SPHK1 Is a Novel Target of Metformin in Ovarian Cancer

This article is featured in Highlights of This Issue, p. 837

Biologically active lipids are key signaling molecules that mediate various aspects of cellular communication and play a role in several oncogenic processes, including cell proliferation, migration, survival, and pluripotency (1). Sphingolipids are a unique group of bioactive lipids thought to regulate tumor progression (2). For example, ceramides and sphingosine induce apoptosis and cell senescence, whereas sphingosine-1-phosphate (S1P) promotes cell growth, proliferation, and migration (3–5). The balance between ceramide/sphingosine and S1P may influence cancer cell fate and is referred to as the sphingolipid rheostat (Fig. 1A; reviewed in ref. 6).

Sphingosine kinases (SPHK), including SPHK1, catalyze the formation of S1P from sphingosine and are pivotal regulators of the balance between ceramides, sphingosine and S1P. Increased expression of SPHK1 promotes tumor growth and metastasis in breast, colorectal, and lung cancers (1, 3, 7, 8), and the role of the S1P pathway is becoming increasingly relevant in ovarian cancer (9, 10). Two older studies reported higher S1P levels in the serum and ascites in ovarian cancer patients compared with controls (11, 12), and a more recent study showed that SPHK1 is highly expressed in tumor stroma of high-grade serous ovarian cancer (13). Because increasing evidence suggests that SPHK1 is tumor promoting, attempts have been made to inhibit tumor growth by targeting this kinase (14). Unfortunately, progress with this approach has been limited by the toxicity of the agents (1). One approach to overcome drug toxicity is to utilize drug repurposing where efforts are focused on agents that already have demonstrated safety and are widely used for noncancer indications (15).

Metformin is a lead candidate for drug repurposing in cancer. The diabetes drug started receiving considerable attention as a potential cancer therapeutic after retrospective studies reported that use of metformin by patients with diabetes was associated with improved survival in several types of cancers, including ovarian cancer (16, 17). Preclinical data then began to accumulate demonstrating anticancer effects of metformin in several types of cancer (18). In ovarian cancer, we demonstrated that metformin inhibits cancer growth in vitro and in multiple mouse models (19, 20). Although the preliminary findings in multiple cancer types are promising, metformin's mechanism of action in cancer has been difficult to define. Early studies focused on metformin's activation of AMPK (21) and alteration of serum insulin levels (22). In a recent study, we focused on defining global metabolic changes and demonstrated that metformin accumulates in human ovarian tumors at concentrations that inhibit cell-intrinsic mitochondrial function including altering levels of tricarboxylic acid cycle intermediates (23).

Preclinical data demonstrating anticancer effects of metformin have led to several prospective clinical trials testing metformin as a treatment for various cancers (24), including our ongoing phase II randomized trial of metformin as upfront treatment for ovarian cancer (NCT02122185). However, if metformin is to advance as a cancer treatment, it is critically important that the molecular mechanisms of action in cancer be defined and that biological predictors of response be identified. Given the effects of metformin on tumor metabolism (23) and the drug's ability to alter lipid signaling (22, 25), in this study, we postulated that metformin's anticancer effects may be mediated, in part, through regulation of the S1P rheostat and that activity of this pathway might predict response to metformin in ovarian cancer.

The findings show that elevated SPHK1 expression and S1P levels promote ovarian cancer growth by increasing multiple oncogenic processes, including proliferation and self-renewal in vitro as well as increased tumor growth in vivo. Moreover, we identified a novel mechanism by which metformin inhibits activity of hypoxia-inducible factors (HIF1α and HIF2α) to repress SPHK1. Together, these findings support a critical role for S1P homeostasis in ovarian cancer tumor progression and show that tumors that possess an activated S1P metabolic profile might be uniquely susceptible to the anticancer effects of metformin.

