Transketolase Regulates the Metabolic Switch to Control Breast Cancer Cell Metastasis via the α-Ketoglutarate Signaling Pathway

Patients with breast cancer have a 5-year survival rate over 90%; however, for patients with distant metastasis, their survival rate decreases to only about 25% because of the lack of effective strategies against breast cancer metastasis and recurrence (1). Tumor cells with altered metabolic program have high requirements of glucose metabolism for rapid proliferation. Despite some studies aiming at elucidating the correlation between aberrant metabolic behavior and tumor progression, how metabolic processes regulate breast cancer cells growth and metastasis is not fully understood.

A number of studies show that oncogenic signaling in cancers drives metabolic reprogramming to generate large amounts of biomass during rapid tumor growth (2). For example, HIF1α elevates the expression of glycolytic enzymes, including aldolase A, phosphoglycerate kinase 1, and pyruvate kinase (3). In addition, a number of studies revealed that genetic defects in TCA cycle enzymes, such as succinate dehydrogenase (SDH) and fumarate hydratase (FH), were also associated with tumor progression (4, 5).

In this study, we used a proteomic approach to identify certain differentially expressed metabolic enzymes involved in tumor progression such as aldolase A (ALDOA), triose phosphate isomerase (TPIS), α-enolase (ENOA), transketolase (TKT), and pyruvate dehydrogenase E1 (ODPB). Among them, TKT is a metabolic enzyme involved in the nonoxidative branch of the pentose phosphate pathway (PPP) and connects PPP with glycolysis. Previous studies revealed that TKT was associated with metastasis of ovarian (6) and esophageal (7) cancers, as well as poor patient survival (6, 7). To date, no study has reported the effect of TKT-regulated metabolic signaling on tumor metastasis in breast cancer.

In this study, we reveal clinical significance and regulatory mechanism of TKT in progression and metastasis of breast cancer via alpha-ketoglutarate (αKG) signaling. TKT plays important roles in regulating dynamic switch of glucose metabolism. The combined therapy based on the new targets TKT or αKG could be developed as an improved therapeutic approach for triple-negative breast cancer (TNBC).

The human breast cancer MDA-MB-231, Hs578T and MCF-7 cells, and mouse breast cancer 4T1 cells were obtained from ATCC. The 4T1 is a highly tumorigenic and invasive cell line capable of metastasizing from the primary mammary gland tumor to liver, lung, lymph nodes, and brain. The highly metastatic cell line MDA-MB-231-IV2-3 was previously established and described (8). All cell lines were cultured in DMEM (Invitrogen) supplemented with 10% FBS (Biological Industries) at 37°C with 5% CO₂. Cell lines were clear of Mycoplasma as determined by the Venor GeM kit (Minerva Biolabs) and were further authenticated in 2017 by Taiwan Bioresource Collection and Research Centre (BCRC) using a short tandem repeat method. For transfection assay, cells were transfected with 20 μmol/L siTKT or 20 μmol/L siRNA control or TKT/pCMV plasmid (1 μg/μL) using Lipofectamine RNAiMAX transfection reagent (Thermo Fisher Scientific).

Cell samples were lysed in lysis buffer containing 7 mol/L urea, 2 mol/L thiourea, 4% w/v CHAPS, 10 mmol/L Tris-HCl pH 8.3, and 1 mmol/L EDTA. Protein lysates were extracted, sonicated, and centrifuged and the protein concentration was determined using Coomassie Protein Assay Reagent (Bio-Rad).

The protein profiles of tumor tissues with 0.5, 1, and 2 cm in size were analyzed using 2D differential gel electrophoresis (DIGE). Protein samples were labeled with cyanine dyes Cy2, Cy3, and Cy5, and all procedures have been described previously (9, 10). The Cy-Dye–labeled 2-DE gels were visualized according to the previous report (10). For protein identification, the peptide mixture was loaded onto a MALDI plate and samples were analyzed using an Autoflex III mass spectrometer (Bruker Daltonics) and parameters were described according to the previous report (10).

Cells were lysed in the lysis buffer containing 7 mol/L urea, 2 mol/L thiourea, 4% w/v CHAPS, 10 mmol/L Tris-HCl (pH 8.3), 1 mmol/L EDTA, and phosphatase and protease inhibitors (Roche). Protein lysates were sonicated and centrifuged and the protein concentration was determined using protein assay kit (Thermo Fisher Scientific). The defined amount of final lysates was resolved in 8%–12% SDS-polyacrylamide gels, transferred onto polyvinylidene difluoride membrane and probed with appropriate antibodies. Antibodies include rabbit polyclonal anti-LDHA (GTX101416, Genetex), rabbit polyclonal anti–αKG dehydrogenase (clone C2C3, GTX105124, Genetex), rabbit polyclonal anti-SDH (GTX113833, Genetex), rabbit polyclonal anti-FH (clone N2C2, GTX110128, Genetex), rabbit polyclonal anti-MDM2 (GTX100531, Genetex), mouse monoclonal anti-TKT (clone 7H1AA1, ab112997, Abcam), and mouse monoclonal anti-PHD2 (clone 366G/76/3, Thermo Fisher Scientific). Mouse monoclonal anti-β-actin (clone SPM161, Santa Cruz Biotechnology) was used as the internal control, and protein expression levels were visualized with the Enhanced Chemiluminescence Detection Kit (Pierce Boston Technology) and exposed to X-ray film. All experiments were repeated three times.

