Snail2 promotes osteosarcoma cell motility through remodelling of the actin cytoskeleton and regulates tumor development

Highlights ► Snail2 is over-expressed in osteosarcoma. ► Snail2 regulates osteosarcoma cell migration and invasion. ► Snail2 also regulates osteosarcoma tumor size. ► Snail2 may represent a therapeutic target for osteosarcoma.


Introduction
In epithelial tumor types (e.g. breast, lung and ovarian) the functions of the Snail zinc finger transcriptio nal repressors have been extensively studied [1]. In this context, the key function of Snail2 is similar to its function in embryonic epithelial tissues, namely the promotion of epithelial-to-mes enchymal transition (EMT) [2]. The mechanis m of action is also similar, utilizing transcriptional repression of epithelial cellular adhesion molecules, including E-cadher in, thus allowing cells to break their cell to cell contacts [3][4][5], which is an early step in the process of EMT.
During embryonic developmen t Snail2 is present in one tissue of mesenchym al origin, namely the developing long bone [6]. However its functions in this tissue are largely unknown. Interestingly expression is lost with age and in post-natal bone in vivo , Snail2 is absent (unpublished observations).
In a recent study, we demonst rated that Snail2 is expressed in long bone canine osteosarcomas ; tumors of mesenchym al origin [7]. Furthermore our study also showed that there was a strong correlation between levels of Snail2 and grade (malignancy) of these osteosarc omas. This suggests that the re-expression of high levels of Snail2 in this tumor type may, in part, be responsible for increasing malignancy . Since osteosarcom as are mesenchym al tumors the function of Snail2 cannot be to drive changes in epithelial cell adhesion during EMT, suggesting that it most likely has other unknown functions in these, and possibly other, mesenchymally derived tumor types.
In order to investiga te the function of Snail2 in osteosarcoma, we generate d stable cell lines in which loss of Snail2 function was achieved using small interfering RNA and gain of function using CMV promote r driven over-expression. The motility of these tumor cells in vitro was assessed using a scratch assay and tumor forming ability together with vascular invasion determined in an in vivo model. Knockdown of Snail2 resulted in reduced motility while over expression of Snail2 resulted in increased motility. These changes in motility were associated with changes in the polymerizat ion of the actin cytoskeleton and in focal adhesion s as well as altered expression of Wnt5a, sFRP2 and osteoblast cadherin (OB-Cad). Reductio n of Snail2 expression also resulted in reduced tumor forming ability in an in vivo assay. These data indicate a role for Snail2 in both motility and tumor formation.
shRNA producing plasmid vector, pLVX-shRNA2 (Clontech), using a previously characterized human Snail2 sequence (5 0 -GGACCACAGTGGCTCAGAA-3 0 ) [8], also present in dog. Control vector contained a target sequence for eGFP. For Snail2 overexpression, the coding sequence of human Snail2, minus the stop codon, was inserted into pcDNA3.1 (Invitrogen) in frame with eGFP, producing a Snail2-eGFP fusion protein. Control vector contained the eGFP coding sequence. Cells were transfected with construct, plated in 100 mm culture dishes and selected with G418 and presence of GFP. Clonal colonies (2-3) of positive cells were ring cloned and individually amplified. A representative clone from each cell line was included. Overexpression and down-regulation of Snail2 was confirmed by immunofluorescent and qRT-PCR analysis. Cells from passages 4-10 were used in subsequent experiments.

Scratch assays
Confluent cell monolayers were wounded with a pipette tip to obtain two perpendicular wounds, forming a cross shape. Wounds were photographed at 0, 24 and 72 h using an inverted microscope (Leica, Solms, Germany). Average distances between wound edges were calculated by measuring the uncovered wound area and dividing by the width of the field of view. Distance migrated was calculated by subtracting the average distance between wound edges from that at time 0. For each experiment a total of 12 wounds were measured per group, and each experiment was repeated three times.

