Research ArticleStructural Biology

Structure of a Pentavalent G-Actin•MRTF-A Complex Reveals How G-Actin Controls Nucleocytoplasmic Shuttling of a Transcriptional Coactivator

See allHide authors and affiliations

Science Signaling  14 Jun 2011:
Vol. 4, Issue 177, pp. ra40
DOI: 10.1126/scisignal.2001750


Subcellular localization of the actin-binding transcriptional coactivator MRTF-A is controlled by its interaction with monomeric actin (G-actin). Signal-induced decreases in G-actin concentration reduce MRTF-A nuclear export, leading to its nuclear accumulation, whereas artificial increases in G-actin concentration in resting cells block MRTF-A nuclear import, retaining it in the cytoplasm. This regulation is dependent on three actin-binding RPEL motifs in the regulatory domain of MRTF-A. We describe the structures of pentavalent and trivalent G-actin•RPEL domain complexes. In the pentavalent complex, each RPEL motif and the two intervening spacer sequences bound an actin monomer, forming a compact assembly. In contrast, the trivalent complex lacked the C-terminal spacer- and RPEL-actins, both of which bound only weakly in the pentavalent complex. Cytoplasmic localization of MRTF-A in unstimulated fibroblasts also required binding of G-actin to the spacer sequences. The bipartite MRTF-A nuclear localization sequence was buried in the pentameric assembly, explaining how increases in G-actin concentration prevent nuclear import of MRTF-A. Analyses of the pentavalent and trivalent complexes show how actin loads onto the RPEL domain and reveal a molecular mechanism by which actin can control the activity of one of its binding partners.


The transcription factor SRF (serum response factor) controls growth factor–responsive and muscle-specific genes as well as genes encoding components of the actin cytoskeleton (1, 2). In fibroblasts, stimuli that activate Rho, such as the mitogenic phospholipid LPA (lysophosphatidic acid), induce assembly of polymeric actin (F-actin) through activation of the Rho effectors mDia and Rho kinase (35). F-actin assembly lowers the concentration of monomeric actin (G-actin) and concomitantly activates SRF (6, 7).

The transcription factors MRTF-A (myocardin-related transcription factor A; also known as MAL) and MRTF-B (also known as Mkl2) (812) are actin-binding proteins that allow SRF to sense Rho-induced depletion of G-actin (13, 14). MRTF-A continuously shuttles between the cytoplasm and the nucleus in unstimulated cells, but accumulates in the nucleus at SRF target genes upon Rho activation (13, 15). The N-terminal sequences of MRTF-A up to and including its actin-binding RPEL domain are sufficient to confer regulated shuttling behavior on a chimeric protein (16). Signal-induced depletion of the G-actin pool results in reduced binding of G-actin to MRTF-A, thereby inhibiting MRTF-A nuclear export, which requires actin binding and the exportin Crm1 (15). In contrast, when the G-actin concentration in unstimulated cells is artificially increased, MRTF-A nuclear import is inhibited (15), because G-actin binding displaces importin α–importin β dimers from a bipartite nuclear localization signal (NLS) in the RPEL domain (17, 18).

The RPEL domain contains three 22–amino acid RPEL motifs (Pfam 02755) separated by 22-residue spacer sequences (Fig. 1A). Each RPEL motif binds autonomously to G-actin, and the integrity of all three RPEL motifs is required to maintain MRTF-A in the cytoplasm in unstimulated cells (16, 19). Structural studies of actin-RPEL peptide complexes reveal that the RPEL motif is L-shaped and comprises two helices, which respectively contact a hydrophobic cleft between subdomains 1 and 3 of the four-lobed actin molecule and a hydrophobic ledge on subdomain 3 (20). Binding of the RPEL motif to G-actin is thus predicted to occur competitively with binding of the major G-actin buffering proteins profilin and thymosin β4, as well as with assembly of F-actin (21). Consistent with this, the RPEL domain can inhibit actin polymerization in vitro (19).

Fig. 1

Crystal structure of a 5:1 G-actin•MRTF-A RPEL domain assembly. (A) Schematic of MRTF-A domain organization and alignment of the MRTF-A sequences used for crystallization studies with the corresponding MRTF-B and myocardin sequences from Mus musculus (MRTF-A: EDL04588; MRTF-A BSAC: NP694629; MRTF-B isoform1: P001116139; MRTF-B isoform2: NP705816; myocardin: AAQ63841); lowercase letters indicate remaining residues from the bacterial fusion protein. Observed RPEL domain secondary structure observed in the pentamer complex is indicated above the sequence. Colored bars indicate the major contacts with RPEL-actins R1, R2, and R3 (pale blue, pale green, and magenta) and spacer-actins S1 and S2 (pale yellow and orange). RPEL motifs (Pfam PF02755) are highlighted in red, spacer sequences in gray, and the two basic elements of the bipartite nuclear import signal, B3 and B2 (17, 18), in dark gray. Abbreviations for the amino acid residues are as follows: A, Ala; C, Cys; D, Asp; E, Glu; F, Phe; G, Gly; H, His; I, Ile; K, Lys; L, Leu; M, Met; N, Asn; P, Pro; Q, Gln; R, Arg; S, Ser; T, Thr; V, Val; W, Trp; and Y, Tyr. (B to D) Crystal form I. Colors are as in (A). (B) A difference mFoDFc electron density map (blue, contoured at 3σ) for the RPEL domain before its inclusion in refinement, overlaid on the refined crank-shaped RPEL domain structure. (C) Structure of a pentavalent actin•RPEL complex determined at 3.5 Å. Ribbon schematic of the actins, with a surface representation for the RPEL domain. Panel is color-coded as in (A) and oriented as in (B) (see also movies S1 and S2). (D) Space-filling representations of the pentavalent actin•MRTF-A RPEL domain complex, color-coded as in (A). (E) Space-filling representation of four actins from an F-actin filament (2224) [Protein Data Bank ((PDB) code 3MFP]. Orientation is such that the “top” (blue) filament actin is in the same orientation as actin R1 (blue) in (D), right. For relative orientations of individual actin subdomains in the F-actin and the actin•MRTF-A complexes, see fig. S2.

