Prostanoid Signaling in Cancers: Expression and Regulation Patterns of Enzymes and Receptors

Simple Summary Neoplastic processes are accompanied by the reprogramming of cell metabolism and alteration of the balance between endogenous bioregulators with signaling functions. Prostanoid signaling is a part of the global arachidonic acid pathway and is associated with cancer progression. It includes prostanoids (prostacyclin, thromboxane, and prostaglandins E2, F2α, D2, H2), prostanoid metabolizing enzymes, and receptors. We focused on a comparative systematic analysis of expression patterns of target genes, encoding prostanoid enzymes and receptors. We also addressed the possible ways of their regulation at different levels in normal and tumor tissues (expression of genes and proteins, mutation and copy number landscape, promoter methylation status, prediction of tissue-specific master regulators, microRNAs). The results of the systematic analysis made it possible to suggest models of regulation of differentially expressed prostanoid enzymes and receptors. The associations between gene expression signatures and patients’ overall survival rates were established which can be valuable for translational biomedicine. Abstract Cancer-associated disturbance of prostanoid signaling provides an aberrant accumulation of prostanoids. This signaling consists of 19 target genes, encoding metabolic enzymes and G-protein-coupled receptors, and prostanoids (prostacyclin, thromboxane, and prostaglandins E2, F2α, D2, H2). The study addresses the systems biology analysis of target genes in 24 solid tumors using a data mining pipeline. We analyzed differential expression patterns of genes and proteins, promoter methylation status as well as tissue-specific master regulators and microRNAs. Tumor types were clustered into several groups according to gene expression patterns. Target genes were characterized as low mutated in tumors, with the exception of melanoma. We found at least six ubiquitin ligases and eight protein kinases that post-translationally modified the most connected proteins PTGES3 and PTGIS. Models of regulation of PTGIS and PTGIR gene expression in lung and uterine cancers were suggested. For the first time, we found associations between the patient’s overall survival rates with nine multigene transcriptomics signatures in eight tumors. Expression patterns of each of the six target genes have predictive value with respect to cytostatic therapy response. One of the consequences of the study is an assumption of prostanoid-dependent (or independent) tumor phenotypes. Thus, pharmacologic targeting the prostanoid signaling could be a probable additional anticancer strategy.

Data from integrative resources, for example, DisGeNET [38] and plenty of literature reports   (Table S1), clearly demonstrate the significance of genes, encoding prostanoid enzymes and receptors in cancers. It is to be noted that there are reports on correlations of gene expression and methylation patterns in different tumors with disease prognosis [45][46][47][48]52,58,[64][65][66]69,70,74,75,[79][80][81]85,86,89,90]. Tumors with over-expressed or under-expressed prostanoid enzymes and receptors [41][42][43]49,55,69,78,82,91] could determine the benefits of pharmacological correction of this metabolic cluster [44,77]. Indeed, exploration of tumor-specific gene expression and regulation patterns of prostanoid enzymes and receptors contributed to the discovery of new candidate biomarkers and drug targets [7,[96][97][98], which are of great value for translational medicine. However, we did not find a bioinformatic pan-cancer and/or cancer-specific analysis of the expression and regulation patterns of the entire pool of prostanoid enzymes and receptors. Nowadays, despite the increased number of literature reports on the prostanoid signaling field, there is Biology 2022, 11, 590 3 of 26 still a lack of a sophisticated understanding of the impact of transcriptomic, proteomic, and metabolomic factors on tumor promotion or suppression.
The aim of this work was to perform a comprehensive analysis of the gene expression and regulation patterns in prostanoid signaling in the most common human cancers. According to the expression patterns, prostanoid-dependent and independent tumors were conditionally selected. Models of tumor-specific regulation of gene expression including a repertoire of master regulators, microRNAs associated with cancers (oncomiRs), the methylation status of gene promoters, protein-protein interactions, and modifying proteins were predicted. The prognostic values of several transcriptomics signatures were also revealed.

Gene Expression Analysis
Differentially expressed genes (DEGs) were selected from the Cancer Genome Atlas (TCGA) [99] twenty-four tumor datasets. "TCGA tumor" and "TCGA normal and GTEx data" datasets were compared using web-based tool GEPIA2 [100] (accessed on 10 November 2021) at p-value < 0.01 and cut-off value log 2 FC = 1. Datasets of hematologic malignancies were not analyzed. Gene co-expression analysis was performed using GEPIA2. The UALCAN browser [101] (http://ualcan.path.uab.edu/, accessed on 13 November 2021) was used to retrieve the list of genes, which are co-expressed with target genes at Pearson correlation coefficient ≥ 0.75.
The prognostic value of gene expression patterns was explored using pan-cancer Kaplan-Meier plotter [102] (https://kmplot.com/analysis/, accessed on 9 December 2021) with the following settings: follow-up period-"60 months"; number of patients with available clinical data-"more than 100", auto-select-"best cut-off". Overall survival analysis was performed using the TCGA data. The predictive value of differentially expressed genes was explored using the ROC-plotter server (http://www.rocplot.org/, accessed on 5 December 2021) with breast, ovarian, glioblastoma, and colorectal tumor datasets [103][104][105] with outlier exclusion in plot settings.

