Cumulative evidence demonstrates that concentration of lactic acid shows a consistent increase in urine (30) or tissue (31,35,36,38) samples of gastric cancer groups, but glucose is considerably depleted compared with those healthy counterparts or non-malignant patients (like chronic superficial gastritis and chronic atrophic gastritis without intestinal metaplasia) (35,36). The high lactate level might be attributed to the special metabolism of most cancer cells, known as 'Warburg effect' we mentioned above (6). Scarce glucose might result from the overexpression of glucose transporters (42) and type II hexokinase (43), which are both confirmed in gastric cancer tissues. Higher fructose-6-phosphokinase (6-FPK) activity can also result in low glucose in gastric cancer tissues (44), as it regulates the output of glucose to glycolysis pathway. The glycolytic switch has been identified to be associated with oncogenic transformation and molecular signal transduction, such as hypoxia-inducible factor pathway, insulin signaling pathway and PI3K-Akt-mTOR pathway (45). Furthermore, overexpression of pyruvate kinase and lactate dehydrogenase is positively associated with tumor proliferation and poor prognosis, downregulation of them in vitro experiment can impair tumor invasion (38,46–49). On the other hand, such special microenvironment might be the requirement of rapid propagation of tumor cells. To our understanding, it has been reported that accumulated lactic acid moderates the activity of proteases that decompose extracellular matrix, which can produce some peptides and amino acids that are consumable for energy generation (44). Acidosis microenvironment is also ascribed to the formation of cancer blood vessels, meeting the plentiful supply of nutrients and leading to tumor invasion and metastasis (50). Moreover, tumor-derived lactate shows strongly negative effects on cytotoxic T-cell/NK cell function (11,51) and blocks differentiation of monocytes to dendritic cells (52), finally leading to tumor immune escape. However, such outcome demands further verification in gastric cancer.

Considering tricarboxylic acid cycle (TCA) intermediates, an increase of five metabolites (α-ketoglutaric acid, malic acid, fumarate, succinate, citric acid) is noticed regardless of blood (26,28), urine (30,32) or tissue (28,35,36,39) samples in gastric cancer. There are some possible reasons that can explain this phenomenon. One account is that cancer cells still use a small portion of glucose for oxidative phosphorylation. Secondly, cancer cells might also utilize fumarate respiration to generate energy under special conditions of glucose deprivation and severe hypoxia in microenvironment (53), and succinate is one of the byproducts in this process except for originating from TCA. Hence, it provides a likely explanation for the accumulation of fumarate and succinate. Another reason is that some amino acids, such as glutamine, threonine, phenylalanine, tyrosine or proline, can be converted into these intermediates involving in TCA (Fig. 2). Additionally, elevated levels of citric acid can be used in the de novo fatty acid synthesis, but it is noted that citrate can also induce apoptosis in two gastric cancer cell lines in vitro experiment (54,55).

Availability of amino acids is pivotal for cellular protein biosynthesis and cytoskeleton formation, while it has been pointed out that amino acids especially those linking to TCA (Fig. 2) are an alternative energy source of cancer cell proliferation (56). By employing metabolomics technologies, levels of various amino acids (including serine, valine, phenylalanine, tryptophan, glycine, and proline) and their primary derivatives (such as kynurenine, kynurenic acid, anthranilic acid and nicotinic acid) are significantly higher in tissue specimens (31,36,37) and gastric content (29,40,41), but decreased concentration in some of them is observed in blood (24). The overexpression of L-type amino acid transporter 1 (LAT1) might be proposed to explain this dissimilarity (57). Free amino acids are greatly assimilated to cancer tissues via LAT1 from bloodstream, resulting in the low accumulation of amino acid in contrast to normal counterparts. Malnourishment may also be a contributing factor to these reduced levels of plasma amino acids. Apart from these, degradation of extracellular matrix mediated by the overexpressed matrix metalloproteinases (MMPs) and activated autophagic degradation of intracellular proteins are considered as the potential source of accumulative amino acids in tumor tissues (58–60).

