Thymoquinone attenuates tumor growth in ApcMin mice by interference with Wnt-signaling

Background Patients with familial adenomatous polyposis (FAP) are at increased risk for the development of colorectal cancer. Surgery and chemoprevention are the most effective means to prevent cancer development. Thymoquinone (TQ) is considered the main compound of the volatile Nigella sativa seed oil and has been reported to possess anticarcinogenic properties. In this study we evaluated the chemopreventive properties of TQ in a mouse model of FAP. Methods APCMin mice were fed with chow containing 37.5 mg/kg or 375 mg/kg TQ for 12 weeks. H&E stained intestine tissue sections were assessed for tumor number, localization, size, and grade. Immunohistochemistry for β-catenin, c-myc, Ki-67 and TUNEL-staining was performed to investigate TQ’s effect on major colorectal cancer pathways. TQ’s impact on GSK-3β and β-catenin were studied in RKO cells. Results 375 mg/kg but not 37.5 mg/kg TQ decreased the number of large polyps in the small intestine of APCMin mice. TQ induced apoptosis in the neoplastic tissue but not in the normal mucosa. Furthermore, upon TQ treatment, β-catenin was retained at the membrane and c-myc decreased in the nucleus, which was associated with a reduced cell proliferation in the villi. In vitro, TQ activated GSK-3β, which induced membranous localization of β-catenin and reduced nuclear c-myc expression. Conclusions In summary, TQ interferes with polyp progression in ApcMin mice through induction of tumor-cell specific apoptosis and by modulating Wnt signaling through activation of GSK-3β. Nigella sativa oil (or TQ) might be useful as nutritional supplement to complement surgery and chemoprevention in FAP.


Background
In developed countries, colorectal cancer is one of the most common types of cancer, ranking 2 nd and 3 rd in women and men respectively. The lifetime risk for colorectal cancer development in the general population is approximately 6% and is responsible for about 8% of all cancer deaths worldwide [1]. Certain familial syndromes are at increased risk for the development of colorectal cancer. This is about 80% in hereditary nonpolyposis colorectal cancer and almost 100% in familial adenomatous polyposis (FAP) [2].
FAP is an autosomal dominantly inherited disease caused by a germline mutation of the adenomatous polyposis coli gene (APC) located on chromosome 5q21 [3,4]. It is characterized by the development of numerous, predominantly large intestinal adenomatous polyps, ultimately leading to colorectal cancer [2]. Besides endoscopic surveillance and prophylactic surgery, chemoprevention is the most effective means to repress cancer development in FAP. Inhibition of polyp progression improves tumor outcome, thereby prolonging survival in FAP patients. The protective effect of non-steroidal anti-inflammatory drugs (NSAIDs) such as aspirin [5], sulindac [6,7] and the selective cyclooxygenase-2 inhibitor celecoxib [8] has been extensively studied in FAP patients. Some also reduce rectal polyp burden after colectomy with ileorectal anastomosis [6]. Celecoxib was shown to be beneficial against the duodenal tumor burden [9]. However, a chemoprevention study dealing with genotypically but not phenotypically affected patients treated with sulindac failed to show prevention of primary adenoma development [7]. Unfortunately, these drugs are associated with gastrointestinal (bleeding, ulceration) and cardiovascular side effects (thromboembolic events and heart failure) in the long-term [10][11][12].
In consideration of such side effects, the use of natural substances such as curcumin, eicosapentaenoic acid, apple polyphenols, capsaicin, or thymoquinone (TQ) for chemoprevention is emerging [13][14][15]. TQ is the main active component of the volatile Nigella sativa (black cumin) seed oil, which is used as a spice in countries with low incidence of colorectal cancer such as Egypt, Pakistan, or India. Traditional medicine has utilized its anti-inflammatory, antioxidant, and anti-carcinogenic properties, supporting TQ as a promising dietary chemopreventive agent [16]. In vitro studies indicate that TQ inhibits tumor cell proliferation in various cancers [17][18][19], including colorectal cancer [20,21]. TQ induces a G1 cell cycle arrest, increases p53 and p21WAF1 protein levels, induces apoptosis in a dose-and timedependent manner, and reduces Bcl-2 protein in HCT116. Further actions of TQ include inhibition of angiogenesis, endothelial cell migration, invasion, and tube formation as demonstrated in HUVECs [18]. In vivo weekly i.p. injections of 5mg/kgbw TQ reduced the number and size of aberrant crypt foci and tumor multiplicity in a chemically-induced colorectal cancer mouse model. The suppression of tumor development was sustainable, as treatment with TQ resulted in a reduction of tumor number even after a 10-week discontinuation. Furthermore, in a HCT116 cell xenograft model, a 3-times weekly i.p. injection of 20 mg/kg TQ reduced the relative tumor size by 29% from 2.8 to 2.0 mm 2 [20]. This study was designed to test the chemopreventive effect of TQ in Apc Min (APC, adenomatous polyposis coli; Min, multiple intestinal neoplasia) mice, which best resemble the FAP phenotype.

