Low-Dose Paclitaxel Reduces S100A4 Nuclear Import to Inhibit Invasion and Hematogenous Metastasis of Cholangiocarcinoma

Cholangiocarcinoma, a malignancy arising from either the intrahepatic or the extrahepatic bile ducts, still carries a severe prognosis. Cholangiocarcinoma is responsible for the 10%–20% of deaths related to primary liver tumors (1). In the Western countries, its incidence has been steadily increasing in the past decades (1, 2), but, unfortunately, the prognosis of cholangiocarcinoma has not changed, with less than 5% of patients surviving up to 5 years from diagnosis (1). At the time of diagnosis, less than 30% of patients are eligible for surgical resection or liver transplantation, the only potentially curative strategies. Thus, in 70% of patients, the stage is advanced, because of the tumor invasiveness and early extrahepatic dissemination. Furthermore, success of curative treatments is hindered by the high rate of recurrence, with a 5-year survival after resection around 20%–40% (3). Combined cisplatin and gemcitabine therapy, the current standard of care for advanced cholangiocarcinoma, increases the patient's overall survival by less than 4 months with respect to gemcitabine alone (4). The lack of effective treatments reflects the deep gap in knowledge on the molecular mechanisms underlying cholangiocarcinoma invasiveness. Better understanding of these mechanisms is needed to predict the invasiveness of the individual tumor and to devise molecular-targeted therapy (5).

Among the biomarkers of increased tumor invasiveness, S100A4 has drawn particular attention in the past few years. S100A4, a low molecular weight, cytoskeleton-associated calcium-binding protein, is normally expressed by mesenchymal (mostly fibroblasts and macrophages), but not by epithelial cells. S100A4 may handle different functions depending upon its subcellular localization. When localized in the cytoplasm, it may interact with cytoskeleton and plasma membrane proteins (including actin, non-muscle myosin-IIA and -IIB, p53, liprin-β1, methionine aminopeptidase-2; ref. 6), thereby contributing to the regulation of cell proliferation, survival, differentiation, as well as cell reshaping and cytoskeletal rearrangement. When translocated to the nucleus, S100A4 may act as a transcription factor for several genes, including those encoding adherence junction proteins, thus controlling cell motility (7). A number of studies have shown that S100A4 is a marker of poor prognosis in breast and colon cancers (7, 8). We have shown that in cholangiocarcinoma patients undergoing surgical resection, nuclear expression of S100A4 in tumor cells is a strong, independent prognostic marker of poor outcome in terms of both metastasization and tumor-related death (9). Furthermore, S100A4 lentiviral silencing significantly reduced motility and invasive capabilities of cholangiocarcinoma cells (9), suggesting that nuclear expression of S100A4 is not merely a marker of cancer invasiveness, but is a key determinant of the metastatic phenotype of cholangiocarcinoma.

Aim of this study was to understand the mechanisms by which nuclear S100A4 induces an invasive phenotype in cholangiocarcinoma and the mechanism regulating nuclear translocation of S100A4. Unfortunately, mechanisms governing S100A4 expression in the nucleus remain elusive, and there are no strategies to selectively target S100A4 nuclear import. In planning our experiments, we came across studies from the early 1990s showing that paclitaxel was able to reduce the total expression of S100A4 in the B16 murine melanoma cells (10, 11).

Paclitaxel is a semisynthetic derivative of taxol, a natural diterpene alkaloid, isolated from the bark of Taxus brevifolia. Because of its antiproliferative and proapoptotic effects, paclitaxel is currently used in chemotherapy protocols for the treatment of ovary, lung, thyroid, and breast carcinomas (12). Our study shows that low-dose paclitaxel inhibits tumor invasiveness and hematogenous metastases by blocking SUMOylation-dependent S100A4 nuclear import in cholangiocarcinoma.

