Introduction

Doxorubicin is a popular DNA-intercalating agent for breast cancer therapy [1, 2]. Its anticancer mechanisms are well known, including direct intercalation between DNA bases, DNA strand breaks [3], free oxygen radical generation, increased lipid peroxidation, membrane structure alterations, and the effects on apoptosis pathways through caspase-3 activation and p21 induction [4, 5]. However, breast cancer patients often encounter drug resistance during continuous dosing of doxorubicin, which poses a great challenge to using doxorubicin in breast cancer treatment [6]. This situation also occurred with cisplation.

The popular MCF-7 breast cancer cell line can produce a stable and resistant cell line, MCF-7/Adr, which is resistant to doxorubicin after chronic exposure to doxorubicin to sensitive breast cancer cell lines [7]. COC1/DPP as well as MCF-7/Adr, which is resistant to cisplation, was also produced by chronic exposure to cisplatin.

By using genomics and proteomics methods, researchers have investigated the mechanisms of drug resistance induced by doxorubicin or cisplatin [8]. The results revealed that some genes [9, 10] and proteins are related to drug resistance [11, 12]. Some cell factors also play important roles in the development of drug resistance, e.g., epithelial-to-mesenchymal transition as well as altered heat shock response elements [13]. Proteomics comparing analyses between drug-resistant and drug-sensitive MCF-7 cells identified that several proteins involved in apoptosis, the cell cycle, glucose metabolism, and fatty acid oxidation contribute to drug resistance [14]. The genomic analysis of drug-resistant cell lines revealed that the drug resistance is related to gene repair [15], cell cycle regulation, and overexpression of multidrug-resistant genes [16]. Although the resistance-related genes and proteins have been investigated and several metabolic pathways involved in MCF-7/Adr drug resistance have been found [17], little is known about the global metabolic pattern and shift between the sensitive and resistant breast tumor cells.

Although tumor cells were characterized by reprogrammed metabolic pathways including enhanced anaerobic glycolysis and reduced tricarboxylic acid cycle (TCA) activity [18], i.e., the Warburg effect [19], alternative utilization of glutamine as an energy supply, and the glycine in rapid tumor cell proliferation [20], the metabolomics characteristics of drug-resistant tumor cells are not well understood [5, 21].

Metabolomics is a quantitative measurement of the dynamic metabolic response of living systems to genetic modifications, physiological and environment changes, and so forth [22]. Metabolomics used for the high-throughput analysis of low molecular mass compounds in cells or biological samples [23] reflects the metabolic shifts involved in biological processes; thus, it might reveal the underlying mechanisms involved in these processes [24]. Studies of the effects of chemotherapeutic agents on drug-resistant MCF-7/Adr cells would provide useful information to understand the mechanisms of drug resistance.

In this study, metabolomics was employed to profile intracellular metabolites in MCF-7/Adr and MCF-7/S cells. Some potential biomarkers were identified and the metabolic pathways were analyzed to illustrate the mechanisms of drug resistance in the cellular metabolic processes. Metabolic patterns and potential markers identified can be used for evaluating the pharmacodynamics and exploring the mechanisms of tumor cells resistant to doxorubicin.

Experimental

Materials

Doxorubicin and cisplatin were purchased from Hisun Pharm Co. (Taizhou, China). Ethanol, acetonitrile, and chloroform (HPLC grade) were purchased from Dima Co. Water used as the HPLC mobile phase was deionized with an Alpha-Q purifier from Millipore (Bedford, USA). MCF-7/S and MCF-7/Adr were obtained from Chinese Academy of Medical Science (the 57th generation; MCF-7/S and MCF-7/Adr were in the same generation). RPMI-1640 medium was obtained from Hyclone, China. Fetal bovine serum was supplied by Hangzhou Sijiqing Co Ltd. Penicillin and streptomycin were purchased from Beijing Dingguo Co. Ltd. The GENiosPro plate reader was purchased from Tecan (Switzerland).