TQ attenuates tumor growth in Apc Min mice
To evaluate the effect of TQ on polyp formation in the APC Min mouse, 4-6 week old female and male animals were randomly divided into 4 groups and treated over a period of 12 weeks. Neither TQ nor piroxicam influenced weight gain and food uptake (Additional file 1: Figure S1). Mouse colonoscopy at week 9 demonstrated a significant reduction of distal large intestinal polyps in the TQ-low and the piroxicam group (p<0.05), with a trend also for TQ-high (p=0.124; Additional file 2: Figure S2).

TQ induces apoptosis in polyps of Apc Min mice
To ascertain the effect of TQ on apoptosis, TUNELstaining of Swiss rolls was performed. Apoptotic cells were analyzed within polyps and normal mucosa of the small intestine. The number of apoptotic cells increased in the neoplastic but not in the normal tissue upon TQ treatment ( Figure 2). The average number of apoptotic cells within polyps was 18±16 per FoV for untreated, 45±18 for TQ-low and 50±30 for TQ-high treated mice (p<0.05 and p<0.01, respectively). However, this effect was not observed in the piroxicam treated group (13±14 per FoV). These results suggest that TQ reduces polyp growth through selective induction of apoptosis.

TQ reduces proliferation in the villi of Apc Min mice
Nuclear Ki-67 staining was assessed to investigate the effect of TQ on cell proliferation. The percentage of Ki-67 positive cells was scored in polyps, normal crypts and villi of APC Min . Neither TQ nor piroxicam altered the number of Ki-67 cells in polyps or crypt cells (Additional file 3: Figure S3). In C57BL/6 wt mice Ki-67 positive cells are specifically restricted to the crypt cells. In APC Min mice, however, we observed nuclear Ki-67 staining also in cells of the intermediate zone of the villi. A reduction of Ki-67 positive cells in TQ-high and to a lesser extent also in TQ-low treated mice was observed. The mean Ki-67 IRS in the villi was reduced from 4.8±1.6 to 4.0±1.7 in TQ-low and to 3.6±1.8 in TQ-high (p<0.05) treated cells. Again piroxicam had no effect (IRS 5.1±1.7).

TQ reduces c-myc expression in the polyps of Apc Min mice
We further analyzed nuclear c-myc expression, which is highly abundant in proliferative tissue and necessary for cell cycle progression [22]. TQ-low (p<0.001) and TQhigh (p<0.01) treatment reduced nuclear c-myc protein in the polyps, an effect which was not observed in piroxicam-treated mice ( Figure 3).

