Exocyst subunit Sec6 is positioned by microtubule overlaps in the moss phragmoplast prior to cell plate membrane arrival

The physical separation of two daughter cells formed during plant cell division occurs through a transient, disk-shaped membrane compartment that expands radially towards the parental cell wall. This membrane compartment is termed the cell plate and its construction culminates in a new cell wall segment dividing two individual plasma membranes (Drakakaki, 2015; Müller and Jürgens, 2015; Smertenko et al., 2017). Cell plate initiation and radial expansion rely on the fusion of vesicles that are supplied mainly by the secretory pathway (McMichael and Bednarek, 2013; Richter et al., 2014; Boruc and Van Damme, 2015). Adaptations of canonical trafficking mechanisms are, however, required because there is no pre-existing target membrane at the site of cell division to which vesicles can fuse. Instead, membrane deposition is thought to be initiated by ‘homotypic’ fusion of vesicles (Smertenko et al., 2017). This raises the question of how vesicle fusion is spatially restricted to the site of cell division, instead of it occurring spuriously throughout the whole cell. It has been proposed that vesicles are transported along polarized microtubules to the center of the phragmoplast, a cytoskeletal apparatus that supports cell plate assembly (Euteneuer et al., 1980; Lee et al., 2001; Otegui et al., 2001). Transport may concentrate vesicles locally to enhance fusion rates but whether transport alone can provide the spatial accuracy required to build a straight and flat cell plate is unknown. Recently, we have identified short stretches of antiparallel microtubule overlap at the midzone of phragmoplasts in the moss Physcomitrella patens as sites where membrane build-up is initiated (de Keijzer et al., 2017). It remained, however, unclear whether there are specific molecules present at overlaps, which then trigger vesicle fusion locally.

In eukaryotic cells the fusion of transport vesicles with endomembrane compartments and the plasma membrane relies on the combined action of fusion and tethering complexes. The force driving the fusion of a vesicle and its destination membrane is almost universally generated by soluble N-ethylmaleimide-sensitive factor attachment receptor (SNARE) complexes, which are typically composed of four membrane-associated proteins (Söllner et al., 1993; Wickner and Schekman, 2008; Südhof and Rothman, 2009). However, tethering – the establishment of the initial physical connection between the two membranes – is orchestrated by a multitude of molecular machinery, often in the form of multimeric protein complexes (Koumandou et al., 2007; Vukašinović and Žárský, 2016). To enable targeted trafficking among the various distinct membrane compartments in eukaryotic cells, membranes acquire different identities that dictate which membrane fusion reactions are allowed. Although the SNARE complex composition bestows some of this specificity (Paumet et al., 2004) other factors, including Rab GTPases (Grosshans et al., 2006; Stenmark, 2009) and the various tethering complexes, play an important role as well (Yu and Hughson, 2010). Indeed, each of the several tethering complexes typically facilitates the docking of vesicles at a specific target membrane (Vukašinović and Žárský, 2016; Yu and Hughson, 2010).

While work in Arabidopsis has uncovered much about the composition, trafficking and function of the SNARE complexes involved in fusing cell plate vesicles (reviewed in Müller and Jürgens, 2015; Jürgens et al., 2015), comparatively little is known about the functioning of tethering factors involved. In Arabidopsis, both the transport protein particle II (TRAPPII) and exocyst tethering complexes contribute to cell plate biogenesis (Thellmann et al., 2010; Fendrych et al., 2010; Qi et al., 2011; Rybak et al., 2014). Both are multimeric protein complexes that function in post-Golgi membrane trafficking (Drakakaki et al., 2012; Fendrych et al., 2010; reviewed by Vukašinović and Žárský, 2016). Interestingly, whereas TRAPPII is associated with the cell plate throughout its formation, the exocyst is present on the cell plate during cell plate initiation and, thereafter, its abundance drops during cell plate expansion and increases again after cell plate attachment (Fendrych et al., 2010; Zhang et al., 2013; Rybak et al., 2014). Nonetheless, the two tethering complexes physically interact and are most likely to function cooperatively during cytokinesis (Rybak et al., 2014). Whereas functional analysis of the TRAPPII complex is expedited by only single genes encoding two unique TRAPPII subunits (Thellmann et al., 2010; Qi et al., 2011), the diverse complement of genes encoding exocyst subunits have made dissecting the role of the exocyst comparatively cumbersome. Nonetheless, several exocyst mutants have been identified that harbor cell plate defects (Fendrych et al., 2010; Wu et al., 2013; Rawat et al., 2017).

