The preprophase band of microtubules performs the crucial function of marking the plane of cell division. Although the preprophase band depolymerises at the onset of mitosis, the division plane is ‘memorized’ by a cortical division zone to which the phragmoplast is attracted during cytokinesis. Proteins have been discovered that are part of the molecular memory but little is known about how they contribute to phragmoplast guidance. Previously, we found that the microtubule-associated protein AIR9 is found in the cortical division zone at preprophase and returns during cell plate insertion but is absent from the cortex during the intervening mitosis. To identify new components of the preprophase memory, we searched for proteins that interact with AIR9. We detected the kinesin-like calmodulin-binding protein, KCBP, which can be visualized at the predicted cortical site throughout division. A truncation study of KCBP indicates that its MyTH4-FERM domain is required for linking the motor domain to the cortex. These results suggest a mechanism by which minus-end-directed KCBP helps guide the centrifugally expanding phragmoplast to the cortical division site.
The preprophase band (PPB) of microtubules was first observed in electron microscopic studies on higher plant cell division performed in the 1960s (Pickett-Heaps and Northcote, 1966). This premitotic ring was seen to predict the edge of the plane where, after nuclear division, the outwardly growing cell plate would make contact with the mother cell wall (Jürgens, 2005; Müller, 2012). Although a few specialized types of higher plant cells were later found to divide without first forming a preprophase band, the presence of this structure in most cells was found to accurately predict all varieties of division plane: curved or straight, periclinal or anticlinal, and symmetric or asymmetric. In these cases, the preprophase band always disappeared before the cytokinetic apparatus could construct the dividing wall in the predicted plane. This gave rise to the idea that the disappearing preprophase band imprinted some kind of ‘memory’ at the cortical division site, acting as a molecular beacon to guide the edge of the phragmoplast to the predicted cortical zone (Gunning and Wick, 1985; Mineyuki, 1999; Van Damme et al., 2007).
Best seen in large vacuolated cells, the division plane is also memorized by the cytoplasm that forms a transvacuolar division disc (the phragmosome), which suspends the chromatin throughout mitosis and cytokinesis. Actin filaments pervade the phragmosome and continue into the preprophase band that marks the perimeter of the disc (Traas et al., 1987). However, by metaphase, actin filaments depolymerize in a narrow ring within the wider ring that was once marked by the preprophase band. This constitutes the actin-depleted zone, which represents a negative marker of the future phragmoplast attachment site (Cleary, 1995). Actin is not, however, completely absent from the vicinity of the wider cortical division zone (CDZ) because F-actin is enriched at either side of the actin-depleted zone to form what has been termed ‘actin twin peaks’ (Sano et al., 2005). Another negative marker of the CDZ is the KCA1 kinesin (Vanstraelen et al., 2006). Like actin, it is found at the plasma membrane of dividing cells but is excluded from the CDZ itself.
The first marker reported to remain at the division site from preprophase to cytokinesis was the plant-specific protein TANGLED (TAN) (Walker et al., 2007). The GTPase-activating protein of the small GTPase Ran (RanGAP1) was also found to remain at the CDZ throughout division (Xu et al., 2008). The recruitment of RanGAP1 to the preprophase band requires FASS/TONNEAU2 (Torres-Ruiz and Jürgens, 1994; Camilleri et al., 2002; Spinner et al., 2013), an upstream regulatory subunit of protein phosphatase 2A found to be essential for preprophase band assembly. Persistence of RanGAP1 at the cortical division site beyond preprophase, however, is dependent on two kinesins, POK1 and POK2 (Müller et al., 2006) – kinesins that are implicated in the association of TAN with the predicted division site. A recent publication shows that the POK1 kinesin too is a continuous marker of the cell division plane and remains attached to the cortex from prophase to cytokinesis (Lipka et al., 2014). FASS/TONNEAU2 itself, or at least its maize homologue DISCORDIA1, remains in the CDZ until metaphase, after which it disappears from the cortex (Wright et al., 2009).
Another class of markers has emerged that does positively mark the CDZ but only fleetingly given that they disappear along with the preprophase band microtubules by metaphase. Such evanescent markers are unlikely therefore to ‘memorize’ the division site until cytokinesis; these include several kinesin-like proteins, cell cycle regulators and mitogen-activated protein kinases (MAPKs) (Van Damme et al., 2007; Rasmussen et al., 2011a; Lipka and Müller, 2012). The CDZ is also a region specialized for membrane transport. A so-called ‘Golgi belt’ – an accumulation of Golgi stacks present in the cytoplasm underlying the division zone – was seen during the mitosis of dividing tobacco BY-2 cells (Nebenführ et al., 2000). Although the presence of the Golgi belt could indicate that division plane alignment requires exocytosis, transient application of the exocytosis inhibitor brefeldin A at prophase suggested that Golgi secretion is not required for marking the cortical division site (Dixit and Cyr, 2002). However, there is now good evidence that clathrin-mediated endocytosis is enhanced in the region of the CDZ at prophase (Karahara et al., 2009). Furthermore, the Arabidopsis TPLATE protein, which shows similarity to adaptin and coatomer proteins and which physically interacts with Arabidopsis clathrin light chain 2, localized to cell plates and, during cytokinesis, specifically to the CDZ. However, TPLATE is excluded from the narrower ring where the cell plate will insert. Knockdown of TPLATE interferes with cell plate insertion and it appears that membrane traffic involving TPLATE is required to complete cytokinesis (Van Damme et al., 2006; Van Damme et al., 2011).
All of this indicates that some proteins mark (positively or negatively) a wider division belt around the cell, whereas others define a more restricted – but more accurately predictive – ring. This distinction has been emphasized by Van Damme et al. (Van Damme et al., 2011) who proposed that the term CDZ be reserved for the wider cortical band while the narrow ring it brackets be known as the cortical division site. This appreciation of two functionally different molecular zones agrees with observations that the preprophase band of microtubules becomes narrower upon the cortex before it disappears (Mineyuki et al., 1989) and also with the fact that the preprophase memory sharpens as the cell progresses from metaphase to cytokinesis, as in the case of TAN (Walker et al., 2007; Rasmussen et al., 2011b). Furthermore, TAN contains two functional domains: one that targets TAN to the CDZ in preprophase and another that facilitates the return of TAN to the cortex at cytokinesis (Rasmussen et al., 2011b). This supports the idea that the CDZ functionally matures as it progresses from prophase to cytokinesis.
