TSO1 functions in cell division during Arabidopsis flower development

One of the fundamental differences between animals and plants is that plant cells are surrounded by a rigid wall, preventing cell migration. The lack of mobility of plant cells leads to the requirement for well organized and regulated cell divisions with specific cell division orientations to achieve directional growth. Like other dicots, Arabidopsis shoot apical meristems and floral meristems are organized into three cell layers (L1, L2, and L3) with cells in L1 and L2 usually dividing anticli-nally (division plane perpendicular to the surface) and cells in L3 dividing in all directions (Poethig 1989; Steeves and Sussex, 1989). Such control of division planes may ensure properly organized meristems and thus proper organ initiation and morphogenesis.

Higher plant cell division exhibits several unique properties: the presence of a preprophase band (PPB), the absence of an obvious microtubule-organizing center (MTOC), and a mechanism for cytokinesis involving vesicle fusion. The division plane in higher plants is first marked in advance by a preprophase band (PPB) of microtubules and actin, which briefly surrounds the cell prior to prophase (Gunning and Wick, 1985) and is not present in animals and fungi. Second, higher plants lack obvious microtubule organizing centers (the centrosome in animals and spindle pole bodies in yeast; Lloyd, 1991; Lambert, 1993). Third, unlike animal cells, where a contractive mechanism is used for cytokinesis (Rappaport, 1986), most higher plants use vesicle fusion for formation of new cell walls and thus for cytokinesis (Pickett-Heaps and Northcote, 1966; Gunning and Wick, 1985). New cell wall and membrane materials are deposited between dividing cells by vesicle fusion via a structure called a phragmoplast. Although the cell division process in higher plants has been well described, little is known about the genes and the control mechanisms under-lying the cell division process.

Genetic approaches may be effective in identifying genes that regulate aspects of cell division. Recently, several genes have been identified and/or isolated that affect plant cytokinesis and preprophase band formation. KNOLLE (KN; Lukowitz et al., 1996) and KEULE (Jürgens et al., 1991) in Arabidopsis and CYD in pea (Liu et al., 1995) have been shown to be involved in cytokinesis. The FASS/TON genes have been implicated in PPB formation (Torres-Ruiz and Jürgens 1994; Traas et al., 1995). The identification, isolation, and analysis of genes in the plant cell division process is beginning to reveal the molecular mechanisms governing cell division, organ development and morphogenesis in higher plants.

In this paper, we report the identification of tso1, an Arabidopsis mutant that affects proper flower organ development. Our analyses show that tso1 mutations cause a disorganized floral meristem structure, and tso1 mutants exhibit similar cellular defects to kn, keule, and cyd mutants. These cellular defects include partially formed cell walls and increased DNA ploidy. Our data suggest that TSO1 may be similarly involved in cytokinesis. However, unlike kn, keule and cyd, which affect embryos and seedlings, tso1 mutations only affect later stages of development, mostly limited to flower development. This suggests either that TSO1 encodes a floral-specific cell division component, or that TSO1 function is redundant in other tissue types.

Plant growth conditions and mutagenesis were described previously (Liu and Meyerowitz, 1995). tso1-1 was identified in the M₂ genera-tion of an EMS mutagenesis of ap2-1 mutants in Landsberg erecta (Ler) background, and subsequently recovered from its M₂ siblings heterozygous for tso1-1. The ap2-1 mutation was subsequently removed from the tso1-1 background by back crosses. To map tso1, tso1-1 heterozygous plants were crossed with wild-type plants of ecotype Columbia (Col). Individual F₂ wild-type plants heterozygous for tso1-1 were identified by the segregation of tso1-1 plants in the F₃ generation. One F₃tso1-1 plant of each F₂ family was analyzed using SSLP (Simple Sequence Length Polymorphism; Bell and Ecker, 1994) and CAPS (Co-dominant cleaved Amplified Polymorphic Sequences; Konieczny and Ausubel 1993) markers. Linkages to two SSLP markers on chromosome 3 (nga162 and nga172) were found, and the distances between tso1 and nga172 and nga162 are estimated to be 30 and 16 map units respectively. Linkage to a CAPS marker residing in the SUPERMAN gene (SUP; Sakai et al., 1995) was also detected, and tso1 was found to be 6 map units north of SUP. None of the existing mutations that map to this region of Chromosome 3 exhibits similar phenotypes to tso1.

