Persistent nuclear actin filaments inhibit transcription by RNA polymerase II

Actin is one of the most evolutionarily conserved and abundant cellular proteins (Pollard and Cooper, 2009). Cytoplasmic actin has been studied for decades and is known to provide mechanical support and to power a host of cellular responses. Actin is also found in the nucleus and is integral to many nuclear processes (reviewed in de Lanerolle, 2012; de Lanerolle and Serebryannyy, 2011; Grosse and Vartiainen, 2013). Actively imported and exported, nuclear actin is a cofactor for several chromatin remodelers, transcription complexes, RNA-binding proteins and all three RNA polymerases. In the cytoplasm, the ability of actin to polymerize and depolymerize, which is crucial for its physiological functions (Pollard and Cooper, 2009), is sensitively regulated. Many of the same regulators of actin polymerization shuttle into the nucleus, suggesting that actin dynamics are also key to its nuclear functions (de Lanerolle and Serebryannyy, 2011; Gettemans et al., 2005; Grosse and Vartiainen, 2013).

New actin probes (Baarlink et al., 2013; Belin et al., 2013) and fluorescence recovery after photobleaching (FRAP) studies (McDonald et al., 2006) have indicated that actin exists in monomers and short polymers throughout the nucleus. Nuclear actin polymerization has been implicated in cellular differentiation (Sen et al., 2015), DNA damage clearance (Belin et al., 2015) and adenoviral replication (Fuchsova et al., 2015). Transient actin filaments, detected in nuclei upon serum stimulation or cell spreading, are reported to regulate the MAL (also known as MRTF) transcription factor pathway (Baarlink et al., 2013; Plessner et al., 2015). However, their effects on general transcription and chromatin remodeling are poorly understood. Notably, nuclear actin filaments are generally not observed and the use of exogenous probes to visualize nuclear actin polymerization might artificially promote actin assembly (Belin et al., 2013; Courtemanche et al., 2016; Du et al., 2015; Spracklen et al., 2014).

The polymerization of nuclear actin into dense, stable and phalloidin-stainable nuclear actin filaments occurs in some pathological conditions, such as skeletal myopathies (Domazetovska et al., 2007b), baculovirus infection (Goley et al., 2006) and idiopathic inflammatory myopathy (Stenzel et al., 2015). Nuclear actin filaments have also been noted following treatment with actin-depolymerizing drugs (Belin et al., 2013; Sen et al., 2015; Yahara et al., 1982) and upon dysregulation of actin-binding proteins such as exportin-6 (Dopie et al., 2015, 2012) and MICAL-2 (Lundquist et al., 2014). Furthermore, acute cellular stress, such as heat shock, as well as huntingtin mutations, lead to the formation of micron long actin and cofilin nuclear rods (Munsie et al., 2011; Nishida et al., 1987). Oocytes present a special situation where nuclear actin filaments confer structural stability to the large oocyte nucleus (Feric and Brangwynne, 2013). Oocytes also require nuclear actin polymerization for gene reprogramming (Miyamoto et al., 2011). However, stable, phalloidin-staining actin filaments are absent in normal mammalian somatic nuclei and their impact upon the nucleus in pathological conditions is unclear.

The role of nuclear actin in regulating general transcription has been studied extensively. Actin associates and co-purifies with RNA polymerases I, II and III and stimulates the activity of all three RNA polymerases in vitro (Hofmann et al., 2004; Hu et al., 2004; Philimonenko et al., 2004). Furthermore, nuclear actin is recruited to promoters with all three RNA polymerases in an activity-dependent manner (Hofmann et al., 2004; Hu et al., 2004; Philimonenko et al., 2004). In addition, in vitro translation studies have shown that β-actin directly interacts with at least three RNA polymerase III subunits: RPC3, RPABC2 and RPABC3 (Hu et al., 2004). Two of these subunits are components of all three RNA polymerases and might constitute conserved sites of nuclear actin binding. RNA polymerase I has been shown to require both actin and nuclear myosin I motor activity (Ye et al., 2008). RNA polymerase II (RNAPII) also requires actin and nuclear myosin I, and actin is part of the pre-initiation complex (Hofmann et al., 2004, 2006). However, the form of actin necessary for RNAPII activity remains unclear (de Lanerolle and Serebryannyy, 2011; Grosse and Vartiainen, 2013).

We report here that formation of stable nuclear actin filaments correlated with altered RNAPII dynamics and localization. Sequestering endogenous monomeric nuclear actin by forming nuclear actin filaments reduced the association of RNAPII with nuclear actin and inhibited the recruitment of RNAPII to activated promoters, ultimately inhibiting transcription and proliferation. Similarly, polymerizing or crosslinking nuclear actin was able to impair the actin–RNAPII interaction and inhibit transcription in vitro. Taken together, our data suggest that maintaining nuclear actin in a non-filamentous state is critical for the functional integrity of the nucleus and shed mechanistic insight into the absence of classical actin filaments in the normal somatic nucleus.

