The Role of Purine Metabolism-Related Genes PPAT and IMPDH1 in the Carcinogenesis of Intrahepatic Cholangiocarcinoma Based on Metabonomic and Bioinformatic Analyses

In this study, we investigated the role of tumor microenvironment and serum differential metabolites in intrahepatic cholangiocarcinoma (ICC) carcinogenesis, providing new evidence for ICC treatment. Serum samples from healthy individuals and ICC patients were collected for metabolomic analysis. The purine metabolites such as inosine, guanosine, hypoxanthine, and xanthine were increased in patient serum. TCGA database samples were collected, and the correlation between purine metabolism-related genes and ICC clinical features was analyzed using R language to obtain the differential genes including PPAT, PFAS, ATIC, and IMPDH2. High PPAT expression was associated with poor ICC prognosis. A PPAT silencing model in HCCC-9810 cells was constructed. The cell phenotype was examined by qRT-PCR, CCK-8, transwell, and flow cytometry, showing a decrease in IMPDH1 expression, colony and invasive cells numbers, and an increase in apoptosis. Guanosine reversed IMPDH1 expression in HCCC-9810 cells, promoting the secretion of inflammatory factors IL-6, IL-8, OPN, VEGF, and VCAM-1 and intensifying epithelial-mesenchymal transition (EMT) progression in the cells. In nude mice, the IMPDH1 inhibitory drug MMF inhibited tumor growth and reduced the expression of tumor stem cell characteristic markers CD133 and SOX2. Guanosine accelerated the malignant progression of ICC inhibition of purine metabolism-related genes, PPAT and IMPDH2, suppressed the malignant phenotype in HCCC-9810 cells, and inhibited tumor growth.