Flowers were collected, fixed, embedded, sectioned and hybridized as described previously (Drews et al., 1991).

Samples were collected, fixed, stained and visualized using the procedure of Running et al. (1995).

Serial 4 μm sections of floral tissues embedded in JB4 resin (Polysciences Inc.) were stained with 4′,6′-diamidino-2-phenylindole (DAPI), and the relative fluorescence units were measured using a MSP 20 microspectrophotometer with digital microprocessor attached to a Zeiss Axiophot microscope equipped with epifluorescence (essentially as described by Friedman, 1991). The area measured was adjusted according to the size of the nucleus, and the background was subtracted each time from an area adjacent to the nucleus measured. The 2C standard was derived by taking an average of readings of several newly divided or dividing nuclei at telophase.

Tissues were fixed in 4% paraformaldehyde and 2.5% glutaradehyde in 25 mM phosphate buffer (pH 7.2-7.4) for 2 hours at room temperature. They were washed, stained en bloc with uranyl acetate (5%), post-fixed in 1% osmium tetroxide, and embedded in LR white resin. 0.08-0.1 μm thick sections were placed on grids, which were stained with lead citrate (33 mg/ml) and uranyl acetate (5%), and visualized using a Philips E420 electron microscope.

During postembryonic development of Arabidopsis, the shoot apical meristem first generates several rosette leaves, then gives rise to cauline leaves with axillary shoots. Subsequently, the apical meristem switches to the production of flowers instead of cauline leaves, and is then termed an inflorescence meristem. The inflorescence meristem gives rise to floral meristems in a spiral phyllotaxy (Fig. 1A,B), and each floral meristem develops 4 types of floral organs arranged in 4 concentric whorls: 4 sepals in whorl 1, 4 petals in whorl 2, 6 stamens in whorl 3, and 2 fused carpels in whorl 4 (Fig. 1A). The tso1-1 mutant was initially identified in an EMS mutagenesis screen by its ‘callus-like’ flowers (Fig. 1D,G and Methods). In this paper, the term ‘callus-like’ describes the dis- organized cell and organ mass and does not imply extra cell proliferation, nor properties of a callus in plant tissue-culture. The mutation was named ‘tso’, which means ugly in Chinese. A second EMS-induced allele, tso1-2, caused a similar but weaker phenotype. Both tso1 alleles are recessive to wild type. All subsequent analyses were performed with the strong tso1-1 allele. TSO1 has been mapped to chromosome 3 using PCR-based markers nga162 (Bell and Ecker, 1994) and SUPERMAN (SUP; Sakai et al., 1995) and found to be 16 map units south of nga162 and 6 map units north of SUP (see Methods).

Unlike floral homeotic mutations that transform the identity of floral organs (reviewed by Weigel and Meyerowitz, 1994), the tso1-1 mutation blocks the formation of floral organs in whorls 2, 3 and 4 (Figs 1I, 2E). tso1 inflorescence occasionally fasciates (Fig. 1G), and the phyllotaxy is also abnormal (compare Fig. 1B with 1E). tso1-1 mutant flowers develop abnormal sepals (Fig. 1H) but do not develop petals, stamens, or carpels, and are consequently completely sterile (Fig. 1D,G,I). Although sepals are formed in tso1-1 flowers, they consist of abnormally large and misshapen (bloated) cells (Fig. 1H) and are also abnormal in number; between five to six sepals are formed in each tso1-1 flower instead of four in wild type (compare Fig. 1A,C with 1F,H).

