The Myb-domain protein ULTRAPETALA1 INTERACTING FACTOR 1 controls floral meristem activities in Arabidopsis

Plants differ from the majority of animals by their ability to produce new organs throughout their life cycle. Continuous organogenesis is allowed by the maintenance of stem cell reservoirs sustained through continuous cell divisions in dynamic structures called meristems (Carles and Fletcher, 2003; Holt et al., 2014; Gaillochet et al., 2015). The shoot apical meristem (SAM) is located at the growing apex and produces cells that will be incorporated into new structures such as lateral organs. Cell divisions constantly replenish the meristem in its central zone (CZ), thus providing a cell reservoir for the adjacent peripheral zone (PZ). PZ cells further differentiate into lateral organ primordia on the meristem flanks (Irish and Sussex, 1992; Leyser and Furner, 1992). Proper, continuous growth of the plant thus requires long-term function of the SAM, itself depending on a coordinated balance between cell proliferation and differentiation (Laux, 2003; Ha et al., 2010).

In Arabidopsis thaliana, SAM homeostasis is governed by a feedback loop signalling network involving the CLAVATA (CLV) and WUSCHEL (WUS) factors (Ha et al., 2010). The WUS homeobox transcription factor is both necessary and sufficient for stem cell specification (Laux et al., 1996; Mayer et al., 1998; Schoof et al., 2000; Lenhard et al., 2002; Gallois et al., 2004; Leibfried et al., 2005). In mutants with defective WUS function, the activity of the SAM stops at early stages of development, correlating with a lack of stem cells. WUS is expressed in a small central meristematic zone called the organizing centre (OC). The WUS protein, after being synthesised in cells of the OC, migrates in the adjacent cells of the CZ where it promotes stem-cell identity and both directly and indirectly regulates transcription of the CLV genes (Busch et al., 2010; Yadav et al., 2011; Schuster et al., 2014). In turn, the CLV3 signalling peptide restricts WUS expression to the OC (Mayer et al., 1998; Brand et al., 2000), via the CLV1-CLV2-CRN receptor protein complexes and the RPK2 receptor-like kinase (Fletcher et al., 1999; Rojo et al., 2002; Lenhard and Laux, 2003; Müller et al., 2008; Ogawa et al., 2008; Wang and Fiers, 2010; Kinoshita et al., 2010). As a consequence, clv loss-of-function mutants produce a widely enlarged SAM and more lateral organs, owing to extreme expansion of the WUS expression domain (Clark et al., 1993; Schoof et al., 2000). Thus, the WUS-CLV negative feedback loop is crucial for the maintenance of meristem size and function.

During the vegetative stage, the SAM produces leaves and axillary meristems that reiterate SAM development, whereas during the reproductive stage, SAM and axillary meristems produce flower meristems (FMs), which develop into flowers (Alvarez-Buylla et al., 2010). In A. thaliana, flowers are made of four whorls, or concentric rings, each carrying a fixed number of organs with distinct identities: four sepals, four petals, six stamens and two fused carpels. Stem cell activity at the FM determines the number of organs carried by each whorl and is regulated by the same WUS-CLV feedback loop, as flowers of clv mutants produce more organs per whorl, whereas wus hypomorphic mutants produce fewer floral organs than do wild-type (WT) plants (Laux et al., 1996; Schoof et al., 2000). Moreover and in contrast to the SAM, FM growth is determinate as organ production ceases once the carpels are initiated in the innermost whorl of the flower (Prunet et al., 2009). FM termination relies on the timely expression of the flower homeotic gene AGAMOUS (AG), which encodes a MADS domain-containing protein that both directly and indirectly represses WUS expression in the centre of the FM, leading to extinction of stem cell activity at stage 6 (Lenhard et al., 2001; Lohmann et al., 2001; Sun et al., 2009, 2014; Liu et al., 2011).

