Villin controls the formation and enlargement of punctate actin foci in pollen tubes

The actin cytoskeleton is a complex dynamic network that undergoes rapid assembly and disassembly in response to various internal and external cues (Staiger, 2000). Well-documented examples in plants include the rearrangement of the actin cytoskeleton in root hairs during Nod factor infection from Rhizobium bacteria (Cárdenas et al., 2003; Cárdenas et al., 1998), rearrangement of the actin cytoskeleton during the cell cycle (Sano et al., 2005; Yu et al., 2006), pathogen infection (Takemoto et al., 2003), stomatal closure and opening (Eun and Lee, 1997; Gao et al., 2008; Hwang and Lee, 2001) and the dramatic depolymerization and rearrangement of the actin cytoskeleton in Papaver rhoeas pollen during the self-incompatibility (SI) response (Geitmann et al., 2000; Poulter et al., 2010; Snowman et al., 2002). These cellular changes are triggered by increases in cytosolic free Ca²⁺ ([Ca²⁺]_cyt) to trigger a Ca²⁺-dependent signaling network involving increases in protons and reactive oxygen species to trigger programmed cell death (PCD) (Thomas et al., 2006; Wang et al., 2019; Wilkins et al., 2014). However, the molecular players and mechanisms underlying the rapid and dramatic actin alterations in pollen tubes in response to SI signaling remain largely unexplored.

The pollen tube is a model cellular system for the study of tip growth and its regulation (Bedinger et al., 1994; Feijo et al., 2001; Hepler et al., 2001; Holdaway-Clarke and Hepler, 2003), and the actin cytoskeleton is known to play a critical role in regulating pollen tube growth (Gibbon et al., 1999; Qu et al., 2017; Vidali et al., 2001). It has been shown that Ca²⁺ signaling and the actin cytoskeleton are inherently correlated during pollen tube growth, as evidenced by the observations that both [Ca²⁺] and F-actin levels oscillate during pollen tube growth (Diao et al., 2018; Fu et al., 2001; Holdaway-Clarke and Hepler, 2003; Hwang et al., 2005). SI is a genetically controlled pollen–pistil recognition system controlled by tightly linked polymorphic S-determinant genes expressed in the pollen and pistil. Interaction of cognate S-determinants results in inhibition of incompatible pollen tube growth. In P. rhoeas, the S-determinants are PrsS (a small secreted signaling ligand, related to cysteine-rich proteins) and PrpS, a small, novel transmembrane protein (Foote et al., 1994; Wheeler et al., 2010). Cognate PrsS and PrpS interact and trigger increases in [Ca²⁺]_cyt in incompatible pollen tubes (Franklin-Tong et al., 1993). The rapid elevation in [Ca²⁺]_cyt in the subapical and shank regions to ∼1.5 µM in incompatible pollen tubes is accompanied by the disappearance of the apical Ca²⁺ gradient (Franklin-Tong et al., 1997). During the SI response, actin filaments undergo dramatic reorganization: the distinctive longitudinal F-actin bundles disappear and this is accompanied by a large reduction in actin polymer level (Snowman et al., 2002). Afterwards, fine fragments of actin reorganize into larger aggregates and form highly stable actin foci that increase in size over time in SI pollen tubes and persist for at least 3 h (Geitmann et al., 2000; Poulter et al., 2011; Poulter et al., 2010). There is evidence that elevations in [Ca²⁺]_cyt are upstream of SI-induced actin cytoskeletal rearrangements (Rudd and Franklin-Tong, 2003; Staiger and Franklin-Tong, 2003). Treatment of pollen tubes with the Ca²⁺ ionophore A23187 mimics the SI response, inducing dramatic depolymerization and actin rearrangement into actin foci in pollen tubes (Geitmann et al., 2000; Snowman et al., 2002). The level of actin depolymerization was measured and shown to be far in excess of that required to arrest pollen tube growth (Snowman et al., 2002). Subsequent studies showed that both actin depolymerization and actin stabilization are sufficient to trigger PCD (Thomas et al., 2006). However, to date, the exact mechanisms involved in mediating the dramatic actin alterations and the molecular players involved in this response are largely unknown.

Actin binding proteins (ABPs) are direct regulators of actin dynamics (Pollard, 2016) and transduce cellular stimuli into alterations of the actin architecture, mediating actin cytoskeleton rearrangements in response to various stimuli (Li et al., 2015). Some of them are directly regulated by Ca²⁺ and several ABPs have been shown to bind directly to and be regulated by second messengers (Drøbak et al., 1994; Gungabissoon et al., 1998; Janmey et al., 1987) or by phosphorylation and dephosphorylation events (Khurana et al., 1997; Zhai et al., 2001). Relevant to the Papaver system, a previous study found that profilins, G-actin sequestering ABPs that are abundant in pollen tubes, do not appear to have sufficient Ca²⁺-stimulated sequestering activity to account for the level of depolymerization measured during SI (Snowman et al., 2002). Moreover, two further ABPs: cyclase-associated protein (CAP) and actin-depolymerizing factor (ADF) have been implicated in the regulation of the formation of actin foci in SI pollen tubes, as they colocalize to the actin foci (Poulter et al., 2011; Poulter et al., 2010). Interestingly, fimbrin, a side-binding ABP (McCurdy et al., 2001) did not colocalize to actin foci (Poulter et al., 2011). Villins are Ca²⁺-responsive ABPs (Kumar et al., 2004; Markus et al., 1997) that mediate actin alterations in pollen tubes in response to the elevation in [Ca²⁺]_cyt, but have not previously been studied in relation to the SI response in Papaver. Villin homologues were originally identified by biochemical means in plants (Yokota and Shimmen, 1999; Yokota et al., 2003) and the function and mechanism of action of villins have been documented extensively in Arabidopsis (Huang et al., 2015). Apart from Arabidopsis VILLIN1 (VLN1), which is a simple actin bundler (Huang et al., 2005), other Arabidopsis villins have the full suite of villin-related actin regulatory activities, including nucleating actin assembly, capping, severing and bundling actin filaments (Bao et al., 2012; Khurana et al., 2010; Zhang et al., 2010, 2011). Micromolar-range free Ca²⁺ is sufficient to trigger the actin-severing activity of Arabidopsis villins (Bao et al., 2012; Khurana et al., 2010; Zhang et al., 2010, 2011), implying that the actin-severing activity of villin is biologically relevant within the apical region of pollen tubes where free [Ca²⁺]_cyt reaches about 1 µM to 10 µM (Messerli and Robinson, 1997; Pierson et al., 1994). Indeed, loss of function of VLN2 and VLN5 causes accumulation of actin filaments at pollen tube tips (Qu et al., 2013). However, to date, whether or how exactly villins contribute to the dramatic actin alterations induced by SI in pollen tubes lacks direct genetic and cytological evidence.

