Identification and validation of hub genes based on machine learning in prostate cancer and AOX1 is an oxidative stress-related biomarker

doi:10.21203/rs.3.rs-1503908/v2

Download PDF

Research

Identification and validation of hub genes based on machine learning in prostate cancer and AOX1 is an oxidative stress-related biomarker

https://doi.org/10.21203/rs.3.rs-1503908/v2

This work is licensed under a CC BY 4.0 License

Version 2

posted

You are reading this latest preprint version

Objectives To explore the potential diagnostic and prognostic biomarkers associated with prostate cancer (PCa).

Methods We downloaded from the Gene Expression Omnibus (GEO) database, six previously published datasets containing gene expression data (GSE8511, GSE14206, GSE46602, GSE55945, GSE69223, and GSE71016). Take GSE8511, GSE14206, GSE46602, and GSE55945 as the training group (127 PCa cases and 52 normal controls) and GSE69223 and GSE71016 as the test group (63 PCa cases and 62 normal controls). Between 127 PCa cases and 52 normal controls, differentially expressed genes (DEGs) were filtered. Following that, candidate PCa biomarkers were discovered using a least absolute shrinkage and selector operation (LASSO) regression model and support vector machine recursive feature elimination (SVM-RFE) analyses. We conducted a difference analysis for these genes in the test group, and a significance level of P< 0.05 was considered statistically significant. To determine the train group’s discriminating ability, the area under the receiver operating characteristic curve (AUC) value was acquired and utilized, with hub genes defined as those having an AUC greater than 85%. The biomarkers’ expression levels and diagnostic utility in PCa were confirmed further in the GSE69223 and GSE71016 datasets. Lastly, to assess the relevant invasion of cells per every sample, the CIBERSORT algorithm and the ESTIMATE technique were utilized.

Results The possible PCa diagnostic biomarkers AOX1, APOC1, ARMCX1, FLRT3, GSTM2, and HPN were found and verified using the GSE69223 and GSE71016 datasets. Immune cell infiltration analysis revealed that AOX1, APOC1, ARMCX1, FLRT3, GSTM2, and HPN were correlated with Mast cells resting, Macrophages M0, Macrophages M2, Neutrophils, NK cells activated, T cells CD4 memory resting, T cells gamma delta, T cells follicular helper, T cells CD4 naive and B cells memory. Meanwhile, AOX1 was identified as an oxidative stress-related biomarker that could use to be a prognostic biomarker. From expreimental validations, we detected that AOX1 was lowly expressed in PCa cell lines. Overexpression of AOX1 clearly reduced the proliferation and migration of the PCa cells, the anti-tumor effect of APX1 may be related to oxidative stress.

Conclusion We used a multi-pronged approach to identify PCa biomarkers and examine the important function of infiltration of cells in PCa. AOX1, APOC1, ARMCX1, FLRT3, GSTM2, and HPN can be used as diagnostic markers of PCa, and AOX1 can be used to predicting the prognosis, which providing new perspectives into the incidence and molecular mechanisms of PCa in the future.

biomarker

prostate cancer

AOX1

CIBERSORT

diagnostic

Prostate cancer (PCa) is a major global health issue. It causes 1.6 million cases and 366000 deaths worldwide each year. It is the second most frequent cancer worldwide and the fifth leading cause of mortality in males [1]. PCa has a rising probability of recurrence due to its biological features, distant micrometastases, and localized residuals, although certain curative therapies exist, such as radical prostatectomy (RP). When used early in the course of recurrence, salvage therapy can significantly lower the risk of distant metastases, lengthen lifespan, and potentially cure the tumor. Therefore, early detection of PCa can significantly improve prognosis and reduce patient mortality [2].