Paraffin-embedded matched normal, primary tumor, and lymph node metastatic tissue sections of breast cancer specimens (n = 11) were provided by Dr. Wen-Hung Kuo, National Taiwan University Hospital (Taipei, Taiwan). Other samples were from commercial tissue arrays (US Biomax; SuperBioChips), including 19 normal, 90 tumors, and 50 lymph node metastatic tissues. The slides were stained with mouse monoclonal anti-TKT antibody (clone 7H1AA1, ab112997; Abcam) using an automatic slide stainer BenchMark XT (Ventana Medical Systems). The staining intensities were evaluated and quantified by one pathologist (Pathology Core Lab, National Health Research Institutes) and two independent investigators. The IHC scores of TKT for each specimen were graded as follows: no expression, weak (+); moderate (++); and strong (+++).The expression levels of TKT in tumor cells were quantified as a percentage. Paraffin-embedded sections of tumor cells with TKTL1 overexpression (Origene, RG205218) were stained with mouse monoclonal anti-TKT antibody (1 mg/mL, 1:75 dilution; clone 7H1AA1, ab112997; Abcam) or rabbit polyclonal anti-TKTL1 antibody (1 mg/mL, 1:75 dilution; clone N1C1, GTX109459; Genetex). We first used the D'Agostino and Pearson omnibus normality test to reveal that the quantitative results of IHC TKT expression were not Gaussian distribution (P = 0.0015). Thus, we used nonparametric Mann–Whitney test to analyze the quantitative results.

Cell proliferation was detected using CellTiter 96 Aqueous One Solution Cell Proliferation Assay (Promega). Assay was performed according to manufacturer's protocol. A total of 1.4 × 10⁴ cells were cultured in a 24-well plate and incubated for different times. CellTiter 96 Aqueous One Solution reagent was added and incubated for 1 hour at 37°C. The quantity of formazan product, proportional to living cell numbers, was measured at 490 nm using 96-well plate reader. Each experiment was performed in triplicate and the shown data were mean ± SD.

MDA-MB-231 and Hs578T cells were treated with 20 μmol/L siTKT or 1 mmol/L αKG, or TKT/pCMV plasmid (1 μg/μL). After 48 hours, these cells (1 × 10⁵ cells) were seeded on Boyden chamber, incubated for 8 hours, and then stained with 0.5% crystal violet dye. Cell invasion and migration were assayed in 8-μm Falcon Cell Culture Inserts with or without Matrigel (BD Biosciences), respectively. All experiments were performed in triplicate.

MDA-MB-231 or MCF-7 cells at densities of 1 × 10⁵ cells were seeded in 6-well plate containing top layer of 0.4% agarose and bottom layer of 0.6% agarose medium. The treatment group was transfected with 20 μmol/L of siTKT for 48 hours. After one month, colonies were stained with p-Iodonitrotetrazolium violet (1 mg/mL) for 48 hours and then counted. Data represent mean ± SD and the experiment was performed in triplicate.

To study the effects of αKG on tumor progression, MDA-MB-231 cells (1 × 10⁶) resuspended in 100-μL PBS were implanted orthotopically in 4th mammary fat pads of 8-week-old female CB17-SCID mice (8). After implantation of MDA-MB-231 cells for 24 hours, αKG reagent was intraperitoneally injected 3 times a week until 3 months. αKG dissolved in PBS was used for injection of 10 mg/kg each time. The tumor volume was calculated by the formula: tumor volume (cm³) = [length (cm) × width (cm)² × 0.5]. To study the effects of TKT knockdown on tumor metastasis, MDA-MB-231-IV2-3 cells were treated with 20 μmol/L siTKT. After 48 hours, MDA-MB-231-IV2-3 cells (1 × 10⁶) resuspended in 100-μL PBS were injected per mouse intravenously via tail veins into 6- to 8-week-old female CB17-SCID mice (BioLASCO). Tumor growth and metastasis to individual organs were observed using live animal bioluminescence imaging (BLI; Caliper IVIS system, PerkinElmer). Tumor volume and weight were also measured at the end point. Cell metastases were quantified by BLI signals of each mouse at the end point. Animal experiments were approved by the Institutional Animal Care and Use Committee (IACUC).

MDA-MB-231 cells were, respectively, transfected with two independent GFP-TKT/pCMV plasmids (Origene, NM001064). After 48 hours, these cells were selected by flow cytometry and transfection efficiency was confirmed by Western blot analysis. Stable shTKT cells were orthotopically injected at 1 × 10⁶ cells per mouse into 4th mammary fat pads of CB17-SCID mice (n = 7) and tumor volumes were recorded once a week during the 70 days period. Tumor volume = 4/3πR³, R = [length (cm) + width (cm)]/2. Animal experiments were approved by IACUC. n = 7.

