Aberrant N-glycan Golgi remodeling and metabolism are associated with epithelial–mesenchymal transition (EMT) and metastasis in patients with breast cancer. Despite this association, the N-glycosylation pathway has not been successfully targeted in cancer. Here, we show that inhibition of the mevalonate pathway with fluvastatin, a clinically approved drug, reduces both N-glycosylation and N-glycan-branching, essential components of the EMT program and tumor metastasis. This indicates novel cross-talk between N-glycosylation at the endoplasmic reticulum (ER) and N-glycan remodeling at the Golgi. Consistent with this cooperative model between the two spatially separated levels of protein N-glycosylation, fluvastatin-induced tumor cell death was enhanced by loss of Golgi-associated N-acetylglucosaminyltransferases MGAT1 or MGAT5. In a mouse model of postsurgical metastatic breast cancer, adjuvant fluvastatin treatment reduced metastatic burden and improved overall survival. Collectively, these data support the immediate repurposing of fluvastatin as an adjuvant therapeutic to combat metastatic recurrence in breast cancer by targeting protein N-glycosylation at both the ER and Golgi.
These findings show that metastatic breast cancer cells depend on the fluvastatin-sensitive mevalonate pathway to support protein N-glycosylation, warranting immediate clinical testing of fluvastatin as an adjuvant therapy for breast cancer.
The first-line therapy for early-stage breast cancer is surgical removal of the tumor, followed by adjuvant therapies (1). Despite aggressive treatment, 15% to 20% of patients with early-stage breast cancer experience recurrence, often as distant metastases (1). Prevention or delay of metastatic recurrence in breast cancer would represent a key advance in the treatment of this disease. Several retrospective studies have indicated that the risk of postsurgical breast cancer recurrence is reduced by 30% to 60% in patients who are taking statins (2–5), a class of approved drugs that lowers serum cholesterol. Increased duration of adjuvant statin use is associated with decreased risk of recurrence (5), suggesting that long-term intake of statins in the adjuvant setting may prolong patient survival. Preclinically, statins have been shown to inhibit metastasis in a broad range of cancers (6–9); however, the precise mechanism remains unclear. Mechanistic understanding of the effect of fluvastatin on metastatic breast cancer cells may provide the essential insight required to guide the design of clinical trials, identify biomarkers of statin response, and provide a starting point for the development of additional agents to target metastatic recurrence.
Statins inhibit the metabolic conversion of 3-hydroxy-3-methylglutaryl coenzyme A (HMG-CoA) to mevalonate (MVA), the rate-limiting step of the MVA pathway (Fig. 1A). The MVA pathway synthesizes cholesterol; farnesyl pyrophosphate (FPP) and geranylgeranyl pyrophosphate (GGPP), required for protein prenylation; coenzyme Q (CoQ), required for the electron transport chain (ETC); and dolichol, required for protein N-glycosylation (Fig. 1A; ref. 10). Statin-triggered tumor cell death can be rescued by exogenous GGPP; therefore, statin activity has been linked to inhibition of prenylated proteins (11, 12). However, recent interrogation has revealed that statins preferentially target cancer cells with enriched mesenchymal features, but this effect is uncoupled from inhibition of RAS family protein prenylation (11). This suggests an alternative mechanism of action of fluvastatin on cells undergoing EMT, which occurs downstream of GGPP. As disseminated primary tumor cells often gain mesenchymal characteristics while losing epithelial features (13), investigating this novel mechanism is of interest as targeting breast cancer cells with mesenchymal phenotypes may have utility in the adjuvant setting to prevent metastatic recurrence.