Elevated amino acids in microenvironment are contributing factors in carcinogenesis. Most strikingly, it is indicated that many cancer cell lines cannot survive in the absence of glutamine (61), because it is required for anabolic growth of mammalian cells through its ability to control the master regulator of protein translation mTORC1 (62). Reprogramming of glutamine metabolism further contributes to the proliferative and metabolic responses regulated by oncogenic transcription factor c-MYC (63). In addition, it is also the nitrogen donor for several key metabolic enzymes and for the de novo synthesis of both purines and pyrimidines (Fig. 2). Serine also participates in the de novo synthesis of nucleotides by serving one carbon unit. Functional genomics further indicates that serine biosynthesis pathway is significant for breast cancer event, which can be attributable to the overexpression of phosphoglycerate dehydrogenase (PHGDH) that controls the flow of intermediates originated from glycolysis (64). Inhibition of PHGDH in cells can result in lower serine and decrease cellular proliferation in vitro. However, this remains unclear in gastric cancer. Tryptophan and its downstream metabolites (mainly including kynurenine, kynurenic acid, anthranilic acid, nicotinic acid) via kynurenine pathway are related to the pathogenesis and prognosis of various malignancies including gastric cancer (65,66). Kynurenine pathway catalyzed by indoleamine-2, 3-dioxygenase (IDO) plays a key role in adapting the tumor microenvironment to favor cancer progression because higher IDO expression is associated with an increase in immunosuppressive T-regulatory cell activity (67), and its immunosuppressive role inhibits T-cell mediated cytotoxicity and cell proliferation of gastric cell lines in vitro (68). Additionally, 3-hydroxyanthanilie acid (downstream metabolites in kynurenine pathway) also has suppressive effects on inflammation and immune response (69). Glycine used in living organism as building blocks of purines is strongly correlated with the rapid proliferation rates, and then antagonizing glycine uptake and its mitochondrial biosynthesis preferentially impair rapidly proliferating cells (70). The indirect anti-angiogenic impact of glycine is also identified in vitro (71,72) and in vivo (73,74), possibly because it might inhibit the proliferation of vascular endothelial cells, finally leading to angiogenesis (74). Elevated proline in tumor tissues might begin with the activation of MMPs and degradation of micro-environmentally extracellular matrix (ECM), subsequently the degradation of collagen catalyzed by proline dehydrogenase (PRODH) that can be regulated under conditions of nutrient stress linked to mTOR signaling system (75). Other elevated amino acids, such as tyrosine, valine and cysteine, can be converted into the TCA intermediates (except for citric acid, isocitrate, succinyl-CoA, oxaloacetate) to generate energy (Fig. 2).

The notable feature of lipid metabolism in cancer cells is an increased rate of lipogenesis and the upregulation of mitochondrial fatty acid β-oxidation, gastric cancer shows a similar tendency and presents typical changes regarding various metabolites involving in lipid metabolism.

Fatty acids, such as hexadecenoic acid, docosahexaenoic acid, eptanoic acid and β-hydroxybutyrate, are significantly larger in gastric cancer tissues than in benign tissues (like chronic superficial gastritis) (28). Octadecanoic acid is also found to be elevated in blood specimens obtained from gastric cancer patients (39). Of them, β-hydroxybutyrate is the common product of fatty acid degradation via β-oxidation, suggesting more intensive decomposition of fatty acids in microenvironment. The accelerated metabolism from lipids to fatty acids and finally ketone bodies consumes fat, which might explain the fact that patients become very thin in later stages of gastric cancer. This signature also has been identified by the xenograft animal models with gastric cancer showing elevated levels of glycerol and hexadecanoic acid (30), resulting from the high activation of adipocyte lipolysis in cancer cells as well as enhanced expression and function of adipocyte hormone-sensitive lipase in cancer cachexia (76). In contrast, some data show that unsaturated fatty acids such as 9-hexadecenoic acid, cis-vaccenic acid, arachidinic acid, hexadecanoic acid and 3-hydroxybutanoic acid are found to be significantly decreased in cancer tissue samples (26). Of note, the level of O-acetylcarnitine, which increases the β-oxidation of fatty acid, shows a declining trend as the early gastric cancer progresses into advanced stage (31). Accordingly, it seems that decreased O-acetylcarnitine might explain the impaired fatty acids β-oxidation in stage III/IV gastric cancer, which is characterized by the decline in unsaturated fatty acids we discussed above (9-hexadecenoic acid, cis-vaccenic acid, arachidinic acid, hexadecanoic acid and 3-hydroxybutanoic acid). However, this discrepancy between different research needs further elucidation with larger samples and different analytical methods. On the other hand, free fatty acids in plasma, including palmitic acid, stearic acid, 9-(Z)-hexadecenoic acid, oleic acid, linoleic acid, docosahexaenoic acid and arachidonic acid, are equivalent in both gastric cancer and gastric benign disorders (25). Therefore, it infers that free fatty acids in blood might be not utilized by tumor cells.

