Cell cycle-independent expression of the Arabidopsiscytokinesis-specific syntaxin KNOLLE results in mistargeting to the plasma membrane and is not sufficient for cytokinesis

Cytokinesis partitions the cytoplasm of the dividing cell, which requires targeting of membrane vesicles to the plane of division. In yeast and animal cells, a contractile actomyosin ring supports ingrowth of the existing plasma membrane, and this cleavage furrow is expanded by the fusion of membrane vesicles that are delivered along furrow microtubule arrays (reviewed by Robinson and Spudich, 2000EF43;Straight and Field, 2000EF56). By contrast, plant cells form the partitioning plasma membrane de novo from the centre to the periphery of the cell (reviewed by Staehelin and Hepler,1996EF54; Heese et al.,1998EF21). Golgi-derived vesicles are transported along the microtubules of a plant-specific cytoskeletal array,the phragmoplast, to the plane of cell division where they fuse with one another to form a transient membrane-bounded compartment, the cell plate,which matures into a cell wall with flanking plasma membranes. The lateral expansion of the cell plate is mediated by the transformation of the phragmoplast from a compact array into a widening hollow cylindrical structure that delivers additional vesicles to the growing edge of the cell plate until the latter fuses with the parental plasma membrane (Samuels et al.,1995EF47). Thus, plant cytokinesis is a special case of vesicle trafficking and fusion.

Mutations in several genes of Arabidopsis, including KNOLLE and KEULE, result in cytokinesis defects, such as enlarged cells with incomplete cell walls and more than one nucleus (Lukowitz et al., 1996; Assaad et al.,1996; Nacry et al.,2000). KNOLLE encodes a cytokinesis-specific syntaxin (Lukowitz et al.,1996; Lauber et al.,1997). KEULE is a member of the Sec1 family of syntaxin-binding proteins that interacts with KNOLLE in vitro and in vivo, and mutations in both genes result in the accumulation of unfused cytokinetic vesicles (Assaad et al.,2000; Lauber et al.,1997; Waizenegger et al.,2000). Whereas the KEULE gene appears to be expressed in both proliferating and non-proliferating cells (Assaad et al.,2000), the expression of KNOLLE is tightly regulated during the cell cycle. KNOLLEmRNA accumulates transiently in proliferating cells, giving a patchy pattern that reflects asynchrony of cell division in the embryo (Lukowitz et al.,1996). KNOLLE protein accumulates only during M phase, initially in patches presumed to represent Golgi stacks, then localises to the forming cell plate during telophase and disappears at the end of cytokinesis (Lauber et al.,1997). The tight regulation of KNOLLE expression is reminiscent of the synthesis and degradation of mitotic cyclins (Ito, 2000). KNOLLE syntaxin appears to be involved in all sporophytic cell divisions as well as in endosperm cellularisation (Lauber et al.,1997).

Syntaxins are components of SNARE complexes that play an important role in membrane fusion events (reviewed by Jahn and Südhof,1999EF27). The SNARE core complex consists of three or four proteins that form a four-helix bundle: a bipartite t-SNARE on the target membrane, which consists of a syntaxin and a SNAP25 protein or two t-SNARE light-chain proteins, interacts with the v-SNARE synaptobrevin on the vesicle membrane (Clague and Herrmann,2000EF10). There are numerous members of each SNARE protein family in yeast, animals and plants that have been implicated in diverse vesicle trafficking pathways between membrane compartments (for reviews on plant SNAREs, see Blatt et al.,1999EF7; Sanderfoot et al.,2000EF48). In general, syntaxins and synaptobrevins involved in a particular pathway appear more closely related to functional counterparts in different organisms than to family members involved in a different pathway within the same organism. The original SNARE hypothesis postulated that specific pairs of cognate syntaxins and synaptobrevins provide specificity to vesicle trafficking(Söllner et al.,1993aEF51;Söllner et al.,1993bEF52). This idea was challenged in recent in vitro interaction studies that provided evidence for promiscuity among interacting SNARE partners (Fasshauer et al.,1999EF15). However, thorough analyses of yeast SNARE interactions in liposome assays have indicated a high degree of specificity of interaction between syntaxins and synaptobrevins(Fukuda et al., 2000EF17; McNew et al., 2000EF36; Parlati et al.,2000EF42).

KNOLLE is a distant member of the plasma membrane subgroup of the syntaxin family but has no close counterpart among yeast or animal syntaxins (Lukowitz et al., 1996; Sanderfoot et al., 2000). However, syntaxins with analogous roles in membrane fusion during cellularisation or cytokinesis have been described in animals. The Drosophila syntaxin 1 gene is required for cellularisation of the blastoderm embryo, as well as for neural development (Burgess et al.,1997). Likewise, the Caenorhabditis syn-4 gene is involved in embryo cleavage divisions but also plays a role in nuclear membrane reformation (Jantsch-Plunger and Glotzer, 1999). In contrast to the other two syntaxins, KNOLLE is required only for de novo formation of the partitioning plasma membrane during cytokinesis, and its expression is tightly regulated during the cell cycle, suggesting a unique role in cytokinesis.

