Altered L-Arginine Metabolic Pathways in Gastric Cancer: Potential Therapeutic Targets and Biomarkers

There is a pressing need for molecular targets and biomarkers in gastric cancer (GC). We aimed at identifying aberrations in L-arginine metabolism with therapeutic and diagnostic potential. Systemic metabolites were quantified using mass spectrometry in 293 individuals and enzymes’ gene expression was quantified in 29 paired tumor-normal samples using qPCR and referred to cancer pathology and molecular landscape. Patients with cancer or benign disorders had reduced systemic arginine, citrulline, and ornithine and elevated symmetric dimethylarginine and dimethylamine. Citrulline and ornithine depletion was accentuated in metastasizing cancers. Metabolite diagnostic panel had 91% accuracy in detecting cancer and 70% accuracy in differentiating cancer from benign disorders. Gastric tumors had upregulated NOS2 and downregulated ASL, PRMT2, ORNT1, and DDAH1 expression. NOS2 upregulation was less and ASL downregulation was more pronounced in metastatic cancers. Tumor ASL and PRMT2 expression was inversely related to local advancement. Enzyme up- or downregulation was greater or significant solely in cardia subtype. Metabolic reprogramming in GC includes aberrant L-arginine metabolism, reflecting GC subtype and pathology, and is manifested by altered interplay of its intermediates and enzymes. Exploiting L-arginine metabolic pathways for diagnostic and therapeutic purposes is warranted. Functional studies on ASL, PRMT2, and ORNT1 in GC are needed.