Our previous work has shown that a large microtubule-associated protein from Arabidopsis, AIR9 (auxin-induced in root cultures 9), seems to have a critical role in cytokinesis and cross-wall maturation (Buschmann et al., 2006; Buschmann et al., 2007). AIR9 colocalizes with the preprophase band but, unlike TAN and RanGAP1, is not present in the CDZ once the preprophase band has depolymerized. However, in contrast to the evanescent class of markers cited above, AIR9 does reappear at the cortex as a tight ring just as the expanding phragmoplast makes contact with the cortical division site. Once the phragmoplast contacts the former preprophase band site, AIR9 invades the nascent cell plate forming a torus. Drug-induced formation of supernumerary branches of the phragmoplast has shown that an AIR9 torus forms only where the cell plate becomes attached at the former preprophase band site, but not at any other cortical site unmarked by a preprophase band. This suggests that AIR9 specifically interacts with proteins constituting the preprophase memory. In addition, a truncation study of AIR9 confirmed that microtubule binding and preprophase band memory interaction are represented by separate regions within the AIR9 protein: the positively charged N-terminus and the large region containing the repeated A9 domains, respectively (Buschmann et al., 2006). Homologues of AIR9 are present in certain bacteria and protozoa (Buschmann et al., 2007). Interfering with the function of the AIR9 homologue from Trypansoma brucei results in defects in nuclear positioning and cleavage furrow placement (May et al., 2012).
To find out more about the functional composition of the molecular memory of the preprophase band, we exploited the ability of AIR9 to ‘remember’ the CDZ and searched for interaction partners that bind its region of repeated A9 domains. One such protein was KCBP, the kinesin-like calmodulin-binding protein, which features a C-terminal motor domain that moves along microtubules in the minus direction (Reddy et al., 1996; Oppenheimer et al., 1997; Song et al., 1997). We find that KCBP, when expressed from the endogenous promoter, labels the cortical division site throughout mitosis. KCBP also contains an N-terminal myosin tail homology 4 and band4.1, ezrin, radixin, moesin (MyTH4-FERM) domain (Richardson et al., 2006). Using truncation studies, we show that loss of this domain abolishes the ability of KCBP to locate to the cortical division site. We consider mechanisms by which microtubule minus-end-directed motors like KCBP could guide the outgrowing phragmoplast to the cortical division site.
The dynamic behaviour of AIR9 during cell division
We previously determined the 187 kDa AIR9 protein from Arabidopsis thaliana (L.) Heynh., which is expressed in cycling cell suspension cultures, is a microtubule-associated protein (Buschmann et al., 2006). To establish that the AIR9 gene is involved in cell division in planta we generated a fusion of the 1852 bp AIR9 promoter with the Escherichia coli β-glucuronidase (GUS) coding sequence. In transgenic Arabidopsis, the AIR9 promoter directed the expression of GUS to root meristems. GUS expression was also seen in the root stele (Fig. 1A). In aerial parts of the plants, the AIR9 promoter was active in young leaves, in trichomes of young and of mature leaves, and in the vasculature of leaves and stems (supplementary material Fig. S1). To verify that AIR9 is expressed in dividing cells we performed whole-mount in situ mRNA hybridization of Arabidopsis roots using an AIR9 antisense probe. The probe stained the meristematic region of the root tip, supporting the conclusion that AIR9 is expressed strongest in dividing cells (Fig. 1B,C).
GFP–AIR9 expression in cycling tobacco BY-2 cells, driven by the CaMV35S promoter, labels the preprophase band and, during the late phragmoplast stage, the cortical division site. GFP–AIR9 subsequently moves into the new cross wall forming a torus (Buschmann et al., 2006). To analyse cell division in Arabidopsis root tips, the same construct was transformed into wild-type Col-0 plants. Spinning disc confocal microscopy confirmed that the localization of the GFP–AIR9 fusion in root meristems followed the same pattern previously reported in tobacco BY-2 cells (supplementary material Fig. S2). However, because of the stronger signal in BY-2 suspension cells, this tobacco cell line was used as the model for the high-resolution analysis of AIR9 behaviour during cell division (Fig. 1D–F; supplementary material Movie 1). Double labelling of tobacco BY-2 cell division with GFP–tubulin showed that mRFP1–AIR9 does not bind to spindle microtubules (Fig. 1G–I). After cytokinesis, GFP–AIR9 tends to label the new cross wall (Buschmann et al., 2006), which was also clearly seen in root tips of transgenic 35S::GFP–AIR9 plants (Fig. 1J). To investigate whether the strong cross-wall label is independent of microtubules, we investigated the effect of microtubule-depolymerizing oryzalin using tobacco BY-2 cells. In this case, GFP–AIR9 was still found on the cross wall, indicating that AIR9 binds some other component of the cortex in the absence of microtubules (Fig. 1K,L).
We previously reported that GFP–AIR9 disappears from the CDZ after preprophase but returns to the cortex just as the cell plate inserts (Buschmann et al., 2006). To clarify whether GFP–AIR9 arrives at this location through the approaching microtubules of the phragmoplast, or whether its localization is independent of the phragmoplast, we applied brefeldin A, because this drug has been shown in BY-2 cells to cause the phragmoplast to stop growing and collapse after prolonged incubation (Reichardt et al., 2007). Interestingly, when brefeldin A was applied during the spindle stage, GFP–AIR9 could be seen to label the cortical division site prematurely during the early phragmoplast stage (arrows point at the dotted cortical AIR9 signal) rather than at the stage of cell plate attachment (Fig. 1M). This demonstrates that cortical GFP–AIR9 is capable of binding the cortical division site independently of the approaching phragmoplast (framed by asterisks). This also emphasizes that AIR9 is capable of recognizing and binding a component of the cortical division site but that this ability is normally suppressed during the spindle and early phragmoplast stages of division (Fig. 1M).
AIR9 physically interacts with KCBP
Based on our previous study and the results presented in Fig. 1, we hypothesized that AIR9 physically interacts with a component of the preprophase memory. To identify this component we initiated a yeast two-hybrid (Y2H) screen. Because the truncation study of AIR9 (Buschmann et al., 2006) had raised the possibility that the 11 repeated A9 domains of AIR9 might be sufficient for interacting with the preprophase memory, this large region (positions 455–1595, i.e. 1141 amino acids) was subcloned and used in a Y2H screen against an Arabidopsis seedling library. Several putative protein interactors were identified but the calmodulin-regulated kinesin KCBP represented the best ‘hit’. We obtained 12 N-terminal fragments of KCBP that, based on sequencing, were likely to be independent clones. Sequence alignment with full-length KCBP allowed us to determine the minimal KCBP sequence required for interacting with AIR9. The minimal KCBP sequence encompassed a part of the FERM domain (i.e. the FERM-M and FERM-C subdomains) including a part of the adjacent coiled coil region (Fig. 2A).
