ADF10 shapes the overall organization of apical actin filaments by promoting their turnover and ordering in pollen tubes

The actin cytoskeleton undergoes constant assembly and disassembly and forms distinct structures within cells to fulfill diverse physiological cellular functions. Pollen tube growth depends on a dynamically remodeled actin cytoskeleton that functions at the heart of the growth regulatory machinery by coordinating various cellular events, such as vesicle trafficking and cell wall construction. Consistent with the zonation of the cytoplasm (Cheung and Wu, 2007), actin filaments assume distinct distribution patterns within different regions of the pollen tube, and these differentially organized filaments carry out distinct cellular functions (Chen et al., 2009; Cheung and Wu, 2008; Fu, 2015; Qu et al., 2015; Ren and Xiang, 2007; Staiger et al., 2010). Actin filaments are highly dynamic within the apical region of the pollen tube (Fu et al., 2001; Gibbon et al., 1999; Qu et al., 2013; Vidali et al., 2001). However, we still have an incomplete understanding of the mechanisms that dynamically remodel apical actin filaments while maintaining their distinct overall organization to meet the demands of rapid and directional pollen tube growth.

Previous studies have revealed that apical actin filaments mainly originate from the membrane within the growth domain of pollen tubes (Cheung et al., 2010; Liu et al., 2015; Qu et al., 2013; Zhang et al., 2016) and are arrayed into a unique ‘apical actin structure’ that is spatially separated from the shank-oriented longitudinal actin cables (Qu et al., 2017). Interestingly, it has been further shown that apical actin filaments have distinct functions in regulating vesicle trafficking within the growth domain of pollen tubes (Qu et al., 2017), although more direct evidence is needed to support this model. Nonetheless, it is interesting to ask how pollen tubes dynamically remodel the apical actin filaments and meanwhile maintain their distinct spatial distribution to drive vesicle trafficking and ensure rapid pollen tube growth. In this regard, members of the actin-depolymerizing factor family (ADFs) are reasonable candidates to drive the turnover of apical actin filaments and consequently influence their spatial distribution pattern.

In support of this hypothesis, ADFs along with their cofactor AIP1 have been shown to accumulate at the subapex of the pollen tube (Chen et al., 2002; Lovy-Wheeler et al., 2006). Historically, however, it has been hypothesized that ADFs are mainly involved in promoting actin polymerization by severing actin filaments, thus generating more naked barbed ends to facilitate the construction of the prominent subapical actin structure (Lovy-Wheeler et al., 2005). In line with this hypothesis, cofilin was reported to promote actin polymerization in vivo (Ghosh et al., 2004). Therefore, to clarify whether and how ADFs regulate the dynamics of apical actin filaments in the pollen tube, it is necessary to carefully examine the precise intracellular localization of ADFs and the effect of loss of ADF function on the dynamics of individual actin filaments.

The biochemical activities and cellular functions of ADFs have been documented extensively in plants (Allwood et al., 2002; Carlier et al., 1997; Henty et al., 2011; Lopez et al., 1996; Smertenko et al., 2001; Tian et al., 2009). In particular, ADFs have been implicated in tip growth of root hairs (Jiang et al., 1997), moss protonematal cells (Augustine et al., 2008) and pollen tubes (Chen et al., 2002; Li et al., 2010; Zheng et al., 2013). The Arabidopsis genome contains a total of 11 ADF genes, which are categorized into four subclasses (Nan et al., 2017; Roy-Zokan et al., 2015). Several Arabidopsis ADFs have been implicated in the regulation of pollen tube growth. For instance, the class III ADF5 is involved in the regulation of pollen tube growth by bundling and stabilizing actin filaments (Zhu et al., 2017). Two class IIa ADF isovariants (ADF7 and ADF10) are expressed specifically in Arabidopsis pollen (Pina et al., 2005; Ruzicka et al., 2007) and they have been shown to have distinct subcellular localization patterns in the pollen tube (Bou Daher et al., 2011; Bou Daher and Geitmann, 2012). This indicates that ADF7 and ADF10 may have distinct functions in regulating actin dynamics in the pollen tube. ADF7 has been implicated in regulating the turnover of shank-oriented longitudinal actin cables (Zheng et al., 2013). However, it remains to be determined how ADF7 and ADF10, either alone or coordinately, regulate the dynamics and construction of distinct actin arrays within pollen tubes.

To determine the role of ADFs in regulating the turnover and construction of apical actin filaments in the pollen tube, we analyzed the function of ADF10 by using the transcription activator-like effector nucleases (TALEN)-mediated knockout approach in combination with state-of-the-art live-cell imaging and in vitro biochemical assays. We demonstrate that ADF10 shapes the overall organization of apical actin filaments mainly by promoting their turnover and ordering. Consequently, ADF10 controls tip-directed transport and accumulation of vesicles in order to drive rapid pollen tube growth. This finding thus provides significant insights into the cellular mechanism underlying actin-mediated regulation of vesicle trafficking and pollen tube growth.