MDA-MB-231 cells were transfected with 20 μmol/L siTKT. After 48 hours, these cells were lysed with 0.1 mol/L Tris-HC1 buffer (pH 7.6), centrifuged, and the supernatant was collected (11). Supernatant (50 μL) was mixed with 200 μL reaction mixture including 14.4 mmol/L ribose-5-phosphate, 190 μmol/L NADH, 380 μmol/L TP, >250 U/L glycerol-3-phosphate dehydrogenase, and >6,500 U/L triose phosphate isomerase (12).

Enzyme activity was detected at 340 nm. One unit of enzyme activity indicates the amount of enzyme catalyzing the oxidation of 1 μmol of NADH per minute.

Oxygen consumption rate (OCR) is an indicator of mitochondrial oxidation and extracellular acidification rate (ECAR) is an indicator of lactate production that is equated to the glycolytic rate. OCR and ECAR were detected by XFe24 extracellular flux analyzer (Seahorse Bioscience). MDA-MB-231 cells (7 × 10⁴ cells) were cultured in X24 culture plate (Seahorse Bioscience). OCR and ECAR were measured in XF base medium (Seahorse Bioscience). OCR was analyzed over time following injection of 1 μmol/L oligomycin, 2 μmol/L carbonyl cyanide-p-trifluoromethoxyphenylhydrazone (FCCP), and 0.5 μmol/L rotenone/antimycin. ECAR was measured over time following injection of 10 mmol/L glucose, 1 μmol/L oligomycin, and 50 mmol/L 2-deoxyglucose (2-DG). For ECAR, glucose (10 mmol/L), oligomycin (1 μmol/L), and 2-DG (50 mmol/L) were used to estimate glycolytic metabolism. Glucose treatment could increase glycolytic metabolism in cells. 2-DG, a synthetic glucose analogue, acted as a competitor for glucose and interfered with glucose metabolism. For OCR, oligomycin (1 μmol/L), carbonyl cyanide-4-(trifluoromethoxy)phenylhydrazone (FCCP; 1 μmol/L), and rotenone/antimycin (0.5 μmol/L) were used to estimate oxidative respiration. For mitochondrial respiration, oligomycin treatment inhibited ATP synthase in mitochondria. FCCP, a proton ionophore in mitochondria, transports protons across cell membranes to disrupt ATP synthesis. Finally, rotenone and antimycin are inhibitors for electric transport chain in mitochondria.

Kaplan–Meier method (log-rank test) was used to analyze survival data. Data were presented as mean ± SD. Student t test was used to compare the differences between two experimental groups and one-way ANOVA was used to compare the differences among multiple groups using Tukey test in GraphPad. χ² test was used to analyze the correlation between TKT levels and clinical factors; *, P < 0.05; **, P < 0.01; ***, P < 0.001. OCR and ECAR data were calculated by paired t test.

Using a proteomic approach and examining tumors of varying sizes, we attempted to identify differentially expressed proteins associated with breast cancer progression. We used syngeneic orthotopic implantation of 4T1 cells in BALB/c mice, and tumors with 0.5, 1, and 2 cm in size were collected for further proteomic analyses. The protein profiles from the tumor with 0.5 cm in size were compared with those tumors with 1 and 2 cm in size by two-dimensional protein gel analysis (Supplementary Fig. S1A–S1C). After spot detection and quantification from the two-dimensional gel images, a total of 21 differentially expressed proteins (P < 0.05) with 1.5-fold changes were chosen for further identification (Supplementary Fig. S1A–S1C) by using MALDI-TOF-MS and MASCOT database (Supplementary Table S1). Three proteins related to glycolysis were upregulated in the bigger tumors; they included ALDOA, TPIS, and ENOA. Other metabolic enzymes included the upregulated TKT involved in PPP and the downregulated ODPB involved in pyruvate oxidation (Supplementary Table S1; Supplementary Fig. S1D). Tumors with 1 and 2 cm in size had 1.5- and 2-fold, respectively, increased expression of TKT when compared with the 0.5-cm tumor (Supplementary Table S1; Supplementary Fig. S1D).

We analyzed TKT expression in normal and tumor tissues according to gene expression arrays from Oncomine database (Bild data). As compared with normal tissues, TKT displayed significantly higher expression in tumor tissues (Fig. 1A, nonparametric Mann–Whitney test, P = 0.03). We also found that the levels of TKT in TNBC patients were significantly higher than those in non-TNBC patients (Fig. 1B, P < 0.001). Kaplan–Meier survival curve (log-rank test) from Curtis 5-year overall survival data showed that patients with higher TKT levels had poorer 5-year survival than those with lower TKT levels (n = 637, Fig. 1C; P = 0.019, χ²= 5.502, HR = 1.3298). The similar result is also observed in different clinical database (n = 158; Fig. 1D; P = 0.003, χ² =8.7476, and HR = 2.3131), suggesting that TKT has a prognostic potential. TNBC is the breast cancer subtype with the poorest outcome; however, very few metabolic enzymes as prognostic indicators for patients with TNBC are known. The role of TKT in patients with TNBC has not been reported; thus, we further analyzed the correlation between TKT expression levels and patients with TNBC 5-year overall survival. Among the 637 cases, there were a total of 106 patients with TNBC. Our analysis showed that patients with TNBC with higher TKT levels had poorer 5-year overall survival than those with lower TKT levels (n = 106; Fig. 1E; P = 0.0006, χ² = 11.7166, HR = 2.3758), showing that TKT might have a prognostic potential in patients with TNBC, and it could play a role in TNBC progression. The clinicopathologic features of TKT in patients with breast cancer from Curtis data showed that TKT levels were significantly associated with some clinical factors, including stage, age, grade, type, TNBC, and tumor size (Supplementary Table S2). We also analyzed TKT expression in normal, primary tumor, and lymph node metastatic tissues by using IHC. First, we checked whether TKT antibody used in the IHC staining cross-reacted with TKTL-1. To address this, we used TKTL1/pCMV plasmid to overexpress TKTL1 in MDA-MB-231 cells. The overexpression efficiency was verified (Supplementary Fig. S2A). The paraffin-embedded sections of tumor cells with TKTL1 overexpression were stained with anti-TKT or anti-TKTL1 antibody. Our results displayed high staining intensity of TKTL1 in tumor cells overexpressing TKTL1 using anti-TKTL1, whereas staining intensity using TKT antibody was insignificant (Supplementary Fig. S2B). These results suggest that TKT antibody used in the IHC staining does not cross-react with TKTL1.

