Sage Journals: Discover world-class research

Abstract

BACKGROUND:

This study investigated the use of serum amino acids and organic acids profiles as the novel metabolites for screening breast cancer (BC) patients.

METHODS:

A total of 116 subjects as training set were divided into the following three groups: BC patients ( $n=$ 34), benign (BE) patients ( $n=$ 38) and controls ( $n=$ 44). The amino acids profiles from three groups were measured using liquid chromatography-mass spectrometry and organic acids profiles in three groups were studied by gas chromatography-mass spectrometry. The resultant study data set was subjected to multivariate statistical analysis to identify important metabolites related with BC and construct the criteria for discriminating BC patients from BE subjects or controls. A test data set derived from 60 patients (30 BC and 30 BE subjects) and 30 controls was used to validate the stability of the different metabolites.

RESULTS:

The serum amino acids and organic acids profiles significantly differed between the BC patients, BE patients and the controls. Our results demonstrate that combinations of three candidate metabolites from taurine, glutamic acid and ethylmalonic acid were found to mirror tumour burden, with AUC values ranging from 0.751 to 0.834 when comparing BC patients to the controls. The areas under the curve from the taurine, glutamic acid and ethylmalonic acid in validated study were 0.901, 0.924 and 0.749, respectively.

CONCLUSIONS:

This study shows that amino acids and organic acids profiles will be a potential screening tool for BC patients. The dysregulated metabolism of amino acids and organic acids in breast cancer might be useful for the diagnosis, therapy, prognosis and understanding the pathogenesis of breast cancer.

Keywords

Breast cancer metabolomics amino acids profiles organic acids profiles serum

1. Introduction

Breast cancer (BC) is one of the most fearful diseases due to its increasing worldwide prevalence, not only in European-Americans, but also in Asian countries [1]. In Europe, 458,337 women were diagnosed with breast cancer in 2012 and 131,259 women died of breast cancer [2]. Worldwide, close to 1.4 million women are diagnosed with breast cancer each year and approximately one-third die from this disease. In China, the crude incidence was estimated to be 42.55 per 100,000 women [3]. Early diagnosis of breast cancer has been clearly shown to reduce mortality and improve survival [4, 5]. Several tumor markers, such as CEA and CA15-3, are known to increase in breast cancer patients [6]. However, the elevated level of the tumor markers can also be due to other causes such as megaloblastic anemia and smoking [7, 8]. With the development of many high-throughput measurement technologies, there have been many studies regarding the diagnosis of BC using high-throughput methods of analysis, such as genomics, proteomics and metabolomics. It has been reported that breast cancer may be associated with specific metabolic signatures that include significant changes in the amino acids profile, as well as large perturbation of the phospholipid metabolism [9]. Here, we focus on metabolomics, namely metabolic profiling.

Metabolomics is the most recently developed of the omics sciences and refers to the study, identification and quantification of metabolites. Metabolites are small molecules ( $<$ 1500 Da) that are the product of metabolic reactions or the consequences of the activity of genes and proteins [10]. In recent years, an increasing number of metabolomics studies focused on the diagnosis of the different cancers including breast cancer have been reported [11, 12, 13, 14]. Shen et al. reported that the altered amino acids and FA $\mathrm{\beta}$ -oxidation were observed in patients with breast cancer [14], while Yang et al. also found that phospholipid composition was altered in the serum of patients with breast cancer [13]. Metabolomics is now considered a powerful technology for the identification of biomarkers and altered metabolic pathways in cancer [15, 16]. Although metabolome analysis is a promising approach in screening for cancer, some practical limitations which regard mainly the profile metabolomic analysis remain. These limitations include the necessity to measure a huge number of lipid metabolites where phospholipids are the abundant endogenous components in serum matrices and can also cause matrix effects and ion suppression in LC-MS analysis [17], data-redundancy problems, including the false-discovery rate and overfitting, and cost constraints. One approach to overcoming these problems is “targeted metabolomics”, which limits the objects of the analysis to those that play roles in general metabolism and share physical similarities. Recently, less study pay attention to the metabolic change of aqueous organic acids in serum because of the above limitation of the metabolomics. Amino acids and organic acids are the important types of the aqueous small molecules in serum. It has reported that metabolism, including that the amino acids metabolism is notably altered in cancer cells [18, 19] and that changes in free amino acids profiles can also occur in breast cancer [20]. Therefore, it is necessary for determination of amino acids profiles and organic acids profiles in breast cancer. These profiles analysis may contribute to the understanding of how amino acids and organic acids react within a biological system and it will be a powerful tool for elucidating the mechanisms of disease, screening novel biomarkers and monitoring clinical pharmacologic therapy.

Thus, the aim of the present study was to investigate whether amino acids profiles and organic acids profiles, established from a simple blood draw from healthy women, benign (BE) patients and BC patients. To this end, a targeted metabolomic analysis, a highly reliable and reproducible technology, was performed in this study. Alterations in the concentrations of certain metabolites detected in the serum of these samples can be candidates as future biomarkers of BC. Likewise, alterations in these metabolites could provide new insights into the pathological heterogeneity of breast tumours.

Table 1
Clinical characteristics of 116 subjects in the discovery set (mean $\pm$ SD) ${}^{\text{a}}$

Parameter	Healthy controls ( $n=$ 34)	BE patients ( $n=$ 38)	BC patients ( $n=$ 44)
Age (years)	45.8 $\pm$ 5.5	45.1 $\pm$ 5.5	46.2 $\pm$ 4.8
Protein (g/d)	84.67 $\pm$ 10.51	82.67 $\pm$ 13.45	85.12 $\pm$ 14.38
Fat(g/d)	105.12 $\pm$ 23.56	101.23 $\pm$ 16.38	102.15 $\pm$ 19.02
Carbohydrate (g/d)	323.23 $\pm$ 78.15	330.12 $\pm$ 87.12	329.01 $\pm$ 85.19
BMI (kg/m ${}^{2}$ )	22.5 $\pm$ 2.3	21.9 $\pm$ 2.5	23.0 $\pm$ 2.1
DBP (mmHg)	75.3 $\pm$ 7.9	75.5 $\pm$ 6.8	76.2 $\pm$ 6.7
SBP (mmHg)	116.0 $\pm$ 9.6	115.5 $\pm$ 8.9	117.5 $\pm$ 9.5
FPG (mmol/L)	5.10 $\pm$ 0.71	4.98 $\pm$ 0.31	4.91 $\pm$ 0.45
TG (mmol/L)	1.74 $\pm$ 0.37	1.70 $\pm$ 0.34	1.64 $\pm$ 0.43
TC (mmol/L)	5.01 $\pm$ 0.33	4.78 $\pm$ 0.47	5.23 $\pm$ 0.82
Age at menarche (in years)	14.4 $\pm$ 1.3	14.3 $\pm$ 1.5	15.1 $\pm$ 1.4
The rate of Postmenopausal women (%)	14.7	13.2	13.6
Age at first birth (in years)	22.2 $\pm$ 10.2	22.1 $\pm$ 11.68	21.5 $\pm$ 8.5
Receptor status, $n$ (%)
ER+/PR+	–	–	22.7% (10/44)
ER+/PR-	–	–	20.4% (9/44)
ER-/PR+	–	–	9.0% (4/44)
ER-/PR-	–	–	18.1% (8/44)
HER2+	–	–	29.5% (13/44)
CA15-3 (U/mL)	6.45 $\pm$ 1.34	7.67 $\pm$ 1.76	20.31 $\pm$ 3.01**
CEA ( $\mu$ g/L)	1.54 $\pm$ 0.39	1.67 $\pm$ 0.42	6.31 $\pm$ 1.71*

${}^{\text{a}}$ : SBP: systolic blood pressure; DBP: diastolic blood pressure; BMI: body mass index; TG: triglycerides; TC: total cholesterol. FBG: fasting plasma glucose; ER: estrogen receptor; PR: progesterone receptor; CA15-3: cancer antigen 15-3; CEA: carcinoembryonic antigen. * $p<$ 0.05, ** $p<$ 0.001, compared with healthy control.