Cell culture

Human MCF-7 breast cancer cells and the doxorubicin-resistant subline (MCF-7/Adr) were grown in RPMI-1640 medium supplemented with 10 % (v/v) fetal bovine serum. COC1 and COC1/DDP cells were cultured in DMEM medium with 100 U/ml penicillin and 0.1 mg/ml streptomycin at 37 °C with 5 % CO2. Both cell lines were cultured in six-well plates and grown to nearly 90 % confluence. MCF-7/Adr cells were incubated with 10 μM doxorubicin for 48 h and then changed to fresh medium without doxorubicin. After another 24 h, the culture medium in each well was collected, and the residues were discarded. The adherent cells were washed twice with PBS immediately before adding 300 μL of trypsin solutions (5 %) to each well. Cells were then collected into 1.5-mL tubes (three replicates were cultured in parallel and the mean of these was obtained).

Cell viability assays

To monitor the cell viability during the stimulations of apoptosis and necrosis, the MTT assay was performed. Cells were grown in 96-well plates. After 24 h, 10 mL of 5 mg/ml solution of MTT (Roche, Indianapolis, IN, USA) in PBS buffer was added to each well and the cells were incubated at 37 °C for 4 h. The medium was then completely removed. To each well, 100 mL of DMSO was added and the plates were incubated for 40 min at room temperature. Optical absorbance of reduced MTT, reflecting the viability of the cells, was measured at 590 nm using a GENiosPro plate reader [25].

Sample preparation

The metabolites from MCF-7/S and MCF-7/Adr cells were obtained as described in elsewhere [10]. Briefly, at the end of the incubation periods, the cells were washed rapidly with precooled PBS and then treated with trypsin. Each digested sample was resuspended in 150 μL of sterilized ultrapure water. The resulting solution was frozen for 10 min in a freezer at −80 °C and then put into a 37 °C water bath; this process was repeated three times. N,N-Dimethyl-l-phenylalanine (N,N-Phe) was spiked to correct the retention time. Cell lysates were added with 600 μL ice methanol and the mixture was then vortexed for 5 min. Thereafter, 400 μL chloroform was added into the sample tube to extract the metabolites, and the mixture was vortexed for 5 min. The sample tube was then stored at −20 °C for 6–8 h. After the tube was centrifuged at 12,000 rpm for 5 min at 4 °C, the supernatant was transferred to a filter cap and centrifuged again at 12,000 rpm for 5 min at 4 °C. The filtrate was collected and freeze-dried by nitrogen.

UPLC–MS and UPLC–MS/MS analysis

The UPLC–MS and UPLC–MS/MS experiments were performed with a Waters Acquity™ ultra-performance liquid chromatography system coupled with a Waters Synapt G2 Q-TOF mass spectrometer (Waters, Manchester, UK). Chromatographic separation was carried out at 25 °C on an ACQUITY UPLC HSS T3 column (2.1 mm × 100 mm, 1.7 μM, Waters). The mobile phase consisted of A (0.1 % formic acid in water) and B (0.1 % formic acid in acetonitrile). The optimized UPLC elution conditions were as follows: 0–1.0 min, 10 % B; 1.0–4.0 min, 10–36 % B; 4.0–7.0 min, 36–85 % B; 7.0–10.0 min, 85–100 % B; 10.0–16.0 min, 100 % B; 16.0–16.1 min, 100.0–10.0 % B; and 16.1–20.0 min, 10.0 % B. The flow rate was 0.5 mL/min. The autosampler was maintained at 4 °C. A sample volume of 5 μL was injected for each run.