Upregulated lipid peroxides are also confirmed in this disease. Accumulation of 4-hydroxyphenylacetate resulting from the oxidative degradation of lipids was observed in the study of Jung et al (31). Elevation of azelaic acid in blood samples, which is the end product of linoleic acid when subjected to peroxide decomposition (77) and can serve as a marker of lipid peroxidation (78), as observed by Yu et al (25).

Based on that indicated above, these signatures show that cancer cells utilize massive fatty acids to meet the demand of cell membrane synthesis, mainly for lipid raft and lipid-modified signaling molecules (79); and a large fraction of their membrane lipids are biosynthesized de novo rather than scavenging from extracellular sources. In de novo lipogenesis, fatty acid synthase (FAS) catalyzes the synthesis of palmitate from acetyl-CoA or malonyl-CoA in the presence of NADPH as a redox equivalent. FAS expression is commonly low in non-proliferating cells that typically import lipids from the extracellular milieu. In contrast, actively proliferating cells, especially tumor cells, have increased demands for lipids, which is highly dependent on de novo synthesis. So FAS is frequently upregulated in many types of tumors (80–82) including gastric cancer (83,84); and increased FAS expression is linked to tumor proliferation, chemoresistance and poorer prognosis in cancers (85–88). Thus, this key enzyme implicated in lipogenesis has been studied as potential target in anti-neoplastic therapy (84). On the other hand, enhancement of fatty acid-β oxidation is also considered to be an important metabolic reprogramming in the early stage of some cancer types (89), as it produces more ATP and acetyl coenzyme A which in turn can accelerate the rate of citric acid oxidation and serve as the energy source (90). Furthermore, production of polyunsaturated fatty acids, to some extent, is also associated with tumor cell proliferation, apoptosis and angiogenesis (91,92).

Tumor cells are in a state of such rapid proliferation and differentiation that frequent nucleotide synthesis and metabolism are upregulated significantly. Accumulation of the end products of nucleotide catabolism is characterized by the higher levels of uric acid or urate (25,30) in gastric cancer patients or animal models. Other purines compounds like hypoxanthine and guanosine were also increased (35,37), but Aa et al showed decreases in uridine (an RNA building block) (28). Nucleotides are also associated with energy metabolism, mainly in the form of ATP and GTP. Of tumor cells, adequate energy should be supplied to meet their proliferation. In this way, it is assumed that nucleotide phosphates should increase in cancer tissues compared with normal tissues. However, Hirayama and colleagues (35), identified that there was no noticeable difference between gastric cancer tissues and adjacent normal tissues with regard to most nucleotide phosphates (ATP, ADP, GTP, and GDP), total adenylate and energy charge. Accordingly, it infers that cancer cells gain growth superiority over their normal counterparts by switching metabolic patterns of energy to anaerobic glycolysis and possibly fumarate respiration that we have discussed above, instead of securing more ATP.

Except for the changed metabolisms mentioned above, other metabolite concentrations also show increased or decreased trend in the development of gastric cancer. Increased level of creatinine, a waste product of muscle metabolism, was detected in urine samples of tumor groups (33), which might be induced by lower total body skeletal mass among cachectic patients (93,94). Changes in inositol level of gastric malignancy patients are investigated in either tumor tissues (28,36,37) or urine samples (30), but its mechanism and significance are poorly understood.