Herein, we show that statin-dependent depletion of dolichol selectively inhibits the viability of EMT-induced invasive breast cancer cells. Dolichol is a group of long-chain isoprenoids that comprises the lipid component of lipid-linked oligosaccharides (LLO), essential for N-linked glycosylation of nascent peptides translated in the secretory pathway (Fig. 1A; ref. 10). The oligosaccharide in LLO is added cotranslationally to asparagine on secretory and membrane proteins at the endoplasmic reticulum (ER), and subsequently processed to more complex structures by glycosidases and glycosyltransferases in the ER and Golgi during transit to the cell surface. Surprisingly, we show that in addition to reducing LLO-dependent N-glycosylation at the ER, fluvastatin treatment also reduced subsequent branching of complex-type N-glycans that occurs in the medial-Golgi. Oncogenic mutations induce N-glycan branching by increasing expression of MGAT4, MGAT5, and metabolic pathways to nucleotide-sugars, which modify receptor kinases that promote epithelial-to-mesenchymal transition (EMT) and metastasis (14–18). Knockout of MGAT5 in mice has been shown to reduce mammary tumor growth and metastases (19), and knockdown of MGAT1 significantly decreased tumor growth and incidence of lung metastases in a prostate cancer xenograft model (20). Moreover, N-glycan branches and the number of glycan-occupied sites in receptors act cooperatively as ligands for multivalent galectins, thereby regulating cell surface residency and signaling (16). To date, however, ER and Golgi levels of protein N-glycosylation in cancer metastasis has not been successfully targeted. Here, we show for the first time that aberrant protein N-glycosylation in metastatic breast cancer cells can be therapeutically targeted by inhibiting dolichol biosynthesis using fluvastatin, using a model of spontaneous postsurgical metastasis that closely follows the course of human breast cancer progression and treatment (21). Our results demonstrate that postsurgical fluvastatin treatment attenuates the development of breast cancer metastases and improves overall survival by >30%. Taken together, our results support the immediate clinical testing of fluvastatin as a safe and effective therapeutic in the adjuvant setting, and support the further development of novel therapeutics to combat metastatic recurrence in breast cancer by inhibiting aberrant protein N-glycosylation.
Materials and Methods
Fluvastatin was purchased from US Biologicals (F5277–76). TGFβ was purchased from PeproTech (100–21). PNGase F was purchased from NEB (P0704). Complete protease inhibitor was purchased from Roche (11697498001). RapiGest SF was purchased from Waters (186001861). Sialidase was purchased from Glyko (GK80040). All other chemicals were purchased from Sigma unless otherwise specified. In the conduct of research involving hazardous organisms or toxins, the investigators adhered to the CDC-NIH Guide for Biosafety in Microbiological and Biomedical Laboratories.
MCF10A cells were a kind gift from Dr. Senthil Muthuswamy. MDA-MB-231 and LM2–4 cells were a kind gift from Dr. Robert Kerbel. All other cell lines were obtained from ATCC. HEK293Tv, LM2–4, MCF-7, MCF10A, and MDA-MB-231 cells were cultured at 37°C in a humidified atmosphere at 5% CO2 in supplemented growth media (21–23). All cell lines were authenticated by short tandem repeat (STR) profiling and tested to be free of Mycoplasma monthly using commercial mycoplasma detection kits. All cell lines were used between 3 to 20 passages after thawing. Transgene expression was stably introduced into MCF10A cells using retroviral insertion with pLPC, a kind gift from Dr. Roberta Maestro, or pBabePuro (22). In the conduct of research utilizing recombinant DNA, the investigators adhered to NIH Guidelines for research involving recombinant DNA molecules.
HeLa Flp-In-TREx cells were transfected with two guide RNA (sgRNA) in the CRISPR/Cas9 px459 vector targeting exon 4 and the flaking intron for removal of 110 bp from the SLC3A2 gene. sgRNA#1: CAGATTCAACCGGAGGTACC, sgRNA#2: CCGCGTTGTCGCGAGCTAC. Deletions were confirmed by sequencing. Inducible expression was restored by transfecting the cells with human SLC3A2 cDNA cloned into the pcDNA5/FRT/TO vector for single site insertion at a preintegrated FRT recombination site. MGAT1 and MGAT5 mutant MDA-MB-231 cells were generated with CRISPR/Cas9 px459 vector using guide RNA for a deletion within the catalytic domain (https://zlab.bio/guide-design-resources). The null mutations were validated by sequencing and LC/MS-MS analysis of glycopeptidase released N-glycan.
3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assays were performed as described previously (23). Cells were seeded at 750 to 5,000 cells/well in 96-well plates and treated in triplicate with 8 doses of drugs or the solvent control for 72 hours. IC50 values were computed using GraphPad Prism with a bottom constraint equal to 0.
Lysates were prepared in RIPA lysis buffer (25 mmol/L Tris pH 7.4, 150 mmol/L NaCl, 0.5% sodium deoxycholate, 1% NP-40, protease inhibitors). Antibodies used were c-MYC (MAb 9E10, in-house), E-cadherin (CST 3195), vimentin (CST 5741), fibronectin (Abcam ab32419), actin (Sigma A2066), tubulin (Millipore CP06), GP130 (SCB sc-655), EGFR (CST 2232), SLC3A2 (SCB sc-7095), and Ku80 (CST 2180).