The MS conditions are as follows: the m/z scan range was from 50 to 1200. In negative ion mode, the capillary and cone voltage were set at −2.5 kV and 45 V, respectively. In positive ion mode, the capillary and cone voltage were set at 3.0 kV and 45 V, respectively. The desolvation gas was set to 600 L/h at a temperature of 400 °C, the cone gas was set to 30 L/h, and the source temperature was set to 120 °C. The data acquisition rate was set to 0.1 s, with a 0.1-s interscan delay. The MSE experiment with two scan functions was carried out with the following parameters: function 1: m/z 50–1200, 0.1 s scan duration, 0.02 s interscan delay, collision energy ramp of 10–30 V; function 2: m/z 50–1200, 0.1 s scan duration, 0.02 s interscan delay, collision energy ramp of 10–20 V. MSE mode enabled almost simultaneous acquisition of both MS and MS/MS data in a single analytical run. All analyses were acquired using the lock spray to ensure accuracy and reproducibility. Leucine–enkephalin was used as the lock mass at a concentration of 300 ng/ml and flow rate of 5 μL/min. Data were collected in continuum mode, the lock spray frequency was set at 10 s, and data were averaged over 10 scans. All the acquisition and analysis of data were controlled by Waters MassLynx v 4.1 software [26].

Quantification of metabolites was performed using LC–MS/MS where a liquid chromatograph is coupled with a triple quadrupole mass spectrometer (Xevo-TQ, Waters). All the acquisition data were analyzed by masslynx 4.1 software.

Data treatment and statistical analysis

The mass data acquired were imported to Markerlynx XS (Waters Corp., Milford, MA, USA) within the Masslynx software for peak detection and alignment. Prior to analysis, the quality of the resulting data should be checked in the metabolic fingerprinting assay. The raw data obtained from UPLC–Q-TOF-MS was converted into network common data using the Masslynx software (Waters Corp, Milford, MA).

For comparison, data were filtered by choosing only the features that were present in all samples in one of the compared groups. To approximate a normal distribution, data were scaled by applying the common logarithm to the abundances after filtering.

Markerlynx XS software was used to preprocess the UPLC–Q-TOF-MS data using several algorithms including the filtration and the normalization. For filtration and detection of ion peaks, the ion chromatograms were extracted by using relative model peaks. The UPLC–Q-TOF-MS data were cut into slices by a peak detection algorithm based on mass unit, and then the retention time in the total ion chromatograph was recorded. The retention time and relative intensity were compared to generate the mass domain bins in which to group peaks. UPLC–Q-TOF-MS data on the retention time and the peak intensity, i.e., in two dimensions, was analyzed by Markerlynx XS Microsoft. The drift retention time was corrected and the peak intensity was compared from run to run. For each group, the median value and the deviation of retention time for each mass unit in that group were calculated.

SIMCA-P software (V11.5, Umetrics, Umea, Sweden) was used to conduct orthogonal partial least-squares discriminant analysis (OPLS-DA) in which the OPLS model comprises two blocks of modeled variation, i.e., the Y predictive (T p) block refers to the class variation and the Y orthogonal (T o) block refers to the uncorrelated variation. The score plots for t p and t o afforded a visual classification of MCF-7/Adr and MCF-7/S. After we obtained the score plot, the data was analyzed by the OPLS-DA model and then the loading plot, which had an S shape. The metabolites that far from the origin were chosen to create a marker list. In this list, the variable importance in the projection (VIP) value and p value were clear.

The differences among samples were evaluated for individual metabolites by using an independent samples t test (p < 0.05) and a VIP value generated by SIMCA-P in the OPLS-DA model of more than 1.0. These metabolites were chosen as potential biomarkers, and the accurate mass of features representing significant differences was searched against METLIN, mass Trix public databases, Madison Metabolomics Consortium Database (MMCD), and Human Metabolite Database (HMDB). To identify the metabolite more accurately, the mass accuracy was set at 30 ppm.

Results and discussion

Persistent resistance of MCF-7/Adr cells to doxorubicin

Drug-resistant cell lines are usually generated from drug-sensitive cells by maintaining a low concentration of drug in the cells and gradually increasing the dose over time. The drug-resistant cell line of MCF-7/Adr is characterized by overexpression of P-glycoprotein and the failure of DNA repair which leads to gene mutation. For the gene mutation the drug’s target is also changed. In this paper, MCF-7/S and MCF-7/Adr cells were maintained in RPMI-1640 medium only supplemented with 10 % (v/v) FBS, and then the effect of different concentrations of doxorubicin on cell viability was examined by the MTT method to verify whether the insensitivity of MCF-7/Adr cells to doxorubicin was persistent and whether they exhibited high tolerance to doxorubicin. As expected, the MCF-7/Adr cells showed higher survival rates compared to MCF-7/S cells after being maintained in doxorubicin culture medium (Fig. 1).

