QUIRKY interacts with STRUBBELIG and PAL OF QUIRKY to regulate cell growth anisotropy during Arabidopsis gynoecium development

In plants, organ morphogenesis requires multiple cellular processes, among which cell division and cell elongation are crucial (Harashima and Schnittger, 2010; Jakoby and Schnittger, 2004; Nath et al., 2003; Ramirez-Parra et al., 2005). The high reproducibility of organ architecture strongly suggests that these processes are coordinated between cells throughout organ development: each cell or group of cells needs positional information to divide, elongate and differentiate properly.

Cell-cell communication is therefore crucial to orchestrate proper development. Various members of the Receptor-like kinase (RLK) family, including STRUBBELIG/SCRAMBLED (SUB/SCM), have been shown to be involved in this process (De Smet et al., 2009; Ingram, 2004; Savaldi-Goldstein and Chory, 2008; Schiefelbein et al., 2009). SUB and SCM were independently isolated upon genetic screens with the aim of identifying mutants that are defective in position-dependent patterning of the root epidermal cell types (SCM) (Kwak et al., 2005) and in ovule development (SUB) (Chevalier et al., 2005). In the Ler background, sub mutants also display aberrant floral phyllotaxis and twisted leaves, stems and floral organs (Chevalier et al., 2005; Fulton et al., 2009). At the cellular level, SUB controls cell shape, cell division planes and, to some extent, cell proliferation during development.

In addition to the molecular control of intercellular communication, mechanical forces exerted by plant cells and tissues are also crucial for the coordination of growth (Hamant and Traas, 2010; Johnson and Lenhard, 2011; Lloyd, 2011; Mathur, 2004; Seifert and Blaukopf, 2010). As plant cells are immotile and glued together through their walls, each cell is inescapably subjected to the influence of its neighbours. Abnormal behaviour in a cell or group of cells might therefore influence the position or behaviour of adjacent cells (Abe et al., 2003; Jarvis et al., 2003; Reinhardt et al., 2003). The control of cell growth in response to mechanical forces largely relies on cell wall and cytoskeleton dynamics. Indeed, cortical microtubules (CMTs), by guiding cellulose synthase distribution, influence the orientation of rigid cellulose microfibrils and thereby growth anisotropy (Paredez et al., 2006).

It therefore remains very challenging to elucidate the intercellular communication mechanisms underlying plant morphogenesis and their possible link with mechanical forces. To address this question, we performed a detailed molecular and functional characterisation of QUIRKY (QKY) during the development of the Arabidopsis female reproductive structure, the gynoecium (Alvarez-Buylla et al., 2010; Roeder and Yanofsky, 2006). The Arabidopsis gynoecium (or pistil) starts forming at stage 6 of flower development and is initially composed of two fused carpels. After fertilisation, which occurs at stage 14, it develops into a fruit, or silique, containing two valves fused to a central replum. Both the pistil and silique display a regular, bilaterally symmetrical shape, suggesting a tight orchestration of cell division and elongation. Both cell division and cell elongation contribute to the development of the carpel, but during development, cell division slows down while cell elongation increases. Until stage 14, epidermal cells mostly divide, whereas after this stage, epidermal valve cells mostly elongate. QKY was previously isolated as part of the SUB genetic pathway (Fulton et al., 2009). Similarly to sub mutants, qky displays abnormal floral phyllotaxis, twisted leaves, stem and floral organs (including siliques) and modified root hair patterning. This mutant also exhibits cell shape and periclinal division plane defects in the L1 and L2 of floral meristems. However, no molecular or cellular functions have been assigned to the protein.

QKY contains four predicted C2 domains in its N-terminal part and two transmembrane regions (TMRs) at both ends of a plant-specific phosphoribosyltransferase C-terminal region (Fulton et al., 2009) (Fig. 1). Proteins hosting C2 domains (Kikkawa et al., 1989) have mainly been characterised in animals and display variable functions and properties. They can be categorised by the number of C2 domains and the presence of additional domains in their sequences. The group that contains multiple (up to six) C2 domains and TMRs is involved in membrane trafficking and its members can be subdivided into five classes based on their structures: synaptotagmins and extended-synaptotagmins, ferlins, tricalbins, vp115 and Multiple C2 domain proteins with two transmembrane regions (MCTPs) (Cho and Stahelin, 2006; Shin et al., 2005). It is likely that QKY belongs to this last category and represents the only plant MCTP reported so far (Fulton et al., 2009). Although the functions of synaptotagmins and ferlins have been fairly well described in various organisms (Bansal and Campbell, 2004; Craxton, 2010; Martens and McMahon, 2008; Posey et al., 2011; Schapire et al., 2009), the roles of MCTPs have been much less studied (Fulton et al., 2009; Maeda et al., 2001; Shin et al., 2005). However, loss of expression of the single MCTP present in Caenorhabditis elegans results in early embryonic lethality, thus demonstrating that MCTPs might be essential to development (Maeda et al., 2001).