To verify the Y2H result, the interaction of AIR9 with KCBP was studied by immunoprecipitation. This was done using a rabbit antibody against AIR9 (Basu et al., 2005) and a rat antibody against cotton KCBP (Preuss et al., 2003). Initial experiments showed that the antibody against cotton KCBP cross-reacted with Arabidopsis KCBP from cycling Arabidopsis cell suspensions [a green cell line of Landsberg erecta (Ler) ecotype]. The same suspension cell line was used in subsequent tests because AIR9 was highly expressed in these cycling cells. For pulldown experiments, cell extracts from the Ler line were exposed to the AIR9 antibody (see Materials and Methods). After pulldown with protein A beads and analysis by SDS-PAGE, a single band of 140 kDa could be detected in western blots (Fig. 2B), indicating that AIR9 interacts with KCBP in cycling cells in suspension.
Next, we asked whether an interaction of AIR9 with KCBP could be observed using purified proteins. For this, full-length GST–KCBP was produced in E. coli. AIR9 was produced as an 1194-amino-acid fragment (amino acids 421–1614) tagged with N-terminal His6, also in E. coli (total 132 kDa). This fragment of AIR9 contains the entire region of the A9 repeated domains (similar to that used in the Y2H experiment). Nickel-column purification of His–AIR9, however, produced two bands – a larger band corresponding to the 132 kDa fragment and a smaller ∼50 kDa band that was also detected by the anti-His antibody and was presumed to be an N-terminal truncation product. The purified, desalted and concentrated KCBP and AIR9 proteins were mixed and incubated with agitation at room temperature. Next, equilibrated glutathione–agarose beads were added and the mixture was further incubated. The beads were then centrifuged, washed, boiled in SDS-buffer and subjected to western blot analysis. Binding in the presence of GST–KCBP was greatly enhanced above the background binding of His–AIR9 to the beads (Fig. 2C). This substantiates that there is a physical interaction between KCBP and AIR9, and suggests the interaction is direct. Interestingly, the smaller ∼50 kDa N-terminal truncation product of the His–AIR9 fragment also interacted with GST–KCBP, suggesting that a KCBP interaction domain resides in this region.
Next, we investigated whether AIR9 and KCBP colocalize during cell division. For initial experiments, a cDNA of KCBP was cloned and tagged with C-terminal or N-terminal GFP. These fusions were expressed in cycling tobacco BY-2 cells using the CaMV35S promoter. We found that both fusions of KCBP localized to the preprophase band in a punctate manner. The tagged KCBP constructs were then co-transformed with mRFP1–AIR9. mRFP1–AIR9 and GFP-tagged KCBP colocalized in the preprophase band of tobacco BY-2 cells (Fig. 2D). Colocalization was also observed on the cross-walls (arrowheads in Fig. 2D) and on the distal half of phragmoplast microtubules (Fig. 2E).
AIR9 recruits KCBP to interphase microtubules in Nicotiana benthamiana
To gain further insight into the interaction of AIR9 with KCBP, fusions of these proteins to fluorescent proteins were expressed in Nicotiana benthamiana (Domin) leaf epidermal cells using Agrobacterium-mediated transient transformation. When GFP–KCBP was expressed alone no, or only a few, microtubules were seen in these non-dividing cells. Instead, most of the label appeared to be cytoplasmic or weakly associated with a reticulate cytoplasmic system (Fig. 3A). When mRFP1–AIR9 was transformed alone, cortical microtubules became visible. mRFP1–AIR9 expression in the leaf cells caused microtubule bundling (Fig. 3B) that was also seen in other experiments involving transient expression of AIR9 (e.g. in Arabidopsis suspension cells; supplementary material Fig. S3A). Surprisingly, co-expression of GFP–KCBP with mRFP1–AIR9 in leaf epidermal cells led to a dramatic shift in the localization of KCBP. In this case, GFP–KCBP could be seen to localize to microtubule bundles that were also labelled by mRFP1–AIR9. Merging of the channels showed extensive colocalization (Fig. 3C–E). This indicates that in epidermal leaf cells of Nicotiana, AIR9 is capable of recruiting KCBP to cortical microtubules. Similar results were obtained using the C-terminal GFP fusion of KCBP (supplementary material Fig. S3). As a control, we investigated the co-expression of GFP–KCBP with another microtubule-associated protein (MAP): the mRFP1-tagged microtubule-binding domain of mammalian MAP4. In to mRFP1–AIR9, mRFP1–MAP4 was not capable of recruiting KCBP to cortical microtubules (Fig. 3; supplementary material Fig. S3).
The phenotype of air9 and zwi mutants
To deepen our understanding of AIR9 and ZWICHEL gene function in cell division, we analysed respective mutants and double mutants. Screens for transposon mutants impaired in gametophytic development have repeatedly identified insertions into the AIR9 single copy gene. We analysed in detail the transposon-induced ungud9 (air9-9) mutant of AIR9 (Lalanne et al., 2004) (Fig. 4A; supplementary material Table S1) and observed a dramatic reduction of pollen germination for ungud9 in vitro (data not shown). The defects in female gametophytic development included aberrations in cell division and ovule differentiation (Fig. 4B; for details see supplementary material Fig. S4A,B). However, when the hitherto unknown 5′ transposon border of ungud9 was isolated it was found that the ungud9 insertion is associated with a deletion of (at least) 11 genes starting from AIR9. By testing polymorphic markers we confirmed the presence of a large deletion using F1 plants of a cross between ungud9 (Ler ecotype) and Col-0 (Fig. 4A). Another transposon-induced mutant, haumea, shows a similar but even larger deletion (approximately 83 kb) that also includes part of the AIR9 gene. haumea too has male and female transmission defects (Page et al., 2004). A third transposon-induced deletion mutant termed GT3730 shows a 5′ deletion of AIR9 sequences but does not affect further genes. The reason for the high frequency of AIR9-associated deletions in transposon lines has to do with the fact that the T-DNA of the transposon starter line DSG-1 resides in AIR9 (Sundaresan et al., 1995; Brodersen et al., 2002). Mechanisms for how this might result in large genomic deletions have been suggested to involve deletion of the negative selection marker Indole Acetic Acid Hydrolase from the DSG-1 T-DNA after inducing transposition (Page et al., 2004). Whereas ungud9 and haumea have obvious defects in gametophytic development, no morphological phenotypes could be observed in the homozygous DSG-1 starter line (e.g. Fig. 4B) nor in the homozygous GT3730 line (data not shown).