2. Materials and methods

2.1 Study samples

The study was approved by the Ethics Committee of Harbin Medical University and was conducted in accordance with the Declaration of Helsinki. Written informed consent was obtained from each participant. Prior to the study, a 3-day food record was obtained from all participants to exclude any significant differences between the three groups due to their dietary intake (Tables 1 and S1). The enrolled patients were not on any medications before sample collection. All blood samples were collected in the morning before breakfast and fasting blood samples were allowed to clot for 30 min to 1 h at room temperature. And then, blood samples were centrifuged at 3,000 g for 10 min at room temperature, stored at $-$ 80 ${}^{\circ}$ C until analysis.

2.1.1 Subjects of the discovery set

A total of 116 subjects (aged 34–56 years) were recruited from The Affiliated Tumor Hospital of Harbin Medical University, Heilongjiang Province in the North of China. Subjects were divided into the following three groups: BC patients ( $n=$ 44), BE patients ( $n=$ 38) and healthy controls ( $n=$ 34). BE patients suffered from Chromic fibroadenosis of breast. For BC patients, the size of malignant tumors was 1.78 $\pm$ 0.67 cm (range: 0.70–3.50 cm). The cancer staging was carried according to American Joint Committee on Cancer (AJCC). Histologically, the tumors were defined as ductal cancers and there was not cancer metastasis in these patients. These patients were the second grade of invasive ductal carcinoma based on the pathological findings. All study subjects underwent the same diagnostic procedures including a physical breast examination, mammography and ultrasonography as detailed by the American Joint Committee on Cancer staging system [21]. In addition, none of the subjects had a family history of BC or had undergone chemotherapy. Oral contraceptive use and female hormone drugs were not taken by all of the subjects. The normal age-matched controls had no evidence of benign or malignant breast disease.

2.1.2 Subjects of the validation set

We tested findings for validation in the independent samples (90 subjects). These subjects were also recruited from The Affiliated Tumor Hospital of Harbin Medical University and were as a subset of the original collection for the validation. The distribution of 90 subjects was as follows: BC patients ( $n=$ 30; age $=$ 44.8 $\pm$ 4.7 years), BE patients ( $n=$ 30; age $=$ 45.6 $\pm$ 3.2 years) and healthy controls ( $n=$ 30; age $=$ 43.5 $\pm$ 3.6 years). All samples were collected and stored according to the above sample collection.

2.2 Chemicals and reagents

Amino acids and biogenic amines standards were purchased from Sigma (St Louis, MO, USA, $\geqslant$ 99% purity): Alanine, Aminobutyric acid, $\gamma$ -aminobutyric acid, Arginine, Asparagine, Creatine, Creatinine, Cystein, Dimethylglycine, Glycine, L- $\alpha$ -Glycerophosphorylcholine, Glutamic acid, Glutamine, Histidine, 4-Hydroxy-L-proline, Isoleucine, Leucine, Lysine, Methionine, Proline, Phenylalanine, Serine, Taurine, Threonine, Trimethylamine-N-oxide, Tryptophan, Tyrosine, Valine, valine-d ${}_{8}$ and phenylalanine-d ${}_{8}$ . Organic acids standard and derivating agent were also purchased from Sigma (St Louis, MO, USA, $\geqslant$ 99% purity): Aconitic acid, Adipic acid, Capric acid, Caproic acid, Caprylic acid, Citrate acid, Ethylmalonic acid, Fumaric acid, Glycolic acid, Glutaric acids, 2-hydroxybutyric acid, 3-hydroxybutyrate, 2-hydroxyisocaproic acid, Isocitric acid, $\alpha$ -ketoglutaric acid, Lactic acid, Malic acid, Malonic acid, Orotic acid, Oxalic acid, Oxaloacetic acid, Phospho(enol)pyruvic acid, Pimelic acid, Pyroglutamic acid, Pyruvic acid, Sebacic acid, Suberic aicd, Succinic acids and succinate-d ${}_{4}$ . methoxylamine hydrochloride, N-Methyl-N-(trimethylsilyl) trifluoroacetamide (MSTFA). The solution MeOH-Succinate-d ${}_{4}$ (100 $\mu$ g/mL) was freshly prepared (Purity: $\geqslant$ 98.0%) in chromatographic grade methanol. Pyridine (chromatographic grade), acetonitrile (chromatographic grade), formic acid (chromatographic grade), sodium formate (chromatographic grade), sodium sulfate (analytical reagent) were purchased from Tianjin Guangfu Chemical Reagent Co (Tianjin, China).

2.3 Preparation of standard solutions and assessment of the method

2.3.1 Solutions preparation

Stock solutions of the 47 standards were prepared at 1000 $\mu$ g/mL in methanol or water and the stock solutions of lactic acid was prepared at 5000 $\mu$ g/mL in water. The concentrations of working solutions from the internal standards are MeOH-100 $\mu$ g/mL (Succinate-d ${}_{4}$ ) and 100 $\mu$ g/mL (valine-d ${}_{8}$ and phenylalanine-d ${}_{8}$ ). All working solutions of 48 standards as the calibrators samples were prepared at concentrations of different concentration (Table S1). All standard solutions were stored at $-$ 20 ${}^{\circ}$ C until required.

2.3.2 Preparation of quality control sample

Quality control (QC) samples were prepared by mixing equal volume of different individual serum (10 healthy control, 10 BE patients and 10 BC patients). One QC sample was injected at the start of analytical batch and then one QC sample at every fifth injection of samples was analyzed throughout the whole analytical workflow. The replicate of QCs ( $n=$ 3) were run for batch of analysis. The repeatability and reliability of the method were performed by PCA analysis and the coefficients of variation (CV%) of QC samples. In addition, the recoveries were determined in triplicate by spiking different concentrations of the analytes into the serum samples. Intra-day and inter-day precision (% CV) at each metabolites from the same pool sample was expected to be not greater than 15% and intra-day and inter-day accuracy was ranged from 80–120% of each metabolites.

2.4 Biochemical measurements

The serum biochemistry assays were performed on full-automated clinical chemistry analyzer (AUTOLAB PM 4000; AMS Corporation, Rome, Italy) using commercial biochemical kits (Nanjing Jiancheng Bioengineering Institute, Nanjing, China), calibrated with control reagents before sample analysis. The analytes measured included glucose (GLU), total cholesterol (TC), triglyceride (TG), High-density lipoprotein cholesterol (HDL-C), and low-density lipoprotein cholesterol (LDL-C). Cancer antigen 15-3 (CA15-3) concentrations were analyzed using EIA kit (ROCHE Diagnostics, Basel, Switzerland) and expressed as Units/L. Carcinoembryonic antigen (CEA) levels were assayed by ELISA kit (ROCHE Diagnostics, Basel, Switzerland ) and expressed as $\mu$ g/L.

2.5 Serum amino acids and biogenic amines profiles analysis

2.5.1 Serum preparation for amino acids and biogenic amines profiles

Serum amino acids and biogenic amines profiles were prepared as previously describe [20]. Briefly, each 50 $\mu$ L serum sample was used for metabolite extraction before UPLC/MS-MS analysis. The metabolite extraction procedure was carried out after adding 250 $\mu$ L of acetonitrile/methanol/formic acid (74.9:24.9:0.2 v/v/v) containing two additional stable isotope-labeled internal standards for valine-d ${}_{8}$ and phenylalanine-d ${}_{8}$ in serum. After vortexing for 1 min, the mixture was kept at room temperature for 10 min, and centrifuged at 14,000 g for 10 min at 4 ${}^{\circ}$ C. The supernatant were transferred to the vial tube. The solution was filtered through a syringe filter (0.22 $\mu$ m) and placed into the sampling vial pending UHPLC/MS-MS analysis.

