Molecular dissection of plant cytokinesis and phragmoplast structure: a survey of GFP-tagged proteins


(fax +32 9 3313809; e-mail


To identify molecular players implicated in cytokinesis and division plane determination, the Arabidopsis thaliana genome was explored for potential cytokinesis genes. More than 100 open reading frames were selected based on similarity to yeast and animal cytokinesis genes, cytoskeleton and polarity genes, and Nicotiana tabacum genes showing cell cycle-controlled expression. The subcellular localization of these proteins was determined by means of GFP tagging in tobacco Bright Yellow-2 cells and Arabidopsis plants. Detailed confocal microscopy identified 15 proteins targeted to distinct regions of the phragmoplast and the cell plate. EB1- and MAP65-like proteins were associated with the plus-end, the minus-end, or along the entire length of microtubules. The actin-binding protein myosin, the kinase Aurora, and a novel cell cycle protein designated T22, accumulated preferentially at the midline. EB1 and Aurora, in addition to other regulatory proteins (homologs of Mob1, Sid1, and Sid2), were targeted to the nucleus, suggesting that this organelle operates as a coordinating hub for cytokinesis.


Cytokinesis is more complex in plants than in animals due to the presence of a rigid external wall. Testimony to the difficulties the cell wall imposes on the division process are the preprophase band (PPB) and the phragmoplast, two cytoskeletal structures that are necessary to assure adequate positioning and assembly of a new cell wall between the separating sister nuclei (Verma, 2001). The PPB is a ring of actin filaments and microtubules (MTs) arranged in bundles at the cell periphery that surround the nucleus temporarily prior to the onset of mitosis (Mineyuki, 1999). The role of the PPB in cell division is unclear but its occurrence seems to correlate with a landmark, presumably inserted into the plasma membrane that dictates the place where the future cell wall will connect with the mother wall. On the contrary, the phragmoplast guides Golgi-derived vesicles containing cell wall synthesis enzyme complexes to the cell center to construct a callose disk that gradually expands toward the mother wall. Currently, our knowledge on the identity of the molecular players that are involved in the functioning of the PPB and the phragmoplast is far from complete. Genetic screens of Arabidopsisthaliana plants have identified several mutants with cytokinesis defects and, in some cases, the affected genes have been isolated (Assaad, 2001; Smith, 2001). Interestingly, the genes identified in this way were plant specific, such as tangled (Smith et al., 1996), or were found to encode proteins with defined activities related to vesicular trafficking and fusion (Assaad, 2001; Bednarek and Falbel, 2002). Kinesins have been proposed to be responsible for the transport of Golgi-derived vesicles to the forming cell plate in addition to a role in maintaining the integrity and organization of phragmoplast MTs (Liu and Lee, 2001). Expansion of the cell plate requires the activation of a MAPK signaling pathway mediated by the formation and targeting of a MAPKKK–kinesin complex to the phragmoplast (Nishihama et al., 2002).

Despite the uniqueness of the PPB and the phragmoplast to plant species (Sawitzky and Grolig, 1995), a certain structural resemblance can be recognized to ring-shaped cytoskeletal organizations that occur in mitotic yeast and animal cells. The organization of the phragmoplast is reminiscent of the mid body in cytokinetic animal cells. Both are composed of opposing bundles of MTs that face one another with their positive ends at the center (Field et al., 1999). However, the so-called actomyosin ring in yeast and animal cells assists the cytokinesis process by recruitment of membrane components and by a myosin-driven contraction of the actin filaments in the ring, leading to a narrowing of the division plane, also referred to as constriction (Bi, 2001). Genetic studies in yeast have identified numerous components of the contractile ring that function in ring assembly, positioning, and contraction (Bi, 2001). Because many of these proteins are evolutionarily conserved in animals, common molecular mechanisms may govern aspects of eukaryotic cell division (Feierbach and Chang, 2001). However, it is not clear to what extent these common factors occur in plants and, if so, what their contribution to cytokinesis might be.

To investigate the alleged similarity of these cytoskeletal structures and to build a comprehensive list of the components involved, the Arabidopsis genome was searched for cytoskeleton and cytokinesis-related genes. The selection of genes and gene families analyzed was restricted to those for which evidence of their implication in cytokinesis was found in the literature. A recent review by Guertin et al. (2002) gives an excellent overview of cytokinesis-related genes in eukaryotes. Here, Arabidopsis actin-binding proteins (profilins, formins, Rho-type GTPases, and myosins), MT-binding proteins (MAP65, EB1, and CLIP170), regulatory proteins, and putative cell cycle-controlled proteins were subcellularly localized in interphase and dividing Bright Yellow-2 (BY-2) cells of tobacco as well as in transformed Arabidopsis plants by GFP tagging.


Selection of putative cytokinesis-related Arabidopsis genes

Because cell division and cytokinesis are recurrent processes in all eukaryotes, accomplished by common underlying mechanisms of signaling, membrane traffic, and cytoskeleton organization (Hales et al., 1999), we explored the Arabidopsis genome sequence for homologs of known cytokinesis genes by amino acid sequence blast searches. Our survey pointed out that approximately one-third of the 53 genes and gene classes searched were conserved in the Arabidopsis genome. The genes that were considered for the analysis are listed in Tables S1–S4 (Supplementary material available with this article online). The poor representation of the known cytokinesis genes probably reflects diversification in cytokinesis and division plane positioning in plants. In particular, the lack of an actin or actomyosin ring in plant cells is remarkable and fully in line with the absence of genes encoding septins, type-II myosins, and IQGAP, which are major components of the actin ring in yeast and the cleavage furrow in animal cells (Guertin et al., 2002). The emergence of an actin ring requires the inactivation of the core cell cycle kinase that is mediated by anaphase-promoting complex-controlled cyclin degradation (Balasubramanian et al., 2000). Polo kinase coordinates the mitotic and cytokinetic events by stimulating the proteolytic activity of the anaphase-promoting complex and takes care of proper positioning and assembly of the actomyosin ring through phosphorylation of the Mid1 protein and association with the septins cdc11 and cdc12 (Paoletti and Chang, 2000). Neither Polo kinases nor genes homologous to the interacting Mid1 were found in the Arabidopsis genome, suggesting that nuclear division and spatial control of cytokinesis are not coupled in plants or rely on an alternative regulatory mechanism.