We previously proposed that G-actin loading onto the RPEL domain enables formation of actin•MRTF-A complexes that exhibit different abilities to interact with importin α–importin β and Crm1 and that formation of these complexes is sensitive to small changes in the availability of G-actin (16, 20). To understand the mechanism of MRTF-A regulation by G-actin, we undertook a structural analysis of G-actin•RPEL domain complexes.


A pentavalent actin•RPEL domain assembly

Crystallization of the RPEL domain of MRTF-A (residues 67 to 199) (Fig. 1A) complexed with G-actin, Mg•adenosine 5′-triphosphate (ATP), and latrunculin B (collectively referred to as actin) gave two distinct crystal forms (Table 1). Crystal form I diffracted to 3.5 Å resolution, and its structure was determined by molecular replacement and refined to an R/Rfree value of 23.7/27.3% with excellent geometry (Table 1), revealing a pentavalent actin•RPEL domain complex (15). Within the complex, the RPEL domain adopts a crank-shaped conformation with a left-handed superhelical twist of 150° along the crank “axis” (Fig. 1B), binding the five actins to generate a compact brick-like shape of dimensions 95 Å × 130 Å × 65 Å (Fig. 1, C and D, and movie S1). The RPEL domain is mostly helical in the complex, although circular dichroism (CD) measurements of the isolated domain in solution suggest that its helical content is less than 10% (fig. S1). Almost half of the RPEL domain surface (5060 Å2) is occluded when five actins are bound. The three individual RPEL motifs adopt L-shaped structures similar to those observed in actin•RPEL peptide complexes (20), except that their N-terminal helices are extended by 5, 13, and 13 residues, respectively, and the two spacers adopt similar conformations despite their divergent sequences (Fig. 1B).

Table 1

Data collection and refinement statistics. Values in parentheses are for the highest-resolution shell.

View this table:

Each RPEL motif engages an actin—the “RPEL-actins” R1, R2, and R3—and the helical N-terminal extensions of RPEL2 and RPEL3 recruit two other actins to the spacer elements—the “spacer-actins” S1 and S2 (Fig. 1B). The relative orientations of the actins are distinct from those in the classical F-actin filament (Fig. 1, C and E, and fig. S2) (2224). In addition to binding its cognate actin, RPEL1 also makes close contact with the adjacent RPEL-actin R2, whereas the opposite faces of the extended N-terminal helices of RPEL2 and RPEL3 simultaneously contact actins S1 and R2 and S2 and R3, respectively (Fig. 1B and fig. S3). Only the RPEL-actins (R1, R2, and R3), which are organized around the crank axis, make appreciable direct contacts with each other. The contacts between these “axial” actins, which bury relatively small surfaces (actin R1-R2, 165 Å2; actin R2-R3, 265 Å2), center around an RPEL-actin subdomain 3 loop (residues 285 to 290) that interacts with subdomains 2 and 4 of the subsequent RPEL-actin (Fig. 1B and movie S1). At this resolution, we cannot see sufficient ordered solvent detail to comment on indirect actin-actin contacts.

In the pentavalent complex, the RPEL-actins are related (R3 onto R2 and R2 onto R1) by a rotation of ~150° and translation of 38.7 Å along an axis roughly coincident with that of the RPEL domain crank (Fig. 2, A and B). This orients the RPEL-actins and the RPEL domain crank in a quasi–double-helical arrangement (movie S2). This rotation-translation operator also relates a trivalent subcomplex containing actins R2, S2, and R3 (and RPEL2 and RPEL3) onto a second trivalent complex comprising actins R1, S1, and R2 (and RPEL1 and RPEL2). These complexes can be directly superimposed with a root mean square deviation (RMSD) of 2.5 Å2 over 959 actin Cα atoms (Fig. 2C). The pentameric complex can thus also be considered as comprising two trivalent subcomplexes, with actin R2 as a shared subunit. The back of the RPEL1 and RPEL2 motifs at the core of each subcomplex remains solvent-exposed (Fig. 1C).

Fig. 2

Relationships between actin subunits within the pentavalent assembly and a trivalent core subcomplex. (A) The three RPEL-actins are colored as in Fig. 1 and the spacer-actins are shown in white to highlight the internal symmetry within the pentameric complex. A noncrystallographic screw axis operator (rotation, ~150°; translation, 38.7 Å) that relates RPEL-actins and the two spacer-actins is shown in black along the length of the RPEL crank. A different operator, corresponding to a rotation of ~40° and translation of 36 Å, superposes actin R1 onto S1 and actin R2 onto S2. (B) View down the crank axis with only the RPEL domain displayed. (C) The noncrystallographic screw axis operator superposes the actin R2-S2-R3 subcomplex onto the actin R1-S1-R2 subcomplex. Color coding as in Fig. 1. (D) Crystal form II. Structure of a trivalent actin•RPEL complex determined at 3.1 Å. Ribbon schematic of the actins, with a surface representation for the RPEL domain, which is ordered up to residue 120, colored as in Fig. 1. The N-terminal basic element of the NLS, B3, is highlighted in dark gray. (E) Size exclusion chromatography (SEC) analysis of actin•RPEL domain complexes. Positions of the peaks for actin•MRTF-A complex migration in buffer alone (molecular mass, ~145 kD; three actins apparently bound) and in buffer containing 4 μM actin (molecular mass, ~215 kD; 4.7 actins apparently bound) are shown. WT, wild type.

