QQT proteins colocalize with microtubules and are essential for early embryo development in Arabidopsis


  • Sylvie Lahmy,

    1. Laboratoire Génome et Développement des Plantes, UMRCNRS 5096, Université de Perpignan, Avenue Paul Alduy, 66860 Perpignan-cedex, France,
    Search for more papers by this author
  • Jocelyne Guilleminot,

    1. Laboratoire Génome et Développement des Plantes, UMRCNRS 5096, Université de Perpignan, Avenue Paul Alduy, 66860 Perpignan-cedex, France,
    Search for more papers by this author
  • Anne-Catherine Schmit,

    1. Institut de Biologie Moléculaire des Plantes du CNRS, 12 rue du général Zimmer, 67084 Strasbourg-cedex, France, and
    Search for more papers by this author
  • Georges Pelletier,

    1. Station de Génétique et Amélioration des Plantes, INRA, route de Saint-Cyr, 78032 Versailles-cedex, France
    Search for more papers by this author
  • Marie-Edith Chaboute,

    1. Institut de Biologie Moléculaire des Plantes du CNRS, 12 rue du général Zimmer, 67084 Strasbourg-cedex, France, and
    Search for more papers by this author
    • Martine Devic and Marie-Edith Chabouté are equal contributors to this work.

  • Martine Devic

    Corresponding author
    1. Laboratoire Génome et Développement des Plantes, UMRCNRS 5096, Université de Perpignan, Avenue Paul Alduy, 66860 Perpignan-cedex, France,
      (fax : +33 4 68 66 84 99; e-mail: devic@univ-perp.fr).
    Search for more papers by this author
    • Martine Devic and Marie-Edith Chabouté are equal contributors to this work.

(fax : +33 4 68 66 84 99; e-mail: devic@univ-perp.fr).


During Arabidopsis embryogenesis, the control of division between daughter cells is critical for pattern formation. Two embryo-defective (emb) mutant lines named quatre-quart (qqt) were characterized by forward and reverse genetics. The terminal arrest of qqt1 and qqt2 embryos was at the octant stage, just prior to the round of periclinal divisions that establishes the dermatogen stage . Homozygous embryos of a weaker allele of qqt1 were able to divide further, resulting in aberrant periclinal divisions. These phenotypic analyses support an essential role of the QQT proteins in the correct formation of the tangential divisions. That an important proportion of qqt1 embryos were arrested prior to the octant stage indicated a more general role in cell division. The analysis of QQT1 and QQT2 genes revealed that they belong to a small subgroup of the large family encoding ATP/GTP binding proteins, and are widely conserved among plants, vertebrates and Archaea. We showed that QQT1 and QQT2 proteins interact with each other in a yeast two-hybrid system, and that QQT1 and QQT2 tagged by distinct fluorescent probes colocalize with microtubules during mitosis, in agreement with their potential role in cell division and their mutant phenotype. We propose that QQT1 and QQT2 proteins participate in the organization of microtubules during cell division, and that this function is essential for the correct development of the early embryo.


Embryogenesis in the flowering plant Arabidopsis involves a regular pattern of cell divisions and cell shape changes that are highly coordinated. After perpendicular shifts during three successive rounds of cell division, the embryo of eight cells (octant) undergoes a periclinal division that results in a 16-cell embryo. This event contributes to the formation of the outer cell layer, the protoderm, precursor of the epidermis, and is considered as the first step of the radial pattern of the embryo (Mayer et al., 1991).

Asymmetric cell divisions contribute to the morphological diversity of organisms and are therefore important to normal development. As a result of the presence of their rigid walls, the location, shape and size of plant cells are determined by a controlled morphogenetic pathway, which leads to functional specificities. The cytoskeleton is widely involved in such control of plant cell growth and spatial regulation (Wasteneys and Yang, 2004). In non-plant cell systems, the connection between the determination of cell polarity and the cell division plane is dominated by cytoskeleton rearrangements (Wodarz, 2002). Two cytoskeletal arrays, the preprophase band (PPB) and the phragmoplast (PHR) have been identified that help to ensure the fidelity of cytokinesis in higher plants. The PPB is a transient ring of cortical microtubules and actin filaments that appears in late S phase, narrows throughout G2 and disappears during prophase when the nuclear envelope breaks down. The phragmoplast constitutes an array of microtubules and actin filaments present at the equator of a dividing cell during the anaphase and telophase transition (Schmit and Lambert, 1988). The first manifestation of the orientation of cell division is a specific positioning of the microtubular preprophase band, which marks the future division plane and the landmarks where the newly built cell plate will fuse with the mother plasma membrane.

Several mutations that affect cytokinesis in the Arabidopsis embryo have been isolated. The first characterized were KNOLLE and KEULE that belong to the syntaxin and Sec-1 families respectively (Assaad, 2001; Lukowitz et al., 1996). Interestingly, knolle and keule mutations were initially described as affecting the radial pattern of the embryo (Mayer et al., 1991). The fass/tonneau (Traas et al., 1995) mutant with defects in the PPB illustrates the importance of division plane specification . The cytokinesis mutants, with the exception of keule, are also impaired in endosperm cellularization (Sorensen et al., 2002). Another class of mutant that includes pilz and titan, encoding tubulin folding cofactors cohesin or condensin, is impaired in tubulin dimer biogenesis, which affects both the embryo and endosperm development (Nacry et al., 2000; Mayer and Jürgens, 2002; Tzafir et al., 2002).

In plants, organogenesis is closely correlated with a correct positioning of the protoderm. The homeobox AtML1 and the lipid transfer protein AtLTP1 represent early molecular markers for the establishment of apical–basal and radial patterns during plant embryogenesis (Lu et al., 1996; Vroemen et al., 1996). AtML1 is expressed in the apical cell after the first division of the zygote, and continues to be expressed in the outer but not in the inner cells, whereas AtLTP1 is expressed specifically in the protoderm soon after its formation. Recently AtDEK1 has been shown to be a key component of L1 specification for progression through embryogenesis, although its expression is not restricted to the epidermis at any stage in development (Johnson et al., 2005). Remarkably, no mutation has so far been identified in Arabidopsis that abolishes this first step of radial patterning.