For tumors, two sequential slices were stained for hematoxylin and eosin (H&E) or Ki67 (Novus NB110–90592). For lungs, two sequential slices were obtained every 200 μm for three depths containing all five lobes, and stained for H&E or hEGFR (Zymed 28005). Metastatic colonies were identified by hEGFR staining and confirmed by H&E. Total hEGFR positivity was computed using ImageScope.
Cell death assay
Cells were seeded at 250,000/plate overnight, then treated with as indicated for 72 hours. Cells were fixed in 70% ethanol overnight, stained with propidium iodide (Sigma), and analyzed for the subdiploid DNA (“pre-G1”) population as previously described (23).
Total RNA was harvested from subconfluent cells using TRIzol Reagent (Invitrogen). cDNA was synthesized from 500 ng of RNA using SuperScript III (Invitrogen). Real-time quantitative RT-PCR was performed using SYBR Green (Applied Biosystems) with the following primers:
Sample preparation for glycopeptide analysis
A total of 1 × 107 cells were harvested after indicated treatment. Cells were lysed in 1 mL IP lysis buffer (1% Triton-100, 20 mmol/L Tris pH 7.5, 150 mmol/L NaCl, 1 mmol/L EDTA, 1 mmol/L EGTA, cOmplete protease inhibitor), and centrifuged at 14,000 rpm for 30 minutes at 4°C. Lysates were normalized to 2.5 mg/mL and 1 mL was incubated with 20 μL of FLAG beads at 4°C overnight. Beads were washed thoroughly in TBS (50 mmol/L Tris pH 7.5, 150 mmol/L NaCl) and 50 mmol/L ammonium bicarbonate and on-bead trypsin digest was carried out using 0.5 μg of trypsin at 37°C overnight. Glycopeptides were extracted using 0.5% formic acid, vacuumed to dry, and desialidated with 0.5 μL of sialidase at 37°C overnight.
Glycopeptide analysis by LC/MS-MS
Peptides were applied to a nano-HPLC Chip using an Agilent 1200 series microwell-plate autosampler, and interface with an Agilent 6550 Q-TOF MS (Agilent Technologies). The reverse-phase nano-HPLC Chip (G4240–62002) had a 40 nL enrichment column and a 75 μmol/L × 150 mm separation column packed with 5 μmol/L Zorbax 300SB-C18. The mobile phase was 0.1% formic acid in water (v/v) as solvent A, and 0.1% formic acid in ACN (v/v) as solvent B. The flow rate at 0.3 μL/min with gradient schedule; 3% B (0–1 minutes); 3%–40% B (1–90 minutes); 40%–80% B (90–95 minutes); 80% B (95–100 minutes), and 80%–3% B (100–105 minutes). Mascot search was used to identify proteins and peptide sequences coverage. Extract glycopeptide were identified by Agilent Masshunter Quanlititive Analysis software by the presence of hexose (Hex) and N-acetylhexosamine (NAc), such as 204 (HexNAc ions), and 366 (HexHexNAc ions). Glycan structures were predicted for extracted glycopeptides by online GlycoMod (http://web.expasy.org/glycomod/). Glycan structure by MS/MS and occupancy of NXS/T(X≠P) N-glycosylation sites were determined manually.
A total of 15 × 106cells were seeded overnight and treated as indicated. Cells were harvested, suspended in 1 mL of HEPES homogenization buffer (0.25 mol/L sucrose, 50 mmol/L HEPES pH 7.5, 5 mmol/L NaF, 5 mmol/L EDTA, 2 mmol/L DTT, cOmplete protease inhibitor), and lysed using a probe sonicator. Homogenate was cleared at 2,000 × g for 20 minutes at 4°C, then ultracentrifuged at 115,000 × g for 70 minutes at 4°C. The pellet was vigorously suspended in 650 μL Tris buffer (0.8% Triton X-114, 50 mmol/L Tris pH 7.5, 0.1 mmol/L NaCl, 5 mmol/L EDTA, 5 mmol/L NaF, 2 mmol/L DTT, cOmplete protease inhibitor). The homogenate was chilled on ice for 10 minutes, incubated at 37°C for 20 minutes, then phase partitioned at 1,950 × g for 2 minutes at room temperature. The upper phase was discarded. Membrane proteins in the lower phase was precipitated with 1 mL acetone at −20°C overnight.
