Metformin Mitigates DPP-4 Inhibitor-Induced Breast Cancer Metastasis via Suppression of mTOR Signaling

Type 2 diabetes (T2DM) is an independent risk factor for certain cancers (1). Diabetic patients with cancer have been shown to have a higher all-cause mortality risk and a higher cancer site–specific mortality risk than their nondiabetic counterparts (2).

Breast cancer is a key cause of death for females worldwide. A meta-analysis revealed that women with diabetes have a 23% higher risk of breast cancer incidence than those without diabetes (3). Another meta-analysis indicated that preexisting diabetes is associated with a poor prognosis in women with breast cancer; patients with diabetes have a reduced overall survival rate and a reduced disease-free survival rate compared with nondiabetic patients (4). It has been shown that more than 90% of breast cancer–related deaths are caused by distant metastasis rather than by the primary tumor itself (5). Importantly, 6% to 10% of breast cancer patients display distant metastasis at the time of cancer diagnosis (6).

Epithelial-to-mesenchymal transition (EMT) is a biological process by which epithelial cells convert to a mesenchymal cell phenotype. Induction of EMT is associated with cancer progression and metastasis (7–9), including breast cancer (10). Overexpression of the EMT transcriptional factor Snail detected by IHC in breast cancer patients is associated with a poor prognosis (11). EMT induction by Snail1 in breast cancer increases metastasis in vivo (12).

Another important aspect of cancer biology in diabetic patients is the influence of the long-term use of antidiabetic drugs. Several types of antidiabetic agents, such as dipeptidyl peptidase (DPP)-4 inhibitors (13), glucagon-like peptide (GLP)-1 agonists (14), sulfonylurea (15), and thiazolidinediones (16), have been shown to be associated with the presence of certain cancers in diabetic patients. However, clinical studies focusing on the relationship between antidiabetic drugs and cancers have been shown by retrospective analyses. Therefore, the safety profile and biological significance of antidiabetic agents in cancer-bearing diabetic patients remain relevant topics for further medical research.

DPP-4 inhibitors are antidiabetic drugs for the treatment of T2DM that act by inducing the accumulation of incretin hormones. In addition to incretin hormones, DPP-4 cleaves numerous substrates, including C-X-C motif chemokine 12 (CXCL12). DPP-4 inhibitor induced both invasion and metastasis of prostate cancer by inhibiting CXCL12 degradation in vitro and in vivo (17). We recently reported that a DPP-4 inhibitor induced EMT in both mouse mammary and human breast cancer cells and increased lung metastasis in an allograft breast cancer model mouse via CXCL12, C-X-C receptor 4 (CXCR4), and the mammalian target of rapamycin (mTOR) pathway (18). Importantly, mTOR suppression was sufficient to alleviate mammary/breast cancer cell EMT induced by DPP-4 inhibition, suggesting that inhibition of mTOR, downstream of the CXCL12/CXCR4 pathway, could be key to combat DPP-4 inhibitor–induced EMT and tumor metastasis.

Metformin, the first-line therapy for T2DM, has also been recognized as an anticancer agent (19, 20). Clinical studies indicate that metformin use decreases cancer incidence and cancer-related mortality (21–23). Metformin may benefit cancer-related survival (24) and decrease all-cause mortality in breast cancer–bearing diabetic patients (25, 26). Another retrospective study also showed that metformin was associated with a 30% reduction in the risk of developing metastasis in breast cancer–bearing T2DM patients (27). Additionally, Bayraktar and colleagues showed that patients with triple receptor–negative breast cancer who are not taking metformin tended to have a higher risk of distant metastasis compared with those taking metformin (28). Furthermore, SEER-Medicare database analysis revealed that metformin use was related to favorable overall survival in breast cancer patients either with or without DPP-4 inhibitors; DPP-4 inhibitors alone displayed a worse trend (29). However, such effects of metformin/DPP-4 inhibitors were distinct from that in both prostate and pancreatic cancers (29). The anticancer effect of metformin is known to be associated with the inhibition of mTOR signaling pathways (30). Furthermore, metformin has been shown to suppress the EMT, proliferation, and migration of breast cancer cells (31).

In this report, we assessed whether metformin could repress DPP-4 inhibitor–induced breast cancer metastasis through suppression of mTOR activation.