Subjects provided informed consent, and fasting serum and tumor samples were collected as part of an ongoing biorepository approved by the University of Chicago Institutional Review Board. Tissue microarrays were constructed from tumor implants collected at multiple abdominal sites during debulking surgery. The intensity, frequency, and distribution of SPHK1 IHC staining were quantified by two subspecialty-trained gynecologic pathologists (R. Lastra and S.M. McGregor). Weak staining was scored for very focal and light intensity of SPHK1, whereas strong staining was scored for high frequency and dark intensity. Serum metabolite analysis was performed as described previously (26). For analysis of serum S1P levels, global metabolic profiling was performed during a prior study by our group (23), and reanalyzed for S1P content between control and metformin-treated groups. Briefly, Q-Exactive mass spectrometer was used for untargeted metabolomics with the SIEVE package (Thermo Fisher Scientific), and ions with larger changes in the treatment group (fold change > 2) were tentatively assigned by database searching using ChemSpider (http://www.chemspider.com/) using the detected mass to charge ratio (m/z) and accurate mass within 5 ppm (26).

Ovarian cancer cells were maintained in standard DMEM supplemented with 10% FB-Essence (VWR Life Science SeraDigm) with 1% penicillin/streptomycin (Corning), 1% MEM nonessential amino acids (Corning), and 1% MEM vitamins (Corning). All cell lines were genotyped to confirm their authenticity (IDEXX Bioresearch short tandem repeat marker profiling every 3 months), and were used between 2 and 12 passages from first thaw. All experiments were carried out in low-glucose DMEM (VWR Life Science) supplemented as above. The pLX304 SPHK1-V5 lentiviral expression vector was purchased from DNASU. The constitutively active HIF1α (P402A/P564A) and HIF2α (P405A/P531A) plasmids were a gift from William Kaelin (27) obtained through Addgene (338299). 5HRE-GFP reporter was a gift from Martin Brown and Thomas Foster obtained through Addgene (28). SPHK1 siRNA was obtained from GE Dharmacon. Metformin, phenformin, CoCl₂, and thiazolyl blue tetrazolium bromide were purchased from Sigma-Aldrich. S1P (Huzzah S1P) was purchased from Avanti Polar Lipids. Hypoxia was induced through 1% O₂/5% CO₂ using Hypoxia Incubator Chamber by Stemcell Technologies.

Cellular lipids were extracted by a modified Bligh and Dyer procedure (29) with the use of 0.1N HCl for phase separation as described in ref. 30. C17-S1P (40 pmol) was used as internal standard and was added during the initial step of lipid extraction. The extracted lipids were dissolved in methanol/chloroform (4:1, v/v), and aliquots were taken to determine the total phospholipid content (31). Analyses of sphingoid base-1-phosphates were performed by electrospray ionization tandem mass spectrometry (ESI-LC/MS/MS). Briefly, API4000 Q-trap hybrid triple quadrupole linear ion-trap mass spectrometer (Applied Biosystems) equipped with a turboionspray ionization source interfaced with an automated Agilent 1100 series liquid chromatograph and autosampler (Agilent Technologies), as performed in ref. 31.

Assays measuring cell migration and invasion were performed as done previously by our lab (32). Briefly, to test migration, a wound closure assay was utilized with cells seeded in a 96-well plate 24 hours prior to the assay. The scratch was made by using Woundmaker (Essen Bioscience), and images were taken at 0 minutes and 16 hours. Migration was quantified using the ImageJ (NIH) to measure relative distance over time. To assess invasion, ovarian cancer cells were seeded onto 0.8-μm pore transwell insert (Corning) with 300 μL serum-free DMEM, as we had done previously (33). In the lower chamber (24-well plate), 700 μL of serum-free low glucose DMEM supplemented with 0.1% bovine-serum albumin was treated with S1P (100 nmol/L) in the presence or absence of metformin (5 mmol/L) for 24 hours. Cells in the top of the chamber were removed using a Q-Tip, and invaded cells were stained using Wright Stain (Sigma) for 20 minutes, fixed with 4% paraformaldehyde (Sigma) for 10 minutes, and then imaged using Zeiss Axiovert 200M (Zeiss) and quantified using ImageJ (NIH).

