Intranuclear Actin Structure Modulates Mesenchymal Stem Cell Differentiation

Abstract

Actin structure contributes to physiologic events within the nucleus to control mesenchymal stromal cell (MSC) differentiation. Continuous cytochalasin D (Cyto D) disruption of the MSC actin cytoskeleton leads to osteogenic or adipogenic differentiation, both requiring mass transfer of actin into the nucleus. Cyto D remains extranuclear, thus intranuclear actin polymerization is potentiated by actin transfer: we asked whether actin structure affects differentiation. We show that secondary actin filament branching via the Arp2/3 complex is required for osteogenesis and that preventing actin branching stimulates adipogenesis, as shown by expression profiling of osteogenic and adipogenic biomarkers and unbiased RNA-seq analysis. Mechanistically, Cyto D activates osteoblast master regulators (e.g., Runx2, Sp7, Dlx5) and novel coregulated genes (e.g., Atoh8, Nr4a3, Slfn5). Formin-induced primary actin filament formation is critical for Arp2/3 complex recruitment: osteogenesis is prevented by silencing of the formin mDia1, but not its paralog mDia2. Furthermore, while inhibition of actin, branching is a potent adipogenic stimulus, silencing of either mDia1 or mDia2 blocks adipogenic gene expression. We propose that mDia1, which localizes in the cytoplasm of multipotential MSCs and traffics into the nucleus after cytoskeletal disruption, joins intranuclear mDia2 to facilitate primary filament formation before mediating subsequent branching via Arp2/3 complex recruitment. The resulting intranuclear branched actin network specifies osteogenic differentiation, while actin polymerization in the absence of Arp2/3 complex-mediated secondary branching causes adipogenic differentiation.

Introduction

Mesenchymal stem cell (MSC) differentiation responds to both chemical and physical cues generated in the extracellular environment. Physical influences rely on the actin cytoskeleton to transmit signals that control gene expression, either through providing structural signaling nodes organized around actin structures [1, 2] or transmission of force to the via actin connections that impact nuclear architecture through linker of nucleus and cytoplasm (LINC) complexes [3]. It has been suggested that the enriched cytoskeletal structure conferred by increased numbers of F-actin stress fibers promotes osteoblastic differentiation of MSC, while concomitantly preventing adipocyte differentiation [4, 5]. Generation of actin structure requires activation of the RhoA system [6], while recruitment of formins mobilizes end-on-end polymerization and initiate secondary filament branching through association of the Arp2/3 complex [7]. It is well known that these members of the actin toolbox are critical for cell mobility and division, and dynamic actin polymerization has also been shown to regulate incoming signal density [2, 8] and fine-tune specific aspects of signal transduction [9]. Importantly, the actin cytoskeleton generates force on the nuclear envelope through nesprin [10], a component of the LINC complex that connects to the internal nuclear leaflet and nucleoskeleton [11]. External force thus not only changes nuclear shape [12] but also influences chromatin structure, thus influencing gene silencing and mobility [13, 14]. Furthermore, actin trafficking into the nucleus may directly affect chromatin nuclear architecture to modulate gene expression.

We previously demonstrated that continuous cytochalasin D (Cyto D) disruption of the actin cytoskeleton primarily leads to osteogenic lineage commitment of bone marrow-derived MSCs (mdMSCs) in vitro and in the tibial marrow space of rodents in vivo [15]. Because our findings with Cyto D challenge a model that invokes a linkage between decreased cytoskeletal rigidity and adipogenesis [5], we established that the osteogenic effect of Cyto D requires actin translocation into the nucleus. Indeed, most cellular actin is found inside the nucleus after Cyto D treatment. Remarkably, osteoblast differentiation is entirely prevented when actin transfer into the nucleus is blocked by silencing either one of the actin cotransporters, cofilin-1, or importin-9. We also noted that a proportion of multipotential MSCs treated with Cyto D entered the adipogenic lineage, both in vitro and within the murine tibial space [15]. As mdMSCs are inclined toward osteogenic differentiation [16], such adipogenesis may represent actin induction of alternate lineage speciation or trans-differentiation.

Intranuclear actin supports gene transcription by enhancing RNA polymerase and regulating the location of heterochromatin [17]. Nuclear actin depletion is associated with cell quiescence with decreased proliferation [18]. Availability of monomeric or polymeric actin in the nucleus may further contribute to specific cell phenotype decisions by interacting with multiple transcription factors [19, 20]. A functional role for actin structure in gene expression is also suggested by the nuclear presence of molecules that regulate actin polymerization [21], and supported by the finding that actin polymers contribute to the ability of MAL (myocardin-like protein 1 (MKL-1)) to regulate myocyte proliferation [19]. Furthermore, depolymerization of the cytoplasmic actin cytoskeleton in MSCs causes a rapid and robust Runx2-dependent osteogenic differentiation that depends on intranuclear actin transport [15]. In this study, we addressed the key question of whether intranuclear actin structure specifically determines the mesenchymal cell phenotype during differentiation.

