Nuclear Actin Network Assembly by Formins Regulates the SRF Coactivator MAL

See allHide authors and affiliations

Science  17 May 2013:
Vol. 340, Issue 6134, pp. 864-867
DOI: 10.1126/science.1235038

Nuclear Actin in Action

Actin polymerization is essential for structures in mammalian cells. Although actin filament network structures are observed in the cytoplasm and at the plasma membrane, monomeric actin is also seen in the nucleus. Baarlink et al. (p. 864, published online 4 April) directly visualized a distinct and dynamic actin network within the nucleus in living cells. The network spanned the entire nucleus and appeared to be enriched along the nuclear cortex. Transient formation of a nuclear actin network may be induced by the transcriptional serum response.


Formins are potent activators of actin filament assembly in the cytoplasm. In turn, cytoplasmic actin polymerization can promote release of actin from megakaryocytic acute leukemia (MAL) protein for serum response factor (SRF) transcriptional activity. We found that formins polymerized actin inside the mammalian nucleus to drive serum-dependent MAL-SRF activity. Serum stimulated rapid assembly of actin filaments within the nucleus in a formin-dependent manner. The endogenous formin mDia was regulated with an optogenetic tool, which allowed for photoreactive release of nuclear formin autoinhibition. Activated mDia promoted rapid and reversible nuclear actin network assembly, subsequent MAL nuclear accumulation, and SRF activity. Thus, a dynamic polymeric actin structure within the nucleus is part of the serum response.

Formins are actin nucleation and elongation factors that assemble linear actin filaments (13). The formins mDia1 and mDia2 can be detected in the nucleus, and mDia2 is exported in an exportin 1 (CRM1)–dependent manner (4, 5). However, whether formins have a function in the nucleus is unclear. Immunoblotting for endogenous proteins revealed the presence of mDia1 and mDia2, as well as β-actin, in nuclear fractions; nevertheless, the absence of α-tubulin demonstrated that the fractions were not contaminated with cytoplasmic content (Fig. 1A). When compartmentalization of FLAG-labeled mDia2 was examined by using leptomycin B (LMB) to block CRM1, we observed nuclear accumulation within 5 min (Fig. 1B), which suggests rapid mDia2 nuclear shuttling.

Fig. 1 Nuclear mDia effects on MAL-SRF transcriptional activity.

(A) Subcellular fractions were immunoblotted for endogenous proteins as indicated. Numbers indicate molecular mass (kD). (B) NIH3T3 cells stably expressing FLAG-mDia2 were treated with 20 nM LMB before fixation. FLAG-immunolabeling reveals nuclear accumulation of mDia2 over time. Scale bar, 20 μM. (C) Nuclear formin negotiates inhibition of MAL by nuclear actin. Cells were transfected with the MAL-SRF reporter 3DA.Luc to assess effects on SRF. SRF activity was measured after expression of nuclear Dia1ct or cytoplasmic NES-Dia1ct with increasing amounts of NLS-actin (0.1, 0.3, and 0.5 μg). Results are means ± SD (n = 3). (D) SRF activity from serum (fetal calf or bovine serum, FCS)– or ΔN-MAL–stimulated cells in the absence or presence of dn.Dia-NLS to suppress endogenous mDia in the nucleus. Results are means ± SD (n = 3).

