Glyoxalase 1 sustains the metastatic phenotype of prostate cancer cells via EMT control

Abstract Metastasis is the primary cause of death in prostate cancer (PCa) patients. Effective therapeutic intervention in metastatic PCa is undermined by our poor understanding of its molecular aetiology. Defining the mechanisms underlying PCa metastasis may lead to insights into how to decrease morbidity and mortality in this disease. Glyoxalase 1 (Glo1) is the detoxification enzyme of methylglyoxal (MG), a potent precursor of advanced glycation end products (AGEs). Hydroimidazolone (MG‐H1) and argpyrimidine (AP) are AGEs originating from MG‐mediated post‐translational modification of proteins at arginine residues. AP is involved in the control of epithelial to mesenchymal transition (EMT), a crucial determinant of cancer metastasis and invasion, whose regulation mechanisms in malignant cells are still emerging. Here, we uncover a novel mechanism linking Glo1 to the maintenance of the metastatic phenotype of PCa cells by controlling EMT by engaging the tumour suppressor miR‐101, MG‐H1‐AP and TGF‐β1/Smad signalling. Moreover, circulating levels of Glo1, miR‐101, MG‐H1‐AP and TGF‐β1 in patients with metastatic compared with non‐metastatic PCa support our in vitro results, demonstrating their clinical relevance. We suggest that Glo1, together with miR‐101, might be potential therapeutic targets for metastatic PCa, possibly by metformin administration.

acids and proteins to form the heterogeneous family of advanced glycation end products (AGEs). 5 MG-derived dicarbonyl adducts exert complex pleiotropic effects, including modulation of protein biological activity 6 and stability, 7 generation of reactive oxygen species and oxidative stress, 8,9 which may culminate in distinct biological outcomes. [9][10][11][12][13][14] Hydroimidazolone (MG-H1) and argpyrimidine (AP) are major AGEs formed by spontaneous reaction between MG and protein arginine residues. 15 The levels of MG-derived AGEs, as well as of AGEs generated by the reaction of the free amino groups of N-terminal amino acids with carbonyl groups of reducing sugars, have been investigated in different human cancer tissues. [16][17][18] Results from these studies suggest that the expression pattern of AGEs can be tumour specific 16 and that their accumulation in cancer tissues may be linked to tumour aggression. 19 Specifically, in PCa, accumulation of a specific AGE (carboxymethyl lysine, CML) has been recently found in malignant compared with normal tissues where it positively correlates with tumour aggressiveness. 20 AGEs function as ligand activators for the transmembrane receptor for AGEs (RAGE) 21 which is overexpressed in a variety of human tumours, including PCa. [20][21][22][23] Intriguingly, recent data demonstrate that RAGE can regulate Glo1 expression at both mRNA and protein levels. 24 Glo1 is overexpressed in several human cancers, 5 where it represents an important tumour survival strategy by preventing accumulation of cytotoxic MG, thereby suppressing MG-mediated glycation reactions leading to AGE formation. In PCa, we and others have previously shown that Glo1 plays a major role in the progression of this neoplasia. 25,26 In particular, in highly invasive and metastatic human PC3 PCa cells, Glo1 acts as a pro-survival factor by eluding apoptosis in a mechanism involving AP and NF-kB signalling pathway. 27 An important determinant of metastasis is the epithelial-tomesenchymal transition (EMT), a dynamic transdifferentiation during which cells acquire a migratory and invasive phenotype. 28 The mechanisms that control the process of EMT in cancer cells are complex and still emerging. In PCa, several key factors contributing to EMT 29 have recently been identified; transforming growth factor beta (TGFb) and miRNAs 30 have been shown to play a crucial role. In addition, we have recently demonstrated that the Glo1/AP axis can control EMT in a human bronchial model. 11 The possible role of the Glo1/ AP axis in the control of EMT in metastatic PCa has never been investigated. Understanding this circuit in the context of PCa might identify Glo1 as a novel therapeutic target for this lethal stage of the disease. Hence, in this study, we studied whether and how Glo1 might sustain PCa metastatic phenotype as part of the molecular events associated with EMT in DU145 and PC3 human PCa cell lines, models of metastatic PCa. Moreover, to examine this regulatory circuit in a clinically relevant setting, we measured the circulating levels of some of the involved molecules in patients with metastatic compared with non-metastatic PCa. Finally, we examined the capacity of metformin, a widely used anti-diabetic drug and an emerging anti-cancer drug, especially in the field of urologic oncology, [31][32][33] to inhibit the EMT-based metastatic phenotype, through control of Glo1 and miR-101 in both PCa cell lines.