2.5.2 UPLC/MS-MS analysis

UPLC/MS-MS analysis was performed by using a Waters ACQUITY UPLC system (Waters Corporation, Milford, MA, USA) coupled to a Waters Xevo TQD Mass Spectrometer (Waters Corporation, Manchester, UK). The UHPLC/MS-MS method was performed as our previously described [20]. In briefly, a part (2 $\mu$ L) of the sample solution was injected into an ACQUITY UPLC™ HILIC column (100 mm $\times$ 2.1 mm i.d., 1.7 $\mu$ m; Waters Corporation, Milford, MA, USA). The flow rate of the mobile phase was 300 $\mu$ L/min. Analytes were recovered from the column with a gradient elution which is described in our previous publication [22].

MS analyses were made use of electrospray ionization (ESI) and Selected Reaction Monitoring (SRM) scans in the positive ion mode. In each transition, cone voltage and collision energies were optimized, the ion spray voltage was 3.2 kV, and the source temperature was 150 ${}^{\circ}$ C. Internal standard peak areas (valine-d ${}_{8}$ and phenylalanine-d ${}_{8}$ ) were monitored for quality control and individual samples with peak areas differing from the group mean by more than two standard deviations were reanalyzed. MarkerLynx Application Manager software (Version 4.1; Waters Corporation, Milford, MA, USA) was used for automated peak integration and metabolite peaks were manually reviewed for quality of integration [21].

2.6 Serum organic acid profiles analysis

2.6.1 Sample preparation for organic acids

Serum organic acids profiles were prepared as previously described [23]. Briefly, Aliquots (100 $\mu$ L) of serum were spiked with internal standard (I.S.) working solution (9 $\mu$ L MeOH-Succinate-d ${}_{4}$ 100 $\mu$ g/mL), and 300 $\mu$ L acetone was then added to deposit protein at $-$ 20 ${}^{\circ}$ C for 2 h. The solutions were centrifuged at 15,000 rpm for 15 min at 4 ${}^{\circ}$ C to collect the supernatant. The supernatant were dried under N ${}_{2}$ and derivatized to rapidly form silyl derivatives. Briefly, in order to protect ketone groups [24], we added 30 $\mu$ L of methoxylamine hydrochloride dissolved in pyridine (30 mg/mL) to each sample, which was then incubated for 1.5 h at 37 ${}^{\circ}$ C with agitation. Then, 45 $\mu$ L of N-Methyl-N-(trimethylsilyl)trifluoroacetamide (MSTFA)/acetonitrile (1:2) was added and samples were agitated for 10 min. The samples were placed in the dark for 1 h at 40 ${}^{\circ}$ C and transferred into a vial before immediate analysis.

2.6.2 GC-MS/MS analysis

GC-MS/MS analysis used using a TRACE gas chromatograph with a Polaris Q mass spectrometer (Thermo Finnigan, Austin, TX, USA). Helium was used as the carrier gas. A split injector (the split ratio being 1:10) at 250 ${}^{\circ}$ C was used to add the sample (1.0 $\mu$ L) onto a J&W DB-5 (30 m $\times$ 0.25 mm I.D, 0.25- $\mu$ m film thickness) capillary column. Organic acids were separated in constant flow by the following oven program: (a) initially 70 ${}^{\circ}$ C for 2 min; (b) temperature was increased up to 116 ${}^{\circ}$ C at the rate of 4 ${}^{\circ}$ C/min and maintained at 116 ${}^{\circ}$ C for 2 min; (c) increased up to 170 ${}^{\circ}$ C at the rate of 4 ${}^{\circ}$ C/min and maintained at 170 ${}^{\circ}$ C for 2 min; (d) increased up to 177 ${}^{\circ}$ C at the rate of 4 ${}^{\circ}$ C/min and maintained at 177 ${}^{\circ}$ C for 2 min; (e) increased up to 180 ${}^{\circ}$ C at the rate of 4 ${}^{\circ}$ C/min and maintained at 180 ${}^{\circ}$ C for 2 min; (f) increased up to 220 ${}^{\circ}$ C at the rate of 4 ${}^{\circ}$ C/min and maintained at 220 ${}^{\circ}$ C for 2 min. The temperature of transfer line was maintained at 250 ${}^{\circ}$ C. The ion trap mass spectrometer was operated under electron bomb ionization (EI) mode. Mass spectra of m/z 30–450 were collected by full scan mode with 0.58 s/scan velocity. Solvent delay time was 5 min. The source temperature was 230 ${}^{\circ}$ C with the electron energy at 70 eV.

2.7 Statistical analysis

2.7.1 Statistical analysis

Statistical analysis was carried out using SPSS 13.0 software (SPSS, Chicago, IL, USA). All data were presented as means $\pm$ SD. If data were not normally distributed, the date were logarithmically transformed to obtain normal distribution before analysis. Continuous variables were analyzed by one-way ANOVA with Tukey’s test. $P<$ 0.05 was considered to be statistically significant. Receiver operator characteristic (ROC) curve analysis was performed to determine the areas under the curve (AUC) as a measure for comparing the predictive ability of some different metabolites.

2.7.2 Multivariate analysis

The amino acids and organic acids profiles data were analyzed by multivariate statistical analysis using SIMCA-P 13.0 software (Umetrics, Umeå, Sweden). All data were standardized to a mean of 0 and a variance of 1, according to the following formula: (X – mean(X))/std(X) for multivariate statistical analysis. Unsupervised principal component analysis (PCA) was used first in all samples to see the general separation. PCA is a bilinear decomposition method used for overview ‘clusters’ within multivariate data. The score plot of PC1 versus PC2 was used to examine separation or clusters for the three groups.

Table 2
Quantitative analysis of amino acids and biogenic amines profiles in three groups

Amino acids and biogenic amines ( $\mu$ mol/L)	Healthy controls ( $n=$ 34)	BE patients ( $n=$ 38)	BC patients ( $n=$ 44)
Alanine	540.65 $\pm$ 105.30	522.30 $\pm$ 80.97	597.73 $\pm$ 132.38* ${}^{{\dagger}}$
Aminobutyric acid	18.86 $\pm$ 5.49	15.37 $\pm$ 6.50*	24.51 $\pm$ 9.16** ${}^{{\dagger}{\dagger}}$
$\gamma$ -aminobutyric acid	48.93 $\pm$ 7.87	46.91 $\pm$ 7.06	41.81 $\pm$ 14.68* ${}^{{\dagger}}$
Arginine	2117.39 $\pm$ 659.51	1723.15 $\pm$ 604.84*	1722.50 $\pm$ 352.86**
Asparagine	24.47 $\pm$ 7.31	21.37 $\pm$ 5.41*	30.15 $\pm$ 10.42* ${}^{{\dagger}}$
Creatine	14.11 $\pm$ 5.58	15.33 $\pm$ 4.93	16.11 $\pm$ 7.51
Creatinine	81.07 $\pm$ 18.58	89.69 $\pm$ 18.51*	102.36 $\pm$ 37.90** ${}^{{\dagger}{\dagger}}$
Cystein	0.64 $\pm$ 0.22	0.81 $\pm$ 0.45	0.84 $\pm$ 0.38*
Dimethylglycine	7.14 $\pm$ 2.94	10.89 $\pm$ 4.70**	9.26 $\pm$ 3.02**
Glycine	366.65 $\pm$ 105.57	312.55 $\pm$ 68.71*	424.49 $\pm$ 104.86* ${}^{{\dagger}}$
L- $\alpha$ -Glycerophosphorylcholine	4.56 $\pm$ 2.56	3.98 $\pm$ 1.43	3.88 $\pm$ 2.73
Glutamic acid	158.79 $\pm$ 51.27	206.61 $\pm$ 45.32**	287.95 $\pm$ 115.06** ${}^{{\dagger}{\dagger}}$
Glutamine	234.81 $\pm$ 194.04	362.72 $\pm$ 119.53**	358.28 $\pm$ 121.67**
Histidine	168.85 $\pm$ 70.03	187.75 $\pm$ 41.88	202.79 $\pm$ 61.39*
4-Hydroxy-L-proline	14.66 $\pm$ 6.25	15.51 $\pm$ 6.41	14.01 $\pm$ 5.10
Isoleucine	32.54 $\pm$ 5.69	29.90 $\pm$ 5.19*	36.12 $\pm$ 7.57* ${}^{{\dagger}{\dagger}}$
Leucine	21.61 $\pm$ 3.09	20.95 $\pm$ 4.43	24.82 $\pm$ 6.01
Lysine	334.81 $\pm$ 83.68	311.24 $\pm$ 84.66	328.65 $\pm$ 63.16
Methionine	3.17 $\pm$ 0.5	4.05 $\pm$ 0.6	6.4 $\pm$ 1.44** ${}^{{\dagger}{\dagger}}$
Proline	183.19 $\pm$ 71.03	201.54 $\pm$ 63.96	218.79 $\pm$ 78.20*
Phenylalanine	108.87 $\pm$ 21.08	106.37 $\pm$ 21.55	131.26 $\pm$ 31.74** ${}^{{\dagger}{\dagger}}$
Serine	163.01 $\pm$ 29.35	160.75 $\pm$ 35.10	185.77 $\pm$ 41.62* ${}^{{\dagger}}$
Taurine	97.57 $\pm$ 25.51	85.93 $\pm$ 31.21*	143.86 $\pm$ 40.15** ${}^{{\dagger}{\dagger}}$
Threonine	194.33 $\pm$ 48.925	204.09 $\pm$ 49.84	193.86 $\pm$ 53.57
Trimethylamine-N-oxide	0.39 $\pm$ 0.11	0.31 $\pm$ 0.06	0.43 $\pm$ 0.12
Tryptophan	75.82 $\pm$ 13.22	81.63 $\pm$ 17.28	77.33 $\pm$ 14.82
Tyrosine	98.36 $\pm$ 26.69	102.77 $\pm$ 23.93	118.03 $\pm$ 38.69* ${}^{{\dagger}}$
Valine	285.95 $\pm$ 36.26	287.76 $\pm$ 52.68*	327.43 $\pm$ 82.18* ${}^{{\dagger}}$
glutamate/glutamine ratio	0.68 $\pm$ 0.26	0.57 $\pm$ 0.37*	0.80 $\pm$ 0.64* ${}^{{\dagger}}$

