Tumor promotion by γ and suppression by β non-muscle actin isoforms

Here we have shown that β-cytoplasmic actin acts as a tumor suppressor, inhibiting cell growth and invasion in vitro and tumor growth in vivo. In contrast, γ-cytoplasmic actin increases the oncogenic potential via ERK1/2, p34-Arc, WAVE2, cofilin1, PP1 and other regulatory proteins. There is a positive feedback loop between γ-actin expression and ERK1/2 activation. We conclude that non-muscle actin isoforms should not be considered as merely housekeeping proteins and the β/γ-actins ratio can be used as an oncogenic marker at least for lung and colon carcinomas. Agents that increase β- and/or decrease γ-actin expression may be useful for anticancer therapy.


INTRODUCTION
At the moment six highly conserved actin isoforms in vertebrates are known: four muscle and two nonmuscle. The main actin isoform studied in the context of carcinogenesis is the α-smooth muscle actin (ACTA2), which is expressed in normal smooth muscle cells, in myoepithelial cells and in myofibroblasts including tumorassociated fibroblasts [1]. Muscle actins are mainly tissue specific, whereas non-muscle β-and γ-cytoplasmic actins (β-and γ-actins hereafter), encoded by ACTB and ACTG1 genes respectively, are ubiquitously expressed in almost all cells [2,3] and can be essential for cell survival [4]. The proportion of β-and γ-actins depends on the cell type [2,[5][6][7]. Their expression differs in various tissues, depending not only on differentiation, but on functional activity of the cell. Actin isoforms expression levels are often associated with different pathological processes [8]. Nonmuscle actin isoforms play crucial roles in cell migration, division and even intracellular signaling [9]. The purpose of this study was to explore the roles of non-muscle cytoplasmiс β-and γ-actin isoforms expression changes in cell transformation and tumor progression. These proteins differ only in four amino acids near the N-terminus [2] and are expressed in normal epithelial cells. Previously, we have shown that β-actin is connected with contraction and adhesion, whereas γ-actin predominantly forms the cortical network necessary for shape flexibility and motile activity of normal fibroblasts and epithelial cells [10].
The majority of studies consider actins to play only an architectonic role. Despite the mechanisms of actindependent migration have been deeply investigated, less is known about possible specific functions of the cytoplasmic actin isoforms in this process. Cell motility can depend on non-muscle β-and γ-actins during embryogenesis and in normal human subcutaneous fibroblasts, with γ-actin determining the directionality of cell movement [10,11]. Partial RNAi suppression of γ-actin expression in SH-EP neuroblastoma cells resulted in a significant decrease in wound healing and transwell migration. Similarly, the knockdown of γ-actin significantly reduced speed of motility and severely affected the cell's ability to explore, which was, in part, due to a loss of cell polarity [12]. Some data on γ-actin regulation of cell migration and ROCK signaling has also been obtained [12,13]. Recent data on cytoplasmic actins as AcGFP fusion proteins overexpressed in colon adenocarcinoma suggest that both actin isoforms have an impact on cancer cell motility, with the subtle predominance of γ-actin [14].
We have previously shown that the relative level www.impactjournals.com/oncotarget of β-actin was decreased in tumors compared with corresponding normal tissue (cervical, breast) while γ-actin was expressed evenly and diffuse in all studied normal and malignant tissues [15][16][17].
The aim of this work was to study the occurrence of the above mentioned actins expression changes in various types of common cancers such as colon and lung. And, most importantly, we aimed to study the role of β-and γ-actins in cell transformation and/or tumor progression as well as to find some proteins through which of actin isoforms could influence these processes.