Early diagnosis is the key element determining the outcome of treatment in cancer research, but current application of cancer biomarkers, endoscopy and imaging is still not satisfactory. Serum biomarkers, like CEA and CA19-9, are not effective given their poor sensitivity or specificity. Inconsistent diagnostic efficacy at endoscopy that results from the variations in skill and experience of endoscopists and pathologist might lead to missed diagnosis at early phase, while positive results displayed on imaging examination (such as barium meal and computer tomography) are prone to advanced stage. Interestingly, utility of various -omics technologies open a new field to discover potential biomarkers for gastric cancer diagnosis, especially based on metabolomics.

Exploration of gastric cancer biomarkers in blood or urine is more appreciated because of its non-invasive priority. Yu et al demonstrated that metabolic profiles were quite different in gastric cancer patients with different pathological types in the Correa model, but intestinal metaplasia shared similar metabolic phenotype (threonate, glutamate and azelaic acid) in plasma with neoplastic groups (25,95). Ikeda et al also identified that there were obvious variations in serum metabolic profiles of gastrointestinal cancers (including esophageal, gastric and colorectal) in contrast to healthy volunteers (27). In particular, changes in the levels of 3-hydroxypropionic acid and pyruvic acid were sufficient to differentiate gastric cancer from esophageal and colorectal cancer, and showed high values for both sensitivity (84.6 and 70.0%) and specificity (71.4 and 90.0%) compared with conventional biomarkers (CA19-9 and CEA) (27). The diagnostic potential of serum metabolic profiles between gastric cancer and non-cancer groups was also confirmed by Song et al, and these alterations occurred at early stage of gastric carcinogenesis (26).

Recently, one urine metabolomics in gastric cancer found that 14 out of 17 metabolites detected from training set (94 urine samples) via GC-MS showed diagnostic value better than classic blood biomarkers on validation set (199 urine samples) (34). Six of them (L-alanine, L-isoleucine, L-serine, L-threonine, L-proline and L-methionine) revealed satisfactory diagnostic values with the area under the ROC of >0.75. Chan et al also revealed that gastric cancer has a unique urine metabolic profiles in contrast to benign gastric diseases and healthy patients, especially 2-hydroxyisobutyrate, 3-indoxylsulfate and alanine, producing a discriminatory model with the area under the curve (AUC) of 0.95 (33). Another study reported that metabolites altered in urinary data of gastric cancer patients was predicted with higher sensitivity than CA19-9 and CEA (31).

In tissue testing, Wu and colleagues indicated that 18 metabolites were detected differently between the malignant tissues and the adjacent non-malignant tissues of gastric mucosa with AUC value of 0.9629 (37), but tissue testing was not a non-invasive approach in contrast to blood or urine testing. Our data, on the other hand, showed that higher levels of tyrosine, phenylalanine and tryptophan in the gastric juice were detected in the early phase of gastric carcinogenesis (40), and the sensitivity and specificity for gastric cancer detection with phenylalanine was 87.9 and 79.4% respectively (41).

Most gastric cancer-related deaths occur as a result of metastasis, even among patients undergoing gastrectomy. Unfortunately, no molecular markers for predicting metastasis and prognosis are accessible.

Based on metabolomics, Wu and colleagues showed that five metabolites (increased L-cysteine, hypoxanthine and L-tyrosine; decreased phenanthrenol and butanoic acid) were detected differently between non-invasive (T1 and T2) and invasive (T3 and T4) groups, furthermore, 4-hydroxyphenyl-acetate, alanine, phenylacetylglycine, mannitol, glycolate and arginine levels were significantly correlated with cancer T stage (37). By establishing animal models with gastric cancer cell line SGC-7901, Chen et al confirmed that metabolites correlated to proline and serine metabolism could distinguish metastatic from non-metastatic specimens with an AUC value of 1.0 (36). Study conducted by Hu et al suggested that decreased levels of alanine, glycerol, L-proline, butanoic acid and L-threonic acid as well as increased levels of butanediotic acid and myo-inositol could detect non-metastatic and metastatic groups (AUC=1.00) (30).

Significantly, Chen and coworkers recently evaluated the prognostic value of 17 urinary metabolites, which have been identified differently between gastric cancer group and normal group, by following up 82 out of 112 gastric cancer cases for 3–5 years after surgery (34). They discovered that patients with higher levels of proline, p-cresol and 4-hydroxybenzoic acid display poor prognosis with median survival time 16, 15 and 15 months, respectively. Furthermore, the concentration of p-cresol closely correlated with gastric cancer stage, which was gradually increased with the stage of the patients.