To determine the function of ADF10 in pollen, we generated ADF10 loss-of-function alleles by using the transcription activator-like effector nucleases (TALEN)-based gene disruption approach. A TALEN-binding site was selected in the second exon of ADF10 (Fig. 1A) and a total of 40 TALEN integrated lines were generated. The Taq^αI site (TCGA) between the left and right arm is intact in wild-type (WT) plants (Fig. 1A), and therefore the amplified PCR product of ∼0.95 kb is digested by Taq^αI into two small fragments of 0.7 kb and 0.26 kb (Fig. 1B). By comparison, in the TALEN line #10 (named as adf10 hereafter), the ∼0.95 kb fragment is not cut because the Taq^αI site is disrupted (Fig. 1B). The mutation in the ADF10 gene in the adf10 line was further confirmed by sequencing of genomic DNA. We next examined the pollen tube phenotype associated with adf10 and found that adf10 pollen tubes appear to be more curved (Fig. 1C). By tracing the pollen tube growth axis starting from the germination aperture to the tip of the pollen tube (Fig. 1D), we generated plots for pollen tubes derived from WT and adf10 plants, which showed that adf10 pollen tubes are more curved than WT pollen tubes (Fig. 1E,F). In addition, we found that the velocity of pollen tube growth is reduced significantly in adf10 pollen tubes (Fig. 1G). Thus, our data suggest that loss of function of ADF10 impairs polarized pollen tube growth.

To determine the subcellular localization of ADF10 in the pollen tube, we generated an ADF10–EGFP fusion construct driven by the ADF10 promoter (ADF10pro:ADF10-EGFP) and transformed it into adf10 mutant plants to generate transgenic plants (ADF10pro:ADF10-EGFP;adf10). We found that pollen derived from ADF10pro:ADF10-EGFP;adf10 is indistinguishable from that derived from WT plants in terms of pollen germination and pollen tube growth (Fig. S1A–D), which suggests that expression of ADF10–EGFP under the control of the endogenous ADF10 promoter can complement the adf10 mutation. This result, along with data showing that ADF10–EGFP has comparable activity to that of ADF10 in vitro (Fig. S1E–I), provides evidence that ADF10–EGFP is functional and can therefore act as a faithful probe to indicate the intracellular localization of ADF10 in pollen tubes. Similar to the previously reported intracellular localization of ADF7 in the pollen tube (Zheng et al., 2013), we found that ADF10 decorates actin filaments throughout the entire pollen tube, including the apical and subapical regions (Fig. 2). ADF10–EGFP also gave very prominent cytosolic signals (Fig. 2), a result that is consistent with the ability of ADF10 to bind to monomeric G-actin (see below). Thus, we show that ADF10 exists in both filamentous and cytosolic forms throughout the entire pollen tube.

We next generated recombinant ADF10 (Fig. 3A) to examine its effect on the regulation of actin dynamics in vitro. We found that ADF10 led to the depolymerization of actin filaments in a dose-dependent manner (Fig. 3B,C). The depolymerizing activity of ADF10 was confirmed through a kinetic 7-chloro-4-nitrobenzo-2-oxa-1,3-diazole (NBD)–actin disassembly assay (Fig. 3D). In addition, we found that ADF10 prefers binding to ADP-bound G-actin rather than ATP-bound G-actin (Fig. 3E), and inhibits actin nucleotide exchange (Fig. 3F). Furthermore, we found that ADF10 is able to generate breaks along actin filaments in a time-dependent manner (Fig. 3G,H) and that it promotes monomer dissociation (Fig. 3I). Based on the sequencing results shown in Fig. 1A, the ADF10 transcript transcribed in adf10 line harbors a deletion of nine nucleotides. The mutated ADF10 transcript will translate a predicted ADF10 mutant protein (ADF10_mut) harboring a deletion of three amino acid residues and replacement of several other amino acids within a region spanning 11 amino acids in total, which covers mostly of the β4 sheet of ADF10 (Fig. S2A). To yield insight into the nature of adf10 line, we generated the recombinant ADF10_mut to test its biochemical activity and found that it fails to shorten and depolymerize actin filaments (Fig. S2C–E), which suggests that adf10 is a null allele even if the ADF10_mut stably exists in cells. Our study suggests that the β4 sheet is crucial for the function of ADF10. Consistent with this, a previous study showed that mutations in several amino acid residues in the β4 sheet of yeast cofilin are lethal in yeast (Lappalainen et al., 1997).

Given that loss of function of ADF7 reduces the rate of actin turnover (Zheng et al., 2013), we sought to examine the effect of loss of ADF10 function on actin turnover in vivo. We initially assayed the effect of latrunculin B (LatB) treatment on pollen germination and found that pollen derived from several TALEN-integrated lines of ADF10 (Fig. 1A) exhibited a LatB-resistant phenotype (Fig. S2F). We next selected a representative adf10 line (TALEN 10#, Fig. S2F) to document the LatB-resistant phenotype of pollen germination and pollen tube growth more carefully. We found that LatB inhibited the germination of pollen derived from WT and adf10 plants in a dose-dependent manner, but adf10 pollen was significantly more resistant to LatB treatment (Fig. 3J,K). In addition, the elongation of adf10 pollen tubes was significantly more resistant than WT pollen tubes to LatB treatment (Fig. 3L). Furthermore, the LatB-induced breakdown of actin filaments was significantly reduced in both adf10 pollen grains and tubes compared to that in WT (Fig. 3M–O). Taken together, these results suggest that ADF10 is a typical actin-depolymerizing factor that prefers binding to ADP–G-actin and inhibits G-actin nucleotide exchange. ADF10 severs and depolymerizes actin filaments in vitro, and promotes actin turnover in pollen.

