Esophageal Squamous Cell Carcinoma Is Accompanied by Local and Systemic Changes in L-arginine/NO Pathway

The L-arginine/NO pathway holds promise as a source of potential therapy target and biomarker; yet, its status and utility in esophageal squamous cell carcinoma (ESCC) is unclear. We aimed at quantifying pathway metabolites in sera from patients with ESCC (n = 61) and benign conditions (n = 62) using LC-QTOF-MS and enzyme expression in esophageal tumors and matched noncancerous samples (n = 40) using real-time PCR with reference to ESCC pathology and circulating immune/inflammatory mediators, quantified using Luminex xMAP technology. ESCC was associated with elevated systemic arginine and asymmetric dimethylarginine. Citrulline decreased and arginine bioavailability increased along with increasing ESCC advancement. Compared to adjacent tissue, tumors overexpressed ODC1, NOS2, PRMT1, and PRMT5 but had downregulated ARG1, ARG2, and DDAH1. Except for markedly higher NOS2 and lower ODC1 in tumors from M1 patients, the pathology-associated changes in enzyme expression were subtle and present also in noncancerous tissue. Both the local enzyme expression level and systemic metabolite concentration were related to circulating inflammatory and immune mediators, particularly those associated with eosinophils and those promoting viability and self-renewal of cancer stem cells. Metabolic reprogramming in ESCC manifests itself by the altered L-arginine/NO pathway. Upregulation of PRMTs in addition to NOS2 and ODC1 and the pathway link with stemness-promoting cytokines warrants further investigation.


Introduction
Squamous cell carcinoma of the esophagus (ESCC) is the dominant type of esophageal cancerand one of the deadliest and understudied malignancies worldwide. The etiology of ESCC differs by geographic region with smoking and heavy alcohol consumption considered synergistic primary causes in Western European countries. New potential factors, such as polycyclic aromatic hydrocarbons; diet-low in antioxidants and contaminated with mycotoxins and nitrosamines; and the oral microbiome, are still emerging. [1,2] High case fatality rates, approaching 90%, result from delayed Figure 1. A simplified overview of the L-arginine/NO pathway. Pathway enzymes are written in italics and metabolites in a straight script; gray color was used to indicate pathway components not determined in the current study. Inhibitory effects are marked in red, with dashed line if the effect is weak. ADMA, asymmetric dimethylarginine; ARG, arginase; CATs, cationic amino acid transporter; DDAH, dimethylarginine dimethylaminohydrolase; DMA, dimethylamine; NOS, nitric oxide synthase; ODC, ornithine decarboxylase; PM, plasma membrane; PRMT, protein methyltransferase.

L-arginine/NO Pathway Metabolites in ESCC and Benign Conditions
Systemic concentration of the pathway metabolites was determined using LC-QTOF-MS (liquid chromatography quadrupole time-of-flight mass spectrometry) in 61 patients with ESCC and 62 individuals with benign conditions of the esophagus. Both groups were well age and sex matched. The ESCC patients had higher arginine and ADMA than individuals with benign conditions of the esophagus. There was no significant difference in the Arg/ADMA ratio, indicative of arginine availability for NO synthesis, but the general arginine bioavailability was greater in ESCC as indicated by the higher Arg/(Cit+Orn) ratio. The other metabolites, that is, citrulline, ornithine, SDMA, and DMA, did not differ between groups (Figure 2).
(a) (b) Figure 1. A simplified overview of the L-arginine/NO pathway. Pathway enzymes are written in italics and metabolites in a straight script; gray color was used to indicate pathway components not determined in the current study. Inhibitory effects are marked in red, with dashed line if the effect is weak. ADMA, asymmetric dimethylarginine; ARG, arginase; CATs, cationic amino acid transporter; DDAH, dimethylarginine dimethylaminohydrolase; DMA, dimethylamine; NOS, nitric oxide synthase; ODC, ornithine decarboxylase; PM, plasma membrane; PRMT, protein methyltransferase.