A trivalent actin•RPEL domain core complex

Insight into the nature of the 3:1 actin•RPEL domain complex previously observed in gel filtration experiments came from the structure determination of crystal form II at 3.1 Å resolution (Table 1, Crystal form II). The actin arrangement within this complex is substantially similar to the actin R1-S1-R2 subcomplex of the pentavalent structure, although the RPEL domain sequences C-terminal to position 120, within RPEL2, are apparently disordered (Fig. 2D). In the trimer complex, actins R1 and S1 make contacts similar to those seen in the pentamer [RMSD = 1.95 Å when superimposed onto actins R1-S1-R2 (862 Cα); 2.62 Å when superimposed onto actins R2-S2-R3 (947 Cα)], and RPEL1 again engages both actins R1 and R2 (Fig. 2D). The small differences in the relative orientation of actin R2 in the two complexes probably reflect loss of contacts with the incompletely ordered RPEL2 motif in the trimer complex (fig. S4).

Relationship between the pentameric and the trimeric complexes

The identification of a 5:1 actin•RPEL domain complex in crystal form I led us to reexamine the behavior of actin•RPEL domain complexes in gel filtration. Consistent with our previous studies (15), actin formed a 3:1 complex with MRTF-A(67–199) in gel filtration experiments (Fig. 2D). However, inclusion of G-actin in the gel filtration buffer enabled detection of a larger actin•RPEL domain complex of molecular mass ~215 kD, consistent with binding of about five actin molecules (Fig. 2E). The higher average temperature factor (~184 Å2) of actin S2 residues compared with that of the other actins (average of 115 Å2) within the pentavalent complex shows that in the crystal, actin S2 residues are highly mobile (fig. S5). Consistent with this, a TLS (translation, libration, and screw displacement) refinement analysis, which measures the overall mobility of actin S2, suggests that this actin has a high degree of flexibility within the crystal (fig. S6). Furthermore, as a peptide, RPEL3 has the lowest actin-binding affinity of the three RPEL motifs [KdRPEL3, 28.9 μM compared with KdRPEL1, 1.0 μM, and KdRPEL2, 1.9 μM (20)]. Together with the gel filtration data, these observations suggest that actins S2 and R3 associate relatively weakly with a more stable trimeric complex containing actins R1, S1, and R2. Consistent with this idea, an RPEL3 mutation that blocks actin binding reduced the apparent stoichiometry of the actin•MRTF-A complex to 3 in the presence of actin in the running buffer (Fig. 2E).

Actin•RPEL domain interactions

Superimposition of RPEL2-spacer2-RPEL3 of the RPEL domain onto RPEL1-spacer1-RPEL2 indicates a similar orientation (RMSD of 1.6 Å2 for 62 Cα atoms) and conservation of most of the actin contacts (Fig. 3A). A distinct feature of RPEL-actin interactions in the complex is the contacts between the extended RPEL motif N-terminal helices and a region at the front of the actin hydrophobic cleft that remains exposed in the actin filament [the “pocket” (21)]. Residues Leu67(RPEL1), Arg111(spacer1), and Arg155(spacer2) make hydrophobic contacts with the pocket of actins R1, R2, and R3, respectively, with Arg111(spacer1) and Arg155(spacer2) also hydrogen bonding to the carbonyl moieties of Gly23(actinR2/R3) (Fig. 3A). The role of these residues in actin binding means that they should be formally considered integral to the RPEL motif, although they are not part of the RPEL motif as defined by the Pfam algorithm (Fig. 1A). The different trajectory of the N-terminal helix of RPEL1 across the hydrophobic cleft compared with that of RPEL2 allows the Leu67(RPEL1) side chain to engage the pocket in a manner similar to that of the longer Arg111(spacer1) and Arg155(spacer2) side chains (Fig. 3, A and B) (20). The other contacts with RPEL-actins R1 and R2 are identical to those previously identified in the actin•RPEL peptide structures (20). The RPEL3 motif structure had not been previously determined because of its lower affinity for actin (20). It makes actin contacts virtually identical to those of RPEL2, with Arg169(RPEL2) ion-pairing with Phe375(actinR3) carboxylate, consistent with biochemical studies (20) (Fig. 3, A and B, and fig. S7A).

Fig. 3

Conservation of actin-RPEL domain interactions. (A) Conservation of crystal form I actin•RPEL domain contacts between the actin R1-S1-R2 and the actin R2-S2-R3 trimer subcomplexes after superposition of RPEL2-spacer2-RPEL3 onto RPEL1-spacer1-RPEL2. RPEL domain and actin subunits are color-coded as in Fig. 1. Selected side chains within RPEL2 and RPEL3 that make crucial actin contacts are shown together with their actin counterparts. Side-chain orientations are generally conserved between subcomplexes, although some interacting side chains are disordered in one of the trimer subcomplexes (see fig. S7). (B) RPEL3 makes identical interactions to RPEL2. Actins R1, R2, and R3 were superimposed, and their cognate RPEL domain sequences are shown with actin R2 in white surface representation. Sky blue and purple patches indicate the actin R2 subdomains 1 to 3 hydrophobic cleft and subdomain 3 ledge, respectively (20, 21). RPEL motifs and their extensions are color-coded according to their bound actin as in Fig. 1 and selected RPEL domain residues are shown. Each RPEL-actin in the complex makes an additional contact with an apolar pocket at the front of the cleft that is not seen in actin•RPEL peptide structures, which is due to the N-terminally extended first RPEL helix (Leu67(RPEL1), Arg111(spacer1), and Arg155(spacer2)). (C) Actin•RPEL domain contacts within crystal form II. RPEL domain and actin subunits are color-coded as in Fig. 1. Selected side chains within the RPEL domain that make crucial actin contacts are shown together with their actin counterparts.

