A variant of LEAFY reveals its capacity to stimulate meristem development by inducing RAX1

Authors

  • Hicham Chahtane,

    1. Laboratoire Physiologie Cellulaire et Végétale, CEA (Commissariat à l'Energie Atomique et aux Energies Alternatives), iRTSV (Institut de Recherches en Technologies et Sciences pour le Vivant), Grenoble, France
    2. Centre National de la Recherche Scientifique, UMR5168, Grenoble, France
    3. Université Joseph Fourier-Grenoble I, UMR5168, Grenoble, France
    4. Institut National de la Recherche Agronomique, USC1359, Grenoble, France
    Search for more papers by this author
  • Gilles Vachon,

    1. Laboratoire Physiologie Cellulaire et Végétale, CEA (Commissariat à l'Energie Atomique et aux Energies Alternatives), iRTSV (Institut de Recherches en Technologies et Sciences pour le Vivant), Grenoble, France
    2. Centre National de la Recherche Scientifique, UMR5168, Grenoble, France
    3. Université Joseph Fourier-Grenoble I, UMR5168, Grenoble, France
    4. Institut National de la Recherche Agronomique, USC1359, Grenoble, France
    Search for more papers by this author
  • Marie Le Masson,

    1. Laboratoire Physiologie Cellulaire et Végétale, CEA (Commissariat à l'Energie Atomique et aux Energies Alternatives), iRTSV (Institut de Recherches en Technologies et Sciences pour le Vivant), Grenoble, France
    2. Centre National de la Recherche Scientifique, UMR5168, Grenoble, France
    3. Université Joseph Fourier-Grenoble I, UMR5168, Grenoble, France
    4. Institut National de la Recherche Agronomique, USC1359, Grenoble, France
    Search for more papers by this author
  • Emmanuel Thévenon,

    1. Laboratoire Physiologie Cellulaire et Végétale, CEA (Commissariat à l'Energie Atomique et aux Energies Alternatives), iRTSV (Institut de Recherches en Technologies et Sciences pour le Vivant), Grenoble, France
    2. Centre National de la Recherche Scientifique, UMR5168, Grenoble, France
    3. Université Joseph Fourier-Grenoble I, UMR5168, Grenoble, France
    4. Institut National de la Recherche Agronomique, USC1359, Grenoble, France
    Search for more papers by this author
  • Sophie Périgon,

    1. Laboratoire Physiologie Cellulaire et Végétale, CEA (Commissariat à l'Energie Atomique et aux Energies Alternatives), iRTSV (Institut de Recherches en Technologies et Sciences pour le Vivant), Grenoble, France
    2. Centre National de la Recherche Scientifique, UMR5168, Grenoble, France
    3. Université Joseph Fourier-Grenoble I, UMR5168, Grenoble, France
    4. Institut National de la Recherche Agronomique, USC1359, Grenoble, France
    Current affiliation:
    1. Laboratoire d'Ecologie Alpine, Centre National de la Recherche Scientifique, UMR 5553, Grenoble, France
    Search for more papers by this author
  • Nela Mihajlovic,

    1. Department of Biology, University of Western Ontario, London, ON, Canada
    Search for more papers by this author
  • Anna Kalinina,

    1. Department of Biology, University of Western Ontario, London, ON, Canada
    Search for more papers by this author
  • Robin Michard,

    1. Laboratoire Physiologie Cellulaire et Végétale, CEA (Commissariat à l'Energie Atomique et aux Energies Alternatives), iRTSV (Institut de Recherches en Technologies et Sciences pour le Vivant), Grenoble, France
    2. Centre National de la Recherche Scientifique, UMR5168, Grenoble, France
    3. Université Joseph Fourier-Grenoble I, UMR5168, Grenoble, France
    4. Institut National de la Recherche Agronomique, USC1359, Grenoble, France
    Search for more papers by this author
  • Edwige Moyroud,

    1. Laboratoire Physiologie Cellulaire et Végétale, CEA (Commissariat à l'Energie Atomique et aux Energies Alternatives), iRTSV (Institut de Recherches en Technologies et Sciences pour le Vivant), Grenoble, France
    2. Centre National de la Recherche Scientifique, UMR5168, Grenoble, France
    3. Université Joseph Fourier-Grenoble I, UMR5168, Grenoble, France
    4. Institut National de la Recherche Agronomique, USC1359, Grenoble, France
    Current affiliation:
    1. Department of Plant Sciences, University of Cambridge, Cambridge, UK
    Search for more papers by this author
  • Marie Monniaux,

    1. Laboratoire Physiologie Cellulaire et Végétale, CEA (Commissariat à l'Energie Atomique et aux Energies Alternatives), iRTSV (Institut de Recherches en Technologies et Sciences pour le Vivant), Grenoble, France
    2. Centre National de la Recherche Scientifique, UMR5168, Grenoble, France
    3. Université Joseph Fourier-Grenoble I, UMR5168, Grenoble, France
    4. Institut National de la Recherche Agronomique, USC1359, Grenoble, France
    Search for more papers by this author
  • Camille Sayou,

    1. Laboratoire Physiologie Cellulaire et Végétale, CEA (Commissariat à l'Energie Atomique et aux Energies Alternatives), iRTSV (Institut de Recherches en Technologies et Sciences pour le Vivant), Grenoble, France
    2. Centre National de la Recherche Scientifique, UMR5168, Grenoble, France
    3. Université Joseph Fourier-Grenoble I, UMR5168, Grenoble, France
    4. Institut National de la Recherche Agronomique, USC1359, Grenoble, France
    Search for more papers by this author
  • Vojislava Grbic,

    1. Department of Biology, University of Western Ontario, London, ON, Canada
    Search for more papers by this author
  • Francois Parcy,

    Corresponding author
    1. Centre National de la Recherche Scientifique, UMR5168, Grenoble, France
    2. Université Joseph Fourier-Grenoble I, UMR5168, Grenoble, France
    3. Institut National de la Recherche Agronomique, USC1359, Grenoble, France
    Current affiliation:
    1. Centre for Molecular Medicine and Therapeutics, Child and Family Research Institute, University of British Columbia, Vancouver, BC, V5Z 4H4, Canada
    • Laboratoire Physiologie Cellulaire et Végétale, CEA (Commissariat à l'Energie Atomique et aux Energies Alternatives), iRTSV (Institut de Recherches en Technologies et Sciences pour le Vivant), Grenoble, France
    Search for more papers by this author
  • Gabrielle Tichtinsky

    1. Laboratoire Physiologie Cellulaire et Végétale, CEA (Commissariat à l'Energie Atomique et aux Energies Alternatives), iRTSV (Institut de Recherches en Technologies et Sciences pour le Vivant), Grenoble, France
    2. Centre National de la Recherche Scientifique, UMR5168, Grenoble, France
    3. Université Joseph Fourier-Grenoble I, UMR5168, Grenoble, France
    4. Institut National de la Recherche Agronomique, USC1359, Grenoble, France
    Search for more papers by this author

For correspondence (e-mail francois.parcy@cea.fr).

