The control of axillary meristem fate in the maize ramosa pathway

Meristems are the reservoirs of stem cells that are responsible for the post-embryonic development of plants (Sablowski, 2007a). After germination, plants continuously form, in the axils of true or modified leaves, axillary meristems that can remain quiescent, or create either branches or flowers. The number, position and fate of these meristems are responsible for the variability observed in the architectures of different plant species (McSteen and Leyser, 2005; Sablowski, 2007b). Plant architecture, which is often modified during the domestication of crop species, is still a major target of selection in current breeding programs, as it influences the productivity and the mechanization requirements of modern farming. In maize, a major crop species, two types of axillary meristems are evident in the very early development of its reproductive structures, the tassel and the ear (Fig. 1A,C). In the tassel, the apical male inflorescence, the first axillary meristems formed are indeterminate, produce long branches and are therefore called branch meristems (Fig. 1E). The majority of axillary meristems, which are formed shortly afterwards, are instead determinate meristems that eventually terminate with the formation of a pair of spikelets (grass-specific structures containing flowers), and are therefore called spikelet-pair meristems (Fig. 1E). In the female inflorescence, the ear, only spikelet-pair meristems form in the early stages of development (Fig. 1G) (McSteen et al., 2000).

In the ramosa1 (ra1), ramosa2 (ra2) and ramosa3 (ra3) mutants of maize, spikelet-pair meristems assume the identity and fate of branch meristems (Vollbrecht et al., 2005; Bortiri et al., 2006; Satoh-Nagasawa et al., 2006), and give rise to highly branched inflorescences. Both RA1 and RA2 encode plant-specific putative transcription factors, a Cys2-His2 zinc-finger protein of the EPF class and a LOB-domain-containing transcription factor, respectively (Vollbrecht et al., 2005; Bortiri et al., 2006). RA3 instead encodes a metabolic enzyme, a trehalose-6-phosphate-phosphatase, and it is proposed to function through the modulation of a trehalose-related sugar signal or through a direct role as a transcriptional regulator (Satoh-Nagasawa et al., 2006). Maize, together with sorghum and sugarcane, belongs to the Andropogoneae tribe, a group of grasses that develop spikelet pairs. In rice and other more distantly related grasses, spikelets are single, and no RA1 homologue has been identified (Vollbrecht et al., 2005). This led to the hypothesis that the ramosa pathway and, in particular RA1, plays a central role in the evolution of grass inflorescence morphologies (Kellogg, 2007; McSteen, 2006; Vollbrecht et al., 2005). Despite genetic and expression analysis that place both RA2 and RA3 genes upstream of RA1 in regulating spikelet-pair meristem identity and fate (Bortiri et al., 2006; Satoh-Nagasawa et al., 2006; Vollbrecht et al., 2005), it is unclear how this recently discovered pathway is operating.

The ra1-RS allele was originally identified while screening a maize population in an undefined genetic background containing active transposable elements of the Spm family. In the course of this research, we determined (by searching for polymorphic markers) that this genetic background is closely related to the inbred line A619. Pollen was collected from ra1-RS homozygous plants and treated with 0.06% ethyl methanesulfonate (EMS) in paraffin oil according to Neuffer (Neuffer, 1994). ra1-RS ears were then fertilized with the treated pollen using a brush. The resulting M1 plants were self-fertilized in order to create M2 families to be screened for enhanced and suppressed ra1-RS phenotypes (30 plants per family). For the double and triple mutant combinations, we used the following alleles: rel2-ref; ra1-RS; ra1-63.3359; ra2-ref introgressed in A619 (Vollbrecht et al., 2005; Bortiri et al., 2006). Tassel and ear branch numbers were determined by counting all primary branches, considered as any branch carrying more than two spikelets. Secondary branches are considered any branch, as defined above, arising from primary branches. The statistical significance of the differences in branch number was analyzed using Student's t-test.