We have addressed the biological significance of the tight regulation of KNOLLE expression by replacing the endogenous 5′ regulatory region with promoters that are active in both proliferating and non-proliferating cells. The transgenic plants were phenotypically normal, although KNOLLE protein accumulated strongly in non-proliferating cells and was mistargeted to the plasma membrane. Conversely, the KNOLLE transgene did not rescue knolle mutant embryos, which correlated with low-level accumulation of mRNA from the KNOLLE transgene in proliferating embryonic tissue,when compared with the activity of the endogenous KNOLLE gene. Our observations suggest that the tight regulation of KNOLLE expression meets two opposing requirements. First, the KNOLLE gene must be strongly expressed to produce sufficient KNOLLE protein during M phase for the efficient execution of cytokinetic vesicle fusion. Second, degradation of KNOLLE mRNA and protein prevents the accumulation of large quantities of useless molecules.

Arabidopsis thaliana ecotypes Wassilewskija (WS),Landsberg/Niederzenz (Ler/Nd) heterozygous for the knolle mutation X37-2 (Lukowitz et al.,1996) and the EMS-induced knolle allele UU1319 (kindly provided by U. Mayer) were grown on soil at 18°C, as described previously (Mayer et al.,1991). The Arabidopsis cell suspension culture (Fuerst et al.,1996) was a gift from the John Innes Centre (Norwich, UK). The in vitro culture was done in petri dishes containing 1% Select Agar (Gibco BRL, Karlsruhe, Germany) and 0.5× or 1× Murashige and Skoog (MS) salts (Ducheva, Haarlem, The Netherlands) at 18°C under constant illumination. To induce root hair formation, 1%sucrose was added. To determine the organisation of vacuoles in root hairs,seedlings were grown on 0.5× MS medium containing no, 1% or 3% sucrose. Callus induction of knolle-X37-2 mutant seedlings carrying the 35S::KN transgene was carried out with modified callus-inducing (1 mg/l 2.4D, 0.25 mg/l kinetin) or shoot-inducing (0.5 mg/l NAA, 0.25 mg/l kinetin) media (Soni et al.,1995).

WS and knolle-X37-2 (Ler/Nd) heterozygous plants were transformed by a modified transformation protocol (Bechtold and Pelletier,1998; Clough and Bent,1998), using a combination of vacuum infiltration and additional dip transformation one week later. One hundred to 150 plants were transformed with an Agrobacterium GV3101 culture bearing the desired transgene. Infiltration medium consisted of 0.5× MS salts, 1× Gamborg B5 vitamins (Ducheva), 5% sucrose, 0.044 μM benzyl aminopurine (Sigma) pH 5.7 with KOH, and 0.005% SILWET L77 (Osi Specialities). T1 seeds were bulk-harvested (each seed representing a single transformation event; Bechtold et al., 2000;Desfeux et al., 2000; Ye et al., 1999), sown on soil and selected for transformants by spraying BASTA® (183 g/l Glufosinate,AgrEvo™; Düsseldorf, Germany; 1:1000)twice. BASTA®-resistant plants were genotyped for KNOLLE by PCR with the primers X37-2C and X37-2D (Lukowitz et al.,1996), which amplify a 0.7 kb fragment from X37-2 and a 1.7 kb fragment from wild type. Seeds containing knolle mutant embryos are shrunken and darker than wild-type seeds. For confirmation of the genotype, mutant seeds were germinated on 0.5× MS salts, 1% Select Agar plates, and seedlings were examined for the knolle mutant phenotype.

Plants heterozygous for kn-X37-2 were transformed with the KNRescue construct (see Fig. 1). Selfing of the T0 plants gave three distinguishable genotypes of BASTA®-resistant T1 progeny bearing the KN transgene: (1) KN/KN, (2) kn/KN and (3) kn/kn. PCR with KNOLLE-specific primers amplified the kn-X37-2 fragment from the genotypes (2) and (3) (Lukowitz et al.,1996). Selfing of T1 plants with genotype (2) or (3) produced 6.25% or 25% knolle mutant T2 seeds, respectively. Reduction of phenotypically mutant seeds from 25% to 6.25% for genotype (2) indicated complementation. One-third of the Basta-resistant T2 plants derived from genotype (3) were homozygous for the transgene and produced only phenotypically normal seeds, although the embryos were homozygous for the kn-X37-2 mutant allele (T3 generation).