Ovarian cancer cells were seeded at 5 × 10² or 1 × 10³ cells, respectively, per 6-well dish. Cells were allowed to grow overnight, and then media were replaced and cells were treated for an additional 9 days prior to analysis. Cells were washed with 1× PBS, fixed with 4% paraformaldehyde for 10 minutes, and then stained with crystal violet (0.5%) for 30 minutes at room temperature. After removing crystal violet, plates were imaged and the number of colonies was manually counted using ImageJ (NIH).

Cells were plated in 96-well plates in quintuplicate overnight and subsequently treated as indicated. Cell viability was determined via MTT assay with 0.5 mg/mL thiazolyl blue tetrazolium bromide (Sigma), as previously described (34). Following log(x) transformation and normalization per data set, IC₅₀ values were determined using nonlinear regression (using log (inhibitor) vs response – variable slope [four parameters]).

Cells were plated in 6-well dishes overnight and transfected with 5HRE-GFP the following day. After 24-hour recovery, cells were treated with CoCl₂ with or without metformin for 4 hours. Cells were imaged using Zeiss Axiovert 200M (Zeiss) and then the number of cells positive was quantified using ImageJ software (NIH).

Cell lysates were prepared in RIPA buffer, and immunoblotting was performed as previously described (19). Nuclear fractionation was performed using the REAP cellular fractionation method, as described in ref. 35. Antibodies used were as follows: SPHK1 (Bethyl Laboratories), SGPL1, and HIF1α (Abcam), HIF2α (Novus Biologicals), phospho-AKT (S473), AKT, SOX2, phospho-AMPK (T172), AMPK, c-MYC, OCT-1, PKM1/2, β-Tubulin, Histone H3, cleaved caspase-3, Snail, V5-tag, and β-Actin (Cell Signaling Technology).

RNA was extracted using TRIzol (Invitrogen) and transcribed into cDNA using cDNA Reverse Transcription Kit (Applied Biosystems). Quantitative real-time reverse transcriptase PCR (quantitative RT-PCR, qRT-PCR) was performed as described (36) using Applied Biosystems 7500 Real-Time PCR System. SPHK1, SPHK2, SGPL1, ASAH1, ASAH2, and GAPDH TaqMan probes were obtained from Applied Biosystems. Relative levels of messenger RNA (mRNA) expression were calculated by the 2^−ΔΔCT method (37), using GAPDH as the housekeeping gene for normalization.

All animal experiments were approved by the University of Chicago Institutional Animal Care and Use Committee. SPHK1 or vector control stably transfected OVCAR5 cells (2.5 × 10⁴) were injected subcutaneously into the left frank of 7-week-old female athymic mice. The tumors were measured with calibers as Tumor volume = 1/2(length × width²). Eigteen days after injection of the cancer cells, the mice were sacrificed and tumors excised. Tumor samples were stained by IHC with antibodies against Ki67 (Thermo Fisher Scientific) and SOX2 (Cell Signaling Technologies). Slides were imaged using Zeiss Axiovert 200M (Zeiss), and the number of cells with positive staining per high power field were recorded.

Specific analyses performed for each experiment are described in the figure legends. In all analyses, data were evaluated using a one-way ANOVA with post hoc Tukey, or two-tailed t test, as appropriate; P < 0.05 was considered statistically significant (GraphPad Prism 7).