Introduction
Intrahepatic cholangiocarcinoma (ICC) is a high-mortality malignancy with a steadily increasing incidence worldwide [1]. ICC is usually asymptomatic in its early stages and is usually diagnosed in an advanced stage when symptoms appear and difused disease is detected, which limits the current potential treatment options based on surgery or liver transplantation [2]. In the clinical setting, primary biliary diseases are clinically distinct but show some similar features, such as chronic cholestasis, idiopathic disease, and high risk of malignant transformation [3]. Terefore, the diferential metabolic phenotypes observed in the sera of patients with malignant and benign cholestatic diseases may provide clinical value and increase available diagnostic modalities [4]. Te role of diferential metabolites in the carcinogenic process of ICC and their regulatory mechanisms needs to be investigated urgently.
Metabolomics is one of the latest histological techniques that uses robust analytical techniques to screen biological samples for low-molecular weight metabolites that are closely associated with functional alterations in the body [5]. Metabolomics allows us to better understand the pathological processes and substance metabolic pathways by analyzing and validating specifc biomarkers of a disease. Even small metabolite changes can help detect early pathological changes with better sensitivity than traditional diagnostic methods [6]. Given the potential of metabolomic analysis for biomarker discovery, our aim was to study the metabolites of interest for analysis and to presuppose that lipids, bile acids, high-energy metabolites, and amino acids could be potential biomarkers for biliary cancers. Purine metabolism is associated with carcinogenic efects and has been reported to be altered in the urinary biochemical profles of patients with hepatocellular carcinoma [7]. Our pre-experimental results revealed that the purine metabolite products such as inosine, guanosine, hypoxanthine, and xanthine were elevated in ICC.
Guanosine monophosphate (GMP) plays a crucial role in energy metabolism, cell signaling, and cell reproduction [8]. Despite the important functions of the genes involved in the purine ab initio synthesis pathway, little is known about their regulation and in vivo expression patterns [9]. Purine metabolite synthesis is dependent on phosphoribosyl pyrophosphate amidotransferase (PPAT) [10]. It has also been reported that genes with oncogenic efects form a metabolically distinct subpopulation with abundant purines, which is dependent on inosine monophosphate dehydrogenase (IMPDH), an enzyme for guanosine triphosphate (GTP) biosynthesis [11]. In this subgroup, induction of IMPDH1 expression and dependence was necessary and sufcient, and IMPDH1 inhibitors were able to inhibit tumor growth in xenograft and in situ genetically engineered mouse models [12]. Screening of purine metabolism-related genes from TCGA database revealed that PPAT and IMPDH1 were associated with ICC survival. Terefore, the next step in studying ICC involves the regulation of PPAT and IMPDH.
PPAT transfers the c-nitrogen of glutamine to 5phosphoribosyl pyrophosphate (PRPP), a key rate-limiting reaction in purine biosynthesis [13,14]. Indeed, PPAT expression is increased in lung adenocarcinoma and correlates with patient prognosis [15]. PPAT has been less studied in ICC, and the link between PPAT and IMPDH1 remains unclear. Enhanced IMPDH1 activity could facilitate the provision of highly depleted GTP nucleotides in highly proliferating stem cells [16]. Lymphocytes undergo intense proliferation in the presence of high guanine nucleotide depletion in response to an antigenic challenge. During the proliferation of antigen-activated T cells, IMPDH1 polymerization forms cytoplasmic structures [17]. IMPDH1 cell cytosolic assembly makes the enzyme more resistant to feedback inhibition of guanosine-di/trisphosphate (GDP/ GTP) transitions, thus promoting the accumulation of guanine nucleotide pools [18]. Tis fnding implies that IMPDH1 can potentially infuence the immune microenvironment. We hypothesized that PPAT and IMPDH1 afect the proliferation, apoptosis, migration, and invasive functions of ICC.
However, the results of previous studies have shown signifcant diferences in purine metabolites between normal patients and patients with ICC. Nevertheless, whether guanosine, a purine metabolite, can regulate ICC function and the immune microenvironment to inhibit tumor progression is unclear. Te present study clarifed the high levels of guanosine in patients with ICC. Expression of PPAT and IMPDH1 was higher in ICC. Exogenous supplementation with mycophenolic acid (MPA), a drug that inhibits IMPDH1, suppresses the malignant phenotype of ICC, inhibits the secretion of infammatory factors, and ultimately alleviates tumor growth in ICC. Te inhibition of genes related to purine metabolism can serve as basis for the study of the disease and development of treatments for ICC. ) and maintained at 37°C in a humidifed incubator containing 5% CO 2 . In cell grouping, NFs, cocultured with HCCC-9810, was used as the NF group. Te iCAF group was cocultured with iCAF and HCCC-9810 was iCAF group. When iCAF and HCCC-9810 cells were cocultured with 10 μM MPA for 72 h, the iCAF + MPA group was added. Te iCAF + MPA + guanosine group was incubated with 10 μM MPA and 50 μM guanosine when cocultured with HCCC-9810 cells for 72 h. PPAT and IMPDH1genes were silenced in HCCC-9810 cells using Nucleofector II. Te control, PPAT, and IMPDH1 siRNA lentiviruses were purchased from HonorGene. Te siRNA sequences are listed in Table 2.

ICC Datasets and Preprocessing.
In total, 44 samples were collected from TCGA-CHOL (https://portal.gdc. cancer.gov/), including 36 tumor and nine normal samples. RNA sequencing (RNA-seq) data were downloaded from TCGA data portal. Te fragment number per million fragments (FPKM) values was then converted to transcriptional/per million-word node (TPM) values. Te microarray dataset GSE72094 was downloaded from the Gene Expression Omnibus (GEO, https://www.ncbi.nlm. nih.gov/geo/). All data were analyzed using R software (version 3.6.1) and R Bioconductor package. Eleven PMrelated genes were selected to create heat maps, and box plots were constructed from the expression data of the 11 genes. Te genes were subjected to survival analysis and clinical (pathologic-T, stage, grade, and sex) correlations.

