Live cell imaging reveals actin-cytoskeleton-induced self-association of the actin-bundling protein WLIM1

As a vital structure of eukaryotic cells, the actin cytoskeleton is a key player in numerous processes, including cell division and elongation, vesicle and organelle trafficking, adhesion and motility, and establishing polarity. In line with this wide range of functions, the actin cytoskeleton exhibits a high degree of structural plasticity allowing a highly dynamic rearrangement of its basic elements, namely the actin filaments. The creation and turnover of actin filaments, as well as their assembly into higher-order structures, are tightly regulated at spatial and temporal levels by a plethora of actin-binding proteins, which control nucleation, polymerization, capping, severing and crosslinking (dos Remedios et al., 2003; Higaki et al., 2007; Staiger and Blanchoin, 2006; Thomas et al., 2009; Winder and Ayscough, 2005). Recent live cell studies combining reliable fluorescent actin markers with novel high-resolution imaging techniques such as spinning-disc confocal microscopy and variable-angle epifluorescence microscopy (VAEM) have provided key insights into actin cytoskeleton dynamics at the cell cortex of plant cells (Augustine et al., 2011; Era et al., 2009; Henty et al., 2011; Henty-Ridilla et al., 2013; Khurana et al., 2010; Konopka and Bednarek, 2008; Li et al., 2012; Smertenko et al., 2010; Staiger et al., 2009; Tóth et al., 2012). In striking contrast to in vitro actin treadmilling (Bugyi and Carlier, 2010; Selve and Wegner, 1986), in vivo remodeling of plant cortical actin arrays was found to follow a so-called ‘stochastic dynamics’ process that is dominated by fast elongation and prolific severing of single filaments (Blanchoin et al., 2010; Okreglak and Drubin, 2010; Staiger et al., 2009). Consistently, single actin filaments have a surprisingly short lifetime, ∼20 seconds in Arabidopsis hypocotyl epidermal cells, and reach a maximum length of 12–15 µm before disappearing. In this context, the crosslinking of actin filaments into thick bundles emerges as a strategy ‘used’ by cells to shape more stable and organized cytoskeletal structures. Indeed, probably as a result of their higher resistance to severing factors, actin bundles exhibit significantly longer lifetimes (Smertenko et al., 2010; Staiger et al., 2009). In addition, actin bundles appear stiffer than single actin filaments and tend to align with the long cell axis (Era et al., 2009; Henty et al., 2011; Staiger et al., 2009; Vidali et al., 2009).

Actin bundles are found in all plant cells and are involved in cytoplasmic streaming and myosin-driven movement of organelles and vesicles (Smith and Oppenheimer, 2005; Thomas, 2012; Thomas et al., 2009; Walter and Holweg, 2008). The role of actin bundles as long-distance tracks has been particularly well described in pollen tubes, where longitudinally arranged actin bundles enable the directional transport of Golgi-derived vesicles from the shank to the apex (Ketelaar et al., 2002; Lenartowska and Michalska, 2008; Miller et al., 1999; Ren and Xiang, 2007; Tominaga et al., 2000). The recent characterization of pollen-enriched villin, fimbrin and LIM proteins further corroborates the importance of actin-bundling processes for proper pollen tube elongation (Papuga et al., 2010; Staiger et al., 2010; Wang et al., 2008; Wu et al., 2010; Zhang et al., 2010). Additional, more specific, roles have been assigned to actin bundling during guard cell and chloroplast movements (Higaki et al., 2010b), cell growth and morphogenesis (Baluska et al., 2001; Higaki et al., 2010a; Nick, 2010; Nick et al., 2009; Smith and Oppenheimer, 2005) and the set-up of plant defense responses against pathogens (Clément et al., 2009; Day et al., 2011; Hardham et al., 2007; Henty-Ridilla et al., 2013; Opalski et al., 2005; Schmidt and Panstruga, 2007; Takemoto et al., 2006).

Actin-bundling proteins exhibit a modular organization and combine one or more actin-binding domains (ABDs), with regulatory domains conferring sensitivity to specific stimuli such as phospholipids or variations in pH or calcium concentration (Bartles, 2000; Furukawa and Fechheimer, 1997; Huang et al., 2006; Li et al., 2012; Papuga et al., 2010; Puius et al., 1998; Wang et al., 2008). Despite the frequent pairwise arrangement of their ABDs, some actin-bundling proteins undergo dimerization to crosslink actin filaments (Mimura and Asano, 1986; Thomas, 2012; Thomas et al., 2009). For instance, dimerization as a prerequisite for actin-bundling activity has been reported for members of the evolutionarily conserved villin and formin families and for the recently described, plant-specific SCAB1 protein (Chhabra and Higgs, 2006; George et al., 2007; Harris et al., 2004; Li and Higgs, 2005; Michelot et al., 2006; Michelot et al., 2005; Mimura and Asano, 1986; Xu et al., 2004; Yokota et al., 1998; Zhang et al., 2012). By contrast, fimbrins use the close proximity of their two ABDs to induce bundle formation and function as monomers in yeast, animals and plants (Klein et al., 2004; Nakano et al., 2001; Volkmann et al., 2001). We previously reported that both LIM domains of plant LIM proteins, which each consist of a double zinc finger motif, display intrinsic actin-binding activity in vitro (Thomas et al., 2007). It is therefore possible that plant LIMs crosslink actin filaments in a monomeric form, although dimerization or oligomerization cannot be ruled out. Indeed, a recent study focusing on the nuclear functions of plant LIM proteins suggested that tobacco WLIM2 (NtWLIM2), which also acts as an actin-bundling protein in the cytoplasm, forms dimers in the nucleus of tobacco cells (Moes et al., 2013). In support of dimerization, the mammalian counterparts of plant LIM proteins, namely cysteine-rich proteins (CRPs), have been reported to dimerize both in vitro and in live cells (Arber and Caroni, 1996; Boateng et al., 2007; Feuerstein et al., 1994). However, the oligomerization status of LIM proteins in the cytoplasm of plant cells and, more particularly, along the actin cytoskeleton has not been directly assessed so far. Despite the importance of actin bundling in plant cells and the increasing number of plant actin-bundling proteins identified, the questions as how actin filaments are brought into contact by actin-bundling proteins and how they are arranged inside the bundle have been rarely addressed. A recent article by Khurana and co-workers (Khurana et al., 2010) reported that Arabidopsis VILLIN1 and VILLIN3 promote actin bundle formation by a so-called ‘catch and zipper’ mechanism. However, whether this applies to other types of plant actin-bundling proteins remains unclear. To the best of our knowledge, the relative orientation of actin filaments within bundles induced by a plant actin-bundling protein was only reported in the case of the non-processive Arabidopsis formin AFH1 (Michelot et al., 2006).

In the present work, we used state-of-the-art imaging approaches to dissect the molecular mechanisms underlying tobacco WLIM1 (NtWLIM1)-induced actin bundle formation in both in vitro reconstituted assays and live cells. Together, our data support a multistep process in which self-association of F-actin-induced NtWLIM1 operates as a driving force for the zippering of NtWLIM1-decorated actin filaments.