In order to provide insight on the mechanisms involved in the radial pattern, we chose a genetic approach and isolated mutant lines bearing embryos that did not perform periclinal divisions leading to the formation of protoderm. Here, we present the characterization of two genes, which we name Quatre-QuarT 1 (QQT1) and QQT2. We show that the proteins encoded by these genes interact with each other and are colocalized with tubulin in dividing cells. Their localization with the preprophase band, spindle and phragmoplast suggests they function in the positioning of the division plane during mitosis, and that they may participate in the patterning of the early embryo at the octant–dermatogen transition.


quatre-quart embryo phenotype

The qqt1 mutant (EIC6 line, INRA Versailles collection; http://arabidopsis.info) was identified by screening T1 populations of T-DNA transformed Arabidopsis plants (Wassilewskija [Ws] ecotype) for an embryo-defective (emb) phenotype. The number of emb lines with embryos arrested at eight cells without initiation of protoderm formation was very low in the collection, and represented only 0.01% of the 16 000 plants screened. The qqt1/QQT1 plant showed a wild-type phenotype for plant size, leaf morphogenesis and flower development. However, the emb phenotype, illustrated by 1/4 white seeds for 3/4 green phenotypically wild-type seeds (126 white seeds, 389 green seeds, χ2 = 0.041), was observed in the siliques after selfing.

At the late heart-cotyledon stages of embryo development, abnormal seeds became readily distinguishable as white seeds compared with green seeds. When wild-type embryos were at the torpedo stage (Figure 1j), the phenotype of qqt1 embryos was determined (Table 1). The terminal arrest of qqt1 embryos was at the octant stage. Mutant embryos arrested at octant stage (Figure 1d,e) were different from wild-type octant embryos (Figure 1a,b). Histological comparison of cells of embryo and suspensor showed that mutant cells were slightly larger and vacuolated compared with wild-type cells, although their nucleus and nucleolus did not obviously differ in either size or organization (Figure 1k,l). In this mutant line, we did not observe embryos performing periclinal divisions leading to the formation of the protoderm. However, 1.7% of the mutant embryos (Table 1) were able to perform additional divisions leading to embryos with abnormal shapes. The size of the qqt1 seed at torpedo stage was larger than that of a wild-type seed at octant stage (Figure 1a,d), and the number of nucleoli greatly exceeded 40. This result showed that endosperm was not blocked after three successive mitoses, unlike cells of the embryo proper. The final size of qqt1 seeds was in agreement with the important role of the endosperm in the control of seed size (Berger, 2003). These observations suggest that endosperm and embryo development in qqt1 were affected in cell division in different ways. Furthermore, at the young torpedo stage, the majority of qqt1 embryos were composed of less than eight cells exhibiting a normal pattern (Table 1). We concluded that qqt1 embryos were also impaired in cell division during early embryo development in addition to their inhability to perform periclinal divisions.

Figure 1.

 Phenotypes of qqt1 and qqt2 mutant embryos.
Images of cleared seeds containing a wild-type embryo at octant stage (a) and at dermatogen stage (c), a qqt1 embryo (d), a qqt2 embryo (h), emb1705-1 embryos (f, g). Panels (b), (e) and (i) are close-up views respectively of (a), (d) and (h). Panel (j) corresponds to the developmental stage of a wild-type embryo in the siliques for (d), (f), (g) and (h). Panel (c) represents a wild-type 16-cell dermatogen embryo. Longitudinal sections of wild-type octant embryo (k), qqt1 embryo (l) and heart wild-type embryo (m). Scale bars in (a), (d), (h) and (j) represent 50 μm; scale bars in (b), (c), (e), (f), (g), (i), (k), (l) and (m) represent 20 μm.
E, embryo; en, endosperm; S, suspensor. Asterisks indicate cell after periclinal division.

Table 1.   Distribution of embryo phenotypes of alleles of At5g22370 (qqt1) and At4g21800 (qqt2)
Homozygous mutant seedsNumber of cells in the embryo proper
<4 cells (%) 4 cells (%)4< n <8 cells abnormal divisions (%) 8 cells (%)8< <16 cells abnormal divisions (%)16 cells abnormal divisions (%)n > 16 cells
  1. *Numbers indicate the percentage of mutant embryos in each class when wild-type seeds are at torpedo stage.

qqt1 (n = 358)8.4049.701723.201.7000
emb1705-1 (n = 187)0.502.105.402132.10380
emb1705-2 (n = 200)24.5039.502015.500.5000
qqt2 (n = 211)1.4012.3011.4070.104.7000