The S100A4-expressing established cholangiocarcinoma cell lines, EGI-1 (both in the nucleus and in the cytoplasm) and TFK-1 (only in the cytoplasm), both obtained from extrahepatic cholangiocarcinoma (9), were purchased from Deutsche Sammlung von Mikroorganismen und Zellkulturen. The primary cholangiocarcinoma cell line CCA-TV3 was isolated from a human sample derived from surgical resection of an intrahepatic mass-forming cholangiocarcinoma, histologically categorized as cholangiocellular carcinoma (grade G3), performed in Treviso Regional Hospital (M. Massani and T. Stecca), as described previously (13). Local regional ethical committee approval was obtained for tissue collection and cell preparation. Cultured cells were grown in RPMI 1640 supplemented with 10% FBS and 1% penicillin at 37°C in a 5% CO₂ atmosphere, and then frozen at low passages (<5). After any resuscitation, cell authentication was performed by checking morphology and by evaluating their immunophenotype as characterized by our previous studies (9, 13), including cytokeratin (K)-7, K19, EpCAM (clone HEA125), E-cadherin, β-catenin, and S100A4. Following experiments were run in cultured cells with <20 passages. Mycoplasma contamination was excluded using a specific biochemical test (Lonza).

In all experiments, cultured cholangiocarcinoma cells were seeded and grown for 24 hours (otherwise differently indicated) before exposure for 24 hours to paclitaxel at low doses (1.5 and 15 nmol/L, diluted in DMSO, Sigma), except for assessment of cell proliferation, viability, apoptosis, and cytoskeletal integrity, where high doses (150 and 1,500 nmol/L) were further tested. Untreated cholangiocarcinoma cells served as controls.

Differential expression of S100A4 was evaluated in cytoplasmic and nuclear cell fractions by Western blotting using the same primary antibody (DAKO, 1:2000) and the NE-PER Kit (Pierce) as detection system, as already performed by us previously (9, 13).

Membrane-type 1-matrix metalloproteinase (MT1-MMP) expression was evaluated by Western blot analysis in total lysate of EGI-1 before and after paclitaxel treatment using an anti-MT1-MMP mAb (Millipore, 1:500). The membrane expression levels of MT1-MMP were then evaluated by assessing the fluorescence intensity profile on cultured cells in five random fields for each experiment (14). See Supplementary Data for details.

Cell proliferation and cell viability were evaluated by BrdUrd (GE Healthcare) and MTS (Promega) respectively, while cell apoptosis was assessed by immunofluorescence for cleaved caspase-3 (Cell Signaling Technology). See Supplementary Data for details.

In paclitaxel-treated cholangiocarcinoma cells, actin filaments were stained by Alexa Fluor 488–conjugated phalloidin (Invitrogen), and then the percentage of cells showing a damaged cytoskeleton on the total cultured cells was evaluated (15). In additional experiments, we assessed the expression of β-tubulin by Western blot analysis in EGI-1 cells before and after exposure to paclitaxel at different doses, in microtubule fractions purified by ultracentrifugation (Cytoskeleton Inc).

For details, see Supplementary Data.

Cell invasion (Boyden chamber) assay was performed as described previously (9, 13).

For details, see Supplementary Data.

As ELISA assessed both pro and active MMP-9 forms (RayBiotech), we performed gelatin zymography in EGI-1 with and without paclitaxel treatment to see whether MMP-9 was actually active. For details, see Supplementary Data.

Cholangiocarcinoma cells were seeded in a 6-well plate and let to grow until confluence, before exposure to paclitaxel. Cell lysates were prepared using CelLytic MT (Sigma) and 200 μL lysates (1 mg/mL) were loaded into column coated with VIVAbind SUMO matrix (VivaBioscience), to capture SUMOylated proteins. SUMOylated protein fraction was then eluted and analyzed by Western blot analysis (1:1,000). Amount of SUMOylated S100A4 was then related to that of total S100A4 (flow through) and compared with controls.

In cholangiocarcinoma cells cultured and exposed to paclitaxel as before, expression levels of the three enzymatic subunits, E1 (activating enzyme), E2 (conjugating enzyme), and E3 (ligase), involved in SUMO modification, were measured by real-time PCR using specific probes (Life Technologies) and compared with untreated cells.