Introduction
Gastric cancer remains the leading cause of cancer-related deaths despite a steady decline in incidence rates, attributed to improved hygiene and diet and the popularization of therapies against Helicobacter pylori [1]. While efficient in reducing the occurrence of non-cardia subtype of gastric adenocarcinoma (GA), they failed to stop the rising incidence of cardia subtype (CA) [2]. Poor survival rates characteristic for gastric cancer result from the disease being diagnosed at advanced stage, when it is not amenable for curative resection, leaving chemotherapy as a major therapeutic option [1]. However, chemotherapy is not only highly toxic but its effectiveness is below expectations. Therefore, "a radical shift toward precision medicine" is advocated. However, this requires unraveling the molecular landscape of gastric cancer in order to establish novel therapeutic targets and biomarkers [2].
Reprogramming of metabolic pathways is a well-recognized hallmark of cancer [3], going well beyond the Warburg effect [4]. Therefore, untargeted and targeted metabolic profiling is viewed as a promising tool in precision medicine [5]. However, recent advances in metabolomics showed that the alterations are neither uniform across cancer types nor constant over time. Rather, they reflect cancer heterogeneity as well as its progression [6].
L-arginine (Arg) metabolic pathways are among those repeatedly found to be deregulated [7][8][9]. Pharmacological manipulation of pathway enzymes is viewed as an attractive therapeutic approach while monitoring enzyme activity and/or metabolite concentration may aid cancer diagnosis and treatment [10][11][12][13]. However, a better understanding of pathway status and function in cancer is needed as even the role of arginine, the pathway precursor, is ambiguous [13]. Arginine is competed for by immune and cancer cells, either contributing to immunosurveillance or supporting tumor growth and metastasis [13]. Still, the amino acid role in gastric cancer might not be unequivocally tumor-supporting. Uncharacteristically, arginine has been shown to inhibit growth of gastric cancer cells in vitro by upregulating caspase 8 expression and consequently inducing apoptosis [14].
Synthesis of NO by NO synthases (NOSs) and synthesis of L-ornithine (ornithine; Orn) by arginases (ARGs), as a precursor of polyamines synthesized by ornithine decarboxylase (ODC), are two main competitive ways of Arg utilization [11]. NO synthesis is inhibited by methylated arginine derivatives such as asymmetric and symmetric dimethylarginines (ADMA and SDMA), products of protein arginine methyltransferases (PRMTs), as they compete with Arg for NOS as well as for amino acid transporters. The ADMA is metabolized to L-citrulline (citrulline; Cit) and dimethylamine (DMA) by dimethylarginine dimethylaminohydrolases (DDAHs). Citrulline can be recycled to arginine in a two-step reaction catalyzed by argininosuccinate synthase-1 (ASS1) and argininosuccinate lyase (ASL) [10,15,16]. A basic overview of key players of arginine metabolic pathways and their interrelationship is depicted in Figure 1. Overview of metabolic pathways of L-arginine. Pathway players not investigated in the current study are indicated by the grey color, while the analyzed pathway metabolites are depicted in rectangular frames, and enzymes in elliptical frames. Inhibitory effects are marked by black blunt-ended arrows-dashed if the effect is weak. ADMA, asymmetric dimethylarginine; SDMA, symmetric dimethylarginine; DMA, dimethylamine; ASL, argininosuccinate lyase; ARG1,2, arginase 1 and 2; ASS1, argininosuccinate synthase 1; CATs, cationic amino acid transporters; DDAH1,2, dimethylarginine dimethylaminohydrolase 1 and 2; NO, nitric oxide; NOS2, nitric oxide synthase 2; ODC1, ornithine decarboxylase 1; PRMT, protein arginine methyltransferase.
Considering the pressing need for new molecular targets and biomarkers for gastric cancer and the growing interest in Arg metabolism in this capacity, our aim was to explore the pathway status in gastric cancer in order to identify aberrations with therapeutic and Figure 1. Overview of metabolic pathways of L-arginine. Pathway players not investigated in the current study are indicated by the grey color, while the analyzed pathway metabolites are depicted in rectangular frames, and enzymes in elliptical frames. Inhibitory effects are marked by black blunt-ended arrows-dashed if the effect is weak. ADMA, asymmetric dimethylarginine; SDMA, symmetric dimethylarginine; DMA, dimethylamine; ASL, argininosuccinate lyase; ARG1,2, arginase 1 and 2; ASS1, argininosuccinate synthase 1; CATs, cationic amino acid transporters; DDAH1,2, dimethylarginine dimethylaminohydrolase 1 and 2; NO, nitric oxide; NOS2, nitric oxide synthase 2; ODC1, ornithine decarboxylase 1; PRMT, protein arginine methyltransferase.
Considering the pressing need for new molecular targets and biomarkers for gastric cancer and the growing interest in Arg metabolism in this capacity, our aim was to explore the pathway status in gastric cancer in order to identify aberrations with therapeutic and diagnostic potential. In the present study, a comprehensive analysis of systemic metabolite concentration (Arg, Cit, Orn, ADMA, SDMA, and DMA) and local enzyme (ASL, ARG1, ARG2, ASS1, DDAH1, DDAH2, NOS2, ODC1, PRMT1, PRMT2, and PRMT5) and transporter (ORNT1) expression was conducted. Metabolomic and transcriptomic data were referred to cancer anatomical subsite and pathology. Patterns of interrelationships of pathway players and their correlation with local and systemic immune, inflammatory, and angiogenic mediators and other molecules relevant for cancer were examined.

Metabolomic Analysis
Biobanked serum samples stored at −80 • C, obtained from 293 individuals, including 153 apparently healthy controls, 50 patients with benign gastric disorders, and 90 patients with histopathologically confirmed gastric adenocarcinoma, were used in metabolomic analysis. Cancer patients and patients with benign gastric disorders (gastritis, cardiospasmus, gastro-esophageal reflux disease) were admitted to the Department of Gastrointestinal and General Surgery of Wroclaw Medical University for the disease diagnosis and/or treatment. Cancer patients underwent standard preoperative evaluation consisting of blood work, physical examination, and imaging techniques (ultrasonography, computed tomography, and magnetic resonance). Cancers were staged clinically using the 7th edition of the Union for International Cancer Control TNM system. Control individuals were recruited from apparently healthy blood donors. Detailed population characteristics are summarized in Table 1.