Constructs for plant transformation were introduced into pBar vectors. pBarA (AJ251013) was used for AP3::KN misexpression and for the rescue of the knolle mutant phenotype by a KNOLLE SacI-SnaIgenomic fragment. pBar-35S (AJ251014) was used for 35S::KNmisexpression. pBar vectors were a gift from G. Cardon (MPI, Cologne,Germany). PCR was carried out according to standard procedures using TaqPlus Precision™ (Stratagene, La Jolla, CA), Expand High Fidelity™(Roche Mannheim, Germany) and Taq DNA Polymerase™ (Roche). The AP3 promoter fragment was amplified by PCR using pD1075:AP3(-650Æ -1, a gift from T. Jack) with the forward primer AP3-98a(5′AATTCTAGACAAGGATCTTTAGTTAAGGC 3) introducing a XbaI 5′ restriction site and with the reverse primers AP3-46a(5′ ATACTGCAGATTTGGTGGAGAGGACAAG 3′) and AP3-47: (5′ATACTGCAGGAAGAGATTTGGTGGAGAGGACAAG 3′)introducing a consensus transcription start (Joshi,1987) and a 3′ Pst1 restriction site. The KNOLLE fragments KN1 (-217 to+1500) and KN2 (-4 to +1500) were PCR amplified with either forward primer KNstart1 (5′TTTCTGCAGCTTTCTCTCATCTCACA AATC 3′) or KNstart2 (5′ATACTGCAGAAGATGAACGACTTGATGACG 3′),introducing a PstI restriction site at the 5′ end, and reverse primer KNstop (5′ ATAGAATTCATGACCTTGTTCCAGAGATTG 3′), introducing an EcoRI cloning site at the 3′ end. AP3::KN1 (KNOLLE coding sequence with intron; see Fig. 1) and AP3::KN2(coding sequence without intron) were used for KNOLLE misexpression, which gave essentially the same results (data not shown). XbaI-AP3::KN-EcoRI was cloned into pBarA. p35S::KNconstructs were obtained by restriction digest of Bluescript SK::KN2 with XhoI, introducing 40 nucleotides non-coding sequence 5′ to the KNOLLE translational start. The 3′ end of the KN2 fragment was a XbaI restriction site within the Bluescript SK vector, introducing another 30 nucleotides. This fragment was cloned into pBar-35S SmaI +XbaI. A genomic SnaI and SacI 4.75 kb fragment containing KNOLLE was directly subcloned into pBarA SacI, SmaI. The constructs were confirmed by sequencing and transformed into Agrobacterium tumefaciens strain GV3101 (GV3101 + pM90: gift from G. Cardon, MPI Cologne, Germany). The p35S::KN transgene was re-amplified from transgenic plants by using the forward primer 35SPromoter2(5′ ACGCACAATCCCACTATCCTT 3′) and the reverse primer 35STerminator2 (5′ AAGAACCCTAATTCCCTTATCTGG 3′) close to the multiple cloning site of pBar-35S, and sequenced. The 35S::GUSreporter construct from the pBIC20 cosmid vector (Meyer et al.,1994) was used to monitor 35S promoter activity in embryos and seedlings. Molecular work was carried out according to standard protocols (Sambrook et al.,1989). Restriction enzymes were purchased from NEB (New England Biolabs, Hitchin, UK); synthetic oligonucleotides were from ARK Scientific (Sigma, Germany).

Sample preparation and in situ hybridisation of KNOLLE antisense riboprobes transcribed in vitro were carried out according to Mayer et al.(Mayer et al., 1998), using a 300 bp KNOLLE fragment from the 5′ end of the coding region(Lukowitz et al., 1996). Paraffin-embedded material was cut into 8 μm sections for embryos and 12.5-17.5 μm for seedlings. Digoxigenin-labelled probes were detected with Boehringer anti-Dig FAB (Roche) coupled to alkaline phosphatase. Western Blue® alkaline phosphatase (Promega, Madison, USA) colour reaction was carried out for 1-4 days. Samples were mounted in 50% glycerol. Images were taken with a Zeiss Axiophot (Carl Zeiss Inc., Thornwood, NY), using a Nikon Coolpix 990 digital camera with 3.34 Mio pixels.

Immunoblotting and immunolocalisation were as previously described (Lauber et al., 1997; Steinmann et al., 1999). For separation of proteins, 12 to 15% polyacrylamide SDS (Sigma) gels were used. Protein extraction was achieved by grinding plant material with sand and boiling in 1× Laemmli buffer, except for cell fractionation experiments. Western analysis with rabbit anti-KNOLLE antiserum was performed as previously described (Lauber et al.,1997). Protein concentrations were estimated by Coomassie Blue staining. Cell fractionation was done as previously described (Lauber et al.,1997). S10 was the supernatant of a 10,000 g precentrifugation, S100 and P100 were the supernatant and the pellet of a 100,000 g centrifugation for 12 hours. Integral membrane proteins were solubilised with Triton X-100(Sigma; Lauber et al.,1997).

Protein concentration was measured using a Bradford assay, and equal amounts of protein were loaded onto the gel. Immunolocalisation was carried out with rabbit anti-KNOLLE antiserum diluted 1:2,000 or with mouse monoclonal anti-plasma membrane H⁺-ATPase antibody diluted 1:500. Root tissue was fixed with 4% paraformaldehyde (Sigma) in MTSB (pH 7.0) for 0.5 hours. Goat anti-rabbit secondary antibody was coupled to Cy3™ (Dianova,Hamburg, Germany) or Alexa-m488 (Molecular Probes, Eugene OR, USA), goat anti-mouse secondary antibody was coupled to Cy3 (Dianova). Instead of MTSB,PBS (pH 7.2) was used in all steps after fixation of the plant material. The primary antibody was incubated for 3 hours at 37°C after blocking for 1 hour with 1% BSA in PBS, the secondary antibody was incubated for 3 hours at 37°C. Nuclei were stained with 1 mg/ml DAPI. After mounting in Citifluor(Agar, Amersham), specimens were analysed with a Zeiss Axiophot epifluorescence microscope equipped with a Nikon Digital camera (3.34 Mio pixels) or with a Leica confocal laser scanning microscope (CLSM) with Leica TCS-NT software. The CLSM standard objective was 63× (water immersion),scanning was carried out with electronic magnification.

