Epithelial ovarian cancer (EOC), the most common subtype of ovarian cancer, is one of the most lethal gynaecological neoplasms in women worldwide, and is characterized by non-specific symptoms and late diagnosis, which result in poor survival rates [1
]. Unfortunately, the current therapies yield modest results. Therefore, it is important to understand the molecular mechanisms that characterize this disease and to develop new strategies to improve patient therapy and survival. EOC is characterized by rapid growth and pronounced angiogenesis, which is attributed to overexpression of many growth factors that favour tumorigenesis, including nerve growth factor (NGF) [4
] and vascular endothelial growth factor (VEGF) [5
Some studies have shown that presence and expression of NGF and its high affinity receptor TRKA increase during EOC progression, and promote tumour cell proliferation and angiogenesis by activation of the phosphoinositide 3-kinases (PI3K)/ Protein kinase B (AKT) and mitogen-activated protein kinases (MAPK)/extracellular signal-regulated kinase (ERK) signalling pathways [4
]. Moreover, results from our group showed that NGF/TRKA is highly expressed in EOC cells compared to non-tumoral ovarian cells, which leads to increased NGF/TRKA signalling [4
]. In addition, NGF stimulation increases the expression of VEGF and the transcription factor c-MYC in EOC cells and explants from EOC biopsies [8
]. NGF/TRKA signalling (via PI3K/AKT and MAPK/ERK pathways) increases the expression of oncoproteins such as survivin and β-catenin, in cancer cells [11
], and these two proteins are also upregulated in EOC [16
Survivin is a protein that inhibits apoptosis, regulates cell division and is involved in tissue healing after injury [18
]. Survivin belongs to the inhibitor of apoptosis (IAP) anti-apoptotic family of proteins and its aberrant expression is associated with increased tumour proliferation, progression, angiogenesis, resistance to therapy, and poor prognosis in several cancers [20
]. In addition, survivin expression is a marker of poor prognosis in ovarian cancer patients [24
] and its expression promotes VEGF-induced tumour angiogenesis via PI3K/AKT [25
]. On the other hand, β-catenin is a transcriptional co-regulator and an adaptor protein for intracellular adhesion [26
]. Binding of β-catenin to the T-cell factor/lymphoid enhancer-binding factor (TCF/Lef) family of transcription factors promotes transcription of many oncogenes following activation of the canonical Wnt signalling pathway [27
]. For instance, the β-catenin/TCF-Lef complex increases the expression of proteins associated with epithelial-mesenchymal transition (EMT) [30
], as well as those relating to cell survival and cell proliferation, including survivin and c-MYC [32
]. Both proteins in turn promote β-catenin/TCF-Lef activity, in an amplification loop that could be involved in tumour growth and EOC progression.
Because NGF/TRKA signalling regulates the expression of several oncoproteins, we propose that this may involve control via microRNAs (miRs), the largest family of non-coding RNAs [34
] that target specific messenger RNAs (mRNAs) to either induce their degradation or block protein translation [34
]. The miR-145 and miR-23b are two oncosuppressor miRs that are known to be downregulated in ovarian cancer tissues [35
] and in-silico analysis revealed that c-MYC and VEGF could be targets for these miRs (Table 1
), which in turn are modulated by NGF/TRKA [35
Among the different drugs that can block the effects of growth factors in cancer cells, metformin has emerged as an interesting candidate. Metformin is a biguanide, which is widely used to treat metabolic disorders, such as type II Diabetes Mellitus, gestational Diabetes, and Metabolic Syndrome [36
]. Interestingly, metformin has been attributed anti-tumour effects because it decreases the mortality due to cancer in diabetic patients [39
]. Furthermore, observational studies in EOC have shown that metformin intake is associated with a decrease in ovarian cancer incidence and mortality [40
]. This is particularly intriguing given that metformin is cheap, readily accessible and safe. Mechanisms proposed to explain metformin’s pleiotropic effects include the down-regulation of several oncogenic proteins in different cancer cells, both by epigenetic [41
], as well as post-transcriptional mechanisms that include miR-mediated regulation [43
We previously reported that the anti-diabetic drug metformin reduced NGF-induced proliferation of EOC cells, as well as the angiogenic potential of endothelial cells [46
]. So, the purpose of this study was to understand better the mechanisms by which metformin blocks NGF-dependent effects in EOC cells and thereby contributes to clarifying the anti-tumour mechanisms of this drug.