Historically, PCa was diagnosed with a blood test of PSA, a digital rectal examination (DRE), and a biopsy of the prostate. However, PSA lacks specificity, leading to over-diagnosis and overtreatment of PCa. There is a rising clinical need for the detection of new prognostic, predictive, and therapeutic response biomarkers that may be utilized to implement a precision medicine strategy for the management of PCa [3]. For example, some studies have shown that the deletion of phosphatase and tensin homolog (PTEN) is linked with poor prognosis in PCa patients [4]. Its loss in biopsy samples predicted an increased risk of castration-resistant prostate cancer (CRPC), metastasis, and PCa-specific mortality [5, 6]. To prospectively analyze the underlying molecular processes and genomic ramifications of the co-deletion of BRCA2 and RB1 in PCa. We can find that the deletion of BRCA2 leads to a castration-resistant phenotype in human PCa cell lines (LNCaP and lapc4) [7]. Therefore, it is feasible to elucidate the molecular pathways underlying the progression of PCa and to investigate innovative diagnostic approaches for this illness. However, so far, no study has coupled the least absolute shrinkage, and selector operation (LASSO) regression model with support vector machine recursive feature elimination (SVM-RFE) studies to discover PCa biomarkers.

In recent years, immunology research has established that immune cell infiltration is critical for the incidence and progression of PCa. For example, Flammiger et al., in prostate cancer specimens, regulatory T cells (Tregs) was detected by forkhead box P3 (FOXP3) immunohistochemistry found that the augmentation of Tregs was related to an advanced and deteriorating prognosis in prostate cancer tissues [8]. Eastham et al. performed a comparative study of normal patients and PCa patients and found that the level of transforming growth factor-beta (TGF-β) in PCa patients increased, promoting both migration and invasion by PCa cells [9]. Tissue microarray analysis confirmed decreased levels of FOXA1 protein and increased TGF-β signaling pathway in CRPC as compared with primary tumors suppress CRPC progression [10]. In addition, some studies have shown that tyrosine hydroxylase 2 (Th2) and central memory T cell (TCM) are associated with PCa recurrence after RP, and they are independent protective factors [11]. Nevertheless, until yet, few researches have utilized the CIBERSORT technique to investigate the infiltration of immune cells and possible PCa biomarkers.

We downloaded six publicly available datasets on PCa from the database of Gene Expression Omnibus (GEO) in this article. A metadata cohort was created by combining the four datasets and taking them as the training group. And we merged the other two datasets into a metadata cohort and took it as the treatment group. In the training group, between 127 PCa cases and 52 normal controls, differentially expressed genes (DEGs) were sieved. The screening and identification of diagnostic biomarkers for PCa were carried out using machine-learning techniques. These candidate genes were identified and validated in the treatment group. Additionally, the CIBERSORT methods were employed to investigate the relationship between biomarkers and PCa-infiltrating immune cells in order to get a better understanding of the molecular immunological process underlying PCa and to provide a practical and conceptual framework for future study.

2.1. Gene expression data acquisition and processing. Using the geo database (https://www.ncbi.nlm.nih.gov/geo/), we searched all the data about PCa. Ultimately, the GEO database was used to download the raw data from the GSE8511, GSE14206, GSE46602, GSE55945, GSE69223, and GSE71016 datasets were downloaded from. We divided these datasets into the training group (GSE8511, GSE14206, GSE46602, and GSE55945; 127 PCa cases and 52 normal controls) and the test group (GSE69223 and GSE71016; 63 PCa cases and 62 normal controls). The datasets are merged into a metadata queue in the training group or the test group, and they are preprocessed and eliminated batching effects using the combat capabilities of the “SVA” package [12] (Table 1).

2.2. Identification of DEGs in PCa. The R (http://www.bioconductor.org/) package “limma” was employed in the usual mode of comparison to detect DEGs between 127 PCa patients and 52 normal controls in the training group [13]. We set the threshold points of DEGs as |log fold change (FC)|>1, adjusted false discovery rate P < 0.05. In the training group, we compare 127 PCa cases with 52 normal controls. The volcanic map was created using the R software program “ggplot2” to detect DEGs. Then, using the R package “heatmap,” a heat map of DEGs was created.