Mutation Status Analysis
The somatic mutation frequency of genes in different tumors was analyzed using eight pan-cancer cBioPortal cohorts with a total of 47,942 clinical cases [106,107].
The web-based tool muTARGET [108] (accessed on 31 October 2021) was used to predict associations between the mutation status of target genes in tumors and the expression of other genes with the following options: mutation type-"all somatic mutations"; p-value cut-off "0.05"; fold change cut-off "2"; FDR cut-off "no FDR filter".

Over-Representation Analysis
The WEB-based GEne SeT AnaLysis Toolkit (Webgestalt server) [109,110] was used to enrich gene sets with KEGG (Kyoto Encyclopedia of Genes and Genomes), Reactome, and Gene Ontology (GO) terms with the following settings: multiple test adjustment-"Benjamini-Yekutieli FDR-controlling method (FDR < 0.1)", minimal number of gene for category-"5", redundancy reduction-"affinity propagation".

Master Regulators
The prediction of tissue-specific master regulators (MRs) for each DEG was performed in the hTFtarget database [111] (http://bioinfo.life.hust.edu.cn/hTFtarget, accessed on 3 December 2021). hTFtarget accumulates pairs of transcription factors/genes from human ChIP-Seq data (7190 experiment samples of 659 transcriptional factors) in 569 conditions (399 types of cell line, 129 classes of tissues or cells, and 141 kinds of treatments). Tumorspecific gene expression patterns and cancer hallmarks of MRs were explored using GEPIA2 and Cancer Gene Census portal [112], respectively.

Promoter Methylation Status
The promoter methylation status of target genes was analyzed using the UALCAN browser. The beta-value indicates the level of DNA methylation ranging from 0 (unmethylated) to 1 (fully methylated). Different beta value cut-offs have been considered to indicate hypermethylation (beta value: 0.7-0.5) or hypomethylation (beta-value: 0.3-0.25) [115,116].

Protein Expression Analysis
Tumor-specific protein expression was explored using The Human Protein Atlas (HPA) portal version 21.0 [117] (https://www.proteinatlas.org/, accessed on 26 November 2021). The UALCAN browser was used to search for differentially expressed proteins (DEPs) within CPTAC datasets (breast, ovarian, colon, lung tumors, and uterine corpus endometrial carcinoma) [118]. Z-values represent standard deviations from the median across samples for the given cancer type. Log 2 spectral count ratio values from CPTAC were first normalized within each sample profile, then normalized across samples.

Post-Translational Modifications
Modifying proteins carrying out post-translational modifications (PTMs) of prostanoid enzymes and receptors were retrieved from the Biogrid portal [119] (accessed on 10 January 2022). Only physical PPIs and matched subcellular localization profiles (according to the COMPARTMENTS database [120]) were used for the final selection of modifying proteins.

Cluster Analysis of Gene Expression Matrix
Principal Component Analysis (PCA) and Classification and Regression Trees (CRT) methods were used to cluster DEGs in different tumors selected by |log 2 FC| tumor/normal > 1 and p < 0.01. For the PCA method, the gene expression matrix was prepared by assigning integer values (+2) or (−2) to the up-or down-regulated genes, respectively. Genes with |log 2 FC| < 1 were assigned zero value. The ClustVis web-based tool was used for PCA analysis and result visualization [123]. Data preprocessing options were the following: data transformation-"no transformation"; row scaling-"unit variance scaling"; PCA methods-"SVD with imputation". Heat map options: clustering distance for rows-"Manhattan"; clustering methods for rows-"average".
The CRT clusterization was performed by IBM SPSS Statistics v. 23 software with the Decision Tree algorithm. CRT classifies cases into groups of dependent variables based on the values of independent variables. CRT splits the data into segments that are as homogeneous as possible with respect to the dependent variable. A terminal node in which all cases have the same value for the dependent variable is a homogenous, "pure" node. The variables used were: Expression, dependent, and Gene, independent. Expression values were: NEGATIVE, POSITIVE, and NODIFF. The definitions of Expression values were as below. NEGATIVE: the gene expression level in the tumor datasets was decreased compared to the normal tissue datasets. POSITIVE: the gene expression level in the tumor datasets was increased compared to the normal tissue datasets. NODIFF: no changes were shown for the gene expression level in tumor and normal tissue datasets. Gene values were gene names. The Decision Tree procedure was used with the settings below: validation method-"cross-validation", minimum cases in parent node-"100", minimum cases in child node-"50", level growth limits-"5", impurity measure-"twoing".

