Differential metabolic responses in breast cancer cell lines to acidosis and lactic acidosis revealed by stable isotope assisted metabolomics

Extracellular acidosis is considered as a hallmark of most human tumors, which plays an important role in promoting tumor malignant and aggressive phenotype in tumorigenesis. Acidosis and lactic acidosis can induce different responses in tumors. Previous studies have associated the response to lactic acidosis of tumors with good survival outcomes. In this study, we investigated the metabolomic changes in triple negative and luminal subtype breast cancer cell lines in response to acidosis and lactic acidosis. Our results showed that acidosis results in the reduction of cell viability and glycolysis in breast cancer cells, which is reversely correlated with the malignancy of cell lines. Under lactic acidosis, this reduction is reversed slightly. Untargeted metabolomic profiling revealed that glutaminolysis and fatty acid synthesis in cancer cells under acidosis are increased, while TCA cycle and glycolysis are decreased. Under lactic acidosis, the pentose phosphate pathway and acetate release are increased in MDA-MB-231 cells. The current results uncovered the different metabolic responses of breast cancer cells to acidosis and lactic acidosis, demonstrating the power of combined untargeted and stable isotope assisted metabolomics in comprehensive metabolomic analysis.


Results
Cell viability. The effects of acidosis on the viability of breast cancer cell lines (MDA-MB-468, MDA-MB-231, SkBr3 and MCF-7) were assessed by using CCK-8 assay. As shown in Fig. 1, acidosis resulted in significant decrease of viability of all four cell lines, especially MCF-7 and SkBr3, which can be ameliorated by the addition of lactate under lactic acidosis. MDA-MB-468 and MDA-MB-231, both of which are basal-like and triple negative cell lines 41 , are less susceptible to acidosis, and lactate could recover the decreased viability to near that under normal pH condition. On the other hand, MCF-7 and SkBr3, both of which are luminal subtype cell lines 41 , are more susceptible to acidosis and lactate can only partially restore the viability. Subtle difference in cell viability between these two cell lines was also observed. The SkBr3 cells is less susceptible to acidosis than MCF-7 cells.
The influence of acidosis on glycolysis. It has been reported that acidosis can decrease glycolysis of cancer cells 28 . Therefore, we compared the glucose consumption and lactate production in culture media under acidosis, lactic acidosis and normal pH. The relative content of glucose consumed and lactate excreted over This tendency of changes in glucose consumption is consistent with that of cell viability. On the other hand, the production of lactate (Fig. 2b), the terminal metabolite of glycolysis, was also significantly decreased in all three cell lines under acidosis, with the most drastic change in MCF-7 cells. Under lactic acidosis, the production of lactate was further reduced as compared with that under acidosis, in contrast to the changes in viability and glucose consumption. This may be due to the increased TCA activity which consumes pyruvate through the  www.nature.com/scientificreports/ production of acetyl-CoA (Ac-CoA), or the increased utilization of lactate by cancer cells under lactic acidosis. These results showed that acidosis can decrease the glycolysis in cancer cells. We performed 13 C NMR experiments on the extracts of MCF-7 and MDA-MB-231 cell lines, cultured in media supplemented with a mixture of U-13 C 6 -Glucose and 1-13 C-Glucose (1:1). As shown in Fig. 2c, the spectra results showed that the intracellular labeled lactic acid increased obviously in lactic acid group which illustrated that lactic acidosis can alleviate inhibition of glycolysis caused by acidosis. The response to acidosis in two triple negative breast cell lines is much weaker than in MCF-7. The extracellular acidification rate (ECAR) is widely used to evaluate the progress of glycolysis. The increased ECAR of cells cultured in lactic acidosis group indicates that lactic acidosis could promote glycolysis (Fig. 2d,e), which is consist with the 13 C NMR results.
