Background
Angiogenesis that mediates the formation of new blood vessels serves as a hallmark for cancers [
1], of which the critical role in cancer progression has now been widely accepted [
2]. Hence, anti-angiogenic drugs (AADs) have been extensively developed whose usage constitutes a major modality of anti-tumor therapy [
3]. However, mechanisms underlying AADs-induced antitumor effects remained unclear. Currently, there are two major hypotheses highly relevant to AADs-related antitumor activities. One offers a possible mechanism that cancer cells are killed through the blocking of blood supply by AADs via the inhibition of tumor angiogenesis [
2]. Up until now, this hypothesized tumor-starving mechanism has not been clinically verified. Another hypothesis involves the remodeling of the remaining abnormal vessels [
4,
5], also known as “vessel normalization”. In the latter hypothesis, the drugs not only suppress both the growth and metastasis of the tumor but also enhance the chemosensitization of cancer cells by improving the vascular maturity and functionality, and ameliorating tumor hypoxia [
6].
Conventional therapies targeting tumor angiogenesis is efficacious (in terms of survival benefit) only for some cancers, such as colorectal cancer and renal cell carcinoma, etc., but not for others (e.g. breast cancer, melanoma) [
6,
7]. Anti-angiogenic benefit in term of survival cannot be seen in all patients with cancers [
8,
9], which have been clinically demonstrated to be responsive to anti-angiogenic therapies. For instance, bevacizumab added to chemo-drug did not significantly improve the overall survival of the patients with metastatic breast cancers [
9]. This is partially due to the lack of vascular parameters available for predicting the treatment efficacy [
10]. Moreover, intrinsic and acquired resistance have been shown to even impair the survival benefit already achieved clinically in some cancer patients [
3,
11]. Thus, there is a pressing need for researchers to develop a more effective treatment regimen.
Population- and clinic-based studies have demonstrated the potential anti-proliferative and anti-metastatic activities of the antidiabetic agent metformin, a member of biguanides, when used in cases with malignant diseases [
12‐
14]. Data from preclinical studies have revealed the pleiotropic effects of metformin [
15,
16]. However, the mechanisms of metformin’s effects in carcinogenesis were not fully understood, and more details concerning metformin’s effects should be further studied. The anti-angiogenesis potential of metformin has recently been reported by several laboratories [
17‐
19]. However, little is known to date about if or how metformin remodels the abnormal tumor vasculature, while inhibiting angiogenesis. Since vascular maturity and functionality are closely associated with hypoxia and metastasis [
20], further researches with a focus on the vascular mechanism would be hugely meaningful. Additionally, biguanides also have the potential to enhance the in vivo toxicity of chemo-drug for cancer treatment [
21,
22], but it was still unclear whether this chemosensitization involves a vascular mechanism.
The aim of the present study was to investigate the effects of metformin on vascular maturity and functionality and angiogenesis. Further results of genetic screening imply the deep involvement of platelet-derived growth factor B (PDGF-B) in metformin-induced vessel normalization.
Methods
Cell culture, proliferation, colony formation and migration assays
HUVECs and murine 4T1 and human MDA-MB-231 metastatic breast cancer cell lines were obtained from the American Type Culture Collection (ATCC, Manassas, VA) and cultured in Dulbecco’s Modified Eagle’s medium (DMEM) (Invitrogen) supplemented with 10% fetal bovine serum (FBS). All cell lines used in the study were not listed in the database of commonly misidentified cell lines maintained by International Cell Line Authentication Committee (ICLAC). Cell line Cross-Contamination was tested using the Short Tandem Repeat (STR) genotyping analysis method. Mycoplasma contamination was tested using Myco-Test Kit of MP Biomedicals (No.093030000) every three months, or when the growth rate and morphology of the cell lines were found to be abnormal. All cell lines were cultured and maintained in an atmosphere consisting of CO2 (5%) and the room air (95%) at 37 °C. In vitro migration was analyzed by inoculating cancer cells in the upper chamber of a transwell (Millipore; 8 μm pore insert). To assess in vitro colonization ability, cancer cells were cultured in a 6-well culture plate, which were finally fixed and stained with crystal violet solution (C0775, Sigma). Cellular proliferation rate was measured by counting the number of cells in the culture dish (6 cm diameter) each day.