The contractile ring structure is composed of actin filaments and contains besides septins, IQGAP, and myosin, also actin-associating proteins, including profilin, formin, fimbrin, and the arp2/3 complex proteins. Although no actomyosin ring is produced in plant cells, actin organization is important for the completion of cytokinesis (Wick, 1991). By using conserved sequences of the respective actomyosin constituents, we identified several putative Arabidopsis homologs (Table S1). Formin and profilin have been assigned functions more related to the dynamics of actin polymerization and are present in Arabidopsis as small protein families consisting of 21 formins and four profilins (Cvrčková, 2000; Deeks et al., 2002). Myosin-like sequences were retrieved, but were neither of type II nor type IV myosins, which are specific for the contractile actomyosin ring. All 17 myosin-like Arabidopsis sequences put forward so far belong to the type VIII and XI class, and perhaps a third, ill-defined subfamily (Reddy and Day, 2001). We also searched for GTPase-activating proteins, because small GTPases of the Rho, Rac, and Cdc42 family and Rho-type GTPase-activating proteins are required for cytokinesis in animal cells (Table S2). The Rho-type Arabidopsis AtROP1 interacts with a callose synthase complex at the phragmoplast (Hong et al., 2001).

Because MTs are prominent components of the PPB and the phragmoplast, we searched for MT-binding proteins. Particularly, MT plus-end-binding proteins of the EB1 and CLIP170 family are important for the establishment of cellular polarity and the dynamics of MT structures. Three EB1-like open reading frames (ORFs) were identified (Table S1). We also looked for MAP65/Ase1/PRC1-like proteins, a new class of MT-binding proteins of which Arabidopsis has nine members (Table S2; Hussey et al., 2002; Schuyler et al., 2003). Two ORFs were found with similarity to the spindle pole body protein Sad1+ from yeast. This protein also associates with MTs and the nuclear membrane, and may therefore have a task in coordinating the cytoskeleton–membrane interphase (Hagan and Yanagida, 1995).

To test and evaluate GFP localization of Arabidopsis proteins for which no function has been assigned based on sequence information, we selected the homologs of 24 tobacco cDNA tags (indicated as T and a number; Table S3) that have been shown to be cell cycle modulated in synchronized BY-2 cells (Breyne et al., 2002). The corresponding ORFs of seven S/G2-phase and 17 G2/M-phase expression tags were selected for cloning and analysis (Table S3).

In total, 103 Arabidopsis ORFs were selected for cloning. The genes selected and yeast cytokinesis-related genes that had no obvious homologs in Arabidopsis are listed in Tables S1–S4. For 75 ORFs, we obtained GATEWAY-adapted PCR fragments by using cDNA from an Arabidopsis cell suspension as template. The PCR fragments were cloned in frame with GFP placed at the C-terminal end, unless stated otherwise. Expression of the fusion proteins was driven by the 35S promoter. For some of the ORFs also N-terminal GFP fusions were generated. Upon introduction of the constructs into BY-2 cells, microscopic analysis identified 56 transgenic lines with GFP fluorescence. The fluorescence patterns of these lines, in addition to information on the different constructs that were produced, can be found via links accompanying Tables S1–S3 at our website (additional data online:

GFP-fusion proteins localized to the nucleus, chromosomes, and the nuclear envelope

The S and G2 phases are marked by changes in gene expression and nuclear DNA structure reorganizations (Heslop-Harrison, 2003). Chromosomes were labeled in dividing Arabidopsis and BY-2 cells transformed with a construct containing the S phase-specific T9 gene (At5g11860) (Figure 1a,c,e). T9-GFP fluorescence was concentrated in five to 10 nuclear dots in Arabidopsis (Figure 1a,b) and many dots in BY-2 (Figure 1f). These nuclear dots corresponded to regions of the nucleus that were also heavily stained with 4,6-diamidino-2-phenylindole (data not shown). According to the size of the nuclei and the timing of appearance, the chromosomal arrangement in five fluorescent dots seemed to occur in G1 cells and in 10 dots in G2 cells (Figure 1a,b). During anaphase, the T9-GFP label comigrated with the chromosomes to the poles (Figure 1c). In the reassembled nuclei, again five dots were observed (Figure 1d).

Figure 1.

Targeting of T9-GFP to chromosomes and chromocenters.
(a) Arabidopsis root epidermal cells showing a metaphase cell with lined-up chromosomes and nuclei containing five fluorescent nuclear dots.
(b) Arabidopsis root epidermal cell containing 10 fluorescent nuclear dots. The nucleus is approximately twice the size of that of G1 nuclei.
(c) Anaphase cell in Arabidopsis root epidermal layer (arrow). Fluorescence is concentrated at the chromosomes.
(d) Same cell as in (c) after completion of cytokinesis (arrows). T9-GFP fluorescence is concentrated in five nuclear dots.
(e) BY-2 metaphase cell showing fluorescence of T9-GFP concentrated at the chromosomes.
(f) Nuclear localization of T9-GFP in BY-2 nucleus and in nuclear foci.
Scale bars = 5 μm (a,b), 10 μm (c,d,f), 20 μm (e).

Several kinases (Aurora, Dyrk), phosphatases (PP2A), and putative mitotic exit network (MEN)/septation initiation network (SIN) components (AtSid1, AtSid2, and AtMob1) were also targeted to the nucleus (additional data online). In yeast, the spindle pole body operates as a signaling center during cytokinesis (Simanis, 2003). MEN/SIN regulators, such as Sid1 and Mob1, as well as Polo kinase temporarily associate with the spindle pole body at some point in the cell cycle. In analogy to this function, centrosomes have recently been implicated in completing cytokinesis (Doxsey, 2001). Because plant cells do not possess a spindle pole body or centrosomes, the nuclear targeting of Sid1 and Mob1 can be taken as an argument to put forward the nucleus as an alternative center for the coordination of cytokinesis.