$P$ -values were calculated by ANOVA analysis. * $p<$ 0.05, ** $p<$ 0.001, BC patients or BE patients were compared with healthy control. ${}^{{\dagger}}p<$ 0.05, ${}^{{\dagger}{\dagger}}p<$ 0.001, BC patients were compared with BE patients.

Then, partial least-squares-discriminant analyses (PLS-DA), which is a supervised classification method, is based on the PLS approach [25]. The standard PLS algorithm can be used for the dependent y vector and class labels can be used. In the two-class case the values of the dependent variable are usually assigned a 1 for one class and a 0 or $-$ 1 for the other class [26] . In the case of more than 2 classes, dummy variables are defined and a PLS algorithm is used [26]. Furthermore, to guard against model over-fitting, the supervised models were validated with a 200 times permutation test. In other words, the order of Y was randomly permuted a number of times (200 times) and separate models were fitted to all the permuted Y’s extracting as many components as was done with the original matrix Y. Metabolites that contributed to sample classification were identified by their values of variable importance in the projection [25] (VIP, $>$ 1.5) and false discovery rate (FDR) in PLS-DA model. Loading plot was used to visualize the variable effects in PLS-DA models. FDR was implemented to correct for multiple testing by R-package fdrtool. The different metabolites were mapped through their respective metabolic pathways using MetaboAnalyst 3.0 (http://www.metaboanalyst.ca).

3. Results

3.1 Method validation

We used QC samples to check the UPLC-MS/MS and GC-MS/MS system for reproducibility and reliability. PCA performed on QC samples and the other groups revealed that the QC samples were clustered in PCA scores plot (Supplemental Fig. S1). Fifty four metabolites from QC samples were also used to assess for method reproducibility and reliability. The relative standard deviations (RSD%) of the metabolites were showed to be 2.44–9.13% (Supplemental Table S1). Intra-day and inter-day accuracies and precisions of this method were determined by analyzing three replicates. The precisions ranged from 1.11% to 14.91%, whereas accuracies ranged from 80.14% to 118.70% (Supplementary Table S1). The above results showed that this method had good repeatability, reliability and stability for metabolite analysis.

Figure 1.

Score plot from PCA analysis of the three groups (R2X ${}_{\text{(cum)}}$ $=$ 0.352, Q2 ${}_{\text{(cum)}}$ $=$ 0.102).

3.2 Clinical characteristics of the subjects

There were no significant differences between the three groups with respect to age, smoking and alcohol consumption (Table 1). Some clinical characteristics including body mass index (BMI), TG, TC, fasting blood glucose (FBG), systolic blood pressure (SBP) and diastolic blood pressure (DBP) did not differ significantly between the three groups (ANOVA, $P>$ 0.05). Moreover, age at menarche, the number of postmenopausal women and age at first birth were not significantly different between the healthy controls, BE patients and BC patients (Table 1). There were higher levels of CA15-3 and CEA in BC group when compared with the healthy controls and BE patients.

3.3 Changes in the amino acids and biogenic amines profiles

We performed serum amino acids and biogenic amines profiles using UPLC-MS/MS in three groups. Twenty-eight metabolites were detected in all subjects (Table 2). The levels of 20 metabolites changed significantly in the BC group when compared to the control group (Fig. 1, $P<$ 0.05). For 18 of these metabolites ( $P<$ 0.05), the mean levels in the serum were significantly higher in the BC patients than the controls, while two metabolites ( $\gamma$ -aminobutyric acid and arginine) were significantly lower in the BC patients than the controls. Significant differences in the levels of the 11 amino acids and biogenic amines were observed between the BE patients and controls ( $P<$ 0.05) (Table 2) (Supplemental Fig. S2). In addition, significant differences in 15 metabolites ( $P<$ 0.05) were observed between the BE and BC individuals (Supplemental Fig. S2).

3.4 Changes in the organic acids profiles

The quantitative results of organic acids profiles analysis in the three groups are presented in Table 2. Comparison of 28 organic acids concentrations measured in serum from the controls, BE patients and BC patients revealed substantial differences. Ten of the organic acids (adipic acid, caproic acid, ethylmalonic acid, glycolic acid, glutaric acids, 2-hydroxybutyric acid, 2-hydroxyisocaproic acid, pyruvic acid, suberic acid, succinic acids) in the BC patients significantly decreased when compared with the healthy control. Six organic acid (adipic acid, caproic acid, glycolic acid, glutaric acids, 2-hydroxybutyric acid, succinic acids) in the BC group were lower than those of BE group. There were significant differences in three organic acids (ethylmalonic acid, 2-hydroxyisocaproic acid, succinic acids) between the healthy control and BE subjects ( $P<$ 0.001). The above results showed that there were different organic acids profiles in three groups.

Figure 2.

PLS-DA scores plot between two sub-groups. Circle (green), healthy control; Triangle (red), BE; Box (blue), BC. (A) PLS-DA scores plot for healthy control and BE groups (R2X ${}_{\text{(cum)}}$ $=$ 0.218, R2Y ${}_{\text{(cum)}}$ $=$ 0.825, Q2 ${}_{\text{(cum)}}$ $=$ 0.424); (B) PLS-DA scores plot for control and BC groups (R2X ${}_{\text{(cum)}}$ $=$ 0.313, R2Y ${}_{\text{(cum)}}$ $=$ 0.912, Q2 ${}_{\text{(cum)}}$ $=$ 0.82); (C) PLS-DA scores plot for BE and BC groups (R2X ${}_{\text{(cum)}}$ $=$ 0.316, R2Y ${}_{\text{(cum)}}$ $=$ 0.892, Q2 ${}_{\text{(cum)}}$ $=$ 0.762).

Figure 3.

Loadings plots of free fatty acids from PLS-DA models. (A) Healthy control and BE; (B) Healthy control and BC; (C) BE and BC Corresponding loadings plot; possible biomarkers were marked with a red.