In search for true air9 phenotypes, additional T-DNA insertion lines were obtained from stock centres (supplementary material Table S1). Homozygous T-DNA insertions into AIR9 showed severe disruption of AIR9 transcription (Fig. 4C); however, no morphological phenotypes have so far been detected. The mutant air9-5, for example, shows apparently normal ovule development (Fig. 4B). When we quantified cell division plane alignment in roots, we found that the accuracy of cross wall placement in air9-5 (also in air9-27) was highly similar, if not identical, to that in wild-type. Because AIR9 interacts with KCBP we were curious to analyse cell division plane alignment in the zwichel (zwi) mutants of KCBP. Such zwi mutants are best known for a defect in leaf trichome branching and expansion (Hülskamp et al., 1994; Oppenheimer et al., 1997). We quantified cross wall placement in the rhizodermis of zwiA (zwiA is a T-DNA insertion mutant of the SALK collection with a strong trichome phenotype; see Materials and Methods). No significant difference of cross wall placement in zwiA with respect to the wild-type could be detected (supplementary material Fig. S4C). We therefore created the double mutant of air9-5 with zwiA. Again, in double homozygous air9-5 zwiA plants, division wall alignment was indistinguishable from wild-type (Fig. 4D; supplementary material Fig. S4C). Taken together, these results suggest that the function of AIR9 and ZWICHEL in plant cell division might be redundant with additional genes. Uncovering these genes will help to establish a complete phenotypic description of air9 and zwi mutations.
GFP–KCBP is a continuous resident of the CDZ in Arabidopsis
We next investigated whether GFP-tagged KCBP behaves similarly to GFP-tagged AIR9 during plant cell division. Previous localization studies of KCBP during cell division have necessarily used fixed material because they were based on immunofluorescence (Bowser and Reddy, 1997; Smirnova et al., 1998; Preuss et al., 2003). Therefore, to study KCBP localization in living cells the Arabidopsis KCBP promoter (Reddy and Reddy, 2004) was isolated and ligated into pGWB6, replacing the CaMV35S promoter. The resulting destination vector was recombined with an entry clone containing the entire Arabidopsis KCBP gene from its start codon, including all introns and its 3′ untranslated region. The genomic GFP-tagged KCBP construct was then used for complementation experiments using zwiA mutants. Transgenic zwiA plants carrying the genomic GFP–KCBP construct were identified by hygromycin and kanamycin selection on agar plates. Many, but not all, transgenic lines showed complementation of the zwi trichome phenotype (Fig. 5A,B). Complemented zwiA plants were investigated with respect to GFP localization in the root. GFP signals were detectable but expression levels appeared to be low. Low magnification microscopy showed that GFP–KCBP expression was strongest in the root tip (Fig. 5C) although GFP–KCBP expression was still detectable in the elongation zone and in root hairs. Closer examination of the root tip showed that GFP–KCBP localizes to cross walls of interphase cells (Fig. 5D), similar to what was seen for GFP–AIR9 in meristems (Fig. 1). In dividing cells, GFP–KCBP localized to preprophase bands during premitosis then remained in the CDZ throughout mitosis and cytokinesis (Fig. 5E–G). GFP–KCBP was also seen towards the poles of the anaphase spindle (Fig. 5F) and on the phragmoplast (Fig. 5G) but was absent from the metaphase spindle.
Detailed analysis of GFP–KCBP localization in tobacco BY-2 cells
Because of the superior characteristics of tobacco BY-2 for the microscopy of cell division, we analysed the localization of GFP–KCBP in this cell line. CaMV35S-driven GFP–KCBP constructs in tobacco BY-2 cells underwent silencing quickly and it was hard to obtain data for the entire process of cell division (however, some movies were obtained by picking early colonies from transformation plates; see e.g. supplementary material Movie 2). By contrast, transformation with the genomic GFP–KCBP construct led to improved fluorescence and cell-cycle-dependent microtubule labelling in tobacco BY-2 cells (Fig. 6). GFP–KCBP expressed from its own promoter could be seen in the preprophase band, where it localized to foci arranged in a belt-like manner (Fig. 6A). At pro-metaphase, when the nuclear envelope disintegrated, GFP–KCBP remained at the cortex but did not label the spindle (Fig. 6B). During metaphase, the GFP–KCBP signal remained in the cortex, becoming slightly narrower. Metaphase spindles were not stained (Fig. 6C). At the anaphase–telophase transition (Fig. 6D), the cortical GFP–KCBP signal narrowed further and GFP–KCBP could then be seen at distal poles of the developing phragmoplast, where most microtubule minus-ends are found (Smertenko et al., 2011). From the phragmoplast stage onwards, the cortical GFP–KCBP signal was arranged like beads on a string (Fig. 6E,F). Throughout phragmoplast development, GFP–KCBP tended to accumulate towards the distal end of the phragmoplast rather than towards the proximal cell plate (see z-projection in Fig. 6F, and single section in Fig. 6G). GFP–KCBP was not found on the cell plate during the phragmoplast stage (Fig. 6G) but after cytokinesis some of the GFP–KCBP label moved into the new cross wall, forming a torus as previously described for AIR9 (arrowhead in Fig. 6H) (Buschmann et al., 2006).
Taken together, the results presented in Fig. 5, Fig. 6 and supplementary material Movie 2 indicate that KCBP is a continuous marker of the cortical division site in higher plants. The 1:1 colocalization of the KCBP signal with the lipid dye FM4-64 during metaphase (Fig. 6I) demonstrates the close association between the kinesin and the plasma membrane at a stage during which no cortical microtubules are present.
Colocalization of KCBP with microtubules during preprophase and cytokinesis
Next, we investigated the relationship between KCBP localization and microtubule positioning by double labelling. For this, we used transformed tobacco BY-2 cells expressing our genomic GFP–KCBP construct and mRFP1–MAP4 to label microtubules (Van Damme et al., 2004). This showed that KCBP and microtubule localizations were similar but not identical during the preprophase band stage. Sites of GFP–KCBP localization were detected where microtubules appeared to be absent and some microtubules were not decorated by GFP–KCBP (Fig. 7A–D).
