Serum proton NMR metabolomics analysis of human lung cancer following microwave ablation

The study was approved by the Institutional Ethics Committee of the Jiangxi Provincial People’s Hospital. A written informed consent was obtained from all participants involved in this study.

Between 20 February 2014 and 30 May 2016 patients who met the following criteria were retrospectively enrolled in the study: (1) pathologically verified peripheral NSCLC, (2) stage IIIB or IV, (3) chemotherapy-naive (except patients with recurrence treated with adjuvant chemotherapy or adjuvant radiation), (4) an Eastern Cooperation Oncology Group (ECOG) performance status of 0 to 2, and (5) adequate pulmonary, cardiac, hepatic, renal and hematological functions to allow anticancer treatment.

Selected characteristics of the NSCLC cases and controls were summarized in Table 1. Blood samples were collected prior to the morning meal, and serums were obtained by centrifugation of the blood samples at 3500 rpm for 10 min at 4 °C. Aliquots of the serum samples were stored at − 80 °C until NMR analysis.

About 1 mL serum samples were deproteinized by methanol with the ratios of serum: methanol as 1: 2 (v/v). The mixtures were vortexed and incubated at − 20 °C for about 30 min, and then were centrifuged into pellet proteins at 12000 g for 30 min. The supernatants were transferred into fresh tubes and lyophilized. The dried samples were dissolved in 600 μL 99.8% D₂O phosphate buffer (0.2 M, pH at 7.0) containing 0.05% sodium salt of 3-trimethylsilylpropionic acid (TSP, w/v), vortexed, centrifuged and decanted to 5 mm NMR tubes.

All ¹H NMR spectra were recorded with a Bruker AV 500 MHz spectrometer. A transverse relaxation-edited Carr-Purcell-Meiboom-Gill sequence [recycle delay-90-(τ-180-τ)n-acquisition] with a total spin echo delay (2nτ) of 40 ms was used to attenuate broad signals from slowly tumbling molecules such as proteins, whereupon the signals of the micro-molecule metabolites were clearly observed. ¹H NMR spectra were measured with 128 scans into 64 K data points with a recycle delay of 3 s over a spectral width of 20 ppm. The spectra were Fourier-transformed after multiplication by an exponential window function corresponding to a line broadening of 0.5 Hz. Resonances were assigned by querying the database HMDB (http://​www.​hmdb.​ca/), MMCD (http://​mmcd.​nmrfam.​wisc.​edu/), and were aided by the Chenomx NMR suite (version 8.0, Chenomx, Inc.).

The ¹H NMR spectra were manually phased and baseline-corrected by MestReNova (version 8.0.1,Mestrelab Research SL). Spectra were aligned to the TSP signal at zero ppm. A linear interpolation method was used to align the spectra using MestReNova. Regions containing residual water and methanol signals were removed, and the spectra were binned into integrated segments with equal widths of 0.01 ppm and adjusted by probabilistic quotient normalization and pareto-scaling prior to multivariate analysis.

First, an unsupervised principal component analysis (PCA) was used for for the metabolomics data overview and the spotting of outliers, and then for the detection of any grouping. Clustering was failed for the dataset (Additional file 1: Figure S1). Then a supervised partial least squares discriminant analysis (PLS-DA) was used to gain valuable insights on group-predictive spectral features. The metabolic profiles could be visualized as score plot, where each point represents a sample. The corresponding loading plot and S-plot were generated to provide information on the metabolites that influence clustering of the samples. In addition, a correlation circle plot produced by sparse PLS regression was applied to illustrate the relationships between certain integral metabolites and groupings.

To quantitatively assess the performance of the model, a repeated two fold cross-validation and permutation testing was carried out. Permutation distribution of the test statistic was computed based on 10,000 times repeated random permutation of the class labels and the significance value was taken to be the fraction of samples not exceeding the test statistic for the original sample. The overall quality of the model was evaluated by the cumulative R², while the predictive ability was assessed by the cumulative Q². Integration areas of the detected metabolites with marked differentiating ability were first tested for distribution normality. The student t-test or rank test was then employed to detect differences in metabolite levels between groups and p < 0.05 was considered to be statistically significant.