Here, to understand whether the exocyst is a cell-plate assembly factor active on microtubule overlaps, we monitored the localization of several GFP-tagged exocyst subunits during cytokinesis in the moss P. patens. Although this representative of a basal land plant lineage has seen a similar degree of gene amplification regarding exocyst subunit genes as Arabidopsis – albeit with a notable smaller radiation of Exo70 paralogs (Cvrčková et al., 2012) – it features a main haploid phase in its life cycle, which expedited genetic analysis. Since cell plate-assisted cytokinesis is a hallmark of all land plants (Buschmann and Zachgo, 2016) and a comprehensive toolset to study cell division exists for P. patens (Yamada et al., 2016), we anticipated that this model plant is particularly useful for studying exocyst functioning in relation to the fusion between cell plate membranes. We show that the exocyst subunit Sec6, but not subunits Sec3 and Sec5, localized to microtubule overlaps at the phragmoplast midzone prior to the arrival of membrane vesicles. The localization of Sec6 depended on the presence and size of the microtubule overlaps. Our findings make the case that, in general, spatial control over homotypic vesicle fusion can be achieved by linking tethering activity to a pre-defined, non-membranous subcellular structure.

The eukaryotic exocyst has been best studied for its role in the targeting and fusion of exocytotic vesicles to the plasma membrane. A key step is thought to be the formation of a fully assembled complex between the vesicle- and plasma membrane-associated subunit subsets (e.g. Yu and Hughson, 2010; Boyd et al., 2004). How exocyst assembly proceeds in absence of a plasma membrane in the context of homotypic fusion is unknown. To characterize assembly dynamics during cytokinesis in P. patens, we generated a time-resolved localization map of selected exocyst subunits in protonemal tip cells that exhibit repetitive cell divisions. Sec6 was included in the screen because it is the only exocyst subunit that is represented by a single gene in the moss P. patens (Cvrčková et al., 2012). The protein was, therefore, expected to be part of all fully assembled exocyst complexes. Moreover, Sec6 from Arabidopsis has been shown to interact with vesicle fusion machinery and might, thus, be an important regulator of cell plate formation (Wu et al., 2013). To investigate whether exocyst complex assembly occurs in sequential steps, we included Sec5 and Sec3 paralogs, which – in yeast – have been proposed to be associated to the vesicle and plasma membrane, respectively (He and Guo, 2009). Sec5 is encoded by four genes and Sec3 by three genes in P. patens (Fig. S1) (Cvrčková et al., 2012). To conduct the screen, a GFP-encoding fragment was integrated at the end of the single Sec6 gene, at the end of all three genes encoding Sec3 paralogs, and at the end of the three Sec5 paralogs that are expressed in protonemal tissue and/or in caulonemal and/or chloronemal cells (Ortiz-Ramírez et al., 2016) (Fig. S1A). We started by analyzing the relative abundance of the different isoforms by imaging interphase apical caulonemal cells (Fig. S1B). Image-based assessment of relative protein abundance matched the data available from gene expression databases (Fig. S1A). All investigated exocyst subunits, except the less strongly expressing Sec3c, localized to plasma membrane foci as reported previously, and some – notably Sec6 and Sec5b – were also abundant in the cytoplasm (Fendrych et al., 2013; Zhang et al., 2013; Vukašinović and Žárský, 2016; Bloch et al., 2016; Synek et al., 2017; van Gisbergen et al., 2018). Of those subunits that showed brightest cortical localization, i.e. Sec3a, Sec5b, Sec5d and Sec6 (Fig. S1), fusion proteins were expressed together with mCherry-tagged α-tubulin, such that exocyst localization dynamics could be related to mitotic progression (Fig. S1, Fig. 1A). All selected exocyst subunits and, additionally, the established cell plate membrane marker SCAMP4 (de Keijzer et al., 2017) were imaged throughout cell division. The onset of anaphase was used as temporal reference to aid the mutual comparison of arrival times at specific cytokinetic structures (Fig. 1A, Movie 1).

