Changes in the expression of cancer- and metastasis-related genes and proteins after metformin treatment under different metabolic conditions in endometrial cancer cells

Research question Hyperinsulinemia and elevated estrogen levels are known risk factors for endometrial cancer (EC) development and are associated with obesity, type 2 diabetes mellitus (T2DM), insulin resistance, among others. Metformin, an insulin-sensitizing drug, displays anti-tumor effects in cancer patients, including EC, but the mechanism of action is still not completely understood. In the present study, the effects of metformin on gene and protein expression were investigated in pre- and postmenopausal EC in vitro models in order to identify candidates that are potentially involved in the drug's anti-cancer mechanism. Design After treating the cells with metformin (0.1 and 1.0 mmol/L), changes in the expression of >160 cancer- and metastasis-related gene transcripts were evaluated with RNA arrays. A total of 19 genes and 7 proteins were selected for a follow-up expression analysis, including further treatment conditions, in order to evaluate the influence of hyperinsulinemia and hyperglycemia on metformin-induced effects. Results Changes in the expression of BCL2L11, CDH1, CDKN1A, COL1A1, PTEN, MMP9 and TIMP2 were analyzed on gene and protein level. The consequences resulting from the detected expression changes as well as the influence of varying environmental influences are discussed in detail. With the presented data, we contribute to a better understanding of the direct anti-cancer activity of metformin as well as its underlying mechanism of action in EC cells. Conclusions Although further research will be necessary to confirm the data, the influence of different environmental settings on metformin-induced effects could be highlighted with the presented data. Additionally, gene and protein regulation were not similar in the pre- and postmenopausal in vitro models.


Introduction
Obesity, hyperglycemia and related type 2 diabetes mellitus (T2DM), as well as other associated diseases such as polycystic ovary syndrome (PCOS) and insulin resistance lead to hyperinsulinemia and/or elevated estrogen levels, which are known risk factors for the development and progression of endometrial cancer (EC) [1][2][3][4].
Metformin, an insulin-sensitizing biguanide agent used in T2DM therapy [5], displayed anti-cancer effects in numerous in vitro studies, but also in diabetic patients with EC and other cancer entities [6][7][8][9][10][11][12]. During T2DM treatment, metformin reduces circulating glucose and insulin levels via a blockage of gluconeogenesis in the liver and via improvement of peripheral insulin sensitivity by opposing the action of glucagon, increasing glucose uptake in the muscle while decreasing glucose absorption in the small intestine, and inhibiting lipolysis in adipose tissue [13][14][15][16]. Metformin's anti-cancer effects target signaling pathways related to cellular growth and proliferation via activation of 5 ′ adenosine monophosphate-activated protein kinase (AMPK), leading to a subsequent induction of tumor suppressors as well as a regulation of growth-related pathways [17]. Additionally, metformin acts independent of AMPK, e.g. via ras-related guanosine triphosphate hydrolases (Rag GTPases) or mitochondrial respiratory-chain complex 1 [17][18][19][20]. However, the mechanism of metformin's direct anti-tumor activity is still not completely understood.
In the present study, the effects of metformin on gene and protein expression were investigated in the human EC cell lines HEC-1A and Ishikawa in order to identify candidates that are involved in the drug's anti-cancer mechanism in EC. HEC-1A and Ishikawa cells represent different EC in vitro model systems, namely a pre-(Ishikawa) and a postmenopausal (HEC-1A) model. HEC-1A cells are characterized by a poor expression of estrogen receptor α (ERα) [21,22] and maintain low cellular estrogen sensitivity due to the expression of ERβ and G protein-coupled estrogen receptor 1 (GPER) [23]. Ishikawa cells are defined by an intact expression of ERα/β, GPER and progesterone receptor (PR), equipping the cell line with an unaffected sensitivity to β-estradiol (E2) [12,[23][24][25][26]. The design of the current in vitro study took different clinical settings into account that are associated with EC development and progression. Firstly, insulin supplementation was incorporated into the study in order to implement a model for insulin resistance and related hyperinsulinemia, both often observed in obese, prediabetic or PCOS patients that are prone to EC development [27,28]. Secondly, cells were treated in the presence of normal (5.5 mmol/L) or elevated glucose concentrations (17.0 mmol/L) to be able to evaluate the influence of hyperglycemia on metformin-induced effects. Thirdly, cells were treated in the presence of E2 in order to mimic increased estrogen levels that contribute to EC development and progression and, like hyperinsulinemia, result from aforementioned diseases [1,2,29].
After treating the cells with metformin for 7 d, changes in the expression of >160 cancer-and metastasis-related gene transcripts were evaluated with the help of RNA arrays, of which 19 candidates, namely APC, BCL2L11, CASP8, CDH1, CDKN1A, CEACAM1, COL1A1, CTNNA1, IGF1, PTEN, RAC1, TGFB1, MMP2/9/10 and TIMP1/2/3/4, were selected for a follow-up quantitative real-time PCR (qPCR) analysis, including further treatment conditions. Afterwards, changes on protein level were additionally investigated by Western blot analysis for those selected 7 candidates that had shown the most prominent and favorable changes in gene expression analysis, namely BCL2L11, CDH1, CDKN1A, COL1A1, PTEN, MMP9 and TIMP2. The consequences for the cells resulting from the detected gene and protein expression changes as well as the influence of varying environmental influences are discussed in detail.