In this regard, the localization of GFP fusions of AtSad1a (At5g04990) and AtSad1b (At3g10730), two Arabidopsis homologs of the spindle pole body protein Sad1 (Hagan and Yanagida, 1995), was informative. They brightly stained the nuclear envelope in BY-2 and Arabidopsis cells (Figure 2a,b). In dividing cells, AtSad1a-GFP strongly accumulated in dots that were associated with the nuclear rim or in the cytoplasmic space close to the plasma membrane (Figure 2b,c). Upon nuclear envelope breakdown, AtSad1a-GFP surrounded the spindle and the phragmoplast. In addition, we reproducibly found that metaphase cells contained one or a few, very bright fluorescent spots at one end of the spindle and one or more spots at the other end, attached to the plasma membrane (Figure 2c). Although plant cells have no centrioles, the spots bear resemblance to Sad1 bodies described in fission yeast mutants affected in meiosis-specific spindle pole body integrity (Jin et al., 2002). In plant cells, the nuclear surface is strewn with MT nucleation centers containing γ-tubulin, which become polarly organized when spindle formation begins (Binarováet al., 2000; Schmit, 2002). These nucleation centers also contain AtSpc98, a homolog of the yeast spindle pole body protein Spc98 (Erhardt et al., 2002). In view of the role of the spindle pole body and the centrosomes in establishing a bipolar spindle and in coordinating events leading up to cytokinesis (Doxsey, 2001), the nucleus and nuclear surface seemingly play an important role in taking on these tasks in plant cells.

Figure 2.

Localization of AtSad1a to the nuclear envelope and Sad1-like bodies.
(a) Confluent labeling of the nuclear membrane by overproduction of the AtSad1a-GFP fusion product. Due to nuclear invaginations, knot-like structures penetrate the nucleus.
(b) Concentration of AtSad1a-GFP in dots that are in the cytoplasm and at the nuclear rim. Cells are leaving the mitotic phase.
(c) BY-2 cell at metaphase. AtSad1a-GFP-labeled dots that are at the polar ends of the spindle (arrow). The distribution is reminiscent of that of the spindle pole bodies and associated Sad1 protein in dividing yeast cells.
Scale bars = 10 μm (a,b), 20 μm (c).

MT-binding proteins MAP65 and EB1

Arabidopsis contains nine MAP65-like genes (Hussey et al., 2002) of which seven (AtMAP65-1, AtMAP65-2, AtMAP65-3, AtMAP65-4, AtMAP65-5, AtMAP65-6, and AtMAP65-8) were cloned and analyzed. Expression of GFP-tagged proteins of AtMAP65-1, AtMAP65-3, AtMAP65-5, and AtMAP65-8 resulted in fluorescent labeling of the cortical array, an MT structure that is unique to plant cells (Hardham and Gunning, 1978). Collapsed Z-stack images of labeled cortical arrays of representative interphase BY-2 cells are shown in Figure 3(a–d). The MAP4 MT-binding domain (MBD) and tubulin TUA6 were used as reference for the fluorescent marking of MTs (Granger and Cyr, 2000; Hasezawa et al., 2000). Striking differences in labeling patterns were observed. The labeling pattern in BY-2 cells containing the AtMAP65-5-GFP construct resembled best the cortical MT array visualized by the MBD-GFP control (Figure 3c). AtMAP65-1-GFP, on the contrary, labeled much thicker bundles of cortical MTs and concentrated in dot-like structures that were attached to the apparent end of the MTs (Figure 3a, arrows). These dots were fixed in position over a period of 20 min and may correspond to anchor points in the cell cortex, perhaps connected to the plasma membrane (Hardham and Gunning, 1978). AtMAP65-3-GFP fluorescence was very pronounced at the centrally located endocytic MTs that formed very thick bundles in comparison with the poorly stained, much thinner cortical MTs. The endocytic MT bundles wrapped around the nucleus and formed extensions that were connected to the cortical MTs of the cell poles (Figure 3b). Similar structures could also be observed in transgenic BY-2 cells transformed with other MAP65 constructs, the GFP-MBD, and TUA6-GFP controls. However, in these cases, the cells were always committed to enter mitosis. The constitutive and high expression of AtMAP65-3-GFP may therefore have led to the persistence or formation of these endocytic MTs. AtMAP65-3-GFP caused extensive bundling of MTs at the center of the cell, leaving cortical arrays within the same cell apparently unaffected (Figure 3b). The endocytic MTs must therefore have a different composition allowing for the differential binding of AtMAP65-3-GFP. Another labeling pattern was seen with the AtMAP65-8-GFP construct. Whereas AtMAP65-3-GFP and AtMAP65-5-GFP labeled the MTs homogenously, AtMAP65-8-GFP fluorescence appeared as dotted lines (Figure 3d).

Figure 3.