The demonstration that the Papaver SI S-determinants have been functionally transferred to Arabidopsis (de Graaf et al., 2012; Lin et al., 2015) suggests that the downstream signals and targets involved in modulating the SI response (e.g. alterations to the actin cytoskeleton and triggering of PCD), might be common and conserved among higher plants. Use of Arabidopsis as an experimental system enables the use of the powerful Arabidopsis genetic approaches and resources to uncover the signaling networks, cellular targets and genes and proteins involved in the SI response in incompatible pollen tubes. It also allows the use of advanced live-cell imaging technologies in conjunction with fluorescent actin markers. Indeed, recent studies suggest that the Arabidopsis pollen tube is a good cellular system in which to perform live-cell imaging of actin dynamics (Qu et al., 2017). As many of the molecular players and mechanisms underlying the dramatic rearrangement of the actin cytoskeleton into the punctate actin foci in response to SI remain to be explored, we have made the first step by treating Arabidopsis thaliana pollen tubes with the Ca²⁺ ionophore A23187. This drug has previously been shown to increase [Ca²⁺]_cyt in pollen tubes (Diao et al., 2018; Franklin-Tong et al., 1996) and also mimics the SI response in Papaver pollen tubes, stimulating the formation of distinctive actin alterations (Snowman et al., 2002), as well as triggering several other SI-related events (Wilkins et al., 2011).

Here, we perform live-cell imaging of actin filament dynamics in response to A23187 in Arabidopsis pollen tubes, and reveal that actin alteration during this process involves the rearrangement, fragmentation and depolymerization of actin filaments. We found that actin foci form at a very early stage and that they gradually enlarge with the incorporation of fragmented actin structures. We further identify that villins are involved in A23187-triggered actin depolymerization and formation of actin foci in pollen tubes, which requires its Ca²⁺-responsive actin-severing activity. Our findings suggest that fragmentation of actin filaments is a key event during the formation of actin foci and it is required for efficient actin depolymerization in pollen tubes. Strikingly, we found that villins are also involved in the enlargement of the distinctive punctate actin foci; this stage of remodeling requires villin actin-bundling activity. Our study thus reveals novel molecular details and insights into mechanisms underlying actin alterations in response to A23187 in pollen tubes and provides evidence that villins are major players in modulating this remodeling.

To initiate a detailed analysis of the molecular mechanisms potentially involved in mediating similar actin alterations to those triggered by the SI response, we used the ionophore A23187 as a tool to investigate actin reorganization in pollen tubes of Arabidopsis thaliana. We treated A. thaliana pollen tubes with the Ca²⁺ ionophore A23187 and fixed them at different time points after treatment and then stained with Alexa Fluor 488-conjugated phalloidin to visualize F-actin. Untreated control pollen tubes had actin filaments arrayed in prominent actin cables oriented longitudinally in the shank region and a bright apical actin structure (Fig. 1A). After treatment with 10 µM A23187 for 2 min, 10 min and 60 min, respectively, dramatic rearrangements of actin filaments were observed in pollen tubes (Fig. 1A). Actin filaments became fragmented, the amount of thick actin bundles was reduced and actin foci started to form in the pollen tubes after being treated with 10 µM A23187 for 2 min (Fig. 1A). The number of actin foci increased and the size of actin foci enlarged in pollen tubes after treatment with A23187 for 10 min and 60 min compared with those treated with A23187 for 2 min (Fig. 1A).

The observed actin rearrangements in Arabidopsis pollen tubes after treatment with A23187 were very similar to those triggered in Papaver pollen tubes during the SI response (Geitmann et al., 2000; Poulter et al., 2010; Snowman et al., 2002) and to those observed in A. thaliana pollen tubes expressing the PRPS1 gene when challenged with recombinant PrsS proteins, reconstituting SI in this species (Fig. 1B). With SI, actin filaments became fragmented and the amount of thick actin bundles was reduced after 2 min (Fig. 1B). The actin remodeling in response to SI was slower compared to A23187, but foci were forming at 60 min and the size of actin foci was increased at 3 h (Fig. 1B). These data clearly show the striking similarity between the dramatic actin remodeling stimulated by A23187 and SI.

To trace the dynamic behavior of actin filaments during this process, we monitored Arabidopsis pollen tubes expressing Lat52:Lifeact-eGFP (Qu et al., 2013) in real-time after treatment with 10 µM A23187. This revealed that, very soon after A23187 treatment, actin filaments undergo rearrangements, exhibited as a conversion of longitudinally arranged actin cables into network-like structures in the pollen tube (Fig. 2A; Fig. S1A). Although the starting time points of actin rearrangement varied between individual pollen tubes, actin rearrangement occurred within 1 min of A23187 treatment. Subsequently, actin filaments became fragmented and these fragmented actin filaments appeared to depolymerize (Fig. 2A,B, Movie 1; Fig. S1A,B). The depolymerization of actin filaments was revealed by determining the percentage of ‘occupancy’ of actin filaments, which was evaluated by the parameter of density of actin filaments, as described previously (Higaki et al., 2010). As shown in Fig. 2C and Fig. S1C, the percentage of occupancy of actin filaments decreased during treatment with A23187, suggesting that actin filaments underwent depolymerization.