TQ translocates β-catenin to the membrane in APC Min polyps
As TQ influenced tumor size and and c-myc expression, we considered how TQ may be implicated in the βcatenin pathway. Loss of APC protein leads to a deregulated WNT/β-catenin pathway, as APC is part of the β-catenin destruction complex. Deregulated β-catenin degradation leads to an accumulation of free β-catenin and nuclear translocation. In the nucleus, β-catenin acts as a transcription factor binding in a complex with TCF/LEF to DNA enhancer sequences leading to the upregulation of genes like the proto-oncogene c-myc [23]. A membranous, cytoplasmic, and nuclear IRS was calculated for β-catenin in the normal mucosa, small and large polyps. Remarkably, β-catenin was translocated to the membrane in large polyps of TQ-high treated APC Min mice (p<0.05; Figure 4). A trend towards this effect was seen for small polyps as well (p=0.064). We found a similar trend for β-catenin shifting to the membrane in piroxicam treated mice (p=0.094). No change in β-catenin expression within the normal epithelium was identified, except for a slight increase in cytoplasmic β-catenin levels in TQ-low treated mice, an effect which was abrogated in small and large polyps. To further investigate β-catenin expression upon TQ treatment, we studied the poorly differentiated colon cancer cell line RKO, which harbors wt APC [24], wt p53 [25] and wt β-catenin alleles, the latter being expressed at low levels [24]. Treatment with TQ resulted in decreased nuclear β-catenin within 4 h and up to 24 h (as shown by Western blot). In parallel, membranous and cytoplasmic β-catenin increased at 4 h ( Figure 5A). In conclusion, TQ reduces nuclear β-catenin and translocates β-catenin to the membrane.

TQ exerts different effects on cell viability on colon cancer cells with various mutational backgrounds
To assess the effect of TQ on cell viability, MTT assays were performed in different human colon cancer cell lines (having various mutational backgrounds) and in HCEC-1CT normal diploid human colon epithelial cells. APC-truncated and p53-mutated cells (DLD1 and HT29) are the most resistant to TQ (IC 50 : 196 μM and 160 μM, respectively). LoVo, which has wt p53 and mutated APC, was the most sensitive, having an IC 50 value of 36 μM. In comparison, cells with wt p53 and wt APC such as HCT116, RKO and HCEC-1CT were less affected by TQ, with IC 50 values of 118, 86 and 79 μM, respectively (Additional file 4: Figure S4A). These results indicate that TQ's effect on cell viability may be influenced by the mutational status of APC and p53. To provide further evidence that TQ induces cell apoptosis rather than cell cytostasis, RKO cells were incubated with increasing TQ concentrations for 24 h and Annexin V/PI staining was performed. Indeed, TQ treatment led to a concentration dependent increase in apoptotic and dead cells measured by flow cytometry (Additional file 5: Figure S5).

TQ acts on the GSK-3β pathway
TQ was reported to inhibit proliferation and angiogenesis by suppressing ERK and AKT phosphorylation in HUVECs [18]. In RKO, ERK1/2 and AKT1 are highly phosphorylated. In our experiments, TQ did not have any effect on the phosphorylation status of ERK1/2 (Thr202/ Tyr204) nor AKT1 (Ser473) from 30' to 12h, despite successful inhibition of phosphorylation of ERK1/2 by the MEK1/2 inhibitor UO126 (Additional file 4: Figure S4C and D). A well-known downstream target of the PI3K/ AKT pathway and, alternatively, the RAS/RAF/MEK/ERK pathway, is the glycogen synthase kinase 3β (GSK-3β). The activity of GSK-3β is inversely correlated with its phosphorylation status at Ser9. When testing for GSK-3β phosphorylation upon TQ treatment, we observed a reduction of p-GSK-3β. GSK-3β appears to be an important molecular target of TQ, which may subsequently affect the stability of c-myc ( Figure 5).