The staining intensity of TKT was evaluated and quantified ranging from no expression to the highest expression by a pathologist and two independent investigators of our team. As summarized in Supplementary Fig. S2C, a high percentage of normal tissues displayed insignificant TKT intensities (60%) or low intensities of TKT (30%) when compared with those of tumor tissues (P < 0.001). Moreover, metastatic lymph node tissues displayed a higher percentage of high intensities of TKT (56%) when compared with primary tumor (25%, P < 0.001). The representative staining photographs are shown in Fig. 1F. The percentage of TKT expression in tumor cells, not including stroma cells, from primary tumor and lymph node metastatic tissue sections was further quantified. Metastatic lymph node tissues displayed a higher percentage of TKT expression in tumor cells when compared with the primary tumor (Supplementary Fig. S2D; P < 0.001). These results showed that TKT expression levels were the highest in lymph node metastases, suggesting that a possible correlation of TKT levels with progression of metastasis in breast cancer.

To further elucidate the functional role of TKT, we manipulated TKT expression by siRNA depletion of TKT in MDA-MB-231 and Hs578T TNBC cells (Fig. 2A). The downregulation of TKT in MDA-MB-231 cells resulted in significantly decreased cell proliferation (Fig. 2B–D). This phenomenon was also observed in Hs578T cells (Fig. 2E–G). The inhibition by TKT knockdown in both cell lines was significantly rescued by TKT/pCMV overexpression (Fig. 2B–G). Cell migration and invasion were carried out by transwell Boyden chamber assays. Downregulation of TKT led to a significant inhibition of invasion (Fig. 2H) and migration (Fig. 2I) of MDA-MB-231 and Hs578T cells, whereas the inhibitory effects were almost completely rescued by TKT overexpression. MDA-MB-231 cells with the inhibited TKT expression displayed reduced ability of colony formation (Fig. 2J). TKT knockdown increased the percentage of cells in the G₂–M phase in MDA-MB-231 and Hs578T cells (Fig. 2K). Taken together, these data suggested that the depletion of TKT impaired tumor cell growth and metastasis-related abilities.

To evaluate whether the depletion of TKT suppressed cancer cell metastasis in vivo, we used tail vein injection of the highly invasive MDA-MB-231-IV2-3 cells (1 × 10⁶ cells) in CB17-SCID mice (n = 8). The highly metastatic MDA-MB-231-IV2-3 sublines derived from the MDA-MB-231 parental line were established and described previously (8). The MDA-MB-231-IV2-3 cells exhibited dramatically higher invasiveness than the MDA-MB-231 parental cells in vitro and they also exhibited more aggressive lung and lymph node metastasis in vivo (8). The data from tail vein injection model showed that knockdown of TKT resulted in greatly decreased lung metastasis of the MDA-MB-231-IV2-3 cells [Fig. 3A (P = 0.005) and B (P = 0.002)] by BLI as also reflected in hematoxylin and eosin staining (Fig. 3C). These findings indicated that knockdown of TKT inhibited lung metastasis of the highly invasive breast cancer cells (Supplementary Fig. S3A–S3F).

To further assess whether decreased lung metastasis by the depletion of TKT resulted from decreased targeting of the tumor cells to lung, the cells transfected with the control or TKT siRNA were injected into CB17-SCID mice through tail vein (n = 8) and after 24 hours, the lung was perfused with PBS to flush out intravascular tumor cells and subsequently the expression levels of human GAPDH reflecting the injected cells in lung tissues were measured. BLI analysis exhibited about equivalent signals in the lungs of siCon or siTKT-transfected cells 30 minutes after injection (Fig. 3D, P = 0.294). The qPCR data confirmed the result (Fig. 3E, P = 0.222). To confirm the inhibitory effects of transient TKT knockdown on tumor growth, two different knockdown stable lines, MDA-MB-231-shTKT1 and MDA-MB-231-shTKT2, as well as MDA-MB-231-shNC line, were established, and each (1 × 10⁶ cells) were implanted orthotopically into the 4th mammary fat pad of CB17-SCID mouse (n = 7). The knockdown efficiency of shTKT was verified (Fig. 3F) and the result showed that tumor sizes in both TKT knockdown groups were significantly smaller than those in the control group (Fig. 3G and H; Supplementary Fig. S3G–S3I). These findings indicated that knockdown of TKT did not inhibit lung targeting (Supplementary Fig. S4A–S4D), but inhibited the subsequent lung colonization ability of the tumor cells.