The same double-expressing tobacco BY-2 cell lines were then used to clarify the relationship between phragmoplast microtubules and GFP–KCBP of the CDZ. In Fig. 7E–H, a phragmoplast is shown that grows out towards the CDZ at one side of the cell. The microtubule label at the edge of the phragmoplast can be seen to connect to the CDZ, thereby bridging GFP–KCBP at the base of the phragmoplast (the distal end) with GFP–KCBP at the cortex (see arrowheads in Fig. 7H,I).
The MyTH4-FERM domain is required for localization to the cortical division plane
The kinesin KCBP from plants is unique because it possesses a calmodulin-binding domain for Ca2+-mediated regulation of its C-terminal motor (Reddy et al., 1996; Narasimhulu and Reddy, 1998; Vinogradova et al., 2009) and because it has an N-terminal MyTH4-FERM domain, which is a protein–protein interaction domain found in myosin-related actin motors (Titus, 2004). Both the C-terminal and N-terminal regions of KCBP contribute to its capacity for microtubule bundling (Kao et al., 2000). To determine which part of the KCBP molecule is responsible for localization to the CDZ, a series of KCBP truncations was made, coupled to N-terminal GFP and expressed in dividing tobacco BY-2 cells (using CaMV35S-driven expression; Fig. 8A). First, we investigated whether Ca2+ regulation was required for KCBP localization. The first fragment, which lacked the C-terminus including the calmodulin-binding helix, located to the CDZ. Even when the entire C-terminal regulatory domain was deleted, including its neck mimic (Vinogradova et al., 2009), localization to the CDZ was observed (Fig. 8B). Therefore, as labelling of the CDZ was routinely observed with these constructs it would appear that the calmodulin-binding domain (and, by extension, Ca2+ regulation per se) is not required for localization of KCBP to the CDZ after preprophase band breakdown.
Next, we tested whether the motor domain of KCBP (including its Ca2+ responsive sequences) was sufficient for localization to the CDZ (fragment Δ8). This proved negative (Fig. 8). Testing additional fragments of KCBP showed that the MyTH4-FERM domain is involved in plasma membrane localization but that this requires the presence of its adjacent coiled coil domain (fragment Δ6) (Fig. 8C). Fragment Δ6 was found to label the cortex at preprophase, mitosis and during the phragmoplast stage. Neither the MyTH4-FERM domain nor the coiled coil region by itself conferred plasma membrane localization during cell division. One possibility is that the MyTH4-FERM domain of KCBP could require oligomerization through the neighbouring coiled coil region to facilitate CDZ localization.
Next, we modelled the structure of the KCBP MyTH4-FERM region using the available crystal structure of the MyTH4-FERM region of human Myosin X (PDB 3PZD) (Wei et al., 2011) and the Phyre2 server. Phyre2 reported a 100% confidence for the model, suggesting that the folding of the KCBP MyTH4-FERM region is highly similar to that of myosin X (Fig. 8D). This high significance suggests functional similarity of the MyTH4-FERM domains of KCBP and myosin X. An alignment of 18 plant KCBPs for modelling sequence conservation in the MyTH4-FERM domain showed that the FERM-C subdomain is especially conserved among plants (Fig. 8E), underlining its importance for KCBP function.
The interaction of AIR9 with KCBP
Our previous work showed that the large microtubule-associated protein AIR9 interacts with the preprophase band of microtubules, but unlike the proteins TAN and RanGAP1 it does not remain present in the CDZ once the band has depolymerized (Buschmann et al., 2006). However, it does return as a tighter cortical ring (the CDS) just as the phragmoplast makes contact with the cortex. AIR9 then moves inwards to decorate the maturing cell plate, forming a torus. Two main findings support the idea that AIR9 directly and specifically interacts with the molecular memory of the preprophase band. First, when phragmoplasts (and cell plates) from tobacco BY-2 cells were fragmented using the herbicide CIPC, only those lobes that contacted the cortex correctly at the exact position of the CDS (i.e. those predicted by the preprophase band) produced an AIR9 torus. Cell plate insertion at aberrant positions away from the preprophase-band-predicted CDS did not produce a torus (Buschmann et al. 2006). Second, as shown in this paper (Fig. 1), when tobacco BY-2 cells expressing GFP-AIR9 were treated with brefeldin A, phragmoplasts became stalled in their expansion at early telophase; however, at the same time GFP–AIR9 could be seen to accumulate precociously at the CDZ. These results support the idea that AIR9 is capable of associating with and recognizing the preprophase memory.
We therefore hypothesized that new components of the preprophase memory could be isolated by using AIR9 as bait in a yeast two-hybrid screen against a cDNA library from Arabidopsis seedlings. Among other cDNAs representing putative AIR9 interactors, the screen yielded several fragments of the minus-end-directed kinesin KCBP (Reddy et al., 1996; Oppenheimer et al., 1997). The interaction of AIR9 with KCBP, as observed for the yeast nucleus, was then confirmed by immunoprecipitation from plant extracts and by pulldowns using bacterially expressed proteins (Fig. 2). When overexpressed in Nicotiana benthamiana epidermal leaf cells, AIR9 is capable of recruiting KCBP to the cortical microtubules, indicating a physical interaction between both proteins (Fig. 3). Furthermore, AIR9 and KCBP were shown to colocalize in the preprophase band, on cross walls and on the phragmoplast (Fig. 2). We emphasize that the localization of both proteins is not identical (e.g. only KCBP remains in the CDZ throughout mitosis and cytokinesis) and the interaction might therefore be temporary and regulated by specific signals, for instance cell cycle progression. Combined, these results support the idea that AIR9 from Arabidopsis interacts with a microtubule motor.
GFP tagging reveals that KCBP localizes to the preprophase memory
The most notable finding of this study, however, is that GFP-tagged KCBP marks the cortical division site throughout division and as such is part of the cortical machinery that memorizes the preprophase band. Previous studies using immunofluorescence to locate KCBP have reported labelling of the preprophase band, the spindle and the phragmoplast but not the CDZ from mitosis onwards (Bowser and Reddy, 1997; Smirnova et al., 1998; Preuss et al., 2003). The absence of CDZ staining in fixation studies invites parallels with studies on TANGLED1, which only localized to the cortical division site after the introduction of live-cell GFP-tagging (Smith et al., 2001; Walker et al., 2007).