Untargeted metabolic profiling of acidosis and lactic acidosis response in breast cancer cells. We evaluated the metabolomic profiles of breast cancer cells cultured in acidosis condition (pH 6.5), lactic acidosis condition (pH 6.5 with 10 mmol/L lactate) and normal condition (pH 7.4). Representative total ion chromatograms (TIC) of intracellular extracts from MDA-MB-231 and MCF-7 cells are shown in Supplementary Fig. S1. Multivariate statistical analyses were conducted on these data sets. Partial least squares discriminant analysis (PLS-DA) indicated that the metabolic profiles of each cell lines at variant culture conditions can be apparently separated ( Supplementary Fig. S2).
There are 58 and 49 metabolites were identified in MDA-MB-231 and MCF-7, respectively. The identified metabolites in MDA-MB-231 and MCF-7 were used to create the heat maps hierarchical clustering (Fig. 3). Figure 3 showed the three groups of MDA-MB-231 were basically discriminated and one of the SA sample was mixed into group LA, while in MCF-7, the groups were clustered well. The groups of SA and LA were similar compare with NP group, while there existed differences.
In order to compare the differences of three groups in MDA-MB-231 and MCF-7, we screen out the significantly changed metabolites. Significantly changed metabolites in all cell lines at different culture conditions were obtained based on the criteria of VIP > 1.0 and P < 0.05, which were summarized in Tables 1, 2 and Supplementary Table S1-S2. Under acidosis condition, sixteen metabolites were significantly changed in MDA-MB-231 cells (Table 1), including 11 decreased amino acid metabolites and 5 increased metabolites related to fatty acid metabolism. In MCF-7 cells, eleven metabolites were significantly decreased including amino acids and inositol, and four were significantly increased including glycolytic metabolites, citric acid and inosine 5′-monophosphate (IMP) (Supplementary Table S1). Under lactate acidosis, fifteen significantly metabolites were decreased and gluconic acid, palmitic acid and phenylalanine were increased in MDA-MB-231 (Table 2) when compared to normal condition. In MCF-7, eight significantly decreased metabolites were obtained including amino acids and inositol and five significantly increased metabolites including glycolytic metabolites, TCA metabolites and IMP (Supplementary Table S2). No significant changes in fatty acid were found in MCF-7 under both acidosis and lactic acidosis, while glycerol-3-P, one of glycolytic metabolites, also a skeleton compound of lipid synthesis, was significantly increased in both cell lines.
Identification of affected pathways in acidosis and lactic acidosis was based on enrichment analysis using MetaboAnalyst (Fig. 4). The results showed that three common pathways were affected by acidosis in two cell lines, which include: citrate cycle (TCA cycle), alanine, aspartate and glutamate metabolism, D-glutamine and D-glutamate metabolism. Pathways modulated by lactic acidosis in two cell lines include inositol phosphate metabolism, phenylalanine metabolism, alanine, aspartate and glutamate metabolism, arginine and proline metabolism. The affected pathways in MDA-MB-231 was more than that in MCF-7, which indicated that MDA-MB-231 was more active when faced with the external disturbance. under acidosis and lactic acidosis conditions. The labeled substrates can produce distinct labeling patterns of intermediates, making an outline of metabolic pathways activity in the given cell lines. As shown in Supplementary Fig. S3, the scheme shows labeled patterns of pyruvate from glucose. When the oxidative branch of PPP is operative, glucose would lose the first carbon as CO 2 and U-13 C 6 -glucose and 1-13 C-glucose converted to U-13 C 3 -pyruvate (M3) and unlabeled pyruvate (M0), respectively. When 1-13 C-glucose metabolized via glycolysis pathway, unlabeled pyruvate and 3-13 C-pyruvate (M1) would be generated. Therefore, the relative contents of M1 pyruvate were used to estimate the flux of pyruvate from PPP. The lower the relative content is, the more active the PPP pathway is. In MDA-MB-231 cells, significantly decreased M1 pyruvate Table 1. Relative concentration of significantly changed metabolites in MDA-MB-231 cultured at pH 6.5 (mean ± SD).   