The tso1-1 mutant floral meristems were further analyzed using Confocal Laser Scanning Microscopy (CLSM; Fig. 2). Wild-type and tso1-1 floral meristems were stained with propidium iodide (PI; a DNA staining dye), and then optically sectioned according to the method of Running et al. (1995). In wild type, the three cell layers (L1, L2, and L3) of a floral meristem are distinct and well defined (Fig. 2A), whereas in tso1-1, the organization of the three layers appears disrupted (Fig. 2D). In addition, there are fewer cells in tso1 floral meristems than in wild type (compare Fig. 2A with 2D, and 2B with 2E), and the nuclei in tso1-1 floral meristems are larger and irregular in shape (Fig. 2D,E), implying a defect in nuclear division. In flowers at stage 6, wild-type floral meristems give rise to organs in the inner whorls via regulated cell division patterns (Fig. 2B). tso1-1 floral meristems give rise to a few incipient organ primordia (or abnormal outgrowths) in the center of the flowers (Fig. 2E; note the area between the brackets), but not to organ primordia. Cells in these incipient organ primordia are abnormally large and vacuolated; they eventually undergo degeneration, exhibiting degraded cytoplasmic materials (data not shown). Although cellular abnormalities can be detected as early as stage 1 floral meristems, these cellular abnormalities are more frequently detected and more severe in meristems of advanced developmental stages (from stage 2 onward).

The effects of tso1 mutations are largely specific to the flowers (Figs 1B, 2). tso1 plants are largely normal until they start to flower. The tissue-specificity of tso1-1 mutations was also examined using CLSM. Roots, cauline leaves, apical meristems (including the early vegetative phase and later the reproductive phase) were examined in tso1-1 plants. Although the cauline leaves of tso1 appear more succulent and slightly darker green than wild type (Figs 1D and 4H), we could not detect any abnormalities or differences from wild type in paradermal sections of tso1-1 cauline leaves (Fig. 2G). 7 day-old root tips as well as adult roots were also examined; no dramatic differences were observed in the roots of tso1-1 from those of wild type (Fig. 2H). tso1-1 vegetative meristems and early inflorescence meristems do not exhibit any discernible defects (Fig. 2I). However, tso1-1 late inflorescence meristems sometimes fasciate, resulting in enlarged stems, increased numbers of flowers, and abnormal phyllotaxy (Figs 1E,G, 2F). In contrast to the wild-type inflorescence meristem (Fig. 2C), the tso1-1 primary inflorescence meristem fasciated and bifurcated to produce a secondary inflorescence meristem, from which a floral meristem would develop (Fig. 2F). However, the cellular defects (i.e. abnormal cell layers, enlarged or irregularly shaped nuclei) in tso1-1 are predominantly found in floral meristems and only rarely in the inflorescence meristems.

To test if the cells in the center of the flower can properly express floral homeotic genes, we examined the expression of the genes APETALA3 (AP3) and AGAMOUS (AG) (Yanofsky et al., 1990; Jack et al., 1992; reviewed by Weigel and Meyerowitz, 1994). AP3 is required to specify petal and stamen development, and AP3 RNA is normally expressed in whorls 2 and 3 of a flower (Fig. 3A). AG is required to specify stamen and carpel development and is normally expressed in whorls 3 and 4 (Fig. 3C). In tso1-1 mutant flowers, these two genes are expressed in proper domains: AP3 in an area corresponding to whorls 2 and 3 (Fig. 3B), and AG in the center of the floral meristem corresponding to whorls 3 and 4 (Fig. 3D). However, the level of their expression appears to be reduced relative to wild type (Fig. 3B,D). This reduction of floral homeotic gene expression may result from a reduced number of cells and/or reduced transcriptional activities in these cells.