Selected effects of ginkolic acid (GA), a well-established SUMOylation inhibitor, on nuclearization of S100A4 (by Western blot analysis), cell viability (by MTS assay), and migration (by wound-healing assay) of cholangiocarcinoma cells were studied and compared with paclitaxel. Given the toxicity of GA, a preliminary dose–response experiment of cell viability was run treating EGI-1 with increasing concentrations of GA (1, 10, 100 μmol/L) for 72 hours (16).

EGI-1 (500,000 cells suspended in PBS 100 μL) were injected into the spleen of SCID mice (6–8 weeks old; Charles River Laboratories), after transduction with a lentiviral vector encoding the firefly luciferase gene to enable detection of tumor engraftment by in vivo bioluminescence imaging. Tumor engraftment was checked at weekly intervals using the Living Image software (Xenogen), and considered positive when reaching an average of at least 10³ p/sec/cm²/sr, according to our previous studies (9, 13). Cholangiocarcinoma cells injected intrasplenically are delivered straight into the liver through the portal venous axis, giving rise to orthotopic cholangiocarcinomas. Procedures involving animals and their care conformed to the institutional guidelines that comply with national and international laws and policies (EEC Council Directive 86/609, OJ L 358, December 12, 1987), and approved by the Ethical Committee of the University of Padua (Padua, Italy).

Once tumor engraftment was confirmed by bioluminescence imaging (time 0), we started metronomic infusion of paclitaxel [diluted in a 50%/50% solution of Cremophor EL (BASF) and ethanol (Carlo Erba; vehicle)] for 2 weeks at the dosage of 2.6 mg/kg/day by intraperitoneal injection using micro-osmotic pumps (Alzet 1004, Durec). Mice were randomly divided into 2 experimental groups: (i) controls (vehicle only, n = 14); (ii) low-dose metronomic (LDM) paclitaxel (n = 10). After the first week of treatment, mice were checked by bioluminescence imaging (time 1) to detect metastatic spread, as performed previously (9, 13). At the end of treatment (time 2), after a further bioluminescence analysis, mice were anesthetized and sacrificed for necroscopic examination and sample harvesting from spleen (to evaluate the tumor mass at the site of injection) and lungs (to evaluate hematogenous metastases). Tissue samples were fixed in buffered formalin and embedded in paraffin for immunohistochemical analysis. The xenograft model is illustrated in Supplementary Fig. S1A.

The tumor growth in the site of injection was evaluated in paraffin-embedded sections obtained from the tumor-bearing spleen and collected at the time of sacrifice. The tumor mass area was measured using an electronic caliper and expressed in mm².

Serial sections from 10 different cutting plans at a 200-μm interval were taken from lungs of sacrificed mice and stained by hematoxylin and eosin (H&E; Supplementary Fig. S1B). In the same section, IHC for human mitochondria (1:100, Millipore) was performed to detect two different types of metastatic invasion, isolated tumor cells (ITC), represented by single cells or small clumps up to 5 cells, and micrometastases (MM), larger clusters containing more than 5 cells (17–19). IHC for human mitochondria provides a useful tool to improve human cancer cell detection in xenograft models. The metastases were expressed as total number of human mitochondria–expressing cells in the 10 cutting plans.

For details, see Supplementary Data.

Results were shown as the mean ± SD. Statistical comparisons were made using Student t test. Statistical analyses were performed using SPSS 20.0 software (IBM Corp.). A two-tailed P value <0.05 was considered significant.

Continuous data were shown as mean ± SD, and categorical data as counts and percentages. Distributions of tumor size, number of MMs and ITCs in lung samples were graphically displayed by box plots and scatter dot plots comparing the two groups (paclitaxel vs. vehicle). One-sample Kolmogorov–Smirnov test was used to check distributions for normality, while Levene statistic was used for homogeneity of variances. Mean differences between the two groups were analyzed with the two independent samples t test or Welch test, according to the distribution characteristics. Welch t test is an adaptation of Student t test and is more reliable when the two samples have unequal variances and unequal sample sizes. Data were collected and reviewed in Microsoft Excel, and statistical analysis was performed using SPSS 20.0 software. All two-tailed P < 0.05 were considered statistically significant.