With the presented data, several genes and proteins were identified that might be involved in the direct anti-cancer mechanism of metformin in EC. However, further pathway analysis will be necessary in order to identify upstream regulators and downstream targets of the identified candidate genes and proteins, e.g. with regards to promising candidates such as the apoptosis-related BCL2L11 protein, the cell cycle mediator CDKN1A, or metastasis-associated MMP9 and TIMP2.
Cells were treated with either 0.1 or 1.0 mmol/L metformin (100 mmol/L stock in H 2 O; Sigma-Aldrich), 100 ng/mL insulin (10.0 μg/mL stock in phosphate-buffered saline (PBS); Sigma-Aldrich) or a combination of metformin and insulin under normo-(5.5 mmol/L glucose) or hyperglycemic conditions (17.0 mmol/L glucose, equivalent to 300 mg/dL) in the presence of 10 nmol/L E2 (all purchased from Sigma-Aldrich) in phenol-red-free medium for 7 d with medium changes and renewed treatments every 2-3 d. The selected metformin concentrations induced no loss of cellular viability in the ATP, and a mild to moderate loss of cellular viability in the MTT (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide) viability assays under similar experimental conditions in an in vitro study recently published by our group [12] and therefore ensured to maintain subtoxic conditions throughout the duration of the treatment. An insulin concentration of ≥50 ng/mL is well established in vitro studies in the field of endocrinology and diabetes research in order to mimic hyperinsulinemia [30][31][32][33] and has also been used by our group before [34][35][36]. Control cells were treated with substance-free medium supplemented with 10 nmol/L E2 as well as H 2 O and PBS (Sigma-Aldrich) as vehicles.

RNA extraction and cDNA synthesis
After 7 d, the culture medium was removed and the cells were washed with PBS twice. RNA was extracted from the aqueous phase after addition of 500 μL/well TRIzol reagent (Invitrogen, Waltham, MA, USA) and 200 μL chloroform (Sigma-Aldrich). Afterwards, total RNA was precipitated with 500 μL isopropanol containing 1,6 μL GlycoBlue co-precipitant (Invitrogen) at − 20 • C for 2 h, washed with 1 mL 75% (v/v) ethanol, and resuspended in 100 μL nuclease-free H 2 O (Sigma-Aldrich). RNA concentration was detected with a NanoDrop spectrophotometer (ND-1000; Thermo Fisher) and samples were stored at − 80 • C until cDNA synthesis.
Reverse transcription of 1.0 μg of the collected RNA was carried out with the reverse transcription system kit (Promega, Fitchburg, WI, USA) using avian myeloblastosis virus (AMV) reverse transcriptase according to the manufacturer's protocol in order to synthesize complementary DNA (cDNA). Generated cDNA samples were stored at − 20 • C until gene expression analysis.

Protein extraction and BCA assay
After 7 d, the culture medium was removed and the cells were washed with PBS twice. Then, cells were lysed with 200 μL/well icecold RIPA buffer (Thermo Fisher, Waltham, MA, USA) supplemented with protease inhibitor cocktail (PIC; Roche, Basel, Switzerland). After gentle shaking on ice for 5 min, samples were collected and centrifuged at 14,000 g at 4 • C for 15 min. Protein extracts were stored at − 80 • C until Western blot analysis.
Protein concentrations were determined with the bicinchoninic acid (BCA) assay (Thermo Fisher) according to the manufacturer's protocol with the help of a microplate reader (AR2001; Anthos Microsystems, Friesoythe, Germany) at λ = 570 nm.

Protein analysis via western blotting
For Western blot analysis, equal amounts of total protein (10-35 μg) were heated to 85 • C for 3 min in Laemmli buffer [37] (0.5 mol/L dithiothreitol (DTT) instead of β-mercaptoethanol), separated by SDS-PAGE using pre-cast Tris-glycine gradient gels (8-16%; Invitrogen) and transferred to polyvinylidene difluoride (PVDF) membranes (Bio-Rad, Karlsruhe, Germany) with a transfer system (Trans-blot Turbo; Bio-Rad). Transferred total proteins were detected with the no-stain protein labeling reagent (Invitrogen) according to the manufacturer's protocol and were used as the loading control. Afterwards, membranes were blocked with 5% (w/v) non-fat milk (Merck, Darmstadt, Germany) in Tris buffer containing 0.05% (v/v) Tween-20 (TBST; Carl Roth, Karlsruhe, Germany) at room temperature for 2 h. After washing for three times with TBST for 10 min, the membranes were incubated with primary rabbit antibodies at 4 • C overnight in TBST supplemented with 5% (w/v) bovine serum albumin (BSA; Biomol, Hamburg, Germany) directed against the following human proteins: BCL2L11 (2933), CDH1 (3195), MMP9 (3852) PTEN (9559), TIMP2 (5738; all from Cell Signaling, Leiden, Netherlands), CDKN1A (ab109520) and COL1A1 (ab138492; both from Abcam, Cambridge, UK) (Fig. S3). Afterwards, membranes were washed again and incubated with goat anti-rabbit IgG secondary antibody conjugated with horseradish peroxidase (HRP; 7074; Cell Signaling) at room temperature for 4 h. Chemiluminescence was detected after washing and subsequent incubation with enhanced chemiluminescence (ECL) substrate solution (Cytiva, Marlborough, MA, USA) for 2 min in the dark and was visualized with an imaging system (iBright FL1500; Thermo Fisher). Semi-quantitative, densitometric Western blot analysis has been done with the ImageJ software [38,39]. Data are presented as expression levels normalized with total protein expression and relative to the expression in untreated reference cells under normoglycemic conditions (fold-change).