Sample Extraction
The extraction and derivatization of metabolites associated with Arg metabolic pathways were carried out using the previously described method [17]. One hundred microliters of calibration standards or serum samples were mixed with 50 µL of borate buffer (pH = 9.2) and 10 µL of internal standard solution of D6-arginine, D6-DMA, D7-ADMA, D6-ornithine (100 µM, 50 µM, 20 µM, 100 µM, and 70 µM, respectively). After mixing, 400 µL of acetonitrile and 10 µL of 10% BCl in acetonitrile were added. Then, the mixture was vortexed for 10 min at 25 • C. The samples were centrifuged at 10,000 rpm for 7 min at 4 • C. Obtained supernatants were diluted 4:1 with water.
Chromatographic conditions used for compound quantification were as follow: Waters HSS T3 chromatographic column (1.8 µm, 1.0 × 50 mm) with mobile phase A = 0.1% formic acid in water and mobile phase B = 0.1% formic acid in methanol, column temperature set at 60 • C, flow rate = 0.220 mL/min and injection volume of 2 µL. Typical elution conditions were held at 5% B from 0.0 to 1 min, to 14% B in 2.5 min, to 60% B in 1.5 min, to 90% B in 0.5 min, held at 90% B for 1.1 min, to 5%B in 0.1 min, held at 5% B for 1.9 min.
MS acquisition was carried out with an electro spray ionization (ESI) ion source operated in a positive mode. Source parameters were as follows: nebulizing and drying gas (nitrogen): 650 L/h and 65 L/h, respectively; spray voltage: 0.5 kV; source temperature: 120 • C, and the desolvation temperature: Phenol-chloroform extraction was used for RNA isolation and RNA isolates were additionally purified with PureLink™ RNA Mini Kit (Thermo-Fisher Scientific) and subjected to on-column genomic DNA digestion with DNase PureLink™ DNase Set (PureLink™ DNase Set, Thermo-Fisher Scientific). RNA concentration, purity and integrity were determined spectrophotometrically using Nanodrop 2000 (Thermo-Fisher Scientific) and by microfluidic electrophoresis using the Experion platform and dedicated RNA StdSens analysis kits (BioRad, Hercules, CA, USA).
The cDNA library was created from 1000 ng of RNA, reversely transcribed in C1000 thermocycler (BioRad) using iScript™ cDNA Synthesis Kit (BioRad) according to the manufacturer's recommendations.
Prior to the statistical analysis, technical qPCR replicates were averaged. Geometric mean of all Cq values for each gene was calculated. Individual sample Cq were subtracted from mean giving ∆Cq, linearized by 2ˆ∆ Cq conversion, and subsequently normalized to GAPDH giving normalized relative quantities (NRQ) [18].

Statistical Analysis
Data were assessed for normality and homogeneity of variances prior to each analysis using the Kolmogorov-Smirnov and Levene tests, respectively, and log-transformation was applied if appropriate. Expression data were examined using t-test for paired samples or Wilcoxon test (paired analysis) and with t-test for independent samples or Mann-Whitney U test and presented as, respectively, geometric means or medians with 95% confidence intervals (CI). In metabolomic analysis, two-group comparisons were conducted using t-test for independent samples with Welch correction in case of unequal variances, while multi-group comparisons using one-way analysis of variance with Scheffe post-hoc test or Kruskal-Wallis H test with Conover post-hoc test. Data were presented as means or geometric means with 95%CI or standard deviation or medians with 95%CI. Correlation analysis was conducted using Spearman rank correlation or Pearson correlation. The following descriptors were used for interpretation of correlation coefficients: <0.1 as negligible correlation; 0.1-0.39 as weak; 0.4-0.69 as moderate; 0.7-0.89 as strong; 0.9-1.0 as very strong correlation (as quoted in [20]). The ROC curve analysis was conducted to determine the strength of association and the diagnostic potential of assessed analyte. The ROC analysis data are reported as area under ROC curve (AUC), representing marker accuracy expressed in %. In addition, marker sensitivity and specificity were calculated. Enter method of logistic regression was applied to calculate predicted probabilities, subsequently used as dependent variable in ROC analysis of diagnostic potential of metabolite panels. Frequency analysis was conducted using Fisher's exact test or Chi-square test. Multivariate analysis (stepwise method) was used to select independent predictors of gene expression.

