major roles in flower and fruit development attributed to poppy FUL-like genes cannot simply be extrapolated to other Ranunculales species (Figure 7). In turn, this raises questions regarding the universality of hypothesized ancestral functions in flower and fruit development of the gene lineage prior to the core eudicot duplication.

Figure 7.

Optimization and mapping of functions reported for AP1/FUL homologs. Based on the results of Pabón-Mora et al. (2012), we hypothesize that the ancestral functions in the AP1/FUL gene lineage include specification of floral meristem and sepal identity (1), because these are functions that are shared with the sister SEPALLATA and AGL6 gene lineages. Control of the reproductive transition (2) appears to have evolved in the ancestor to the AP1/FUL lineage. Before the diversification of the Papaveraceae, the genes acquired functions in cauline leaf development (3), branching (6) and fruit development (5), although acquisition of these functions may have occurred earlier than shown. After diversification of the core eudicots, some of these functions (1 and 6) were retained by both the euFUL and the euAP1 clades, whereas others (2, 3 and 5) were exclusively retained by members of the euFUL clade. Before the divergence of Aquilegia, AP1/FUL genes may have lost their roles in floral meristem and floral organ identity and in fruit development (−1, 5), and acquired a role in leaf morphogenesis (7). A role in petal identity (4) appears to have been independently acquired in poppy and Arabidopsis. Closed circles and numbers indicate gain of function; open circles and numbers preceded by a minus sign indicate loss of function. Plus symbols indicate that the function has been recorded for that gene. *indicates genes of Arabidopsis and snapdragon (Antirrhinum majus) that have not been functionally characterized.

If AqcFL1 is genuinely decoupled from floral development, this might be related to altered protein interaction partners. Protein interaction data show that AqcFL1A interacts with SOC1, AGL24, AGL6 and SEP homologs, but not with PI, AP3 or AG homologs. This is in agreement with the interactions reported for AP1 in Y2H experiments (Riechmann et al., 1996; Pelaz et al., 2001; DeFolter et al., 2005). Although pairwise interactions between AP1 and AP3, PI and AG have been detected by in vitro immunoprecipitation, such dimers lack DNA-binding activity in electrophoretic mobility shift assays (Riechmann et al., 1996) and have not been detected in Y2H assays, therefore their biological significance is unclear. Thus, our data suggest that the lack of a role for AqcFL1 in flower development is not the result of significant differences in dimerization capability in comparison to AP1. AqcFL1A interactions differ from those of FUL in that AqcFL1A does not interact with AqcAG1. This may explain the apparent absence of AqcFL1 involvement in fruit development; however, interactions have not yet been tested with AqcAG2. Untangling the functional distinctions among FUL-like homologs will require more protein interaction data from Aquilegia as well as other basal eudicots, along with silencing of additional loci, such as SEP genes and AGL6.

Our results do show that AqcFL1 plays a critical role in the repression of axillary meristems in the rosette. Similar roles have been reported for other FUL-like and euFUL genes, although often this repression of axillary meristem growth is associated with cauline and not rosette leaves (Ferrandiz et al., 2000; Berbel et al., 2001, 2012; Fornara et al., 2004; Pabón-Mora et al., 2012). Nevertheless, the role of AP1/FUL genes in the regulation of branching appears to be ancestral for eudicot (basal and core) AP1/FUL genes (Figure 7). In addition, aqcfl1 plants exhibit changes in the inflorescence architecture. In Aquilegia, the inflorescence meristem becomes a terminal floral meristem after formation of the initial cauline leaves. These cauline leaves each subtend axillary meristems that repeat the same developmental pattern, producing cauline leaves before becoming a determinate floral meristem. aqcfl1 plants have fewer flowers per inflorescence, sometimes only a single terminal flower. This is consistent with a premature transition from an inflorescence meristem to a terminal flower. The reduced number of flowers per inflorescence stem indicates that axillary meristems either do not form or do not grow out, possibly due to a reduced ability to specify or maintain inflorescence identity in these meristems. In fact, this phenotype has been observed in a euFUL mutant, veg1, in pea (Pisum sativum) (Berbel et al., 2012). This phenotype in Aquilegia is consistent with overlapping expression in axillary meristems of the putative inflorescence identity factors AqcFL1 and AqcAGL24.1/2 (Ballerini and Kramer, 2011), and with the interaction between AqcFL1A and AqcAGL24.1. Thus the Aquilegia FUL-like genes appear to function in a manner that is more akin to Arabidopsis FUL in maintaining inflorescence identity (Melzer et al., 2008) than to Arabidopsis AP1, which represses AGL24 and promotes floral meristem fate (Yu et al., 2004).