Our results show that KCBP can now be added alongside TANGLED1, RanGAP1 and the POK1 kinesin as part of the molecular machinery that ‘memorizes’ the narrow cortical division site to which the outgrowing phragmoplast is directed in cytokinesis (Walker et al., 2007; Xu et al., 2008; Lipka et al., 2014). In this context, it is interesting that both KCBP (this study) and TANGLED1 show remarkably similar behaviour during the transition from metaphase to cytokinesis (Fig. 6). In metaphase, both proteins form a wide band (similar to the preprophase band or CDZ), but this band then becomes narrower so that the distribution during cytokinesis resembles a fine string of beads (Walker et al., 2007; Rasmussen et al., 2011b). It will be interesting to learn how this concentrating mechanism is achieved at the molecular level.
Attempts have been made to reveal the function of KCBP in cell division by manipulating the activity of the kinesin (Vos et al., 2000; Hepler et al., 2002). The interaction of Ca2+/calmodulin with KCBP interferes with microtubule binding of the motor domain (Deavours et al., 1998; Narasimhulu and Reddy, 1998), and Vos et al. (Vos et al., 2000) used microinjection of Tradescantia stamen hair cells to activate KCBP throughout cell division using an antibody directed against its calmodulin-binding domain. This led to the conclusion that the motor activity of KCBP is normally downregulated during metaphase but is reactivated during anaphase when its minus end activity draws microtubules together at the spindle poles. These findings are consistent with our localization results using GFP–KCBP (Figs. 5–7; supplementary material Movie 2).
The interwoven roles of F-actin and microtubules in cell division alignment
Several previous studies visualizing F-actin, and/or using actin-directed drugs, have suggested that F-actin is not only involved in phragmoplast integrity (Molchan et al., 2002; Smertenko et al., 2011) but also in phragmoplast guidance (Lloyd and Traas, 1988; Sano et al., 2005; Kojo et al., 2013). It is not, however, immediately clear how a role for actin in memorizing the division plane (as it must do in occupying the phragmosomal plane during mitosis) is compatible with the absence of F-actin from the narrow cortical ADZ (Cleary, 1995), to which the phragmoplast appears to be attracted. Defining a role for F-actin has also remained problematic because cell division appears to occur normally in several Arabidopsis mutants affected in the actin cytoskeleton (Nishimura et al., 2003). However, recent research has shown that myosin VIII and F-actin are required for phragmoplast guidance in the moss Physcomitrella. In that study, myosin VIII was not only shown to locate to the cortical division site from metaphase onwards but also to the phragmoplast midzone where it was positioned at the microtubule plus ends (Wu and Bezanilla, 2014). Future research might show whether the observations made for the protonemal cells of Physcomitrella, which do not contain preprophase bands, hold true for higher plants.
Concerning the relative contributions of actin and microtubules to division plane determination, it is interesting that the kinesin analysed in this study, KCBP, is a chimera of myosin with a kinesin-14 motor domain (Reddy and Day, 2000). Fig. 8 illustrates the high degree of similarity between the N-terminal MyTH4-FERM domain of KCBP and MyTH4-FERM domains found in metazoan myosin X. Sequence similarity searches suggest that KCBP is the only protein containing a MyTH4-FERM domain in Arabidopsis (data not shown) and Myosin VIII from Physcomitrella does not have a MyTH4-FERM domain. The MyTH4-FERM domain of myosins is required for cargo binding and several interacting proteins have been characterized, including plasma-membrane-bound receptors (Hirano et al., 2011; Wei et al., 2011; Wu et al., 2011). In this study, we show that the MyTH4-FERM domain of KCBP is involved in the localization of a kinesin-14 motor to the CDZ (Fig. 8). A similar region of KCBP is required for AIR9 interaction (Fig. 2). As AIR9 is not part of the prophase memory itself it cannot be responsible for holding KCBP in the CDZ. One possibility is that KCBP is attached through its MyTH4-FERM domain to a continuous resident of the preprophase band memory such as a membrane protein and/or a class of specialized lipids. The MyTH4-FERM domain of KCBP contains an additional microtubule-binding domain (in addition to its motor) (Narasimhulu and Reddy, 1998; Kao et al., 2000). It therefore seems possible that AIR9 is involved in regulating microtubule binding versus CDZ localization of the MyTH4-FERM domain of KCBP. Finally, it is noteworthy that some animal myosins with MyTH4-FERM domains are critically involved in cell division. Myosin X, which exhibits a MyTH4-FERM domain highly similar to KCBP (Fig. 8D), is believed to physically connect microtubules with F-actin and is required for spindle assembly and nuclear anchoring during mitosis in Xenopus (Weber et al., 2004).
A hypothesis for the role of KCBP in cytokinesis
What is clear is that the MyTH4-FERM domain of KCBP facilitates the localization of a minus-end-directed microtubule motor to the cortical division site, that is to the exact position where the cell plate will fuse with the plasma membrane. We have observed that so-called peripheral or exploratory microtubules that extend from the phragmoplast connect with KCBP at the cortical division site (Fig. 7H,I). Previous studies have suggested that microtubules of the phragmoplast are turned over by treadmilling and that the plus-ends of all microtubules are directed towards the cell plate (Euteneuer and McIntosh, 1980; Asada et al., 1991). However, a recent investigation demonstrated that phragmoplast microtubules show dynamic instability rather than treadmilling (Smertenko et al., 2011). Furthermore, according to this new study, nucleation of new microtubules mainly occurs on extant microtubules (thereby preserving microtubule polarity), although there is also likely to be a subset of freely nucleated microtubules with opposing polarity. Because the majority of microtubules nevertheless points towards the cell plate, polar transport of cell wall material to the midline is possible (Smertenko et al., 2011). Such microtubule nucleation on the phragmoplast, together with microtubules emanating from the nuclear surfaces in late cytokinesis would explain the origin of exploratory microtubule plus ends that connect the nucleus and phragmoplast to the cortical division site (Chan et al., 2005; Dhonukshe et al., 2005). The presence of KCBP at the cortical division site therefore suggests a mechanism in which this minus-end-directed microtubule motor forms a bridgehead through which the plus ends of exploratory microtubules attach to the cortex. By processing towards the minus ends of these microtubules KCBP would effectively ‘reel in’ the edges of the phragmoplast. Such a model is not without precedent because cytoplasmic dynein is known to fulfil a similar function in orientating the spindles of animals and yeast. In Saccharomyces, dynein becomes attached to the plasma membrane where this minus-end-directed microtubule motor ‘reels in’ the astral microtubules and helps align the spindle poles (McNally, 2013). Similar mechanisms involving dynein and cortical pulling are at work during cell division in animals and are involved in determining the site of cytokinesis (furrowing) (Kotak and Gönczy, 2013). Flowering plants lack all types of dynein heavy chain and we hypothesize that other minus-end-directed motors, such as KCBP, have taken over some of the molecular functions performed elsewhere by this protein (Lawrence et al., 2001; Frey et al., 2010).