We show here that the morphology of gynoecium epidermal cells in qky is highly heterogeneous. Groups of cells arranged sporadically along the gynoecium display aberrant sizes and shapes that often correlate with CMT disorganisation. Such data, together with a strongly reduced twisting phenotype observed in the qky botero double mutant, suggests that the twisting phenotype results from defects of cell growth anisotropy. We also show that QKY localises at the plasma membrane and interacts in a complex with SUB and a PB1-domain containing protein named PAL OF QUIRKY (POQ). Such data reinforce the role of QKY in the cell-cell communication process and generate new insights on the role of SUB.

In an attempt to identify novel factors involved in carpel architecture and/or development, 4000 M2 families of γ-ray-mutagenised wild-type C24 plants were screened for flower morphological defects. Among the mutants observed, one displayed siliques with an irregularly twisted shape instead of the bilaterally symmetrical shape of wild type (compare Fig. 1A with 1D). In our conditions, 97% of mature siliques (n=40, four independent individuals) displayed this torsion phenotype. qky siliques were also shorter than wild type (Fig. 1G). The mutation responsible for the observed phenotype was mapped via map-based cloning analysis to QKY, which had been recently identified in a screen for strubbelig-like mutants (slm) (Fulton et al., 2009). Our initial mutant, and five T-DNA lines that we subsequently isolated, were therefore named accordingly (qky-10 to qky-15, Fig. 1H). All of the mutations identified, except qky-15, in which the T-DNA insertion exactly maps to the gene start codon, led to shorter proteins lacking the PTR_C region and the two predicted TMRs. All qky mutant alleles are recessive.

Sixteen sequences clustering with QKY were identified in the Arabidopsis genome (supplementary material Fig. S2A,B). T-DNA insertion mutants were isolated for nine paralogues, but none of these exhibited altered pistil architecture. Furthermore, double mutants between the closest paralogues of QKY (At3g03680 and At5g17980) and qky-11 did not show an increase of the gynoecium phenotype compared to the qky-11 single mutant. These data suggest that there is no functional redundancy between QKY and other MCTPs, at least regarding pistil architecture.

All the qky alleles isolated to date show a phenotype similar to that previously described, with shorter and irregularly left-handed or right-handed twisted siliques containing fewer seeds that are, in a noticeable proportion, wrinkled and smaller than those of the wild-type (Fig. 1D; supplementary material Fig. S3) (Fulton et al., 2009). However, with the exception of a slight flowering-time delay (supplementary material Fig. S4), and slightly aberrant floral phyllotaxis, which is probably due to a slight twist of the stem, qky alleles do not exhibit any phenotypic abnormalities: the floral formula is unchanged, floral organs and rosette leaves do not display any bending, twisting or developmental alterations (compare Fig. 1B,C with 1E,F).

To investigate the origin of the twisted pistil phenotype, and to better understand the function of QKY, we next performed a detailed analysis of valve epidermal cells. Valve epidermal cells were first observed using a scanning electron microscope (SEM). Cell shape and size were monitored in the upper part of wild-type and qky-10 mutant siliques, at stages 12, 14 and 16 of flower development (Alvarez-Buylla et al., 2010; Roeder and Yanofsky, 2006).

In wild-type plants, the three developmental stages investigated display epidermal cells organised in parallel lines. The large majority of these cells have a regular morphology and are almost perfectly rectangular, especially at stage 12 (Fig. 2A,G). From stage 14 on, cells appear longer, as they start elongating, and slightly more heterogeneous in shape and size (Fig. 2C,E,H,I). Conversely, cell shape and size in qky-10 are much more heterogeneous. Some cells appear contorted (Fig. 2B,D,F). Such aberrant cell morphology has been observed in 90% (n=40) of the qky-10 siliques that twist and at all three developmental stages analysed. Nevertheless, cell distortions do not affect all cells, but only groups of cells arranged sporadically on the pistil. Measurement of epidermal cell surfaces (in zones where the phenotype is visible) also shows a statistically relevant higher proportion of larger cells in the mutant samples, at stages 12 and 14, compared with the wild-type samples (Fig. 2G-I). However, differences between the mutant and wild-type samples lessen during development and become statistically insignificant (according to a two-sample Kolmogorov-Smirnov test) at stage 16.