The A. thaliana ULTRAPETALA1 (ULT1) factor is a key negative regulator of stem cell activity in the shoot apical and floral meristems (Fletcher, 2001; Carles et al., 2004, 2005). In particular, ult1 loss-of-function mutants produce more flowers and supernumerary floral organs of each type. In these mutants, the inflorescence and flower meristem sizes are enlarged as a result of the expansion of the WUS expression domain (Carles et al., 2004, 2005). Thus, ULT1 maintains shoot and floral meristem function by restricting WUS expression to the OC. Moreover, FM termination is delayed in ult1 mutants (Fletcher, 2001; Prunet et al., 2008), leading to the production of additional organs inside the innermost whorl, such as stamens and carpelloid structures. This phenotype correlates with a delay in AG induction, which, in turn, delays WUS extinction in the centre of the FM (Fletcher, 2001; Carles et al., 2004). Thus, ULT1 acts as a negative regulator of WUS expression in the FM, and this function is required for both meristem homeostasis and termination (Carles et al., 2004). The ULT1 gene encodes a plant-specific protein composed of a SAND (Sp100, AIRE-1, NucP41/75 and DEAF-1) putative DNA-interacting domain and a B-box, which is likely to be involved in protein-protein interaction (Carles et al., 2005). It has been proposed that ULT1 directly activates AG at the centre of the FM, via interaction with the chromatin activator ARABIDOPSIS HOMOLOG OF TRITHORAX 1 (ATX1) (Alvarez-Venegas et al., 2003; Ding et al., 2011), leading to a reduction in trimethyl marks at lysine 27 of histone 3 (H3K27me3, repressive chromatin mark) at the target locus (Carles and Fletcher, 2009, 2010; Engelhorn et al., 2014a). Thus, ULT1 acts as a positive regulator of AG that contributes to the extinction of WUS expression in stage 8 flowers.

However, several questions remain open concerning the precise molecular mechanism(s) through which ULT1 regulates flower meristem activity. How does ULT1 repress WUS for SAM and FM homeostasis at early stages of flower development? Could ULT1 directly function in WUS repression? How does ULT1 interact with DNA and what defines its target gene specificity? In order to elucidate the answers to these questions, we searched for novel ULT1 interacting partners. Here, we identify and characterise ULT1 INTERACTING FACTOR 1 (UIF1), a Myb domain-containing transcription factor. We found that loss of function in UIF1 causes phenotypes similar to ult1 mutant phenotypes, such as production of supernumerary floral organs. We show that UIF1 and ULT1 function in the same pathway for the regulation of organ number in the flower. Moreover, UIF1 displays DNA-binding activity and specifically binds to WUS regulatory elements. Because UIF1 protein displays a transcriptional repressor activity, we propose that UIF1 and ULT1 work together in FM homeostasis, via direct regulation of WUS expression.

To identify novel interactors of ULT1, we performed a yeast two-hybrid (Y2H) screen, using an A. thaliana inflorescence and flower cDNA library and the full-length ULT1 open reading frame (ORF) as bait. Upon selection, we isolated a clone corresponding to At4g37180.1 as annotated in the tair database (https://www.arabidopsis.org/servlets/TairObject?id=126622&type=locus), which we named ULT1 INTERACTING FACTOR 1 (UIF1) (Fig. 1A). A targeted binary Y2H test using the full-length ULT1 and UIF1 proteins confirmed the interaction (Fig. 1B; Fig. S1A). The UIF1-ULT1 interaction was next validated in planta by a bimolecular fluorescence complementation (BiFC) assay, yielding an interaction signal in the nucleus of transformed cells (Fig. 1C; Fig. S1B). The 1804-bp long UIF1 gene (At4g37180) encodes a 356-aa long Myb domain-containing protein of unreported function, belonging to the GARP G2-like subfamily of transcription factors (Fig. 2A,B). The Myb-like putative DNA-binding domain of UIF1 (aa 209-268) contains a SHLQKYR motif at its C-terminus (Fig. 2B); this motif is found in other plant proteins (AtCCA1; AtLHY; LeMYB1; ZmMRP-1; MybSt1) and reported as a specific DNA-binding domain (Wang et al., 1997; Schaffer et al., 1998; Rose et al., 1999; Gómez et al., 2002; Baranowskij et al., 1994). UIF1 protein also contains two distinct EAR-like motifs (LxLxL, with L indicating leucine and x indicating any aa) at its N- and C-terminal ends, as well as a predicted nuclear localisation signal (NLS) (Hiratsu et al., 2002; Ohta et al., 2001) (Fig. 2B). We found six homologues for UIF1 in A. thaliana (Fig. S2A,B), which share 86-91% similarity over their Myb domain and 43-63% similarity over the full-length protein. Another globular domain rich in hydrophobic amino acids (aa 45-86) such as L and isoleucine (I) shows 78-81% similarity between all homologues (Fig. S2A). We identified orthologues of UIF1 among monocots and dicots (Fig. 2B), which share strong similarity both over the hydrophobic and Myb domains (68-83% and 88-96%, respectively). Interestingly, although the N-terminal EAR motif is conserved among all UIF1 orthologues, UIF1 differs by the presence of an additional LDLEL EAR-like motif at its C-terminus end (Fig. 2B). Phylogenetic reconstructions generated with UIF1 orthologues in eudicots suggest that, unlike At1g13300 and At3g25790 on one hand, and At1g68670 and At1g25550 on the other hand, the UIF1 gene does not result from duplication in the Brassicaceae lineage (Fig. S2C).