MATERIALS AND METHODS
An entry clone containing the AIR9 promoter sequence was created by amplifying a 1852 base pair sequence immediately upstream of the AIR9 translational start (this promoter sequence includes the AIR9 5′ UTR and its hypothetical intron) with the oligonucleotides AIR9prom_attB1and AIR9prom_attB2, and recombining the obtained DNA fragment with pDONR207. AIR9 promoter::GUS fusions were constructed in a GATEWAY-compatible binary vector (pGWB3).
The genomic construct of KCBP labelled with N-terminal GFP was assembled using GATEWAY-compatible pGWB6. First, the CaMV35S promoter was replaced with the ∼0.75 kb KCBP promoter (Reddy and Reddy, 2004) using restriction enzyme sites for SfbI and XbaI present in pGWB6. Next, using the GATEWAY LR reaction, the entire KCBP coding region, including all introns and its 3′ UTR (contained in an entry clone derived from pDONR207), was recombined into pGWB6 carrying the KCBP promoter.
Further expression clones for CaMV35S-driven GFP-fusions (including the KCBP cDNA and its truncations) were generated using entry clones (derived from pDONR207) and destination vectors pGFP-N-BIN and pGFP-C-BIN (originally generated by Ben Trevaskis, MPI, Germany). The CaMV35S-driven mRFP1-fusions used in the experiments are derived from pH7WGR2 and pH7WRG2 destination vectors of the Ghent clone series (Karimi et al., 2002).
Oligonucleotide sequences used for the PCR-based cloning of promoters, full and partial cDNAs and genomic fragments will be quickly provided on request.
RNA in situ hybridization of whole tissue samples
Arabidopsis seedlings were grown on wet Whatman filter paper for 3–5 days under long day conditions. The preparation and testing of the AIR9 antisense probe was performed as described previously (Zachgo, 2002). Fixation of the seedlings and the following whole-mount in situ hybridization was performed according to Hejátko et al. (Hejátko et al., 2006) but with the tissue incubation with Proteinase K (125 mg/ml) prolonged to 30 min and an additional cell wall digestion step using 1% (w/v) cellulase and 0.2% (w/v) Macerozyme (Althoff et al., 2014).
Reverse-transcriptase polymerase chain reaction
cDNAs were generated based on total RNA preparations (Qiagene) from agar-grown plants using the Superscript II Reverse Transcriptase (Invitrogene) according to the manufacturer's instructions. Oligonucleotide sequences for the amplification of AIR9 fragments will be given on request.
AIR9 and KCBP interaction studies
Immunoprecipitations were based on protein extracts from a cycling Arabidopsis Ler suspension culture (Korolev et al., 2005). Cells were harvested 3 days after sub-culturing and ground with mortar and pestle using liquid N2. Ground cells were further homogenized in one volume immunoprecipitation buffer (50 mM Tris-HCl pH 8, 150 mM NaCl; 5 mM EDTA; 2% (v/v) Triton X-100, proteinase inhibitors by Roche) and incubated on ice for 30 min. Extracts were centrifuged (>20,000 rpm in a SS34 rotor, Sorvall) for 20 min at 4°C and the supernatant was collected. After one repeated centrifugation, the supernatant was cleared through a 0.2 µm filter. 500 µl of the supernatant were then mixed with 20 µl of a polyclonal rabbit antibody against AIR9 (Basu et al., 2005) and rotated end over end at 4°C for 1 h. Next, 50 µl of IP-buffer equilibrated protein A beads (Sigma P3476) were added to the mixture. This was further incubated under slow rotation for at least 1 h or overnight. The beads were then washed three times with cold IP buffer aided using a table-top centrifuge. Finally, the washing liquid was removed entirely before 50 µl of 2x SDS loading buffer were added to the beads. This was boiled for 5 min, cooled down and centrifuged. 15 µl of the supernatant were then subjected to SDS-PAGE using 7.5% (w/v) gels. Western blotting was performed according to standard methods, but without methanol. The blots were then probed with the rat antibody against cotton KCBP. Detection was achieved using Abcam (Cambridge, UK) secondary antibodies and a chemiluminescence kit (Thermo Scientific, Germany).
The reciprocal experiment was based on bacterially expressed proteins. The 132 kDa His-AIR9 fragment, which includes the 11 repeated A9 domains of the full-length protein, was expressed in BL21 (DE3) cells at 20°C overnight using the pET30A vector system and IPTG induction (Basu et al., 2005). The His-AIR9 fragment was purified from sonicated extracts using Nickel columns by Macherey and Nagel (Düren, Germany) according to the manufacturer's instructions. The protein was further purified and concentrated using Amicon columns (Millipore; size exclusion limit<50 kDa). GST-tagged KCBP was expressed using the pDEST15 vector and IPTG induction in BL21 DE3 cells at 20°C overnight. GST-KCBP was purified from sonicated extracts using glutathione-agarose beads (Macherey and Nagel) according to the manufacturer's instructions. The GST-KCBP protein was concentrated and purified from free glutathione using Amicon columns (size exclusion limit<50 kDa). The interaction of the 132 kDa His-AIR9 fragment with GST-KCBP was probed in PBS buffer (including protease inhibitors) at room temperature. After 50 min of gentle agitation equilibrated glutathione-agarose beads were added and the mixture was incubated for a further 30 min. The beads were then washed four times and resuspended in two-fold SDS-buffer. After polyacrylamide gel electrophoresis and western blotting His-tagged AIR9 protein was detected using antibodies by Abcam (Cambridge, UK) and a chemiluminescence kit (Thermo Scientific, Germany).
The yeast-2-hybrid screen against a cDNA library (young seedlings from Arabidopsis) using the 11 repeated A9 domains of AIR9 as bait was performed by the Hybrigenics company (Paris, France).