As previously shown (9), EGI-1 cells constitutively expressed S100A4 in the nucleus, and therefore represent a good model to study the effects of its nuclear downmodulation. Treatment with low-dose paclitaxel (1.5 and 15 nmol/L) induced a significant and marked reduction in S100A4 nuclear expression (of about 60 and 80%, with respect to controls, respectively; Fig. 1A), without modifying its cytoplasmic expression (Fig. 1B). This finding was further confirmed in primary cholangiocarcinoma cell lines (CCA-TV3) obtained from a surgical sample similarly expressing S100A4 in the nucleus; paclitaxel induced a dose-dependent reduction in the nuclear levels of S100A4 of 38% (1.5 nmol/L) and 62% (15 nmol/L) as compared with controls, again without affecting the cytoplasmic fraction (Fig. 1C and D).

After exposure to paclitaxel at 1.5 and 15 nmol/L, both EGI-1 (Fig. 2A) and CCA-TV3 cells (Supplementary Fig. S2A) showed a significant dose-dependent reduction in cell motility, compared with controls. Cell motility and cell invasiveness of EGI-1 and CCA-TV3 were also significantly reduced in a dose-dependent manner by low-dose paclitaxel (Fig. 2B and Supplementary Fig. S2B, respectively). Interestingly, by comparing paclitaxel 1.5 and 15 nmol/L effects on cholangiocarcinoma cells, we found that the degree of cell motility inhibition paralleled the extent of S100A4 nuclear reduction (Fig. 2A and B).

In contrast, low-dose paclitaxel did not induce significant change in cell proliferation (Fig. 3A for EGI-1, Supplementary Fig. S3A for CCA-TV3), cell viability (Fig. 3B for EGI-1, Supplementary Fig. S3B for CCA-TV3), and apoptosis (Fig. 3C for EGI-1, Supplementary Fig. S3C for CCA-TV3). Cell proliferation, viability, and apoptosis were instead strongly affected by high-dose paclitaxel (Fig. 3A–C and Supplementary Fig. S3A–S3C) in both cholangiocarcinoma cell lines. These data indicate that in cholangiocarcinoma cells, nuclear expression of S100A4 exerts clear promotile and proinvasive effects, without influencing the proliferation/apoptosis balance.

To better understand the mechanisms promoting cell motility and invasiveness dependent upon S100A4 nuclearization, we turned to study the effects of low-dose paclitaxel on the activity of small Rho GTPases, the expression of MT1-MMP, and the secretion of MMP-9. Small Rho GTPases are key effectors of cell motility by inducing the formation of stress fibers (Rho-A), lamellipodia (Rac-1), and filopodia (Cdc-42; ref. 20). MMP-9, on the other hand, is a proteolytic enzyme secreted by many cancer cell types: it potently stimulates matrix degradation, facilitating the invasive migration of tumoral cells from the primary site of growth (21). Its activation depends upon the expression of membrane-anchored MMP, particularly of MT1-MMP, whose expression at the surface of cancer cells is critical for breaking the basement membrane (22). As EGI-1 phenocopied the behavior of primary CCA-TV3 cell line, the following experiments were performed in EGI-1 only. As compared with controls, EGI-1 treated with low-dose paclitaxel showed a significant reduction in Rho-A (Fig. 4A) and Cdc-42 (Fig. 4B) GTP levels, but not in Rac-1 (Supplementary Fig. S4). As evaluated by Western blot analysis and immunofluorescence on cultured EGI-1, MT1-MMP expression significantly decreased after paclitaxel treatment (inhibition of 40% and 46% with 1.5, and of 50% and 49% with 15 nmol/L, for Western blot analysis and immunofluorescence, respectively; Fig. 4C and D). Similarly, MMP-9 secretion and activation were inhibited by challenging cholangiocarcinoma cell cultures with paclitaxel, as shown by ELISA (inhibition of 57% with 1.5, and of 72% with 15 nmol/L) and gel zymography (Fig. 4E and F).