Data Mining
Molecular profiling data on 19 target genes, encoding prostanoid enzymes and receptors, from 24 solid tumors were analyzed using a data mining pipeline ( Figure 2). The Cancer Genome Atlas (TCGA) portal was the main data-source for the gene expression and co-expression patterns as well as promoter methylation status. Tumor-specific protein expression data were retrieved from Clinical Proteomic Tumor Analysis Consortium (CPTAC) portal. Several web-based tools (GEPIA2, UALCAN, cBioPortal) were adapted for comparative statistical analysis of tumor and normal TCGA or CPTAC datasets to identify DEGs and DEPs. hTFtarget and CSmirTar portals were used for the prediction of master regulators and microRNAs for a subset of target genes, respectively. Finally, predictive and prognostic values of gene expression patterns were investigated using ROCplotter and KM-plotter, respectively. Over-representation analysis was performed in the WebGestalt server.
The CRT clusterization was performed by IBM SPSS Statistics v. 23 software with the Decision Tree algorithm. CRT classifies cases into groups of dependent variables based on the values of independent variables. CRT splits the data into segments that are as homogeneous as possible with respect to the dependent variable. A terminal node in which all cases have the same value for the dependent variable is a homogenous, "pure" node. The variables used were: Expression, dependent, and Gene, independent. Expression values were: NEGATIVE, POSITIVE, and NODIFF. The definitions of Expression values were as below. NEGATIVE: the gene expression level in the tumor datasets was decreased compared to the normal tissue datasets. POSITIVE: the gene expression level in the tumor datasets was increased compared to the normal tissue datasets. NODIFF: no changes were shown for the gene expression level in tumor and normal tissue datasets. Gene values were gene names. The Decision Tree procedure was used with the settings below: validation method-"cross-validation", minimum cases in parent node-"100", minimum cases in child node-"50", level growth limits-"5", impurity measure-"twoing".

Data Mining
Molecular profiling data on 19 target genes, encoding prostanoid enzymes and receptors, from 24 solid tumors were analyzed using a data mining pipeline ( Figure 2). The Cancer Genome Atlas (TCGA) portal was the main data-source for the gene expression and co-expression patterns as well as promoter methylation status. Tumor-specific protein expression data were retrieved from Clinical Proteomic Tumor Analysis Consortium (CPTAC) portal. Several web-based tools (GEPIA2, UALCAN, cBioPortal) were adapted for comparative statistical analysis of tumor and normal TCGA or CPTAC datasets to identify DEGs and DEPs. hTFtarget and CSmirTar portals were used for the prediction of master regulators and microRNAs for a subset of target genes, respectively. Finally, predictive and prognostic values of gene expression patterns were investigated using ROCplotter and KM-plotter, respectively. Over-representation analysis was performed in the WebGestalt server.

Gene Expression Analysis
The gene expression landscape of the prostanoid signaling in different tumors is shown in Figure 3. At least four groups of tumors can be conditionally distinguished according to a similar pattern of DEGs. In Figure 3, it was shown that group I consists of predominantly up-regulated DEGs in GBM, LIHC, PAAD, and THYM tumors. Group II has predominantly down-regulated DEGs in ACC, BRCA, KICH, KIRC, LUAD, PRAD, SKCM, THCA, UCES, and UCS tumors. Group III consists of both up-and down-regulated genes in COAD, ESCA, KIRP, LUSC, OV, READ, STAD, and TGCT. Group IV with HNSC and LGG tumors (not shown in Figure 3) does not contain DEGs.