Finally, although this study was unable to assess whether AqcFL1 plays a role in the transition to reproduction, the increase in the expression level of AqcFL1 after vernalization suggests that it may also regulate, or be regulated by, phase transition. This result contrasts with the constitutively high expression of AqcFL1 orthologs found both before and after vernalization in A. formosa and Aquilegia vulgaris (Ballerini and Kramer, 2011; E. Kramer, unpublished data). Further studies are required to determine whether this is due to simple factors such as differences in the precise leaf developmental stages that were sampled, or to differences between species in terms of architecture and flowering time control. The available data suggest that the role of FUL-like genes in reproductive competency was most likely acquired before the monocot/eudicot split (Murai et al., 2003; Kobayashi et al., 2012; Pabón-Mora et al., 2012), and has been maintained in some euFUL core eudicot genes such as Arabidopsis FUL, pea VEG1 and petunia PGF (Figure 7) (Fornara et al., 2004; Melzer et al., 2008; Berbel et al., 2012). Alternatively, this phase transition function may have evolved independently in cereal grasses (Murai et al., 2003; Trevaskis et al., 2003, 2007; Preston and Kellogg, 2007, 2008; Kobayashi et al., 2012) and in basal and core eudicots (Figure 7).

AqcFL1 plays a major role in leaf morphogenesis

Our results show that leaf defects in plants with reduced aqcfl1 expression may be broadly divided into: (i) broader cauline leaves with vascular defects, (ii) a decreased number of leaflets and secondary leaflets, as well as reduced marginal lobing, in rosette and basal-most cauline leaves, and (iii) longer petiolules in cauline leaves. Other ap1/ful mutants such as ful mutants in Arabidopsis and ful-like mutants in poppy also exhibit enlarged cauline leaves, often with multiple mid-veins and defects in the distribution of the peripheral vasculature (Gu et al., 1998; Pabón-Mora et al., 2012). Nevertheless, the specific role of AqcFL1 in leaf segmentation and marginal lobing, in both rosette and cauline leaves, is unique among AP1/FUL gene functions reported so far, and appears to have been acquired after divergence of the basal eudicot lineages that include Papaveraceae and Ranunculaceae, possibly in the lineage leading to Aquilegia (Figure 7).

The gene networks regulating leaf shape and dissection have been well studied in core eudicot model species, but less so in basal eudicots. In most angiosperms with compound leaves that have been studied, class 1 KNOTTED1-like HOMEOBOX (KNOX1) genes have been shown to play a role in promoting leaf complexity through reactivation of marginal meristems, which results in leaflet growth (Bharathan et al., 2002; Hay and Tsiantis, 2010). One exception is the recruitment of orthologs of the floral meristem gene LEAFY (LFY) in place of KNOX1 in pea and close relatives (Hofer et al., 1997; Gourlay et al., 2000; Champagne and Sinha, 2004; Champagne et al., 2007). In both KNOX- and LFY-dependent compound leaves, it appears that a conserved genetic mechanism controlled by the NAM/CUC3 genes regulates the localized repression of growth that is required for proper compound leaf dissection (Blein et al., 2008). Our results show that AqcFL1 is expressed in the expanding lamina and may function similarly to KNOX1 and LFY homologs in re-establishing and/or maintaining marginal meristem activity, resulting in the formation of leaflets, secondary leaflets and lobes during plastochrons P1–6. In addition, because of the striking similarity of leaves in plants that are down-regulated for AqcFL1 and NAM/CUC3 in Aquilegia (Blein et al., 2008), it is tempting to speculate that the genes work together to regulate proper compound leaf development and marginal lobing in Aquilegia. In such a scenario, AqcFL1 may be a positive regulator of NAM/CUC3 genes and vice versa in a feedback mechanism similar to the one described for KNOX1 or LFY and NAM/CUC3 genes in other flowering plants (Blein et al., 2008; Efroni et al., 2010; Hay and Tsiantis, 2010).