The selected exocyst subunits showed disparate localization patterns during cell division. Sec3a and Sec5b localized to the phragmoplast midzone during cell plate initiation (Fig. 1A, Movie 1). Unlike the exocyst subunits studied in Arabidopsis (Fendrych et al., 2010; Zhang et al., 2013; Rybak et al., 2014), no sharp drop in intensity occurred during cell plate expansion. However, similar to the observations in Arabidopsis, the amount of Sec3a and Sec5b on the new wall facet transiently increased after cell plate attachment, while that of Sec5d deviated from this localization and was absent from the cell plate until late cytokinesis. Moreover, Sec6 was detected already at the center of the division apparatus from spindle formation onwards (Fig. 1A, white arrowheads; Movie 1). Unlike the other studied exocyst subunits, the intensity of Sec6 at the phragmoplast midzone peaked during cell plate initiation and expansion, and dropped after cell plate attachment. Imaging of Sec6 in conjunction with FM4-64, an amphiphilic dye for the endomembrane system, showed that a distinct pattern of bright Sec6 sections was visible in the phragmoplast midzone before initial deposits of cell plate membrane appeared ∼2 min after anaphase onset (Fig. 1C, Movie 2; de Keijzer et al., 2017). Interestingly, all four selected subunits were already stably associated with the cell cortex before cell plate attachment at a site where the cell division plane intersected with the parental cell wall (Fig. 2A,B). To our knowledge such cortical localization pattern has not been observed during cytokinesis for any labeled subunit in Arabidopsis (Fendrych et al., 2010; Zhang et al., 2013; Rybak et al., 2014). The early cortical labeling was punctate, with several puncta making up a 2–3 μm wide band that formed a continuous ring at the cell cortex (Fig. 2A). The band was mobile on the cortex until cell plate attachment and its movement was synchronous with rotation of the phragmoplast in these cells (Fig. 2B,D).

To further characterize the observed differences in localization, we identified those moments – with respect to anaphase onset – at which subunits first became visible at the phragmoplast midzone and the cortex (Figs 1B and 2C). This analysis showed that all studied subunits arrived at the cortex at around the same time (5–10 min post anaphase onset). The average time of arrival at the phragmoplast midzone, however, diverged for the studied subunits. Sec6 was already visible when the spindle midzone transformed into the phragmoplast midzone at the onset of anaphase. Sec3a and Sec5b, however, appeared simultaneously with the initial membrane accumulations, and Sec5d only appeared ∼23 min later (Fig. 2, Movies 1, 8, 9). The more weakly expressing paralogs Sec3b-GFP and Sec5a-GFP also appeared at the midzone at the time of initial membrane accumulation (Movie 2). Thus, whereas sharing localization at the cortex and late cell plate with the other studied exocyst subunits, Sec6 exhibited localization to the phragmoplast and spindle midzones apparently independent of the other studied exocyst complexes and membrane compartments.

The early localization pattern of Sec6 showed strong resemblance to sites at which microtubules from opposite poles form antiparallel overlaps in the spindle and phragmoplast midzone (Ho et al., 2011; Kosetsu et al., 2013; de Keijzer et al., 2017). To investigate their interdependency, we generated a moss line expressing Sec6-mCherry together with a Citrine-tagged version of the antiparallel microtubule bundling protein MAP65 (Kosetsu et al., 2013). Sec6 localization, indeed, closely followed the distribution of MAP65-Citrine labeled regions (Fig. 3, left panels; Movie 3). Strikingly, this behavior was even visible during prophase, when the nuclear envelope is still intact and no bipolar spindle has yet formed (Fig. 3, arrowheads). We have previously generated cells lacking the regulator of microtubule dynamics Kin4-Ic (Δkin4-Ic cells), which show delayed shortening of microtubule overlaps at anaphase onset (de Keijzer et al., 2017). In this genetic background a similarly strong colocalization of MAP65 and Sec6 was observed during prophase and spindle stages. The width of the Sec6 patches after anaphase onset was larger in Δkin4-Ic cells, which is in accordance with the larger overlaps. As cytokinesis progressed, Sec6 increasingly localized to only the central region of MAP65, resembling membrane depositions that also gradually became more confined to overlap centers (Fig. 3, arrows in right panel; Movie 3; de Keijzer et al., 2017). This behavior was best visible before overlaps shortened in Δkin4-Ic cells, i.e. ∼5 min after anaphase onset. Thus, Sec6 localizes to MAP65 regions but, after cell plate initiation, other factors might additionally affect its localization.