In vivo cell invasion assays
Fertilized white leghorn chicken eggs (Henry Stewart, UK) were incubated at 37°C. On day 9 of development [10] the chorio-allantoic membrane (CAM) surface was gently lacerated with filter paper, and a plastic ring (6 mm inner diameter) placed on this region. 25 ll of medium containing 3 Â 10 5 control or Snail2 knockdown cells was added to the ring and the eggs re-incubated for a further 7 days before CAMs were excised and fixed in 4% PFA. Tumor size and cell motility were assessed using a Nikon SMZ1500 microscope and DS-2Mv digital fluorescent camera (Nikon Instruments Inc., Japan). Tumor areas from two separate experiments were measured using ImageJ software (NIH, Maryland, US).

Quantitative real-time RT-PCR
Total RNA was isolated from cells using an RNeasy Plus Mini Kit (Qiagen). cDNA was synthesized using Superscript II Reverse transcriptase (Invitrogen Ltd., Paisley, Scotland, UK) and random hexamer primers. Quantitative real time RT-PCR (qRT-PCR) was carried out as previously described [11] using QuantiTect SYBR Green PCR kit and Opticon 2 LightCycler (MJ Research, Waltham, MA). Primers used were against Snail2, OB-Cadherin, Wnt5A, sFRP2 and the housekeeping genes b-actin, GAPDH and 18S (sequences and conditions in supplemental information ). A relative standard curve was constructed for Snail2, OB-cadherin, Wnt5A, sFRP2 and the housekeeping genes using serial dilutions of their amplicons, and these standard curves were included in each run. Standards were run in duplicate and samples in triplicate. The expression levels for all the genes analyzed were normalized to b-actin, GAPDH and 18S.

Statistical analysis
Data are presented as mean ± standard deviation. Statistical comparison of each Snail2 modified cell line and their appropriate control cell line was performed using the Student's t test in Microsoft Excel. In all cases, P < 0.05 was considered significant.

Results
3.1. Generatio n of stable Snail2 over-expressing /knock-down osteosarco ma cell lines To investigate the functional role of Snail2 in osteosarc oma tumorige nesis, stable cell lines were produced which either overexpresse d or had reduced levels of Snail2. Overexpression and knock-down were confirmed by immunohistoc hemistry and qRT-PCR . Antibody labeling showed that levels of nuclear Snail2 protein were increased in both D-17 and Saos-2 overexpress ing cells compare d to controls (Fig. 1a). In contrast, knockdown cells showed reduced levels of nuclear Snail2 expression compared to controls in both cell lines (Fig. 1a). Analysis of Snail2 expression by qRT-PCR in D-17 cell lines matched the results seen for immunostainin g ( Fig. 1b and c). Snail2 transcript levels were increased and decreased for overexpres sion and knockdown lines respectively, and these changes were maintain ed over time in culture ( Fig. 1b and c). In Saos-2 Snail2 overexpress ing cells, increase in levels of Snail2 transcrip ts were only evident at later passage numbers (Fig. 1d), even though both low and high passage cells produced exogenous GFP tagged Snail2 from the inserted vector. Equally while immunosta ining showed a decrease in Snail2 protein levels in shRNA cells (a decrease similar to that seen in another study using the same target sequence in an ovarian cell line [8]), qRT-PCR analysis showed no decrease (Fig. 1e). This may be due to an imperfect match between target and sequence which can produce inhibition without mRNA cleavage. Cell lines are known to possess SNPs and other genetic changes that may affect perfect matching and cleavage [12][13][14] and thus result in the apparent discrepancy between the mRNA and protein levels.

Snail2 modifies osteosarcoma cell morphology
Control cell lines maintained the same osteoblast-li ke phenotype as parental Saos-2 (human Fig. 2a: A and B and E and F) and D-17 (canine Fig. 2b: A and B and E and F) osteosarcom a cell lines. However , D-17 Snail2 shRNA cells showed a loss of normal osteoblast morphology, losing their characterist ic spindle shape and becoming more polygonal or stellate (compare Fig. 2b: E and G), while Saos-2 Snail2 shRNA cells appeared relatively unchanged (compare Fig. 2a: E and G). Cell morphology was largely normal in D-17 cells overexpressing Snail2 (compare Fig. 2b: A and C), however Saos-2 cells overexpress ing Snail2 showed an abnormal amoeboid appearance (compare Fig. 2a: A and C).