For drug treatment, KR62436 hydrate (KR, K4264) and metformin (M0605000), purchased from Sigma, were used in all in vitro and in vivo experiments. The experimental metastatic breast/mammary cancer cell lines (MDA-MB-231 and 4T1) were purchased from the American Type Culture Collection (ATCC). MDA-MB-231 cells (HTB-26, RRID: CVCL_0062) were cultured in Leibovitz L-15 medium (#30-2008, ATCC) with 10 % fetal bovine serum (FBS). This medium contains 2 mmol/L L-glutamine and 900 mg/L galactose. 4T1 cells (CRL-2539, RRID: CVCL_0125; GFP-expressing cells) were cultured in RPMI-1640 medium (#30-2011, ATCC) with 10% FBS. This medium was modified to contain 2 mmol/L L-glutamine, 10 mmol/L HEPES, 1 mmol/L sodium pyruvate, 4,500 mg/L glucose, and 1500 mg/L sodium bicarbonate. MDA-MB 231 cells were maintained in a 0% CO₂ atmosphere at 37°C; 4T1 cells were maintained in a 5% CO₂ atmosphere at 37°C. These cells were used for up to 10 passages after reviving from frozen vials and were regularly stained every 3 months with DAPI (Vector Labs) to test for Mycoplasma contamination. Additionally, negativity for Mycoplasma was confirmed by PCR (test no. 18Q397, November 21, 2018, ICLAS Monitoring Center, Japan). The identities of the cells were confirmed by cytochrome c oxidase I (COI) assay for human MDA-MB-231 breast cancer cells and mouse 4T1 mammary tumor cells. When the cells reached 70% to 80% confluence, the DPP-4 inhibitor KR (50 μmol/L) and/or metformin (10 mmol/L) were added to the experimental medium for 48 hours. Cells were harvested for subsequent analyses. All analyses were performed within 6 months of obtaining the cells from the ATCC.

Matrigel Basement Membrane (BD, No. 356234) was thawed on ice overnight at 4°C. The cell culture inserts (0264, Falcon) were coated with Matrigel solution and incubated at 37°C for 1 hour. MDA-MB-231 and 4T1 cells were incubated on the upper layer of a cell culture insert with serum-free medium, and complete medium was added to the 6-well plates below the cell culture insert. Following treatment with KR (50 μmol/L), metformin (10 mmol/L), or KR and metformin for 48 hours, the cells that migrated through the membrane were stained with hematoxylin and counted. Migration was analyzed by microscopy (Carl Zeiss Microscopy, Germany). Photographs at 200 × magnification were captured from 6 different areas and quantified.

All animal experiments were performed in the animal facility of Kanazawa Medical University (detail in Supplementary materials). Female BALB/c mice (8 weeks old) (CLEA Japan Inc., Japan, RRID: MGI: 3586693) were analyzed in the following experiments. We designed the following two animal experiments. In the first experiment, cultured 4T1 cells (5 × 10⁵ cells in 20 μL of PBS) were injected into the mammary fat pad of each mouse. When the primary tumor size reached approximately 300 mm³, the mouse breast cancer model was treated with DPP-4 inhibitor KR (20 mg/kg/day by oral administration), metformin (200 mg/kg/day by intraperitoneal injection), or KR and metformin. PBS was given as control for each drug either oral administration or intraperitoneal injection. After 7 days of intervention, the mice were sacrificed, and the primary tumors and lungs were removed. The protocol used in the present study was changed from that used in our previous study for the following reasons: (i) We treated the tumor-bearing mice with drugs for 7 to 10 days to see the effects of the drug compounds. (ii) The IACUC policies of our university changed such that the new tumor size threshold for sacrificing tumor-bearing mice was 2,000 mm³ (previously 3,000 mm³). According to the preliminary analysis using a new batch of 4T1 cells, if the intervention was initiated when the tumor size reached 500 mm³, the tumor size would exceed approximately 2,000 mm³ in 4 to 5 days (18). Therefore, we initiated DPP-4 inhibitor KR and/or metformin intervention when the primary tumor volume reached 300 mm³. In the second experiment, control shRNA- or shRNA-DPP-4–transduced 4T1 cells were injected into the mammary fat pad of each mouse. When the primary tumor size reached approximately 300 mm³, the control shRNA-transduced 4T1-bearing mice were treated with PBS as the DPP-4 knockdown control (sh-control mice). The shRNA-DPP-4–transduced mice (sh-DPP-4 mice) were treated with PBS or metformin intraperitoneally (200 mg/kg/day). Seven days after initiating treatment, the mice were sacrificed, and the primary tumors and lungs were analyzed. Primary tumor volume was measured using the formula V = 1/2(L × W²), where L is length and W is width of the primary tumor. All animal experiments were approved by the IACUC of Kanazawa Medical University (protocol numbers 2018-16 and 2019-20, lentivirus shRNA in vivo experiment protocol numbers 2018-11 and 2019-7).

