|Home | About | Journals | Submit | Contact Us | Français|
Vertebrate nonmuscle cells express two actin isoforms: cytoplasmic β- and γ-actin. Because of the presence and localized translation of β-actin at the leading edge, this isoform is generally accepted to specifically generate protrusive forces for cell migration. Recent evidence also implicates β-actin in gene regulation. Cell migration without β-actin has remained unstudied until recently and it is unclear whether other actin isoforms can compensate for this cytoplasmic function and/or for its nuclear role. Primary mouse embryonic fibroblasts lacking β-actin display compensatory expression of other actin isoforms. Consistent with this preservation of polymerization capacity, β-actin knockout cells have unchanged lamellipodial protrusion rates despite a severe migration defect. To solve this paradox we applied quantitative proteomics revealing a broad genetic reprogramming of β-actin knockout cells. This also explains why reintroducing β-actin in knockout cells does not restore the affected cell migration. Pathway analysis suggested increased Rho-ROCK signaling, consistent with observed phenotypic changes. We therefore developed and tested a model explaining the phenotypes in β-actin knockout cells based on increased Rho-ROCK signaling and increased TGFβ production resulting in increased adhesion and contractility in the knockout cells. Inhibiting ROCK or myosin restores migration of β-actin knockout cells indicating that other actins compensate for β-actin in this process. Consequently, isoactins act redundantly in providing propulsive forces for cell migration, but β-actin has a unique nuclear function, regulating expression on transcriptional and post-translational levels, thereby preventing myogenic differentiation.
Vertebrates express six highly conserved actin isoforms (1) in complex developmental and tissue-specific patterns (2). The major actin isoforms expressed in nonmuscle cells are β- and γ-cytoplasmic actin (further referred to as β- and γ-actin). Remarkably, in warm blooded vertebrates these isoforms differ only in four amino acids at the N terminus (1). The conserved nature of these substitutions can be interpreted in a scenario in which these isoforms perform redundant functions. Yet, spatial and temporal segregation of these isoforms in the cytoplasm has been observed (3), suggesting specific roles. γ-actin displays a more ubiquitous distribution, whereas β-actin is preferentially located at the leading edge of newly formed cellular compartments and protrusions (4–8). Given this preferred localization and its ubiquitous expression it is generally accepted that β-actin specifically functions in generating cell protrusion. Consistent with this view is that overexpression of β-actin increases cell speed by increasing areas of protrusion and retraction (4, 9–11). It is however unclear if other actin isoforms are equally capable of generating cell protrusion and productive cell migration.
More recently the presence of actin in the nucleus was recognized, and a role for actin in modulating transcription is increasingly appreciated (reviewed in (12, 13, 14)). Antibodies against β-actin block transcription (15) and nuclear translocation of β-actin is involved in macrophage differentiation (16). Whereas this univocally demonstrates that β-actin is involved in controlling gene transcription, it is unclear to what extent this occurs and if other actin isoforms can compensate for this nuclear function.
Genetic evidence suggests that β-actin is an essential gene. Three knock-out models are available (17–19) and in all cases whole-body knock-out results in embryonic lethality, albeit the stage where it happens is different (in one model after E8.5, (18) and in the two other ones at E10.5 (17, 19)). We exploited β-actin−/− (knock-out, KO1) mouse model from (17) to create a unique model system: mouse embryonic fibroblasts (MEFs) devoid of β-actin. This enabled us to address long term effects of lack of β-actin function on cellular properties. In line with (18) we find compensatory expression of other actin isoforms. This compensation allows cells to adhere to substrates, to make protrusions, and, when ROCK is inhibited, even to migrate, which indicates that other actin isoforms can act redundantly for the cytoplasmic functions of β-actin; i.e. providing force for cell protrusion. Increased expression of other actin isoforms, upon loss of β-actin, is accompanied by a larger change in the genetic program as evidenced by a differential proteome study. Pathway analysis suggested augmented contractility and TGFβ activation. This changed program, resulting from β-actin deletion, occurs despite the presence of other actins in the nucleus suggesting that the nuclear function of β-actin is more unique.
Creation of the heterozygous β-actin KO mice has been described (17). The β-actin knockin (KI) mice were created by recombinase mediated cassette exchange with a pCEHyg-H-ACTb insertion plasmid containing the human β-actin cDNA (see supplemental Fig. S1B). MEFs were derived from littermates in case of wild type (WT)10, KO1, KO2, and heterozygous (HET)8 cells. MEFs were prepared from individual 10.5-day-old embryos. Head and organs enriched in blood vessels were removed and tissue was minced and dissociated in 25 ml ice-cold 0.25% trypsin solution for about 12 h. The solution was warmed to 37 °C and then shaken for a few seconds. After carefully washing, the MEFs were propagated in Dulbecco's modified Eagle medium with high glucose, 10% fetal bovine serum, 870 mg/l glutamine, 0.1 mm beta-mercaptoethanol, and 100 units/ml penicillin and 100 μg/ml streptomycin. The cells were immortalized with the pSV51 plasmid expressing SV40 largeT-antigen (20) (http://www.belspo.be/bccm/lmbp.html; accession number LMBP1829) at passage 2. The same protocol was used to prepare fibroblasts from homozygous β-actin KI embryos. Embryonic stem (ES) cells were prepared from embryos at the morula stage from crosses of β-actin EGFP with β-actin Plap mice (17). To reduce β-actin expression we used shRNAmir V2MM-75091 (Open Biosystems) against the 3′UTR of β-actin encoding a short hairpin inserted in the retroviral vector pSM2c. RHS1704 was used as nonsilencing control. Five micrograms endotoxin-free plasmids were nucleofected (Amaxa kit MEF2) into 2 × 106 cells and selection with 0.4 μg/ml puromycin was maintained during 11 days. Two human β-actin constructs, one containing the coding sequence with only the ZIP-code (21) and with the full-length 3′UTR, were cloned in pMSCV-puro (Clontech). Virus was produced in Human Embryonic Kidney HEK293 cells and KO1 MEFs were transduced and selected with 0.4 μg/ml puromycine during 10 days. Cells were either used for WB or for migration experiments.