www.nature.com/scientificreports/ content indicated that PPP was enhanced in lactic acidosis group, while M1 pyruvate only showed slight change in MCF-7 cells. As shown in Fig. 5a, the scheme shows the fate of carbons originated from glucose. U-13 C 6 -glucose and 1-13 C-glucose can be metabolized to M3 pyruvate and M1 pyruvate, which can be further converted into M2 acetyl-CoA and M1 acetyl-CoA which participates in the TCA. The mass isotopologue distribution (MID) of fumaric acid and malic acid can be used to indicate the conversion rate of glucose into TCA cycle. In both cell lines, the significantly higher M0 and lower M2 components of citrate, fumaric acid and malic acid of all four metabolites in lactic acidosis group suggested the contribution of other sources such as lactate and a decreased conversion rate of glucose into TCA cycle (Fig. 5b). Glucose can be converted into α-ketoglutarate and glutamate (Fig. 5a). The percentage of M2 glutamate (Fig. 5b) was lower in LA, indicating a decreased conversion flux from glucose to glutamate under lactic acidosis condition. In Fig. 5b, M1 metabolites were increased in LA, while M1 pyruvate was decreased in LA ( Supplementary Fig. S3), which indicated that pyruvate in LA converted into other pathways decreased such as lactate (Fig. 2). For MDA-MB-231 cells, only slight changes of these metabolites were observed in acidosis condition whereas for MCF-7 cells, M2 malic acids were increased (Fig. 5b), which indicated that TCA pathway was not sensitive to acidic environment in MDA-MB-231 an increased conversion rate of glucose into TCA cycle in MCF-7.
Pyruvate enters TCA cycle not only via pyruvate decarboxylation but also via carboxylation. Pyruvate can be converted to oxaloacetate under the catalysis of pyruvate carboxylase (PC). M3 pyruvate was converted to M3 oxaloacetate and produced M3-M5 citrate (Supplementary Fig. S4). As shown in Supplementary Fig. S4, significantly decreased M3-M5 citrate in lactic acidosis group indicated that pyruvate carboxylation was decreased in both cell lines, while in acidosis group increased M3-M5 citrate in MCF-7 showed that pyruvate carboxylation was increased slightly. 13 C NMR experiments of MDA-MB-231 cell extracts cultured with a 1:1 mixture of U-13 C 6 -glucose and 1-13 C-glucose were also carried out, and the spectral assignments are shown in Supplementary Fig. S5. We can also see that the content of unlabeled acetate was increased in SA and LA ( Supplementary Fig. S6). Significantly increased unlabeled acetate in lactic acidosis group may be caused by other carbon sources such as lactate. In MCF-7 the labeled acetate was significantly decreased in acid media, while this change is not obvious in MDA-MB-231. Studies have showed that acetate was released more in acidosis via histone deacetylation to regulate intracellular pH 12 . The unlabeled acetate originated from histone deacetylation, the metabolism of other carbon sources and the undialyzed serum. ATP-citrate lyase (ACLY) could make citrate cleavage, with subsequent metabolites of OAA and acetyl-CoA 42 . Acetyl-CoA can be further converted to acetate. The level of labeled acetate illustrated that acidosis, especially lactic acidosis decreased glucose-to-acetate metabolic switch in MCF-7, while MDA-MB-231 was not sensitive to this stimulation.
In order to evaluate the effects of acidosis and lactic acidosis on PPP, MCF7 and MDA-MB-231 cells were cultured in media supplemented with 1,2-13 C 2 -glucose for 48 h, and the 13 C NMR data of media were collected. The 1,2-13 C 2 -glucose is metabolized into 2,3-13 C 2 -pyruvate through glycolysis and 3-13 C-pyruvate through PPP, followed by the production of 2,3-13 C 2 -lactate with a doublet and 3-13 C-lactate with a singlet (Fig. 6a). The relative 13 C fractional enrichments of excreted 3-13 C-lactate were evaluated by the integrating signals relative to labeled lactate C3. As shown in Fig. 6b, increased percentage of 3-13 C-lactate verified that PPP in both cell lines are metabolically more active in lactic acidosis.