We previously reported that NtWLIM1 promotes the formation of actin bundles in vitro and in live cells (Thomas et al., 2006; Thomas et al., 2007). Here, we aimed to characterize NtWLIM1 bundling activity in a more quantitative manner as well as the underlying molecular mechanisms. Fluorescently labeled actin filaments (1 µM) were co-polymerized with 0–4 µM of NtWLIM1 and subsequently imaged by TIRF microscopy. The corresponding pictures were pseudo-colored as a function of pixel fluorescence intensities (see examples shown in Fig. 1A–C), whose distributions were illustrated by histograms (supplementary material Fig. S1A–C). In control experiments (no NtWLIM1), single actin filaments formed a randomly organized array (Fig. 1A). In the presence of NtWLIM1, filaments assembled into long and rather straight bundles (Fig. 1B,C).

Noticeably, actin bundles grew larger with increasing NtWLIM1 concentrations. Fluorescence-intensity-based measurements revealed that a maximal average number of 3–5 filaments per bundle was reached at an actin-to-NtWLIM1 ratio of 1∶3 (Fig. 1D). Consistently, the same actin-to-NtWLIM1 ratio yielded the highest skewness value (degree of asymmetry of pixel intensity distribution; Higaki et al., 2010b; supplementary material Fig. S1D).

The distribution of NtWLIM1 along emerging actin bundles was assessed using a NtWLIM1 protein labeled with HiLyte Fluor™ 488 (subsequently referred to as LIM1-488). Remarkably, LIM1-488 promoted the formation of actin bundles in a similar manner to unlabeled NtWLIM1 (supplementary material Fig. S2A–E and data not shown). As shown in supplementary material Fig. S2A–C, LIM1-488 was distributed along the whole length of actin bundles. To test the ability of NtWLIM1 to bind to individual filaments, we lowered the LIM-488-to-actin ratio (1∶1) and thereby kept a large fraction of filaments in a non-bundled form (Fig. 1B). For a higher resolution, samples were immobilized onto NEM-myosin-coated coverslips and imaged by TIRF microscopy. LIM1-488 decorated a mixture of bundles and fine fibers, which we assumed to be individual, non-crosslinked actin filaments (supplementary material Fig. S2D; asterisks and arrowheads, respectively). Interestingly, as revealed by Rhodamine-Phalloidin co-labeling (supplementary material Fig. S2E), the fluorescent signal intensity due to LIM1-488 directly correlated with bundle thickness. A close-up of the LIM1-488-decorated and Rhodamine-Phalloidin co-labeled fine fibers (supplementary material Fig. S2F,G, respectively) confirmed that they exhibited a shape and intensity that was similar to those of control filaments polymerized without LIM-488 (supplementary material Fig. S2H). We therefore conclude that NtWLIM1 can efficiently bind to individual actin filaments, and that this event probably precedes actin filament crosslinking.

The formation of actin bundles induced by NtWLIM1 was monitored by time-lapse TIRF microscopy. In the absence of NtWLIM1, growing actin filaments occasionally co-aligned in a transient manner, but never formed bundles of two or more filaments over a long distance (>2 µm) and/or time (>5 second; supplementary material Movie 1). The addition of NtWLIM1 promoted lateral contacts between filaments and their subsequent crosslinking in parallel bundles. In two representative examples, the extremity of an individual filament contacted a close-by fine bundle in its terminal or more central region, respectively, and subsequently fused with this bundle by a zippering-like process (Fig. 2A,B; supplementary material Movies 2,3). NtWLIM1-induced bundles were observed to elongate by two distinct mechanisms. On the one hand, they grew at their terminal regions by the addition of pre-formed short filaments (Fig. 2C; supplementary material Movie 4). Interestingly, these short filaments sometimes undertook several docking attempts before finally joining the targeted bundle (Fig. 2D; supplementary material Movie 5). These ‘touch and flip’ events probably reflect local steric and/or electrostatic constraints, which favor a preferential relative orientation between the added filament fragments and growing bundles. On the other hand, bundle elongation was achieved through polymerization of the inside filaments (Fig. 2E, arrowheads; supplementary material Movie 6). As shown by the kymograph in Fig. 2F, filaments elongate with similar growth rates both inside and at the extremities of bundles. In addition, we calculated similar average growth rates for NtWLIM1-bundled actin filaments and individual actin filaments polymerized alone (0.6±0.1 µm/minute), indicating that WLIM1 does not significantly alter the actin polymerization rate.

Advanced live cell imaging approaches have recently enabled high-resolution monitoring of actin cytoskeleton remodeling in plant cells (Henty et al., 2011; Li et al., 2012; Smertenko et al., 2010; Staiger et al., 2009). To validate and extend our in vitro biochemical data on LIM-protein-mediated actin bundle assembly, we applied variable-angle epifluorescence microscopy (VAEM) on Nicotiana benthamiana seedlings transiently expressing NtWLIM1 fused to GFP (GFP-NtWLIM1). In particular, we wanted to answer the following questions: does NtWLIM1 bind to individual actin filaments in live cells? If so, do such filaments undergo bundling through mechanisms similar to those observed in in vitro reconstituted assays? In hypocotyl epidermal cells, GFP-NtWLIM1 decorated single actin filaments, which elongated by fast polymerization and disintegrated by severing and depolymerization events (data not shown). Moreover, we noticed pair-wise interaction between single filaments or fine bundles, which subsequently zippered into thicker fibers. Because the biolistic approach used to express NtWLIM1-GFP in tobacco hypocotyl only resulted in a low number of transformed cortical epidermal cells (the cell type amenable to VAEM analysis) and variable transgene expression levels, we switched to transgenic Arabidopsis seedlings stably expressing the Arabidopsis WLIM1 protein fused to GFP (GFP-AtWLIM1) to conduct more quantitative and statistically relevant analyses. Noticeably, AtWLIM1 shares 86% sequence similarity (79% sequence identity) with NtWLIM1 and exhibits equivalent actin-binding and -bundling activities in vitro (Papuga et al., 2010; Thomas et al., 2006). As in the tobacco system, GFP-AtWLIM1 extensively decorated actin cytoskeleton components in Arabidopsis hypocotyl cells, ranging from individual filaments to thick bundles (Fig. 3A; supplementary material Movie 7). In agreement with previous imaging studies using standard fluorescent F-actin reporters (Smertenko et al., 2010; Staiger et al., 2009) (Henty et al., 2011; Li et al., 2012; Staiger et al., 2009; Tóth et al., 2012), actin cytoskeleton remodeling (Fig. 3A) was dominated by fast polymerization and prolific severing (Fig. 3B). Tracking of individual filaments revealed that many of them were readily destroyed by severing and usually exhibited lifetimes <30 seconds. To check whether the binding of AtWLIM1 to single actin filaments affects the dynamics of these filaments in vivo, we quantified well-defined stochastic dynamics parameters such as filament elongation rates, severing frequency and convolutedness (Staiger et al., 2009) in the cortex of hypocotyl epidermal cells from 5-day-old Arabidopsis seedlings expressing GFP-AtWLIM1 or the widely used GFP-fABD2 actin marker (supplementary material Table S1, Movie 8). Both filament elongation rates and severing frequency showed statistically significant but modest differences in GFP-AtWLIM1 and GFP-fABD2 seedlings. In addition, filament convolutedness was not statistically different in GFP-AtWLIM1 and GFP-fABD2 plants. Together, these findings indicate that overexpression of WLIM1 is not detrimental to single actin filament dynamics. In addition, every optical section analyzed over a period of 200 seconds revealed filaments that interacted with each other and subsequently zippered into bundles (Fig. 3C; supplementary material Movie 9). Since several zippering events could be observed over the whole time of observation, we concluded that these events are relatively frequent and represent a major mode of actin bundle formation in vivo. Statistical quantification analyses of these events in GFP-AtWLIM1 hypocotyl epidermal cells yielded a bundling frequency of 2.0×10⁻⁴±6.7×10⁻⁵ zippering events⋅µm⁻²⋅second⁻¹ (supplementary material Table S1). A parallel analysis conducted in the hypocotyls of seedlings expressing GFP-fABD2 (Sheahan et al., 2004; Staiger et al., 2009; Tóth et al., 2012) indicated a threefold lower frequency of filament bundling processes (6.9×10⁻⁵±2.6×10⁻⁵ zippering events⋅µm⁻²⋅second⁻¹). Together, these data are consistent with an in vivo actin bundling activity of AtWLIM1. Compared with individual filaments, GFP-AtWLIM1-labeled actin bundles exhibited less-convoluted shapes and longer lifetimes (Fig. 3A; supplementary material Movie 7). As in NtWLIM1-based in vitro bundling assays, bundles grew by both elongation of internal filaments and fusion/recycling of nearby-severed filament fragments. In addition to straight bundles, some ring-shaped bundles were observed to form through actin filament circularization (supplementary material Movie 10). Similar ring-shaped bundles have been previously referred to as acquosomes and have been proposed to serve as actin storage organelles (Smertenko et al., 2010). However, because we never observed acquosomes in vitro, their formation probably requires additional or simply factors other than LIM proteins.