To further understand the extent to which the presence of microtubule overlaps is a prerequisite for the localization of Sec6 during cytokinesis, we silenced the three MAP65 genes expressed in protonema. Earlier studies have demonstrated that, upon knockdown of MAP65, antiparallel microtubule overlap formation in the phragmoplast is severely compromised, ultimately leading to phragmoplast collapse and cytokinesis failure. Nonetheless, during early cytokinesis cell plate membrane does aggregate and phragmoplast microtubules remain organized as two opposing sets (Kosetsu et al., 2013). Whereas in control cells Sec6-GFP appeared to be regularly distributed over the division plane during early cytokinesis, in MAP65-silenced cells Sec6-GFP formed discontinuous patterns (Fig. 4; Movie 4). There was no obvious correlation between the appearance of gaps in the Sec6 localization pattern and the presence of microtubules; areas devoid of Sec6 were visible both in regions populated and unpopulated by microtubules (Fig. 4B). Localization of Sec6 to the phragmoplast midzone, therefore, not just correlates with the presence of bipolar microtubules but requires an ordered array of microtubule overlaps.

A physical interaction between Sec6 and Sec1/Munc18 (SM) family protein KEULE has been proposed to be an important regulatory step in cytokinetic vesicle fusion in Arabidopsis (Wu et al., 2013). KEULE contributes to fusion by preventing the important cytokinetic SNARE-component KNOLLE from refolding into its closed, non-fusion-competent conformation, thereby allowing the formation of fusogenic trans-SNARE complexes among vesicles delivered to the cell plate (Park et al., 2012; Karnahl et al., 2017; Jürgens et al., 2015). Whether the Sec6 interaction regulates this specific activity is presently unknown. Nonetheless, cytokinesis defects were found in Sec6 mutants that resemble these encountered in plants lacking KEULE (Wu et al., 2013; Assaad et al., 1996), hinting that a cooperative Sec6-KEULE interaction regulates cytokinesis. On the basis of these findings, we hypothesized that Sec6 on overlaps in moss regulates vesicle fusion activity in space through an interaction with KEULE. In search of a gene homologous to KEULE, we identified seven loci in the P. patens genome that are predicted to encode SM family proteins. Of those loci, two comprised genes with a predicted exon number identical to that of the Arabidopsis KEULE (AtKEULE) gene, i.e. Pp3c17_24130 and Pp3c16_570; and their expected gene products exhibited an overall similarity to that of AtKEULE that was at least twice as high as for the other predicted SM genes (Fig. 5A). Analysis of gene expression levels revealed that the gene encoded at locus Pp3c17_24130 was ubiquitously expressed in moss tissues, whereas the other gene (at Pp3c16_570) was only expressed in rhizoids (Fig. 3A and Fig. S2) (Ortiz-Ramírez et al., 2016). We, therefore, focused on the former gene and tentatively named it P. patens KEULE (PpKEULE). Since the position of introns in genes contains information on the evolutionary trajectory of the gene family they belong to (e.g. Rogozin et al., 2003; Garcia-España et al., 2009; Javelle et al., 2011), we compared the intron–exon structure between AtKEULE and PpKEULE. The intron positions were highly similar for both species (Fig. 5B). Introns in the PpKEULE gene were slightly longer; however, this is in line with increased average intron length in moss compared to that of Arabidopsis (Rensing et al., 2005). The concurrent intron–exon structure of the genes of both species indicates that PpKEULE is likely to be orthologous to AtKEULE.

In Arabidopsis, the interaction domain of KEULE with Sec6 was narrowed down to a C-terminal portion of the protein designated C1 (Wu et al., 2013). The high degree of sequence conservation allowed us to delineate the same domain in PpKEULE and test it for its ability to interact with Sec6. In a reciprocal yeast two-hybrid assay with P. patens Sec6, KEULE (PpKEULE) and the C1 domain of KEULE (PpKEULE-C1), we found a strong interaction between Sec6 and both PpKEULE and PpKEULE-C1 when Sec6 was fused to the activation domain of the Gal4 transcription factor. When Sec6 was fused to the DNA-binding domain of Gal4 only a weak interaction with the PpKEULE-C1 domain was found (Fig. 5C). Possibly, in this case, the binding partners are in a less favorable configuration to reconstitute Gal4 function since a similar result was reported for the interaction of Arabidopsis proteins (Wu et al., 2013). Because PpKEULE interacts with Sec6 in a fashion indistinguishable from that of its ortholog in Arabidopsis, the two proteins might form a conserved module regulating cell plate membrane fusion. To investigate where and when in P. patens Sec6 and PpKEULE might interact, we visualized Sec6-mCherry together with endogenous PpKEULE tagged with GFP (Fig. S3). We chose to fuse GFP to the C-terminal end, since this yielded a functional fusion protein for AtKEULE (Steiner et al., 2016). Although absent from Sec6-labeled regions during prophase and metaphase, PpKEULE-GFP became rapidly enriched at the site of Sec6-mCherry at the onset of cytokinesis around the time that localized membrane deposition is first observed at overlaps (Fig. 5D; Movie 5). During radial expansion of the phragmoplast, appearing sites of Sec6 localization at the leading zone, which are likely to correspond with newly formed microtubule overlaps with yet little accumulated vesicles (de Keijzer et al., 2017), had low PpKEULE-GFP signal associated (Fig. 5D, arrowheads). At the cortical zone marked by the labeled exocyst subunit no corresponding PpKEULE signal was found (Fig. 5D, arrow). Taken together, the colocalization pattern suggested that physical association between Sec6 and PpKEULE in a cytokinetic context is both spatially and temporally regulated.