These observations are very similar to previous observations of actin alterations and measurements of depolymerization during the SI response in Papaver pollen tubes (Snowman et al., 2002). Accompanied by the fragmentation and depolymerization of actin filaments, actin foci began to form. Soon after actin foci formed, they enlarged during treatment with A23187 (Fig. 2D), quantified by an increase in fluorescence intensity (Fig. 2E and Fig. S1D). Again, this is consistent with previous observations and measurements of formation of actin foci during SI in Papaver pollen tubes (Poulter et al., 2010), except the timing of actin remodeling was faster with A23187.

In terms of the formation of actin foci in response to A23187 treatment of pollen tubes, we observed several distinct patterns of actin reorganization. One of the patterns started with the fragmentation of large actin bundles. After they were split into several pieces, these fragmented actin structures were used as templates to recruit small actin structures to promote the formation of the punctate actin foci structures (Fig. 2F,G). Another pattern of actin foci formation started with the greater fragmentation of actin bundles into tiny structures first, and these structures subsequently came together in a manner resembling some sort of aggregation or fusion, to form the large actin foci (Fig. 2H,I). These detailed observations suggest that the formation and enlargement of these actin foci mainly result from the crosslinking of fragmented actin structures. The common feature of these patterns of actin remodeling is that actin filaments or bundles need to be fragmented first, prior to the formation of actin foci. This suggests that A23187 triggers fragmentation or severing of actin structures, which appears to be a prerequisite for the formation of these punctate actin foci. Thus, our study reveals novel key information about the underlying actin alterations involved in the formation of these distinctive punctate actin foci in pollen tubes in response to A23187.

To ask which ABP(s) might participate in this A23187-triggered actin alteration process in pollen tubes, the Ca²⁺-sensitive gelsolin and villin family proteins were considered good candidates, as A23187 is known to trigger increases in [Ca²⁺]_cyt. In this regard, VLN2, VLN3 and VLN5 could potentially be involved in this actin alteration process, as they are expressed in pollen (https://www.genevestigator.com/gv/index.jsp; Honys and Twell, 2003; Pina et al., 2005) and shown to be versatile Ca²⁺-responsive actin regulatory proteins involved in fragmenting actin filaments (Bao et al., 2012; Khurana et al., 2010; Zhang et al., 2010). We determined the localization of VLN2, VLN3 and VLN5 before and after the addition of A23187 in pollen tubes. We found that VLN2-GFP, VLN3-GFP and VLN5-GFP associated with filamentous structures resembling actin to different extents in normally growing A. thaliana pollen tubes in the absence of A23187 (Fig. 3A; Qu et al., 2013). After the treatment of pollen tubes with A23187, within a few minutes the VLN2-GFP, VLN3-GFP and VLN5-GFP formed dot-like structures (Fig. 3A; Movies 2–4), which started in the apical region and propagated backwards to the shank in pollen tubes and the number of dots increased over time (Fig. 3A). By monitoring the dot-like structures formed by the GFP fusion proteins in real time, we established that their size increased over time (Fig. 3B; Fig. S2). The changes in the pattern of GFP fusion proteins resembled that of actin after treatment with A23187. In support of this notion, examination of fixed A23187-treated pollen tubes at a later time point revealed that the structures decorated with VLN-GFP fusion proteins in these pollen tubes colocalized with phalloidin-stained punctate actin foci (Fig. 3C). This demonstrated that the VLN-GFP fusion proteins colocalize with F-actin. Taken together, these data implicate the involvement of villins in these actin remodeling processes in pollen tubes.

To determine whether and to what extent villins contribute to actin alterations in response to A23187 in A. thaliana pollen tubes, we visualized the actin cytoskeleton in villin loss-of-function mutant pollen tubes treated with A23187 and compared it with that in WT pollen tubes. We found that actin alterations in villin loss-of-function mutant pollen tubes were perturbed when compared with that in WT pollen tubes (Fig. 4A). In particular, the loss of function of VLN5 had a more profound effect on A23187-induced changes in the actin cytoskeleton in pollen tubes when compared with that in WT or vln2 single mutant pollen tubes (Fig. 4A). Specifically, at 60 min after A23187 treatment, we found that there were more fragmented actin bundles in vln5 pollen tubes, whereas most actin filaments were converted into actin foci in WT or vln2 single mutant pollen tubes (Fig. 4A). This suggests that loss of function of VLN5 impairs the fragmentation of actin filaments or actin bundles after treatment with A23187. In addition, we found that the relative amount of fragmented actin bundles versus actin foci is substantially higher in the vln2 vln5 double mutant pollen tubes when compared with that in vln5 pollen tubes (Fig. 4A). Strikingly, we found that the formation of actin foci was severely impaired in vln2 vln3 vln5 triple mutant pollen tubes (Fig. 4A), suggesting that these villins are redundantly required for these actin alterations in pollen tubes in response to A23187.

To quantitatively determine the amount of actin filaments in pollen tubes derived from WT and villin loss-of-function mutants, we skeletonized actin filament staining images (representative skeletonized images after treatment with A23187 for 60 min are shown in Fig. 4B). We found that the density of actin filaments was significantly higher in the villin loss-of-function mutant pollen tubes compared with that in WT pollen tubes, with the exception of vln2 (Fig. 4B,C). These data, showing that loss of function of villins impaired both actin depolymerization and the formation of punctate actin foci in pollen tubes in response to A23187 suggests that villins are involved in mediating these processes, acting in concert to remodel actin to form these structures. Moreover, this suggested that villins are also involved in the highly similar actin remodeling observed during the SI response in Papaver.