S1P levels are determined by activity of a rheostat whereby ceramide is converted to sphingosine, and phosphorylation of sphingosine generates S1P, which can then be degraded (reviewed in refs. 2, 6; Fig. 1A). Clinically, if S1P activity contributes to ovarian cancer growth, identifying a safe and readily available agent that modulates this pathway could benefit patients. The biguanide metformin has been shown to affect intracellular and circulating levels of lipids in diabetes (22, 25, 38), metabolic syndrome (39), and cardiovascular disease (40). Therefore, we asked if metformin modulates the S1P rheostat in ovarian cancer. Using serum from patients with stage III/IV ovarian cancer, serum metabolites were measured using metabolomics (26). The analysis demonstrated that patients with ovarian cancer who were using metformin for the treatment of type II diabetes mellitus had significantly lower serum S1P levels than patients not using metformin (Fig. 1B). To further characterize the effects of metformin on the sphingolipid rheostat, an ovarian cancer cell line was treated with metformin or control and lipidomic profiling was performed. Mirroring the findings in the patient serum, metformin-treated cells had markedly reduced S1P and sphingosine levels, whereas ceramide levels were increased (Fig. 1C). Further lipidomic analysis of the components of the S1P pathway indicated that dihydroceramide (DH-ceramide) and dihydrosphingosine (DH-sphingosine) were unaltered by metformin treatment, whereas dihydro-S1P (DH-S1P) was markedly decreased (Fig. 1D). Combined, these findings support the hypothesis that metformin shifts the rheostat toward reduced S1P production. To determine whether S1P does indeed promote ovarian cancer, cancer cell migration and invasion, critical events in ovarian cancer progression (41), were tested. Wound closure assays demonstrated that exogenous S1P promoted migration of ovarian cancer cells and that this effect was attenuated by pretreatment with metformin (Fig. 1E). Similarly, S1P promoted invasion of the HeyA8 ovarian cancer cell line and metformin treatment blunted this effect (Fig. 1F). Together, these data suggest that S1P production and S1P-driven signaling are novel targets of metformin.

Given the observation that metformin shifts the rheostat from S1P toward ceramide, we evaluated the effect of metformin on the regulatory steps of the S1P synthesis pathway (shown in Fig. 1A). The metabolic profile of decreased S1P and DH-S1P levels suggested that metformin might inhibit SPHK1, the key regulatory enzyme that catalyzes the conversion of DH-sphingosine to DH-S1P. To test if SPHK1 is a target of metformin, ovarian cancer cell lines were treated with metformin or another biguanide, phenformin, and SPHK1 expression was evaluated. The results demonstrate that both drugs significantly decreased SPHK1 mRNA levels in a dose-dependent manner in two different cell lines (Fig. 2A). Protein expression of SPHK1 was also reduced by both drugs (Fig. 2B). The effects of biguanides on other key regulators of the S1P rheostat, including ASAH1, SPHK2, and SGPL1, were inconsistent and only observed at very high doses (Supplementary Fig. S1). This evaluation of the regulators of S1P rheostat suggested that metformin targets SPHK1. If SPHK1 is important in metformin's anticancer effect, there should be strong evidence that SPHK1 promotes ovarian cancer growth. To understand if this was the case, we sought to test if SPHK1 is tumor promoting in ovarian cancer by directly evaluating the effect of overexpression of SPHK1 on several hallmarks of cancer growth by using ovarian cancer cell lines that possess otherwise nearly undetectable endogenous levels of SPHK1 (Supplementary Fig. S2). The results show that cells with ectopic SPHK1 expression had a higher rate of migration (Fig. 2C) and proliferated more rapidly than control transfected cells (Fig. 2D). Consistent with recent reports of S1P's role in AKT-dependent cell proliferation (8, 42), we also found markedly increased activation of AKT (phospho-S473) in SPHK1-overexpressing cells (Fig. 2E). Next, using a colony formation assay, we tested whether SPHK1 promotes clonogenic growth. SPHK1-overexpressing cells formed significantly more colonies (Fig. 2F), and SOX2, a key factor involved in stem cell–like phenotypes including self-renewal (43, 44), was markedly upregulated at both the mRNA and protein levels (Fig. 2G and H, respectively). The ability of SPHK1 expression to promote proliferation, migration, clonogenicity, and SOX2 expression was recapitulated in an additional cell line representative of high-grade serous ovarian cancer (Supplementary Fig. S3). Finally, in a xenograft mouse model of ovarian cancer, overexpression of SPHK1 resulted in increased tumor burden compared with control mice (Fig. 2I). Mirroring the findings in cell lines, IHC analysis of mouse tumors identified increased expression of Ki67 (Fig. 2J) and SOX2 (Fig. 2K) in the SPHK1-overexpressing tumors compared with controls. Altogether, these data suggest that SPHK1 promotes tumorigenicity in ovarian cancer through induction of an aggressive phenotype characterized by increased motility, proliferation, and self-renewal.