Pathway Status in Gastric Cancers and Benign Disorders
The concentration of all metabolites differed significantly between groups ( Figure 2). Arginine, citrulline, and ornithine were significantly higher and SDMA and DMA were significantly lower in controls than in patients with cancer or benign gastric disorders, while ADMA was significantly higher, only compared to patients with benign conditions. Cancer patients had lower SDMA than those with benign disorders and those with CA also had lower citrulline, ornithine, and ADMA. Cancer patients with tumor location in gastric cardia had lower citrulline than those with non-cardia gastric tumors.
The concentration of all metabolites differed significantly between groups ( Figure 2). Arginine, citrulline, and ornithine were significantly higher and SDMA and DMA were significantly lower in controls than in patients with cancer or benign gastric disorders, while ADMA was significantly higher, only compared to patients with benign conditions. Cancer patients had lower SDMA than those with benign disorders and those with CA also had lower citrulline, ornithine, and ADMA. Cancer patients with tumor location in gastric cardia had lower citrulline than those with non-cardia gastric tumors.

Association with Cancer Pathology
Arginine, SDMA, and DMA were not associated significantly with cancer pathology. Ornithine and ADMA were weakly negatively correlated with overall TNM stage and with depth of tumor invasion. Citrulline was significantly lower in patients with lymph node and distant metastases and ornithine was also lower in the case of lymph node involvement ( Table 3). Table 3. Effect of cancer pathology on metabolites associated with arginine metabolism.

Diagnostic Significance of the Pathway
Receiver operating characteristic (ROC) curve analysis was employed to determine the strength of detected associations and diagnostic potential of metabolites associated with arginine metabolism, individually and as a panel. As a discriminator between healthy individuals and cancer patients, DMA and ornithine had high (over 80%) accuracy and their sensitivity was superior over specificity. The "all metabolites" panel had 91% accuracy and was characterized by high sensitivity and specificity. The panel also had superior 70% accuracy, accompanied by fair specificity and sensitivity, in discriminating cancer patients from those with benign gastric disorders. Citrulline was the only metabolite with statistically significant power in discriminating CA and GA patients and the panel had only slightly better accuracy (66% compared to 63% of citrulline) but had improved specificity (Table 5).

Pathway Enzymes in Gastric Cancers
To discern whether altered status of Arg metabolic pathways at systemic level is underlined by local alterations in expression patterns of genes encoding key pathway enzymes, the expression of ASL, ARG1, ARG2, ASS1, DDAH1, DDAH2, NOS2, ODC1, ORNT1, PRMT1, PRMT2, and PRMT5 was examined.

Effect of Cancer Anatomical Site and Pathology on Pathway Enzymes
The upregulation of NOS2 and the downregulation of ORNT1 expression in tumors was markedly higher in CA than in GA. In turn, the downregulation of ASL and PRMT2 in tumors was significant solely in CA (Figure 3).
Using a two-fold change in expression as an arbitrary threshold for up-or downregulation, 52% of patients downregulated and 17% upregulated ASL in tumors (p = 0.005). 17% of patients had downregulated ASS1 and 38% had it upregulated (p = 0.076). ORNT1 was upregulated in 11% of patients as opposed to 59% with downregulated enzyme expression in tumors (p < 0.001). Tumor DDAH1 was upregulated in 11% of patients and downregulated in 33% (p = 0.050) and PRMT2 was upregulated in 15% of patients and downregulated in 33% (p = 0.120). Tumor NOS2 expression was upregulated in 67% of patients and downregulated in 18.5% of patients (p < 0.001).