For map-based cloning of REL2, a back-cross population was generated by crossing rel2;ra1-RS plants and the inbred line B73. Close linkage to Simple Sequence Repeat (SSR) marker umc1962 was used to develop additional markers as described elsewhere (Gallavotti et al., 2008) (see Fig. S3 and Table S4 in the supplementary material). For double and triple mutant combinations, we genotyped all plants with gene-specific markers (see Table S5 in the supplementary material).

For in situ hybridization, we used an antisense probe designed on the 3′ untranslated region of the REL2 gene amplified using the following primers: REL2-3′UTR-F ATCTGATCAGCCAACAAGGTG and REL2-3′UTR-R ATCCATCAAATAGCCCCAAAC. RT-PCR and in situ hybridization expression analysis were performed as previously described (Gallavotti et al., 2008). REL2 and KNOTTED1 (KN1) RT-PCR primer sequences are provided in Table S6 in the supplementary material.

For scanning electron microscopy immature inflorescences were fixed in FAA (50% ethanol, 3.7% formaldehyde, 5% acetic acid), dehydrated in a graded ethanol series, critically point dried and coated with a gold/palladium mixture. Images were taken with a Quanta 600 scanning electron microscope at 20 kV. We also used a Hitachi S-3500N environmental scanning electron microscope on freshly dissected inflorescences at 5 kV.

For histological sections of the pulvinus (Galinat, 1959), tassel branch nodes were collected prior to anthesis, immediately after the tassel fully emerged from the flag leaf, in order to avoid complete lignification of the tissue. They were fixed in FAA, dehydrated in a graded ethanol series, treated with Histoclear and embedded in paraplast. Sections (8 μm) were placed on slides, dewaxed and stained with a Saffranin O/Alcian Blue solution (0.01%/0.05%), followed by rinsing in water. The sections were finally mounted in Permount (Fisher Scientific).

The following constructs were used for protein-protein interaction assays: REL2ΔWD40 corresponds to +1 to +1022 nucleotide of the REL2-coding sequence; for REL2ΔCTLH and REL2ΔCTLHΔWD40s, nucleotides +109 to +276 were deleted from the coding sequence; for RA1ΔEAR and RA1Δ2EAR, the coding sequence was truncated at positions +496 and +298, respectively; for RA1 mEAR, the central EAR motif was changed to LDFEFRF (original sequence LDLELSL) and the C-terminal EAR motif was changed to LDFQFRF (original sequence LDLQLRL); for RA1-RS and RA1-RSenh, the coding sequence starts at position +28; for RA1-63.3359, the coding sequence includes an extra 17 amino acids at the C terminus (VLSQTEERTATMGTCSA).

For targeted yeast 2-hybrid assays, prey genes were cloned in the pAD-GAL4 2.1 vector, whereas bait genes were cloned in the pBD-GAL4 CAM vector (Stratagene). Interactions were assayed by growth on histidine-lacking media and by β-galactosidase activity, according to manufacturer's instructions (Stratagene). β-Galactosidase activity was quantified as previously described (Reynolds and Lundblad, 1987). Transient assays in tobacco were performed as previously described (Szemenyei et al., 2008). Constructs were prepared as N-terminal fusions in the SPYNE (N-terminal YFP fragment) and SPYCE (C-terminal YFP fragment) vectors of the Bimolecular Fluorescence Complementation (BiFC) system (Walter et al., 2004). Pull-down assays were performed using the MagneGST Pull-Down System (Promega) following manufacturer's instructions. For in vitro transcription/translation reactions, samples were incubated with bound GST (Glutathione S-transferase) fusions for ∼2-4 hours, rotating at 4°C. For pull-down assays from leaf plant extracts, tobacco leaf were injected with Agrobacterium expressing the viral suppressor p19 (Voinnet et al., 2003) together with a 2X35S::RA1:MYC:NYFP construct, and after 4 days the leaves were collected and ground for crude protein extraction. Incubations were performed overnight, rotating at 4°C. To express GST:REL2ΔWD40s fusion protein, the REL2 gene was cloned as C-terminal fusions in pGEX-2TK (GE Healthcare). A 3×HA epitope tag (HA3) was added as a C-terminal fusion to RA1 in the pCITE vector (Novagen) and used for in vitro transcription/translation reactions.