Principal Component Analysis (PCA).
In the model calculation, we frst fnd a straight line so that the residual sum of squares of all samples from this line is the smallest, the vector sum of squares projected in the direction of this number axis is the largest, and the direction of this line also refects the maximum diference between samples; thus, the frst principal component (PC1) is obtained. Te second principal component (PC2) was obtained by fnding the second line with the most signifcant diference along the vertical direction of the previous principal component line. Te main parameter used to judge the quality of the PCA model was R2X, which represented the interpretation rate of the original data after dimensionality reduction. Te closer the value is to one, the better it is. It is generally believed that R2X greater than 0.5 indicates a better model efect.

Partial Least-Squares Discriminant Analysis (PLS-DA)
. PLS-DA is a discriminant analysis method based on the classical partial least-squares regression model. For PLS-DA, samples must be specifed and grouped during analysis, after which the model automatically adds an implicit dataset Y. Tis method of model calculation forces each group to be classifed, which is conducive to identifying similarities and diferences between diferent groups.

Orthogonal Partial Least-Squares Discriminant Analysis
(OPLS-DA). OPLS-DA is a combination of orthogonal signal correction (OSC) and PLS-DA. Variance factors, irrelevant or orthogonal to the Y variable in the X data variable, are removed. OPLS-DA divides the diferences in data table X into two parts according to the diferences in data table Y. Te frst part represents the diferences related to Y and the second part represents the diferences unrelated to Y (orthogonal and vertical). OPLS-DA can distinguish between these two parts.

Liquid Chromatography-Tandem Mass Spectrometry (LC-MS/MS).
To perform LC-MS/MS, we frst extracted metabolites from the serum. Te supernatant was centrifuged at 1150g for 10 min at 4°C. Te supernatant (100 μL) was   Table 3. Using 2 μg of total cDNA as a template, the relative transcription level of the target gene was calculated using the relative quantitative method (2 −△△Ct ):

Transwell
Preparation of Cell Suspension. Te intervened cells were digested with 0.25% trypsin digestion solution, the cell suspension was prepared with serum-free basal medium, and the cell density was adjusted to approximately 1 × 10 5 / mL. Te culture solution in the small chamber was discarded, and the cells were washed twice with PBS. Te cells in the upper chamber were wiped with wet cotton swabs, fxed in acetone: methanol (1 : 1) for 20 min, washed twice with PBS, stained with 0.5% crystal violet for 5 min, and washed more than three times with water. Te cells were observed under an inverted microscope and photographed.

Colony Formation Assay
Preparation of Cell Suspension. After 24 h of intervention, the cultured cells were digested with 0.25% trypsin digestion solution, the cell suspension was prepared with serum-free basal medium, and the cell density was adjusted to approximately 1 × 10 5 /mL. Inoculation of Cells. 1000 cells/2 mL of each group were inoculated in a 6-well plate, shaken to disperse, and incubated at 37°C in an incubator with 5% CO 2 . Te cells were incubated at 37°C and 5% CO 2 , and the solution was changed every 2-3 d.

Statistical Analyses.
Te measurement data are expressed as mean ± standard deviation (SD). Te count information was presented as a percentage. Each test was repeated independently three times. All data were analyzed by using GraphPad Prism 8.0 software (La Jolla, CA, USA). Kolmogorov−Smirnov test and exploratory descriptive statistics test were used to analyze whether the data conformed to a normal distribution and homogeneity of variance. Te measurement data obeyed the normal distribution and homogeneity of variance. Te data were analyzed by the parametric test. Te unpaired Student's t-test was used to compare the data of two groups that were not one-to-one correspondence. One-way ANOVA and Tukey's posthoc test were used to compare data among three groups. Te degrees of freedom, the sum of squares, and the mean of the sum of squares were used to present the analysis results. Te difference was statistically signifcant at P < 0.05.