2.3. Functional enrichment analysis. Gene module-related functions were found using functional enrichment studies done using the R package “cluster profile” employing the gene ontology (GO), the disease ontology (DO) ontologies, and the Kyoto encyclopedia of genes and genomes (KEGG) [14]. We achieved the gene set enrichment analysis (GSEA) of the training group by applying signal pathway differences. By the use of the ‘cluster Profiler’ and ‘enrich plot’ programs, the gene expression matrix was subjected to GSEA analysis, and the reference gene was ‘c2.cp.kegg.v7.4.symbols.gmt’ [15]. KEGG GSEA analysis was applied to the PCa and normal cases of the training group, respectively. Significant saturation was defined as P < 0.05.

2.4. Screening characteristic related biomarkers via machine learning. Two machine learning methods were employed to assess PCa-related biomarkers in order to uncover significant prognostic factors. The LASSO technique was used in conjunction with the R package “glmnet” to identify genes that were clearly connected with the distinction between PCa and normal patients [16]. Also, as a surveillance machine learning technique for support vectors, the support vector machine (SVM) identifies the optimal variables by eliminating the SVM’s feature vectors. To minimize overfitting, the best genes were chosen using a recursive feature elimination (RFE) approach. Thus, in order to determine the genes set with the highest discriminatory power, we used SVM-RFE to choose the proper characteristics. To perform classification analysis on the biomarkers selected for PCa diagnosis, the SVM-RFE classifier from the R package “e1071” was used [17]. Finally, we use the Venn diagram to find the overlapping genes of the two algorithms, and these genes will be further verified in the test group.

2.5. Diagnostic value of the biomarkers in PCa. To assess the differences of these genes in the test group and verify the predictive value of established biomarkers, we used the R package “ggpubr”, and P < 0.05 was deemed significant statistically [18]. Then, we use the R package “proc” to generate an ROC curve in the training group (127 PCa cases and 52 normal controls) [19], with hub genes defined as those value “>85%” (AUC). To evaluate the diagnostic impact of PCa on normal samples, which were then verified in the test group, the AUC value was used (63 PCa cases and 62 normal controls).

2.6. Assessment of immune cell infiltration. To classify 22 different kinds of immune cell matrix, the CIBERSORT method was used. To assess the presumed abundance of immune cells, with 1,000 permutations, a reference set of 22 immune cell subtypes was utilized [20]. Regarding p < 0.05, the invasion of the immune cell matrix was generated. To illustrate the association within 22 different kinds of immune cell infiltration and to create a correlation between heatmap and boxplot, the program ‘corrplot’ was utilized [21]. Violin plots were created by the use of the R package “vioplot” to illustrate the variations in the infiltration of immune cells between PCa and normal samples [22].

2.7. Analysis of correlations between identified genes and immune cell infiltration. An R software Spearman’s rank correlation analysis was performed to detect the correlation between the levels of expression of the identified biomarkers and the level of infiltrating immune cells [23]. The resultant correlations were conducted and displayed using the “ggpubr” package’s charting approach [24].

2.8. Cell culture and transfection. The PCa cell lines (LNCaP, PC3 and DU145) and the normal prostate cell line (RWPE-1) were cultured in RPMI-1640 medium. Oe-AOX1 and its negative control (Oe-ctrl) were synthesized by GenePharma (Shanghai, China). LNCaP and PC3 cells were equably plated in 24-well plates. When the two cell lines reached about 80-90% confluence, they were transfected according to the instructions.

2.9. Western blot analysis. Cell proteins were electrophoresed on 12% SDS-PAGE gel and transferred to the PVDF membranes. The membranes were blocked with silk milk (5%) at room temperature for 1h. Afterward, the membranes were incubated with primary antibodies anti-AOX1 (ab92519; 1:500; Abcam; USA) and β-actin (ab8226; 1:2,500; Abcam; USA) at 4 °C overnight, and secondary antibodies (ab6721; 1:3,000; ab6728; 1:3,000; Abcam; USA) at room temperature for 1h. The membranes were visualized by enhanced ECL detection kit (Beyotime, China).