Effect of Cancer Anatomical Site and Pathology on Pathway Enzymes
The upregulation of NOS2 and the downregulation of ORNT1 expression in tumors was markedly higher in CA than in GA. In turn, the downregulation of ASL and PRMT2 in tumors was significant solely in CA (Figure 3).
(e) (f) Anatomical site affected the expression of ARG1, in both normal and tumor tissue, which was higher by six-fold in CA than in GA (Figure 4).  Tumor expression of NOS2 was decreased in patients with lymph node (by 6.3-fold) and distant metastasis (by 19.7-fold) compared to those without metastases ( Figure 5). Anatomical site affected the expression of ARG1, in both normal and tumor tissue, which was higher by six-fold in CA than in GA (Figure 4).  Anatomical site affected the expression of ARG1, in both normal and tumor tissue, which was higher by six-fold in CA than in GA (Figure 4).  Tumor expression of NOS2 was decreased in patients with lymph node (by 6.3-fold) and distant metastasis (by 19.7-fold) compared to those without metastases ( Figure 5). In addition, ASL downregulation in tumors compared to normal tissue (N/T expression ratio) was more pronounced in M1 than M0 cancers (by 3.2-fold) ( Figure 5).

Interplay between Pathway Enzymes
The expression of NOS2 was not significantly correlated with any other pathway gene in normal mucosa (Table 7), while its association with expression of other genes in tumors was moderate (Table 8). ARG1 was rather poorly related with other pathway enzymes as well. The strongest correlations were between PRMTs and DDAHs, both in tumors and normal tissue, although those in normal tissue were generally stronger. The most marked cancer-related difference seems to be associated with ODC and ARG2 expression patterns, mostly changed from very strong to strong (with DDAHs) or moderate (with PRMTs), and with loosening of ASS1 and ASL association with each other as well as other pathway enzymes. ns Data presented as a heatmap of Spearman correlation coefficients (ρ) with statistical significance indicated as follows: 1 p ≤ 0.0001; 2 p ≤ 0.001; 3 p ≤ 0.01; 4 p < 0.05 and strength of association color-coded ranging from red (strong positive), to yellow, green, and blue (strong negative). ASL, argininosuccinate lyase; ARG, arginase; ASS1, argininosuccinate synthase 1; DDAH, dimethylarginine dimethylaminohydrolase; NOS, nitric oxide synthase; ODC, ornithine decarboxylase; ORNT1, ornithine translocase 1; PRMT, protein arginine methyltransferase; ns, not significant (p ≥ 0.05). In addition, ASL downregulation in tumors compared to normal tissue (N/T expression ratio) was more pronounced in M1 than M0 cancers (by 3.2-fold) ( Figure 5).

Interplay between Pathway Enzymes
The expression of NOS2 was not significantly correlated with any other pathway gene in normal mucosa (Table 7), while its association with expression of other genes in tumors was moderate (Table 8). ARG1 was rather poorly related with other pathway enzymes as well. The strongest correlations were between PRMTs and DDAHs, both in tumors and normal tissue, although those in normal tissue were generally stronger. The most marked cancer-related difference seems to be associated with ODC and ARG2 expression patterns, mostly changed from very strong to strong (with DDAHs) or moderate (with PRMTs), and with loosening of ASS1 and ASL association with each other as well as other pathway enzymes. Data presented as a heatmap of Spearman correlation coefficients (ρ) with statistical significance indicated as follows: 1 p ≤ 0.0001; 2 p ≤ 0.001; 3 p ≤ 0.01; 4 p < 0.05 and strength of association color-coded ranging from red (strong positive), to yellow, green, and blue (strong negative). ASL, argininosuccinate lyase; ARG, arginase; ASS1, argininosuccinate synthase 1; DDAH, dimethylarginine dimethylaminohydrolase; NOS, nitric oxide synthase; ODC, ornithine decarboxylase; ORNT1, ornithine translocase 1; PRMT, protein arginine methyltransferase; ns, not significant (p ≥ 0.05). Data presented as a heatmap of Spearman correlation coefficients (ρ) with statistical significance indicated as follows: 1 p ≤ 0.0001; 2 p ≤ 0.001; 3 p ≤ 0.01; 4 p < 0.05 and strength of association color-coded ranging from red (strong positive) to yellow, green, and blue (strong negative). ASL, argininosuccinate lyase; ARG, arginase; ASS1, argininosuccinate synthase 1; DDAH, dimethylarginine dimethylaminohydrolase; NOS, nitric oxide synthase; ODC, ornithine decarboxylase; ORNT1, ornithine translocase 1; PRMT, protein arginine methyltransferase; ns, not significant (p ≥ 0.05).

