Metformin Suppresses Cancer Stem Cells through AMPK Activation and Inhibition of Protein Prenylation of the Mevalonate Pathway in Colorectal Cancer

Simple Summary Tumor suppressing effect of metformin has been reported, and one of mechanism of this effect is suppression of cancer stem cells (CSCs). However, detailed mechanism of metformin-induced CSC-inhibitory effect has not been known. We demonstrated that the CSC-suppressive effect of metformin was associated with AMPK activation/mTOR inhibition and repression of protein prenylation through suppression of mevalonate pathway in colorectal cancer. Further studies would be needed to investigate cross-reactions with other mechanisms of the antitumor effect of metformin, and clinical impact of metformin should be considered as chemopreventive or adjunctive treatment for colorectal tumor. Abstract Metformin is a well-known AMPK (AMP-activated protein kinase) activator that suppresses cancer stem cells (CSCs) in some cancers. However, the mechanisms of the CSC-suppressing effects of metformin are not yet well understood. In this study, we investigated the CSC-suppressive effect of metformin via the mevalonate (MVA) pathway in colorectal cancer (CRC). Two colorectal cancer cell lines, HT29 and DLD-1 cells, were treated with metformin, mevalonate, or a combination of the two. We measured CSC populations by flow cytometric analysis (CD44+/CD133+) and by tumor spheroid growth. The expression of p-AMPK, mTORC1 (pS6), and key enzymes (HMGCR, FDPS, GGPS1, and SQLE) of the MVA pathway was also analyzed. We investigated the effects of metformin and/or mevalonate in xenograft mice using HT29 cells; immunohistochemical staining for CSC markers and key enzymes of the MVA pathway in tumor xenografts was performed. In both HT29 and DLD-1 cells, the CSC population was significantly decreased following treatment with metformin, AMPK activator (AICAR), HMG-CoA reductase inhibitor (simvastatin), or mTOR inhibitor (rapamycin), and was increased by mevalonate. The CSC-suppressing effect of these drugs was attenuated by mevalonate. The results of tumor spheroid growth matched those of the CSC population experiments. Metformin treatment increased p-AMPK and decreased mTOR (pS6) expression; these effects were reversed by addition of mevalonate. The expression of key MVA pathway enzymes was significantly increased in tumor spheroid culture, and by addition of mevalonate, and decreased upon treatment with metformin, AICAR, or rapamycin. In xenograft experiments, tumor growth and CSC populations were significantly reduced by metformin, and this inhibitory effect of metformin was abrogated by combined treatment with mevalonate. Furthermore, in the MVA pathway, CSC populations were reduced by inhibition of protein prenylation with a farnesyl transferase inhibitor (FTI-277) or a geranylgeranyl transferase inhibitor (GGTI-298), but not by inhibition of cholesterol synthesis with a squalene synthase inhibitor (YM-53601). In conclusion, the CSC-suppressive effect of metformin was associated with AMPK activation and repression of protein prenylation through MVA pathway suppression in colorectal cancer.

Based on our previously reported data [8], we selected HT29 and DLD-1 cells from the available metformin-sensitive CRC cell lines. We confirmed decreased mRNA expression of the CSC markers Lgr5, CD44, and CD133 ( Figure 1A). In addition, in another public transcript data set (GSE76342) for the colon cancer cell line LoVo with or without metformin treatment, we found that metformin inhibited expression of the CSC markers Lgr5, ASCL2, EPHB3, OLFM4, BMI1, Lrig1, TERT, CD44, and CD133 ( Figure 1B). Because (E,F) In tumor sphere cultures treated with metformin (10 mM), simvastatin (2 µM), AICAR (1 mM), or rapamycin (200 nM) for 7 days, the number of tumor spheres (≥200 µm in diameter) was counted and compared to the counts in control cultures. Data are expressed as the mean ± standard error of three independent experiments; * p < 0.05, ** p < 0.01, and *** p < 0.001 (compared with control).
In addition, in another public transcript data set (GSE76342) for the colon cancer cell line LoVo with or without metformin treatment, we found that metformin inhibited expression of the CSC markers Lgr5, ASCL2, EPHB3, OLFM4, BMI1, Lrig1, TERT, CD44, and CD133 ( Figure 1B). Because one of the well-known pathways of metformin action is the AMPK-dependent mTOR pathway, we examined the expression of activated AMPK and mTOR signaling molecules after treatment with metformin. Metformin treatment increased p-AMPK and decreased p-S6 expression ( Figure 1C). Using flow cytometric analysis, we confirmed that the CSC population was significantly decreased by Cancers 2020, 12, 2554 4 of 15 metformin, AICAR (AMPK activator), simvastatin (HMG-CoA reductase inhibitor), and rapamycin (mTOR inhibitor) ( Figure 1D). In addition, in tumor sphere culture experiments, tumor sphere formation was significantly decreased by treatment with these drugs ( Figure 1E,F).