Double mutants homozygous for tso1 and floral homeotic mutations exhibit additive phenotypes. For example, weak ap2-1 mutants develop leaves in whorl 1 characterized by branched trichomes (Fig. 4A; Bowman et al., 1991), while strong ap2-2 mutants develop carpels in whorl 1 characterized by stigmatic tissues and thick carpel walls (Fig. 4D; Bowman et al., 1991). In ap2-1 tso1-1 double mutants, whorl 1 organs are leaf-like with branched trichomes (Fig. 4B,C), while in ap2-2 tso1-1 double mutants, whorl 1 organs exhibit carpelloid features such as thick carpel walls and the development of ovule primordia (Fig. 4E,F). Organs interior to the first whorl in either of these double mutants fail to form as in tso1-1 mutants (Fig. 4B,E). In addition, whorl 1 organs in these double mutants exhibit abnormal cellular morphology such as large and misshapen cells characteristic of tso1-1 (Fig. 4C,F). Hence, these double mutants express additive phenotypes: the corresponding ap2 phenotypes in organ identity transformation and the tso1-1 phenotypes in inner whorl organ formation and in cellular morphology. Since AP2 normally functions to repress AG expression in whorls 1 and 2, the ap2-2 mutation results in ectopic AG expression in whorls 1 and 2 (Bowman et al., 1991; Drews et al., 1991). The carpelloid first whorl organs in the ap2-2 tso1-1 double mutants suggest that AG still functions in the tso1-1 background, but its function in whorls 2, 3 and 4 is masked by the callus-like defects caused by tso1-1. Thus floral homeotic genes and TSO1 likely act independently of each other.

LEAFY (LFY) specifies the identity of floral meristems and, thus, is a member of the meristem identity genes (Weigel et al., 1992). The weak lfy mutant, lfy-5, enhances the fasciation occasionally observed in tso1 mutants (Fig. 4G,H,I). The weakly expressed fasciation in tso1 mutants is significantly enhanced in lfy-5 tso1-1 double mutants both in the percentage of plants that fasciate and in the extent of fasciation. More cauline leaves and secondary shoots are made in the double mutants than either single mutant, perhaps as a result of fasciation in the shoot apical meristems. Thus, in lfy5 tso1-1 double mutants, many more leaves are made at the expense of floral tissues (Fig. 4H). Consequently, callus-like tissues are reduced.

The confocal microscopy analyses indicate that cells in tso1-1 floral meristems appear to have more highly-stained and larger nuclei, suggesting that tso1 mutants are defective in nuclear division or cytokinesis, but are capable of DNA synthesis. The DNA content in individual cells was measured through microscopic spectrophotometry. Plastic histological sections of floral tissues were stained with DAPI, and the relative fluorescence units were measured using two different methods. First, nuclei were measured from single 4 μm sections of wild-type and of tso1-1 floral meristems. Since the nuclei in wild-type Arabidopsis floral meristem cells are, on average, 2-3 μm in diameter (based on TEM measurements), about half of the nuclei are encompassed entirely within the 4 μm histological sections. Thus, only large and intact nuclei in each 4 μm were measured. The range of readings are summarized in Fig. 5. While the nuclei of wild-type cells in a stage 5 flower exhibit 2C and 4C values, some of the tso1-1 nuclei exhibit 8C and even 16C levels of DNA. Since the first method may underestimate certain nuclei that are particularly large and thus may reside in several sections, in the second method, the total fluorescence units of each nucleus were derived by measuring the same nucleus in successive sections to include all nuclear materials. The C value here reflects the combined readings from 2 to 3 successive 4 μm sections. Only those nuclei that are easily traceable in successive sections were measured; thus the readings do not reflect the distribution of C value in the population of cells. While wild-type cells exhibit C values of 2C and 4C, tso1-1 cells exhibit C values between 2C and 63C. Specifically, C values of 2C, 6C, 6.6C, 9.2C, 10C, 12C, 12.8C, 40C and 63C were found in 16 tso1-1 nuclei measured using the second method.