3.5 Multivariate data statistical analysis

Amino acids and biogenic amines, and organic acids profiles were analyzed by PCA. The score plot of PCA represented the distribution of all samples (Fig. 1A). The first two components of the model (PC1 and PC2) explained 35.2% of the variance (R2 $=$ 0.352, and Q2 $=$ 0.102). As shown in Fig. 1A, many samples in the different groups were scattered and could not be completely separated.

After crude screening by PCA, a PLS-DA model was constructed for classification of the three groups. In order to screen the important metabolites, three sub-comparisons using PLS-DA were performed to compare the healthy controls and BE patients, the healthy controls and BC patients, and BE and BC patients. The results of PLS-DA showed that the samples in BC patients versus the healthy controls and BE individuals (Fig. 2B and C) were separated clearly; there was also acceptable separation in the comparison of healthy controls and BE individuals (Fig. 2A). There were high Q2 values of in different models (Q2 $>$ 0.5). Therefore, these models were suitable for the recognition analysis of two sub-comparisons. Based on the permutation test, all of the R2 ${}_{\text{(cum)}}$ and Q2 ${}_{\text{(cum)}}$ values were lower than the original values in the validation plot (Supplemental Fig. S3). These results assured the validity of the PLS-DA models between the three sub-comparisons.

Table 3
Quantitative analysis of organic acids profiles in three groups

Organic acids ( $\mu$ mol/L)	Healthy controls ( $n=$ 34)	BE patients ( $n=$ 38)	BC patients ( $n=$ 44)
Aconitic acid	0.30 $\pm$ 0.09	0.21 $\pm$ 0.14	0.23 $\pm$ 0.11
Adipic acid	7.17 $\pm$ 2.05	6.90 $\pm$ 2.43	5.43 $\pm$ 1.24* ${}^{{\dagger}}$
Capric acid	4.30 $\pm$ 1.56	14.23 $\pm$ 1.28	4.60 $\pm$ 1.49
Caproic acid	417.57 $\pm$ 108.11	391.78 $\pm$ 137.21	340.55 $\pm$ 53.34** ${}^{{\dagger}}$
Caprylic acid	73.84 $\pm$ 6.54	72.83 $\pm$ 1.11	73.57 $\pm$ 2.29
Citric acid	0.94 $\pm$ 0.30	0.91 $\pm$ 0.29	0.88 $\pm$ 0.29
Ethylmalonic acid	1124.58 $\pm$ 257.68	875.01 $\pm$ 225.15**	854.74 $\pm$ 252.99**
Fumaric acid	5.58 $\pm$ 2.01	5.28 $\pm$ 2.19	5.91 $\pm$ 1.81
Glycolic acid	64.63 $\pm$ 7.55	62.98 $\pm$ 13.96	55.08 $\pm$ 7.47** ${}^{{\dagger}{\dagger}}$
Glutaric acid	4.51 $\pm$ 1.90	4.60 $\pm$ 2.69	3.57 $\pm$ 1.67* ${}^{{\dagger}}$
2-hydroxybutyric acid	117.57 $\pm$ 55.32	118.99 $\pm$ 30.95	95.81 $\pm$ 29.29* ${}^{{\dagger}}$
3-hydroxybutyrate	251.69 $\pm$ 110.36	211.22 $\pm$ 58.83	240.19 $\pm$ 102.16
2-hydroxyisocaproic acid	3625.62 $\pm$ 1565.91	2899.80 $\pm$ 1085.98*	3037.07 $\pm$ 1220.10*
Isocitric acid	0.0016 $\pm$ 0.0003	0.0014 $\pm$ 0.0005	0.0012 $\pm$ 0.0004
$\alpha$ -ketoglutaric acid	224.54 $\pm$ 7.84	222.81 $\pm$ 1.17	222.75 $\pm$ 2.17
Lactic acid	6625.81 $\pm$ 1399.37	6051.09 $\pm$ 1424.17	6443.81 $\pm$ 968.86
Malic acid	116.42 $\pm$ 47.49	119.21 $\pm$ 72.41	117.08 $\pm$ 61.94
Malonic acid	1.56 $\pm$ 0.26	1.53 $\pm$ 0.13	1.51 $\pm$ 0.14
Orotic acid	10.01 $\pm$ 1.59	10.08 $\pm$ 2.48	10.40 $\pm$ 3.28
Oxalic acid	80.47 $\pm$ 1.14	80.18 $\pm$ 0.70	80.09 $\pm$ 0.71
Oxaloacetic acid	73.78 $\pm$ 13.69	68.73 $\pm$ 19.49	73.37 $\pm$ 14.96
Phospho(enol)pyruvic acid	6.15 $\pm$ 0.77	5.98 $\pm$ 0.55	6.52 $\pm$ 1.74
Pimelic acid	83.66 $\pm$ 3.25	83.65 $\pm$ 3.94	81.49 $\pm$ 1.82
Pyroglutamic acid	236.16 $\pm$ 91.62	234.68 $\pm$ 137.33	262.19 $\pm$ 171.99
Pyruvic acid	142.57 $\pm$ 70.15	126.79 $\pm$ 38.07	118.64 $\pm$ 30.04*
Sebacic acid	0.59 $\pm$ 0.18	0.42 $\pm$ 0.26	0.61 $\pm$ 0.35
Suberic acid	21.67 $\pm$ 13.37	21.01 $\pm$ 15.78	16.21 $\pm$ 9.73*
Succinic acid	17.41 $\pm$ 5.53	16.14 $\pm$ 4.93*	13.79 $\pm$ 4.67** ${}^{{\dagger}}$

Table 4

Important metabolites based on the statistical analysis in three groups

Groups	Biomarkers	VIP values ${}^{\text{a}}$	$P$ -values ${}^{\text{b}}$	FDR	FC ${}^{\text{c}}$	AUC ${}^{\text{d}}$
Healthy control-BE patients	Glutamic acid	2.76	$<$ 0.001	0.033	1.30	0.868
	Glutamine	1.75	0.001	$<$ 0.001	1.54	0.757
	Dimethylglycine	1.99	$<$ 0.001	$<$ 0.001	1.53	0.770
Healthy control-BC patients	Taurine	2.04	$<$ 0.001	$<$ 0.001	1.47	0.834
	Glutamic acid	1.57	$<$ 0.001	0.045	1.81	0.751
	Ethylmalonic acid	1.54	$<$ 0.001	0.042	0.76	0.779
BE patients-BC patients	Taurine	2.21	$<$ 0.001	$<$ 0.001	1.67	0.862
	Glycine	1.69	$<$ 0.001	$<$ 0.001	1.36	0.844
	Aminobutyric acid	1.61	$<$ 0.001	0.037	1.59	0.789

${}^{\text{a}}$ Variable importance in the projection (VIP) was obtained from PLS-DA with a threshold of 1.5. ${}^{\text{b}}$ The $p$ -value was calculated by ANOVA analysis. ${}^{\text{c}}$ Fold change was calculated from the arithmetic mean values of each group. ${}^{\text{d}}$ The areas under the curve (AUC) of receiver operator characteristic curve analysis.