Materials and Methods

Reagents

Fetal bovine serum was from Atlanta Biologicals (Atlanta, GA). Culture media, trypsin-EDTA reagent, antibiotics, Cyto D were from Sigma-Aldrich, St Louis, Mo; CK666 was from R&D, Minneapolis, MN.

RNA Interference

The cells were transfected with siRNA (50 nM) in serum-free OptiMEM overnight before replacing the medium and adding reagents for cell treatment. siRNAs were as follows-for importin 9: 5′-CCCAGCUCUUCAACCUGCUUAUGGA and control (nucleotide change within same sequence) 5′-CCCTCTCCTAACCGTTCATTGAGGA; for Pparg: 5′-CCGACACCTGCAGATTGATATTGAG and control 5′-CCGCCACGTAGAGTTATATTACGAG; for MKL-1: 5′-CAACCCAAGTCTGCCAGCGAGAAAT and control 5′-CAAACGACTGTCCGAGCGAACCAAT; for mDia1: 5′-CCGACACCTGCAGATTGATATTGAG and control 5′-CCGCCACGTAGAGTTATATTACGAG; for mDia2: 5′-CAGAGTCCATGATTCAGAACTTAAT and control 5′-CAGCCTGTATTAGACCAATTAGAAT.

Cells and Culture Conditions

Mouse mdMSCs were harvested from murine marrow using a published protocol [22, 23]. These MSCs were maintained in minimal essential medium (MEM) containing 10% fetal bovine serum and 100 μg/ml penicillin/streptomycin. For experiments, the cells, used at passages 6–12, were plated at a density of 10,000 cells per square centimeter in six-well culture plates (Fisher) in MEM and cultured for 1 day before application of treatments. Adipogenic (0.1 μM dexamethasone, 5 μg/ml insulin, and 50 μM indomethacin) or osteogenic (50 μg/ml ascorbic acid, 10 μM β-glycerophosphate) additions were added to media as indicated.

Immunofluorescence

For microscopy, cells were fixed with 4% paraformaldehyde × 10 minutes, permeabilized in 0.1% Triton-X 100 × 5 minutes, blocked sequentially with 0.2 M glycine × 10 minutes, separated by three 10-minute phosphate-buffered saline (PBS) washes between steps. Silicone membranes were cut from plates and transferred to six-well plate surface. Actin stress fibers were visualized with Alexa Fluor 488-conjugated phalloidin (Invitrogen). After 3 × 10 minute washes, membranes were sealed with mounting medium on glass. Cells were imaged on an Olympus BX61 inverted microscope system using filters: Alexa Fluor 488 Phalloidin: Semroc 3540B. For confocal and three-dimensional pictures, cells were cultured on m-slide (Ibidi) and treated with EverFluor-TMR-conjugated-Cyto D (SETAREH Biotech) or transfected with YFP-NLS-actin (Addgene) plasmids and imaged with an Olympus IX70 confocal microscope.

Image Analysis

Following Cyto D treatments, cells were fixed and immunostained against Lamin B1 (ab16048, Abcam, Cambridge UK) and F-actin (phalloidin). Using Zeiss LSM 710 confocal microscope, the entire height of individual cells was imaged at intervals of 0.36 mm using the same parameters. Deconvoluted confocal image stacks (Post Auto Quant) were imported into NIH imageJ software (https://imagej.nih.gov/ij/). Using Lamin B1 as a landmark, nuclear height was quantified via counting the number of stacks between first and last slices with detectable, in-focus Lamin B1 signal using cross-sectional images. Next, the entire nuclear section was collapsed into a single image using “Maximum Intensity Projection.” Nuclear area was measured via tracing the outer circumference of Lamin B1. Nuclear intensity was reported as the mean actin intensity within the area traced by Lamin B1.

Real-Time Reverse Transcriptase Polymerase Chain Reaction

Total RNA was isolated with the RNeasy mini kit (Qiagen) and treated with DNase I. Reverse transcription of 1 μg of RNA in a total volume of 20 μl was performed with iScript cDNA Synthesis Kit (Bio-Rad, Hercules, CA) before real-time polymerase chain reaction (PCR) (Bio-Rad iCycler). The 25-μl amplification reactions contained primers (0.5 μM), dNTPs (0.2 mM each), 0.03 units Taq polymerase, and SYBR-green (Molecular Probes, Waltham MA) at 1:150,000. Aliquots of cDNA were diluted 5– 5,000-fold to generate relative standard curves to which sample cDNA was compared. Fabp2, Adipoq, Pparg, Alpl, Sp7, Runx2, Bglap, and 18S primer sets were those previously reported. mDia1 forward primer: 5′-TCCAAGCTGACAGGAGA GGT-3′ and reverse primer: 5-GGGGGAGGTGGAATAACAGT-3′. mDia2 forward primer: 5′-TCAAACCTTCGGATTTGACC-3′ and reverse primer: 5′-TAAATGTGCCAAATCGTCCA-3′. Standards and samples were run in triplicate. PCR products were normalized to 18 S amplicons in the room temperature (RT) sample, and standardized on a dilution curve from RT sample.