Gene Expression Analysis
The gene expression landscape of the prostanoid signaling in different tumors is shown in Figure 3. At least four groups of tumors can be conditionally distinguished according to a similar pattern of DEGs. In Figure 3, it was shown that group I consists of predominantly up-regulated DEGs in GBM, LIHC, PAAD, and THYM tumors. Group II has predominantly down-regulated DEGs in ACC, BRCA, KICH, KIRC, LUAD, PRAD, SKCM, THCA, UCES, and UCS tumors. Group III consists of both up-and down-regulated genes in COAD, ESCA, KIRP, LUSC, OV, READ, STAD, and TGCT. Group IV with HNSC and LGG tumors (not shown in Figure 3) does not contain DEGs. Groups I and II with contrasted gene expression patterns are the most interesting for analysis. Up-regulation of genes, encoding enzymes, which produce pro-neoplastic prostanoids, can lead to tumor promotion and prostanoid-dependent tumor phenotype. Down-regulation of such genes may indicate tissue-specific responses to neoplastic transformation or cancer-driven metabolism reprogramming aimed to reduce the accumulation of tumor-suppressive prostanoids. To test these assumptions indirectly, correlations Groups I and II with contrasted gene expression patterns are the most interesting for analysis. Up-regulation of genes, encoding enzymes, which produce pro-neoplastic prostanoids, can lead to tumor promotion and prostanoid-dependent tumor phenotype. Down-regulation of such genes may indicate tissue-specific responses to neoplastic transformation or cancer-driven metabolism reprogramming aimed to reduce the accumulation of tumor-suppressive prostanoids. To test these assumptions indirectly, correlations between the DEGs and disease prognosis were explored. We also considered the theory that the overexpressed genes ( Figure 3) could be critical for tumor cell viability. In other words, how sensitive tumor cells are to knockouts and knockdowns of genes in prostanoid signaling. We analyzed the data of pan-cancer Clustered Regularly Interspaced Short Palindromic Repeats (CRISPR) and RNA interference (RNAi) screens on the DepMap portal [124]. A general rule is that the lower the gene effect score, the more its dependency in a cell line, so a score close to −1 corresponds to genes that can be essential for a cell line viability [124]. Table S2 shows the gene effect scores. Knockouts/knockdowns of several target genes with scores below −0.5 (an accepted cutoff value) can be related to tumor cell  (Table S2): TBXAS1 (gallbladder adenocarcinoma, OCUG1 cell culture), PTGDR (lung cancer, CORL279), PTGER4 (lymphoma, C8166) and PTGES3 (brain cancer, ONS76). Incidentally, PTGES3, being the most common DEG in various tumors (Figure 3), tends to be an essential gene in tumor cell lines (Table S2). At the same time, there is an absence of a significant effect in cell lines caused by the "separate switching off" of most of the target genes. Thus, it allows us to consider these genes as non-essential for tumor cell viability.
Cluster analysis was used to find outmatched expression patterns of target DEGs within a set of tumors. Comparing the results obtained by two clustering methods, the Classification and Regression Trees (CRT) (Appendix B, Figure A1) and the Principal Component Analysis (PCA) (Figure 4), subclusters with matched gene expression patterns (PTGIS-PTGDS, PTGES2-PTGDR2, TBXA2R-PTGER4-PTGER1-PTGER-PTGDR, and PTGES3-PRXL2B) were identified. tanoid signaling. We analyzed the data of pan-cancer Clustered Regularly Interspaced Short Palindromic Repeats (CRISPR) and RNA interference (RNAi) screens on the Dep-Map portal [124]. A general rule is that the lower the gene effect score, the more its dependency in a cell line, so a score close to −1 corresponds to genes that can be essential for a cell line viability [124]. Table S2 shows the gene effect scores. Knockouts/knockdowns of several target genes with scores below −0.5 (an accepted cutoff value) can be related to tumor cell viability (Table S2): TBXAS1 (gallbladder adenocarcinoma, OCUG1 cell culture), PTGDR (lung cancer, CORL279), PTGER4 (lymphoma, C8166) and PTGES3 (brain cancer, ONS76). Incidentally, PTGES3, being the most common DEG in various tumors (Figure 3), tends to be an essential gene in tumor cell lines (Table S2). At the same time, there is an absence of a significant effect in cell lines caused by the "separate switching off" of most of the target genes. Thus, it allows us to consider these genes as non-essential for tumor cell viability.
Cluster analysis was used to find outmatched expression patterns of target DEGs within a set of tumors. Comparing the results obtained by two clustering methods, the Classification and Regression Trees (CRT) (Appendix B, Figure A1   Next, we searched for other genes whose expression patterns correlate with the target genes in each of the three PCA clusters (Figure 4). It should be said that the total number of found co-expressed genes for the first cluster was about 10 times higher than for the second cluster. Gene sets co-expressed with TBXAS1 and PTGDS in a subset of tumors (KIRC, LIHC, GBM, PCPG, OV, ACC, LGG, UVM, ESCA, CHOL) are enriched in immune reaction and phagocytosis pathway terms (Table S3). Gene sets co-expressed with PTGIS in DLCA, CHOL, COAD, READ, PRAD, TGCT tumors, and PTGES in ACC tumors are enriched in "muscle contraction" and "extracellular matrix organization" terms. Gene sets co-expressed with PTGES3 in THCA and SKCM tumors are enriched with the "mRNA processing" term (Table S3). Gene sets co-expressed with genes, encoding prostanoids receptors TBXA2R, PTGFR, and PTGER3, are involved in immune responses and structural processes (elastic fibers formation, collagen polymerization/depolymerization) as well as the organization of the extracellular matrix (Table S3) in COAD and TGCT tumors. These findings indicate the leading cellular processes that accompany prostanoid signaling in different tumors. These terms are in good agreement with the known effects of prostanoids on blood vessel smooth muscle proliferation in the regulation of cardiovascular homeostasis [125,126], actin cytoskeleton reorganization via stimulation of stress fiber formation [127], and the immune response modulation [127][128][129]. Enrichment analysis also shows the association of prostanoid enzymes and receptors with focal adhesion kinase 1 (PTK2), which participates in the neoplastic transformation via activation of Wnt/β-catenin signaling [130].

Promoter Methylation
Differences in the transcript accumulation in normal and tumor tissues may be the result of several competing factors at the transcriptional and post-transcriptional levels such as the methylation status of gene promoters, combination of transcriptional master regulators, and transcript stability. Analysis of promoter methylation status of the target genes show PTGDR, PTGER3, PTGIR, and TBXA2R down-regulation in a number of tumors which may be partly due to the elevation of promoter methylation (Table S4). On the other hand, for up-regulated CBR3 and AKR1C3 genes in LUSC tumors, there is an agreement with promoter hypomethylation. The down-regulation of PTGDS in ESCA and HNSC tumors was accompanied by a statistically significant decrease in promoter methylation while remaining in the hypermethylation range.