The observed abnormalities in leaf margin development, leaf vascular patterning and released suppression of axillary rosette buds are also suggestive of auxin transport inhibition (Bennett et al., 1998; Mattsson et al., 1999; Tsiantis et al., 1999). Likewise, dwarfism of aqcfl1 inflorescences may be linked to a block in polar auxin transport (Tsiantis et al., 1999). However, no specific auxin feedback loop has been previously implicated in the regulation of any AP1/FUL gene. Alternatively, the shorter inflorescence axis and internodes observed in aqcfl1 plants may be associated with defects in cell elongation and proliferation due to reduced levels of gibberellins (Ross et al., 1997; Sakamoto et al., 2004). This is consistent with the described positive feedback loop between the up-regulation of genes involved in the metabolism of gibberellin and AP1 expression (Kaufmann et al., 2010). Furthermore, both auxin and gibberellin have been implicated in the control of marginal serration as well as KNOX gene expression (Jasinski et al., 2005; Barkoulas et al., 2008; Yanai et al., 2011). Therefore, AqcFL1 regulation of the gibberellin pathway may have downstream effects on multiple aspects of leaf margin development. It is possible that the minor roles in leaf development played by other AP1/FUL homologs are an effect of this gibberellin regulatory feedback loop, a conserved genetic module that may have been uniquely exploited in Aquilegia for a role in leaf development.

Our functional characterization of AqcFL1 suggests conserved ancestral roles for the AP1/FUL gene lineage in the repression of axillary meristems, as well as in the regulation of cauline leaf growth during flowering. At the same time, it shows that AP1/FUL genes have been recruited for divergent functions in basal eudicots, including an unexpected new role in leaf morphogenesis in Aquilegia and a function in inflorescence development. This contrasts sharply with functional data previously reported for AP1/FUL genes. The present results may be consistent with an enhancement of the role of AqcFL1 in controlling hormonal pathways, particularly those involving gibberellin or auxin, but it will also be important to more fully explore the interaction repertoire of AqcFL1 proteins.

Experimental Procedures

Plant material and growth conditions

Columbine (Aquilegia coerulea cv. Origami) seeds were obtained from Burpee (http://www.burpee.com). Seeds were germinated, vernalized and grown as described by Kramer et al. (2007).

In situ hybridization, scanning electron microscopy and anatomy

mRNA in situ hybridization was performed as described by Pabón-Mora et al. (2012) with the following modifications. Developing apices before and after vernalization were collected from wild-type Aquilegia plants growing at the Nolen Greenhouses (The New York Botanical Garden). DNA templates for sense and antisense RNA probe synthesis were obtained by PCR amplification of a 400 bp fragment of the K and C domains of AqcFL1 from cDNA using the primers listed in Table S3. Slides were hybridized overnight at 52°C.

For scanning electron microscopy and anatomy studies, all tissues were fixed in 70% ethanol, dehydrated through an ethanol series, and processed by critical point drying or wax infiltration as described by Pabón-Mora et al. (2012).

Quantitative RT-PCR

To test for levels of AqcFL1 transcript in wild-type organs, we extracted RNA from a series of dissected organs including the shoot apex (shoot apical meristem with young leaves not yet expanded) as well as expanding leaves before and after vernalization, and dissected floral organs from pre-anthesis buds as well as fruits. Samples were pooled from two or three individuals. Total RNA was prepared using Trizol reagent (Invitrogen, http://www.invitrogen.com) and subsequently treated with DNase (Ambion, http://www.invitrogen.com/ambion). Total RNA (0.5 μg) was used for cDNA synthesis with SuperScript III (Invitrogen). The resulting cDNA was diluted 1:10. PCR product was amplified using locus-specific primers designed using Primer ExpressTM version 3.0 (Applied Biosystems; www.appliedbiosystems.com) (Table S3). Quantitative RT-PCR was performed as described by Pabón-Mora et al. (2012). AqIPP2 (isopentyl pyrophosphate:dimethylallyl pyrophosphate isomerase) has been shown to have little quantitative transcriptional variation across tissues and developmental time points, and was therefore used as the constitutive reference transcript (Sharma et al., 2011). The level of AqcFL1A and AqcFL1B mRNA was analyzed relative to AqIPP2 using the 2−ΔΔCt method (Livak and Schmittgen, 2001). Standard deviation is reported for three technical replicates of each sample. Quantitative RT-PCR reactions were performed using the 7300 qPCR system and SDS software (Applied Biosystems).