This suggests that the qky phenotype results both from abnormal division patterns and impaired cell expansion. The fact that such defects occur sporadically along the gynoecium might explain the irregularity of the twisting phenotype observed.

How does the function of QKY relate to the epidermal cellular defect and thereby to the twisting of the pistil? Because of the role of microtubules in cell shape, and knowing the strong morphogenetic defects of most microtubule-related mutants, we investigated whether microtubule behaviour is impaired in qky. We crossed the qky-10 mutant to the microtubule reporter line p35S::GFP-TUA6 (Ueda et al., 1999). Microtubule organisation was monitored in the upper half of the gynoecium at stage 12 of flower development. Fig. 3 reports z-stack projections, all obtained with identical experimental parameters.

In GFP-TUA6 transgenics, the GFP signal is homogeneous among all carpel epidermal cells (Fig. 3A). Microtubules are oriented transversely, which is compatible with the direction of cell elongation (Lloyd, 2011). Conversely, in GFP-TUA6 qky-10 plants, the signal is heterogeneous and reveals defects of CMT network organisation (Fig. 3B,C). These defects are highly variable. For the strongest defects (illustrated in Fig. 3B), the GFP signal appears homogeneously distributed into the cytoplasm with very few and thicker bundles of CMTs, or CMTs form fewer and thicker bundles that are oriented in all directions. When defects are weaker (illustrated in Fig. 3C) CMTs appear normal, albeit often misoriented, and one can still observe a homogenously distributed GFP signal in the cytoplasm. All possible intermediates can be observed. Similarly to data reported in the previous section, defects are not observable in all cells but rather in multiple groups of cells. Interestingly, such defects are observed in zones where the gynoecium twists, strongly suggesting a causal link between microtubule defects and abnormal cell shapes (illustrated in Fig. 3B,D).

To check whether QKY belongs to a pathway with known regulators of organ twisting, we crossed qky-10 to spiral1-3 (spr1-3) (Nakajima et al., 2004), spiral2-2 (spr2-2) (Shoji et al., 2004), lefty1 and lefty2 (Thitamadee et al., 2002). Mutations in SPR1 and SPR2, which encode two plant-specific microtubule-associated proteins (Furutani et al., 2000) and in LEFTY1 and LEFTY2, which encode α-tubulin 4 and 6, respectively (Thitamadee et al., 2002), result in an alteration of microtubule polymer structure and therefore in the twisting of elongating organs.

All double mutants analysed display an additive phenotype with clockwise or anticlockwise twisted cotyledons, leaves or petals, as in the single spr or lefty, and randomly twisted carpels, as in the qky plants (supplementary material Fig. S5A-D). This suggests that pistil twisting in qky is induced by spiral- and lefty-independent pathways. One hypothesis could be that twisting is induced by the growth heterogeneity in the qky pistils.

This last hypothesis was addressed by introgressing the bot1-7 mutation in qky-11. BOTERO encodes a katanin protein involved in the normal orientation of CMTs and anisotropic cell expansion. Mutations in this gene do not trigger strong organ twisting, but rather result in small plants with enlarged organs (Bichet et al., 2001). Recently, bot1-7 has been shown to display more homogeneous growth in the shoot apical meristem when compared with the wild type (Uyttewaal et al., 2012).