Subcellular localization of cytoskeleton proteins in interphase cells.
(a–d) Projections of Z-stack confocal images, approximately two-third in the Z-axis of transgenic BY-2 cells.
(a) Labeling of cortical microtubules (MTs) by AtMAP65-1-GFP and concentration at MT ends, suggesting a preference for distal end labeling. Arrows indicate dot-like structures at MT ends.
(b) Association of AtMAP65-3-GFP with thick bundles of endocytotic MTs that emanate from the nucleus to connect with the cortical array at the cell poles. The cortical MTs were much less stained and were visualized by increasing the laser intensity of the confocal microscope. The cortical MTs appeared thinner than those labeled with AtMAP65-1-GFP.
(c) Labeling of the cortical array by AtMAP65-5-GFP as a fine mesh of transverse MTs. The labeling pattern was very similar to that of control cells expressing MBD-GFP.
(d) Punctate accumulation of AtMAP65-8-GFP along MTs. The density of visualized MTs was low compared with that of control lines, suggesting that AtMAP65-8-GFP was associated with a subpopulation of MTs in the cortical array.
(e) Bright fluorescence of AtMAP65-8-GPF at separate foci near the spindle poles during mitosis (arrows). Some of the spindle MTs were also labeled.
(f) AtEB1b-GFP in Arabidopsis leaf epidermis. The plus-ends polymerize more frequently than the minus-ends and become visible as comet-like structures. Two stomata are shown in the left bottom corner.
(g) Labeling of the post-cytokinetic walls in BY-2 cell files by AtFH6-GFP.
(h) Optical section through an Arabidopsis root tip. AtFH6-GFP was primarily targeted to the transverse walls.
Scale bars = 20 μm (a–d, f and h), 10 μm (e).

The punctate labeling of cortical MTs was also observed in cells expressing AtEB1a and AtEB1b GFP-fusion proteins (Figure 3f; Chan et al., 2003 [AtEB1a: At3g47690]; Mathur et al., 2003 [AtEB1b: At5g62500]). AtEB1a and AtMAP65-8 concentrated at the spindle poles, suggesting that they are associated with the minus-ends of MTs or the MT-organizing centers (Figure 3e; Chan et al., 2003). The mouse adenomatous polyposis coli-binding protein EB1 localizes to centrosomes where it could serve as an anchor point for MT minus-ends (Askham et al., 2002).

However, the localization patterns of EB1 and the yeast homolog BIM1 have also established them as MT plus-end-binding proteins (Mimori-Kiyosue and Tsukita, 2003; Mimori-Kiyosue et al., 2000; Schwartz et al., 1997). At the plus-ends of MTs they recruit interacting factors involved in an MT-capturing mechanism that influences MT dynamics (Korinek et al., 2000; Lee et al., 2000). Fluorescence of the dynamic plus-ends of MTs was visible as comet-like spots in Arabidopsis epidermal cells that expressed AtEB1a-GFP or AtEB1b-GFP (Figure 3f; additional data online; Chan et al., 2003). Although the existence of a capturing mechanism in plants has not been reported, it is likely that the plus-end-bound AtEB1 proteins are involved in such a process (Gundersen, 2002).

Remarkably, in contrast to AtEB1a and AtEB1b, AtEB1c did not or very weakly label cortical MTs in interphase cells (data not shown). Only with moderate to high laser excitation intensities, AtEB1c was seen to bind MTs present at the cell tip, and to some extent with the PPB MTs in prophase cells (Figure 5a). The majority of the AtEB1c protein however was concentrated in the nucleus (Figure 5a). During mitosis AtEB1c was released from the nucleus upon nuclear envelope breakdown and subsequently reentered the nucleus at telophase (Figure 5a).

Figure 5.

Time-lapse analysis of cytokinesis in BY-2.
Fluorescence of dividing BY-2 cells was imaged at time intervals indicated.
(a) Dynamics of AtEB1c-GFP localization. The highest concentration of the protein is present in the nucleus during interphase, but a small amount is observed at the cell poles (full arrowheads). Upon envelope breakdown, AtEB1c was released and then immediately associated with the spindle microtubules (MTs) and subsequently with the phragmoplast MTs. At the center of the developing cell plate, the MTs depolymerized. Some of the fluorescence remained at the nuclear surface. When the nuclear membrane resealed, AtEB1c-GFP went partially back into the nucleus and partially moved to the junctions of the cell plate and mother wall (open arrowheads) and to the cell tip (full arrowheads).
(b) Dynamics of AtMAP65-5-GFP localization. AtMAP65-5-GFP was excluded from the nuclear space until formation of the spindle. The phragmoplast MTs were much more strongly labeled than the spindle. In the second phase of cell plate formation, which corresponds to a centrifugal expansion of the phragmoplast, fluorescence persisted as focal points (plasmodesmata) at the matured cell plate. When cytokinesis was terminated some of the label was still present in the cross wall (arrow). Parallel lines indicate the position of the preprophase band.
Scale bars = 20 μm.

Actin-binding proteins

The actin-binding proteins that were analyzed did not clearly label actin cables or the fine actin mesh. Differential interference contrast images of cells expressing actin-binding GFP-fusion proteins often showed accumulation of granular structures in the cytoplasm that are also observed in damaged or stressed cells. Specific localization patterns were obtained for GFP fusions of formin, profilin, and myosin. GFP-fusion constructs of the type-I formins AtFH6 (At5g67470) resulted in fluorescent labeling of cross walls of filamentous BY-2 cells (Figure 3g). In Arabidopsis roots and stems, AtFH6-GFP fluorescence was present at the cell periphery predominantly at the cross walls (Figure 3h). Type-I formins are structurally different from their animal and fungal counterparts in that they harbor an N-terminal signal sequence followed by a transmembrane domain, which could target these proteins to the secretory system (Cvrčková, 2000; Deeks et al., 2002). AtFH6-GFP fluorescence appeared exclusively at the periphery of the cell and may therefore be associated with the plasma membrane.

Phragmoplast proteins

In total, 15 fusion proteins labeled the cytokinetic apparatus (Figure 4; see additional data online). These proteins categorized into functional groups as MT-binding, actin-binding, signaling, and novel proteins. These proteins associated with separate subregions of the phragmoplast and/or the cell plate, demonstrating the existence of distinct structural components of the cytokinetic apparatus. The most important findings are summarized here.

Figure 4.