Detection of Diferential Metabolic Small Molecules in
Serum by Metabonomics. In order to investigate whether the metabolism of patients with ICC was abnormal, serum samples from clinical ICC patients and healthy persons were collected for metabolic analysis. Principal component analysis (PCA) (Figure 1(a)), partial least-squares discriminant analysis (PLS-DA) (Figure 1(b)), and orthogonal partial least-squares discriminant analysis (OPLS-DA) (Figure 1(c)) showed signifcant diferences in metabolites between patients with ICC and who were healthy. Serum purine metabolism, aminoacyl-tRNA biosynthesis, and phenylalanine metabolism were signifcantly enriched in patients with ICC ( Figure 1(d)). Volcano plots were used to show diferential metabolites at high and low levels ( Figure 1(e)). Te results of Figure 1(f ) show that purine metabolism metabolites such as inosine, guanosine, hypoxanthine, and xanthine were increased in ICC patients' serum, and inosine monophosphate, adenosine diphosphate ribose, ADP, and DGDP were signifcantly decreased. L-Methionine, L-Tyrosine, and L-Proline were increased and L-Histidine was decreased in aminoacyl-tRNA biosynthesis. Among them, L-Tyrosine is also one of the important products of the metabolic pathway of phenylalanine metabolism. Terefore, the diferential purine metabolism may have a regulatory role in the development of ICC.

Purine Metabolism Genes Are Upregulated in ICC.
To further observe the efect of purine metabolism in ICC, we constructed a clinical model of purine metabolism-related genes and ICC (Figure 2(a)). Compared with the normal group, the tumor group showed higher expression of purine metabolism-related genes such as PFAS, ATIC, IMPDH2, GART, GMPS, IMPDH1, ADSS, ADSL, PRPS1, PPAT, and PAICS (Figure 2(b)). Next, according to the 11 genes survival curve analysis, only the high and low expression of PPAT had survival diferences in ICC (Figure 2(c)). Te expression of PPAT was signifcantly higher in the tumor group compared to the normal group, which was statistically signifcant. In ICC patients, PPAT expression difered less among age, sex, grade, pathologic_M, pathologic_N, path-ologic_T, and stage (Figure 2(d)). It suggested that PPAT may be an important gene afecting ICC.    Journal of Oncology silencing PPAT model was successfully constructed (Figure 3(a)). Te level of IMPDH1 decreased in the presence of PPAT inhibition, which implied a regulatory relationship between IMPDH1 and PPAT (Figure 3(b)). We further examined the cell function situation. Compared with the si-NC group, the number of colonies (Figure 3(c)) and the number of migration and invasion cells (Figures 3(e) and 3(f )) decreased and the apoptosis rate increased in the si-PPAT1 and si-PPAT2 groups (Figure 3(d)). Te level of protein of N-cadherin, vimentin, SOX2, and CD133 decreased and the E-cadherin increased in the si-PPAT1 and si-PPAT2 groups (Figures 3(g) and 3(h)). In conclusion, PPAT was able to inhibit the cell growth and migration of HCCC-9810.  (Figures 4(a) and 4(b)). Compared with the si-NC group, the number of colonies (Figure 4(c)) and the number of migration and invasion cells (Figures 4(e) and 4(f )) decreased and the apoptosis rate increased in the si-IMPDH1-1 and si-IMPDH1-2 groups (Figure 4(d)). Te level of protein of Ncadherin, vimentin, SOX2, and CD133 decreased and the Ecadherin increased in the si-IMPDH1-1 and si-IMPDH1-2 groups (Figures 4(g) and 4(h)). In conclusion, IMPDH1 was able to inhibit the cell growth and migration of HCCC-9810.

Purine Metabolism Could Mediate iCAF to Regulate
HCCC-9810 Cells. Te abovementioned results suggested that PPAT has an inhibitory efect on IMPDH1, which inhibited the cellular function of HCCC-9810. We further explored whether MPA, an inhibitor of IMPDH1, and guanosine, a purine metabolite, could afect iCAF to regulate HCCC-9810 cells. First, we characterized NF cells and iCAF with morphological and iCAF cellular markers α-SMA (84.87%), CD34 (0.02%), and CD45 (0.1%) results to determine that the cells could be used for subsequent experiments ( Figure S1). Compared with the NF group, HCCC-9810 cells in the iCAF group showed increased secretion of infammatory factors IL-6, IL-8, OPN, VEGF, and VCAM-1 ( Figure 5(a)) and rapid cell proliferation ( Figure 5(b)), which indicated that HCCC-9810 cells tended to become malignant under the infuence of iCAF cells. Furthermore, western blot was used to detect epithelial-mesenchymal transition EMT indicators, and the data results showed that N-cadherin and vimentin expression increased and E-cadherin expression was inhibited ( Figure 5(c)). ICAF promoted EMT progression in HCCC-9810 cells. However, MPA drug was able to inhibit the secretion of infammatory factors IL-6, IL-8,