| Materials
Tissue culture media, foetal bovine serum, penicillin/streptomycin, BCA kit and Laemmli buffer were from ThermoFisher Scientific (Milan, Italy); Roti-Block was from Roth (Germany). Primary antibodies against Glo1, MMP-2, MMP-9, Smad4, lamin b and b-actin were from DBA (Milan, Italy); primary antibodies against E-cadherin (E-cad), zonula occludens-1 (ZO-1), vimentin (Vim), N-cadherin (N-cad), Snail, TGF-b1, phospho-Smad2, phospho-Smad3 and total Smad2 or Smad3 were from Cell Signalling Technology (Danvers, MA, USA). Anti-AP mAb was from Antibodies-online GmbH (Aachen, Germany). The ELISA kit for human TGF-b1 was purchased from R&D Systems (Milan, Italy). The ELISA kit for MG-H1 was from DBA (Milan, Italy). Aminoguanidine bicarbonate (AG), a scavenger of free MG 34 (1 mmol/L, for 3 hours), and metformin (5, 10 and 20 mmol/L for 48 hours) were from Sigma-Aldrich (Milan, Italy). The TGF-b type I receptor inhibitor 35,36 SB431542 (0.5 lmol/L for 1 hour) was from Tocris (Milan, Italy). Control cells for the experiments with agents dissolved in non-aqueous solvents showed no significant difference with respect to control cells in RPMI-1640 medium; therefore, all the relative treatments were compared with these latter controls. The biochemical evidence supporting the efficacy of the inhibitors and scavenging agents used in this study was always tested in preliminary experiments, whenever appropriate (data not shown).

| Cell lines and cell culture conditions
The human brain metastasis-derived DU145 and bone metastasisderived PC3 cancer cell lines were obtained from the American Type Culture Collection (ATCC) and cultured as per the suppliers' recommendations at 37°C and 5% CO 2 .

| RNA isolation, reverse transcription and
qRT-PCR analysis RNA isolation, reverse transcription and qRT-PCR analysis were performed as previously described. 10,27 Briefly, total cellular RNA was iso-

| Glyoxalase 1 enzyme activity
The activity of Glo1 was assessed as previously described in lysates either from cell lines 40 or from red blood cells. [41][42][43] Glo1 enzymatic activity was assayed by an established method. 44 Briefly, the assay solution contained 0.1 mol/L sodium phosphate buffer, pH 7.2, 2 mmol/L MG and 1 mmol/L GSH. The reaction was monitored spectrophotometrically by following the increase in absorbance at 240 nm and 25°C. One unit of activity is defined as 1 lmol of S-Dlactoylglutathione produced per minute. blotting and spectrophotometric assay, or qRT-PCR, respectively ( Figure 1A). The same experimental design was used for glyoxalase 2 (Glo2) gene silencing (data not shown). Mock transfection was performed to control potential effects due to the transfection reagent. As no significant differences were found between siCtr-or mock-treated and non-transfected cells on the biological phenomena under investigation ( Figure S1) in either DU145 or PC3 cells, the observed changes were shown with respect only to siCtrexposed cells.

| Transwell migration and invasion assays
Transwell migration and invasion assays were carried out as described by Malkoski et al. 46 Briefly, transwell assays were con-  The histograms indicate mean AE SD of three different cultures, and each was tested in triplicate. Distance scale is 10 lm. siCtr: control (nonspecific siRNA). **P < .01, ***P < .001 vs siCtr cells

| Immunofluorescence microscopy
Immunofluorescence for F-actin detection was performed as previously described. 11,49 Briefly, the cells grown on a coverslip were 2.14 | Affinity purification of MG-modified proteins and amino acid sequence analysis Affinity purification of MG-modified proteins and amino acid sequence analysis was performed according to Sakamoto et al. 53 Briefly, DU145 and PC3 cells (10 9 cells) were lysed in buffer A (50 mmol/L Hepes-KOH, pH 7.0, 0.1% CHAPS, 2 mmol/L EDTA, 10 mmol/L dithiothreitol, 10% glycerol) by Dounce homogenization.
The lysate was clarified through successive centrifugation steps (a 1000 9 g spin followed by a 10 000 9 g spin) at 4°C. Mouse mono-