TQ downregulates c-myc protein expression
Active GSK-3β is assumed to be an important kinase phosphorylating c-myc on Thr58 for subsequent ubiquitination by the F-box protein Fbw7 [26], a component of the SCF-class ubiquitin ligase (E3) complex, and degradation. Indeed, treatment with TQ reduces c-myc expression at 2 h and 24 h ( Figure 5A-C). The subcellular distribution of c-myc showed reduced expression of nuclear c-myc upon TQ treatment ( Figure 5A), which is in line with the reduced nuclear c-myc expression in polyps of APC Min mice ( Figure 3). To further support the mediation of c-myc reduction at protein level, c-myc mRNA was quantified. In fact, the expression of c-myc mRNA was not altered upon TQ treatment (Additional file 4: Figure S4B).
TQ reduces GSK-3β phosphorylation via inhibition of the MEK1/2 pathway rather than PI3K As a next step we verified if the reduced GSK-3β phosphorylation upon TQ treatment was dependent on the PI3K pathway, which is upstream from GSK-3β. We used the PI3K inhibitor LY294002 together with TQ to test for synergistic or additive effects. Treatment with TQ and LY294002 for 2 h and 4 h reduced the levels of GSK-3β phosphorylation in an additive manner (see also Figure 5D, E, Table 1 and Methods section for statistical explanation) indicating that TQ's effect on GSK-3β phosphorylation is independent of PI3K. As expected, total GSK-3β levels remained stable over time. Also, total c-myc protein was reduced more effectively at the 4 h point in time, if both substances have been added ( Figure 5D).
We then investigated the RAS/RAF/MEK pathway, which also influences GSK-3β Ser9 phosphorylation. MEK1/2 inhibition with UO126 and concomitant incubation with TQ did not show additional reduction of GSK-3β phosphorylation ( Figure 5E), assuming that both compounds are utilizing the same pathway. As stated above, pharmacological inhibition of MEK1/2 in RKO cells resulted in an abrogation of p-ERK1/2 phosphorylation that was not seen on treatment with TQ. Therefore, we assume that TQ carries out its function by inhibiting another kinase in this pathway.