The absence of classical actin filaments, despite the identification of many actin regulatory proteins in the nucleus, suggests that nuclear actin polymerization is tightly regulated. Therefore, we asked how stabilizing actin filaments in the nucleus would affect nuclear functions. We stabilized nuclear actin filaments by mutating actin (V163M-α-actin) or by bundling nuclear actin (with a fragment of supervillin, amino acids 1–1010) (Fig. 1A–C). V163M, one of many naturally occurring point mutations in α-skeletal muscle actin that cause the human disease intranuclear rod myopathy, is hallmarked by the formation of nuclear actin filaments (Domazetovska et al., 2007b) (Fig. 1B). Supervillin, a membrane-associated protein with multiple endogenous nuclear localization sequences (NLS), is an actin-bundling protein that regulates p53 levels and nuclear androgen receptor activity in prostate cancer (Fang and Luna, 2013; Ting et al., 2002; Wulfkuhle et al., 1999). Expressing a fragment of supervillin (Fig. 1C) resulted in the polymerization of endogenous actin into stable nuclear actin filaments that can be stained with phalloidin. Because of their association with several human diseases, these proteins represent pathologically relevant models for studying actin polymerization in the nucleus.

To quantify the effects of nuclear actin filaments on the dynamics of the nucleoplasmic pool of actin, we performed FRAP studies on COS7 cells using the EYFP–NLS-β-actin construct as a reporter as previously described (McDonald et al., 2006). Because targeting EYFP–β-actin to the nucleus can result in actin filaments in some cells (Fig. 1D), we photobleached ∼1 µm² areas devoid of any filaments in nuclei with and without nuclear actin filaments (Fig. 1E). Quantification of the recovery dynamics showed that nuclei with actin filaments had significantly increased half-lives of recovery and lower mobile fractions (Fig. 1F; Table S1A). Similar results were obtained from FRAP analyses on nuclei of cells transfected with wild-type EYFP–β-actin alone or in combination with the mCherry–supervillin fragment (Fig. S1A,B). Wild-type EYFP–β-actin did not form nuclear actin filaments alone but was readily incorporated into nuclear actin filaments when co-transfected with the supervillin fragment. This allowed us to analyze actin dynamics in nuclei both without nuclear actin filaments and with nuclear actin filaments induced by nuclear actin bundling. Analysis of the data in Fig. 1F and Fig. S1A showed that the curves can be superimposed (R²=0.88 and 0.89, when the curves calculated in Fig. 1F for EYFP–NLS-β-actin are fitted to the data in Fig. S1A). To ensure that transfection with supervillin fragment predominately affected the nuclear rather than the cytoplasmic actin pool, we performed FRAP assays on diffuse wild-type EYFP–β-actin in the cytoplasm (Fig. S1C). As previously reported, diffuse nuclear and cytoplasmic actin populations showed similar recovery kinetics (Ho et al., 2013; Johnson et al., 2013; McDonald et al., 2006). Importantly, unlike the nucleus (Fig. S1A), co-transfection with the mCherry–supervillin fragment and the formation of nuclear actin filaments had no noticeable effect on wild-type EYFP–β-actin in the cytoplasm (Fig. S1C). These results agree with previous FRAP studies on nuclear actin following treatment with actin-polymerizing drugs (McDonald et al., 2006). The FRAP data support a model wherein the formation of stable nuclear actin filaments decreases the mobility of the remaining, non-filamentous actin in the nucleus, without significantly affecting the cytoplasmic pool of actin.

This model presumes that nuclear actin filaments incorporate endogenous actin. To investigate this possibility, immunoprecipitation assays were performed with an antibody specific to endogenous β-actin (antibodies validated in Fig. S1D) on extracts from COS7 cells expressing EGFP (control), V163M-α-actin–GFP or EYFP–NLS-β-actin. Blots probed for β-actin or GFP (Fig. 2A) demonstrated that endogenous β-actin associates with both exogenous V163M-α-actin–GFP and EYFP–NLS-β-actin. To establish that endogenous actin is directly incorporated into nuclear actin filaments, cells were transfected, fixed with ethanol to expose the actin epitope buried in filaments, and stained with an antibody that preferentially recognizes endogenous actin but not exogenous fluorescent actin (Fig. S1D,E). Fig. S1E shows the presence of endogenous actin in nuclear actin filaments formed by V163M-α-actin–GFP or EYFP–NLS-β-actin, establishing that endogenous actin directly incorporates into nuclear actin filaments. Furthermore, because phalloidin staining is highly specific to actin, phalloidin staining in supervillin-transfected cells (Fig. 1B) can only result from the inclusion of endogenous β-actin into nuclear actin filaments. Indeed, co-transfecting cells with wild-type EYFP–β-actin, which does not form nuclear actin filaments when expressed alone, and the mCherry–supervillin fragment showed that exogenous β-actin is also readily incorporated into nuclear actin filaments (Fig. S1B).

One explanation for the results above is that nuclear actin filaments increase nuclear actin polymerization and thus, sequester and decrease the concentration of non-filamentous actin in the nucleus. We confirmed this possibility by performing molecular brightness (aggregation) analysis. This fluorescence correlation spectroscopy method distinguishes the brightness of a particle from the number of particles in a given volume and, therefore, measures the degree of protein aggregation (Berland et al., 1995; Chen et al., 2003; Digman et al., 2008). Diffuse areas of fluorescence in nuclei expressing EGFP–NLS alone (as an ideal monomer control), EYFP–NLS-β-actin (both nuclear actin filament positive and negative), and EGFP–β-actin in mCherry–supervillin fragment co-expressing nuclei were analyzed (Fig. 2B). We found that EGFP–NLS and EYFP–NLS-β-actin (nuclear actin filament negative) behaved predominantly as monomers; however, EYFP–NLS-β-actin (nuclear actin filament positive) and EGFP–β-actin in supervillin-fragment-containing nuclei exhibited significantly increased molecular brightness. Thus, EYFP–β-actin in the nucleus behaves primarily as a monomer (as compared to EGFP) and exogenous β-actin in nuclei with nuclear actin filaments exhibits an overall increased polymerization state. We also performed 100,000 g spins on extracts from purified nuclei expressing EYFP–NLS-β-actin, V163M-α-actin–GFP or EGFP, based on the prior demonstration that polymerized actin is found in the pellet (Brotschi et al., 1978). Sedimentation assays show that the majority of nuclear actin remains in the soluble fraction. Furthermore, actin is enriched in the pellet fraction in nuclei containing actin filaments as compared to in EGFP-transfected cells (Fig. 2C). These data support the notion that pathogenic formation of nuclear actin filaments increases the polymerization and sequestration of nuclear actin.