Summary

In indeterminate inflorescences, floral meristems develop on the flanks of the shoot apical meristem, at positions determined by auxin maxima. The floral identity of these meristems is conferred by a handful of genes called floral meristem identity genes, among which the LEAFY (LFY) transcription factor plays a prominent role. However, the molecular mechanism controlling the early emergence of floral meristems remains unknown. A body of evidence indicates that LFY may contribute to this developmental shift, but a direct effect of LFY on meristem emergence has not been demonstrated. We have generated a LFY allele with reduced floral function and revealed its ability to stimulate axillary meristem growth. This role is barely detectable in the lfy single mutant but becomes obvious in several double mutant backgrounds and plants ectopically expressing LFY. We show that this role requires the ability of LFY to bind DNA, and is mediated by direct induction of REGULATOR OF AXILLARY MERISTEMS1 (RAX1) by LFY. We propose that this function unifies the diverse roles described for LFY in multiple angiosperm species, ranging from monocot inflorescence identity to legume leaf development, and that it probably pre-dates the origin of angiosperms.

Introduction

In plants bearing raceme inflorescences such as Arabidopsis thaliana, flowers arise on the flanks of the shoot apical meristem. They initially emerge as small groups of cells that acquire first meristematic identity and then floral identity, to eventually produce flower buds. Meristematic identity is first evident in the shape and organization of this domed group of dividing cells (stages 1–2), and later in the molecular markers that they express, such as CLAVATA3 (CLV3), WUSCHEL (WUS) and UNUSUAL FLORAL ORGANS (UFO) (stages 2–3). Floral meristem identity genes determine floral identity (Liu et al., 2009; Wagner, 2009; Irish, 2010; Moyroud et al., 2010; Pose et al., 2012). This class includes LEAFY (LFY) and APETALA1 (AP1), which control the development of flowers and their patterning through induction of organ identity genes such as AGAMOUS (AG) and APETALA3 (AP3). LFY and AP1 act very early in meristem development (stages 0–1) and share many direct target genes (Kaufmann et al., 2010; Moyroud et al., 2011; Winter et al., 2011). Although determination of floral meristem identity has been intensively studied and is well understood, how the initial group of cells acquires its meristematic identity remained unknown. It is known that auxin maxima determine where floral meristems arise (Kuhlemeier, 2007; Vernoux et al., 2010), but the link between auxin signaling and induction of meristematic features has not yet been fully established.

Several observations indicate that LFY is a possible candidate to mediate this action. First, LFY expression is controlled by auxin as shown in pin mutants (Vernoux et al., 2000; Lebedeva et al., 2005; Blazquez et al., 2006). Second, scattered data suggest that LFY may play a role in conferring meristematic identity to the floral meristem (Moyroud et al., 2009, 2010). Finally, LFY is able to induce meristematic identity in other developmental contexts, for instance to create compound leaves in some legumes, tillers in rice (Oryza sativa), or complex inflorescences in rice and maize (Zea mays) (Moyroud et al., 2009, 2010). LFY may thus provide the link between auxin signaling and meristematic growth. However, LFY is not required for meristem development per se: in the lfy single mutant, flowers are replaced by cauline leaves that bear a meristematic structure in their axil. This structure gives rise to shoots or shoot/flower intermediates. It is only when the lfy mutation is combined with defects in genes such as PINOID (PID) (Ezhova et al., 2000), CLV1 (Clark et al., 1993), FILAMENTOUS FLOWERS (FIL) (Chen et al., 1999; Sawa et al., 1999a), ENHANCED RESPONSE TO ABA (ERA) (Running et al., 1998; Yalovsky et al., 2000), FUSED FLORAL ORGANS (FFO) (Levin et al., 1998), FLOWERING LOCUS T (FT), FWA (Ruiz-Garcia et al., 1997) or BLADE ON PETIOLE1/2 (BOP1/2) (Norberg et al., 2005) that leaves lacking axillary meristems develop at some positions on the inflorescence (referred to as empty cauline leaves).

Here we provide direct evidence in Athaliana that LFY has the capacity to trigger meristem growth. We obtained this evidence by altering the properties of the LFY protein. LFY has two conserved functional domains: an N-terminal dimerization domain and a C-terminal DNA-binding domain (DBD) (Hamès et al., 2008; Siriwardana and Lamb, 2012). The DBD dimerizes upon DNA binding. Taking advantage of its crystallographic structure, we engineered a form of the LFY DBD (called LFYHARA) that has a compromised dimerization interface. When expressed in plants, this allele shows reduced floral function, thereby revealing the ability of LFY to stimulate the emergence of meristems.

Results

The HARA mutation reduces the affinity of LFY for DNA without affecting its sequence specificity

Two amino acid residues (H387 and R390) involved in dimerization of the LFY DBD (Figure 1a) have been identified previously, but their importance was only established in vitro using the isolated LFY DBD (Hamès et al., 2008); their role within the context of a full-length LFY protein, which contains the conserved N-terminal dimerization domain (Siriwardana and Lamb, 2012), has not been determined. We thus analyzed the properties of the near full-length LFYHARA(∆40) protein (see Experimental procedures), in which both the H387 and R390 residues are mutated to alanine, thereby suppressing the hydrogen bonds between the DBD monomers (Figure 1a). Using EMSA analysis (Figure 1b), we found that LFYHARA(∆40) binds AP1bs1, a binding site present in the AP1 promoter (Benlloch et al., 2011), but with weaker affinity than wild-type LFY(∆40). This result confirms the importance of contacts between the DBDs even in the presence of the N-terminal dimerization domain. To test whether the HARA mutations affect LFY DNA-binding specificity, we performed a high-throughput Systematic evolution of ligands by exponential enrichment (SELEX) experiment (Zhao et al., 2009). The resulting logo, representing the DNA-binding preferences of LFYHARA(∆40) (Figure 1c,d), was found to be extremely similar to the one obtained for LFY(∆40), indicating that the HARA mutations do not greatly affect LFY DNA-binding specificity. We also tested the importance of LFY DBD dimerization in a yeast one-hybrid assay. Because LFY displays no transcriptional activity in yeast, full-length LFY and LFYHARA proteins were fused to the VP16 transcriptional activation domain (Parcy et al., 1998; Lohmann et al., 2001), and their activity were tested using AGAMOUS (AG) regulatory sequences cloned upstream of the lacZ reporter gene. As shown in Figure 1(e), the activity of LFYHARA was dramatically reduced compared to LFY, showing that HARA mutations strongly weaken the ability of LFY to bind to AG regulatory sequences.