Precipitated proteins were suspended in 60 μL of suspension buffer (0.25% RapiGest SF, 50 mmol/L ammonium bicarbonate, 5 mmol/L DTT). The completely dissolved solution was heated for 3 minutes at 85°C. Approximately 30 μg proteins was mixed with 0.5 μL of PNGase F, 0.7 μL of sialidase, and 20 μL of 50 mmol/L ammonium bicarbonate, and incubated at 42°C for 2 hours followed by 37°C overnight. Released N-glycans were extracted with 4 to 5 volumes of 100% ethanol at −80°C for 2 hours. The supernatant containing released N-glycans was speed vacuumed to dry.
Homemade porous graphitized carbon (PGC) microtips containing 10 mg PGC in a bed volume of 50 μL was washed with 500 μL of ddH2O, 500 μL of 80% acetonitrile (ACN), and equilibrated with 500 μL 0.1% trifluoroacetic acid (TFA). N-glycan pellets were dissolved in 50 μL of 0.1% TFA and slowly loaded into microtips. Microtips were washed with 500 μL 0.1% TFA. N-glycans were eluted several times with 500 μL of elution buffer (0.05% TFA, 40% ACN). The eluted N-glycans were speed vacuumed to dry.
Global glycan analysis by LC/MS-MS
Analysis of the eluted N-glycans was modified from a previous method (24). Total glycan samples were applied to a nano-HPLC Chip using an Agilent 1200 series microwell plate autosampler, and interface with an Agilent 6550 Q-TOF MS (Agilent Technologies). The HPLC PGC-Chip (G4240–64010) had a 40 nL enrichment column and a 75 μmol/L × 43 mm separation column packed with 5 μmol/L porous graphitized carbon as stationary phase. The mobile phase was 0.1% formic acid in water (v/v) as solvent A, and 0.1% formic acid in ACN (v/v) as solvent B. The flow rate at 0.3 μL/minute with gradient schedule; 5% B (0–1 minutes); 5%–20% B (1–15 minutes); 20%–70% B (15–16 minutes); 70% B (16–19 minutes), and 70%–5% B (19–20 minutes). Free glycans released by PNGase F were identified by Agilent Masshunter Quanlititive Analysis software in the presence of hexose and N-acetylhexosamine. Glycan structures were predicted by online GlycoMod (http://web.expasy.org/glycomod/). Agilent Masshunter Quantitative Analysis software was used to quantify the extracted glycan peaks.
Animal work was carried out with the approval of the Princess Margaret Cancer Centre Ethics Review Board in accordance with the regulations of the Canadian Council on Animal Care. In conducting research using animals, the investigators adhered to the laws of the United States and regulations of the Department of Agriculture. Female SCID mice were obtained from the in-house breeding colony at the Princess Margaret Cancer Centre and at 6 to 8 weeks of age. All mice were maintained under specific pathogen-free conditions with a 12-hour light/dark cycle. Food and water were provided ad libitum.
LM2–4 cells (1 × 106cells in 50 μL) were implanted subcutaneously in female SCID mice (6–8 weeks), obtained in-house from the University Health Network animal colony. Primary tumors were measured every two days and calculated by (width × width × length)/2. After surgical removal of the primary tumors, animals were monitored daily for endpoint, including signs of metastatic load in the lung (labored breathing). Treatment was given daily orally with PBS or 50 mg/kg/day fluvastatin. Necropsy was performed at endpoint where any tissue with evidence of metastatic disease is rapidly excised and fixed in formalin for histopathology.
Quantification and statistical analysis
Statistical analysis was performed using GraphPad Prism 6 and R software. Statistical testing and significance are performed as indicated in the legend of each figure. Histopathologic analyses were independently reviewed by two personnel blinded to group allocation at the time of analysis. Quantification of histochemical analyses was performed using ImageScope software. In vitro experiments were not feasible for randomization or blinding due to the nature of the experiments.
EMT sensitizes breast cancer cells to the inhibition of dolichol synthesis
To delineate the mechanism of statin action on mesenchymal-enriched breast cancer cells, fluvastatin was chosen for our studies based on its favorable pharmacokinetic properties and promising anti–breast cancer activities in the preclinical and clinical preoperative settings (23, 25). We used MCF10A breast epithelial cells as our model system, which allowed for the evaluation of EMT in an isogenic panel of cells in the absence of gross genetic instability (26). Ectopic expression of the EMT-inducing transcription factor SNAIL triggered EMT in MCF10A cells, as shown by downregulation of E-cadherin and upregulation of fibronectin (Fig. 1B). Treatment with fluvastatin readily induced cell death in SNAIL-overexpressing cells, but not vector control cells, as assessed by quantification of DNA content following cell fixation and propidium iodide staining (Fig. 1C). Fluvastatin-induced cell death in SNAIL-overexpressing cells was fully rescued by coadministration with MVA or GGPP, but not FPP (Fig. 1C). FPP and GGPP at the concentrations used have previously been shown to enter the cells and rescue protein prenylation (12, 27). This preferential rescue of statin-induced cell death in tumor cells by GGPP has also been reported in several other cancer cell lines (28, 29), together suggesting that disruption of biological processes downstream of GGPP is critical for statin-induced cell death.