Snail2 regulates osteosarcoma cell motility
Scratch assays were performed to investigate changes in cell motility in Snail2 overexpress ing and siRNA osteosarcom a cell lines. Knock-down of Snail2 in both canine and human osteosarcoma cells resulted in a significant reduction in motility from the wound edge compared to control cells ( Fig. 3a and b). Overexpression of Snail2 significantly increased motility in Saos-2 but not D-17 osteosarcom a cells ( Fig. 3a and b). Actin staining of Snail2 modified cells at the wound edge revealed morphological modifications. Control cells at the leading edge showed prominent cytoplasm ic protrusions into which actin stress fibers extended to the tip, in the direction of migration ( Fig. 3c; panels E and G (Saos2), M and O (D-17). Saos2 Snail2 overexpressing cells had a markedly different morphology and actin distribution. Cells at the leading edge had increased numbers of smaller protrusions containing condensed actin at the tip. In general, the actin cytoskeleton appeared disorgani zed in these cells with few, if any, stress fibers ( Fig. 3c; panel F). In comparison, Saos-2 Snail2 knockdow n cells at the leading edge had less well defined protrusions compared to controls and more prominent and intensely labeled stress fibers ( Fig. 3c; panel H).
D-17 Snail2 overexpress ing cells had similar morphology to controls, however there appeared to be less prominent stress fibers ( Fig. 3c; panel N). In D-17 Snail2 knockdown cells, the vast majority did not form any protrusions . In those that did, actin fibers did mRNA expression levels measured by qRT-PCR. The expression levels were normalized to housekeeping genes, and results are expressed as mRNA copy numbers. In D17 cells Snail2 mRNA transcript levels were increased in Snail2-GFP (b) and decreased in shRNA-Snail2 (c) cells compared with controls (GFP and shRNA-GFP). These were stable over 12 passages. In Saos-2 cells, transcript levels for Snail2 were increased by passage 8 (d) but were not decreased with shRNA for Snail2 at either passage analyzed (e). Results are shown as mean ± SD of three replicates. not fully extend into these structures and were not always arranged in the direction of migration. (Fig. 3c; panel P).

Altered gene expression following Snail2 modulation
qRT-PCR analysis showed that Saos-2 cells overexpressing Snail2 had markedly reduced levels of OB-Cad expression (Fig. 4a), while decreasing levels of Snail2 increased OB-Cad expression. Furthermore Wnt5a showed an increase (Fig. 4c) while its antagonist sFRP2 decreased in Snail2 overexpress ing osteosarcoma cells compared to controls (Fig. 4b). Expression of these two genes was not affected by knockdown of Snail2 (Fig. 4b and c).
Thus stable overexpres sion of Snail2 in Saos-2 cells leads to repression of OB-cadherin, which likely results in weaker cell-cell adhesion combined with an increase in pro-migratory noncanonical Wnt signaling due to increased Wnt5a expression which is compounded by decrease d expression of its antagoni st sFRP2.