3.6 Potential different metabolites

In this work, a PLS-DA model and univariate statistical analysis was used for classification of healthy control, BE and BC patients and the identification of potential biomarkers. Firstly, the possible biomarker candidates were filtered using VIP values (VIP $>$ 1.5) in the PLS-DA model. Secondly, ANOVA and FDR analysis ( $P<$ 0.05) were used to analyze the possible biomarker candidates. Finally, the fold change (FC $>$ 1.2 or $<$ 0.83) and the ROC analysis (AUC $>$ 0.75) of these selected metabolites were performed (Table 4). It is interesting that we identified taurine as potential metabolites for identification of BC patients versus non-BC groups (healthy control and BE subjects). Regarding the controls and BC patients, three important metabolites (Glutamic acid, taurine and ethylmalonic acid) were selected as the potential metabolites (Table 4). The results of Table 4 show that three metabolites including glutamic acid, ethylmalonic acid and dimethylglycine were used to distinguish the controls to BE patients. And for BE versus BC, the VIP values of taurine, glycine and aminobutyric acid were more than 1.5. And there were significant differences in the concentration of these metabolites between BE and BC individuals, so taurine, glycine and aminobutyric acid were selected as potential metabolites distinguished with BE from BC patients. Moreover, the potential metabolites can also be identified from the PLS-DA model loadings plot. The loadings plot (Fig. 3) identified the variables that strongly contribute to the separation of classes. In turn, these metabolites might be biomarker candidates. From the loadings plot, we can also detect the potential metabolites furthest from the origin. The potential metabolites found in the loadings plot were in agreement with those identified by the above results. Thus, patients who have higher level of taurine, glutamic acid and ethylmalonic acid are likely to be diagnosed with BC patients, and it is necessary to determinate dimethylglycine levels to help discriminate the healthy control and BE patients.

Table 5
Pathways modulated by breast cancer condition

	Total	Expected	Hits	$P$	FDR	Impact
Aminoacyl-tRNA biosynthesis	75	0.8413	15	$<$ 0.0001	$<$ 0.0001	0.11268
Alanine, aspartate and glutamate metabolism	24	0.26921	6	$<$ 0.0001	$<$ 0.0001	0.48623
Taurine and hypotaurine metabolism	20	0.22435	4	$<$ 0.0001	0.001	0.38489
Glycine, serine and threonine metabolism	48	0.53843	5	$<$ 0.0001	0.002	0.35798
Arginine and proline metabolism	77	0.86373	6	$<$ 0.0001	0.002	0.2709
Cysteine and methionine metabolism	56	0.62817	5	$<$ 0.0001	0.004	0.19337

Figure 4.

Quantitative analysis of taurine, glutamic acid and ethylmalonic acid in the validation study. $P$ -values were calculated by ANOVA analysis. ** $P<$ 0.001. $P$ values were calculated by comparing with controls. ${}^{{\dagger}{\dagger}}P<$ 0.001, compared with BE samples.

Figure 5.

ROC analysis for discrimination of BC and the healthy control by important metabolites in the validation set.

Figure 6.

The topology analysis of pathway for the important metabolites.

3.7 The important metabolites in the validation set

We measured the important metabolites which were able to distinguish BC patient from the other subjects in the discovery set (Table 4) from the independent healthy controls, BE patients and BC patients to validate the above results. Compared with healthy controls and BE patients, the increased taurine, glutamic acid and decreased ethylmalonic acid in BC patients was observed in Fig. 4. The results of ROC analysis (Fig. 5) showed that the AUC values for taurine, glutamic acid and ethylmalonic acid were 0.901, 0.924 and 0.749 for BC patients and the controls. The sensitivity and specificity for prediction were higher than 70% in the ROC curve (data not shown). The above results suggested that BC patients with higher level of taurine glutamic acid and the lower ethylmalonic acid were likely to be diagnosed with BC regardless of the controls.

3.8 Pathway analysis

In this work, 29 differential metabolites were imported into the MetaboAnalyst 3.0. As a result, the web-based software generated 6 metabolic pathways with a FDR, $P$ value $<$ 0.01 and impact $>$ 0.1, which were supposed to be significantly altered in BC patients (Table 5). In these important metabolic pathway, amino acids metabolism and taurine metabolism were the most important metabolic pathways disturbed BC patients (Fig. 6).

4. Discussion

In this study, we demonstrated that the targeted metabolomics technology could be used to discriminate BC patients, BE patients and the controls based on serum data analyzed by UPLC-MS/MS and GC-MS/ MS. Recently, untargeted metabolomics technology has been used to analyze significant metabolic variations in BC patients. Lécuyer et al. [27] found that elevated plasma levels of lipoproteins and lipids in BC patients were observed based on the NMR metabolomics. Furthermore, Takayama et al. [28] found that polyamines have been significantly associated with rapid tumor growth and appears to be useful for the diagnosis of breast cancer patients. The above studies showed that there was the alternation of lipids and polyamines metabolism. Our results showed that some amino acids and organic acids were differently expressed in the three groups, and glutamic acid, taurine and ethylmalonic acid were identified as the potential metabolites of BC patients. Moreover, the determination of the glutamic acid, ethylmalonic acid and dimethylglycine is helpful for distinguishing the healthy control from BE patients.

One remarkable observation in this work is that the serum amino acids profiles were significantly changed in BC patients. Amino acids play essential physiological roles both as basic metabolites and metabolic regulators. Serum amino acids abundantly circulate as a medium linking all organ systems and amino acids profiles have been known to be influenced by metabolic variations in specific organ systems induced by specific diseases [29, 30]. Dowling et al. reported that the amino acids profiles characterized higher glutamate, isoleucine and proline changed in breast cancer [31]. Consistent with this study, we also found that there were the changed amino acids profiles and increased glutamate and glutamine in BC patients. Simultaneously, the result of the pathway analysis showed that the metabolic change of the alanine, aspartate and glutamate metabolism were observed in BC patients (Fig. 6). Glutamate is a product of the glutamine under the action of the phosphate-dependent glutaminase (GLS), an enzyme found within the inner mitochondrial membrane. It has been shown that there is the overexpression of GLS in both solid tumours and cell line models of cancer [32, 33]. Other groups have reported on elevated glutamate levels and associated signaling pathways in malignancies [34], and the increased rates of glutamine metabolism through the accelerated hydrolysis of glutamine to glutamate, as catalyzed by mitochondrial GLS activity in breast cancer [31]. Our findings support the metabolic changes in cancer metabolism resulted in a shift to increased rates of glutamine metabolism.

We also found taurine was also importantly different metabolite distinguishing BC patients from non-breast cancer groups (The controls and BE patients). Meanwhile, taurine and hypotaurine metabolism is importantly metabolism in breast cancer. The taurine and hypotaurine metabolism has been shown to be relevant to women cancers, such as ovarian and breast cancer [35, 36]. Here, we also have discovered that taurine and hypotaurine metabolism is also dysregulated in the blood samples of breast cancer. Many lines of evidence suggest taurine and hypotaurine metabolism is a critical pathway in tumor development [35, 36]. Taurine, which is a sulfuric amino acid, can be synthesized by the human body from cysteine. Taurine has many diverse biological functions that serve as stabilizers of cell membranes and anti-oxidants. Much evidences support the role of reactive oxygen species (ROS) and oxidative stress in development of breast cancer [37]. ROS, by attacking DNA, can induce DNA damage which in turn can either inhibit or induce replication errors and genomic instability. All these effects are known to be associated with carcinogenesis [38]. In this study, the elevated taurine, as an anti-oxidants, has high an AUC of 0.941 which diagnosed the BC patients in the validation set. The elevated content in taurine is also essential for the antioxidant glutathione [39] and taurine was increased in plasma from cancer subjects could be involved in protecting cancer cells from excessive damage by oxidative stress [40]. These studied further confirmed our results. In short, the taurine and hypotaurine metabolism significantly changed in BC patients, which is also confirmed by the other studies.

Previous metabolomics studies have shown that certain metabolites can differentiate breast cancer patients from normal control [41, 42]. For example, glutamate was found enriched and the glutamate-to-glutamine ratio was significantly correlated with estrogen receptor status in breast cancer patients [41]. Our result confirmed that there were higher glutamate and glutamate-to-glutamine ratio in breast cancer than that of the healthy control (Table 2). Moreover, histidine, glucose, and lipids were strongly correlated with breast cancer relapse with a predictive accuracy of 75% based on the serum profiles [42]. However, although TCA cycle key enzyme genes may serve as potential biomarkers to predict the outcomes of lung cancer [43], less attention is paid to the serum organic acids profiles in breast cancer. Our data suggested that the alterations of organic acids profiles could reflect underlying metabolic changes taking place during BC development. The present study has highlighted ethylmalonic acid as the organic acid with the greatest potential to act as biomarkers of BC. Animal experiments suggested that ethylmalonic acid is related to an inhibition of creatine kinase activity and induction of lipid peroxidation in hippocampus [44]. Notably, the serum ethylmalonic acid in cancer is not reported in the literature. So the reason of decreased ethylmalonic acid in BC patients will be investigated in the future study. Together, the significant increase in the serum levels of these biomarkers observed in BC patients could possibly indicate that breast cancer is accompanied by, or precipitated by, an underlying metabolic disorder affecting organic acids metabolism.