Metformin Suppresses Key Enzymes of the Mevalonate Pathway
Some reports have shown potent effects of metformin on lipid and cholesterol biosynthesis in cancer cells [11,12]. Therefore, we postulated that metformin may inhibit HMG-CoA reductase (HMGCR), as well as several important enzymes of the MVA pathway, in CRC. In the same data set shown in Figure 1B, we found that metformin reduced key enzymes of the MVA pathway, including HMGCR, mevalonate kinase (MVK), phospho-mevalonate kinase (PMVK), mevalonate decarboxylase (MVD), FDPS, GGPS, and SQLE ( Figure 2A).
Cancers 2020, 12, x 4 of 16 one of the well-known pathways of metformin action is the AMPK-dependent mTOR pathway, we examined the expression of activated AMPK and mTOR signaling molecules after treatment with metformin. Metformin treatment increased p-AMPK and decreased p-S6 expression ( Figure 1C). Using flow cytometric analysis, we confirmed that the CSC population was significantly decreased by metformin, AICAR (AMPK activator), simvastatin (HMG-CoA reductase inhibitor), and rapamycin (mTOR inhibitor) ( Figure 1D). In addition, in tumor sphere culture experiments, tumor sphere formation was significantly decreased by treatment with these drugs ( Figure 1E,F).

Metformin Suppresses Key Enzymes of the Mevalonate Pathway
Some reports have shown potent effects of metformin on lipid and cholesterol biosynthesis in cancer cells [11,12]. Therefore, we postulated that metformin may inhibit HMG-CoA reductase (HMGCR), as well as several important enzymes of the MVA pathway, in CRC. In the same data set shown in Figure 1B, we found that metformin reduced key enzymes of the MVA pathway, including HMGCR, mevalonate kinase (MVK), phospho-mevalonate kinase (PMVK), mevalonate decarboxylase (MVD), FDPS, GGPS, and SQLE ( Figure 2A). To demonstrate increased activity of the MVA pathway in CSCs of CRC, we cultured tumor spheres and found significantly elevated protein and mRNA levels of key enzymes of the MVA pathway, including HMGCR, FDPS, GGPS1, and SQLE, in tumor spheroids compared to 2D adherent cultured cells ( Figure 2B,C). These findings suggest that the MVA pathway may have an important role in CRC tumorigenesis by promoting CSCs of CRC. In addition, metformin significantly reduced expression of key enzymes of the MVA pathway that were upregulated in 3D tumor spheroid cultures of CRC cells ( Figure 2B,C).
We asked whether other AMPK/mTOR regulators can inhibit enzymes of the MVA pathway. Thus, we confirmed significant decreases in mRNA levels of these key enzymes of the MVA pathway in tumor spheres after 7 days of treatment with metformin, AICAR, or rapamycin. We also tested simvastatin on tumor spheres of CRC cells, and found that reduction of these key enzymes of the MVA pathway by simvastatin was relatively weak and inconsistent compared to the effects of metformin ( Figure 2D). Moreover, treatment with mevalonate promoted expression of these key enzymes of the MVA pathway ( Figure 2D), resulting in an increased proportion of CSCs among CRC cells ( Figure 2E). In addition, mevalonate treatment attenuated the suppressive effect of metformin on CSCs of CRC and the formation of tumor spheroids ( Figure 2E,F). This reversal effect of mevalonate was also observed in treatment with AICAR, simvastatin, and rapamycin in both cell lines ( Figure 2E).
Taken together, these results demonstrate that the MVA pathway was activated in CSCs and that metformin, AMPK activation, and mTOR inhibition suppressed CSCs through inhibiting the expression of key enzymes of the MVA pathway. To demonstrate increased activity of the MVA pathway in CSCs of CRC, we cultured tumor spheres and found significantly elevated protein and mRNA levels of key enzymes of the MVA pathway, including HMGCR, FDPS, GGPS1, and SQLE, in tumor spheroids compared to 2D adherent cultured cells ( Figure 2B,C). These findings suggest that the MVA pathway may have an important role in CRC tumorigenesis by promoting CSCs of CRC. In addition, metformin significantly reduced expression of key enzymes of the MVA pathway that were upregulated in 3D tumor spheroid cultures of CRC cells ( Figure 2B,C).
We asked whether other AMPK/mTOR regulators can inhibit enzymes of the MVA pathway. Thus, we confirmed significant decreases in mRNA levels of these key enzymes of the MVA pathway in tumor spheres after 7 days of treatment with metformin, AICAR, or rapamycin. We also tested simvastatin on tumor spheres of CRC cells, and found that reduction of these key enzymes of the MVA pathway by simvastatin was relatively weak and inconsistent compared to the effects of metformin ( Figure 2D). Moreover, treatment with mevalonate promoted expression of these key enzymes of the MVA pathway ( Figure 2D), resulting in an increased proportion of CSCs among CRC cells ( Figure 2E). In addition, mevalonate treatment attenuated the suppressive effect of metformin on CSCs of CRC and the formation of tumor spheroids ( Figure 2E,F). This reversal effect of mevalonate was also observed in treatment with AICAR, simvastatin, and rapamycin in both cell lines ( Figure 2E).
Taken together, these results demonstrate that the MVA pathway was activated in CSCs and that metformin, AMPK activation, and mTOR inhibition suppressed CSCs through inhibiting the expression of key enzymes of the MVA pathway.