Figure 1.

Biochemical characterization of the LFYHARA and LFYHARAKARA proteins.
(a) Details of the structure of a LFY DBD dimer (Hamès et al., 2008) showing the hydrogen bonds (dotted lines) between monomers (shown in blue and green) involving His387 (H387) and Arg390 (R390).
(b) Electrophoretic mobility shift assay (EMSA) using the same increasing quantities of LFY(∆40), LFYHARA(∆40) and LFYHARAKARA(∆40) mixed with the AP1bs1 binding site (Benlloch et al., 2011).
(c, d) Logos illustrating position-specific scoring matrixes obtained by high-throughput SELEX experiments performed with LFY(∆40) (c) and LFYHARA(∆40) (d).
(e) Yeast one-hybrid assay. LFY proteins were fused to the viral VP16 activation domain. Their ability to bind the regulatory sequences of AG was followed using the activity of the reporter gene AG:LACZ (β-galactosidase activity). C: Control (LFY unfused to VP16); WT: wild −type LFY-VP16; H387A: LFYHA-VP16; R390A: LFYRA-VP16; H387A R390A: LFYHARA-VP16. Error bars indicate the standard error of the mean (n = 10).
(f) Detail of the LFY/DNA interaction (Hamès et al., 2008) showing hydrogen bonds (dotted lines) between Lys307 (K307) and Arg237 (R237) and the DNA bases.

LFYHARA is impaired in its floral meristem identity function

We next tested the functional relevance of LFY DBD dimerization in plants. LFY and LFYHARA coding sequences were placed under the control of the LFY endogenous promoter (ProLFY) in a lfy-12 null mutant background. In 9 of 10 lfy-12−/− primary transformants, the ProLFY:LFY construct rescued the lfy mutant phenotype almost fully: flowers were present all along the main inflorescence and most displayed a wild-type appearance and correctly arranged floral organs (Figure 2c,e and Table S1). Although the ProLFY:LFYHARA transgene also induced flower formation in eight of nine lfy-12−/− primary transformants (Figure 2d and Table S1), flower development was systematically abnormal compared to wild-type or ProLFY:LFY plants. Petals and stamens were either missing or distorted, frequently resulting in sterile flowers (Figure 2e and Table S1). Occasionally, the shoot apical meristem produced a few fertile flowers on the lower positions of the inflorescence, and shoot-like structures at higher positions on the stem (Figure S1), revealing the weakened ability of LFYHARA to trigger flower development. These results indicate that the capacity of LFY to control flower formation is reduced (but not abolished) when the two amino acids, H387 and R390, are mutated.

Figure 2.

ProLFY:LFY and Pro35S:LFY transgenic Arabidopsis.
(a–d) Inflorescences of wild-type Col-0 (a), lfy-12 mutant (b), ProLFY:LFY lfy-12 (c) and ProLFY:LFYHARA lfy-12 (d) plants.
(e) Corresponding floral phenotypes. Some floral organs of ProLFY:LFY lfy-12 flowers are distorted. Petal and stamen number are frequently reduced in the flowers of ProLFY:LFYHARA lfy-12 plants.
(f–h) Close-up view of the rosette of 5-week-old wild-type and Pro35S:LFY plants. The axils of wild-type plants do not show any macroscopic structures (f), whereas flowers and shoots developed at the axils of Pro35S:LFY (g) or Pro35S:LFYHARA (h) rosette leaves, respectively.
(i, j) Wild-type plant (i) and Pro35S: LFYHARA plant (j) cultivated for 2 months under short-day conditions.
(k) Percentage of rosette leaves bearing a flower or a shoot at their axil just before decapitation and 10 days after decapitation of the primary inflorescence. The number of individuals tested is indicated for each genotype, i.e. wild-type plants, lfy-12 mutants and two independent homozygous Pro35S:LFYHARA lines (#4 and #7). Error bars represent 95% confidence intervals.

LFYHARA over-expression triggers precocious meristem development in the axil of rosette leaves

Upon floral transition, several partially independent pathways converge to build flowers at the inflorescence meristem (Fornara et al., 2010). To assess the function of LFYHARA independently of these other pathways, we ectopically expressed LFY and LFYHARA under control of the 35S CaMV promoter (Pro35S). Under long-day conditions, 5-week-old wild-type plants displayed no visible axillary structures at the base of rosette leaves (Figure 2f). As previously described (Weigel and Nilsson, 1995), over-expression of the wild-type LFY cDNA induced formation of ectopic flowers in the axils of rosette and cauline leaves, as well as termination of the main shoot in a solitary flower (Figure 2g). In contrast, Pro35S:LFYHARA plants displayed no or very few ectopic flowers; instead, they showed precocious emergence of axillary shoots compared to wild-type plants (Figure 2h and Table S2). This effect was even more evident under short-day conditions, where Pro35S:LFYHARA plants became so bushy that the primary shoot was indistinguishable from the numerous secondary inflorescences (Figure 2i,j).

The growth of rosette axillary shoots in Pro35S:LFYHARA plants may be due to accelerated development of their meristems or insensitivity to the apical dominance imposed by the shoot apical meristem. In order to distinguish between these possibilities, we performed a decapitation experiment: we pruned the apical meristem to reduce apical dominance and counted visible axillary shoots after 10 days. In wild-type and lfy-12 mutant plants, decapitation resulted in a two- to threefold increased proportion of rosette leaves bearing visible shoots at their axil (Figure 2k). Pro35S:LFYHARA plants also responded to decapitation to a similar extent, despite the fact they already displayed a high number of axillary shoots before decapitation (Figure 2k). This result shows that Pro35S:LFYHARA plants are still sensitive to apical dominance. Next we studied the time course of axillary meristem initiation. To do this, we monitored the expression of the CLV3:GUS meristematic marker (Brand et al., 2002) during seedling development (Figure 3a–c). We found that, on average, the CLV3:GUS signal appeared 3.5 days earlier at the axil of rosette leaves in Pro35S:LFYHARA seedlings than in wild-type seedlings (Figure 3a). In representative 16-day-old wild-type seedlings, only the apical meristem was stained (Figure 3b), whereas in Pro35S:LFYHARA seedlings, several axillary meristems also showed CLV3:GUS expression (Figure 3c).

Figure 3.