Discussion
It was found that viability of different breast cancer cell lines under acidic milieu was discrepant. The cells viability was decreased under acidosis and lactate could reduce the acidosis damage to cells. Utilization and excretion of lactic acid depend on the concentration of lactic acid intracellular and extracellular and the expression of MCTs. In MDA-MB-231 and MDA-MB-468, both being basal-like and triple negative cell lines, expression of MCTs was different. MCT1 is expressed in MDA-MB-468 but could barely be detected in MDA-MB-231 43,44 . Several studies have been published on the acidosis and its effects on phenotype in tumorigenesis, however, little is known about the cellular metabolic response to acidosis and lactate acidosis in most solid tumors. Here we tested the relative content of glucose and lactate in the medium, which are the main metabolites of glycolysis, and employed untargeted and stable isotope-assisted metabolomics to investigate metabolites and pathways changes. In this study, decreased glucose and lactate in the medium reflected decreased glycolysis in acidosis and the results of 13 C NMR and glycolysis stress indicated that lactate acidosis could increased glycolysis. TCA cycle plays a key role in generating energy and providing precursors of certain amino acid as well as reducing agent NADH used in many other biochemical reactions. With the previous studies, our present data showed that levels of TCA cycle-related metabolites citric acid, fumaric acid and malic acid were decreased in low pH in MDA-MB-231 ( Supplementary  Fig. S7). There was a statistically significant difference (P < 0.05) in the content of malonate between SA and NP, while between LA and NP, the malonate occurs at elevated concentration with no significant difference. Hence, a slight elevation of TCA cycle metabolites in LA was observed in comparison with SA group. In MCF-7 only slight changes with no significant differences were observed (Supplementary Fig. S8). Malonate that is analogue of succinate, a succinate dehydrogenase inhibitor of TCA cycle, which increases in low pH, could be an influence factor of making TCA decrease. Citric acid usually used to support de novo synthesis of fatty acid, other TCA cycle metabolites are used for biosynthesis of various macromolecules.
ATP-citrate lyase (ACLY) could make citrate cleavage, with subsequent metabolites of OAA and acetyl-CoA 42 . The obtained acetyl-CoA is a key metabolite, which connect glycolysis, TCA and fatty acid metabolism. Acetyl-CoA can also make modification for histone acetylation, which is of great importance for formulation and development of tumors. Acetyl-CoA furthermore converted to acetate to be detected. Studies have showed that acetate was released more in acidosis via histone deacetylation to regulate intracellular pH 12  www.nature.com/scientificreports/ results, on one hand, the unlabeled acetate originated from histone deacetylation, on the other hand, it comes from metabolism of other carbon sources. The level of labeled acetate illustrated that acidosis, especially lactate acidosis decreased glucose-to-acetate metabolic switch in MCF-7, which may suggest that acidosis and lactic acidosis could decrease ACLY expression, while MDA-MB-231 was not sensitive to this stimulation. Pentose phosphate pathway is another important metabolic pathway for cell survival. PPP can produce carbon skeleton of ribose or derivatives, such as ATP, CoA, NAD+, FAD, RNA or DNA, which are crucial constituent part of biological molecules. For another, PPP is the main source of NADPH. NADH is mainly used to provide ATP, while unlike NADH, NADPH can provide hydrogen ions in reducing biosynthesis 45,46 . Therefore, reduced NADPH can favor fatty acid and steroid biosynthesis 47 . In our study, PPP was increased in lactic acidosis with the help of stable isotope-assisted metabolomics in MDA-MB-231 and MCF-7. The cell viability enhanced in lactic acidosis compared to that in acidosis can be accounted for the increased PPP flux.
Lipids are of importance to the membrane formation, energy storage and signal transduction. Dysregulation of de novo lipogenesis is increasingly recognized as a hall mark of cancer. Increased lipid synthesis fuels cancer energy and contributes to production of membranes for cell proliferation under metabolic stress, such as hypoxia and acidosis 48,49 . Studies have shown that fatty acid synthase (FAS) overexpression promoted apoptosis and proliferation in cancer cells 50 . We can observe that the fatty acids (stearic acid, palmitic acid, oleic acid, myristic acid) and cholesterol increased in low pH with significant differences in MDA-MB-231 cells, while that of MCF-7 showed minor changes (Supplementary Figure S7, S8). Some studies show that acidosis drives FA metabolism reprogramming in cancer cells, which can be a reason that MDA-MB-231 survived in acidosis according to the viability tests. This metabolic disturbance makes cancer cells overcome the metabolic stress 51 .