GGPP is required for three biological processes: protein prenylation, synthesis of CoQ used in the ETC, and synthesis of dolichol required for protein N-glycosylation (Fig. 1A; ref. 10). We tested whether inhibiting any of these pathways individually using specific inhibitors (Fig. 1A) could phenocopy statin treatment and preferentially kill breast cancer cells with mesenchymal phenotypes. EMT sensitized cells to fluvastatin, as indicated by a lower IC50 value in SNAIL-overexpressing cells (Fig. 1D). Consistent with our previous finding (11), EMT did not sensitize cells to geranylgeranyltransferase inhibitors, GGTI-298 or GGTI-2133 (Fig. 1D), indicating that fluvastatin-induced cell death in this context is independent from inhibition of protein prenylation. The IC50 for 2-thenoyltrifluoroacetone (2-TTFA) and rotenone, both inhibitors of the ETC, were similar in both vector and SNAIL-overexpressing cell lines (Fig. 1D), indicating that EMT does not sensitize cells to inhibition of the ETC. Instead, inhibition of LLO assembly downstream of dolichol synthesis by tunicamycin phenocopied fluvastatin treatment, as evidenced by a lower IC50 in SNAIL-overexpressing cells (Fig. 1D).
These observations were validated in MCF10A cells overexpressing additional inducers of EMT (SLUG, TWIST, ZEB1), as well as two independent breast cancer cell lines, MDA-MB-231 and MCF-7 (Supplementary Fig. S1A–S1D). Ectopic expression of TWIST or ZEB1 induced EMT in MCF10A cells, as indicated by downregulation of E-cadherin and upregulation of fibronectin or vimentin (Supplementary Fig. S1A). SLUG did not induce EMT in the MCF10A cell system, likely arising from a relatively small increase of SLUG expression in our experiment (Supplementary Fig. S1A) and indicating that a critical level of SLUG expression is needed to induce EMT (30). Consistently, the mesenchymal TWIST- and ZEB1-expressing cells became more sensitive to fluvastatin and tunicamycin compared with the vector control (Supplementary Fig. S1B). The IC50 for geranylgeranyltransferase inhibitor (GGTI) and ETC inhibitors were unaffected by EMT (Supplementary Fig. S1B). Similarly, immunoblotting for E-cadherin and vimentin indicated that MCF-7 cells were epithelial and MDA-MB-231 cells were mesenchymal (Supplementary Fig. S1C). MDA-MB-231 cells were 50-fold more sensitive to both fluvastatin and tunicamycin than MCF-7 cells, which could not be phenocopied by GGTI-298, GGTI-2133, 2-TTFA, or rotenone (Supplementary Fig. S1D). Together, these data indicate that breast cancer cells with mesenchymal phenotypes are more sensitive to inhibition of dolichol synthesis and function, by either fluvastatin or tunicamycin.
As tunicamycin is an inhibitor of the first enzyme downstream of dolichol, leading to LLO synthesis, and elicits ER stress as a result of blocking N-glycosylation (31), we tested whether its effect is an indirect consequence of ER stress. To this end, we treated cells with thapsigargin, a dolichol-independent inducer of ER stress. Treatment with tunicamycin or thapsigargin upregulated ER stress markers ERdj4 and BiP, in both vector and SNAIL-overexpressing cells following 24 hours of treatment (Supplementary Fig. S2A). In contrast, treatment with fluvastatin displayed only a moderate increase the mRNA expression of ERdj4 and BiP in SNAIL-overexpressing MCF10A cells compared with the vector control cells, after up to 72 hours of treatment (Supplementary Fig. S2B). These data indicate that mesenchymal breast cancer cells are sensitized to fluvastatin treatment by inhibition of N-glycosylation, and while tunicamycin also inhibits N-glycosylation, the effect is accompanied by elevated levels of ER stress leading to greater toxicity in normal cells, which has limited its clinical development as an anticancer therapeutic (32). In contrast, fluvastatin produces a milder effect on protein N-glycosylation by dampening the dolichol synthesis pathway further upstream, which alleviates the induction of a strong ER stress response.
