Rac1-Regulated Endothelial Radiation Response Stimulates Extravasation and Metastasis That Can Be Blocked by HMG-CoA Reductase Inhibitors

Radiotherapy (RT) plays a key role in cancer treatment. Although the benefit of ionizing radiation (IR) is well established, some findings raise the possibility that irradiation of the primary tumor not only triggers a killing response but also increases the metastatic potential of surviving tumor cells. Here we addressed the question of whether irradiation of normal cells outside of the primary tumor augments metastasis by stimulating the extravasation of circulating tumor cells. We show that IR exposure of human endothelial cells (EC), tumor cells (TC) or both increases TC-EC adhesion in vitro. IR-stimulated TC-EC adhesion was blocked by the HMG-CoA reductase inhibitor lovastatin. Glycyrrhizic acid from liquorice root, which acts as a Sialyl-Lewis X mimetic drug, and the Rac1 inhibitor NSC23766 also reduced TC-EC adhesion. To examine the in vivo relevance of these findings, tumorigenic cells were injected into the tail vein of immunodeficient mice followed by total body irradiation (TBI). The data obtained show that TBI dramatically enhances tumor cell extravasation and lung metastasis. This pro-metastatic radiation effect was blocked by pre-treating mice with lovastatin, glycyrrhizic acid or NSC23766. TBI of mice prior to tumor cell transplantation also stimulated metastasis, which was again blocked by lovastatin. The data point to a pro-metastatic trans-effect of RT, which likely rests on the endothelial radiation response promoting the extravasation of circulating tumor cells. Administration of the widely used lipid-lowering drug lovastatin prior to irradiation counteracts this process, likely by suppressing Rac1-regulated E-selectin expression following irradiation. The data support the concern that radiation exposure might increase the extravasation of circulating tumor cells and recommend co-administration of lipid-lowering drugs to avoid this adverse effect of ionizing radiation.


Introduction
Ionizing radiation (IR) is frequently used in cancer therapy to achieve local tumor control. Despite of the enormous merit of radiotherapy in the treatment of malignant diseases, it is well known to cause not only tumor cell death but also normal tissue damage that results in inflammation and fibrosis [1,2]. Another side effect of IR rests on its ability to change the geno-and phenotype of tumor cells that have survived radiation exposure, leading to increased malignancy. Thus, in vitro studies demonstrated a gain of motility, adhesion and invasiveness of tumor cells upon irradiation, which are based on complex changes in gene expression, among others the up-regulation of matrix metalloproteinases (MMP) [3,4,5,6,7]. A variety of preclinical in vivo studies suggest that IR-induced stress responses of surviving tumor cells might promote their invasive potency [8,9,10,11]. Moreover, proangiogenic effects of ionizing radiation have been reported [12], although a report exists claiming the opposite to be the case [13]. Occasional observations in patients are assisting the concern of adverse pro-metastatic radiation effects [14,15,16]. Therefore, failure to achieve local tumor control is suspected to promote the dissemination of single tumor cells from the primary tumor and their subsequent invasion into the normal tissue. The clinical importance of the pro-metastatic radiation effect is controversially discussed [17,18].
Apart from tumor cells, the normal tissue is also exposed to radiation, e.g. during prophylactic cranial irradiation (PCI) or when large parts of the body are irradiated, e.g. for the treatment of Hodgkins disease or for palliative purposes. Low dose exposure of normal cells is even unavoidable in the course of local tumor irradiation. It is conceivable, yet poorly investigated to date, that irradiation of the normal tissue might also stimulate metastasis. This could occur for example by enhanced extravasation of tumor cells that are circulating in the blood at the time of radiotherapy. This possible side effect of radiotherapy, which has not been addressed yet, might be evoked by upregulation of endothelial adhesion factors required for the binding and extravasation of malignant cells. In this process, endothelial selectin (E-selectin), which is upregulated by IR both on the level of the mRNA and protein [19,20], is of particular importance [7,21] as it mediates the interaction of the endothelium with tumor cells that specifically express sialylated carbohydrate structures operating as E-selectin ligands [21,22,23,24,25].
A key response of both normal and tumor cells to IR is the activation of the transcription factor nuclear factor-kappaB (NF-kB) [26,27], which regulates cell survival [28,29] and inflammation [30,31]. Bearing in mind that the expression of a variety of cell adhesion molecules [32,33], including E-selectin [34,35], depends on the transcription factor NF-kB, pharmacological interference with IR-induced activation of NF-kB is a rational strategy for modulating non-beneficial radiation responses of normal tissue. Previously, we showed that HMG-CoA reductase inhibitors (statins), which are widely used in the clinic for lipid lowering reasons, are able to attenuate the endothelial activation of NF-kB provoked by inflammatory cytokines and IR [7,36]. This statin effect rests on the inhibition of the activity of Rashomologous (Rho) GTPases [37] that are key regulators of NF-kB [38].