S100A4 is normally associated to cytoskeletal fibers. To study whether low-dose paclitaxel altered cytoskeletal integrity, we performed phalloidin fluorescence in dose-dependent experiments, ranging from 1.5–15 nmol/L to 150–1500 nmol/L. In contrast with high-dose paclitaxel, which induced actin fiber changes (shortening, thickening, fragmentation often leading to dense coiling, accumulating in the perinuclear area) in 16% (150 nmol/L) and 32% (1,500 nmol/L) of cultured EGI-1, more than 90% of cells treated with low-dose paclitaxel showed preserved cytoskeletal structure (Supplementary Fig. S5). These data indicate that effects of low-dose paclitaxel on cell motility and invasiveness are not due to a cytoskeletal damage. Next, we sought to understand the mechanisms regulating the nuclear import of S100A4.

Posttranslational modification by small ubiquitin-like modifier (SUMO) of target proteins is an important mechanism directing their intracellular shuttling. We showed that EGI-1 contained much higher amounts of SUMOylated S100A4 than TFK-1, a cholangiocarcinoma cell line expressing S100A4 only in the cytoplasm, where instead un-SUMOylated S100A4 was detected (Supplementary Fig. S6A). Furthermore, after treatment with low-dose paclitaxel, EGI-1 showed a marked reduction selectively in the SUMOylated fraction of S100A4 as compared with controls, whereas the S100A4 unSUMOylated levels remained unchanged (Fig. 5A). SUMO inhibition by paclitaxel was not associated with decreased mRNA expression levels of the three SUMO E components of the SUMOylating complex (Supplementary Fig. S6B–S6D). To study whether the inhibition of SUMOylation halted cell-invasive capabilities, EGI-1 were treated with GA, a specific natural inhibitor of the E1 subunit. Preliminary dose–response experiments to assess toxicity levels of GA on EGI-1 cells identified 1 μmol/L as the dose devoid of effects on cell viability (not shown), and therefore used onwards. Consistent with our hypothesis, GA significantly reduced cell motility of EGI-1 with respect to controls (Fig. 5B), an effect associated with a significant reduction (37%) in the nuclear expression of S100A4 to an extent comparable with paclitaxel 1.5 nmol/L, without affecting the cytoplasmic levels of S100A4 (Fig. 5C and D).

Altogether, the in vitro data suggest that the SUMOylation-dependent nuclear import of S100A4 is indeed a mechanistic determinant of the invasive phenotype of cholangiocarcinoma cells, which can be inhibited by paclitaxel at nanomolar doses without altering the cytoskeletal integrity, as well as the proliferation and/or apoptosis activities. To test in vivo whether targeting nuclear S100A4 by low-dose paclitaxel is therapeutically relevant to reduce cholangiocarcinoma invasiveness, we moved to the experimental model of cholangiocarcinoma generated by EGI-1 cell xenotransplantation in the SCID mouse (9, 13).

To reproduce the small nmol/L doses of paclitaxel able to hamper S100A4 nuclear entry in cholangiocarcinoma cells in vitro, a low-dose metronomic regimen was chosen for drug infusion in xenografted mice (LDM paclitaxel). To assess specific effects on cancer invasiveness, the treatment was started upon successful EGI-1 cell engraftment as confirmed by bioluminescence imaging (on average 29.75 ± 4.53 days after intrasplenic injection). Notably, before starting LDM paclitaxel, the levels of photon emission were comparable between the two groups. With respect to control animals, LDM paclitaxel neither significantly reduced the photon emission from the spleen through the treatment time (Supplementary Fig. S7A and S7B) nor the size of the splenic tumor mass at the time of sacrifice (Supplementary Fig. S7C). IHC in tissue sections obtained from the tumor mass confirmed that LDM paclitaxel was effective in decreasing S100A4 expression in the nucleus of engrafted EGI-1 cells compared with controls (Fig. 6A); however, the expression of p-Hist3 (proliferation marker) and CC3 (apoptosis marker) did not significantly differ between LDM paclitaxel–treated and control mice (Fig. 6B and C). These findings indicate that LDM paclitaxel reproduces the downmodulating effects on S100A4 nuclear expression by EGI-1 shown in vitro, without causing a dysregulation in the proliferation/apoptosis balance of cholangiocarcinoma cells. In contrast, LDM paclitaxel significantly halted metastatic dissemination. IHC for human mitochondria revealed that both ITCs and MMs in the lung were significantly reduced by paclitaxel treatment (Fig. 7A and B). These data further prove the functional impact of S100A4 nuclear expression on cholangiocarcinoma biology as a mechanism specifically driving hematogenous metastasization, without stimulating tumorigenesis.