Chemicals and reagents
Metformin (No.13118), cyclophosphamide (CTX, No.13849), cisplatin (CPT, No.13119) and imatinib (No.13139) were purchased from Cayman Chemical. The active 4-hydroxy-CTX (4-OH-CTX) was obtained by the ozonization of cyclophosphamide using the method described previously [
23,
24].
Protein array assay
To analyze the proteome profile of angiogenesis-related factors, mouse angiogenesis array (ARY015, R&D System) was incubated with fresh cellular lysates containing 1 mg protein according to the manufacturer’s instructions. The RapidStep™ ECL reagent (No.345818, Millipore) was used, and spot intensities were measured using the NIH ImageJ software (Bethesda, MD) for data capture.
Mouse models
All animals were obtained from Animal Experiment Center of Xi’an Jiaotong University. To assess the in vivo tumor growth, 5Х105 4T1 and 2Х106 MDA-MB-231 cells were suspended in 100 μL precoated PBS, and then injected into the fat pad of the 4th left mammary gland of mice (4T1: female BALB/c; MDA-MB-231: female nude mice). After 6–8 weeks, the mice without obvious abnormality in appearance were randomly divided into different groups (8 mice per group) when the mean tumor volume reached about 50–100 mm2. For the observation of chemosensitization, tumor-bearing mice were pretreated with metformin (orally) for 5 days before receiving the intraperitoneal injection of CTX. Before the experiment started formally, the proper sample size that ensures adequate power for a statistical difference was estimated using the following formula: N = 2·[(u0.05) + u(0.10))·S/X]2, “S” indicates standard deviation of the overall sample, and “X” indicates the difference of the mean tumor weight between two groups. Tumor volume was measured with a caliper every two or three days and calculated using the formula V = 0.523•[a2•A] (“a” indicates the minor tumor axis; “A” indicates the major tumor axis). To observe the change of lung metastasis (4 T1) from primary tumors, tumor-bearing mice were fed for at least 28 days after inoculation.
Flow cytometry analysis
Necrotic and late apoptotic cells were labeled with propidium iodide (PI) at a concentration of 5 μg/mL, and PI+ dead cells were identified by Flow Cytometry (BD Bioscience, USA).
PDGF-B knockdown by shRNA
Lentivirus-mediated PDGF-B silencing was performed by transfecting 4T1 cells with control shRNA (against scrambled sequence) or mouse PDGF-B shRNA. The transfection procedure was carried out according to the manufacture’s protocol. Positively transfected cells were selected using puromycin, and the silencing efficiency was investigated with the quantitative real-time polymerase chain reaction (PCR). The primers for detecting mRNA level of mouse PDGF-B (NM_011057.3 → NP_035187.2, CCDS: CCDS27656.1) were as follows: the forward primer: 5`-TCTCTGCTGCTACCTGCGTCT-3`, the reverse primer: 5`- CAGCCCCATCTTCATCTACGG -3`.
Transcriptome sequencing assay
100 mg tissue samples from the 4T1 tumors in metformin-treated or untreated mice were extracted quickly and then put on the ice, and each sample was immediately cryopreserved with liquid nitrogen. Messenger RNA (mRNA) extraction, cDNA synthesis, PCR enrichment, library construction, quality control and sequencing were performed by Beijing Biomarker Corporation (China, Beijing). 3 independent samples in each group were used for gene expression analysis. Heatmaps were presented to show the change of gene expression levels using Prism 7.0 (GraphPad, USA).