Here we addressed the questions of whether (i) IR can enhance the binding of tumor cells to the endothelium thus assisting metastasis and (ii) statins are able to mitigate IR-stimulated metastatic processes. We show that IR stimulates tumor cellendothelial cell adhesion in vitro, which is based on the upregulation of pro-adhesive factors in both tumor and endothelial cells. We extended these studies to an in vivo model of metastasis. The data revealed that whole body irradiation of mice that had been transplanted with tumorigenic cells by intravenous injection leads to a significant increase in lung metastases as compared to non-irradiated animals. We further provide evidence that the IRstimulated extravasation and metastasis can be antagonized by pre-administration of the lipid lowering drug lovastatin.

Results and Discussion
Pro-adhesive radiation responses of human tumor and endothelial cells are affected by lovastatin in vitro Previously, we reported a pro-adhesive IR response of human endothelial cells (EC), which is due to the upregulation of endothelial adhesion molecules, in particular E-selectin, via NF-kB [7]. Here, we expand on this observation showing that irradiation of HT29 human colon carcinoma cells (TC) increases their binding to human umbilical vein endothelial cells (HUVEC) in vitro. Tumor cell-endothelial cell (TC-EC) adhesion following irradiation of tumor cells occurred in a dose dependent manner (Fig. 1A). Irradiation of both tumor and endothelial cells exerted additive effects (Fig. 1B). Similar pro-adhesive effects were obtained after irradiation of tumorigenic rodent cells (i.e. CHO-K1 cells; see Figure S1). Whereas pre-treatment of HUVEC with the lipid lowering drug lovastatin attenuated TC-EC adhesion, pre-treatment of tumor cells with lovastatin was ineffective (Fig. 1C). This indicates that the anti-adhesive effect of lovastatin mainly rests on a reduced upregulation of endothelial adhesion factors following radiation exposure. Similar to IR, treatment of HT29 cells with the pro-inflammatory cytokine TNFa, which is Figure 1. Irradiation of both endothelial and tumor cells stimulates TC-EC adhesion that is attenuated by lovastatin. A: Human colon carcinoma cells (HT29) were irradiated with 2-10 Gy. 4 h later, adhesion to primary human endothelial cells (HUVEC) was analyzed as described in Methods. Relative cell adhesion in untreated control was set to 1.0. *p#0.05, **p#0.01 (n = 15). B: Human colon carcinoma cells (HT29) or endothelial cells (HUVEC) were irradiated with 5 Gy. 4 h later, TC-EC adhesion was analyzed as described in Methods. **p#0.01 (n = 8). C: Human colon carcinoma (HT29) and/or endothelial cells (HUVEC) were left untreated or were pretreated overnight with lovastatin (10 mM) (Lova). Afterwards, cells were irradiated (10 Gy) (IR) and TC-EC interactions were assayed after further incubation period of 4 h as described in Methods. See also Figure S1 and S2. ***p#0.001; **p#0.01 (n = 8). D: Logarithmically growing human endothelial (HUVEC) or tumor cells (HT29) were left untreated or were pretreated overnight with lovastatin (Lova). Afterwards cells were irradiated (IR) with 10 Gy and harvested 2 h or 4 h later for analysis of mRNA expression of various cell adhesion molecules. GAPDH mRNA expression was determined as internal loading control. doi:10.1371/journal.pone.0026413.g001 another prototypical NF-kB activating agent, also augmented TC-EC adhesion (see Figure S2A). We also show that TC-EC adhesion following TNFa treatment of HUVEC [7], was further enhanced when the colon carcinoma cells were irradiated (see Figure S2B and S2C). Whether treatment of either HUVEC or HT29 cells with TNFa plus IR ameliorates TC-EC adhesion as compared to a single treatment remains to be tested. Regarding HUVEC, such additivity might be expected since we observed TNFa and IR to have additive effects on the expression level of Eselectin protein [7]. Overall, the data indicate that activation of the transcription factor NF-kB, either by TNFa or ionizing radiation, is of relevance for TC-EC adhesion.