The relative stability of the trimer complex likely reflects differences in the way RPELs 1 and 2 approach actins R2 and R3 in the R1-S1-R2 and R2-S2-R3 subcomplexes of the pentamer complex. In the pentavalent complex, residue Gln90(RPEL1), oriented by Arg82(RPEL1), interacts with the carbonyl group of Met305(actinR2) and approaches the amino group of the ATP adenine moiety and the carbonyl group of Glu214(actinR2) (Fig. 3A). In the trimer complex, actin R2 makes closer contacts with RPEL1, Gln90(RPEL1) interacts with the Glu214(actinR2) carbonyl group and the ATP amino group, and Gln80(RPEL1) and Ala89(RPEL1) make additional hydrogen bonds with Ser239(actinR2) and Arg335(actinR2), respectively (Fig. 3C and fig. S4). The absence of analogous interactions between RPEL2 and actin R3 (Fig. 3A) may thus underlie the weak binding of actins S2 and R3 in the pentamer complex.

The spacer-actin contacts are similar to those of the RPEL-actins, although the RPEL domain residues involved are found in differing secondary-structure elements in the two settings (Fig. 4A). The spacer sequences emerge at right angles to the RPEL1 α2 and RPEL2 α4 helices and execute another sharp turn to align the extended N-terminal RPEL2 and RPEL3 helices with the crank axis (Figs. 2B, 3A, and 4B). Phe102(spacer1) and Leu146(spacer2) contact the hydrophobic pocket of actins S1 and S2, facilitated by hydrogen bonding between Ala100(spacer1), Ala101(spacer1) main-chain amide groups, and the Asp25(actinS1) carboxylate group, and between Glu143(spacer2) and the main chain atoms of Asp25(actinS2) and Lys113(actinR2). As the extended helix approaches the RPEL motif itself, Leu109(spacer1) and Leu153(spacer2) make hydrophobic interactions with the hydrophobic cleft of actins S1 and S2 and Arg113(spacer1) ion pairs with the C-terminal Phe375 carboxylate group of actin S1 (Fig. 4B and fig. S7B). These actin contacts are similar to those made by the well-characterized actin-binding WH2 motif, although this sequence is oriented with opposite N- to C-terminal polarity across the cleft, and the ion pair with the actin C terminus is missing (Fig. 4C) (25, 26).

Fig. 4

An actin-binding element within the RPEL spacers. (A) Structural alignment of RPEL and spacer sequences highlighting crucial contacts with actin. The sequences of spacer1 and spacer2 are aligned with those of RPEL2 and RPEL3, such that conserved contacts with the hydrophobic cleft are in register. Contacts are schematized and color-coded for the different actins as in Fig. 1. Blue bars indicate conserved contacts; helix boundaries are also indicated. (B) Actin contacts made by spacer1 and spacer2. The spacer trajectories are similar despite divergent sequences. Arg113(spacer1) engages the actin S1 C-terminal Phe375(actinS1) carboxylate through an ion-pairing interaction. The orientation of Arg157(spacer2) indicates that it makes a similar contact with Phe375(actinS2), although the C-terminal helix of actin S2 is poorly ordered (see fig. S7). (C) Structural superposition of the RPEL1 and RPEL2 motifs, spacer 1, and the WAVE2 WH2 motif (red; PDB code 2A40) onto actin. The WH2 peptide traverses the actin hydrophobic cleft with an opposite polarity to RPEL motifs, but makes similar contacts through Leu449(WAVE2), Phe447(WAVE2) displayed on a strand, and Ile443(WAVE2), Leu439(WAVE2), and Arg436(WAVE2) displayed on a short helix. Ion pair and hydrogen-bonding contacts with the acidic 22 to 28 loop in actin appear to orient the pocket-contacting residue in each structure.

On the opposite face of the N-terminally extended RPEL2 and RPEL3 helices, Arg111(spacer1) and Arg155(spacer2) contact the hydrophobic pocket of actins R2 and R3, indicating that an RPEL motif can bridge two actins (Fig. 3A). Under conditions of peptide excess, a spacer1-RPEL2 peptide formed a 2:1 complex with actin, formation of which was abolished by alanine substitutions at the spacer-actin contact residues Phe102 and Leu109 (fig. S8, A and B), suggesting that these interdigitated interactions facilitate cooperative actin binding. The analogous spacer2-RPEL3 peptide did not form a dimeric complex under the same conditions, consistent with the lower actin-binding affinity of RPEL3 (fig. S8, A and B). Nevertheless, aspartate substitution of Leu153(spacer2) (L153Dspacer2) did reduce the apparent molecular weight of the actin•MRTF-A(67–199) complex in gel filtration in actin buffer, indicating that this contact is required for formation of the pentamer complex (fig. S8C).