To investigate the functional importance of Sec6 for correct KEULE localization and cell plate assembly in general, we sought out ways to reduce or abolish cellular Sec6 levels. We failed to isolate a Sec6 knockout mutant, suggesting that Sec6 is an essential protein for cell proliferation – as it has been reported in RNA interference (RNAi) studies by van Gisbergen et al. (2018). We, therefore, introduced inducible small interfering RNA (siRNA) of Sec6 in cells expressing either the cell plate membrane marker SCAMP4-mCherry or PpKEULE-GFP. RNAi of Sec6 caused an ∼80% reduction of moss colony expansion over a 5-day growth period and an absence of caulonemal cells (Fig. S4A,C). Continuation of colony expansion suggested that some functional exocyst complexes remain present after initiating RNAi. This was in agreement with an incomplete (∼3-fold) reduction of transcript levels after RNAi (Fig. S5B). When using RNAi in which different regions of the Sec6 gene were targeted in chloronemal cells, we reproducibly found a more-bulbous tip shape, suggesting that exocyst functioning in polarized tip growth was severely disrupted but not completely abolished (Figs S4C and S6). Three days after induction of knockdown using RNAi, the final cell walls separating two daughter cells showed several morphological defects (Fig. S4D). Incomplete cytokinesis could, however, not be concluded without ultrastructural analysis of the formed cell plates. To understand whether defects arose due to exocyst malfunctioning during cell plate initiation and expansion, we observed cells expressing SCAMP4-mCherry as a membrane marker throughout cytokinesis (Fig. 6A). Although the time of initial membrane appearance was similar, the rate at which membrane material accumulated was delayed after initiating RNAi (Fig. 6B and Movie 6). RNAi experiments, thus, suggested that reduction of Sec6 functioning at the midzone decreased membrane build-up during early cytokinesis, a time at which Sec6 levels peaked in unperturbed cells (Fig. 1A). Surprisingly, the velocity of cell plate expansion appeared similar under both induced and non-induced conditions (Fig. 6D). To test whether silencing of Sec6 affects KEULE levels at the phragmoplast midzone, we imaged PpKEULE-GFP after initiation of RNAi. A delay in the recruitment of PpKEULE-GFP to the midzone was observed after induction with similar dynamics as SCAMP4-mCherry (Fig. 6E,F; Movie 7). This suggested that KEULE accumulation to the phragmoplast midzone was, at least partly, dependent on Sec6, which is in agreement with the observed interaction between KEULE and Sec6 (Fig. 5).

Our cytokinesis-oriented localization screen of exocyst subunits in moss revealed two localization patterns that have not been previously described in plants. First, exocyst subunits localized to MAP65-labeled regions in the midzone of the spindle and phragmoplast. Second, they formed a cortical ring prior to cell plate attachment. Interestingly, the two localization patterns overlapped in time yet differed in subunit composition. In the midzone, Sec6 arrival preceded that of Sec5 and Sec3 homologs, although this was not evident for the cortical localization. The systematic design of the conducted screen, thus, provides some of the clearest evidence to date for sequential assembly of the exocyst complex. So, although biochemical purification and live cell data show that the eight exocyst subunits in yeast form a stable complex, our work shows that subunits can as well have individual localization patterns (Hála et al., 2008; Heider et al., 2016; Picco et al., 2017).