Transmission electron microscopy (TEM) analysis has revealed more details of the subcellular defects in tso1 floral meristematic cells (Fig. 6). While wild-type floral meristem cells have completely formed cell walls and smooth and round nuclear membranes (Fig. 6A), tso1 floral meristem cells exhibit a variety of defects. First, cell walls are frequently incompletely formed (Fig. 6C,D,G). We often observed that cell walls project from one side and fail to connect to the other side. Occasionally, cell walls project from both sides and fail to connect to each other. Second, some nuclei are abnormally large, abnormal in shape (Fig. 6B,C,D,E), and the nuclear membrane invaginates and exhibits a wavy (convoluted) shape (Fig. 6D,E; indicated by arrowheads). Cells that exhibit these subcellular defects are more frequently found in the L1 than in the L2 and L3 layers. In L1, 18 of 106 cells or 17%, in L2, 2 of 78 cells or 2.5%, and in L3, 3 of 85 cells or 3.5% exhibited obvious subcellular defects. Abnormal cells are also frequently found in tso1-1 mutant sepals. Metaphase cells with condensed chromosomes were found in tso1 floral meristems (Fig. 6F-H), suggesting proper chromosome condensation and nuclear envelope breakdown. However, this may not be true with all the tso1 mutant cells. By analysis of serial sections, we observed that the chromosomes in different planes appear to lie in different subcellular locations: at the metaphase plate (Fig. 6F) in one section and near one pole of the cell (Fig. 6G) in another section, suggesting different spindle orientation in different planes of the same cell. Serial sections of cells that appear multinucleate revealed that some of the nuclei are linked by membrane materials (such as the one shown in Fig. 6C). However, in other cases, the multiple nuclei are completely separated from each other, nonetheless remaining in the same cell (data not shown).

We have identified and characterized a novel genetic locus, TSO1, in Arabidopsis. Tso1 mutants are defective in nuclear division (mitosis) and cytokinesis in floral cells. In tso1 floral meristems and sepals, we observed abnormalities including increased cell size, decreased cell number, disorganized layer structure, increased DNA content per cell, and partially formed cell walls. These defects are different from those of floral homeotic mutants, which act to alter floral organ identity. Based on the additive phenotypes of tso1 and floral homeotic double mutants, TSO1 and floral homeotic genes probably act independently of each other.

What is the primary role of TSO1 in floral meristem cell division? The subcellular phenotypes suggest a defect in mitosis, cytokinesis, and/or failed coordination between mitosis and cytokinesis. These defects can be caused by abnormal cytoskeletal structure, microtubule organizing activity, vesicle fusion during cell plate formation, or failed metaphase/anaphase check points. For instance, incomplete separation of nascent chromosomes could lead to a failure in the formation of two separate nuclei, subsequently blocking completion of cytokinesis. Alternatively, tso1 mutant cells could fail to monitor the metaphase/anaphase check point, and proceed to cytokinesis without completing mitosis.

The effects of tso1 mutations are similar to mutants found in the fission yeast Schizosaccharomyces pombe (reviewed by Fankhauser and Simanis, 1994) and in other plants (see below). The S. pombe mutants cdc3, cdc4, cdc8 and cdc12 arrest as swollen, elongated cells with two or four nuclei, and contain disorganized septal materials. These S. pombe genes encode profilin (cdc3), a novel form of tropomyosin (cdc8), and an EF-hand protein (cdc4) (see review by Fankhauser and Simanis, 1994). Another class of S. pombe mutations uncouple the normal dependency of mitosis and cytokinesis. This so called cut phenotype is seen when cytokinesis occurs despite the fact that nuclear division is blocked, resulting in cleavage of the undivided nucleus by the septum (Uzawa et al., 1990; Samejima et al., 1993). CUT1 encodes a protein with ATP-binding domains, is a rare component of the insoluble nuclear fraction, and may play a key role in coupling chromosome dis-junction with other cell cycle events (Uzawa et al., 1990). CUT7 encodes kinesin-like protein (Hagan and Yanagida, 1990). The Arabidopsis mutants kn (Lukowitz et al., 1996), keule (Jürgens et al., 1991), and the pea mutant cyd (Liu et al., 1995) also exhibit similar subcellular defects to tso1-1, including partially formed cell walls, polyploid cells, and multinucleate cells. However, unlike tso1 mutations, kn, keule, and cyd affect embryos and seedlings. KN protein is similar to syntaxins, a protein family involved in vesicular trafficking, implying a role of KN in cell plate formation (Lukowitz, et al., 1996). The molecular isolation and analyses of TSO1 will reveal if TSO1 shares homology with any of these genes at the molecular level.