Requirement for spacer-actin binding in MRTF-A regulation

To investigate the functional role of spacer-actin binding, we examined derivatives of MRTF-A or the MRTF-A–PK (pyruvate kinase) fusion protein MRTF-A(2–204)–PK, which has similar shuttling behavior (15, 16). The functional analysis is complicated by the presence of B2, the C-terminal basic element of the NLS, within the spacer2 sequence (Fig. 5A) (15, 18). To assess whether the mutations in the spacer sequence compromise nuclear import, we analyzed nuclear accumulation after serum stimulation or leptomycin B treatment, which respectively reduces or blocks nuclear MRTF-A nuclear export (15).

Fig. 5

Functional analysis of RPEL spacer interaction with actin. (A) Summary of mutants analyzed by immunofluorescence and reporter assays. Spacer1 and spacer2 sequences are shown with the B2 NLS element highlighted in gray and actin contacts schematized as in Fig. 4C. (B) Mutational analysis of MRTF-A. Upper panel, subcellular distribution in unstimulated cells. Colored boxes indicate the actin whose contacts are affected by the mutation. Error bars represent SEM for 3 to 16 independent experiments. At least 100 cells were counted per condition. “XXX” refers to a previously described mutant of MRTF-A, 123-1A, bearing alanine substitutions at Arg81(RPEL1), Arg125(RPEL2), and Arg169(RPEL3) (15, 16). Lower panel, mutational analysis of SRF reporter activity. Activity of a cotransfected SRF reporter gene in unstimulated cells expressing the indicated mutants was analyzed. Reporter activation was normalized to reporter activation conferred by SRF-VP16. Error bars represent SEM for 3 to 19 independent experiments. *P < 0.05; **P < 0.01; ***P < 0.001. (C) Binding of actin to spacers does not affect nuclear import. Subcellular localization of MRTF-A(2–204)–PK fusion protein derivatives in unstimulated cells (−), or cells stimulated for 30 min with 15% fetal calf serum (FCS), or treated for 30 min with leptomycin B (LMB) was analyzed as in (B). The effect of alanine substitutions at Lys152 and Lys154 (K152A-K154A) in the NLS B2 element (see Fig. 1A) was described previously (15, 18). (D) Localization analysis of MRTF-A spacer mutants in unstimulated cells. Representative cells are shown for WT MRTF-A, L109D-L146D-K149D (mutant 5 + 16), F102D-Q105D-L146D-K149D (mutant 8 + 16), and R81A-R125A-R169A (XXX).

We first examined residues contacting the actin S1 hydrophobic cleft (Fig. 5A). Alanine or aspartate substitutions at residues Phe102(spacer1), Gln105(spacer1), or Leu109(spacer1) caused substantial nuclear accumulation of MRTF-A in unstimulated cells; combined substitutions did not further enhance nuclear accumulation (Fig. 5B). Aspartate or alanine substitution at Leu109(spacer1) also caused increased nuclear accumulation of MRTF-A(2–204)–PK in unstimulated cells; this reflected defective nuclear export, because the fusion protein accumulated in the nucleus normally after serum stimulation or leptomycin B treatment (Fig. 5C and fig. S9). Similar effects were seen with mutations affecting actin S2 binding. MRTF-A nuclear accumulation was not affected by aspartate substitutions at Leu146(spacer2) or Lys149(spacer2) mutations unless they were combined, but it was significantly increased upon alanine substitution of Leu153(spacer2) (Fig. 5B). Although the L153Aspacer2 mutation is in the NLS B2 element, it increased nuclear accumulation of MRTF-A(2–204)–PK in unstimulated cells without compromising nuclear import (Fig. 5C). In contrast, the L153Dspacer2 mutation decreased MRTF-A nuclear accumulation (Fig. 5B), reflecting impaired nuclear import caused by the acidic substitution in the NLS basic element B2 (fig. S9) (18). Simultaneous disruption of both actin S1 and S2 contacts, by combining mutations in spacer1 and spacer2, resulted in effectively complete nuclear accumulation of MRTF-A in unstimulated cells, even though the RPEL motifs remained intact (Fig. 5, B and D).

The C-terminal spacer sequences interact with actins S1 and R2 and with actins S2 and R3. Alanine substitution at Arg111(spacer1) or Arg113(spacer1), which disrupts contacts with the actin R2 hydrophobic pocket and actin S1 C-terminal carboxylate, respectively, caused substantial MRTF-A nuclear accumulation (Fig. 5B). In contrast, the corresponding alanine substitution mutations at Arg155(spacer2) or Arg157(spacer2) caused MRTF-A cytoplasmic localization (Fig. 5B). These contacts also form part of the importin α–importin β binding site in the NLS (18), and, accordingly, the R155A-R157Aspacer2 double mutation, but not the R111A-R113Aspacer1 double mutation, significantly inhibited nuclear import of MRTF-A(2–204)–PK (fig. S9).

We also examined SRF reporter activation by MRTF-A mutants in which actin binding to the spacer sequences was disrupted. All the spacer1 mutations that caused significant nuclear accumulation that were tested also significantly increased activity of a cotransfected SRF reporter gene. In contrast, mutations in spacer2 that increased MRTF-A nuclear accumulation in unstimulated cells had variable effects on reporter activity (Fig. 5B). Combination of mutations in spacer1 and spacer2 did not increase reporter activity to the same extent as disruption of all three RPEL motifs (Fig. 5B). Thus, it remains possible that the residual actin-binding capability of the RPEL motifs can exert a repressive effect on transcription.