We then investigated whether formation of nuclear actin filaments decreased the monomeric actin pool in the nucleus. Transfected cells were fixed with formaldehyde, which preserves the filamentous actin structure and limits access to the actin epitopes buried within actin filaments (Gonsior et al., 1999), and stained with an antibody that recognizes endogenous non-filamentous nuclear actin (Cisterna et al., 2006; Sacco-Bubulya and Spector, 2002). Because this antibody recognizes actin filaments poorly after formaldehyde fixation, the nuclear fluorescence results from binding to non-filamentous actin. Fluorescence intensity quantification showed that cells with nuclear actin filaments had less non-filamentous actin staining in the nucleus (Fig. 2D). These data establish that formation of persistent nuclear actin filaments increases nuclear actin polymerization (Figs 1E,F, 2B–D; Fig. S1A–C) and sequesters endogenous actin (Figs 1A–D, 2A; Fig. S1E), thereby decreasing the pool of monomeric actin in the nucleus.

To quantify the overarching effects of forming persistent nuclear actin filaments and sequestering nuclear actin, COS7 cells were treated with BrU or BrdU, nucleotide analogs that preferentially incorporate into mRNA or DNA, respectively (Lin et al., 2008). Nuclei with actin filaments showed a decrease in BrU incorporation, indicating that the formation of nuclear actin filaments coincides with reduced levels of global transcription (Fig. 3A). Importantly, both the polymerization-resistant actin mutant NLS-R62D-β-actin and wild-type actin, which increase cytoplasmic actin levels, had no effect. Co-transfection with the mCherry–supervillin fragment to induce nuclear actin filaments and either wild-type or EYFP-NLS-R62D-β-actin showed that increasing nuclear but not cytoplasmic levels of monomeric actin was able to rescue the effects of nuclear actin filaments on transcription (Fig. 3B). This suggests that the reduction in transcription is specific to the increase in nuclear actin polymerization. This finding is consistent with work by Dopie et al. (Dopie et al., 2012), which showed that knockdown of the actin import factor (Imp9, also known as IPO9) decreased nuclear actin levels and, correspondingly, general transcription. Our results are also consistent with other reports that have found that depleting or polymerizing nuclear actin leads to decreased general transcription (Daugherty et al., 2014; Dopie et al., 2012; Spencer et al., 2011).

We next used a SRF-dependent luciferase construct to determine whether the residual transcription activity in these cells could be from genes that are upregulated upon nuclear actin filament formation. Previous reports have shown that nuclear actin filaments sequester monomeric nuclear actin and activate the MAL–SRF pathway (Baarlink et al., 2013; Ho et al., 2013; Kokai et al., 2014; Vartiainen et al., 2007). Similarly, we find EYFP–NLS-S14C-β-actin and EGFP–supervillin fragment expression both increase induction of the SRF-dependent reporter (Fig. S2A). These data suggest that nuclear actin filament formation has an inhibitory effect on general transcription, but this does not preclude specific gene stimulation by transcription factors that are activated by the depletion of nuclear actin.

Analysis of BrdU incorporation and cell number counts consistently demonstrated that proliferation of nuclear actin filament positive cells is reduced >50% (Fig. 3C; Fig. S2B). However, there was no increase in apoptosis in cells with nuclear actin filaments as determined by a TUNEL assay (Fig. S2C). Therefore, the global decreases in transcription and proliferation in cells with nuclear actin filaments likely underlie the importance of maintaining a monomeric pool of nuclear actin.

To determine whether the decrease in general transcription was related to changes in RNAPII, we assessed how formation of nuclear actin filaments affects RNAPII localization and dynamics. Structured illumination microscopy (SIM) on cells stained for active RNAPII revealed relatively evenly distributed RNAPII in small transcription factories throughout the nucleus in non-transfected cells (Fig. 4A). However, the presence of stable nuclear actin filaments coincided with the appearance of unusually large factories or clusters (Fig. 4B–D). We noted a significant difference in mean transcription factory size and an increase in the maximum range compared to nuclei without actin filaments (Fig. 4E). Furthermore, co-transfecting cells with the mCherry–supervillin fragment to induce nuclear filaments and EYFP–NLS-R62D-β-actin to increase nuclear actin levels restored transcription factory size to that of non-transfected cells (Fig. 4E; Fig. S3A). We generated a COS7 cell line stably expressing the large subunit of RNAPII conjugated to GFP to determine whether altering nuclear actin polymerization affects RNAPII dynamics. These cells exhibit many small (<0.5 µm), and a limited number of medium sized (0.5–1 µm), transcription factories (Fig. 4F). GFP–RNAPII cells transfected with the mCherry–supervillin fragment, in contrast, have small and large (>1 µm) factories. FRAP studies on small factories in control cells and cells containing nuclear actin filaments showed similar recoveries (Fig. 4F). However, FRAP studies on large factories in actin-filament-containing nuclei showed that they are consistently less dynamic than small factories (Fig. 4F; Table S1B). These data suggest a role for nuclear actin in regulating transcription factory dynamics.