On the whole, live cell analyses are highly consistent with our in vitro data and support a model in which LIM proteins bind to both individual filaments and bundles and promote the zippering of filaments or bundles that come into contact with each other. Furthermore, the unaffected dynamics of single AtWLIM1-decorated filaments support the idea that WLIM1 proteins promote actin bundling directly through their actin-crosslinking activity.

Depending on their respective functions, actin bundles can be either of uniform polarity with the barbed ends of all filaments pointing in the same direction (parallel filaments) or of mixed polarity (anti-parallel filaments). To measure the polarity of NtWLIM1-induced actin bundles, we first measured actin bundle polarity using a dual labeling fluorescence microscopy assay (Harris et al., 2006). In brief, actin (1 µM) was co-polymerized with NtWLIM1 (2 µM) and subsequently labeled by Alexa-Fluor-488-phalloidin. These bundles were incubated with profilin-bound actin monomers (mixed at a profilin-to-actin ratio of 4∶1) in order to inhibit actin polymerization at the pointed end of filaments and resume filament elongation exclusively at the barbed end (Kovar et al., 2006; Pantaloni and Carlier, 1993; Pollard and Cooper, 1984; Pring et al., 1992). Finally, these newly polymerized bundle sections were labeled with Rhodamine-Phalloidin. Data revealed that 75% (n = 82) of NtWLIM1-induced bundles resumed bi-directional growth as indicated by red filaments emerging from each extremity of the green initial bundle (Fig. 4A, left). The rest of the bundles (25%) showed a unipolar conformation, with only one of their extremities labeled in red (Fig. 4A, middle). By contrast, control experiments with the human actin-bundling protein fascin yielded almost exclusively unipolar bundles (90%, n = 20) (Fig. 4A, right), a result consistent with the high selectivity previously reported for fascin (Breitsprecher et al., 2011; Courson and Rock, 2010; Ishikawa et al., 2003; Skau et al., 2011).

To further examine actin filament orientation in NtWLIM1-triggered bundles, we traced fast barbed-end elongation of individual and bundled filaments using real-time TIRF microscopy (Fig. 4B–D). The majority of bundles (≈90%) contained several filaments that grew in opposite directions (Fig. 4C). In addition, we regularly observed that anti-parallel filaments, although elongating toward each other, crossed and continued their growth along the bundle axis (Fig. 4D; supplementary material Movie 11). In accordance with the above dual-fluorescence labeling experiments, some NtWLIM1-induced bundles (≈10%) exhibited an exclusively parallel filament orientation (data not shown). Together, these data indicate that NtWLIM1 has no or only weak selectivity for actin filament polarity and predominantly generates bundles of mixed polarity.

To extend the above data, we tracked the orientation of elongating filaments after they zippered into fine growing bundles in Arabidopsis hypocotyl cells expressing GFP-AtWLIM1 (supplementary material Movie 12). Data show that 55% of filaments that zipper into a thin bundle were elongating in the same direction, whereas the rest of the analyzed population (45%) grew in opposite directions. These findings are therefore consistent with the above-characterized weak intrinsic selectivity of NtWLIM1 for actin filament polarity. It should be noted that we could not reliably characterize the relative orientation of actin filaments in thick bundles. We assume that most of these bundles have a mixed polarity, although this awaits confirmation.

The dimerization of NtWLIM1 in live cells as well as its potential implication in actin bundling (Thomas et al., 2007) were directly assessed by two complementary approaches, namely bimolecular fluorescence complementation (BiFC) (Hu et al., 2002; Hu and Kerppola, 2003; Kodama and Hu, 2012) and fluorescence resonance energy transfer based on fluorescence lifetime imaging microscopy (FLIM-FRET) (Becker, 2012; Ishikawa-Ankerhold et al., 2012). In the first set of experiments, two BiFC constructs, consisting of complementary N-terminal and C-terminal fragments of the enhanced yellow fluorescent protein (eYFP) fused to NtWLIM1 (YN-NtWLIM1 and YC-NtWLIM1, respectively) were co-expressed in tobacco BY-2 cells and checked for their association. An intense signal of reconstituted eYFP indicative of NtWLIM1–NtWLIM1 interaction was observed in almost 90% of cells (Fig. 5Aa). Remarkably, as revealed by treatment with the F-actin-disrupting drug Latrunculin B (Lat B), NtWLIM1 BiFC complexes sharply decorated the actin cytoskeleton (supplementary material Fig. S3A–D). This contrasted with the dual diffuse cytoplasmic and cytoskeletal localization observed when NtWLIM1 was fused to full-length eYFP (Fig. 5Ab). Together, these data support the idea that NtWLIM1 complexes concentrate along the actin cytoskeleton and might hint at the existence of NtWLIM1 monomeric and di/oligomeric pools. However, we cannot exclude the possibility that the diffuse cytoplasmic localization observed for eYFP-NtWLIM1 results from the interference of the full-length eYFP tag with NtWLIM1 self-association and/or by the elevated levels of overexpressed eYFP-NtWLIM1 and resulting cytoplasmic accumulation of the fusion protein occurring on top of its appropriate cytoskeletal localization. No significant fluorescence could be detected in control cells transformed with the two empty BiFC vectors, nor with YN- or YC-NtWLIM1, on the one hand, and the complementary empty vector on the other hand (data not shown). To confirm that the formation of NtWLIM1 BiFC complexes was not only due to the close vicinity of both NtWLIM1 BiFC constructs on the actin cytoskeleton, we performed additional control assays with complementary BiFC constructs respectively fused to WLIM1 and another actin filament-binding protein, namely the actin binding domain 2 of Arabidopsis FIMBRIN1 (fABD2) (Ketelaar et al., 2004). In contrast to the strong signal typically observed for NtWLIM1–NtWLIM1 complexes, no fluorescence could be detected in 75% of cells co-expressing YN-NtWLIM1 and YC-fABD2 (Fig. 5Ac). The rest of cells (25%) exhibited significant but relatively weak fluorescence (Fig. 5Ad). This BiFC signal could be entirely abolished by pre-treating cells with Lat B (data not shown), indicating that it was due to unspecific BiFC complexes that formed along the cytoskeleton. However, as previously stated, both the frequency and intensity of unspecific actin-associated BiFC complexes were much lower than those observed for NtWLIM1–NtWLIM1 BiFC complexes, supporting the specificity of the latter.