To examine the effect of loss of function of ADF10 on the organization of actin filaments in the pollen tube, we used Alexa-Fluor-488-conjugated phalloidin staining to reveal the organization of actin filaments in fixed pollen tubes. We found that the region occupied by the brighter apical actin filaments at the subapex became enlarged in adf10 pollen tubes compared to that in WT pollen tubes (Fig. 4A,C). This was confirmed by measuring the fluorescence intensity of Alexa–Fluor-488–phalloidin staining (Fig. 4B,D). Visualization of actin filaments in transverse sections of pollen tubes showed that the inner region, which contains fewer actin filaments, extended much further toward the tube base in adf10 pollen tubes (Fig. 4C, right two panels) when compared to the WT pollen tubes (Fig. 4A, right two panels). Analysis of transverse sections showed that the inner region with fewer actin filaments stopped at ∼5 µm from the tip in WT pollen tubes (Fig. 4A, right two panels), but is still visible at ∼8.5 µm away from the tip in adf10 pollen tubes (Fig. 4C, right two panels). Thus, in adf10 pollen tubes, the apical actin filaments extend a lot further toward the tube base and consequently generate a comparatively large inner region with fewer apical actin filaments than in WT pollen tubes.

Given that apical actin filaments are involved in regulating the apical accumulation of vesicles, we sought to examine the distribution of YFP–RabA4b-decorated vesicles in adf10 pollen tubes. We found that the ‘V’-shaped conical region in which vesicles accumulate was enlarged in adf10 pollen tubes compared to that in WT pollen tubes (Fig. 5A; Movie 1). This result was supported by visualizing transverse sections from the tip to the base (Fig. 5B). We quantified the area in which tip-directed vesicles accumulate by measuring the V-shaped region in mutant and WT pollen tubes. We determined the angle formed at the base of the V, and the distances from the top of the V to the bottom of the V and the extreme tip of the pollen tube (Fig. 5C). The results showed that the V angles are reduced significantly in adf10 pollen tubes (Fig. 5D). In addition, the depth of the V and the distance of the V from the tip were increased significantly in adf10 pollen tubes (Fig. 5E). The apical vesicle distribution pattern mainly results from the backward flow of vesicles (Parton et al., 2001; Qu et al., 2017), and apical actin filaments appear to act as a physical barrier to prevent the backward movement of vesicles to shape the apical vesicle distribution pattern (Qu et al., 2017). We wanted to determine whether the abnormal apical vesicle accumulation pattern in adf10 pollen tubes results from the impaired function of apical actin filaments in preventing the backward movement of vesicles. We therefore performed fluorescence recovery after photobleaching (FRAP) analysis of YFP–RabA4b-positive vesicles with the tip kept unbleached and found that the recovered vesicles naturally form a V-shaped conical distribution pattern in WT pollen tubes (Fig. S3A, left panel) (Qu et al., 2017). By comparison, the region where recovered vesicles accumulated was larger in adf10 pollen tubes (Fig. S3A, right panel), which suggests that the function of apical actin filaments as the physical barrier is impaired in the mutants. Given that apical actin filaments are also involved in preventing the apical invasion of large organelles (Qu et al., 2017), we sought to examine whether the movement of large organelles was altered in adf10 pollen tubes. We visualized YFP–ARA7-decorated endosomes, which were previously shown to be excluded from the tip of pollen tubes (Zhang et al., 2010b). We found that the apical region without YFP–ARA7-positive vesicles was slightly enlarged in adf10 pollen tubes compared to WT pollen tubes (Fig. S3B). This result is supported by measurements showing that the position at which large organelles reverse their direction of travel in adf10 pollen tubes is slightly but significantly further away from the pollen tube tip (Fig. S3C,D; Movie 2). This suggests that in adf10 pollen tubes, the apical actin filaments largely maintain their ability to prevent large organelles reaching the tip of the tube.

To determine whether the alteration in the organization of apical actin filaments affects their role in driving tip-directed vesicle transport in adf10 pollen tubes, we performed FRAP analysis of pollen tubes harboring YFP–RabA4b. The results showed that tip-directed vesicle transport was reduced within the growth domain of adf10 pollen tubes, as evidenced by the significantly reduced recovery rate of YFP–RabA4b fluorescence within the growth domain of adf10 pollen tubes compared to that found for WT pollen tubes (Fig. 5F–H). To assay the vesicle transport function of apical actin filaments more specifically, we performed another type of FRAP experiment in which a small region of the pollen tube tip was monitored after photobleaching (Fig. 5I). The results showed that the recovery rate is reduced in adf10 pollen tubes compared to that in WT pollen tubes (Fig. 5J,K). This suggests that loss of function of ADF10 impairs the function of apical actin filaments as the molecular tracks for myosin motors. Given that ADF10 decorates actin filaments throughout the entire pollen tube (Fig. 2), we wondered whether loss of function of ADF10 also alters the function of actin filaments as molecular tracks within the shank region. We found that the velocity of YFP–ARA7-decorated endosomes was reduced significantly in adf10 pollen tubes compared to that in WT pollen tubes (Fig. S3E). This suggests that loss of function of ADF10 also impairs the function of shank-oriented actin filaments as molecular tracks. Taken together, these results suggest that loss of function of ADF10 reduces the velocity of tip-directed vesicle transport and affects the apical vesicle accumulation pattern.