Proteins in the total cell lysates (MDA-MB231 and 4T1 cells) were harvested using RIPA lysis buffer containing phenylmethylsulfonylfluoride, protease inhibitor cocktail, and sodium orthovanadate (Santa Cruz Biotechnology) on ice. The amount of protein in each sample was quantified using a standard BCA assay kit (Thermo Fisher Scientific). Lysates were boiled in sodium dodecyl sulfate (SDS) sample buffer at 95°C for 5 minutes, resolved on an SDS-polyacrylamide gel (5%–20%), and transferred onto PVDF membranes (Pall Corporation, Pensacola). The membranes were blocked and incubated with antigen-specific primary antibodies at 4°C overnight and then incubated with peroxidase-conjugated secondary antibodies for 1 hour at room temperature. Bands were detected by using an enhanced chemiluminescence detection system (Pierce Biotechnology) and visualized using an ImageQuant Las 400 camera system (GE Healthcare Life Sciences). Antibodies against the following antigens were used for Western blots: E-cadherin (1:2,000, GeneTex, GTX11512, RRID:AB_381324), α-SMA (1:1,000, Abcam, ab5694, RRID:AB_2223021), CXCL12 (1:1,000, Abcam, ab18919, RRID:AB_444705), CXCR4 (1:1,000, Abcam, ab1670, RRID:AB_302470), DPP-4 (1:500, Sigma, SAB2500328, RRID:AB_10603974), mTOR (1:1,000, CST, 2983, RRID:AD_2105622), phospho-mTOR (1:1,000, CST, 2971, RRID:AB_330970), AMPK (1:1,000, CST, 2532, RRID:AB_330331), phospho-AMPK (1:1,000, CST, 2535, RRID:AB_331250), IGF-R1 (1:1,000, CST, 3027, RRID:AB_2122378), Sp1 (1:500, Sigma, 07-645, RRID:AB_310773), Sp3 (1:250, Santa Cruz Biotechnology, sc-28305, RRID:AB_628274), Sp4 (1:500, Santa Cruz Biotechnology, sc-390124, N.A.), and β-actin (1:10,000, Sigma, A2228, RRID:AB_476697).

Paraffin slides of primary tumor specimens in vivo experiments were deparaffinized in xylene and rehydrated by transferring to decreasing concentrations of ethanol. Next, the slides were incubated in 10 mmol/L citrate buffer, pH 6.0, for 30 minutes (microwave). All sections were incubated with 3% H₂O₂ for 10 minutes to block endogenous peroxidase activity. The specimens were then blocked with normal goat serum for 1 hour and incubated for 1 hour at room temperature with antiphosphorylation-mTOR (1:100, Abcam, ab109268, RRID:AB_10888105), antiphosphorylation-p70 S6 kinase (Thr389) (1:50, Sigma, SAB4503957, N.A.) and anti-Snail (1:500, Abcam, ab180714, RRID:AB_2728773) antibodies. Normal rabbit serum was used to block nonspecific background, and samples were incubated for 1 hour at room temperature with a specific primary antibody against CXCR4 (1:100, Abcam, ab1670, RRID:AB_302470). PBS displaced the primary antibody for the negative control. The corresponding secondary antibodies were added for 30 minutes at room temperature. The IHC mentioned above was performed using a Vectastain ABC rabbit/goat IgG Kit (Vector Laboratories). Then, the slides were incubated with DAB substrate solution (SK-4100, Vector Laboratories) and counterstained with hematoxylin. The slides were incubated with 4 concentrations of ethanol (75%, 95%, 100%, and 100%) for dehydration and hyalinized in xylene. The images were analyzed by a NanoZoomer Digital Pathology System. Photographs at 200× magnification were obtained from 5 different areas of each sample. The images were analyzed with NIH ImageJ Fiji software (USA, RRID:SCR_002285). The pixel intensity values were divided into 4 zones and scored on a scale of 1 to 4.

Lung specimens were treated with 10% formaldehyde and then fixed in Bouin buffer solution (10% formaldehyde: 0.9% picric acid: 5% acetic acid, 15:5:1) for 24 to 48 hours. After Bouin staining, the number of metastatic nodules was counted to quantitate surface lung metastasis.

For each lung, 5-μm sections were prepared and stained with a standard H&E protocol. Lung metastatic burden was determined by counting individual lung nodules on one section.

The levels of CXCL12 in plasma and primary tumor lysates of 4T1 tumor–bearing mice were measured using a mouse CXCL12 ELISA kit (DY460, R&D Systems). CXCL12 was quantitated according to the manufacturer's instructions, and absorbance at 450 nm was measured using a microplate reader (SoftMaxPro, ver. 6.4, Molecular Devices). The results are presented in pg/mL (plasma) or ng/mg (primary tumor).

Mouse shRNA vectors were constructed from the pSIH vector (System Biosciences) using synthetic oligonucleotides against mouse DPP-4 antisense sequences (sh-DPP-4-1: 5′-TAGAAGGAGTATTCAATGAGC-3′ or sh-DPP-4-2: 5′-AATAGTCAGCTA-GTGAATACG-3′) with a loop 5′-CTCGAG-3′. Lentivirus experiments were performed as we previously described (18) and approved by the Kanazawa Medical University Safety Committee for Recombinant DNA experiment (protocol number: 2018-11 and 2019-7).