In combination with bot1-7, the qky-11 mutant displays a botero phenotype, with shorter and enlarged organs. In addition, gynoecium twisting is strongly reduced, as only 3.7% of qky-11 bot1-7 mature siliques (n=164, four independent individuals) display an obvious torsion phenotype, compared with 80% (n=40, three independent individuals) of qky-11 single mutants. In addition, in the qky-11 bot1-7 double mutants that display gynoecium twisting, the phenotype is much weaker compared with the qky-11 single mutant (exemplified in supplementary material Fig. S6A-C). As fertility is also strongly reduced in double mutants, the monitoring of the phenotype was carried out on fertilised siliques containing at least a few green seeds. We also evaluated the percentage of twisting at an earlier stage of development (stage 15), at the beginning of the phase in which cell elongation predominates. At that stage, 23.1% (n=121, three independent individuals) of qky-11 mutants harbour siliques with an obvious twisting phenotype (illustrated in supplementary material Fig. S7). At the same stage, 5.2% (n=424, four independent individuals) of double mutant siliques display a similar phenotype, whereas no such phenotype is apparent in bot1-7. However, the possibility that the qky-11 bot1-7 double mutant or the bot mutation alone produce very slight torsion cannot be excluded. Interestingly, gynoecium cells of bot1-7 display an abnormal shape similar to the one observed in qky (supplementary material Fig. S8). However, in bot1-7, cells with abnormal shape are regularly arranged, all along the gynoecium. It is therefore tempting to hypothesise that the qky phenotype results from cellular anisotropic growth defects in certain zones of the gynoecium only. As the rest of the gynoecium of these mutants grows normally, the entire structure becomes twisted during development.

RT-PCR experiments indicate that QKY is ubiquitously transcribed (Fig. 4A). Consultation of AtGenExpress and the eFP Browser data set confirmed this broad expression pattern (Schmid et al., 2005; Winter et al., 2007), as did the beta-glucuronidase (GUS) expression profile obtained from pQKY::GUS fusion reporter lines (Fig. 4B-E). Interestingly, GUS staining first appears at the tip of rosette leaves, suggesting that QKY is preferentially expressed in differentiating rather than dividing cells (Donnelly et al., 1999). In situ hybridisation did not show any interpretable signal, despite several attempts and the use of several different probes.

In an attempt to determine QKY subcellular localisation, we generated transgenic plants stably expressing the QKY-eGFP fusion protein under the control of the cauliflower mosaic virus 35S promoter (p35S::QKY-eGFP, Fig. 4F,G; supplementary material Fig. S9A) and of its own promoter (pQKY::QKY-eGFP, supplementary material Fig. S9B). In leaf, gynoecium and root tissues, the QKY-eGFP signal seems to be associated with the plasma membrane. Interestingly, the signal appears discontinuous along the membrane, with highly labelled spots separated by unlabelled zones (Fig. 4F,G). We observed a similar subcellular localisation pattern in leaf protoplasts isolated from the p35S::QKY-eGFP transgenic line, in which the GFP signal is localised at the plasma membrane similarly to that observed in protoplasts isolated from the p35S::LTi6b-eGFP (plasma membrane marker) line. Hence, the signal in p35S::QKY-eGFP is distinct from that observed in the cytoplasm of protoplasts isolated from p35S::eGFP (compare Fig. 4H with 4I,J). Note that, in protoplasts isolated from the p35S::QKY-eGFP line, the GFP signal remains discontinuous along the plasma membrane. However, both p35S::QKY-eGFP and pQKY::QKY-eGFP plants produce twisted pistils, showing that a modified QKY (here by addition of an eGFP moiety) is directly responsible for the twisted phenotype. As a matter of fact, transgenic qky-10 plants expressing QKY-eGFP did not recover a wild-type appearance.

To further determine the relevance of the membrane localisation of QKY, the QKY-eGFP fusion protein (under the control of the 35S promoter) was transiently expressed in tobacco epidermal cells by infiltration. Fig. 4K shows a GFP signal localised at the plasma membrane, with a discontinuous and irregular pattern along the membrane similar to the one observed in stable transgenics. Co-infiltration with the p35S::VPS29-mCherry construct, which labels the cytoplasm (and endosomes) in infiltrated tobacco leaves (Zelazny et al., 2013), shows that QKY-eGFP does not localise to the cytoplasm (Fig. 4L). Finally, plasmolysis treatment of infiltrated leaf cells shows that QKY-eGFP remains localised at the membrane even after its dissociation from the cell wall, and, labelling of the plasma membrane of infiltrated leaf cells with the FM4-64 compound (Grebe et al., 2003) shows a colocalisation of the two signals (supplementary material Fig. S9C-E). Altogether, these data strongly suggest that QKY is localised at the plasma membrane.

We next evaluated the properties of the C2 and TMRs of QKY with respect to their subcellular localisation. For this purpose, the first two of the four C2 domains and the PTR_C/TM domain (F1, F2 and F3 subdomains, respectively; Fig. 1H), were fused to eGFP and evaluated for their subcellular localisation. When fused to C2 domains, eGFP localises to the plasma membrane (Fig. 4M,N). Conversely, when fused to the PTR_C/TM domain, eGFP remains cytosolic (Fig. 2O), even if we cannot formally exclude an additional membrane localisation for this domain.