Metformin Reduced CSCs through Inhibition of Protein Prenylation, rather than Cholesterol Synthesis, in the MVA Pathway
We further investigated the specific final processes of the MVA pathway associated with metformin-induced suppression of CSCs by performing experiments using inhibitors of protein prenylation, such as farnesylation and geranylgeranylation, and of cholesterol synthesis. We used the farnesyl transferase inhibitor FTI-277, the geranylgeranyl transferase inhibitor GGTI-298, and the squalene synthase inhibitor YM-53601. To evaluate the CSC-specific effects and exclude the possibility of direct cellular toxicity of these inhibitors, we selected concentrations of inhibitors showing CSC-suppressive effects without significant cell death. For example, higher concentrations of YM-53601 (>10 µM) induced significant cell death without changes in CSC population (data not shown). FTI-277 and GGTI-298 induced suppression of the CSC population in a dose-dependent manner, similar to metformin treatment ( Figure 3A). Moreover, combined treatment of FTI-277 and GGTI-298 decreased the CSC population even further ( Figure 3A). However, YM-53601 did not show a significant effect on CSC populations ( Figure 3A). Furthermore, in tumor sphere formation assays ( Figure 3B,C), treatment with FTI-277, GGTI-298, or YM-53601 produced results similar to those observed in flow cytometric analyses of CSC populations. In addition, to confirm the decrease of protein prenylation by metformin, we performed Western blot analysis to identify the change of shifted prenylated protein bands of RAS and Ral A by farnesylation and geranylgeranylation, respectively. Although the shifted bands were weak in some conditions, we found the decrease of shifted prenylated protein bands of RAS and Ral A (Supplementary Figure S1). These results demonstrated that metformin may suppress CSCs through inhibition of protein prenylation, via farnesylation and geranylgeranylation, rather than through cholesterol synthesis by SQLE.