For gene-expression profiling analysis, primary tumors from our previous report were used (18). The same procedure for the tumor injection was performed as above. However, different from the method stated in the Animal experiments section above, chemical treatment (PBS, KR, and CXCR4 inhibitor AMD3100 [A5602, Sigma, 7.5 mg/kg/day, intraperitoneal injection]) was initiated when the primary tumor measured approximately 500 mm³. Seven days after initiating treatment, mice were euthanized, and the primary tumor tissues were dissected. cRNAs from primary tumors were prepared and applied to a Mouse Gene ST 2.0 microarray (Thermo Fisher Scientific) according to the manufacturer's instructions (n = 3 in each group). Genes that were upregulated or downregulated were extracted using the GeneSpring GX software package version 14.9.1 (Agilent Technologies). A volcano plot was generated, and Student t test and fold change calculation were performed. The fold change threshold was set to 2.0, and the unpaired t test threshold was P = 0.05. Multiple testing correction was not performed. The genes extracted as a result are displayed in a heat map.

GraphPad Prism software (ver. 8.2.0) (RRID:SCR_002798) was used for all statistical analyses. One-way ANOVA with Tukey multiple comparison test and unpaired t tests (only CXCL12 ELISA of primary tumor in drug administration experiment) were used to determine statistical significance, which was defined as a P < 0.05.

All data that support the conclusions are available from the corresponding author upon request.

We previously reported that DPP-4 inhibition induces breast cancer EMT and metastasis through inhibition of CXCL12 cleavage and consequent stimulation of the CXCR4-mediated mTOR pathway in both allograft mammary tumor model mice and in vitro human breast and mouse mammary cancer cell lines (18). In our previous report, CXCR4 inhibitor AMD3100 (AMD) significantly suppressed DPP-4 inhibitor KR62436 (KR)-induced mammary tumor metastasis into the lung. To confirm whether the AMD-reduced metastasis in KR-treated mice was associated with suppression of the mTOR signaling pathway and potentially EMT, we used gene microarray analysis of primary tumors in control, KR-treated, and KR + AMD-treated mice obtained in our previous report (18). We investigated genes associated with mTOR and EMT-related pathways. mTOR pathway–regulated genes [particularly autophagy regulators (Bnip3, Adm) and endoplasmic reticulum stress/molecular chaperones (Hspa1a, Fabp5, Ddit3)] were significantly altered by KR; AMD normalized the gene profile of KR-treated mice to that of the control (Supplementary Fig. S1). Additionally, EMT-related pathway genes [particularly cell adhesion associates (Glycam), autophagy regulators (Bnip3, Adm), inducible transcriptional factors (Atf3, Egr1, Fos), growth factors/chemokines/receptors (Areg, Ftl1, CXCL15, Pf4), inflammation-related genes (Tnfrsf9, Nfkbiz), and molecular chaperones (Hspa1a)] also fluctuated similarly with KR or KR + AMD (Supplementary Fig. S1). These unbiased approaches revealed the relevance of mTOR suppression for mammary tumor metastasis and EMT in mice treated with a DPP-4 inhibitor.

The antidiabetic drug metformin is also known to be a potent mTOR inhibitor (32). First, to confirm whether metformin can attenuate the induction of EMT in response to a DPP-4 inhibitor, we treated metastatic mammary/breast cancer cell lines with the DPP-4 inhibitor KR with or without metformin. Compared with the control, KR decreased the epithelial marker E-cadherin and increased the mesenchymal marker α-SMA, suggesting that EMT occurred in KR-incubated MDA-MB-231 cells, a human breast cancer cell line. Metformin alone increased E-cadherin and decreased α-SMA expression levels compared with those in the control. Under coincubation with KR, metformin also suppressed KR-induced EMT in MDA-MB-231 cells (Fig. 1A). Additionally, 4T1 cells, which are murine mammary tumor cells, were treated with KR and displayed a reduction in the levels of E-cadherin and an increase in the levels of α-SMA compared with those in the control. Metformin suppressed both the basal and KR-induced EMT phenotype in 4T1 cells (Fig. 1B). Transwell migration assays revealed that KR accelerated cell migration in 4T1 and MDA-MB-231 cells; metformin treatment, in the presence or absence of KR, suppressed cancer cell migration (Fig. 1C).

To further explore the mechanism by which metformin suppressed KR-induced EMT and cell migration, we examined the CXCL12/CXCR4–mTOR pathway by Western blot analysis of 4T1 cells. KR significantly increased both CXCL12 and CXCR4; metformin did not alter the basal or KR-increased levels of CXCL12 and CXCR4 (Fig. 1D). Phosphorylation of mTOR was significantly increased by KR; metformin significantly reduced basal and KR-induced mTOR phosphorylation in 4T1 cells (Fig. 1E). Metformin increased the level of AMPK phosphorylation in control and KR-incubated 4T1 cells (Fig. 1E).

