Identification of a functional switch for actin severing by cytoskeletal proteins.

Actin severing is vital for the organization of the actin cytoskeleton during cell motility. Severing of F-actin by the homologous proteins villin and gelsolin requires unphysiologically high calcium concentrations (20-200 microM). Here we demonstrate that high calcium releases an autoinhibited conformation in villin that is maintained by two low affinity calcium binding sites (aspartic acids 467 and 715) that interact with a cluster of basic residues in the S2 domain of villin. Mutation of either of these sites as well as tyrosine phosphorylation alters the conformation of villin resulting in a protein that can sever actin in nanomolar calcium. These results suggest that tyrosine phosphorylation rather than high calcium may be the mechanism by which villin and other related proteins sever actin in vivo.

Villin, gelsolin, fragmin, and adseverin are homologous, calcium (Ca 2ϩ )-regulated, actin-severing proteins that are involved in actin remodeling and regulation of cell motility. Gene knock-out experiments have implicated the actin severing properties of villin and gelsolin in their motogenic effect (5,6). However, all these proteins require very high calcium concentrations for actin severing (between 1 and 200 M) (7)(8)(9)(10)(11)(12)(13). Activation of gelsolin by Ca 2ϩ may be modulated by other factors as well (7,9,14). Low pH has been reported to override the Ca 2ϩ requirements of gelsolin (7); however, the concentration of H ϩ (pH Ͻ 6.0) may also never be reached inside the cell. Thus understanding the regulation of actin severing by villin and its related proteins is important. Villin and gelsolin contain segments that display internal homology with each other. In villin, gelsolin and adseverin/scinderin there are six such segmental repeats (S1-S6) with S1 most closely related to S4, S2-S5, and S3-S6. The two halves of the protein, the aminoterminal (S1-S3) and carboxyl-terminal (S4 -S6) are joined by a long linker region. Interaction of actin with villin is strongly Ca 2ϩ -dependent (15). Changes in the conformation of villin protein following its association with calcium have led to a hypothetical mechanism referred to as the "hinge mechanism" (16). According to this model, villin bound to Ca 2ϩ adopts a local secondary structure that allows another region of the villin protein to rotate out such that it is now exposed to the solution interface (16). In gelsolin, the high calcium concentrations requirements have been attributed to an analogous model referred to as the "tail helix unlatching" that is required to expose the actin side binding and severing domain (S1-S2 and S4) in this protein (10). However, no biochemical or functional studies are available identifying these latch/ hinge or low affinity Ca 2ϩ binding sites in either of these proteins.
The Ca 2ϩ concentrations required for actin severing by villin are in the micromolar range, which suggests that functional regulation of villin cannot be modulated by calcium, since intracellular Ca 2ϩ levels may never reach 200 M (17). We have suggested previously that tyrosine phosphorylation of villin could modulate its ability to regulate the cortical cytoskeleton in submicromolar concentrations of Ca 2ϩ (1). To resolve this paradox, we sought first to identify low affinity Ca 2ϩ binding residues that could allow villin to adopt a head to tail latched/ hinge mechanism and secondly to determine whether such a mechanism could be released by tyrosine phosphorylation as opposed to high Ca 2ϩ .