GUS staining was as previously described (Sundaresan et al.,1995EF57). Plant tissue was incubated in the detection solution (500 mM NaPO₄ buffer pH 7.0,500 mM EDTA pH 8.0, 150 mM potassium ferrocyanide(K₄Fe(CN)₆ 3H₂O), 5% Triton X100, 40 mM X-Gluc in dimethylformamide) in the dark at 37°C for several hours until the blue colour became apparent. After transfer to water, the stained tissue was examined by bright-field light microscopy.

Wild-type and transgenic seedlings grown on agar plates were stained with the lipophilic steryl-dyes FM4-64 and FM1-43 (1 mg/mL; Molecular Probes,Eugene OR, USA). Both dyes show a Stoke shift when integrated into membranes,depending on the membrane composition. Whole seedlings were stained alive or after fixation with 4% paraformaldehyde (in MTSB) for 5 minutes and then destained for 15 minutes in water. For confocal laser scanning microscopy,FM4-64 and FM1-43 were excited by the laser at 568 nm and 488 nm,respectively, and emitted light at >600 nm and 500-530 nm plus 580-630 nm. In live root hair cells, FM4-64 stains predominantly endomembranes and also the plasma membrane, whereas the less hydrophobic dye FM1-43 preferentially stains the plasma membrane. FM4-64 endomembrane labelling correlated with the number and size of vacuolar structures in root hairs grown on 0.5× MS medium by varying sucrose content: increasing sucrose concentration (0% to 3%)resulted in enlarged but fewer vacuoles.

Roots of 5-day-old plants grown on 0.5× MS salts, 1% sucrose and 1%Select Agar were fixed with 4% formaldehyde in MTSB (pH 7.0) for 60 minutes and embedded in 1% agarose. After infiltration with 20% (w/v)polyvinylpyrrolidone (MW 10,000, PVP-10; Sigma) in 1.8 M sucrose (Tokuyasu,1989EF59) and freezing in liquid nitrogen, cells were sectioned at -85°C (400 nm, semithin) or at-100°C (100 nm, ultrathin) with a Leica Ultracut S/FCS. Cryosections were transferred to poly-L-lysine-coated (Sigma) coverslips for immunofluorescence or collected on electron microscopy copper grids. After blocking (1% skim milk/0.5% BSA in PBS, pH 7.2), labelling with rabbit anti-KNOLLE antiserum(1:1000) was performed for 60 minutes followed by incubation with Cy3-conjugated goat anti-rabbit secondary antibody (Dianova) or protein A-15 nm gold for 60 minutes (Slot and Geuze,1985EF50). For immunofluorescence,sections were stained with DAPI and embedded in Mowiol 4.88 (Hoechst,Frankfurt/Main, Germany; Rodriguez and Deinhardt,1960EF44) containing DABCO (25 mg/ml; Sigma; Langanger et al.,1983EF30). For electron microscopy, grids were stained with uranyl acetate and embedded in methyl cellulose (Sigma) according to Griffiths (Griffiths,1993EF19). Cy3-labeled cryosections were viewed with a Zeiss Axioplan, gold-labelled cryosections with a Philips 201 electron microscope at 60 kV accelerating voltage. For ultrastructural analysis, 5-day-old roots were cryofixed in liquid propane,freeze-substituted in acetone containing 0.5% glutaraldehyde and 0.5% osmium tetroxide and embedded in Spurr. Ultrathin sections were stained with uranyl acetate and lead citrate.

All images shown were processed with Photoshop 5.0 and Illustrator 8.0(Adobe Mountain View, CA).