To reveal the underlying defects in the organization of apical actin filaments in adf10 pollen tubes, we traced the dynamics of actin filaments decorated with Lifeact–EGFP (Qu et al., 2013; Vidali et al., 2009). Consistent with previous observations (Liu et al., 2015; Qu et al., 2017, 2013; Zhang et al., 2016), we found that the actin filaments in WT pollen tubes polymerized from the apical membrane and grew out fairly straight into the cytoplasm, where they were subjected to frequent severing (Fig. 6A). By comparison, in adf10 pollen tubes, the actin filaments grew out from the apical membrane toward the cytoplasm in a more disorganized fashion and were subjected to less-frequent severing (Fig. 6A). Consequently, the organization of apical actin filaments appears to be relatively random in adf10 pollen tubes compared to that seen in WT pollen tubes (Fig. 6A, schematic diagram in far-right panels). This was supported by measurements showing that the average angles formed between apical actin filaments and the tube growth axis increased substantially in adf10 pollen tubes compared to that for WT pollen tubes (Fig. 6B). Measurement of the parameters associated with individual apical actin filaments showed that the depolymerization rate and severing frequency of actin filaments were reduced significantly in adf10 pollen tubes compared to that for WT pollen tubes (Fig. 6C). Consequently, the maximal filament life-time was significantly longer in adf10 pollen tubes (Fig. 6C). However, no overt difference was detected in the elongation rate and maximal filament length of apical actin filaments in adf10 pollen tubes compared to WT pollen tubes (Fig. 6C). Interestingly, we found that apical actin filaments that formed large angles with the tube growth axis tended to be severed more frequently in WT pollen tubes (Fig. 6D). However, in adf10 pollen tubes, the severing frequency of actin filaments that form large angles with the tube growth axis was significantly reduced (Fig. 6D), and the maximal filament life-time was significantly increased in the mutants (Fig. 6E). To some extent, this explains why apical actin filaments are distributed relatively randomly in adf10 pollen tubes and implies that ADF10 is able to promote the ordering of apical actin filaments. Interestingly, we found that the extent of actin filament bundling increases in adf10 pollen tubes compared to that in WT pollen tubes (Fig. S4A). This was quantitatively supported by measurements showing that the bundling frequency of actin filaments increases in adf10 pollen tubes while the debundling frequency decreases (Fig. S4B). This helps to explain, at least in part, why the dynamics of actin filaments are reduced in adf10 pollen tubes. Taken together, our results suggest that ADF10 regulates the organization of apical actin filaments mainly by promoting their turnover and ordering in the pollen tube.

Here, we demonstrate that ADF10 shapes the overall organization of apical actin filaments in Arabidopsis pollen tubes. ADF10 enhances the severing and depolymerization of actin filaments, thus promoting their turnover and ordering (Fig. 6F). Consistent with the fact that ADF10 is a typical actin-depolymerizing factor (Fig. 3), its loss of function caused a reduction in actin filament severing and depolymerizing events in the pollen tube (Fig. 6). In line with this finding, the amount of apical actin filaments increases in adf10 pollen tubes and the apical actin structure extends a lot further toward the tube base. In addition, the inner region with fewer actin filaments, which corresponds to the vesicle accumulation region, is enlarged in adf10 pollen tubes compared to in WT pollen tubes (Fig. 4A,C, Fig. 5A,B). Interestingly, although the amount of internal apical actin filaments increases in the inner region of adf10 pollen tubes, their role in preventing the backward movement of vesicles is compromised compared to what is seen for WT pollen tubes (Fig. S3A). This suggests that proper organization of the inner apical actin filaments is crucial for their role in preventing the backward movement of vesicles and supports the notion that internal apical actin filaments play an active role in regulating vesicle accumulation at the pollen tube tip. Furthermore, loss of function of ADF10 alters the arrangement of apical actin filaments at the cortex and impairs the associated tip-directed vesicle transport (Fig. 5J,K). This present study supports our recently proposed model regarding the role of apical actin filaments in driving tip-directed transport and apical accumulation of vesicles (Qu et al., 2017).

Previous observations have revealed that actin filaments mainly polymerize from the apical membrane within the growth domain of pollen tubes (Cheung et al., 2010; Liu et al., 2015; Qu et al., 2017, 2013; Zhang et al., 2016) and are arrayed into a unique structure, designated as the ‘apical actin structure’, which regulates vesicle trafficking at the pollen tube tip (Qu et al., 2017). It has been shown that rapid actin polymerization occurs concurrently with pollen tube growth from the apical membrane, and the elongation rate of the filaments reaches ∼0.25 µm/s (Qu et al., 2013; Zhang et al., 2016), which is several-fold faster than the pollen tube growth rate. A remaining question is how the continuous membrane-originated actin polymerization is balanced with maintaining the overall organization of apical actin filaments to generate the apical actin structure. We speculated that an ADF protein would be a perfect candidate for this task for the following two reasons. The first is that, since apical actin filaments were assumed to be nucleated by the membrane-anchored barbed end actin polymerase, the formins (Cheung et al., 2010; Li et al., 2017; Liu et al., 2015), the extending portion of apical actin filaments within the cytoplasm contains the ADP–G-actins that are preferred by ADF. The second is that the base of the apical actin structure corresponds to the alkaline subapical region reported previously (Feijó et al., 1999) where the activity of ADF proteins is favored. However, we still lacked direct genetic and cytological evidence to support the role of ADF proteins in regulating the dynamics of actin filaments at the pollen tube tip.