Proteins targeted to the phragmoplast.
Confocal projections (a and b) and single optical sections (d–h) of BY-2 phragmoplasts. Arrowheads indicate the position of the equatorial plane.
(a) Labeling by AtEB1a-GFP of MT plus-ends that are assembled at the midline. In addition, brightly fluorescent dots occur in the cytoplasm and at the poles of the daughter nuclei farthest from the equatorial plane.
(b) Strong AtMAP65-1-GFP fluorescence concentrated at the phragmoplast MTs and dots surrounding the nuclei. In contrast to AtEB1a (a), AtMAP65-1-GFP was absent from MTs at the midline.
(c) Concentration of AtMAP65-3-GFP exclusively at the plus-ends of MTs in the midline.
(d) Accumulation of AtMAP65-5-GFP at the cell plate. An expanding phragmoplast is shown with the cell plate at the center and phragmoplast MTs at the border (brace). Fluorescence is most pronounced in the region of the maturing cell plate as tubular structures, possibly plasmodesmata crossing the cell plate.
(e) End labeling of MT by AtMAP65-8-GFP that coresides with the MT-organizing centers located at the nuclear surface. The minus-ends of the phragmoplast MTs were also labeled.
(f) Concentration of actin-binding protein myosin ATM1-GFP (AT3g19960) at the maturing cell plate. Fluorescence was evenly distributed, suggesting its association with a membrane compartment.
(g) Concentration of T22-GFP at the cell plate and at the midline of the expanding phragmoplast ring.
(h) Concentration of At-Aurora 1-GFP at the young cell plate and at the position where formerly the spindle poles were.
Scale bars = 20 μm.

The AtEB1a, AtEB1b, and AtEB1c proteins localized to the phragmoplast MTs (Figure 4a; additional data online). AtEB1a showed a preference for binding MTs near the midline, suggesting they were concentrated at the plus-ends of the antiparallelly arranged MTs, which can best be seen in a collapsed image of a Z-stack series of the phragmoplast (Figure 4a). During cytokinesis, all three AtEB1-GFP fusions accumulated in highly fluorescent dot-like structures appearing at the surface of the daughter nuclei and some dispersed throughout the cytoplasm (Figures 4a and 5a; additional data online). The dots that were in close contact with the nuclei were usually larger than the cytoplasmic ones and they mainly concentrated at the position where formerly the poles of the spindle were located. After completion of cytokinesis, the fluorescent dots disappeared. Subsequently, part of the AtEB1c-GFP protein was redistributed to the extreme ends of the BY-2 cell (Figure 5a). AtEB1c could therefore play a role in the guidance of MTs toward the cell tips in agreement with functioning in the capturing of MTs as proposed for the yeast EB1 homolog BIM1 (Korinek et al., 2000; Lee et al., 2000).

AtMAP65-GFP proteins were also associated with phragmoplast MTs but each had its own specific localization pattern. AtMAP65-3 was strictly targeted to the midline and did not bind MTs or MT ends that were located outside the midline, in agreement with the immunolocalization data recently reported by Müller et al. (2004; Figure 4c). On the contrary, AtMAP65-1 and AtMAP65-8 were excluded from the phragmoplast midline and labeled either along the MTs (AtMAP65-1; Figure 4b) or at the outer extreme ends of the phragmoplast MTs near the nuclear surface where the MT minus-ends are located (AtMAP65-8; Figure 4e; Chan et al., 2003). As AtMAP65-8-GFP also concentrated at foci at either side of the spindle in metaphase cells (Figure 3e), this protein seems to associate with MT minus-ends. In addition to association with MTs, AtMAP65-1-GFP and AtMAP65-8 were concentrated in dots that surrounded the nuclei and in the cytoplasm similar to those observed with the AtEB1 constructs (Figure 4b,e). AtMAP65-5-GFP showed a complex localization pattern throughout mitosis and in particular during development of the phragmoplast. AtMAP65-5-GFP fluorescence was more intensely associated with the phragmoplast than with the PPB and the spindle (Figure 5b). It labeled the MTs from a young, emerging phragmoplast but did no longer bind along the MTs when the disk-shaped phragmoplast transformed into a ring-shaped, centrifugally expanding structure (Figure 5b). The MTs of the ring-shaped phragmoplast were not or poorly labeled, but instead AtMAP65-5-GFP was concentrated in the cell plate and associated with the surface of the separated nuclei (Figures 4d and 5b). During the phragmoplast expansion phase the central MTs are depolymerized, callose is deposited, and plasmodesmata are formed (Heinlein, 2002; Laporte et al., 2003; Samuels et al., 1995). Because AtMAP65-5 remained present in the cell plate after division was completed and appeared to traverse the newly formed cross wall (Figures 4d and 5b), it is possible that AtMAP65-5 is an integral part of plasmodesmata. Plasmodesmata are pierced with MTs and indeed incorporate MT-binding proteins (Boyko et al., 2002). For instance, GFP-tagged grapevine fanleaf viral movement protein accumulates in the developing cell plate of dividing BY-2 cells leading to the formation of fluorescent tubular structures within the immature cross wall reminiscent of the structures labeled by AtMAP65-5-GFP (Laporte et al., 2003).

In addition to microtubular structures, cell plate formation also relies on actin and actin-dependent transport and on fusion of Golgi-derived vesicles (Wick, 1991). We found that the GFP fusions of the actin-binding protein myosin (At3g19960) and an unknown protein T22 (At3g01780) were primarily localized at the developing cell plate (Figure 4f,g). T22 was most strongly associated with the leading edges of the cell plate and followed the borders of the expanding phragmoplast (Figure 4g). In contrast, myosin was prominent throughout the cell plate and remained associated with the cross wall (Figure 4f). T22 was removed readily from the post-cytokinetic wall, indicating that T22 and myosin were differentially processed after completion of cytokinesis (data not shown).