Lovastatin mitigates IR-induced expression of cell adhesion molecules in endothelial cells
To ascertain which cell adhesion molecules are most relevant for the IR-stimulated increase in TC-EC adhesion and the inhibitory effect of lovastatin, the mRNA expression levels of a subset of cell adhesion molecules were analyzed. In endothelial cells, IR increased the mRNA levels of E-selectin, VCAM and ICAM-1 (,5.2-fold) (Fig. 1D) and IR-induced expression was largely blocked by lovastatin (Fig. 1D). We assume that protein expression is altered accordingly since we found in a previous study that changes in the mRNA levels of E-selectin and ICAM-1 are accompanied by congruent alterations in the amounts of protein [7]. In tumor cells, IR increased the mRNA expression of genes coding for CD44 (,2.2-fold), LAMP2 (,1.7-fold) and CXCR4 (,3.1-fold), which are described as E-selectin specific counter-ligands [39,40] (Fig. 1D). Notably, their mRNA expression level was not affected by lovastatin (Fig. 1D). This is in line with the observation that lovastatin pre-treatment of tumor cells did not impact TC-EC adhesion. Whether CD44, LAMP2 and/or CXCR4 are functionally involved in TC-EC adhesion stimulated by ionizing radiation is still unclear. Recently it has been reported that IR promotes the sialylation of integrin beta 1 [6]. Since sialylated proteins are discussed as E-selectin counter-ligands [25], it is conceivable that stimulation of TC-EC adhesion following irradiation of tumor cells might result from changes in the posttranslational modification of pro-adhesive molecules rather than an increase in de novo protein synthesis. Overall, the data show that simultaneous irradiation of TC and EC has an additive effect on TC-EC adhesion in vitro. This is likely due to a concomitant upregulation of adhesion molecules in endothelial cells and corresponding counter-ligands in tumor cells.

Ionizing radiation stimulates tumor cell extravasation and formation of lung metastases in mice
Binding of circulating tumor cells to the endothelium is a prerequisite for their extravasation in vivo. Hence, the in vitro data are alarming because they implicate that radiotherapy might promote metastasis in vivo. To scrutinize this hypothesis we performed experiments using different mouse models. We showed that total body irradiation (TBI) of Balb/c mice results in the upregulation of both E-selectin and ICAM-1 mRNA levels ( Fig. 2A), which was accompanied by activation of NF-kB (Fig. 2B). This is fully in line with other experimental systems demonstrating activation of NF-kB and subsequent expression of E-selectin and ICAM mRNA and protein by IR [20,34,35,41]. Next, we determined whether TBI promotes extravasation and metastasis. To this end, tumorigenic Chinese hamster (CHO-K1) cells, which were shown to have significant metastatic potential [42], were injected into the tail vein of immunodeficient mice (Rag 22/2 Balb/c strain). In our hands, CHO-K1 cells turned out to be the most appropriate cell line in this mouse model. Other cell lines tested were not metastatic at all or showed too high basal level of metastasis (see Figure S4). Immediately after injection, mice were subjected to TBI with a single dose of 4 Gy and colonization of the lung with metastases was analyzed 3-4 weeks later in comparison with non-irradiated animals. As shown in Fig. 2C, TBI caused a dramatic increase in the formation of lung metastases and, to a lesser extent, also increased the formation of liver metastases. Apparently, TBI increases the likelihood of extravasation of circulating tumor cells into the lung vasculature and subsequent formation of lung metastases. The pro-metastatic effect of TBI was further illustrated using CHO-K1 cells for transplantation that overexpress the red fluorescent protein (Fig. 2D) or the luciferase protein (Fig. 2E). Lung metastasis was also quantitated after intravenous injection of b-Gal overexpressing CHO-K1 cells followed by the analysis of bgalactosidase activity in protein extracts prepared from the lung of irradiated versus non irradiated mice. These studies revealed that TBI following transplantation of tumorigenic cells stimulated lung metastasis by about 4-5-fold (Fig. 2F). The pro-metastatic effect of TBI was also demonstrated by calculating the tumor area in lung sections, which was significantly enhanced as compared to nonirradiated mice (Fig. 2G). The data clearly show that TBI of mice harbouring circulating tumor cells leads to a massive colonization of the lung with metastases. Since extravasation of circulating tumorigenic cells is a prerequisite for the formation of lung metastases and, as shown above, TC-EC adhesion is stimulated by IR, we conclude that TBI promotes the process of tumor cell extravasation. With respect to the clinical situation, the data point to the possibility that the well established and highly appreciated therapeutic effect of radiotherapy (i.e. killing of tumor cells) might be influenced by an increased probability of circulating tumor cells, which have survived irradiation, to extravasate and develop tumors at secondary sites.