Cytoplasmic actins expression differs in normal and carcinoma cells of human lung and colon
We have studied the distribution of β-and γ-actins in normal cells compared with malignant human lung and colon epithelial cells. First, we studied β-and γ-actin expression in matching pairs of neoplastic and normal tissues (20 non-small cell lung cancer (NSCLC) and 15 colon cancer). Significant decrease of β-actin staining was observed in NSCLC compared with non-malignant tissue. Quantification of relative fluorescent signal revealed 4-times lower intensity of β-actin staining in cancer lesions compared with matching normal tissues. γ-Actin staining was doubled in carcinoma compared with normal tissues ( Figure 1A, lung). Similar results were obtained for colon cancer ( Figure 1A, colon): 5 times lower intensity of β-actin and about double enhancement of γ-actin staining in neoplastic vs normal cells.

Ras-transformation reduces β-actin and stimulates expression of γ-actin
Activation of the Ras pathway is one of the most frequent molecular abnormalities of various malignancies, including lung and colon cancers. We introduced N-RasD 13 [18] into human normal spontaneously immortalized keratinocytes HaCaT. Exogenous N-Ras-expressing cells developed a transformed fibroblast-like phenotype that can be described as EMT III type [19,20]. Control epithelial cells had well-developed β-actin circular bundles connected with intercellular contacts ( Figure 1С). Ras-induced ERK1/2 activation was accompanied with a significant β-actin decrease and γ-actin increase while the total actin amount remained unchanged ( Figure 1B, Sup. Figure 1).

Non-muscle actin isoforms in phenotype of normal HaCat cells
We created stable HaCaT derivatives with overexpressed or silenced β-or γ-actins to check their presumable transforming potency ( Figure 1D, 1E). Downregulation of β-actin by shRNA or γ-actin overexpression induced a spread and fan-shaped phenotype. Downregulation of γ-actin or β-actin overexpression induced a more contractile phenotype ( Figure 1F). The total amount of cytoplasmic actin remained unchanged. Any isoform expression alteration led to the opposite effect on the other isoform both on protein ( Figure 1D) and on mRNA levels ( Figure 1E).
Changes of actins expression can induce phenotypical traits of transformed cells as well as influence their physiological properties. Basal weak invasion of HaCaT cells in matrigel assay increased significantly upon γ-actin overexpression ( Figure 1G, left). On the other hand, β-actin overexpression significantly reduced HaCaT proliferation in culture. shRNA β-or γ-actin down-regulation reduced proliferation ( Figure 1G, right), while γ-actin overexpression stimulated it. Despite its pro-invasive and pro-proliferative effect in immortalized keratinocytes, γ-actin overexpression was not sufficient to induce HaCaT tumorigenicity in a xenograft model. So to study the potential γ-actin oncogenesis-promoting role we further used lung adenocarcinoma A549 and colon carcinoma HCT116 cell lines as well as their xenografts.

Actin isoforms down-and up-regulation alter cancer cells phenotype
We obtained A549 (Figure 2, left) and HCT116 ( Figure 2, right) derivatives with overexpressed or silenced β-and γ-actins. shRNAs depletions were specific for each of cytoplasmic actin isoforms ( Figure 2A). β-Actin depletion in A549 and HCT116 as in HaCaT cells was accompanied by γ-actin up-regulation. γ-Actin down-regulation resulted in β-actin enhancement. The compensatory mechanism for cytoplasmic actin isoforms expression acts both on mRNA and protein levels ( Figure  2A, 2B).
Parental A549 and HCT116 cultures have moderate or low levels of β-actin staining, mainly presented by diffuse or disorganized β-actin bundles and evenly moderate level of γ-actin cortical staining. In A549 ( Figure 2C, left) and HCT116 ( Figure 2C, right) cells further shRNA β-actin down-regulation induced a spread and motile cell phenotype and γ-actin down-regulation induced a more contractile one. β-Actin overexpression inhibited spread and motile phenotype of both cell lines, which lead to a "more normal" epithelial phenotype. γ-Actin overexpression induced a "more transformed" fanshaped scattered phenotype. This demonstrates that γ-actin Oncotarget 14558 www.impactjournals.com/oncotarget  13 . Graphs represent relative actins expression (Mean ± SD). C. Immunofluorescent staining for β-actin (green) and γ-actin (red) of N-RasD 13 transformed HaCaT cells. Scale bars represent 10 µm. D. WB analysis of HaCaT cells with exogenous expression of β-or γ-actins and corresponding shRNAs. shC is sh to GFP. Graphs represent relative actins expression (Mean ± SD). E. Semi-quantitative RT-PCR analysis of β-or γ-actins expression in cells with corresponding shRNAs. Graphs represent folds of RNA expression in comparison to control cells (Mean ± SD). F. Immunofluorescent staining of HaCaT cells with altered β-or γ-actins expression. Scale bars represent 10 µm. G. HaCaT cells with silenced or exogenously expressed non-muscle cytoplasmic actins that crossed the matrigel-coated membranes (left). Graphs represent mean ± SD. HaCaT proliferation dynamics with exogenous expression or silenced of β-or γ-actins (right). Error bars are SD.