Discussion
In this study we demonstrate that TQ, the main active component in the essential oil of Nigella sativa seeds, reduces the number of large polyps in the small intestine of APC Min mice. This effect was mainly due to induction of tumor cell-specific apoptosis. In addition, we found a shift of β-catenin to the membrane and a reduced nuclear c-myc expression in polyps. In RKO, TQ activates GSK-3β, which in turn might mediate phosphorylation and subsequent degradation of c-myc protein (model see Figure 6).
Black cumin seed oil is commonly used in traditional medicine and as a spice in countries with low prevalence for colorectal cancer [16]. It has been described to interfere with several tumor pathways. Gali-Muhtasib and others showed that TQ reduces the number of aberrant crypt foci and inhibits tumor growth in a 1,2-dimethylhydrazine-induced and a xenograft colon carcinogenesis mouse model, mainly by induction of apoptosis. Furthermore this group demonstrated that TQ decreases the invasive potential of a mouse colorectal cancer cell line [20]. Other publications showed that the antiproliferative and apoptotic effects of TQ are either mediated by p53dependent or independent mechanisms. The former could be shown for colorectal cancer cells, the latter only for osteosarcoma cells [27] and a myeloblastic leukemia cell line [28]. p53-null HCT116 cells were less sensitive to TQ-induced growth arrest and apoptosis compared to p53-wt HCT116 cells [21]. The reduced Figure 4 β-catenin expression in untreated, TQ-low, TQ-high, or piroxicam treated APC Min mice (n=8). Membranous (A), cytoplasmic (B), and nuclear (C) β-catenin IRSs were calculated for epithelial cells of the normal mucosa, small and large polyps. Bar graphs display mean β-catenin IRSs ± SD. TQ-high treated mice (E) showed an increase in membranous β-catenin expression in large polyps (ANOVA, Dunnett, 2-sided, *p<0.05) and a trend for an increase in small polyps (p=0.064) when compared to untreated (D) APC Min mice. Membranous β-catenin expression in piroxicam treated mice did not reach significance (p=0.094) due to a too small number of large polyps, although a membranous relocalization of the protein was present as in the TQ-high treated mice. Cytoplasmic expression of β-catenin increased upon TQ-low and TQ-high (p=0.064) treatment in the normal mucosa. Magnification: 40x (left panel) and 400x (right panel). Panel F shows the negative control sample without the primary antibody (F). apoptosis in p53 −/− HCT116 cells is due to an upregulation of the DNA damage sensor CHK1, which is normally transcriptionally repressed by p53 [29]. This is in line with our observation that cells having functional p53, such as LoVo, HCEC-1CT, RKO and HCT116, are more sensitive to TQ induced cell death compared to cells with p53 mutations, as shown for DLD1 and HT29. Also, we could confirm in our model that one of the main chemopreventive pathways of TQ is the cancer cell-specific induction of apoptosis which is Figure 5 Effect of TQ on β-catenin and the glycogen synthase kinase 3β (GSK-3β) pathway in RKO. Separating for membranous, cytoplasmic and nuclear protein fractions showed reduction of nuclear β-catenin and c-myc upon 30 μM TQ for up to 24 h and a short term increase of membranous and cytoplasmic β-catenin. α-tubulin is used as a total protein and cytoplasmic marker, fibrillarin as nuclear marker, and Na-K-ATPase as membranous marker (A). TQ reduces the amount of p-GSK-3β Ser9, the inactive form of the protein, at concentrations of 10 and 30 μM, shown by Western blot (B) and fluorescence microscopy (C; top panel, green: p-GSK-3β, blue: DAPI; LSM 700). Total GSK-3β protein levels are constant over time (B; C, bottom panel). Total c-myc levels were reduced at 30 μM TQ up to 24 h, an effect that was not seen for 10 μM (B). 30 μM TQ and 30 μM PI3K inhibitor LY294002 additively reduced the amount of p-GSK-3β in total protein lysates. 5 independent Western blots had been carried out and the ratio of p-GSK3-3β levels and total GSK3-3β protein has been calculated with densitometry analysis (D, E and see also Table 1 for the statistical calculations). Treatment with TQ increased total β-catenin protein but reduced the amount of c-myc protein (D). Inhibition of MEK1/2 with 30 μM UO126 reduced p-GSK-3β, while concomitant incubation with TQ did not lead to an additive reduction of p-GSK-3β (F).
likely mediated by p53. TQ was shown to induce oxidative stress and subsequently H2AX phosphorylation in HCT116 cells, which was diminished in p53 −/− HCT116 [29]. The authors concluded that the changes in H2AX expression and the increased apoptosis correlated with enhanced mitochondrial ROS production. In addition to TQ's tumor cell-specific induction of apoptosis we observed that TQ reverses the Ki-67 positivity found in the cells of the intermediate zone of APC Min villi. This is indicative of lowered proliferation in the normal tissue. However, the number of Ki-67 was not altered in polyps. This result is in agreement with previous studies that assessed Ki-67 staining in HCT116 xenograft tumors [20]. RKO cells were treated with TQ, PI3K inhibitor or with a combination of both substances for 2 h or 4 h to test for syngergistic or additive effects on the dephosphorylation of GSK-3β. 5 different Western blots were performed and the relative ratios of p-GSK-3β and total GSK-3β protein levels were calculated using the program Image J. Mathematically, when effects are measured as proportions, the additive effect of two substances is the multiplication of the two values resulting from the single treatments. The mathematical product of both treatments is only minimally deviating from the actual values for the double treatment resulting from the Western blots. The null-hypothesis of no mean difference cannot be rejected at either time point (T-test, p-value *0.56 (2 h) and **0.34 (4 h)). While this result is in itself no proof of additivity, a closer inspection of the calculated differences shows that the assumption of additivity is indeed reasonable: In experiments 1, 3 and 5 the differences are very close to zero. Experiments 2 and 5 show stronger deviations, however in opposite directions and the resulting mean differences are close to zero for both time points. In vitro analysis of TQ`s influence on the Ser9 phosphorylation status of GSK-3β, an important kinase for cell differentiation and apoptosis, showed reduced phosphorylation and activation of the protein. There is increasing evidence that c-myc Thr58 is phosphorylated by active GSK-3β, thereby regulating its stability by mediating Fbw7-driven turnover [26]. Enhanced GSK-3β activity by TQ might lead to phosphorylation and subsequent degradation of c-myc. The cause of reduced Ser9 phosphorylation in RKO cells is still elusive. Concomitant treatment with TQ and the PI3K inhibitor LY294002 leads to an additive reduction of p-GSK-3β, assuming that, in colon cells, TQ does not utilize the PI3K/AKT pathway, as shown for prostate cancer [18]. Instead, inhibition of MEK1/2 with UO126 reduced p-ERK1/2 and p-GSK-3β levels and concomitant treatment with TQ did not reveal an additive effect on GSK-3β phosphorylation, indicating that TQ mediates its effect over the RAS/RAF/MEK pathway rather than ERK1/2. As shown in vivo and in vitro TQ leads to membranous β-catenin translocation, which is probably also mediated by GSK-3β activation. This is in line with a study conducted by Li et. al. showing that active GSK-3β interacts with MUC1, thereby releasing MUC1 bound β-catenin, which then associates with membranous E-cadherin [30] ( Figure 6).