Tumor-Suppressing Effect of Metformin in a Murine Xenograft Model Was Reversed by Additional Treatment with Mevalonate
We performed in vivo xenograft experiments to evaluate the effect of mevalonate on metformin-induced tumor suppression. At 14 days after implantation of HT29 cells, we treated mice with mevalonate, metformin, or a combination of mevalonate and metformin, by intraperitoneal injection for 21 days ( Figure 4A). In the metformin-treated group, tumor growth was suppressed by 20% compared to the control group. Treatment with mevalonate alone showed a trend toward further tumor growth relative to the control group; the addition of mevalonate to metformin treatment induced a significant increase in tumor growth compared to metformin treatment alone, indicating that mevalonate could reverse the tumor-suppressing effect of metformin ( Figure 4B-D). In immunohistochemical (IHC) staining for the CSC markers CD44 and CD133, the metformin treatment group showed significantly decreased IHC scores for CD44 and CD133; combination treatment of metformin and mevalonate induced significant increases in both CSC markers compared to metformin treatment alone ( Figure 4E,F). Moreover, key enzymes of protein prenylation of the MVA pathway, FDPS and GGPS1, were suppressed by metformin, and this suppressive effect was reversed by addition of mevalonate ( Figure 4G,H). From these results, we confirmed that mevalonate could reverse the CSC-suppressing effect of metformin in the in vivo xenograft model, and metformin-induced CSC suppression could be associated with suppression of key enzymes of the MVA pathway. In addition, we performed Ki67 staining to show antiproliferative effect of metformin, and found a significant decrease of Ki67 staining by treatment of metformin (Supplementary Figure S2). Therefore, the tumor-suppressing effect of metformin would not depend on only suppression of CSCs. However, because the CSCs are important for tumor initiation, persistent progression, and metastasis, and the effect of metformin was more prominent on CSC population in our previous data [8], we focused on CSC suppression by metformin.

Discussion
Several previous studies revealed the inhibitory effect of metformin on CSCs [8,9], but the detailed mechanism has not been well defined. In this study, we identified that metformin serves as a negative regulator of the MVA pathway and exerts CSC-suppressive effects in CRC through the inhibition of protein prenylation processes of the MVA pathway. Regarding the relationship among tumor growth, the cholesterol pathway, and metformin, Sharma at al. showed increased cholesterol levels and elevated expression of certain cholesterol regulatory genes in malignant breast tumor tissues, such as HMGCR, LDLR, and SREBP1 [12]. They also demonstrated that metformin inhibited cholesterol levels in breast cancer MDA-MB-231 cells, with decreased expression of cholesterol regulatory genes. In addition, Wahdan-Alaswad et al. suggested that metformin suppressed triple-negative MDA-MB-231 cells by reducing fatty acid synthesis [10,11]. Furthermore, they showed that metformin targets the cholesterol biosynthesis pathway and stabilizes GM1 lipid rafts, reducing membrane EGFR (epidermal growth factor receptor) signaling and its activation in triple-negative breast cancer. Moreover, they demonstrated that the combination of metformin with the statin-mimetic methyl-β-cyclodextrin (MβCD) synergistically attenuates cholesterol biosynthesis and cell proliferation. In particular, they suggested that metformin might act as an HMGCR inhibitor. In our analysis using their public RNA-seq data for the metformin-treated cell line, we found that expression of most genes involved in the MVA pathway was reduced by metformin treatment (Figure 2A). Several other studies also demonstrated that metformin may inhibit HMGCR and SREBP1 in different cell types [18][19][20]. However, the detailed relationship between metformin-induced regulation of the MVA/cholesterol pathway and CSCs has not been defined.
Our study aimed to investigate how metformin affects the MVA pathway and whether this may be related to the inhibitory effect on CSCs. We showed that metformin inhibited CSCs in CRC cells, with concomitant decreases in MVA pathway enzymes, including HMGCR, FDPS, GGPS1, and SQLE. We observed that the inhibitory effect of metformin on CSCs and MVA pathway enzymes was reversed by addition of mevalonate. The elevated expression of these MVA pathway enzymes was related to increased CSCs of CRC, suggesting a close relationship between CSCs and activated MVA pathway.
Interestingly, we demonstrated that CSCs could be reduced through the inhibition of protein prenylation, but not by inhibition of cholesterol synthesis. The prenylation of Ras and Ral/Rho proteins via MVA pathway-dependent farnesylation and geranylgeranylation, respectively, is important for their function. The Ras, Ral, and Rho families are associated with many tumor characteristics, such as invasive growth, cell survival, and three-dimensional growth, [21,22], and play critical roles in tumor development, progression, and metastasis, including activation of CSCs [23][24][25]. Therefore, inhibition of protein prenylation via the MVA pathway might be an important target for CSC suppression. In addition, Freed-Pastor and Mizuno et al. reported that mutant p53 upregulated expression of MVA pathway enzymes [17]. The tumor characteristics of metabolic subtypes of CRC, such as consensus molecular subtype 3 or KRAS mutations, may also be related to MVA pathway activation.
Metformin is a well-known AMPK activator, and activation of AMPK inhibits mTOR signaling and energy-consuming pathways [26,27]. Ching and Abraham et al. reported that activation of AMPK inhibits acetyl-CoA carboxylase (ACC) activity, which is related to fatty acid synthesis [28]. Many studies have reported that metformin induces anticancer activity by AMPK activation and mTOR inhibition. Thus, we also investigated the relationship between AMPK/mTOR regulators and the MVA pathway, and confirmed that an AMPK activator and an mTOR inhibitor reduced CSC populations; these effects were reversed by adding mevalonate. Moreover, the AMPK activator and mTOR inhibitor decreased the expression of mediator enzymes of the MVA pathway, suggesting regulation of this pathway by AMPK/mTOR signaling. In addition, we found that metformin-resistant cells (SW620) did not show significant change of AMPK/mTOR signals and reduction of key enzymes expression of the MVA pathway by treatment of metformin (Supplementary Figure S3).
Statin is a drug that lowers cholesterol through HMGCR inhibition and can decrease cancer risk [29]. We also showed that statin was a potent inhibitor of HMGCR and had a CSC-suppressive effect. However, the inhibitory effects of statin on key enzymes of the MVA pathway were weaker and less consistent than those of metformin, suggesting the importance of additional effects of metformin on the MVA pathway via AMPK activation and mTOR inhibition. Based on the mechanism of CSC suppression by inhibition of protein prenylation, these candidate drugs, such as metformin and AMPK activators, could be useful for adjunctive treatment in chemotherapy for advanced CRC or in chemoprevention for high-risk CRC groups if they have a wide margin of safety.
In summary, we investigated the detailed relationship between metformin-induced suppression of CSCs and the MVA pathway. Metformin-induced CSC suppression could occur through AMPK activation and protein prenylation inhibition, rather than through inhibition of cholesterol synthesis via the MVA pathway ( Figure 5). However, because metformin has many molecular mechanisms of antitumor effect, we could not elucidate the detailed interaction between prenylation-dependent and other direct and indirect mechanisms of metformin-induced antitumor or CSC suppression. Further studies would be needed to investigate these cross-reactions among direct and indirect mechanisms of the CSC-suppressing effect of metformin. consistent than those of metformin, suggesting the importance of additional effects of metformin on the MVA pathway via AMPK activation and mTOR inhibition. Based on the mechanism of CSC suppression by inhibition of protein prenylation, these candidate drugs, such as metformin and AMPK activators, could be useful for adjunctive treatment in chemotherapy for advanced CRC or in chemoprevention for high-risk CRC groups if they have a wide margin of safety.
In summary, we investigated the detailed relationship between metformin-induced suppression of CSCs and the MVA pathway. Metformin-induced CSC suppression could occur through AMPK activation and protein prenylation inhibition, rather than through inhibition of cholesterol synthesis via the MVA pathway ( Figure 5). However, because metformin has many molecular mechanisms of antitumor effect, we could not elucidate the detailed interaction between prenylation-dependent and other direct and indirect mechanisms of metformin-induced antitumor or CSC suppression. Further studies would be needed to investigate these cross-reactions among direct and indirect mechanisms of the CSC-suppressing effect of metformin.

Cell Lines and Culture Conditions
HT29 and DLD-1 colorectal cancer cell lines were purchased from the American Type Culture Collection (Manassas, VA, USA). Cell lines were maintained in Dulbecco modified Eagle's medium (DMEM; Invitrogen, Carlsbad, CA, USA) supplemented with 10% fetal bovine serum (Gibco, Franklin Lakes, NJ, USA) and 1% penicillin/streptomycin (Invitrogen, Carlsbad, CA, USA) at 37 • C in 5% CO 2 .

Tumor Sphere Culture Assay
HT29 and DLD-1 cells (2000 or 4000 cells per well) were plated in 24-well ultra-low adhesive plates (Corning Incorporated, Corning, NY, USA) in sphere formation medium with drugs for 5 or 7 days. The sphere formation medium was serum-free DMEM-F12 supplemented with B27 (Life Technologies, Carlsbad, CA, USA), 20 ng/mL epidermal growth factor, 10 ng/mL basic fibroblast growth factor (R & D Systems, Minneapolis, MN, USA), 1% penicillin/streptomycin, and 2 mM L-glutamine (Life Technologies, Carlsbad, CA, USA). Cells were incubated in a 5% CO 2 chamber at 37 • C, and 500 µL of the culture medium was changed every 48 h. After 5 or 7 days, the number of tumor spheres was counted under a microscope (Olympus, Tokyo, Japan, BX51 microscope).

Microarray Data Analysis Using a Public Data Set
We performed a microarray data analysis of GSE76342 using the limma package in R. Limma is an R/Bioconductor software package that provides an integrated solution for analyzing data from gene expression experiments. CEL files have been deposited to the Gene Expression Omnibus database (http://www.ncbi.nlm.nih.gov/geo) under the accession number GSE76342.

Flow-Cytometric Analysis and Fluorescence-Activated Cell Sorting (FACS)
HT29 and DLD-1 were plated in 6-well plates at a density of 2 × 10 5 cells/well and treated with metformin and other reagents. After 48 h, the prepared cells were detached with Accutase (Millipore, Billerica, MA, USA) and resuspended in FACS buffer (1 × PBS, 1% bovine serum albumin, and 2 mM ethylenediaminetetraacetic acid). Primary antibodies against CSC markers (PE-conjugated anti-CD133 and FITC-conjugated anti-CD44) were added. Specific information of antibodies used in the study was given in Supplementary Table S2. Samples were incubated for 10 min at 4 • C, washed with FACS buffer, and subjected to flow cytometry for analysis using a FACSVerse (BD Biosciences, San Diego, CA, USA) coupled to a computer with BD FACSuite software.

RNA Extraction and qPCR
Total RNA was isolated using TRIZOL Reagent (Invitrogen). Equal amounts of cDNA were synthesized using the reverse transcription 5× Master Pro Mix (ELPISBIO, Daejeon, Korea), and mixed with 2× SYBR Green with high ROX (ELPISBIO). qPCR reactions were performed using the gene-specific primers in Supplementary Table S1. All qPCRs performed using SYBR Green were conducted at 55 • C for 10 min, 95 • C for 10 min, and 40 cycles of 95 • C for 15 s and 60 • C for 1 min. The specificity of the reaction was verified by melt curve analysis.

Western Blotting
Prepared cells were lysed using a protein extraction solution (iNtRON Biotechnology, Gyeonggi, Korea). After protein quantification, 20 µg portions of protein extracts were fractionated using 12% or 15% sodium dodecyl sulfate polyacrylamide gel electrophoresis and transferred to polyvinylidene fluoride membranes (Bio-Rad, Hercules, CA, USA). After blocking with 10% skim milk, membranes were incubated with primary antibodies overnight at 4 • C. Subsequently, membranes were incubated with secondary antibodies for 1 h at room temperature. Proteins were detected using an ECL Western blotting detection kit (Amersham Biosciences, Freiburg, Germany) and light was captured on Kodak image film.

In Vivo Mouse Xenograft Experiments
Six-week-old male BALB/c athymic nude mice were purchased from Central Lab (Seoul, Korea) and acclimated for 1 week. Mouse experiments were performed in accordance with protocols approved by the Committee on Care and Use of Laboratory Animals of Yonsei University College of Medicine (Seoul, Korea) and according to institutional guidelines and policies.
HT29 cells were suspended in Matrigel (BD Bioscience) at a density of 1 × 10 6 cells/200 µL, diluted 1:1 in PBS, and subcutaneously injected into both flanks of the mice. After 2 weeks, the implanted mice were randomly divided into four groups (5 mice per group): control, metformin only, mevalonate only, and combination (metformin and mevalonate). All drugs were injected intraperitoneally (metformin: 250 mg/kg in 200 µL PBS; mevalonate: 10 mg/kg in 200 µL PBS) on a daily basis for 21 days. Vehicle control (200 µL of PBS) was injected intraperitoneally into the mice in the control, metformin-only, and mevalonate-only groups. Tumor sizes were measured each day using calipers, and tumor volumes were calculated based on the following formula: tumor volume = length × (width) 2 / 2. All mice were sacrificed 21 days after the first drug treatment and the tumor masses were dissected. The dissected tumors were placed in 4% paraformaldehyde (PFA) for immunohistochemistry (IHC) and were collected for further analysis.

Immunohistochemistry
IHC for CD133, CD44, GGPS1, and FDPS was performed on 4 µm sections of formalin-fixed, paraffin-embedded, dissected tumor samples. The paraffin-embedded sections were deparaffinized in xylene and rehydrated in gradually decreasing concentrations of ethanol. Antigen retrieval was performed using sodium citrate buffer (10 mM, pH 6.0) in a heated pressure cooker for 5 or 7 min. After incubation with 3% hydrogen peroxide for 30 min to block endogenous peroxidase activity, sections were incubated in a blocking reagent for 30 min at room temperature. Sections were incubated with primary antibodies overnight at 4 • C, followed by secondary antibody for 30 min at room temperature. After slides were developed with a Vectastain ABC kit (Vector Laboratories, Burlingame, CA, USA), immunodetection was performed using DAB solution (Dako, Carpinteria, CA, USA). After counterstaining with hematoxylin, IHC staining was evaluated by light microscopy and immunoactivity was assessed based on the proportion of immunostained tumor cells. Additional information on antibodies is shown in Supplementary Table S2. We measured the intensity of CSC markers (CD44 and CD133) and IHC scores using IHC profiles based on the Image J program [30]. The IHC score calculation was score = (number of pixels in a zone × score of the zone)/total number of pixels in the image. The score of the zone was assigned as 4 for high positive zones, 3 for positive zones, 2 for low positive zones, and 1 for negative zones.

Statistical Analysis
Statistical analyses were performed using IBM SPSS Statistics version 20.0 (IBM Co., Armonk, NY, USA). For the evaluation of two data sets, unpaired Student's t tests or Mann-Whitney tests were performed. To evaluate more than two groups, one-way or two-way analysis of variance was applied. Every experiment was conducted at least in triplicate to ensure reliability. All calculated p-values were two-sided and p < 0.05 was considered statistically significant.

Conclusions
We investigated mechanism of metformin-induced CSC-inhibitory effect in the field of tumor metabolism, and demonstrated that the CSC-suppressive effect of metformin was associated with AMPK activation/mTOR inhibition and repression of protein prenylation through suppression of mevalonate pathway in colorectal cancer. In the future, clinical usefulness of metformin might be considered as chemopreventive or adjunctive treatment for colorectal tumor.
Supplementary Materials: The following are available online at http://www.mdpi.com/2072-6694/12/9/2554/s1, Figure S1: Western blot analysis to identify change of shifted prenylated protein of RAS and Ral A, Figure S2: IHC to evaluate expression of Ki67 in xenograft tumor, Figure S3: Western blot and qPCR analysis of major signals and key enzymes in metformin-resistant SW620 cells, Table S1: Sequences of the primers used in qPCR, Table S2: List of antibodies, Table S3: List of drugs used in experiments.
Author Contributions: Y.S. and T.I.K. designed the research and analyzed data. Y.S. performed the majority of the experiments. J.K. and T.I.K. provided a critical review. Y.S. drafted and J.K., S.J.P., J.J.P., J.H.C., and W.H.K. contributed to parts of the manuscript. Y.S. and T.I.K. finalized the article. All authors have read and agreed to the published version of the manuscript.