L-arginine/NO Pathway Metabolites in ESCC and Benign Conditions
Systemic concentration of the pathway metabolites was determined using LC-QTOF-MS (liquid chromatography quadrupole time-of-flight mass spectrometry) in 61 patients with ESCC and 62 individuals with benign conditions of the esophagus. Both groups were well age and sex matched. The ESCC patients had higher arginine and ADMA than individuals with benign conditions of the esophagus. There was no significant difference in the Arg/ADMA ratio, indicative of arginine availability for NO synthesis, but the general arginine bioavailability was greater in ESCC as indicated by the higher Arg/(Cit+Orn) ratio. The other metabolites, that is, citrulline, ornithine, SDMA, and DMA, did not differ between groups ( Figure 2).
There were 16 patients with resected tumors of esophagus admitted three months postsurgery for the esophagoplasty. We compared the systemic concentration of metabolites between those patients and the ESCC group and found that ESCC patients tended to have higher arginine (124. individuals with benign conditions of the esophagus. Both groups were well age and sex matched. The ESCC patients had higher arginine and ADMA than individuals with benign conditions of the esophagus. There was no significant difference in the Arg/ADMA ratio, indicative of arginine availability for NO synthesis, but the general arginine bioavailability was greater in ESCC as indicated by the higher Arg/(Cit+Orn) ratio. The other metabolites, that is, citrulline, ornithine, SDMA, and DMA, did not differ between groups (Figure 2   (g) arginine-to-ADMA ratio (Arg/ADMA); (h) arginine-to-(citrulline+ornithine) (Arg/(Cit+Orn). Data presented as the means with 95% confidence interval (red triangles with whiskers as well as numeric data) and analyzed using a t-test for independent samples. The analysis of the metabolite concentration against ESCC advancement showed that only citrulline was significantly associated with the disease overall stage (TNM) as well as all its components: primary tumor extension, lymph node involvement, and distant metastasis. The metabolite concentration dropped along with increasing stage (ρ = −0.55, p < 0.0001), being significantly lower in stage IV than I or II, and along with primary tumor extension (ρ = −0.47, p = 0.0001), being significantly lower in T4 than T1 or T3. Citrulline was significantly lower in ESCC patients with lymph node or distant metastases ( Figure 3). lower in stage IV than I or II, and along with primary tumor extension (ρ = −0.47, p = 0.0001), being significantly lower in T4 than T1 or T3. Citrulline was significantly lower in ESCC patients with lymph node or distant metastases (Figure 3). . Data presented as the means with 95% confidence interval (red triangles with whiskers as well as numeric data) and analyzed using one-way ANOVA. a, significantly different from T1; b, significantly different from T2; c, significantly different from T3; d, significantly different from T4.
Except for arginine elevation in patients with T3 cancers, no other metabolite was significantly associated with the disease advancement. Consequently, arginine-based ratios, Arg/ADMA and Arg/(Cit+Orn), were significantly higher in T3 cancers as well ( Figure 4). Owing to the gradual citrulline depletion with the disease advancing, Arg/(Cit+Orn) correlated with TNM (ρ = 0.27, p = 0.036) and T (ρ = 0.29, p = 0.023). The association between the arginine and T stage remained significant (p = 0.029) following the removal of two outlying observations in T3 cancers, although the difference between T2 and T3 lost significance.  . Data presented as the means with 95% confidence interval (red triangles with whiskers as well as numeric data) and analyzed using one-way ANOVA. a, significantly different from T1; b, significantly different from T2; c, significantly different from T3; d, significantly different from T4.
Except for arginine elevation in patients with T3 cancers, no other metabolite was significantly associated with the disease advancement. Consequently, arginine-based ratios, Arg/ADMA and Arg/(Cit+Orn), were significantly higher in T3 cancers as well ( Figure 4). Owing to the gradual citrulline depletion with the disease advancing, Arg/(Cit+Orn) correlated with TNM (ρ = 0.27, p = 0.036) and T (ρ = 0.29, p = 0.023). The association between the arginine and T stage remained significant (p = 0.029) following the removal of two outlying observations in T3 cancers, although the difference between T2 and T3 lost significance. associated with the disease advancement. Consequently, arginine-based ratios, Arg/ADMA and Arg/(Cit+Orn), were significantly higher in T3 cancers as well ( Figure 4). Owing to the gradual citrulline depletion with the disease advancing, Arg/(Cit+Orn) correlated with TNM (ρ = 0.27, p = 0.036) and T (ρ = 0.29, p = 0.023). The association between the arginine and T stage remained significant (p = 0.029) following the removal of two outlying observations in T3 cancers, although the difference between T2 and T3 lost significance.  (b) arginine-to-ADMA ratio (Arg/ADMA); (c) arginine-to-(citrulline+ornithine) (Arg/(Cit+Orn)). Data presented as the means with 95% confidence interval (red triangles with whiskers as well as numeric data) and analyzed using one-way ANOVA. a, significantly different from T1; b, significantly different from T2; c, significantly different from T3; d, significantly different from T4.

L-arginine/NO Pathway Metabolites as Differential Markers
Pathway metabolites, which were significantly different in patients with ESCC and benign esophageal conditions, were analyzed as potential differential markers using receiver operating curve (ROC) analysis. The individual performance of the arginine, ADMA, and Arg/(Cit+Orn) ratio was very comparable, although arginine and Arg/(Cit+Orn) had superior specificity and ADMA-sensitivity. Logistic regression (stepwise method) was applied to identify the independent predictors of ESCC. The ADMA and Arg/(Cit-Orn) were selected, and predicted probabilities were used in a ROC curve analysis to assess the diagnostic power of their concomitant assessment. The overall accuracy of the ADMA and Arg/(Cit-Orn) panel improved only minimally ( Figure 5). presented as the means with 95% confidence interval (red triangles with whiskers as well as numeric data) and analyzed using one-way ANOVA. a, significantly different from T1; b, significantly different from T2; c, significantly different from T3; d, significantly different from T4.

L-arginine/NO Pathway Metabolites as Differential Markers
Pathway metabolites, which were significantly different in patients with ESCC and benign esophageal conditions, were analyzed as potential differential markers using receiver operating curve (ROC) analysis. The individual performance of the arginine, ADMA, and Arg/(Cit+Orn) ratio was very comparable, although arginine and Arg/(Cit+Orn) had superior specificity and ADMA-sensitivity. Logistic regression (stepwise method) was applied to identify the independent predictors of ESCC. The ADMA and Arg/(Cit-Orn) were selected, and predicted probabilities were used in a ROC curve analysis to assess the diagnostic power of their concomitant assessment. The overall accuracy of the ADMA and Arg/(Cit-Orn) panel improved only minimally ( Figure 5).  (Table 1). ADMA was positively correlated with all other metabolites in both ESCC and patients with benign conditions; although in ESCC, the correlation was stronger with arginine and SDMA and in patients with benign conditions with citrulline. Arginine was correlated positively with citrulline in both (c) arginine-to-(citrulline+ornithine) ratio (Arg/(Cit+Orn)); (d) panel consisting of ADMA and Arg/(Cit+Orn). Graphs present ROC curve (solid curve) with 95% confidence interval (dotted curves) and an optimal cut-off value (red dot). Performance of a chance marker devoid of diagnostic power (AUC = 0.5) is indicated by a diagonal line. AUC, area under ROC curve; Sens., sensitivity; spec., specificity.