In order to map the regions necessary for ULT1-UIF1 interaction, we tested truncated versions of ULT1 and UIF1 by Y2H and BiFC (Fig. 1A,C). We showed that in the absence of the B-box domain (Carles et al., 2005), ULT1 (ULT1 1-193) does not interact with UIF1 in Y2H (Fig. 1B) or in BiFC (Fig. 1C). Furthermore, the ULT1 B-box domain alone (aa 116-237) could interact with UIF1 and restore yeast growth on selective medium (Fig. 1A,B). Similarly, we showed that deletion of the N-terminal domain of UIF1 (UIF1 93-364), which contains the (L/I)-rich motif, resulted in the loss of interaction with ULT1 (Fig. 1A,C). Taken together, these results suggest that UIF1 can interact with the ULT1 protein and that ULT1 B-box and UIF1 N-terminus are required for this interaction.

To investigate the functional significance of UIF1-ULT1 interaction, we analysed the temporal and spatial expression pattern of UIF1 in A. thaliana WT tissues. Quantitative real-time RT-PCR (RT-qPCR) analysis showed that UIF1 is expressed in all tested tissues: roots, 6-day-old (do) seedlings, inflorescences, closed flowers and open flowers, with highest levels in 6-do seedlings and inflorescences (Fig. 2C). This profile is similar to that of ULT1 and was confirmed by microarray data grouped in the public eFP Browser database (http://bbc.botany.utoronto.ca/efp/cgi-bin/efpWeb.cgi; Fig. S3A). In particular, ULT1 and UIF1 expression patterns overlap in vegetative shoot apex, floral buds and floral organs especially in stamens and carpels (in stage 12-15 flowers). At the cellular level, RNA in situ hybridisation experiments on inflorescence and flower tissue sections (Fig. 2D-F; Fig. S3B,C) showed that UIF1 is expressed throughout the inflorescence meristem (Fig. 2D) and in floral primordia from stage 2 onwards, being restricted later to stamens and to the adaxial side of carpel primordia (Fig. 2E). In mature flowers, UIF1 transcripts were also found in ovules (Fig. 2F). This pattern largely overlaps with that of ULT1 transcripts (Carles et al., 2005). Finally, at the subcellular level, we found that the GFP-UIF1 fusion protein localises in the nucleus and in cytosolic foci in the cell (Fig. 2G), a pattern very similar to that reported for the ULT1 protein (Carles et al., 2005).

Altogether, similarities in expression patterns and protein subcellular localisation further support a molecular interaction for ULT1 and UIF1 and a binomial function in inflorescences.

To assess whether UIF1 functions in the same developmental processes as ULT1, we analysed two independent A. thaliana T-DNA insertion lines (Fig. 3A). The uif1-1 allele contains a T-DNA insertion 132 bp upstream of the stop codon, creating an early stop codon 72 bp after the insertion (Fig. 3A). In the uif1-3 allele, the T-DNA is located in the promoter region, 255 bp upstream of the start codon (Fig. 3A). RT-PCR analyses on inflorescences showed that for both homozygous mutants, no cDNA could be detected for the full-length UIF1 transcript (Fig. 3B). In the uif1-1 line, the transcript detected upstream of the T-DNA insertion site would result in the synthesis of a truncated UIF1 protein lacking the predicted NLS domain and the C-terminus EAR-like motif (Fig. 3A). Whether the truncated transcript detected in uif1-1 translates into the production of a truncated protein or no protein at all remains to be addressed. We found that both uif1 mutant alleles produce flowers with more sepals and more petals than the WT control (Fig. 3C,D,G,H,K,L,O,P; Table 1), a phenotype reminiscent of that observed in ult1 mutants (Fletcher, 2001; Carles et al., 2005). In addition, flowers of both uif1 mutants display loss of organ boundary and identity (Table 2), producing fused sepals (Fig. 3G) as well as petaloid stamens (Fig. 3I,M,Q; Fig. S4A,B) and branching stamens fused along the filament (Fig. 3I,J,N; Fig. S4C). Altogether, these observations suggest that UIF1 regulates floral morphogenesis and controls organ number similarly to ULT1. Moreover, UIF1 regulates additional features in the flower, such as cell fate and organ identity.

Whereas ult1 mutant flowers display supernumerary organs of each type (Fletcher, 2001; Carles et al., 2005), uif1 mutants consistently produce supernumerary sepals and petals but to a lesser extent than ult1 mutants (Fig. 4; Table 3). To test whether ULT1 and UIF1 are involved in a same regulatory pathway to control organ number, uif1-3 was crossed to the ult1-3 null mutant. There was no significant difference between uif1 ult1 and ult1 mutant phenotypes for additional sepals and petals (Fig. 4; Table 3), supporting the hypothesis that UIF1 and ULT1 function in the same pathway for regulating organ numbers in the perianth. However, modified organs, such as branched stamens, were observed in flowers of uif1 ult1 and uif1 mutants but never in ult1-3, supporting an additional role of UIF1, independent of ULT1 function, in the control of flower organ identity.