QQT1 gene isolation

Cosegregation of qqt1 phenotype and T-DNA insertion was verified on a kanamycin-resistant population of 234 plants. Southern blot analysis of the mutant genomic DNA using probes corresponding to the T-DNA revealed a single insertion. The plant genomic sequences flanking the left and right borders of the T-DNA were identified by PCR walking (Devic et al., 1997). The T-DNA has integrated in the second intron of At5g22370 (Figure 2a), 90 nucleotides upstream from the beginning of the exon and resulted in a deletion of 24 bp without further rearrangements. The QQT1 gene encodes a protein of 298 amino acids of unknown function. Two insertion lines were found in the Seedgenes collection (http://www.seedgenes.org) in which the T-DNA was inserted in the same intron (Figure 2a). Mutant embryos of the SALK_059852 (emb1705-2) line revealed that the terminal stage of arrest was octant stage, like qqt1 embryos, and that most embryos were impaired throughout early embryogenesis (Table 1). Crosses demonstrated that qqt1 and SALK_059852 were allelic (136 wild-type seeds, 44 emb seeds, χ2 = 0.01). Analysis of embryos of a third insertion line, emb1705-1, highlighted less severe defects in embryo development (Table 1). A few mutant embryos were arrested at early stages (8%) and at the octant stage (21%), but most embryos were able to divide further (70.1%). The eight embryonic cells achieved periclinal but incorrect divisions as compared with wild type (Figure 1c,f,g). The 8–16 cells of the embryo proper rapidly degenerated, as suggested by the Nomarski images of seeds. On the contrary, suspensor cells either continued to divide and form a line of between seven and nine cells similar to mature wild-type suspensors (Figure 1f), or cells proliferated abnormally and resembled suspensors of some globular arrested mutants (Figure 1g; Yeung and Meinke, 1993). In addition, emb1705-1 seeds at mid-embryogenesis were yellow–green because of proplastids differentiation in the endosperm occurring simultaneously with the phenotypically wild-type seed. We concluded that emb1705-1 represents a weaker allele of QQT1. In this less severe allele, the mutation again affected the embryo proper and the endosperm in distinct ways.

Figure 2.

QQT1 and QQT2 gene structure and alignment of the deduced amino acid sequences.
(a) Schematic representation of QQT1 and QQT2 genes. Solid grey rectangles are exons and connecting lines are introns. Numbers refer to nucleotide positions on the chromosomes. Black diamonds correspond to insertion of T-DNA in Versailles line and grey diamonds in other T-DNA insertion lines.
(b) Alignment of deduced amino acid sequences of QQT1, QQT2 and At4g12790. The solid line overlies the ATP/GTP binding site, the dotted line indicates the predicted leucine repeat motif of QQT1. Identical amino acids are boxed, similar amino acids are shaded in grey. QQT1 and At4g12790: 36% identity, 59% similarity; QQT1 and QQT2: 23% identity, 50% similarity.

In order to confirm that the qqt1 phenotype resulted in insertion of the T-DNA at the At5g22370 locus, a complementation experiment was performed by transforming qqt1/QQT1 heterozygous plants with a construct containing QQT1 cDNA under the control of 1 kb of its own promoter region (ProQQT1:QQT1). Four independent transformants were obtained and T2 plants were analysed after selection on antibiotics. We found that 100% of the seeds in the siliques showed a wild-type phenotype (100 < n < 150 per T2 plant), in agreement with effective complementation. Together these experiments demonstrated that disruption of At5g22370 by a T-DNA insertion was responsible for the qqt1 phenotype.

Identification of the QQT2 gene

Database searches revealed that QQT1 was highly conserved in higher plants. In addition, QQT1 protein was similar over its entire length to vertebrate proteins of unknown function, showing 72% similarity (55% identity) to human (AAH08634) and mouse (Q8VEJ1) homologues. The QQT1 protein presented 70% similarity (50% identity) with the YOR262W (Q08726) protein of baker’s yeast (Saccharomyces cerevisiae). Null mutation of YOR262W also led to a lethal phenotype in yeast (Winzeler et al., 1999). Interestingly, YOR262W has been shown to interact with YJR072C in a double hybrid assay (Uetz et al., 2000). We identified At4g21800 as a putative homologue of YJR072C (75% similarity, 57% identity) in Arabidopsis. We used reverse genetics to study its function and searched for T-DNA insertion lines in the different collections. We identified an insertion line (SALK_026078) at the Salk Institute Genomic Analysis Laboratory site (http://signal.salk.edu/). A T-DNA has inserted in the sixth intron of At4g21800 (Figure 2a). Plants showed an emb phenotype (73 white seeds, 303 green seeds, χ2 = 2.34) similar to that of qqt1 embryos (Figure 1d,e,h,i). Therefore we named this gene QQT2. The main difference of the qqt2 phenotype compared with that of qqt1 was that fewer embryos were arrested before octant stage (25%), and that the majority reached the terminal octant stage (70.1%, Table 1). ProQQT2:QQT2 was used to complement the qqt2 phenotype. Six independent transformants obtained after selection on antibiotics were analyzed and presented 99–100% of phenotypically wild-type seeds (100 < n < 200). These results confirmed that the qqt2 phenotype resulted from the insertion of T-DNA in At4g21800. In addition, Nomarski analysis of 100 white seeds obtained from crosses between qqt1 and qqt2 showed that embryos were arrested before four cells (26.3%), at four cells (50%), between four and eight cells (13.2%) and at eight cells (10.5%), indicating that the qqt1qqt2 phenotype was not more severe than the phenotype of the qqt1 single mutant.

Taken together, these results support the idea that these two genes influence the same aspects of the process of cell division during early embryo development. To test whether these mutations could also have a gametophytic effect, reciprocal crosses of qqt1 and qqt2 alleles were performed. Such crosses confirmed that these mutations produced mainly sporophytic effects (Table 2).

Table 2.   Reciprocal crosses of qqt mutants
Number of seedsQQT1/qqt1 selfingQQT1/qqt1 × WsWs × QQT1/qqt1QQT2/qqt2 selfingQQT2/qqt2 × WsWs × QQT2/qqt2
Seeds examined515220143376272143
Aborted seeds126007300