2.10. RT-qPCR analysis. RNA was extracted using TriZol (Beyotime, China) from PCa cells. Extracted RNA was reverse transcribed into complementary DNA. SYBR Green qPCR Master Mix was used to detect the quantified expressions using the 2^-ΔΔCq method.

2.11. Cell proliferation assay. 96 well plates were taken to seed PCa cells (LNCaP and PC3). After incubation at 37°C and 5% CO₂ for six times (0-96h), a CCK−8 reagent test kit (Tiangen), was mixed at 10 µl/ well, and then PCa cells were evenly incubated for 1h at 37°C and 5% CO₂. Finally, the OD value (at 450nm) was read on the microplate reader for analysis.

2.12. Clone formation assay. The cultured cells in the logarithmic growth phase were diluted, and seeded into dishes containing culture medium at the appropriate gradient density. Cells were cultured at 37℃ with 5% CO₂ for 2 weeks. Then, they were washed twice with PBS. Paraformaldehyde was added and fixed for 15 minutes. And colonies were stained with 0.1% crystal violet for 15 minutes. Colonies were washed off with water. Clones greater than 10 cells were counted.

2.13. Measurement of malondialdehyde (MDA). Cells were lysed and centrifuged at 10,000g for 10min. MDA Assay Kit (S0131, Beyotime, China) was used to measure the MDA content of cells. The samples were tested at 532 nm with microplate reader and compared to the standard curve of MDA.

2.14. Measurement of Glutathione (GSH) and ROS. The contents of GSH and ROS were detected using corresponding kits according to the reference protocols provided by the manufacturer.

2.15. Statistical Analysis. All statistical analyses were performed utilizing the Perl version 5.32.1 and R software version 4.1.2. P <0.05 was judged statistically significant.

3.1. Identification of DEGs in PCa. The DEGs of 127 PCa cases and 52 normal controls from the GEO databases (GSE8511, GSE14206, GSE46602, and GSE55945) were identified by R package “limma.” There were 37 DEGs, 17 of which were up-regulated and 20 of which were downregulated. Log fold change (FC) > 0 represents that the gene is up-regulated in the training group, and log FC< 0 represents that the gene is down-regulated, which can be shown in the volcano plot (Figure 1(a)). The levels of expression of 37 DEGs were illustrated by the heat map (Figure 1(b)).

3.2. Functional enrichment analysis. To evaluate the 37 DEGs’ probable biological activities, the GO, KEGG, DO, and GSEA analyses were performed utilizing the R package “cluster profile”. It was illustrated by the GO results that the majority of these genes were associated with basement membrane organization, cornification, and positive regulation of secretion by cell (Figure 2(a)). KEGG enrichment analysis identified genes to be involved in drug metabolism, including those involved in cytochrome P450, nicotinate and nicotinamide metabolism, and retinol metabolism (Figure 2(b)). The results of DO showed that diseases enriched by DEGs were primarily related to chronic myeloproliferative diseases, epidermolysis bullosa, integumentary system disease, vesiculobullous skin disease, peripheral primitive neuroectodermal tumor, and prostate cancer (Figure 2(c)). The GSEA results in the PCa group demonstrated that the enriched pathways mainly involved bladder cancer, cell cycle, purine metabolism, ribosome, and toll-like receptor signaling pathway (Figure 2(d)). The GSEA results in the control group demonstrated that the enriched pathways mainly involved glutathione metabolism, focal adhesion, cytochrome P4 metabolism of xenobiotics, cytochrome P450 metabolism of drugs, and vascular smooth muscle contraction (Figure 2(e)).