Co-Expression with Markers of Proliferation, Survival, Inflammation, Angiogenesis, Metabolic Reprogramming, and Epithelial-Mesenchymal-Transition
The expression of pathway enzymes was-to varying degrees-interrelated with the expression of key genes involved in cancer growth and progression. In univariate analysis (Table 9), all genes except for ASL and ORNT1 were positively correlated with proliferation index Ki67 and anti-apoptotic BCLXL. Ki67 was an independent predictor of ARG2 and PRMT1 expression, while BCLXL was a predictor of DDAH1 and PRMT5 in multiple regression. CDKN1A, encoding cell cycle regulator p21, was positively correlated with ARG2, ODC1, and PRMT5 and was an independent predictor of ODC1 in multiple regression. Except for ASS1, the expression of all pathway enzymes was correlated with mediators of inflammation: CCL2 or PTGS2 (encoding COX2) or both, with PTGS2 being an independent predictor of ORNT1 and CCL2 of PRMT2 and PRMT5. As a marker of metabolic reprogramming, GLUT1 was positively correlated with ASS1, of which it was an independent predictor in multiple regression, along with ARG2, ODC1, PRMT1, and PRMT5. Angiogenic VEGFA positively correlated with ODC1, ORNT1, PRMT1, and PRMT5 and epithelial mesenchymal markers were related to the expression of ARG2, DDAH1, DDAH2, ODC1, and PRMTs. As PRMT5 expression was independently associated with both BCLXL and CCL2, the respective partial correlation coefficients were: r p = 0.79, p = 0.003 and r p = 0.61, p = 0.035. ARG1 and NOS2 did not display any significant correlations.