Fig. 1
figure 1

Sensitivity of MCF-7/S and MCF-7/Adr cells to doxorubicin treatment. (As the MTT results show, the IC50 values for doxorubicin against the MCF-7 and MCF-7/Adr cell lines were 0.8 μM and 1.4 μM, respectively)

Full size image

Optimization of experimental conditions

The extracts of cells were measured both in positive and negative ion ESI modes in to optimize the experimental conditions. More metabolites with lower noise level and matrix effects were detected using positive ion ESI. A 20-min UPLC gradient program was employed to separate the metabolites (see “UPLC–MS and UPLC–MS/MS analysis”). The autosampler compartment was set at 4 °C throughout the analysis to minimize thermal degradation of the metabolites. As shown in Fig. 2a, a good number of metabolites were eluted and detected within 18 min with better separation.

Fig. 2
figure 2

Total ion chromatography of metabolites and OPLS-DA score plots. a Total ion chromatography of metabolites. Score plot of MCF-7/S and MCF-7/Adr cultured doxorubicin in b positive ion mode and c negative ion mode. Score plot of MCF-7/S and MCF-7/Adr cultured without doxorubicin in d positive ion mode and e negative ion model. Squares MCF-7/Adr, triangles MCF-7/S

Full size image

Data analysis

The cancer cells produced lots of metabolites. The total ion chromatography in positive mode in Fig. 2a shows that many endogenous metabolites were obtained. Figures 2b and c show that the MCF-7/S group and MCF-7/Adr group were distinguished efficiently, indicating a lot of different endogenous metabolites between these groups. We used the SIMCA-P software to find the biomarkers that were far away from the origin on the basis of the loading plot, which was obtained from SIMCA-P. The compounds that had changed dramatically were identified as potential biomarkers. A different significance t test was used in this paper (p < 0.05). Further identification of the biomarkers was based on LC–MS/MS and the mass spectrometry database search. To explain the metabolic changes of drug-resistant cells, we also compared the metabolites between MCF-7/Adr cells and MCF-7/S cells which were cultured in the medium without doxorubicin. As show in Fig. 2d, e, the metabolites were not changed dramatically.

Metabolite identification

In this study, METLIN, MMCD, and HMDB were used for mass-based metabolite identification. METLIN is a Web-based database that can facilitate the identification of metabolites by using accurate mass data. It includes an annotated list of structural information for known metabolites. The database has links to other public databases such as the Kyoto Encyclopedia of Genes and Genomes (KEGG) and PubChem. Further MS/MS data were obtained for metabolite identification (Fig. 3). For example, the ion at m/z 308.09 with a corresponding compound elution time of 5.76 min was initially assigned as glutathione through a database search, and its identity was further confirmed by MS/MS with the daughter ions at m/z 233.0610, 179.0501, 162.0235, and 76.0129. In total, 24 compounds were putatively identified (Table 1), and these compounds belong to different metabolic pathways. Most of them are related to energy and amino acid metabolism activities that are closely related to drug resistance development [19, 20].

Fig. 3
figure 3

MS/MS spectra of the endogenous metabolites

Full size image
Table 1 Metabolites with significant differences between MCF-7/S cells and MCF-7/Adr cells
Full size table

To verify the changes of metabolites, we quantified the concentrations of several important metabolites by triple quadrupole mass spectrometry using the standard and calculated the accurate concentration by calibration curves. The analytical conditions are shown in Table 2. The concentrations of some important metabolites changed dramatically (Fig. 4). For example, some amino acids, ketoglutarate, NADH, glutathione, and creatine were upregulated dramatically, and l-cysteine and tyrosine were downregulated. These amino acids are related to protein synthesis; ketoglutarate and NADH are related to the energy metabolism; glutathione may be associated with the antioxidant activity. (The result was the mean of three repeats of each cell line).