Next, we wondered whether the severing activity of villins is involved in actin alterations in response to A23187 in pollen tubes_. We decided to abolish the severing activity of villins using VLN5 as the representative protein. The severing-deficient version of VLN5, designated VLN5^SD, was generated as described previously (Fig. 5A; Revenu et al., 2007). We initially performed dilution-mediated actin depolymerization experiments to determine the actin-depolymerizing activity of recombinant VLN5 and VLN5^SD proteins (Fig. 5B; Zhang et al., 2010). We found that VLN5 promotes actin depolymerisation, whereas VLN5^SD lacked such activity (Fig. 5C). Strikingly, in contrast, VLN5^SD stabilizes actin filaments from dilution-mediated actin depolymerization (Fig. 5C). In addition, through direct visualization of individual actin filaments by total internal reflection fluorescence microscopy (TIRFM), we found that while VLN5 efficiently severed actin filaments, VLN5^SD lacked such actin-severing activity (Fig. 5D,E). These data demonstrate that VLN5^SD fails to depolymerize and sever actin filaments. However, we found that VLN5^SD retained a similar actin-bundling activity compared to VLN5 at 0.4 µM (Fig. 5F,G). Interestingly, VLN5^SD had a slightly higher (P=0.03) actin-bundling activity than VLN5 at a lower concentration (0.2 µM; Fig. 5G). Nonetheless, these data together demonstrate that VLN5^SD retains virtually normal actin bundling activity but is defective in severing actin filaments.

We next transformed VLN5 and VLN5^SD into the vln2 vln5 double mutant under the control of the native VLN5 promoter (Fig. S3A) to determine how the actin-severing activity of VLN5 might be involved in regulating actin alterations in response to A23187 in pollen tubes. We found that the introduction of VLN5 can rescue, whereas the introduction of VLN5^SD cannot rescue, the defective fragmentation of actin filaments or bundles and the formation of actin foci in vln2 vln5 mutant pollen tubes (Fig. 5H). In addition, we found that the introduction of VLN5^SD cannot rescue the defects in actin depolymerization in response to A23187 treatment in vln2 vln5 mutant pollen tubes by determining the percentage of occupancy of actin filaments (Fig. 5I,J). This suggests that the severing activity of VLN5 (which is known to be Ca²⁺-dependent; Zhang et al., 2010) contributes to its role in driving A23187-triggered actin depolymerization. Our data show that VLN5^SD cannot fully rescue the defects in A23187-triggered actin alterations compared to VLN5 in pollen tubes, particularly in terms of actin filament fragmentation and the formation of actin foci (Fig. 5K). This suggests that the actin-severing activity of VLN5 is required for the fragmentation and depolymerization of actin filaments in response to A23187 in pollen tubes.

To determine whether the bundling activity of villins might be involved in driving actin alterations responsible for the formation of the large punctate actin foci in pollen tubes in response to A23187, we used VLN5 as the representative villin to address this question. We decided to remove the bundling activity of VLN5 by deleting its linker and headpiece domain, as previously done for VLN3 (van der Honing et al., 2012). The truncated version of VLN5 with the deletion of linker and headpiece domain was designated VLN5ΔHL (Fig. 6A,B). We found that both VLN5ΔHL and VLN5 can bind to F-actin using high-speed F-actin co-sedimentation experiments (Fig. S4A–D). However, using low-speed F-actin co-sedimentation experiments, we found that the actin-bundling activity of VLN5ΔHL was significantly reduced compared with VLN5 (Fig. 6C,D), especially at higher concentrations, when the VLN5ΔHL mutant had only 12% bundling activity (P=0.000004) compared with VLN5 (Fig. 6D). The decrease in the bundling activity of VLN5ΔHL was also confirmed by direct visualization of filamentous actin structures with fluorescence light microscopy (Fig. 6E). However, using direct visualization of actin filaments with TIRFM, we found that VLN5ΔHL severed actin filaments just as efficiently as VLN5 (Fig. 6F,G). This demonstrates that its severing activity was not affected.

We next transformed VLN5ΔHL into the vln2 vln5 mutant under the control of VLN5 native promoter (Fig. S3B), and found that actin filaments or bundles could be fragmented and converted into actin foci (Fig. 6H). Interestingly, we found that the size of actin foci was smaller in VLN5p:VLN5ΔHL (vln2 vln5) pollen tubes when compared with that in VLN5p:VLN5 (vln2 vln5) pollen tubes under the same conditions (Fig. 6H,I). Visualization of the actin cytoskeleton in pollen tubes at different time points after treatment with A23187 showed the increase in the fragmentation of actin structures in pollen tubes within 2 min and subsequent accumulation of punctate actin foci (Fig. 6J), further demonstrating that VLN5ΔHL can fragment actin structures in vivo. Quantification of the occupancy of actin structures showed that VLN5ΔHL was able to depolymerize actin filaments in response to A23187 treatment of pollen tubes (Fig. 6K). However, we do not currently know why VLN5ΔHL cannot depolymerize actin filaments as efficiently as VLN5 in pollen tubes after A23187 treatment (Fig. 6K). Pollen tubes with the VLN5ΔHL retained the ability to promote the formation of actin foci, but failed to enlarge actin foci in response to A23187 in pollen tubes. This was confirmed by direct visualization of actin filaments decorated with Lifeact-GFP after treatment with A23187 in VLN5p:VLN5ΔHL (vln2 vln5) pollen tubes (Fig. 6L,M). These findings were also supported by data showing that VLN5ΔHL colocalizes with actin foci in A23187-treated pollen tubes (Fig. S4E), but the enlargement of VLN5ΔHL-decorated structures was impaired severely (Fig. S4F). These observations demonstrated that the initial stages of actin foci formation in the VLN5p:VLN5ΔHL (vln2 vln5) pollen tubes compared with either WT or VLN5p:VLN5 (vln2 vln5) pollen tubes are normal, while the enlargement of actin foci is impaired (Fig. 6N). Because VLN5ΔHL has dramatically reduced bundling activity due to removal of the linker and headpiece domains, these data implicate that the bundling activity of VLN5 is required for the enlargement of actin foci in response to the elevation in [Ca²⁺]_cyt in pollen tubes. Thus, together, our data suggest that the formation of the actin foci involves at least two steps, both involving villin proteins: (1) the severing of actin filament bundles, which will allow rapid/efficient depolymerization and (2) bundling of actin filament fragments to form large actin foci.