| Plasmid constructs and luciferase reporter assay
To construct the Glo1 expression plasmids, the wild-type or mutant 3 0 UTR of Glo1 gene was cloned into the pEZX vector. All clones were   Moreover, Glo1 silencing reduced MMP-2 and MMP-9 protein expression and activity, while these remained unaffected in siCtrtreated DU145 or PC3 cells ( Figure 1D). Finally, Glo1 silencing also resulted in the reorganization of the cytoskeleton ( Figure 1E). In particular, immunofluorescence staining of F-actin showed a clear decrease in staining intensity and reduction of actin stress fibres in siGlo1-treated compared with siCtr-treated DU145 and PC3 cells ( Figure 1E). Collectively, our findings showed that Glo1 sustained the metastatic phenotype of the human DU145 and PC3 PCa cell lines by EMT control.

| Glo1-dependent control of EMT occurs via
hydroimidazolone (MG-H1) and AP depletion in DU145 and PC3 cells To investigate a possible mechanism by which Glo1 sustained the metastatic phenotype of DU145 and PC3 cells via EMT control, we focused on MG-H1 and AP, AGEs generated from the spontaneous modification of arginine residues 15 by MG, the substrate of Glo1.
Besides, AP, like other AGEs, has been reported to be involved in EMT control in a context-dependent manner. 11,60 With an ELISA kit specific to MG-H1 and an antibody specific to AP-modified residues to use in Western blotting, we detected the intracellular levels of these two adducts. As illustrated in Figure 2A, compared with siCtr cells, siGlo1 treatment induced MG-H1 and AP accumulation in both DU145 and PC3 cells. In particular, as to AP immunodetection, we observed a major band with an approximate molecular weight of 40 kD whose intensity was markedly increased by siGlo1 treatment.
To further test the hypothesis that MG-H1 and AP play a key role in the phenotypic changes associated with siRNA-mediated Glo1 depletion, we next examined the effect of AG, a scavenger of free MG, 61 on various aspects of the EMT phenotype lost upon siGlo treatment.
Thus, as depicted in Figure 2, upon Glo1 silencing, AG exposure was able to restore the EMT-associated phenotype, at the level of or secretion ( Figure 3C) of TGF-b1 and Smad4 activation ( Figure 3D) were all significantly inhibited compared with control (siCtr) cells.
Moreover, pre-treatment with AG restored TGF-b1 expression (Figure 3A,B,C) and Smad4 activation ( Figure 3D) to the control levels, indicating that in metastatic PC3 cells, Glo1-dependent MG-H1 and AP depletion was able to drive TGF-b1/SMAD4 signalling pathway.
No differences were observed between siCtr and untreated (Ctr) cells (data not shown). We then wanted to assess whether TGF-b1/ Smad4 signalling was required for the phenotypic changes to PCa cells upon Glo1 knock-down by the TGFb1, inhibitor SB431542.
Altogether, our results showed that Glo1-dependent MG-derived MG-H1 and AP depletion triggered EMT via TGF-b/Smad signalling pathway in metastatic PCa cell lines. showed that like TGFb1, these EMT-associated proteins were not subject to MG modification to form AP (data not shown).

| Loss of tumour suppressor miR-101 induces Glo1-dependent EMT in DU145 and PC3 cells
Loss of tumour suppressor microRNAs (miRNAs or miR) is an established mechanism in cancer progression. It has been reported that reduced expression or loss of miR-101 is associated with metastasis of PCa cells. 69,70 We have recently demonstrated that miR-101 down-regulates Glo1 expression. 10 (Table 2).