Nuclear and Cytoplasmic Protein Fractionation

Cells were washed with 1× PBS, the cell pellet resuspended in 0.33 M sucrose, 10 mM Hepes, pH 7.4, 1 mM MgCl₂, 0.1% Triton X-100 (pellet vs.buffer, 1:5), and placed on ice for 15 minutes. After 3,000 rpm for 5 minutes, the supernatant was collected (cytoplasmic fraction). The pellet was resuspended in 0.45 M NaCl and 10 mM Hepes, pH 7.4, and placed on ice for 15 minutes. After centrifugation at 12,000 rpm for 5 minutes, the nuclear fraction supernatant was collected.

Nuclear G-Actin/F-Actin Assay

To assay actin, 10⁷ cells per condition were collected for fractionation of nuclear proteins and assay using “G-actin/F-actin Assay Kit” (Cytoskeleton, Inc., Denver CO). As per instructions, 100 ml LAS2 buffer was added to nuclear pellets for homogenization through aspiration in a 25-G syringe; lysates were incubated at 37°C for 10 minutes then centrifuged at 350g for 5 minutes at RT. Supernatant, containing G-actin, was then collected after 100,000g at 37°C for 1 hour. F-actin depolymerization buffer of 100 ml is added to the pellet to extract F-actin.

Immunoblot

Whole cell lysates were prepared with lysis buffer (150 mM NaCl, 50 mM Tris HCl, 1 mM EGTA, 0.24% sodium deoxycholate, 1% Igepal, pH 7.5) containing 25 mM NaF and 2 mM Na3VO4, aprotinin, leupeptin, pepstatin, and phenylmethylsulfonyl fluoride were added before each lysis. Fractionated or whole lysate proteins of 5–20 μg were loaded onto a 7%–10% poly-acrylamide gel for chromatography and transferred to polyvinylidene difluoride membrane. After blocking, primary antibody was applied overnight at 4°C including antibodies against Bglap, Pparg, PARP1 (Cell Signaling, Danvers Mass); Runx2, mDia2, MKL-1, importin-9, Arp3 (Abcam), mDia1 (BD), LDHA (Millipore, St Louis, Mo), Fabp4 (ProSci, Poway, CA), actin, beta-tubulin (Santa Cruz, Dallas TX), Adipoq (Affinity Bioreagents, Golden, CO). Secondary antibody conjugated with horseradish peroxidase was detected with ECL plus chemiluminescence kit (Amersham Biosciences, Piscataway NJ). The images were acquired with an HP Scanjet and densitometry determined using NIH ImageJ, 1.37v.

Alpl Assay

The cells were scraped into Alpl lysis buffer (10 mM Tris-HCl, pH 8.0; 1 mM MgCl₂; 0.5% Triton X-100 in PBS). The samples were sonicated for 5 minutes. The Alpl activity was measured with Alkaline Phosphatase Assay Kit (Abcam).

RNA Seq

High-resolution RNA-sequencing (RNA-seq) was performed using three distinct mouse mdMSC preparations that were each treated in triplicate for 24 hours with Cyto D and/or CK666. The biological triplicates were pooled before RNA-seq to yield technically robust datasets for each of the three MSC preparations. RNA-seq analysis of MSCs was performed as described previously [24, 25] with oligo dT purified mRNA and indexed cDNAs (standard TruSeq Kits 12-Set A and 12-Set B) on the Illumina platform (Illumina’s RTA version 1.17.21.3). Raw reads were mapped using a well-established pipeline (Bioinformatics Core standard tool) that includes analysis using the MAPRSeq v.1.2.1 system, alignment with TopHat 2.0.6, gene counting with the HTSeq application, as well as normalization and expression analysis using EdgeR. Gene expression is expressed in reads per kilobase pair per million mapped reads (RPKM). The sequencing data are available through the Center for Biotechnology Information (Gene Expression Omnibus accession GSE89132).

Statistical Analysis

Results are expressed as mean ± SEM. Statistical significance was evaluated by one-way analysis of variance or t test as appropriate (GraphPad Prism). All experiments were replicated at least three times to assure reproducibility.

Results

Cyto D Actin Depolymerization Increases F-Actin within the Nucleus

Actin monomer transport into the nucleus is highly regulated [26], perhaps in part because actin is known to support gene transcription [27, 28]. As previously shown [15], Cyto D causes depolymerization of the cytoplasmic cytoskeleton, with monomeric and dimeric actin transfer into the nucleus (Fig. 1A). We validated the observation that Cyto D does not enter the nucleus [29] by tracking EverFluor-TMP-conjugated-Cyto D; labeled Cyto D remained in the cytoplasm and was restricted from nuclear entry (Fig. 1B). Because Cyto D is restricted to the cytoplasm, its continued presence in the cell does not preclude actin polymerization within the nucleus.