Inhibition of Purine Metabolism Gene IMPDH1 Inhibited the Development of ICC in Mice.
Te abovementioned results indicated that MPA could inhibit the proliferation of HCCC-9810 in vitro, and we further investigated the efect of MMF, a prodrug of MPA, in ICC rats in vivo. We constructed a nude mouse model of ICC, and the tumor volume and tumor weight of nude mice were signifcantly reduced after the addition of MMF (Figure 6(a)). Compared with the ICC group, the expression levels of PPATand IMPDH1 were decreased in the ICC + MMF group (Figures 6(b) and 6(c)). Te next observation of tumor tissue morphology showed that a large number of vacuoles appeared in the tumor tissues after drug administration, probably due to more cell death in the tumor tissues ( Figure 6(d)). Te positivity of apoptosis-associated gene Caspase 3 and proliferationassociated gene Ki67 in tumor tissues was further examined. Te data results showed that Caspase 3 expression was elevated and Ki67 expression was inhibited by the intervention of MMF. Tis indicates that MMF can promote the initiation of tumor cell death program (Figure 6(e)). Te tumor stem cell markers CD133 and SOX2 were reduced by MMF intervention (Figure 6(f )). IHC was used to detect epithelial-mesenchymal transition EMT indicators, and the data results showed that N-cadherin and vimentin expression inhibited and E-cadherin expression was increased by MMF intervention (Figure 6(g)). In summary, inhibition of the purine metabolism gene IMPDH1 inhibited the growth of tumor in mice.