For western blots we used mouse monoclonal anti-HA (Covance) and anti-GST (Cell Signaling Technology) antibodies at 1:2000 and 1:6000 dilutions, respectively, and a rabbit polyclonal anti-MYC (Cell Signaling Technology) antibody at 1:3000 dilution. For secondary antibodies we used 1:5000 to 1:10000 dilutions. For detection we used the ECL Plus Western Blotting Detection System (GE Healthcare).

The in planta repression assay and the TOPLESS promoter have been described by Szemenyei et al. (Szemenyei et al., 2008).

To uncover new genes in the ramosa pathway, we carried out an EMS-based mutagenesis on a weak allele of ra1, designated ra1-RS (Vollbrecht et al., 2005). ra1-RS tassels have an increase in branching compared with normal tassels as a result of the indeterminacy of spikelet-pair meristems (Fig. 1A,B,E,F) (Vollbrecht et al., 2005). Ear development is also affected, showing disorganized rows of seeds on the mature cob and the occasional formation of basal branches from the outgrowth of spikelet-pair meristems (Fig. 1C,D,G,H). Masculinization at the tip of the ear is also frequently observed (Fig. 1D). We screened ∼1800 M2 families for an increase in tassel and ear branching, and we identified several mutations enhancing inflorescence branching. Among these, one recessive mutation affecting both tassel and ear branching was named ramosa1 enhancer locus2 (rel2). rel2;ra1-RS double mutant plants showed highly branched ears and had tassels with distinctive upright branches (Fig. 2A). In developing ears, spikelet-pair meristems were indeterminate and converted into elongated branches (Fig. 2C). rel2 and rel2;ra1-RS tassels also showed an increase in the number of primary branches (see Fig. S1 in the supplementary material). The rel2 single mutant retained upright primary branches in the tassel but did not significantly affect ear branching (Fig. 2A,B). To determine whether the enhancement of ear branching was specific to the weak ra1-RS allele, which encodes a predicted RA1 protein that lacks the first nine amino acids (Vollbrecht et al., 2005), we created a double mutant with another characterized weak allele of ra1, designated ra1-63.3359 (Vollbrecht et al., 2005). The C terminus of the predicted RA1-63.3359 protein has an extra 17 amino acids (Vollbrecht et al., 2005). rel2;ra1-63.3359 mutant ears showed a strong enhancement of branching (Fig. 2D). Analysis of this cross revealed that in +/rel2;ra1-63.3359/ra1-63.3359 and +/rel2;ra1-RS/ra1-63.3359 plants, ear branching was also enhanced when compared with either single ra1-63.3359 mutants or +/rel2;ra1-RS/ra1-RS plants (Fig. 2E,F; see Fig. S1 in the supplementary material). Plants carrying the ra1-63.3359 allele were therefore more sensitive to the lack of a functional copy of REL2. In a specific sensitized genetic background, therefore, the rel2 mutation shows a semi-dominant effect.

The ra2 mutant was reported to enhance the ra1-RS mutant phenotype (Vollbrecht et al., 2005). We therefore crossed the rel2-ref mutant with an allele of ra2 with a weaker phenotype. The resulting rel2;ra2 double mutant plants showed an enhancement of ear branching relative to either single mutant (Fig. 2G; see Table S1 in the supplementary material). This enhancement was significant, although not as severe as in rel2;ra1-RS double mutants. We then generated rel2;ra1;ra2 triple mutant plants. In these plants, the ears were characterized by a massive proliferation of branches. This was due to the indeterminacy of all spikelet-pair meristems, including those borne on branch meristems, resulting in the formation of primary, secondary and, occasionally, tertiary branches (Fig. 2H, see Table S2 in the supplementary material). Similar defects were also observed in ra1;ra2 double mutants (Vollbrecht et al., 2005) (see Table S3 in the supplementary material) and in strong ra1 alleles, such as ra1-r (see Fig. S2 in the supplementary material), suggesting that the rel2;ra1-RS ear phenotype can still be enhanced by the ra2 mutation.