Sequence alignment of human villin and gelsolin as well as the crystal structure of gelsolin (18,19) allowed us to identify three well conserved aspartic acid (Asp) residues (Asp 86 , Asp 467 , and Asp 715 ) that could constitute the low affinity Ca 2ϩ binding sites in villin and function as latch residues. Based on the crystal structure of gelsolin, these three residues could form salt bridges with arginine and lysine residues in the S1/S2 domain of villin (as predicted for gelsolin) thus forming a latch between the S1-S3 or S2-S6 or the linker region between S3-S4 and S2 domain of villin, thus inhibiting access to the actin binding sites in the villin core. Binding of Ca 2ϩ to these low affinity binding sites could release the autoinhibited conformation thus regulating actin severing. On the basis of this, mutants of villin lacking these putative latch residues were first assessed for their ability to bind Ca 2ϩ (17). Using sitedirected mutagenesis, each of these aspartic acid residues was mutated to a leucine and expressed as glutathione S-transferase fusion protein (20). Ca 2ϩ binding was determined by recording quenching of the intrinsic tryptophan fluorescence of wild-type and mutant villin proteins. Villin has an emission maximum of 337 nm, and Ca 2ϩ induces a dose-dependent (0 to 1.0 mM) and saturable quenching of the intrinsic tryptophan fluorescence of villin without shifting its emission maxima (Fig.  1A). In contrast, the villin latch mutants VIL/D467L and VIL/ D715L show no Ca 2ϩ -dependent quenching of tryptophan fluorescence even at the highest concentration of Ca 2ϩ used ( Fig.  1, B and C). VIL/D86L behaved like VIL/WT (supplemental Fig.  S1). These data suggest that Asp 467 and Asp 715 are low affinity Ca 2ϩ binding residues that could function as latch residues. To determine the effect of tyrosine phosphorylation on Ca 2ϩ sensitivity of villin, Ca 2ϩ -dependent change in tryptophan fluorescence was monitored for full-length tyrosine-phosphorylated human villin expressed as a glutathione S-transferase fusion protein (VILT/WT) (1). Tyrosine-phosphorylated villin does not show a Ca 2ϩ -dependent change in tryptophan fluorescence (Fig. 1D). Thus tyrosine phosphorylation of villin may result in a change in the conformation of villin protein such that it is insensitive to Ca 2ϩ -induced changes in its structure.
To determine whether the latch mutants are functional and to determine the Ca 2ϩ requirements of these mutants as well as of tyrosine-phosphorylated villin to sever actin, we assessed their ability to sever actin in vitro by pyrene-actin depolymerization assays (4). VIL/WT severs actin efficiently at 200 M ( Fig. 2A) consistent with previous reports (11). Interestingly, tyrosine phosphorylation of villin allows it to sever actin very effectively at nanomolar Ca 2ϩ levels ( Fig. 2A). Tyrosine phosphorylation abolishes the high Ca 2ϩ requirements for actin severing by villin and reduces the Ca 2ϩ dependence by more than 4000-fold. In contrast, VIL/D467L and VIL/D715L mutants are constitutively active at physiological Ca 2ϩ concentrations and can sever actin just as effectively at much lower Ca 2ϩ concentrations of 50 nM (Fig. 2B). None of the villin proteins,

Identification of a Functional Switch for Actin Severing 24916
wild-type or mutant, sever actin in the absence of Ca 2ϩ , consistent with previous observations that villin requires calcium for its association with actin (21). Therefore, actin severing remains a Ca 2ϩ -dependent process, suggesting the presence of additional high affinity Ca 2ϩ binding sites in villin. The latch mutants and tyrosine phosphorylation had no effect on the actin capping activity of villin (supplemental Fig. S2). Together these data suggest that binding of Ca 2ϩ to low affinity Ca 2ϩ binding sites, namely Asp 467 and Asp 715 as well as tyrosine phosphorylation of villin regulate the Ca 2ϩ -dependent actin severing activity of villin.