Statistical analysis
RNA arrays were only performed once in order to screen for genes involved in cancer mechanisms and tumor metastasis that were affected by metformin treatment, and thus no statistical analysis was performed with the resulting data.
For qPCR analysis, ΔC T and ΔΔC T values were calculated according to equations (1) and (2), respectively: and ΔΔC T = ΔC T, treatment -ΔC T, NG/HG control A mixed effects model analysis was performed with the ΔC T values followed by Tukey's (comparing metformin/insulin effects under identical glucose conditions) or Š ídák's (comparing identical treatments between different glucose levels) multiple comparison post-hoc tests. In order to ensure reliability of the qPCR data, samples with C T values ≥ 34.0 were excluded due to their low expression levels.
For western blotting, mixed effects model analysis and subsequent Tukey's or Š ídák's multiple comparison post-hoc test were performed with protein expression levels normalized to the total protein amount and relative to protein levels in untreated cells under

Table 1
Metformin-induced changes in gene expression as detected after metformin treatment in transcriptomic analysis. HEC-1A and Ishikawa cells were treated with 1.0 mmol/L metformin under hyperglycemic conditions and gene expression was compared to untreated control cells. For the identification of target genes, the expression of >160 genes was investigated with the TaqMan "Human Molecular Mechanisms of Cancer" and "Human Tumor Metastasis" RNA arrays. Interesting target genes were ≥ 2-fold up-or downregulated in a favorable, tumor-suppressing way due to the metformin treatment. normoglycemic conditions. Real-time PCR and Western blot data were presented as aligned dot plots including the means of at least three independent experiments. Statistical analyses were carried out with the help of Prism 9 (GraphPad Software, La Jolla, CA, USA). A value of p ≤ 0.05 was considered statistically significant; p < 0.1 were additionally mentioned in the text and p < 0.15 were displayed as values in the plots. Unless stated otherwise, p values mentioned in the text refer to the untreated normo-or hyperglycemic control sample. Single missing samples and values appeared due to random reasons (e.g. cell handling errors, low RNA quality, and pipetting or other errors during cDNA synthesis, qPCR, BCA assay, SDS PAGE or western blotting).

Screening for metformin-regulated target genes in transcriptome analysis
From >160 different genes investigated in the TaqMan "Human Molecular Mechanisms of Cancer" and "Human Tumor Metastasis" RNA arrays, 11 and 21 genes, respectively, showed a ≥ 2-fold up-or downregulation and were regulated by metformin in a favorable, tumor-suppressing way ( Fig. 1 and Table 1). Further genes were ≥ 2-fold up-or downregulated due to metformin treatment, however, regulation was rather unfavorable and tumor-promoting (Table 1); these genes were not considered for further analyses.
Although COL1A1 expression was not detected in HEC-1A cells and was upregulated in Ishikawa cells, which is rather unfavorable, the gene has been selected for the follow-up qPCR analysis, because it is directly connected to selected MMP2/9/10 as well as indirectly related to selected TIMP1/2/4. More importantly, our group observed a favorable downregulation of COL1A1 expression in an analysis of the HEC-1A proteome after metformin treatment in a recent study [34]. Additionally, TIMP3 has also been selected for further analysis due to the connection to the selected MMP2/9/10 and TIMP1/2/4 genes. Furthermore, CDKN1A has been selected for further analysis, because it was regulated in a favorable way in both cell lines, although upregulation was only 1.8-fold. An overview of selected genes/proteins and their relation to cancer and metastasis is given in Table 2.
Changes in the Expression of Selected Genes after Metformin Treatment under Different Metabolic Conditions in Real-Time PCR Analysis.
After selecting 19 target genes from the transcriptome analysis, a gene expression analysis has been carried out including further treatment conditions: metformin (0.1 and 1.0 mmol/L), insulin (100 ng/mL) and a combined administration, all under normo-(NG; 5.5 mmol/L) and hyperglycemic (HG; 17.0 mmol/L) conditions (Fig. 2). In the postmenopausal model cell line HEC-1A, COL1A1, MMP10 and TIMP3 genes could not be detected, whereas CEACAM1 could not be found in the premenopausal Ishikawa cell line. Unless stated otherwise, p values refer to the untreated normo-or hyperglycemic control sample; p values < 0.1 were mentioned in the text.