Fluorescent protein microscopy was performed using a spinning disc confocal by VisiTech (equipped with a Hamamatsu Orca ER detector) and various confocal microscopes from Zeiss and Leica, including Meta710-Zen and SP5 instruments, respectively. Living Arabidopsis seedlings and tobacco BY-2 cells (20 µl) were observed by sealing them between a coverslip and a piece of the gas-permeable Biofoil 25 membrane (VivaScience, Hannover, Germany).
In order to examine the placement of cross-walls in the Arabidopsis rhizodermis whole plants (7–8 days after germination) were transferred to Paul's medium containing 100 µg/ml propidium iodide and incubated in this solution for at least 30 min. Roots were observed with a confocal microscope using a x40 oil immersion objective.
Scanning electron microscopy of Arabidopsis leaf trichomes
Scanning electron microscopy was performed as recently described (Sambade et al., 2014). For the analysis of ovules in wild-type and hemizygous ungud9 plants unfertilized flowers were fixed in 9:1 ethanol:acetic acid solution (v/v) for at least 30 min. Next, the specimen were transferred to pure ethanol and then to 70% (v/v) ethanol (each step 30 minutes). The flowers were then plunged into clearing solution, which consisted of 8:2:1 chloral hydrate:H20:glycerol (w/v/v). After an overnight incubation they were examined using a x60 oil immersion objective using Nomarski optics.
Plant growth and transformation
Tobacco BY-2 cultures were cultivated as described previously (Koroleva et al., 2006) and Arabidopsis seedlings were grown on Paul's medium (Buschmann et al., 2011). Tobacco BY-2 cells were transformed as previously described (Korolev et al., 2005). Columbia-0 suspension cells were transiently transformed as described by Mathur et al. (Mathur et al., 1998). The transient transformation of Nicotiana benthamiana epidermal leaf cells through infiltration with Agrobacterium was performed as previously described (Niehl et al., 2012). Transformation of Arabidopsis was achieved by the floral dip method (Clough and Bent, 1998).
In situ β-glucuronidase (GUS) analysis
GUS staining based on β-glucuronidase expression (Jefferson et al., 1987) driven by the AIR9 promoter was performed as follows. Plant material was vacuum-infiltrated with fixation buffer [1% formaldehyde (w/v), 50 mM sodium phosphate buffer (pH 7), 0.05% (v/v) Triton-X-100] and incubated at room temperature for 40 min. Plant tissues were then washed three times with 50 mM sodium phosphate buffer (pH 7) and vacuum-infiltrated with GUS staining buffer [1 mM 5-bromo-4-chloro-indolyl glucuronide, 1 mM potassium hexacyanoferrate-(II), 1 mM potassium hexacyanoferrate-(III), 10 µM EDTA, 50 mM sodium phosphate buffer, 0.1% (v/v) Triton-X-100]. Samples were incubated at 37°C for 24 h. Tissues were then cleared using 70% (v/v) and 99% (v/v) ethanol washing steps. For photography, specimens were transferred to 40% (v/v) glycerol using two intermediate steps: a 70% (v/v) ethanol wash followed by an ethanol and glycerol (35% and 20%, respectively, v/v) wash. GUS-stained plants were photographed with a Nikon Eclipse 800 microscope equipped with a digital camera.
Arabidopsis mutant lines
In order to obtain strong zwichel mutants in Col-0 background, we ordered T-DNA insertion lines for the ZWICHEL (KCBP) gene from NASC Nottingham. The following lines were obtained: N531704 (dubbed zwiA) with an insertion into the third exon, N639201 (zwiB) with insertion into the ninth exon, N652109 (zwiC) with insertion into the twentieth exon. The zwiC insertion had only a weak trichome phenotype; however, the zwiA trichome phenotype was strong and comparable to zwi9311-11 in RLD2 background (Luo and Oppenheimer, 1999). zwiA was used for the complementation test using the genomic GFP-KCBP fusion.
Mutants of AIR9 were obtained from NASC, from the Arabidopsis TILLING project and from the laboratories of David Twell (University of Leicester, UK) and Ueli Grossniklaus (Zurich, Switzerland). See supplementary material Table S1 for details on air9 mutants analysed in this study.
Computational protein sequence and structure analyses
The structure of the amino acid fragment 110–620 from A. thaliana KCBP was modelled using Phyre2 one-to-one threading (Kelley and Sternberg, 2009) with the structure of human Myosin X (chain A from PDB 3PZD) (Wei et al., 2011). A selection of 17 BLAST orthologs of A. thaliana KCBP was collected from the NCBI Entrez Protein database. A multiple sequence alignment of A. thaliana KCBP and orthologs was constructed using ClustalW (Larkin et al., 2007). The alignment and the structure models were used as an input to display residue conservation in KCBP from plants mapped on to the predicted structure using the ConSurf web tool (Ashkenazy et al., 2010).
Plasmids pGWB3 and pGWB6 were used by kind permission of Tsuyoshi Nakagawa (Department of Molecular and Functional Genomics, Shimane University, Japan). The 35S::mRFP1-MAP4 vector was kindly provided by Daniel Van Damme (VIB Department of Plant Systems Biology, Ghent University, Belgium). We are very grateful to Bo Liu (Department of Plant Biology, University of California Davis, USA) for providing the KCBP-directed antiserum and to Anireddy S.N. Reddy for advice on the cell biology of KCBP. Special thanks to Nora Gutsche for advice on handling bacterially expressed proteins. We further thank Phon Green for lab assistance and Kim Findlay and Grant Calder for help with microscopes. Seeds of air9 deletion mutants were kindly provided by David Twell (Department of Biology, University of Leicester, United Kingdom) and Ueli Grossniklaus (Institue of Plant Biology, University of Zurich, Switzerland).
H.B., J.H.D. and C.W.L. designed the research. H.B., J.D., S.K., M.A.A-N., D.B.S. and E.J.P. performed the research. H.B., M.A.A-N., S.Z., M.H. and C.W.L. analysed the data. H.B. and C.W.L. wrote the article. M.H. and J.H.D. contributed to writing.
This work was supported by the Biotechnology and Biological Sciences Research Council (BBSRC) [grant number BB/E022634/1 to C.W.L. and J.H.D.]; by an EMBO short-term fellowship [grant number ASTF 169.00-2011 to H.B.] ; and by Deutsche Forschungsgemeinschaft [grant number ZA 259/6 to S.Z.]. J.D. was supported by the Leonardo Da Vinci programme of the European Union.
The authors declare no competing or financial interests.