QQT1 and QQT2 proteins belong to the same family and are conserved among species

QQT1 contains an ATP/GTP binding site motif A (P-loop) spanning from G-9 to T-16 (Figure 2b). Leucine-rich repeats were predicted from L-74 to L-102. When the deduced amino acid sequence of QQT1 was aligned against protein databases, only two sequences matched in the Arabidopsis genome. The first was At4g12790 and the second At4g21800 (QQT2). The Pfam 03029 motif including the ATP/GTP binding site was conserved in the three members (Figure 2b). QQT2 differed from the two other members by the presence of terminal extensions at the N- and C-termini, and one insertion (from N-47 to I-68). The three proteins constitute a small subgroup in the large family of conserved hypothetical ATP-binding proteins. A phylogenic tree (Figure 3) built using the darwin program (Gonnet et al., 2000) indicated that the genomes of rice and other plants, of yeast, human and of many other eukaryotes, but not of prokaryotes, possess three genes, putative orthologues of QQT1, QQT2 and At4g12790. The three proteins and their orthologues were similar over their entire length.The origin of these genes is ancient as they are present in Archae, but as a single copy. No function has been assigned to the proteins encoded by QQT1 and At4g12790 and their respective orthologues in other species. YLR072C, the yeast QQT2 orthologue, has been shown to be present in the pre-60S ribosomal particles and has been renamed Npa3 for Nucleolar preribosomal-associated protein (Dez et al., 2004). However, as the yeast Npa3p is localized in the cytoplasm, there is no evidence for a direct role in ribosome biogenesis.

Figure 3.

  Phylogenic tree of protein homologues to QQT1, QQT2 and At4g12790.1
Full-length protein sequences were used for alignments. Branch lengths are proportional to calculated sequence divergence. Accession numbers are indicated for proteins. Hs, Homo sapiens; Os, Oryza sativa; Sc, Saccharomyces cereviseae.

To assess the function of the third member of the family, we analysed the SAIL449A09 insertion line where the T-DNA is inserted in the eigth exon. We did not observe a qqt phenotype, probably indicating that either At4g12790 did not participate in the function of At5g22370 and At4g21800 at octant–dermatogen transition or was not essential.

QQT1 interacts with QQT2 in a yeast two-hybrid assay

As an interaction between YOR262W and YJR072C has been shown in yeast (Uetz et al., 2000), a two-hybrid assay was performed to test the existence of an interaction between the two Arabidopsis proteins. A full-length QQT2 was used as bait because, as for the yeast protein, QQT1 could activate the yeast reporter gene on its own when fused to the GAL4-BD domain (data not shown). When yeast cells were transformed with empty vectors and either AD-QQT1 or BD-QQT2, limited growth was observed on the selective medium (Figure 4). However, cell growth was significantly enhanced when the two proteins were present in the same cell demonstrating a clear interaction between QQT2 and QQT1 (Figure 4).

Figure 4.

 QQT1 interacts with QQT2 in the yeast two-hybrid system.
The coding sequences were cloned into either the GAL4 activation domain (AD) or the GAL4 DNA binding domain (BD) vectors. The left panel represents growth on minimal medium containing histidine (+His) such that interaction is not required for growth (control plate); the right panel is a histidine-free (–His) minimal medium with 20 mm 3-aminotriazole (+3AT), such that interaction is required for cell growth (test plate). Two independent transformed cells were tested.

Developmental expression of QQT1 and QQT2 in planta

To analyse in detail the expression of QQT genes during the octant–dermatogen transition, we performed in situ hybridization with QQT1 and QQT2 probes on seeds at different stages of development. A uniform signal was detected throughout the embryo at octant stage with the QQT1 probe (Figure 5a). After dermatogen stage, the expression became stronger in the inner cells of the embryo from globular to torpedo stage. These inner cells will define the provascular tissues (Figure 5b–e). No staining was observed at later stages in embryogenesis. With the QQT2 probe, a uniform staining was observed throughout the embryo from octant to heart stage. No staining above background was detected in later stages of embryogenesis (Figure 5i). Generally, the intensity of the signal for QQT2 was lower than for QQT1 (Figure 5b-d,g-h).

Figure 5.

 Expression of QQT1 and QQT2 during embryogenesis.
(a)–(e) In situ hybridizations using QQT1 antisense probe. Expression was detected as dark coloration through octant (a), globular (b), heart (c) and torpedo (d, e) stages in the embryo proper and suspensor. The expression was strongest in the provascular regions (b)–(e). (f) Sense probe. E, embryo; en, endosperm; S, suspensor.
(g)–(i) In situ hybridizations using QQT2 antisense probe. Expression was uniform from dermatogen (g) to late heart (h), and undetectable at cotyledon stage (i). (j) Sense probe. Scale bars: 20 μm.

To determine the expression of QQT1 and QQT2 during vegetative growth, we used a GUS reporter gene assay. The promoter fragment of 1036 bp used for the complementation of QQT1/qqt1 plants, and which thus probably contained the sequences necessary to direct a correct gene expression, was used to drive the expression of GUS. The construction was made by insertion of the uidA Open Reading Frame (ORF) at the initiation codon of QQT1 in the second exon. Therefore, the promoter region of QQT1 also included the putative regulatory elements located in the first exon and first intron in the 5′untranslated region (UTR) (Figure 2a). QQT1 promoter activity was analysed in seedlings and adult plants of five independent transgenic lines. GUS activity was detected in all the vascular tissues of the plant (Figure 6a–e). In root, the staining was also enhanced in root tips (Figure 6a). Root sections showed that staining was strong in the stele and pericycle, and in the emerging secondary root primordium (Figure 6b). Other primordia, like floral primordia, were also stained (Figure 6e). QQT1 was highly expressed in apical and root meristematic regions (Figure 6a,d). No activity was detected either in pollen grains or ovules (data not shown). Similarly, the 0.8-kb promoter region of QQT2 was joined to the GUS reporter gene. As the QQT2 gene organization was similar to that of QQT1, the first exon and intron of the 5′UTR were included in the construction (Figure 2a). Seven independent transformants were examined. In most plant organs, e.g. root, leaf and flower, GUS expression was restricted to individual cells (Figure 6f–i). Intriguingly, frequently these stained cells were close to the vasculature (Figure 6i), although the vascular bundles were not stained. This expression pattern was reminiscent of the patchy staining pattern reported for the promoters of cell cycle proteins (Barroco et al., 2005).

Figure 6.