Finally, we identified three Aurora kinases in Arabidopsis. The protein corresponding to At4g32830 (At-Aurora 1) labeled spindle MTs (additional data online) and the cell plate (Figure 4h). Higher eukaryotes have A-, B- and C-type Aurora kinases of which the B-type is essential for cytokinesis (Carmena and Earnshaw, 2003). Overexpression of GFP-tagged At-Aurora 1 led to the formation of binucleated cells that were not observed in wild-type BY-2 or transgenic calli transformed with either of the two other Aurora proteins (Figure 6a). Because chromosome separation appeared normal in the binucleated cells, At-Aurora 1-GFP must have acted after chromosome segregation to inhibit cytokinesis in a dominant negative manner. Binucleated cells were also observed in transgenic cultures overproducing AtEB1c-GFP, but not in cultures producing AtEB1a-GFP or AtEB1b-GFP (Figure 6b).

Figure 6.

Cytokinesis defects in transgenic BY-2 cells.
(a) Binucleated cell in At-Aurora 1-GFP line. Two nuclei lie adjacent to each other without a separating wall. At-Aurora 1-GFP is present in the cytoplasm and the nucleus and is concentrated at the nuclear rim.
(b) Binucleated cell in AtEB1c-GFP cell line. The brightness of the image was increased to reveal fluorescence associated with endocytic microtubules (MTs). A cell plate is missing in-between the two separated nuclei. Instead, a few MTs connect the nuclei to each other. Fluorescent dots at the cell periphery indicate the presence of a cortical array. Dense material at one side of the cell is seen (arrow) that could be the remnants of an unsuccessful attempt to make a cell wall.
Scale bars = 20 μm.


Because eukaryotic cells are highly compartmentalized, the distribution and localization of a protein is intrinsically bound to its function. With this principle in mind, screening methods using random cDNA-GFP constructs to approach plant gene function have been developed before (Cutler et al., 2000; Escobar et al., 2003). The screening of such GFP-fusion libraries requires a substantial input of effort with a difficult-to-predict valuable outcome. To increase the effectiveness of a GFP-based screening method we have opted for a more selective cloning of genes related to cytokinesis and cell division processes.

As the development of a cytokinesis program must have had its origin early on in the evolution of the eukaryotic lineage, we anticipate ample conservation of the implicated molecular components that have previously been identified through genetic analysis of yeast and animal cells (Feierbach and Chang, 2001). However, merely a third of the genes searched for have obvious homologs in the Arabidopsis genome. Outstanding absentees are Polo kinase and several of the structural components of the actomyosin ring, an actin structure not formed by plants. The lack of a Polo kinase is surprising, especially because one of its roles is to trigger cytokinesis directly through the activation of the SIN pathway in fission yeast (Song and Lee, 2001; Tanaka et al., 2001). Most of the components that are part of the MEN/SIN signaling machinery have homologs in Arabidopsis, suggesting that at least this end of the regulatory pathway has been conserved. Further support for the conservation of a SIN-like pathway in plants follows from the complementation of a loss-of-function mutation in Saccharomyces pombe of the MAP3K kinase SIN element cdc7p by a Brassica napus homolog (Jouannic et al., 2001).

Our survey has generated a useful set of marker GFP-fusions that are particularly relevant for the study of cell division-related processes. For example, the T9 protein is a unique marker that labels the nucleus and concentrates at nuclear sites of condensed chromatin also referred to as chromocenters (Fransz et al., 2002). These areas correspond to the centromeric regions of Arabidopsis interphase chromosomes (Fransz et al., 2002; Talbert et al., 2002). The association of T9-GFP to the chromocenters can be exploited to monitor chromatin organization according to the cell cycle phase, either in single cell cultures or in cells imbedded in the context of a whole tissue. The amino acid sequence of T9 (At5g11860) indicates similarity with a nuclear LIM-interacting protein (NLI interacting factor 1 or NIF1), a human protein for which the function remains to be identified (Jurata et al., 1996).

The AtSad1a and AtSad1b GFP fusions are interesting because of a potential connection between the nuclear membrane and MT nucleation. Moderate overexpression of Sad1 in yeast leads to association with the nuclear envelope (Hagan and Yanagida, 1995). Fission yeast Sad1 protein is a constitutive spindle pole body (SPB) component essential for spindle formation and function. Plants do not possess SPB or centrosome-like structures, but nevertheless harbor genes homologous to the SPB elements Spc97, Spc98, and Sad1 (Erhardt et al., 2002; this work). The AtSpc98 protein colocalizes with γ-tubulin at the cell cortex and the nuclear periphery where it stimulates tubulin polymerization (Erhardt et al., 2002). So-called Sad bodies that are not the products of ordinal SPB duplication have recently been found associated with the nucleus in zygotic yeast cells (Goto et al., 2001). The occurrence of the Sad bodies, which contained other SPB components, correlates with particular centromere and telomere arrangements in meiotic prophase nuclei, indicating that the functioning of these proteins is not restricted to MT nucleation (Goto et al., 2001; Jin et al., 2002). The AtSad-GFP concentrates in highly intense fluorescent dots located in the cytoplasm as well as dots at either side of the spindle, sometimes associated with the cell periphery. Because of the occurrence of fluorescent dots in the cytoplasm, we suspect that the function of AtSad, as in yeast, is implicated in a process other than that of MT nucleation.

Despite poor conservation of cytokinesis genes, our survey has identified 15 GFP-fusion proteins that are targeted to the cytokinetic apparatus. The proteins involved are members of the protein families EB1-like, MAP65, myosin, formin, Aurora, Rho-like GTPases Rop, and a novel protein of unknown function provisionally nominated T22. The fluorescence pattern in BY-2 cells of each of these proteins was highly distinctive and marked different subregions of the cytokinetic apparatus. EB1 and MAP65 proteins highlighted MTs with a preference for either the plus-ends, the minus-ends, or along the length of the polymers.