Recent reports indicated the involvement of Warburg effect in tumor metastasis and suggested that the molecules participating in metabolic modulation were potential targets for antimetastasis therapy (13). To address TKT-regulated metabolic pathways in breast cancer cells, we manipulated TKT expression by siRNA treatment of MDA-MB-231 cells for 48 hours and then cell lysates were harvested for identifying altered metabolites by LC-MS/MS (Waters Corporation). The differentially expressed metabolites in siTKT-treated cells were identified when comparing with the siRNA control cells. Knockdown of TKT increased some TCA-cycle intermediates including αKG (Fig. 4A) and malate, while decreased succinate and fumurate (P < 0.05). Reports indicated that the alternation of metabolites in the TCA cycle was associated with tumor formation (4). For example, succinate and fumarate accumulated in the mitochondria leaked out to the cytosol because of inactivation of the tumor suppressors SDH and FH, resulting in promoting cancer formation (14). Currently, the potential role of αKG and the relationship between TKT and αKG in TNBC are still unclear. Our findings that TKT might play an important role in metastasis and its knockdown led to increased αKG prompted us to further investigate the potential effect of αKG in oncogenic behavior of cancer cells.

We further found that TKT overexpression attenuated αKG levels (Fig. 4B, P < 0.001), which was consistent with the result from TKT knockdown. The physiologic concentration of αKG in healthy brain tissues ranges from 1 to 3 mmol/L, whereas its concentration is decreased to 100 to 300 μmol/L in gliomas (15). IDH1-mutated tumor cells exhibited decreased αKG, leading to increased HIF1α levels (16). The similar results were observed in αKG derivatives treatments in IDH1-mutated gliomas (17) or SDH-deficient tumor cells (18). Despite its tumor suppressor role of artificial αKG derivative in cancers, many studies revealed that non-αKG derivative could attenuate cell proliferation of colon cancer (19) and reduces the levels of VEGF and erythropoietin through decreasing HIF1α, thereby, inhibiting angiogenesis ability of the Hep3B hepatoma cells (20). These findings suggested the potential tumor suppressing role of αKG. Furthermore, G-protein–coupled receptor GPR99 was reported to function as a receptor for the TCA cycle intermediate αKG (21). Although, previous studies indicated that αKG–dependent dioxygenases signaling pathways functioned as tumor suppressors (22), the regulatory role of αKG in breast cancer is unclear. Treatment of αKG resulted in significantly decreased MDA-MB-231 cell growth when compared with the control (Fig. 4C). Furthermore, treatment of αKG led to a significant inhibition of cell invasion (Fig. 4D) and migration (Fig. 4E). MCF-7 cells with the inhibited TKT expression displayed reduced ability of colony formation (Supplementary Fig. S4E). TKT overexpression promoted cell proliferation in MDA-MB-231 (Fig. 4F–H) and Hs578T (Supplementary Fig. S4F–S4H) cells, whereas its effect was substantially reversed by αKG treatment. These findings indicate that αKG can impair metastatic-related abilities of breast cancer cells. We further verified that the promotion of TKT on invasion (Fig. 4I) and migration (Fig. 4J) in MDA-MB-231 and Hs578T cells was substantially reversed by αKG treatment, suggesting that TKT regulated invasion and migration of tumor cells via αKG signaling. In this study, we observed the cellular levels of αKG were increased after the treatment of αKG (Supplementary Fig. S5A and S5B).

We next assessed the effect of this metabolic pathway on tumor growth and metastasis using a mouse model. A total of 1 × 10⁶ MDA-MB-213 cells were implanted orthotopically into mammary fat pads of CB17-SCID mice (n = 10). One day after implantation, intraperitoneal αKG (10 mg/kg) administration was started three times a week for 3 months. BLI data revealed that αKG treatment led to a significant reduction of primary tumor growth (Fig. 4K, P < 0.001). There were significant differences in the weights (Fig. 4L, P = 0.024) and sizes (Fig. 4M, P = 0.004) of primary tumors between control and the αKG–treated groups after 3 months. Individual organ metastases were also examined, and we found that αKG treatment significantly diminished lung and lymph node metastases (Fig. 4N, P < 0.05). Overall, our data for the first time demonstrated that TKT-mediated αKG signaling suppressed growth and metastases of breast cancer.