Fluvastatin inhibits ER-associated protein N-glycosylation and Golgi-associated N-glycan remodeling
Dolichol is a group of hydrophobic long-chain isoprenoid molecules that constitutes the lipid portion of LLOs, an essential component for protein Asn(N)-glycosylation (10) that occurs on newly synthesized peptides at the consensus sequence NXS/T(X≠P) (Fig. 2A). As dolichol is technically difficult to directly quantify and its only known function is in glycosylation, we validated fluvastatin inhibition of dolichol synthesis by evaluating whether fluvastatin treatment could reduce protein N-glycosylation. To this end, we used SLC3A2 as a molecular biomarker of protein glycosylation. SLC3A2 is a single-pass transmembrane glycoprotein with four N-glycosylation sites, all modified at the ER and remodeled at the Golgi with complex-type N-glycan structures (33). We expressed SLC3A2 in a doxycycline-inducible manner in HeLa cells, where endogenous SLC3A2 has been knocked out. With fluvastatin treatment, doxycycline-induced FLAG-SLC3A2 displayed lower molecular weight immunoblot bands, intermediate in size compared with that of N-glycopeptidase-treated samples, indicating reduced occupancy of N-glycan sites consistent with suppression of dolichol and, in turn, LLO and N-glycosylation (Fig. 2B). To examine site occupancy more directly, three peptides containing N-glycosylation sites at N365, N381, and N424 in FLAG-SLC3A2 were detected and quantified by LC-MS/MS (Supplementary Table S1). With fluvastatin treatment, an increase in the unoccupied fraction of peptides containing Asn365 and Asn381 was observed (Fig. 2C). Interestingly, site occupancy of Asn424 remained unaffected by fluvastatin treatment (Fig. 2C), indicating that N-glycosylation sites on the same protein can differ in sensitivity to reduced dolichol levels. These results and immunoblotting for additional N-glycosylated receptors (Supplementary Fig. S3) are consistent with a partial reduction in N-glycosylation in response to fluvastatin treatment.
Complex type N-glycans are a major subset of post-Golgi structures on mature cell surface glycoproteins. These N-glycans can be further subdivided by N-acetylactosamine branching and fucose (F, Fuc) at the core region and on the peripheral branches (Fig. 2D, highlighted in box). Analysis of SLC3A2 glycopeptides revealed that, in addition to partial inhibition of glycan transfer from LLO to the protein substrates by oligosaccharyltransferase (OST), fluvastatin treatment altered the Golgi dependent profile of residual N-glycans measured in a site-specific manner. Notably, a significant reduction in branched complex N-glycans was observed at N381, N424, and N365 sites (Fig. 2E–G; Supplementary Table S1). Immunoblotting for three additional membrane proteins (EGFR, GP130, and SLC3A2) in vector and SNAIL-overexpressing MCF10A cells revealed that these proteins became under-glycosylated after 48 to 72 hours of fluvastatin treatment in both cell lines to a similar extent (Supplementary Fig. S3A). Treatment with thapsigargin for up to 72 hours did not result in under-glycosylation of EGFR, GP130, or SLC3A2 in either vector or SNAIL-overexpressing cells, although a slight reduction in total glycoprotein levels were observed (Supplementary Fig. S3B). In contrast, these receptors were markedly under-glycosylated with 24 hours of tunicamycin treatment. As cancer cell metastasis requires increased expression of tetra-antennary complex type N-glycans (14–18), we examined whether the transition to EMT was accompanied by increased expression of these glycan structures. To this end, we profiled N-glycans released from membranes of control and SNAIL-overexpressing MCF10A cells treated with and without fluvastatin (Supplementary Table S2). The MCF10A glycome consists of 32% high mannose type N-glycans and 59% complex type N-glycans. The latter can be further subdivided based on branching and fucosylation (F, Fuc) status at the core region and the antennae (Fig. 3A). In MCF10A cells, complex type N-glycans were commonly expressed in the unfucosylated and singly fucosylated (core) forms, with a small amount of doubly fucosylated (core and antennae) structures (Fig. 3A). With induction of EMT, the expression of 12 N-glycans structures were significantly upregulated, all of which belonged to the complex type subgroup; 15 structures were downregulated, including all 9 of the doubly fucosylated complex structures detected (Fig. 3B; Supplementary Table S2). We then examined the effect of fluvastatin treatment on N-glycan profiles and found that 6 of the 12 complex type N-glycans that were upregulated following induction of EMT were specifically inhibited by fluvastatin treatment in SNAIL-overexpressing cells, but not in control cells (Fig. 3C, black arrowheads). Of these, the singly fucosylated triantennary (N2FM3+N3H3) and singly fucosylated tetra-antennary (N2FM3+N4H4) structures, each representing approximately 10% of the total surface glycome, were both upregulated in SNAIL-overexpressing cells, and significantly reduced in response to fluvastatin treatment (Fig. 3D and E).