Growth kinetics in cell culture and xenograft nude mouse model: γ-actin stimulates carcinoma cells proliferation in vitro and in vivo
Overexpression of γ-actin in A549 and HCT116 cells led to a significant acceleration of proliferation. β-Actin exogenous expression slightly slowed down cell division while silencing of studied actins led to significant alterations in proliferation in vitro ( Figure 3A). The latter phenomenon could be explained by impaired cytokinesis in β-actin-depleted cells [10] and by the necessity of both actin isoforms for mitosis. Full inactivation β-or γ-actins inhibits cells division. Subcutaneous A549 and HCT116 tumors with overexpressed γ-actin grew faster compared with control cells. β-Actin overexpression slightly slowed down xenografts growth. Silencing of β-and γ-actins led to A549 and HCT116 subcutaneous growth complete arrest ( Figure 3B; Sup. Figure 2). Overexpression of actin isoforms remained unchanged after 21 days of subcutaneous growth in both cell lines ( Figure 3C).

γ-actin overexpression stimulates invasion
In Boyden chambers assay mirroring cellular characteristics linked to malignant features cells with down-regulated γ-actin demonstrated lower invasiveness in vitro compared with controls. Similar to γ-actindeficient cell cultures, invasion was almost blocked in β-actin overexpressing A549 and HCT116. On the contrary, γ-actin stimulated invasion of both cell lines; HCT116 cells with down-regulated β-actin also invaded more effectively ( Figure 3D).

Reciprocal regulation between β-and γ-actins and ERK1/2
Both cell proliferation and invasion in vitro are stimulated by growth factors activating the canonical MAPK pathway. We supposed that changing actin isoforms ratio could influence the intensity of this signaling. Indeed γ-actin predominance in A549 and HCT116 was associated with ERK1/2 activation (according to western blot analysis, Figure 4A).
We confirmed phospho-ERK1/2 and γ-actin binding by co-immunoprecipitation ( Figure 4C). For these experiments we used A549 cells with silenced βor γ-actins to minimize the influence of actin isoforms on biochemical data. Using PLA (Proximity Ligation Assay) [21,22] we verified the pERK1/2−γ-actin colocalization ( Figure 4D). PLA using γ-actin and pERK1/2 demonstrated highly specific and strong signals as multiple cytoplasmic dots in control and β-actin-deficient cells ( Figure 4D). Comparative fluorescent signals of pERK1/2−γ-actin PLA dots in control and actins-depleted A549 cells quantification is shown in Figure 4D. β-Actin and pERK1/2 antibodies gave fluorescent signals on the level of background.
So these experiments showed for the first time that active ERK1/2 interacts with γ-actin in carcinoma cells. Moreover, ERK1/2 activation leads not only to stimulation of cell proliferation and morphological changes in intercellular architecture, but also to β-actin down-and γ-actin up-regulation. So we assume that alterations in β/γactin ratio may partly explain the phenotypical changes upon ERK1/2 activation.