In support of the above speculation, we found that ADF10 is distributed throughout the entire pollen tube including the subapical region (Fig. 2). Furthermore, in adf10 pollen tubes, the apical actin filaments extend a lot further toward the tube base, which consequently induces the extension of the whole apical actin structure (Fig. 4). Further observations reveal that the altered organization and dynamics of actin filaments at the subapex mainly result from the reduction in filament severing and depolymerizing activities (Fig. 6C,D). This is supported by in vitro data showing that ADF10 is a typical actin-depolymerizing factor that severs and depolymerizes actin filaments (Fig. 3A–I). Given that apical actin filaments are mainly polymerized from the membrane (Cheung et al., 2010; Liu et al., 2015; Qu et al., 2017, 2013; Zhang et al., 2016), we propose that ADF10 shapes the organization of apical actin filaments predominantly by limiting the length of membrane-originated actin filaments. We cannot completely rule out the possibility that incorporation of the longer-lived actin filaments in the cytosol at the subapex contributes to the extension of the whole apical actin structure in adf10 pollen tubes (Fig. 4). Consistent with our observation that ADF10 also decorates shank-oriented actin filaments (Fig. 2), the dynamics of shank-oriented actin filaments are also reduced in adf10 pollen tubes (Fig. S5). Accordingly, the bundling frequency increased and the debundling frequency decreased in the shank of adf10 pollen tubes (Fig. S4). In addition, we found that the percentage of apical filaments forming large angles with the tube growth axis is substantially increased in adf10 pollen tubes (Fig. 6B). This supports the hypothesis that ADF10 may be involved in regulating the organization of apical actin filaments by promoting their ordering.

Therefore, our data suggest that ADF10 is involved in regulating the dynamics of apical actin filaments mainly by promoting their turnover and ordering. Overall, our findings differ from previous speculation that ADF promotes actin polymerization to facilitate the construction of the actin fringe structure (Lovy-Wheeler et al., 2006). That hypothesis was mainly based on observations that the actin fringe structure is even more prominent than the shank-oriented longitudinal actin cables (Lovy-Wheeler et al., 2005) and that ADF accumulates at the alkaline subapex (Lovy-Wheeler et al., 2006) where the actin depolymerizing activity of ADF is favored. Within this framework, Lovy-Wheeler et al. speculated that ADF generates more barbed ends to enhance actin polymerization and consequently promotes the formation of the prominent actin fringe structure (Lovy-Wheeler et al., 2006). Given that actin filaments were assumed to polymerize locally at the subapex (Lovy-Wheeler et al., 2005), that was a reasonable hypothesis. However, our study suggests that ADF10 mainly limits the length of membrane-originated actin filaments to shape the formation of an apical actin structure that is spatially distinct from the shank-oriented longitudinal actin cables.

Given that ADF10 and ADF7 belong to class IIa and are highly similar to each other (Ruzicka et al., 2007; Bou Daher et al., 2011; Zheng et al., 2013), and that they are both expressed specifically in pollen (Pina et al., 2005; Ruzicka et al., 2007), it will be interesting to determine how they may coordinate to regulate actin dynamics in the pollen tube. However, the failure to obtain adf7 adf10 pollen tubes (Fig. S6) prevents us from performing such an analysis. Nonetheless, the data suggest that, to some extent, ADF7 and ADF10 act redundantly in the regulation of actin turnover in the pollen tube.

There has been a long debate regarding how apical actin filaments regulate vesicle trafficking within the growth domain of the pollen tube. Our very recent observations suggest that apical actin filaments play an active role in regulating both the tipward and backward movement of vesicles within the growth domain of the pollen tube (Qu et al., 2017). This contrasts with previous theories in which apical actin filaments were simply viewed as molecular tracks for myosin XI (Lovy-Wheeler et al., 2005) or as the filter to allow the apical invasion of small vesicles (Cheung et al., 2008; Kost et al., 1998). The results presented here, describing the organization of apical actin filaments and the tip-directed transport and accumulation of vesicles in adf10 pollen tubes, support our recent hypothesis to some extent (Qu et al., 2017).