Discussion
Metabolomic profiling is viewed as a promising approach to discover novel biomarkers facilitating cancer diagnosis and differentiation, monitoring treatment efficacy, and prognostication [5]. In the present study, the analysis of metabolites associated with arginine metabolism clearly indicated that amino acid deficiency occurs at systemic level already in patients with benign gastric disorders, and is further exacerbated in cardiac cancer. Moreover, depletion of citrulline and ornithine was aggravated in cancer patients with lymph node and distant metastases. Consistently, amino acid concentrations were negatively correlated with a set of immune modulators and proinflammatory cytokines as well as cancer-promoting growth factors. A gastric cancer-related systemic drop in arginine and citrulline, but not ornithine, has previously been shown by Miyagi et al. [21], and plasma-free amino acid profiling has been successfully explored as a diagnostic tool. Therefore, we assessed the individual discriminative power of arginine, ornithine, and citrulline and found ornithine to be superior cancer marker with overall accuracy exceeding 80%. Citrulline, in turn, was the only metabolite able to discriminate cancer patients by anatomical subsite of primary tumor, although its power was only moderate. Contrary to amino acids, the other evaluated metabolites-dimethylarginines and DMA-were elevated in patients with benign gastric disorders and SDMA and DMA also in cancer patients. Ornithine, ADMA, and SDMA were moderately efficient in discriminating patients with cancer from those with benign gastric disorders. Among assessed metabolites, DMA was the best individual marker of cancer presence. Still, we demonstrated that concomitant quantification of all metabolites was superior to individual determinations in terms of diagnostic power in overall cancer detection, differentiation between benign and cancerous gastric diseases, and distinguishing CA from GA with, respectively, 91%, 70%, and 66% overall accuracy.
Systemic arginine depletion was accompanied by upregulated local NOS2 expression while the expression levels of ARGs were unaltered between tumors and adjacent mucosa. This observation is consistent with that of Wang et al. [22] who showed upregulated NOS2 in gastric tumors. However, it does not confirm previous findings on the upregulation of ARG in breast [23] and that of ODC in gastric [24] tumors. As NOS2 overexpression has been associated with transformed epithelial cells [22] and that of ARG and ODC with tumor-infiltrating macrophages [23,25], possible low content of these immune cells in tumors analyzed in the present study might account for lack of ARG and ODC upregulation. Unlike immunochemistry, the RTqPCR technique is fully quantitative but does not allow for determining the cellular source of expression. Noteworthily, lack of gene upregulation in tumors compared to adjacent tissue may not indicate lack of cancer-related gene upregulation. It has been repeatedly demonstrated that tumor-adjacent tissue might already have upregulated gene expression, even indicating comparative gene downregulation in tumors, despite lack of morphological and histological changes in its architecture [26][27][28][29]. Such apparent downregulation, resulting from less pronounced upregulation in tumors, has also been noted for ARG1 expression in the colon [7]. No normal gastric mucosa was available in the current study to confirm the speculation, but the notion is supported by higher ODC activity in non-transformed mucosa from GC patients than normal mucosa from healthy individuals reported by others [24].
We confirmed, on a larger set of samples, our previous observation [19] that NOS2 upregulation is greater in cardia subtype of gastric cancer. The downregulation of ORNT1, ASL, and PRMT2 was more evident in cardia subtype as well. Considering the tumorsupporting consequences of enzyme deregulation, this finding might shed some light on molecular background of more aggressive phenotype and worse prognosis of cardia than non-cardia gastric cancer [30].
Elevated concentration of dimethylarginines, more so in benign disorders than in cancer, is in line with inflammatory character of SDMA and the role attributed to ADMA in gastric injury. ADMA has been shown to induce inflammatory response and oxidative stress in gastric mucosa [31,32] and mediate cell migration and invasion via Wnt/β-catenin signaling pathway [33]. The impact of SDMA accumulation is mostly unknown, but in colorectal tumors, it has been linked with greater metastatic potential [34]. Here, both ADMA and SDMA were positively correlated with systemic hepatocyte growth factor (HGF), pivotal for gastric cancer development and progression [35]. SDMA was also correlated with stroma-derived (SCF), the signaling of which is involved in viability and self-renewing properties of cancer stem cells [36]. The interrelationship between intermediates of arginine metabolic pathways and stemness-promoting cytokines, recurring in various cancers [9], is intriguing and worth exploration.
Corroborating findings of others [32,37], systemic elevation of ADMA was accompanied by DDAH1 downregulation. A proneoplastic role has generally been attributed to DDAHs [12] and the only evidence of DDAH1 acting as a tumor suppressor has been shown in gastric cancer [12,37]. Still, a co-expression pattern of DDAH1 observed here might imply a tumor-promoting role. DDAH1 expression was positively correlated with genes encoding proliferation marker Ki67, anti-apoptotic BCLXL, and mesenchymal marker claudin-2 (CLDN2), with BCLXL being an independent predictor of DDAH1 expression. Similar correlation patterns were observed for DDAH2, a dominant endothelial isoform.