γ-actin selectively interacts with Arp2/3 complex
Actin polymerization at the leading edge is crucial for any type of cell migration [23]. We have previously demonstrated that both β-and γ-actins are enriched in protrusions with active actin branching and nucleation at the leading edge of normal subcutaneous fibroblasts [10]. It has been previously shown that ERK-MAPK signaling drives lamellipodial protrusions with WAVE2 regulatory complex activation [24]. ERK pathway activation during migration leads to actin reorganization by WAVE2/Arp2/3 polymerization complex [25][26][27]. Our data on spreading fibroblasts with specific N-WASP inhibitor [10] and preliminary experiments on carcinoma cells using Arp2/3 inhibitor CK666 (not shown) allowed us to hypothesize diverse polymerizing systems/pathways for β-and γ-actins. As ERK1/2 activation is associated with γ-actin at the leading edge we assumed that Arp2/3 polymerization complex could predominantly interact with γ-actin during actin branching and migration.
Using LSM of cofilin1 in A549 cells with silenced β-or γ-actins ( Figure 6A) we have shown that γ-actin predominance induces enhancement of cofilin1 staining (especially at the leading edge). Co-localization of cofilin1 with γ-actin (especially in β-actin-depleted cells) was demonstrated (Sup. Figure 5). Selective silencing of actin isoforms led to modulations of cofilin1 expression that correlated with γ-actin expression ( Figure 6B). IP experiments confirmed cofilin1−γ-actin interaction ( Figure 6C). γ-Actin/cofilin1 PLA demonstrated specific and strong signals at the leading edge of control and of β-actin-deficient cells ( Figure 6D). β-Actin/cofilin1 PLA gave fluorescent signals on the level of background.
Oncotarget 14565 www.impactjournals.com/oncotarget dephosphorylation by phosphatases. Existence of different phosphatases in various complexes contributes to diversity of their effects [38].
A preliminary screening has revealed that Protein Phosphatase 1 (PP1) is involved in γ-actin-dependent signaling. PP1 is one of the main dephosphorylating enzymes. In particular, PP1 can dephosphorylate cofilin1 and thus activate it [39]. PP1α confocal LSM of control A549 cells showed cytoplasmic staining. β-Actin downregulation in A549 induced a moderate enhancement of PP1α cytoplasmic staining ( Figure 7A). PP1α colocalized with γ-actin at the cellular leading edge. Western blot analysis ( Figure 7B) confirmed reduction of PP1α expression as a result of γ-actin down-regulating. Co-IP showed γ-actin (not β-actin) binding with PP1α ( Figure   7C). γ-Actin−PP1α PLA demonstrated specific and strong signals in the cytoplasm of control and, more obviously, of β-actin-deficient cells. β-Actin−PP1α PLA demonstrated low, but significant fluorescence in control and in γ-actindeficient cells (Figure 7D, 7E).