The N-terminal domain of the MRTF proteins contains three actin-binding RPEL motifs required for the regulation of MRTF activity in response to signal-induced changes in actin dynamics. We previously proposed a mass action model in which the concentration of G-actin determines the formation of different export- or import-competent actin•MRTF-A complexes (16). Here, we identified a pentavalent complex in which the crank-shaped RPEL domain recruits actin through the three RPEL motifs and the two intervening spacers (actins R1, R2, R3, S1, and S2, respectively). The pentavalent complex has internal symmetry and can be considered as two trivalent subcomplexes comprising actins R1, S1, and R2 and actins R2, S2, and R3, which are related by a rotation-translation axis. Functional analysis shows that the spacer-actin binding is required to maintain MRTF-A in the cytoplasm in unstimulated cells. The residues involved in formation of the pentamer complex are mostly conserved in MRTF-B, suggesting that it may interact with actin in a similar manner. Tandemly repeated WH2 domains can also assemble complexes containing multiple actins, and the similarity of the actin-actin interactions in these complexes to those in F-actin may underlie their filament-nucleating activity (2729). The actin orientations and interactions in the RPEL assemblies are not related to those in F-actin (2224), and it therefore would appear unlikely that they will facilitate filament nucleation.

Biochemical data suggest that binding of actins S2 and R3 of the pentamer complex is relatively unstable and that the complex dissociates into a trivalent complex on gel filtration. Indeed, we characterized a trivalent actin•MRTF-A complex containing actins R1, S1, and R2, which is structurally closely related to the N-terminal trimeric subcomplex of the pentamer. The unfolding of spacer2 and RPEL3 sequences in the trimer complex most likely results from the absence of actins S2 and R3. The actin-binding properties of the RPEL1-spacer1-RPEL2 sequences, rather than those of RPEL3, specify the signal-regulated shuttling behavior of MRTF-A as opposed to the constitutively nuclear localization of myocardin (16), suggesting that formation of the trimer of actins R1, S1, and R2 is essential for actin-based regulation of the MRTFs.

The apparent actin-binding affinity of the intact RPEL domain is considerably greater than that of the individual RPEL motifs (19, 20), indicating that actin binding must occur cooperatively. Actin binding must stabilize RPEL domain secondary structure, which is 60% helical in the pentamer complex, although substantially unstructured in solution. Within the N-terminal trimer subcomplex, such stabilizing interactions are likely to include the RPEL1-actin R2 contacts, which will stabilize the crank orientation, and the stabilization of the extended RPEL2 N-terminal helices by binding of actins S1 and R2. Indeed, we observed that two actin molecules can bind cooperatively to a spacer1-RPEL2 peptide in gel filtration. In contrast, binding of actins S2 and R3 of the pentamer complex is less stable. The RPEL3 motif has a weaker binding affinity than the others, actins S2 and R3 dissociate readily from the pentamer complex in solution, and dimeric complexes of actin on spacer2-RPEL3 peptides were not observed. We suggest that the trimer complex will readily form in cells and that recruitment of further actins to generate the pentamer complex will be sensitive to actin concentration. Further work will be necessary to test this hypothesis.

Our data show that the cytoplasmic localization of MRTF-A in unstimulated cells requires the integrity of both spacer sequences and that binding of actin to spacer sequences is required for MRTF-A export under these conditions. A simple model might be that the pentamer complex represents a structure specifically required for effective recruitment of Crm1, either to the RPEL domain itself, its associated actins, or to other sequences in the MRTF-A N-terminal region (fig. S10). We cannot exclude the possibility that the trimer complex also facilitates Crm1 recruitment, although the exposure in this complex of B2, the C-terminal basic element of the bipartite NLS, instead suggests that it is likely to recruit importin α–importin β heterodimers. We propose that the actin concentration in unstimulated cells is not sufficient to saturate the RPEL domain with actin, so that export-competent complexes, such as the pentamer, would be in equilibrium with free MRTF-A or import-competent complexes, such as the trimer. As a result, the protein continuously shuttles between cytoplasm and nucleus (fig. S10). According to this view, signal-induced depletion of the G-actin pool would result in reduced pentameric actin•MRTF-A complex formation, thereby decreasing nuclear export rates and promoting MRTF-A nuclear accumulation.

In contrast, artificial increase of the actin concentration in unstimulated cells inhibits MRTF-A nuclear import and leads to its retention in the cytoplasm upon cell stimulation (6, 15). We propose that under these conditions, the RPEL domain will be effectively saturated with actin and that the pentamer complex will predominate (fig. S10). In the pentamer complex, actin binding occludes both basic elements of the bipartite NLS (18), which fold as helices rather than the β strand conformation typical of importin α–importin β binding sites. This prevents binding to importin α–importin β, thereby rendering the pentamer incompetent for import (fig. S10). Formation of the pentamer complex therefore provides a molecular explanation of the observed inhibitory effect of actin binding on nuclear import of MRTF-A (18). Together, the actin•MRTF-A complexes described here provide the first molecular picture of G-actin acting as operator rather than as operand in its transactions with a binding partner.