Figure 1.

Cyto D treatment is associated with intranuclear actin polymerization. Except where indicated, marrow-derived mesenchymal stromal cells (MSCs) were cultured in growth medium and treated with Cyto D (0.1 mg/ml) for 3 days. (A): Control and Cyto D-treated MSCs stained with phalloidin (green), day 3. (B): Cells were cultured with 50 nM EverFluor-TMR-conjugated-Cyto D (red) for 24 hours and stained with phalloidin (green). Scale bars = 25 mM. (C): After transfection with YFP-NLS-actin plasmid for 48 hours and ±mDia1 or control siRNA for 24 hours, MSCs were treated with ±Cyto D for 24 hours (upper panel); 20 images were analyzed for each condition with ImageJ program (lower panel); ***, p < .001. (D, E): Nuclear G- and F-actin immunoblot analysis using G/F actin assay kit (Cytoskeleton, Inc.) for control and Cyto D-treated MSCs (D) or for Cyto D-treated MSCs ±mDia1 siRNA (E). Abbreviations: CTL, control; Cyto D, cytochalasin D; siCTL, Control siRNA; simDia, siRNA targeting mDia1.

To characterize actin polymerization within nuclei of Cyto D-treated cells, we overexpressed YFP-NLS-actin [30] in MSC. After Cyto D treatment, YFP-labeled actin was found within the nucleus including the appearance of filamentous formation (Fig. 1C). Confocal images of control cells show intranuclear actin distributed smoothly throughout a flat nucleus, whereas intense actin structures occupy most of the nuclear space in a Cyto D-treated cell. Nuclear shape was significantly altered by Cyto D treatment (Fig. 1C). Confocal images allowed quantification of nuclear height, area, and actin label intensity, shown graphically below the selected images. After treatment with Cyto D, nuclear height more than tripled, potentially through the loss of cytomechanical constraints on the apical cell surface via actin-LINC connections, or due to swelling of the nucleus with actin, shown by increased signal of intranuclear labeled actin (Fig. 1C, graphs below).

To evaluate whether preventing actin filament formation affects nuclear dimensions, we used siRNA to silence a formin involved in actin nucleation and elongation, mDia1 (Diaph1), which has been found in the nucleus and shown to be necessary for the effects of MKL-1 in NIH3T3 cells [19]. The nuclear height after Cyto D treatment in mDia1 silenced cells is similar to that in mDia1 replete cells (left graph, Fig. 1C), suggesting that the increased nuclear height is due to release of cytomechanical forces on the nucleus. In contrast, while Cyto D-treated mDia1 silenced cells have greater nuclear height, they do not have more nuclear actin signal (far right graph, Fig. 1C). This observation suggests that nonpolymerized actin moves freely between cytoplasm and nucleus in the absence of mDia1, as would be expected since active actin transport mechanisms recognize mono- and dimeric actin [31].

Quantification of nuclear F-/G-actin ratio confirms that nuclear actin increases after Cyto D and supports that much of increase in nuclear actin is in the form of F-actin (Fig. 1C). We next compared MSCa treated with a siRNA control or targeted against mDia1, followed by Cyto D induction of nuclear actin transport: in mDia1-silenced cells, there was a decrease in both G- and F-actin (Fig. 1E), indicating that not only was actin polymerization prevented but that actin monomers had also access to the known exportin-6 transport mechanism to the cytoplasm [32].

Cyto D Induction of Osteogenesis and Adipogenesis Are Both Dependent on Intranuclear Actin

We next set out to clarify whether Cyto D-associated adipogenesis, which accompanies osteoblast differentiation when Cyto D is administered for several days [15], was also dependent on actin transfer into the nucleus. MSCs in basal growth medium respond to 72 hours of Cyto D treatment with an adipogenic program of gene expression (e.g., Fab4, Pparg, and Adipoq), apart from the prominent osteogenic program (e.g., Alpl, Sp7, and Bglap), shown in Figure 2A. Furthermore, Cyto D enhances adipocyte differentiation in response to adipogenic differentiation medium, as reflected by greater expression of adipogenic mRNAs and proteins including the master adipogenic transcription factor, Pparg, after 3 days of treatment (Fig. 2B, 2C). When cultured in adipogenic medium, expression of osteogenic genes after Cyto D treatment is prevented (Fig. 2B). Similarly, when grown in osteogenic medium, MSCs show no adipogenic response to Cyto D (Fig. 2B). Thus, Cyto D enhances the activity of regulatory factors that induce differentiation along one pathway, but does not allow lineage switching. When nuclear actin transport was prevented by silencing of the actin co-transporter importin-9 (Ipo9, Fig. 2D), adipogenesis was prevented both in the absence and presence of Cyto D (Fig. 2E, 2F). This result complements our findings for osteogenesis [15], demonstrating that Cyto D-induced differentiation of MSCs in either the adipogenic or osteogenic lineage is dependent on nuclear actin transport.