An unmet need in cholangiocarcinoma, as in other malignancies, whose dismal prognosis relates to limited therapeutic approaches, is the development of biomarkers able to identify patients most likely to take advantage of curative treatments. Biomarkers may also represent disease-relevant targets for therapeutic interventions (9). Our previous studies showed that nuclear expression of S100A4 in cancer cells of resected cholangiocarcinoma (nearly a half) identified a more invasive clinical phenotype, characterized by increased metastasization and reduced survival after surgery (9). Notably, a worse prognosis after surgery was still observed even when S100A4 nuclear expression was scattered, limited to less than 30% of the neoplastic ducts (9). The aim of the current study was to understand the mechanisms by which nuclear expression of S100A4 promotes cancer invasiveness, and to elucidate the mechanism regulating nuclear translocation of S100A4. We also wanted to understand whether these mechanisms are putative target for therapeutic intervention.

We initially found that paclitaxel given in vitro at low doses (1.5 and 15 nmol/L) was able to effectively and selectively downregulate S100A4 expression in the nucleus of cholangiocarcinoma cells, leaving cytoplasmic expression unaffected. Then, by combining in vitro and in vivo techniques, we showed that: (i) paclitaxel -induced downregulation of nuclear S100A4 was associated with a reduction in motility and invasiveness of cholangiocarcinoma cells, in activity of Rho-A and Cdc-42, and in secretion of MMP-9; (ii) at the doses able to downregulate nuclear expression of S100A4, paclitaxel did not affect cell proliferation, apoptosis, and cytoskeletal architecture; (iii) the nuclear translocation of S100A4 was regulated by SUMOylation, a posttranslational mechanism that was affected by paclitaxel; (iv) in SCID mice xenografted with human nuclear S100A4–expressing cholangiocarcinoma cells, downregulation of nuclear S100A4 by LDM paclitaxel was associated with a reduction in hematogenous metastasization.

The biological functions of S100A4 are largely unknown. They depend on its interacting partners, which are mainly located in the cytoplasm, where S100A4 is commonly expressed. In the nucleus, interacting partners have not been characterized yet, opening the possibility that nuclear S100A4 may act independently as transcription factor (7).

Earlier studies showed that in mouse melanoma cells, taxol at conventional doses, reduced the total amount of S100A, an effect associated with maintenance in the G₀ phase of the cells and increased expression of p53 (10, 11). We initially found that in cholangiocarcinoma cells, low doses of paclitaxel (1.5–15 nmol/L) induced a marked reduction in S100A4 selectively in the nucleus, without altering its cytoplasmic expression. This effect, obtained in an established cholangiocarcinoma cell line (EGI-1) was reproduced in a primary cholangiocarcinoma cell line derived from a patient undergoing surgical resection. The nuclear downregulation of S100A4 was biologically relevant as it associated with a strong inhibition of the motile and invasive properties displayed by cholangiocarcinoma cells in culture. Interestingly, at the same small doses, paclitaxel did not exert antiproliferative and proapoptotic functions, nor it affected cell viability or the integrity of the actin cytoskeletal filaments of cultured cholangiocarcinoma cells. All these cytotoxic effects were instead induced by paclitaxel at higher doses (150–1,500 nmol/L), coupled with pronounced antiproliferative and proapoptotic activities, in line with the mechanism of action supporting the current indications of paclitaxel for the chemotherapeutic treatment of several aggressive carcinomas, from ovary to breast and thyroid cancer (23, 24). The lack of proproliferative stimuli when S100A4 translocates into the nucleus has been recently shown also in colorectal cancer cells (25). In this study, although S100A4 translocated into the nucleus in a cell-cycle–dependent fashion, being most prominent in the G₂–M phase, lentiviral silencing of nuclear S100A4 did not induce changes in cell proliferation. Our findings are consistent with these observations, and confirm that nuclear S100A4 confers specific proinvasive functions.