Next, we investigated whether the anti-EMT effect of metformin was relevant for tumor metastasis in vivo utilizing an allograft model of 4T1 mammary tumors in BALB/c mice (Fig. 2A). In the treatment period, body weight decreased in all groups of mice (Fig. 2B). KR-treated mice displayed faster primary tumor growth than control mice. Although metformin alone displayed no alteration in tumor growth, metformin significantly suppressed KR-induced primary tumor growth (Fig. 2C). At the time of sacrifice, KR-treated mice exhibited increased volume and weights of primary tumors compared with control mice. Metformin alone did not affect the primary tumor size or weight; metformin on top of KR significantly suppressed KR-induced tumor growth (Fig. 2D and E). H&E staining of primary tumors revealed a poorly differentiated and more invasive phenotype in KR-treated mice compared with control mice. Metformin alone did not affect the pathologic findings of the primary tumor; metformin inhibited KR-induced pathologic alterations (Fig. 2F).

Furthermore, the lung surfaces of KR-treated mice had more visible metastatic nodules than those of the control mice, as we previously reported (Fig. 3A). Compared with the control, metformin alone had no impact on 4T1 mammary tumor lung metastasis; metformin significantly suppressed KR-induced 4T1 tumor metastasis (Fig. 3B). Histologic analysis of the lungs showed that the metastatic tumor area of KR-treated mice was significantly larger than that of control mice (Fig. 3C). Metformin alone did not affect the metastatic area; metformin significantly suppressed KR-induced lung metastasis of 4T1 tumors to basal levels.

Systemic administration of DPP-4 inhibitors induces dynamic alterations in physiologic environments in either an enzymatic or nonenzymatic manner. Furthermore, DPP-4 inhibitors may display nonspecific effects as chemical compounds in cancer biology. Therefore, we next established and utilized shRNA-mediated DPP-4 knockdown tumor–bearing mice (sh-DPP-4 mice) for the in vivo experiment (Fig. 4A). First, we confirmed DPP-4 knockdown efficiency by Western blot analysis in vitro and found that DPP-4 protein levels were reduced in sh-DPP-4 by 70% compared with the control (Fig. 4B). We utilized sh-DPP-4-2 clone for all DPP-4 knockdown analysis. In vivo, when compared with the sh-control mice, sh-DPP-4 mice displayed a reduced trend of body weight; metformin treatment inhibited body-weight reduction in sh-DPP-4 mice (Fig. 4C). In the treatment period, sh-DPP-4 mice displayed accelerated primary tumor growth compared with sh-control mice (Fig. 4D). At the time of sacrifice, the primary tumor volume and weight were significantly increased in sh-DPP-4 mice compared with sh-control mice (Fig. 4E and F). Unexpectedly, in contrast to the effects of metformin on KR-increased primary tumors (Fig. 2C and E), metformin had no significant effect on primary tumor growth in sh-DPP-4 mice (Fig. 4F). H&E staining of primary tumors revealed a more invasive phenotype in sh-DPP-4 mice; metformin inhibited these histologic alterations of primary tumor in sh-DPP-4 mice (Fig. 4G).

Similar to the results in the KR administration experiment (Fig. 3), shRNA-mediated DPP-4 knockdown tumor transplantation also induced mammary tumor metastasis to the lungs; metformin suppressed this metastasis to a level similar to that in sh-control mice (Fig. 5A and B). Furthermore, metformin treatment reduced the enlarged metastatic area in the lungs of sh-DPP-4 mice (Fig. 5C).

To determine whether the effect of metformin on DPP-4 deficiency–induced mammary tumor metastasis is attributed to suppression of the mTOR pathway in vivo, we next performed IHC analysis of primary tumors. Similar to the in vitro experiment (Fig. 1D), the primary tumor in KR-treated mice displayed increased CXCR4 expression, whereas metformin had no impact on the expression level of CXCR4 either with or without KR (Fig. 6A). Compared with the control, the primary tumors of KR-treated mice had a significant increase in phosphorylated mTOR (p-mTOR; Fig. 6B) and phosphorylated-p70 S6 kinase (p-p70S6K), molecular target of mTOR complex 1 (Fig. 6C). Metformin alone did not affect either p-mTOR or p-p70S6K levels; metformin treatment reversed the KR-mediated induction of both p-mTOR and p-p70S6K (Fig. 6B and C). We also examined the alteration of Snail, an EMT-transducing factor, by immunolabeling in primary tumors. KR significantly increased Snail expression compared with the control (Fig. 6D). Metformin alone did not affect Snail expression compared with the control; metformin significantly suppressed Snail levels in primary tumors of KR-treated mice (Fig. 6D). To investigate the direct effect of DPP-4 suppression, we also examined the level of CXCL12, a substrate of DPP-4, in primary tumor and plasma samples of the in vivo experiment. As expected by in vitro experiments, KR increased CXCL12 levels in primary tumors; metformin did not alter the levels of CXCL12 in the primary tumors of either control or KR-treated mice (Fig. 6E). The plasma CXCL12 level was insignificantly elevated by KR but exhibited no alteration in any group (Fig. 6E).