 ProQQT1:GUS and ProQQT2:GUS expression in plants.
roQQT1:GUS – (a) roots, (b) root section at secondary root emergence, (c) vascular tissue in rosette leaf, (d) shoot apical meristem in a 10-day-old seedling, (e) floral primordia.
ProQQT2:GUS – (f) meristematic zone of primary root, (g) primary root, (h) shoot apical meristem in 10-day-old seedling, (i) rosette leaf. Black arrows indicate isolated stained cells.
Scale bars in (a) and (c)–(e) represent 100 μm; scale bars in (b), (f)–(i) represent 20 μm. fp, floral primordia; lv, leaf vein; sam, shoot apical meristem; st, stele.

Intracellular localization of QQT1 and QQT2

To obtain further information on their function and to establish their intracellular localization, QQT1 and QQT2 fused with fluorescent proteins were expressed in Nicotiana tabacum cv. Bright-Yellow-2 (BY-2 tobacco) cells. At interphase, the fluorescence was mainly present in the cytoplasm with a weak signal in the nucleus (Figure 7a,b). During cell division, QQT1 protein fusion appeared as filaments staining the PPB, the spindle and the PHR during mitosis (Figure 7d–f). In the same manner, QQT2-GFP labelled the spindle (Figure 7g). As a control, BY-2 cells transformed with a vector containing GFP alone did not present such a staining network (Figure 7c,h). When BY-2 cells were cotransformed with QQT1-GFP and QQT2-RFP fusions, the signal of QQT2-RFP fusion was much lower compared with QQT2-GFP (Figure 7g,j), but was still higher than the cell autofluorescence level (verified in non-transformed BY-2 cells). Both proteins were colocalized on spindle microtubules (Figure 6i–k), whereas colocalization observed at interphase did not reveal labelling at the cortical microtubule array (Figure 7l–n). The common localization between QQT1 and QQT2 supported the results obtained in the two-hybrid assay. To confirm that localization of QQT1 and QQT2 corresponded to cytoskeleton filaments, we transformed BY-2 cells with a GFP-tubulin fusion (Kumagai et al., 2001) and either QQT1 or QQT2 fused with red fluorescent protein (RFP). Colocalization was observed between either QQT1 or QQT2 and tubulin on spindle and phragmoplast (Figure 8a–l). Furthermore, treatment of BY-2 cells with oryzalin resulted in a loss of organization of the fluorescent labelling pattern in cells expressing QQT1-GFP (Figure 8m–o) and GFP-tubulin (Figure 8p–r). These results showed that QQT1 and QQT2 were associated with microtubules during cell division.

Figure 7.

 Localization of QQT1 and QQT2 in tobacco BY-2 cells.
(a) QQT1-GFP, (b) QQT2-GFP and (c) GFP alone in interphase. (d) QQT1-GFP labelled the preprophase band (PPB), (e) the spindle (SP) and (f) the phragmoplast (PHR). (g) QQT2-GFP labels the SP. (h) GFP alone during mitosis.
(i)–(k) QQT1-GFP and QQT2-RFP present a common localization in BY-2 cells during mitosis, probably on the tubulin network, and (l)–(n) in the cytoplasm at the level of cortical tubulin arrays. QQT-GFP (green channel), QQT2-RFP (red channel), (k), (n) are merged images where the yellow colouration corresponds to QQT1-GFP and QQT2-RFP colocalization. Scale bars: 10 μm.

Figure 8.

 QQT1 and QQT2 colocalized with GFP-tubulin in tobacco BY-2 cells.
(a)–(c) Cells in metaphase; (d)–(f) cells in telophase. GFP-tubulin (green channel), QQT1-RFP (red channel). (c) and (f) are merged images.
(g)–(i) Cells in metaphase; (j)–(l) cells in telophase. GFP-tubulin (green channel), QQT2-RFP (red channel). (i) and (l) are merged images. PHR, phragmoplast; SP, spindle.
BY-2 cell treated with 10 μm oryzalin (m–o); QQT1-GFP, tubulin-GFP (p–r). Images of the same cells were taken at 0, 5 and 10 min after addition of the drug.
Scale bars: 10 μm.


In an attempt to identify the factors involved in protoderm formation, we used forward and reverse genetics to isolate mutant embryos with arrest at the octant stage as their terminal phenotype. Acquisition of protoderm identity is a key element in the body organization of the embryo as radial patterning will generate tissue-specific cell fates in a position-dependent manner (Mayer and Jürgens, 1998). Therefore, qqt mutants, the final arrest of which was mainly at the octant stage, constitute an attractive means to discover gene function involved in cell division during early patterning of the embryo.