Among putative molecular players mediating the effects of nuclear S100A4 on cell motility and invasiveness, we focused on the small G proteins (GTPases) belonging to the Rho family and on MMPs. The small Rho GTPases are recognized as key effectors able to activate invasion and metastasis programs (20, 26). In EGI-1 cells, inhibition of S100A4 nuclear import by low-dose paclitaxel significantly reduced the activation of Rho-A and Cdc-42. In cancer cells, Cdc-42 is one of the factors involved in the formation of specialized plasma-membrane actin-based microdomains combining adhesive properties with matrix-degrading activities, called invadopodia, which support cancer invasion by dismantling the basement membrane and then by invading the stromal environment mostly composed of fibrillar type I collagen. A main component of invadopodia is the transmembrane metalloproteinase MT1-MMP, whose function is essential for the in situ to invasive carcinoma transition in breast cancer (14). MT1-MMP promotes the activation of several soluble MMPs, such as MMP-2 and MMP-9. Our data indicate that EGI-1 cells constitutively expressed MT1-MMP and MMP-9, but not MMP-2, and when exposed to low-dose paclitaxel, they showed a significant dose-dependent reduction in the surface expression of MT1-MMP along with the ability to secrete MMP-9. Interestingly, the interplay between Rho-A and Cdc-42 regulates the delivery and accumulation of MMPs at the invading surface (27). Furthermore, as MMP-9 secretion can be also modulated by the activation of the Rho-A/ROCK signaling (21), and the membrane translocation of MT1-MMP is directly regulated by Cdc-42 (28), we suggest the cooperation of Rho-A/MMP-9 and Cdc-42/MT1-MMP pathways to form invadopodia activated by nuclear S100A4.

Cytoskeletal damages induced by paclitaxel, a well-established depolymerizing agent (29), could in theory, influence the nuclear translocation of S100A4. From this point of view, it is important to underline that structural cytoskeleton alterations were absent with low-dose paclitaxel, in contrast with what coherently observed with high-dose paclitaxel.

The mechanisms regulating S100A4 entry into the nucleus remain enigmatic. Notably, S100A4 does not possess nuclear docking sites or nuclear import sequences. In human chondrocytes, nuclearization of S100A4 was induced by IL1β through a SUMOylation-dependent mechanism (30). SUMOylation is a posttranslational mechanism similar to ubiquitination, operated by SUMO proteins, and involved in several biological functions, including protein stability, DNA repair, cell-cycle regulation, apoptosis, nuclear transport, and gene transcription (31). The SUMOylation processes are catalyzed by three enzymes, SUMO E1, E2, and E3, acting as a pacemaker of the cascade reaction leading to the modification of their substrate protein (32). Components of the SUMOylation machinery have been found deregulated in several human cancers, and are emerging as relevant players in tumor invasiveness and in epithelial–mesenchymal transition (33). We found that EGI-1 (expressing S100A4 the nucleus) contained higher amounts of SUMOylated S100A4 with respect to TFK-1 (expressing S100A4 only in the cytoplasm). Then, in EGI-1 treated with low-dose paclitaxel, downregulation of nuclear S100A4 was associated with a significant, dose-dependent reduction in the S100A4 SUMOylated fraction, without changes in the un-SUMOylated subset. These effects occurred without affecting the expression levels of E1, E2, E3 subunits of the SUMOylating complex, thus suggesting a functional inhibitory mechanism, similar to GA. GA is a natural compound derived from Gingko biloba that specifically inhibits the first step of the SUMOylation reaction by directly binding E1 and inhibiting the formation of the E1-SUMO intermediate (16). Of note, also the reduction in nuclear levels of S100A4 induced by GA significantly inhibited EGI-1 cell motility, to an extent comparable with paclitaxel. This observation confirms the relevance of SUMOylation in mediating the pro-oncogenic functions of S100A4 when translocated into the nucleus.