To check whether the Myb-containing UIF1 protein is able to bind DNA on specific target sequences, a protein binding microarray (PBM) experiment (Berger and Bulyk, 2009; Godoy et al., 2011; Franco-Zorrilla et al., 2014) was performed. Analysis of DNA fragments bound by MBP-UIF1 recombinant protein yielded a series of motifs recognised with high affinity (E-score >0.45), resembling DNA sequences recognised by other transcription factors belonging to the same GARP G2-like subfamily (Franco-Zorrilla et al., 2014; Weirauch et al., 2014) (Fig. 5A). Position weight matrices of the three top-scoring motifs were used to scan regulatory sequences in putative target genes of UIF1 using Morpheus (http://biodev.cea.fr/morpheus), an algorithm that predicts binding sites in given sequences by computing affinity scores related to the used matrix (Fig. 5A). Because uif1 mutants phenocopy ult1 mutants, we assessed whether WUS could be a direct target of UIF1 (Carles et al., 2005). Hence, binding sites for UIF1 were searched and found in the promoter of WUS. The top two sites (Fig. S5; Fig. 5A) lie in specific regulatory regions as reported in the WUS promoter scanning study of Bäurle and Laux (2005). The WUS-1 site (position −1182) is located in proximity to the quantitative element required for enhanced expression levels in the meristem, and the WUS-2 site (position −2819) is embedded in the region for expression intensity in FM and ovule. Protein-DNA interactions were tested by an electrophoretic mobility shift assay (EMSA) using purified recombinant His-tagged GST-UIF1 fusion protein (Fig. 5A,B; Fig. S6). We found that UIF1 specifically binds the GTAGATTCCT motif (WUS-1), as interaction is lost upon mutation of either of two single bases at positions 5 or 8 in the sequence. Similar results were obtained when using the full-length protein or the isolated Myb domain of UIF1 fused to GST (Fig. 5B; Fig. S6). This suggests that UIF1 might directly regulate WUS.

We showed in this study that UIF1 physically interacts with ULT1, previously reported to control AG expression. Moreover, we found that some of the uif1 phenotypes, such as organ fusions, are reminiscent of those observed in the cup-shaped cotyledon (cuc) mutant backgrounds (Aida et al., 1997). This prompted us to seek putative UIF1-binding sites at the AG and CUC loci (Takada et al., 2001; Aida et al., 1997; Vroemen et al., 2003). Three UIF1 binding sites were found in AG regulatory sequences (Fig. 5A). Among these, the best-scored AAGAATCTTT site, present in the large regulatory intron of AG, was confirmed by EMSA (Fig. 5B; Fig. S6). We found that UIF1 specifically binds this motif, as interaction is lost upon mutation of a single base (c7a) in the sequence. UIF1-binding sites were also found in the promoters of the CUC genes: two in CUC1, three in CUC2 and two in CUC3 (Fig. 5A). Specific binding to motifs in the CUC genes was confirmed by EMSA (Fig. S6) and interaction with the UIF1 Myb domain was strongly reduced upon a single base mutation (a5t). These data indicate that UIF1 probably has a direct effect on AG and CUC gene regulation.

To test whether UIF1 has a role in regulating transcription, we performed a dual luciferase reporter assay (DLRA) in onion epidermis cells (Fig. 5C). For this experiment, UIF1 protein was fused to the Gal4 DNA-binding domain (G4DBD-UIF1), which binds to the upstream activation sequence (UAS) site located upstream of the CaMV35S minimal promoter driving the expression of firefly luciferase. The DLRA test showed that firefly luciferase expression was repressed by a mean of 2.7-fold in the presence of UIF1 (G4DBD-UIF1) compared with the inert control (G4DBD-GFP), indicating that UIF1 might function as a transcriptional repressor. Even though repression was not as efficient as for the EAR-containing protein IAA12/BODENLOS (BDL; 10.8-fold repression reduction; Szemenyei et al., 2008), we tested whether UIF1 activity was due to EAR motifs mapped at the N-terminus (EAR^N) and C-terminus (EAR^C) of the protein (Fig. 2B). We used two modified versions of UIF1: a truncated version lacking EAR^C (UIF1-dCT) or the truncated version with additional mutations in EAR^N by substituting leucines with alanines (UIF1-ANANA-dCT). Although deletion of EAR^C did not significantly affect UIF1 activity, additional mutation in EAR^N completely abolished its activity (Fig. 5C). Thus, the conserved EAR^N motif of UIF1 is necessary for transcriptional repression.