Characterization of axillary meristem development in Pro35S:LFY, Pro35S:LFYHARA and Pro35S:AP1 plants.
(a–c) CLV3:GUS expression in Pro35S:LFYHARA and wild-type seedlings grown under long-day conditions. (a) Kinetics of CLV3:GUS activity in axillary meristems during seedling development. The mean number of stained axillary meristems is indicated as green, blue and red circles for wild-type plants and Pro35S:LFYHARA lines (#4 and #7), respectively (n = 5–12). A mathematical model was constructed to describe these data (see Experimental procedures). The corresponding green, blue and red fitting curves are shown. According to this model, the plants entered the phase of axillary meristem production at day 18.3 (17.2–19.5) for wild-type plants versus day 15.0 (14.2–15.8) and 14.9 (14.1–16.8) for Pro35S:LFYHARA lines #4 and #7, respectively (95% confidence intervals in parentheses). Representative 16-day-old wild-type (b) and Pro35S:LFYHARA (c) seedlings are shown.
(d–g) Development of the axillary meristem in Pro35S:LFY, Pro35S:AP1 and wild-type (Col) plants grown under short-day conditions. (d–f) Transverse sections of individual shoots illustrating the stages of axillary meristem development (arrowheads) according to Stirnberg et al. (2002). Scale bar = 20 μm. (g) Mean number of node intervals, or plastochrons, between the shoot apex and the first stage 1 axillary meristem (top to stage 1), between the first stage 1 and the first stage 2 axillary meristems (stage 1 to 2), and between the first stage 2 and the first stage 3 axillary meristems (stage 2 to 3). Error bars indicate the standard error of the mean (n = 3–12).

Wild-type LFY over-expression also accelerates axillary meristem development

To determine whether the wild-type LFY gene (and not only the LFYHARA allele) influences rosette axillary meristem development, we analyzed Pro35S:LFY plants. We used short-day conditions to delay precocious flowering, and included Pro35S:AP1 plants, which share many features with Pro35S:LFY plants, such as conversion of axillary meristems into flowers (Mandel and Yanofsky, 1995). To determine the developmental rate of meristem initiation and growth at the axil of rosette leaves, we scored the developmental stage of the axillary meristem at each internode as described previously (Stirnberg et al., 2002). Stage 1 is when a group of densely stained cells become visible at the leaf axil; stage 2 is when meristematic dome forms at the leaf axil; stage 3 is when axillary meristems commence primordial initiation (Figure 3d–f). Plastochron formation was used as the developmental time unit. Stage 1 was observed after 17 plastochrons in Col plants, but only after nine plastochrons (on average) in Pro35S:LFY plants. The difference between Pro35S:LFY and Col is even more striking for progression to stages 2 and 3. Axillary meristems at stage 2 were frequently absent in Pro35S:LFY plants, and axillary meristems at stages 1 and 3 were seen in adjacent plastochrons. This contrasts with the slower development of axillary meristems in Col and Pro35S:AP1 plants, where it took over five plastochrons (on average) to progress from stage 1 to stage 2, and 7.5 additional plastochrons to reach stage 3 (Figure 3g). This experiment showed that axillary meristems develop faster in Pro35S:LFY plants than in wild-type or Pro35S:AP1 plants. Given that the plastochron rate is not slower in 35S:LFY plants than in wild-type plants (Figure S2), this clearly demonstrates that LFY accelerates axillary meristem development. A further inspection of Pro35S:LFY plants revealed the occasional presence of meristems in the axils of cotyledons (Table S3), a structure commonly observed in the Pro35S:LFYHARA plants (Figure S3) or in mutants such as branched1 (Aguilar-Martinez et al., 2007), but extremely rare in wild-type plants.

The meristematic function of LEAFY requires DNA binding

The experiments described above suggest that LFY may have two distinct and sequential functions: first, LFY may trigger the development of meristems, and then convert them into flowers. LFY DNA binding is essential for the later function, but whether it is required for the first role has never been tested. To answer this question, we took advantage of the crystallographic structure of LFY DBD (Hamès et al., 2008) and specifically disrupted the interaction of LFY with DNA: in addition to H387 and R390 involved in dimerization, we mutated the two key amino acids (K307 and R237), which contact DNA bases directly, into alanines (Figure 1f). We confirmed in vitro that the resulting LFYHARAKARA mutant protein had lost its ability to specifically bind DNA sequences (Figure 1b). Next, we ectopically expressed LFYHARAKARA in transgenic Arabidopsis: these plants were indistinguishable from wild-type plants (Table S2), demonstrating that, as for the floral function, the meristematic function of LFY requires LFY DNA binding. This result suggests that LFY fulfils this function by directly regulating downstream target genes.

The LFY meristematic function involves up-regulation of RAX1

We next assessed through which pathway LFY mediates this function. Among genes bound by LFY in ChIP experiments (Moyroud et al., 2011; Winter et al., 2011), we found several that are known to contribute to meristem or axillary meristem development: REGULATOR OF AXILLARY MERISTEMS 1 (RAX1) (Keller et al., 2006; Muller et al., 2006), participating in axillary meristem development, TERMINAL EAR LIKE2 (TEL2) (Anderson et al., 2004), which is probably involved in meristem maintenance, ARABIDOPSIS RESPONSE REGULATOR (ARR7), which is involved in cytokinin signaling (Lee et al., 2007), and GROWTH REGULATOR FACTOR5 (GRF5) (Kim et al., 2003; Horiguchi et al., 2005). To determine whether these genes are genuinely regulated by LFY (and not just bound), we analyzed their transcript levels in various tissues and genetic backgrounds where LFY expression or activity was modified relative to wild-type plants. We used inflorescences from lfy-12 and ProLFY:LFY-VP16 plants (Parcy et al., 1998) to assess regulation in floral tissues. We also used young seedlings of Pro35S:LFY transgenic plants to monitor gene expression in a context that minimizes the influence of inflorescence-specific redundant pathways. In this assay, known LFY direct target genes such as AP1, an induced target, or TERMINAL FLOWER1 (TFL1), a repressed one, served as positive controls (Figure 4a) and behaved as expected from the published data (Parcy et al., 1998, 2002; Ratcliffe et al., 1999). We found that RAX1 and TEL2 were up-regulated in ProLFY:LFY-VP16 and Pro35S:LFY, showing that LFY induces expression of these two genes in both floral and seedling tissues. ARR7 was up-regulated in all three backgrounds (seedlings of Pro35S:LFY and inflorescences of ProLFY:LFY-VP16 and lfy-12), suggesting that it is regulated by LFY (repressed in inflorescence tissues and induced in seedlings). However, GRF5 was not affected in any of the three genetic backgrounds, despite strong binding by LFY. Among these candidates, RAX1 was particularly interesting for several reasons: (i) the RAX1 locus is bound by LFY in three regions (promoter, coding sequence and 3′ untranslated region) in both Pro35S:LFY seedling and wild-type inflorescence samples (Figure 4b) (Moyroud et al., 2011; Winter et al., 2011), (ii) the spatio-temporal expression pattern of RAX1 in floral meristems correlates with LFY expression (Keller et al., 2006; Muller et al., 2006), (iii) RAX1 is required for the emergence of axillary meristems in various conditions and genetic backgrounds (Keller et al., 2006; Muller et al., 2006; Yang et al., 2012), and (iv) RAX1 up-regulation in isolated Pro35S:LFY leaves (Figure 4e) indicates that LFY is sufficient for this induction: this regulation does not require any additional flower-specific co-factor and is direct enough to occur independently of meristem development. To determine whether RAX1 contributes to meristem stimulation in Pro35S:LFYHARA plants, we characterized Pro35S:LFYHARA rax1-3 plants. Under long-day conditions, the number of shoots emerging from rax1-3 Pro35S:LFYHARA rosette leaves was markedly decreased compared to Pro35S:LFYHARA plants (Figure 4c,d), showing that LFY induces the formation and growth of axillary meristems at least in part through induction of the RAX1 gene.