Immunofluorescence, Histomorphometry and H&E Staining
Mouse tissues were fixed in 4% PFA for 12 h at 4 °C, and sequentially dehydrated in the 20 and 30% sucrose solutions, respectively. For the 2D and 3D confocal imaging, tissue samples were cut into 6 μm-thick and 40 μm-thick sections, respectively. The prepared sections were stored in a − 80 °C Laboratory Freezer (DW-86L728J, Haier). Single or double immunostaining was performed with the following antibodies. Primary antibodies: CD31 (anti-rabbit, ab28364, Abcam; anti-Rat, ab7388, Abcam), PDGF-B (BA0519–2, Boster), α-SMA (BM0002, Boster), VE-cadherin (No.138101, BioLegend), cl-PARP (#9542, Cell Signaling), PCNA (BM3888, Boster), cl-Caspase-9 (#9542, Cell Signaling), cl-Caspase-3 (#9661, Cell Signaling); NG-2 (R&D, MAB6689). Secondary antibodies: Alexa fluor 488-conjugated Goat anti-Rabbit antibody (A-11008, Invitrogen), Alexa fluor 488-conjugated Goat anti-Rat antibody (A-11006, Invitrogen), DyLight 550-conjugated Donkey anti-Rat antibody (SA5–10027, Invitrogen), Alexa fluor 546-conjugated Goat anti-Rabbit antibody (A-11010, Invitrogen), Alexa fluor 546-conjugated Donkey anti-Mouse antibody (A-10036, Invitrogen), Alexa fluor 647-conjugated Donkey anti-Mouse antibody (A-31571, Invitrogen), Alexa fluor 546-conjugated Goat anti-Rat antibody (A-11081, Invitrogen). Sections were washed with 0.1% PBST, blocked with 5% BSA in PBST at 37 °C for 1 h, and permeabilized with the 0.2% triton X-100 solution for 15–30 min.
For fluorescent 3D-reconstruction, 40 μm-thick sections were treated with the 0.1% trypsin retrieval solution at 37 °C for 15–20 min to get enhanced signal for signal detection. Sections were then incubated with primary antibodies diluted in 5% BSA (0.1% PBST) at 4 °C for no less than 24 h, followed by staining with the appropriate, fluorescently conjugated secondary antibodies. Nuclei were counter-stained by DAPI (2-5 μg/mL) at the room temperature for 15 min before the fluorescent imaging. The fluorescent single- or multi-layer images (2.5–3.5 μm per layer) were obtained using confocal laser scanning microscopy (Leica, German). Software of LAS AF Lite (Leica, German) was used to perform the 3D-reconstrution of CD31 or CD31/α-SMA fluorescent signaling. Furthermore, 5–10 fields (20 x magnification) per tumor were randomly selected and analyzed [
25].
To observe the perfusion status, perfused vessels were labeled by the intravenous injection of 20 mg/kg Rhodamine-labeled lectin (RL-1102, Vector Labs) 15 min before the intracardiac perfusion of 40 ml 4% paraformaldehyde with a flux of 10 ml/min. Tumors were then extracted, fixed in 4% PFA for 1 h at 4 °C, sequentially dehydrated, embedded in OCT (Tissue-Tek #4583, Sakura Finetek, USA), and cut into 6 μm-thick sections. All perfused or un-perfused vessels were immunostained with anti-CD31 antibodies (ab28364, Abcam) followed by staining with Alexa fluor 488-conjugated Goat anti-Rabbit secondary antibodies (A-11008, Invitrogen). For the observation of the vascular leakage of tumors, 100 mg/kg Fluorescein Isothiocyanate (FITC)-conjugated Dextran (70kD, No.53471, Sigma) was intravenously injected into tumor-bearing mice 10 min before tumors were harvested. To detect tumor hypoxia, 60 mg/kg PIMO (Hypoxyprobe Inc.) was intravenously injected into the tail vein of tumor-bearing mice 90 min before tumors were wholly extracted. PIMO+ hypoxic cells in tumor sections were then immunostained with anti-PIMO antibody (Hypoxyprobe Inc.) according to the manufacturer’s instructions.
The H&E-stained paraffin sections (4 μm) of those fixed tumors or lung tissues were assessed for tumor necrosis, hemorrhage and metastatic nodules. To observe the chemosensitization, tumor sections from Cisplatin-treated mice were stained with anti-Cisplatin-modified DNA antibody (GTX17412, Genetex). Paraffin-embedded sections were sequentially incubated with secondary antibodies (SV0002, Boster), 3, 3′-diaminobenzidine (DAB, ZLI-9017, ZSGB-Bio) and the hematoxylin (H9627, Sigma) solution.