Recently, AtEB1a was reported to be a MT minus-end binding protein and AtEB1a associates with slowly moving dots dispersed in the cortical array that are believed to be MT minus-ends based on their mobility properties (Chan et al., 2003). As AtEB1a was also present at fast-growing MT tips, it was proposed that AtEB1a has the capacity to bind both MT plus- and minus-ends. Our data do not confirm the association of AtEB1a-GFP with MT minus-ends perhaps because a different suspension cell type was studied (Arabidopsis versus BY-2). Alternatively, the expression levels in stably transformed BY-2 cells may have been more moderate compared with a transient expression system used by Chan et al. (2003) and have disallowed detection of minus-end association. For instance, in the continuously propagated AtEB1a-GFP and AtEB1b-GFP BY-2-expressing cell lines, we did not observe label along MTs and the fluorescence at the MT ends were 1–2 μm in size rather than 2–5 μm as was reported (data not shown; Chan et al., 2003; Mathur et al., 2003). In agreement with the findings of Mathur et al. (2003) who showed that AtEB1b, in addition to the MT plus-ends, associates with the endoplasmic reticulum and the endomembrane system we found background fluorescence in the cytoplasm that remained upon continuous propagation. Given that AtEB1a and AtEB1b were primarily plus-end MT-binding proteins, the localization of AtEB1c-GFP in the nucleus was unexpected. The AtEB1c-GFP MT-binding capacity is revealed during mitosis when it is released from the nucleus after envelope breakdown and it labels the spindle and phragmoplast. Whether AtEB1c has a preference for MT plus-ends was difficult to detect because of the three-dimensional organization and density of MTs in the spindle and the phragmoplast structures. Because AtEB1c gets redistributed to the extreme tips of the newly born cells when cytokinesis is completed, it is possible that AtEB1c plays a role in polarized cell growth similar to the MT plus-end-binding protein CLIP170 in fission yeast (Feierbach et al., 2004).

AtMAP65-3-GFP was associated with phragmoplast MTs near the midline and at the spindle midzone, suggesting that AtMAP65-3 binds to overlapping MT plus-ends. In addition, we show that AtMAP65-3-GFP labeled the cortical array although the fluorescence was much weaker compared with that of other MAP65-GFP fusions and the signal coming from endocytic MTs. Immunolocalization of AtMAP65-3 with a specific antibody does not confirm localization to the cortical array or the spindle and showed an exclusive association with the phragmoplast (Müller et al., 2004). As AtMAP65-3 transcription peaks at mitosis, protein synthesis is likely cell cycle controlled and limited to the mitotic phase (Menges et al., 2003). Continuous expression driven by the 35S promoter may therefore lead to artificial accumulation of AtMAP65-3-GFP at the cortical MTs. As AtMAP65-3-GFP was not concentrated at the plus-ends of cortical MTs, suggesting that the antiparallel arrangement of the MT plus-ends, as it occurs in the phragmoplast midline and the spindle midzone, is needed for strong MT binding. Biochemically purified Nicotiana tabacum NtMAP65 protein cross-links in vitro stabilized MTs (Chan et al., 1999). Antibody raised against purified NtMAP65 protein binds to a small family of MAP65 proteins collectively called NtMAP65-1 that were localized to the midline in cytokinetic BY-2 cells, in line with a role in stabilization of the antiparallel phragmoplast MTs (Smertenko et al., 2000). A similar stabilizing function was proposed for Ase1 in yeast, a map65 homolog that is essential for the establishment and maintenance of bipolar organization of the spindle (Schuyler et al., 2003). In analogy to the function of ASE1, NtMAP65-1 and AtMAP65-3 could act as anchors for MT plus-ends at sites where they interdigitate, thereby maintaining phragmoplast bipolarity. The absence of AtMAP65-1-GFP, AtMAP65-5-GFP, and AtMAP65-8-GFP from the midline indicates that cross-linking of MT plus-ends is not a common feature of all members of the MAP65 family.

The localization of AtMAP65-8 was most remarkable, as it concentrated at the outermost side of the phragmoplast as well as at the spindle poles. Here, MT minus-ends congregate in multiple sites of MT-organizing centers from which MTs emanate. The phragmoplast MTs are also polarly organized and nucleate from sites near the surface of the nucleus that faces the equatorial plane. The minus-ends of MTs are anchored to the organizing centers so that new MTs polymerizing at the plus-end originate from a fixed position. In the cortical array, AtMAP65-8-GFP clearly stained along MTs, but in contrast to GFP-MBD and TUA6-GFP controls, the label was heterogeneously distributed. The MAP65-8-GFP protein also concentrated at relatively immobile foci in the cortical MT array, which may correlate with the peripherically organizing centers in interphase cells previously described based on γ-tubulin localization experiments (Figure 3d) (Canaday et al., 2000). We therefore propose that AtMAP65-8 is associated with MT-organizing centers both in interphase and in dividing cells.

Although localization at the phragmoplast in itself is no good proof, other evidence supports a role for the phragmoplast proteins in cytokinesis. Several of the phragmoplast-targeted proteins were transcriptionally activated at mitosis. For instance, the expression of Aurora and T22 peaks in synchronized BY-2 cell cultures 7 h after aphidicoline release, along with the expression of B-type cell cycle-dependent kinase (cluster 1; Breyne et al., 2002). The RNA levels of AtMAP65-3, AtEB1c, and Aurora significantly increased during mitosis according to microarray analysis of synchronized Arabidopsis cell cultures (Menges et al., 2003). We noticed that in several transgenic BY-2 lines aberrant cells carrying multiple nuclei occurred, indicative of erroneous cytokinesis events. Cytokinesis defects were most pronounced in cell lines expressing the cell cycle-controlled genes AtEB1c, T22, and Aurora. The Aurora BY-2 lines that most frequently produced binucleated cells gradually died and could not be maintained for extended periods because of accumulative defects in the division process probably due to the constitutive expression of the fusion protein. Despite constitutive expression driven by the 35S promoter, some of the candidate cytokinesis proteins were primarily targeted to the phragmoplast structure and not to other mitotic configurations. For example, AtMAP65-1-GFP and AtMAP65-3-GFP were abundantly present in the phragmoplast and much less in the spindle. The localization of AtMAP65-3 is in good agreement with the cytokinetic phenotype of the Arabidopsis root morphogenic mutant pleiade that carries a null mutation in the AtMAP65-3 gene (Müller et al., 2002). Pleiade roots develop expanded irregularly shaped cells containing multiple nuclei probably as a consequence of the formation of distorted phragmoplasts with a widened midline (Müller et al., 2004).