In addition to BiFC analyses, NtWLIM1 self-interaction was evaluated using FLIM-based FRET measurements (Becker, 2012; Ishikawa-Ankerhold et al., 2012). Basically, the extent of quenching of a donor fluorophore lifetime by an acceptor fluorophore allowed us to quantitatively evaluate the physical interaction between the respectively fused candidate proteins. C-terminal fusions of NtWLIM1 to AmCyan and eYFP were used as donor and acceptor constructs (AmCyan-NtWLIM1 and eYFP-WLIM1, respectively). For AmCyan-NtWLIM1 alone, we measured a fluorescence lifetime of 2.83±0.08 nseconds (Fig. 5Ba–b′,C), a value similar to those previously published for AmCyan-based donor constructs (Ismail et al., 2010). In the presence of eYFP-NtWLIM1, the overall fluorescence lifetime of AmCyan-NtWLIM1 significantly decreased, indicating the presence of NtWLIM1 complexes (Fig. 5B). Importantly, particularly low values were reached along the actin cytoskeleton, supporting the asymmetrical subcellular distribution of NtWLIM1 complexes (Fig. 5Bc–d′, arrowheads). Along actin bundles, we measured an average lifetime of 2.43±0.08 nseconds (Fig. 5C) corresponding to a FRET efficiency of 14%. As a control, we performed additional FLIM-FRET analyses on cells co-expressing AmCyan and eYFP fusions of NtWLIM1 and fABD2. Although eYFP-fABD2 decorated the actin cytoskeleton similar to AmCyan-NtNtWLIM1 (supplementary material Fig. S4A–C), no significant change of AmCyan-NtWLIM1 fluorescence lifetime was noticed (Fig. 5Be–f′). Indeed, we measured an AmCyan-NtWLIM1 lifetime of 2.75±0.07 nseconds, corresponding to a FRET efficiency of 2.8% (Fig. 5C), a value below the typical 5% significance threshold (Ciubotaru et al., 2007). Together, these data confirm the BiFC analyses and provide compelling evidence that NtWLIM1 self-associates along the actin cytoskeleton.

Both BiFC and FLIM-FRET analyses revealed NtWLIM1 self-association and highlighted the preferential localization of NtWLIM1 complexes to the actin cytoskeleton. Because the two LIM domains of NtWLIM1 (supplementary material Fig. S5A) were previously shown to function as autonomous actin-bundling modules in vitro (Thomas et al., 2007), we hypothesized that they hold the ability to self-associate. Therefore, the single LIM domains of NtWLIM1, namely LIM1 and LIM2, were fused to either GFP (donor) or mRFP (acceptor) and assessed for their interaction ability in BY-2 cells using FLIM-FRET analyses. Consistent with the lack of actin-binding activity previously reported for LIM1 and LIM2 in live cells (Thomas et al., 2007), FLIM-FRET constructs exhibited a diffuse cytoplasmic distribution (supplementary material Fig. S5B). Donor constructs (GFP-LIM1 and GFP-LIM2) yielded an unquenched GFP fluorescence lifetime average of 2.45±0.04 nseconds and 2.42±0.02 nseconds, respectively (Fig. 6Aa–b′ and data not shown, respectively). When GFP-LIM1 was co-expressed with mRFP-LIM1 or mRFP-LIM2, GFP fluorescence lifetime average decreased to 2.36±0.06 nseconds and 2.33±0.09 nseconds, respectively (Fig. 6Ac–d′ and e–f′). Similarly, when GFP-LIM2 was co-expressed with mRFP-LIM2, the overall GFP fluorescence lifetime decreased to 2.34±0.06 nseconds (Fig. 6Ag–h′). In all cases, FRET efficiencies were below the typical 5% threshold, suggesting no significant interaction between LIM domains (Fig. 6B). However, we noticed on lifetime images that obvious GFP quenching occurred in small restricted cytoplasmic areas (Fig. 6Ad,f,h, arrowheads), a phenomenon that was not observed with donor constructs alone (Fig. 6Ab). We therefore suspected that single LIM domains could sporadically assemble into weak complexes that accumulate in the whole cytoplasm. Supporting such a scenario, quantitative analyses conducted on restricted positive cytoplasmic areas chosen from lifetime images yielded FRET efficiencies ranging from 7.7% to 9.3% for the different LIM domain combinations (Fig. 6B). By contrast, only background FRET efficiencies were obtained from co-expression of GFP-LIM1 with the mRFP fusion of either fABD2 or the C-terminal domain of NtWLIM1 (2.4% and 3.3%, respectively).

Although the above data supported transient interactions between LIM domains, this required further confirmation. Of interest, BiFC complexes have been reported to over-stabilize protein interactions, thereby displacing the equilibrium toward the dimeric form of the proteins fused to the complementary eYFP fragments. This property, which is sometimes considered to be a disadvantage, provides the opportunity to reveal transient and weak interactions (Ding et al., 2006; Hu et al., 2002; Kerppola, 2008; Kerppola, 2009; Kodama and Hu, 2010; Robida and Kerppola, 2009; Shyu and Hu, 2008). We therefore conducted additional BiFC experiments to assess self-association of single LIM domains. Various combinations of LIM1 and LIM2 BiFC constructs were tested. They all yielded an intense and diffuse eYFP signal in the cytoplasm of BY-2 cells revealing, on the one hand, the formation of LIM1 and LIM2 homomers in 100% and 80% of transformed cells, respectively (supplementary material Fig. S6Aa,b; Fig. S6B) and the formation of LIM1–LIM2 heteromers in about 100% of transformed cells (supplementary material Fig. S6Ac; Fig. S6B). Remarkably, no eYFP fluorescence could be detected in cells co-transformed with either LIM1 or LIM2 BiFC construct and a complementary fABD2 BiFC construct (supplementary material Fig. S6Ad; Fig. S6B). In conclusion, these data are highly consistent with our FLIM-FRET analyses and confirm that LIM domains can form homo- and heteromeric complexes, although they are much less stable than full-length NtWLIM1 complexes.

The difference in the ability to self-associate between single LIM domains and full-length NtWLIM1 is intriguing. Because, contrary to NtWLIM1, single LIM domains do not bind to the actin cytoskeleton in live cells and because NtWLIM1 complexes concentrate along actin bundles, we wondered whether the weak interaction between single LIM domains could be related to their inability to bind the actin cytoskeleton. The induction of protein self-association by actin binding has already been reported for the actin-bundling protein vinculin and, recently, the association of villin with F-actin has been shown to promote villin dimerization (George et al., 2013; Johnson and Craig, 2000). Based on these observations, we wondered whether the actin cytoskeleton could directly promote the formation and/or stabilization of NtWLIM1 dimers or oligomers.