Antibodies used in this study are pan-actin mAb (clone C4), β-actin mAb (clone AC-15), α-smooth muscle actin (α-SMA) mAb (clone 1A4), vinculin mAb (hVIN1), tropomyosin mAb (clone TM311), ADF pAb (clone GV-13), α-tubulin mAb (clone T6199) and B23 mAb (clone FC82291) (all from Sigma), γ-actin pAb (Chemicon and generous gift from J. Ervasti, Department of Biochemistry, University of Minnesota), γ-smooth muscle actin (γ-SMA) mAb, phospho-LIM kinase 1, and 2 pAb (Abcam, Cambridge, MA), glyceraldehyde-3-phosphate dehydrogenase (GAPDH) pAb (US Biological, Swampscott, MA), MLC2 pAb, phospho-Ser19 MLC2 mAb, MYPT1 pAb, and phospho-Thr853-MYPT1 pAb, β-tubulin mAb (9F3) and cofilin, phospho-cofilin, LIM-kinase 1, and 2 (all from Cell Signaling), Filamin-1 pAb (Santa Cruz, Santa Cruz, CA), SM22 pAb and calponin pAb (generous gift from M. Gimona, Cytoskeleton Group, University of Salzburg) and LPP pAb (ImmunoGlobe). Secondary Antibodies for WB were goat anti mouse, anti rabbit or anti sheep antibodies coupled to IRDye 800 or IRDye 680 (Li-Cor).
Immunohistochemistry on whole mount embryos was performed as previously described (22). Briefly, embryos were fixed in methanol:DMSO (4:1) overnight at 4 °C, treated with methanol:DMSO:H2O2 (4:1:1) for 5–10 h at room temperature to block endogenous peroxidase activity and stored in methanol at −20 °C. The embryos were subsequently rehydrated in 50% methanol in phosphate buffered saline PBS (PBSMT) and incubated with the primary antibodies in PBSMT overnight at 4 °C. Following washes in PBSMT for 5 h at room temperature, embryos were incubated with an anti-mouse peroxidase-labeled secondary antibody (1/500, NIF825, Amersham Biosciences,Uppsala, Sweden) in PBSMT overnight at 4 °C. Following washes in PBSMT for 5 h at room temperature and brief washes with PBS with 0.1% Triton X-100, the embryos were developed with 3.3-diaminobenzidine tetrahydrochloride (Vector laboratories, Burlingame, CA). The reaction was stopped by fixing the embryos in 4% paraformaldehyde in PBS at room temperature for 1 h.
Total cell or embryonic lysates were prepared in 7 m urea, 2 m thio-urea, 0.5% TritonX-100, 40 mm dithiothreitol and protease inhibitors (1 μg/ml Leupeptin, 1 μg/ml Antipain, 1 μg/ml Aprotinin, 1.6 μg/ml Benzamidine). The amount of loaded protein was 6 μg for one-dimensional and 30 μg for two-dimensional (2D) polyacrylamide gels. For 2D gel electrophoresis the samples were first separated on Immobiline Drystrip gels (pH range 4–7, Amersham Biosciences) and subsequently separated according to their molecular weight in 10% polyacrylamide gels. The actin isoform expression patterns or the proteins indicated in the figures were analyzed by Western blotting (WB) using the appropriate set of primary and secondary antibodies. The protein bands/spots were quantified on an Odyssey scanner. ROCK inhibitor (Calbiochem, Merck KGaA, 10 μm) treated cells were lysed in the same buffer as above, except that 50 μm of ROCK inhibitor was added as well as 1 mm Na-orthovanadaat and 10 mm of NaF. Blots were probed with the antibodies indicated in the figures. Nuclear and cytoplasmic fractions were prepared using the nuclear extract kit from Active Motif.
Target specific primers (see supplemental Table S4) were designed using Lightcycler Probe Design Software 2.0 (Roche) and synthesized by Proligo (Sigma). Primer sets were validated in silico by N-blasting against the mouse non redundant nucleotide collection at NCBI. Total RNA was isolated from three different cell preparations for each cell line using RNeasy Midi (Qiagen, Dorking, Surrey, UK), followed by DNaseI treatment. cDNA was prepared with the Transcriptor First Strand cDNA Synthesis Kit (Roche). All qRT-PCR reactions were performed on a Lightcycler 480 (Roche) using Fast Start SYBR Green Master mix (Roche). The specificity of each amplification reaction and the absence of primer dimer formation were additionally verified via an evaluation of the melting curve of the amplified product and via gel electrophoresis. Amplification efficiencies for each primer set (target and reference genes) were determined on an equivalent mixture of WT, KO, HET, and KI MEFs. The absence of amplification from putative contaminating DNA in the RNA preparation was ensured using a control sample of RNA from which no cDNA was synthesized. qRT-PCR reactions on biological triplicates were performed in duplicate on a LightCycler 480 (Roche) using the Fast start SYBR Green master mix (Roche).
We normalized these values using a normalization factor calculated by qBASE software (23) (http://medgen.ugent.be/qbase/) that was based on qRT-PCR results of a set of reference genes analyzed for each sample. The combination of pgk1 (phosphoglycerate kinase 1), ubc (ubiquitin-C), s18, and gapdh as reference genes resulted in the lowest variation of reference gene relative quantities across the samples. The normalized Relative Transcript level was calculated.
The adhesion and contractility assays were performed as previously described (24, 25). Briefly, 104 cells were seeded in triplicate for 60 min into 96-well plates coated with collagen (10 μg/cm2). After removal of unattached cells, the adherent cells were fixed, stained with crystal violet and quantified by absorbance at 595 nm. Alternatively cells were allowed to adhere for three hours, fixed and processed for immune fluorescence. Transforming growth factor β (TGFβ) activity was measured using the mink lung epithelial cell (MLEC) luciferase assay (26). Briefly, 2000 to 3000 WT or KO1 MEFs were cultured on plastic. On day 4, 20,000 MLECs were added and luciferase activity was measured on day 5. Three independent experiments were done (two tetraplicates and one duplicate). The average fold of TGFβ activation per technical replicate was calculated and these were used to calculate the overall average fold activation of KO1 MEFs and standard error of the mean (S.E.). Real-time monitoring of cell adhesion using impedance technology was performed on an xCELLigence plate E device (Roche). The xCELLigence apparatus was run according to the manufacturer instructions. Wells were either not pretreated, coated with 10 μg/cm2 collagen in PBS, or coated with 6.25 μg/cm2 fibronectin in DMEM. Each well was seeded with 104 cells and this was done in triplicate for each condition. The impedance was measured continuously for 12 h and expressed as Cell Index (CI).