Interrelationship between Systemic Concentrations of L-arginine/NO Pathway Metabolites
The interrelationship between systemic concentrations of L-arginine/NO pathway metabolites in ESCC patients and individuals with benign conditions of the esophagus was compared (Table 1). ADMA was positively correlated with all other metabolites in both ESCC and patients with benign conditions; although in ESCC, the correlation was stronger with arginine and SDMA and in patients with benign conditions with citrulline. Arginine was correlated positively with citrulline in both groups but with SDMA only in ESCC. Citrulline correlated with arginine and tended to with ADMA in ESCC; while in benign conditions, its correlation with ADMA was stronger, and there was positive correlation with DMA as well. DMA was the most strongly correlated with SDMA in both groups and with ADMA in ESCC rather than with citrulline like in the benign group. Ornithine correlated weakly with ADMA in both groups and with SDMA in ESCC patients. SDMA was correlated with ADMA and SDMA in both groups but only with arginine and ornithine in ESCC.

Pairwise Analysis of Enzyme Expression in Tumor and Adjacent Tissue
Quantitative (real-time) polymerase chain reaction (PCR) with SYBR Green chemistry was used to determine the relative expression level of ARG1 and ARG2, DDAH1 and DDAH2, NOS2, ODC1, and PRMT1 and PRMT5 in 40 patient-matched samples from esophageal tumors and adjacent, macroscopically normal mucosa.

Pairwise Analysis of Enzyme Expression in Tumor and Adjacent Tissue
Quantitative (real-time) polymerase chain reaction (PCR) with SYBR Green chemistry was used to determine the relative expression level of ARG1 and ARG2, DDAH1 and DDAH2, NOS2, ODC1, and PRMT1 and PRMT5 in 40 patient-matched samples from esophageal tumors and adjacent, macroscopically normal mucosa.

Association between Fold Change in Enzyme Expression and ESCC Pathology
Potential relationship between fold change in enzyme expression (tumor-to-adjacent) and ESCC pathology was examined (Table 3).
Only the fold change in DDAH2, ODC1, and PRMT5 expression differed significantly with overall TNM stage, with tumor-to-adjacent ratio decreasing in a stepwise manner along with increasing stage. The same pattern was present for an association between depth of tumor invasion (T) and fold change in DDAH2 and ODC1. Fold change in ARG2 (by 1.9-fold), DDAH2 (by 3.8-fold), NOS2 (by 5.0-fold), ODC1 (by 2.2-fold), and PRMT5 (by 2.3-fold) tended to decrease, and that of PRMT1 was significantly lower (by 1.6-fold) in cancers metastasizing into lymph nodes (N≥1). Fold change in ODC1 was significantly lower (by 4.1-fold) in cancers metastasizing to distant organs as well (Table 3). Table 3. Association between expression ratio (tumor-to-adjacent) of L-arginine/NO pathway enzymes and esophageal squamous cell carcinoma pathology. We investigated whether inverse relationship between expression ratios (tumor-to-adjacent) and cancer advancement results from changes in gene expression in tumor (Table 4) or/and in adjacent noncancerous tissue (Table 5).
This analysis revealed that a negative correlation between ODC1 expression ratio and TNM resulted from the concomitant downregulation of enzyme expression in tumors along with the increasing stage (Table 4) and its upregulation in noncancerous adjacent tissue ( Table 5). The (h) PRMT5. Data analyzed using a t-test for paired samples and presented as means of normalized relative quantities (NRQ) with 95% confidence interval. ARG, arginase; DDAH, dimethylarginine dimethylaminohydrolase; NOS, nitric oxide synthase; ODC, ornithine decarboxylase; PRMT, protein arginine N-methyltransferase.