Lovastatin counteracts IR-induced metastasis in mice
In search of drugs counteracting the putative pro-metastatic effect of IR, we explored the HMG-CoA reductase inhibitor lovastatin. The rationale behind this is that statins, which are widely used as lipid-lowering drugs, block the Rac1-regulated and NF-kB-dependent expression of E-selectin following TNFa or IR exposure of endothelial cells in vitro [7,36]. Lovastatin mitigates TC-EC interaction in vitro (see Fig. 1) and inhibits activation of NF-kB following irradiation in vitro [7] and in vivo [43]. Moreover, statins alleviate normal tissue damage (i.e. pro-inflammatory and pro-fibrotic radiation responses) that results from radiotherapy as acute or delayed side effect [44,45,46,47]. In agreement with our in vitro data, lovastatin impaired the TBI-induced upregulation of cell adhesion molecules in vivo (Fig. 3A). Most importantly, shorttime pre-treatment of Rag 22/2 Balb/c animals for 2 days with lovastatin clearly attenuated the TBI-induced formation of lung metastases (Fig. 3B). We should note that lovastatin does not increase the sensitivity of CHO-K1 cells to irradiation [48]. Assaying b-galactosidase activity in lung extracts, which was taken as indication of the tumor burden, we calculated that lovastatin reduced the TBI-stimulated formation of lung metastases by about 75% (Fig. 3C). Amazingly, under these experimental conditions of short-time pre-treatment with lovastatin, the statin had a weak prometastatic effect on its own (see Figure S3). Extending the pretreatment period with lovastatin for 14 days, the prometastatic statin effect was largely vanished and TBI-stimulated lung metastasis was completely blocked (Fig. 3D). Pre-and posttreatment with lovastatin had the same anti-metastatic effect as pre-treatment alone (Fig. 3D), showing that (i) pretreatment with lovastatin is sufficient to block metastasis and (ii) post-treatment with lovastatin for an extended period of time does not result in adverse effects with respect to metastasis.
In order to elucidate whether TBI was effective in stimulating metastasis when the animals, but not the tumor cells were irradiated, mice were irradiated first and thereafter tumor cells were transplanted. As shown in Fig. 4A, TBI of animals followed by transplantation of non-irradiated tumor cells also resulted in enhanced formation of lung metastases as compared to mice that have not been pre-irradiated. When the animals were pre-treated with lovastatin before they were subjected to TBI and then received non-irradiated tumor cells, the formation of lung metastases was reduced (Fig. 4B). These findings show that the pro-metastatic effect of TBI is a trans-effect, i.e. it is due to a radiation response of the normal tissue that stimulates the metastatic properties of the tumor cells. It is reasonable to suggest that this occurs via transient activation of pro-adhesive endothelial functions. Pre-treatment of mice with lovastatin is sufficient to block this pro-adhesive radiation response of the endothelium and, therefore, antagonizes IR-mediated extravasation of tumor cells and formation of lung metastases.
Whether TBI promotes lung metastasis of human tumor cells could not be addressed in the Rag 22/2 Balb/c mouse model since none of the human tumor cell lines tested (see Figure S4) formed lung metastases in this host. Therefore, we transplanted human HT29 colon carcinoma cells into CB-17 SCID mice where they formed lung metastases with a low basal frequency. Because SCID mice are radiosensitive, which is due to a defect in DNA repair [49], it was necessary to irradiate the animals with lower IR doses. It can be assumed that this dose reduction has a negative impact on the pro-metastatic response of the animals and, therefore, the SCID-based model was applied for confirmatory purpose only. Despite these limitations, the results fully confirmed the data Gy. 4 h after total body irradiation (TBI) mRNA expression of cell adhesion molecules was analyzed in large blood vessels from pulmonary and abdominal artery by RT-PCR. GAPDH mRNA expression was measured as internal loading control. As a positive control, isolated blood vessels were incubated with TNFa (10 ng/ml, 30 min, 37 o C). B: Balb/c mice were subjected to TBI (6 Gy). 4 h after radiation (IR), phosphorylation of NF-kB inhibitor IkBa (p-IkBa), which is indicative of activation of NF-kB, was analyzed in liver extracts by Western blot analysis. ERK2 protein expression was determined as internal loading control. C: 2610 6 CHO-K1 cells were injected into the tail vein of immunodeficient Rag 22/2 Balb/c mice. Immediately after injection, mice were irradiated (4 Gy) (TBI). 3-4 weeks after TBI, the formation of metastases was analyzed in lung and liver. TC, tumor cells injected; TBI, total body irradiation. Shown are representative morphological and histopathological pictures (2-4 animals were analyzed per group). D: Red fluorescent protein overexpressing CHO-K1 cells were used for injection into the tail vein of immunodeficient Rag 22/2 Balb/c mice. Formation of lung metastases was analyzed 3-4 weeks after TBI with 4 Gy. Con, non-irradiated control; IR, TBI with 4 Gy. Shown are representative pictures (2-3 animals were analyzed per group). E: Four weeks after i.v. injection of luciferase overexpressing CHO-K1 cells followed by TBI (4 Gy), luciferase activity in the lung was monitored by use of a life imaging instrument as described in Methods. Shown are representative pictures (2-3 animals were analyzed per group). F: 3-4 weeks after i.v. injection of b-galactosidase (b-Gal) overexpressing CHO-K1 cells followed by TBI (4 Gy), relative b-Gal activity in lung extracts of irradiated mice was related to that of non irradiated control mice, which was set to 100%. **p#0.01 (n = 4-8 mice). G: Formation of lung metastasis was quantitated in the lung of irradiated versus non-irradiated mice by calculating the % tumor area as related to the total lung area. Con, non-irradiated control; IR, TBI with 4 Gy. ***p,0.001 (n = 4-8 mice). doi:10.1371/journal.pone.0026413.g002 obtained with the CHO-K1/Rag 22/2 system. Thus, we observed that TBI with 2.5 Gy promotes lung metastasis after i.v. injection of HT29 cells into CB-17 SCID mice (Fig. 4C). Moreover, lovastatin pre-treatment largely neutralized TBI-stimulated formation of lung metastases in the HT29 cells/SCID mouse model (Fig. 4D). Taken together, we were able to show using two experimental in vivo model systems (i.e. rodent CHO-K1 cell/ Rag 22/2 Balb/c mice and human HT29 cell/CB-17 SCID mice) that TBI stimulates the extravasation of circulating tumor cells and subsequent formation of lung metastases and that this effect is attenuated by lovastatin. The in vivo data thus support the notion of a pro-metastatic potency of IR and the benefit of lovastatin in blocking extravasation and metastasis.