In most plant cells a cortical ring of microtubules, the preprophase band (PPB), is thought to demarcate a functionalized section of the cortex, called the cortical division zone (CDZ), to which cell plate expansion is directed during cytokinesis (van Damme, 2009; Müller et al., 2009; Stöckle et al., 2016). Known components of the CDZ do, however, assemble into ring-like structures in mutants that are unable to form a PPB (Spinner et al., 2010). Other cells, including moss protonemal cells, lack a PPB altogether but cytokinesis does involve the formation of a cortical band with established CDZ components, including myosins and kinesins, prior to cell plate attachment (Schmiedel et al., 1981; Doonan et al., 1985; Otegui and Staehelin, 2000; Hiwatashi et al., 2008; Wu et al., 2011; Nakaoka et al., 2012; Miki et al., 2014; Wu and Bezanilla, 2014; Lipka et al., 2014; Schaefer et al., 2017; Kosetsu et al., 2017). Given the early cortical localization of the exocyst observed in moss, it will be interesting to find out whether exocyst and vesicle trafficking have a role in the establishment or maintenance of the CDZ in absence of a PPB, or whether its function is limited to attachment of the cell plate. Analogies may be present in fission yeast cytokinesis, where exocyst localization and vesicle fusion to the division site appear to precede ingression of the cleavage furrow (Wang et al., 2016). It is surprising that no exocyst subunits have been reported as components of the CDZ in higher plants (Vukašinović and Žárský, 2016; Boruc and van Damme, 2015). Characterization of exocyst localization in plants is, however, impeded by the large number of subunits homologs, particularly the Exo70 subunit.

Local accumulations of exocyst subunits were first described for their role in polarized exocytosis at the plasma membrane (TerBush and Novick, 1995; Hazuka et al., 1999). However, exocyst subunits also localize to internal membrane compartments to assist in autophagosome formation (Kulich et al., 2013; Bodemann et al., 2011). Moreover, in budding and fission yeast it has been demonstrated that the deliberate targeting of exocyst subunits to mitochondria reroutes cellular trafficking towards these organelles (Luo et al., 2014; Wang et al., 2016). These observations lead us to propose that, in moss, a local pool of exocyst subunits on microtubule overlaps stimulates the immobilization and fusion of vesicles to initiate cell plate assembly. However, in all mentioned examples, local pools of exocyst subunits are bound to membranes, whereas – in our study – Sec6 localized to non-membranous microtubule overlaps. It is, therefore, not self-evident how Sec6 on overlaps can initiate membrane-membrane interactions necessary for homotypic vesicle fusion, even when it seeds the assembly of a complete exocyst complex. Possibly, vesicles arrive at the phragmoplast midzone in a fusion-incompetent state and formation of enabled SNARE complexes involves interactions between SNAREs, KEULE and exocyst subunits. In fact, Sec6 appears to be at the center of an interaction network because Sec6 homologs in diverse systems interact with both SNARE and SM proteins separately from other exocyst subunits (Hong and Lev, 2014; Hashizume et al., 2009; Morgera et al., 2012; Dubuke et al., 2015; Sivaram et al., 2005, 2006). Cytosolic levels of Sec6 might be insufficient to establish these regulatory molecular interactions on vesicles, whereas concentrated Sec6 on overlaps could generate a local pool of fusion-competent vesicles near overlaps. Our finding, thus, suggests how vesicle fusion activity may be regulated in space, an aspect that – so far – remains unresolved. Interestingly, PpKEULE does not localize to cortical exocyst subunits prior to cell plate attachment. It might, thus, be exclusively required to regulate vesicle fusion activity in the phragmoplast midzone during cytokinesis.

Unique about Sec6 during moss cytokinesis was its initial localization to the phragmoplast midzone in complete absence of membranes. It is possible that other exocyst subunits share this localization pattern but not Sec5 and Sec3 – which arrived later, around the time of initial membrane accumulation. Colocalization of MAP65 and Sec6, at a stage during which MAP65 is localized to visible regions of microtubule overlap in the phragmoplast midzone (Fig. 3; Kosetsu et al., 2013), demonstrated that Sec6 first associates along microtubule overlaps. Observations in mammalian cells suggest that microtubule overlaps have a more common role in controlling vesicular transport. Through interaction with centralspindlin the protein MICAL3 first localizes to the spindle midzone starting in anaphase and later targets Rab8a-positive vesicles to the midbody (Liu et al., 2016). Centralspindlin does, however, transfer from microtubule overlaps to an adjacent exocyst-containing ring-shaped bulge (Hu et al., 2012; Gromley et al., 2005). It is, therefore, unclear how intimate the spatial association is in these cells between microtubule overlaps and vesicles (Green et al., 2013). Moss was not shown to have centralspindlin but, MAP65 itself or other midzone-associated proteins, might have a function in Sec6 recruitment to microtubule overlaps. Physical interactions between cytoskeletal regulators and exocyst subunits have been reported in various other systems (van Gisbergen et al., 2018; Gromley et al., 2005; Zuo et al., 2006; Oda et al., 2015). Our MAP65 silencing and Δkin4-Ic experiments clearly showed that the length and presence of microtubule overlaps affected Sec6 localization. Sec6, however, also localized to regions of the phragmoplast midzone that were devoid of MAP65 during later stages of cytokinesis, and Sec6 localized predominantly to the center of long MAP65 regions in Δkin4-Ic cells (Fig. 3). We therefore hypothesize that Sec6 and, possibly, other proteins that are initially on microtubule overlaps are gradually incorporated into another structure to enable vesicle fusion events away from microtubule overlaps. This structure might represent the cell plate assembly matrix (CPAM), which is visible in EM micrographs and from which microtubule overlaps are excluded (Seguí-Simarro et al., 2004). Profound defects in cell plate assembly in cells lacking Kin4 do, however, show that insufficient confinement of initial cell plate depositions around microtubule overlaps is not resolved by successive microtubule-independent processes (de Keijzer et al., 2017).