Master Regulators
Transcriptional master regulators (MRs) mean DNA-binding and chromatin remodeling proteins [131], which are capable of regulating the expression of genes-of-interest. We selected only tissue-specific MRs that were predicted for ≥80% of all the target DEGs. Table 1 shows the distribution of 21 MRs that could potentially be responsible for the differential expression of target genes in tumors. Master regulators such as BRD4, CTCF, EP300, FOXA1, and SPI1 are expressed in breast, brain, colorectal, kidney, pancreatic, prostate, skin, and stomach tumors (Table 1) enriched with cancer hallmarks and encoded by cancer driver genes. It is noteworthy that most of the found MRs are overexpressed in tumors. This speaks in favor of the involvement of the predicted MRs in the neoplastic transformation that is also stressed by the participation of AR, EP300, MAX, RELA, SP1, and SPI1 proteins in cancer-related pathways (Table S5).

Protein Expression Patterns
According to The Human Proteome Atlas (HPA), positive immunohistochemical staining (IHC) is shown for the majority of prostanoid enzymes in tumors (Table S8). IHC protein expression data on prostanoid receptors, with the exception of PTGER4, are not yet available in HPA (Table S8). This group can be characterized as low expressed in nor- Thus, it can be assumed that the accumulation of oncomiRs may be in inverse proportion with that of target transcripts. This can be traced in contrasting pairs such as miR-19a/PTGER2 (pancreatic tumor); miR-421/PTGER2 (uterine tumor); miR-590/PTGFR (uterine and breast tumor); miR-20a/PTGER3 (uterine tumor) ( Figure 5).

Protein Expression Patterns
According to The Human Proteome Atlas (HPA), positive immunohistochemical staining (IHC) is shown for the majority of prostanoid enzymes in tumors (Table S8). IHC protein expression data on prostanoid receptors, with the exception of PTGER4, are not yet available in HPA (Table S8). This group can be characterized as low expressed in normal and tumor tissues. PTGDR, PTGDR2, PTGER1, PTGER3, and PTGFR genes have the median expression value of about 1 FPKM (Fragments Per Kilobase Million), while other genes are expressed in the range of 1-5 FPKM, regardless of cancer specificity. A concordance between the transcripts and protein accumulation in normal and tumor tissues was performed by comparing the TCGA and CPTAC data in BRCA, COAD, LUAD, OV, and UCES tumors ( Table 2). From Table 2, it follows that there is a concordance between the transcripts and total protein accumulation. However, up-regulation of PRXL2B is not accompanied by an increase in the protein content in BRCA and OV tumors. There is an inverse relationship between the up-regulated PRXL2B and PTGDS genes and the protein accumulation in COAD and OV tumors, respectively. The general observation is that a decrease in mRNA accumulation leads to a decrease in total protein, which is quite logical. It is assumed that when transcript level increases, but protein content remains unchanged or reduced, there is a translational and/or post-translational regulation.

Prognostic Value of Transcriptomic Signatures
We explored the target genes, encoding the prostanoid enzymes and receptors, in terms of their disease prognostic value. Survival analysis was performed, and all relevant results obtained are presented in Table 3. In particular, Figure S1 shows overall survival curves in groups with high and low gene expression (without subgroup analysis) followed by an assessment of tumor-specific signatures (Table 3). A "pure" tumor specificity of signatures is achieved in the subgroups stratified by gender, stage, grade, and mutation burden. This, however, is balanced by a reduction of clinical cases in a subgroup and the power of a statistical test. Most of the transcriptomic signatures (without subgroup analysis) still show acceptable tumor specificity with the exception of the signature containing PTGDS, CBR3, PTGIR, PTGFR, PTGDR2, and PTGER3 genes in HNSC tumor, which is similar to other tumors (BRCA, CESC, LUAD, SARC, and UCES).

Predictive Value of Prostanoid Enzymes and Receptors Genes
We explored the associations between gene expression of each target gene and responses to anticancer drug treatment of breast, ovarian, glioblastoma, and colorectal tumors using the Receiver Operating Characteristic plotter (ROC-plotter) [103]. It was found that only for breast or ovarian tumors, ROC curves for (PTGIS, PTGES, and PTGER4) or (TBXAS1, PTGES, TBXA2R, and PTGDR2), respectively, had Area Under Curve (AUC) values > 0.75 at the tumor/normal fold changes ≥ 2 ( Figure S2A,B). PTGIS, PTGES, and PT-GER4 up-regulation in the group of responders with HER2-positive breast cancer, known by its aggressive behavior [132,133], is associated with complete tumor response to treatment with fluorouracil, epirubicin, and cyclophosphamide (FEC). On the other hand, the same amplitude of down-regulation of PTGES and PTGDR2 gene expression in the group with serous ovarian cancer grade G3 correlates with relapse-free survival at 6 months and the response to taxanes and platinum treatment, respectively ( Figure S2B). Thus, we observed the associations between the gene expression patterns and responses to cytostatic therapy, while there were no associations in the case of targeted therapy (Trastuzumab, Tamoxifen, Avastin, and aromatase inhibitors).