Association between Fold Change in Enzyme Expression and ESCC Pathology
Potential relationship between fold change in enzyme expression (tumor-to-adjacent) and ESCC pathology was examined (Table 3). Table 3. Association between expression ratio (tumor-to-adjacent) of L-arginine/NO pathway enzymes and esophageal squamous cell carcinoma pathology. Only the fold change in DDAH2, ODC1, and PRMT5 expression differed significantly with overall TNM stage, with tumor-to-adjacent ratio decreasing in a stepwise manner along with increasing stage. The same pattern was present for an association between depth of tumor invasion (T) and fold change in DDAH2 and ODC1. Fold change in ARG2 (by 1.9-fold), DDAH2 (by 3.8-fold), NOS2 (by 5.0-fold), ODC1 (by 2.2-fold), and PRMT5 (by 2.3-fold) tended to decrease, and that of PRMT1 was significantly lower (by 1.6-fold) in cancers metastasizing into lymph nodes (N≥1). Fold change in ODC1 was significantly lower (by 4.1-fold) in cancers metastasizing to distant organs as well (Table 3).
We investigated whether inverse relationship between expression ratios (tumor-to-adjacent) and cancer advancement results from changes in gene expression in tumor (Table 4) or/and in adjacent noncancerous tissue (Table 5).  This analysis revealed that a negative correlation between ODC1 expression ratio and TNM resulted from the concomitant downregulation of enzyme expression in tumors along with the increasing stage (Table 4) and its upregulation in noncancerous adjacent tissue ( Table 5). The expression of DDAH2 did not change in tumors but tended to increase along with increasing TNM in the adjacent tissue (Table 5), contributing to a significantly negative correlation between expression rate and stage (Table 3). Additionally, ARG2 expression in adjacent tissue tended to increase with advancing disease ( Table 5).
The clear tendency of expression ratios being lower in N≥1 than N0 cancers also seem to result from combined slightly lower enzyme expression in tumors (Table 4) and higher in adjacent noncancerous tissue ( Table 5) derived from of N≥1 patients.
Only lower ODC1 expression ratio in M1 than M0 cancers (Table 3) resulted from significantly lower gene expression in tumors from M1 patients (by 3.5-fold) ( Table 4). The presence of distant metastases was, in turn, associated with significantly higher expression of NOS2 in tumors (by 5.1-fold) ( Table 4). It was also associated with a tendency towards a higher expression of ARG1 and DDAH1 in noncancerous tumor-adjacent tissue (Table 5).

Interrelationship between Local Expression Levels of L-arginine/NO Pathway Enzymes
Correlation patterns in the enzyme expression in tumor (Table 6) and noncancerous adjacent tissue were compared (Table 7). In tumor tissue, both ARGs were correlated positively with both DDAHs, while in adjacent tissue, ARG1 was correlated solely with DDAH1. In tumors, NOS2 correlated with both PRMTs, while in adjacent tissue, only with PRMT1 and additionally with ODC1. In tumors, ODC1 correlated with both ARGs, DDAH2, and PRMT5, while in adjacent tissue, solely with ARG2 isoenzyme, with DDAH2, NOS2, and both PRMTs. The expression pattern for DDAH1 was similar, although the associations with ARG1 and PRMTs were weaker in tumors. In turn, DDAH2 in tumors was positively correlated with all genes, while in adjacent tissue, there was no correlation with ARG1 and NOS2. The expression of PRMT5 in tumors correlated with NOS2 but not ARG1, while in adjacent tissue, with ARG1 but not NOS2. The expression of PRMT1 in tumors did not correlate with ODC1, while the association was present in tumor-adjacent tissue.  Data presented as Spearman rank correlation coefficient rho (ρ). ARG, arginase; DDAH, dimethylarginine dimethylaminohydrolase; NOS, nitric oxide synthase; ODC, ornithine decarboxylase; PRMT, protein arginine N-methyltransferase; 1 , p < 0.05; 2 , p < 0.01; 3 , p < 0.001; -, no significant association or tendency (p > 0.1).

Correlation between Local Expression Levels of L-arginine/NO Pathway Enzymes and Circulating Cytokines, Chemokines, and Growth Factors as Well as Systemic Concentrations of Pathway Metabolites
The correlation pattern between enzyme expression in tumor (Table 8) and noncancerous tumor-adjacent tissue (Table 9) and systemic concentrations of cytokines, chemokines, and growth factors was examined (n = 36). There was a negative correlation between PRMT1 expression and ADMA in tumors. Also, ARG1 tended to positively correlate with ADMA and arginine and ARG2 with citrulline and SDMA. Ornithine negatively correlated with DDAH1 (Table 8). In noncancerous tissue, ADMA tended to correlate with ARG1 (Table 9).
Regarding circulating cytokines, ARG1 in tumors positively correlated with HGF and IL-18 and ARG2 with SDF-1α. DDAH1 displayed a negative correlation with FGF2, while DDAH2 did not show any significant associations. NOS2 positively correlated with IL-4, IL-5, and IL-6 and ODC1 with MCP-1. No significant associations could be observed for PRMTs (Table 8).