E-selectin is involved in IR-induced tumor cell extravasation and metastasis
Sialyl-Lewis X structures are known as major counter-ligands of the endothelial adhesion molecule E-selectin [25,50]. They have been shown to be of importance for the progression of colon cancer [23]. To examine whether these tetrasaccharide carbohydrates are involved in IR-stimulated TC-EC adhesion in vitro and metastasis in vivo, we made use of the Sialyl-Lewis X mimetic glycyrrhizic acid (GL) [51], which is the major sweet tasting compound of liquorice root. As shown in Fig. 5A, glycyrrhizic acid largely blocked IR-induced binding of tumor cell to endothelial cells, indicating that E-selectin and its ligand Sialyl-LewisX are important for radiation-promoted TC-EC adhesion in vitro. Next, we investigated the involvement of the E-selectin counter-ligand Sialyl-Lewis X on TBI-driven extravasation and metastasis in vivo. Fully in line with our in vitro data we observed that pre-treatment of mice with GL largely attenuated the TBIinduced extravasation and formation of lung metastases (Fig. 5C). The data support the concept that E-selectin and its ligands are essential for TBI-stimulated tumor cell extravasation and metastasis. The Ras-homologous small GTPase Rac1 is required for IR-driven pro-metastatic processes The E-selectin gene is known to be regulated by the transcription factor NF-kB [35]. As mentioned above, the Ras-homologous GTPase Rac1 is required for the activation of NF-kB by TNFa and IR [7,36]. Therefore, it is reasonable to speculate that Rac1 signaling might interfere with the radiationinduced TC-EC interaction in vitro and metastasis in vivo. In line with this, we found that pharmacological inhibition of Rac1 by NSC23766 [52] impairs the radiation-stimulated increase in TC-EC adhesion in vitro (Fig. 5B). Most importantly, pretreatment of mice with NSC23766 also blocked TBI-induced tumor cell extravasation and formation of lung metastases in vivo ( Fig. 5C and Fig. 5D). Therefore, we suggest that TBI-induced extravasation of circulating tumor cells rests on Rac1/NF-kBdependent upregulation of endothelial adhesion molecules, in particular E-selectin, and involves tumor specific E-selectin ligands such as Sialyl-LewisX carbohydrates. Since Rho GTPases, including Rac1, are known as highly relevant targets of statins, we speculate that the observed protective effect of lovastatin against TBI-induced extravasation and metastasis rests on the inhibition of Rac1-NF-kB-regulated upregulation of E-selectin.