Therefore, in A. thaliana, UIF1 could be a transcriptional repressor of WUS during FM development. To investigate this, we performed RT-qPCR analyses in inflorescences of the two mutant alleles, uif1-1 and uif1-3. In order to exclude WUS RNA derived from anthers, we used dissected inflorescences, harbouring FM younger than stage 5 (Fig. S7). We systematically found a significant increase in WUS expression in the mutants (Fig. 5D). This potential, direct negative regulation of WUS by UIF1 could explain the production of supernumerary organs by uif1 mutant flowers, in a ULT1-dependent manner (Figs 4, 5). Indeed, we previously showed that the WUS expression domain was laterally enlarged in ult1 mutants, in a Ler background (Carles et al., 2005). Interestingly, and as expected in view of the relative phenotypic intensities, uif1 mutant alleles show a weaker increase in WUS expression relative to the ult1-3 null mutant in a Col-0 background (Fig. 5D).

Here, we report in vivo physical interaction between ULT1 and the novel Myb transcription factor UIF1 (Figs 1, 2). UIF1 was isolated from a Y2H screen against ULT1. The interaction was further confirmed in plants, where BiFC took place in the nucleus. We show that the UIF1 N-terminus region is necessary and that the ULT1 C-terminus region containing the B-box is sufficient to mediate the interaction. This is strongly supported by reports in plants showing that the B-box domain is involved in protein-protein interactions (Datta et al., 2006, 2007; Gangappa et al., 2013). In particular, the formation of heterodimers both within and outside the B-box-containing protein family play important roles in regulating transcription and fine-tuning plant growth and development (Gangappa and Botto, 2014).

Both ULT1 and UIF1 genes are expressed from early stages of A. thaliana vegetative development and have overlapping expression patterns in inflorescence and floral meristems. However, the levels of expression of ULT1 and UIF1 differ in floral organs or developmental stages. Interestingly, the differences seem to be associated with the developmental status of organs, ULT1 being expressed at higher levels in dividing tissues and UIF1 being expressed at higher levels in differentiating tissues.

Similarly to ult1 mutants, uif1 mutants display supernumerary floral organs (Fig. 3). In ult1 mutants, increase of floral organ numbers correlates with inflorescence and floral meristem enlargement (Fletcher, 2001) and with lateral enlargement of the WUS domain of expression (Carles et al., 2005). Flowers of ult1 uif1 double and ult1 single mutants produce a similar number of additional sepals and petals (Fig. 4), indicating that ULT1 and UIF1 function in the same regulatory pathway for the production of perianth founder cells. The low penetrance of uif1 phenotypes, together with their moderate intensity, do not allow measurement of meristem size differences, nor enlargement of the WUS expression domain. However, we found a slight increase in WUS transcript level in both uif1 mutant alleles (Fig. 5). Together, these findings indicate that ULT1 and UIF1 might function together in OC size regulation.

ULT1 was previously proposed to act through WUS to regulate meristem homeostasis (Carles et al., 2004), functioning as a negative regulator of WUS expression in inflorescence and floral meristems (Carles et al., 2005). Although ULT1 protein contains a SAND domain shown to act as a DNA-binding module in animals (Bottomley et al., 2001; Surdo et al., 2003) and although ULT1 protein shows general affinity for DNA, no evidence for ULT1 binding at the WUS locus was reported thus far, nor that ULT1 protein could recognise specific DNA motifs (Carles and Fletcher, 2009; Engelhorn et al., 2014a). In the present study, we show that UIF1 is able to bind DNA and can recognise motifs in the promoter sequence of WUS. Scanning of the entire promoter using UIF1 PBM-generated matrices led to relevant binding sites in regions previously reported as general, meristem or stem cell quantitative elements (QEs) (Fig. S5; Bäurle and Laux, 2005). We showed by EMSA that UIF1 specifically binds to a site proximal to the meristem QE and that a single base mutation abolishes the binding (Fig. 5; Fig. S6). Because UIF1 protein possesses a repressor activity (Fig. 5), a reasonable hypothesis is that UIF1 binds to WUS and directly contributes to repress its transcription. This hypothesis is supported by the increased WUS expression in the inflorescences of both uif1 mutant alleles. Moreover, UIF1 might provide DNA-binding specificity and contribute to ULT1 recruitment at the WUS locus for further repression via chromatin modifications. The respective mutant phenotypes support this scenario, as ult1 defects are stronger than those of uif1.