Our findings provide the first indication that molecules that drive membrane fusion processes are, at least temporarily, associated with microtubule overlaps during plant cytokinesis. The formation of intracellular septa is complex, requiring spatial control over vesicle fusion to produce a new plasma membrane. The involvement of multiple tethering complexes has been reported in several eukaryotes (see Neto and Gould, 2011; Wang et al., 2016; Rybak et al., 2014). Redundancy in the functioning of these complexes might, however, hinder their functional characterization. It will be interesting to learn how the positioning of tethering complexes can fine-tune cell plate formation at the ultrastructural level. Differential localization of TRAPPII and exocyst complexes has been reported during fission yeast cytokinesis, and the two complexes in Arabidopsis were alternatingly present at the developing cell plate (Wang et al., 2016; Rybak et al., 2014). The molecular and light microscopic tools available in moss, together with the identification of overlaps as sites of vesicle accumulation, will help to resolve how multiple tethering complexes together shape the cell plate.

All plasmids used throughout this study are listed in Table S2. To construct GFP- and/or mCherry-tagged constructs of Sec6, Sec3a, Sec3b, Sec3c, Sec5a, Sec5b, Sec5d and PpKEULE, regions of ∼1 kb before and after the stop codon of their encoding gene were amplified by PCR using primers listed in Table S1. The PCR fragments were digested with restriction enzymes indicated in Table S1 and ligated into correspondingly digested peGFP-NPTII or pmCherry-LoxP-BsdR vector (de Keijzer et al., 2017).

For RNAi induced by β-estradiol to target the Sec6 transcript, we used the system published by Nakaoka et al. (2012). A ∼500 bp fragment of the coding sequence of Sec6 (Fig. S5) was amplified by PCR from a cDNA library derived from protonemal tissue and cloned into gateway entry plasmid pENTR-D-TOPO. The fragment was subsequently introduced into silencing vector pGG626 via a Gateway LR reaction.

To generate the expression constructs used for yeast two-hybrid assays, first the coding sequences of Sec6, PpKEULE, and the PpKEULE-C1 domain were PCR-amplified from a P. patens cDNA library derived from protonemal tissue. The PCR products were then introduced into the pENTR-D-TOPO vector and subsequently subcloned into destination plasmids pDEST22 and pDEST32 (Invitrogen) via Gateway LR reactions.

For transient small interfering RNA (siRNA) constructs used in RNAi, the region used for induced silencing as described above and regions located at the 5′-untranslated region and the 3′-end of the coding sequence were amplified and cloned into pENTR-D-TOPO (Fig. S6). The fragments were then introduced into the siRNA vector pUGGi (Bezanilla et al., 2005) via a Gateway LR reaction.

P. patens tissues were routinely grown on BCDAT plates under continuous light. Plasmids were linearized and introduced into the P. patens genome by homologous recombination using PEG-mediated protoplast transformation (Nishiyama et al., 2000). Correct insertion events were characterized by PCR (Fig. S3). RNAi was initiated using β-estradiol treatment and knockdown was verified by quantitative RT-PCR 3 days (Fig. S5). Characteristics of generated moss lines and their use throughout the study are summarized in Table S3. For imaging, protonemal tissue was grown for 5–7 days on BCD medium in glass-bottom dishes (Yamada et al., 2016).

For transient silencing by RNAi, siRNA constructs were transformed into the SNAP-TM-mCherry moss line expressing nuclear targeted GFP-GUS (Van Gisbergen et al., 2018). The nuclear GFP signal was used as a visual marker to identify silenced tissues. Filaments with decreased GFP intensity were selected for analysis (Fig. S6).