Statistical analysis
Quantitative analysis was performed using the Prism 5.0 or 7.0 software (GraphPad, San Diego, CA). All quantitative data were represented by mean ± SEM. Kolmogorov-Smirnov normality test was performed to analyze the normal distribution, and coefficient of variation (CV) was used to estimate the variation of data within each group. When CV was greater than 15%, the data was considered to be abnormal. Bartlett’s test was performed for investigating the homogeneity of variances between the groups. For any set of data which was not normally distributed, nonparametric Wilcoxon or Kruskal-Wallis test was performed to investigate the statistical difference between two or multiple independent samples. The statistical significance between two groups and multiple groups was defined as P < 0.05 by two-tailed student’s t test or one-way ANOVA t-test. Two-way ANOVA analysis was performed when an additional factor or variant was involved in the experiment.
Discussion
Bevacizumab, an AAD approved at an early time, did not significantly improve the overall survival of patients with metastatic breast cancers [
9,
39]. Bevacizumab was initially designed to neutralize the VEGF, thus inhibiting VEGFR-2-mediated angiogenesis and tumor growth. However, as previously reported, metastatic breast cancer cells had high expressions of several other pro-angiogenic factors in addition to VEGF [
40], such as FGF-2, Ang-2 and PDGF-B [
37]. Thus, in theory, AADs designed for targeting a single factor may be not enough [
41], which offers an explanation for AAD’s ineffectiveness in treating cancers derived from some organ systems. Our laboratory previously found that metformin inhibited the expression of VEGF, Ang-2 and FGF-2 in a metastatic breast cancer model [
37]. Besides, PDGF-B was screened out by transcriptome sequencing, and angiogenesis was greatly inhibited by PDGF-B knockdown. This result was validated by the clinical data that the PDGF-B expression level was positively associated with the CD31 expression level. Due to these findings, metformin should be considered as a reagent with a broad range of targeted factors possibly more available than the conventional AADs.
Another disadvantage of traditional AADs is the excessive pruning of the tumor vasculature [
20], thus leading to the hypoxia-mediated tumor cell dissemination [
42,
43]. Compared to AADs, metformin has more angiogenic targets, such as VEGF, PDGF-B and FGF-2. Therefore, metformin is assumed to prune the breast cancer vasculature more excessively. However, to date, it has not been reported that metformin aggravated tumor hypoxia [
44]. Consistently, metformin ameliorated the hypoxia of two metastatic breast cancers, while MVD was greatly reduced. These evidences suggest there exists a mechanism independent of affecting the vasculature, which is accountable for the ameliorated hypoxia. Metformin is an AMPK activator that induces energetic stress [
45]. In this condition, tumor cells were metabolically reprogrammed to consume less oxygen, thus counteracting or reversing the vascular pruning-induced hypoxia [
44].
Currently, the issue of tumoral PDGF-B’s effect on the vascular maturity has become controversial [
46]. Platelet-derived growth factor B (PDGF-B), a member of the PDGFs family [
46], binds to its receptors, such as PDGF receptor β (PDGFR-β), to induce the cell survival, proliferation and migration [
47]. It is now widely accepted that the endothelial PDGF-B regulates the recruitment of PCs [
47]. Thus, the downregulation of PDGF-B should reduce the vascular maturity. However, PDGF-B was highly expressed by some tumors [
48,
49], which were inversely characterized by an excessively angiogenic and immature vasculature. Further blockade of PDGF-B/PDGFR-β significantly increased the vascular maturity of tumors with high PDGF-B expression [
49], but reduced that of tumors with low PDGF-B expression. These evidences indicated that high and low expression of PDGF-B in tumors might have opposite effects on vascular maturity. To date, this mechanism has not been reported in studies on metastatic breast cancers. Herein, it was found that high PDGF-B expression was detected in the metastatic breast cancer model. By reducing tumoral PDGF-B, the metformin treatment resulted in the suppression of angiogenesis and a more mature vasculature of metastatic breast cancers, thus limiting the distant metastasis and improving chemosensitization. These evidences are further supported by the poor prognosis of patients with breast cancers of high PDGF-B expression. These data indicate that the downregulation of PDGF-B in tumors with high expression of PDGF-B inhibits angiogenesis and improves the vascular functionality and maturity.