Immunofluorescence and phase contrast microscopy were done with an Olympus inverted XI71 or XI81 microscope. Images were recorded using a CCD digital camera (SPOT-RT monochrome; Progress Control) controlled by the AnalySIS docu software (Olympus) or on an Olympus-CellM system. Fluorescence images were recorded with a 60X LCPlanFL objective (N.A. = 0.70, Olympus). Cells were seeded on glass cover slips coated with 10 μg/cm2 collagen and fixed after 3 h or 24 h. For visualizing actin isoforms and vinculin with specific antibodies and for monitoring F-actin by phalloidin staining, established procedures using paraformaldehyde fixation and Triton-X100 permeabilization were followed. The primary antibodies used are described above. The secondary antibodies and the phalloidin antibodies are Alexa fluor conjugates (Molecular probes). For ROCK or blebbistatin (Calbiochem) treated samples the cells were first allowed to adhere 5 h and then incubated with 10 μm of inhibitor and fixed after 21 h. Phase contrast images were recorded with a 10 × UPlanFL objective (N.A. = 0.30, Olympus).
Quantitative videomicroscopy for the determination of random 2D cell migration on collagen for 4 h was done as described (27) (for a detailed description of hull area in μm2 see Supplemental Material and Methods and Fig. 4 in reference (28)). For ROCK or blebbistatin treatment we used 10 μm of inhibitor. Kymographs were recorded with a 60x OiPH-UPLFLN objective (N.A. 1.25, Olympus). Images were taken every 10 s for 10 min and analyzed with the Kymograph Plugin for ImageJ (http://rsb.info.nih.gov/ij and http://www.embl.de/eamnet). For scratch wound assays we plated 5.104 cells per well of a 48-well dish 24 h before making uniform wounds with the Cell Scratcher (Peira Scientific Instruments). Wound closure was imaged every 15 min for 24 h. The images were analyzed using an algorithm described in (29).
COFRADIC proteomics was done as described previously (30, 31). β-actin WT and β-actin KO cells were grown in culture medium supplemented with 12C, 14N- or 13C, 15N-methionine (Cambridge Isotope Laboratories, Andover, MA) for at least six doubling times (32). Cells were collected and flash frozen in liquid nitrogen. Cell pellets containing 2–4 106 cells were lysed in 0.7% 3-[(3-cholamidopropyl)dimethylammonio]propanesulfonate, 50 mm HEPES, pH 7.4, 100 mm NaCl, complete protease inhibitor mixture tablet (ROCHE). Lysates were cleared by centrifugation. 250 μg of total protein was processed further. Following trypsin digestion, methionyl peptides were isolated by combined fractional diagonal chromatography (COFRADIC) as described previously (33). Such isolated peptides were sampled by LC-MS/MS using an Ultimate 3000 HPLC system (Dionex, Amsterdam, The Netherlands) in-line connected to a LTQ Orbitrap XL mass spectrometer (Thermo Electron, Bremen, Germany). Peptides were first trapped on a trapping column (PepMap™ C18 column, 0.3 mm internal diameter (I.D.) x 5 mm length (Dionex)) and following back-flushing from the trapping column, the sample was loaded on a 75 μm I.D. × 150 mm length reverse-phase column (PepMap™ C18, Dionex). Peptides were eluted with a linear gradient of 1.8% solvent B' (0.05% formic acid in water/acetonitrile (2/8, v/v)) increase per minute at a constant flow rate of 300 nl/min.
The mass spectrometer was operated in data-dependent mode, automatically switching between MS and MS/MS acquisition for the six most abundant ion peaks per MS spectrum. Full scan MS spectra were acquired at a target value of 1E6 with a resolution of 60,000. The six most intense ions were then isolated for fragmentation in the linear ion trap. In the LTQ, MS/MS scans were recorded in centroid mode at a target value of 5000 ion counts. Peptides were fragmented after filling the ion trap with a maximum ion time of 10 ms and a maximum of 1E4 ion counts. From the MS/MS data in each LC run, Mascot generic files (mgf) were created using the Mascot Distiller software (version 184.108.40.206, Matrix Science) and were searched against the SwissProt database (v14.07) restricted to Mus musculus taxonomy (15813 entries) using the Mascot search engine (version 2.2.1). The following MASCOT parameters were set: the protease setting was trypsin (cleavage between Lys or Arg and Pro was tolerated) with a maximum of one allowed missed cleavage, whereas acetylation of a protein's N terminus, deamidation of Asn and Gln, formation of carbamidomethyl Cys, oxidation of carbamidomethyl Cys and pyroglutamate were considered as possible modifications, and tolerances for the precursor ion mass and fragment ions masses were set to ± 10 ppm and 0.5 Da respectively. Determination of the light (12C, 14N-Met) and heavy (13C, 15N-Met) labeled sulfoxide methionyl peptides for further quantification was established by using the quantitation option in Mascot. Finally, peptide hits of which the MASCOT ion score of the MS/MS spectrum exceeded MASCOT's identity threshold score set at 99% confidence and which were ranked one were withheld and considered identified. A false discovery rate of 0.25% was determined by the method described by Käll et al. (34). The total list of identified peptides is available via the PRIDE data repository (http://www.ebi.ac.uk/pride/; PRIDE experiment accession number: 18462–18463 (35) and includes the percent coverage and the number of different peptides per protein (see also Suppl_Table_PRIDE_18462 and Suppl_Table_PRIDE_18463). Quantification was performed using Mascot distiller software (version 220.127.116.11, Matrix Science). Ratios are calculated from a least square fit of the area below the light and heavy isotopic envelope situated in the elution peak of the precursor determined by the Distiller software (XIC threshold 0.3, XIC smooth 1, Max XIC width 250). To validate the calculated ratio, the standard error on the least square fit has to be below 0.16 and correlation coefficient of the isotopic envelope should be above 0.97. All data management was done by ms_lims (36). Only these proteins which were quantified by at least 2 different peptides, were used for quantification (supplemental Table S2C). If more than one MS/MS spectrum was linked to a peptide, the mean of all its calculated ratios was considered as the ratio value for that peptide. The average ratio values of different peptides representative for a protein were used as the ratio value for altered expression of that protein. Robust statistics (37) was applied to the base-2 logarithm of the ratios to identify proteins that significantly deviate from the median ratio. For both conditions WT/KO1 and WT/KO2, a skewed normal distribution with Huber scale of 0.68 and 0.88 on a 95% confidence level respectively and a median of respectively non log 1.112 and 1.105 (supplemental Table S2C). For pathway analysis (see below) we only considered the intersection of proteins that were deregulated in both KO cell lines. Because these are derived from littermates the intersection can be considered as being derived from two biological replicates. Expression levels of selected proteins were validated by standard qRT-PCR or WB or both (see above and supplemental Table S5 for a guide to the respective figures).