Growth factor
One shortcoming of the study is the limited effect of TQ on tumor multiplicity. In this respect our results deviate from a chemically-induced colorectal cancer model [20]. TQ may have less activity in genetically-driven cancer models such as the APC Min mouse than in chemicallyinduced models. Another contributing factor might be the administration route of TQ (as part of the chow versus intraperitoneal injections). Also differences in the mutational status of APC may partially explain this phenomenon. We have therefore used RKO cells (APC wt) for in vitro studies.
The TQ concentration in seeds is between 0.12 and 0.25% (w/v) [31,32]. In the essential oil, which is about 0.41 -0.44% of the total seed, the TQ concentration is between 28 -57% [31]. The human equivalent dose of 375 mg/kg TQ in the diet corresponds to 3.6 mg/kg body weight, resulting in a total amount of 219 mg for a body weight of 60 kg [33]. Studies using TQ in H. pylori infection [34] or diabetes [35] report an uptake of up to 3 g Nigella sativa seeds/day for one to three months, respectively. Both studies report amelioration of the disease but no severe side effects or adverse effects on renal or hepatic functions, respectively. Assuming a seeds content of 0.2% TQ, this results in an effective dose of 6 mg TQ per day. In mice, no adverse effects except for a decrease in glucose levels was observed with TQ concentrations up to 90 mg/kg body weight [36]. The corresponding concentration in our study was 45 mg/kg body weight for TQ-high. Thus the concentrations used are likely well tolerated without major toxicity. A dose dependent decrease of GSH levels in vital organs was found only with higher concentrations (2-3 g/kg, p.o.), most probably due to formation of TQ conjugates with GSH (glutathionyl-dihydrothymoquinone), which might affect bioavailability as well as antioxidant properties of TQ [36,37].
In summary, TQ interferes with polyp progression in APC Min mice by inducing tumor cell-specific apoptosis and by modulating Wnt signaling through GSK-3β activation, β-catenin translocation and reduction of nuclear c-myc. TQ or Nigella sativa seed oil might be a useful nutritional supplement to complement current chemoprevention in FAP.