QQT1 and QQT2 are essential for cell division during embryo development

QQT1 and QQT2 are well-conserved genes across phyla, and homologues of the QQT1 and QQT2 proteins are found in many species. According to the phylogenetic tree (Figure 3), these two proteins may result from a duplication of the single gene in Archae. The third member of this gene family, At4g12790, originate from the duplication of QQT1 (At5g22370) during the duplication of the Arabidopsis genome. Even though their functions remain unknown, they also have an essential role in the development of yeast and Caenorhabditis elegans as mutations in these genes give rise to lethal phenotypes (Walhout et al., 2002; Winzeler et al., 1999). As discussed by Tzafrir et al. (2004), EMB genes are more likely to have counterparts among essential genes of yeast rather than other Arabidopsis genes, indicating a common function for cell maintenance. This is certainly the case for QQT1 (emb1705; Tzafrir et al., 2004 and this study) and for QQT2 (this study). Furthermore, qqt1, the newly characterized allele of emb1705, reconciles the discrepancies observed in the phenotypes of emb1705-1 and emb1705-2. The mutants qqt1 and emb1705-2 exhibited an embryo phenotype similar to each other, but also to that of the loss of function of their interacting protein partner QQT2. We deduced from these data that the failure of qqt1 and emb1705-2 embryos to attain and to complete the octant–dermatogen transition most probably resulted from a total loss of QQT1 function, and hence that emb1705-1 was a weak allele. The site of T-DNA integration cannot easily explain this difference in phenotype severity because all three alleles have a T-DNA located in the second exon (although the sequence of the integrative molecule was different in each case). It is conceivable that on rare occasions in emb1705-1 seeds, the T-DNA can be spliced together with the intron allowing cells to produce a small quantity of QQT1. The availability of this weak allele allows us to test our hypothesis that QQT1 could be required for performing periclinal divisions. In emb1705-1, tangential divisions were observed but not in the correct position, resulting in an unequal partitioning of the cell content, whereas suspensor cells were not affected. The suspensor was still able to divide and sometimes to proliferate. The pattern of QQT1 during early embryogenesis is also suggestive of a participation in the determinism of cell fate after the asymmetric divisions. Periclinal divisions will define an outer and an inner cell layer leading to different cell fates. In contrast to the AtML1 gene, the expression of which was restricted to the outer cells (Lu et al., 1996), QQT1 mRNA was preferentially found in the inner cells after the dermatogen stage, as is the case for MONOPTEROS (Hardtke and Berleth, 1998). During vegetative growth, QQT1 was persistently expressed in the vascular tissues that were defined in the embryo. Loss of QQT2 function seems to be more drastic in the initiation of periclinal divisions than QQT1, as most embryos reached octant stage and only 2% could perform additional, but not periclinal, divisions. Even though the phenotype of qqt1 and qqt2 indicated either a specific or, at least, a greater requirement for these proteins for octant–dermatogen transition, we also found a significant proportion of embryos arrested before the eight-cell stage, indicating there is also a requirement for normal divisions and a role in early patterning of the embryo proper.

QQT1 and QQT2 are linked to mitotic cytoskeleton

The intracellular localization of QQT1 and QQT2 reinforced the idea of their role in cell division. We observed that QQT1- and QQT2-GFP proteins labelled components of the mitotic microtubule arrays: the PPB, the spindle and the phragmoplast. However, the proteins were not detected at interphase on the cortical microtubule arrays, suggesting a possible role in the control of cell division. Whether the association with microtubules was direct or not remains to be determined.

Cell division is an essential process in embryogenesis, and is highly regulated spatially and temporally. The cytokinesis mutants identify components of the machinery for generating a cell plate that physically separates the forming daughter cells. The prototype gene is KNOLLE and encodes a cytokinesis-specific syntaxin required for vesicle fusion at the plane of division (Jürgens, 2005). It shares the same altered phenotype (enlarged cells and nuclei, incomplete cell walls) with KEULE, encoding a SEC1 homologue also required for vesicle fusions and which interacts with the KNOLLE syntaxin (Assaad, 2001). The terminal step of cell division is cytokinesis but other cellular processes occur prior to cytokinesis, such as the changes in microtubule organization through the cell cycle. The pilz and titan mutants (Mayer and Jürgens, 2002; Tzafrir et al., 2002) illustrate the importance of microtubule biogenesis in cell division. These genes encode tubulin cofactors, cohesin and condensin. The mutants are characterized by embryo lethality and a non-cellularized endosperm except for kiesel (Steinborn et al., 2002). Clearly, qqt seeds did not exhibit any of the cytokinesis or the pilz/titan mutant characteristics either in the embryo cells nor in the endosperm nuclei, including incomplete cell walls formation and large nuclei and nucleoli (Figure 1). Therefore, we deduced from these observations that the first role of QQT proteins in cell division occurred probably before chromosome segregation and cell plate formation.

Microtubule-associated proteins (MAPs; Sedbrook, 2004) and motor proteins like kinesins (Reddy and Day, 2001), also play crucial roles in microtubule organization. The first either stabilize or destabilize microtubules, the second possess a motor domain to either convert energy of ATP into movements along microtubules or to control the (de)polymerization of microtubules (Ambrose et al., 2005). Interestingly, Muller et al. (2006) have described two kinesins interacting with TAN1 in maize that participate in the spatial control of cytokinesis. Among several genes involved in cell division (Lee and Liu, 2004; Mayer and Jürgens, 2002), mutants of MAP families illustrate the importance of the microtubule machinery in cell division: mutations in MOR1 (Kawamura et al., 2006), a member of the MAP215 family, affected spindles and phragmoplasts and in some cases prevent completion of cell division in Arabidopsis roots. The phenotype of MAP mutants thus resembles that of cytokinesis mutants with either enlarged multinucleated cells and incomplete cell walls or displaying aberrant planes of cell division and organ growth. Again, most of these features have not been observed in qqt mutants, with the exception of the abnormal planes of divisions of the cell of the embryo proper seen in the emb1705-1 allele of QQT1. The cell and nucleus structures in qqt embryos are more similar to cells either arrested in G1 or at the start of the M phase (Ronceret et al., 2005, M. Devic and A. Ronceret, unpublished data) rather than to cells that have initiated and aborted division. Taken together, these phenotypic comparisons point to a role of the QQT proteins in the early steps of cell division.