To answer this question, we first investigated the assembly state of recombinant NtWLIM1 in the absence of F-actin using size-exclusion chromatography and dynamic light-scattering experiments. Collectively, these approaches provide strong evidence that NtWLIM1 exhibits a single, lower, oligomeric state (data not shown). However, the lack of knowledge about the spatial conformation of NtWLIM1 precludes a definite conclusion on the precise oligomeric state of NtWLIM1. We thus performed sedimentation velocity analytical ultracentrifugation experiments. The data established that NtWLIM1 has an elongated confirmation and behaves as a monomer (supplementary material Fig. S7; Table S2).

None of the above approaches is appropriate to examine the oligomeric state of NtWLIM1in the presence of F-actin. As an alternative, we determined the NtWLIM1-to-actin molar ratio at saturation in low-speed cosedimentation assays conducted at fixed actin levels and increasing NtWLIM1 concentrations (Fig. 7A,B). From four independent experiments, we calculated an average of 1.93±0.19 NtWLIM1 bound per actin subunit. This result was very consistent with live cell data and further indicates that NtWLIM1 switches from monomeric to a di- or oligomeric state when associated with F-actin. In an attempt to directly characterize NtWLIM1 complexes, the protein was incubated with F-actin and the samples were subsequently mixed with a β-mercaptoethanol-deprived buffer and analyzed by SDS-PAGE. The NtWLIM1 species formed under these non-reducing conditions were analyzed by western blotting. In control samples without F-actin, NtWLIM1 was mostly detected at the expected monomeric size (arrow at ∼25 kDa; Fig. 7C, lane 1). However, an additional very faint band of approximately twice the size (double arrow at ∼50 kDa) was scarcely detectable supporting a predisposition of NtWLIM1 for self-association. Consistent with this idea and the promoting effect of F-actin previously indicated by cosedimentation data, the higher band was markedly enhanced in the presence of F-actin (Fig. 7C, lane 2). In a control blot, actin was detected at its expected size, i.e. ∼42 kDa (supplementary material Fig. S8). The ability of NtWLIM1 to self-associate was further tested by incubating the protein with the chemical crosslinker DFDNB. As expected, DFDNB promoted the formation of 50 kDa NtWLIM1 complexes (Fig. 7C, lanes 3 and 4). Additional higher bands of low intensity suggest that NtWLIM1 can reach different oligomeric states although unspecific chemical crosslinking of NtWLIM1 cannot be excluded. Together, these data strongly suggest that NtWLIM1 self-association is promoted and/or enhanced by F-actin.

To confirm that efficient self-association of NtWLIM1 relies on an intact actin cytoskeleton in live cells, additional FLIM-FRET analyses were conducted on BY-2 cells treated with the actin-disrupting drug Lat B. Data revealed that the donor lifetime average (AmCyan-NtWLIM1) decreased in the presence of eYFP-NtWLIM1 acceptor from 2.77±0.03 nseconds to 2.64±0.06 nseconds, corresponding to a FRET efficiency below the significance threshold (4.75%; Fig. 8A,a,b′ versus c,d′ and Fig. 8B). However, similar to what we observed with single LIM domains, some restricted cytoplasmic areas exhibited a greater reduction in AmCyan fluorescence lifetime (Fig. 8A; arrowheads in d). Quantitative analysis of AmCyan fluorescence lifetime limited to those areas yielded a lifetime average of 2.49±0.06 nseconds, corresponding to a significant FRET efficiency of 9.9% (Fig. 8B). By contrast, no such FRET-positive areas were observed in control experiments conducted with eYFP-fABD2 (Fig. 8B,e,f′ and Fig. 8C). Thus, when deprived of an intact actin cytoskeleton, NtWLIM1 exhibits a reduced ability to form complexes.

Actin bundles define a ubiquitous and highly abundant type of cytoskeletal elements. In mammals, they play particularly important roles in the formation and function of various protrusive and contractile cellular structures such as filopodia, microvilli, stereocilia, nerve growth cones and stress fibers. In plants, the functional portfolio of actin bundles has steadily expanded over the past years (Higaki et al., 2010a; Higaki et al., 2010b; Thomas, 2012; Thomas et al., 2009). In addition to their general role as tracks for myosin-dependent transport, actin bundles were recently implicated in a series of more-specific processes, including stomata and chloroplast movements, polarization and recycling of auxin transporters or alteration of turgor pressure by modifications in the cell wall, vacuole and transvacuolar strand (Higaki et al., 2010a; Higaki et al., 2011; Higaki et al., 2010b; Nick, 2010; Nick et al., 2009; Staiger et al., 1994; Szymanski and Cosgrove, 2009).

Contrasting with our increasing knowledge about the functions of actin bundles, relatively little is known regarding their mode of formation. We recently characterized a novel family of small actin-bundling proteins that are widely and abundantly expressed in plants: the LIM proteins (Papuga et al., 2010; Thomas et al., 2006; Thomas et al., 2007). Noticeably, counterparts of plant LIM proteins are found in mammals, suggesting that this subset of LIM domain proteins triggers basic, evolutionarily conserved, functions (Weiskirchen and Günther, 2003). Like plant LIM proteins, the so-called cysteine-rich proteins (CRPs) promote actin bundle assembly in both reconstituted in vitro assays and live cells (Jang and Greenwood, 2009; Kihara et al., 2011; Ma et al., 2011; Tran et al., 2005). In addition, recent data support the idea that CRP-bundling activity is involved in the stabilization of stress fibers in smooth muscle cells and the formation of dendritic filopodia (Kihara et al., 2011; Ma et al., 2011). However, the molecular mechanisms underlying the actin-bundling activity of plant LIM proteins and mammalian CRPs remain unknown. Here, we provide evidence that NtWLIM1 promotes local interaction between adjacent actin filaments and their subsequent zippering into tight bundles. A similar two-step bundling process was recently described for the Arabidopsis villin proteins VLN1 and VLN3 (Khurana et al., 2010), fission yeast fimbrin Fim1 (Skau et al., 2011), mouse TRIOBP (Kitajiri et al., 2010) and human fascin (Breitsprecher et al., 2011), suggesting that it defines a basic mechanism ‘employed’ by structurally unrelated actin-bundling proteins. In addition, in vitro reconstituted assays and live cell analyses conducted with NtWLIM1 and AtWLIM1, respectively, show that both proteins can effectively bind to individual unbundled actin filaments, suggesting that WLIM1 binding chronologically precedes formation of actin bundles.

In contrast to fascin, which almost exclusively assembles unipolar bundles (90%), NtWLIM1 exhibits weak or no intrinsic selectivity regarding polarity of actin filaments and induces a majority (75–90%) of mixed polarity bundles in vitro. Data from VAEM analyses confirm the weak selectivity of WLIM1 proteins for actin filament polarity in live cells. So far, few studies have addressed the orientation of actin filaments within bundles and they have been limited to fixed tissues. Apart from pollen tubes and root hair cells, where unipolar bundles have been characterized by decoration with myosin S1 (Lenartowska and Michalska, 2008; Tominaga et al., 2000; Yokota and Shimmen, 1999; Yokota et al., 2003), the polarity of bundles remains largely unknown. With regard to these and our present data, we assume that the formation of unipolar bundles in vivo (e.g. in pollen tubes) requires factors other than LIM proteins: nucleators such as formins (Michelot et al., 2006) and/or other, more selective, actin-bundling proteins. Consistent with this hypothesis, plant LIM proteins are highly expressed in various types of cell (Arnaud et al., 2012; Eliasson et al., 2000; Papuga et al., 2010; Wang et al., 2008) and are accordingly expected to contribute to the formation of structurally different (unipolar and mixed polarity) bundles.