DISCUSSION
Until recently non-muscle cytoplasmic β-and γ-actins were considered only to play structural roles in cellular architecture and motility. They were viewed as products of housekeeping genes and β-actin was commonly used as internal control in various biochemical Oncotarget 14566 www.impactjournals.com/oncotarget experiments. The difference between β-and γ-actins was poorly studied because of lack of specific antibodies distinguishing these two proteins. Embryonic lethality, with γ-cytoplasmic actin null mice (Actg1-/-) dead within 48h after birth [11], complicated the study of non-muscle actin isoforms. Selective siRNA-mediated knockdown of γ-cytoplasmic actin as compared to β-actin, induced epithelial-to-mesenchymal transition of various epithelial cells, which manifested in increased expression of contractile proteins along with inhibition of genes responsible for cell proliferation [40]. Some data indicated a role of β-and γ-actins in carcinogenesis: components of the Arp2/3 complex were up-regulated in colorectal cancers [41], increased ACTG1 and reduced ACTB levels were identified in osteosarcoma analysis [42].
In our experiments we have observed that stable shRNA-mediated knockdown of γ-actin or enhancement of β-actin induces normalization of carcinoma cells phenotype. We have showed for the first time that β-and γ-actins have distinct roles in tumorigenesis. γ-Actin, as opposed to β-actin, is significantly increased in colon and lung carcinomas when compared with normal tissues. Oncogenic Ras enhances γ-actin expression and reduces β-actin in immortalized HaCaT cells. γ-Actin overexpression in these cells leads to significant phenotypic changes and to invasion induction reminiscent of Ras activation. These alterations are not sufficient to induce tumor growth of HaCaT cells in immunocompromised mice. Nonetheless we have shown that the increase of γ-actin level is a basic characteristic of cell transformation and tumor progression.
We have also discovered the existence of a ratio between these two proteins. A shift in the amount of one isoform is compensated by a reciprocal change in another form that can be explained by transcriptional regulation, at least partially. This phenomenon needs further investigation. The relationship of quantity occurred both in normal and in neoplastic cell lines with total actin amount remaining unchanged upon of β-or γ-actin overexpression or silencing.
We have discovered that γ-actin overexpression (i) accelerates proliferation of carcinoma cells both in vitro and in vivo in subcutaneous xenographts and (ii) results in formation of locomotor phenotype and increased invasion in vitro. On the contrary, β-actin overexpression suppresses the above properties and leads to more epithelial phenotype/differentiation.
To investigate the mechanism of these phenomena we studied the role of individual actin isoforms alterations in ERK1/2 MAP-kinases stimulation of cell Figure 8: (SCHEME). In cancer cells the ratio of non-muscle cytoplasmic β-and γ-actins shifts towards γ-actin predominance. γ-Actin (unlike β-actin) can interact with both structural (components of Arp2/3 and WAVE2 complexes, cofilin1 that is dispensable for movement and cellular architecture) and signaling proteins (ERK1/2, PP1). Upon malignant transformation γ-actin becomes overexpressed. The cell acquires a more mesenchymal phenotype. It becomes more invasive and grows faster both in vitro and in vivo. β-Actin predominance in a transformed cell has an opposite effect: a more "normal" epithelial phenotype, impaired invasion and growth. γ-Actin could be considered as a weak oncogene and β-actin as an anti-oncogene.
Oncotarget 14567 www.impactjournals.com/oncotarget proliferation and migration [43][44][45]. We showed that γ-actin predominance led to ERK1/2 activation. We also discovered ERK1/2−γ-actin binding and, even more, upregulation of γ-actin with down-regulation of β-actin upon ERK1/2 activation and vice versa. So we can conclude that there is a positive feedback mechanism between MAP-kinases activation and γ-actin amount, which enhances malignant traits of neoplastic cells. Taking into account that activation of ERK1/2 MAP-kinases in response to activation of various oncogenes is observed in a majority of human neoplasias, we can assume that γ-actin predominance may present a universal malignant feature.
We also discovered γ-actin selective binding to (i) Arp2/3 complex member p34-Arc, (ii) WAVE2, a cofactor essential for actin polymerization, and (iii) cofilin1 required for proper actin branching at the leading edge. It is worth to mention that cofilin1 displays the above properties in a dephosphorylated state and that one of its putative phosphatases, namely PP1, also binds to γ-actin. All these interplays take place at the leading edge of the cell, as registered by PLA. So we can assume that these interactions could serve as a basis for the structural explanation of γ-actin necessity for migration and invasion. Moreover, γ-actin with ERK1/2 complex could act as a scaffold for further "signalosome" assembly.
We discovered γ-actin dual role in tumorigenesis. Structurally it supports the leading edge formation. It also plays a signal-conductive role. γ-Actin predominance enhances not only the probability of interactions necessary for cellular movement but also the intensity of mitogenic signaling. Vice versa: oncogenic stimuli up-regulate γ-actin in a transformed cell.
To conclude, we have been the first to demonstrate the many-sided role of actin isoforms in cell behavior. The ability of γ-actin to bind key signaling molecules such as PP1α and ERKs and to modify their activity as well as to regulate actin network system via selective interaction with proteins of Arp2/3-complex, WAVE2 and cofilin1 allows to consider it an oncogene involved in tumorigenesis. This suggestion is strongly supported by γ-actin up-regulation in all studied samples of human lung and colorectal carcinomas along with a positive feedback loop with ERK1/2 activation. The results of our experiments are shown in Figure 8. We expect that restoring normal ratio of actins isoforms by diminishing γ-actin expression might provide a new route to improve anticancer therapy. New methods selectively targeting the actin cytoskeleton could be very perspective in this context [46,47].
A549 cells were incubated in medium containing