Snail2 changes osteosarcom a cell focal adhesions and the cytoskeleton
In D-17 cells, downregul ation of Snail2 was associate d with a disorganized cytoskeletal architectur e ( Fig. 3c; panel L), however the number and size of focal adhesion s appeared normal (Fig. 5B). Conversely in D-17 cells with increased Snail2 the actin cytoskeleton appeared normally organized ( Fig. 3c; panel J), but had fewer stress fibers. In contrast, more focal adhesions were evident (Fig. 5D) In Saos2 cells, down-regul ation of Snail2 had little effect on cytoskeletal architectur e ( Fig. 3c; panel H) but resulted in greater numbers of larger focal adhesions (Fig. 5F). Conversely cells with increased levels of Snail2 had fewer actin cables but had a cortical distribut ion of actin as well as condensed actin appearing at points of cell-cell contact ( Fig. 3c; panel F), similar to the distribution of actin seen in amoeboid cell migration [15]. Furthermore, paxillin was not detectable by immunosta ining (Fig. 5H) indicating lack of focal adhesion s, which may explain the observed weak adherence to substrate in this cell line. This suggests that one function of Snail2 in osteosarcoma cells is to regulate and/or organize cell adhesion and the actin cytoskeleton. In order to determine whether Snail2 regulates the actin cytoskeleton directly through Wnt5a, we treated Snail2 overexpress ing cells with a Wnt5a blocking antibody [9]. Visualization of the actin cytoskeleton showed that neither short (4 h) nor long (24 h) treatment with this antibody had any observable effect on the actin cytoskeleton (Supplemental Fig. 1). This would suggest that cytoskeletal rearrangement is not a direct response to upregulation of Wnt5a.

Knock-down of Snail2 inhibits tumor development and cell invasiveness
The chick chorio-allantoic membran e (CAM) assay was used to study tumor formation and invasiven ess in an in vivo model. Control cells formed prominent tumors and were able to invade the stroma, enter and migrate along the vasculature of the CAM ( Fig. 6a; A  Rhodamine-Phalloidin stained cultures of wounded human (A-H) and canine (I-P) Snail2 expressing, knockdown and control osteosarcoma cells. Human (Saos-2) Snail2 overexpressing cells showed a modified morphology with condensed regions of actin cytoskeleton at the tips of many cells (B and F) compared to their controls (A and E), however they still migrated as a coherent group. Human Snail2 knockdown cells (D and H) showed similar cell morphology and actin cytoskeleton to that seen in controls (C and G). In canine (D17) Snail2 overexpressing cells (J and N) the cell morphology and actin cytoskeleton was comparable to that seen in the controls (I and M). Canine Snail2 knockdown cells (L and P) did not migrate as a coherent group but as small groups of cells which lacked a leading edge and directionality when compared with controls (K and O). Scale bars = 10 lm.   Fig. 6a; A, B, and G), whereas the majority of Saos-2 Snail2 knock-down cells did not form a tumor (n = 7/12), and those that did formed markedly smaller and less dense tumors (1.01 mm 2 ± 0.74; P < 0.001) ( Fig. 6a; D, E, and G). Similarly, control D-17 cells develope d a primary tumor at the site of implantation in all cases (3.77 mm 2 ± 2.97; n = 9 Fig. 6a; I, J, and N). Almost 40% of canine knockdown cells failed to form a tumor and those that did (n = 7/11) formed markedly smaller and less dense tumors (1.28 mm 2 ± 1.27; P < 0.05) ( Fig. 6a; L, M, and N).
Histological examination of CAM graft tumors derived from control D-17 and Saos-2 cells revealed a population of mesenchymal osteosarc oma cells separated by chondroid/oste oid matrix, indicative of chondroblas tic osteosarcoma (Fig. 6b; A and C). Conversely, tumors from D-17 and Saos-2 knock-down cells revealed a population of early mesenchymal cells contiguous with each other separated by minimal matrix (Fig. 6b; B and D).