In support of the direct involvement of apical actin filaments in regulating vesicle trafficking at the pollen tube tip, we found that the abnormal organization of apical actin filaments at the tip of adf10 pollen tubes alters the apical accumulation pattern of vesicles and reduces their rate of turnover (Fig. 5). Specifically, we found that vesicles are actively transported to the tip along the cortex in both WT and adf10 pollen tubes, while the velocity of tip-directed transport of vesicles decreases substantially in adf10 pollen tubes (Fig. 5J,K), suggesting that the function of cortical apical actin filaments as molecular tracks is impaired in adf10 pollen tubes. This could be due to the changes in the extent of actin filament bundling at the cortex (Fig. S4), which consequently impairs the function of actin filaments as molecular tracks for myosin motors. Given that ADF10 also binds to actin filaments (Fig. 3), loss of ADF10 function may alter the composition of actin-binding proteins (ABPs) on apical actin filaments at the cortex, which may also impair their role as tracks for myosin motors. In addition, this study provides good evidence to support the notion that internal apical actin filaments act as a physical barrier to prevent the backward movement of vesicles. We found that although the amount of internal apical actin filaments increases in adf10 pollen tubes (Figs 4 and 6), their role in preventing the backward movement of vesicles or restricting the backward moved vesicles is impaired (Fig. S3A). This suggests that proper organization of the internal apical actin filaments is crucial for their role as the physical barrier that blocks the backward movement of vesicles. Interestingly, we found that the role of apical actin filaments in preventing the apical invasion of large organelles is not compromised in adf10 pollen tubes (Fig. S3B–D). These results suggest that the two functions of the physical barrier – preventing the backward movement of small vesicles and the apical invasion of large organelles – depend on different properties of the barrier in terms of organization and/or bundling status. Taken together, these findings support our recently proposed model in which apical actin filaments with distinct spatial organization perform distinct functions in the regulation of vesicle trafficking (Qu et al., 2017).

To generate recombinant ADF10, the coding sequence of ADF10 was amplified by RT-PCR from Arabidopsis pollen cDNA from the wild type (Col-0). Given that ADF10 shares high nucleotide sequence identity with ADF7 (Zheng et al., 2013), amplification of ADF10 cDNA was performed with two sequential steps. First, ADF10 cDNA was amplified with the primer pair ADF10 UTR_For/ADF10 UTR_Rev (Table S1). The PCR product was used as the template for the next round of PCR using the primer pair ADF10_For/ADF10_Rev (Table S1). After verifying the authenticity of sequence, it was ligated into pGEX-KG to generate ADF10-pGEX-KG. The generation of ADF10_mut-pGEX-KG was performed with the same strategy using the pollen cDNAs amplified from the 10# ADF10 TALEN line as the template. The expression constructs were subsequently transformed into E. coli, BL21 DE3 strain. The expression of ADF10 or ADF10_mut was induced by the addition of 0.4 mM IPTG. The procedure for purifying recombinant ADF10 or ADF10_mut was the same as that described previously for the purification of ADF7 (Zheng et al., 2013). The strategy for generating the ADF10–EGFP_V10 recombinant protein is exactly the same as that described for generating ADF7–EGFP_V10, as reported previously (Zheng et al., 2013). Briefly, cDNA of ADF10–EGFP_V10 was amplified from total RNA extracted from gADF10-EGFP_V10;adf10 plants with the primer pair ADF10_For/ADF10_Rev (Table S1). The error-free PCR product was moved into the pGEX-KG vector. Since EGFP interferes with the NBD in NBD–actin binding assays due to overlap of the wavelengths, we deleted the key amino acids (T65/Y66/G67) within the fluorophore center of EGFP using the primer pair EGFP-DEL_For/EGFP-DEL_Rev (Table S1) to generate the pGEX-KG-ADF10-EGFP_V10-del plasmid. The purification of GST–ADF10–EGFP_V10-del was performed through glutathione affinity chromatography according to the manufacturer's instructions (GE Healthcare, 17-0756-01) and the GST moiety was cleaved by the digestion with thrombin. The purified protein ADF10–EGFP_V10-del is simply referred to as ADF10–EGFP in the main text. ADF1 was purified according to Carlier et al. (1997). Actin was purified from rabbit skeletal muscle as described previously (Pollard, 1984; Spudich and Watt, 1971) and labeling of actin with Rhodamine or Oregon Green was performed according to a previously published method (Amann and Pollard, 2001).

The high-speed F-actin co-sedimentation assay was performed according to our previously published methods (Wu et al., 2010). Briefly, preassembled actin filaments (3 µM) were incubated with various concentrations of ADF10 at room temperature for 1 h, and the reaction mixtures were subjected to centrifugation for 30 min at 100,000 g. The pellet and supernatant fractions were separated on 15% SDS-PAGE gels, and the amount of actin in the supernatant and pellet was analyzed through densitometry.

The nucleotide exchange rate of ADP–actin monomers was determined by measuring the increase in fluorescence incorporation of 1, N⁶-ethenoadenosine 5′-triphosphate (ε-ATP, Jena Bioscience). The experiment was performed as described by Chaudhry et al. (2010). Briefly, ADP-actin monomers at 2 µM were incubated with or without ADF10 in Tris/NaCl buffer (20 mM Tris-HCl pH 8.0 and 50 mM NaCl) for 3 min at room temperature. Then 50 mM ε-ATP was added to the solution and the fluorescence of the reaction was monitored with a QuantaMaster Luminescence QM 3 PH fluorometer (Photon Technology International, Lawrenceville, NJ) for 400 s with the excitation and emission wavelengths set at 350 and 410 nm, respectively. Quantification of the rate of ε-ATP incorporation was performed as described by Kovar et al. (2000). Briefly, the data were fitted to a single exponential function by using Origin 8.6 software (OriginLab, Northampton, MA). The Δfluorescence (arbitrary units/s, k) was calculated by determining the slope of the fitted curve at the half-maximal level.