One possibility is that villin adopts an autoinhibited conformation that both high calcium or tyrosine phosphorylation could release allowing villin to sever at physiological Ca 2ϩ levels. The very high Ca 2ϩ levels required to sever actin may in fact function to maintain the villin protein in an autoinhibited conformation, and tyrosine phosphorylation may be the physiologically relevant agonist for the activation of the severing

Identification of a Functional Switch for Actin Severing 24917
activity of villin. To determine whether high Ca 2ϩ or tyrosine phosphorylation could change the conformation of the villin protein, we measured changes in the secondary structure of the villin protein using circular dichroism spectroscopy in the farultraviolet region (2). The CD measurements were made with samples containing VIL/WT, VILT/WT, as well as the latch mutants in the absence or presence of different concentrations of calcium (0 -1.0 mM). The CD spectra were fitted to the K2d program to quantify the structural changes in the villin protein as described before (2). The circular dichroism spectrum of human villin shows significant dose-dependent changes in the presence of Ca 2ϩ (Fig. 3A). Binding with Ca 2ϩ results in a significant increase in the ␣-helical and a decrease in the ␤-sheet content of the protein (supplemental Table S1). These data agree with a previous observation made by Hesterberg and Weber (16), where the ␣-helical content of villin protein was increased at 50 M Ca 2ϩ . A comparison of VIL/WT structure with the structure of the latch mutants VIL/D467L and VIL/D715L shows an increase in the ␣-helical and decrease in the ␤-sheet content of the villin protein, consistent with a change in the secondary structure of the protein (Fig. 3, B and C). Mutation of the latch residues also makes these proteins insensitive to Ca 2ϩ -induced structural changes (insets in Fig. 3, B and C). Tyrosine-phosphorylated villin likewise shows an increase in the ␣-helical content and a decrease in the ␤-sheet structure ( Fig. 3D and supplemental Table S1). Unlike VIL/WT, tyrosine-phosphorylated villin shows no Ca 2ϩ -dependent change in the secondary structure ( Fig. 3D and supplemental Table S1). Thus tyrosine phosphorylation also makes villin insensitive to Ca 2ϩ -induced structural changes. These data show that both high Ca 2ϩ as well as tyrosine phosphorylation induce global changes in the secondary structure of villin, which regulate the ligand binding properties of villin (Ca 2ϩ ) as well as the actin-modifying functions of villin (actin severing). Limited proteolysis was also used to compare the conformational changes induced by Ca 2ϩ during unlatching of the hinge mechanism. S. aureus V8 cleaves villin to generate large amino-terminal fragments, the villin core (87 kDa) and the villin headpiece of 8.5 kDa. Further proteolysis produces fragments of 51, 44, and 30 kDa (supplemental Fig. S4). V8 maximally cleaves VIL/WT in the presence of calcium but not in the presence of EGTA. In contrast, VILT/WT, VIL/D467L, and VIL/D715L are cleaved at these sites even in the presence of EGTA. These data then support the idea that V8 digests villin in a Ca 2ϩ -dependent manner, and tyrosine phosphorylation or latch site mutations in villin generate conformational changes that are comparable with those generated by Ca 2ϩ in full-length villin.
Villin contains a cluster of arginine and lysine residues in S2 that have been identified as the F-actin side binding and actin severing domain in villin (22). Since we speculate that if Asp 467 and Asp 715 could form a latch with this domain in villin, then mutation in the S2 domain should similarly release the inhibitory effect of the latch. As shown in supplemental Table S1 deletion of all the positive residues in this domain (amino acids 138 -146) results in a villin mutant protein (⌬PB2) that adopts a different conformation compared with VIL/WT. ⌬PB2 has a higher ␣-helical content, lower ␤-sheet structure, and an increase in the random coil content of the protein, consistent with significant secondary structural changes in the villin mutant protein (supplemental Table S1). These data are in agreement with the idea that Asp 467 and Asp 715 form a latch with the cluster of basic residues in S2 domain of human villin. We have demonstrated previously that mutation of R138A, R145A, and R146A in villin does not alter the conformation of the protein (2). This suggests that perhaps no single residue in this domain but rather the cluster of positive charge in PB2 contributes to the formation of the latch. These data show that both tyrosine phosphorylation as well as mutation of latch residues result in a rearrangement of the villin conformation that no longer requires high Ca 2ϩ levels for its activity. We speculate that this new conformation of villin is a more open structure where the villin F-actin side binding site is exposed allowing the villin protein to bind and sever F-actin in physiologically relevant Ca 2ϩ concentrations (Fig. 3 E). Such a model would agree with the reported Ca 2ϩ induced increase in the Stoke volume of the villin protein (16).
Regulation of actin severing by tyrosine phosphorylation is likely to extend to other proteins of the villin family including gelsolin (Fig. 3F), which have been shown to be tyrosinephosphorylated in vitro (23,24) and in vivo (23,25). Since both villin and gelsolin are tyrosine-phosphorylated we hypothesize that tyrosine phosphorylation of this family of proteins may abrogate the high Ca 2ϩ requirements for actin severing and may in fact be a molecular mechanism shared by this family of proteins. The reduced calcium affinity of villin may possibly be physiologically relevant, since the intestine is the primary site of calcium absorption, allowing villin to maintain actin bundles not only under physiological calcium concentrations but also at elevated calcium concentrations (less than 100 M).