To identify regulated pathways, robust statistics (37) was applied. The validated quantitative proteomics data were subdivided in three. The first subdivision contained the proteins that were up-regulated in the heavy configuration (L/H < 0.5, in β-actin KO MEFs). The second subdivision holds the proteins that were down-regulated in the heavy configuration (L/H > 2). The last subdivision contains the rest of the proteins and were considered as unchanged protein levels (0.5 < L/H < 2). The p value for every pathway found in the different parts was calculated via the hyper geometric test, using the proteome size (the number of proteins in the proteome that are linked to a KEGG pathway), the sample size (the number of proteins in the subdivision that are linked to a pathway), the in-proteome pathway links (the number of proteins that are linked to a specific pathway), and the in-sample pathway links (the number of proteins in the subdivision that are linked to the same specific pathway). A pathway was considered regulated if the calculated p value for one subdivision was larger than 0.975 or smaller than 0.025. Only KEGG pathways that were regulated in both KO cell lines derived from the first subdivision (condition L/H <0.5) are used in the creation of the Fig. 3C. Overrepresented and underrepresented pathways are respectively colored in a shade of red or green reflecting the calculated p value (see scale).
The IPA reference set was Ingenuity Knowledge Base (Genes Only) in version 8.0. Direct and indirect relationships were considered from all data Sources, all species (human, mouse, rat) and all tissues and cell lines, and all molecules and or relationships were considered as filter. Threshold was 0.05 and Fisher's exact test p value was used as a scoring method. The following data sets were analyzed: the intersection of the identified proteins with significantly altered expression (twofold up or down) in both β-actin KO MEFs relative to WT, the union of these proteins of both cell lines. The union of quantified proteins in both KO cell lines with unchanged expression relative to WT was used as a reference set (supplemental Table S2A).
Homozygous β-actin KO mice are embryonic lethal (17–19) indicating that β-actin is an essential gene. In all three models embryos survive past the gastrulation stage suggesting that compensatory expression of other actin isoforms may rescue at earlier stages of development. Consistent with this we observed a switch to γ-actin expression in embryonic stem (ES) cells lacking β-actin whereas in WT ES cells β-actin is the only isoform expressed (Fig. 1A). We currently cannot explain the variation in stage of embryonic lethality between the models but the observation that in the β-actin KO model described in (17) embryos can be retrieved at E10.5 allowed us to evidence actin isoform switching at this stage in mouse. In whole lysates of E10.5 β-actin+/+ (WT) embryos both β- and γ-actin are present as is a minute amount of α-smooth muscle actin (α-SMA). In contrast in E10.5 β-actin KO embryos the expression levels of both α- and γ-actin isoforms are strongly increased (Fig. 1B). This actin isoform switch was further documented by whole-mount immunohistochemistry on E10.5 embryos. We show ubiquitous expression of β-actin in a WT embryo and its complete lack in a β-actin KO littermate (Fig. 1C). γ-actin, which is also ubiquitous in WT, is increased in the β-actin KO embryos (Fig. 1D). α-SMA expression in WT embryos is restricted to the heart and the myotome cells of the somites (2, 38). In strong contrast in the β-actin KO embryo we observed ectopic expression of α-SMA in the entire embryo (Fig. 1E). Similar iso-actin changes were confirmed in two β-actin KO MEF cell lines isolated from E10.5 KO littermates: further referred to as KO1 and KO2 (Fig. 1F). In addition to α-SMA and γ-actin, these KO cell lines, display increased expression of γ-smooth muscle actin (γ-SMA). Heterozygous β-actin MEFs (HET8) have near WT levels of β-actin, and have moderately increased expression of α-SMA. Remarkably, the total actin levels appear very similar in the different MEF lines. Isoform specific quantitative PCR (qRT-PCR) confirmed this isoactin switch at the transcript level (supplemental Fig. S1A). We conclude that β-actin is an essential gene for development at the level of the organism, but appears dispensable for cell survival. Deletion of the β-actin gene turns on a rescue program by inducing the expression of other actin isoforms. Using recombinase mediated cassette exchange (17) we created a β-actin knock-in (KI) (β-actinhuman β-actin/human β-actin) mouse expressing the human β-actin cDNA (supplemental Fig. S1B). In this viable mouse, expression of β-actin and α-SMA have been restored to WT levels, whereas the γ-actin level remains higher than in the WT embryos (Figs. 1C–1E). MEFs from β-actin KI express WT levels of β- and γ-actin and α- and γ-SMA (Fig. 1F, supplemental Fig. S1A). Additionally, we silenced β-actin in WT MEFs. As observed by others silencing β-actin expression is difficult (3). Despite only a moderate reduction (~20%) of the amount β-actin, a clear increase in α-SMA actin expression is visible (Fig. 1G), corroborating a role for β-actin in regulating the expression of actin isoforms.