Fluorescence of candidate cytokinesis proteins was mostly confined to particular subregions of the phragmoplast MTs or the cell plate, in line with the functional divergence of the respective proteins. For example, the novel protein T22 preferentially concentrated at the rim of the expanding phragmoplast where Golgi-derived vesicles arrive to build the new callose-containing cell wall (Samuels et al., 1995). Myosin-GFP fluorescence, on the contrary, was mostly pronounced at the developing cell plate after removal of the phragmoplast MTs.

In conclusion, the GFP-fusion proteins identified in this study represent valuable markers for future investigations in plant cytokinesis and will be helpful in unraveling structure/function relationship of the phragmoplast in live plant cells.

Experimental procedures

Cell suspension cultures and transformation

Arabidopsis thaliana (L.) Heyhn. Landsberg erecta cells were grown in 4.43 g Murashige and Skoog (MS) medium (Duchefa, Haarlem, The Netherlands), 30 g sucrose, 500 μg NAA and 50 μg kinetin per liter on a rotating platform (150 rpm) at 27°C in the dark. Nicotiana tabacum L. Bright Yellow-2 (BY-2) cells were grown in 4.302 g MS, 0.2 g KH2PO4, 30 g sucrose, 0.02 mg 2,4-dichlorophenoxyacetic acid (auxin), 0.05 mg thiamine, 5 mg myo-inositol per liter, pH 5.8 under the same incubation conditions. Transformation procedure was as described (Geelen and Inzé, 2001). Multiple transgenic BY-2 calli were transferred to fresh medium and inspected for GFP fluorescence with an epifluorescence microscope (Axioskop; Zeiss, Heidelberg, Germany). Arabidopsis was transformed by the floral dip method (Clough and Bent, 1998). Primary seed transformants were selectively grown on 4.43 g l−1 MS, 6.5 g l−1 agar, pH 5.7 containing 75 μg ml−1 kanamycin.

Sequence identification and gene cloning

The Arabidopsis genome was interrogated by blast amino acid sequence searches ( Sequences were selected on an individual basis by considering overall similarity levels and the conservation of crucial amino acid positions. The putative start and stop codons of the identified genes were used for primer design (additional data online). For cloning, RNA was extracted from cells after 48, 60, and 72 h subculturing, following the method described by Leyman et al. (2000). Poly(A)+ RNA was purified using oligo dT(25)-coated Dynabeads (Dynal, Oslo, Norway). ORFs were cloned by a one-step RT-PCR reaction (Titan one tube RT-PCR kit; Roche Diagnostics, Brussels, Belgium) with high-fidelity DNA polymerase and 0.1 μg purified poly(A)+ as template in 50 μl. Primers contained attB1 and attB2 extensions for Gateway® (Invitrogen, Carlsbad, CA, USA) conversion and additional modifications for replacement of the stop codon by a tyrosine codon in frame with the enhanced fluorescent protein EGFP. PCR products were gel-purified (High pure PCR purification kit; Roche Diagnostics), and subcloned into pDONR207 via BP reaction cloning (Gateway®). Recombined plasmids (entry clones) from eight colonies were analyzed by restriction digest (AvaI) and sequence analysis. The inserts of the entry clones were transferred to the plant Gateway ‘destination vector’ pK7WGF2 (Karimi et al., 2002) to generate EGFP C-terminal fusions downstream of the strong constitutive 35S promoter. Plasmids were checked by EcoRV restriction digest and transferred to Agrobacterium tumefaciens strain LBA4404.


Approximately 10 transgenic BY-2 calli of 1 μm were inspected for fluorescence under a coverslip with an Axioskop (Zeiss) fluorescence microscope. In those cases where no EGFP fluorescence could be detected, instability of the fusion protein, lack of expression due to silencing, or counterselection were assumed (Joubès et al., 2003). GFP-positive calli and Arabidopsis seedlings were analyzed by confocal microscopy (Zeiss 100M, equipped with LSM510 software version 3.2). A 63X water corrected objective (numerical aperture of 1.2) was used to scan the samples. The images were captured with the LSM510 image acquisition software (Zeiss). Projections were obtained from approximately 30 serial optical sections, 0.5 μm apart to cover two-thirds of the cell depth. Images were exported as TIFF files and processed with Adobe Photoshop (version 7). Three-dimensional and time-lapse confocal microscopy were carried out on BY-2 cells fixed with poly-l-lysine to the bottom of an 8-well coverglass chamber (Lab-Tek, Naperville, IL, USA) containing 100 μl BY-2 culturing medium. Under these conditions, BY-2 cells carried on dividing and could be monitored overnight.

Supplementary material

The following material is available from

Table S1Arabidopsis genes similar to yeast genes that have been implicated in cytokinesis

Table S2Arabidopsis genes that encode potential cytoskeleton-associated proteins

Table S3Arabidopsis genes that are the putative orthologs of cell cycle-modulated genes previously discovered by cDNA-AFLP analysis in synchronized BY-2 cell cultures (Breyne et al., 2002)

Tables S1 and S3 include a list of gene name and proposed function with corresponding degree of similarity from the BLAST results, and Tables S1, S2, and S3 provide the MIPS code, cDNA length, PCR amplification result, Gateway cloning, and epifluorescence detection for the selected genes.

Table S4 List of yeast genes for which no homologous sequence was found in the Arabidopsis genome

Additional data online

A database of localization patterns linked with Tables S1, S2 and S3 can be found at


We thank Martine De Cock for help with the manuscript. D.V.D. and D.G. are postdoctoral and predoctoral fellows of the Fund for Scientific Research-Flanders respectively. F.-Y.B. is supported by the Centre National de la Recherche Scientifique, France.