Quantitative RT-PCR was also used to evaluate the reduction of transcript in the down-regulated plants. Levels of AqcFL1 transcript were measured in 1–5 leaves in each of 45 plants showing AqANS silencing. Down-regulation in floral organs including sepals, petals, stamens and carpels was confirmed for a subset of 15 samples of those showing down-regulation in leaves. All putative down-regulated samples were analyzed relative to four to eight samples of wild-type leaves, sepals, petals and carpels. Independent wild-type samples were used for initial comparison, but pooled samples were used for the wild-type expression level presented in Figure 2. RNA extraction, cDNA preparation, quantitative RT-PCR and data analysis were performed as described above.

RT-PCR

Expression of AqcFL1, AqcSOC1, AqcSOC1.2, AqcAGL24.1, AqcAGL24.2 and AqcAGL6 was assayed in all floral organs, leaves and fruits using the same cDNA used in the quantitative RT-PCR experiments. Primers amplified the full-length sequence and were the same as those subsequently used for Y2H experiments. Reactions were run for 30–33 cycles at an annealing temperature of 55°C. ACTIN was used as a loading control. The PCR product was run on a 1% agarose gel stained with ethidium bromide, and digitally photographed using a UVP DigiDoc-it® Darkroom (http://www.uvp.com/digidocit.html) equipped with a Canon PC1089 digital camera (www.canon.com).

TRV-VIGS

A 400 bp fragment conserved between AqcFL1A and AqcFL1B, including the K and C domains of the protein, was amplified from floral bud cDNA using primers that added BamHI and XbaI restriction sites to the 5′ and 3′ ends of the PCR product. This fragment was cloned into a previously constructed vector (Gould and Kramer, 2007) carrying a 290 bp fragment of the Aquilegia AqANS gene inserted at the EcoRI and BamHI restriction sites, creating the construct TRV2-AqcFL1-AqANS. Infiltration was performed as described by Gould and Kramer (2007).

Yeast two-hybrid analyses

The full coding sequences of AqcFL1A, AqcSOC1.1, AqcSOC1.2, AqcAGL24.1 and AqcAGL24.2 and all but the conserved MADS domain sequences of AqcSEP1, AqcSEP2A, AqcSEP3, AqcPI, AqAP3-1, AqAP3-2, AqAP3-3, AqAG1 and AqAGL6 were cloned from a mixture of leaf and floral cDNA and fused in-frame into the pGADT7 and pGBKT7 vectors (Clontech; http://www.clontech.com/). Transformants were selected on non-restrictive drop-out agar plates lacking Leu and Trp (–LW). Four colonies of appropriate construct pairs were grown overnight in selective liquid culture supplemented with 10 mm 3-amino-1,2,4-triazole, and then diluted to an absorbance at 600 nm of 0.5. Each AD or BD construct was tested alone or in combination with the complementary empty plasmid to assess background levels of auto-activation. In every case, auto-activation was eliminated by use of 10 mm 3-amino-1,2,4-aminotriazole. Higher 3-amino-1,2,4-aminotriazole concentrations were also tested, but results are only shown for 10 mm. Interactions were tested for growth on selective medium (SD) lacking histidine, tryptophan, leucine and adenine in various combinations (–HWL, –AWL and –HAWL).

Acknowledgements

We thank Barbara Ambrose for her comments on the manuscript and discussion of data. We thank Marc Hachadourian and the staff at the Nolen Greenhouses at The New York Botanical Garden for growing some of the Aquilegia plants. N.P.M. thanks the Estrategia de Sostenibilidad 2013–2014 at the Universidad de Antioquia (Medellín, Colombia). This work was supported by Andrew and Judith Economos and the US National Science Foundation (grant number IOS-0923748).

Ancillary