The activity of several actin-crosslinking proteins was previously suggested to rely on protein di- or oligomerization (Bachmann et al., 1999; Johnson and Craig, 2000; Sanders et al., 1996) and, more recently, dimerization of human villin was reported along actin-bundle-rich structures in living cells and was suggested to regulate actin crosslinking and filopodial assembly by villin (George et al., 2013; George et al., 2007). Using a series of biochemical approaches, we provide evidence that NtWLIM1 exhibits an almost exclusive monomeric conformation in the absence of actin, but reaches a dimeric, and possibly higher oligomeric, state(s) following incubation with F-actin. In support of a close relationship between NtWLIM1 state and F-actin, the sharp cytoskeletal localization of NtWLIM1 BiFC complexes indicates that NtWLIM1 self-associates along actin filaments or bundles. In addition, a strong FRET efficiency indicative of substantial NtWLIM1 self-association was exclusively detected along actin filaments, validating the accumulation of NtWLIM1 complexes at this location. The disruption of actin filaments by Lat B was sufficient to decrease the overall FRET efficiency below the typical significance threshold, pointing to the implication of actin filaments in NtWLIM1 complex formation or stabilization in vivo. Interestingly, we nevertheless observed low, but significant, FRET efficiency in restricted cytoplasmic areas, suggesting that NtWLIM1 sporadically self-associates in the cytoplasm but does not form stable complexes without actin filaments. Consistent with this hypothesis, weak NtWLIM1 complexes in F-actin-free samples could be resolved by western blot following chemical crosslinking.

Together, our results not only draw a simple parallel between WLIM1 self-association and cytoskeletal localization, but also, and above all, disclose a causal link between interaction of WLIM1 with actin filaments and formation or stabilization of WLIM1 di- or oligomers. Similar scenarios of actin-binding-induced protein oligomerization have been reported for vinculin and villin – two non-related actin-bundling proteins (George et al., 2013; Johnson and Craig, 2000). These findings, in concert with our results, hint at a mechanism common to structurally different actin-bundling proteins that allows the control of actin-bundling by actin itself.

We previously reported that, although individual LIM domains of NtWLIM1 can bind to and bundle actin filaments in vitro, they retain an affinity that is too low to interact with the actin cytoskeleton in live cells (Thomas et al., 2007). On the basis of these data and the here-established role of actin filaments in NtWLIM1 self-association, it can be predicted that if individual LIM domains hold the ability to self-associate, the resulting complexes should exhibit low stability. In line with this assumption, FRET analyses conducted with various combinations of individual LIM domains revealed discrete cytoplasmic areas exhibiting positive but weak FRET efficiencies (ranging from 7% to 9%) which were strikingly similar to the FRET efficiency calculated for full-length NtWLIM1 in Lat-B-treated cells (9.9%). Noticeably, self-interaction of LIM domains was confirmed in BiFC assays, which upon eYFP complementation generate irreversible complexes and thereby allow the characterization of weak protein interactions. In conclusion, NtWLIM1 self-association relies, at least in part, on LIM-domain-based interactions, which might, in turn, be facilitated by actin filaments. We thus end with a multi-step mechanism in which LIM domains sequentially function as actin-binding, protein dimerization and actin-bundling modules. Based on our data, we propose a model for actin crosslinking by WLIM1 dimers or oligomers (Fig. 9). In the cytoplasm, WLIM1 mainly exhibits a monomeric conformation. Although dimers and possibly higher-order oligomers sporadically assemble in the cytosol, they are not stable and the equilibrium is thus displaced toward the monomeric state. Upon binding to actin filaments through its LIM domains, WLIM1 becomes more competent for dimerization or oligomerization. Such an increase in competency of self-association might result from a conformational change of LIM domains induced by their interaction with actin filaments. Through its LIM-domain-triggered self-interaction, WLIM1 brings neighboring actin filaments or bundles into contact and promotes their zippering without marked selectivity for the polarity of actin filaments.

Considering the extremely fast cytoskeletal turnover observed in plant cells as well as the ubiquitous and high expression of plant LIM proteins, such a mechanism might correspond to a default cellular program by which actin filaments increase their own stability and lifetime. Our data might have wider implications and, for instance, could lead to a better understanding of why the human LIM protein counterpart CRP3 (also known as muscle LIM protein) dimerizes or oligomerizes in the cytoplasm but remains exclusively monomeric in the nucleus, and how shuttling of CRP3 between both subcellular compartments is regulated (Boateng et al., 2007; Boateng et al., 2009). These are particularly important questions because abnormalities in the subcellular distribution of CRP3 have been associated with the pathogenesis of heart failure. Future work should confirm and characterize the conformational changes triggered by LIM domain binding to actin filaments and leading to protein dimerization or oligomerization. More than 60 genes containing one or more LIM domains have been identified in the human genome and the list of their specific partners is continuously expanding (Kadrmas and Beckerle, 2004). Thus, it would be highly worthwhile to examine whether a conformational change of LIM domains upon partner interaction is a general process by which LIM-domain-containing proteins trigger functional switches between their partners and/or regulate their own activity.

Coding sequences of NtWLIM1, its two LIM domains (LIM1 and LIM2) and fABD2 were amplified by PCR (for primers and templates see supplementary material Table S3). For BiFC analyses, coding sequences were ligated into vectors pSPYNE(R)173 and pSPYCE(MR) (Waadt et al., 2008), which allow the expression of proteins fused to the N-terminal part (YN, amino acids 1–173) or C-terminal part (YC, amino acids 156–239) of eYFP, respectively. Constructs used for FLIM-FRET analyses were mainly derived from previously described in-house plasmids pNTL2103 and pNTL3103 (Thomas et al., 2006; Thomas et al., 2007), where existing GFP fusion constructs of NtWLIM1, LIM1, LIM2 and the C-terminal domain were changed into the reporter gene fusions of interest (AmCyan-, eYFP- or mRFP-fusions, respectively). The FLIM control plasmid expressing eYFP-fABD2 was a derivative of the pSPYNE(R)173 vector (Waadt et al., 2008), in which the initial N-terminal eYFP fragment was replaced with full-length eYFP and the fABD2 coding sequence (amino acids 325–687) was subcloned in-frame with eYFP.