In HEC-1A cells, metformin slightly downregulated gene expression of APC (FC NG, 0.1 = 0.78; FC NG, 1.0 = 0.68) and more Table 2 Genes selected for follow-up qPCR analysis and their function and relation to cancer as well as metastasis.  Glucose-induced changes in gene expression between identically treated samples were observed, but were not significant (Fig. S1). When untreated controls under normo-and hyperglycemic conditions were compared, glucose supplementation substantially values followed by Tukey's (analysis of metformin and insulin effects under identical glucose conditions) or Š ídák's (analysis of glucose effects between identical treatments; see Figure S1) multiple comparison post-hoc tests; *p ≤ 0.05, **p ≤ 0.01 (metformin effect, black); # p ≤ 0.05 (glucose effect, blue); p < 0.15 additionally displayed as values. Glucose-induced changes in gene expression were observed, but were not significant (Fig. S1). Comparing the untreated controls under normo-and hyperglycemic conditions, glucose supplementation downregulated APC (FC HG, control = 0. After gene expression analysis, 7 genes were selected for analysis on protein level by western blotting (Fig. 3). PTEN could not be detected in the premenopausal Ishikawa cells, but this protein is known to be absent in this cell line [40,41]. Unless stated otherwise, p values refer to the untreated normoglycemic control sample; p values < 0.1 were mentioned in the text.  An overview of the metformin-induced changes on the expression of the 7 selected key molecular targets on gene and protein level is provided in Table 3  Under hyperglycemic conditions, downregulation of BCL2L11 (FC HG, 0.1+ins = 1.07; FC HG, 1.0+ins = 0.80) was enhanced at 1.0 mmol/L, while CDH1 (FC HG, 0.1+ins = 1.00; FC HG, 1.0+ins = 0.81) and COL1A1 (FC HG, 0.1+ins = 0.73; FC HG, 1.0+ins = 0.49) were now also downregulated, which was in contrast to identical treatments at normal glucose levels, but advantageous in case of COL1A1. MMP9 expression (FC HG, 0.1+ins = 0.61; FC HG, 1.0+ins = 0.71) was opposed when 0.1 mmol/L metformin was combined with insulin, leading to a more desirable effect. CDKN1A (FC HG, 0.1+ins = 1.09; FC HG, 1.0+ins = 0.85) as well as TIMP2 levels (FC HG, 0.1+ins = 1.57; FC HG, 1.0+ins = 1.73) were similar to levels observed after single compound administration.
An overview of the metformin-induced changes on the expression of the 7 selected key molecular targets on gene and protein level is provided in Table 4 for the premenopausal Ishikawa cell lines (see Graphical Abstract for a schematic overview).

Table 4
Metformin-induced changes on the expression of 7 key molecular targets on gene and protein level under normo-and hyperglycemic conditions as well as under the influence of hyperinsulinemia for the premenopausal Ishikawa cell line.

Discussion
Transcriptomic analysis of EC cells was used as a screening tool in order to identify target genes that are substantially regulated by metformin administration. The screening was performed with a single selected condition, i.e. metformin treatment at 1.0 mmol/L under hyperglycemic conditions against untreated control cells. This condition was selected in order to incorporate both potential influencing factors that might affect gene expression in our in vitro model, namely the insulin-sensitizing and glucose-regulating drug metformin on the one hand as well as glucose itself on the other hand. The selected metformin concentration led to the inhibition of proliferation, migration and clonogenicity in both cell lines under similar experimental conditions in an in vitro study recently published by our group [12]. As anti-tumor effects on EC cells were the scope of our investigation, the genes used in the screening were all related to human cancer and metastasis. The following 19 potential target genes were selected for a subsequent gene expression analysis including further treatment conditions: APC, BCL2L11, CASP8, CDH1, CDKN1A, CEACAM1, COL1A1, CTNNA1, IGF1, MMP2/9/10, PTEN, RAC1, TGFB1, TIMP1/2/3/4. All genes were ≥ 2-fold up-or downregulated in transcriptomic analysis and the regulations were considered favorable. The following 7 proteins were additionally analyzed for changes in protein expression: BCL2L11, CDH1, CDKN1A, COL1A1, PTEN, MMP9, TIMP2.
Subsequent target gene expression analysis included further treatment conditions to allow for a more detailed evaluation of the influences on metformin-induced effects and to identify optimal conditions for the drug's anti-tumor effects to occur. Metformininduced effects on the expression of 19 selected genes and 7 proteins are discussed below taking into consideration the influences of various treatment conditions. The APC protein acts as a tumor suppressor and downregulation or loss of function of the APC gene are associated with EC [42,43]. In both cell lines, favorable upregulation of APC expression was observed at 1.0 mmol/L metformin only under hyperglycemic conditions, with the effects being more prominent and independent of the presence of insulin in HEC-1A cells. On the other hand, unfavorable downregulation occurred under normal glucose levels after metformin treatment and particularly during hyperinsulinemia, indicating an increased risk for EC promotion especially in the premenopausal Ishikawa model. In patients, however, genetic aberrations leading to nonfunctional APC had no influence on recurrence and metastasis in stage I EC [43] and therefore the predictive value of altered APC expression is at least questionable.