Materials and Methods

Protein preparation

Actin was prepared from rabbit skeletal muscle as described previously (20). G-actin bound to Mg•ATP and latrunculin B was prepared as described previously (20). MRTF-A RPEL domain (residues 67 to 199) was expressed as a glutathione S-transferase (GST) fusion protein. MRTF-A DNA sequence encoding the RPEL domain was inserted between the Bam HI and the Eco RI sites of a pET-41a(+) plasmid (Novagen) modified to contain a 3C protease site and lacking the restriction sites 5′ of Bam HI. Protein expression was induced at 25°C in Escherichia coli Rosetta (DE3) pLysS. Bacteria were harvested by centrifugation and resuspended in lysis buffer: 50 mM tris-HCl (pH 8.0), 300 mM NaCl, 1% Triton X-100, 1 mM dithiothreitol (DTT), 1 mM EDTA, 0.5 mM AEBSF [4-(2-aminoethyl) benzenesulfonyl fluoride], and benzamidine (15 μg/ml). The fusion protein was batch-adsorbed onto a glutathione-Sepharose affinity matrix, and MRTF-A(67–199) was recovered by cleavage with 3C protease at 4°C overnight in 50 mM tris-HCl (pH 8.0), 100 mM NaCl, and 1 mM DTT. MRTF-A(67–199) was then purified by size exclusion chromatography (SEC) with a Superdex 200 column equilibrated and run in 20 mM tris-HCl (pH 8.0) and 20 mM NaCl. Peptides were synthesized by the London Research Institute (LRI) peptide synthesis core facility.

CD measurements and spectra deconvolution

CD spectra were recorded with an Aviv 202SF spectrophotometer with a 0.1-mm path length cell at 25°C. Data were recorded every 0.2 nm with a data acquisition time of 1 s in the range of 185 to 300 nm. MRTF-A(67–199) was stored in 10 mM tris (pH 8) and 10 mM NaCl and concentrated to a final concentration of 1 mg/ml (60 μM). Each spectrum was the average of five repeated scans. The composition of the secondary structure of each peptide was analyzed from CD spectra with DICHROWEB server (30) and the algorithm CONTIN (31).

Size exclusion chromatography analysis of actin•RPEL domain complexes

To analyze the stoichiometry of actin•MRTF-A complexes, we gel-filtered complex preparations in the presence or absence of 4 μM latrunculin B actin in running buffer with a Superdex 200 column calibrated with globular proteins of known molecular weight. The column was equilibrated in 20 mM tris (pH 8), 100 mM NaCl, 3 mM MgCl2, 0.2 mM EGTA, 0.2 mM ATP, 0.3 mM TCEP [(tris(2-carboxyethyl)phosphine)], and 5% (v/v) glycerol.

Crystallization, data collection, and refinement of the pentameric G-actin•RPEL domain complex

To grow crystals, we prepared actin•MRTF-A complexes by mixing G-actin and purified MRTF-A(67–199) at a 6:1 molar ratio and further purified them by SEC. This complex was concentrated to 110 mg/ml and crystallized at 20°C with the sitting-drop vapor-diffusion method. Sitting drops of 6-μl volume consisted of a 1:1 (v/v) mixture of protein and a well solution containing 0.1 M MES (pH 5.3), 5% PEG 10000 (polyethylene glycol, molecular weight 10,000), and 0.2 M magnesium acetate. Crystals appeared after 2 or 3 days and reached their maximum size after 2 weeks (1.7 mm × 0.3 mm). Crystals were cryoprotected in 0.1 M MES (pH 5.2), 10% PEG 10000, 0.2 M magnesium acetate, and 20% glycerol. Crystals were flash-frozen in liquid nitrogen, and x-ray data sets were collected at 100 K at the I03 beamline of the Diamond Light Source Synchrotron (Oxford, UK). The structure was solved at 4.25 Å resolution by molecular replacement (Crystal form I, Table 1). Although the actin•MRTF-A complex migrated with 3:1 stoichiometry in gel filtration (Fig. 2D), the structure revealed a 5:1 actin to MRTF-A stoichiometry. Similar crystals diffracting to 3.5 Å resolution were obtained with an MRTF-A(67–199) derivative containing an RPEL3 high-affinity variant (G171E/P172R; Crystal form I, Table 1). An RPEL3 peptide with this variant has a sixfold higher affinity for actin than wild-type RPEL3 (20).

Data collection and refinement statistics are summarized in Table 1 (Crystal form I). The data set was indexed with MOSFLM and scaled and merged with SCALA (32). Molecular replacement was carried out with the atomic coordinates of actin extracted from the structures of RPEL1•actin and RPEL2•actin (20) in PHASER (33, 34). Molecular replacement identified five actin molecules, and a difference density map revealed electron density for the RPEL domain residues 67 to 183 (Fig. 1B). The pentameric actin•RPEL domain structure was refined to R/Rfree values of 23.7/27.3% with excellent geometry (Crystal form I, Table 1). Refinement was carried out using tight noncrystallographic symmetry constraints in PHENIX (35). Model building was carried out in Coot (36). Average B-factors agree with the measured Wilson B. Model validation used MolProbity and Polygon (37, 38), and figures were prepared with the graphics program PyMOL (

TLS refinement, as implemented in PHENIX (35), was used to complete the refinement of the complex and to account for the inherent anisotropy within the x-ray data (39). We used one TLS group for each actin subdomain and one for the RPEL domain. The eigenvalues of the translations, librations, and screw tensor calculated during the TLS refinement showed a consistently higher anisotropic displacement for actin S2 within the crystal structure (fig. S6). Actin S2 shows a high degree of mobility within the crystal despite the formation of numerous contacts with a neighboring asymmetric unit as a result of crystal packing (60 symmetry contacts, distance <4 Å) and is thus likely to be even more unstable in solution.