Several recent studies have shown that binding of some transcriptional factors to their targets may occur prior to recruitment of the chromatin-modifying machinery, defining them as pioneers for transcriptional activation of target genes (Smale, 2010; Magnani et al., 2011; Zaret and Carroll, 2011; Iwafuchi-Doi and Zaret, 2014). In particular in plants, a pioneer function was proposed for the MADS domain-containing proteins functioning in transcriptional activation of flower morphogenesis genes (Pajoro et al., 2014). It is very likely that, via early binding to target loci, these pioneer factors could function to increase accessibility of the locus, by recruiting chromatin remodellers and modifiers (Smaczniak et al., 2012). UIF1 could be another of these TFs that recruits chromatin regulators to their targets for further regulation of chromatin accessibility. Of relevance to the important function of ULT1 in AG temporal activation at the centre of the FM, it cannot be excluded that UIF1 could be required to bring ULT1 to AG cis-regulatory elements in the proximal promoter and large intron region (Carles and Fletcher, 2009; 2010). Indeed, we found that UIF1 can specifically bind to AG cis-regulatory sequences located in both the promoter and the large intron, regions that we previously reported as bound by ULT1 in chromatin immunoprecipitation experiments (Carles and Fletcher, 2009).

How can we reconcile this hypothesis with a potential transcriptional repressor function, as shown by DLRA experiments? UIF1 might sit on the AG locus at early stages of FM development, preventing its expression, and upon an unknown signal around stage 3, bind to ULT1 protein, leading to further recruitment of chromatin factors, such as ATX1, for enrolment of the RNA PolII complex and transcriptional initiation. An alternative hypothesis is that UIF1 could carry a dual function on transcriptional regulation, depending on its interactors, acting either as repressor or as activator, as already reported for several transcription factors in A. thaliana, such as WUS, AP2 and FIL (Ikeda et al., 2009; Yant et al., 2010; Bonaccorso et al., 2012).

The uif1 mutants display organ fusions at their sepals and stamens. Furthermore, stamens of these mutants partially lose their identity and tend to become petaloid. Such phenotypes were observed in uif1 and uif1 ult1 plants but never in ult1 mutants. This supports additional roles of UIF1 in the control of floral organ identity and patterning that are independent of ULT1 function. Interestingly, the organ fusion phenotypes are reminiscent of those observed in cuc mutants (Aida et al., 1997; Takada et al., 2001). CUC genes encode NAC domain transcription factors that specify boundaries between meristematic and primordia cells, as well as lateral organ separation (Aida et al., 1997; Takada et al., 2001; Vroemen et al., 2003; Breuil-Boyer et al., 2004; Cheng et al., 2012). Thus, mutations in these genes result in sepal fusions and stamens branching with variable strengths ranging from partial to complete fusion depending on genetic backgrounds, such as single or multiple mutants in pairwise combination (Aida et al., 1997; Takada et al., 2001; Vroemen et al., 2003). The fact that CUC promoters contain UIF1-binding sites therefore suggests that UIF1 directly influences CUC gene expression (Fig. 5; Fig. S6). However, further investigations would be required to address this hypothesis.

The fact that AG also contains UIF1-binding sites might explain the loss of stamen identities and the development of petaloid stamens observed in uif1 mutants. Indeed, a decrease of AG expression in the developing third whorl would promote an expression of A-class genes in this domain and therefore the development of petals in place of stamens. The decrease of AG expression is probably very mild, thus explaining the low rate of partial homeotic transformations observed in uif1 backgrounds. However, we cannot exclude the possibility that this phenotype also results from boundary defects between whorls 2 and 3. Such hypotheses would suggest again that UIF1 may be either an activator or a repressor of transcription, promoting here activation of AG expression but repressing WUS expression in other domains.

The A. thaliana Columbia-0 (Col-0) and Landsberg erecta (Ler) ecotypes were used as the WT strains for the uif1 and ult1 alleles, respectively. Tobacco (Nicotiana benthamiana) plants and onion epidermis were used for transient experiments. The uif1-1 (SAIL_806_F06) and uif1-3 (SALK_024632) T-DNA insertional mutant lines were backcrossed two times to Col-0 before analysis. Although a third mutant allele initially named uif1-2 (SALK-073584) putatively contains an insertion in the UIF1 gene, we could not confirm the presence of a T-DNA at the locus in this line. Double mutant plants are in a mixed Ler (from ult1-3)/Col (uif1 alleles) background. As controls, we used single mutants in a mixed background obtained in the F2 generation from the same cross. A. thaliana plants were cultivated under long-day conditions (16 h light/8 h dark) at 18°C/16°C degrees. Tobacco plants were cultivated under long-day conditions (16 h light/8 h dark) at 22°C degrees. DNA material and cloning procedures are detailed in supplementary Materials and Methods.

Sequences of hydrophobic and Myb conserved domains were submitted to GlobPlot (http://globplot.embl.de/cgiDict.py) for prediction of structured domains and to BlastP (http://blast.ncbi.nlm.nih.gov/) for calculation of percentages of similarity. Sequences were aligned using the MultAlin program (Corpet, 1988) and designed under ESPript3.0 (Robert and Gouet, 2014).