3.3. Screening diagnostic feature biomarkers for PCa. To select underlying diagnostic biomarkers, two distinct approaches are applied. The LASSO logistic regression approach was utilized to detect twenty-one genes as possible PCa biomarkers from robust DEGs (Figure 3(a)). The SVM-RFE technique was used to determine 28 genes from the DEGs (Figure 3(b)). Finally, we utilized the Venn diagram to overlap the gene markers provided by the two methods and the sixteen related biomarkers, including AOX1, HPN, GSTM2, APOC1, ARMCX1, FLRT3, MSMB, KRT15, GDF15, DLX1, CD177, NTNG2, CPLX3, ACSM1, ERG, and CD38, were obtained (Figure 3(c)).

3.4. Identification and validation of diagnostic feature biomarkers for PCa. To produce more accurate and reliable results, the GSE69223 dataset and GSE71016 dataset are used to ascertain the degree to which a gene is expressed in these six genes. The levels of expression of ACSM1, APOC1, DLX1, GDF15, and HPN were evidently greater in PCa tissues than in normal tissue (Figure 3(d)-3(h); all P < 0.05). On the other hand, the opposite outcome was seen for AOX1, ARMCX1, CD177, FLRT3, GSTM2, KRT15, and NTNG2 (Figure 3(i)-3(o); all P<0.05). Additionally, there was no signiﬁcant change in the amounts of CD38, CPLX3, ERG, and MSMB between PCa tissues and the normal tissue (all P>0.05). We drew ROC curves of the above twelve meaningful genes in the training group, with hub genes defined as those having an AUC greater than 85%. Then, we identified six PCa-related diagnostic genes, and the AUC of AOX1 was 0.921 (95% CI 0.878-0.956), APOC1 was 0.853 (95% CI 0.782-0.919), ARMCX1 was 0.883 (95% CI 0.834-0.928), FLRT3 was 0.854 (95%CI 0.796-0.904), GSTM2 was 0.877 (95%CI 0.823-0.923) and HPN was 0.871 (95%CI 0.817-0.921).

Then, a robust discrimination was proved in the GSE69223 and GSE71016 datasets, and the AUC of AOX1 was 0.726 (95% CI 0.641-0.810), APOC1 was 0.701 (95% CI 0.607-0.792), ARMCX1 was 0.722 (95% CI 0.625-0.808), FLRT3 was 0.719 (95% CI 0.625-0.804), GSTM2 was 0.696 (95% CI 0.600-0.785) and HPN was 0.789 (95%CI 0.707-0.863) (Figure 4(a)-4(f)).

3.5. Assessment of immune cell infiltration. Following that, we utilized the CIBERSORT technique to visualize the invasion of 22 distinct immune cell kinds in the training group (Figure 5(a)). Additionally, the CIBERSORT method demonstrated the invasion of 22 distinct immune cell types, and the heatmaps showed strong positive relationships between T cells CD4 memory resting and plasma cells (r=0.54) and strong inverse relationships between T cells CD4 memory resting and macrophages M1 (r= -0.51) (Figure 5(b)). We studied the component of immune cells in PCa and normal tissues. T cells CD8 in PCa was remarkably higher compared with the normal controls as indicated in the findings (P = 0.032), while mast cells resting was lower than the normal controls (P = 0.005; Figure 5(c)).

3.6. Correlation analysis between PCa-related biomarkers and immune infiltrating cells. The correlation analysis revealed that AOX1 was positively associated with mast cells resting (R = 0.4, P = 0.036) and negatively associated with macrophages M0 (R = -0.46, P = 0.013; Figure 6(a)). APOC1 was positively associated with macrophages M0 (R = 0.63, P < 0.001), neutrophils (R = 0.4, P = 0.035), and macrophages M2(R = 0.4, P = 0.036) and negatively associated with mast cells resting (R = -0.59, P = 0.001; Figure 6(b)). ARMCX1 was positively associated with NK cells activated (R = 0.38, P = 0.045) and mast cells resting (R = 0.38, P = 0.049; Figure 6(c)). FLRT3 was positively associated with T cells CD4 memory resting (R = 0.41, P = 0.032) and negatively associated with T cells gamma delta (R = -0.43, P = 0.023; Figure 6(d)). GSTM2 was positively associated with T cells follicular helper (R = 0.43, P = 0.022) and negatively associated with T cells CD4 naive (R = -0.41, P = 0.029; Figure 6(e)). HPN was positively associated with B cells memory (R = 0.47, P = 0.011) and macrophages M0 (R = 0.41, P = 0.032) and negatively associated with mast cells resting (R = - 0.41, P = 0.031; Figure 6(f)).