Figure 4.

Identification of RAX1 as a LFY target involved in the stimulation of the axillary meristem in Pro35S:LFYHARA plants.
(a) Expression of AP1, TFL1, RAX1, TEL2, ARR7 and GRF5 in wild-type, lfy-12 mutant and ProLFY:LFY-VP16 inflorescences, and in Pro35S:LFY and wild-type whole seedlings. Expression relative to wild-type is shown. Error bars indicate standard errors, and asterisks indicate statistically significant differences compared with wild-type (P value < 0.05) according to a randomization test (Pfaffl et al., 2002).
(b) LFY and AP1 binding profiles at the RAX1 locus. Coding and untranslated regions of the RAX1 gene are shown as blue or white boxes, respectively. ChIP-Seq read coverage for LFY (Moyroud et al., 2011) and AP1 (Kaufmann et al., 2010) is shown in red and green, respectively, and ChIP/Chip data for LFY binding in inflorescences and seedlings (Winter et al., 2011) are indicated in blue and orange, respectively. The LFY-binding sites (LFY bs) predicted by the LFY binding model (Moyroud et al., 2011) are indicated as red bars (for scores above -20). The figure was created using the Integrated Genome Browser (Nicol et al., 2009).
(c, d) A mutation in the RAX1 gene partially reduces the number of precocious axillary structures in Pro35S:LFYHARA. (c) Rosette axillary structures of Pro35S:LFYHARA plants (n = 35) and Pro35S:LFYHARA rax1-3−/− plants (n = 38). Errors bars indicate standard errors of the mean. (d) Representative Pro35S:LFYHARA and Pro35S:LFYHARA rax1-3−/− rosettes were dissected to show all axillary structures growing at the base of the rosette leaves.
(e) RT-PCR analysis of RAX1 expression in rosette leaves of 1-month-old wild-type and Pro35S:LFY plants grown under long-day conditions. The SAND gene (At2g28390) was used as a constitutive control.

Discussion

LFY promotes the development of axillary meristems

LEAFY is a prominent floral meristem identity gene. It is highly expressed in the youngest incipient floral primordium to which it confers a floral fate. How meristems are turned into flowers is relatively well understood, but the molecular mechanism controlling their early emergence on the flanks of the shoot apical meristem has remained more elusive.

In this paper, we combine a variety of approaches to demonstrate that LFY is able to stimulate the growth of axillary meristems. We took advantage of knowledge of the LFY DBD 3D structure to modify its properties in a subtle and precise manner. By altering the capacity of the LFY DBD to dimerize on DNA, we engineered a modified version, LFYHARA, that partially fulfils the LFY floral function, revealing the ability of LFY to stimulate axillary bud outgrowth. This property is obvious in plants ectopically expressing LFYHARA: their architecture is profoundly altered (particularly under short-day conditions), and the CLV3 meristematic marker is precociously expressed in the leaf axils.

Various lines of evidence demonstrate that this property is a native characteristic of wild-type LFY rather than a neomorphic feature of the LFYHARA allele. First, the accelerated meristem development was also observed in plants ectopically expressing the wild-type LFY gene. These plants sometimes develop meristems in the axil of the cotyledons, a feature that is not present in the wild-type plants. Second, we established that LFYHARA not only binds to the AP1bs1 sequence in EMSA, but to the same set of DNA sequences as LFY in high-throughput SELEX. Third, we showed that a LFYHARA version that does not bind DNA (LFYHARAKARA) loses its capacity to stimulate meristem growth. It is therefore likely that LFYHARA acts by binding to genuine LFY targets.

LFY stimulates meristem growth through induction of RAX1

Our data indicate that the RAX1 gene is an important direct LFY target that is required for stimulation of axillary meristem growth. Indeed, RAX1 is strongly bound by LFY in ChIP experiments (both in Pro35S:LFY seedlings and wild-type inflorescences) (Moyroud et al., 2011; Winter et al., 2011), and its transcript level is increased in whole seedlings or isolated leaves of Pro35S:LFY plants and in ProLFY:LFY-VP16 inflorescences. Moreover, the rax1 mutation reduces the emergence of axillary shoots in Pro35S:LFYHARA plants, demonstrating that LFY induces the growth of axillary shoots through activation of RAX1.

LFY meristematic function is cryptic but important during flower meristem development

The lfy mutation compromises the emergence of a wild-type floral meristem but does not lead to a naked pin-like shoot because leaf primordia with associated axillary meristems grow instead of flowers. Similarly, during plant vegetative growth, axillary meristems emerge in the absence of LFY expression: LFY is therefore not required for meristem development on the flanks of the shoot apical meristem as other pathways exist that also drive this process. This redundancy may explain why the role of LFY in the emergence of floral meristems has remained unnoticed. However, when the lfy mutation is combined with defects in other genes, the importance of LFY in this process becomes clear. FIL and CLV1 are two genes that act during floral meristem development (Clark et al., 1993; Chen et al., 1999; Sawa et al., 1999a,b). fil and clv1 single mutants show some floral meristem initiation defects, with occasional empty cauline leaves or filamentous structures. The lfy mutation greatly enhances the fil and clv1 floral defects. In double mutant plants, most floral primordia are replaced by empty cauline leaves or filaments, which are determinate structures lacking meristematic potential (Sawa et al., 1999a). PINOID (PID) is another regulator (from the auxin signaling cascade) that contributes to meristem development (Bennett et al., 1995; Christensen et al., 2000). pid floral meristem defects are also strongly enhanced in the pid lfy double mutant even under growth conditions where the pid single mutant only displays a very mild phenotype (Ezhova et al., 2000). Another striking example comes from analysis of BOP genes (Norberg et al., 2005): empty cauline leaves replace some flowers in the bop1 bop2 lfy triple mutant but not in the bop1 bop2 double mutant. Thus, LFY does stimulate floral meristem outgrowth, but this role is masked by redundant pathways.