Table 2 MS conditions for analysis of the metabolites
Full size table
Fig. 4
figure 4

Concentration changes of relative metabolites in MCF-7/S cells and MCF-7/Adr cells. (Column are the concentrations of metabolites and error bars are SD values of 3 repeats)

Full size image

As shown in Fig. 5, the metabolism of MCF-7/Adr cells changed dramatically. Some secondary metabolites related to energy metabolism and the ammonia cycle, such as pyruvic acid and spermine, increased in MCF-7/Adr cells. But some of the amino acids were reduced in MCF-7/Adr cells, which might be the result of high-speed protein synthesis [13].

Fig. 5
figure 5

Hierarchically clustered heat map of metabolism levels in MCF-7/Adr cells and MCF-7/S cells. Rows represent different cell lines, columns represent different reactions. Different color represents the concentration differences of different reactions, e.g., red represents a high concentration and blue represents a low concentration. Red band MCF-7/Adr, green band MCF-7/S

Full size image

To study the common mechanism of drug resistance, we also identified the different metabolites between COC1 cells and COC1/DDP cells which are resistant to cisplatin; the method used to identify the compounds was similar to before. The compounds which have metabolic significance are listed in Table 3. We putatively identified 15 compounds as the most different metabolites by LC–MS/MS. Most of these compounds were similar to the metabolites found in MCF-7/Adr cells.

Table 3 Metabolites with significant differences between COC1 cells and COC1/DDP cells
Full size table

Table 3 shows that there were some similar metabolites between doxorubicin-resistant cells and cisplatin-resistant cells. These metabolites were amino acids, energy metabolites, or cofactors. Most of the amino acids are related to the biosynthesis of proteins. The glutamine, α-ketoglutarate, and NADH are related to glucose metabolites. They are important metabolites for cell proliferation and survival. Metabolites were analyzed by MATELAB 7.0 and the resultant heat map clearly shows the changes of metabolites (Fig. 5).

In this study, the data were input into a Web analysis databank (http://www.metaboanalyst.ca/faces/ModuleView.xhtml) and the metabolic pathways were analyzed to find which are largely involved. As shown in Fig. 6, the biosynthesis of amino acyl-tRNA was enhanced the most, reflecting that the biosyntheses of proteins were enriched including enzymes, transport proteins, and other membrane proteins. Several metabolic pathways related to energy metabolism were also enhanced, including phenylalanine metabolism, pantothenate and CoA biosynthesis, butanoate metabolism, and the citrate cycle. In addition, reactions related to the urine cycle and sulfur transfer, such as arginine, proline, cysteine, and methionine metabolism, were also enhanced. These metabolic pathways might directly or indirectly reflect the drug resistance of tumor cells.

Fig. 6
figure 6

Changes of MCF-7/Adr metabolic pathways (left) and overview of metabolites that were enriched in MCF-7/Adr cells (right). 1, 2 Aminoacyl-tRNA biosynthesis, 3 phenylalanine metabolism, 4 arginine and proline metabolism, 5 pantothenate and CoA biosynthesis, 6 butanoate metabolism, 7 citrate cycle (TCA cycle), 8 thiamine metabolism, 9 alanine, aspartate, and glutamate metabolism, 10 cysteine and methionine metabolism, 11 phenylalanine, tyrosine, and tryptophan biosynthesis, 12 valine, leucine, and isoleucine biosynthesis, 13 beta-alanine metabolism, 14 propanoate metabolism

Full size image

In the study of the metabolic difference between COC1 and the cisplatin-resistant cell line COC1/DDP, similar metabolic pathways changed. The enrichment of the biosynthesis of amino acyl-tRNA might reflect the enrichment of protein biosynthesis; the pantothenate and CoA biosynthesis might reflect the glucose metabolism (Fig. 7).