For binary interaction tests, ULT1 and UIF1 full-length and truncated versions were expressed as Gal4 DNA-BD (BD: binding domain) or Gal4 AD (AD: activation domain) fusion proteins. ULT1 and UIF1-transformed yeast strains were mated and the resulting diploids were selected for the presence of both plasmids with medium lacking Leu and Trp (−LT). The selected strains were then dotted on −LT or –LTAH medium (lacking Leu, Trp, His and Ade) on which only the diploids presenting protein interactions can grow. For more details on the Y2H screen and protein extraction for western blot on mated Y2H diploids, see supplementary Materials and Methods.

For BiFC tests in tobacco leaf epidermis cells, constructs were transfected by agro-infiltration into leaves of 4-week-old N. benthamiana (Hamilton et al., 2002; Walter et al., 2004). When OD_600 nm=0.5 was reached, Agrobacteria cultures were pelleted and resuspended in 10 mM MgCl₂ and then treated with 100 µM acetosyringone for 2 h. Tobacco leaves were infiltrated with Agrobacteria solutions for co-transformation allowing expression of proteins fused to the N-terminal or C-terminal part of the split YFP (NY or YC). Cultures were grown separately and mixed at equal density before co-infiltration. For subcellular localisation, the pEZS-UIF1 constructs were transformed into onion epidermis cells using the Biolistic PDS-1000/He particle delivery system (Bio-Rad) (Sanford et al., 1993). The corresponding GFP-UIF fusion construct is suitable to address subcellular localisation of UIF1 protein as GFP-UIF can complement uif1 mutants, indicating that the GFP does not affect UIF1 function. Tissues were observed 3 days after infiltration (tobacco) or 24-36 h after bombardment (onion), by epifluorescence microscopy under an Axioscope A1 (Carl Zeiss), using a 46 YFP filter (EX BP 500/20, BS FT 515, EM BP 535/30), a 49 DAPI filter (EX G 365, BS FT 395, EM BP 445/50) or a 38 Endow GFP filter (EX BP 470/40, BS FT 495, EM BP 525/50). Images were acquired using an AxioCam MRc camera equipped with the ZEN lite Module Hardware.

Micro-projectiles were prepared with 50 ng of pRLC (Renilla luciferase construct), 400 ng of pBB168 (firefly luciferase construct; Blanvillain et al., 2011) and 15 µg of G4DBD-UIF1, G4DBD-UIF1-dCT or G4DBD-UIF1-ANANA-dCT. Onion epidermal cells were transfected as described above and kept in the dark for 20 h at 21°C, before being ground in liquid nitrogen. Proteins were extracted in 1 ml of PBLuc buffer [200 mM NaPO₄, pH 7, 4 mM EDTA, 2 mM dithiothreitol, 5% glycerol, 10 mg/l bovine serum albumin and EDTA-free protease inhibitor cocktail Complete (Roche)] and assayed (5-20 µl) using the Dual-Luciferase Reporter Assay System (Promega). Luciferase activities (mean±s.d.) were determined from seven independent biological replicates for each tested combination and normalised to the G4-DBD-GΔFP control set at 1.

Total RNA extraction and RT-PCR experiments were performed as previously described (Engelhorn et al., 2014b), with 35 cycles of PCR (except for EF1α, 25 cycles). RT-qPCR experiments were performed in a 10 µl final volume, using 5 µl of SYBR Green Mix buffer (Bio-Rad) and 0.4 µM primers. The Bio-Rad CFX 384 real time system was used with the following program: 98°C, 2 min; 39 cycles (98°C, 5 s; 60°C, 5 s). Relative quantification (RQ) values were calculated using the 2−ΔΔCT method (Livak and Schmittgen, 2001). ΔCt values were calculated using the EF1α or SAND genes as endogenous controls. Mean values±s.d. were calculated from three biological replicates (each quantified in three technical replicates). Sequences of UIF1 and WUS primers are listed in Table S1, and those for ULT1, EF1α (At5g60390) and SAND (At2g28390) were already described (Carles and Fletcher, 2009; Czechowski et al., 2005).

RNA in situ hybridisation was performed as previously described (Carles and Fletcher, 2009). Inflorescences were harvested 4 weeks after bolting. UIF1 sense and antisense probes were generated by T7 RNA polymerase activity from a 1-kb insert (PCR product amplified from UIF1 cDNA, using the ISH_UIF_F1/ISH_UIF_R1 primers; Table S1), cloned into the Zero Blunt PCR vector (Invitrogen).

Counting of floral organs was carried out on the ten first flowers of first ten bolting plants for each genotype. We used the χ² test (Plackett, 1983) to verify that each population is homogeneous and to assess whether percentage values of additional organs counted between uif1ult1 and ult1 mutant lines are significantly different.