The ra2 mutant also has a similar upright branch phenotype to rel2 (Bortiri et al., 2006), independent from ra1 (Fig. 3G). We therefore analyzed cross-sections of the tassel branch nodes in both rel2 and ra2 mutants. At the node where the tassel branch is connected with the central axis (spike), a group of cells forms a specialized structure called the pulvinus (Galinat, 1959), which swells at anthesis allowing branches to separate from the central spike and help the spreading of the pollen by the wind. Staining of these sections with Saffranin O and Alcian Blue, which stains lignified and non-lignified cell walls red and blue, respectively, revealed a group of red staining cells within the nodes of the rel2 and rel2;ra1-RS mutants that was absent from either wild-type or ra1-RS plants (Fig. 3A-F). On their adaxial side, a variable number of small and unorganized vascular bundles is observed. In wild type or ra1-RS mutants, those additional vascular bundles were absent (Fig. 3E,F). However, blue staining of the pulvinal cells revealed that those cells seem unaffected. The presence of patches of lignified cells at the base of tassel branches in the rel2 mutants is likely to cause a lack of flexibility in this area, forcing the tassel branches to develop upright. Given the similar upright branch phenotype of ra2 mutants, we also sectioned and stained ra2 tassels. More severe defects, such as misplaced pulvinal cells, were observed in ra2 tassel branch nodes (Fig. 3G-I). The different structural and cellular organization in the two mutants suggests that rel2 and ra2 differentially affect the growth of tassel branches.

We isolated the REL2 gene by positional cloning and identified nonsense mutations in three independent alleles of rel2 (rel2-ref, rel2-SLO73, rel2-SLO335) (Fig. 4A; see Fig. S3 in the supplementary material). REL2 is composed of 25 exons, and encodes a protein of 1141 amino acids with 66% identity to the transcriptional co-repressor TOPLESS (TPL) of Arabidopsis (Long et al., 2006) (Fig. 4A). TPL belongs to a family of five redundant members in Arabidopsis, and has been shown to function in the auxin signaling pathway for the maintenance of apical fate during embryogenesis (Szemenyei et al., 2008). Transcriptional co-repressors lack DNA-binding ability but are recruited by specific transcription factors to regulate gene expression of their target genes (Liu and Karmarkar, 2008). The domain organization of the REL2 protein resembles the one described for TPL (Long et al., 2006; Szemenyei et al., 2008) (Fig. 4A). The N terminus is characterized by two domains, the LisH (lyssencephaly type1-like homology) and the CTLH (C-terminal to LisH) domains. Following a proline-rich region of unknown function, eleven WD40 repeats are found in the C-terminal REL2 sequence. Notably, the CTLH domain of TPL is necessary for the interaction with the EAR repressor motif of the AUX/IAA proteins (Szemenyei et al., 2008), a class of transcriptional repressors of early auxin responsive genes (Overvoorde et al., 2005). AUX/IAA proteins form heterodimers with the ARF transcription factors (Hardtke et al., 2004; Overvoorde et al., 2005), and, by recruiting TPL, they repress transcription of ARF target genes (Smith and Long, 2010; Szemenyei et al., 2008). RT-PCR (see Fig. S3 in the supplementary material) and in situ hybridizations of REL2 in developing maize inflorescences showed a ubiquitous signal, with stronger expression in meristematic tissues, such as inflorescence, branch and spikelet-pair meristems (Fig. 4B,C,E,F), in developing floral organs and in the vasculature (Fig. 4D).