In endothelial cells and glioma (45, 46), SPHK1 expression is induced by hypoxia, and SPHK1 has been shown to possess two hypoxia-response elements (HRE) in its promoter which are targeted by HIFs (46); therefore, we asked if hypoxia-inducible factors (e.g., HIF1α, HIF2α) upregulate SPHK1 in ovarian cancer. To answer this, three ovarian cancer cell lines were transfected with the constitutively stable forms of either HIF1α (P402A/P564A) or HIF2α (P405A/P531A; ref. 32) and SPHK1 expression was evaluated. In all three cell lines tested, induction of either HIF1α or HIF2α significantly increased SPHK1 protein expression (Fig. 3A). Exposure to metformin prevented the induction of SPHK1 protein and mRNA expression by ectopic expression of HIF 1α (left) or HIF 2α (right; Fig. 3B; Supplementary Fig. S4). Likewise, exposure to metformin prevented the induction of SPHK1 protein by hypoxia (Fig. 3C). Interestingly, HIF expression itself was not robustly repressed by metformin across both cell lines tested; therefore, we next asked if metformin might instead alter the activity of HIFs through reducing its nuclear localization. Cell fractionation demonstrated that, following HIF stabilization with CoCl₂ (47), metformin significantly reduced HIF1α in both the nucleus and the cytosol as well as nuclear HIF2α (Fig. 3D; Supplementary Fig. S5). These findings were confirmed using cells with constitutive expression of HIF1α or HIF2α. Here, it was noted that metformin reduced HIF1 or HIF2 solely in the nuclear fraction (Fig. 3E). Finally, to directly test the impact of metformin on HIF transcriptional activity, ovarian cancer cells were cotransfected with the HRE plasmid reporter, 5HRE (28), concurrently with either HIF1α or HIF2α stable mutants and then subsequently treated with metformin. Figure 3F shows that in normoxia, when HIF is stabilized using constitutively expressing mutants, metformin significantly inhibited HIF transcriptional activity. Combined these findings indicate that metformin suppresses hypoxia-induced SPHK1 expression.

Although there is enthusiasm for metformin's potential as a cancer therapeutic, it is unlikely to be effective for all patients. In ovarian cancer cell lines, we have found the cytotoxic effects of metformin to be variable, with some cell lines being very sensitive to metformin (HeyA8 and TYKnu) and others to a lesser extent (CAOV3, OVCAR5, Kuramochi, and SNU119; Fig. 4A). Clinically, it is important to understanding a priori which patients will have the greatest anticancer effects from metformin. Interestingly, when evaluating metformin-sensitive and -resistant cell lines, we noted that, compared with metformin-resistant cells, metformin-sensitive cells expressed high levels of SPHK1 and low levels of the enzyme responsible for S1P degradation, S1P lyase (SGPL1; Fig. 4B/C). This finding indicated that an SPHK1^high/SPGL1^low profile may predispose ovarian cancer cells to the cytotoxic effects of metformin. Supporting this concept, in three different cell lines the IC₅₀ for metformin was reduced by >50% in SPHK1-overexpressing cells compared with controls (Fig. 4D). Likewise, the induction of apoptosis by metformin was greater in SPHK1-overexpressing than in control transfected cells, as indicated by increased cleavage of caspase-3 (Fig. 4E). Given the marked sensitization of ovarian cancer tumor cells to metformin by ectopic expression of SPHK1, we next asked if decreased levels of SPHK1 were linked to resistance to the drug, and whether this was associated with hypoxia-dependent regulation of the kinase. To answer this question, SPHK1 was silenced in the metformin-sensitive cell line TYKnu, which expresses relatively high levels of SPHK1 compared with other ovarian cancer cell lines. Knockdown of SPHK1 using siRNA markedly increased the IC₅₀ of metformin compared with control cells as measured by MTT assay, indicating that reduced SPHK1 increases resistance to the drug (Fig. 4F–G; Supplementary Fig. S6). Intriguingly, hypoxia dramatically sensitized ovarian cancer cells to metformin which, importantly, was potently abrogated in cells with reduced SPHK1 expression (Fig. 4F; Supplementary Fig. S6), suggesting that SPHK1 expression downstream of hypoxic signaling is central to metformin's impact on cell viability in this setting. In addition, knockdown of SPHK1 in these cells prevented the induction of caspase-3 cleavage by metformin (Fig. 4H). Interestingly, regulation of metformin sensitivity by SPHK1 expression appeared to be independent of other putative markers of metformin's efficacy, as neither overexpression nor knockdown of SPHK1 altered expression of metformin's membrane transporter OCT1, its canonical target AMPK, or glycolytic enzymes that may mediate its cytotoxic effects such as c-MYC or PKM1/2 (ref. 20; Supplementary Fig. S7). Taken together, these data indicate that in ovarian cancer cells, SPHK1 expression independently modulates sensitivity to the cytotoxic effects of metformin.