In summary, we provide evidence that IR largely stimulates EC-TC adhesion in vitro and that whole body irradiation of mice increases the probability of extravasation and subsequent formation of lung metastases of transplanted circulating tumorigenic cells of rodent and human origin. The pro-metastatic radiation effect appears to rest on the upregulation of both endothelial and tumor cell-specific adhesion factors (see Fig. 6). Importantly, the radiation response is a trans-effect since irradiation of mice prior to transplantation of non-irradiated tumor cells also stimulated metastasis. From this we infer that irradiation of the normal tissue is sufficient to trigger metastasis. Searching for pharmacological strategies counteracting the adverse radiation effect, we show that pre-administration of lovastatin, which is nowadays widely used as lipid-lowering drug, can antagonize pro-adhesive radiation effects in vitro and the pro-metastatic effects of TBI in vivo. The antimetastatic statin effect is very likely due to inhibition of IR-induced normal tissue responses, in particular the IR-stimulated Rac-1regulated increase in the expression of endothelial cell adhesion molecules (see Fig. 6). Thus, the proposed pharmacological strategy suitable to counteract pro-metastatic effects of RT is based on the inhibition of Rac1 signaling, either by statins (e.g. lovastatin) or Rac1-specific inhibitors, such as NSC23766 (see Fig. 6). Alternatively, radiation-stimulated binding of tumor cells to the endothelium can be blocked by Sialyl-LewisX mimetic drugs, which act as E-selectin antagonists (see Fig. 6). However, neither Sialyl-LewisX mimetics nor Rac1 inhibitors have thus far been established for clinical use. Therefore, we propose to focus on statins that are clinically well established, very well tolerated and widely used for lipid lowering purpose. We consider them as first choice drugs for short-term clinical application.
Regarding speculations about the clinical relevance of the findings, it is important to consider the question whether the local concentration of circulating tumor cells in the lung vasculature is high enough to achieve significant extravasation in humans, in particular because radiation to the lung is kept to a minimum in the clinic. Besides, although extravasation is essential for metastasis, additional factors (e.g. motility, proliferation, vascularization) are required to complete the formation of metastases. Despite the complexity of the metastatic process, the data support the concern raised in previous reports [11,12,14,16,18] that RT might stimulate metastasis. This notion should not be taken as an argument against the well-known therapeutic benefit of RT. It just raises the point that, if tumor cells are not fully eradicated by RT, there is the possibility of an increased level of extravasation and metastasis in the survivors. This also pertains to prophylactic irradiation. Although there is no clinical evidence that prophylactic irradiation increases metastasis, a pro-metastatic potential cannot be excluded. Thus, the data reported here support a previous note that calls for a critical reassessment of the benefit of prophylactic irradiation [17,53]. Apparently, further preclinical studies are required using clinically more appropriate experimental settings, for example fractionated local irradiation of the lung with low doses. Irrespective of this, we propose on the basis of the data that the inclusion of statins into current radiotherapeutic regimen is a reasonable strategy for improving the long-term clinical outcome of radiation-based tumor therapy. Since statins are widely used for lipid lowering purposes, the question regarding the clinical relevance of our data might easily be addressed in retrospective and/or prospective clinical studies in which statins are administered concomitantly with RT.

Materials
For treatment of cells and animals with ionizing radiation (i.e. crays), a 60 Co source (Atomic Energy of Cananda Ltd.) was used. Total RNA was isolated by use of E.Z.N.A. Total RNA Kit (Omega Bio-Tek, Norcross, USA). cDNA synthesis was performed using iScript TM cDNA Synthesis Kit (Biorad, Munic, Germany). TNFa, Calcein AM solution, glycyrrhizic acid and DNA oligos originate from SIGMA-Aldrich (Taufkirchen, Germany). dNTPs, the PCR Y-buffer, and nitrocellulose were from Perkin Elmer (Rodgau, Germany). Orange Loading Dye Solution was obtained from Fermentas (St. Leon-Rot, Germany). The Absolute QPCR SYBR Green Fluorescein Mix was from Thermo Fisher (Schwerte, Germany). ERK2 antibody used in this study was obtained from Santa Cruz (Heidelberg, Germany), antibody for phosphorylated IkBa (p-IkBa) originated from New England Biolabs (Frankfurt am Main, Germany). Hyperfilm ECL was from GE Healthcare (München, Germany). Taq DNA Polymerase was a kind gift of H. Kleinert (Mainz, Germany). Lovastatin and Rac1 inhibitor NSC23766 were purchased from Calbiochem-Novabiochem (Bad Soden, Germany) and dissolved in ethanol for the in vitro analyses. For the in vivo studies lovastatin was solubilized in 0.9% saline and sonicated prior to oral administration or injection.