CTNNA1 has been shown to interact with APC [44] and inhibits cell proliferation, invasion and epithelial-mesenchymal transition (EMT) [45]. While CTNNA1 expression was negatively affected by metformin and a combined treatment together with insulin under normoglycemia in Ishikawa cells, desired upregulation was induced under any tested condition under the influence of high glucose levels in HEC-1A cells. Similar to the regulation of APC, our results suggest a positive influence of glucose on the metformin-induced regulation of CTNNA1. CTNNA1, CTNNB1 and CDH1 expressions were found to be suppressed in EC as well as other gynecological cancers, and further decreased with advanced invasion, contributing to cell-to-cell junctional dysfunction [46,47]. Loss of CDH1 (E-cadherin) and CTNNB1 expressions resulted in increased cell motility and advanced cancer stages [48,49] and are associated with EMT, a key event during EC development [50]. Inhibition of EMT with up to 5.0 mmol/L metformin for 24-48 h was shown to act via CDH1 upregulation in Ishikawa as well as HEC-50 cells in vitro and was also detected in tumors of diabetic EC patients [51,52]. In our study, CDH1 transcript expression was not changed in Ishikawa cells during hyperglycemia, but metformin-mediated upregulation occurred in the postmenopausal HEC-1A cell line at 1.0 mmol/L. Unfavorable CDH1 downregulation was detected in both cell lines under normoglycemic conditions, once again indicating a positive impact of increased glucose levels on metformin-induced expression changes of cancer-related genes. However, deviating results were obtained for CDH1 expression on protein level. In Ishikawa cells, CDH1 levels were not substantially changed under any tested condition, whereas in HEC-1A cells, upregulating metformin and insulin effects occurred exclusively under the influence of normal glucose levels. Conclusively, no positive impact of a high glucose environment on metformin-induced CDH1-mediated metastasis inhibition was observed. However, it has to be noted that CDH1 expression was higher in the untreated hyperglycemic control compared to its normoglycemic counterpart, suggesting a positive effect of glucose per se.
TGF-β1 represses CDH1 as well as CTNNA1 expression [53] and is known as the primary inducer of EMT, but also plays an essential role in reproductive physiology [54,55]. TGF-β1 stimulated migration, invasion and growth in endometrial HEC-1A, HEC-1B, KLE and Ishikawa cells in vitro [56][57][58]. In our study, TGFB1 transcripts were downregulated by the biguanide and also by insulin or a combined treatment in both cell lines in a normoglycemic environment. Our results suggest the involvement of the TGF-β pathway in the anti-metastatic mechanism of metformin, which has been observed before [59,60].
The tumor suppressor PTEN is connected to WNT/CDKN1B signaling via regulation of the PI3K/AKT pathway and is also involved in morphogenesis and growth arrest by interacting with CDH1 [61,62]. PTEN is mutated or lost in a large proportion of tumor entities and PTEN loss is also frequent in EC [41,63]. PTEN upregulation suppresses proliferation and inhibits invasion [64,65] and was shown to be induced by treatment of the Ishikawa cell line with 0.06 mmol/L metformin after 24 h in a study by Pabona and colleagues [66]. In the present study, PTEN levels were downregulated by metformin and insulin supplementation in Ishikawa and even more substantially in HEC-1A cells under normoglycemia. In contrast, unfavorable PTEN downregulation was prevented in both cell lines in a hyperglycemic environment. On protein level, PTEN could not be detected in the Ishikawa cell line, which is in accordance with known characteristics of the cell line [40,41]. In HEC-1A cells, however, PTEN expression was upregulated by metformin especially when combined with insulin in a normoglycemic environment, indicating an anti-proliferative effect, which was not predictable on transcript level. A hyperglycemic environment generally upregulated PTEN expression, but the effects of metformin were opposed, suggesting a metformin-induced negative regulation of PTEN with high glucose levels. Ambiguous results were obtained in other studies, where PTEN downregulation was observed after treatment with 10.0 mmol/L for 24 h in 3T3-L1 preadipocyte cells [67] and with 1.0-10.0 mmol/L for 24-72 h in Huh7.5 hepatocytes [68], whereas upregulation occurred with 27.0 mmol/L after 24 h in hepatocellular carcinoma cell lines MHCC97H and Hep3B [69].
PTEN expression not only affects cellular proliferation and growth, but the tumor suppressor also inhibits migration via suppressing effects on various MMPs, including MMP2 and MMP9 [69][70][71]. MMP2 and MMP9 downregulation and protein inhibition are desired effects, as their expression levels are closely related to invasion and metastasis [72,73], which also applies for endometrial tumors [74][75][76]. In a study with the breast cancer cell line MCF-7, metformin displayed inhibiting effects on MMP9 expression at 5.0 and 10.0 mmol/L after 24 h, whereas MMP2 levels were not affected [77]. In the present study, MMP2 expression was slightly downregulated in Ishikawa cells irrespective of the applied treatment condition under normoglycemic conditions, while metformin upregulated MMP2 levels in HEC-1A cells. However, MMP9 expression was downregulated after metformin administration in Ishikawa and particularly in HEC-1A cells in a normoglycemic environment and in the latter also in the presence of elevated glucose concentrations. Favorable metformin-induced downregulation of MMP9 was confirmed on protein level for both cell lines, suggesting MMP9 as an actively involved protein in the anti-metastatic mechanism of the drug. A negative influence of glucose on the metformin-mediated effects on MMP9 expression was only observed in the Ishikawa premenopausal model. For MMP10, another metastasis-promoting matrix metalloproteinase [78,79], downregulation was achieved in Ishikawa cells with 1.0 mmol/L metformin during hyperglycemia, while 0.1 mmol/L metformin was more favorable under normoglycemic conditions, particularly in combination with insulin; the MMP10 transcript could not be detected in HEC-1A cells.