Our results show that metformin, strongly decreases c-MYC, β-catenin and VEGF expression in EOC cells with little effect on non-tumour ovarian cells. Importantly, metformin blocks the NGF-induced increase in c-MYC, survivin and VEGF, as well as the increase in MYC and β-catenin/TCF-Lef transcriptional activity in ovarian cancer cells. Additionally, metformin treatment of EOC cells increased the levels of miR-145 and miR-23b by blocking the NGF-induced decrease in these miRs, suggesting that the anti-tumour effects of metformin could be mediated by miR regulation. Consistent with the in-vitro results, the comparison within a small group of patients with borderline ovarian tumours of users vs. non-users of metformin revealed that metformin intake correlated with decreased immunodetection of survivin, c-MYC and β-catenin in tissue sections. Thus, the results obtained by analysis of human tissue samples confirmed our observations in EOC cell lines.
In summary, these results contribute to a better understanding of the tumour suppressor effects of metformin, by showing that the drug decreases the expression of several important oncoproteins, including c-MYC, VEGF and β-catenin, and likely does so via a pathway involving miR modulation. Thus, metformin holds considerable promise as a possible complementary therapy in EOC treatment.
In the present study, we sought to shed light on the tumour suppressor mechanisms triggered by metformin in EOC cells. Previous in-vitro studies by our group showed that metformin blocks the pro-angiogenic and proliferative effects of NGF/TRKA [46
]. In the current study we provide evidence that this effect can be attributed to a decrease in the expression of several important oncoproteins, including c-MYC, VEGF and survivin. In addition, a decrease in the transcriptional activity of c-MYC and β-catenin/TCF-Lef was observed in EOC cells. These changes coincide temporally with increased levels of miR-23b and miR-145 in ovarian cells, likely reflecting the reduced ability of NGF to decrease these oncosuppressor miRs in the presence of metformin. Our in-silico studies show that the proteins VEGF, c-MYC and survivin may represent miR-23b and miR-145 targets. Thus, the observed up-regulation of these miRs in the presence of metformin likely explains the changes in these proteins due to metformin treatment, as well as the pleiotropic effects of metformin described by other authors, given that one miR may potentially regulate the expression of hundreds of cell proteins [54
EOC is the leading cause of death due to gynaecological neoplasia in developed countries and therapeutic success has not improved substantially in the last decades [2
]. A better understanding of the molecular mechanisms underlying EOC is necessary in order to identify new therapeutic targets and develop new therapeutic strategies. The current results identify the use of metformin as a potentially attractive adjuvant therapy in the context of EOC.
A limitation of this study is that the metformin concentrations used here (10 mM, 48 h) were considerably higher than the plasma concentrations of this drug generally described in the literature [59
]. However, bearing in mind the pharmacodynamics, as well as studies suggesting that metformin may accumulate in rodent tissues [60
] it is possible that the elevated concentrations employed here are relevant. The concentration of metformin reached in tissues, like the ovary, is controversial; however, a recent study identified micromolar metformin concentrations in mouse ovarian cancer tumours [62
]. Another limitation to our in-vitro experiments is that only relatively short time points were evaluated. In patients, metformin can accumulate in ovarian tissues, because patients consume this drug for extended periods of time. It is also for this reason that we decided to test the higher dose in our short term in vitro experiments.
In future experiments, it will be important to corroborate these in-vitro results in mouse models. There are several studies in which the anti-tumor effects of metformin have been evaluated in-vivo but, unfortunately, working with NGF is more difficult, because the half-life in circulation is only 2.3 h following intravenous injection [63
]. For this reason, the use of mini-osmotic pumps is recommended to guarantee stable levels of circulating NGF for several days [63
]. We plan to set up such experiments in the future.
Given these limitations, we also compared β-catenin, c-MYC and survivin levels in samples from a small group of patients with serous and mucinous borderline ovarian tumours that were either users or not of metformin. The results are consistent with our in-vitro experiments. Specifically, in the biopsies from patients who had taken metformin levels of β-catenin, c-MYC and survivin were lower compared to those found in biopsies from patients that were not under treatment with this drug (see Figure 6
). These findings support the notion that our in-vitro experiments represent an adequate proxy to define how mechanistically metformin acts in ovarian cancer patients. It is important to note that the proteins analysed by immunohistochemistry (IHC) were detected in the early stages of EOC development (borderline ovarian tumours). This because we only had access to a reduced number of ovarian cancer patients that were users of metformin. In any case, the results shown here are promising and indicate that metformin intake may serve to prevent/limit the progression of ovarian cancer, as suggested previously by others [39
]. However, to consolidate these promising preliminary results, further patient-based studies involving many more patients will be necessary.