Compared with control shRNA transduction, DPP-4 knockdown in 4T1 mammary tumor cells induced increases in CXCR4, p-mTOR, p-p70S6K, and Snail (Fig. 7A–D). Treatment with metformin abolished this DPP-4 knockdown–induced alteration in all molecules, similar to the level of the control (Fig. 7A–D). We also examined the CXCL12 level of primary tumors and plasma in shRNA-mediated DPP-4 knockdown tumor–bearing mice. The lentiviral system has been shown to display various off-target effects and may affect the CXCL12 signaling pathway as well (33). In our analysis, we also confirmed that primary tumors in mice treated with the lentiviral transduction system displayed approximately 10 times higher levels of CXCL12 than primary tumors in mice without the lentiviral system (Fig. 7E). Despite such high levels in this set of animals, the primary tumors in sh-DPP-4 mice displayed elevated CXCL12 levels compared with those in control mice; metformin suppressed this elevation of CXCL12 in the primary tumors of sh-DPP-4 mice (Fig. 7E). The plasma levels of CXCL12 remained unchanged in all groups of mice in this set of experiments (Fig. 7E).

Preclinical analysis demonstrated that metformin inhibited the mTOR pathway via suppression of specificity protein (Sp) transcription factors and the consequent decreased level of insulin-like growth factor-1 receptor (IGF-1R) in pancreatic cancer cells (34). Western blot analysis revealed that KR increased Sp4 protein level in 4T1 cells (Supplementary Fig. S2). Metformin alone did not alter the level of Sp4; metformin significantly suppressed KR-induced Sp4 expression in 4T1 cells (Supplementary Fig. S2). KR insignificantly induced Sp1 level compared with control; metformin significantly increased the expression of Sp1 either with or without KR (Supplementary Fig. S2). The level of Sp3 exhibited no alteration in all groups (Supplementary Fig. S2). IGF-1R level displayed no alteration in KR-incubated 4T1 cells compared with control; metformin suppressed expression of IGF-1R either in the presence or absence of KR (Supplementary Fig. S3).

As a follow-up to the results of our previous report (18), we performed a further investigation to determine whether metformin, an antidiabetic and potent anticancer agent, can combat cancer metastasis induced by DPP-4 inhibitor use. The current data presented here strongly suggest that metformin exerts an antimetastatic effect on mammary tumor via suppression of the mTOR pathway, especially combined with DPP-4 inhibitor therapy.

Several studies have shown an association between DPP-4 inhibitor use and cancer progression (13, 35). Abrahami and colleagues reported that DPP-4 inhibitors were significantly associated with an increased risk of cholangiocarcinoma among adults with T2DM compared with other antidiabetic drugs (36). Recently, Lee and colleagues reported that DPP-4 inhibitor use can increase pancreatic cancer risk (37). These cohort studies showed the possibility that DPP-4 inhibitors can influence the onset of certain cancers in diabetic patients; however, a meta-analysis of clinical studies demonstrated no evidence of an association between the use of DPP-4 inhibitors and an increased risk of site-specific cancer (38). Similarly, the link between DPP-4 inhibitors and the risk of cancer metastases in T2DM patients is still controversial. Wang and colleagues showed that DPP-4 inhibitors affected the proliferation and migration of various cancer cells, resulting in an increased risk of metastasis (39). On the other hand, an observational study indicated that the use of DPP-4 inhibitors was not associated with a higher risk of metastasis in patients with T2DM and breast cancer (40). Further studies investigating cancer risk with DPP-4 inhibitor treatment are absolutely needed, but even such clinical analysis could be biased for the following reasons: (i) DPP-4 inhibitors are often used as second- or third-line drugs in the treatment of T2DM, and diabetic patients who are prescribed DPP-4 inhibitors have much higher blood glucose levels, supporting the aggressive behavior of cancer cells. (ii) The number of patients using DPP-4 inhibitors is often limited. (iii) The period of follow-up is short to confirm the safety of DPP-4 inhibitors. (iv) For retrospective analysis, cancer stage and its features, such as receptor expression, are usually not documented. Therefore, a basic study is also required, as well as larger clinical studies, to provide evidence for the safety profile of DPP-4 inhibitor use in patients with diabetes and cancer.

We reported that DPP-4 inhibition accelerated 4T1 mammary tumor metastasis via CXCL12/CXCR4/mTOR activation-induced EMT. Metformin has been shown to inhibit the mTOR pathway by both AMPK-dependent and AMPK-independent mechanisms (41–43). The metastasis rate is lower in metformin users than in metformin nonusers among diabetic patients with breast cancer (27). Therefore, we hypothesized that metformin, a potent mTOR inhibitor, can combat breast cancer metastasis induced by DPP-4 inhibitors. We found that metformin increased the phosphorylation of AMPK and suppressed mTOR activation, regardless of combination with a DPP-4 inhibitor, in mammary tumor cells in vitro. Whether inhibition of mTOR by metformin depends on AMPK activation is unknown, and the AMPK dependence or independence of mTOR suppression by metformin is beyond the scope of our study. Importantly, we showed that metformin inhibited DPP-4 deficiency-induced mTOR activation without affecting the CXCL12 level in vitro and in vivo, suggesting that the antimetastatic effect of metformin shown here depends on mTOR inhibition rather than suppression of CXCL12/CXCR4. Mechanistically, Nair and colleagues beautifully reported that Sp transcriptional factors and Sp-regulated IGF-1R are the targets of the anti-mTOR effect of metformin (34, 44). Our data also suggested that the suppression of Sp4 and IGF-1R could be participating in the molecular mechanisms of anti-mTOR effects of metformin in mammary tumor cells in part.