To verify that REL2 functions as a transcriptional co-repressor, we tested whether or not REL2 was capable of rescuing the embryonic defects of the temperature-sensitive tpl-1 mutant (Long et al., 2006; Szemenyei et al., 2008). tpl-1 embryos show a range of phenotypes, such as single fused cotyledons (Fig. 4G) and, in the most severe cases, a root in place of a shoot – hence the name (Long et al., 2002). We transformed tpl-1 plants with a construct expressing a REL2-YFP fusion protein under the control of the endogenous TPL promoter (pTPL::REL2:YFP). Seven independent lines showed a complete rescue of the tpl-1 embryonic defects, and nuclear localization of the REL2-YFP fusion protein, consistent with a role in transcriptional regulation (Fig. 4G-J). We subsequently tested whether REL2 was capable of repressing the transcription of a β-glucoronidase (GUS) reporter gene in an in planta repression assay (Szemenyei et al., 2008) and we observed decreased GUS activity (Fig. 4K,L). Taken together, these results demonstrate that REL2 is a transcriptional co-repressor.

In our screen for the enhancement of inflorescence branching, we also identified an intragenic enhancer of the ra1-RS mutation (Fig. 5). We discovered a C475T transition in the ra1-RS coding sequence that caused the replacement of one leucine with a phenylalanine in the conserved C-terminal EAR-like repressor motif (LxLxLx) of the RA1 protein (Vollbrecht et al., 2005) (Fig. 5A). EAR repressor motifs have been described in several families of transcriptional regulators, such as the zinc-finger SUPERMAN (SUP) and the AUX/IAA proteins (Hiratsu et al., 2004; Tiwari et al., 2004). This mutant, renamed ra1-RSenh, showed a dramatic increase in ear branching owing to the conversion of spikelet-pair meristems into branch meristems (Fig. 5B,C), indicating that the leucine in the C-terminal EAR motif is crucial for function and suggesting that a transcriptional repressor mechanism regulates the establishment of spikelet-pair meristem determinacy. The RA1 protein also carries another potential EAR motif (LDLELSL, amino-acids 100-106), distal to the Zn-finger domain.

The discovery that REL2 encodes a transcriptional co-repressor with a CTLH domain, and that the RA1 protein carries two putative EAR motifs, at least one of which is important for RA1 function, prompted us to investigate whether the REL2 and RA1 proteins physically interact. We first tested this hypothesis using a targeted yeast 2-hybrid assay, and we detected interaction between RA1 and full-length REL2, as well as a truncated version of REL2 (REL2ΔWD40). The presence of the CTLH domain of REL2 was required, as evidenced by the loss of interaction when expressing REL2 proteins that lacked this domain (REL2ΔCTLH, RELΔWD40ΔCTLH) (Fig. 6A,B). Mutations specifically affecting the C terminus of the RA1 protein (RA1-RSenh, RA1-63.3359, RA1ΔEAR) resulted in a weaker interaction with REL2 when compared with RA1 or RA1-RS proteins (Fig. 6A,B; see Fig. S4 in the supplementary material). The interaction was completely abolished only if RA1 was missing the C terminus (RA1Δ2EAR) or when mutations were introduced in both EAR motifs simultaneously (RA1mEAR) (Fig. 6B). To test the interaction in planta, we used a bimolecular fluorescence complementation (BiFC) system for transient expression in tobacco leaves (Walter et al., 2004), and verified that REL2 and RA1 interact in the nucleus and confirmed that the interaction depends on the two EAR motifs (REL2-CYFP/RA1mEAR-NYFP) (Fig. 6C). We also tested the REL2/RA1 interaction by pull-down assays, using beads bound to a GST:REL2ΔWD40 bacterial expressed protein or to GST alone. Both in vitro expressed RA1:HA3 (Fig. 6D) and in planta expressed RA1:MYC:NYFP fusion proteins (Fig. 6E) were recovered using GST:REL2ΔWD40 bound beads, but not with GST alone. These results show that REL2 and RA1 can physically interact in both in vitro and in vivo assays. We also tested REL2 together with RA2 or RA3 proteins in yeast 2-hybrid and BiFC assays, but no interaction was detected (data not shown).