Because of the tight regulation of KNOLLE expression in proliferating cells, we considered the possibility that its deregulated expression might be deleterious, as has been shown for other cell cycle-regulated genes (Ito,2000) and syntaxins (Zhou et al., 2000). We therefore generated transgenic plants that expressed KNOLLE protein under the control of the flower-specific APETALA3 (AP3) promoter(Fig. 1; AP3::KN1). The AP3 promoter is active in the petal and stamen primordia of the flower (Jack et al., 1992;Jack et al., 1994). If misexpression of KNOLLE interfered with cellular or developmental processes,this would be recognised by floral abnormalities of viable transgenic plants. Surprisingly, the transgenic plants were morphologically normal, although KNOLLE protein accumulated to high levels in petals(Fig. 2A). In addition, KNOLLE protein was detected in the microsomal fraction(Fig. 2A), as described for endogenous KNOLLE, an integral membrane protein (Lauber et al.,1997; see Fig. 2D). To determine possible effects of excessive amounts of KNOLLE protein in other organs and developmental stages, we generated 35S::KN transgenic plants(Fig. 1; Benfey et al.,1989). The 35S::KNtransgenic plants were also morphologically normal, although misexpressed KNOLLE protein was detected at very high levels in protein extracts from mature leaves and stem, both of which consist of differentiated cells that normally contain no KNOLLE protein (Fig. 2B). By semi-quantitative analysis of protein extracts, 35S::KN transgenic seedlings contained at least 100-fold more KNOLLE protein than did wild-type seedlings (Fig. 2C). Furthermore, in leaves of 35S::KN plants, KNOLLE protein was localised to the membrane fraction from which it could be removed by Triton-X100 (Fig. 2D), as reported for endogenous KNOLLE protein (Lauber et al.,1997). This subcellular localisation indicated that the membrane anchor of misexpressed KNOLLE protein was functional. For comparison, truncated KNOLLE protein from the knolle allele UU1319 that shows the complete loss-of-function phenotype lacks the membrane anchor and was detected in the soluble fraction(Fig. 2D; see Materials and Methods). In summary, misexpression from two different promoters resulted in high-level accumulation of membrane-integrated KNOLLE protein in non-proliferating cells, but did not interfere with essential cellular or developmental processes. Because it had no deleterious effect, misexpression of KNOLLE offered the unique possibility of exploring how the cell deals with a syntaxin that cannot be trafficked to its proper destination, which in this case is the cell plate.

To determine the fate of misexpressed KNOLLE protein, 35S::KNtransgenic embryos and seedlings were analysed by whole-mount immunolocalisation. Whereas no distinct pattern of abnormal localisation was detected in embryos (data not shown), deviation from the wild-type pattern was observed in roots of transgenic seedlings carrying two copies of the 35S::KN transgene outside the meristematic region of the root tip(Fig. 3). A clear difference between 35S::KN and wild-type roots was noted in the cells of the adjacent elongation zone and in the more mature part of the root, but not in the root tip, where cell divisions take place. The expanding cells of the central cylinder and the tip-growing root hairs strongly accumulated KNOLLE protein in transgenic, but not in wild-type seedlings (compare Fig. 3D,G with Fig. 3F,H). The activity of the 35S promoter was independently monitored by the expression of a GUS reporter gene (Fig. 3I-L). GUS staining was observed in root hairs, the central part of the root and the root tip, as described previously for the tobacco seedling root (Benfey et al.,1989). Although the 35S promoter was active in the proliferating cells of the root tip,KNOLLE protein failed to accumulate to high levels, presumably owing to cell cycle-dependent degradation. In summary, KNOLLE protein expressed from the transgene accumulated in root cells that were no longer dividing.

The subcellular localisation of misexpressed KNOLLE protein was analysed in detail in readily accessible root hair cells by confocal laser scanning microscopy (Fig. 4). These non-proliferating epidermal cells form a local outgrowth, the root hair, which undergoes tip growth by targeting membrane vesicles from the trans-Golgi to the apical plasma membrane (for a review, see Yang,1999; see also Fig. 5J,K). In a sense, this preferential vesicle targeting to the growing tip of the root hair resembles the vesicle targeting to the cell division plane during cytokinesis. Growing and mature root hairs were immunostained with anti-KNOLLE antiserum and with a monoclonal antibody directed against the plasma membrane H⁺-ATPase(PM-ATPase; Lauber et al.,1997). KNOLLE protein accumulated strongly in growing and mature root hairs of transgenic seedlings,in contrast to the wild-type control (Fig. 4B-D,E-G; compare with Fig. 4N). In young root hairs, KNOLLE was concentrated at the tip(Fig. 4B,E,H), whereas older root hairs also accumulated KNOLLE away from the tip(Fig. 4C-D,F,G,K). For comparison, we used the lipophilic fluorescent dye, FM4-64, which labels predominantly the membrane of the vacuole in yeast (Vida and Emr,1995). In root hairs, FM4-64 labelled predominantly endomembranes and to some extent the plasma membrane(Fig. 4O). Most of the FM4-64 label was located below the KNOLLE-positive apical region in young root hairs(compare Fig. 4P with 4E). In older root hairs, however, the FM4-64 label resembled KNOLLE-positive aggregates (Fig. 4P,Q; compare with Fig. 4F,G). To determine the fate of misexpressed KNOLLE protein more precisely, root hairs of 35S::KN seedlings were simultaneously immunostained for PM-ATPase(Fig. 4H-M). In young root hairs, both proteins co-localised to the apical tip region, not only at the surface but also internally (Fig. 4E-G; additional optical sections not shown). Away from the tip region, both proteins were strictly localised to the surface, resembling the plasma membrane staining with the lipophilic fluorescent dye, FM1-43 (data not shown; see Materials and Methods). Older root hairs also displayed almost perfect co-localisation of KNOLLE and PM-ATPase(Fig. 4K-M). Thus, KNOLLE protein was targeted like a protein destined to the apical plasma membrane of the growing root hair.