Discussion
In humans and canines approximat ely 80% of osteosarcomas originate in the appendicular skeleton [16,17] and these are usually more aggressive than those originating at other skeletal sites [18]. Long bone osteosarcomas cause local skeletal and soft tissue destruction and are highly metastati c [19] with 20% of human patients showing clinically detectable metastases on initial presentation [20]. Despite advances in clinical managemen t of osteosarcom a, the prognosis for osteosarcom a patients remains poor, with a reported 5 year survival rate of approximat ely 60% in metastatic disease [21], even when using chemoth erapy as an adjuvant therapy [22].
To date there are no reliable biomarker s that predict clinical outcome or could be used in diagnosis to tailor treatments to individual patients. Therefore in order to improve diagnosis and treatment a better understand ing of the biology of this disease is critical. In a recent study we demonstrated a correlation between Snail2 expression and tumor grade [7], which suggests that Snail2 may be involved in osteosarc oma progression, and therefore identifies it as a possible target for therapeutic intervention.
Snail2 is an established mediator of malignancy in epithelial tumors where it induces epithelial -mesenchymal transition (EMT), promoting both onset and progression of the disease [1]. However in osteosarcom a there is no requiremen t for EMT, as these tumors are mesenchym al in origin [23], so the function of Snail2 in these tumors remains unknown . We therefore generate d osteosarcoma cell lines, with both increased and decreased levels of Snail2 protein, to investigate this.
We assessed the ability of these cells to migrate both in vitro and in vivo . In the osteosarcom a cell lines tested, decreasing the levels of Snail2 significantly decrease d cell motility, demonstrat ing that Snail2 is required for migration. Conversely, increasing levels of Snail2 promoted motility in human Saos-2 cells but not in D-17 canine cells, demonst rating intrinsic differences between these cell lines. Saos-2 cells are derived from a primary osteosarcoma and are not highly metastati c. In contrast, D-17 cells are derived from a metastatic osteosarcoma of the lung and therefore may already be migrating at their maximum rate, which cannot be enhanced by additional Snail2 activity. Alternatively Snail2 activity could be saturated in these cells. In addition to migration, the CAM assay assessed the ability of Snail2 knockdown cells to invade stroma and intravasate, key steps in cancer cell metastasis. Control Saos-2 and D-17 cells were able to invade and intravasate while Snail2 knockdown attenuated these abilities. While the exact mechanis m underlying these changes is not clear, it may be related to the control of MMP expression as Snail2 mediates the upregulation of MMP2 and MMP9 in a range of other cancers [24][25][26].
To further investigate the molecular mechanisms driving changes in motility, we examine d members of two key pathways linked to Snail expression in both developmen t and cancer. Snail2 is well known to inhibit expression of adhesion molecules such as E-cadher in, increasing cell migration due to reduced cell-cell adhesion [27]. E-cadherin is not highly expresse d in osteosarcoma cells [28], however Snail2 may also regulate expression of other cadherins, such as mesenchymal cadherin (OB-cadherin/CDH11) [29], known to be expressed in osteosarcom as [29]. We have shown that expression of OB-cad is indeed inhibited by Snail2, which would promote cell migration and that knockdown of Snail2 allows for upregula tion of OB-cad. In keeping with this, it has previously been shown that overexpres sion of OB-cadherin in osteosarcomas inhibits migration and reduces metastasis [30]. Furthermore a higher level of expression of OB-cadherin/C DH11 has been correlated with increased patient survival in osteosarcoma [31].
The second molecule investiga ted, Wnt5a, has been described as a tumor suppressor [32]. However, more recent evidence suggests it is a potent inducer of cell motility in a number of tumor types and cell lines, including osteosarc oma [33]. A recent paper  has implied a link between another member of the Snail2 family (Snail1) and Wnt5a expression in Saos-2 cells [34]. Overexpressi on of Snail2 increases Wnt5a expression in Saos-2 cells and it is likely that this has a role in the increased motility observed. Previous analysis of the role of Wnt5a in gastric cancer cell lines indicated that loss of Wnt5a resulted in changes in the actin cytoskeleton [35]. In order to confirm whether Snail2 modulate s cytoskeletal structure directly through Wnt5a, Saos-2 overexpressing cells were treated with a Wnt5a blocking antibody [9]. This resulted in no observable changes in cytoskeletal arrangement, suggesting that the effect of Snail2 on cytoskeletal reorganizati on is not a direct response to Wnt5a. It is therefore possible that Wnt5a is not the driving force for cytoskeletal changes and that the effect of Snail2 on the cytoskeleton is mediated by other mechanis ms such as Rho-GTPas es. In this context, Snail1 induced motility has been very recently reported to be mediated by Rho GTPases [36]. This raises the potential that Snail2 may also promote motility via Rho GTPases in osteosarcom a cells.