Discussion
In the era of precision medicine, metabolic reprogramming of cancer cells is considered a promising source of potential biomarkers and novel molecular targets for antineoplastic strategies [15]. In this respect, the L-arginine/NO pathway becomes an area of active investigation. Its focus is shared between the arginine/ornithine/polyamines rout, the interest in which has recently rekindled, and arginine methylation, its methylated derivatives, and their metabolism, an understudied topic in cancer context. The present study provides a comprehensive overview of pathway status in ESCC, including local enzyme expression and systemic metabolites, with reference to a broad spectrum of inflammatory, immune, and angiogenic mediators. Corroborating our own preliminary observations [6] as well as findings of others [16,17], ODC1 expression in ESCC tumors was markedly upregulated as compared to adjacent noncancerous tissue. As a rate-limiting enzyme of the polyamine pathway and a down-stream target of c-Myc, ODC is implicated in facilitating tumor growth [18]. Moreover, its overexpression can induce neoplastic transformation, making ODC1 an oncogene [19]. Accordingly, ODC1 expression in the colon is upregulated in conditions associated with increased risk of cancer [8]. In turn, the enzyme inhibitors used as chemoprevention, alone or in combination with nonsteroidal anti-inflammatory drugs, have shown promising results in animal models of CRC [20] as well as in clinical trials [18,21]. They reduced the number and size of tumors in test animals or prevented formation of sporadic adenomas, respectively. The enzyme has recently been shown to promote ESCC as well. It induces cell proliferation and survival, as the gene silencing or enzyme inhibition with the substrate analogue, difluoromethylornithine, results in cell arrest in the G2/M phase and triggers apoptosis [17]. The advocated enzyme involvement in the initial phases of carcinogenesis would explain a significant enzyme upregulation in tumors and yet a negative correlation of expression rate with ESCC advancement observed here. However, He et al. [17] reported ODC protein to be more pronouncedly expressed in samples obtained from patients with stage III or N1 cancers than with stage II or N0. When ODC1 expression patterns were inspected more thoroughly, we found that ODC1 expression in patients with more advanced cancers decreased in tumors but increased in adjacent tissue. As the changes affect both normal tissue and the tumor, and they occur in opposite directions, unsurprisingly, systemic ornithine did not differ with respect to ESCC stage.
Likewise, a negative correlation between the disease advancement, particularly lymph node involvement, was observed for DDAHs and PRMTs. Similarly to ODC1, it resulted from gene downregulation in tumors with concomitant upregulation in tumor-adjacent noncancerous tissue. The phenomenon of so-called "tumor molecular margin", describing alterations in molecular landscape in the tissue surrounding tumor, is increasingly recognized and has previously been documented also in ESCC [22]. This phenomenon is of clinical relevance as it is held responsible, at least in part, for synchronous tumors and cancer recurrence following treatment [23,24]. Moreover, it is considered to better represent dysregulations leading to neoplastic transformation than changes encountered in already transformed tissue. As such, studying the tumor molecular margin is deemed better suited to aid the discovery of potential targets for chemoprevention [24]. Regarding L-arginine/NO pathway enzymes, our group showed DDAHs and PRMTs in CRC to be apparently downregulated, as their relative expression was lower in tumors than in adjacent tissue, while they were in fact upregulated in both. There were no normal esophageal mucosa samples available for comparison in the present study, but the observation on adjacent tissue not being "inert" but displaying an increasing enzyme expression with advancing disease seems to indicate that ESCC might be like CRC in this respect. Still, a lack of control mucosa samples from healthy individuals should be recognized as a limitation of the current study.
The analysis of RNA-seq expression data gathered by The Cancer Genome Atlas (TCGA) Research Network and the Genotype-Tissue Expression (GTEx) project indicated DDAH1 upregulation in ESCC [25], although confirming individual studies, analyzing DDAH1 association with reference to ESCC pathology, seem to be missing. The DDAH2 status, as well as the precise role of either isoform in cancer, remains largely unknown. Here, DDAH1 was slightly downregulated in tumors, and DDAH2 did not show significant difference as compared to the surrounding tissue. Their expression levels were interrelated, but there was variation in their correlation pattern with other genes as well as circulating cytokines, supporting the notion that neither their distribution nor function is completely overlapping [25]. Both enzymes catabolize ADMA, but their spatial transcriptional patterns differ and they may be altered in cancer in a different, sometimes opposing, manner [25]. The overexpression of DDAH1 and DDAH2 is linked with the promotion of angiogenesis by the indirect stimulation of VEGF-A expression and NO synthesis, via degrading NOS inhibitor-ADMA. Consequently, their targeting is viewed as an emerging anticancer strategy [25,26]. However, neither DDAH1 nor DDAH2 correlated with circulating VEGF-A in the present study. On the contrary, DDAH1 in tumors exhibited a negative correlation with another angiogenic factor, FGF2, and DDAH2 in tumor-adjacent tissue-a positive association with an angiostatic IP-10. Moreover, the expression of DDAH2 only tended to weakly correlate with NOS2 in tumors, although in prostatic cancer cell lines, its expression was accompanied by the upregulation of iNOS and VEGF-A [27]. As mentioned, ADMA is considered to inhibit angiogenesis, but its role in cancer is unclear. Extensively investigated in cardiometabolic diseases [28], it has been understudied in cancer. However, limited evidence indicates that ADMA accumulate in tumor and surrounding tissue [29] and may support tumor growth by protecting cancer cells from nutritional stress and drug-induced death [30]. Here, consistently with a cancer-promoting role, its systemic concentrations were elevated. Moreover, ADMA levels positively correlated with immune mediators known to support cancer development, the most strongly with GROα (CXCL1), the cytokine released from tumors as well as tumor-associated macrophages and implicated in the recruitment of tumor-associated neutrophils [31] and promotion of metastasis [32], respectively, and associated with poor overall survival [33].