Figure 2.

Intranuclear actin is required for both Cyto D-induced adipogenesis and osteogenesis. (A, B): Reverse transcriptase polymerase chain reaction (RT-PCR) from control and Cyto D-treated mesenchymal stromal cells (MSCs) demonstrate expression of adipogenic (Fabp4, Adipoq, and Pparg) and osteogenic (Alpl, Sp7, and Bglap) genes in growth medium; *, p < .05 (A) or in adipogenic medium (B). (C): Western blot analysis for MSCs in adipogenic medium. (D): Importin-9 silencing analyzed by RT-PCR for ±Cyto D-treated cells. (E): Western blot and (F) RT-PCR analysis for ±Cyto D-treated cells ±importin-9 silencing. Abbreviations: A, adipogenic medium; CTL, control; Cyto D, cytochalasin D; O, osteogenic medium; siCTL, control siRNA; silpo9, siRNA targeting importin 9.

Adipogenesis requires expression of the master transcription factor Pparg, such that silencing of this factor prevents the adipogenic response to Cyto D (Fig. 3A), similar to the requirement of Runx2 for Cyto D -induced osteogenesis [15]. Interestingly, smooth muscle transcription factor MKL-1 (myocardin, MAL) affinity for actin monomers led to its calponin domains being used to generate monomeric actin probes [33]; increased cytoplasmic actin monomers have recently been shown to sequester MKL-1 outside of the nucleus [28]. The latter may relieve suppression of PPARg to promote adipogenesis of MSCs [34]. Similarly, the WW-domain containing transcriptional co-activators yes associated protein 1 (YAP) and tafazzin (TAZ) interact with polymeric actin, which directs their localization outside of the nucleus [35]. Our results show that MKL-1 is retained within the nucleus at 24 hours after Cyto D treatment (Fig. 3B, 3C). Cyto D increases the association of MKL-1 with actin but does not alter interactions of Pparg with actin based on immunoprecipitation assays (Fig. 3D). Although silencing of MKL-1 enhances adipogenesis, shown by modest upregulation of adiponectin and Fabp4 proteins, Cyto D is able to induce adipogenic differentiation in the absence of MKL-1 (Fig. 3E). Thus, MLK-1 translocation to the cytoplasm is not critical for Cyto D -induced adipogenesis.

Figure 3.

Intranuclear actin induction of adipogenesis requires nuclear PPARg and is independent of MKL-1. (A): Analysis for ± Cyto D cells ±Pparg targeted or control siRNA. (B): MKL-1 (red) and phalloidin (green) staining for cells ± Cyto D 1-day, scale bars = 25 mM. (C): Nuclear and cytoplasmic IB, ±Cyto D. (D): IP actin for ±Cyto D cells, with IB for actin, MKL-1, and Pparg as shown. (E): IB for ±Cyto D cells, ±MKL-1 silencing. Abbreviations: CTL, control; Cyto, cytosol; Cyto D, cytochalasin D; IB, immunoblot; IP, immunoprecipitate; MKL-1, myocardin related protein; Nuc, nucleus; siCTL, control siRNA; siPparg, siRNA targeting Pparg.

Intranuclear Branched Actin Is Critical for Osteogenesis, and Its Inhibition Induces Adipogenesis

The function of molecules necessary for creating a branched network, including Arp2/3, cofilin, and profilin, which are found in the nucleus [21], should not be affected by Cyto D as it is restricted from nuclear entry (Fig. 1B). Therefore, we addressed the key question as to whether the nuclear structure of actin might contribute to lineage-specification of MSC differentiation.

We validated that Arp3 is found in the nuclei of MSCs both before and after Cyto D treatment (Supporting Information Fig. S1). Preventing Arp2/3 function is lethal in developing organisms [36], but its function can be studied using CK666 [37, 38], specific enzymatic inhibitor that inhibits the active site of Arp2/3 by blocking a key conformational change required for activation [39].

In MSCs cultured in growth medium, inhibition of the Arp2/3 complex using CK666 completely prevents Cyto D-induced osteogenesis at 72 hours (Fig. 4A). Furthermore, CK666 promotes adipogenic differentiation when administered either alone or together with Cyto D. The increase in expression of osteogenic genes Alpl and Sp7 upon Cyto D treatment was largely prevented in the presence of CK666, and protein expression confirmed that this extended to protein translation (Fig. 4B, 4C). Strikingly, CK666 alone suffices to activate expression of adipogenic genes (e.g., Fabp4 and Adipoq; Fig. 4A) and proteins (e.g., Pparg; Fig. 4D), and this effect of CK666 is potentially enhanced by Cyto D.