Although MMP2 and MMP9 belong to the group of gelatin binding MMPs, collagens serve as substrates for both proteinases as well. In a recent study by our group, HEC-1A cells were treated with 0.5 mmol/L metformin for 7 d and cellular lysates were analyzed for proteomic changes. Among other proteins, COL1A1 was significantly downregulated by metformin [34]. COL1A1, together with COL1A2, forms type I collagen, which is the major component of the extracellular matrix (ECM) in connective tissues and COL1A1 displayed tumor-promoting effects in various cancer cells [80][81][82]. However, in EC patients, the survival outcome was better with high expression of COL1A1 transcripts, whereas COL1A1 protein expression appeared to be increased in EC tissues versus normal endometrium [83]. In the present study, COL1A1 was not detectable in HEC-1A cells and was negatively regulated in Ishikawa cells during normoglycemia by metformin and particularly insulin. Surprisingly, downregulation was reduced when both substances were combined. On protein level, favorable COL1A1 downregulation was observed with metformin and insulin alone and again particularly after combination in a normoglycemic milieu in Ishikawa cells. During hyperglycemia, a combined treatment also reduced COL1A1 expression, while single treatments did not induce any desired effects. However, elevated glucose concentrations per se led to a favorable reduction of COL1A1 levels in untreated cells. In HEC-1A cells, COL1A1 downregulation exclusively occurred after metformin treatment under hyperinsulinemic conditions during normoglycemia, confirming the positive effect of insulin on metformin-mediated COL1A1 regulation.
TIMPs are closely related to MMPs as they act as specific inhibitors on the metastasis-associated proteinases, and therefore, upregulation of TIMP2, TIMP3 and TIMP4 is considered advantageous, while downregulation seems to be favorable for TIMP1 [84]. However, loss of TIMP1 and associated loss of PTEN promoted invasion and migration in prostate cancer in vivo [85], while TIMP1 overexpression was associated with cancer progression and poor prognosis in numerous clinical studies [86,87]. TIMP1 levels were also found to be increased in serum and in flushings from women with EC [88,89]. Therefore, in the context of the present study, a TIMP1 downregulation seems to be more favorable, which was only detected after metformin treatment in Ishikawa cells at normal glucose concentrations and was not affected by additional insulin supplementation. TIMP2, TIMP3 and TIMP4 all inhibit metastasis-related proteinases MMP2 and MMP9, of which MMP9 has already been identified as an interesting target for metformin in the present study. For TIMP2, increased as well as decreased levels were detected in clinical studies [86], but in relevant EC patients, overall survival correlated with enhanced TIMP2 levels [90]. In a study by Dai et al. with HEC-1A cells, inhibition of TIMP2 at mRNA and protein level enhanced MMP2 expression, indicating an unfavorable metastasis-promoting effect [91]. In the present study, TIMP2 levels were upregulated by metformin in HEC-1A and to a lesser extent in Ishikawa cells only in a hyperglycemic environment. During normoglycemia, TIMP2 expression was not affected in the HEC-1A cell line and decreased in the premenopausal Ishikawa cell line, indicating a positive influence of glucose on the metformin-induced TIMP2 regulation. On protein level, TIMP2 was upregulated by 1.0 mmol/L metformin in both cell lines in the presence of elevated glucose concentrations, while metformin induced an unfavorable TIMP2 downregulation during normoglycemia, confirming the results from gene expression analysis and, again, suggesting a positive impact of hyperglycemia. Downregulation of TIMP3 is associated with cancer progression and poor prognosis and TIMP3 silencing has been found in multiple human cancers [86]. In the context of EC, TIMP3 downregulation was found in high stage endometrioid EC [92] and stimulated growth and invasion in HEC-1B and Ishikawa EC cells in vitro [93]. Metformin also led to an increased expression of TIMP1 and TIMP3 in osteoarthritis chondrocytes co-cultured with metformin-treated adipose tissue-derived human mesenchymal stem cells [94]. In the present study with EC cell lines, low-concentration metformin as well as insulin alone induced TIMP3 downregulation in Ishikawa cells, while 1.0 mmol/L metformin slightly enhanced TIMP3 expression both under normo-and hyperglycemic conditions; the TIMP3 gene could not be detected in HEC-1A cells. Like TIMP2, TIMP4 was found to be up-as well as downregulated in tumors depending on the type of cancer [86]. In EC, TIMP4 transcript levels were lower [95], whereas high TIMP4 protein expression was detected in endometrial tumor tissue in correlation with myometrial invasion, suggesting a key role in endometrial tumor progression [96]. In the present study, 1.0 mmol/L metformin induced TIMP4 upregulation in HEC-1A and Ishikawa cells in a hyperglycemic environment, suggesting a positive influence of glucose, as seen for TIMP2 before. Supplementation with insulin further increased TIMP4 expression only in HEC-1A cells irrespective of changes in glucose concentration.
As a key regulator of the actin cytoskeleton, RAC1 is involved in actin reorganization, which is required for proliferation and migration. RAC1 downregulation was shown to reduce metastasis [97][98][99], while RAC1 activation promoted tumor growth [100]. Metformin reduced RAC1 protein expression and cell migration in prostate cancer cell lines in a study by Dirat et al. [101] as well as in a keratinocyte cell line in a study by Hakimee and colleagues [102]. In the present study, RAC1 transcript expression has been lowered by 1.0 mmol/L metformin in a hyperglycemic environment in HEC-1A cells, while the same effect was observed in Ishikawa cells only at normal glucose levels, indicating a varying impact of glucose in pre-and postmenopausal settings.