In conclusion, we compared the metabolic changes of two acidosis modes on MDA-MB-231 cells and MCF-7 cells in our study. The activity of breast cancer cells was reduced after exposure to acute acid stimulation, while exposure to lactic acidosis was associated with an enhancement of cell viability compared to acidosis. Acidosis decreases the glycolysis as revealed by downregulated lactic acid in media, while lactic acidosis could promote glycolysis as we observed the elevated ECAR in lactic acidosis. The M1 pyruvate and labeling pattern of lactate C3 indicated that PPP flux increased with the lactic acidosis stimulating. Both acidic irritations can promote fatty acid synthesis in MDA-MB-231. These results indicated that cell metabolic changes existed in acidosis, while both the acidosis caused different metabolic reprogramming. Addition of lactate can rescue the low viability of cancer cells in acute acid environment. All these metabolic changes in cancer cells caused by lactate in acidosis can provide valuable information, which could help to get known about the underlying metabolism of cancer cells.

Materials and methods
Cell culture and extraction. MCF-7, MDA-MB-468, MDA-MB-231, and SkBr3 cells were obtained from Hunan Fenghui Biotechnology Co., Ltd. All cells were maintained in Dulbecco's Modified Eagle Medium (DMEM, Gibco) containing 10% Fetal Bovine Serum (FBS, Gibco) and 1% penicillin-streptomycin (HyClone). Cells were subcultured to 10 cm plates by rinsing with PBS (Gibco) and detaching with 0.5% trypsin-EDTA (Sigma). Cells were grown in an incubator (SANYO, MCO-20AIC) containing 5% CO 2 and ambient oxygen at 37 °C. Simple acidosis (SA) was generated via media pH adjusted to pH 6.5 using HCl, while lactic acidosis (LA) was generated via the addition of 10 mmol/L sodium lactate (Sigma), then media pH adjusted to pH 6.5 using HCl (Lactate, with a relatively weak dissociation, has an limited impact on the buffer capacity). The media contain 10 mmol/L glucose. All groups were cultured for 48 h. www.nature.com/scientificreports/ For 13 C-glucose labeling experiments, cells were plated in high glucose DMEM overnight in 10 cm or 15 cm dishes, and then rinsed with PBS and switched to media containing 5 mmol/L U-13 C 6 -glucose and 5 mmol/L 1-13 C-glucose or 10 mmol/L 1,2-13 C 2 -glucose (CIL, Cambridge Isotope Laboratories) under control, acidosis or lactic acidosis condition and cultured for 48 h.
Culture media and cell pellets were collected after cells were rinsed twice in 1 × PBS, quenched and scraped in 1 mL of 80% methanol/H 2 O. The cell suspension was ultrasonicated in ice bath twice with 20 min each time and vortexed for 30 s, and then centrifuged at 13,000×g for 20 min at 4 °C. The supernatant was collected into 2 mL vial and added with 20 µL of 0.2 mg/mL ribitol (Sigma-Aldrich) as an internal standard. The procedures for derivatization of the supernatant were performed as previously reported 7 . For GC-MS, six replicates of each sample were performed and for NMR, three replicates of each sample were performed in the subsequent detection.
Cell treatment and cell viability assay. Cell viability of cancer cell lines (MCF-7, MDA-MB-231, MDA-MB-468 and SkBr3) was assessed in acidosis and lactic acidosis by colorimetric assay with Cell Counting Kit-8 (CCK-8 solution, Beijing Applygen Technologies Inc.). Cancer cells were plated in 96 well culture plates at a concentration of 5 × 10 3 cells per well and allowed to attach for 12 h. Cells were treated with acidosis medium (pH 6.5) and lactic acidosis medium (10 mmol/L lactate, pH 6.5) with a volume of 200 μL and incubated for 48 h. After the addition of the CCK-8 solution (10 μL/well), the reaction was incubated for 2 h and a reading of optical densities (OD) was taken at 450 nm using an ELISA reader (Biotek, Elx808, USA). The viability (%) was calculated based on the following formula: Viability (%) = (OD tr -OD bl ) / (OD ntr -OD bl ). (OD tr is the mean OD of the treated cultures, OD ntr is the mean OD of the non-treated cultures and OD bl is mean OD of the blank-wells).