The expression of various PRMT isoenzymes in cancer is also largely unknown, but gaining attention as potential targets for chemoprevention [34,35]. Their targeting is also investigated as potential strategy for overcoming cancer chemoresistance [36]. It becomes increasingly apparent that the role of PRMTs is not limited to NO synthesis. The enzymes have been implicated in the global regulation of RNA splicing and translation [37]. Both PRMT1 and PRMT5 support tumor growth and are, therefore, overexpressed in a number of cancers [34]. Corroborating our observation, the expression of PRMT1 is upregulated in ESCC [38,39] and head and neck tumors [40]. Mechanistically, it facilitates cell proliferation and migration of oral and esophageal squamous cell carcinoma cells [38,40]. In addition, Zhao et al. [39] showed PRMT1 to be preferentially expressed in esophageal tumor initiating cells and function to enhance the self-renewal features, tumorigenicity, and chemoresistance of ESCC. The role of PRMT5 in cancer is even more poorly understood. Mostly oncogenic, owing to promoting cancer cell proliferation and migration, in breast cancer it prevents metastasis, while data regarding prostate cancer are contradictory (reviewed in [41]). Only recently, PRMT5 has been shown to regulate Hsp90A, a known cancer-related chaperone providing protection for a number of oncoproteins [42]. There seems to be no previous reports on PRMT5 expression in ESCC. We found it upregulated in tumors as compared to adjacent tissue. In cancers metastasizing to lymph nodes, its expression increased in tumor adjacent tissue. The expression of PRMT5 was tightly associated with that of PRMT1, more so in adjacent tissue than tumors. Of the other pathway enzymes, PRMT5 expression was associated closely with that of ARG2 and DDAHs, in tumors more so with DDAH2 than DDAH1. In turn, a correlation with NOS2 could be observed in ESCC tumors. Reflecting enzyme ability to induce expression of proinflammatory genes [43], both PRMTs were positively correlated with circulating MIP-1β.
Arginases compete with NOS enzymes for arginine and, as NOS activity is considered cancer-promoting, strategies based on upregulating endogenous arginase activity or introduction of exogenous arginine-consuming enzymes are currently being tested [11,12]. Moreover, arginases synthesize ornithine and, therefore, supply substrates for polyamine biosynthesis [44]. Accordingly, ARG2 expression level, and in tumors also that of ARG1, was positively correlated with ODC1, a key regulatory enzyme of the pathway. However, arginine is necessary for the proper functioning of immune cells, and its depletion is a strategy of immune evasion employed by cancer cells [11,45,46]. Moreover, arginase-2 is expressed by immunosuppressive M2 polarized macrophages and dendritic cells as well as myeloid-derived suppressor cells [47]. In the sample set investigated, ARG1 was substantially while ARG2 borderline significantly downregulated in tumors. Consistent with counter-regulatory mechanisms in the ARG/NOS interplay [44], NOS2 was markedly upregulated and ARG1 tended to positively correlate with the systemic concentration of ADMA, a NOS inhibitor. Moreover, NOS2 was the only pathway enzyme clearly expressed at a higher level in metastasizing tumors. In line with its proinflammatory character, NOS2 expression correlated positively with circulating immune and inflammatory mediators, whilst ARG1, a hepatocyte isoform [44], correlated with HGF, positively in tumors but negatively in adjacent tissue. Consistent with this observation, HGF/c-met signaling axis was shown to induce macrophage M2 polarization as its inhibition upregulated NOS2 and other M1 polarization markers and downregulated ARG1 [48]. The expression of ARG2, but not ARG1, has been shown to be upregulated by hypoxia and implicated in mediating hypoxia-induced cancer cell proliferation [49]. Moreover, cancer-associated fibroblasts and tumor-associated M2 macrophages from hypoxic regions overexpress ARG2 as well [50]. In macrophages, HIF-1α controls iNOS while HIF2α-arginase [51]. Supporting the ARG2 link with hypoxia and angiogenesis, ARG2 expression in evaluated tumors positively correlated with SDF-1α, a hypoxia-induced proangiogenic factor [52], although its relation to VEGF-A tended to be negative.
Cancer-associated upregulation of arginase -1 contributes to a diminished extracellular arginine pool [11]. Here, consistently with decreased ARG1 and ARG2 expression, systemic arginine was elevated. Likewise, the arginine-to-citrulline and ornithine ratio was higher, indicative of increased arginine bioavailability in ESCC. A weak positive correlation between amino acid bioavailability and the disease stage resulted from gradually decreasing citrulline concentration, implying accelerating activity of the arginine-citrulline pathway in parallel with ESCC advancement. Arginine can be regenerated from citrulline by argininosuccinate synthetase (ASL) and argininosuccinate lyase (ASS1), and ASS1 is a rate-limiting enzyme of this pathway [53]. As ASS1 expression, especially leukocyte enzyme, is controlled by inflammatory and immune mediators [53], the systemic citrulline concentration in our patients was tightly but inversely related to their level. The expression of ASS1 is upregulated also by p53, in order to ensure cell survival under genotoxic stress conditions [54]. Still, ASS1 is reported to be epigenetically silenced in some cancer types [55], making them sensitive to antineoplastic therapy based on arginine-deprivation. Nonaffected arginine bioavailability, due to the efficient arginine-citrulline recycling pathway demonstrated here, would indicate that ESCC is not a suitable candidate.
Elevated plasma concentrations of arginine, dimethylarginines (pooled ADMA and SDMA), and N-acetylated derivatives of putrescine and ornithine have been observed in children with eosinophilic esophagitis [56]. In line with the implicated close relationship between eosinophils and the pathway metabolites, both ARGs tended to positively correlate with IL-3 in tumors. The expression of ARG1 in adjacent tissue also correlated with other eosinophil-associated cytokines such as IL-13 and GM-CSF and that of NOS2 with IL-5. In addition, ADMA positively correlated with IL-3 and SDMA also with IL-13 and GM-CSF.
Correlation analysis showed that DDAH2, NOS2, and ODC1 expression in normal tissue was positively correlated with circulating LIF. In addition, LIF concentration was correlated also with SDMA and, consistently with a possible positive effect on DDAH2, with DMA. The cytokine is regarded as an oncogene, facilitating the self-renewal of tumor-initiating cells, supporting cancer-associated fibroblasts, and inducing resistance to radio-and chemotherapy [57][58][59]. It has also been found to hamper anti-PD1 therapy by upregulating MIG in tumor-associated macrophages [60]. Consistently, the ADMA and SDMA concentration in our ESCC patients was correlated with that of MIG as well. Intriguingly, various components of the L-arginine/NO pathway were related to the factors associated with cancer cell renewal. The expression of ODC1 was associated with SCF, as was arginine and, more so, SDMA and DMA concentration. The SCF and its receptor c-Kit (CD117) ensure cancer stem cell viability and self-renewing properties. It might be membrane-bound and expressed in cancer cells or tumor-associated macrophages and cancer-associated fibroblasts, from which a soluble form of SCF can be released [61]. The expression of DDAH2 and systemic SDMA and DMA were, in turn, associated with circulating SCGFβ, a growth factor for primitive hematopoietic progenitor cells, which is overexpressed in cancer by circulating cancer cells [62] and drug-resistant cancer stem cells [63].
Metabolomics is gaining interest as a potential tool in biomarker discovery [64], especially in that it enables the concomitant quantification of a predetermined panel of metabolites. The utility of individual intermediates of the L-arginine/NO pathway as biomarkers for various conditions has repeatedly been demonstrated [56,[65][66][67]. In cancer, the simultaneous quantification of SDMA, citrulline, and DMA proved excellent in detecting CRC, and the changes in the early postoperative period in arginine, ADMA, and SDMA were an accurate marker of surgical complications [7]. In ESCC, however, neither individual metabolites nor a panel consisting of ADMA and Arg/(Cit+Orn) index possessed satisfactory diagnostic power to be of clinical use.