Total RNA purification and RT-PCR reaction 1 mg of purified total RNA was used for cDNA synthesis. Reverse transcription was performed at 25uC for 5 min, 42uC for 30 min and 85uC for 5 min. For PCR reaction 30 cycles were performed (denaturation: 94uC, 1 min; annealing: 60uC, 2 min; polymerization: 68uC, 2 min). PCR products were separated onto agarose gels and DNA was visualized by ethidium bromide staining. The sequence of the sense (s) and antisense (as) primers used for amplification reactions were as follows: E-selectin, f:  TCTCTCAGCTCTCACTTTG, r: TTCTTCTTGCTGCACC-TCT (384 bp)

Preparation of protein extracts, SDS PAGE and Western Blot analysis
For preparation of total protein extracts 15 mg of the respective organs were disrupted by Ultra-Turrax in lysis buffer (20 mM TRIS, pH 7,5, 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 2,5 mM sodium pyrophosphate, 1 mM Na 3 VO 4 , 1 mg/ml Complete EDTA free (Roche), 1 mM PMSF, 1% Triton X-100). Subsequently, samples were sonicated. After centrifugation (10 min, 10.000 x g, 4 o C), the pellet was discarded and the supernatant used for protein determination by Bradford and subsequent Western Blot analysis. Protein extracts were separated by SDS PAGE. Subsequently, proteins were transferred onto nitrocellulose membranes (Perkin Elmer) using a Protean Mini Cell (BioRad). After completion of the transfer, membranes were blocked in 5% non-fat milk in TBS/0.1% Tween 20 for at least 60 min. Incubation with the primary antibody (as indicated) was conducted overnight at 4uC. Incubation with peroxidase conjugated anti-mouse or anti-rabbit secondary antibody (1:2000) (Rockland) was performed for 60 min at room temperature. Bound antibodies were then visualized using a chemiluminescence reaction and Hyperfilm ECL (GE Healthcare). For densitome- trical quantification of the autoradiographies, the Multi Analyst software (BioRad) was applied.

Analysis of cell adhesion
Tumor cell -endothelial cell adhesion (TC-EC adhesion) was analyzed by an ELISA-based method as described [7]. Briefly, cells were pretreated or not with lovastatin (10 mM) for 16h before TNFa treatment or irradiation was performed. Tumor cells were labeled with the fluorescent dye Calcein AM (5 mM, 30 min, 37uC) in RPMI containing 10% FCS. Cells were washed with PBS and resuspended in RPMI containing 10% FCS. 5610 5 labeled cells were added to confluent monolayers of HUVEC (human umbilical vein endothelial cells), which have been stimulated or not 4 h earlier with TNFa or c-rays, for 2 h at 37uC. After removal of nonadherent cells by washing with PBS twice, the attached cells were quantified by fluorimetric analysis (excitation wavelength: 495 nm, emission: 517 nm). Relative adhesion of non-stimulated tumor cells to non-stimulated endothelial cells was set to 1.0. Data are shown as mean values 6 SD from 2-4 independent experiments each performed in quadruplicate (n$8). For statistical evaluation Students t-test was used.

In vivo analyses
For in vivo experiments, immunodeficient Rag 22/2 Balb/c mice (both provided by our local animal facility) were used. Mice were bred in our local specific-pathogen free animal housing facility and were 3-4 months of age at the start of the experiments. To analyze the effect of lovastatin on early radiation responses, Balb/c mice were pre-treated with lovastatin p.o. (10 mg/kg BW) for 2 days. At day 3 the mice were irradiated with 6 Gy and sacrificed 4 h or 24 h later. Blood vessels (isolated from pulmonary and abdominal artery) and organs were isolated for the analysis of mRNA and protein expression. For analyzing the effect of total body irradiation (TBI) on the extravasation and lung colonization of circulating cells the CHO-K1 system, which is an accepted model of metastasis [42], was routinely used. Cells were left untransfected or were stably transfected with either a b-Gal (pcDNA3.1His/lacZ (Invitrogen, Karlsruhe, Germany)), luciferase expression vector (pGL2-basic (Promega, Mannheim, Germany)) or pTurboFP635-C Vektor (Evrogen, Moscow, Russia) and injected into the lateral tail vein of immunodeficient Rag 22/2 Balb/c mice (2610 6 cells/ 0.25 ml PBS/mouse). Lovastatin pre-treatment of the mice (10 mg/kg, p.o. or i.p.) was performed for different period of time (2d -14d). Immediately after injection of the tumor cells, animals were irradiated (total body irradiation) with 4 Gy. TBI was used as a proof-of-principle approach to assess the role of radiation on metastasis. 4 weeks later animals were sacrificed and organs, in particular the lung, were examined macroscopically for the presence of metastases. Lung tissue from each animal as well as any other tissue containing macroscopically detectable metastases were fixed in formaldehyde or frozen in liquid nitrogen for pathological examinations and biochemical analyses, respectively. Glycyrrhizic acid was applied twice, i.e. 24 h (10 mg/kg; i.p.) and 0.5 h (5 mg/kg; i.p.) before TBI. The Rac inhibitor NSC23766 (5 mg/kg; i.p) was given 48 h and 24 h before TBI. Data shown are mean values 6 SD based on the results of at least 2-4 independent experiments with 2-4 animals per group in each experiment (total number of animals 4-8). Statistical significance was calculated using Students t-test. For confirmation of the results, a second independent model system, consisting of immunodeficient CB-17 SCID mice that were irradiated with 2.5 Gy after injection of human colon carcinoma cells (HT29), was used. Under our experimental conditions, the mice behaved normal throughout the experiment and did not show major signs of toxicity such as loss of weight or scrubby coat. All animal work performed in this study was conducted according to the national guidelines and was reviewed and approved by an institutional review board/ethics committee headed by the local animal welfare officer (Prof. Kempski) of the University Medical Center (Mainz, Germany). The animal experiments were additionally approved by the responsible national authority, which is the National Investigation Office Rheinland-Pfalz (Koblenz, Germany). The Approval ID assigned by this authority is the following: AZ 23177-07/G09-1-023.