CDKN1A acts as a tumor suppressor via G 1 cell cycle arrest, leading to growth arrest, senescence or apoptosis, but CDKN1A inhibits apoptosis and promotes cell proliferation in p53-deficient tumors [103][104][105]; the HEC-1A as well as the Ishikawa cell lines were found to be TP53-as well as p53-positive [106][107][108][109]. Additionally, 17-45% of EC tissue samples were p53-positive in various studies [110]. In the present study, metformin upregulated CDKN1A transcripts in HEC-1A and Ishikawa cells in the presence and absence of increased insulin levels only under hyperglycemic conditions, while normoglycemia had an unfavorable impact. On protein level, CDKN1A expression was only minimally affected under any test condition in the premenopausal Ishikawa model. In HEC-1A cells, metformin-induced favorable CDKN1A upregulation occurred only during normoglycemia, but not hyperglycemia, which was in contrast to CDKN1A transcript regulation. Favorable metformin effects in a normoglycemic environment are particularly interesting for clinical settings, as metformin will normalize glucose levels in patients with hyperglycemia due to its known indirect effects that are induced via blockage of gluconeogenesis. However, high glucose levels per se led to an increase in CDKN1A expression in untreated control samples.
CEACAM1 plays a role in adhesion and in pathways related to survival, differentiation as well as growth and it has been suggested to act as a tumor suppressor [111]. CEACAM1 loss is associated with poor prognosis in gastric cancer patients [112], while overexpression suppressed proliferation, induced cell apoptosis and inhibited migration in multiple myeloma cell lines in vitro [111]. On the other hand, CEACAM1 was found to be pro-angiogenic in vivo [113] and to stimulate cellular metastasis in various cancer types [114]. In EC and other malignancies, CEACAM1 expression was downregulated [115,116]. In the present study, favorable CEACAM1 upregulation was observed after metformin treatment in HEC-1A cells under hyperglycemic conditions, which was even increased when the biguanide was given together with insulin. It has to be noted that CEACAM1 levels were substantially lower under any tested condition in a hyperglycemic environment when compared with the respective normoglycemic counterpart, indicating a remarkable negative impact of glucose on CEACAM1 expression. In the Ishikawa premenopausal model, CEACAM1 expression could not be detected.
IGF1 triggers proliferation and enhances survival upon binding to the IGF1 receptor (IGF1R) [117]. The IGF system plays an important role in estrogen-induced EC [118,119] and high IGF1 levels stimulate proliferation, migration as well as invasion, and thus promote tumor growth, angiogenesis and metastasis [120]. In HEC-1A cells, IGF1 expression was reduced by metformin with and without additional insulin in a hyperglycemic milieu in our study, while decreased IGF1 levels were observed under any tested condition in normoglycemic media in Ishikawa cells, suggesting a varying influence of glucose on metformin-induced effects in a pre-versus a postmenopausal model.
Upon activation during early stage of the extrinsic apoptosis pathway, the initiator caspase CASP8 activates effector caspases 3 and 7. CASP8 is also associated with the intrinsic apoptosis pathway via activation of BH3-interacting domain death agonist (BID) [121]. In tumor cells, CASP8 induces apoptosis and inhibits proliferation [122] and metformin was found to increase CASP8 expression in A498 renal carcinoma cells at 7.5 mmol/L after 24 h [123] as well as in several pancreatic cancer cell lines at 30.0 mmol/L in vitro [124]. In the present study, CASP8 expression was decreased after metformin treatment and in particular after insulin supplementation in normoglycemic media, while no CASP8 regulation was detected in a hyperglycemic milieu, indicating a negative metformin effect on apoptosis induction at normal glucose levels.
BCL2L11 is involved in the intrinsic apoptosis pathway as a pro-apoptotic regulator and overexpression inhibited tumor growth and increased apoptosis induction [125]. In several studies, metformin increased BCL2L11 expression and, for instance, inhibited growth of EC cell line RL95-2 after treatment with 4.0 mmol/L for 48 h [126], induced apoptosis at 1.0 mmol/L after 48 h in H1975 and PC-9 lung cancer cell lines [127] or inhibited proliferation of esophageal cancer cell lines Eca109 and EC9706 at 10.0-20.0 mmol/L after 24 h [128]. In the present study, BCL2L11 has been upregulated after metformin treatment with and without insulin in a hyperglycemic milieu in HEC-1A, but not in Ishikawa cells. The BCL2L11 protein was upregulated by metformin and insulin alone in HEC-1A cells irrespective of the glucose concentration, but a combined treatment induced the desired effect only during normoglycemia. In Ishikawa cells, only treatment with metformin or insulin alone led to favorable BCL2L11 upregulation.