Discussion
Pan-cancer analysis showed several interesting gene expression patterns in prostanoid signaling that were used to model tissue-specific regulation patterns (Table S10). Transcriptomic data were retrieved from the TCGA portal, which actually contains data on most tumor types and subtypes. At the same time, the availability of proteomic information in the CPTAC portal is more limited. For this reason, we compared the gene expression patterns of target genes, their predicted master regulators, and oncomiRs in several tumors. It could be suggested that the up-regulation of AKR1C3, CBR1, and CBR3 genes from PCA cluster 2 ( Figure 4) in LUSC tumors may occur due to promoter hypomethylation and gene expression changes of master regulators FOXA2, LMNB1, SPI1, miR-511 (Table S10). In contrast, down-regulation of co-expressed genes CBR1 and CBR3 in ESCA tumors can be associated with elevated accumulation of miR-335 (for CBR1) and a decrease in promoter methylation status similar to that in LUSC. The up-regulation of all ten genes, encoding prostanoid enzymes, in PAAD tumors compared to normal tissue is noteworthy. This expression pattern is not found in any other tumors, except for THYM tumors, where only seven of ten genes are up-regulated. It follows from Table S10 that at least half of DEGs are characterized by an increase in copy number variations (amplification type) indexed in the cBioPortal UTSW Nat. Commun. 2015 cohort. We have found no statistically significant changes in the promoter methylation status in PAAD tumors compared to normal tissue. However, up-regulated PTGDS, AKR1C3, and CBR3 genes were hypermethylated, an epigenetic situation that also occurs in tumors [134]. The accumulation of PTGES3 and PRXL2B transcripts in tumor tissue correlates with a simultaneous reduction of oncomiRs miR-223, miR-19a, miR-605, and miR-486, miR-211, miR-423, respectively. It should be noted that some of those are tumor suppressors [135][136][137]. Up-regulation of ten DEGs in PAAD tumor is in concordance with the up-regulation of master regulators CTCF, IRF1, and KLF4, while POLR2A and STAG1 remain unchanged. It is well known that transcriptional activation of master regulators is critical for tumor progression, in particular, for pancreatic [138] and colorectal cancers [139,140].
Prostacyclin, a metabolite produced by prostacyclin synthase (PTGIS), is historically believed to exert tumor-suppressive effects [141,142] and lowers its level along with downregulation of PTGIS associated with an aggressive tumor phenotype and a poor disease prognosis [43]. Figure 3 shows a simultaneous decrease in the expression of both PTGIS and PTGIR in eight tumor types, and therefore, we analyzed the possible causes of such changes (Table S10). It can be pointed out that down-regulation of PTGIS in KIRP, LUAD, THCA, and UCES tumors is accompanied by an increase in miR-34a levels, which mainly plays a considerable role in inhibiting tumor progression in thyroid tumors [143]. In LUSC tumors, there is a decrease in miR-34c, which, like miR-34a, possesses antitumor activity [144]. On the other hand, miR-326 expression is suppressed in lung cancer tissues. This oncomiR, as shown in [145], inhibits lung cancer cell proliferation, and colony formation and provokes apoptosis. It should also mention that the down-regulation of PTGIS is comparable to the lowering of a corresponding protein product in LUAD and UCEC tumors ( Table 2). As for the predicted transcriptional master regulators of PTGIS and PTGIR (Table S10), there are fundamentally distinct tumor-specific patterns. Transcript and total protein accumulation of master regulator ZBTB7A as well as PTGIS and PTGIR were reduced in UCES tumors, which was markedly related to the stage and prognosis of this tumor type [146]. Thus, the down-regulation of PTGIS and PTGIR in eight different tumors in our model may be due to the contribution of master regulators and oncomiR combinations at the transcriptional and post-transcriptional levels under conditions of unchanged promoter methylation status and copy number variations (deletion type) of target genes.