To further elucidate the role of the plasma membrane protein QKY, we searched for putative QKY-interacting proteins in a yeast two-hybrid screen using the full-length QKY protein as bait. Upon selection, we isolated one relevant clone corresponding to At5g16220, which we named PAL OF QUIRKY (POQ).

To investigate the functional significance of this putative protein-protein interaction between QKY and POQ, we first investigated the temporal and spatial expression of POQ and its expressed protein. RT-PCRs show that POQ, comparably to QKY, is expressed in all tissues (Fig. 4A). However, consultation of the AtGenExpress data (Schmid et al., 2005) revealed that their expression patterns are exact opposites: when POQ expression is high that of QKY is low, and vice versa (supplementary material Fig. S10). This suggests an in vivo balance between the levels of transcription of the two genes. Subcellular localisation experiments in Nicotiana benthamiana cells suggest that POQ localises at, or close to, the plasma membrane and in small cytoplasmic compartments (Fig. 5A).

QKY-POQ interaction was next substantiated by bimolecular fluorescence complementation (BiFC) assay (Desprez et al., 2007). Full-length QKY and POQ proteins were fused to NE- and CE-YFP and co-infiltrated in tobacco epidermal cells. YFP expression localised at, or close to, the plasma membrane, was reproducibly obtained in multiple independent experiments and with three independent combinations of protein fusions (exemplified in Fig. 5B with the QKY-NE_YFP/CE_YFP-POQ combination). It is worth noting that the QKY-POQ interaction occurs only at the plasma membrane and not in intracellular compartments. This interaction was additional confirmed by in vitro pull-down assays. Full-length POQ and QKY were fused to GST and 6His, respectively. After immobilisation of GST-POQ on Glutathione resin, 6His-QKY was loaded on the columns. The eluted fraction contains proteins, the sizes of which correspond to those of expected GST-POQ and 6His-QKY recombinant proteins, suggesting that these proteins interact in vitro (Fig. 5C). Taken together, these results suggest that QKY can form heterodimers with the POQ protein.

POQ contains two exons, encoding a protein of 476 amino acid residues, that includes an octicosapeptide/Phox/Bem1p (PB1) domain (Fig. 5D,F) (Sumimoto et al., 2007; Terasawa et al., 2001). Composed of ∼90 amino acids, PB1 domains display the topology of a ubiquitin-like β-grasp fold with two α helices and a mixed, five-stranded β-sheet. PB1 domains are classified into groups harbouring an acidic motif termed OPCA (type I), an invariant lysine residue on the first β strand (type II), or both (type I/II) (Ponting et al., 2002; Sumimoto et al., 2007). Protein alignment suggests that the POQ PB1 domain is a type I domain (Fig. 5G).

Five allelic lines (poq-1 to 5, Fig. 5D) were grown in standard conditions. Analysis of these six single mutants revealed that poq resembles the wild type in that its organs do not twist (Fig. 5E).

Our biochemical and cell biology data suggest the existence of a QKY/POQ complex. However, phenotypic similarities and genetic interactions between qky and sub (Fulton et al., 2009) also suggest that SUB could be part of this complex. To further test this hypothesis genetically, we generated single, double and triple mutant combinations of these genes and studied the resulting phenotypes. Genetic crosses were performed using poq-3 (Fig. 5D), sub-6/scm-2 and qky-11 (Fig. 1H), such that mutant combinations were all in the same Col-0 background and all mutations resulted from T-DNA insertion.

Genetic interactions between QKY, POQ and SUB were evaluated by calculating the percentage of pistils showing an obvious twisting phenotype (illustrated in supplementary material Fig. S7) at stage 15 of flower development (Fig. 6A). poq-3 behaves as a wild-type plant, showing no twisting. Conversely, 24.2% of qky-11 and 25.6% sub-6 single mutants display pistil deviations. Analysis of double mutants shows a significant increase of the torsion phenotype compared with either single mutant. However, the level of enhancement observed in qky-11 poq-3 (44%) and sub-6 poq-3 (42.1%) remains significantly lower than that observed in qky-11 sub-6 (61.3%). Further confirming the genetic interaction, the qky-11 poq-3 sub-6 triple mutant presents a much stronger phenotype compared with each of the double mutants. This genetic interaction was also confirmed by measuring the number of seeds per silique and the number of aborted seeds per silique (supplementary material Fig. S11).