The actin cytoskeleton is a major target of the signaling networks in eukaryotic cells (Li et al., 2015). How the actin cytoskeleton responds to upstream signaling events and achieves reorganization to meet cellular demands is a fascinating topic with many unanswered questions. ABPs are direct regulators of actin dynamics that transduce cellular stimuli into alterations of the actin architecture (Hussey et al., 2006; Smith and Oppenheimer, 2005; Staiger and Franklin-Tong, 2003). The SI response in Papaver rhoeas pollen tubes has a relatively well-characterized Ca²⁺-mediated signal transduction network that induces dramatic alterations to the actin cytoskeleton in incompatible pollen tubes involving the formation of distinctive punctate actin foci (Geitmann et al., 2000; Poulter et al., 2010; Snowman et al., 2002) necessary and sufficient to trigger PCD in this system (Thomas et al., 2006; Wang et al., 2019; Wilkins et al., 2014). These actin foci are rather unusual as they are highly stable; once formed, they are resistant to disassembly by 1 µM Latrunculin B (LatB) under conditions that remove virtually all detectable F-actin in normally growing pollen tubes (Snowman et al., 2002; Thomas et al., 2006). This sets them apart from many other actin-based structures reported. Currently, there are no explanations for the formation or unusual dynamic properties of these structures. The molecular players and mechanisms underlying the actin alterations in pollen tubes in response to SI signaling remains largely unexplored, except for a study showing that ADF and CAP colocalize with these actin foci during their formation; somewhat surprisingly, fimbrin, which crosslinks and stabilizes actin filaments in plant cells (Kovar et al., 2000b) did not colocalize with the SI-induced actin foci (Poulter et al., 2010).

Here, we performed live-cell imaging of actin dynamics to investigate the formation of actin foci in Arabidopsis pollen tubes, using the Ca²⁺ ionophore A23187 as a trigger. This drug, which has been shown to raise [Ca²⁺]_cyt artificially in pollen tubes (Diao et al., 2018; Franklin-Tong et al., 1996), mimics the SI response in Papaver pollen tubes as it induces similar alterations to the actin cytoskeleton in Papaver pollen tubes to those seen in the SI response (Snowman et al., 2002). Because A23187 has previously been demonstrated to increase [Ca²⁺]_cyt, we infer that A23187 triggers increases in [Ca²⁺]_cyt in these A. thaliana pollen tubes, even though we have not assessed this here. However, regardless of this, A23187 provides a useful, simple tool to trigger the formation of punctate actin foci in order to study both the nature of their formation and ABPs involved in this process. Use of Arabidopsis enabled us to investigate the role of Ca²⁺-responsive villins using cytological and genetic approaches not available in Papaver. Thus, we show for the first time that villins are involved in mediating actin alterations very similar to those observed during the SI response in Papaver pollen tubes. Moreover, we provide important insights into the mechanistic details of what is involved in the formation of the distinctive punctate actin foci. Treatment with A23187 triggers the formation of punctate actin foci in A. thaliana pollen tubes, which is one of the characteristic features shown in Papaver SI pollen tubes (Rudd and Franklin-Tong, 2003; Staiger and Franklin-Tong, 2003), suggesting that A. thaliana might have all the components necessary for an SI-triggered Ca²⁺-dependent signaling network in Papaver pollen tubes.

Based on our findings, we propose a simple model describing actin alterations in response to the A23187 treatment of the pollen tube incorporating the involvement of the Ca²⁺-responsive ABP, villin (Fig. 7). Specifically, we have imaged at least four distinct configurations of actin: (1) normal actin configuration of longitudinal bundles in normally growing pollen tubes (Fig. 7A); (2) fragmentation of longitudinal actin bundles involving the severing activity of villins, which will allow more effective depolymerization that drives the breakdown of filaments (Fig. 7B); (3) the formation of actin foci from actin fragments necessitates a bundling activity, which involves villins (Fig. 7C); (4) enlargement of the foci, which also involves villins (Fig. 7D). Because the deletion of the linker and headpiece domains prevents the enlargement of actin foci, this provides evidence that this later stage is achieved by the crosslinking of fragmented actin structures.

As A23187 has been shown to increase [Ca²⁺]_cyt (Diao et al., 2018; Franklin-Tong et al., 1996), these data together suggest that an elevation in [Ca²⁺]_cyt stimulates the initial fragmentation and depolymerization of actin filaments in pollen tubes. This could cause disruption of the normal actin configuration and a reduction in the amount of actin filaments observed here. Our data are consistent with previous findings from Papaver showing a severe reduction in the number of actin filaments in pollen tubes after SI induction (Geitmann et al., 2000; Snowman et al., 2002; Thomas et al., 2006). An attempt was previously made to assess the potential contribution of profilin to actin depolymerization in response to the elevation in [Ca²⁺] in vitro, as actin monomer sequestering activity of profilins is stimulated by Ca²⁺ (Kovar et al., 2000a) and the concentration of profilins is at almost equal molar concentrations to actin in pollen (Gibbon et al., 1999; Snowman et al., 2002; Vidali and Hepler, 1997). However, this study showed that the elevation in [Ca²⁺] could only account for a 10–16% reduction in the amount of actin filaments (Snowman et al., 2002), implying that other ABPs contribute to this actin depolymerization process, as the SI response can cause 60–70% depolymerization of actin in pollen (Snowman et al., 2002; Thomas et al., 2006).