Actin dynamics directly control activation of serum response factor (SRF), a transcription factor essential for early embryogenesis (6), through the cofactor megakaryocytic acute leukemia protein MAL (also known as MRTF-A or MKL1) (7). This occurs via release of actin-MAL interactions for nuclear accumulation of MAL (8). MAL-SRF–dependent transcription is critical for numerous cytoskeletal processes, such as cell contraction, adhesion, and motility (7). MAL is also regulated through actin inside the nucleus, although the underlying mechanism is unclear (8). We hypothesized that nuclear mDia participates in this process. To compare the abilities of cytoplasmic and nuclear mDia to activate MAL-SRF, we fused a CRM1-dependent nuclear export signal (NES) to a constitutively active version of mDia1 (Dia1ct) that displays predominant nuclear localization (9). Nuclear mDia-driven SRF activity was at least as potent as its cytoplasmic counterpart (fig. S1, A and B). To confirm nuclear activity of Dia1ct, we compared actin assembly rates from cytoplasmic and nuclear fractions from Dia1ct-expressing cells. Only Dia1ct-expressing nuclear extracts displayed increased actin assembly over mock transfections (fig. S1C). We then assessed whether nuclear formin–mediated SRF regulation requires actin dynamics by expressing a nonpolymerizeable R62D actin mutant (in which aspartic acid replaces arginine 62) fused to a nuclear localization signal (NLS) (10). Indeed, Dia1ct could not induce SRF activity in cells expressing R62D actin, while the SRF-VP16 fusion protein was unaffected (fig. S1D). Of note, nuclear confined Dia1ct increased the expression of the SRF target genes vinculin, SRF, and Acta2 comparable to levels of a constitutively active ΔN-MAL (11) (fig. S1E). To characterize the impact of cytoplasmic versus nuclear Dia1ct on MAL-SRF, we analyzed SRF activity under increasing amounts of NLS-actin. Only nuclear mDia could efficiently counteract increased concentrations of nuclear G-actin for SRF activity (Fig. 1C), which suggested that nuclear actin polymerization promotes G-actin release from MAL. Indeed, endogenous MAL interaction with NLS-actin was efficiently perturbed in the presence of nuclear Dia1ct but not NES-Dia1ct (fig. S1F). Furthermore, Dia1ct-expressing cells displayed a highly organized nuclear actin network (fig. S2) in contrast to expression of a nuclear localized VCA-domain, a known activator of Arp2/3 (fig. S2E) (12). We then inhibited mDia selectively in the nuclear compartment by fusing an NLS to dominant-negative mDia (13). This construct, dn.Dia-NLS, which efficiently localized and exclusively functioned in the nucleus (fig. S1G), robustly interfered with serum-induced SRF activity but did not affect the actin-independent mutant ΔN-MAL (Fig. 1D). Consistently, replacement of endogenous mDia2 with a derivative unable to enter the nucleus could not fully restore serum-induced SRF activity (fig. S3). Thus, nuclear function of mDia2 is required to efficiently activate SRF by serum.

Physiological nuclear actin network assembly has not been reported in mammalian cell culture, although the presence of monomeric actin in the nucleus is well documented (1416). Given the predominance of cytoplasmic actin over nuclear actin, we fused an NLS to the F-actin probe LifeAct, tagged with green fluorescent protein (GFP), to restrict detection to the nuclear compartment. In serum-starved cells, LifeAct-GFP-NLS displayed a diffuse fluorescence signal (Fig. 2A). When cells were stimulated with 20% serum, we observed the rapid assembly of an endogenous nuclear actin network within 15 s (Fig. 2A and movie S1). This was confirmed by glutaraldehyde fixation and phalloidin staining on cells with no ectopic protein expression (Fig. 2B and fig. S4). Serum-induced nuclear actin assembly was abrogated by combined small interfering RNA (siRNA) treatment against mDia1 and mDia2 (Fig. 2B). Serum-stimulated actin polymerization, as well as its mDia dependency, was further confirmed by actin assembly assays on nuclear fractions and by using a small-molecule inhibitor (17) to block mDia1 and mDia2 activity (Fig. 2, C and D). Consistently, RNA interference of mammalian mDia1 or mDia2 also inhibited basal actin assembly rates of nuclear extracts, whereas knockdown of the formin FHOD1 or inhibition of Arp2/3 had no effect (fig. S5 and S6).

Fig. 2 Dynamics, serum-responsiveness, and mDia-dependence of nuclear actin assembly.