It is generally accepted that polymerization of β-actin generates the protrusive force for cell migration. We therefore tested the impact of β-actin deletion on random 2D cell migration. Unexpectedly, we still observed 2D random migration for a significant portion of KO MEFs (10% of KO1 and 40% of KO2 compared with nearly 80% of WT and 90% of KI) (Fig. 2A and supplemental Movies S1–S5). This finding indicates that lack of β-actin impairs but does not completely abolish cell migration. Kymograph analysis of lamellipodial activities (see supplemental Movies S6–S8) revealed 62–70% reduced protrusion lengths in β-actin KO MEFs, and a decrease in persistence of their protrusions by 50% (Figs. 2B–2D and supplemental Table S1). However the number of formed protrusions and retractions is similar to WT. Remarkably, the protrusion rates measured for β-actin KO MEFs are similar to those of WT cells (Fig. 2E, supplemental Table S1) suggesting that continuous actin polymerization close to the membrane still occurs. This is further supported by two observations: the protrusive phase of cell spreading during adhesion occurs faster in β-actin KO cells and protrusions that are formed contain filamentous actin and α-SMA (supplemental Fig. S2A and data not shown). Consistently, adherent KO cells display increased localization of other actin isoforms at the leading edge of protrusions in β-actin KO cells, in addition to incorporation in stress fibers (supplemental Fig. S2B). We also investigated whether α-SMA and γ-actin were present in nuclear fractions. Both isoforms where present in the nucleus in β-actin KO MEFs (supplemental Fig. S2C).
We attempted rescuing migration by stably reintroducing β-actin in KO1 MEFs which had the more severe migration phenotype (Fig. 2A). Two different constructs were used: the coding sequence with the ZIP-code (ZIP), which is important for proper localization of the β-actin mRNA (21), and the coding sequence with the full length 3′ untranslated region (UTR). Despite the fact that β-actin was expressed to levels similar as in WT MEFs (Fig. 2G), none of these constructs was capable of fully restoring cell migration because the transfected KO1 cells migrated shorter distances than the WT MEFs (evidenced by a reduced hull area Fig. 2F). This indicates that simple reintroduction of β-actin is not sufficient to alleviate the migration defect. Remarkably, the expression level of α-SMA did not decrease after re-expression of β-actin (Fig. 2G), suggesting that the observed isoform switching in KO cells is the result of long-term adaption or a unidirectional genetic reprogramming (at least toward actin isoform switching).
Given the severe migration defect, despite normal lamellipodial protrusion rates (Fig. 2E, supplemental Table S1) and the increased occurrence of other actin isoforms at the leading edge (supplemental Fig. S2B), and in view of the notion that the migration defect of the KO cells cannot be restored by exogenous β-actin (Fig. 2F), it appeared unlikely that solely actin isoform switching caused the migration defective phenotype. We therefore investigated whether the β-actin KO MEFs underwent more profound reprogramming. To obtain more global views on this, we applied a differential mass spectrometry approach combining stable isotope labeling with amino acids in cell culture (SILAC) and combined fractional diagonal chromatography (COFRADIC) (33, 39).
We compared protein expression levels between WT and KO cells and found that approximately one fourth of the 2308 bona fide quantified proteins (out of 3630 identified proteins) exhibited altered expression (supplemental Fig. S3A–S3C). This substantiates that deletion of β-actin has a major impact on overall gene expression in cells. This was further confirmed by micro-array experiments (Lambrechts et al., unpublished). In KO1 and KO2 MEFs, respectively, 419 and 398 proteins, were at least twofold up- or down-regulated (supplemental Fig. S3A–S3C, supplemental Table S2A). Of the 661 identified proteins with altered expression in both cell lines (defined as the union), 156 were in common (defined as the intersection, supplemental Fig. S3C, supplemental Table S2B). We applied Ingenuity Pathway analysis (IPA) (http://www.ingenuity.com), Gene Ontology (GO) and KEGG pathway analysis (http://david.abcc.ncifcrf.gov), and Software Tool for Researching Annotations of Proteins (STRAP) (40) (http://www.bumc.bu.edu/cardiovascularproteomics/strap) to the intersection and/or the union.
IPA identifies overrepresented networks, canonical pathways or functions (Figs. 3A, ,33B). Both the intersection (all 156 proteins mapped in IPA) and the union (658 of the 661 identified proteins mapped in IPA) were analyzed. Additionally, the union of quantified proteins in both KO cell lines with unchanged expression relative to WT (1499 of the 1509 identified proteins mapped in IPA) was used as reference set (supplemental Table S2A). The results for the intersection and union of proteins with changed expression were comparable with respect to identified physiological system and development functions (supplemental Fig. S3D) and canonical pathways (supplemental Fig. S3E). This suggests that β-actin KO MEFs, although having partly different expression profiles and not being phenotypically identical, underwent relatively similar differentiation fates. Moreover, the top ranking functions and canonical pathways generated by the intersection protein set, were distinct from these of the reference set with unaltered expression (supplemental Fig. S3F–S3G). We further only considered the proteins with changed expression common to both cell lines (i.e. the intersection set) because these most likely present proteins whose expression is directly or indirectly controlled by ablation of β-actin. The highest ranked IPA network was enriched for actin cytoskeleton associated proteins (supplemental Table S3A). The highest ranking physiological system and development function in IPA was, not unexpectedly, “Connective Tissue Development and Function” (supplemental Table S3B, and 30 deregulated proteins are linked with connective tissue disorders, supplemental Table S3C) but also “Skeletal and Muscular System Development and Function” (Fig. 3A, supplemental Fig. S3D, supplemental Table S3B, see also below). The highest scoring canonical pathway was Rho signaling (see also below, Fig. 3A and Fig. 4, supplemental Fig. S3E, supplemental Table S3D). Identified groups of cellular molecular functions included cell cycle (note the altered expression of 39 proteins associated with cell division, a.o. several septins) and cellular assembly and organization and cell morphology (supplemental Table S3E).
In a separate analysis of general molecular function within GO and KEGG pathways, actin cytoskeleton associated pathways are overrepresented, whereas metabolic pathways are underrepresented (Fig. 3C). The results provided by STRAP also indicated that cytoskeletal proteins were enriched in the intersection dataset (Fig. 3D). These analyses are in line with our findings that the β-actin deletion has major impact on cell migration (Fig. 2), and as evidenced further below, on cell morphology and cell adhesion and contractility (Fig. 5 and supplemental Fig. S2A).