Immunofluorescent and confocal laser scanning microscopy
Cells on cover slips were fixed in 1% PFA in prewarmed DMEM/Hepes for 15 min and treated for 5 min with MeOH at -20°C. Cells were incubated with primary and secondary antibodies. DAPI (Life technologies) was applied for nuclear staining.
Cell cultures growth rate 5x10 4 А549, 2,5x10 4 НСТ116 and 2x10 4 HaCaT cells were seeded into 6-well plates and cell count was performed each two days using the hemocytometer (three wells per time point). The measurement proceeded till monolayer formation.

Boyden chamber cell migration assay
was performed using transwell Matrigel-coated chambers with 8-µm pore-size membranes (BD Biosciences) according to manufacturer instructions with 5x10 4 A549, 2,5x10 4 HCT116 and 2x10 4 HaCaT cells. The migration activity was quantified by blind counting of the migrated cells of at least 10 fields per chamber.

Nude mice assay
Nude mice were inoculated with 10 6 cells as previously described [49][50][51]. Tumor sizes were measured every 3 days and their volumes were calculated as (width2) x (length) x 0.5. After 3 weeks of observation, explanted tumors were isolated and analyzed. The animal experimental protocols were approved by the Committee for Ethics of Animal Experimentation and the experiments were conducted in accordance with the Guidelines for Animal Experiments in Russian Blokhin Cancer Research Center.

Patients and tissue material
CRC samples and paired non-cancerous tissues were obtained from the patients of Russian Blokhin Cancer Research Center undergone radical surgery for colorectal cancer and lung cancer. Informed consent has been obtained from patients and the project was approved by the institutional review board. All patients had a definite pathological diagnosis (adenocarcinomas G2 grade) and did not receive chemotherapy or radiotherapy before surgery. Formol-fixed, paraffin-embedded samples of human tissues were obtained from surgical material and paraffin archive blocks of the Division of Clinical Pathology, Russian Blokhin Cancer Research Center. Animal tissue samples were rinsed in ice-cold PBS, fixed for 24 hrs in 4 % formaldehyde and placed in 70 % EtOH until embedded in paraffin. The histological www.impactjournals.com/oncotarget sections (5 μm) were deparafinised in о-xylol twice for 5 min and then three times for 5 min in 96 % EtOH before immunofluorescent staining.

Immunofluorescent staining of histological sections
Antigen retrieval was achieved by heating at 95°C in target retrieval solution (pH = 6.0) (S1699, Dako) for 40 min. The sections were incubated with primary antibodies at room temperature for 1 hour, followed by incubations with FITC-/TRITC-conjugated secondary antibodies for 30 min at room temperature. Quantitative analysis for the intensity of immunoflurescence was performed using ImageJ software (http://rsbweb.nih.gov) by measuring the specific color intensity. For each sample at least 5 sections were examined, in each section at least 15 fields of vision were analyzed.

In situ proximity ligation assay (PLA)
PLA was conducted according to manufacturer's instructions (Sigma-Aldrich). In brief, PFA/methanol-fixed cells were incubated with pairs of primary antibodies, washed and incubated with secondary antibodies conjugated with oligonucleotides (PLA probe MINUS and PLA probe PLUS). A proximity-dependent ligation and amplification of DNA reporter molecule followed. The resulting rolling circle products (RCPs) were visualized by fluorescence microscopy upon hybridization with a complementary fluorescence-labeled oligonucleotide probe.

Statistical analysis
Statistical analysis was done with unpaired Student's t tests, and data are expressed as SD as indicated in figure legends. P values ≤ 0.05 were considered to be significant. All the experiments were performed for at least three times.

DISCLOSURE OF POTENTIAL CONFLICTS OF INTERESTS
Authors do not have any conflicts of interests.