A gelsolin-like protein from Papaver rhoeas pollen (PrABP80) stimulates calcium-regulated severing and depolymerization of actin filaments.

The cytoskeleton is a key regulator of plant morphogenesis, sexual reproduction, and cellular responses to extracellular stimuli. During the self-incompatibility response of Papaver rhoeas L. (field poppy) pollen, the actin filament network is rapidly depolymerized by a flood of cytosolic free Ca2+ that results in cessation of tip growth and prevention of fertilization. Attempts to model this dramatic cytoskeletal response with known pollen actin-binding proteins (ABPs) revealed that the major G-actin-binding protein profilin can account for only a small percentage of the measured depolymerization. We have identified an 80-kDa, Ca(2+)-regulated ABP from poppy pollen (PrABP80) and characterized its biochemical properties in vitro. Sequence determination by mass spectrometry revealed that PrABP80 is related to gelsolin and villin. The molecular weight, lack of filament cross-linking activity, and a potent severing activity are all consistent with PrABP80 being a plant gelsolin. Kinetic analysis of actin assembly/disassembly reactions revealed that substoichiometric amounts of PrABP80 can nucleate actin polymerization from monomers, block the assembly of profilin-actin complex onto actin filament ends, and enhance profilin-mediated actin depolymerization. Fluorescence microscopy of individual actin filaments provided compelling, direct evidence for filament severing and confirmed the actin nucleation and barbed end capping properties. This is the first direct evidence for a plant gelsolin and the first example of efficient severing by a plant ABP. We propose that PrABP80 functions at the center of the self-incompatibility response by creating new filament pointed ends for disassembly and by blocking barbed ends from profilin-actin assembly.


INTRODUCTION
.9830, respectively). Mass spectra were acquired in an automated data-dependent mode (IDA, information-dependent acquisition) in positive mode over 25 min. De novo sequencing was performed with BioAnalyst software (ABI) and MS BLAST searches were performed at www.bork.emblheidelberg.de.
High-speed and Low-speed Co-sedimentation Assays-High-and low-speed co-sedimentation assays were used to examine the actin-binding and actin-crosslinking properties of PrABP80, respectively (36). Actin Nucleation Assay-Actin nucleation was carried out essentially as described by Schafer et al. (39).

Identification of PrABP80 as a Gelsolin-like Protein by Mass Spectrometry-To identify
PrABP80, the polypeptide was isolated from Coomassie-stained polyacrylamide gels, digested with trypsin and subjected to mass spectrometry (MS). Peptide mass fingerprinting with Q-TOF yielded profiles that matched villin/gelsolin proteins in current databases. To obtain sequence information, tryptic digests of PrABP80 were performed and seven fragments analyzed by ESI-MS/MS ( Table 1). The amino acid sequences were most similar to 135-ABP from lily pollen and Arabidopsis villins (Fig. 2). PrABP80 peptides shared 36-80% amino acid sequence identity with 135-ABP or AtVLN3, verifying that it is villin-or gelsolin-like. The peptides also shared 11-38% identity with human gelsolin. At 80-kDa, the purified protein is significantly smaller than the predicted M r for plant villins, which range from 107,000 to 135,000. Moreover, none of the sequences obtained matched the C-terminal villin-headpiece for the known plant villins. Along with the biochemical properties (see below), these data suggest that PrABP80 is pollen gelsolin and not a villin-like protein.
The purified protein was recognized by an affinity-purified antibody (Fig. 1B, lane 2), raised against the G1-G3 domains of recombinant AtVLN1, suggesting that it is related to the gelsolin/villin family. It also cross-reacted with lily villin, 135-ABP, antiserum (not shown).
To demonstrate that PrABP80 exists within cells, a pollen extract was prepared by grinding ungerminated pollen under liquid nitrogen and proteins extracted with Laemmli buffer. Western blots using either antibody revealed a prominent pair of bands at ~135 kDa, as well as a distinct polypeptide of 80 kDa that comigrates with purified, PrABP80 (Fig. 1B, lane 1). The higher Mr polypeptides are presumed to be similar to lily villins characterized by Shimmen and coworkers. bundling, low-speed pelleting assays were performed (Fig. 3B). In the absence of PrABP80, very little F-actin sedimented at 13,500 g (Fig. 3B, lane 2). A considerable amount of polymerized actin sedimented in the presence of AtVLN1 (Fig. 3B, lane 6); however, very little actin sedimented in the presence of PrABP80 (Fig. 3B, lane 4). The lack of PrABP80 filament bundling or crosslinking was confirmed by fluorescence microscopy, as described below. Unlike AtVLN1 or lily villins (28,31), the 80-kDa ABP from poppy does not have bundling activity, which is consistent with it being a gelsolin-like protein. previously (19).