Differential metabolites were subjected to pathway analysis by MetaboAnalyst [13], which combines results from powerful pathway enrichment analysis with the pathway topology analysis, to identify the most relevant pathways involved in the NSCLC patients.

In order to excavate anything gene as potential biomarkers, and to integrate gene and metabolomics information to provide a better understanding of the efficacy of MWA in the treatment of NSCLC, the identified metabolites in metabolomics were then mapped to the KEGG pathway for biological interpretation of higher-level systemic functions. The metabolites and corresponding pathways were visualized using KEGG Mapper tool (http://​www.​genome.​jp/​kegg/​mapper.​html).

Representative ¹H NMR spectra of serum samples were shown in Fig. 1, with metabolites assigned. The variations in serum metabolites between groups were summarized in Table 2. The means and standard deviation values of metabolites for each group were provided in the Additional file 1: Table S1. Compared with the control group, the NSCLC group showed elevated levels of lactate, alanine, glutamate, proline, glycoprotein, phenylalanine, tyrosine and tryptophan, and decreased levels of glucose, taurine, glutamine, glycine, phosphocreatine and threonine, which could be partially or completely reversed by MWA treatment.

The validity of the PLS-DA model against over-fitting was assessed by the parameters R² (0.68), and the predictive ability was described by Q² (0.45). Theoretically, the closer the R² and Q² value to 1, the better the PLS-DA model is. A permutation test (n = 1000) was then performed to assure the predictive capacity of the PLS-DA model. The observed statistic P values via permutation testing were 0.036 which was less than 0.05, thus confirming the validity of the PLS-DA model. The PLS-DA score plot (Fig. 2) revealed satisfactory discrimination between the three groups. MWA treatment could reverse the disturbed metabolic profile towards the control group. The corresponding color-coded coefficient loading plot on the first component (Fig. 3) and S-plot (Fig. 4) visualized the contribution of each metabolite to the separation between groups. Significantly altered metabolites between groups were the pseudo peaks in warm color in the loadings plot, and those points in the upper right and lower left quadrants of the S-plot. Peaks in the positive axis and points in the upper right quadrants means the metabolites increased in the NSCLC group. The select metabolites were graphed as scatter plots in order to show the interindividual variation (Additional file 1: Figure S2). A concentric correlation circle concerning the correlations between the identified m a status similar to healthy controls etabolites and the three groups were plotted as scatter plot, provided additional information on the endogenous metabolites among groupings (Fig. 5).

The evidently disturbed metabolites were subjected to pathway analysis using a web-tool MetaboAnalyst [13]. Consequently, four metabolic pathways, including taurine and hypotaurine metabolism; d-glutamine and d-glutamate metabolism; glycine, serine and threonine metabolism; alanine, aspartate and glutamate metabolism, were filtered out as the most important pathways related with the metabolic disturbances in NSCLC patients (Fig. 6). The P values table generated by MetaboAnalyst in association with Fig. 6 was provided in the Additional file 1: Table S2. These metabolic alterations and the associated pathways provided insights into the mechanisms involved in the development and progression of NSCLC.

The identified metabolites were mapped on the KEGG reference pathway diagram concerning central carbon metabolism in cancer (Fig. 7), where red and blue nodes represented increased and decreased metabolites in NSCLC patients, respectively. Some oncogene and tumor suppressor gene were found, such as the tumor suppressor gene of p53 [14] SIRT3 [15] and SIRT6 [16], and oncogene of Ras [17] PI3K [18], AKT [19] and c-Myc [20]. However, these genes were not found via any genetic exploration in this study, but are instead potentially linked to the measured downstream metabolic changes, which might reflect the potential treatment mechanism of MWA on NSCLC.