Plant development relies on the activity of meristems, highly organized and regulated groups of stem cells responsible for the formation of all post-embryonic organs. Whereas the shoot and the root apical meristems, formed during embryogenesis, establish the main axis of plant growth (apical to basal), axillary meristems are responsible for the formation of secondary axes of growth, such as branches and flowers. The outcome of axillary meristem activity is a crucial component of what determines the architectural variation observed in different plant species.

In this paper, we described one molecular mechanism by which an axillary meristem can acquire a determinate fate and result in the formation of a pair of spikelets, the defining units of grass inflorescences and the source of maize productivity. Based on our genetic and molecular data, we propose a model for maize spikelet-pair meristem determinacy in which RA1 recruits REL2 to the promoter of its target genes via an interaction involving two EAR repressor motifs and the CTLH domain of the REL2 protein (Fig. 7). The physical interaction between the C-terminal end of the RA1 protein and REL2 is supported by genetic evidence such as the haploinsufficiency of REL2 observed in the ra1-63.3359 mutant, and by the highly branched phenotype of the ra1-RSenh mutant. Although the targets of RA1 binding remain unknown, several mutations in the zinc-finger DNA-binding domain have been identified and show a strong ra1 phenotype (Cassani et al., 2006; Vollbrecht et al., 2005) (see Fig. S2 in the supplementary material). The reported absence of RA1 expression in the tassel branch meristems (Vollbrecht et al., 2005) is in agreement with this model, whereby no repression mechanism is in place and indeterminacy is maintained. It is likely that, as in Arabidopsis (Long et al., 2006), other REL2-like proteins act in a partially redundant manner in this pathway, which could explain why rel2 single mutants do not show enhanced ear branching. In rel2;ra1-RS mutants, the expression of REL2 is unchanged (see Fig. S3 in the supplementary material). This suggests that a truncated REL2-REF protein could still be present and interact with the RA1-RS protein, but the complex would have impaired repressing function. The enhancement of the rel2;ra1-RS ear mutant phenotype by the ra2 mutation indeed suggests that the REL2-REF/RA1-RS complex is still partially functional. Similarly, in the ra1-RSenh mutant, the interaction between REL2 and the mutant RA1-RSenh protein is likely not completely lost (Fig. 6), and no enhancement of secondary branches is indeed observed (data not shown).

Transcriptional repression is rapidly emerging as a common theme in the control of developmental processes in plants (Liu and Karmarkar, 2008; Krogan and Long, 2009). In this paper, we show that the regulation of axillary meristem fate can occur by a transcriptional repression mechanism involving the interaction of an EAR-containing zinc-finger transcription factor and the REL2 transcriptional co-repressor. Two fundamental and distinct pathways, the control of meristem fate (described here) and the establishment of embryo polarity (Smith and Long, 2010; Szemenyei et al., 2008), employ the same type of transcriptional co-repressor, the recruitment of which is guided by different and unrelated transcription factors via EAR-repressor motifs. A similar mechanism has also been recently described in Arabidopsis for the signaling pathway of the plant hormone jasmonate (Pauwels et al., 2010). What seems to differ are the number and nature of the proteins involved in the assembly of the repressor complex, as well as the mode of interaction. Whereas in the case of the auxin signaling pathway three proteins are involved (TPL, AUX/IAAs and ARFs) (Szemenyei et al., 2008), in the jasmonate signaling pathway four proteins have been reported as being part of the repressor complex (Pauwels et al., 2010). However, in the ramosa pathway, RA1 and REL2 may be the only proteins required, and, only in this case, two EAR motifs are implicated in the interaction.