IPA identified Rho signaling as the highest ranked canonical pathway (Fig. 3B, supplemental Fig. S3E). Rho is a key regulator of cell migration and stress fiber contraction in addition to its implication in other biological processes such as cytoskeletal reorganization (with increased expression of FAK), actin polymerization, stress fiber contraction and cytokinesis (Fig. 4) (41, 42). Several upstream regulators and downstream targets that were identified in the proteomics experiment display altered expression in the KO MEFs (Fig. 4). Additionally we probed expression levels of known RhoA regulators that were not recovered from the proteomics experiment. RhoA and ROCK 1 and 2 protein expression levels are similar between WT and KO cells (Supplemental Fig. 4A) and RhoA, B, and C mRNA expression are not increased (Supplemental Fig. 4B). However, several nucleotide exchange factors that render Rho in an active state have increased expression (RAPGEF2, ARHGEF2/GEFH1, and ARHGEF10). In contrast, the expression levels of negative regulators of Rho-ROCK activity (43–45): Rho GDI 1 and 2, RhoU/Wrch1, Rnd2 (KO1, KO2) and Rnd1 (KO1) or Rnd3 (KO2) are significantly reduced (Supplemental Fig. 4A,B). Interestingly, two other upstream activators of Rho-ROCK (in different signaling pathways) are also up-regulated in KO cells: p120 catenin and serum deprivation response protein (SDRP) (Fig. 4) (supplemental Table 2B). N-cadherin-dependent RhoA activation is perturbed by p120 catenin knockdown (46) and SDPR interacts with MURC which increases Rho-ROCK dependent expression of atrial natriuretic peptide (47). Also the ROCK downstream targets, Myosin regulatory light chain 2, smooth muscle isoform (Myl9): MLC2 and Myosin phosphatase target subunit 1: MYPT1 (Fig. 5A, supplemental Table 2B), that regulate actomyosin contraction, are significantly overexpressed which parallels observations in contractile fibroblasts (48). These expression data are thus most consistent with increased Rho-ROCK activity.
Rho-ROCK activity is positively correlated with stress fiber formation and with focal adhesion maturation. Therefore it was not surprising that upon visible inspection KO MEFs had increased stress fiber and focal adhesion content (Fig. 5B, Supplemental Fig. 2B, also see refs 21–24 and 29–31 in (49)). A similar observation was made in (18). We therefore investigated cell spreading and adhesion more quantitatively. β-actin KO MEFs have more and larger (supermature) focal adhesions (FA) than WT cells (Fig. 5B-C) and adhere faster and stronger to different substrates (Fig. 5D-E, supplemental Fig. S5A-C). In addition KO1 MEFs gain a larger surface area (supplemental Fig. S5D). The combination of increased stress fibers and focal adhesions allows increased contractility downstream of Rho-ROCK (see below for a molecular explanation). We therefore tested contractility of KO1 MEFs in stress-relaxed collagen gel. Indeed, their contractility index was doubled compared with WT cells (Fig. 5F).
From these expression values and the observed phenotypes we developed a molecular model (Fig. 6) that can be tested. The model first implicates increased Rho-ROCK activity and predicts that ROCK targets should display augmented phosphorylation. This post-translational response can be monitored by the phosphorylation status of known ROCK effectors or of their downstream targets (50, 51). Indeed, we observed increased phosphorylation of LIM kinases 1 and 2 in the β-actin KO1 and KO2 cells compared with the WT10 cells (supplemental Fig. S6A). The phosphorylated/activated LIM kinases in turn phosphorylate and inactivate cofilin and actin-depolymerising factor ADF (supplemental Fig. S6B). Phosphorylation of these actin dynamizing proteins (52) results in decreased actin filament turnover. Additionally increased cofilin phosphorylation is correlated with augmented stress fiber formation in fibroblasts (49) as observed here (Fig. 5B). Consistent with our data, silencing of β-actin mRNA in HeLa cells was shown to result in enhanced phosphorylation of cofilin and ADF (53).
Two other ROCK targets: MLC2, MYPT1 also display increased phosphorylation (Figs. 5A, ,77A, upper panel). Phosphorylated MLC2 and MYPT1 oppositely regulate myosin activity. Phosphorylation of MLC2 directly activates actomyosin contraction whereas phosphorylation of MYPT1 inhibits the phosphatase activity that dephosphorylates and thus inactivates myosin. The increase in phosphorylation in both proteins in the KO MEFs explains the increased actomyosin contractility we observed (Fig. 5F).
Increased contractility of (myo)fibroblasts via Rho-ROCK requires and results in sustained TGFβ1 production (54), as is also the case for cells developing to a smooth muscle phenotype (55, 49) (see also network 4 cellular development, cellular growth and proliferation, connective tissue development, supplemental Table S3A). Therefore, our molecular model secondly predicts increased TGFβ production. We measured TGFβ1 activity with a reporter cell co-culture assay employing the β-actin KO1 cells. Indeed these produce more active TGFβ1 compared with WT cells (Fig. 7B). TGFβ1 plays a predominant role in the development of the connective tissue and muscular systems (identified in IPA-analysis above Fig. 3A, supplemental Fig. S3D) as well as during differentiation of myofibroblasts and smooth muscle cells. The observed transcriptional response in β-actin KO cells is consistent with this. A group of highly up-regulated proteins includes α- and γ-SMA, various other myofibroblast or smooth muscle marker proteins such as SM22/transgelin, SM calponin-1, tropomyosin-2, LPP, palladin, filamin-A, and other actin filament associated or stabilizing proteins (supplemental Table S3B). We validated the mass spectrometry results of smooth muscle marker proteins with Western blots and/or qRT-PCR and consistently found increased expression of these F-actin binding proteins in the β-actin KO cells (Fig. 7C and supplemental Table S5, supplemental Fig. S6D) (24, 56). Some of these (α-SMA, γ-SMA, transgelin, SM calponin 1) are known to be induced by TGFβ1, as are other proteins with increased expression we identified in the proteomics experiment: α-skeletal muscle actin, collagen1A1, β-catenin, fibronectin 1, heme oxygenase 1, tumor protein 53, insulin-like growth factor binding protein 3, GEF-H1 (which regulates α-SMA expression induced by TGF-β1 (57)), and TGFβ1 induced transcript 1 protein (also known as Hic5/ARA55) (supplemental Table S2b). Focal adhesion protein Hic5 (~6.5-fold increased in the KO1 cells) is a known positive regulator of a TGFβ1 autocrine loop (58). Forced expression of Hic5 led to the formation of ROCK-dependent actin stress fibers suggesting that this protein also increases Rho-ROCK activity (59). We conclude from these data that the observed altered phenotypes of β-actin KO cells are mediated by changes of both protein expression and altered posttranslational actin cytoskeletal activity regulation, in part downstream of TGFβ1 i.e. via Rho-ROCK (Fig. 6).