PrABP80 Inhibits the Addition of Profilin-Actin Complexes to Filament Barbed
Ends-Gelsolin caps the newly-created barbed end after it severs an actin filament (49). The activity of PrABP80 was also tested with a dynamic elongation assay from F-actin nuclei. The elongation rate depends on the availability of barbed ends, under these experimental conditions: 0.4 µM pre-formed F-actin was incubated with varying concentrations of ABP for 5 min and polymerization initiated with addition of 1 µM G-actin (5% pyrene labeled) at a free [Ca 2+ ] of 160 µM. The G-actin was saturated with 4 µM human profilin to prevent polymerization from pointed ends. The results showed that the initial elongation rate was decreased with substoichiometric amounts of PrABP80 (Fig. 5A), just like AtCP (19). By fitting the data to Eq. 2 (Experimental Procedures), an apparent K d value for binding barbed ends of 5.5 nM was calculated (Fig. 5B). For comparison, a representative experiment with AtCP gave a K d of 23 nM (Fig. 5B).

Downloaded from
Similar experiments were performed in the presence of 5 mM EGTA and variable amounts of Ca 2+ , to examine the Ca 2+ -dependence of capping. Free Ca 2+ between 10 nM and 10 µM had no effect on the affinity of PrABP80 for the barbed end of filaments (  (Table 2). Therefore, PrABP80 and GS binding to the barbed-end of actin filaments is of high affinity and independent of Ca 2+ .

PrABP80 Severing Activity can be Observed with Fluorescence Microscopy-Gelsolin is able
to sever phalloidin-stabilized F-actin (50). Initially, we tested whether PrABP80 can sever populations of pre-polymerized F-actin stabilized with rhodamine-phalloidin. As shown in Fig.  9, PrABP80 significantly reduced the length of actin filaments (mean = 0.4 ± 0.4 µm) after incubation in the presence of 160 µM free Ca 2+ for 30 min (Fig. 9D) compared with controls ( Fig. 9A, mean = 5.1 ± 3.7 µm). After incubation with 500 nM AtCP (Fig. 9B), the mean length of actin filaments, 4.9 ± 3.4 µm, was not significantly different from the control. The reduction in filament length by PrABP80 was dependent on Ca 2+ ; reactions performed in the presence of 5 mM EGTA and 16 nM free Ca 2+ had a mean filament length of 4.7 ± 3.2 µm (Fig. 9C). These results indicated that PrABP80 not only caps ends, but also severs actin filaments.
The severing activity of PrABP80 was also demonstrated by monitoring the timecourse of actin filament length reduction directly. As shown in Fig. 10