Mice, genotyping, treatment, and colonoscopy
Heterozygous male C57BL/6J-Apc Min /J (APC Min ) mice (The Jackson Laboratory, Bar Harbor, ME) were bred with wild-type female C57BL/6J mice. Apc Min mice harbor a nonsense mutation at nucleotide 2549 in the murine homolog of the human APC gene [38]. Genotyping was performed from mouse tail DNA using primerintroduced restriction analysis-polymerase chain reaction with APC primers (APC for : 5'-TCT CGT TCT GAG AAA GAC AGA AGC T-3'; APC rev : 5'-TGA TAC TTC TTC CAA AGC TTT GGC TAT-3') and HindIII digestion according to Musteanu et.al [39]. 4-6 week old female and male APC Min mice were housed at the Institute of Biomedical Research (Medical University Vienna, Vienna). Mice were kept under 12 hour light/dark cycles. Chow and water were available ad libitum. Animals were weighed weekly and the amount of food intake was documented. All animal experiments were performed in accordance with the Austrian and European law, defined by the Good Scientific Practice guidelines of the Medical University Vienna (animal ethics approval number: BMWF-66.009/0113-II/10b/2010).
The animals were randomly divided into 4 groups and treated over a period of 12 weeks. Chemopreventive substances were added to a commercial rodent diet (C1000, Altromin, Lage, Germany) as follows: 37.5 mg/kg chow of TQ (274666, Sigma Aldrich) referred to as TQ-low (n=13) and 375 mg/kg chow of TQ referred to as TQhigh (n=16). 200 mg/kg chow piroxicam (P5654, Sigma-Aldrich) was used as a positive control (n=15) [40,41]. Mice fed the diet alone served as a negative control (n=17). Mice gaining less than 1.5 g of weight were excluded for analysis, as the chemopreventive substance was taken up through the diet. Mice with breast cancer, which is associated with the APC Min phenotype [42], were euthanized and excluded from analysis. At week 9 of treatment mice underwent colonoscopy. Briefly, mice were anesthetized with an i.p. injection of ketamine and xylazine, the colonoscope (Karl Storz, Tuttlingen, Germany) connected to an airpump (Eheim, Deizau, Germany) was inserted and the colon inflated [43]. During insertion and withdrawal of the colonoscope up to 3 cm the number of polyps was determined in real-time using a standard monitor and a video was recorded. Mice were euthanized, the intestine was dissected, flushed with PBS and 10% neutral buffered formalin, and coiled up to a Swiss roll [44]. Prior to paraffin embedding the intestine was fixed in neutral buffered formalin for 24 h.
Immunohistochemical staining was performed on paraffin-embedded mouse intestine using antibodies against β-catenin, Ki-67, c-myc and isotype specific (anti-mouse/anti-rabbit) biotinylated secondary antibody and standard staining procedures (details for antibodies see Additional file 6: Table S1). Slides were dried, dewaxed in xylene and rehydrated using a decreasing alcohol series. After blocking of endogenous peroxidase with 15% H 2 0 2 in methanol, antigen retrieval was performed in 10 mM citrate buffer, pH 6. Subsequently, slides were blocked in 2% horse serum, 3% BSA in TRIS buffer and endogenous IgG was blocked with Vector M.O.M. Blocking Reagent (MKB-2213, Vector Laboratories). Primary antibodies were incubated at 4°C overnight, followed by biotinylated secondary antibody and avidinbiotin-HRP complex (Vectastain ABC Kit, PK-6100; Vector Laboratories). Staining was visualized using 3,3′diaminobenzidine (32750; Fluka) and nuclear counterstaining was performed using hematoxylin. Slides were dehydrated and embedded in Histofluid (6900002; Marienfeld, Lauda Koenigshofen, Germany). Images were recorded at 40× to 400× magnification using an Olympus BH-2 microscope and an Olympus E330 digital camera. Immunoreactivity was independently scored by two investigators. A standardized immunoreactivity scoring system was modified to evaluate both the intensity of immunohistochemical staining and the proportion of cells stained [46]. The staining intensity was classified into 0 (no staining), 1+ (weak), 2+ (moderate), 3+ (strong) and the percentage of positive cells was recorded (0-100%), resulting in a highest value of 300. Total counts were divided by 25 to reach a maximum immunoreactivity score (IRS) of 12, to be comparable to other publications using a scoring system reaching a maximum of 12 [46]. IRS was calculated for small polyps, large polyps and normal mucosa of 6 different fields of view (FoV). C-myc staining was evaluated by scoring the overall staining intensity per polyp (n=16) as follows: negative (0), weakly (1+) or moderately positive (2+).

Immunocytochemistry
Fluorescence immunocytochemistry was performed using antibodies against p-GSK-3β and GSK-3β. Cells were fixed, permeabilized, blocked and incubated with the primary antibody overnight at 4°C. For protein visualization secondary AlexaFluor 488 antibody was used (Additional file 6: Table S1). Nuclear counterstaining was performed using Vectashield mounting medium with DAPI. Images were scanned 400x magnification on a LSM 700 (Zeiss).
Annexin V/propidium iodide staining 1.5×10 5 RKO cells were seeded in a 6-well plate and grown for 24 h. Cells were treated with 0-120 μM TQ for 24 h or 5 μg/ml Actinomycin D (positive control) for 16 h and processed for Annexin V-FITC/propidium iodide (PI) staining according to the manufacturer's protocol (eBioscience, BMS500FI). Annexin/PI positive cells were measured with a Cell Lab Quanta SC Flow Cytometer (Beckman Coulter) and analyzed with Quanta Analysis software.
Cell viability assay 1.5×10 4 cells/well were seeded in a 96-well microtiter plate and grown for 24 h. After a 24 h-treatment with 0-500 μM TQ, cells were incubated for 3 h with 0.5 mg/mL MTT (Sigma, M5655). The cells were subjected to a 1:1 ethanol/DMSO treatment to dissolve formazan crystals. The intensity of the solubilized crystals was measured colorimetrically at 570 nm (Anthos 2010). Each measurement was performed in biological quadruplicates.