Ectopic expression of LFY at a time when these pathways are less active revealed the importance of RAX1 induction by LFY for precocious emergence of meristems in the axils of rosette leaves. There is evidence that this induction also occurs in flowers: (i) RAX1 mRNA is present in early stage 1 flower meristems just after LFY expression starts (Keller et al., 2006; Muller et al., 2006), (ii) LFY strongly binds the RAX1 locus at three locations in inflorescence tissues (Winter et al., 2011), and (iii) expression of LFY-VP16 induces RAX1 expression in inflorescences (Figure 4a). However, expression of RAX1 is not lower in the single lfy mutant, probably because it is also induced by other regulators such as the LATERAL ORGAN FUSION (LOF) genes (Lee et al., 2009). Similarly, the single rax1 mutation does not affect development of flowers but only suppresses some axillary meristems at the base of leaves from plants grown under short-day conditions. Genes such as LATERAL SUPPRESSOR (LAS) (Greb et al., 2003) and REGULATOR OF AXILLARY MERISTEM FORMATION (ROX) (Yang et al., 2012) have been shown to act in parallel with RAX1 to stimulate meristem growth. Interestingly, these two genes are required for axillary meristem emergence but not for flowers, although they are both expressed in floral primordia: they are thus prominent candidates to act redundantly with LFY during the reproductive phase.

Down-regulation of ARR7 expression by LFY (as indicated by analysis of the lfy-12 mutant) may also contribute to meristem emergence. ARR7 is a negative regulator of cytokinin signaling (Lee et al., 2007). Cytokinin is a phytohormone that is known to promote axillary meristem growth (Tantikanjana et al., 2001) and is directly repressed by WUS during early flower development (Leibfried et al., 2005). As LFY and WUS interact to induce AG expression (Lohmann et al., 2001), it is possible that they also collaborate to repress ARR7. A direct interaction between LFY and the ARR7 promoter was observed both in vitro (Leibfried et al., 2005) and in vivo in seedlings ectopically expressing LFY (Moyroud et al., 2011; Winter et al., 2011). In addition to the cytokinin pathway, auxin signaling may also be important downstream of LFY: several genes from the auxin pathway are bound by LFY in ChIP experiments (Moyroud et al., 2011; Winter et al., 2011), and LAS has been shown to act upstream of auxin components (Greb et al., 2003). Meristem emergence may thus result from both convergence of LFY and LAS action on hormones and stimulation of the RAX1 pathway.

AP1 and LFY share the floral function, but only LFY stimulates meristem development

We observed that constitutive AP1 expression only mildly affects meristem development compared to LFY. The fact that AP1 does not participate in meristem outgrowth is supported by the phenotypes of double mutants such as ap1 clv1, ap1 fil or ap1 bop1 bop2 that do not display the empty cauline leaves or filamentous structures observed in double or triple mutants that include the lfy mutation (Clark et al., 1996; Chen et al., 1999; Xu et al., 2010). This may seem surprising given that LFY and AP1 share many floral functions and bind to a common set of target genes, including the RAX1 locus (Figure 4). However, the fact that RAX1 is not induced but instead is repressed by AP1 (Kaufmann et al., 2010) provides a striking explanation for the differential effect of these two floral meristem identity genes on meristem stimulation: AP1 probably contributes to the loss of meristematic character of the floral meristem established by LFY.

A developmental scenario for meristem emergence and floral fate determination

The data provided here and elsewhere suggests the following sequence of events for early flower development (Figure 5). The site of leaf or flower meristem emergence is determined by the auxin maximum (Kuhlemeier, 2007; Vernoux et al., 2010). After the floral transition, auxin accumulation stimulates LFY expression in the floral anlage (stage 0). LFY contributes to outgrowth of the meristem and expression of meristematic genes such as WUS, CLV3 and UFO by inducing RAX1 and probably repressing ARR7 together with WUS. Together with some of its own targets such as AP1 or LMI1 (Saddic et al., 2006), LFY also confers floral fate to the nascent meristem (stage 2). Acquisition of floral identity involves the induction of floral homeotic genes as well as loss of inflorescence meristem genes through repression of RAX1, SUPPRESSOR OF OVEREXPRESSION OF CONSTANS1 (SOC1), AGAMOUS-LIKE24 (AGL24) and SHORT VEGETATIVE PHASE (SVP) by AP1 (Liu et al., 2007; Kaufmann et al., 2010) and of WUS by AG (Lohmann et al., 2001). This scenario provides another illustration of the importance of the timing of these feed-forward loops for proper development of the flower. Controlling both the emergence of the meristem and its floral determination by the same regulator (LFY) may be an efficient way to couple both events and ensure they do not occur independently, when environmental conditions affect the cell division rate for example. Firm demonstration of the role of the LFY-RAX1 module in flowers will require further investigation in genetic contexts with reduced redundancy between pathways.

Figure 5.

Speculative model in which the links between LFY, RAX1 and ARR7 are placed in the context of flower meristem development and integrated with other known regulation pathways. The numbers on the inflorescence section indicate floral stages.

The action of LFY on meristems explains the lfy phenotypes in various plants

Homologs of LFY are found in all angiosperms, but the phenotypes of lfy loss-of-function mutants differ in some species. Moyroud et al. (2009, 2010) proposed that a meristematic function of LFY provides a common explanation for the phenotypes observed in various species, such as the reduced number of flowers and tillers of the rice floricaula leafy (rfl) mutant (Kyozuka et al., 1998; Rao et al., 2008; Ikeda-Kawakatsu et al., 2012) or simple leaf development when the LFY ortholog is mutated in pea (Pisum sativum) or Medicago truncatula (Hofer et al., 1997; Champagne et al., 2007; Wang et al., 2008). It will be interesting to determine whether the molecular mechanism unraveled in Arabidopsis is shared by other species, and whether homologs of RAX1 contribute to rice tiller or inflorescence development or to leaf dissection in some legumes. In tomato, two RAX1 co-orthologs, named BLIND (BL) and POTATO LEAF (C), control shoot architecture and leaf development (Schmitz et al., 2002; Busch et al., 2011). Mutations of these genes reduce plant branching and leaf complexity. Interestingly, mutations in FALSIFLORA (FA), the LFY ortholog in tomato (Solanum lycopersicum), also affect both processes (Molinero-Rosales et al., 1999). These common phenotypes, together with the overlapping expression patterns of FA and BL/C, are consistent with possible regulation of BL/C genes by FA in tomato. In the moss Physcomitrella patens, LFY homologs were shown to act in the first cell division after fertilization (Tanahashi et al., 2005). It is therefore possible that the ancestral function of LFY was to control of cell division, and that LFY specialized in the emergence of meristems before being co-opted to confer floral fate to the meristems it produced.