Based on the observations that actin filament fragmentation events observed are very frequent after A23187 treatment, the ADF/cofilin, villin and gelsolin actin-severing protein families are likely contributors. As SI in Papaver triggers both increases in [Ca²⁺]_cyt (Franklin-Tong et al., 1997; Franklin-Tong et al., 1993; Wu et al., 2011) and acidification of cytoplasm (Wilkins et al., 2015) in SI-induced pollen tubes, ADF is an unlikely candidate for this actin fragmentation process, as the severing activity of plant ADFs is reduced by Ca²⁺-dependent protein kinases (Allwood et al., 2001; Dong and Hong, 2013) and they prefer alkaline conditions for severing (Maciver and Hussey, 2002). We considered the Ca²⁺-responsive villins or gelsolins as likely candidates for this depolymerisation and fragmentation process, as a previous study demonstrated that a gelsolin purified from Papaver pollen fragments actin filaments and acts in concert with profilins to promote substantial actin depolymerization (Huang et al., 2004). Thus, we wondered whether Arabidopsis villins might perform similar functions by fragmenting actin filaments and capping the barbed end of actin filaments to promote actin depolymerization via acting in concert with profilin in response to elevations in [Ca²⁺]_cyt in pollen tubes. Since the Arabidopsis genome only encodes villin-like genes (Klahre et al., 2000), the pollen-expressed villins are likely to be involved in this actin alteration process. Here, we have established a clear role for villin in the early actin-remodeling events involving severing activity in response to A23187. As the severing activity of villins is Ca²⁺ responsive (Bao et al., 2012; Khurana et al., 2010; Kumar et al., 2004; Markus et al., 1997; Wu et al., 2015; Yokota et al., 2005; Zhang et al., 2010, 2011) and A23187 is known to increase [Ca²⁺]_cyt (Diao et al., 2018; Franklin-Tong et al., 1996), this suggests that severing is an important early event triggered by increases in [Ca²⁺]_cyt, responsible for breaking down long actin filament bundles to generate more filament ends for monomer dissociation. We propose that villin activity is likely to act in concert or cooperatively with the Ca²⁺-responsive monomer-sequestering activity of profilin (Kovar et al., 2000a).

Later, another activity is evident, as the fragmented actin filaments or actin bundles were observed to be brought together to form the punctate actin foci. We provide evidence that villins mediate the formation and enlargement of the actin foci in pollen tubes. Our observations suggest that this mainly results from the crosslinking of fragmented actin structures; this is further supported by our findings that villins associate with the actin foci during their formation and the actin-bundling activity of villins is required for the enlargement of actin foci. Thus, our data also implicate a role for villins in a bundling activity involved in the formation of punctate foci in pollen tubes. It was previously shown that the SI-induced actin foci are highly stable, being resistant to 1 µM LatB (Poulter et al., 2010). Our findings that the actin foci are decorated by villins helps explain why the actin foci are very stable, as it has been shown that VLN1-decorated actin filaments are resistant to the action of ADF and LatB (Huang et al., 2005).

However, exactly how and why these structures form, as well as what their function is, is still largely unknown. It is worth mentioning that a previous study revealed ADF colocalizing with the actin foci (Poulter et al., 2010), which makes sense if there is a change in activity from severing to bundling as a result of a change in pH. Therefore, it is possible that ADF and villins may act in concert or cooperatively. Moreover, CAP may also be involved in this remodeling, together with villins, as CAP was previously colocalized with these actin foci (Poulter et al., 2010). In budding yeast (Saccharomyces cerevisiae), the formation of F-actin ‘bodies’ is dependent on Srv2p/CAP activity (Gourlay et al., 2004); moreover, this study showed that a decrease in actin dynamics can induce cell death involving reactive oxygen species (ROS). As it has been shown that both actin depolymerization and stabilization can stimulate PCD in pollen tubes (Thomas et al., 2006), this further suggests that mechanistically, the stabilization of the actin cytoskeleton somehow plays a role in modulating PCD in these pollen tubes. Indeed, it has been shown that SI stimulates increases in ROS and that A23187 also does so; downstream, actin foci are formed and PCD involving a DEVDase activity is triggered (Wilkins et al., 2011). However, exactly what physiological functions these foci perform and how they achieve these outcomes is not known. Nonetheless, our study yields significant insights into the mechanisms involved in the formation and enlargement of actin foci in pollen tubes.

In summary, our study uncovers molecular details and the mechanism underlying actin alterations in response to A23187 and assumed accompanying elevation in [Ca²⁺]_cyt in pollen tubes. Because the events observed here are so similar to those observed during an authentic SI response, this study provides significant insights into the cellular mechanisms of the SI response in Papaver pollen tubes. Nevertheless, it will be important in the future to study the SI response in Arabidopsis lines expressing PrpS to investigate and confirm that these events are involved in authentic SI responses and to uncover further details of the mechanisms and ABPs involved in this actin remodeling.

Arabidopsis thaliana Columbia ecotype (Col-0) was used as the wild-type plant. The information of T-DNA insertion mutants of VLN2, VLN3 and VLN5 as well as the transgenic plants harboring GFP fusion constructs: pK7FWG2-VLN2p:VLN2-EGFP;vln2, pCambia1301-VLN3p:VLN3-EGFP;vln3 and pK7FWG2-VLN5p:VLN5-EGFP;vln5 have been described previously (Bao et al., 2012; Zhang et al., 2010). Arabidopsis transgenic plants harboring VLN5 that was deficient in either severing or bundling actin filaments were generated as described below. The At-ntp303p::PrpS₁-GFP transgenic lines are described in de Graaf et al. (2012). Briefly, Col-0 A. thaliana was transformed with the Ti vector pGreen0029-ntp303p::PrpS₁-GFP. The ntp303 is a pollen-specific promoter. This line is referred in the text as At-PrpS₁ for simplicity. Arabidopsis plants were grown on half-strength Murashige and Skoog medium or in soil at 22°C under a photoperiod regime of 16 h light and 8 h dark.

To generate the recombinant VLN5^SD protein (severing-deficient VLN5; Fig. 5A), we made a prokaryotic expression construct to produce the necessary 934 bp fragment, as described previously (Revenu et al., 2007). Briefly, this was amplified using pET23d-VLN5 (Zhang et al., 2010) as the temperate with primer pair VLN5CDS_F and VLN5CDS_R (Table S1). The amplified fragment was subsequently moved into pMD19-T, which was used as the template for subsequent PCR amplification using primer pairs VLN5^SD_1F and VLN5^SD_1R, VLN5^SD_2F and VLN5^SD_2R (Table S1). The fragment digested with KpnI/MluI was moved into pET23d-VLN5 restricted with the same enzymes to generate the pET23d-VLN5^SD. To construct the pET23d-VLN5ΔHL plasmid, the PCR product was amplified with primer pair VLN5ΔHL_F and VLN5ΔHL_R (Table S1) using pET23d-VLN5 (Zhang et al., 2010) as the template and the error-free PCR fragment restricted with SacI/EcoRI was subsequently moved into pET23d to generate pET23d-VLN5ΔHL.