The lack of effect seen with Wnt5a neutralizing antibody does not rule out a role for Snail2 in maintaining and/organizi ng the cytoskeleton in link with cell motility. Therefore we examined in detail the cytoskeleton and focal adhesion s.
In Saos-2 cells with decreased levels of Snail2, there were more prominent actin cables however they had fewer cellular protrusions, indicating a less migratory phenotype despite having more focal adhesions. In contrast, cells overexpressing Snail2, had regions of condensed actin in numerous small protrusions , reminiscent of structures described as invadopodia in metastati c cells [37] and indeed these cells were highly motile. Strikingly, they lacked paxillin containing focal adhesions.
Similar to Saos-2 cells, D-17 cell lines had fewer cellular protrusions when Snail2 levels were decreased, but no change in focal adhesions. Unlike the human cells, there was little effect on either the actin cytoskeleton or focal adhesion s when Snail2 levels were increased and these cells did not have increased motility. This suggests that Snail2 has a prominent role in directing actin polymerization to form cell protrusions. The formation of cellular protrusions is largely controlled by actin related protein complexes (Arp 2/3) and actin severing proteins at the leading edge [38]. While Snail2 has been linked to disorganization of the actin cytoskeleton in pancreati c cancer cells [26], its role in cellular protrusions (i.e. lamellopidia, filopodia) has not been explored and this warrants further investiga tion.
Changes in focal adhesions will also alter cell migration dynamics. However the changes in focal adhesion in the human cells would initially appear to be contradic tory to their migratory phenotype. It has previously been shown that knockdown of paxillin in highly metastatic osteosarcom a sub-lines M112 and 132 inhibits migration [39], which is in direct contrast to our observati on that loss of paxillin in Saos-2 cells correlates with increased motility. However fibroblasts derived from paxillin knockout mice retained the ability to migrate [38], suggestin g that paxillin expression is not a direct correlate with migration. However, loss of paxillin/focal adhesions may explain changes in morphology of Snail2 overexpressing cells, as paxillin deficient cells have previously been reported to have similar disorganized cortical cytoskeleton and delayed spreading in culture [40].
Another finding pertinent to the non-canonic al Wnt signaling pathway is that overexpress ion of Snail2 reduces the expression of sFRP2, an inhibitor of Wnt5a signaling [41]. sFRP2 has not previously been identified as a target for Snail2 transcriptio nal repression in either physiological or pathological settings. The net result of downregul ating this gene would be potentiation of the Wnt5a signal. Indeed, in cervical cancer it has been shown that expression of sFRP2 attenuates Wnt signaling and suppresses cancer cell growth [42]. Knockdown of Snail2 did not affect either sFRP2 or Wnt5a expression. This may suggest that these genes are not direct targets of Snail2.
A decrease in Snail2 expression also resulted in the generation of smaller tumors in the CAM assay. Reducing Snail2 expression reduced osteosarcom a cell migration and increased OB-cad expression, thus increasing cell-cell adhesion, which should promote tumor formation. However, cell-cell adhesion is not the only factor that drives tumorige nesis and the tumor microenvironmen t and matrix scaffold composition (such as collagen and fibronectin) is paramou nt for tumor formation [21,43]. Thus Snail2 may also regulate the expression of these proteins in osteosarcom as. Further studies are required to determine if this theory is correct.
Collectiv ely, this study shows for the first time the requiremen t for Snail2 for motility and tissue invasion in human and canine osteosarc oma cells. Furthermore we also show that decreasing the levels of Snail2 impaired tumor developmen t in vivo . Thus, the clinical benefits of selective ly blocking Snail2 in patients with osteosarc oma may be twofold, as it may decrease metastasis, which is a leading cause of death, and also inhibit tumor growth.