In stamen hairs of Tradescantia, Ca²⁺ depletion can result in partial cell wall formation (Paul and Goff, 1973; Jürgens, et al. 1994). Similarly, caffeine treatment of plant cells can also mimic this effect (Paul and Goff, 1973; Liu et al., 1995). This is intriguing for Ca²⁺ has long been implicated as a second messenger in both mitosis and cytokinesis. Ca²⁺ has also been implicated in the microtubule organizing center (Lambert, 1993). If similar mechanisms are employed in Arabidopsis, TSO1 could also be involved in regulating intracellular Ca²⁺ concentration and/or Ca²⁺ signal transduction, or it could be a target of caffeine.

In most higher plant cells, new cell plate formation occurs from the center of the dividing cell and then expands outward toward the parental walls. This outward growth of cell walls is achieved by fusion with vesicles containing new cell wall materials. The partially formed cell plates (or wall stubs) observed in tso1 mutant cells are always attached to parental walls on one side. These cell wall stubs appear to imply an inward cell wall growth. However, the cell wall stubs could also reflect remnants of cell plates that are stabilized by fusion with the parental walls and the unfused cell walls could be rapidly degraded. These cell wall stubs are a common phenomenon, shared among tso1, kn, keule, and cyd mutant cells, as well as caffeine-treated plant cells.

Meristem identity genes such as LFY and AP1 (Irish and Sussex, 1990; Weigel et al., 1992; Bowman, et al., 1993) positively regulate the initiation of the floral program. Thus, any floral tissue-specific genes are directly or indirectly regulated by meristem identity genes even if these floral-specific genes perform diverse functions. Likewise, TSO1, as a floral-specific gene, is similarly regulated by the meristem identity genes even though TSO1 may act in a very different pathway/process from the meristem identity genes. Thus, only if the floral meristem identity is determined and expressed, can the tso1 mutant effects be manifested.

Two possible mechanisms could be responsible for the largely floral tissue-specific effect of tso1: (1) a redundant gene in tissue-types other than flowers or (2) a floral meristem-specific cell division function. In light of the second mechanism, it is likely that different cell division functions may be necessary for different tissue types. For example, floral meristems play fundamental roles for plant reproduction. Mechanisms must exist to ensure proper cell proliferation, meristem architecture, cell-cell communication, organ initiation, differentiation and morphogenesis in the floral meristems. The unique properties of floral meristems may thus require floral-specific cell division regulators or floral-specific components of cell division machineries. Thus further genetic and molecular analyses of TSO1 may shed light on these questions regarding tissue-specific cell division control and plant development.

We thank Patrick Koen and Jean Edens for their assistance in TEM, Joshua Levin for the tso1-2 allele, William Friedman for advice on microscopic spectrophotometric measurements and the use of his facilities, Glen Turner, Michael Chu and Mintao Fan for contributions to double mutant construction, confocal microscopy and mapping respectively. We also thank Eric Baehrecke and members of the Meyerowitz lab for critical reading of the manuscript. Z. L. was a DOE-energy Biosciences Research Fellow of the Life Science Research Foundation, and was previously a post-doctoral fellow of Damon Runyon-Walter Winchell Cancer Research Fund. M. P. R. was a Howard Hughes predoctoral fellow. This work was supported by NIH grant GM 45697 to E. M. M.