Sample Collection
Biobanked material, collected in the Department of Gastrointestinal and General Surgery of Wroclaw Medical University between 2010 and 2015 (prior the introduction of radiotherapy as a treatment preceding surgery), was used in the present study.
Whole blood was drawn from patients by venipuncture into serum separator tubes following overnight fasting and prior to any treatment. Collected blood was clotted for 30 min at room temperature and centrifuged (1500× g for 10 min at room temperature). Obtained sera were aliquoted and stored at −45 • C until examination.
Patient-matched tumor and macroscopically normal mucosa were rinsed with PBS and immersed in RNAlater (Ambion Inc., Austin, TX, USA). Solution-soaked tissue samples were then stored at −80 • C until RNA isolation.

Study Population-Metabolomic Analysis
The study population for metabolomic analysis consisted of 123 patients admitted to the Department of Gastrointestinal and General Surgery of Wroclaw Medical University for the diagnosis and treatment of histologically confirmed ESCC (n = 61; curative surgery or palliative treatment) or benign esophageal conditions (n = 62), including esophagoplasty following esophagectomy (n = 16) or thermal burn (n = 1), achalasia (n = 25), stenosis (n = 2), esophageal lipoma (n = 2), and Zenker's diverticulum (n = 16). Cancer patients underwent a standard diagnostic procedure, including blood work, physical examination, and imaging techniques, such as ultrasonography, computed tomography, and magnetic resonance. Cancers were staged clinically using the 7th edition of the Union for International Cancer Control TNM system. Detailed population characteristics are presented in Table 10.

Study Population-Transcriptomic Analysis
The study population for transcriptomic analysis consisted of 40 ESCC patients undergoing curative tumor resection in the Department of Gastrointestinal and General Surgery of Wroclaw Medical University. Patients with any severe systemic illness, with gross metastatic disease, or subjected to radio-or chemotherapy were not included. Patients were subjected to standard preoperative evaluation (blood work, physical examination, and imaging techniques, such as ultrasonography, computed tomography, and magnetic resonance). Cancers were staged pathologically using the 7th edition of the Union for International Cancer Control TNM system. In all cases, the resection margins were confirmed to be tumor-free. Detailed population characteristics are depicted in Table 11.

Ethical Considerations
The sample collection was approved by the Medical Ethics Committee of Wroclaw Medical University (#KB 28/2011 and #KB 784/2012). The study was conducted in accordance with the Helsinki Declaration of 1975, as revised in 1983, and informed consent was obtained from all study participants.