To further explore TKT-regulated downstream pathways in breast cancer metastasis, the effects of TKT on the αKG and TCA-cycle enzymes were examined. Previous studies indicated that accumulation of αKG enhanced the activity of PHD and subsequent destabilization of its downstream target HIF1α (23). To assess the relationship between TKT and HIF1α in MDA-MB-231 cells, the impact of TKT on PHD2 was investigated. Results revealed that downregulation of TKT enhanced PHD2 expression (Fig. 5A) and this phenomenon was also observed in the αKG–treated cells (Fig. 5B). Moreover, knockdown of TKT reduced HIF1α expression (Fig. 5A), suggesting that TKT affected HIF1α expression via the PHD2 signaling pathway. HIF1α has been reported to be associated with tumor metastasis (24) and is known to be a transcription factor regulating the expression of LDHA (25). Other studies revealed that knockdown of LDHA inhibited breast cancer metastasis (26). Currently, the relationship between TKT and LDHA is not known; thus, we further assess the effect of TKT on LDHA expression. Our data showed that knockdown of TKT inhibited LDHA expression (Fig. 5A) and this phenomenon was also observed in αKG–treated cells (Fig. 5B). These results suggested that TKT decreased LDHA expression and promoted HIF1α degradation through the αKG signaling pathway, leading to the inhibition of breast cancer metastasis. Our data suggest that a regulatory network of those metabolites and their corresponding catalyzing enzymes are involved in the regulation of breast cancer metastasis.

Previous study indicates that L-2HG dehydrogenase (L2HGDH) and D-2HG dehydrogenase (D2HGDH) prevent oncometabolites L-2HG and D-2HG from accumulating in normal cells, respectively, by converting them back to αKG (22). We have found that TKT depletion enhanced the levels of L2HGDH and D2HGDH (Fig. 5A). Overall, these results indicate that TKT depletion enhances L2HGDH and D2HGDH levels, resulting in the increase of αKG and PHD2 levels and thereby promoting HIF1α degradation.

Our data showed that knockdown of TKT decreased the expression levels of metabolites succinate and fumarate (Fig. 5C). Previous studies indicated that the inactivation mutations in SDH and FH led to abnormal accumulation of metabolites succinate and fumarate in TCA cycle, which in turn inhibited PHD and increased HIF1α in tumors (4, 5). The correlation between TKT, SDH, and FH in breast cancer is still unclear; thus, we investigated the effects of TKT knockdown on the expression levels of SDH and FH. We found that knockdown of TKT increased the levels of SDH and FH (Fig. 5A), leading to decreased levels of succinate and fumarate and thus stabilizing the PHD2-regulated signaling pathway.

Previous studies reported αKG–dependent dioxygenases signaling pathways functioning as tumor suppressors (22). In addition, SDH and FH have been reported to be targets of αKG–dependent dioxygenases, including JmjC domain-containing histone demethylase (KDMs) and DNA demethylases (27). These studies suggest that TKT may control transcriptional regulation of SDH and FH via αKG–dependent dioxygenases. To elucidate the potential underlying mechanism, we detected the effects of TKT depletion or αKG treatment on RNA levels of SDH and FH. Our results showed that TKT depletion (Fig. 5D, P < 0.001) or αKG treatment (Fig. 5E, P < 0.01) indeed increased RNA levels of SDH and FH, suggesting regulation at the transcriptional level. Overall, the likely regulatory mechanism of TKT via αKG signaling in breast cancer metastasis is depicted in Fig. 5F.

Tumor cells predominantly metabolize glucose through glycolysis instead of oxidative phosphorylation in TCA cycle to rapidly produce ATPs and nucleic acid building stones for supporting their high rate of growth (28). The effect of TKT on metabolic activities in cancers was unclear; thus, we examined the relationship among glycolysis, mitochondrial metabolism, and TKT signaling. The knockdown efficiency of TKT in MDA-MB-231 cells was initially estimated (Fig. 6A). TKT knockdown (Fig. 6B, P < 0.001) or αKG treatment (Fig. 6C, P < 0.001) exhibited decreased ECAR. TKT knockdown (Fig. 6D, P < 0.001) or αKG treatment (Fig. 6E, P < 0.001) elevated OCR. These results demonstrated that reduced TKT led to switch of glucose metabolism from glycolysis to mitochondrial respiration via the αKG signaling pathway.

To further confirm whether knockdown of TKT drove the switch of glucose metabolism from glycolysis to TCA cycle, we used mass spectrometry to measure expression levels of metabolites in glycolysis and TCA cycle. Reduction of TKT diminished the levels of glycolytic metabolites including glucose-6-phosphate (G6P), pyruvate, and lactic acid, while increased the TCA-cycle metabolites including αKG and malate (Supplementary Fig. S5A). We treated the cancer cells with αKG and observed a similar result like TKT knockdown (Supplementary Fig. S5B), suggesting that reduction of TKT drove the switch of glucose metabolism from glycolysis to mitochondrial metabolism at least in part through the αKG signaling pathway.

To further verify this, the effect of decreased TKT on the expression levels of metabolic enzymes in TCA cycle was evaluated. We found that the depletion of TKT resulted in increased expression levels of metabolic enzymes in TCA cycle including aconitase, αKG dehydrogenase, SDH, FH, and malate dehydrogenase (Supplementary Fig. S6A). The similar result was obtained in αKG–treated cells (Supplementary Fig. S6B). In contrast, the depletion of TKT resulted in decreased levels of glycolytic enzymes including PKM2, HK, and PFK (Supplementary Fig. S6C), and the similar results were also observed in αKG–treated cells (Supplementary Fig. S6C). Taken together, these results indicate that reduced TKT leads to the alteration of glucose metabolism by switching it from glycolytic to mitochondrial metabolism via the elevation of metabolic enzymes in TCA cycle through the αKG signaling pathway. As tumor cells depend on glycolysis for their rapid growth, inhibition of TKT or addition of αKG could be used as a modality for developing cancer therapeutics not only for breast cancer including triple-negative breast cancer as shown in this study, but for other types of cancer as well.