Given the prominent role of Rho-ROCK signaling on the organization of the actin cytoskeleton in the β-actin KO1 MEFs, we investigated if we could restore random cell migration by treating the cells with ROCK inhibitor Y27632 in order to relieve the effect of ROCK on stress fibers. β-actin KO cells treated with 10 μm Y27632 lead to strongly decreased phosphorylation of MYPT1 (Fig. 7A, lower panel) and have less stress fibers (Fig. 8A) although the expression level of α-SMA is not affected (supplemental Fig. S7A). In addition KO MEFs gained a migratory phenotype as clearly evidenced by an increase in hull area and an increased number of migrating cell (supplemental Movies S9, S10) (Figs. 8B, ,88C). Similar results were obtained with the myosin II inhibitor blebbistatin. Interestingly, blebbistatin treated KO MEFs migrated even further than untreated WT cells (Fig. 8B), univocally showing that other actin isoforms can generate protrusive force for productive cell migration. This was further corroborated by wound healing assays showing that the KO MEFs close wounds in a similar time span than WT cells (supplemental Fig. 7B) and thus, in this type of migration assay, are capable to migrate employing other actin isoforms even in the absence of inhibitors. Collectively our data establish that the random migration defect of untreated β-actin KO cells is because of the reprogramming of the cells, resulting in sustained TGFβ1 signaling and altered ROCK signaling and not because of lack of β-actin polymerization capacity.
We here described the effects of genetic deletion of β-actin on fibroblast fate. With respect to the traditional role of β-actin we find that the KO MEFs display altered cell morphology and strongly impaired migration. These, but not other, phenotypes are very similar to the ones described for a different β-actin KO cell model that was published prior to submission of the present manuscript. Similar to our initial observation MEFs without β-actin generated in (18) had impaired migration and increased expression of α-SMA and γ-cytoplasmic actin. These main phenotypes are thus independent from the different approaches taken but other aspects of MEF behavior appear differently affected. Bunnell and coworkers also describe deleterious effects on cell survival, a zero growth rate mainly caused by a high rate of apoptosis and multinucleated cells. Moreover, they see a down-regulation of cell cycle regulators. We observe the opposite in the MEFs used in our study. Cell cycle regulators appear to be up-regulated (supplemental Table S3E) and no deleterious effects are seen on cell survival. In addition, the MEFs in (18) also show different kinetics of lamellipodial protrusion and retraction. We isolated MEFs directly from β-actin KO embryos whereas Bunnell and coworkers prepared MEFs from WT embryos and excised the β-actin gene by tamoxifen induced Cre expression. Apparently, this difference in approach has consequences for the outcome of some of the observed phenotypes. One explanation is that the MEFs used in the present study had a longer adaptation period in the embryonic environment prior to their isolation at E10.5) for the absence of beta-actin. One is therefore looking at two types of findings: acute findings in the manuscript of Bunnell and coworkers and more chronic findings in this study. In view of this disparity between possible acute and chronic effects, a detailed comparison between the directed PCR array approach taken by Bunnell and coworkers (18) and our proteomics approach is premature. Importantly, we add to the findings of Bunnell by showing that, isoform switching also occurs in the KO embryos, that α-SMA expression increases upon silencing of β-actin, and that reintroducing β-actin does not alleviate the impaired migration nor reduces α-SMA levels. More importantly and surprisingly, when the MEFs are brought under appropriate conditions cell migration can be restored.
Our observation of actin isoform compensation is in accordance with these on muscle cell development and on other actin isoform KO mice (18, 60–63). Deletion of one of the α-muscle actin isoforms consistently resulted in the up-regulation of one or more of the other muscle α-actins. In the γ-actin KO mouse (63) a limited increase of β-actin expression occurs, analogous to the twofold increase of γ-actin that we observed in the β-actin KO embryos (Fig. 1). In contrast to these previous studies we find an inverse relation between the expression levels of a cytoplasmic form (β-actin) and a muscle isoform (α-SMA). This may parallel events occurring during embryonic muscle differentiation. During (skeletal) muscle development myoblasts repress β-actin and transiently express α-SMA (38, 64) prior to sarcomeric actin expression (65). This switch to α-SMA expression is also reminiscent of fibroblast to myofibroblast transdifferentiation. Such reprogramming, including increased expression of several smooth muscle cell markers, results in a.o. increased contractility and cell substrate adhesion, and in an arrest of cell migration. Notably sustained TGFβ1 activity (Fig. 7B) in the β-actin KO cells, likely produced by a contractility dependent positive autocrine feedback loop (58), and augmented Rho-ROCK activation (Fig. 5 and and7,7, supplemental Fig. S6), results in these cytokeletal phenotypes. These signaling pathways have previously been shown to be interconnected and to function both in differentiation toward myofibroblasts or smooth muscle cells and in cytoskeletal rearrangements (for references see review (66)). The arrest in cell migration that we initially observed is thus an indirect effect caused by the reprogramming of the KO cells, also resulting in altered posttranslational signaling, at least in part mediated by the TGFβ and Rho-ROCK signaling pathways.