Hyperthermia induced by MWA have been shown to change cell membrane integrity and fluidity, which were first considered to be the main cause of cell death [5, 22, 23]. Phosphocholine and choline, components of the plasma membranes, are often regarded as markers of the integrity of membranes. In our experiment, the slightly lowered levels of choline and O-phosphocholine in NSCLC group indicated an accelerated use of them for cancer cell proliferation. Compared with NSCLC, their levels in MWA group have no significant change. Ablation change the fluidity and permeability of cell membrane, thereby leading to concretion and cytolysis, subsequently causing intracellular metabolites shifts.

Mitochondrial dysfunction has been well correlated with heat-induced injury [24]. Some ultrastructural changes was observed previously, such as dilatation of the mitochondria with rupture or loss of the cristae [24]. The level of taurine in NSCLC patients was decreased, which was in accordance with previous reports [11, 25‐28]. As an antioxidant, taurine may be utilized as an antioxidant defense system against the oxidative stress involved in cancer process [29]. Ablation disrupted the mitochondrial membrane potential, resulting in the change in the redox status of cells [30], inducing cancer cell killing, thus leading to the reduced consumption of taurine.

Glycine was found to be decreased, which was in accordance with the previous study [31]. As a simple, nonessential amino acid, glycine was involved in the production of DNA, phospholipids and collagen as well as the release of energy. Given that cancer cells reprogram their metabolisms comprehensively, the decreased level of glycine in NSCLC patients may be related with the accelerated DNA synthesis. The level of glycine was increased after hyperthermic ablation, indicating that hyperthermia treatment caused nucleotide damage. Previous research offers some support for this view, which indicated that hyperthermia could cause damage to DNA [30] and lead to inhibition of nucleolar RNA synthesis [24, 32, 33]. One possible mechanism of heat-induced DNA injury was the denaturation of some key replication enzymes, such as DNA polymerase α and β, which is responsible for DNA replication and repair synthesis, respectively [34]. Another potential mechanism could be attributed to the formation of endogenous reactive oxygen species (ROS) as a consequence of hyperthermic ablation [5], which subsequently results in DNA damage because DNA is particularly vulnerable to ROS-induced damage. As a result, imbalance of glycine related to DNA damage and repairs was found [35].

The level of lactate was increased in NSCLC patients, which was in accordance with the previous studies [11, 31, 36‐38]. As previously reported, cancer cell metabolism involves primarily the conversion of glucose ultimately to lactate by an enzyme-catalyzed anaerobic fermentation rather than the oxidation of glucose ultimately to carbon dioxide and water as occurs with normal cells, known as the Warburg effect [39]. Recent study has proposed that lactate is also a tricarboxylic acid (TCA) cycle carbon source for NSCLC and sustain tumor metabolism in vivo [40].Therefore, the elevated levels of lactate found in the serum of NSCLC patients could be attributed to the enormous domestic for cell proliferation. Hyperthermia caused extensive damage of the ultrastructural of mitochondria, ruptured glycolysis and energy metabolism of cancerous cells, which blocking the generation of lactate as a fuel of TCA cycle.

Levels of serum alanine was elevated in NSCLC group, which was consistent with previous report [36]. Alanine is a glycogenic (glycogen-producing) amino acid that can be converted to pyruvate and tricarboxylic acid cycle intermediates, and then to glucose by gluconeogenesis, functioning as an energy source to meet the huge demand of energy consumed in various metabolic activities in tumor cells. Elevated levels of serum alanine may facilitate energy synthesis in the cells and provide enough energy for cell growth. Tumor cells also utilized glutamine as another energy supply [41], which was called glutaminolysis. The decreased level of glutamine has been reported previously [42, 43], which was possibly due to its consumption for energy supply. After MWA treatment, the elevated level of alanine and decreased level of glutamine were recovered to the status of normal control group. Destroy of mitochondria and coagulative necrosis of cancer cells caused by hyperthermia might be responsible for the reversion of the above metabolites which are related with energy metabolism.

However, the roles of the significant pathways mentioned above in relation to MWA treatment were only hypothesis, without any complementary techniques as a backing. The model without vigorous explaination and prediction ability limits us to draw any definitive conclusions, which may be attributed to the small sample size and the heterogeneity of patients. Future studies with a larger sample size are planned to more clearly characterize metabolic profiles of NSCLC patients.