| Metformin affects DU145 and PC3 cell metastatic phenotype, inhibits Glo1 and induces miR-101 expression
Metformin is the most widely used anti-diabetic drug in the world, 31 but there is increasing evidence of its potential efficacy as an anticancer drug. 31 In particular, recent studies have suggested that metformin may have benefits especially in the field of urologic oncology. 32,33 It has been reported that Glo1 can be inhibited by metformin in endometrial cancer cells 72,73 and that miR-101, typically lost in pancreatic cancer, can be re-expressed following metformin treatment, and in this setting, it is associated with inhibition of cell proliferation, cell migration and invasion, and self-renewal capacity of cancer stem cells. 74 Therefore, we wanted to investigate whether metformin could also down-regulate Glo1 and up-regulate miR-101 in our models and whether these changes might be causally linked with the rescue of the metastatic phenotype of DU145 and PC3 cells in terms of EMT, migration and invasion. In PC3 cells, we found that metformin significantly inhibited Glo1 expression ( Figure 8A) and increased miR-101 expression ( Figure 8B) in a dose-dependent manner compared with untreated control cells. Moreover, this was associated with a reversal of EMT ( Figure 8C), decreased MMP-2/MMP-9 expression ( Figure 8D), as well as decreased migratory and invasive potential (Figure 8E). More importantly, following metformin administration, miR-101 inhibition restored Glo1 activity and migration/invasion (Figure 8F), and similarly, Glo1 overexpression did ( Figure 8G). Similar results were obtained in DU145 cells ( Figure S7).

| DISCUSSION
The dissemination of PCa is a common, incurable aspect of the advanced disease and the primary cause of death in PCa patients.
Prevention and treatment of this terminal phase of PCa require improved understanding of its molecular aetiology to obtain more insights into how to decrease morbidity and mortality in this disease.
In the present study, we have, for the first time, demonstrated that  MMPs are a family of endopeptidases required for extracellular matrix degradation. Among MMPs, MMP-2 and MMP-9 play important roles for basement membrane type IV collagen degradation during cancer progression, especially for promoting tumour migration and invasion. 88 These two MMPs are well expressed in human PCa and PCa cell lines 89,90 where their expression levels positively correlate with metastatic disease. 59 In the present study, we also demonstrated, for the first time, that Glo1/MG-H1-AP/TGF-b1/Smad axis controls MMP-2 and MMP-9 expression and activity.
MicroRNAs (miRNAs) are small non-coding RNAs modulating gene expression at both post-transcriptional and post-translational levels.
Growing evidence suggests that miRNAs are important regulators of EMT 91,92 and that loss of tumour suppressor miRNA is an established mechanism in cancer progression. 69 In particular, the tumour suppressor miR-101 is down-regulated in PCa and this attenuated expression is an important event in oncogenesis. 70 Moreover, the loss or reduced expression of miR-101 is associated with PCa cell metastasis. 69,70 Here, we demonstrated that reduced miR-101 expression, by main- define a novel mechanism, based on miR-101/Glo1/MG-H1-AP/TGFb/Smad axis, in the molecular aetiology of metastatic PCa (Figure 9), further extending our knowledge of the mechanisms underlying PCa metastasis and suggesting miR-101 and Glo1 that orchestrate the mechanism, as novel potential therapeutic targets for metastatic PCa.
In this regard, it is significant that we show that metformin is able to inhibit Glo1, reactivate miR-101, and inhibit EMT, migration and invasion of metastatic PCa cells, opening new avenues of investigation for a novel potential mechanism, which up to now remain poorly explained, 74,106,107 by which this drug might control metastatic phenotype in PCa. Recently, there has been debate as to how data from in vitro studies of metformin translate into in vivo activity in clinical trials, as the majority of work in vitro has been performed with concentrations varying from 1 to 100 mmol/L, predominantly between 1 and 20 mmol/L. [108][109][110][111][112] Such concentrations would exceed levels measured in the blood of diabetic patients, 113 although many organs such as liver, kidney or small intestine are exposed to much higher concentration of metformin compared with levels measured in the serum. 114 Importantly, plasma membrane monoamine transporter or equilibrative nucleoside transporter ENT-4 facilitates metformin absorption from the lumen. 115 Other studies show the importance of the growth. 109 Similarly, other studies showed that treatment of cancer cells with metformin at low concentrations (30 lmol/L) decreased their invasive capacity. 119,120 In conclusion, we demonstrate here that Glo1 sustains metastatic

CONFLI CT OF INTEREST
The authors confirm that there are no conflict of interests.

AUTHOR CONTRI BUTION
CA, RC, FR and MP performed the research; CA designed the research study; CA and VT analysed the data; CA wrote the manuscript.