Figure 4.

Branching of intranuclear actin via Arp2/3 favors osteogenesis over adipogenesis. ±Cyto D cells ±CK666 (100 mM) treatment analyzed by (A) reverse transcriptase polymerase chain reaction, (B, D) Western blot, and (C) alkaline phosphatase assay. (E): Heatmap analysis of conditions: control, +Cyto D, +CK666, + Cyto D/+CK666. (F): Expanded gene heat map for osteogenic and adipogenic genes upregulated by Cyto D and CK666, respectively. Abbreviations: CTL, control; Cyto D, cytochalasin D.

Arp2/3 inhibition will cause a loss of branched networks in the cytoskeleton as well as in the nucleus, which may be partially responsible for the enhanced adipogenesis seen at the 72 hours time point when cell proteins indicate adipogenic maturation. However, Arp2/3 inhibition enhances adipogenesis in the presence of Cyto D (Fig. 4A), when the majority of actin is no longer localized to the cytoplasm. This finding suggests that prevention of intranuclear secondary branching further accelerates the adipogenic program with a concomitant block in osteogenesis.

To query early molecular mechanisms that are subject to activation by Cyto D or CK666, we performed high-resolution mRNA expression profiling using next generation RNA-sequencing (RNA-seq) of MSCs that were treated for 24 hours with Cyto D and/or CK666 (Supporting Information Results; Figures). Expression values were normalized as reads per million after correction for gene length (RPKM) and filtered for genes that are expressed at clearly detectable levels (RPKM > 0.3), exhibited a statistically significant difference (p < .01) and a biologically relevant change (fold change >2) in pairwise comparisons between experimental groups (Fig. 4E, 4F, Supporting Information Fig. S2; Table 1). Heatmaps generated by hierarchical clustering revealed that the expression profiles of the four treatment groups readily separated into three main clades that differ in cellular phenotypes (top of heatmap) and four gene clades that encompass genes responding to either Cyto D or CK666 in comparison with control MSC (Fig. 4E). For example, Cyto D-treated MSCs are phenotypically different from control MSCs, and CK666-treated MSCs differ from control MSCs regardless of Cyto D treatment (Fig. 4E). Based on gene ontology analysis (Supporting Information Figures), we conclude that Cyto D activates an osteogenic program of many genes (Clade III) that are related to extracellular matrix deposition and cell signaling, but this program is selectively suppressed by CK666. Thus, loss of actin branching suppresses osteogenesis.

CK666, in contrast, induces an adipogenic program (Clades I and II), and only a subset of these genes are also upregulated by Cyto D treatment at 24 hours (Clade II). Hence, CK666 is a potent adipogenic stimulant that dominates over effects of Cyto D. Clade IV is a heterogeneous set of genes that is downregulated by both Cyto D and CK666 and includes genes involved in cell adhesion (e.g., ITGA5, ITGA11), cell signaling (e.g., CTGF), and cytoskeleton (e.g., ACTA1, ACTA2, ACTGA). RNA-seq analysis of representative known and novel biomarkers for osteogenesis (e.g., Sp7, Dlx5, Runx2, Atoh, Nr4a3, Sfn5) or adipogenesis (e.g., Fabp4, Adipoq, Plin1) clearly supports the general interpretation of the RNA-seq data that CK666 alone suffices to induce adipogenic gene expression, while Cyto D activates a program of known and novel genes linked to osteogenesis (Fig. 4F).

To evaluate whether such gene profiles were associated with quantifiable alterations in intranuclear actin, we performed confocal assessment of nuclear structure in the presence and absence of Arp2/3 complex inhibition and Cyto D. Cyto D increases nuclear height by nearly threefold in this experimental set, and this effect was maintained in the presence of CK666 (Fig. 5). In fact, the nuclei of cells where Arp2/3 complex was inhibited with CK666 also showed a significant increase in nuclear height, presumably due to reduced cytoplasmic cytoskeletal tension. However, CK666 treated nuclei lacked comparable increases in nuclear actin intensity in the absence of Cyto D -directed nuclear actin transfer (far R graph, Fig. 5). When Cyto D was added to CK666 inhibited cells, nuclear height rose still further (L graph, Fig. 5) and was accompanied by increased nuclear actin intensity. The increased nuclear actin intensity incumbent on Cyto D addition was decreased in the presence of CK666, perhaps due to loss of YFP-NLS-actin intensity that could be ascribed to loss of branching density [40]. Our data thus suggest that the presence of actin in the nucleus is insufficient to initiate osteogenesis in the absence of Arp2/3-directed secondary branching.

Figure 5.

CK666 limits Cyto D-induced increase in nuclear actin intensity. Mesenchymal stromal cells transfected with YFP-NLS-actin plasmid were treated with Cyto D and/or CK666, both causing increased height as shown in representative confocal images, Scale bars = 25 mM; 20 images were analyzed for each condition with ImageJ program, shown in the lower panel; *, p < .05; **, p < .01; ***, p < .001. Abbreviations: CTL, control; Cyto D, cytochalasin D.