DISCUSSION
Here we report the first molecular, biochemical and cytological evidence for the existence of a gelsolin-like protein in plants. An 80-kDa polypeptide from Papaver rhoeas pollen was purified to near homogeneity and its sequence determined by mass spectrometry. The actin-binding activity of gelsolin is typically held to be Ca 2+ -regulated, but the details of this regulation are complex and somewhat controversial. Part of the complexity is due to the fact that an individual gelsolin polypeptide can bind multiple Ca 2+ molecules and additional sites are coordinated with bound actin. Type-2 sites reside wholly within the gelsolin polypeptide, whereas type-1 sites require residues on both gelsolin and a bound actin. X-ray crystallographic studies indicate that gelsolin has up to six type-2 sites, one for each gelsolin-homology domain, and two type-1 sites (59)(60)(61)(62). In the absence of Ca 2+ , the six homologous domains are folded into a compact, globular structure with the actin-binding sites buried (63). An overwhelming amount of biochemical and structural data lend support to a 'tail latch' mechanism, whereby calcium binding induces a conformational change that releases the C-terminal helix of G6 from its contact with G2 and exposes the actin-binding site(s) (  To date, only limited and circumstantial evidence for severing activity by plant ABPs exists in the literature. Weak severing activity is known for non-plant ADF/cofilin proteins (71).
Severing has never been demonstrated directly for a plant ADF, although it is often inferred from depolymerization and polymerization assays (72,73). The relative contribution of severing versus enhancing the off-rate at pointed-ends to ADF/cofilin depolymerizing activity remains controversial (74)(75)(76). There is limited biochemical evidence for a Ca 2+ -dependent ABP from Mimosa pudica, the sensitive plant, that is related to gelsolin/fragmin-like proteins (77). When However, the three other peptide sequences failed to align with plant villins.
The biochemical evidence for a plant gelsolin is somewhat surprising, based solely on analysis of the Arabidopsis genome database (6,78). Although both gelsolin and villin are constructed from the same core of 6 tandem gelsolin subdomains, villin contains an additional actin-binding module (20,21). This 8.5-kDa villin 'headpiece' (VHP) confers Ca 2+ -independent actin bundling activity upon the villin family members, a biochemical property that is missing from gelsolins. All 5 gelsolin/villin-related genes in Arabidopsis (AtVLN1-5) contain a VHP, suggesting that they encode villin-like proteins (6,33). Gelsolin could be generated as an alternative splicing product from any of the AtVLN transcripts. Indeed evidence for a CapG-like splice variant for AtVLN1 (33; unpublished data) and a G1-G5 protein from AtVLN3 (AY093052) exists in current databases. to ionophores or extracellular signals (13,83,84). The identification of PrABP80 lends some credence to these hypotheses, but further experimentation is required to verify that this particular class of ABP is involved in actin reorganization during each of these responses.
Actin depolymerization is a hallmark of the SI response of field poppy and can be mimicked by drugs that raise [Ca 2+ ] i (13,14). The rapid and sustained 50-80% reduction in polymer levels is unrivaled by most eukaryotic cells. Reconstitution experiments with native pollen profilin, a G-actin binding protein that has increased sequestering activity at elevated Ca 2+ , fail to account for the level of depolymerization observed in vivo (14). We proposed previously that profilin acts in cooperation with other ABPs to mediate this depolymerization. Two proteins, capping protein (19) and PrABP80, may be such partners of profilin. The former is predicted to assist profilin in maintaining a large pool of unpolymerized actin by blocking assembly of the profilin-actin complex onto filament barbed ends, and allowing profilin to function as a simple actin sequestering protein (18). PrABP80 could also assist filament depolymerization by Ca 2+activated severing and the creation of new pointed ends for depolymerization by profilin. The feasibility of this mechanism is supported by experiments in which equimolar amounts of profilin were added to pre-existing actin filaments along with nM amounts of PrABP80. Up to 50% actin depolymerization was observed with just 10 nM PrABP80, but only in the presence of µM free Ca 2+ . This is remarkably similar to the situation that occurs during SI. To test this molecular model thoroughly, we will need to determine the cellular concentrations for CP and  Letters are single amino acid code for fragments that were sequenced by tandem mass spectrometry (MS/MS). The presence of the corresponding b or y fragment ions (or both) (see (85) for fragment ion nomenclature) for boldface amino acid residues were observed in the experimental MS/MS spectrum and are within 175 ppm mass tolerance. Amino acid residues given in plain text did not fit these criteria.
b Sequences shown were the best match primary amino acid sequence from 135-ABP (GenBank # AAD54660). The entire de novo sequence was used in BLAST searches (http://dove.embl-heidelberg.de/Blast2/msblast.html). c The gelsolin subdomain from which the best match was obtained (e.g. G1-G6) is given. These were determined by domain analysis of the 135-ABP sequence at SMART (http://smart.embl-heidelberg.de/).  (7) 2.00 ± 0.79 nM (8) a Free calcium concentration in the presence of EGTA was calculated using the 'EGTA' program. b Mean K d values (in nM) for binding to filament barbed ends from a representative experiment (±SD) of the elongation assay performed at different free calcium levels are given. Sample size (n) for these experiments is the number of PrABP80 and human plasma gelsolin (GS) concentrations used to determine a K d value.

Supplemental Data
Supplemental Movie 1. Timelapse Movie of the Time Series Shown in Figure 10 Shows Potent Severing of Actin Filaments by PrABP80.
Fluorescence images were collected at 500 msec intervals and playback is at 2 frames per second, therefore the sequence is at real-time. Data collection began approx. 2 sec. after addition of rhodamine-phalloidin labeled actin filaments to PrABP80, and the total time elapsed is 11 sec.
Numerous breaks along the backbone of several filaments are obvious within seconds, and by the Shown are the results from another independent severing experiment analyzed by fluorescence microscopy. Data collection and display are as described for Movie 1, and the elapsed time is 21 sec.