It is possible that changes in proline might be essential in tumor metastasis. As we have mentioned above, proline in tumor tissues might result from the degradation of collagen (73). This process mainly begins with the activation of MMPs and degradation of microenvironmental ECM, which partially accounts for the tumor invasion and metastasis (96). In this respective, elevated proline serving as metastatic biomarker for gastric cancer is possible, but further research is necessary.

Chemosensitivity prediction that aims to maximize the therapeutic response and minimize adverse effects is a difficult task in the treatment of advanced tumors. One of classical approaches for predicting the activity of anticancer agents is cell culture testing, which is mainly based on clone formation, cell metabolic activity assays, proliferation and tumor growth in vitro experiments. However, it must be noted that these methods still fail to fully reproduce the tumor microenvironment, although current patient-derived primary cell culture or patient-derived tumor xenograft models are able to retain cellular heterogeneity of original tumors (97).

Lu et al, in particular, suggested that some conventional cytotoxic anticancer agents (vincristine, taxol, 5-fluorouracil, doxorubicin, cisplatin, camptothecin) lost their efficacy apparently when cultured PNAC-1 cells (pancreatic cancer) in vitro were deprived of glucose (98). Similarly, a recent study also identified that high glucose conditions promoted SGC-7901 proliferation in vitro and reduced chemosensitivity in vivo or in vitro (99). We could speculate that responses of gastric cancer against anticancer drugs in actual microenvironment in vivo might be considerably different from what we expect in culture condition. Therefore, utilizing metabolomics is considered to be a promising tool to assess the sensitivity of chemotherapy in virtual conditions and discovering therapeutic targets regarding specific tumor metabolism (20,21).

Wang et al applied high performance liquid chromatography coupled with a quadrupole time-of-flight mass spectrometer to predict chemotherapy response in a human xenograft model of gastric cancer administered with cisplatin plus 5-fluorouracil (5-FU) (100). Consequently, 1-acyl-lysophosphatidylcholine and polyunsaturated fatty acid were proposed to surveil gastric cancer chemosensitivity, since 1-acyl-lysophosphatidylcholine can regulate the activity of enzymes like phospholipase A2 (PLA2) and lysophosphatidylcholine acetyltransferases. PLA2 catalyzes the production of arachidonic acid that is likely to promote cell cycle arrest and apoptosis dependent on ceramide pathway (101,102), while lysophosphatidylcholine acetyltransferases catalyzes phospholipid synthesis linked to tumor cell proliferation. Another study suggested that proline was reduced while glutamate increased dramatically, and PRODH (catalyzes the metabolic production of glutamate from proline proceeds) mRNA expression was upregulated 2-fold after 5-FU administration; but they were less affected in 5-FU-resistant cells (103). Thus PRODH might make it possible to be a marker for assessing intracellular dynamic responses to 5-FU. Additionally, Kim and colleagues utilized ¹H-NMR to investigate the metabolic changes in urine sample following Adriamycin (ADR) treatment for gastric adenocarcinoma in an animal model (104). This study revealed that levels of trimethylamine oxide, hippurate and taurine, which all decreased in tumor group without treatment, were increased dramatically after ADR disposal; while 2-oxoglutarate, 3-indoxylsulfate, trigonelline, trimethylamine and citrate recovered to those of normal group (104). Alterations in these metabolites might be ascribed to the pharmacological activity of ADR that activates apoptotic process of gastric cancer cells via ADR-induced genotoxic stress.

In another study, dysregulation of pyruvic acid efflux in gastric cardia cancer was observed with the combination of proteomics and metabolomics (38). Furthermore, Cai et al also found that downregulation of lactate dehydrogenase A (LDH-A) and overexpression of pyruvate dehydrogenase B (PDH-B) could force pyruvic acid into the Krebs cycle rather than the glycolysis process in gastric cancer cell line AGS, consequently inhibiting cell growth and migration (38). In view of the above, LDH or PDH might serve as a therapeutic target in gastric cancer treatment.