The fine regulation of the balance between differentiation and stem cell proliferation determines whether a meristem acquires a determinate or an indeterminate fate (Sablowski, 2007a; Sablowski, 2007b). In Arabidopsis, the homeodomain protein WUSCHEL (WUS) is well known to maintain the pools of meristematic stem cells (Mayer et al., 1998; Sablowski, 2007b). During reproductive development, in the determinate floral meristems the function of WUS eventually has to be terminated to form all floral organs. The floral homeotic gene AGAMOUS (AG) is indeed necessary for the termination of WUS activity (Lenhard et al., 2001; Lohmann et al., 2001). Recently, it has been proposed that AG activates the transcription of the zinc-finger transcription factor KNUCKLES (KNU), which contains a C-terminal EAR motif that in turn negatively regulates WUS. The timing of KNU expression has been associated with the regulation of meristem fate, as delayed KNU expression results in the formation of indeterminate meristems, whereas ectopic expression results in premature termination of floral meristems (Sun et al., 2009). This effect has also been reported for RA1, where changes in the timing of expression have been correlated to changes in the inflorescence architecture of some maize alleles and of related grasses, such as sorghum (Vollbrecht et al., 2005). These similarities may suggest that KNU functions similarly to RA1, and therefore, expanding the analogy, that the REL2/RA1 complex may repress WUS-like activity in meristems, promoting differentiation at the expense of meristem maintenance. Although the maize orthologs of WUS are not expressed in the same domain as RA1 (Nardmann and Werr, 2006), RA1 was shown to function non cell-autonomously by sector analysis (Vollbrecht et al., 2005). It is important to note, though, that the architecture of maize and Arabidopsis inflorescences is very distinct. Unlike in Arabidopsis, where floral meristems form directly at the flanks of the inflorescence meristem, maize spikelet-pair meristems give rise to a series of axillary meristems (spikelet and floral meristems) that eventually form flowers (McSteen et al., 2000). Whereas in weak ra1 mutants, spikelet meristems are also slightly indeterminate, causing the formation of ears with unorganized rows (Vollbrecht et al., 2005; Cassani et al., 2006), determinate floral meristems eventually form. A series of genes, mostly transcription factors, that specifically regulate the identity and fate of these additional meristems have been characterized (Chuck et al., 1998; Chuck et al., 2002; Chuck et al., 2007; Chuck et al., 2008; Thompson et al., 2009), highlighting the existence of multiple regulatory pathways controlling the determinacy of most reproductive axillary meristems.

The activity and development of axillary meristems determine the diversity observed in the body plans and in the inflorescences of different plant species, therefore affecting productivity and the needs of modern agriculture. The identification of REL2 and of its physical interaction with RA1 sheds light on the molecular mechanisms that underlie different inflorescence architectures in grasses, and provides useful knowledge for modern breeding and for the improvement of crop species.

We are grateful to Mary Galli for helpful discussions on protein-protein interactions and on the manuscript; to Yunde Zhao for critical reading of the manuscript; to Mike Hannon for help with the tobacco transient assays; to Darren Hall, Victoria Shar, Kathryn Mara, Craig Gaines, Edsel Sandoval and Lana Blank for help with mapping and genotyping; to Steve Moose, Torbert Rocheford and Matt Ritter for growing M2 families; to members of the Yanofsky lab for helpful discussions; and to David Baulcombe at the Sainsbury Laboratory and Tracey Bettinson at Plant Bioscience Limited for the p19 vector. This work was supported by grants from the National Science Foundation (NSF DBI-0604923) to R.J.S., D.J. and E.V., and from the National Institute of Health (GM072764) to J.A.L. REL2 sequences are deposited at the NCBI database (accession numbers GQ927145 and GQ927146). Deposited in PMC for release after 12 months.