The CXCL12/CXCR4 axis is known to be related to poor prognosis and high recurrence and metastasis rates in cancer patients. CXCL12 secreted in an autocrine manner interacts with CXCR4-positive esophageal cancer cells and maintains the ability of high-level invasion and metastasis (45). Another report showed that the CXCL12/CXCR4 autocrine loop and continuous upregulation of the CXCL12/CXCR4 pathway may be important for cancer metastasis (46). DPP-4 inhibitors can systemically inactivate DPP-4 and inhibit CXCL12 cleavage; thus, DPP-4 inhibitors have the potential to stimulate both autocrine and paracrine secretion of CXCL12 in cancer cells. Therefore, the CXCL12/CXCR4 autocrine loop is strongly activated by pharmacologic DPP-4 inhibition, and the CXCL12/CXCR4 pathway mainly contributes to tumor progression in DPP-4 inhibitor–treated mice. Metformin could inhibit CXCL12/CXCR4-induced tumor proliferation and metastasis via suppression of the downstream mTOR pathway or other pathways. However, DPP-4 knockdown by shRNA stimulates only autocrine CXCL12 secretion because DPP-4 is active except in primary tumors, and the tumor progression mechanism induced by DPP-4 knockdown would not strongly depend on the CXCL12/CXCR4 pathway; therefore, the antitumor growth effect of metformin was limited in the DPP-4 knockdown study (Visual Overview). Furthermore, DPP-4 knockdown might inhibit apoptosis and contribute to the growth of tumors, similar to a study in neuroblastoma cells (47). Because this mechanism could be independent of the anti-mTOR effect of metformin, metformin likely has no influence on tumor growth induced by DPP-4 knockdown itself.

Unexpectedly, our study showed that metformin alone suppressed neither mammary tumor proliferation nor metastasis. Such results were potentially due to the short duration and the late initiation of metformin intervention to see the apparent antitumor effect. Furthermore, as we confirmed in IHC analysis of primary tumors, metformin suppressed aberrant activation of the mTOR signaling pathway in DPP-4–deficient cells, whereas metformin alone had no impact on the basal mTOR pathway. Interestingly, several papers have demonstrated that metformin preferentially targets cancer stem cell–like cells that are closely associated with the EMT phenotype (48). Additionally, mTOR activation is a potent inducer of cancer stemness (49). Even though the mechanistic roles of DPP-4 inhibitors in EMT and/or cancer stemness are largely dependent on cell types (50–52), our current study suggested that metformin can act as an antimetastatic agent against cells with aberrant mTOR activation induced by DPP-4 inhibitor use in certain cancer cells. In addition, antitumor effects of metformin have been shown to be different in the presence or absence of high fat diet (53–57); our experimental model also potentially contributed to the lack of metformin's effect on the tumor progression.

EMT is a critical step for tumor metastasis. It is known that metformin reverses EMT in breast cancer cells (31, 58). A clinical study also showed that higher expression of α-SMA in human breast cancer, an indicator of EMT induction in epithelial cells, is associated with a higher rate of metastasis and a poor overall survival rate (59). In addition, metformin is known to inhibit EMT through AKT-mediated mTOR pathway suppression in glioblastoma cells (60). Our analysis clearly demonstrated that metformin decreased EMT associated with mTOR suppression, further supporting the potential of metformin as a fundamental anti-EMT and antimetastasis drug to combat DPP-4 inhibitor–induced EMT. Furthermore, metformin alone also showed an anti-EMT effect in an in vitro study; therefore, metformin may have antimetastatic potential even administered alone in selected types of cancer. Further study is needed to confirm whether earlier and longer intervention with metformin alone can suppress cancer metastasis compared with the control.