There are limitations of this study. Firstly, although our results have been validated in the independent samples, the number of studied subjects was relatively small. Thus, care must be exercised in the application of our findings to larger populations of BC patients. Secondly, the relevance of the change of organic acids profiles in the pathoetiology of BC could not be explained. Furthermore, multiple controls, such as other types of women cancer including cervical cancer, ovarian cancer, and so on, would be needed to evaluate whether the selected amino acids and organic acids biomarkers is specific to BC. Thirdly, further studies are required to explore the metabolic mechanisms of organic acids profiles in breast cancer and to validate these results in the human larger population.

In conclusion, we performed a targeted metabolomic investigation for an understanding of breast cancer metabolism. Metabolism of BC patients, as expected, differs a lot from that of BE patients and the controls. Metabolic reprogramming in BC patients, mainly characterized by changed taurine and hypotaurine metabolism, disturbed metabolism of organic acids, abnormal amino acid metabolism, is found to be related with BC development and progression. Taurine, glutamic acid and ethylmalonic acid as different metabolites are identified that could facilitate early diagnosis and clinical staging for EOC. In all, targeted metabolomics studies from breast cancer patients could be useful to describe diagnostic and/or prognosis biomarkers, as well as for monitoring treatment.

Footnotes

Acknowledgments

Natural Science Foundation of Heilongjiang Province for the returned students (No. LC2016030), The Youth Science Foundation of Liande Wu in Harbin Medical University (No. WLD-QN1704), Program for the innovation talent in Harbin Science and Technology Bureau (No. 2017RAQXJ169).

Conflict of interest

The authors declare they have no actual or potential competing conflict of interest.

References

Jemal

Siegel

and Ward

, Cancer statistics, 2010, CA Cancer J Clin 60 (2010), 277–300.

Ferlay

Steliarovafoucher

Lortettieulent

Rosso

Coebergh

J.W.

Comber

Forman

and Bray

, Cancer incidence and mortality patterns in Europe: Estimates for 40 countries in 2012, European Journal of Cancer 49 (2013), 1374–1403.

Chen

Zheng

Zhang

Zhao

and He

, The incidences and mortalities of major cancers in China, 2009, Chin J Cancer 32 (2013), 106–12.

Breen

, Effect of screening and adjuvant therapy on mortality from breast cancer – NEJM, New England Journal of Medicine 353 (2005), 1784–1792.

Rosmawati

N.H.

, Knowledge, attitudes and practice of breast self-examination among women in a suburban area in Terengganu, Malaysia, Asian Pacific Journal of Cancer Prevention Apjcp 11 (2010), 1503.

Keshaviah

Dellapasqua

Rotmensz

Lindtner

Crivellari

Collins

Colleoni

Thürlimann

Mendiola

and Aebi

, CA15-3 and alkaline phosphatase as predictors for breast cancer recurrence: A combined analysis of seven International Breast Cancer Study Group trials, Annals of Oncology 18 (2007), 701–708.

Symeonidis

Kouraklis-Symeonidis

Apostolopoulos

Arvanitopoulou

Giannakoulas

Vassilakos

and Zoumbos

, Increased serum CA-15.3 levels in patients with megaloblastic anemia due to vitamin B12 deficiency, Oncology 67 (2004), 359–67.

Kashiwabara

Nakamura

and Yokoi

, Chronological change of serum carcinoembryonic antigen (CEA) concentrations and pulmonary function data after cessation of smoking in subjects with smoking-associated CEA abnormality, Clin Chim Acta 303 (2001), 25–32.

Corona

Polesel

Fratino

Miolo

Rizzolio

Crivellari

Addobbati

Cervo

and Toffoli

, Metabolomics biomarkers of frailty in elderly breast cancer patients, J Cell Physiol 229 (2014), 898–902.

10.

Roig

Rodríguezbalada

Samino

Lam

E.W.

Guaitaesteruelas

Gomes

A.R.

Correig

Borràs

Yanes

and Gumà

, Metabolomics reveals novel blood plasma biomarkers associated to the BRCA1-mutated phenotype of human breast cancer, Scientific Reports 7 (2017), 17831.

11.

Asiago

V.M.

Alvarado

L.Z.

Shanaiah

Gowda

G.A.

Owususarfo

Ballas

R.A.

and Raftery

, Early detection of recurrent breast cancer using metabolite profiling, Cancer Research 70 (2010), 8309.

12.

Hart

C.D.

Vignoli

Tenori

G.L.

Van To

Adebamowo

Hossain

S.M.

Biganzoli

Risi

Love

R.R.

Luchinat

and Di Leo

, Serum metabolomic profiles identify ER-positive early breast cancer patients at increased risk of disease recurrence in a multicenter population, Clin Cancer Res 23 (2017), 1422–1431.

13.

Yang

Cui

Zhang

Bai

Han

Shi

and Liu

, Comprehensive lipid profiling of plasma in patients with benign breast tumor and breast cancer reveals novel biomarkers, Anal Bioanal Chem 407 (2015), 5065–77.

14.

Shen

Yan

Liu

Ambrosone

C.B.

and Zhao

, Plasma metabolomic profiles in breast cancer patients and healthy controls: By race and tumor receptor subtypes, Transl Oncol 6 (2013), 757–65.

15.

Beger

R.D.

, A review of applications of metabolomics in cancer, Metabolites 3 (2013), 552–574.

16.

Claudino

W.M.

Goncalves

P.H.

L.A.

Philip

P.A.

and Sarkar

F.H.

, Metabolomics in cancer: A bench-to-bedside intersection, Critical Reviews in Oncology/Hematology 84 (2012), 1–7.

17.

Sitter

Bathen

T.F.

Singstad

T.E.

Fjøsne

H.E.

Lundgren

Halgunset

and Gribbestad

I.S.

, Quantification of metabolites in breast cancer patients with different clinical prognosis using HR MAS MR spectroscopy, NMR in Biomedicine 23 (2010), 424–431.

18.

Hirayama

Kami

Sugimoto

Sugawara

Toki

Onozuka

Kinoshita

Saito

Ochiai

and Tomita

, Quantitative metabolome profiling of colon and stomach cancer microenvironment by capillary electrophoresis time-of-flight mass spectrometry, Cancer Research 69 (2009), 4918–4925.

19.

Borgan

Sitter

Lingjærde

O.C.

Johnsen

Lundgren

Bathen

T.F.

Sørlie

Børresen-Dale

A.L.

and Gribbestad

I.S.

, Merging transcriptomics and metabolomics – advances in breast cancer profiling, Bmc Cancer 10 (2010), 628.

20.

Miyagi

Higashiyama

Gochi

Akaike

Ishikawa

Miura

Saruki

Bando

Kimura

and Imamura

, Plasma free amino acid profiling of five types of cancer patients and its application for early detection, Plos One 6 (2011), e24143.

21.

Singletary

S.E.

Allred

Ashley

Bassett

L.W.

Berry

Bland

K.I.

Borgen

P.I.

Clark

Edge

S.B.

and Hayes

D.F.

, Revision of the American Joint Committee on Cancer staging system for breast cancer, Journal of Clinical Oncology Official Journal of the American Society of Clinical Oncology 20 (2002), 3628.

22.

Liu

Feng

Guo

Jiao

and Sun

, Targeted metabolomic analysis reveals the association between the postprandial change in palmitic acid, branched-chain amino acids and insulin resistance in young obese subjects, Diabetes Research and Clinical Practice 108 (2015), 84–93.

23.