Quantitative real time-PCR
Total RNA was isolated with TRIZOL reagent (Biorad) and was reverse transcribed to cDNA using the Thermoscript RT-PCR System (11146-024; Invitrogen) according to the manufacturer's protocols. Quantitative real time-PCR was carried out in duplicates using Fast SYBR Green Master Mix (AB; 4385612) and c-myc primers (Additional file 6: Table S2). Data were normalized to two endogenous controls GAPDH and β-actin (QIAGEN). Relative expression levels of the transcripts were calculated using the comparative CT method [51].

Statistics
Statistical analysis was performed using SPSS software version 17.0. Polyp number, size, and the number of apoptotic cells (TUNEL assay) were analyzed using univariate analysis of variance (ANOVA). Results were corrected by Dunnett (2-sided). Paired T-test was used to analyze the number of apoptotic cells in the normal mucosa compared to neoplastic tissue within one group. Immunohistochemistry data for β-catenin, Ki-67 and cmyc staining were analyzed using ANOVA. Results were corrected by Dunnett (2-sided). To investigate the additive or synergistic nature of the effects of TQ and PI3Kinhibitor on GSK-3β dephosphorylation the expected effects under the assumption of additivity were compared to the actual effects measured in the double treatment experiments. Since the effects are measured as proportions, an additive nature of effects means that the expected double effect is the product of the single effects. This product was calculated for each experiment and point in time and compared to the measured effect of the double treatment by calculating the difference. The resulting differences are reported for all experiments. The mean and the standard deviation of the differences were calculated for 2 h and 4 h points in time. Paired T-tests were performed to test the null-hypotheses of the mean differences being zero. p-values were considered as statistical significant if less than 0.05 (*p<0.05; **p<0.01; ***p<0.001). Data are expressed by mean and the 95% confidence interval for the mean (95% CI) or mean and standard deviation.
Additional file 2: Figure S2. Colonoscopy and small intestinal tumor number. Number of polyps/mouse detected during colonoscopy after 9 weeks of treatment, reaching 2-3 cm into the colon (A). Representative images of normal mucosa (I) and polyps of different size (II-IV) are shown (B). For TQ-low (n=14) and piroxicam (n=14) a significantly reduced number of polyps compared to untreated (n=17) mice was found via colonoscopy. For TQ-high (n=17) a trend for reduction of polyps was seen. Every dot represents the number of polyps of a single mouse. *p<0.05; ANOVA, Dunnett 2-sided. Total number of polyps in the SI of APC Min mice (C). Bars show mean number (± SD) of SI polyps/mouse. Piroxicam decreased the total number of SI polyps. For TQ-high there was a trend for reduction of total SI polyps. ***p<0.001; ANOVA, Dunnett 2-sided. Additional file 4: Figure S4. MTT assay. TQ exerts different effects on cell viability in cancer cell lines having various mutational backgrounds. Colon cancer cells and 1CT normal diploid human colon epithelial cells were incubated with TQ at different concentrations in quadruplicates for 24 h. The absorbance values (relative OD), expressed as means, compared to untreated cells were measured by MTT assay and IC50 concentrations were calculated (A). TQ does not exert its effect on the p-ERK1/2 and p-AKT1 (Ser473) pathway, but inhibition of MEK1/2 with the compound UO126 at 30 μM abrogated p-ERK1/2 phosphorylation (Thr202/Tyr204) between 30' and 12 h and reduced to a much lower extent p-AKT (Ser473), shown for whole cell lysates of RKO cells (B and D). Colon cancer cell lines RKO and HT29 were treated with 30 μM TQ for indicated times or left untreated (control). C-myc transcript levels stay constant upon TQ treatment. Relative mRNA expression levels of c-myc were calculated with GAPDH and β-actin as endogenous controls using