Discussion
In this study, PCA showed statistically signifcant diferences in metabolites between the ICC and control samples. Interestingly, 60 diferential metabolites were found in the samples obtained from ICC patients compared to those in the control group. Notably, guanosine, a purine metabolite, was found to be harmful to the prevention and treatment of ICC. In this study, R-linguistic analysis was used to establish a prognostic model for ICC, and the purine metabolism-related genes PPAT and IMPDH1 were screened. In vitro and in vivo experiments verifed that inhibition of PPAT and IMPDH1 could inhibit the progression of ICC.
Considering the diagnostic uncertainty of most ICC cases, metabolomic approaches may not only elucidate the disease pathogenesis of ICC but may also provide information for diagnosis, prediction, or prognosis [19].
Guanine nucleotides are building blocks of DNA and RNA and are also used by the guanine binding protein family (gproteins) for a large number of cellular functions, such as cytoskeletal rearrangement, membrane transport, protein synthesis, and signal transduction [20]. Te biosynthetic pathway of GTP is important for the progression of many tumors, such as glioblastoma [21], small cell lung cancer subgroup [22], solid and serous malignancies, and clear cell renal cell carcinoma [23]. GTP is a well-known biological metabolite in the context of cancer. Te biosynthesis of GTP and its intracellular levels are well-coordinated and regulated, at least in part, by IMPDH1. In in vitro experiments, exogenous guanosine efectively promoted the expression of PPAT and IMPDH1 in HCCC-9810 cells and stimulated the secretion of the infammatory factors IL-6, IL-8, OPN, VEGF, and VCAM-1. Under PPAT silencing, the expression of IMPDH1 was subsequently reduced, and this fnding suggests that there may be an interaction between IMPDH1 and PPAT. We will consider further studies on the regulatory mechanisms between PPAT and IMPDH1.
As for the role of IMDPH in tumors, recent studies have shown that high IMPDH2 expression was associated with cancer progression [24]. Tumors from mice and humans express high levels of IMPDH1 and IMPDH2, the ratelimiting enzymes for de novo guanine nucleotide synthesis [25]. Te gene encoding these enzymes was MYC. Drug  inhibition of IMPDH1 can reduce the expression of RNA polymerase I-dependent ribosomal RNA and efectively inhibit the growth of lung cancer cells in vitro [26]. According to recent fndings, including the regulation of IMPDH1 through cell cytosolic assembly, metastable inhibition of these enzymes is possible and key diferences exist between IMPDH1 isoforms [17]. Specially designed molecules may be developed to target IMPDH1 with minimal immunosuppressive efects but maximal tumorigenic inhibition, possibly by modulating the aggregation of specifc IMPDH1 isoforms or drugs with diferent tissue penetration capabilities [27]. We used MPA as an inhibitor of IMPDH1 in vitro and found that HCCC-9810 cell growth, migration, and invasion were signifcantly inhibited.
As cancer therapy is moving toward personalization, one approach is to use anticancer drugs encapsulated in nanoparticles, such as liposomes or nanopolymers, and these are taken up by ICC via endocytosis, leading to higher intracellular concentrations and enhanced anticancer drug efcacy [28]. Arsenate plasmonic complex-based chemotherapy is paving the way for efcient nanomedicineenabled boronate afnity-based arsenic chemotherapeutics for on demand site-specifc cancer combination treatment of glioblastoma [29]. Te application of nanomedicine is also expected to bring new hope to liver cancer [30] and pancreatic ductal adenocarcinoma treatment [31,32]. In addition, several targeting strategies have been proposed to deliver drugs specifcally to ICC. For this purpose, bile acid derivatives have been used as "Trojan horses" to enhance the uptake by cancer cells of antitumor fractions in enterohepatic circulation, such as cisplatin, which are chemically bound to bile acids and are recognized and transported through the plasma [33,34]. We used oral MMF in in vivo experiments in nude mice with tumorigenesis, where tumor volume and weight were reduced, tumor tissue cell growth was inhibited, and cancer cell stemness was reduced. We were unable to conduct clinical trials owing to funding and conditions. In future, we will try to overcome this limitation and perform more in-depth studies.
In conclusion, inhibition of the production of the purine metabolite guanosine inhibits the development of ICC. In vitro, we verifed that the purine metabolism-related genes PPAT and IMPDH1 were highly expressed in HCCC-9810 cells. Silencing PPAT and IMPDH1 efectively inhibited the malignant progression of HCCC-9810 cells. In vivo, MMF was shown to inhibit IMPDH1 expression, inhibit tumor growth, and promote apoptosis of HCCC-9810 cells. Silencing the purine metabolism gene IMPDH1 has potential value in the treatment of ICC.

Data Availability
All data included in this study are available from the corresponding authors upon request.

Ethical Approval
Tis research was approved by Medical Ethics Committee of Hunan Provincial People's Hospital (2022-13) and informed consent was obtained in writing from each patient. Animal experiments were approved by the Animal Ethics Committee of Hunan Provincial People's Hospital (2022-05).

Conflicts of Interest
Te authors declare that they have no conficts of interest.

Authors' Contributions
Wen Hao-quan and Wang Xiao-hui conceptualized and supervised the study and developed methodology and software. Liu Chang-jun, Ma, Zhong-zhi, and Gong Wei-zhi involved in data curation and wrote the original draft. Mao  IHC was used to detect the positive rate of proliferation (Ki67) and apoptosis (caspase 3) genes in tissue. (f ) IF was used to detect stem cell characteristics (CD133 and SOX2) markers in tissue. (g) IHC was used to measure the positive rate of E-cadherin, N-cadherin, and vimentin in tissue. * P < 0.05 compared with the ICC group. Te measurement data are expressed as mean ± SD. Te unpaired t-test was used to analyze comparisons between two groups. n � 6.
Xian-Hai visualized and investigated the study. Liu Changjun, Ma, Zhong-zhi, Gong Wei-zhi, and Mao Xian-Hai developed software and validated the study. Liu Chang-ju reviewed and edited the article. All the authors above approved the version of the manuscript to be published.