As early as in 1970’s, biguanides were reported to potentiate the antitumor effects of CTX and other chemo-drugs in vivo [
21]. As shown in a recent clinical trial [
50], diabetic patients receiving metformin had a greater response rate to chemo-drug than non-diabetic patients. Despite the increasing efforts that have been made [
22], it was still unclear whether or how metformin sensitizes cancer cells to chemo-drugs. Recently, metformin has been demonstrated to directly enhance the toxicity of chemo-drugs [
51‐
53]. However, concentrations used in those studies were higher than the blood concentration of patients. In the current paper, metformin was not found to significantly inhibit the proliferation and migration of 4 T1 cancer cells at the blood concentrations in vitro, indicating an indirect chemosensitization mechanism. Furthermore, our results suggest metformin-mediated chemosensitization resulted from the enhancement of drug delivery rather than the direct enhancement of the toxicity of chemo-drug.
Interestingly, metformin pretreatment results in a response of metastatic breast cancer cells to CTX at a lower dose, which was further supported by the increased CPT nucleus adduct
+ cells. Critically, those CTX-induced apoptotic cells were located to the regions adjacent to vessels. Thus, metformin’s chemosensitization effect may be due to the increased delivery of chemo-drugs to the deep tumor. Vessel normalization activity increases the tumor oxygen and reduces the interstitial fluid pressure [
36,
54], thus enhancing the generation of oxygen radicals and promoting egress of cytotoxic agents to the perivascular region in tumors. Consistent with the results from other laboratories [
44], metformin treatment increased tumor oxygenation, reduced tumor hypoxia and improved radiotherapy response. Furthermore, metformin-induced chemosensitization may be contributed by the reduction in tumoral PDGF-B, which mediates resistance by PDGFR-β signaling and increasing IFP [
55]. It has also been reported that metformin’s chemosensitization effect was contributed by increased uptake of chemotherapeutic by tumor cells [
56].
Given that metformin has long been used for the treatment of type 2 diabetes mellitus, the results of this research is remarkable in terms of drug re-positioning (DR) [
57]. DR is a screening for anti-cancer therapeutic effects of conventionally administered medications for non-malignant disorders, which has attracted a great deal of attention as the safety and frequency of side effects of these medicines have been already proven. For a typical instance, ticlopidine (purinergic receptor P2Y12 inhibitor), which is an anti-coagulant drug to prevent the transient ischemic attack (TIA) and stroke, and has been shown to be effective for low-grade glioma and high-grade astrocytoma. This P2Y12 inhibitor increases the intracellular cAMP level and promotes the autophagy flux [
58]. Notably, tricyclic antidepressants such as imipramine promote autophagy in glioma cells synergistically with this drug by further elevating the intracellular cAMP concentration [
59].
Metformin activates AMPK signal pathway, which not only decreases insulin resistance in type 2 diabetes mellitus but also blocks AMPK-mediated mTOR activation even in cancer stem cells (CSCs) [
60]. mTOR signal is regulated by amino-acid transporters [
61], characterized by the L-type amino acid transporter 1 (LAT1; SLC7A5) and the glutamine/amino acid transporter (ASCT2; SLC1A5), which explains why the AMPK-mTOR axis functions as a sensor of the dynamic change in the nutrient/growth factor microenvironment. Particularly, the leucine uptake via LAT1 activates the mTOR signal pathway leading to poor prognosis. Because EpCAM is a functional CSC marker that forms a complex with amino-acid transporters such as LAT1 [
62], it is reasonable that the LAT1 expression level would be positively correlated with poor prognosis. Therefore, the LKB1-AMPK-mTOR axis is orchestrated by the amino-acid concentration in the tumor microenvironment, and this axis promotes the metabolic reprogramming of cancer cells in response to the microenvironment [
58,
63].
More clinical and pre-clinical evidences should be provided to validate the vascular mechanism, and if metformin targets a broad range of angiogenesis-related factors. As metformin is a drug widely prescribed for metabolic disorder, further efforts should also be devoted to investigating the involvement of the metabolic mechanism and its contribution to ameliorated hypoxia.