To understand whether these preclinical findings are relevant to patients with ovarian cancer, we analyzed the expression of SPHK1 in patient pathology specimens (n = 389) from common anatomic sites of tumor growth. The results show that SPHK1 is strongly expressed in the cytoplasm of cells in >40% of primary ovarian tumors, omental metastases, and peritoneal metastases (Fig. 5A), indicating that SPHK1/S1P signaling may be an important pathway throughout ovarian cancer tumor progression. Given that the prior experiments indicated that SPHK1 activity may be associated with metformin response, we asked if metformin would attenuate the cancer-promoting effects of SPHK1. The results show that colony formation promoted by overexpression of SPHK1 was abrogated by metformin (Fig. 5B). Conversely, silencing SPHK1 markedly reduced the ability of ovarian cancer cells to form colonies and treatment with metformin did not further inhibit clonogenicity in these cells (Fig. 5C). Consistent with a reduction in colony formation by metformin, expression of the self-renewal factor SOX2 and its downstream target Snail induced by SPHK1 was prevented by the drug (Fig. 5D). In a final experiment, we tested the necessity of targeting SPHK1 to metformin's ability to prevent invasion. In addition, based on the previous observations, we asked if metformin's effect on ovarian cancer invasion was dependent on the upregulation of SPHK1 by hypoxia. The results show that ovarian cancer cell invasion was significantly increased under hypoxia and that the hypoxia-induced increase in cancer cell invasion was prevented by concurrent treatment with metformin (Fig. 5E). Importantly, silencing SPHK1 inhibited invasion of cancer cells in normoxia and significantly reduced invasion induced by hypoxia (Fig. 5E; Supplementary Fig. S8). Notably, silencing SPHK1 prevented the ability of metformin to reduce hypoxia-driven invasion in these cells (Fig. 5E). Combined, these findings support the concept that the tumor-promoting effects of SPHK1 in ovarian cancer can be targeted using metformin and directly relate to the efficacy of the drug in preventing tumor growth (Fig. 5F).

The biguanide metformin has been shown to affect intracellular and circulating levels of lipids in diabetes (22, 25, 38) and significant effort is under way to understand if the diabetes medication has potential as cancer treatment (48). Here, we show that the S1P rheostat is a novel target of metformin in ovarian cancer. As depicted in Fig. 1, serum levels of S1P are lower in patients with ovarian cancer that were taking metformin and, in vitro, metformin shifts the sphingolipid rheostat away from oncogenic S1P. Increasing evidence supports an important role of bioactive lipids such as S1P and the upstream kinase, SPHK1, in cancer (3) and the experiments reported here demonstrate that metformin suppresses migration and chemotaxis induced by ectopic SPHK1 and downstream S1P signaling.