Experimental procedures

Plant material and growth conditions

The Pro35S:AP1 seeds were a gift from Detlef Weigel (Max Planck Institute, Tübingen, Germany). The ProCLV3:GUS line (Brand et al., 2002) and the rax1-3 mutant T-DNA line (SALK_071748) (Muller et al., 2006) were obtained from the Nottingham Arabidopsis Stock Centre (http://arabidopsis.info). All mutants and transgenic lines used are in the A. thaliana Columbia-0 accession.

Seeds were sown on soil or surface-sterilized and grown in Petri dishes on Murashige and Skoog basal salt mixture medium (Sigma-Aldrich, www.sigmaaldrich.com). Plants were transferred to soil and grown at 22°C under long-day conditions (16 h of 100 μE light) or short-day conditions (8 h of 100 μE light). The ProLFY:LFY-VP16 plants have been described previously (Parcy et al., 1998).

Plant genotyping

Plants carrying the lfy-12 mutation were genotyped using the endpoint genotyping program with a Light Cycler 480 (Roche, http://www.roche.com) as described by Benlloch et al. (2011) or using the following dCAPS primers: forward (5′-TCAAGCACCACCTCCGGTTCCACCTCCA-3′) and reverse (5′-CGGACGAAAACCCTACGCTGAACCACCA-3′). Only the wild-type 85 bp PCR product is cleaved by the Fnu4HI restriction enzyme. The RAX1 locus was identified by PCR using primers oHC150 (5′-TGTGAAAAGACCAACCTCACC-3′) and oHC151 (5′-TCGGACATTTCAGTTTGGAAG-3′) for RAX1, and the rax1-3 allele was identified using primers oHC150 and oHC152 (5′-ATTTTGCCGATTTCGGAAC-3′).

Plant vectors and transformation

The pFP100 binary vector (Bensmihen et al., 2004), which contains the ProAt2S3:GFP in planta selection marker, was used as a backbone for the following binary constructs: ProLFY:LFY (pETH29), ProLFY:LFYHARA (pETH39), Pro35S:LFY (pCA26) and Pro35S: LFYHARA (pSP3). Details of the cloning procedures can be provided upon request. Agrobacterium tumefaciens C58 pMP90 was used for stable transformation of wild-type or lfy-12 heterozygous plants.

Yeast vectors and one-hybrid assay

For yeast experiments, the p424 vector with the GAL1 promoter was used to clone LFYHA-VP16 (pETH32), LFYRA-VP16 (pETH46) and LFYHARA-VP16 (pETH31) fusions either by conventional cloning or by using the megaprimer strategy (Kirsch and Joly, 1998). Yeast vectors containing LFY and LFY:VP16 (pFP13 and pFP14, respectively) and the AGAMOUS:LACZ reporter construct (pFP50) have been described previously (Parcy et al., 1998; Lohmann et al., 2001). Detailed maps for all constructs are available on request. For one-hybrid experiments, effectors (pFP13, pFP14, pETH31, pETH32 or pETH46) and reporter plasmids (pFP50) were co-transformed into Saccharomyces cerevisiae strain EGY48 (Invitrogen, www.lifetechnologies.com). β-galactosidase measurements were performed as described previously (Lohmann et al., 2001), except that they were adapted to the Safire2 microplate reader (Tecan, www.tecan.com).

Expression and purification of recombinant LFY proteins

The pET-30a vector (Novagen, www.merckmillipore.com) was used to clone recombinant near full-length LFY [LFY(∆40), pETH79], LFYHARA(∆40) (pETH156) and LFYHARAKARA(∆40) (pETH180). Because it greatly facilitated expression and purification of these LFY recombinant proteins, the non-conserved first 39 amino acids were omitted in these constructs. The proteins were expressed in the Escherichia coli Rosetta™2 (DE3) strain (Novagen, http://www.emdmillipore.com/), and purified on Ni Sepharose™ high-performance resin (GE Healthcare, www.gehealthcare.com) as described in Methods S1.

EMSA experiments

Electrophoretic mobility shift assay were performed as previously described (Moyroud et al., 2011). AP1 oligonucleotides (5′-GTTGGGGAAGGACCAGTGGTCCGTACAATGT-3′ and 5′-ACATTGT ACGGACCACTGGTCCTTCCCCAA-3′) labeled with the Cy5 fluorophore were prepared as described by Moyroud et al. (2011). The binding reactions were performed over 15 min using 100, 500 or 1500 nm LFY in 20 μl binding buffer (10 mm HEPES, 1 mm spermidine, 14 mm EDTA, 30 μg BSA, 0.25% CHAPS, 3% glycerol, 3 mm tris(2-carboxyethyl)phosphine (TCEP), pH 8, and 28 ng/ml fish sperm DNA), and loaded on a native gel (6% acrylamide, 0.5 × TBE). Gels were electrophoresed at 90 V for 70 min at 4°C, and scanned on the Typhoon 9400 scanner (Amersham Biosciences, www.gehealthcare.com).

Systematic evolution of ligands by exponential enrichment experiments

Systematic evolution of ligands by exponential enrichment experiments were performed using fluorescent 73-mers and LFY(Δ40) or LFYHARA(Δ40) proteins. Initially, a random library was synthesized using 73-mers 5′-TGGAGAAGAGGAGAGATCTAGC(N)30CTTGTTCTTCTTCGATTCCGG-3′ as template with a fluorescent TAMRA-labelled forward primer (SElex-F, 5′-TGGAGAAGAGGAGAG ATCTAG-3′) and a non-labeled reverse primer (SElex-R, 5′-CCGG AATCGAAGAAGAACAA-3′), as previously described (Moyroud et al., 2011).