(A) Live NIH3T3 cells transfected with LifeAct-GFP-NLS (25 ng/6-well) were kept in serum-free medium for 24 hours and monitored before and during serum (FCS) stimulation. Individual frames reveal LifeAct-GFP-NLS probed endogenous F-actin distribution at indicated time points. Scale bar, 10 μm. See also movie S1. (B) Depletion of endogenous mDia impairs nuclear actin filament formation. NIH3T3 cells were treated with siRNA against mDia1 and mDia2, kept in serum-free medium for 16 hours before stimulation with 20% FCS for 20 s and instant glutaraldehyde fixation. Probing endogenous F-actin by using fluorescently labeled phalloidin reveals native nuclear actin filaments in 72% of control siRNA-treated cells and 7% of cells silenced for both mDia1 and mDia2 (>30 cells per condition). Scale bar, 10 μm. DAPI, 4′,6′-diamidino-2-phenylindole. See also fig. S4. (C and D) The capacity of nuclear extracts to stimulate actin assembly was quantified by using pyrenyl-actin assembly assays. Results are shown as means ± SD (n = 3). (C) NIH3T3 cells were starved for 16 hours and instantly lysed at indicated time points after stimulation with 20% FCS. After subcellular fractionation, equal amounts of nuclear extracts were measured in pyrenyl-actin assembly assays (see also fig. S5G). (D) Serum-stimulated nuclear actin assembly is mDia-dependent. Nuclear extracts, obtained directly after serum-stimulation, were subdivided to compare the effects of addition of either dimethyl sulfoxide or the mDia-inhibitor K216-0385 (compound 2.4) on actin assembly. Results are means ± SD (n = 3). DMSO, dimethyl sulfoxide.

mDia2 activity is controlled through autoinhibition mediated by the key residue M1041 in the diaphanous autoregulatory domain (DAD) (18, 19). To spatially activate endogenous mDia2 in the nucleus, we made use of DAD expression as a tool to release mDia autoinhibition (fig. S7A) (19). Cells expressing the DAD construct had a nuclear actin network that was constantly and dynamically reorganized over time (Fig. 3A and movie S2). We also observed what appeared to be bending or buckling of filaments, suggesting that the assembled filaments might be under tension and organized into a distinct network spanning the nucleus. We subsequently assessed this process of endogenous formin activation in a spatiotemporal manner. A NES-like sequence within the mDia2-DAD region requires residue Leu1168 (fig. S7A) (4). This allowed us to rapidly accumulate mCherry-DAD in the nucleus after adding LMB to monitor the de novo generation of nuclear actin filaments. Nuclear translocation of DAD readily induced the formation of an actin filament network inside the nucleus (Fig. 3B and movie S3), which was confirmed by phalloidin staining in the absence of LifeAct (Fig. 3C). No actin network could be detected when a mutated M1041A-DAD was translocated that is unable to release endogenous mDia autoinhibition (Fig. 3B). We then asked whether endogenous nuclear formins can activate SRF. We generated a NES mutant of mDia2-DAD (DAD-L1168G) for continuous nuclear accumulation and mDia activation (fig. S7B). Expression of this mutant activated MAL-SRF, whereas the double mutant DAD-L1168G/M1041A defective in diaphanous inhibitory domain (DID) binding did not (Fig. 3D). This effect was abrogated by the presence of dn.Dia-NLS, which also blocked endogenous nuclear actin network production by mDia activation through mCherry-DAD (Fig. 3D and fig. S7C). Thus, nuclear mDia activation is necessary for actin assembly and MAL-SRF activity, and under certain circumstances, it can be sufficient.

Fig. 3 Activation of nuclear formins and effects on actin network assembly and dynamics for SRF activity.

(A) Live HeLa cells expressing LifeAct-GFP-NLS together with FLAG-tagged mDia2-DAD were monitored over time. Individual frames at higher magnification (white rectangle) reveal newly formed actin filaments (arrow heads) and dynamics at indicated time points in seconds (movie S2). Scale bar, 10 μM. Yellow lines delineate cell borders. (B) Spatiotemporal induction of a nuclear actin network by localized release of autoinhibited endogenous mDia through mCherry-DAD derivatives, as indicated, monitored using LifeAct-GFP-NLS. Nuclear accumulation of mCherry-DAD variants (shown in insets) was mediated by LMB addition, as indicated (movie S3). Scale bar, 10 μM. (C) HeLa cells expressing mCherry-DAD were treated with or without LMB (20 nM for 60 min) before fixation and staining of endogenous F-actin by using fluorescently labeled phalloidin. Scale bar, 10 μM. (D) Nuclear mDia2-DAD stimulates SRF and nuclear actin network formation. SRF activation by nuclear trapped DAD-L1168G was compared with the ability of nuclear accumulated DAD (mCherry-DAD + LMB) to induce nuclear actin network formation in the presence or absence of dn.Dia-NLS to inhibit endogenous mDia in the nucleus. Image examples are shown in fig. S7.