It is interesting to note that we worked here with three different cell lines, which represent different stages in epithelial ovarian cancer progression. HOSE cells are a well-known model of immortalized, but non-tumoral ovarian surface epithelial cells [7
], while the A2780 cell line was obtained from primary ovarian adenocarcinoma of a patient without treatment [49
], and finally the SKOV3 cell line was isolated from the ascites of a woman with EOC [50
]. SKOV3 cells are resistant to several cytotoxic drugs and are highly migratory and invasive [51
], similar to metastatic cells. Not surprisingly, some notable differences in the pattern of responses to metformin were apparent. For instance, in HOSE cells the oncoproteins studied were essentially not affected by metformin treatment, while A2780 cells were the most sensitive to metformin treatment. Unfortunately, SKOV3 cells were less sensitive to metformin than A2780 cells. Moreover, metformin increased the oncosuppressor miR-23b and miR-145 by five times or more in A2780 cells, while in SKOV3 cells, levels of miR-145 increased only 1.7 times. These findings indicate that SKOV3 cells may be more resistant to metformin treatment and that the therapeutic benefit of metformin might be temporarily limited. However, other reports showed that metformin reduced the migration and invasiveness of SKOV3 cells [68
] and we were able to replicate these findings using our experimental conditions in vitro (Supplementary Figure S9
). These observations may be taken to suggest that metformin has more pronounced effects on the behaviour of EOC cells with elevated metastatic potential.
Moreover, we provide evidence that NGF/TRKA stimulation decreases the levels of the oncosuppressors miR-145 and miR-23b, but the magnitude of the NGF effect varies between the EOC cell lines A2780 and SKOV3. These differences may relate to the baseline levels of the miRs. For example, miR-145 is considered a suppressor of cell migration and invasion [70
] and the levels of miR-145 are lower in SKOV3 cells than A2780 cells (Supplementary Figure S9
). This may explain why SKOV3 cells are more migratory and display an elevated metastatic potential compared to A2780 cells [51
]. In addition, the low basal levels of miR-145 likely explain why NGF stimulation leads to only a moderate decrease in the levels of this miR in SKOV3 cells. Alternatively, A2780 cells express higher basal levels of miR-145 and in these cells NGF induced a strong decrease in miR-145 content. Additionally, the differences in the responses of EOC cells to NGF stimulation may also reflect differential expression of neurotrophins and their receptors. For instance, Pro-NGF also promotes signalling by binding directly to TRKA or P75 (low affinity receptor of NGF)/sortilin in breast cancer and melanoma cells [72
]. As shown in Supplementary Figure S10
, NGF levels are lower in SKOV3 cells than A2780 cells. On the other hand, Pro-NGF levels are higher in SKOV3 cells. These results are in agreement with a previous study [74
], which described that SKOV3 cells express less NGF and TRKA receptor, but elevated protein levels of the P75 receptor compared with A2780 cells. These findings could explain why SKOV3 cells required higher concentrations of NGF (150 ng/mL vs. 100 ng/mL used in A2780 cells) to increase proliferation/migration, given that SKOV3 cells express fewer TRKA receptors. Furthermore, the higher levels of Pro-NGF and the P75 receptor in SKOV3 cells may explain why SKOV3 cells migrated more and displayed an elevated metastatic potential compared to A2780 cells.
Another possible limitation of this study is that the ovarian cell lines used are not representative of high grade serous (HGS) ovarian carcinoma [75
], which is described as the most malignant form of the disease. However, a recent report shows that non-HGS EOC cell lines (as A2780 and SKOV3) migrate and invade to a greater extent than those derived from HGS carcinomas [51
], suggesting that these non-HGS cells may in fact have a higher metastatic potential than cells derived from HGS carcinomas. Importantly, our results show that metformin reduces significantly the migration and invasion of these aggressive cell lines. In addition, metformin has been shown to improve the chemosensitivity of ovarian cancer cells with a resistant phenotype [76
]. In this case, miR modulation may provide a possible explanation. For instance, miR-145 is known to regulate several transporters of the ABC1 family and SLC1A2 [77
], which are important regulators of drug entry in cancer cells. Thus, the results presented here open up many interesting possibilities concerning the mode of metformin action that merit more detailed analysis in the future.
Another interesting point is that metformin treatment increased miR-145 and miR-23b in all the ovarian cells used, including HOSE cells. That may be because HOSE are not entirely normal, but rather immortalized epithelial ovarian cells [66
]. HOSE cells are significantly different from EOC cells and they are considered as non-tumoral model or used as a model of early stages of tumoral development [7
]. That said, it is highly likely that they have different sets of miRs compared to primary epithelial ovarian cells.
AKT and ERK signalling pathways activated by NGF/TRKA (see Supplementary Figure S5
) favour cancer cell growth [7
]. Specifically, β-catenin translocation and β-catenin/TCF-Lef transcriptional activity are associated with AKT activation and positive feedback loops have been shown to connect these components; for instance, survivin and β-catenin/TCF-Lef promote AKT signalling in cancer cells [25
]. Alternatively, ERK activation is associated with an increase in c-MYC protein levels and its transcriptional activity, as well as an increase in VEGF levels in several cancer models [11
]. c-MYC appears to be a key target of metformin, which could explain several of the anti-tumor effects observed. c-MYC can increase the transcription of the β-catenin and VEGF genes, and, at the same time, β-catenin/TCF-Lef activity depends on c-MYC transcription in colorectal cells [79
], suggesting that all these proteins collaborate in signalling loops to favour the development and progression of cancer cells.