Crystallization, data collection, and refinement of the trimeric G-actin•RPEL domain complex

Actin•MRTF-A complexes were prepared by mixing purified MRTF-A(67–199) and actin in a 4:1 molar ratio and further purified by SEC as previously described. Crystals were grown by concentrating the complex to 30 mg/ml and crystallizing at 20°C with the sitting-drop vapor-diffusion method. Drops (2 μl) consisted of a 1:1 (v/v) mixture of protein and a well solution containing 0.1 M BTP (bis-triazinylpyridine, pH 8.25), 20.5% PEG 3350, and 0.2 M sodium nitrate. Crystals appeared after 2 days and reached their maximum size after 10 days. Crystals were then dehydrated to improve the diffraction resolution from 6 to 3.1 Å by a serial transfer of the protein crystals to 20-μl droplets of cryoprotective agent of increasing concentration up to 0.1 M BTP (pH 8.25), 0.2 M sodium nitrate, 35% PEG 3350, and 10% ethylene glycol. The final 20-μl drop containing the protein crystal was then equilibrated against air for 3 hours. Crystals were then flash-frozen in liquid nitrogen, and x-ray data were collected at 100 K at the ID14-2 beamline of the European Synchrotron Radiation Facility (Grenoble, France). The data were then refined similarly to that of the pentameric actin•RPEL domain (see above). Data collection and refinement statistics are summarized in Table 1 (Crystal form II).

Immunofluorescence microscopy

Immunofluorescence microscopy was performed as described previously (15, 16, 20). NIH 3T3 cells (150,000 cells per well in a six-well dish) were transfected with 100 ng of expression plasmids encoding C-terminally hemagglutinin (HA)2–tagged MRTF-A or N-terminally Flag-tagged MRTF-A(2–204)–PK (16). After transfection, cells were maintained in medium containing 0.5% serum for 20 hours, followed by stimulation with 15% serum or 20 nM leptomycin B for 30 min. The primary antibody was anti-HA (3F10; Roche) or anti-Flag (F7425; Sigma). The localization of each MRTF-A derivative was scored as predominantly nuclear, pancellular, or predominantly cytoplasmic in 100 cells. Statistical significance was assessed with Fisher’s exact test on the sums of the individual cell localization counts, corrected for multiple testing with a Benjamini-Hochberg multiple testing correction assuming a false discovery rate of 1%.

Luciferase reporter assay

Luciferase reporter assays were performed as described previously (16, 20). NIH 3T3 cells (30,000 cells per well in a 24-well dish) were transfected with SRF reporter p3DA.luc (8 ng), reference reporter ptk-RL (20 ng), and SRF-VP16 (40 ng), MRTF-A (10 ng), or MRTF-A derivative (10 ng). After transfection, cells were maintained in medium containing 0.5% serum for 22 hours. Firefly luciferase activity was measured, normalized to Renilla luciferase activity (Dual-Luciferase Reporter Assay System; Promega), and calculated relative to that obtained with SRF-VP16, which was set as 1.0. Statistical significance was assessed with the nonparametric Wilcoxon test, applying a Benjamini-Hochberg multiple testing correction assuming a false discovery rate of 1%.

Supplementary Materials

Fig. S1. CD spectrum for MRTF-A RPEL domain.

Fig. S2. Arrangement of actin molecules and structural subdomains in the pentavalent actin•RPEL assembly compared with those in the actin filament.

Fig. S3. Individual RPEL motifs and their extensions engage one or two actins within the pentameric complex.

Fig. S4. Comparison of pentavalent and trivalent actin•RPEL domain complexes.

Fig. S5. Actin S2 exhibits higher temperature factors than other actins in the pentavalent actin•RPEL domain assembly.

Fig. S6. TLS refinement shows higher mobility of actin S2 in the crystal lattice.

Fig. S7. Comparison of the interactions of RPELs and spacers close to the C terminus of actin in the pentameric actin•RPEL domain assembly.

Fig. S8. Size exclusion chromatography of G-actin•RPEL peptide and actin•RPEL domain complexes.

Fig. S9. Nuclear import properties of spacer mutants.

Fig. S10. Relationships of the actin•RPEL assemblies to import and export.


Movie S1. The pentavalent actin•RPEL domain assembly.

Movie S2. RPEL-actins R1, R2, and R3 align around the crank axis, in intimate contact with the RPEL domain.

References and Notes

  1. Acknowledgments: We thank C. Moore for early electron microscopy experiments and T. Daviter for access to the CD facilities within the Biophysics Centre, Institute of Structural Molecular Biology, University of London, Birkbeck. We thank N. O’Reilly (LRI peptide synthesis core facility) for peptides and G. Kelly (LRI Bioinformatics and Biostatistics department) for statistics advice. We thank members of the Treisman and McDonald laboratories, C. Hill, and M. Way for assistance, helpful discussions, and comments on the manuscript. Funding: S.M. was funded by a Cancer Research UK (CR-UK) fellowship, C.A.L. was funded by a Marie Curie Intra-European fellowship within the Seventh European Community Framework Programme, and S.G. was funded by a Boehringer Ingelheim Fonds predoctoral fellowship. Work in the R.T. and N.Q.M. laboratories is supported by CR-UK core funding to the LRI. Author contributions: S.M. and S.G. defined protein domains, purified proteins, and established crystallization conditions. S.M. determined and refined the structures of the complexes and conducted gel filtration studies. C.A.L. constructed mutant proteins and carried out the cell-based functional assays. R.T. and N.Q.M. planned the project and designed the experiments. N.Q.M., S.M., and R.T. wrote the paper. Competing interests: The authors declare that they have no competing financial interests. Accession numbers: Coordinates have been deposited at the Protein Data Bank with accession codes 2YJF (pentavalent actin•RPEL domain complex) and 2YJE (trivalent actin•RPEL domain complex).
View Abstract

Stay Connected to Science Signaling

Navigate This Article