The effect of ADF10 or ADF10_mut on shortening actin filaments was determined through fluorescence light microscopy as described previously (Shi et al., 2013). Briefly, 4 µM pre-polymerized actin filaments were incubated with 2.5 µM ADF10 or ADF10_mut for 3 min at room temperature. An equimolar amount of Alexa-Fluor-488–phalloidin was added to label actin filaments and stop the reaction. Finally, actin filaments were diluted to 10 nM with fluorescence buffer (10 mM imidazole-HCl pH 7.0, 50 mM KCl, 1 mM MgCl2, 100 mM DTT, 100 μg/ml Glc oxidase, 15 μg/ml Glc, 20 μg/ml catalase and 0.5% methylcellulose) and visualized with an Olympus IX71 microscope equipped with a ×60 oil objective. The image acquisition was performed using a Retiga EXi Fast 1394 CCD camera (QImaging) with Image-Pro Express 6.3 software. The length of individual actin filaments was quantified with ImageJ software (http://rsbweb.nih.gov/ij/; version 1.38).

A dilution-mediated actin depolymerization assay was performed to determine the activity of ADF10 or ADF10_mut. Briefly, 5 µM pre-polymerized actin filaments (20% NBD labeled) were incubated with various concentrations of ADF10 or ADF10_mut for 2 min, which were subsequently diluted to 0.2 µM with Buffer G (5 mM Tris-HCl, pH 8.0, 0.2 mM CaCl₂, 0.02% NaN₃, 0.2 mM ATP and 0.2 mM DTT). The changes in NBD fluorescence were monitored for 1200 s with a QuantaMaster Luminescence QM 3 PH Fluorometer (Photon Technology International, Inc.) with the excitation and emission wavelengths set at 475 nm and 530 nm, respectively.

The effect of ADF10 on the behavior of single actin filaments was visualized by total internal reflection fluorescence microscopy (TIRFM) roughly according to the previously published method (Amann and Pollard, 2001). The flow cells were made as described previously (Amann and Pollard, 2001). Preassembled Rhodamine-labeled actin filaments (50% labeled; 0.5 µM) were injected into the perfusion chamber, followed by injection of various concentrations of ADF10. The time-lapse images were collected with an Andor iXon3 DU888 EMCCD camera with microManager software (www.micro-manager.org). The frequency of actin filament severing was quantified by counting the number of breaks per filament length per time (breaks/µm/s) as described previously (Shi et al., 2013). More than 20 actin filaments whose lengths were >10 μm were selected for actin filament severing frequency quantification. The actin monomer dissociation rate (subunits/µm) was determined by assuming that there are 334 subunits/µm, as described previously (Zheng et al., 2013).

The TALEN expression cassette was constructed with a TALEN kit according to the manufacturer's instructions (Sidansai Company, Shanghai, China). Briefly, TALEN_For and TALEN_Rev primers (Table S1) were designed based on the sequence within the second exon of ADF10, followed by module assembly and TALEN vector ligation. The sequenced TALEN left and right vectors were ligated into pCambia1301 which was cut with the restriction enzymes HindIII+AscI+SacI. The resulting product was named as ADF10 TALEN L3+R8-1301. Then the TALEN expression cassette was transformed into Arabidopsis plants by the argo-bacteria-mediated flower dipping method (Clough and Bent, 1998). To identify the TALEN lines, total DNA was extracted from the leaves of transformed plants and used as the template for PCR amplification using primers ADF10 Talen PCR_For and ADF10 Talen PCR_Rev (Table S1). The 0.95 kb PCR product was digested with Taq^αI (TAKARA) at 65°C for 1.5 h and the PCR products that were resistant to enzymatic digestion (retaining the 0.95 kb fragment) were subjected to sequencing to confirm the TALEN-mediated mutations. In order to determine the functional redundancy between ADF10 and ADF7, we attempted to generate adf7 adf10 double mutants by crossing the ADF10 TALEN knockout mutant (adf10) with an adf7 mutant (salk_024576, Zheng et al., 2013).

To complement ADF10 TALEN knockout mutants, we sought to remove the FOK1 gene via self-segregation. The FOK1 gene was identified by PCR with the primer pair FOK1_For/FOK1_Rev (see Table S1). We found that ∼10 plants out of a total of 80 T2 plants lacked the FOK1 gene. To identify TALEN mutants, the DNA was digested with Taq^αI (TAKARA) as described above. Approximately 14 plants out of a total of 64 T3 ADF10 TALEN lines were ADF10 TALEN positive. The identified ADF10 TALEN lines without the FOK1 gene were selected for the subsequent complementation analysis. To complement the ADF10 TALEN lines, an EGFP fusion construct with EGFP inserted after Val10 (V10) of ADF10 (designated as gADF10-EGFP_V10) was generated using the same strategy as described previously for ADF7 (Zheng et al., 2013). Briefly, the upstream fragment of ADF10 genomic DNA, EGFP with linkers and the downstream fragment of ADF10 genomic DNA were amplified with the primers described in Table S1. After the amplified fragments were verified by sequencing, they were assembled into the binary vector pBINPLUS to generate the construct gADF10-EGFP_V10. The construct was transformed into adf10 plants by the argo-bacteria-mediated flower dipping method (Clough and Bent, 1998).