The temporal relationship between transcription factory size and nuclear actin polymerization was investigated by transfecting COS7 cells with the EGFP–supervillin fragment (Fig. S3B) or EYFP–NLS-β-actin (Fig. S3C). Cells were fixed over 32 h, stained for RNAPII, and RNAPII clustering was correlated with construct expression (Fig. S3B–D). As construct expression increased with time so did the percentage of transfected cells with visible nuclear actin filaments. This experiment demonstrated the presence of RNAPII clustering in transfected cells before the formation of nuclear actin filaments (Fig. S3B–D). Furthermore, we found RNAPII clusters in no more than 4% and 1% of cells with cytoplasmic expression of the EGFP–supervillin fragment or EYFP–NLS-β-actin, respectively. This strongly suggests that RNAPII clustering is not due to changes in cytoplasmic actin. Additionally, treatment with the actin-depolymerizing drug latrunculin B or the myosin II inhibitor blebbistatin did not have an effect on RNAPII localization (Fig. S3E). The major implication of these data is that polymerizing nuclear actin increases RNAPII clustering, but filaments, per se, are not needed to cluster RNAPII.

The effect of nuclear actin filaments on gene transcription, in vivo, was investigated next using the U2OS-263 system (Fig. 4G). U2OS-263 cells have 200 repeats of an inducible gene cassette stably integrated at a single locus (Darzacq et al., 2007). Each repeat encodes an mRNA with a 24-nucleotide stem-loop repeat in its 3′ untranslated region. U2OS-263 cells were co-transfected with MS2–GFP, an RNA-stem-loop-binding protein, to track transcription, and the mCherry–supervillin fragment to form nuclear actin filaments. Upon induction, the gene cassette is transcribed, and MS2–GFP accumulates at the transcribed cassette and serves as a direct measure of transcript synthesis. FRAP assays were performed to assess whether the rates of transcription of the cassette differ in control cells and cells containing nuclear actin filaments. Quantification of the accumulation of MS2–GFP fluorescence after bleaching showed reduced MS2 fluorescence recovery in nuclei expressing the mCherry–supervillin fragment as compared to nuclei without actin filaments (Fig. 4G; Table S1C). These results are consistent with a slower rate of transcription (Darzacq et al., 2007).

The results above suggested that nuclear actin filaments impact RNAPII clustering, perhaps by altering the actin–RNAPII interaction. We investigated this possibility by performing GFP pulldowns on cells expressing polymerization-promoting (S14C) or polymerization-resistant (R62D) EYFP–NLS-β-actin mutants (Posern et al., 2002). Western blot analyses showed a higher ratio of RNAPII to EYFP in pulldowns of cells expressing EYFP–NLS-R62D-β-actin compared to cells expressing EGFP or EYFP–NLS-S14C-β-actin (Fig. 5A). This preferred association of RNAPII with a polymerization-resistant β-actin is in agreement with the poor colocalization between nuclear actin filaments and RNAPII (Fig. 4A), as well as between actin polymers and RNAPII (Belin et al., 2013). Furthermore, immunoprecipitation experiments showed a reduced association between RNAPII and β-actin in cells with nuclear actin filaments versus EGFP-transfected control cells (Fig. 5B). These data further support the notion that the incorporation of endogenous actin into nuclear actin filaments depletes the available non-polymerized nuclear actin pool, thereby reducing the actin–RNAPII interaction.

We have previously shown that the actin–RNAPII complex is specifically recruited to the MHCIITA (also known as CIITA) promoter upon induction with interferon-γ (Hofmann et al., 2004). Therefore, we asked whether the recruitment of RNAPII to the MHCIITA promoter is impaired by sequestering nuclear actin. ChIP-qPCR assays showed significant decreases in RNAPII recruitment in cells with nuclear actin filaments compared to cells expressing EGFP alone (Fig. 5C). Reporter gene experiments showed that induction of the MHCIITA promoter with interferon-γ was significantly reduced in cells with nuclear actin filaments (Fig. 5D), in agreement with the effects on general transcription (Fig. 3A). We complemented these studies by directly inhibiting RNAPII with α-amanitin or 5,6-dichloro-1-b-dribofuranosylbenzimidazole (DRB), chemical inhibitors of RNAPII that result in the formation of enlarged transcription factories (Bregman et al., 1995) (Fig. S4A). Although some actin still associates with these enlarged factories (Fig. S4B), co-immunoprecipitation experiments revealed an overall decrease in the association between RNAPII and actin (Fig. S4C). Whereas the large transcription factories formed by direct RNAPII inhibition are morphologically distinct from RNAPII clusters induced by nuclear actin filaments (note SC-35 colocalization with RNAPII in Fig. S4A but not Fig. S4D), these results closely parallel the effects of nuclear actin filaments on RNAPII. They also support the importance of the actin–RNAPII association and further affirm the idea that nuclear actin might be a dynamic regulator of RNAPII localization.