To translate the results of the in vitro experiments, we turned to an experimental model of cholangiocarcinoma generated by xenotransplantation of human EGI-1 cells into the spleen of SCID mice, as we performed in previous studies (9, 13). Because of its short half-life (34), paclitaxel was administered to xenografted SCID mice with a LDM infusion. Continuous delivery of paclitaxel by LDM ensured the achievement of constant low concentrations of paclitaxel comparable with the small doses used for the in vitro experiments (34). To evaluate specific effects on tumor invasion, LDM paclitaxel was started only after human cholangiocarcinoma cell engraftment was confirmed by bioluminescence imaging. At the end of treatment, histologic evaluation of the splenic tumor showed in the LDM paclitaxel–treated group, a significant reduction in the amount of nuclear S100A4-expressing EGI-1 cells. In line with what observed in vitro, downregulation of S100A4 was not associated with reduced tumor cell proliferation (p-Hist3) or with increased apoptosis (CC3) of cholangiocarcinoma cells. Therefore, our experimental model is ideal to study whether the pharmacologic targeting of nuclear S100A4 is a useful strategy to inhibit invasiveness and metastasization. We found that LDM paclitaxel did not affect the growth of the tumor at the site of injection, but caused a significant reduction in both ITCs and MMs in the lungs, where EGI-1 cells metastasize following a hematogenous route through the portal vein and the hepatic veins. This finding is of great clinical value, because the lung is the site most frequently involved in the extrahepatic progression of cholangiocarcinoma, fostered by a specific mechanism of vascular encasement by tumor cells (5).

It is important to underline that paclitaxel may exert additional functions not related to modulation of S100A4 nuclear transport that potentially may contribute to its anti-invasive effects. These are largely dependent upon paclitaxel ability to promote microtubule polymerization and stabilization, which inhibits mitosis and leads to apoptosis. While these mechanisms of action are well evident at the conventional doses, at the much lower doses used in the current study, they do not seem to occur. Recent studies performed in ovarian carcinoma cells indicate that alternatively, paclitaxel may inhibit the expression of other critical molecular factors of tumor progression, such as hypoxia-inducible factor-1α and VEGF (35).

In conclusion, this study unveils a specific role of SUMOylation-dependent nuclear import of S100A4 in cholangiocarcinoma on hematogenous metastasization. The small GTPases Cdc-42 and Rho-A, in concert with MT1-MMP and MMP-9, are the molecular effectors mediating the proinvasive functions promoted by S100A4 nuclearization. Bearing in mind the potential toxic effects of paclitaxel in patients with overt cirrhosis and cholestasis (36), these mechanisms represents a promising therapeutic target aimed at preventing metastatic dissemination after detection of the tumor.

S. Indraccolo reports receiving a commercial research grant from Oncomed Pharmaceuticals, Inc. No potential conflicts of interest were disclosed by the other authors.

Conception and design: M. Cadamuro, C. Spirli, L. Fabris, M. Strazzabosco

Development of methodology: M. Cadamuro

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): G. Spagnuolo, L. Sambado, S. Indraccolo, G. Nardo, A. Rosato, S. Brivio, T. Stecca, M. Massani, N. Bassi

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): E. Novelli

Writing, review, and/or revision of the manuscript: M. Cadamuro, S. Indraccolo, C. Spirli, L. Fabris, M. Strazzabosco

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): T. Stecca

Study supervision: N. Bassi, L. Fabris, M. Strazzabosco

Other (immunohistochemical analysis and quantitative immunohistochemical evaluation): G. Spagnuolo

Other (cell, culture, Western blotting): L. Sambado

Other (Western blotting, scratch assay): S. Brivio

Other (immunohistochemistry): C. Caslini

This work was supported by grant from the Progetto di Ricerca Ateneo 2011 (Grant #CPD113799/11 to L. Fabris and M. Cadamuro) and Grant Associazione Chirurgica Tarvisium (M. Massani, T. Stecca, and N. Bassi), grant Associazione Italiana Ricerca sul Cancro (AIRC #IG14295 to S. Indraccolo), Projects CARIPLO 2011-0470 and PRIN 2009ARYX4T_005 (M. Strazzabosco), NIH grants (DK079005 to M. Strazzabosco; RO1DK 101528 to C. Spirli), and grants from the Silvio O. Conte Digestive Diseases Research Core Centers (DK034989 to M. Strazzabosco and C. Spirli).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.