UIF1 protein (fused to the 6His-MBP tag for PBM experiments, or to the 6His-GST for EMSA experiments) or the UIF1 Myb domain (fused to 6His tag for EMSA experiments) were expressed in Escherichia coli strain Rosetta2 (DE3) (Novagen Merck Millipore), from the pETH447, pETH380 or pETH455 vector, respectively. After induction by 0.4 mM isopropyl β-D-1-thiogalactopyranoside, cells were grown overnight at 17°C. Prior to EMSA experiments, recombinant proteins were purified via their 6His tag. Pellets of 0.5 l of culture were sonicated in 50 ml lysis buffer A [500 mM NaCl, 20 mM Tris-HCl pH 8, 1 mM Tris (2-carboxyethyl)phosphine hydrochloride], containing one protease inhibitor cocktail tablet Complete EDTA-free (Roche) and centrifuged for 30 min at 13,000 g. Clear supernatants were loaded on 1 ml Ni-Sepharose resin (GE Healthcare). Resin was washed with 20 volumes of buffer A containing 30 mM imidazole and eluted with buffer A containing 350 mM imidazole. The fractions containing recombinant proteins were pooled and dialysed overnight against buffer A.

A pellet corresponding to 25 ml of an induced E. coli culture was stored at −80°C and resuspended in 1 ml 1× binding buffer prior to the DNA-binding assay. Synthesis of double-stranded microarray, protein incubation and immunological detection of DNA-protein complexes were performed as described (Godoy et al., 2011). We used the nPBM11 design containing 167,773 different oligonucleotide probes (Franco-Zorrilla et al., 2014) synthesised in an Agilent's SurePrint G3 4×180k format (Agilent Technologies). DNA microarray was scanned in a DNA Microarray Scanner at 2 µm resolution and quantified with Feature Extraction 9.0 software (Agilent Technologies). Normalisation of probe intensities and calculation of E-scores of all the possible 8-mers were carried out with the PBM Analysis Suite (Berger and Bulyk, 2009). Perl scripts were modified to adapt them to nPBM11 microarray dimensions and to input files generated by Feature Extraction software. Position weight matrices corresponding to top three motifs recognised with high affinity (UIF1_1ary, E-score=0.4832; UIF1_2ary, E-score=0.4814; and UIF1_3ary, E-score=0.4804) were obtained from analysis of all DNA fragments bound by the UIF1 using PBMs. These matrices were used for prediction of UIF1 binding sites in candidate targets with the Morpheus program (http://biodev.cea.fr/morpheus; Moyroud et al., 2011), using a cut-off score of −3. A score was assigned to each binding site found in candidate target genes (WUS, AG, CUC1-3), and only those with a score ranging from −2 to 0 (best score value) were considered for further EMSA analyses. [The best-scored AG-0 sequence is found at the 5.23×10⁻⁶ frequency in genome-wide regulatory regions (from a scan through the TAIR10-3 kb upstream database).] For more details of the generation of these scores, see supplementary Materials and Methods.

For EMSA, dsDNA probes were prepared by incubating complementary single-stranded oligonucleotides (Eurofins Genomics) in annealing buffer (10 mM Tris-HCl pH 7.5, 150 mM NaCl and 1 mM EDTA). dsDNA (4 pmol) with a protruding G were fluorescently labelled with Cy5–dCTP (8 pmol) (GE Healthcare) using 1 unit of Klenow fragment polymerase (Ozyme) in 1×Klenow buffer for 1 h at 37°C. Binding reactions were performed in 20 μl binding buffer (50 mM NaCl, 10 mM Tris-HCl pH 7.5, 1% glycerol, 0.5 mM EDTA, 1 mM MgCl₂, 0.05 mM ZnCl₂ and 1% NP-40) supplemented with 28 ng/μl fish sperm DNA (Roche) and 10 nM double-stranded DNA probe. Binding reactions were loaded onto native 6% polyacrylamide gels 0.5× TBE (45 mM Tris, 45 mM boric acid and 1 mM EDTA, pH 8) and electrophoresed at 90 V for 75 min at 4°C. Gels were scanned on a Typhoon 9400 scanner [excitation light 649 nm, emission filter 670 nm band-pass filter (670 BP 30); Molecular Dynamics].

The authors thank Caroline Barette for assistance in DLRA fluorimetry dosages; Grégoire Denay and Laura Grégoire for help with binding site cut-off scores; and Aurélie Vialette-Guiraud for her help with sequence alignments and phylogenetic reconstructions. We thank Julia Engelhorn and Gilles Vachon for critical reading of the manuscript and acknowledge the Nottingham Arabidopsis Stock Centre (NASC) for providing insertion lines.