Actin Polymerization Is Critical for Adipogenesis

The action of formins precedes and regulates the binding of nucleation factors that recruit Arp2/3. To investigate whether formin function contributes to MSC differentiation, we silenced mDia1 and mDia2 by RNA interference. Depleting mDia1 inhibits both adipogenesis and osteogenesis (Fig. 6A). Activation of the osteogenic gene expression program in response to Cyto D in the presence and absence of mDia1 was confirmed by alkaline phosphatase assay and protein immunoblots (Fig. 6B). In contrast, mDia2 knockdown efficiently blocked adipogenesis, but had little effect on Cyto D -induced osteogenesis (Fig. 6C, 6D). Thus, the majority of Arp2/3 complex recruitment to actin filaments may be mediated by mDia1. To assess why osteogenesis depends on mDia1, we performed cell fractionation analysis, which showed that in control cells, mDia1 is largely localized to the cytoplasm, while mDia2 is largely intranuclear (Fig. 6E). Furthermore, mDia1 was trafficked into nucleus in response to Cyto D treatment, while mDia2 remained within the nucleus (Fig. 6E, 6F). As such, the formin mDia1 leaves the cytoplasm in response to Cyto D. The latter suggests that although mDia2 may not play a role in Arp2/3 complex recruitment and activity, both formins work together to promote primary actin filament formation in the nucleus.

Figure 6.

Preventing end-on-end polymerization by silencing formin affects Cyto D-induced differentiation. ±Cyto D, ±CK666 (100 mM) analyzed by (A, C) reverse transcriptase polymerase chain reaction; (B, D) Alpl activity assay and Western blot. (E): Nuclear and cytoplasmic IB for ± Cyto D cells showing formins. (F): Densitometry from four experiments ±Cyto D stained for mDia1; *, p < .05; **, p < .01. Abbreviations: CTL, control; Cyto, cytosol; Cyto D, cytochalasin D; Nuc, nucleus; siCTL, control siRNA; simDia, siRNA to mDia1 (or 2).

Discussion

Nuclear actin has both global and specific effects on gene expression [41], regulating chromatin remodeling, as well as transcript elongation [42] such that actin depletion is associated with cell quiescence [18]. Our understanding of the role of polymerized actin in controlling gene expression has previously concentrated on actin networks outside the nucleus. For instance, the sequestration of the transcription factor MKL-1 (MAL) by actin monomers in the cytoplasm [28] removes suppression of the master adipocyte transcription factor PPARg [34]. Obversely, cytoplasmic F-actin promotes the nuclear localization of the transcriptional cofactor YAP [35]. Along with regulating the location of transcription factors, actin polymers within the cytoplasmic cytoskeleton control nuclear shape and may thus transmit forces which influence chromatin architecture, function, and interactions between distinct chromatin domains [17]. Our data support that the cytoskeleton is largely responsible for controlling nuclear shape, exemplified by the increase in nuclear height measured after disruption of the cytoskeleton via Cyto D or depletion of the formin mDia1. Our findings also conform to previous work showing that alteration in substrate stiffness, which alters cytoplasmic cytoskeletal structure, can modulate stem cell lineage, for example, a less stiff substrate leads to decreased actin structure and promotion of adipogenesis [4, 43]. The idea that cytoskeletal forces might contribute to nuclear actin structure arises after evidence demonstrating that cell spreading leads to nuclear actin polymerization [44]. Indeed, such actin filaments have been described within nuclei and could participate in organizing the chromatin machinery [26, 33]. Our data here indicate, for the first time, that nuclear actin structure is at least as important as cytoplasmic actin structures in the regulation of MSC cell fate (Supporting Information Fig. S3).

Nuclear actin structure, which is influenced by mechanical force [40], formin translocation [45] and alterations of LINC complex associations [14], might be expected to affect chromatin organization and determination of lineage. Our results indicate that once actin is intranuclear, polymerization is a pre-requisite for initiation of MSC differentiation. Importantly, while end-on-end actin fiber growth supports both osteogenesis and adipogenesis, branching of actin polymers is specifically necessary for osteogenic differentiation. This finding suggests that access of both Runx2 and Pparg to their respective cistromes is largely dependent on intranuclear actin structure. The inner nuclear membrane scaffold, which is composed of LaminA/C, may function as a suitable substrate for actin polymerization because it contains two actin binding domains and anchors other actin binding proteins (e.g., emerin) [46]. This leads to our speculation that to deploy the gene expression required for osteogenesis, formation of a branched actin network would depend on nucleation factors associated with the inner leaflet of the nuclear membrane. The subsequent branched actin network might then alter the localization or availability of chromatin that is bound by activated Runx2 and rearrange critical transcriptional machinery [47].