To reveal the ultrastructural localisation of KNOLLE protein in non-proliferating cells of 35S::KN transgenic seedlings, root cryosections were prepared for immunogold-labelling electron microscopy(Fig. 5; see Materials and Methods). In the central cylinder of the root, expanding cells displayed cytoplasmic patches of KNOLLE immunofluorescence that resembled those in dividing cells (Fig. 5D,F;compare with Fig. 5A). However,KNOLLE was also located at the plasma membrane, whereas dividing cells accumulated KNOLLE in the forming cell plate(Fig. 5F, arrowhead; compare with Fig. 5B,C). By immunogold labelling, KNOLLE was detected in Golgi stacks, the trans-Golgi network and at the plasma membrane of expanding root cells(Fig. 5G-I). We also analysed tip-growing young root hairs, which accumulate vesicles underneath the apical plasma membrane (Fig. 5J,K;Galway et al., 1997). Anti-KNOLLE immunogold label was concentrated at the tip of the root hair,with vesicles giving the strongest signal(Fig. 5L). In summary, these data support the light microscopy observation that KNOLLE protein is mistargeted to the plasma membrane in non-proliferating cells of 35S::KN transgenic plants. In addition, this is the first time that KNOLLE protein has been localised in cells at the ultrastructural level.

Expression of the 35S::KN transgene lacked any detectable biological effect in the wild-type genetic background. We therefore examined whether the transgene could functionally replace the endogenous KNOLLE gene. We transformed plants heterozygous for the kn-X37-2 mutation (Lukowitz et al.,1996) with the 35S::KN transgene and analysed their progeny for the occurrence of mutant seeds. As a control, we transformed kn-X37-2/KN heterozygous plants with a genomic DNA fragment that differed from the 35S::KNtransgene in the 5′ region, whereas their 3′ regions were nearly identical (see Fig. 1). The control construct gave 22 kn-X37-2/KN heterozygous and 13 kn-X37-2/knX37-2 homozygous independent transformants, all of which produced viable and fertile kn-X37-2/knX37-2 homozygous plants (for details, see Materials and Methods). By contrast, none of the 18 kn-X37-2/KN heterozygous independent transformants bearing the 35S::KN transgene gave rise to kn-X37-2/knX37-2 homozygous plants. Thus, the 35S::KN transgene did not rescue kn-X37-2mutant embryos. The transgene also did not promote growth of kn-X37-2mutant seedlings on callus-inducing medium (data not shown, see Materials and Methods). Sequencing of the transgene after re-isolation from the transgenic plants did not show any deviation from the wild-type sequence (see Materials and Methods). We therefore checked kn-X37-2 mutant seedlings carrying the 35S::KN transgene for KNOLLE protein accumulation by whole-mount immunolocalisation (Fig. 6). KNOLLE protein was detected in root hair cells of those knolle mutant seedlings. Thus, the 35S::KN transgene produced KNOLLE protein in knolle mutants, but failed to rescue cytokinesis-defective kn-X37-2 mutant embryos and seedlings.

To determine why 35S::KN did not rescue the knolle mutant phenotype, although transgenic plants expressed high levels of KNOLLE protein(see Fig. 2B,C), we analysed the transcript accumulation by in situ hybridisation of a KNOLLEantisense riboprobe to sections of 35S::KN transgenic embryos that carried two functional copies of the endogenous KNOLLE gene(Fig. 7). Up to the torpedo stage of embryogenesis, KN mRNA accumulated in a patchy pattern that was indistinguishable from the wild-type control(Fig. 7A,B; Lukowitz et al.,1996). At the bent-cotyledon stage, transgenic embryos showed diffuse expression predominantly in the cotyledonary primordia, whereas only a few cells in the wild-type control embryos were labelled (compare Fig. 7D,E,G with Fig. 7C). The intensity of diffuse staining was stronger in embryos with two copies of the transgene than in those with only one, indicating that the staining was due to transgene expression (compare Fig. 7E,G with Fig. 7D). Transgene expression was strongest in presumptive vascular cells of the cotyledons(Fig. 7E,F) and in adjacent internal cell layers (Fig. 7G,H). Within these stained tissues, individual cells gave stronger signals that resembled in intensity the stained cells in the hypocotyl, and probably reflect the expression of the endogenous KNOLLE gene (Fig. 7H,arrowheads). In summary, mRNA transcribed from the 35S::KN transgene accumulated detectably only at advanced stages of embryogenesis, at which its level of accumulation was lower than that of the endogenous KNOLLEmRNA. This result was consistent with the observation that developing 35S::GUS embryos showed comparable temporal and spatial distribution and intensity of GUS expression (data not shown), and that KNOLLE protein immunolocalisation did not reveal any difference between 35S::KNtransgenic and wild-type embryos.

To compare KNOLLE mRNA accumulation from 35S::KNtransgene expression with the immunolocalisation of misexpressed KNOLLE protein, we examined seedling roots by in situ hybridisation. Unlike the situation in the embryo, most cells of the seedling are mitotically inactive. Exceptions are the meristems of the shoot and the root as well as the primordia of leaves and lateral roots. The mature root of wild-type seedlings showed little or no distinct staining (Fig. 7I). By contrast, the root of transgenic seedlings gave strong signals in the expanding central cells(Fig. 7J) and also in tip-growing root hairs (Fig. 7K). These observations were consistent with the immunolocalisation of KNOLLE protein in transgenic roots and with the 35S::GUS expression pattern (see Fig. 3). In addition, strong expression of the 35S::KN transgene was observed in lateral root primordia (Fig. 7K), with patches of dark staining above a lighter background, resembling the situation in embryogenesis. Thus, the activity of the 35S::KN transgene yielded high levels of KNOLLE mRNA mainly in non-proliferating cells that did not express the endogenous KNOLLE gene.