Protein-Protein Interactions and Post-Translational Modifications
Since some of the prostanoids are quite metastable (short-lived) metabolites, spatial clustering or compartmentalization of prostanoid enzymes can be required. It is realized through either direct PPIs between enzymes or the involvement of common protein partners as well as post-translational modifications (PTMs). Previously, using the affinity purification and mass-spectrometry approach, we revealed that the PTGES3 protein could be a potential protein partner of PTGIS [147]. PPIs subnetworks with PTGIS and PTGES3 and their protein partners retrieved from STRINGdb are shown in Figure S3. Overrepresentation analysis (ORA) indicates that a subset of the PTGIS's protein partners is enriched with steroid and cholesterol biosynthesis (FDFT1, HSD17B7, LSS, SC5D proteins) pathway terms. Functional "bridges" between cholesterol synthesis and prostanoid pathways (changes in prostacyclin levels in the presence of statins), as described in [148,149], can be mediated via modulation of PTGIS gene expression. But so far, we have not found studies that would evidence the functional value of PPIs with enzymes involved in prostanoid and cholesterol synthesis. The PTGES3 s protein partner subset is enriched with pathway terms such as "protein processing in endoplasmic reticulum", "pathways in cancer" (KEGG); "cellular responses to external stimuli", "aryl hydrocarbon receptor signaling" and "TNF alpha Signaling Pathway" (Reactome). In addition, comparing the data obtained in [147] and STRINGdb, we found that at least heat shock 70 kDa protein 4L (HSPA4L, Uniprot ID: O95757) and calreticulin (CALR, Uniprot ID: P27797) with chaperone activity are common protein partners of PTGIS and PTGES3 proteins.
Amino acid sequences of prostanoid enzymes and receptors contain multiple sites for reversible post-translational modifications such as ubiquitination, phosphorylation/dephosphorylation, and glycosylation. In that context, the gene expression patterns of modifying proteins (ubiquitin ligases, protein kinases/phosphatases, and glycosyltransferases) were analyzed. The spectrum of potential modifying proteins, which physically interact with prostanoid enzymes and receptors, and their tumor-specific gene expression patterns are shown in Table S11. Up-regulation of ubiquitin-protein ligases SIAH2, MARCH2, MARCH3, UBE2W, OTUB1, and VHL, which regulate the stability of mature proteins, as well as the protein kinases STK24, MAP4K1, MAP4K4, PRKCD, CSK, PINK1, PRKAB1, and STK39, was found. It is known that mitogen-activated protein kinases (MAP4K1 and MAP4K4) and the insulin resistance pathway (PRKCD and PRKAB1) may be associated with tumor progression through modulation of gene expression responsible for cell cycle, proliferation, and growth [150]. We have schematically shown in Figure 6 the associations between gene expression patterns of CBR1 and PTGIR proteins and their modifying enzymes. These examples demonstrate tumor-specific multiple modes of post-translational regulation for prostanoid enzymes and receptors, so the presence of "active combinations" of modifying enzymes depending on their expression levels in tumors can be suggested. However, the direct involvement of modifying proteins in post-translational modifications of target enzymes or receptors is still not sufficiently studied. It is known that PTGES protein stability is positively regulated through interaction with ubiquitin-specific peptidase 9 X-Linked (USP9X) [57]. Post-translational regulation via phosphorylation has been described for PTGES [151]. Prostanoid receptors have rather long cytoplasmic tails with potential phosphorylation sites [152], and protein kinase C-dependent phosphorylation has been described for the prostacyclin receptor [153].
The limitations of the study are related to the use of publicly available data from TCGA, CPTAC, and other repositories and web-based tools for the analysis of datasets. The results of the study have more fundamental rather than translational value. The identified transcriptional signatures, with the participation of prostanoid signaling genes with differential expression in tumor/normal tissues, are exploratory. Thus, to further establish the clinical relevance of such signatures, additional rounds of experimental validation should be required.

Conclusions
We investigated the highly heterogeneous gene and protein expression landscape of prostanoid enzymes and receptors in 24 different tumors and suggested the models of These examples demonstrate tumor-specific multiple modes of post-translational regulation for prostanoid enzymes and receptors, so the presence of "active combinations" of modifying enzymes depending on their expression levels in tumors can be suggested. However, the direct involvement of modifying proteins in post-translational modifications of target enzymes or receptors is still not sufficiently studied. It is known that PTGES protein stability is positively regulated through interaction with ubiquitin-specific peptidase 9 X-Linked (USP9X) [57]. Post-translational regulation via phosphorylation has been described for PTGES [151]. Prostanoid receptors have rather long cytoplasmic tails with potential phosphorylation sites [152], and protein kinase C-dependent phosphorylation has been described for the prostacyclin receptor [153].
The limitations of the study are related to the use of publicly available data from TCGA, CPTAC, and other repositories and web-based tools for the analysis of datasets. The results of the study have more fundamental rather than translational value. The identified transcriptional signatures, with the participation of prostanoid signaling genes with differential expression in tumor/normal tissues, are exploratory. Thus, to further establish the clinical relevance of such signatures, additional rounds of experimental validation should be required.