Our results suggest that the elevated sensitivity of mesenchymal breast cancer cells to fluvastatin is due to the dual effect of fluvastatin on protein N-glycosylation: (i) decreasing the level of N-glycosylation at the ER by inhibiting dolichol synthesis; and (ii) decreasing the complex branching of N-glycans that occurs at the Golgi. Of note, the second effect occurs on those N-glycans that are transferred to proteins in the presence of fluvastatin, the mechanism of which remains to be explored. To test this model, we evaluated whether fluvastatin at concentrations that partially inhibit both NXS/T(X≠P) site occupancy and Golgi N-glycan branching, may display synergy with loss of the branching enzymes MGAT1 or MGAT5. MGAT1 knockout blocks all branching, whereas MGAT5 knockout eliminates the last branch to be added (Supplementary Fig. S4A–S4C). Consistent with this hypothesis, the IC50 for fluvastatin treatment was inversely proportional to levels of complex-type branched N-glycans (MDA-MB-231 wild-type > MGAT5 deficient > MGAT1 deficient cells; Supplementary Fig. S4D). The order of interaction between fluvastatin treatment and these Golgi enzymes is consistent with the known effects of mutating these enzymes in cancer models (19, 20, 34). Taken together, our data suggest that fluvastatin treatment impairs the EMT-driven expression of complex type branched N-glycans on multiple cell surface glycoproteins associated with EMT and metastasis (16–18).
Postsurgical adjuvant fluvastatin treatment delays metastatic outgrowth and prolongs survival
As the transition to a more mesenchymal state is associated with cancer metastasis (35), we evaluated the efficacy of fluvastatin treatment against a postsurgical metastatic breast cancer model in vivo. We used the LM2–4 model of postsurgical advanced metastatic breast cancer, derived from the MDA-MB-231 cell line, which spontaneously metastasizes to the mouse lung (21). After subcutaneous implantation, we allowed LM2–4 xenografts to reach approximately 500 mm3, then excised the primary tumors to mimic first-line surgical treatment (Fig. 4A; Supplementary Fig. S5A; ref. 21). After surgery, mice were randomly assigned to receive PBS (vehicle control) or 50 mg/kg fluvastatin orally, daily (Fig. 4A). Adjuvant fluvastatin treatment significantly prolonged overall survival by >30% in this mouse model (Fig. 4B).
To evaluate the potential antimetastatic activity of fluvastatin, we analyzed lung samples at three time points during the course of the experiment: (i) at time of surgery; (ii) at 8 to 9 days postsurgery; and (iii) at endpoint (Fig. 4A). Metastases to the mouse lung were identified by lesions that stained positive for human EGFR (hEGFR) and confirmed by H&E (Supplementary Fig. S5B). At time of surgery, most mice (7/9) did not have any observable metastases, and 2 of 9 mice had very small lung lesions (Fig. 4C). At 8 to 9 days postsurgery, adjuvant fluvastatin treatment effectively inhibited metastatic outgrowth from disseminated breast cancer cells (Fig. 4D). Finally, at endpoint, fluvastatin treatment decreased the proportion of mice with heavy (>50 colonies/slice) or medium (5–50 colonies/slice) metastatic burden, while increasing the proportion of mice with light metastatic burden (<5 colonies/slice; Fig. 4E–G). Consistently, autopsy at endpoint indicated that the majority of PBS-treated mice reached endpoint due to lung metastases, whereas fluvastatin-treated mice largely reached endpoint from primary tumor regrowth (Supplementary Fig. S5C). We have thus demonstrated, using an in vivo postsurgical metastatic breast cancer model that closely follows the course of human disease, that adjuvant fluvastatin use can delay the development of metastases and prolong overall survival.