It is intriguing that the total levels of actin are similar over the various cell lines used (Fig. 1F, Fig. 2C in (18)), which is as yet unexplained. Apparently the cells are capable of monitoring the total amount of actin and transmit a feedback signaling loop to other actin genes. In view of the observation that the G/F ratio's in the β-actin excised MEFs are decreased (18), it is tempting to speculate that the myocardin-related transcription factor A (MRTF-A/MAL), which acts downstream of TGFβ and the serum response factor (67, 68), is part of the sensor relay system which is acting here because expression of α-SMA (and the other smooth muscle markers) was proposed to be downstream of G-actin MRTF-A (69). This assumes isoform specificity of the MRTF-actin interaction and this has to our knowledge not been reported. The picture is likely to be more complex because only 17 of the proteins with changed expression identified here, are also found deregulated in a micro-array study on gene expression downstream of the G-actin MRTF-A interaction (69). Whereas this may somehow emphasize the importance of this signaling pathway it is certainly not the only pathway being deregulated as the majority of the up-regulated proteins are encoded by genes not known to be under control of the MRTF-family. Only ~10% of the proteins in the intersection (i.e. these with common altered expression) of the β-actin KO cells have CArG boxes in their promoter sequences that are recognized by MRTF-A (70 and Table I in (12)). Thus in view of the wider reprogramming, other signaling factors must be involved. We propose that combinatorial alterations, in particular of transcription factors or trans-acting factors and/or nuclear actin or actin binding proteins (see below), cause the induction of the smooth muscle phenotype in response to genetic deletion of β-actin.
Our results on gene reprogramming and cell behavior have important implications for understanding β-actin functions. Although β-actin partakes in cell migration (4, 9–11) and deleting it results in impaired migration (Fig. 2, and Fig. 4 in (18)) we stress that motile phenomena still occur in β-actin MEFs. Dynamic protrusions, with F-actin at the leading edge (Fig. 2) and cell adhesion (Fig. 5D and supplemental Fig. S5A, S5B), a process requiring actin polymerization for the phase of continuous protrusion (71), still occurs in untreated β-actin KO cells (supplemental Fig. S2A). Importantly, if increased contractility or ROCK activity is inhibited the cells regain migratory capability (Fig. 8) and KO cells migrate in wound healing assays (supplemental Fig. S7B). Therefore it appears that the β-actin KO cells have sufficient polymerization capacity provided by the other actin isoforms and the block of productive translocation (Fig. 2) results from altered signaling, likely including the ROCK substrates LIM kinases that affect the activity status of the actin dynamizing proteins from the ADF/cofilin family (72) in conjunction with the increased adhesion and stress fiber formation (Fig. 5 and supplemental Figs. S2B and S5). Collectively this does not preclude a fundamental role for β-actin in cell migration, it is indicating that this is not a unique role.
Our proteomics results show that lack of β-actin signaling to the transcription machinery not only transports to actin genes but leads to substantial genetic reprogramming in the β-actin KO MEFs. A glimpse of the onset of this reprogramming was also observed in (18) by a directed approach. This implies that β-actin exerts an important and regulatory nuclear function, i.e. for instructing gene expression. This is consistent with previous reported roles for nuclear actin: interaction with RNA-polymerases or other components of the basal transcription machinery (15, 73, 74), reviewed in (12, 14)), chromatin remodeling or establishing long-range chromatin organization (reviewed in (13)).
Increased nuclear β-actin accumulation was shown to be involved in regulating transcription during PMA induced macrophage differentiation (16). More recent papers establish roles for actin in a nuclear network with the estrogen receptor or in Oct4/Pou5f1 transcriptional reactivation (75, 76). Experiments in these studies were carried out with anti β-actin antibodies or β-actin mutants, and therefore strongly suggests this isoform is responsible for the observed transcriptional regulation. Whereas in all studies the presence of nuclear β-actin modulates transcription of (reporter) genes, we here have a converse example where depletion of β-actin alters the transcriptional program, in this case toward a myogenic program. Thus it appears that β-actin may have both activating as well as repressing gene regulatory activities. Although the latter remains to be proven we note that more proteins appear to be up-regulated by the absence of β-actin than down-regulated. In the three studies other actin isoforms (i.e. the highly similar γ-actin) were not investigated so it is unclear if β-actin is the sole actin isoform responsible for the observed transcriptional response. We cannot exclude that the other actin isoforms may have beneficial effects on expression of the up-regulated genes because in our study the pool of β-actin is replaced by mainly γ-actin and α-SMA and they are present in nuclear fractions (supplemental Fig. S2C). Given the many nuclear processes that actin as a protein is involved (see above), further research is needed to elucidate how actin isoform specific gene regulation is brought about.
Added to isoform specificity, there may be another layer of complexity to the transcriptional response. Recent studies indicate that both monomeric and polymeric forms of actin participate in gene regulation and that the presence of specific actin binding proteins in the polymerase complexes may modulate this state (75–78). We observed quite a number of (smooth muscle) actin binding proteins with changed expression levels or activity (because of altered ROCK signaling). Unbalancing actin binding protein activities by transcriptional or posttranslational events in the β-actin KO MEFs has clear consequences on the cytoskeleton. Such unbalanced activity may, however, also occur in the nucleus. For instance it was shown that cofilin was present in complex with polymerase II-associated actin and that its gene silencing leads to a transcriptional block (77). We observe increased phosphorylation and thus inactivation of cofilin activity. In addition to the observed cytoplasmic phenotypes it could also affect transcription because phosphorylation of cofilin should preclude its binding to actin in the polymerase II complex. Evidently, sorting out this complexity (the presence or not of actin binding proteins in the nucleus and their activity status in conjunction with how this influences activity of monomeric or polymeric β-actin in transcription) will require dedicated research, in which the model system we present here may be of use.
In summary, we show that the function of β-actin in providing the propulsive force is redundant because other actin isoforms are also capable of driving cell migration. Rather the role of β-actin in regulating cell homeostasis appears more unique. This is an isoform specific function because the observed increased expression of other actin isoforms, in particular of γ-actin that only differs in four amino acids in the N terminus, does not counteract the loss of β-actin in this process. In line with this, ablation of γ-actin (the only other ubiquitously expressed actin isoform at WT stages E8.5–11.5 (2)) does not result in embryonic lethality (63).
This article contains supplemental Figs. S1 to S7, Tables S1 to S5 and Movies S1 to S10.
Supporting Online Material Supplemental Table 1*, Table IIA-D, Table IIIA-E, Table IV*, Table V* (*in TondeleirSuppl.PDF) Supplemental Fig. S1, S2, S3A–G, S4, S5, S6, S7, with descriptive legends for supplementary figures. This file also contains supplemental tables 1, 4 and 5. Movies S1, S2, S3, S4, S5, S6, S7, S8, S9, S10
1 The abbreviations used are: