There is increasing preclinical evidence suggesting that metformin, an antidiabetic drug, has anticancer properties against various malignancies, including colorectal cancer. However, the majority of evidence, which was derived from cancer cell lines and xenografts, was likely to overestimate the benefit of metformin because these models are inadequate and require supraphysiologic levels of metformin. Here, we generated patient-derived xenograft (PDX) lines from 2 colorectal cancer patients to assess the properties of metformin and 5-fluorouracil (5-FU), the first-line drug treatment for colorectal cancer. Metformin (150 mg/kg) as a single agent inhibits the growth of both PDX tumors by at least 50% (P < 0.05) when administered orally for 24 days. In one of the PDX models, metformin given concurrently with 5-FU (25 mg/kg) leads to an 85% (P = 0.054) growth inhibition. Ex vivo culture of organoids generated from PDX demonstrates that metformin inhibits growth by executing metabolic changes to decrease oxygen consumption and activating AMPK-mediated pathways. In addition, we also performed genetic characterizations of serial PDX samples with corresponding parental tissues from patients using next-generation sequencing (NGS). Our pilot NGS study demonstrates that PDX represents a useful platform for analysis in cancer research because it demonstrates high fidelity with parental tumor. Furthermore, NGS analysis of PDX may be useful to determine genetic identifiers of drug response. This is the first preclinical study using PDX and PDX-derived organoids to investigate the efficacy of metformin in colorectal cancer. Mol Cancer Ther; 16(9); 2035–44. ©2017 AACR.

Anticancer drug development is often impeded by a lack of suitable preclinical models that are highly predictive of therapeutic efficacy in humans. Most studies utilizing cell lines and xenografts are useful in the laboratory setting, but may not translate well to the clinical setting (1). The failure of many cell line xenografts to accurately predict the clinical efficacy of anticancer agents is due to their inability to reflect the complexity and heterogeneity of human tumors (2). Substantial genetic divergence exists between primary tumor and the corresponding cell line derived from that tumor, as compared with a direct transplant of tumor into mice (3–5). Patient-derived xenografts (PDX), in which patient-derived tumor fragments are directly implanted into immunocompromised mice through serial passaging, are an increasingly recognized tool for drug candidate evaluation. These PDXs are unlikely to have undergone extensive selection pressure typically associated with cell line establishment as manifested through genetic and molecular changes (3), and are thus valuable for drug studies.

We evaluated the performance of 5-fluorouracil (5-FU), the first-line treatment for colorectal cancer (6), along with the antidiabetic drug metformin. Epidemiologic studies have suggested the beneficial effects of metformin in cancer treatment among diabetic patients (7, 8). This was supported by preclinical studies, which demonstrate that metformin has antineoplastic activity in breast, colon, lung, and prostate cancer (9–14), spurring significant interest to repurpose metformin as an anticancer agent. Nonetheless, preclinical studies often involve supraphysiologic concentrations of metformin that are in excess of the therapeutic levels achieved in patients. In humans, the initial dose of metformin is 500 to 1,000 mg/daily, and subsequently, a maintenance dose of 2,000 mg/daily is required. Studies performed on cell lines require elevated doses of metformin (2–50 mmol/L) to elicit cellular response, while in vivo studies require up to 4 times the maximum clinical dose in humans of 2,550 mg/day (15). Thus, it is anticipated that good relevant models are necessary for physiologic modeling of metformin.

To the best of our knowledge, this is the first study evaluating the use of metformin on colorectal cancer PDX, a highly relevant system for drug candidate testing. Using next-generation sequencing (NGS), we interrogated the molecular profiles of these PDX to evaluate selection pressures initiated during the course of drug treatment. In addition, we aimed to identify molecular signatures that influence response to metformin. In parallel, we utilized colorectal cancer organoids to elucidate the mechanistic effect of metformin.

Cell culture maintenance, proliferation, and clonogenic assays

Cell lines were obtained in 2012 from ATCC directly while isogenic lines HCT116 p53+/+ (wild-type, WT) and p53−/− (knockout, KO) were kindly provided by Dr. Bert Vogelstein (John Hopkins University, MD, USA). Independent authentication of cells was not performed. Cells were cultured in DMEM with 10% fetal bovine serum and maintained in 37°C incubator supplemented with 5% CO2. The MycoAlert Mycoplasma Detection Kit (Lonza) was used to test for Mycoplasma contamination at defined intervals, and the tests indicate that cell lines were free of Mycoplasma throughout the study.

The CellTiter96 AQueous One Solution Assay (Promega) was used to assess proliferation of cells incubated with metformin (0–20 mmol/L) and/or 5-FU (10 μmol/L) in the presence of high (4,500 mg/L) or low (1,000 mg/L) glucose. For clonogenic assay, cells harvested from exponential phase cultures were plated in six-well plates. Seeding densities varied from 100 to 300 cells per well. After 7 days, cells were washed with PBS, fixed with methanol and stained with Coomassie blue. Assays were done in triplicate. Colonies were counted by two independent investigators.

Microsatellite instability (MSI) analysis

MSI was determined using the MSI Analysis Kit version 1.2 (Promega). Briefly, DNA from matched normal and tumor specimens was amplified using a panel of markers designed to amplify targeted microsatellite regions (BAT-25, BAT-26, NR-21, NR-24, and MONO-27). Fragment analysis was performed to assess the size of microsatellite repeats and determine MSI status.

Development of PDX

Fresh colorectal tumors were obtained from consenting patients at Concord Cancer Hospital in accordance with protocols approved by the Institutional Review Board. Tumors were transplanted into severe combined immunodeficient (SCID) female mice aged 4 to 8 weeks (InVivos Singapore) under the approved protocol by the Institutional Animal Care and Use Committee. Tumor specimens were cut into 2 to 4 mm3 pieces and implanted subcutaneously in the left flank of the mice. PDX was maintained by passaging into subsequent generations when tumor volumes reached approximately 1,000 to 1,500 mm3. For treatment studies, tumors were expanded in the left flanks of 5 to 12 mice (at least 5 evaluable tumors per arm). Mice were grouped into three arms: (i) control, (ii) metformin, or (iii) metformin and 5-FU when tumor volumes reached 100 to 200 mm3. Mice were treated daily with metformin dissolved in drinking water (150 mg/kg, orally) and/or weekly with 5-FU (25 mg/kg, intraperitoneally) for 24 days. Mice were monitored daily for signs of toxicity and tumor size was evaluated twice weekly by caliper measurements using the following formula: tumor volume in mm3 = [length × width2]/2. Tumor growth inhibition (TGI) was calculated as shown: TGI = [(tumor volume of control on day 24 − tumor volume of treated on day 24)/(tumor volume of control on day 24 − tumor volume of control on day 0)] × 100.

At endpoint, tumors were harvested. A portion of each tumor was fixed for histologic sections and hematoxylin and eosin (H&E) staining, while the remaining portions were snap-frozen or used for organoids derivation.

Organoids derivation and cellular respiration assay

Tumors from control mice were used for ex vivo measurement of cellular respiration. After removal of necrotic tissue, tumors were washed in HBSS, minced and digested in digestion media (RPMI media with 5X Penicillin-Streptomycin-Amphotericin B, 10 mmol/L HEPES, 300 U/mL Collagenase, 100 U/mL Hyaluronidase, and 10 μmol/L Y-27632) for 2 hours at 37°C. Digestion media were removed and digested tissues were resuspended in culture media (Advanced DMEM/F12 with 5× Penicillin–Streptomycin–Amphotericin B, 1× Glutamax, 1× B27, 1× N-2, 1 mmol/L N-Acetylcysteine, 50 ng/mL rhEGF, and 10 μmol/L Y-27632), and filtered through 100- and 40-μm filter. Organoids were collected from the 40- to 100-μm fraction, and the numbers were estimated.

Oxygen consumption rate (OCR) of organoids were measured using Seahorse Bioscience XF96 analyzer as described previously (16). Five hundred organoids were seeded overnight in a 96-well cartridge and pretreated with 0.1 mmol/L or 1.0 mmol/L metformin for 24 hours. Before measurement, the cells were washed and medium was replaced. Reagents from the Cell Mito Stress Kit (Seahorse Bioscience) were injected into each well sequentially and serve as control.

Next-generation sequencing

A customized Ion AmpliSeqColon Panel was designed to amplify 513 regions covering 40 genes implicated in colorectal cancer (Supplementary Table S1). Briefly, the customized panel was designed based on the multidimensional analysis of genomic profiles in colorectal cancer published by The Cancer Genome Atlas Network (17) and molecular subtype signatures (18–20). Hotspot mutations in genes of interest and other genes related to probable actionable drug targets were identified and ranked. Ten nanograms of DNA was used to generate libraries using the Ion Ampliseq library preparation kit v2.0. The barcoded libraries were diluted to 7 pmol/L for template preparation using the OneTouch 2 instrument and Ion PGMTemplate OT2 200 Kit. The resulting pooled libraries were quality control checked using the Ion Sphere quality control kit, and were then sequenced on the Ion Torrent PGM using a PGM 200 sequencing kit v2 and 318 Chip v2 (Life Technologies).

NGS variant analysis

Point mutations were identified with the Torrent Suite Software v3.0 and the Ion Torrent Variant Caller v4.0 Plugin using the somatic high stringency parameters and the targeted and hotspot pipelines. A 5% allele frequency threshold and 500× minimum coverage was set to report de novo mutations. All the variants identified were established by visualizing the data through IGV 2.3 (Broad Institute). Concurrently, orthogonal verification of variant calls made was performed using the iCMDB (Vishuo Biomedical) platform which provides variant annotation, as well as Sanger sequencing for variant validation.

Western blotting

One hundred milligrams of snap-frozen tissues was lysed in RIPA buffer containing protease and phosphatase inhibitors. Protein extracts (50 μg) were electrophoresed on 4% to 12% Bis-Tris gels and transferred to PVDF membranes. Membranes were blocked with 5% bovine serum albumin in Tris-buffered saline and incubated overnight at 4°C with 1:1,000 dilutions of anti-phospho-mTOR (Ser2448), anti-mTOR, anti-phospho-p70S6K (Thr389), anti-p70S6K, anti-phospho-AMPKα (Thr172), and anti-AMPK antibodies (Thermofisher). Membranes were then washed and incubated with a 1:2,000 dilution of horseradish peroxidase-conjugated goat anti-rabbit secondary antibody. Immunoreactive bands were detected by chemiluminescence using the ECL Prime Western Blotting Kit (GE Healthcare). Intensity of each immunoreactive band was quantified by densitometry using Image J software, and expressed relative to the control mice after normalization to ß-actin.

Statistical analysis

Sample size determination for in vivo studies was performed using preliminary data generated from cell line xenografts a priori. Briefly, the number of samples required per treatment group is 4, at 95% confidence level and 80% power, for a desired effect size of at least 1,000 mm3. For all experiments, Mann–Whitney and Kruskal–Wallis tests were used to determine the differences between two groups, or three and more groups, respectively. P values <0.05 were recognized as statistically significant.

Growth inhibition of 5-FU and metformin on cancer cell lines

Metformin has been reported to preferentially inhibit the growth of p53-mutant cells by forcing a metabolic burden that result in selective toxicity (9, 21, 22). However, recent evidence suggests that metformin requires p53 activity for metformin-induced growth inhibition (23, 24). To investigate whether metformin affects tumor proliferation in a p53-dependent manner, we screened a panel of colorectal cancer cell lines with known p53 statuses (Table 1), including a pair of isogenic cell lines HCT-116 (p53-WT and p53-KO). Increasing concentration of 5-FU at micromolar levels results in a growth inhibition of all colorectal cancer cells (Fig. 1A) while supraphysiologic concentrations of metformin were required to achieve 50% cell growth inhibition (relative to untreated cells) at 48 hours (Fig. 1B).

Table 1.

p53 and microsatellite status of cell lines and PDX

Samplep53 statusMSI/MSS
DLD-1 Mutant MSI 
HCT116 Wild-type MSS 
SK-CO-1 Wild-type MSS 
SW-480 Mutant MSS 
PDX FIT-CRC-086 Mutant MSS 
PDX FIT-CRC-104 Wild-type MSI 
Samplep53 statusMSI/MSS
DLD-1 Mutant MSI 
HCT116 Wild-type MSS 
SK-CO-1 Wild-type MSS 
SW-480 Mutant MSS 
PDX FIT-CRC-086 Mutant MSS 
PDX FIT-CRC-104 Wild-type MSI 
Figure 1.

In vitro effects of 5-fluorouracil (5-FU) and metformin on colorectal cancer cells. 5-FU, the first-line treatment for colorectal cancer, inhibits cell proliferation at micromolar levels (A). Millimolar levels of metformin are required to inhibit cell proliferation at 48 hours, with pronounced inhibition observed in combination with 10 μmol/L 5-FU (B). Glucose deprivation accentuates growth inhibition by metformin (C). Inhibition of colorectal cancer cell proliferation to 5-FU (D) and metformin is independent of p53 status, in both HCT116 p53 knockout (p53 KO) and HCT116 p53 wild-type (p53 WT) cell lines, but with stronger degree of growth inhibition observed under glucose-deprived conditions in the presence of metformin (E). The absolute number of colonies is presented as an average of triplicates (F).

Figure 1.

In vitro effects of 5-fluorouracil (5-FU) and metformin on colorectal cancer cells. 5-FU, the first-line treatment for colorectal cancer, inhibits cell proliferation at micromolar levels (A). Millimolar levels of metformin are required to inhibit cell proliferation at 48 hours, with pronounced inhibition observed in combination with 10 μmol/L 5-FU (B). Glucose deprivation accentuates growth inhibition by metformin (C). Inhibition of colorectal cancer cell proliferation to 5-FU (D) and metformin is independent of p53 status, in both HCT116 p53 knockout (p53 KO) and HCT116 p53 wild-type (p53 WT) cell lines, but with stronger degree of growth inhibition observed under glucose-deprived conditions in the presence of metformin (E). The absolute number of colonies is presented as an average of triplicates (F).

Close modal

Under glucose-deprived conditions, the degree of growth inhibition by metformin is higher in colorectal cancer lines (Fig. 1C). There was no differential growth inhibition in response to 5-FU (Fig. 1D) and metformin between the paired isogenic p53 lines (P = 0.206) as was previously reported, although cells with p53-WT were inhibited to a greater extent under low glucose conditions (P = 0.06; Fig. 1E). The clonogenic ability of p53-WT and p53-KO cells were similar upon metformin treatment (P = 0.743); nonetheless, cell growth for both lines were severely impaired under glucose-deprived conditions (Fig. 1C and D).

Approximately 15% of colorectal cancers display microsatellite instability and are classified as microsatellite instable (MSI) or microsatellite stable (MSS). Characterization of colorectal cancer cells for MSI/MSS status indicates that there was no difference in response to metformin between MSI and MSS cell lines. Essentially, our results indicate that there is non-differential growth inhibition to metformin, regardless of microsatellite or p53 status.

PDX generation

Two primary colorectal cancer tumors, FIT-CRC-086 and FIT-CRC-104, with different clinicopathologic characteristics (Table 2), were engrafted into 4 to 6 weeks old female SCID mice. Growth rates ranged from 8 to 15 weeks for the initial transplantation to reach a tumor volume of 1,000 to 1,500 mm3, and 3 to 8 weeks for subsequent passages. Histopathology of patient's primary tumors and various xenograft passages showed that the PDX models retained the original morphology of the human primary tumor (Fig. 2A).

Table 2.

Clinicopathologic information of patients with xenografts generated

PatientHistologyTumor siteTumor gradeTumor typeTumor size (largest dimension in cm)AJCC stagepTpNpMLymph node invasion
FIT-CRC-086 Adenocarcinoma Rectum Ulcerated IVA 1a 0/7 
FIT-CRC-104 Mucinous adenocarcinoma Ascending colon Localized IIA 0/39 
PatientHistologyTumor siteTumor gradeTumor typeTumor size (largest dimension in cm)AJCC stagepTpNpMLymph node invasion
FIT-CRC-086 Adenocarcinoma Rectum Ulcerated IVA 1a 0/7 
FIT-CRC-104 Mucinous adenocarcinoma Ascending colon Localized IIA 0/39 
Figure 2.

Histomorphologic features of parental tumor are retained in patient-derived xenografts after serial passaging. Representative tumor cells are indicated by solid arrows; mucins are indicated by diamond arrowheads (A). Gross morphology of PDX-FIT-CRC-086 tumors treated with metformin and 5-FU (B). Tumor growth profile of PDX-FIT-CRC-086 (C) and PDX-FIT-CRC-104 (D). Metformin triggers AMPK activation and inhibits mTOR signaling in PDX (E). Protein levels were normalized to ß-actin and fold changes are quantified relative to vehicle control (F).

Figure 2.

Histomorphologic features of parental tumor are retained in patient-derived xenografts after serial passaging. Representative tumor cells are indicated by solid arrows; mucins are indicated by diamond arrowheads (A). Gross morphology of PDX-FIT-CRC-086 tumors treated with metformin and 5-FU (B). Tumor growth profile of PDX-FIT-CRC-086 (C) and PDX-FIT-CRC-104 (D). Metformin triggers AMPK activation and inhibits mTOR signaling in PDX (E). Protein levels were normalized to ß-actin and fold changes are quantified relative to vehicle control (F).

Close modal

Non-differential inhibition of metformin on PDX

To evaluate tumor response in vivo, we utilized two PDX models with differing p53 and MSI/MSS status (Table 1). Mice treated with metformin had a lower tumor burden as compared with control group, regardless of p53 nor microsatellite status (Fig. 2B–D). Metformin was able to slow tumor growth initially, but at 12 days posttreatment initiation, PDX tumors eventually progressed. After 24 days of drug treatment, the average tumor volume of the metformin-treated group was reduced by 50% (P = 0.056) and 65% (P < 0.05) for p53-wild-type and p53-mutant PDX, respectively. A systemic effect of metformin on cell growth was observed in the PDXs independent of tumor p53 status, with more pronounced growth inhibition being observed in the PDXs with mutant-p53.

When metformin was administered concurrently with a clinically acceptable dose of 5-FU, only an additional 4% (P = 0.767) tumor inhibition was observed in PDX-FIT-CRC-104; in contrast, a pronounced tumor growth inhibition of 85% (P = 0.054) was observed in PDX-FIT-CRC-086.

Metformin activates the AMPK pathway

To determine whether metformin influenced the AMPK and mTOR signaling pathway in the PDX materials, we evaluated the phosphorylation of AMPK, mTOR, and p70S6K (Fig. 2E). Metformin treatment activated AMPK as determined by phosphorylation at Thr172, with a corresponding inhibition of mTOR signaling and downstream target phosphorylation of p70S6K. In tumors treated with metformin and 5-FU, AMPK was activated; interestingly, mTOR signaling appears to be activated and not suppressed, although phosphorylation of p70S6K was inhibited (Fig. 2F).

Metformin inhibits oxygen consumption rate (OCR)

For ex vivo measurement of oxygen consumption, organoids derived from PDX were treated with metformin (Fig. 3A). Drugs affecting mitochondrial functions were used in parallel to demonstrate normal mitochondrial function. Metformin inhibited OCR of both FIT-CRC-086 and FIT-CRC-104 (Fig. 3B and C, respectively) in a concentration-dependent manner. Consequent injection of oligomycin did not suppress the OCR further in these metformin-treated organoids, while injection of FCCP, the mitochondrial uncoupler, rescues the respiration inhibition, albeit to a lesser extent in the metformin-treated organoids as compared with control.

Figure 3.

Organoids derived from PDX under 10× (left image) and 40× (right image) (A). Pretreatment with 0.1 mmol/L metformin or 1.0 mmol/L metformin for 24 hours inhibits oxygen consumption rate (OCR) relative to vehicle control for both FIT-CRC-086 (B) and FIT-CRC-104 (C), with corresponding increase in glycolysis for FIT-CRC-086 (D) and FIT-CRC-104 (E). Pretreatment with 0.1 mmol/L metformin and 1.0 mmol/L metformin inhibits OCR as early as 4 hours in both FIT-CRC-086 (F) and FIT-CRC-104 (G). 1.0 μmol/L oligomycin, 1.0 μmol/L carbonyl cyanide-p-trifluoromethoxyphenylhydrazone (FCCP), and 0.5 μmol/L rotenone/antimycin A were used as mitochondrial stress controls.

Figure 3.

Organoids derived from PDX under 10× (left image) and 40× (right image) (A). Pretreatment with 0.1 mmol/L metformin or 1.0 mmol/L metformin for 24 hours inhibits oxygen consumption rate (OCR) relative to vehicle control for both FIT-CRC-086 (B) and FIT-CRC-104 (C), with corresponding increase in glycolysis for FIT-CRC-086 (D) and FIT-CRC-104 (E). Pretreatment with 0.1 mmol/L metformin and 1.0 mmol/L metformin inhibits OCR as early as 4 hours in both FIT-CRC-086 (F) and FIT-CRC-104 (G). 1.0 μmol/L oligomycin, 1.0 μmol/L carbonyl cyanide-p-trifluoromethoxyphenylhydrazone (FCCP), and 0.5 μmol/L rotenone/antimycin A were used as mitochondrial stress controls.

Close modal

Correspondingly, metformin inhibition on cellular respiration resulted in an increase in glycolysis, as measured by the extracellular acidification rate (ECAR; Fig. 3D and E). It was evident that metformin inhibits mitochondrial function as early as 4 hours (Fig. 3F and G).

Molecular profiling of circulating biomarkers in mice

To determine if tumor burden in vivo can be assessed using alternative approaches that are viable in the clinical setting, we investigate the use of cfDNA in our models (25). CfDNA is a heterogeneous mixture of normal and tumor-associated cell populations (26, 27) which makes cfDNA analysis technically challenging. In xenografts, both human and mouse-derived cell populations are present, and the ability to distinguish these two cell populations is critical in order to assess the PDX response to treatment using cfDNA. Here, using a limited amount of cfDNA, our assay can distinguish human and mice-derived DNA efficiently (Supplementary Fig. S1), with limit of detection of 10% human DNA in a background of mice DNA. No cfDNA was detected at week 0, at the point of tumor implantation. Subsequent evaluation of the cfDNA at different time points in the course of treatment indicates that human-derived DNA could be detected in the mice blood, suggesting that cfDNA could be a viable noninvasive biomarker for tumor burden.

Mutation profile of primary tumor and corresponding xenografts

Human samples and PDX were analyzed for variants/mutations by NGS using a panel of 40 genes commonly implicated in colorectal cancer. The mean sequencing depth across all experiments was 1,385X for FIT-CRC-086 and 2,534X for FIT-CRC-104. A summary of annotated variants is shown in Fig. 4A.

Figure 4.

Molecular characterization of germline DNA, parental tumor, serial passages of xenografts and xenografts treated with drugs. Variant allelic frequency is reflected on the scale of 0 to 1 (A). An example of a COSMIC variant detected by next-generation sequencing (NGS) and independently verified by Sanger sequencing (B), two variants that have not been reported in COSMIC are detected by NGS and Sanger sequencing (C).

Figure 4.

Molecular characterization of germline DNA, parental tumor, serial passages of xenografts and xenografts treated with drugs. Variant allelic frequency is reflected on the scale of 0 to 1 (A). An example of a COSMIC variant detected by next-generation sequencing (NGS) and independently verified by Sanger sequencing (B), two variants that have not been reported in COSMIC are detected by NGS and Sanger sequencing (C).

Close modal

For FIT-CRC-086, mutations in APC (K1350*), KRAS (G12V), TP53 (R248Q), and SMAD4 (L495R) were observed for parental tumor and xenografts, but were not present in germline or normal colon. Likewise, in FIT-CRC-104, mutations were observed in KRAS (A146T) and MLH1 (R100*) in parental tumor. Essentially, mutations remain conserved across PDX passages. An example of a variant conserved in tumor and PDX as depicted by NGS are corroborated by Sanger sequencing in Fig. 4B. Interestingly, we also observed the presence of variants that are not reported in COSMIC database. Using Sanger sequencing, we verified the non-COSMIC variants that were detected (Fig. 4C). Nonetheless, non-COSMIC variants were not present in every PDX. Within the same treatment subset, individuals exhibit heterogeneity. Whether these new variants were a function of clonal selection of a subset of tumor cells warrants further investigation.

One of the key objectives of performing NGS is to identify biomarkers that may reflect response to metformin. Our analysis indicates that there are no key colorectal cancer genes that are associated with metformin's activity.

Metformin has been extensively reported as a promising anticancer agent in cancer prevention and therapy (9–12, 21, 22, 28, 29). In the present study, we evaluated the effects of metformin using preclinical models. Similar to reported studies, we observed that the antiproliferative properties of metformin manifest at supraphysiologic conditions in cell lines (30). A recurrent feature is that these cells are typically maintained in non-physiologic conditions that are optimized only for in vitro growth and proliferation. The excessive concentration of growth factors, insulin and glucose may account for the elevated doses of metformin required to elicit cellular responses and trigger metabolic stress. As such, it is virtually impossible to translate these supraphysiologic conditions in the clinical setting, especially when the pharmacokinetics of the drug is different in culture settings and in the human body.

Consequently, we observed that low glucose conditions accentuated the growth inhibition by metformin. This is not surprising—glucose deprivation is a distinctive feature of the tumor microenvironment caused by the imbalance between nutrient supply and oxygen consumption rate (31). In a study exploring the role of the tumor microenvironment in potentiating the effect of metformin, metformin is synthetically lethal with glucose withdrawal in cancer cells (32). Essentially, this suggests that the clinically irrelevant, supraphysiologic levels needed to observe metformin's anticancer effects as reported by many in vitro studies merely underlie the artefactual experimental conditions that poorly reflect actual glucose-starved in vivo conditions.

Most preclinical in vivo models have generally proven to be sub-optimal for directing clinical application of new anticancer therapies, largely due to their inability to reflect the complexity and heterogeneity of human tumors. PDX, where surgically resected tumor samples are engrafted directly into immunocompromised mice, offer several advantages (33, 34). In our study, we have shown that PDX tumors maintained the molecular, genetic, and histologic heterogeneity typical of tumors of origin (35–39). PDX provides an excellent platform to study cancer biology, analyze cancer therapeutics, and novel drug combinations. In our study, we demonstrated metformin efficacy in our PDX models. To the best of our knowledge, ours represents the first study reported for colorectal cancer PDX using metformin. Administering metformin at physiologic levels of 150 mg/kg per day in our mice, which is equivalent to initial clinical dose of 500 to 1,000 mg/daily in human, is sufficient to inhibit tumor growth in PDX. Furthermore, we observed that the response to metformin may be independent of p53 status and is inconsistent with the observations reported earlier (9, 24). Conversely, studies conducted in breast cancer indicate that metformin's effects are independent of p53 status (40, 41). This contraindication suggests that p53 may not be a reliable companion biomarker to predict response to metformin.

Studies from a recent phase II clinical trial evaluating metformin and 5-FU in patients with refractory colorectal cancer showed only a mild median progression-free survival (PFS) of 1.8 months and median overall survival (OS) of 7.9 months, with a majority of patients (78%) having disease progression before 8 weeks (42). For 11 out of 50 patients (22%) having stable disease at the end of 8 weeks, their median PFS and OS were 5.6 months and 16.2 months, respectively. Likely, the modest benefits of metformin and 5-FU observed in this trial may be due to the nature of the refractory colorectal cancer cases and that the benefits may be significant in nonrefractory colorectal cancer cases.

Interestingly, in our study, we observed that combination of 5-FU with metformin leads to a pronounced growth inhibition in the p53-mutant PDX but not in p53-wild-type PDX. It may be possible that p53 status plays a role in differential response to metformin and 5-FU; however, it is also likely that 5-FU is non-beneficial in the p53-wild-type PDX because this PDX has MSI. Indeed, clinical observations have suggested that patients having a deficient mismatch repair system or MSI do not benefit from 5-FU–based adjuvant therapy in colorectal cancer (43, 44). Thus, our observations using PDX are in line with observations obtained in the clinic, thus, emphasizing the crucial role PDXs play as a clinically relevant model.

To demonstrate the integrity of PDX as suitable models, we performed high-throughput NGS to characterize the genetic profiles of parental tumor and PDX. We demonstrate that genes such as TP53, APC, KRAS, and SMAD4, were mutated in the parental tumor and conserved throughout serial passaging of the xenografts. More importantly, in an effort to characterize biomarkers that can serve as companion marker to predict metformin response, we compared the genetic differences between the two PDX lines. Our NGS analysis suggests that there is lack of novel actionable genes that may reflect metformin's activity based on genotype. Nonetheless, the key limitation in this particular analysis is likely due to limited variation in genotype between the two PDX lines.

To gain insight into possible mechanisms of metformin action, we evaluated the frequently implicated AMPK signaling pathway (45). Indeed, our data showed that metformin triggered phosphorylation of Thr172 on AMPKα, which consequently led to downregulation of mTOR and its target p70S6K, whose phosphorylation at Thr389 is necessary for protein synthesis (46). Moreover, we observed that metformin inhibits oxygen consumption in organoids derived from PDX, as early as 4 hours, with a concomitant increase in glycolysis. Taken together, it is likely that treatment with metformin activates the AMPK pathway, which alters the mTOR signaling, and executes a metabolic change in the cells, which ultimately affects cell growth in the PDX.

We acknowledged that there are several limitations to our current study. Firstly, we restricted the use of PDX to subcutaneous tumor models. Subcutaneous models are inferior to orthotopic models because the former do not represent appropriate sites for human tumors, and may not be accurately predictive for drug evaluation (47). However, orthotopic tumor models are time-consuming and technically challenging. Furthermore, it is difficult to monitor tumor growth, and sophisticated micro-imaging techniques are required to visualize the tumor. In anticipation of this limitation (should the need arise to monitor tumor burden in orthotopic models), we assessed noninvasive circulating biomarkers (cfDNA) in our subcutaneous PDX models. Using established techniques that we have utilized previously for clinical cancer patients (48, 49), we demonstrated that it is feasible to prove the presence of human-derived DNA circulating in plasma.

Utilizing NGS approach could aid in predictive biomarker(s) identification as a result of drug selection. Because this is a pilot study involving metformin and PDX, banking on molecular profiles derived from 2 colorectal cancer patients may be inadequate for biomarker discovery process. Expanding the repertoire of PDX lines with different genetic make-up is likely to shed more information in future. While the creation of PDX is of clear interest as a means to better define optimal therapy for colorectal cancer, it is undeniable that the high-throughput data are challenging to manage and visualize. Thoughtful interpretation is necessary to convert the knowledge derived from existing data in order to understand treatment outcomes.

In essence, our study demonstrated the utility of PDX in assessing drug efficacy via molecular analysis, metabolic profiling, and bioinformatics approaches. This is the first study that describes the use of metformin in colorectal cancer PDX and organoids, providing direct evidence of metformin as an anticancer agent in colorectal cancer.

M.-H. Tan has ownership interest in a patent filed for method for KRAS mutation detection, owned by his employer, and licensed to industry. No potential conflicts of interest were disclosed.

Conception and design: N.-A. Mohamed Suhaimi, W.M. Phyo, H.Y. Yap, M.-H. Tan

Development of methodology: N.-A. Mohamed Suhaimi, W.M. Phyo, W.J. Tan

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): N.-A. Mohamed Suhaimi, W.M. Phyo, H.Y. Yap, S.H.Y. Choy, Y. Choudhury, W.J. Tan, L.A.P.Y. Tan, R.S.Y. Foo, S.H.S. Tan, P.K. Koh, M.-H. Tan

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): N.-A. Mohamed Suhaimi, W.M. Phyo, X. Wei, L.A.P.Y. Tan, M.-H. Tan

Writing, review, and/or revision of the manuscript: N.-A. Mohamed Suhaimi, W.M. Phyo, Y. Choudhury, W.J. Tan, L.A.P.Y. Tan, P.K. Koh, M.-H. Tan

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): W.J. Tan, L.A.P.Y. Tan, Z. Tiang, C.F. Wong, M.-H. Tan

Study supervision: M.-H. Tan

We would like to thank Concord Cancer Hospital for the provision of the tissue samples, and Vishuo Biomedical for the support in the bioinformatics analysis. We would like to thank Chua Teck Chiew Timothy, Ma Janise Ann Kabiling Lalic, Mukta Pathak, and Hannes Martin Hentze from A*STAR Biological Resource Centre for their help and support in PDX work. We would also like to thank Chit Fang Cheok for the helpful discussion on the in vitro work and Shengyong Ng for the assistance in organoids derivation and culture.

This work is supported by the Institute of Bioengineering and Nanotechnology and Agency for Science, Technology and Research, Singapore. M.H. Tan received grants from Biomedical Research Council (Diagnostics Grant & Strategic Positioning Fund SPF 2012/003 and SPF2015/002).

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

1.
Voskoglou-Nomikos
T
,
Pater
JL
,
Seymour
L
. 
Clinical predictive value of the in vitro cell line, human xenograft, and mouse allograft preclinical cancer models
.
Clin Cancer Res
2003
;
9
:
4227
39
.
2.
Östman
A
,
Augsten
M
. 
Cancer-associated fibroblasts and tumor growth – bystanders turning into key players
.
Curr Opin Genet Dev
2009
;
19
:
67
73
.
3.
Ertel
A
,
Verghese
A
,
Byers
S
,
Ochs
M
,
Tozeren
A
. 
Pathway-specific differences between tumor cell lines and normal and tumor tissue cells
.
Mol Cancer
2006
;
5
:
55
.
4.
Stein
WD
,
Litman
T
,
Fojo
T
,
Bates
SE
. 
A Serial analysis of gene expression (SAGE) database analysis of chemosensitivity: comparing solid tumors with cell lines and comparing solid tumors from different tissue origins
.
Cancer Res
2004
;
64
:
2805
16
.
5.
Leupin
N
,
Kuhn
A
,
Hügli
B
,
Grob
TJ
,
Jaggi
R
,
Tobler
A
, et al
Gene expression profiling reveals consistent differences between clinical samples of human leukaemias and their model cell lines
.
Br J Haematol
2006
;
135
:
520
3
.
6.
de Gramont
A
,
Figer
A
,
Seymour
M
,
Homerin
M
,
Hmissi
A
,
Cassidy
J
, et al
Leucovorin and fluorouracil with or without oxaliplatin as first-line treatment in advanced colorectal cancer
.
J Clin Oncol
2000
;
18
:
2938
47
.
7.
Evans
JM
,
Donnelly
LA
,
Emslie-Smith
AM
,
Alessi
DR
,
Morris
AD
. 
Metformin and reduced risk of cancer in diabetic patients
.
BMJ
2005
;
330
:
1304
5
.
8.
Morales
DR
,
Morris
AD
. 
Metformin in cancer treatment and prevention
.
Annu Rev Med
2014
;
1
13
.
9.
Buzzai
M
,
Jones
RG
,
Amaravadi
RK
,
Lum
JJ
,
DeBerardinis
RJ
,
Zhao
F
, et al
Systemic treatment with the antidiabetic drug metformin selectively impairs p53-deficient tumor cell growth
.
Cancer Res
2007
;
67
:
6745
52
.
10.
Zi
FM
,
He
JS
,
Li
Y
,
Wu
C
,
Yang
L
,
Yang
Y
, et al
Metformin displays anti-myeloma activity and synergistic effect with dexamethasone in in vitro and in vivo xenograft models
.
Cancer Lett
2015
;
356
:
443
53
.
11.
Memmott
RM
,
Mercado
JR
,
Maier
CR
,
Kawabata
S
,
Fox
SD
,
Dennis
PA
. 
Metformin prevents tobacco carcinogen–induced lung tumorigenesis
.
Cancer Prev Res
2010
;
3
:
1066
76
.
12.
Lengyel
E
,
Litchfield
LM
,
Mitra
AK
,
Nieman
KM
,
Mukherjee
A
,
Zhang
Y
, et al
Metformin inhibits ovarian cancer growth and increases sensitivity to paclitaxel in mouse models
.
Am J Obstet Gynecol
2015
;
212
:
479
.
e1-479.e10
.
13.
Ben Sahra
I
,
Laurent
K
,
Loubat
A
,
Giorgetti-Peraldi
S
,
Colosetti
P
,
Auberger
P
, et al
The antidiabetic drug metformin exerts an antitumoral effect in vitro and in vivo through a decrease of cyclin D1 level
.
Oncogene
2008
;
27
:
3576
86
.
14.
Zordoky
BNM
,
Bark
D
,
Soltys
CL
,
Sung
MM
,
Dyck
JRB
. 
The anti-proliferative effect of metformin in triple-negative MDA-MB-231 breast cancer cells is highly dependent on glucose concentration: implications for cancer therapy and prevention
.
Biochim Biophys Acta
2014
;
1840
:
1943
57
.
15.
Kinaan
M
,
Ding
H
,
Triggle
CR
. 
Metformin: an old drug for the treatment of diabetes but a new drug for the protection of the endothelium
.
Med Princ Pract
2015
;
24
:
401
15
.
16.
Krishnamurthy
S
,
Ng
VW
,
Gao
S
,
Tan
MH
,
Yang
YY
. 
Phenformin-loaded polymeric micelles for targeting both cancer cells and cancer stem cells in vitro and in vivo
.
Biomaterials
2014
;
35
:
9177
86
.
17.
Muzny
DM
,
Bainbridge
MN
,
Chang
K
,
Dinh
HH
,
Drummond
JA
.,
Fowler
G
, et al
Comprehensive molecular characterization of human colon and rectal cancer
.
Nature
2012
;
487
:
330
7
.
18.
De Sousa E Melo
F
,
Wang
X
,
Jansen
M
,
Fessler
E
,
Trinh
A
,
de Rooij
LPMH
. 
Poor-prognosis colon cancer is defined by a molecularly distinct subtype and develops from serrated precursor lesions
.
Nat Med
2013
;
19
:
614
8
.
19.
Sadanandam
A
,
Lyssiotis
CA
,
Homicsko
K
,
Collisson
EA
,
Gibb
WJ
,
Wullschleger
S
, et al
A colorectal cancer classification system that associates cellular phenotype and responses to therapy
.
Nat Med
2013
;
19
:
619
25
.
20.
Budinska
E
,
Popovici
V
,
Tejpar
S
,
D'Ario
G
,
Lapique
N
,
Sikora
KO
, et al
Gene expression patterns unveil a new level of molecular heterogeneity in colorectal cancer
.
J Pathol
2013
;
231
:
63
76
.
21.
Sarfstein
R
,
Friedman
Y
,
Attias-Geva
Z
,
Fishman
A
,
Bruchim
I
,
Werner
H
. 
Metformin downregulates the insulin/IGF-I signaling pathway and inhibits different uterine serous carcinoma (USC) cells proliferation and migration in p53-dependent or -independent manners
.
PLoS One
2013
;
8
:
e61537
.
22.
Ben Sahra
I
,
Laurent
K
,
Giuliano
S
,
Larbret
F
,
Ponzio
G
,
Gounon
P
, et al
Targeting cancer cell metabolism: the combination of metformin and 2-deoxyglucose Induces p53-dependent apoptosis in prostate cancer cells
.
Cancer Res
2010
;
70
:
2465
75
.
23.
Li
P
,
Zhao
M
,
Parris
AB
,
Feng
X
,
Yang
X
. 
p53 is required for metformin-induced growth inhibition, senescence and apoptosis in breast cancer cells
.
Biochem Biophys Res Commun
2015
;
464
:
1267
74
.
24.
Yi
G
,
He
Z
,
Zhou
X
,
Xian
L
,
Yuan
T
,
Jia
X
, et al
Low concentration of metformin induces a p53-dependent senescence in hepatoma cells via activation of the AMPK pathway
.
Int J Oncol
2013
;
43
:
5
.
25.
Mohamed Suhaimi
NA
,
Tan
MH
. 
Circulating tumor cells and circulating tumor DNA in colorectal cancer
.
Expert Rev Precis Med Drug Dev
2016
;
1
:
181
94
.
26.
Jahr
S
,
Hentze
H
,
Englisch
S
,
Hardt
D
,
Fackelmayer
FO
,
Hesch
RD
, et al
DNA fragments in the blood plasma of cancer patients: quantitations and evidence for their origin from apoptotic and necrotic cells
.
Cancer Res
2001
;
61
:
1659
65
.
27.
Tomochika
S
,
Iizuka
N
,
Watanabe
Y
,
Tsutsui
M
,
Takeda
S
,
Yoshino
S
, et al
Increased serum cell-free DNA levels in relation to inflammation are predictive of distant metastasis of esophageal squamous cell carcinoma
.
Exp Ther Med
2010
;
1
:
89
92
.
28.
Hirsch
HA
,
Iliopoulos
D
,
Tsichlis
PN
,
Struhl
K
. 
Metformin selectively targets cancer stem cells, and acts together with chemotherapy to block tumor growth and prolong remission
.
Cancer Res
2009
;
69
:
7507
11
.
29.
Quinn
BJ
,
Kitagawa
H
,
Memmott
RM
,
Gills
JJ
,
Dennis
PA
. 
Repositioning metformin for cancer prevention and treatment
.
Trends Endocrinol Metab
2013
;
24
:
469
80
.
30.
Martin-Castillo
B
,
Vazquez-Martin
A
,
Oliveras-Ferraros
C
,
Menendez
JA
. 
Metformin and cancer: doses, mechanisms and the dandelion and hormetic phenomena
.
Cell Cycle
2010
;
9
:
1057
64
.
31.
Gatenby
RA
,
Gillies
RJ
. 
Why do cancers have high aerobic glycolysis?
Nat Rev Cancer
2004
;
4
:
891
9
.
32.
Menendez
JA
,
Oliveras-Ferraros
C
,
Cufí
S
,
Corominas-Faja
B
,
Joven
J
,
Martin-Castillo
B
, et al
Metformin is synthetically lethal with glucose withdrawal in cancer cells
.
Cell Cycle
2012
;
11
:
2782
92
.
33.
Tentler
JJ
,
Tan
AC
,
Weekes
CD
,
Jimeno
A
,
Leong
S
,
Pitts
TM
, et al
Patient-derived tumour xenografts as models for oncology drug development
.
Nat Rev Clin Oncol
2012
;
9
:
338
50
.
34.
Hidalgo
M
,
Amant
F
,
Biankin A
V
,
Budinská
E
,
Byrne
AT
,
Caldas
C
, et al
Patient-derived xenograft models: an emerging platform for translational cancer research
.
Cancer Discov
2014
;
4
:
998
1013
.
35.
Garralda
E
,
Paz
K
,
López-Casas
PP
,
Jones
S
,
Katz
A
,
Kann
LM
, et al
Integrated next-generation sequencing and avatar mouse models for personalized cancer treatment
.
Clin Cancer Res
2014
;
20
:
2476
84
.
36.
Mattie
M
,
Christensen
A
,
Chang
MS
,
Yeh
W
,
Said
S
,
Shostak
Y
, et al
Molecular characterization of patient-derived human pancreatic tumor xenograft models for preclinical and translational development of cancer therapeutics
.
Neoplasia
2013
;
15
:
1138
50
.
37.
Klinghammer
K
,
Raguse
JD
,
Plath
T
,
Albers
AE
,
Joehrens
K
,
Zakarneh
A
, et al
A comprehensively characterized large panel of head and neck cancer patient-derived xenografts identifies the mTOR inhibitor everolimus as potential new treatment option
.
Int J Cancer
2014
;
2948
.
38.
Seol
HS
,
Kang
H
,
Lee
SI
,
Kim
NE
,
Kim
TI
,
Chun
SM
, et al
Development and characterization of a colon PDX model that reproduces drug responsiveness and the mutation profiles of its original tumor
.
Cancer Lett
2014
;
345
:
56
64
.
39.
Hao
C
,
Wang
L
,
Peng
S
,
Cao
M
,
Li
H
,
Hu
J
, et al
Gene mutations in primary tumors and corresponding patient-derived xenografts derived from non-small cell lung cancer
.
Cancer Lett
2015
;
357
:
179
85
.
40.
Rattan
R
,
Giri
S
,
Hartmann
LC
,
Shridhar
V
. 
Metformin attenuates ovarian cancer cell growth in an AMP-kinase dispensable manner
.
J Cell Mol Med
2011
;
15
:
166
78
.
41.
Pandiri
I
,
Chen
Y
,
Joe
Y
,
Kim
HJ
,
Park
J
,
Chung
HT
, et al
Tristetraprolin mediates the anti-proliferative effects of metformin in breast cancer cells
.
Breast Cancer Res Treat
2016
;
156
:
57
64
.
42.
Miranda
VC
,
Braghiroli
MI
,
Faria
LD
,
Bariani
G
,
Alex
A
,
Evangelista
J
, et al
Phase 2 trial of metformin combined with 5-fluorouracil in patients with refractory metastatic colorectal cancer
.
Clin Colorectal Cancer
2016
;
15
:
321
328.e1
.
43.
Sargent
DJ
,
Marsoni
S
,
Monges
G
,
Thibodeau
SN
,
Labianca
R
,
Hamilton
SR
, et al
Defective mismatch repair as a predictive marker for lack of efficacy of fluorouracil-based adjuvant therapy in colon cancer
.
J Clin Oncol
2010
;
28
:
3219
26
.
44.
Ribic
CM
,
Sargent
DJ
,
Moore
MJ
,
Thibodeau
SN
,
French
AJ
,
Goldberg
RM
, et al
Tumor microsatellite-instability status as a predictor of benefit from fluorouracil-based adjuvant chemotherapy for colon cancer
.
N Engl J Med
2003
;
349
:
247
57
.
45.
Viollet
B
,
Guigas
B
,
Sanz Garcia
N
,
Leclerc
J
,
Foretz
M
,
Andreelli
F
. 
Cellular and molecular mechanisms of metformin: an overview
.
Clin Sci (Lond)
2012
;
122
:
253
70
.
46.
Zakikhani
M
,
Blouin
MJ
,
Piura
E
,
Pollak
MN
. 
Metformin and rapamycin have distinct effects on the AKT pathway and proliferation in breast cancer cells
.
Breast Cancer Res Treat
2010
;
123
:
271
9
.
47.
Killion
J
,
Radinsky
R
,
Fidler
I
. 
Orthotopic models are necessary to predict therapy of transplantable tumors in mice
.
Cancer Metastasis Rev
1998
;
17
:
279
84
.
48.
Lim
LS
,
Hu
M
,
Huang
MC
,
Cheong
WC
,
Gan
ATL
,
Looi
XL
, et al
Microsieve lab-chip device for rapid enumeration and fluorescence in situ hybridization of circulating tumor cells
.
Lab Chip
2012
;
12
:
4388
.
49.
Mohamed Suhaimi
NA
,
Foong
YM
,
Lee
DYS
,
Phyo
WM
,
Cima
I
,
Lee
EX
, et al
Non-invasive sensitive detection of KRAS and BRAF mutation in circulating tumor cells of colorectal cancer patients
.
Mol Oncol
2015
;
9
:
850
60
.