Most plant organs expressed QQT1 and QQT2, but strongest expressions were observed in dividing zones such as shoot apex (Figure 6). Data available on calli, cell suspensions and shoot apex at Genevestigator (http://www.genevestigator.ethz.ch/; Zimmermann et al., 2004) confirmed that expression was higher in dividing tissues. These patterns of expression are consistent with a role in cell division. In addition, GUS expression of QQT1 in all the vascular tissues of the plant, from procambial regions of the embryo to the differentiated organs, was reminiscent of the pattern of expression reported for AtMAP65-1 (Smertenko et al., 2004). The presence of a mitosis-activation motif (MSA; 998-bp upstream from the initiation codon) in the promoter of QQT1 could activate gene expression at the start of mitosis. The expression pattern of QQT2, although outside of the dividing zones, differed from QQT1 but was similar to the expression profile pattern of genes involved in cell division. QQT2 expression was restricted to a few cells, similarly to the patchy pattern of cell cycle genes, such as cyclin B and D (Barroco et al., 2005). The similitudes and divergences in the expression pattern between QQT1 and QQT2 suggest that a transient interaction between the proteins may be required at mitosis, but that the proteins may also act independently. Additional insight is provided from the study of the yeast QQT orthologues. The cytoplasmic function of YLR072C, orthologue of QQT2, renamed Npa3 because of its presence in the pre-60S particles, is probably involved in ribosomes biogenesis and cell growth (Dez et al., 2004). The interaction of YJR072C with Pcl1, a G1-cyclin, emphasizes a role in cell division control (Kenery et al., 2004). Pcl1 participates in the regulation of PHO85, a cylin-dependent kinase (CDK), and is involved in entry into the mitotic cell cycle or START, and regulation of morphogenesis (Moffat and Andrews, 2004). At the protein level, both QQT1 and QQT2 belonged to the ATP-binding protein family, and QQT1 was predicted as a P-loop ATPase. As some P-loop ATPases are necessary in C. elegans for meiotic spindle formation (Lupas and Martin, 2002), QQT1 and QQT2 may act as ATP/GTP providers to proteins associated with microtubules, which require energy for functioning. Taken together, the results in yeast and C. elegans further consolidate our hypothesis of a role of the QQT proteins in the early events of cell division.

Are QQT1 and QQT2 participating in periclinal divisions?

The importance of the PPB in defining the site at which a new cell wall will link up with the parental wall is well established (Otegui and Staehelin, 2000). The phenotype of qqt1 and particularly qqt2 embryos was an arrest at octant stage and a failure either to initiate or to perform correct periclinal divisions. The octant–dermatogen transition represents the first periclinal division occurring in the homozygous state. QQT1 and QQT2 localization in the PPB and the qqt embryo phenotypes suggest that they could be involved in the proper execution of this process. Furthermore, the spindle polarity is controlled by CDK activities in budding yeast (Huisman and Segal, 2005). Either an arrest before periclinal division or incorrect divisions after the octant stage could be explained by the inability of the cells to deliver the correct signal indicating a change in cell polarity, positioning the plane division as periclinal and at the appropriate site. The analysis of the phenotype of the less severe allele of qqt1 supports this hypothesis. The emb1705-1 embryos were able to perform periclinal divisions, although these were abnormal and probably unequal, as no further divisions were observed in the embryo proper whereas suspensor cells continued to divide. Given that suspensor cells were still able to divide and to proliferate, we argue that only cells of the embryo were severely impaired in division. It remains to be determined if the occurrence of these improper tangential divisions results in inappropriate cell fates in the embryo proper, in order to correlate the failure of cell division to defect in embryonic patterning.

Experimental procedures

Plant material and growth conditions

Arabidopsis plants were grown either in soil or in Petri dishes under constant illumination at 22°C. EIC6 mutant line was obtained from INRA Versailles. Mutant lines from the Salk collection (Alonso et al., 2003) and the Seedgenes collection were obtained from Nottingham Arabidopsis Stock Centre (http://arabidopsis.info/). Seedlings were selected on Murashige and Skoog medium containing kanamycin at 100 mg l−1 and/or hygromycin at 30 mg l−1 and transferred to soil after 2 weeks.

Study of embryo development

Siliques from heterozygous plants at different stages of maturity were opened under the binocular to remove the seeds. Seeds were treated for either clearing or histological section as described in Ronceret et al. (2005).

Isolation of the qqt1 mutant and identification of the QQT1 gene

Plant genomic sequences adjacent to the T-DNA insertion in EIC4 were isolated by the PCR walking method (Devic et al., 1997). To clone the full-length QQT1 cDNA we first designed two specific primers, QQT1fw (5′-aagcatggtgtttggacaagtagtaatag) and QQT1rv (5′-ccttcagaagaggctttagtttag), to amplify a cDNA fragment. 5′ and 3′ untranslated sequences of QQT1 were obtained by RACE PCR with a 5′/3′ RACE kit (First choice RLM-Race kit; Ambion, http://www.ambion.com/) using a cDNA library from flower buds built according to manufacturer’s instructions. For 5′UTR, nested primers were N1 (5′-ccattaggaccaagcgagtg) and N2 (5′-caccacactcataaggtaatgc) and for 3′UTR, E1 (5′-cttgtcatacttggagcacc) as detailed in the protocol.

Complementation of the qqt1 and qqt2 mutations

A promoter fragment of 1036 bp of the QQT1 gene, defined as ProQQT1 was amplified from genomic DNA (Col0) with primer (5′-actcgaggatccgtcgcaccacggcgc-3′) including XhoI and primer (5′-accatggttctgaagacgaagaag-3′) including NcoI. The dual 35S promoter was removed from pRTL2-GUS (a gift fom J. Carrington; Oregon State University, Corrallis, USA) and replaced by the Xho1-Nco1 promoter sequence to create ProQQT1:GUS.

The QQT1 cDNA was amplified by PCR from flower buds first-strand cDNA with primer (5′-ccatggtgtttggacaagtag-3′) containing NcoI and primer (5′-ctctagaggttcatacaaaccgtc-3′) containing Xba1. This fragment was cloned into ProQQT1:GUS after removing the Nco1-Xba1 GUS fragment. The ProQQT1:QQT1 cassette was inserted into pBIB-HYG, a binary vector carrying hygromycin resistance (Becker et al., 1992). Heterozygous plants were transformed by the floral-dip protocol (Clough and Bent, 1998). Transgenic plants were selected on hygromycin and kanamycin. The T2 progeny of the four ProQQT1-QQT1 individual primary transformants were genotyped by PCR with QQT1fw and QQT1rv primers. In addition, four siliques per plant were opened and seeds counted. For QQT2, a 3-kb genomic sequence containing ProQQT2 (728 bp), the QQT2 coding region and 500-bp downstream from the stop codon was amplified using 3B1-qqt2 (5′-cacaaatcagccctaactctg-3′) and Gatrevqqt2 (5′-atttgtgtatccccgaatgacg-3′) fused respectively with attB1 and attB2. The PCR products were recombined into pDONR221 vector (Invitrogen, http://www.invitrogen.com/). The fragment was sequenced and shuttled into pHGWFS7 (Karimi et al., 2002). Heterozygous SALK_026078 plants were used for transformation. Analysis of the progeny was performed as described for QQT1.