We next determined whether reducing the monomeric concentration of actin, by crosslinking or polymerizing actin, is sufficient to inhibit transcription in vitro. First, we pre-incubated HeLa nuclear extracts with vehicle or the purified actin-crosslinking domain (ACD) of V. cholera MARTX. ACD toxin catalyzes inter-molecular amide bond formation between actin monomers, covalently crosslinking monomeric actin (Kudryashov et al., 2008). Western blots of ACD-treated nuclear extracts showed a ladder of actin crosslinked into oligomers of increasing molecular weights and a concomitant decrease in the amount of non-crosslinked actin (Fig. 6A). In vitro transcription assays on nuclear extracts incubated with ACD showed significantly reduced levels of transcription compared to vehicle (Fig. 6B). Immunoprecipitation of RNAPII showed that the amount of uncrosslinked actin bound to RNAPII is decreased following ACD treatment and RNAPII binds crosslinked actin poorly if at all (Fig. 6C). Second, we performed in vitro transcription assays in HeLa nuclear extracts pre-incubated with phalloidin. Sedimentation assays performed on purified HeLa nuclear extract incubated with phalloidin showed a dramatic shift from the soluble to the insoluble pool of nuclear actin (Fig. 6D). Furthermore, extracts pre-treated with 10 µM phalloidin exhibited significantly decreased rates of in vitro transcription (Fig. 6E). Correspondingly, we found adding phalloidin decreased the amount of actin co-immunoprecipitated with RNAPII (Fig. 6F). Taken together, these in vitro effects parallel those of forming nuclear actin filaments in cells and demonstrate that oligomeric actin structures or polymerized actin decreases the amount of native actin available to interact with RNAPII. This decreased association of actin with RNAPII coincides with a decrease in transcription independent of nuclear actin filament formation.

It is now established that actin is an abundant nuclear protein (de Lanerolle and Serebryannyy, 2011; Gettemans et al., 2005; Grosse and Vartiainen, 2013). However, the form and functions of nuclear actin in normal mammalian nuclei remain controversial. Taking advantage of an actin mutation that promotes nuclear actin filament formation and disease, along with approaches to induce nuclear actin filaments both in cells and in vitro, we show that forming nuclear actin filaments inhibits transcription by RNAPII. Formation of nuclear actin filaments sequesters a pool of non-polymeric actin required for the proper localization and activation of RNAPII. Overall, our data appear to rationalize the general absence of stable actin filaments in mammalian somatic nuclei. Although our study aimed to understand the role of persistent or stable nuclear actin filaments on transcription, recent studies have reported that nuclear actin polymers and filaments can be transiently induced by specific stimuli (Baarlink et al., 2013; Belin et al., 2013, 2015; McDonald et al., 2006; Plessner et al., 2015). Thus, these studies, together with our own, suggest the exciting possibility that nuclear actin might exist in a dynamic equilibrium between transiently formed polymers and monomers that locally regulate transcription.

Our findings are consistent with previous work demonstrating a role for nuclear actin in mRNA transcription (Egly et al., 1984; Hofmann et al., 2004; Scheer et al., 1984). We find that sequestering actin in nuclear actin filaments (Fig. 2) correlates with reduced MHCIITA promoter recruitment (Fig. 5C) and lower rates of gene transcription (Figs 3A,B, 4G and 5D). In agreement, RNAPII preferentially interacts with the non-polymerizing R62D-β-actin mutant rather than the polymerization-promoting S14C-β-actin mutant (Fig. 5A). Moreover, the actin–RNAPII interaction is impaired in the presence of nuclear actin filaments (Fig. 5B), actin polymerization by phalloidin, and actin crosslinking (Fig. 6). Notably, we find expression of nuclear localized R62D-β-actin, but not cytoplasmic R62D-β-actin, is able to restore transcription in nuclei with filaments (Fig. 3B). Similarly, depleting nuclear actin through changes in growth factors or extracellular molecules has been reported to inhibit RNAPII activity, and this inhibition can be rescued by expressing NLS-R62D-β-actin (Spencer et al., 2011). These data, which support the notion that monomeric actin is important in maintaining transcription, are consistent with the previous demonstration that monomeric nuclear actin interacts with chromatin remodeling complexes (Kapoor and Shen, 2014; Serebryannyy et al., 2016) and the transcription factor MAL (Vartiainen et al., 2007).

Our in vitro transcription assays similarly demonstrate that polymerizing actin represses RNAPII (Fig. 6). Importantly, these assays investigate the effects of polymerizing and crosslinking actin independently of nuclear actin filament formation, changes in nuclear architecture or cytoplasmic actin. Crosslinked actin does not interact with RNAPII and less actin associates with RNAPII when actin is polymerized with phalloidin (Fig. 6). It is telling that crosslinking actin with ACD or polymerizing actin with phalloidin led to significant decreases in mRNA transcription (Fig. 6B,E). Similarly, Egly et al. have reported that adding actin to in vitro transcription assays stimulates mRNA transcription, but actin pre-treated with phalloidin abolishes the increase seen with actin alone, strongly suggesting that polymerized actin inhibits transcription (Egly et al., 1984). Because treatments that increase nuclear actin polymerization inhibit RNAPII activity, these results denote that nuclear actin monomers or highly dynamic polymers facilitate transcription.

One of our most striking observations was the reorganization of RNAPII into larger clusters in nuclei with actin filaments (Fig. 4A–F). Is this due to the effect of nuclear actin filaments on the availability of monomeric actin in the nucleus or to a secondary effect? Our data argue in favor of the former. Co-expressing NLS-R62D-β-actin, to increase the levels of monomeric nuclear actin, with the mCherry–supervillin fragment to induce filaments, decreased the size of RNAPII clusters (Fig. 4B; Fig. S3A), strongly suggesting a role for monomeric actin in transcription. Furthermore, RNAPII clusters induced by nuclear actin filaments exhibit multiple similarities to those found with direct RNAPII inhibition (Bregman et al., 1995), including a decrease in the actin–RNAPII interaction and changes in the organization of RNAPII (Fig. S4). Interestingly, direct pharmacological inhibition of RNAPII in Xenopus oocyte nuclei increases nuclear actin polymerization (Scheer et al., 1984), and nuclear actin polymerization might be necessary for RNAPII interchromatin granule clusters formed during transcription inhibition in mammalian cells (Wang et al., 2006). Another possibility is that the constructs used to polymerize nuclear actin alter actin polymerization in the cytoplasm. However, RNAPII clustering is very rare when these constructs are only expressed in the cytoplasm, and depolymerizing the actin cytoskeleton or inhibiting myosin II did not change the effect of nuclear actin filaments on RNAPII clustering or transcription (Figs 3, 4; Fig. S3). Thus, the reorganization of RNAPII into large factories in nuclei with actin filaments is likely due to the polymerization and subsequent sequestering of nuclear actin, not cytoplasmic influences or the filaments themselves.