The influence of glucose levels on metformin-induced anti-tumor effects seems to be crucial [129], which has also been shown for some of the abovementioned genes and proteins in the present study. Therefore, our results confirmed the impact of varying concentrations of glucose on the effectiveness of metformin's anti-cancer activity. In clinical settings, metformin-treated diabetic breast cancer patients had a better clinical outcome compared to non-treated patients [130] and the drug significantly improved overall and progression-free survival of patients with T2DM [131]. In the context of EC, there was a lower risk for EC development in women with T2DM when metformin was applied [132]. However, metformin did not lower the risk for tumor development in patients with T2DM in another study, although the drug decreased the risk for diabetic patients to develop other gynecological cancers, particularly of the cervix [133]. Data from other clinical studies have also been inconsistent, likely due to inhomogeneities between patient groups with regard to age, body mass index (BMI), cancer subtypes and cancer stages, as well as preexisting metabolic diseases. Such deviating results were found for overall and progression-free survival, tumor recurrence and lower risk for cancer development in diabetic EC patients after metformin administration [134][135][136].
In summary, metformin induced favorable regulation of cancer-and metastasis-related genes and proteins in the present study and was able to reduce or cancel negative insulin-induced effects, particularly in a normoglycemic environment, which could be relevant for prediabetic patients with insulin resistance. However, the current research was designed as an in vitro study and therefore the predictability of in vivo effects is limited. Metformin concentrations applied in vitro (0.1 and 1.0 mmol/L) may deviate greatly from clinical doses as well as plasma or tissue levels (typical plasma levels during T2DM therapy are between 1 and 20 μmol/L [137,138]) reached in patients and may even exceed lethal doses, if applied in vivo. A possible explanation could be high concentrations of C. Lange et al. nutrients and growth factors in the culture media, combined with a fluctuating expression of OCT1 (organic cation transporter 1), which is responsible for the cellular uptake of metformin [139]. Also, factors such as cellular uptake, interactions between cell types, tumor microenvironment, or drug stability are different in an in vivo setting, but these are general drawbacks of in vitro models. Yet, results of the current screening approach were intended to identify potential key targets that contribute to the anti-cancer activity of metformin in pre-and postmenopausal EC under consideration of different environmental influences. Nevertheless, further research will be necessary in order to confirm the results and their potential clinical relevance in vitro studies with concentrations in a clinically applicable low micromolar range and in vivo studies.
Elevated glucose levels tend to have a positive impact on metformin-mediated effects or prevented negative effects that have been observed under normoglycemia for single molecular targets. However, these effects did not transfer to an overall cellular level, where effects on proliferation, viability, clonogenicity or migration were similar or less favorable under high glucose conditions in a recent study of our group with HEC-1A and Ishikawa EC cell lines [12]. An unfavorable impact of high glucose levels on metformin-induced effects on proliferation, cell cycle arrest or apoptosis induction has also been observed in other in vitro studies, e.g. in breast cancer cells [140][141][142]. Special attention should be given to candidates that were positively regulated by metformin in a normoglycemic environment, as metformin will normalize glucose levels in hyperglycemic and diabetic patients due to its indirect effects induced via blockage of gluconeogenesis.
Metformin-induced changes in gene regulation did not necessarily reflect altered protein expression and are therefore not suitable as general predictive markers. Furthermore, metformin regulated several genes and proteins differentially in pre-and postmenopausal EC models. It also has to be noted that no validation of the functional importance of the identified target genes has been carried out in the present in vitro study; knock-down or knock-out of the selected candidate genes would help evaluate the effect of metformininduced expression changes on cellular functions, e.g. proliferation, migration or viability.

Conclusions
In the present study, genes and proteins were identified that might be involved in the anti-cancer mechanism of metformin in EC. However, further research will be needed in order to perform pathway analysis, including upstream regulators and downstream targets of the selected candidates. The following conclusions were drawn based on the presented data: Firstly, metformin-induced regulations were glucose-dependent and were interestingly more favorable in a hyperglycemic environment for some molecular targets. Secondly, favorable gene regulations could not be easily extrapolated to protein level and could not reliably predict protein regulation as nothing is known about parameters such as promoter strength, translation efficiency or transcript half-life. Thirdly, metformin reduced or canceled unfavorable regulations that have been induced by elevated insulin levels in a hyperinsulinemic setting in some cases. And finally, in the context of EC, metformin-induced effects varied between pre-and postmenopausal cell lines, indicating a variable sensitivity to the drug due to hormone-induced differences and suggesting deviating outcomes with varying tumor types and tumor stages. With the presented data, we contribute to a better understanding of the anti-cancer activity of metformin as well as its underlying mechanism of action in EC cells. Although further research will be necessary to confirm the data, the influence of different environmental settings on metformin-induced effects could be highlighted with the presented in vitro results.

Author contribution statement
Carsten Lange: Conceived and designed the experiments; Performed the experiments; Analyzed and interpreted the data; Wrote the paper. Jana Brüggemann; Theresa Thüner: Performed the experiments; Analyzed and interpreted the data; Wrote the paper. Julia Jauckus: Performed the experiments; Analyzed and interpreted the data. Thomas Strowitzki: Contributed reagents, materials, analysis tools or data. Ariane Germeyer; Conceived and designed the experiments; Analyzed and interpreted the data; Contributed reagents, materials, analysis tools or data; Wrote the paper.

Data availability statement
Data included in article/supp. material/referenced in article.

Appendix A. Supplementary data
Supplementary data to this article can be found online at https://doi.org/10.1016/j.heliyon.2023.e16678.