Fig. 7
figure 7

Changes of COC1/DDP metabolic pathways (left) and overview of metabolites that were enriched in COC1/DDP cells (right). 1 Aminoacyl-tRNA biosynthesis, 2 cysteine and methionine metabolism, 3 valine, leucine, and isoleucine biosynthesis, 4 arginine and proline metabolism, 5 glycine, serine, and threonine metabolism, 6 pantothenate and CoA biosynthesis, 7 glutathione metabolism, 8 taurine and hypotaurine metabolism, 9 alanine, aspartate, and glutamate metabolism, 10 valine, leucine, and isoleucine degradation, 11 d-glutamine and d-glutamate metabolism, 12 sulfur metabolism, 13 citrate cycle (TCA cycle), 14 thiamine metabolism, 15 beta-alanine metabolism

Full size image

In this study, the intracellular metabolites of MCF-7/Adr or COC1/DDP cells dynamically changed within a few hours after drug treatment. In particular, several amino acids including proline, glutamate, and others or some related metabolic intermediates were dramatically upregulated or downregulated. They were continuously altered in MCF-7/Adr cells after doxorubicin treatment.

Energy is a primary requirement for life as every process needs energy [27]. In this study, we found that more energy and related substances are needed by the drug-resistant cells (Fig. 8). 3-Phosphoglycerate as a biochemically significant 3-carbon metabolic intermediate in glycolysis. The content changes of 3-phosphoglyceric acid could reflect the changes of energy metabolism in tumor cells. l-Carnitine is a quaternary ammonium compound biosynthesized from the amino acids lysine and methionine. Glutamate and glutamine are important substances during energy metabolism processes. They can interconvert to each other. Glutamate is a key compound in cellular metabolism. It can enter TCA cycle through alpha-ketoglutarate to fuel the TCA cycle and is an important carbon and nitrogen source. Glutamine can be converted into glutamate to provide the source for biomass synthesis.

Fig. 8
figure 8

Correlation metabolic pathways in response to MCF-7/Adr cells with multidrug resistance. The metabolites marked in red denote that they had been identified in the paper. The arrow direction denotes the metabolite’s synthetic direction

Full size image

It was shown that drug-resistant cancer cells were dependent on arginine for optimal growth [28]. When arginine was replaced with ornithine, a urea cycle intermediate and precursor of polyamine synthesis, the arginine failed to restore growth to the same level. Arginine was also a substrate of nitric oxide synthesis; elevated NO upregulated HIF-1 that is related to signaling transduction pathways [29, 30].

Polyamine was an important factor of microbial growth which was firstly studied in 1958 and can promote growth of Haemophilus parainfluenzae [31]. Polyamine was also reported to play a key role in the prevention of nerve damage and duodenal mucosal repair [32]. Spermine was recommended for the treatment of human prostate cancer in individual use or in combination with other anticancer drugs [33]. One report indicates that polyamines can function as growth factors, antioxidants, and stabilizers of DNA, RNA, membranes, metabolic regulators, nutrients as well as secondary messengers [34]. In addition, spermidine and spermine have been shown to bind to the phosphate backbone of polynucleic acids [35]. This interaction is mostly based on electrostatic interactions between positive ammonium groups of the polyamines and the negative phosphates of the polynucleic acids [36].

The intracellular antioxidative system was also an important factor in the protection of cells from chemical-induced oxidative damage of DNA and cytotoxicity [37–39]. The antioxidative system mainly exhausted glutathione (GSH), which is a tripeptide containing sulfhydryl group (–SH) [40]. Glutathione also possessed high antioxidative activity and thus can eliminate superoxide anion free radicals and hydroxyl free radicals. On the basis of this character, glutathione could promote the ability of cancer cells to survive under high oxidative stress, and also elevate the effect of drug resistance [41–43].

Conclusions

The use of metabolomics in exploring drug resistance is still a challenge in spite of some progresses being achieved. In this research, metabolomics analysis was performed in resistant cancer cells to understand the mechanisms of action. Comparison of the metabolites of cisplation-resistant cells with the metabolites of doxorubicin-resistant cells revealed that some compounds are the same. The results indicate that drug resistance of cancer cell is related not only to the biosynthesis of proteins but also survival modes, such as antioxidation, the enhancement of energy metabolism, and accelerating the growth of cancer cells. The results also show that metabolomics can serve as an important tool in investigating the mechanisms of drug resistance by metabolic pathway analysis.