Rieraborrull

Rodríguezgallego

Hernándezaguilera

Luciano

Ras

Cuyàs

Camps

Seguracarretero

Menendez

J.A.

and Joven

, Exploring the process of energy generation in pathophysiology by targeted metabolomics: Performance of a simple and quantitative method, Journal of the American Society for Mass Spectrometry 27 (2016), 168–177.

24.

Yang

Kasumov

Kombu

Zhu

Cendrowski

A.V.

David

Anderson

Kelleher

and Brunengraber

, Metabolomic and mass isotopomer analysis of liver gluconeogenesis and citric acid cycle – II. Heterogeneity of metabolite labeling pattern, Journal of Biological Chemistry 283 (2008), 21978–21987.

25.

Barker

and Rayens

, Partial least squares for discrimination, Journal of Chemometrics 17 (2003), 446–452.

26.

Westerhuis

J.A.

, Assessment of PLSDA cross validation, Metabolomics 4 (2008), 81–89.

27.

Lécuyer

Victor

A.B.

Deschasaux

Bouchemal

Nawfal

M.T.

Vasson

M.P.

Rossary

Demidem

Galan

and Hercberg

, NMR metabolomic signatures reveal predictive plasma metabolites associated with long-term risk of developing breast cancer, International Journal of Epidemiology (2018).

28.

Takayama

Tsutsui

Shimizu

Toyama

Yoshimoto

Endo

Inoue

Todoroki

Min

J.Z.

and Mizuno

, Diagnostic approach to breast cancer patients based on target metabolomics in saliva by liquid chromatography with tandem mass spectrometry, Clinica Chimica ACTA; International Journal of Clinical Chemistry 452 (2016), 18.

29.

Holm

Sedlaczek

and Grips

, Amino acid metabolism in liver disease, Current Opinion in Clinical Nutrition and Metabolic Care 2 (1999), 47.

30.

Hong

S.Y.

Yang

D.H.

and Chang

S.K.

, The relationship between plasma homocysteine and amino acid concentrations in patients with end-stage renal disease, J Ren Nutr 8 (1998), 34–39.

31.

Dowling

Henry

Meleady

Clarke

Gately

O’Byrne

Connolly

Lynch

Ballot

and Gullo

, Metabolomic and proteomic analysis of breast cancer patient samples suggests that glutamate and 12-HETE in combination with CA15-3 may be useful biomarkers reflecting tumour burden, Metabolomics 11 (2015), 620–635.

32.

Bhattacharya

and Maity

, Localization of phosphate dependent glutaminase in ascites fluid of ovarian cancer patient, Pathology Oncology Research Por 6 (2000), 217–223.

33.

Katt

W.P.

Ramachandran

Erickson

J.W.

and Cerione

R.A.

, Dibenzophenanthridines as inhibitors of glutaminase C and cancer cell proliferation, Molecular Cancer Therapeutics 11 (2012), 1269.

34.

Koochekpour

Majumdar

Azabdaftari

Attwood

Scioneaux

Subramani

Manhardt

Lorusso

G.D.

Willard

S.S.

and Thompson

, Serum glutamate levels correlate with gleason score and glutamate blockade decreases proliferation, migration, and invasion and induces apoptosis in prostate cancer cells, Clinical Cancer Research an Official Journal of the American Association for Cancer Research 18 (2012), 5888.

35.

Roy

Mondal

Wang

Khurana

Giri

Hoffmann

Jung

D.B.

Kim

S.H.

and Chini

E.N.

, Erratum to: Loss of HSulf-1 promotes altered lipid metabolism in ovarian cancer, Cancer and Metabolism 2 (2014), 1–1.

36.

Jové

Collado

Quiles

J.L.

Ramíreztortosa

M.C.

Sol

Ruizsanjuan

Fernandez

Cabrera

C.D.L.T.

Ramíreztortosa

and Granadosprincipal

, A plasma metabolomic signature discloses human breast cancer, Oncotarget 8 (2017), 19522–19533.

37.

Gurer-Orhan

Ince

Konyar

Saso

and Suzen

, The role of oxidative stress modulators in breast cancer, Current Medicinal Chemistry 24 (2017).

38.

Valko

Rhodes

C.J.

Moncol

Izakovic

and Mazur

, Free radicals, metals and antioxidants in oxidative stress-induced cancer, Chemico-Biological Interactions 160 (2006), 1.

39.

Stipanuk

M.H.

and Ueki

, Dealing with methionine/ homocysteine sulfur: Cysteine metabolism to taurine and inorganic sulfur, J Inherit Metab Dis 34 (2011), 17–32.

40.

Jove

Collado

Quiles

J.L.

Ramirez-Tortosa

M.C.

Sol

Ruiz-Sanjuan

Fernandez

de la Torre Cabrera

Ramirez-Tortosa

Granados-Principal

Sanchez-Rovira

and Pamplona

, A plasma metabolomic signature discloses human breast cancer, Oncotarget 8 (2017), 19522–19533.

41.

Budczies

Pfitzner

B.M.

Györffy

Winzer

K.J.

Radke

Dietel

Fiehn

and Denkert

, Glutamate enrichment as new diagnostic opportunity in breast cancer, International Journal of Cancer Journal International Du Cancer 136 (2014), 1619–1628.

42.

Tenori

Oakman

Morris

P.G.

Gralka

Turner

Cappadona

Fornier

Hudis

Norton

Luchinat

and Di Leo

, Serum metabolomic profiles evaluated after surgery may identify patients with oestrogen receptor negative early breast cancer at increased risk of disease recurrence. Results from a retrospective study, Mol Oncol 9 (2015), 128–39.

43.

Guo

Chen

Zhou

Wang

Huang

Yang

and Xing

, Genetic variants in genes of tricarboxylic acid cycle key enzymes are associated with prognosis of patients with non-small cell lung cancer, Lung Cancer 87 (2014), 162–8.

44.

Schuck

P.F.

Ferreira

G.C.

Viegas

C.M.

Tonin

A.M.

Busanello

E.N.

Pettenuzzo

L.F.

Netto

C.A.

and Wajner

, Chronic early postnatal administration of ethylmalonic acid to rats causes behavioral deficit, Behavioural Brain Research 197 (2009), 364–370.

Taurine,glutamic acid and ethylmalonic acid as important metabolites for detecting human breast cancer based on the targeted metabolomics

Abstract

BACKGROUND:

METHODS:

RESULTS:

CONCLUSIONS:

Keywords

1. Introduction

Table 1 Clinical characteristics of 116 subjects in the discovery set (mean ± SD) a

2.1 Study samples

2.1.1 Subjects of the discovery set

2.1.2 Subjects of the validation set

2.2 Chemicals and reagents

2.3 Preparation of standard solutions and assessment of the method

2.3.1 Solutions preparation

2.3.2 Preparation of quality control sample

2.4 Biochemical measurements

2.5 Serum amino acids and biogenic amines profiles analysis

2.5.1 Serum preparation for amino acids and biogenic amines profiles

2.5.2 UPLC/MS-MS analysis

2.6 Serum organic acid profiles analysis

2.6.1 Sample preparation for organic acids

2.6.2 GC-MS/MS analysis

2.7 Statistical analysis

2.7.1 Statistical analysis

2.7.2 Multivariate analysis

Table 2 Quantitative analysis of amino acids and biogenic amines profiles in three groups

3.1 Method validation

3.3 Changes in the amino acids and biogenic amines profiles

3.4 Changes in the organic acids profiles

Table 3 Quantitative analysis of organic acids profiles in three groups

Table 5 Pathways modulated by breast cancer condition

3.8 Pathway analysis

4. Discussion

Footnotes

Acknowledgments

Conflict of interest

References

Table 1
Clinical characteristics of 116 subjects in the discovery set (mean $\pm$ SD) ${}^{\text{a}}$

Table 2
Quantitative analysis of amino acids and biogenic amines profiles in three groups

Table 3
Quantitative analysis of organic acids profiles in three groups

Table 5
Pathways modulated by breast cancer condition