Protein sequence alignment of ADF10 with ADF10_mut was performed with DNAMAN software (Lynnon Biosoft). A predicted three-dimensional structure of ADF10 was generated using the crystal structure of ADF1 (PDB ID: 1F7S) as the template to predict the secondary structure of ADF10 to facilitate the sequence alignment.

The (x_beg, y_beg), (x_end, y_end) and (x, y) coordinates in the equation refer to the coordinates of the beginning, the end and the turning points in a pollen tube as originally measured by ImageJ. The plots were subsequently generated with the calibrated coordinates. The coordinates at the beginning and end of the pollen tube were removed during the generation of the plot to avoid overcrowding on the x-axis. More than 120 pollen tubes were measured and plotted for each genotype.

Actin filaments in fixed pollen grains and tubes were revealed by staining with Alexa-Fluor-488–phalloidin as described previously (Zhang et al., 2010a). To determine the effect of latrunculin B (LatB) treatment on the actin cytoskeleton, 100 nM or 200 nM LatB in liquid germination medium was applied onto the surface of solid germination medium for 30 min. The subsequent procedure for actin filament staining was exactly the same as that for non-treated samples. The samples were observed under an Olympus IX83 microscope (Olympus Corporation, Japan) and z-series images were acquired by laser scanning confocal microscopy (Olympus FV1200) with a step size set at 0.5 µm. Samples were excited under a 488-nm argon laser with the emission wavelength set at 500–550 nm. Max-intensity z-projections with optical z-sections at 1-µm intervals were displayed. Transverse sections of the pollen tube starting from the tube apex were generated with ImageJ. Three-dimensional graphs of the intensities of each pixel in these transverse sections were generated using the ‘Surface plot’ tool in ImageJ. A pseudo Rainbow RGB color was applied to the surface plots, with the lower to higher fluorescence intensities represented by colors ranging from dark blue to bright red, respectively. The amount of actin filaments was quantified by determining the fluorescence intensity of Alexa-Fluor-488–phalloidin-staining as described previously (Ye et al., 2009).

Actin filaments were revealed by decoration with Lifeact–EGFP as described previously (Qu et al., 2013). Briefly, pollen tubes (average length 150–200 μm) expressing Lat52:Lifeact-EGFP were observed under an Olympus FV1200 laser scanning confocal microscope or a spinning disk confocal microscope equipped with a Yokogawa CSUX1 scanning head. The time-lapse z-series images were acquired with an Andor iXon3 DU888 EMCCD camera under the spinning disk confocal microscope. The time-intervals were set at 3 s between the time-lapse z-series images and the step size was set at 0.7 μm. Determination of the dynamic parameters of individual actin filaments was performed according to the previously published method (Qu et al., 2013). The determination of bundling and debundling frequency of actin filaments was performed as described previously (Zheng et al., 2013).

Pollen tubes (average length 150–200 μm) expressing Lat52:YFP-RabA4b were observed under an Olympus FV1200 confocal microscope equipped with a 100× oil objective. z-stack slices were scanned and captured with a step size of 0.7 μm at the fastest speed. Generation of transverse sections and the relative 3D images were performed exactly as described for actin filaments in fixed pollen tubes and the pseudocolor ‘red-hot’ was applied. The detailed analysis of the apical accumulation pattern of YFP–RabA4b-labeled vesicles is described in the Fig. 5C legend. To analyze the dynamics of YFP–RabA4b-labeled vesicles, FRAP analysis was performed as described previously (Chang and Huang, 2015). After bleaching, the fluorescence was recorded at 2-s intervals. Recovery rates were acquired by measuring the mean gray values of selected areas with ImageJ, and data sets were plotted as fitted curves as described previously (Chang and Huang, 2015).

Pollen tubes (average length 150–200 μm) expressing Lat52:YFP-ARA7 were observed with a 100× oil objective under a spinning disk confocal microscope equipped with a Yokogawa CSUX1FW spinning disk scanhead. Images were acquired with an Andor iXon3 DU888 EMCCD camera at 2-s time-intervals and a step size of 0.7 μm. A 488-nm laser was used as the excitation light, and the emitted fluorescence was collected using a 525/50 nm band-pass filter. Three-dimensional time-lapse images in hyperstacks were projected with maximum intensity. Analysis of the parameters of ARA7-positive vesicles was performed by using the ‘MtrackJ’ plugin of the ImageJ software (Meijering et al., 2012). More than 20 tracks were traced for each pollen tube, and over 11 tubes were included in the analysis. The furthest position reached by each ARA7-positive vesicle in the pollen tube was recorded manually and the vertical distances from the tube tips and the midlines were measured and used as x- and y-coordinates in the axis graphs.

We thank Yan Zhang (Shandong Agricultural University, Shandong, China) for providing the transgenic lines expressing Lat52:YFP-RabA4b and Lat52:YFP-ARA7 in pollen.