Histology and quantitative assessment of lung metastasis
Formaldehyde-fixed paraffin-embedded samples were cut into sections of 4 mm thickness and stained with haematoxylin and eosin (HE staining). To quantify the amount of metastases the expression of b-galactosidase activity was quantified in lung extracts using the Beta-Glo assay (Promega, Mannheim, Germany). To this end, lung tissue was homogenized with an ultraturrax and sonicated in sonication buffer (20 mM Tris-HCl, pH 8,5, 1 mM EDTA, 1 mM b-mercaptoethanol and 5% glycerol). Cell debris was removed by centrifugation (109, 10000xg, 4uC). Supernatant was used for measuring the protein concentration according to Bradford before Beta-Glo Assay was performed according to the manufacturers protocol (Promega, Mannheim, Germany). In the reaction, 6-O-b-galactopyranosylluciferin serves as a substrate which is cleaved by b-galactosidase to yield luciferin which is metabolized by luciferase. Bioluminescence was detected using a luminometer (Berthold detection systems, Pforzheim, Germany). The b-gal activity was expressed as relative light units (RLU) and was set to 100% in non-irradiated controls. The fraction of tumor tissue in the lung was also quantitated in HE-stained lung sections ($3 sections per lung/animal) by calculating the percentage of tumor tissue in relation to the normal tissue by use of imaging technique (Photoshop 7.0).

In vivo imaging
Luciferase overexpressing CHO-K1 cells (2610 6 ) were injected into the lateral tail vein of Balb/c Rag 22/2 mice. 4 weeks after TBI with 4 Gy, 100 ml of luciferin solution (15 mg/ml in PBS (Promega)) was injected intraperitoneally. Ten minutes later mice were anesthetized by isoflurane and placed ventral side up in an IVIS Lumina II live image instrument (Caliper, Rüsselsheim, Germany). Images of luciferase activity were continuously acquired until maximal activity was detected. The acquired images were analyzed using the Living Image 3.0 software. Figure S1 IR stimulates the adhesion of CHO-K1 cells to HUVEC. Chinese hamster ovary cells (CHO-K1) and/or endothelial cells (HUVEC) were left untreated or were pretreated overnight with lovastatin (10 mM) (Lova). Afterwards, cells were irradiated (10 Gy) (IR) and TC-EC interactions were assayed after further incubation period of 4 h as described in Methods. 1 irradiation after lovastatin pretreament. ** p#0.01 (n = 4). (TIF) Figure S2 Ionizing radiation promotes TNFa-provoked TC-EC adhesion. A: TNFa treatment of both tumor cells (HT29) or endothelial cells (HUVEC) promotes TC-EC adhesion. Pretreatment with TNFa (10 ng/ml) was performed for 4 h. *p#0.05; **p#0.01 (n = 8). B, C: Endothelial cells (HUVEC) were left untreated or were treated with TNFa (10 ng/ml) for 4 h. Human colon carcinoma cells (HT29 (B) or DLD-1 (C)) or were irradiated with 2-10 Gy. 4 h after irradiation, tumor cells were added to the TNFa pre-treated monolayer of HUVEC and cell adhesion was measured as described in Methods. *p#0.05; **p#0.01 (n = 9-15). (TIF) Figure S3 Effect of lovastatin on the formation of lung metastases. 2610 6 CHO-K1 cells were injected into the tail vein of Rag 22/2 BALB/c mice which had been pretreated with lovastatin p.o. or i.p. for different periods of time. The formation of lung metastases was analyzed three weeks later. +, weak effect; ++, stronger effect. As further control, physiological NaCl solution was administered p.o. Data shown are based on the morphological analysis of n = 3-4 animals per experimental condition. (TIF) Figure S4 Cell line specificity of formation of lung metastases. 2610 6 cells were injected into the tail vein of Rag 22/2 BALB/c mice. Afterwards mice were irradiated with 4 Gy (total body irradiation). The formation of lung metastases was analyzed three weeks later. Control, non-irradiated; IR, total body irradiation; -, no lung metastases detectable; +, weak effect; +++, strong effect. Data shown are based on the morphological analysis of n = 3-4 animals per cell line used. (TIF)