A previously described pQE60 construct allowing the production of recombinant 6×His-tagged NtWLIM1 protein (Thomas et al., 2006) was expressed in E. coli and NtWLIM1-6×His was purified under native conditions according to the manufacturer's instructions (Qiagen, Hilden, Germany) and dialysed against 50 mM Tris-HCl, pH 8, 200 mM NaCl, 5 mM β-mercaptoethanol. For dye-coupling, the protein was purified and buffer-exchanged with 10 mM MOPS, pH 7.4, 150 mM NaCl using a Superdex 75 10/300 column (GE healthcare). Fractions containing NtWLIM1-6×His were combined and concentrated using centrifugal filters (Ultracell 10K; Amicon, Merck, Darmstadt, Germany). NtWLIM1-6×His was labeled using the AnaTag™ Hilyte Fluor™ 488 Protein labeling Kit (Anaspec, Fremont) according to the manufacturer's instructions using a protein-to-dye ratio of 1∶2. Micro Bio-Spin® Chromatography Columns (Bio-Gel P-6; Bio-Rad, Nazareth Eke, Belgium) were used to remove unincorporated dye and to exchange the buffer to 10 mM MOPS, 150 mM NaCl, pH 7.4. Labeling efficiency and protein concentration were determined spectrophotometrically (Implen NanoPhotometer™). Lyophilized rabbit muscle actin (Cytoskeleton, Denver, CO) was reconstituted at 116 µM. Rabbit muscle actin-Alexa-Fluor-488 (200 µM) was purchased from Molecular Probes (Life technologies, Merelbeke, Belgium). Before each experiment, a 5 µM solution of G-actin in G Buffer (2 mM Tris-HCl, pH 8.0, 0.2 mM CaCl₂, 0.4 mM DTT, 0.2 mM ATP) was prepared. After overnight incubation on ice, G-actin was centrifuged for 30 minutes at 100,000 g to remove actin oligomers.

Low-speed cosedimentation assays were performed with NtWLIM1-6×His purified under native conditions and dialysed against 10 mM Tris-HCl, pH 7, 150 mM NaCl, 1 mM DTT, 50 µM ZnCl₂. Following protein pre-clarification at 200,000 g, 4 µM of rabbit muscle actin (Cytoskeleton) were polymerized in the presence of increasing NtWLIM1 concentrations (0–24 µM) for 1 hour at room temperature in 5 mM Tris-HCl, pH 7, 0.2 mM CaCl₂, 50 mM KCl, 2 mM MgCl₂, 0.4 mM ATP and 0.4 mM DTT. Samples were then centrifuged for 30 minutes at 12,500 g. The resulting pellet and supernatant fractions were analyzed by SDS-PAGE and the respective amounts of NtWLIM1 were quantified using ImageJ (http://rsbweb.nih.gov/ij/). To calculate the molar NtWLIM1-to-actin ratio at saturation, the ratio of bound NtWLIM1 versus total actin concentration was plotted against the respective total NtWLIM1 concentration and the data points of one experiment were fitted to a hyperbolic function (SigmaPlot®10 software). The value (± s.d.) indicated in the Results corresponds to the mean of four independent experiments.

F-actin was obtained following a 30 minute polymerization of 1 µM G-actin in 1× KMEI buffer (10 mM Imidazole, pH 7, 50 mM KCl, 1 mM EGTA, 1 mM MgCl₂). To localize NtWLIM1 on actin bundles, F-actin was mixed with Hilyte-Fluor-488-labeled NtWLIM1 protein (LIM1-488) at a 1∶2-ratio. After 15 minutes, the sample was labeled with Rhodamine-Phalloidin and diluted 20-fold in fluorescence buffer (1× KMEI supplied with 100 mM DTT, 0.5% methylcellulose, 20 µg/ml catalase, 100 µg/ml glucose oxidase, 15 mM glucose) before imaging. Orientation of actin filaments in WLIM1-induced actin bundles was determined by a dual-labeling fluorescence assay (Harris et al., 2006). In brief, actin monomers (1 µM) were polymerized for 10 minutes in the presence of NtWLIM1 (2 µM) or fascin (250 nM). After addition of Alexa-Fluor-488-phalloidin (Life technologies, Belgium) during 5 minutes, actin filaments were washed in polymerization buffer and centrifuged for 5 minutes at 55,000 g. A second elongation step of 15 minutes was performed with a mix of actin monomers (1 µM), profilin (4 µM), NtWLIM1 protein (2 µM) and Rhodamine-Phalloidin. Bundles were diluted in two volumes of fluorescence buffer, placed on a poly-L-Lysine-coated slide and imaged by confocal microscopy. Bundles formed during the first co-polymerization step are imaged in green, filament elongations arising from the second step are shown in red. The experiment was performed three times and 82 filaments were observed.

After purification of NtWLIM1-6×His, elution buffer was exchanged using size exclusion chromatography to buffer A, B or C using a Superdex 75 10/300 column (GE healthcare). Sedimentation velocity AUC experiments were performed in an XL-I analytical ultracentrifuge (Beckman-Coulter) at 4°C and 50,000 rpm using double-sector charcoal-filled Epon centerpieces (1.2 cm) with sapphire windows. Measurements were carried out in PBS buffer (buffer A) or 50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 10 µM ZnCl₂ and 5 mM 2-mercaptoethanol (buffer B) or 50 mM Tris-HCl, pH 7.5, 50 mM NaCl, 10 µM ZnCl₂ and 5 mM 2-mercaptoethanol (buffer C) with 400 µl of each sample. Absorbance scans were taken at 280 nm every 5 minutes for 16 hours. The partial specific volume of NtWLIM1-6×His was calculated from the amino acid composition to be 0.7177 ml·g⁻¹ using the program SEDNTERP (SEDNTERP, Sedimentation Interpretation Program. Philo, J., Hayes, D. and Laue, T. Alliance Protein Laboratories, Thousand Cloaks, CA). The sedimentation data were analyzed with program SEDFIT using the continuous c(s) distribution analysis (Schuck, 2000). To analyze NtWLIM1 self-association under non-reducing conditions, the protein was first eluted from a Ni-NTA matrix, then buffer exchanged by dialysis to 10 mM MOPS, 150 mM NaCl, pH 7.7. Actin was exchanged for modified G-Buffer (5 mM MOPS, pH 7.5, 2 mM ATP, 0.2 mM CaCl₂). NtWLIM1 was copolymerized with 8 µM actin (final buffer composition: 7 mM MOPS, pH 7.5, 100 mM KCl, 5 mM ATP, 2 mM MgCl₂, 0.2 mM CaCl₂). For crosslinking experiments 1 mM stock solution of 1,5-difluoro-2,4-dinitrobenzene (DFDNB; Thermo Scientific) was freshly prepared in DMSO. The final concentration of DMSO in the sample was 2%. Proteins were separated by non-reducing SDS-PAGE and subsequently analyzed by western blot using a poly-histidine or actin antibody (Sigma, H1029 and abcam, ab7813, respectively).

BY-2 (Nicotiana tabacum L. cv. Bright Yellow-2) cell suspensions were maintained in the dark on a rotary shaker (27°C, 130 rpm). Every week, 3 ml of 7-day-old BY-2 cell culture were transferred into 80 ml fresh BY-2 medium.

For biolistic transformation, 3 ml of 3-day-old BY-2 cells were filtered onto a Whatman filter and placed on 0.8% BY-2 agar plates. Particle preparation and biolistic assays were performed as previously described (Thomas et al., 2007) with the following modifications: 3 mg of gold particles (<10 µm, Sigma-Aldrich, Diegem, Belgium) prepared in 50% glycerol were coated with 5 µg of each plasmid in the presence of 1 M CaCl₂ and 16 mM spermidine in a final volume of 60 µl. After ethanol precipitation, coated particles were resuspended in 25 µl of absolute ethanol. Biolistic transformation was achieved with 6 µl of DNA particle solution using the particle delivery system Biolistic PDS-1000/He from Bio-Rad. Each transformation was performed under vacuum at 1100 psi. Transformed cells were incubated in the dark at 27°C for 24 hours and then observed by confocal microscopy. To disrupt actin cytoskeleton structures, BY-2 cells were incubated 5 hours prior to bombardment on BY-2 agar plates supplied with 2.5 µM Lat B. After transformation, cells were kept on the same Lat B agar plates for 24 hours.