3.7. Experimental verification of AOX1 in PCa. Oxidative stress is closely related to cancer. To further identify whether these six PCa-related diagnostic genes are related to oxidative stress, we downloaded the gene sets of oxidative stress genes from the website GeneCards (https://www.genecards.org/). After taking the intersection, only AOX1 among the DEGs was classified as an oxidative stress gene (Figure 7(a)). The expression of AOX1 was verified. We obtained several PCa cell lines (LNCaP, PC3 and DU145) for experimental validation, with normal prostate cell line (RWPE-1) as the ctrol group. In Figure 7(b) and Figure 7(c), not only mRNA level but also protein level showed the same significant decrease (P < 0.05) trend of AOX1 in PCa cell lines. Subsequently, in order to detect the specific role of AOX1 in the progression of PCa, we applied the functional overexpression (oe-AOX1) into PCa cell lines (LNCaP and PC3). Figure 7(d) showed that the overexpression transfection was clearly successful in PCa cell lines. CCK-8 detection reveled that overexpression of AOX1 could significantly inhibit the proliferation activity of LNCaP and PC3 cells (Figure 7(e) and 7(f)). Similarly, the colony formation assays clearly revealed that the clone capacity of LNCaP and PC3 cells were inhibited by the overexpression of AOX1 (Figure 7(g) and 7(h)). Moreover, we detected the levels of MDA, ROS, and GSH in LNCaP and PC3 cells. The results showed a significant increase in ROS and MDA levels, while an obviously decrease in GSH level (Figure 7(i)-7(k)). To sum up, AXO1 acted as the role of cancer suppressor during the progression of PCa, which may be partly achieved by triggering oxidative stress.

Nowadays, PCa remains to be one of the leading reasons of cancer mortality in males. In view of the recent success of immunotherapy in many hematological and solid malignancies, this territory is being explored widely in PCa[25]. More and more researchers recognize that immune cell infiltration is associated with cancer and other illnesses[26]. Therefore, immunotherapy may become a feasible tool against PCa. To identify tumor-infiltrating immune cells and their correlation with the prognosis of gastric cancer, colorectal cancer, breast cancer, and osteosarcoma, the CIBERSORT technique has been successfully applied[20, 27-29]. However, its importance in PCa has not yet been elucidated. The purpose of this research was to establish the value of immune cell infiltration in PCa and determine potential diagnostic biomarkers for PCa.

To the extent that we are aware, this is the first retrospective research to screen the diagnostic biomarkers of PCa by combining the LASSO and RVM-RFE algorithm and utilizing the CIBERSORT algorithm to analyze the immune cell invasion of PCa. We downloaded six datasets from the GEO database. Two are merged as the test group, and the other four are merged as the training group. In the training group, we identified 37 DEGs, including 17 up-regulated and 20 downregulated genes. The results of GSEA in PCa cases demonstrated that the enriched pathways mainly involved bladder cancer, cell cycle, purine metabolism, ribosome, and toll-like receptor signaling pathway.