Taken together, our data point to a model in which a dynamic pool of nuclear actin affects the localization and dynamics of active RNAPII, perhaps bound to other actin-binding proteins such as cofilin (Dopie et al., 2012; Obrdlik and Percipalle, 2011). They suggest a model in which polymerizing nuclear actin sequesters and decreases the monomeric pool of actin. This, in turn, attenuates the actin–RNAPII interaction and RNAPII activity and decreases transcription and proliferation (Figs 3, 5 and 6). This model is consistent with studies in which injecting actin antibodies or depleting nuclear actin inhibited transcription (Hofmann et al., 2004; Rungger et al., 1979; Scheer et al., 1984; Spencer et al., 2011). In support of this idea, monomeric actin (Posern et al., 2002; Vartiainen et al., 2007) and transient actin filaments (Baarlink et al., 2013; Plessner and Grosse, 2015) have been implicated in signal transduction whereas stable nuclear actin filaments are associated with pathologies (Domazetovska et al., 2007b; Munsie et al., 2011; Stenzel et al., 2015). Because actin filament formation is related to the monomer concentration (Pollard and Cooper, 2009), regulating the concentration of nuclear actin might be one mechanism for preventing the formation of persistent filaments and avoiding their negative effects.

This model is distinct from models describing cytoplasmic actin, which are dominated by the polymerization of actin into filaments that interact with myosins to generate force (Pollard and Cooper, 2009). However, actin evolved long before the first myosin (Goodson and Hawse, 2002; Hofmann et al., 2009; Pollard and Cooper, 2009) and actin must have had essential, conserved functions that predated myosin-mediated force production. Importantly, our data suggest roles for nuclear actin that are independent of its ability to form canonical filaments. It is tempting to speculate that a pool of non-polymerized actin might be able to interact more adroitly with multiple binding partners than filamentous actin. Although this possibility is intriguing, additional experiments are needed to fully understand the role of actin polymerization on the nucleus.

COS7 and HeLa cells were obtained from the ATCC and cultured in Dulbecco's modified Eagle's medium (DMEM; Corning) supplemented with 1% penicillin–streptomycin and 10% fetal bovine serum (Invitrogen).

Primary antibodies were purchased as shown: RNAPII antibodies, H5, H14 (western blotting, 1:5000) (Covance), 4H8 (immunofluorescence, 1:150; western blotting, 1:5000; immunoprecipitation, 5 μg) (Abcam and Active Motif); β-actin antibodies Ac-15 (immunofluorescence, 1:200; western blotting, 1:10,000; immunoprecipitation, 5 μg) and Ac-40 (immunofluorescence, 1:100; western blotting, 1:1000) (Sigma-Aldrich) and pan-actin C4 (immunofluorescence, 1:100; western blotting, 1:5000) (EMD Millipore); non-specific IgG (immunoprecipitation, 5 μg) (Santa Cruz Biotechnology); BrdU/BrU (immunofluorescence, 1:200) (BU-1, EMD Millipore), GFP (immunoprecipitation, 3 μg) (Abcam) and SC-35 (immunofluorescence, 1:100) (SRp30b, EMD Millipore).

Secondary antibodies (1:100) were used as follows: goat IgGs conjugated to Dylight 488 (Thermo Scientific), Texas Red (Jackson Labs), Cy3 (Jackson Labs) and Cy5 (Jackson Labs), as well as Rhodamine–phalloidin (1:400) (Cytoskeleton Inc.). Mounting medium containing DAPI (Vectashield) was used for immunocytochemistry.

The following constructs were as previously described: EGFP–supervillin fragment (Wulfkuhle et al., 1999), α-actin–GFP and the V163M mutant (Domazetovska et al., 2007a), the EYFP–NLS-β-actin construct and mutations of it (Chang et al., 2011; Posern et al., 2002), GFP–RNAPII (Sugaya et al., 2000) and the MHCIITA promoter luciferase construct (Hofmann et al., 2004). To generate the mCherry–supervillin fragment, the original construct was cut using EcoRI, purified, and inserted into the pmCherry-C1 vector backbone (Clontech). The U2OS-263 system was used as described (Darzacq et al., 2007).

Cell transfections were carried out using Polyjet (SignaGen), ExtremeGene HD (Roche) or JetPrime (Polyplus) reagents. Transfection conditions were optimized to obtain comparable levels of EGFP between constructs to minimize non-specific effects.

COS7 cells were transfected with the GFP–RNAPII (RPB1) construct and selected with 0.5 μg/ml α-amanitin (Santa Cruz Biotechnology) added directly into the medium. Colonies were selected and grown continually in DMEM supplemented with 0.5 μg/ml α-amanitin. Where indicated, 5,6-dichloro-1-b-dribofuranosylbenzimidazole (DRB, Calbiochem) or α-amanitin was added directly into the medium for 3 h at 100 mM and 10 μg/ml, respectively.