The next task was to assess whether cell-shape defects were also increased in double and triple mutants. Observations of epidermal valve cells using SEM show an obvious and strong alteration of cell morphology in all mutant genetic backgrounds (Fig. 6Ba-h). This observation suggests that the twisting phenotype, in all genetic backgrounds, results from defects of cell growth anisotropy.

The genetic interactions reported above prompted us to investigate the potential interaction between full length QKY, SUB and POC. This was achieved by targeted two-hybrid tests. A five-by-five matrix: (1) confirmed the interaction between QKY and POQ; (2) revealed the capacity of POQ to form homodimers; and (3) showed a direct interaction between QKY and SUB, but also between POQ and SUB (Fig. 6C). Therefore, these three proteins have the capacity, at least in vitro, to interact two by two and perhaps to form a macromolecular complex.

QKY encodes a Multiple C2 domain protein with two transmembrane regions (MCTP, Fig. 1) (Fulton et al., 2009) and was previously isolated as a potential member of the SUB genetic pathway (Fulton et al., 2009). Sharing also similar phenotypes with sub, QKY was proposed to be involved in cell-cell communications that control cell patterning, organ shape and development. Similarly to sub mutants, the twisting phenotype observed in qky depends on the ecotype. Although qky alleles isolated in the Ler background display strong twisting of several different organ types (Chevalier et al., 2005; Fulton et al., 2009), this phenotype seems to be restricted to the pistil in alleles isolated in Col-0, C24 and WS-4 (Fig. 1). Despite being a very likely candidate (Meng et al., 2012; Shpak et al., 2004; van Zanten et al., 2009), it appears that the ER locus is probably not responsible for the pleiotropic effect observed in the Ler qky or sub mutants (Vaddepalli et al., 2011). Deciphering the SUB/QKY pathway would help us to better understand the ecotype specificity of these mutations.

Several independent experiments suggest that QKY is likely to be localised at the plasma membrane (Fig. 4; supplementary material Fig. S9). This is fully in accordance with the structural properties of QKY and with the subcellular localisation of most of the C2 proteins similar to QKY (Cho and Stahelin, 2006; Schapire et al., 2008; Yamazaki et al., 2008). Interestingly, the subcellular localisations of QKY subdomains indicate that, whereas the C2-contaning N-ter subdomain localises at the plasma membrane, the TMR-containing C-ter subdomain does not (Fig. 4). Such data suggest that C2 domains from QKY belong to the category of C2 domains capable of binding to the plasma membrane (Cho and Stahelin, 2006; Shin et al., 2005), but also that TMRs by themselves are not capable of anchoring QKY to the plasma membrane. These observations highlight a difference of localisation properties between plant and vertebrate MCTP subdomains, as it was shown that TMRs of MCTP2 are themselves competent to anchor to the plasma membrane (Shin et al., 2005).

Relevant qky mutants reported to date show similar phenotypes and most of their predicted proteins lack TMRs (Fulton et al., 2009; this work). However, our data suggest that the importance of the C-ter domain cannot be restricted to plasma membrane targeting as previously suggested (Fulton et al., 2009). On the contrary, this domain may be additionally important for QKY topology or conformation and its removal impairs either the function of QKY or the ability of QKY to interact with its protein partners.

Our data show that QKY, SUB and POQ dimerise combinatorially in vitro (Fig. 6). The plasma membrane localisation of these three proteins (Figs 4, 5) (Yadav et al., 2008), their genetic interactions (Fig. 6) and the phenotypic similarities between qky and sub (Fig. 1) (Fulton et al., 2009) all suggest that these three proteins form a complex in vivo, which might be localised at plasma membrane. Structural and possible functional properties of each protein may provide information about the functioning of this complex.