Docetaxel and doxorubicin are commonly used drugs for TNBC, but their efficiencies are limited as a result of the development of drug resistance. Oxythiamine inhibits TKT and thus could lead to downregulation of glycolysis, not targeted by the above two drugs. Thus combinatory treatment of oxythiamine together with the two drugs may enhance the killing effect of cancer cells. Oxythiamine, an antimetabolite thiamine analogue, induces cell apoptosis and suppresses tumor cell growth in cancers by targeting TKT (29, 30). Although some studies indicate that oxythiamine can suppress tumor progression, the effects of oxythiamine in breast cancer are unclear. In this study, we first assessed the effect of oxythiamine on TKT activity according to previous study (12). Tumor cells were treated with 5 mmol/L oxythiamine for 48 hours. Our results revealed that TKT activity was significantly reduced by oxythiamine treatment (Fig. 7A, P < 0.01). In addition, we found oxythiamine treatment elevated the levels of αKG in MDA-MB-231 (Fig. 7B, P < 0.001) and Hs578T (Fig. 7C, P < 0.001) cells as expected, suggesting that oxythiamine suppressed tumor growth could in part through the αKG signaling pathway. Then, we analyzed whether oxythiamine treatment affected growth of breast normal cells. The results showed that cell viabilities of nontumorigenic human breast epithelial cell line H184 for 24 (P = 0.16), 48 (P = 0.08), and 72 hours (P = 0.07) were not significantly decreased by 5 mmol/L oxythiamine treatment when compared with those without oxythiamine treatment (Fig. 7D), indicating there was no significant side effects of oxythiamine in human breast normal cells. We observed that docetaxel or doxorubicin treatment increased αKG levels (Fig. 7E, P < 0.001). Moreover, previous studies reported that docetaxel or doxorubicin treatment attenuated HIF1α levels (31, 32), further supporting our findings that TKT affects HIF1α expression via αKG signaling. Thus, we tested the inhibitory effects of oxythiamine in combination with docetaxel and/or doxorubicin on cell proliferation. We treated TNBC cell lines MDA-MB-231 (Fig. 7F–H) and Hs578T (Fig. 7I–K) with 5 mmol/L oxythiamine, 1 μmol/L docetaxel, 1 μmol/L doxorubicin and oxythiamine in combination with docetaxel and/or doxorubicin for 24, 48, and 72 hours. Treatment of oxythiamine had significant inhibitory effects for 24 (Fig. 7F and I) and 48 hours (Fig. 7G and J) in both cell lines. Although treatment of docetaxel or doxorubicin had inhibitory effects of TNBC cells, the killing effects of oxythiamine combining with docetaxel or doxorubicin could be strengthened in TNBC cells, In addition, combining of the three drugs had maximum killing effects (>90% decrease) for 72 hours in both TNBC cell lines (Fig. 7H and K). These findings indicate that oxythiamine could enhance the two-drug sensitivities of TNBC cells.

Increasing evidence suggests that some pivotal genes, including HIF1α, are able to regulate certain enzymes to induce metabolic reprogramming in cancers. HIF1 has been reported to induce glycolytic enzymes, including aldolase A, phosphoglycerate kinase 1, and pyruvate kinase (3). HIF1α regulates dynamic switch from oxidative to glycolytic metabolism by activating glucose transporters and glycolytic enzymes (33). Certain metabolic enzymes involved in glucose transport, glycolysis and lipid metabolism are targets of HIF1α (34). In our study, we found that TKT depletion promoted HIF1α degradation via αKG signaling. These results suggest that TKT-mediated signaling pathways may collaborate to regulate dynamic switch of glucose metabolism. Xu and colleagues (11) reported that TKT reduced oxidative stress and played important roles in glycolysis and glutathione synthesis in hepatocellular carcinoma (HCC) cells. TKT knockdown attenuated NADPH production and led to the increase of reactive oxygen species (ROS; ref. 11). TKT knockdown decreased glucose flux, and purine metabolites including AMP, ADP, ATP, and GTP (11). Together, these results provide evidence that TKT may play an important role in metabolic reprogramming in tumors.

The emerging evidence demonstrates that several TCA cycle enzymes are tumor suppressors, such as SDH and FH, and their genetic defects are associated with tumorigenesis. The inactivation mutations in SDH and FH leads to abnormal accumulation of metabolites succinate and fumarate in TCA cycle, and the subsequent inhibition of PHD and enhancement of HIF1α pathways in tumors (4, 5). Here, we have demonstrated that reduction of TKT augments levels of SDH, FH, and PHD2, but decreased levels of HIF1α. In addition, levels of oncometabolites succinate and fumarate are significantly reduced by TKT knockdown, which is likely due to increased levels of SDH and FH, which in turn affects PHD2 stabilization and HIF1α degradation. HIF1α is a transcription factor regulating the expression of LDHA (25) and its knockdown inhibits breast cancer metastasis (26). We have also noticed that knockdown of TKT decreases levels of LDHA, suggesting that reduction of TKT resulted in decreased HIF1α and LDHA via elevated levels of SDH and FH, leading to the inhibition of tumor metastasis.