Further investigation of the effects of metformin on the S1P rheostat revealed that inhibition of SPHK1 is the central mechanism by which metformin represses S1P. Functional assays indicated that ectopic SPHK1 expression alone was sufficient to recapitulate the protumorigenic effects of S1P, suggesting SPHK1′s pivotal role in S1P-mediated oncogenesis. Although earlier studies have shown S1P and SPHK1 to increase proliferation and migration in ovarian cancer in vitro (11, 49), here we demonstrate that SPHK1 directly induces activation of AKT and enhances SOX2 expression, thereby promoting ovarian cancer tumorigenicity in vivo. Together, these data support the S1P pathway and SPHK1 in particular, as potential targets for ovarian cancer treatment. An IHC analysis of over 300 ovarian cancer patient samples confirmed expression of SPHK1 in all the metastatic sites affected by ovarian cancer; however, to our knowledge, no significant attempts have been made to target SPHK1 for ovarian cancer treatment. There are phase I clinical trials targeting SPHK1 in other solid tumors (14, 50, 51) and, in a mouse model, pharmacologic inhibition of SPHK1 with fingolimod decreased mouse ovarian tumor weight (10). Unfortunately, traditional inhibitors of SPHK1 may prove to have limited clinical utility as the inhibitors induce hemolysis and hepatoxicity in preclinical models (1) and use of fingolimod in humans is limited by immunosuppression (52).

To increase safety, decrease cost, and expedite clinical use of new cancer therapeutics a drug-repurposing paradigm is attractive. In this approach, anticancer properties are identified in agents that are already widely used for noncancer indications (18). A cancer therapeutic directed against S1P would, ideally, shift the S1P rheostat toward reduced S1P production (1). This concept is supported by preclinical studies in multiple cancer types that demonstrate increased tumor growth when S1P is elevated either due to increased synthesis as a result of SPHK1/2 upregulation (7, 8) or due to decreased S1P degradation as a result of SGPL1/2 inhibition (53). In the current study, we demonstrate that metformin reduces S1P levels by downregulating SPHK1. Given the limitations of other compounds that target SPHK1, the data presented here demonstrate the potential to expedite the exploitation of the S1P pathway through repurposing metformin as an SPHK1 inhibitor in ovarian cancer.

To understand the mechanism by which metformin inhibits SPHK1, we focused on HIFs as it has been reported that metformin inhibits hypoxia-induced HIF1α expression through repression of complex I–dependent ROS production (54) and prior reports suggested that SPHK1 expression and activity may be induced by hypoxia (45, 46, 55). The results show that in ovarian cancer, HIF stabilization potently induces SPHK1 mRNA and protein expression and that metformin inhibits hypoxia-induced SPHK1 upregulation. It is notable that metformin did not reliably repress HIF expression in the cell lines tested, and metformin, while inconsistent with some studies (47), has been reported by others to be ineffective at reducing HIF expression at the dose and duration used here (56). However, our data do support that, in the current system, metformin abrogated the nuclear accumulation and concomitant transcriptional activity of HIFs in both normoxic and hypoxic conditions in vitro, lending to a novel mechanism by which the drug alters HIF signaling. Although metformin was capable of repressing HIF-induced SPHK1 expression, it is still unclear whether HIFs specifically are necessary targets of the drug's action on SPHK1, as indirect targets or cofactors, such as ARNT1, CBP, and p300 (57), could be involved in its transcriptional activity. Further, our data indicate that metformin may, in some cases, inhibit translocation or import of HIFs into the nucleus; however, the exact mechanism by which the drug elicits this effect remains undetermined. Also, given the importance of SOX2 as a key regulator in pluripotency and tumor development (58), the finding that SPHK1 promotes self-renewal through the upregulation of SOX2 lends to a new understanding of the contribution of sphingolipid signaling to tumor progression. The ability of metformin to abrogate SOX2 expression and the associated clonogenicity of cells with ectopic SPHK1 expression identifies a new mechanism by which the diabetic drug inhibits vital oncogenic processes in ovarian cancer.

New therapeutic approaches for ovarian cancer are desperately needed as improvement in outcomes for patients have been constrained by limited understanding of the drivers of ovarian carcinogenesis and the introduction of very few new treatments in 20 years (59). The results reported here demonstrate that targeting SPHK1 expression and downstream S1P signaling may be a novel approach to ovarian cancer treatment. Demonstrating that SPHK1 is directly enhanced by hypoxia identifies a novel, actionable, mechanism by which SPHK1 and concomitant sphingolipid signaling may be targeted in ovarian cancer. Importantly, our data support that metformin can be repurposed to inhibit SPHK1, thereby repressing S1P oncogenic signaling in ovarian cancer.