Additionally, decreased expression of oncosuppressor miRs has been reported as the consequence of AKT/ERK signalling. For example, in bladder cancer, ERK-mediated activation by the EGFR decreases miR-23b [80
]. In addition, in breast cancer cells, the activation of AKT down-regulates miR-145 levels [81
]. These findings coincide with our current results in EOC cells. Interestingly, previous results from our group showed that miR-23b and miR-145 are downregulated in ovarian cancer samples [35
]. Here we observed that NGF stimulation decreased the levels of both miRs in ovarian cells, while metformin treatment had the opposite effect. Recent studies have shown that metformin increases the expression of the endoribonuclease Dicer [82
] in hepatocellular carcinoma cells, which likely increases the production of some oncosuppressor miRs that are downregulated in cancer cells. Our present results indicate that such mechanisms may also be in place in EOC cells. However, more work in future studies is required to corroborate this possibility.
An interesting point is that metformin blocks the tumour-promoting effects of NGF, but these apparently do not depend on decreased AKT signalling. With respect to this point, it should be noted that AMPK-activation by metformin triggers the unfolded protein response (UPR)-mediated cell death with a compensatory activation of AKT in lymphoblastic leukaemia lymphoblasts [83
]. We obtained comparable results in EOC cells, where metformin treatment induces AKT signalling, as shown in Supplementary Figure S11
. Metformin treatment did not decrease AKT phosphorylation in the amino acid residue threonine 308, a key phosphorylation site implicated in AKT activation by PI3K [84
]. In addition, our results showed that metformin increased the phosphorylation in serine 473, a substrate for mammalian target of rapamycin 2 (mTORC2) [85
]. Metformin is a known inhibitor of mTORC1 and selective mTORC1 inhibition can lead to mTORC2 and AKT activation as a compensatory response [85
]. This mechanism may explain how EOC cells become resistant to metformin treatment. Another important point is that metformin treatment increases AKT ubiquitination and degradation, which coincides with the reduced AKT levels observed in our experimental conditions (Supplementary Figure S11A–F
). However, we did observe that metformin treatment blocked the increase in ERK activation by NGF (Supplementary Figure S11G
). So, anti-proliferative effects and miR upregulation by metformin may be attributable to the decrease in ERK activation observed in EOC cells.
Another key point is that the combination of metformin + NGF might have adverse effects, possibly as the consequence of excessive reactive oxygen species (ROS) accumulation in EOC cells, which is known to promote cell cellular senescence and cell death [86
]. The use of metformin in millimolar concentrations is known to produce a significant increase in ROS in human breast cancer cells [87
], hepatocellular carcinoma [88
] and colorectal cancer cells [89
]. In addition, NGF reportedly also increases ROS production in neuronal cells [90
]. So, such excessive ROS accumulation in the groups M + N may have reduced responses below what was to be expected in some of the assays we report on here.
Interestingly, survivin levels were not affected by metformin treatment in EOC cells. However, c-MYC and β-catenin (that control survivin transcription) levels decreased following metformin treatment. This may again point towards the presence of adaptive mechanisms in EOC cells that favour resistance to metformin treatment. For instance, survivin mRNA may possess aberrant alternative polyadenylation sites that prevent inhibition by miRs and thus increase survivin expression [91
With respect to the anti-angiogenic proprieties of metformin, a key finding here is that metformin decreased VEGF levels. This point is relevant because one characteristic of EOC is the high degree of angiogenesis. The enzyme-linked immunosorbent assay (ELISA) kit used in this study recognises the VEGF isoforms 121 and 165 (secreted VEGF) and our results showed that metformin treatment decreased all three VEGF transcripts in ovarian cells (VEGF 121, 165 y 189). This finding is consistent with the functional experiments performed using endothelial cells in a matrigel assay. There, conditioned medium from EOC cells treated with metformin decreased the angiogenic score of EA.hy926 endothelial cells, showing that metformin not only reduced VEGF expression but also the angiogenic potential of conditioned media from EOC cells.
In summary, the present study sheds light on some mechanisms by which metformin prevents the tumour-promoting effects of NGF in EOC cells. These involve the decrease in the presence of oncoproteins, such a c-MYC and VEGF, as well as the transcriptional activity of MYC and β-catenin-TCF-Lef, which were paralleled by an upregulation of miR-23b and miR-145 following metformin treatment (Figure 7
). Taken together, these results strongly suggest that metformin should be considered as a future alternative in the treatment of EOC. As suggested by a reviewer of this manuscript, additional data related to EOC biopsies will be necessary in order to corroborate in vitro findings, but due to the COVID-19 situation, such experiments had to be delayed and preliminary results were presented in this communication.