To generate VLN5, VLN5^SD and VLN5^ΔHL recombinant proteins, the plasmids pET23d-VLN5, pET23d-VLN5^SD and pET23d-VLN5^ΔHL were transformed into the Escherichia coli BL21 DE3 strain. After induction with the addition of Isopropyl β-D-1-thiogalactopyranoside (final concentration, 0.4 mM) at 16°C for ∼12–16 h, the cells were collected by centrifugation. They were subsequently lysed by sonication, and the protein in the supernatant was precipitated with (NH₄)₂SO₄ (final concentration, 45%). After centrifugation (10,000 g, 30 min, 4°C), the pellet was resuspended in Binding Buffer (5 mM imidazole, 250 mM KCl, 25 mM Tris-HCl, pH 8.0). After clarification at 10,000 g for 30 min at 4°C, the supernatant was loaded onto a Nickel Sepharose column, and purification was performed according to manufacturer's instruction. The purified protein was dialyzed against 5 mM Tris-HCl, pH 8.0. The protein was aliquoted and freshly frozen in liquid nitrogen and stored at −80°C. Skeletal muscle actin was purified according to previously published methods (Pollard, 1984; Spudich and Watt, 1971). To monitor the kinetic actin assembly and disassembly, actin was conjugated with pyrene iodoacetamide, as described previously (Pollard, 1984). To directly visualize the dynamics of individual actin filaments, actin was labeled with Oregon Green, as described previously (Amann and Pollard, 2001).

Pollen was collected from freshly opened Arabidopsis thaliana flowers, as described previously (Chang and Huang, 2015; Chang and Huang, 2017). Total pollen RNA was extracted by Eastep^® Super Total RNA Extraction Kit (Promega) following the manufacturer's instructions. 5 µg RNA was used for reverse transcription by MMLV reverse transcriptase (Promega) to synthesize cDNA. To determine the transcript levels of VLN5^SD and VLN5ΔHL, transcripts of VLN5^SD and VLN5ΔHL were amplified with primer pair qVLN5 F and qVLN5 R (Table S1). eIF4A was amplified with primer pair eIF4A F and eIF4A R (Table S1) as the internal control. Amplification was completed using GenStar qPCR Green Mix (Genstar), running on an Applied Biosystems^® 7500 fast Real-Time PCR System. The transcription levels of genes were calculated using the 2^ΔCt method (Livak and Schmittgen, 2001), in which ΔCt=Ct(eIF4A) – Ct(gene).

In vitro germination of A. thaliana pollen from wild-type and villin loss-of-function mutants was performed as described previously (Ye et al., 2009). For live-cell imaging, A. thaliana lines expressing Lat52:Lifeact-eGFP were used (Qu et al., 2013). Briefly, the newly open flowers were collected and dipped on the surface of pollen germination medium [GM: 1 mM CaCl₂, 1 mM Ca(NO₃)₂, 1 mM MgSO₄, 0.01% (w/v) H₃BO₃, 18% (w/v) sucrose; solidified with addition of 0.8% (w/v) agarose] and incubated at 28°C under high humidity for ∼3 h to allow growth of pollen tubes. After the average length of pollen tubes reached ∼150 μm, A23187 (10 or 50 μM in liquid GM) was added onto the surface of the solid GM containing pollen tubes. After incubation for various periods of time at 28°C, pollen tubes were fixed and subjected to staining with Alexa Fluor 488-conjugated phalloidin, as described below. Alternatively, to visualize the dynamics of actin filaments in pollen tubes after treatment with A23187, pollen tubes expressing Lat52:Lifeact-eGFP were observed immediately after the application of A23187 with a spinning disk confocal microscope, as described below.

Full details of the At-PrpS₁-GFP line and reconstituting the SI response in an in vitro bioassay using this line are described in de Graaf et al. (2012). Briefly, we grew the pollen from this line in vitro to obtain an average length of about 150 μm, before adding recombinant PrsS₁ (Foote et al., 1994) at 20 ng μl⁻¹ to give a cognate PrpS₁–PrsS₁ interaction to trigger an SI response. After incubation for various periods of time at 25°C, pollen tubes were fixed and subjected to staining with Alexa Fluor 488-conjugated phalloidin, as described below.

For the A23187 experiments, actin filaments in A. thaliana pollen tubes were revealed by staining with Alexa Fluor 488-phalloidin, as described previously (Wu et al., 2010). Briefly, when the average length of pollen tubes reached ∼150 µm, 300 μM N-(maleimidobenzoyloxy)-succinimide (MBS) in liquid GM was added and incubated for 1 h at 28°C. The samples were subsequently washed with TBS-T (200 mM NaCl, 400 mM sucrose and 0.05% Nonidet P-40 in 50 mM Tris-HCl, pH 7.5) three times. After washing, 200 nM Alexa Fluor 488-phalloidin (Invitrogen) was added to the sample and incubated at 4°C, overnight. The samples of phalloidin-stained pollen tubes were observed under an Olympus FluoView FV1000 confocal microscope equipped with a ×100 UAPON objective (numerical aperture of 1.49). The z-series images were collected with the size of z-step set at 0.7 µm.

For the SI experiments, actin filaments in A. thaliana pollen tubes were revealed by staining with Alexa Fluor 488-phalloidin according to Wilkins et al. (2011). Briefly, the SI response was stopped by adding MBS in liquid GM to a final concentration of 400 µM and incubated for 6 min at room temperature. The samples were fixed with 2% formaldehyde for 30 min, followed by washes in TBS (50 mM Tris-HCl, pH 7.4, 200 mM NaCl). Pollen tubes were permeabilized by treatment in TBS+0.1% Triton X-100 for 40 min. F-actin was labeled by adding 66 nM Rhodamine-phalloidin for >30 min at room temperature. Phalloidin-stained pollen tubes were observed under a Leica SP8 inverted scanning confocal microscope equipped with a ×100 CS2 objective (numerical aperture of 1.40). The z-series images were collected with the size of z-step set at 0.7 μm.