Yeast two-hybrid assays

Full-length QQT1 cDNA was amplified by PCR using primer B1qqt1 (5′-taatggtgtttggacaagtagtaatagg-3′) and B2qqt1 (5′-tgtcttgtatttcctcatcttcc-3′) and transferred into pDONR207 (Invitrogen). After sequencing, the fragment was recombined into pDEST 22 (Invitrogen) fused with the activation domain of GAL4. Full-length QQT2 cDNA was recombined into pDEST32 (Invitrogen) in fusion with the GAL4 DNA binding domain. Plasmids were cotransformed into yeast strain MAV203 (Invitrogen) using the lithium acetate procedure, and plated onto yeast minimum medium lacking either Leu and Trp, or Leu, Trp and His. Either 10 or 20 mm of 3-amino-1,2,4-Triazole (3-AT) was added in the selective medium.

Histological detection of GUS expression

The ProQQT1:GUS cassette was introduced into the Sal1-Xba1 sites of pBIB-HYG.

For analysis of ProQQT2:GUS, a fragment of 739 bp of the promoter sequence was amplified with primers 3B1-qqt2 (5′-aatgatttgtatatacttaagcttttcc-3′) and B2-proqqt2 (5′-cagtggcgaaatttgcttttttc-3′). This insert was transferred after sequencing to pKGWFS7 to get GUS fusion. Tissues from T2 ProQQT1:GUS and ProQQT2:GUS plants were incubated in 100 mm sodium phosphate buffer (pH 7.5), containing 1 mm 5-bromo-4-chloro-3-indolyl-β-d-glucuronide (Duchefa, http://www.duchefa.com/), 0.5 mm potassium ferricyanide, 0.5 mm potassium ferrocyanide and 0.1% Triton X-100 for 36 h at 37°C followed by washes in 70% ethanol. Several independent lines were analyzed (five for QQT1, seven for QQT2). Roots of ProQQT1:GUS transgenic plants were dehydrated, embedded in Technovit 7100 resin (Kulzer, http://www.kulzer.com ) after GUS staining. Sections of 10 μm were observed.

In situ hybridization

Seeds at different developmental stages were fixed and embedded in paraffin. Sections of 7 μm were hybridized according to a previously described method (Albert et al., 1997). A cDNA QQT1 fragment of 400 bp was amplified using primers QQT1fw and QQT1rv and introduced into the pGEM-T vector (Promega, http://www.promega.com). A QQT2 fragment of 418 bp was amplified with BPQQTse (5′-tatgtctctaccatttggtgcc-3′) and BPQQTre (5′-gcaaacttgtgatctgcgac-3′). Images were taken with a Zeiss Axioskop2 (Zeiss, http://www.zeiss.fr) using a camera Leica DC350FX (Leica, http://www.leica.fr).

Protein fusion and expression in BY-2 tobacco cells

Because the size of QQT1 is smaller than the nuclear exclusion limit of GFP, the QQT1 sequence was fused to GUS-GFP downstream from the CaMV35S promoter at the Nco1 site into the pGG vector provided by F. Barnèche, Cell Signalling Laboratory, Naples, Italy (personal gift). QQT2-GFP fusion was obtained by amplifying cDNAQQT2 using primers B1BP-fw (5′-tgatggatcctatggagtcgtctagtg-3′) and B2BP-rev (5′-taggtagtaatgcttcgtctcg-3′), recombined in pDONR221 (Invitrogen) and transferred to PK7FWG2. QQT1-RFP and QQT2-RFP were obtained by recombination into pH7RWG2 (Karimi et al., 2002). A binary vector containing a fusion construct of GFP and tubulin (sGFP-TUA) was kindly provided by Prof. Hasezawa (University of Tokyo, Japan). Cell transfornmation was performed as described in Chaboute et al. (2000). Three independant transformations were performed in order to analyse the intercellular localization of QQT proteins as protein-GFP expression decreased rafter a few weeks in the cell lines. Oryzalin was added to BY-2 cells just before kinetics were recorded. Irradiation occured every five minutes to avoid photobleaching. At least 10 independent cells were recorded.

Confocal microscopy analysis was performed with a Zeiss LSM510. GFP fluorescence was monitored with a 505–550 nm band-pass filter under 488-nm excitation. RFP fluorescence was followed with a long-pass filter under 543-nm excitation. Typically, sections of 1-μm thickness were taken. Images were processed with ImageJ (http://rsb.info.nih.gov/nih-image ) and Adobe Photoshop (http://www.adobe.com) to adjust contrast.


We first thank Nicole Bechtold and Roger Voisin for their work in the T-DNA insertion collection (Versailles). We are indebted to Dr J. Mutterer for his technical assistance in confocal analysis and to Roxanna Isac for her help in BY-2 cell maintenance. We want to thank Prof. Hasezawa for sending us GFP-tubulin. We do not forget M. Laudie and C. Berger for all their sequencing work. We are grateful to Y. Meyer for help with the phylogenic tree construction, and to T. Roscoe and R. Cooke for critical comments on the manuscript. Plant Gateway vectors were provided by Functional Genomics Division of the Department of Plant Systems Biology (VIB-Ghent University).