Conclusions
We investigated the highly heterogeneous gene and protein expression landscape of prostanoid enzymes and receptors in 24 different tumors and suggested the models of tumor-specific regulation. Nine bioinformatic web-based tools (GEPIA2, UALCAN, cBioPortal, hTFtarget, CSmirTar, ONCOmir, muTARGET, Biogrid, and ClustVis) were used for the analysis of differentially expressed genes, proteins, microRNAs, methylation and mutation patterns, as well as protein-protein interactions. Four groups of tumors were prioritized according to the profiling of the entire pool of differentially expressed target genes. The high correlation of co-expression was shown in the sub-cluster with AKR1C3, CBR1, and CBR3 genes. Down-regulation of PTGDR, PTGER3, PTGIR, and TBXA2R genes in a number of tumors can be linked with promoter methylation status. Tissue-specific master regulators BRD4, CTCF, EP300, FOXA1, and SPI1, overexpressed in tumors, were found for target genes. Predicted microRNAs such as miR-149-3p, miR-19a-3p, miR-338-5p, miR-421, miR-423-5p, and miR-508-5p can be involved in the post-transcriptional regulation of at least two different target RNAs. The highest concordance between expression data of TCGA and CPTAC databases was achieved for PTGIS and PTGES genes in four tumors: BRCA, COAD, LUAD, and UCES. Mutation frequency of TBXAS1, PTGIS, AKR1C3, PTGDR, PTGFR and PTGER3 genes in melanoma were 5-6%, 5-7%, 2.8-5%, 3-4%, 7-9%, 2.1-5%, respectively. One of the conclusions of the study is the assumption of the presence of prostanoid-dependent tumor phenotypes. It can be demonstrated by the total up-regulation of prostanoid synthesis enzymes in GBM, PAAD, and THYM tumors. Down-regulation of the PTGIS and PTGIR genes in eight different tumors may be associated with a more aggressive tumor phenotype due to the abolishment of prostacyclin's tumor-suppressive effects. Models of CBR1 and PTGIR post-translational regulation models were mediated via neddylation and ubiquitination/deubiquitination as well as phosphorylation depending on tumor types. We have found several multigene cancer-specific transcriptomic signatures (in BLCA, CESK, SARC, LUAD, LIHC, and KIRP tumors) associated with patients' overall survival rates (prognostic value). There are associations between the expression pattern of five target genes and the tumor response to cytostatic therapy. For example, differential expression of the PTGES gene was predictive in BRCA and OV tumors. From our point of view, for the first time, a systemic pan-cancer analysis of the molecular features of the expression of genes involved in prostanoid signaling was carried out. The new results obtained will be contributed to the insight into prostanoid signaling in a cancer context.

Supplementary Materials:
The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/biology11040590/s1, Table S1. The associations of prostanoid metabolizing enzymes and prostanoid receptors with neoplastic transformation. Table S2. Cell lines dependency on target genes' knockouts and knockdowns. Table S3. Over-representation analysis of genes showing similar expression patterns with target genes involved in prostanoid signaling. Table S4. Concordance between changes of gene expression and promoter methylation patterns. Table S5. Over-representation analysis of master regulators using the KEGG database as a data source. Table S6. A list of predicted regulatory oncomiRs for target genes. Table S7. Expression ratio (tumor/normal) of predicted master oncomiRs that can regulate mRNAs of genes involved in prostanoid signaling. Table S8. Antibody staining of proteins involved in prostanoid signaling according to the Human Proteome Atlas portal. Table S9. Prediction of genes, which expression patterns can be associated with genetic polymorphism in target genes in melanoma tumors. Table S10. Cancer-specific expression and regulation patterns of prostanoid-metabolizing enzymes and receptors. Table S11. Gene expression changes by modifying proteins retrieved from the Biogrid database and physically interacting with the proteins involved in prostanoid signaling. Figure S1. Kaplan-Meier plots: overall survival vs. expression levels of target genes in several tumors. Figure S2. Predictive value of gene expression of prostanoid-metabolizing enzymes and prostanoid receptors in breast ( Figure S2A) and ovarian ( Figure S2B) cancers. Figure S3 Table A1. Literature reports on the involvement of prostanoids in neoplastic transformation.

Prostanoids Description References
TXA 2 TXA 2 impacts the interface of platelet-tumor cell crosstalk and serves as a link between platelets in ovarian cancer. [20] TXA 2 Inhibition of TXA 2 synthesis reduced human umbilical vein endothelial cells migration stimulated by VEGF or bFGF. The development of lung metastasis in mice models was significantly inhibited by thromboxane synthase inhibitors. [21] TXB 2 High TXB 2 urinary level was associated with (i) prostate cancer in African American men (OR 1.50, 1.13-2.00), but not European American men (OR 1.07, 0.82-1.40); (ii) metastatic prostate cancer (OR 2.60, 1.08-6.28) compared with low levels of TXB 2 .
[22] TXB 2 TXB 2 was much higher in the non-small cell lung carcinoma tissue than in normal tissues and advanced-stage cancers had higher levels of TXB 2 thus supporting the role of TXB 2 in tumor growth promotion. [23] 11-dihydro-TXB 2 In 10 patients with colorectal cancer, the urinary excretion of 11-dehydro-TXB 2 was significantly higher than in 10 controls. Enhanced platelet activation occurs in colorectal cancer patients and low-dose aspirin might restore anti-tumor reactivity.
[24] PGE 2 PGE 2 promotes gastrointestinal tumor progression and metastasis by (i) direct effect on tumor cell proliferation, survival, and migration/invasion; (ii) tumor-associated immunosuppression; (iii) by silencing certain tumor suppressor and DNA repair genes via DNA methylation.

PGD 2
Signaling between PGD 2 and PTGDR2 has the ability to restrict the self-renewal of gastric cancer cells in vitro and suppress tumor growth and metastasis in vivo. The study showed a novel function of PGD 2 /PTGDR2 signaling in cancer stem cells regulation that is critical for tumor neovascularization and invasiveness. [37]