Based on the two algorithms, we screened sixteen genes as possible PCa-related biomarkers using robust DEGs. Then, the differences among these sixteen genes in the test group were analyzed, and that twelve genes were found to be statistically significant (P < 0.05). Finally, we drew the ROC curves of these twelve genes, and we identified the final six PCa-related diagnostic biomarkers (AOX1, APOC1, ARMCX1, FLRT3, GSTM2, and HPN). To assess the predictive efficacy of these six diagnostic biomarkers, we calculated their ROC curves in the test group. Hydrogen peroxide is produced by aldehyde oxidase, which, under some conditions, has the ability to accelerate the creation of superoxide. Aldehyde oxidase is a candidate for amyotrophic lateral sclerosis. Aldehyde oxidase 1 (AOX1) is a monooxygenase that contains flavin, which is included in the detoxification of a range of aldehydes and nitrogen-containing heterocyclic compounds. Its most often seen substrates are alien organisms that contain sulfur and nitrogen[30]. Li et al. examined the connection between a total of 6126633 single nucleotide polymorphisms (SNPs) and PCa-specific survival time utilizing a cox proportional hazard model and used quantitative expression trait locus analysis to test the association between positive candidate SNPs and gene expression. At the AOX1 gene, the SNP rs73055188 is related to a particular survival time for PCa, and the expression level of AOX1 is associated with the biochemical recurrence of PCa[31]. As an important lipoprotein, apolipoprotein is involved in maintaining lipoprotein structure, controlling the interaction of lipoproteins with cell receptors, and the activation of intravascular lipoprotein metabolism enzyme. Apolipoprotein C-I (APOC1) is the most minimal apolipoprotein found on chromosome 19 (only 6.6 kDa). It is a component of lipoproteins with a high triglyceride content and lipoproteins with high density, and it is very important for the metabolism of plasma lipoprotein[32]. Su et al. use RNA interference (RNAi) to inhibit the expression of APOC1 in PCa cell lines. Then it was found that APOC1 mediated the survival, apoptosis, and cell cycle distribution of PCa cells by regulating surviving/Rb/p21/caspase-3 signaling pathway[33]. Armadillo repeat-containing X-linked 1 (ARMCX1), a mammalian-specific gene, encodes mitochondrial localization protein. It has a vital role in mitochondrial transport regulation in injured axons and may have an effect on neuronal survival and regrowth[34]. At present, there is a lack of research between ARMCX1 and PCa. FLRT3 is a fibronectin leucine-rich transmembrane protein that is involved in the control of neuronal cell development and morphogenesis[35]. Balabhadrapatruni et al. used Methods of gene knockdown to show that prolyl 4-hydroxylase subunit alpha 1 (P4HA1) expression is necessary to proliferate and invade the PCa cells and can be used as a feasible therapeutic target. They also verified the results of FLRT3 in benign prostate, PCa, and metastatic PCa tissues by western blot analysis and found a negative correlation between P4HA1 and FLRT3. It was finally confirmed that FLRT3 promoted the ability of P4HA1-mediated PCa cells to migrate and invade [36]. As a member of the superfamily of multifunctional enzymes, by accelerating the binding of glutathione (GSH) to electrophilic molecules, glutathione S-transferases contribute significantly to the defense against a variety of toxic and carcinogenic compounds. Glutathione S-transferase mu 2 (GSTM2) is an important phase II antioxidant enzyme in eukaryotes, which is involved in the production of prostaglandin E2 (PGE2)[37]. Several studies on GSTM2 have been carried out in PCa. Albawardi et al. found that GSTM2 was amplificated in PCa based on SNP microarrays [38]. In addition, three other studies have shown that GSTM2 is highly methylated in PCa[39-41]. Hepsin (HPN) is a type II transmembrane, but not secreted, protein[42]. Ma et al. discovered that the amount of HPN protein expression was significantly greater in PCa with a Gleason score > 6 than in patients with PCa who have a Gleason value = 6 through single-cell RNA sequencing, which established that HPN might be used as a biomarker for the diagnosis and categorization of PCa[43]. The above evidence shows that except for armcx1, other genes contribute significantly to the progression of PCa.