DDAH activity yields citrulline and DMA. The status and role of DMA in gastric cancer is largely unknown. Nonetheless, its accumulation is likely disadvantageous as DMA is a precursor of nitrosodimethylamine-a suspected carcinogen-and can be effectively transported from blood into gastric fluid [38]. DMA in our patients was positively correlated with SDMA and with SCGFβ, a recently discovered secreted sulfated glycoprotein of unknown status and role in gastric cancer, which, however, is a marker of drug-resistance in lung and liver cancers [39].
ASS1 and ASL are involved in intracellular de novo synthesis of arginine from citrulline. Counterintuitively, however, number of cancers downregulate ASS1 and become arginine auxotrophic. The proposed advantage of ASS1 downregulation for tumors is the promotion of proliferation under normal conditions, greater invasiveness under hypoxia, and a buildup of glutamine under acidic conditions [40,41]. Herein, ASS1 expression was not significantly affected but, in line with its proposed pro-survival role in gastric cancer, it was positively correlated with Ki67 and BCLXL. Less is known about ASL, which is the only enzyme able to synthesize arginine endogenously and its silencing also results in arginine auxotrophy [15]. To the best of our knowledge, ASL expression in gastric cancer has not been investigated. Here, we found it to be downregulated in tumors by two-fold on average. Moreover, ASL downregulation was more pronounced in cancers metastasizing to distant organs and ASL expression in tumor was inversely correlated with depth of invasion. Downregulation of ASL is of clinical relevance as, if confirmed on a larger set of samples, it implies that gastric cancer, particularly the one located in cardia, might still be sensitive to arginine-deprivation therapies, despite ASS1 overexpression.
Among cancer-related metabolic aberrations, dysregulation of urea cycle enzymes is quite prevalent and linked with worse overall prognosis but better response to immunotherapy based on checkpoint inhibitors [42]. As the enzymes of the cycle compete for nitrogenous substrates with others, loss-of-function mutations in genes encoding ASS1 or ASL or ornithine translocase (ORNT1) facilitate pyrimidine synthesis by dihydrooratase, and results, as a consequence, in increased cell proliferation [42]. Therefore, downregulation of ASL combined with even more accentuated lower expression of ORNT1 might potentially translate into metabolic rewiring promoting pyrimidine synthesis. In line with Lee et al.'s [42] observations, determining expression level of ASL and ORNT1 might therefore help identify gastric cancer patients more likely to benefit from immune checkpoint inhibitor therapy.
Patients with benign gastric disorders had significantly higher systemic concentrations of dimethylarginines and SDMA was higher also in cancer patients. ADMA and SDMA are products of, respectively, type I and type II PRMTs, but dysregulation of dimethylarginines was not reflected locally and PRMT1 (prototypical type I enzyme) and PRMT5 (prototypical type II enzyme) expression levels were unaltered. PRMT upregulation in certain cancer types [9] has been noted and evoked an interest in PRMTs and their inhibitors as potential antineoplastic strategy [43,44]. Still, unaltered PRMT expression agrees well with the housekeeping nature of those enzymes, further underscored by their tight interrelationship observed here. It is worth mentioning, however, that lack of PRMT1 and PRMT5 elevation in tumors might not necessarily mean that enzyme expression is not affected by cancer. As we have previously shown in colorectal cancer, PRMT1 and PRMT5 expression can be upregulated both in colonic tumors and adjacent tissue [7].
Unlike main isoforms, tumor PRMT2 expression was clearly downregulated in CA and reflected the depth of tumor invasion. Contrary to PRMT1 and PRMT5, little is known about PRMT2 and its potential substrates. In fact, enzyme has even been suspected of lack of methyltransferase activity [45]. Nonetheless, it has been shown that PRMT2 may act as a coactivator for various receptors. However, its partners are implicated in opposing activities, either facilitating or inhibiting tumor growth [45]. While PRMT2 status and role in gastric cancer does not seem to be previously investigated, our observation on diminished PRMT2 expression in cardia tumors is in line with antitumor activity played by peroxisome proliferator-activated receptor γ (PPARγ) in gastric carcinogenesis [46], even though the receptor is reportedly upregulated in gastric tumors [47]. Moreover, it also agrees well with growth inhibition exerted by PRMT2 in breast cancer [48].

Conclusions
Metabolic reprogramming in gastric cancer is manifested by aberrant metabolism of arginine, reflecting cancer subtype and pathology, as well as by altered interplay of pathway intermediates and enzymes. Quantifying metabolites associated with arginine metabolism for diagnostic purposes holds promise, but requires validation prior to clinical application. Exploiting upregulation of NOS2 and downregulation of ASL, PRMT2, and ORNT1 for therapeutic purposes requires confirmation on a larger set of samples. Previously not investigated status of ASL, PRMT2, and ORNT1 warrants further functional studies on the role and clinical significance of enzyme downregulation in gastric cancer.