Consistent with its role as a cytokinesis-specific syntaxin, Arabidopsis KNOLLE protein accumulates during M phase, relocates to the plane of cell division during telophase and disappears at the end of cytokinesis (Lauber et al.,1997). The transient accumulation of KNOLLE protein closely follows that of KNOLLE mRNA(Lukowitz et al., 1996),suggesting that synthesis and degradation of both mRNA and protein are regulated in a cell cycle-dependent manner. We have addressed the biological significance of this tight regulation.

By expressing KNOLLE in non-proliferating cells, we created an abnormal situation in which vesicles budding from the trans-Golgi could deliver KNOLLE to several potential target membranes. However, KNOLLE co-localised with the plasma membrane H⁺-ATPase in tip-growing root hairs, thus behaving like an integral plasma membrane protein. This result was confirmed by ultrastructural immunolocalisation of KNOLLE protein not only in root hairs but also in expanding cells of the central cylinder of the root. Thus, the plasma membrane appears to be the destination of KNOLLE protein in non-proliferating cells.

Mistargeting of KNOLLE to the plasma membrane did not interfere with essential cellular processes that are required for normal plant development. Tip-growing root hairs were morphologically indistinguishable between 35S::KN transgenic and wild-type plants, although only the former accumulated large amounts of KNOLLE protein in the apical growth zone. This observation suggests that KNOLLE did not obviously interfere with the interaction of SNARE complex partners involved in apical membrane fusion,which would be consistent with recent findings of yeast SNARE pairing specificity (McNew et al.,2000). Although we cannot rule out the fact that KNOLLE interacts with non-cognate SNARE partners without obvious deleterious effects, it is also conceivable that in non-proliferating cells, KNOLLE might be a biologically inactive passenger protein on vesicles destined to the plasma membrane.

Why does KNOLLE traffic to the plasma membrane in non-proliferating cells?As eukaryotic cells express several syntaxins, each of which resides in a distinct membrane compartment, there must be a sorting mechanism to ensure that each syntaxin is delivered to its proper destination. For example, the Arabidopsis syntaxin AtPEP12 is targeted to the vacuolar pathway (da Silva Conceicao et al., 1997). Although the mechanism of syntaxin sorting is unknown in plants, recent observations suggest that sorting occurs during vesicle budding from the trans-Golgi donor membrane in yeast. An acidic dileucine motif of the vacuolar t-SNARE Vamp3p appears essential for sorting mediated by the adaptor protein complex AP-3 (Darsow et al.,1998), whereas Golgi-associated coat proteins with homology to gamma adaptin appear to interact with a different sorting motif of Pep12p for its targeting to late endosomes (Black and Pelham,2000). By contrast, no AP3-dependent sorting motif has been identified in the plasma membrane syntaxins, Sso1 and Sso2 (Tang and Hong,1999), and KNOLLE lacks the consensus acidic di-leucine motif. If post-Golgi trafficking to the plasma membrane were a default pathway in the absence of specific targeting cues,mistargeting of KNOLLE protein might reflect the lack of such a sorting signal. This does not exclude the possibility that normal targeting of KNOLLE to the plane of cell division may involve a hypothetical sorting-signal receptor that is not present in non-proliferating cells. The existence of an active sorting mechanism for proteins destined to the cell plate has been hypothesised based on the behaviour of GFP-KOR, a GFP fusion to the Arabidopsis endo-1,4-β-D-glucanase KORRIGAN (Nicol et al.,1998), in a heterologous expression system (Zuo et al.,2000). KORRIGAN and KNOLLE share a YVDL sequence that may act an AP-dependent sorting motif, although its physiological significance and specificity remain to be determined (for a review, see Bonifacino and Dell'Angelica,1999).

Independently of specific sorting signals, a general redirection of membrane flow may be involved in plant cytokinesis. Supporting evidence comes from two recent observations. The Arabidopsis putative auxin efflux carrier PIN1, an integral membrane protein, which is normally located in the basal plasma membrane of non-proliferating vascular cells, also accumulates at the forming cell plate (Steinmann et al.,1999). Furthermore, a secreted enzyme, cell wall-associated endoxyloglucan transferase, has been reported to traffic to the plasma membrane during interphase and to the cell plate during cytokinesis via the endoplasmic reticulum-Golgi pathway in tobacco BY-2 cells(Yokoyama and Nishitani,2001). These observations suggest that the vast majority of vesicles budding from the trans-Golgi during M phase traffic to the plane of cell division. These vesicles may incorporate any membrane or soluble cargo protein that passes through the Golgi stacks at that time and lacks a retention signal. Accordingly, KNOLLE protein would not need a sorting motif. Whatever the underlying mechanism, only KNOLLE protein that is synthesised during M phase can be targeted to the plane of cell division. Consequently, the level of KNOLLE expression during M phase may be a crucial parameter for cytokinetic vesicle fusion.