Cells were plated on glass coverslips >24 h before fixation or transfection. Cells were fixed in 4% paraformaldehyde (PFA) for 10 m permeabilized with 0.3% Triton X-100 (Sigma-Aldrich) in PBS for 7 min. For ethanol fixation, cells were briefly fixed in cold absolute alcohol and washed thoroughly. Cells were washed with PBS and incubated in 2% BSA in PBS for 1 h at room temperature and stained using a humidity chamber. Primary antibody was added for 1 h at room temperature or overnight at 4°C. Cells were washed with PBS and secondary antibody was added for 1 h at room temperature. Cells were washed and mounted using Vectashield containing DAPI.

Confocal images were obtained using a Zeiss LSM 710 confocal microscope at the UIC Confocal Microscopy Facility or at the UCI Laboratory for Fluorescence Dynamics. Images were analyzed using Zeiss Zen software. Structured illumination microscopy was performed using a Nikon N-SIM super resolution microscope through the Nikon Imaging center at Northwestern University.

EYFP–NLS-β-actin, EYFP–β-actin and GFP–RNAPII FRAP was performed using a Zeiss LSM 710 confocal microscope with a 63×1.46 NA oil alpha Plan-Apochromat objective. COS7 cells were transfected or pre-treated with the indicated drugs and incubated at 37°C and 5% CO₂ during image acquisition. To perform FRAP, five pre-bleach images were taken before nuclei were bleached in ∼1 μm circles at 100% 488 nm laser power for 25 iterations. Images were collected at 0.5 s for 60–120 s.

MS2–GFP FRAP was performed on U2OS-263 cells co-transfected with the mCherry–supervillin fragment, the doxycycline-sensitive transcriptional activator chimera rTetR and MS2–GFP for 48 h (Darzacq et al., 2007). Images were acquired using a Zeiss LSM 710 confocal microscope with a 40×1.2 NA water alpha Plan-Apochromat objective. Five pre-bleach images were acquired at 256×256 resolution, then MS2 foci were bleached in ∼2 μm circles at 100% 488 nm laser power for 100 iterations. Images were acquired every 600 ms for 500 s.

Fluorescence intensity was normalized to pre-bleach intensity and corrected for non-specific photobleaching and background using the Zeiss FRAP module. Images were acquired from at least two separate experiments. Graphing analysis and curve fitting was performed using Graphpad Prism and FRAPAnalyser (Halavatyi et al., 2009).

Cells were collected in PBS and lysed in 10 volumes of 50 mM Tris-HCl at pH 7.5, 2 mM EDTA, 150 mM NaCl, and 1% Triton X-100 (immunoprecipitation buffer) followed by brief sonication. Where indicated, cells were first chemically crosslinked in 1 mM Dithiobis (succinimidyl propionate) before lysis. Extracts were incubated with the indicated antibodies overnight at 4°C. Protein G magnetic beads (25 μl of a 50% solution; Thermo Scientific) were added and the mixture was incubated for another 2 h at 4°C. The beads were washed extensively in immunoprecipitation buffer, and eluted by boiling in SDS sample buffer. Immunoprecipitation assays with phalloidin were filtered to remove excess phalloidin before antibody was added, and beads were additionally washed with TBS to reduce non-specific binding before elution. In the GFP pulldowns, either anti-GFP antibody or GFP-Trap magnetic beads were used. GFP-Trap beads (20 μl of a 50% solution; Chromotek) were incubated in cell lysate for 4 h at 4°C instead of antibody or protein G.

Nuclei were purified by incubating cells in hypotonic buffer (10 mM HEPES pH 7.9, 1.5 mM MgCl₂, 10 mM KCl, 0.5 mM DTT) followed by mechanical perturbation through a 25 g needle. Isolated nuclei were then lysed in transcription buffer (20 mM HEPES pH 7.9, 20% glycerol, 100 mM KCl, 0.2 mM EDTA, 0.5 mM PMSF and 0.5 mM DTT) with a 30 g needle and cleared by centrifugation (10 min at 10,000 g). To perform sedimentation assays, extracts were spun at 100,000 g for 1 h at 10°C. Soluble fractions were directly boiled in hot SDS. Pellets were sonicated in hot SDS to ensure complete suspension.

Brightness of the expressed constructs was extracted from a number and brightness analysis performed as previously described (Digman et al., 2008). Time-sequences (100 frames) of raster scan images were collected using an Olympus FV1000 Confocal Microscope, using the following parameters: pixel dwell time, 8 µs; frame size, 256×256 pixels; photon counting detection; excitation wavelength 488 nm (1 and 2) and/or 515 nm (2 and 3). Power level at the sample were: 1.2 µW at 488 nm and 3.3 µW at 515 nm. Emission filters were: DM 488/453/633, BA 505–605 bp (at 488 nm excitation) and DM 405-400/515, BA 535–565 bp (at 515 nm excitation).

The molecular brightness for each pixel of the image was calculated according to published protocols and software (Globals for Images). Briefly, molecular brightness is defined as B=σ²/(S*), where σ² is the variance of the fluorescence signal measured in a pixel over time, is the average intensity and S (S=5) is a correction factor to account for the fact that the detector is not an ideal photon-counting module. The average brightness from all the pixels of the image of the nucleus that do not contain any filament or bright structure is used to provide the B-value of a given cell. Under the experimental conditions employed, the brightness of monomeric EGFP and EYFP appeared to be comparable (Storti et al., 2012).