Proteins of the Sec1/Munc18 (SM) family were identified by BLAST using the AtKEULE protein sequence as input against the predicted proteins of P. patens genome assembly versions 1.6 and 3.1 (Fig. 5A), and by keyword search in the Phytozome 10.3 browser (www.phytozome.org). Expression in protonema was assessed by verifying the presence of ESTs derived from protonemal tissue and using the Physcomitrella eFP browser (Ortiz-Ramírez et al., 2016).

Yeast two-hybrid assays were performed with a split Gal4 transcription factor system using the His3 gene as reporter (James et al., 1996). For this, pDEST22/32-based constructs (Table S2) were transformed into yeast strain PJ69-4a or PJ69-4α by PEG-mediated transformation. Positive transformants with minimal background reporter activity were selected on synthetic drop-out (SD) medium lacking Leu and His or Trp and His, with different concentrations of 3-amino-1,2,4-triazole (3-AT) to increase histidine-dependent growth stringency. Selected clones were then allowed to mate and resulting diploids were selected on plates lacking Leu and Trp. With surviving cells a yeast two-hybrid assay was then performed on SD medium lacking Leu, Trp and His and in the presence of increasing amounts of 3-AT (1, 2.5 or 15 mM).

All live cell imaging was performed on a Roper spinning disk microscope system composed of a Nikon Ti eclipse body, Yokogawa CSU-X1 spinning disc head and Photometrics Evolve 512 camera or Prime 95B camera (Movies 8 and 9). All imaging was conducted with a 100× Plan Apo VC oil immersion objective (NA 1.40), using a 1.2× post-magnification fitted before the camera. The GFP and citrine probes were excited using 491 nm light generated by a Cobolt Calypso50 laser and their emitted light was bandpass filtered at 497–557 nm. For FM4-64 and mCherry 561 nm excitation light generated by a Cobolt Jive50 laser was used in combination with bandpass filtering at 570–620 nm. During image digitization a camera electron multiplication gain of 300 was employed and typical exposures were 800–1000 ms for GFP, citrine and FM4-64, except for Movies 8 and 9, where we used 3000 ms exposure times. For the mCherry probe, exposure times of 1000–2000 ms were used. FM4-64 was dissolved in dH₂O at a final concentration of 10 μM and added to cells at the moment of nuclear envelope breakdown as described (Kosetsu et al., 2013).

FIJI (Schindelin et al., 2012) was used for all image analysis and processing. Figures were prepared in Adobe Illustrator CC 2015. The arrival time of labeled exocyst subunits to the phragmoplast midzone (Fig. 1B) was taken as the time at which structural features with a residence time of more than one time frame first appeared in the generated maximum projections of three image planes. The appearance was abrupt and synchronously over the full length of the midzone. The arrival time of labeled exocyst subunits to a cortical band (Fig. 2C) was defined as the time at which stable cortical patches, visible for more than one time frame, first became visible in the region where the division plane intersected the parental cell wall. These patches were distinctly less dynamic compared to cortical localizations of the exocyst elsewhere in the cell.

To quantify the accumulation of cell plate membrane material and PpKEULE upon RNAi to silence Sec6, we imaged SCAMP4-mCherry and PpKEULE-GFP throughout cytokinesis at 30 s intervals. Images were taken in the central plane of the dividing cell at three confocal planes spaced 0.5 µm apart. Maximum projections made along the z-axis were then used to measure the average intensity of SCAMP4-mCherry and KEULE-GFP using a 5-pixel wide line manually drawn along the division plane for each time point. Simultaneously, the mean level of cytosolic background fluorescence was recorded in an area right next to the cell division site. For each time point, the background fluorescence level was subtracted from the obtained level of fluorescence at the division plane. The expansion rate of the cell plate (Fig. 6D) was calculated as described by de Keijzer et al. (2017).

For gene annotation, Cosmoss (www.cosmoss.org/) and the Phytozome 10.3 browser (www.phytozome.org) were used. Gene expression data were obtained from the Physcomitrella eFP browser (bar.utoronto.ca/efp_physcomitrella/cgi-bin/efpWeb.cgi).

We thank Henk Kieft for technical assistance, Richard Immink and Marco Busscher (WUR-Bioscience, Wageningen, The Netherlands) for assistance with Y2H experiments and Magdalena Bezanilla (Darthmouth College, Hanover, NH) for the kind gift of the pUGGi plasmid.