LC-QTOF-MS Analysis of Selected L-arginine/NO Pathway Metabolites
The stock solutions of DMA, arginine, ADMA, SDMA, citrulline, ornithine, and isotope labeled standards were prepared in water and stored at −20 • C. Standard calibration curves were prepared by diluting the stock solutions in water in the following concentration ranges: 3.0-150 µM for ornithine, 5.0-250 µM for arginine, 0.05-2.5 µM for ADMA and SDMA, 1.0-50 µM for citrulline, and 0.14-7.0 µM for DMA.
Serum samples and calibration standards were prepared using previously validated and described methods [7,8,[68][69][70][71]. Analysis of L-arginine/NO pathway metabolites as benzamide derivatives allows for the chromatographic separation of highly polar substances using a reversed phase system with relatively high retention. Importantly, it also enabled the chromatographic separation of ADMA and SDMA. The extraction and derivatization procedure was conducted as follows: 100µL aliquots of the calibration standards or serum were mixed with 10 µL internal standard solution (50 µM D6-DMA, 20 µM D7-ADMA, 100 µM D7-arginine, and 70 µM D6-ornithine, respectively) and 50 µL borate buffer (0.025 M Na 2 B 4 O 7 ·10H 2 O, 1.77 mM NaOH, pH = 9.2) and vortexed (1 min, 25 • C). The samples were derivatized with 10% BCl in acetonitrile (400 µL) at 25 • C for 10 min. The mixture was centrifuged at 10,000 RPM for 7 min in 4 • C. Two microliters of supernatants diluted with water in a 1:5 ratio was injected into an LC system consisting of a Waters nanoACQUITY UPLC and a nanoAcquity HSS T3 column (C18-phase, internal diameter 1 mm, length 50 mm, particle size 1.8 µm). They were eluted at a flow rate of 0.22 mL/min with a linear gradient of 0.1% formic acid in water (A) and 0.1% formic acid in methanol (B) performed as follows: 3% B for 1.5 min, from 3% to14% B in 2.0 min, from 14% to 60% B in 1.5 min, from 60% to 90% B in 0.5 min, 90% B for 1.0 min, and from 90% to 3% B in 0.10 min; conditioning time was 1.9 min.
The eluates were subjected to analysis by a quadrupole-time-of-flight mass spectrometer (Xevo G2 QTOF MS, Waters) equipped with an electrospray ionization (ESI) source. A voltage of 0.5 kV was applied to the electrospray capillary. The source temperature was at 120 • C, and the desolvation temperature was 450 • C. Nitrogen was used as the desolvation gas (650 L/h) and as the cone gas (65 L/h). The total ion current was recorded in positive ionization mode, with a scan range from 140 to 600 m/z.  The 27-plex included the following analytes: eotaxin, IL-1β, IL-1ra, IL-2, IL-4, IL-5, IL-6, IL-7, IL-8,  IL-9, IL-10, IL-12p70, IL-13, IL-15, IL- The concentration of IL-1α, IL-12p40, MCP-3, M-CSF, and TNF-β was below the limit of detection in most of the samples, and they were, therefore, excluded from analysis. All analyses were conducted in duplicates and following the manufacturer's instructions. Standard curves were drawn using 5-PL logistic regression, and the data were analyzed using BioPlex Manager 6.0 software.
Quantitative PCR (qPCR) was conducted on a CFX96 Real-Time PCR thermocycler (BioRad) using SsoFast EvaGreen ® Supermix (BioRad). The following cycling conditions were applied: 30 s activation at 95 • C, 5 s denaturation at 95 • C, annealing/extension for 5 s at 61 • C, 45 cycles, followed by melting step (60-95 • C with fluorescent reading every 0.5 • C). The reaction mixture consisted of cDNA (2 µL; diluted 1:5), 2×SsoFast EvaGreen ® Supermix (10 µL), 10 nM forward and reverse target-specific primers (1 µL of each), and water up to 20 µL. Primer specificity was assured in the melting curve analysis and an electrophoresis in a high-resolution agarose (SeaKem LE agarose from Lonza, Basel, Switzerland) in TBE with SYBR Green (Lonza) detection. Primers were synthesized by Genomed (Warsaw, Poland), and their sequences are presented in Table 12.  1 , primer sequences were as proposed by Origene (www.origene.com); 2 , primers were designed using Beacon Designer Probe/Primer Design Software (BioRad), validated in silico by Blast analysis, and their specificity tested by means of melting curve analysis and an electrophoresis in a high-resolution agarose. Forward and reverse primer sequences are denoted by "F" and "R", respectively.
Prior analysis, technical replicates were averaged. The geometric mean of all Cq values in a given sample set was obtained and subtracted from the individual sample Cq (∆Cq) then linearized by 2ˆ∆ Cq conversion and normalized to GAPDH. The obtained values were referred to as a normalized relative quantity (NRQ) [72] and subjected to statistical analysis.

Statistical Analysis
Normality of distribution was tested using a Kolmogorov-Smirnov test and homogeneity of variances using Levene's test. Data were analyzed using a t-test for independent samples, with Welch correction in case of unequal variances, or one-way ANOVA with a Tukey-Kramer posthoc test and presented as means with 95% confidence interval (CI). Transcriptional data were log-transformed and presented as geometric means with 95% CI. Paired data were analyzed using a t-test for paired samples. Logistic regression (stepwise model) with variables entered into the model if p < 0.05 and removed if p > 0.1 was used to select independent predictors of ESCC. The ROC curve analysis was applied to determine diagnostic power. The area under ROC curve (AUC), indicative of overall marker accuracy, as well as marker sensitivity and specificity, corresponding with optimal cut-off value, were calculated. The Fisher's exact test was applied for frequency analysis. Correlation analysis was conducted using Spearman rank correlation (ρ) or Pearson correlation (r), depending on the data character. The following rule of thumb was used to interpret the size of the correlation: 0.80 to 1.00 as very strong correlation, 0.60 to 0.80 as moderate correlation, 0.30 to 0.60 as fair correlation, <0.30 as poor correlation [73]. All calculated probabilities were two-tailed. The p-values ≤ 0.05 were considered statistically significant. The entire analysis was conducted using MedCalc Statistical Software version 19.2 (MedCalc Software Ltd., Ostend, Belgium; https://www.medcalc.org; 2020).