Extracellular acidification.
To measure extracellular acidification rate (ECAR), 20,000 cells from each cell line were seeded into each well of a XF24 microplate in a total of 100 μL for 4 h. Cells were subsequently treated with pH 7.4 and pH 6.5 medium (with or without 10 mmol/L lactate) for 48 h. ECAR were measured using a Seahorse Bioscience Extracellular Flux Analyzer (XF24). The data were analyzed using the Seahorse XF software.

Sample derivatization and GC-MS analysis.
The sample derivatization method is the same as previously described 7 . The extracts were added with 40 µL pyridine solution of 20 mg/mL methoxyamine hydrochloride (Sigma-Aldrich), vortexed for 30 s and incubated at 30 °C and 130 rpm for 90 min in a gas bath shaker, followed by addition of 40 µL MSTFA (Sigma-Aldrich) with 1% chlorotrimethylsilane (TMCS, Sigma-Aldrich) and incubated at 37 °C for further 30 min. The derivatized samples were stored at 4 °C before detection.
The samples were analyzed with an Agilent 6890 GC system coupled with a Pegasus 4D time-of-flight mass spectrometer (Leco Corporation, St Joseph, MI, USA). The column was DB-5 MS (30 m × 250 µm i.d., 0.25 µm, Agilent J&W Scientific, Folsom, CA, USA), with an oven program set at 70 °C for 1 min, increased to 280 °C with a ramp rate of 5 °C/min holding for 15 min. Samples (1 μL) were injected with no split. The temperature of the inlet, interface and ion source was set as 250, 250, 220 °C, respectively. The solvent acquisition delay was 300 s. The mass spectrometer was operated in full scan mode (m/z 50-800) with an acquisition rate of 10 spectra/s. Electron impact ionization was set at 70 eV 7,52 . 1 H and 13 C NMR parameters. 13 C NMR experiments were performed at 298 K using a Bruker AVANCE 950 MHz NMR spectrometer equipped with Cryo probes. 13 C spectra were acquired with a relaxation delay of 3 s, 5120 scans, and 32 k data points. 1 H experiments were performed at 298 K using a Varian UNITYINOVA 600 MHz spectrometer. 1 H NMR spectra of cell media were acquired using a standard PRESAT pulse sequence with a relaxation delay of 2 s, 32 scans, 32 k data points, and a spectral width of 7800 Hz. Spectra were manually phased and baseline-corrected.
For cell extracts, the samples were dissolved in 400 μL D 2 O and 5 μL TSP D 2 O solution (1 mg/mL) for 13 C experiments, and for culture media, 500 μL of the media was mixed with 80 μL TSP D 2 O solution (1 mg/mL) for experiments. Data processing and statistical analyses. Peak picking and alignment of the acquired GC-MS data were conducted by the Statistical Compare feature of ChromaTOF software (Version 4.5, Leco Corp.). Peaks with a signal-to-noise ratio (S/N) greater than 50 were considered and normalized to the intensity of ribitol. Metabolites were identified according to NIST library and the library built with various metabolites standards in our laboratory. The processed data were imported into SIMCA 13.0 software (Umetrics) for multivariate pattern recognition analysis. Principal Components Analysis (PCA) and Projection to Latent Structures Discriminant Analysis (PLS-DA) were applied to compare the chromatograms of samples treated with Pareto scaling (Par) 7 . Pathway analysis on the modulated metabolites was performed using Metaboanalyst 4.0 tool 53 .
For NMR spectra, peaks were normalized to the intensity of TSP. Metabolites were identified according to the library built in our laboratory.
For all other graphs, samples sizes are indicated in the figure legends; P values were derived by two-tailed t test. P values are indicated (*P ≤ 0.05, **P ≤ 0.01, and ***P ≤ 0.001) and error bars are standard errors of the mean.