POQ encodes a type I PB1-domain-containing protein (Fig. 5) (Ponting et al., 2002; Sumimoto et al., 2007). Arabidopsis contains 37 PB1 domain-containing proteins and a PB1 domain protein has also been recently reported from alfalfa (Long et al., 2012). However, to our knowledge, the biological functions of the PB1 domain remain unknown in plants. The structure of PB1 domains is highly conserved throughout evolution, suggesting a conserved biochemical function of this domain between plants, animals and fungi (Sumimoto et al., 2007). PB1 domains are scaffold modules and serve to organise functional multiprotein platforms. These complexes, depending on their composition, ensure the specificity and fidelity of a variety of biological events such as cell polarity, but also cell signalling (Moscat et al., 2006; Nakamura et al., 2010; Terasawa et al., 2001; Wilson et al., 2003). Most, but not all, of the proteins involved in such signalling platforms interact via their PB1 domains, and PB1-containing proteins also interact with proteins that do not contain PB1 domains [for example, ERK5 (Nakamura and Johnson, 2007; Seyfried et al., 2005)]. Accordingly, POQ may therefore promote the formation of a multiprotein complex involved in cell signalling through the SUB Receptor-like kinase.

In addition to plasma membrane staining, POQ-eGFP seems also to be localised in intracellular compartments (Fig. 5). Such a localisation suggests a role in membrane trafficking and receptor endocytosis such as recycling. Interestingly, similar expression and function have been shown for p62, a PB1-containing-protein involved in signalling cascades and whose depletion impairs the trafficking of the TrkA receptor into endocytic vesicles (Geetha et al., 2005; Sanchez et al., 1998). Such a hypothetical function is coherent with the fact that, among the large family of C2-domain proteins, MCTPs, to which QKY belongs, are mostly membrane trafficking proteins (Bansal and Campbell, 2004; Cho and Stahelin, 2006). Furthermore, such a function would be coherent with the fact that SUB could undergo internalisation and recycling (Yadav et al., 2008).

Phenotypic characterisation of the qky mutant revealed that its gynoecium contains clusters of cells with aberrant shape and size (Fig. 2). This phenotype seems to result from abnormal division patterns. In this respect, this phenotype is reminiscent of the misoriented cell divisions and aberrant cell morphology observed in ovule integuments and in the two outermost layers of flower meristem in sub and qky mutants (Chevalier et al., 2005; Fulton et al., 2009). However, the twisting phenotype seems also to result from cell elongation defects, as the percentage of siliques displaying a torsion phenotype increase during silique development.

Phenotypic characterisation of qky also shows that a large proportion of distorted cells display a disorganised CMT network (Fig. 3), suggesting that their aberrant shape and size result from cytoskeleton defects. This phenotype partially resembles that of botero (bot), in which all cells are distorted and in which CMT defects are also observed (supplementary material Fig. S8) (Bichet et al., 2001; Burk et al., 2001; Stoppin-Mellet et al., 2006). The function of BOT, which encodes a katanin, is to increase the ability of microtubules to self-organise in parallel arrays and thus to promote anisotropic growth in differentiating cells. As many other examples highlight the morphological consequences of microtubule network perturbations (Buschmann and Lloyd, 2008; Hamant and Traas, 2010; Mathur, 2004; Mathur and Chua, 2000; Uyttewaal et al., 2012), we can make the assumption that similar events occur in qky, though stochastically, only affecting certain patches of epidermal cells. Local changes of anisotropic growth during elongation stages of gynoecium development [from stage 12/14 on (Ferrandiz et al., 2010; Roeder and Yanofsky, 2006)] cause nonhomogenous cell elongation along the apicobasal axis, resulting in a twisting of the gynoecium. This hypothesis is substantiated by the fact that the botero mutation strongly reduces the qky-11 twisting phenotype. Interestingly, it has been very recently shown that SUB/SCM is required to maintain the balance between cell proliferation and cell differentiation during leaf development (Lin et al., 2012). The reason why only a few clustered cells are affected in qky is an interesting question and remains to be investigated. However, the fact that SCM/SUB is involved in position-dependent cell-type patterning (Kwak et al., 2005) is already a beginning of an answer to this question.

Taken together, our data shed light on one of the major issues of developmental biology, which is to identify the components that bridge the biological signals and the effectors that define the shape of cells and organs. Previous data published by the Schiefelbein and Schneitz laboratories showed that SUB/SCM encodes an atypical receptor kinase, able to sense external signals and influence cellular crosstalk, position-dependent cell-type patterning and proper cell elongation and differentiation. Here, the functional characterisation of QKY shows that SUB/SCM forms, together with QKY and POQ, a multiprotein signal transduction complex. This complex influences cell growth anisotropy, and probably does so through cell wall and/or CMT biogenesis and performance. Local malfunctioning of this complex results in local changes of growth anisotropy and thereby changes the global shape of the gynoecium. Such a mechanism has been shown and recently modelled in the leaf (Kuchen et al., 2012; Nath et al., 2003).