For each selection cycle, 1 μm protein was mixed with 10 nm fluorescent dsDNA in 225 μl Selex buffer (20 mm Tris pH 8, 250 mm NaCl, 2 mm MgCl2, 5 mm TCEP, 60 μg/ml fish sperm DNA and 1% glycerol). After a 15 min incubation, 25 μl Ni-NTA magnetic beads (Qiagen, www.qiagen.com), previously equilibrated in Selex buffer, were added to the reaction mix to immobilize DNA/protein complexes via the His tag of the protein. After 30 min incubation, the reaction mix was placed on a tube magnet for 1 min, and the supernatant was removed from the separated beads to eliminate the unbound DNA. Five washes were subsequently performed at 4°C, each of them consisting of adding 250 μl Selex buffer containing 20 μg/ml fish sperm DNA, followed by 2 min incubation and 1 min on the tube magnet to discard the supernatant. Finally, the magnetic beads were resuspended in 50 μl Selex buffer without fish sperm DNA. Selected 73-mers were amplified by PCR as described previously (Moyroud et al., 2011) using 1 μl of the magnetic beads solution as template. PCR products were quantified as previously described (Moyroud et al., 2011), and the selection cycle was repeated three times, each time using the newly synthesized fluorescent DNA as a library. The 73-mers libraries obtained after four cycles of selection were subsequently used for barcoding and high-throughput sequencing.

Barcoding for Illumina sequencing of the SELEX libraries

Oligonucleotides allowing hybridization with the sequencing primer fused to a 6 bp barcode (one unique barcode per library) were added by PCR (PCR#1) to 73-mers of each library selected, using Phusion® DNA polymerase (Ozyme, http://www.ozyme.fr) and 0.5–2 μl of the magnetic beads solution as template, extending the 73-mers into 146 bp dsDNA. Three PCR cycles (PCR#2) were subsequently performed, using 5 μl of PCR#1 product as template, Phusion® DNA polymerase (Ozyme, http://www.ozyme.fr) and Illumina oligonucleotides (http://www.illumina.com/, allowing fixation of the dsDNA to the Illumina sequencing chip. PCR products from PCR#2 (171 bp) were purified using a NucleoSpin Extract II kit (Macherey-Nagel, http://www.mn-net.com), and sequenced on an Illumina platform at the Max Planck Institute for Developmental Biology, Tübingen, Germany.

The 2000 unique most frequent sequences were aligned using MEME software version 4.3.0 (Bailey and Elkan, 1994) using default parameters and imposing symmetry to construct the LFY(Δ40) and the LFYHARA(Δ40) matrix.

Gene expression analysis

RNA extractions were prepared from 50 mg plant tissues using and RNeasy plant RNA extraction kit (Qiagen). Extractions were obtained from 21-day-old seedlings (Col-0 and Pro35S:LFY plants cultivated under short-day conditions) or young apices from which older flowers were removed (Col-0, lfy-12 and ProLFY:LFY-VP16 plants cultivated under long-day conditions). Total RNAs were treated with Ambion® turbo DNA-free™ DNase (Invitrogen) to reduce genomic DNA contamination. Reverse transcription was performed using MMLV reverse transcriptase (Sigma-Aldrich) from 1 μg of total RNA.

Primers used for real-time PCR were designed using Quantprime software (www.quantprime.de/) and are listed in Table S4. PCR reactions were performed in a 15 μl final volume using 0.1 μm for each primer, 1 × mix SYBR Green JumpStart® (Sigma-Aldrich) and 3.125 ng cDNA in a Rotor Gene 3000 (Qiagen) with default cycle parameters. The relative expression of a target gene was calculated by the ΔΔCt method as described by Pfaffl (2001) using the reference gene At2g28390 as previously described (Czechowski et al., 2005; Guenin et al., 2009). Statistical analyses were performed using REST 2009 software (Pfaffl et al., 2002).

Decapitation experiment

Plants were grown for 1 month under short-day conditions and then transferred to long-day conditions. The primary inflorescence was cut once it reached 10–15 cm, i.e. after 20 days for Pro35S:LFYHARA or 26 days for wild-type and lfy-12 plants. The number of secondary inflorescences was scored on the day of decapitation and 10 days later. Parametric statistics for the percentage of rosette leaves bearing a shoot or flower were applied after arcsine root transformation; 95% confidence intervals were found with the usual t distribution and then back-transformed to percentages.

Histology

Tissues were fixed in formaldehyde/acetic acid/ethanol (FAA), and infiltration with paraffin was performed in an ASP300 embedding automat (Leica, http://www.leica.com). Thin sections (9–12 μm) were prepared on an EG1160 microtome (Leica) and transferred to glass slides. Staining was performed with toluidine blue. Images were taken on an Axioplan2 (Zeiss, www.zeiss.com) equipped with an AxioCam HRc digital camera (Zeiss). GUS staining was performed as described previously (Parcy et al., 1998).

Model describing the appearance of the axillary meristem staining

We assumed that the probability of a plant having entered the phase of meristematic development (production of axillary meristem at the axil of rosette leaves) was a sigmoid function of time, characterized by its mean dswitch and its temporal uncertainty τ; thus this probability on day d was modeled as P (meristematic | d) = σ ((d – dswitch)/τ), where σ is the logistic function: σ (x) = 1/[1 + exp (−x)]. In the meristematic development phase, we assumed that a Poisson process of rate λ governed the appearance of meristems: thus the probability of observing n meristems t days after the switch to the meristematic phase is π (n, λ, t) = (λt)n/n! exp (−λt); before this meristematic phase, P (= 0) = 1. With these assumptions, the overall probability of observing > 0 meristems on day d is

display math

By maximizing the likelihood of the experimental meristem counts in this model, we derived the parameters dswitch, τ and λ for each strain. Likelihood ratio tests were then used to test the statistical significance of the parameter differences across strains. As τ and λ were not significantly different across strains (P > 0.1 for both), the model illustrated on Figure 3(a) was constrained to use a single value for each of these two parameters, with three values of dswitch (one for each strain).

Acknowledgements

We thank R. Dumas, R. Benlloch (UPSC, Umeå, Sweden), O. Nilsson (UPSC, Umeå, Sweden), A. Mathelier (CMMT, Vancouver, Canada), W. Wasserman (CMMT, Vancouver, Canada), P. Baraduc (CNC, Lyon, France) and M. Nanao (EMBL, Grenoble, France) for help with experiment or data analysis, D. Weigel (MPI, Tuebingen, Germany) and N. Warthman (MPI, Tuebingen, Germany) for SELEX sequencing, P. Cubas (CNB, Madrid, Spain) and P. Doerner for discussion, and V. Pautot (INRA, Versailles, France), L. Yan (Dept. of Organismic and Evolutionary Biology Harvard University, USA) and P. Laufs (INRA, Versailles, France) for critical reading of the manuscript. This work is supported by grants from the Charmful ANR (Agence Nationale de la Recherche) program and by the Natural Sciences and Engineering Research Council of Canada to V.G.

Ancillary