Transfected cells were stained with anti-RNAPII antibody. Individual nuclei images were thresholded to identify single fluorescence foci using the find maximum function, and the relative area of these foci (as a measure of number of pixels) was obtained using ImageJ.

HeLa cells transfected with EGFP or actin constructs were treated with 100 ng/ml recombinant Interferon-γ (Invitrogen) 16 h before harvesting. Cells were then collected in cold PBS and lysed in 5 mM PIPES pH 8.0, 85 mM KCL, 0.5% NP-40, 0.5 mM PMSF plus protease inhibitor cocktail and phosphatase inhibitor cocktail (Sigma-Aldrich). Nuclei were pelleted and re-suspended in formaldehyde buffer [5 mM PIPES pH 8.0, 85 mM KCL, 1% formaldehyde (Sigma-Aldrich)], 0.5 mM PMSF plus protease inhibitor cocktail and phosphatase inhibitor cocktail for 13 min at room temperature. The reaction was quenched with 67 mM glycine for 5 m. Fixed nuclei were pelleted and washed in 5 mM PIPES pH 8.0, 85 mM KCl, 0.5 mM PMSF plus protease and phosphatase inhibitor cocktails. Nuclei were pelleted by centrifugation and resuspended in 0.5% SDS, 10 mM EDTA, 50 mM Tris-HCl pH 7.5, 0.5 mM PMSF plus protease and phosphatase inhibitor cocktails. Following sonication to generate 200–600-bp DNA fragments, nuclear extracts were diluted 10× with antibody binding buffer (0.01% SDS, 1.1% Triton X-100, 1.2 mM EDTA, 16.7 mM Tris-HCl pH 8.1, 167 mM NaCl, 0.5 mM PMSF plus protease and phosphatase inhibitor cocktails). 5 μg anti-RNAPII (4H8) antibody was added to each sample and incubated overnight at 4°C. Antibody-bound protein–DNA complexes were captured by adding 25 μl of Dynabeads Protein G (Thermo Scientific) for 2 h at 4°C. The complexes were washed (1 ml) once with low-salt buffer (0.1% SDS, 1% Triton X-100, 2 mM EDTA, 20 mM Tris pH 8.1, 150 mM NaCl), once with high-salt buffer (same buffer made 500 mM NaCl), once with LiCl wash buffer (0.25 M LiCl, 1% NP-40, 1 mM EDTA, 10 mM Tris pH 8.0), and twice with TE buffer. 10% Chelex-100 (Biorad) was added to input samples and washed beads to recover DNA and boiled for 15 min at 100°C. 20 μg/ml Proteinase K (Thermo Scientific) was added for 30 min at 55°C with shaking. Proteinase K was inactivated by boiling for 10 min. Quantitative PCR (qPCR) analysis was conducted using iQ Sybr Green reagent (Biorad). Primers for the MHCIITA promoter (5′-GCCTCCAGTCGGTTCCTCACAG and 5′-CACGGTTGGACTGAGTTGGAGAGA) were used. qPCR was run at 95°C for 3 min then 40 cycles of 95°C for 20 s and 63°C for 40 s. Ct values were normalized to total input.

The MAL/SRF luciferase reporter (Baarlink et al., 2013) and the MHCIITA promoter luciferase construct (Hofmann et al., 2004) were used as described previously. Luciferase activity (RLU) was normalized to luciferase activity in EGFP co-transfected cells. For MHCIITA promoter induction, COS7 cells were treated with 500 U Interferon-γ (Sigma-Aldrich) overnight. RLU was normalized to protein concentration and cell transfected with MHCIITA-driven luciferase alone.

COS7 cells were grown on glass coverslips and transiently transfected for 48 h then incubated with 10 µM BrdU for 5 h or 2 mM BrU for 2 h. Cells were fixed in 4% PFA and permeabilized in 0.3% Triton X-100. In the case of BrdU incorporation, cells were then incubated in 2 M HCl for 30 m and neutralized in phosphate buffer. Cells were subsequently stained with anti-BrdU antibody and imaged with an epifluorescent microscope. Fluorescence intensity was quantified using Image J. Fluorescence was normalized to non-transfected cells in each frame.

In vitro transcription assays were performed as previously described (Hofmann et al., 2004). Briefly, purified HeLa nuclear extract (30 µg, Promega) in transcription buffer (20 mM HEPES pH 7.9, 20% glycerol, 100 mM KCl, 0.2 mM EDTA, 0.5 mM PMSF, and 0.5 mM DTT) with 1 mM ATP and 4 mM MgCl₂ was pre-incubated with 0.5 µM purified ACD (Kudryashov et al., 2008) or buffer alone for 1 h at room temperature or with 10 µM phalloidin for 30 min at 4°C. After pre-incubation, RNAse T1 was added, along with 200 ng P41 DNA template containing the adenoviral major late promoter and G-less transcribing region and 0.4 mM of ATP, UTP and 0.02 mM CTP. ³²[P]αCTP was then added to the reaction and incubated for 30 min at 30°C. mRNA transcript was purified using a Trizol–choloroform extraction and run using 6 M urea PAGE.

We thank Elizabeth Luna (University of Massachusetts Medical School) for the EGFP–supervillin constructs, Kathryn North (University of Sydney) for the α-actin–GFP constructs, Marc Vigneron (Université de Strasbourg) for the GFP–RNAPII construct, Maria Vartiainen (University of Helsinki) for the MAL/SRF construct, Teng-Leong Chew (Janelia Farm) for help with SIM microscopy, and the UIC Research Resource Center for help in imaging and DNA sequencing.