Next, we devised an optogenetic tool to activate endogenous formins in the nucleus of living cells, which circumvented the requirement for serum stimulation or LMB treatment. We generated, screened, and characterized fusions of the LOV (light, oxygen, or voltage) Jα-domain of Avena sativa phototropin-1 (AsLOV2) (20) to mDia2-DAD (LOV-DAD) to achieve release of endogenous mDia2 autoinhibition and subsequent SRF activity upon 400- to 500-nm illumination (fig. S8). To restrict this construct to the nucleus, we introduced the point mutation L1168G into the NES-like sequence of mDia2-DAD and fused it to an NLS to obtain nuc.LOV-DAD (fig. S8G). Illumination of cells expressing mCherry-nuc.LOV-DAD rapidly formed a nuclear actin cytoskeleton (Fig. 4A). Actin filaments disassembled when illumination was paused and were rebuilt in response to a second round of illumination (Fig. 4A). Cells expressing mCherry-nuc.LOV-DAD also showed MAL nuclear accumulation, as well as SRF activity, when exposed to blue light, which was dependent on the light-sensitive residue C450 in nuc.LOV-DAD and the presence of mDia1/2 (Fig. 4, B to D). Thus, photoactivation of endogenous mDia in the nucleus promotes nuclear actin polymerization and MAL-SRF activity.

Fig. 4 Effects of nuclear activation of endogenous mDia by a photoactivatable DAD (nuc.LOV-DAD) on nuclear actin filament formation and activation of MAL-SRF.

(A) Photoactivation of nuclear mDia induces reversible nuclear actin filament formation. An NIH3T3 cell expressing mcherry-nuc.LOV-DAD was repeatedly (every 6 s) irradiated with 488 nm to simultaneously activate nuc.LOV-DAD and to visualize redistribution of LifeAct-GFP-NLS. Scale bar, 5 μM. (B) Photoactivation of nuclear mDia triggers nuclear accumulation of MAL. HeLa cells expressing MAL-GFP together with mcherry-nuc.LOV-DAD were irradiated with blue light (every 2 s) to simultaneously photoactivate nuc.LOV-DAD and to follow redistribution of MAL-GFP over time. Note the nuclear accumulation of MAL-GFP (white asterisks). Scale bar, 10 μM. (C) Stimulation of SRF activity by photoactivation of nuc.LOV-DAD in contrast to the light-insensitive nuc.LOV-DAD C450A as assessed by reporter gene assays. Results are means ± SD (n = 3). (D) SRF activation by nuc.LOV-DAD photoactivation is sensitive to mDia1/2 siRNA treatment. Results are means ± SD (n = 4).

Here, we have observed the regulated assembly of nuclear actin filaments in mammalian cells. This nuclear actin network was serum-responsive, mDia-dependent, and formed upon activation of mDia. Indeed, endogenous formin autoinhibition was released inside the nucleus, which suggests that nuclear formin regulation is dynamically and tightly controlled. Thus, the entire process from actin polymerization to SRF-dependent gene expression can occur in the nucleus. Moreover, nuclear formin function represents an integral part of the physiological serum response.

Supplementary Materials

Materials and Methods

Figs. S1 to S8

References (2124)

Movies S1 to S4

References and Notes

  1. Acknowledgments: We thank J. V. Small for advice on rapid actin fixation, L. O. Essen and C. Taxis for advice on using the LOV-domain, M. Innocenti for mDia2 antibodies, B. Di Ventura for critical reading of the manuscript, and laboratory members for discussions. This work was funded by grants from the Deutsche Forschungsgemeinschaft (GR 2111/2-1 and SFB 593) and the Deutsche Krebshilfe e.V. (108293) to R.G. The data presented in this manuscript are tabulated in the main paper and in the suppplementary materials.
View Abstract

Stay Connected to Science

Navigate This Article