We suggest the possibility that combination therapy with metformin and DPP-4 inhibitor could be much safer than DPP-4 inhibitor alone for T2DM patients with certain types of cancers, especially CXCR4-positive cancer. Interestingly, a recent meta-analysis of the cardiovascular outcomes of three major DPP-4 inhibitors revealed that baseline metformin users experienced a significant trend toward improvements in major cardiovascular events by DPP-4 inhibitors (HR 0.92; 95% CI, 0.84–1.01), whereas baseline metformin nonusers exhibited a reverse trend (HR 1.10; 95% CI, 0.97–1.16; ref. 61). Although the underlying mechanisms are unknown, this result suggests that the pharmacologic effects and biological influences of DPP-4 inhibitors might be completely different depending on whether they are coprescribed with metformin. More recently, the potential favorable effect of a combination therapy of metformin and DPP-4 inhibitor vildagliptin in patients with T2DM was reported; vildagliptin use was associated with trends of increased incidence of pancreatic, prostate, and breast cancers, even combined with metformin use (62). Diabetic patients are often prescribed two or more antidiabetic drugs in the clinic; the influence and interaction of antidiabetic agents in cancer biology are absolutely required further investigation.

A limitation of this study is that BALB/c mice are diabetes resistant; whether this mechanism is relevant for diabetic conditions is not clear. Indirect effects of antidiabetic drugs on cancer progression must be considered in T2DM patients. In this study, the indirect effects are unknown, but the direct effect of metformin on cancer biology, especially when using DPP-4 inhibitor, was significant. Further work is needed to establish relevant recommendations regarding the use of antidiabetic drugs in patients with cancer at specific sites. Moreover, it is important to explore possible mechanistic links between metformin and DPP-4 inhibitor-induced tumorigenesis in humans. Finally, contrary to our study, some studies have shown that DPP-4 inhibitors may reduce certain cancer expansion and metastasis via diverse pathways (63, 64). DPP-4 inhibitors could either favorably or unfavorably influence cancer biology, and physicians need to be aware of potential adverse effects.

In conclusion, our findings demonstrate that metformin significantly inhibits DPP-4 deficiency–induced mammary tumor metastasis by suppressing the mTOR signaling pathway in vitro and in vivo, suggesting the potential of metformin to combat the undesirable effects of DPP-4 inhibitors in patients with certain types of cancers.

K. Kanasaki reports grants from JSPS (23790381 and 26460403), Kanazawa Medical University (S2016-3, S2017-1), Japan Diabetes Foundation (Lilly-Incretin Basic Research Grant), and MEXT Fellowship Program (142309) and other from Boehringer Ingelheim (Japan), Mitsubishi Tanabe, Ono, Taisho-Toyama, and Kyowa-Hakko (establishing the Division of Anticipatory Molecular Food Science and Technology), Boehringer Ingelheim (consultancy agreement), Boehringer Ingelheim (lecture fee), Tanabe-Mitsubishi (lecture fee), Sanofi (lecture fee), Taisho (lecture fee), Lilly (lecture fee), Dainippon Sumitomo (lecture fee), Ono (lecture fee), and Kyowa-Hakko (lecture fee) outside the submitted work. D. Koya reports other from Boehringer Ingelheim Japan (donation course), Taisho Pharmaceutical Holdings (donation course), Ono Pharmaceutical Co., Ltd. (donation course), and Mitsubishi Tnanabe Phrma (donation course) outside the submitted work. No disclosures were reported by the other authors.

E. Kawakita: Data curation, software, formal analysis, investigation, visualization, methodology, writing–original draft, writing–review, and editing. F. Yang: Data curation, supervision, and investigation. A. Kumagai: Supervision and investigation. Y. Takagaki: Supervision, funding acquisition. M. Kitada: Supervision. Y. Yoshitomi: Investigation. T. Ikeda: Investigation. Y. Nakamura: Investigation. Y. Ishigaki: Investigation. K. Kanasaki: Conceptualization, software, supervision, funding acquisition, validation, writing–original draft, project administration, writing–review, and editing. D. Koya: Supervision and funding acquisition.

The authors received lecture honorarium from Dainippon-Sumitomo Pharma, MSD, Astellas, AstraZeneca, Ono, Taisho-Toyama, Tanabe-Mitsubishi, Eli Lilly, Boehringer Ingelheim, Novo Nordisk, Kyowa-Hakko, Sanofi, Novartis, and Takeda. This study was partially supported by grants from the Japan Society for the Promotion of Science awarded to K. Kanasaki (23790381 and 26460403), D. Koya (25282028 and 25670414), and Y. Takagaki (18K16214). This study was partially supported by a Grant for Collaborative Research awarded to D. Koya (C2011-4 and C2012-1), a Grant for Promoted Research awarded to K. Kanasaki (S2016-3 and S2017-1), and a Grant for Assist KAKEN awarded to Y. Takagaki (K2017-16) from Kanazawa Medical University. This study was also partially supported by the Lilly-Incretin Basic Research Grant from the Japan Diabetes Foundation awarded to Y. Takagaki. F. Yang is supported by the Japanese Government MEXT (Ministry of Education, Culture, Sports, Science, and Technology) Fellowship Program (142309). Boehringer Ingelheim (Japan), Mitsubishi Tanabe, Ono, Taisho-Toyama, and Kyowa-Hakko contributed to establishing the Division of Anticipatory Molecular Food Science and Technology. K. Kanasaki is under a consultancy agreement with Boehringer Ingelheim.

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.