It has been widely accepted that reduced cytoskeletal structure leads to adipogenesis [2, 4, 5], but evidence for the obverse condition—that an increased cytoskeleton results in osteogenesis—is lacking. Our data indicate instead that the cellular location and secondary branching of actin polymers is a principal regulator of osteogenesis. The Arp2/3 complex is recruited to actin filaments by a membrane bound nucleation factor to allow development of a branched actin network [36]. Induction of such secondary branching has been shown to increase with application of compressive mechanical force [40] and external cues that induce lamellipodia [38]. We found that inhibition of Arp2/3 complex activity with the specific inhibitor CK666 entirely prevents osteogenesis. In the absence of Cyto D, CK666 inhibition of Arp2/3 leads to a small but significant increase in nuclear height, consistent with both decreased cytoskeletal tension [11] and adipogenesis. Cyto D causes the majority of cellular actin to traffic to the nucleus, while both formins and the Arp2/3 complex act together to build an intranuclear branched actin network. In the absence of this branched actin structure, both the Runx2 dependent and independent cistromes necessary for osteogenesis are silenced and impels MSC differentiation definitively toward an adipogenic phenotype.

Our data demonstrate that formin-dependent actin structure is critical for differentiation of multipotential MSCs. Inhibition of actin polymerization through depletion of either mDia2 (Diaph2), which is largely restricted to the nucleus in our MSC, or the cytoplasmic mDia1 that moves into the nucleus after Cyto D treatment, prevented adipogenesis. These results indicate that the level of internal nuclear polymerization is relevant for the adipogenic program. In contrast, while mDia1 silencing significantly blocked Cyto D-induced osteogenesis, loss of mDia2 did not affect this differentiation program. Both forms of mDia bind Rho-GTPase to unlatch the diaph autoregulatory domain thus opening the FH2 region to binding by membrane associated nucleation factors which recruit the Arp2/3 complex [7]. mDia2-dependent filament assembly and bundling, however, has been shown to be more subject to inhibition; for example, the formin inhibitory protein, DIP, appears to specifically inhibit mDia2 and not mDia1 [48]. mDia2 might thus be less able to participate in generation of primary filaments and subsequent Arp2/3 complex recruitment than mDia1, contributing to the differential effect of formin isoforms on osteogenesis and adipogenesis. Interestingly, in myocyte precursors, it is mDia2 rather than mDia1 that shuttles from cytoplasm to nucleus [45], indicating that cell-specific formin location and trafficking may also contribute to isoform specific effects.

The mechanisms by which nuclear actin structures regulate MSC differentiation remain speculative. Heterochromatin formation during differentiation may silence genes involved in alternative linages and genes required for maintaining stemness. Mechanical factors transferred via the cytoskeleton affect polycomb-mediated gene silencing [49], which is known to be involved in skeletal determination [25], and nuclear actin structure might differentially diffuse incoming forces. The mdMSC used in our study are known to be biased toward osteogenesis and requires a resetting of the heterochromatin landscape to allow expression of adipogenic genes [16]. It is likely that advancement of either osteoblastic or adipogenic gene expression requires rearrangements of chromatin associated with intranuclear actin structures. The “choice” of osteogenic versus adipogenic differentiation might further depend on competition of assembly factors within the nucleus [50].

Interestingly, a recent article proposed that Cyto D treatment inhibited the silencing of genes that induce differentiation in epithelial stem cells. Because increasing intranuclear actin via inhibition of exportin-6 also promoted differentiation of epithelial stem cells, the authors concluded that differentiation was dependent on the availability of G-actin to initiate RNA polymerase II [14]. Our work suggests the alternate interpretation that epithelial stem cell differentiation may be due to the presence of polymeric actin at specific promoters.

Conclusion

In conclusion, we find that once actin is transported into the nucleus, actin alters not only nuclear shape and size but also gene expression. We have demonstrated that actin filament formation has profound effects to initiate and enhance MSC differentiation and that branched and non-branched actin filaments confer alternate lineage speciation. While our data invoke actin transfer due to cytoplasmic actin polymer disruption, the ability of Arp2/3 complex inhibition to profoundly alter differentiation in the absence of Cyto D suggests that regulation of trafficking and activity of components of the actin tool box within the nucleus is critical to control of stem cell fates. The results of our study thus permit a reappreciation of the important roles of mechanotransduction and actin polymerization during development.

Significance Statement.

Actin structure within the nucleus plays a key role in directing stem cell differentiation. We show that formation of a branched actin network, which is dependent on the nuclear presence of the mDia1 formin and the Arp2/3 complex, is required for mesenchymal stem cells to progress into the osteoblast lineage. Inhibition of Arp2/3 complex results in unbranched actin polymers in the nucleus, which strongly promotes adipogenesis. These findings provide a fundamentally new paradigm for understanding how stem cells adopt distinct cell fates while also defining a physiological role for the intranuclear actin toolbox.