To visualize the dynamics of actin filaments, Arabidopsis pollen tubes expressing Lat52:Lifeact-eGFP were observed under a Olympus DSU-IX81 spinning disk confocal microscope equipped with a ×100 oil objective (1.4 numerical aperture). The time-lapse z-series images were collected at 2 s intervals with the z-step set at 0.5 µm. The excitation wavelength was set at 488 nm, and the emission wavelength was set at 505–545 nm.

Image processing and measurements were performed using ImageJ software (v. 1.51 s, http://imagej.nih.gov/ij). To analyze the size of the actin foci in pollen tubes, z-stack images of actin foci were selected using the ‘rectangle’ tool and the mean gray value of the selected region was calculated using the ‘Analyze-Measure’ plugin. To illustrate the actin distribution after A23187 treatment (Fig. 4B), z-stack images were thresholded to create a binary image, followed by a skeletonization procedure (Process-Binary-Skeletonize) so all actin filaments were shown as black, rather than gray scale. Output images were subsequently made into full projections that were subsequently used for the quantification of the ‘occupancy’ of actin filaments in pollen tubes. The regions of interest (ROI) in pollen tubes were selected using the built in ‘Polygon selection tool’. The total number of pixels and the number of black pixels in the images was counted. Occupancy is defined as the ratio that the number of black pixels divided by the total number of pixels in the ROI. Thus, number of black pixels gave a quantification of actin microfilaments.

To measure the size of actin foci, z-stack images were thresholded to create a binary image, then the region of interest in pollen tubes were selected using the Polygon selection tool in ImageJ. The outlines of foci were analyzed by the ‘Analyze particles’ plugin in ImageJ, and the size of the actin foci was measured using the ‘Measure’ plugin in ImageJ. Foci with area smaller than 0.04 μm² were dropped from the measurements in order to avoid the background noise. In addition, foci with area larger than 2 μm² were also discounted in order to avoid the inclusion of multiple foci that cannot be separated automatically.

High-speed and low-speed F-actin co-sedimentation assays were performed according to previously published methods (Kovar et al., 2000b). Preassembled actin filaments at 3 µM were incubated with various concentrations of VLN5 or its variants in 1×KMEI reaction system (50 mM KCl, 1 mM MgCl₂, 1 mM EGTA and 10 mM imidazole, pH 8.0) at room temperature for 30 min, the mixtures were subsequently centrifuged at 13,600 g at 4°C for 30 min for low-speed F-actin co-sedimentation experiments. For high-speed F-actin co-sedimentation experiments, the mixtures were centrifuged at 160,000 g at 4°C for 1 h. The supernatants and the pellets were resolved by SDS-PAGE and stained with Coomassie Brilliant Blue R. The relative amount of protein in supernatants and pellets was quantified by densitometry.

Dilution-mediated actin depolymerization assay was performed as previously described (Huang et al., 2003; Shi et al., 2013). Briefly, preassembled actin filaments (50% pyrene labeled) at 5 μM were incubated with VLN5 or its variants for 5 min at room temperature, the mixture was then diluted 50-fold with G buffer (5 mM Tris-HCl, pH 8.0, 0.2 mM ATP, 0.1 mM CaCl₂, 0.5 mM DTT and 0.1 mM imidazole) to induce actin depolymerization. Actin depolymerization was traced by monitoring the changes in pyrene fluorescence with a QuantaMaster Luminescence QM 3 PH Fluorometer (Photon Technology International, Inc.) with the excitation and emission wavelengths set at 365 nm and 407 nm, respectively.

The experiment was performed according to previously published methods (Zhang et al., 2016; Zhang et al., 2019). Briefly, F-actin at 2 μM was incubated with 500 nM VLN5 or its protein variants in the presence of 2 μM Rhodamine-phalloidin (Sigma-Aldrich) in 1× KMEI for 30 min at room temperature. The mixtures were subsequently diluted to 50 nM in TIRF buffer before observation under a BX53 microscope (Olympus) equipped with a 60×, 1.42 numerical aperture oil objective. The images were acquired with Olympus DP80 camera, using Cell Sens Standard 1.12 software.

This experiment was performed according to previously described methods (Jiang and Huang, 2017; Kuhn and Pollard, 2005). Briefly, 50% Rhodamine-labeled rabbit muscle actin was preclarified at 200,000 g for 2 h at 4°C, and then this actin at 10 μM was preassembled at room temperature in the dark. The prepared flow chambers were coated with 10 nM NEM-myosin for 2 min, followed by consecutive washes with HS-BSA buffer (50 mM Tris-HCl, pH 7.5, 600 mM NaCl, 1% BSA) and LS-BSA buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1% BSA) for 2 min. They were then washed with TIRF buffer (50 mM KCl, 1 mM MgCl₂, 1 mM EGTA, 10 mM imidazole, 100 mM DTT, 0.2 mM ATP, 15 mM glucose, 0.5% methylcellulose, 20 mg ml⁻¹ catalase, 100 mg ml⁻¹ glucose oxidase, pH 7.4). After moving to the microscope stage, preassembled actin filaments at 500 nM in TIRF buffer were injected into the flow cell and incubated for 5 min in the dark. After finding the microscopic field of interest, VLN5 and its protein variants or the control buffer diluted in TIRF buffer containing 1.0 μM free Ca²⁺ was injected into the flow chambers. Actin filaments were observed by Olympus IX-71 microscope equipped with a ×100, 1.45 numerical aperture Planapo objective (Olympus) by TIRF illumination. After finding the focal plane, time-lapse images were captured at 3 s intervals by Hamamatsu ORCA-EM-CCD camera (model C9100-12) driven by Micro-Manager software (www.micro-manager.org; v1.4; MMStudio). The severing activity of VLNs was quantified by measuring the average actin filament severing frequency (at least 15 actin filaments ≥5 μm were chosen from each treatment), which was defined as the number of breaks per unit filament length, per unit time (breaks μm⁻¹ s⁻¹) (Khurana et al., 2010; Zhang et al., 2010).

We thank the members of the Huang laboratory for helpful discussions. We thank Hua Zhang for generating the prokaryotic expression construct of severing deficient VLN5.

The peer review history is available online at https://jcs.biologists.org/lookup/doi/10.1242/jcs.237404.reviewer-comments.pdf