The 35S::KN transgene yielded approximately hundred-fold accumulation of KNOLLE protein in seedlings, when compared with the wild-type control. This result is consistent with the common use of the 35S promoter for transgene overexpression in plants (Holtorf et al.,1996; Lermontova and Grimm,2000; Sentoku et al.,2000). However, the 35S::KN transgene did not complement knolle mutant embryos. One difference between developing embryos and seedlings is that embryo cells are proliferating, whereas most cells in the seedling are not. As shown by in situ hybridisation and immunostaining in seedling roots, 35S::KNtransgene expression resulted in the stable accumulation of KNOLLE mRNA and protein in non-proliferating cells. Comparative in situ hybridisation and immunostaining of 35S::KN transgenic and wild-type embryos revealed the relative strength of the 35S promoter in proliferating cells. Expression of the 35S::KN transgene, if at all detectable,supplemented the wild-type patchy pattern of KNOLLE mRNA accumulation by low-level accumulation of KNOLLE mRNA in the primordia of the cotyledons. The 35S promoter activity was independently assessed in 35S::GUS embryos, which gave a similar developmental expression pattern and level of expression as 35S::KN (data not shown). Our results are consistent with previous results demonstrating 35Spromoter activity only from the heart stage on in 35S::GUS transgenic tobacco embryos (Odell et al.,1994). Furthermore, no additional KNOLLE protein accumulation was detected in 35S::KNtransgenic embryos, when compared with wild-type embryos. Taken together,these observations suggest that the expression level of the 35S::KNtransgene in proliferating cells was insufficient to rescue knollemutant embryos.

The difference between the 35S::KN and the KNOLLE rescue(KNRescue) constructs was confined to the 5′ region, whereas both constructs contained the same genomic 3′ sequence. Thus, any difference in expression pattern and intensity between the two constructs can be attributed to different 5′ sequences, the KNOLLE cis-regulatory region as opposed to the 35S promoter. The KNOLLE 5′sequence appears to integrate signals that link KNOLLE expression to the cell cycle as indicated by the patchy pattern of mRNA accumulation. Promoter elements conferring M phase-specific transcription have been identified in mitotic cyclin genes (Ito et al.,1998). The KNOLLE5′ UTR should also contain a sequence that enables translation of the mRNA during M phase, when most protein expression is shut down (for a review,see Sachs, 2000). However,KNOLLE expression is not strictly linked to karyokinesis, as KNOLLE protein accumulates between non-mitotic nuclei during endosperm cellularisation(Lauber et al., 1997; Otegui and Staehelin, 2000).

In contrast to the KNOLLE promoter, the 35S promoter appears to be active in a cell cycle-independent manner. KNOLLE mRNA accumulated stably in non-proliferating cells of 35S::KN transgenic seedlings but only transiently in proliferating cells. Instability of short-lived mRNAs has been attributed to specific sequences in the 3′UTR (Gutierrez et al., 1999;Sachs, 2000). In the case of KNOLLE mRNA, an as yet undefined degradation signal appears to be linked to the M phase and/or cytokinesis. The lack of strong accumulation of KNOLLE mRNA in proliferating cells of 35S::KN transgenic embryos and seedlings suggests that the activity of the 35S promoter is too low in proliferating cells or does not counteract efficiently the cell cycle-dependent mRNA degradation. By contrast, the endogenous KNOLLEpromoter is strong enough to yield high levels of mRNA and protein accumulation in proliferating cells, although it is only active during a brief period of the cell cycle. Thus, during its period of activity, the endogenous KNOLLE promoter is clearly stronger than the 35S promoter,and this difference appears to be crucial for the execution of cytokinesis.

In summary, there is no obvious need for the observed tight regulation of KNOLLE expression, provided enough KNOLLE protein is available during M phase to ensure efficient vesicle fusion during cell-plate formation. This demand is met by the strong KNOLLE promoter, which is highly active during the period preceding cytokinesis. If there were no cell cycle-dependent degradation of KNOLLE mRNA and protein, both would stably accumulate. Although mistargeting of KNOLLE protein to the plasma membrane appears not to be harmful, it is also not useful. Perhaps proliferating cells gain some selective advantage from linking the degradation of KNOLLE mRNA and protein, as well as the promoter activity, to the cell cycle.

We thank W. Lukowitz for kindly providing the cosmid 74-4, G. Cardon for the pBarA and pBar-35S vectors and the Agrobacterium strain GV3101,T. Jack for the pAP3 vector, U. Mayer for the knolle allele UU1319,and W. Michalke for the monoclonal anti-plasma membrane H⁺-ATPase antibody. We also thank S. Mangold and T. Pacher for help with the immunoblotting, and N. Geldner, M. Heese, M. Lenhard, Jaideep Mathur, U. Mayer and K. Schrick for critical reading of the manuscript. This work was supported by grant SFB 446/B-8 from the Deutsche Forschungsgemeinschaft.