Metastatic recurrence is the main cause of breast cancer deaths (1). Since statins are already clinically approved, inexpensive, and have excellent safety profiles that permit their long-term use, these drugs are ideal candidates for repurposing as metastasis prevention agents. Identifying the mechanism of the antimetastatic breast cancer activity of statins, also provides an opportunity to identify novel actionable biomarkers that distinguish patients who will benefit from statin treatment. Here, we show that sensitivity to fluvastatin in the context of breast cancer cell EMT is mediated by inhibition of protein N-glycosylation, providing a mechanistic explanation for previous observations showing statin treatment can block N-glycosylation of specific membrane glycoproteins such as P-gp (36), IGFR (37), EpoR (38), and FLT3 (39). Surprisingly, we also show that fluvastatin exposure impaired Golgi pathway biosynthesis of complex type tri- and tetra-antennary N-glycans associated with breast cancer EMT and metastasis (16–18, 40). The cooperative effects of NXS/T(X≠P) site number and Golgi-associated N-glycan branching is important for cell surface retention and signaling by growth factor receptors (EGF, TGFβ, FGF), and thereby EMT (16). Indeed, we observed cooperative inhibition of fluvastatin and loss of N-glycan branching enzymes MGAT1 or MGAT5 in MDA-MB-231 breast cancer cells. Adjuvant use of fluvastatin delayed breast cancer metastasis and prolonged survival by >30% in a postsurgical model of breast cancer metastasis, supporting the immediate evaluation of fluvastatin in the adjuvant breast cancer space, as well as further development of glycosylation inhibitors to prevent metastatic recurrence in breast cancer (41).
Altered protein N-glycosylation, notably the upregulation of tri- and tetra-antennary complex type glycans, is pivotal to EMT (18, 42, 43) and is a potent modulator of metastatic potential (14–18, 40). High levels of tri- and tetra-antennary complex N-glycans are associated with disease progression and poor prognosis in breast and colon cancer patients (44, 45). Here, we demonstrate that EMT-associated upregulation of complex N-glycans can be targeted by inhibiting the MVA pathway using fluvastatin. The assembly of each N-glycan requires 8 dolichol molecules (46). However, dolichol cannot be efficiently recycled (47) and accumulates with aging (48), indicating that cells must continuously synthesize dolichol. Our results show that fluvastatin treatment can exploit this metabolic vulnerability in metastatic breast cancer cells, reducing protein N-glycosylation on glycoproteins critical to metastasis.
Strong epidemiologic evidence has shown that the risk of postsurgical breast cancer recurrence is reduced by 30% to 60% in patients who are taking statins (2–5). Here, we used a mouse model of postsurgical metastatic breast cancer that closely mimics first-line treatment and disease progression (21), to test the efficacy of fluvastatin when used in the adjuvant setting to prevent metastasis, where long-term use of this safe and inexpensive drug could have considerable clinical benefit. Adjuvant fluvastatin treatment effectively delayed metastasis and prolonged survival by >30% at a daily dose of 50 mg/kg in the mouse, equivalent to a well-tolerated daily dose of 4 mg/kg in human patients (49). Our results support the immediate clinical evaluation of fluvastatin at this well-tolerated dose in the adjuvant setting in patients with breast cancer patients. Moreover, this work reinforces that targeting aberrant tumor metabolism is a feasible strategy for the development of novel and effective anticancer agents.
No disclosures were reported.
R. Yu: Investigation, writing–original draft. J. Longo: Investigation, writing–review and editing. J.E. van Leeuwen: Investigation, writing–review and editing. C. Zhang: Investigation, writing–review and editing. E. Branchard: Investigation, writing–review and editing. M. Elbaz: Investigation, writing–review and editing. D.W. Cescon: Investigation, writing–review and editing. R.R. Drake: Investigation, writing–review and editing. J.W. Dennis: Investigation, writing–review and editing. L.Z. Penn: Supervision, investigation, writing–review and editing.
The authors thank Drs. Robert Kerbel, Roberta Maestro, and Senthil Muthuswamy for providing reagents, Dr. Meegan Larsen for pathology support, Dr. Thomas Kislinger for helpful discussion, Aaliya Tamachi for technical assistance, and all members of the Penn lab for helpful discussion and critical review of the manuscript. This work was supported by funding from the Canada Research Chairs Program (L.Z. Penn), Canadian Institutes of Health Research (L.Z. Penn), Terry Fox Research Institute (L.Z. Penn and D.W. Cescon), CIHR Canada Graduate Scholarship (R. Yu and J. Longo), and Ontario Student Opportunity Trust Fund (J.E. van Leeuwen). This work was also supported by the Office of the Assistant Secretary of Defense for Health Affairs through the Breast Cancer Research Program under Award No. W81XWH-16-1-0068. Opinions, interpretations, conclusions, and recommendations are those of the author and are not necessarily endorsed by the Department of Defense.
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.