qky-10 was isolated after γ-ray irradiation (200 gray) of wild-type Arabidopsis C24 seeds (Rotman et al., 2003). The mutant was backcrossed three times to wild type. The list of mutations isolated in QKY, QKY paralogues, POQ and SUB is given in supplementary material Fig. S1. spiral and lefty mutants, p35S::GFP-TUA6 reporter line, were kindly provided by T. Hashimoto (Nara Institute, Nara, Japan). The botero mutant was obtained from O. Hamant (ENS de Lyon, Lyon, France). For phenotyping purposes, measurements were made using the NIH ImageJ program (http://rsbweb.nih.gov/ij) and statistical analyses performed using R (http://www.R-project.org).

QKY was identified by analysing the F2 progeny of the qky-10 mutant (C24) crossed to a Col-0 wild-type plant following a previously described procedure (Prunet et al., 2008). qky-10 was localised on the right arm of chromosome I, flanked by SSLP markers nga111 (map position 115.55 cM) and AthATPase (117.86 cM). For the fine mapping of QKY, new SSLP or CAPS markers were designed. The primer sequences, restriction enzymes and restriction sites generated across this interval are available upon request.

RT-PCR analyses were performed as described (Prunet et al., 2008). For PCR reactions, the annealing temperature was 55°C for all primer pairs and 28 cycles of amplification were performed for all genes. Computer-based sequence analysis was performed using VectorNTI Suite (Informax) and Sequencher (Gene Codes Corporation) software. Multiple protein alignments were obtained using ClustalW2 (Larkin et al., 2007) and edited with SeqVu (The Garvan Institute of Medical Research).

As bait, the full-length QKY cDNA was cloned into pAS2.1 (Clontech). The library screened was the A. thaliana inflorescence cDNA library in the pADGAL4-2.1 vector (CD4-30) (Fan et al., 1997). The two-hybrid screen was performed by yeast mating (Fromont-Racine et al., 1997), using strains PJ69-4a and Y187, and with triple selection (-Leu, -Trp, -His) on minimal media. Interacting partners were confirmed by re-transformation in PJ69-4a, mating with Y187 carrying either pAS2.1 or pAS2.1-QKY plasmids, and selection for expression of HIS3 reporter. For targeted two-hybrid tests, QKY, POQ and SUB full-length cDNAs were cloned in the pPC86- and pPC97-based vectors (Walhout and Vidal, 2001).

QKY and POQ were recombined into the Gateway vectors developed by F. Parcy (CEA, Grenoble, France). Infiltrations and observations were performed as previously described (Depège-Fargeix et al., 2011). Protoplasts were isolated from leaves according to (Yoo et al., 2007).

QKY and POQ cDNAs were recombined allowing the synthesis of GST- or 6His-tag fusion proteins and expressed in E. coli BL21. The GST-proteins (the ‘preys’) were immobilised on glutathione sepharose beads (Amersham). The 6His-proteins (the ‘baits’) were loaded on the corresponding agarose columns. Interaction complexes were eluted and pulled-down proteins were resolved by 6% SDS-PAGE and detected by western blotting using the appropriate mouse antibodies (anti-His, Sigma, and anti-GST, Santa Cruz) and a horseradish peroxidase-conjugated anti-mouse IgG secondary antibody (Promega). The horseradish peroxidase reaction was developed with an enhanced chemiluminescence system (Pierce).

Samples were observed using a Leica MZ12 stereomicroscope coupled to a DC320 digital camera and using Hitachi S-3000N and Hirox SH-3000 scanning electron microscopes. Fluorescent protein patterns were monitored using Zeiss LSM-510 and -700 confocal microscopes. All figures were composed with Adobe Photoshop 7.0 (Adobe Systems).

We are grateful to Olivier Hamant, Yvon Jaillais, Benoit Landrein and Charles Scutt for constructive discussion, Jérémy Just for help with statistics, and Nigel Briggs for English language consultancy. We thank Isabelle Anselme-Bertrand (CMES, University Jean Monnet, Saint-Etienne) for help during SEM imaging, Claire Lionnet (PlaTIM, SFR BioSciences Lyon-Gerland) for help during confocal microscopy and Cyril Py for excellent technical help. We also thank the anonymous reviewers for their constructive comments.