Although there is significant enthusiasm for the potential of metformin as a cancer treatment, it is unlikely that that drug is going to be effective for all patients and, clinically, it is critically important to be able to identify in advance the patients most likely to benefit from this approach. The findings reported here show that SPHK1 expression levels within the tumor cell directly influence the IC₅₀ of metformin's cytotoxic effects as well as the induction of apoptosis by the drug, with elevated SPHK1 increasing sensitivity and reduced SPHK1 levels conferring resistance to metformin. Importantly, the influence of SPHK1 on metformin's effects was found to be independent of previously described putative markers that indicate sensitivity to the drug. Together, these results suggest that high SPHK1 expression in the tumor, or perhaps even serum S1P levels, may predict increased response to metformin and support a decision for its use therapeutically. This finding could be relevant to the >55 clinical trials testing metformin as treatment for breast, endometrial, prostate, pancreas, and lung cancers (24), and future large clinical trials of metformin could consider incorporating high expression of SPHK1 as a predictive biomarker.

Although this report strongly indicates SPHK1 expression level predicts metformin sensitivity or resistance, studies identifying serum S1P levels as a potential, and importantly a more accessible, biomarker are needed. Indeed, one limitation of the current study analyzing our clinical samples is that no cancer-free subjects were assessed, and whereas patients were matched on age and cancer stage, there was no inclusion of diabetics on other drugs, which may affect these results. Furthermore, although SPHK1 and S1P had similar effects on tumor cell behavior in vitro, it is unclear whether SPHK1 in ovarian cancer tumors directly relates to circulating S1P levels in patients. Although suppression of SPHK1 using inhibitors has been shown to reduce serum S1P in xenograft models in vivo (60), it is clear that other sources of the metabolite result from SPHK1-dependent secretion into plasma, such as from erythrocytes (61), and thus secretion from tumor cells may not entirely account for circulating S1P levels in patients with cancer. Future studies are warranted to evaluate the relationship between SPHK1 and S1P within the tumor and their impact on systemic S1P levels in ovarian cancer, as well as to further characterize whether the circulating metabolite alone may directly provide an indication for metformin's clinical efficacy. Lastly, it is worth noting that the metformin dose used in the present study is significantly lower than doses used in a large number of prior studies, but still is higher than what is observed physiologically in humans (20). These constraints of preclinical investigation make prospective clinical testing of metformin in patients imperative. Looking ahead, recognizing the importance of bioactive lipid pathways in carcinogenesis and targeting those pathways with safe, readily available repurposed drugs might prove to be an unexpected path forward in the journey to improve outcomes for women battling this insidious disease.

No potential conflicts of interest were disclosed.

Conception and design: P.C. Hart, E. Lengyel, I.L. Romero

Development of methodology: P.C. Hart, T. Chiyoda, J.W. Locasale, I.L. Romero

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): P.C. Hart, T. Chiyoda, M. Weigert, R. Lastra, S.M. McGregor, J.W. Locasale, I.L. Romero

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): P.C. Hart, T. Chiyoda, M. Weigert, J.W. Locasale, E. Lengyel, I.L. Romero

Writing, review, and/or revision of the manuscript: P.C. Hart, T. Chiyoda, M. Weigert, R. Lastra, S.M. McGregor, E. Lengyel, I.L. Romero

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): P.C. Hart, X. Liu, M. Curtis, C.Y. Chiang, R. Loth, I.L. Romero

Study supervision: J.W. Locasale, I.L. Romero

This work was supported by grants from the NIH and University of Chicago Cancer Center Support Grant (CA014599) and the Mayo Clinic Ovarian Cancer SPORE (P50CA136393-06A1). I.L. Romero is supported by grants from the NIH (2K12HD000849-26), and the American Board of Obstetrics and Gynecology. E. Lengyel is supported by grants from the NCI (5R01CA111882-07 and 1R01CA169604-01A1). P.C. Hart is a recipient of the Colleen's Dream Foundation Young Investigator Award and the NIH Loan Repayment Program. T. Chiyoda is a recipient of JSPS and Kanae Foundation Fellowships for Research Abroad.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.