Cells were observed with a Zeiss LSM 510META laser-scanning confocal microscope. For BiFC experiments, cells were mounted in BY-2 medium between slide and coverslip 24 hours after bombardment. The same settings were used for all samples with a gain set at 56%. Observations were performed using a 63× oil immersion (NA 1.4) Plan Apochromat objective. Complemented eYFP and Alexa-Fluor-488-phalloidin were excited with the 488 nm line of an Argon laser and emitted light was collected with a 505–530 nm band pass filter. RFP and Rhodamine-Phalloidin were excited with the 543 nm helium neon laser line and emitted light was collected by a 580–650 nm band-pass filter. Stacks with 0.4 µm optical sections were captured and processed for deconvolution using Huygens essential software (SVI, Netherlands).

For TIRF microscopy, glass flow chambers of ∼50 µl were prepared as previously described (Breitsprecher et al., 2009). Chambers were perfused with 2.5 nM NEM-myosin in myosin buffer (10 mM imidazole, pH 7.0, 0.5 M KCl, 10 mM MgCl₂). After 1 minute, the chamber was washed with myosin buffer, then 1% BSA was flowed into the chamber and incubated for 4 minutes. The chamber was finally washed with fluorescence buffer. A mixture of Alexa-Fluor-488-labeled and unlabeled G-actin (1 µM) with or without actin binding protein (diluted in protein dialysis buffer) was prepared in 40 µl of fluorescence buffer; allowed to flow into the chamber and imaged immediately. Actin filaments were imaged by TIRF microscopy on a Zeiss inverted microscope equipped with an alpha-Plan Apochromat 100× /1.46 TIRF objective. Excitation ray 488 nm was provided by an argon laser and emission light was collected with a BP filter 525/50. Time-lapse images were acquired at 5 second intervals over 25 minutes with a Zeiss Axiocam HRm camera. Typical exposure time was 700 mseconds with a laser filter of 5%. Pictures were analysed via ImageJ and, if necessary, x-y drift during TIRF timelapses was corrected by the TurboReg plugin (http://bigwww.epfl.ch/thevenaz/turboreg/). Kymographs were built along actin filaments or bundles with the MultiKymograph plug-in (http://www.embl.de/eamnet/html/body_kymograph.html). Skewness was measured after 40 minutes of G-actin polymerization with WLIM1 protein concentrations increasing from 0 to 4 µM. Measurements were performed on three independent experiments with 40 pictures using the plugin Kbi_Filter2d (ThinLine) (Higaki et al., 2010b). For better visualization of skewness, representative pictures of one ratio are presented with a custom LUT Fire. To quantify actin bundle thickness, we supposed that the Alexa Fluor 488 fluorescence signal was proportional to the amount of actin present within analyzed bundles (Breitsprecher et al., 2011; Smertenko et al., 2010). The absolute signal of each bundle was measured and normalized to the fluorescence intensity of a single actin filament to obtain the number of filaments per bundle.

The cortical actin cytoskeleton in epidermal cells from Arabidopsis hypocotyls expressing Arabidopsis GFP-AtWLIM1 (Papuga et al., 2010) was examined using time-lapse VAEM (Konopka and Bednarek, 2008). Transgenic seeds were stratified for 2 days at 4°C, then exposed for 24 hours to white light and finally grown in the dark for 6 days (both at 21°C). Seedlings were mounted in water between slide and coverslip and analyzed using the above-described TIRF microscope platform and imaging equipment. Images were captured at 1–2 second intervals and filament bundling was examined.

Parameters of actin filament dynamics were measured in the cortex of epidermal cells from the middle third of 5-day-old, dark-grown Arabidopsis GFP-fABD2 (Sheahan et al., 2004) and GFP-NtWLIM1 hypocotyls (Papuga et al., 2010). Seedlings were imaged over 200 seconds with 1.3 second intervals and image sequences were imported into ImageJ. At least 50 filaments from 10 cells in at least five seedlings were tracked to calculate convolutedness and severing frequency according to Staiger et al., 2009. Briefly, severing frequency of filaments was evaluated by measuring the maximal length of one filament and then counting breaks along this filament over time. Only long filaments (>10 µm) were considered and tracked as long as possible to document severing events. Severing frequency was calculated as the number of breaks per length of the original filament per time (breaks/µm/second). Filament flexibility was described by convolutedness, which indicates how much a given line deviates from a straight line. Convolutedness was defined as the ratio of free-hand-traced filament length to the longest side of the bounding rectangle (series of ≥10 frames were observed). Frequently-appearing acquosomes were not taken into consideration. Elongation rates were determined by tracking filament assembly over at least four time frames. The frequency of actin filament bundling was calculated as the average of catch and zip events counted within ROI-squares of 196 µm² over a time span of 200 seconds. To quantify actin-bundling events, a total number of 36 ROI-squares from at least 12 cells and six seedlings were analyzed.

Lifetime imaging was done on a confocal microscope Leica SP2 with a time-correlated single photon counting module TCSPC-730 (Becker & Hickl). BY-2 cells were observed in BY-2 medium with a 40× water-immersion objective. Samples containing AmCyan fusion proteins were excited by a white laser at 470 nm (75% laser power), samples with eGFP constructs with the same laser at 488 nm (20% laser power). To avoid a pulse pile-up effect, laser power was adjusted to an average photon-counting rate of 10⁵–10⁶ photons⋅seconds⁻¹. Fluorescence was detected with a 480–500 nm band pass filter. Typically, the samples were continuously scanned for about 120 seconds to achieve sufficient photon statistics for the fitting of the fluorescence decays. Data were analyzed using a commercial software package (SPCImage V2.8; Becker and Hickl). Curves were fit to a mono-exponential decay with a χ² close to 1. FRET efficiency was calculated as % FRET = 1−(τ_DA)/(τ_D), where τ_DA is fluorescence lifetime of donor in the presence of acceptor; τ_D, the fluorescence lifetime of donor alone. Fluorescence lifetime was collected along bundles that showed colocalization of AmCyan and eYFP and, for each cable, three points were randomly selected; 30 actin bundles from 10 cells were analyzed. For Lat-B-treated samples and for single-LIM-domain analyses, regions of interest (ROIs) showing homogenous fluorescence intensity were selected (three ROIs/cell) from eight cells.

We thank Christopher Staiger and Jessica Henty (Purdue University, USA) for TIRF microscopy and VAEM imaging training, and for the NEM-myosin supply. We are grateful to Jörg Kudla (University of Münster, Germany) for providing us with BiFC plasmids and David Kovar (University of Chicago, USA) for the gift of human fascin. We thank Annette Kuehn and Stéphanie Kler (Laboratory of Immunogenetics and Allergology, CRP-Santé, Luxembourg) for support in size exclusion chromatography and Petr Nazarov (Microarray Center, CRP-Santé, Luxembourg) for advice on statistical analyses. We thank Catherine Birck for her expertise and assistance in AUC and DLS measurements (Structural Biology and Genomics Platform, IGBMC, Illkirch, France).