Previous reports indicate that a glycolytic enzyme pyruvate kinase M2 (PKM2) is a transcriptional coactivator for HIF1, amplifying HIF1 activity via a positive feedback regulation, and thereby promoting cancer progression (35). To date, the underlying mechanism of TKT-mediated regulation of PKM2 via αKG signaling is unclear. We found that TKT depletion or αKG treatment reduced PKM2 levels (Supplementary Fig. S6C) and promoted HIF1α degradation. A significant positive correlation existed between TKT and PKM2 (r = 0.4635, P < 0.0001, Supplementary Fig. S6D). Patients with breast cancer (N = 3951, P < 0.001, Supplementary Fig. S6E) including patients with TNBC (N = 255, P = 0.045, Supplementary Fig. S6F) with higher TKT and PKM2 levels had poorer recurrence-free survival (RFS) than those with lower TKT and PKM2. We also observed that patients with breast cancer (N = 3951, P < 0.001, Supplementary Fig. S6G) including patients with TNBC (N = 255, P = 0.0049, Supplementary Fig. S6H) with higher TKT, PKM2, and HIF1α levels had poorer RFS than those lower TKT, PKM2 and HIF1α. On the other hand, a study indicated that p53 induced tumor suppressor MDM2 E3-ubiquitin-mediated degradation of HIF1α (36). To date, the underlying mechanism of TKT-mediated regulation of MDM2 via αKG signaling is not known. We found that TKT depletion or αKG treatment enhanced MDM2 levels (Supplementary Fig. S6I) and promoted HIF1α degradation. A significant negative correlation existed between TKT and MDM2 (r = −0.2618, P < 0.0001, Supplementary Fig. S6J). Patients with breast cancer with higher MDM2 levels had better RFS than those with lower MDM2 (N = 3951, P = 0.0019, Supplementary Fig. S6K). As both PKM2 and MDM2 could regulate HIF1α stability, our results suggest that aside from the TKT/αKG–mediated regulation of PHD2 and HIF1α degradation, PKM2 and MDM2 could also play a role in TKT-mediated control of HIF1α stability.

αKG functions as a cosubstrate for Fe (II)/αKG–dependent dioxygenases, including KDMs and the TET (ten-eleven translocation) family of DNA hydroxylases (27). They catalyze hydroxylation in diverse substrates including proteins, alkylated DNA/RNA and 5-methylcytosine (5mC) of genomic DNA (27). TET family of DNA hydroxylases catalyzes a three-step oxidation reaction to convert 5mC to 5-carboxylcytosine (5caC) and subsequent decarboxylation of 5caC, leading to DNA demethylation (27). SDH and FH have been reported to be the targets of αKG–dependent dioxygenases, including KDMs and DNA demethylases (27). Our results showed that TKT depletion or αKG treatment increased RNA levels of SDH and FH. Together, these studies suggest that TKT may control transcription of SDH and FH via αKG–dependent dioxygenases signaling.

TKT inhibitor oxythiamine had been reported to have anticancer activity (29, 30). For example, oxythiamine in combination with sorafenib had enhanced effects on HCC cell growth by in vivo assay (11). Despite its potential therapeutic development, at present, the targeted therapy of TKT against TNBC cells has not been reported. Our results showed that the combinations of oxythiamine with docetaxel and doxorubicin had maximum inhibitory effects in TNBC cells, suggesting combinatory drug treatment as a novel therapy against TNBC. Our study for the first time revealed that oxythiamine treatment elevated the levels of αKG in TNBC cells, suggesting that oxythiamine suppressed tumor cell growth via αKG signaling pathway. Together, it is feasible to develop a combinatory drug treatment with the conventional therapeutic drugs to improve treatment benefits for TNBC.

No potential conflicts of interest were disclosed.

Conception and design: L.-H. Wang, C.-W. Tseng, H.-L. Chan, and K.-J. Chang

Development of methodology: L.-H. Wang, C.-W. Tseng, S.-H. Chan, and H.-L. Chan

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C.-W. Tseng, W.-H. Kuo, S.-H. Chan, and H.-L. Chan

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): L.-H. Wang, C.-W. Tseng, W.-H. Kuo, S.-H. Chan, and H.-L. Chan

Writing, review, and/or revision of the manuscript: L.-H. Wang and C.-W. Tseng

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): L.-H. Wang, C.-W. Tseng, S.-H. Chan, and H.-L. Chan

Study supervision: L.-H. Wang, C.-W. Tseng, and K.-J. Chang

We thank the Protein Chemistry Core Lab, Pathology Core Lab, and Cell Sorter Core Lab of the National Health Research Institutes for mass spectrometric analysis, H&E and IHC staining, and technical assistance of cell cycle, respectively. This study has been supported by Ministry of Science and Technology (MOST), Taiwan (MOST 104-2320-B-039-055-MY3, MOST 104-2320-B-039-054-MY3, and MOST 106-2811-B-039-004), and National Health Research Institutes (NHRI06A1-MGPP09-014) grants.

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.