We further evaluated the immune cell infiltration of PCa using CIBERSORT to investigate the function of immune cell infiltration in PCa. An increase in CD8 T cell infiltration and a reduction in mast cell infiltration during rest are considered to be associated with the occurrence and progression of PCa. Correlation analysis of biomarkers associated with PCa and immune invading cells revealed that AOX1, APOC1, ARMCX1, GSTM2, and HPN were significantly correlated with mass cells resting. And HPN, AOX1, and APOC1 were significantly correlated with macrophages M0. Florent et al. performed immunohistochemistry on the tumors of 51 patients with node-positive PCa. The presence of a large density of CD8 + T cells in tumors was discovered to be related to an increased risk of clinical progression in patients with node-positive PCa[44]. Mast cells are involved in a variety of disorders, including hypersensitivity, inflammation, and fibrosis. Notably, mast cells play a critical function in tumor progression as well. CIBERSORT algorithm was used to analyze 52 normal prostate tissues and 497 primary tumors of PCa patients from TCGA. The fraction of static mast cells in PCa was shown to be significantly different than that in normal tissues. Increased resting mast cell count is connected with a bad prognosis. Radiotherapy and targeted molecular treatment can have an effect on resting mast cell immune infiltration[45]. Somaiyeh et al. found that etoposide-treated cancer cells M0 macrophages, THP-1 cells treated with toll-like receptor 4 (TLR4) agonists cultured with the supernatant of P human prostate cancer cell line (PC3) cells had an obvious protective effect on etoposide-induced PCa cells apoptosis by enzyme-linked immunosorbent assay with flow cytometry (ELISA)[46]. Numerous studies, as well as our recent findings, indicate that numerous kinds of invasive immune cells play a crucial factor in PCa and should become the key to future research. Numerous invasive immune cell types play an important role in PCa and should be the subject of future investigation.

Recently, emerging evidence has revealed that oxidative stress is closely related to occurrence and progression of cancer[47, 48]. In this work, we intersected the gene AOX1 between 6 hub genes and 664 oxidative stress-related genes. Xiong et al. has identified that AOX1 was decreased and acted as a tumor suppressor gene in clear cell Renal Cell Carcinoma (ccRCC) and PCa[49, 50]. In this work, we illustrated that AOX1 was down-regulated in PCa cells. Subsequently, functional experiments were executed using oe-AOX1 transfection, revealing that the proliferation and migration of PCa cells were inhibited following the overexpression of AOX1, which were consistent with previous reports on AOX1. More importantly, overexpression of AOX1 induced ROS and MDA accumulation, and restored the GSH content. In short, the anti-cancer effect of AOX1 may work through the activation of oxidative stress.

Our investigation contains several limitations that should be considered. The primary constraint is the tiny sample size of the published datasets. Our findings need to be validated in other datasets and clinical trials to determine whether AOX1, APOC1, armcx1, FLRT3, gstm2, and HPN can be used as biomarkers of PCa. CIBERSORT algorithm was based on limited retrospective gene data. Even though some earlier study findings are compatible with ours, the immune cell infiltration analysis in PCa is currently restricted, and our conclusion should be verified by a prospective study with larger sample size. Moreover, we preliminarily verified the anti-oncogenic role of AOX1 in the malignant progression of PCa and remarkably identified its partly potential mechanism by triggering the occurrence of oxidative stress in vitro, but more rigorous tests are still needed for deep verification. A preprint has previously been published [51].

In summary, we found that AOX1, APOC1, ARMCX1, FLRT3, GSTM2, and HPN were PCa-related biomarkers. T cells CD8、mast cells resting、Macrophages M0、NK cells activated、Neutrophils、Macrophages M2、B cells memory、T cells CD4 memory resting、T cells gamma delta、T cells follicular helper and T cells CD4 naive may be related with the onset and progression of PCa. Further exploring the interaction between immune cell infiltration will help to improve the immunomodulatory treatment of PCA patients. Additional research addressing the association between PCa and immune cell infiltration will aid in improving the immunomodulatory treatment of patients with PCa. Through experimental validation, tumor suppressor function of AOX1 in PCa and identified its potential anti-tumor mechanism by triggering oxidative stress, which may be beneficial to further elucidate the pathogenesis of PCa and provide a new horizon for clinical treatment.

Data Availability
All data used in this work can be acquired from the Gene Expression Omnibus (GEO; https://www.ncbi.nlm.nih.gov/geo/).

Conflicts of Interest

The authors declare that they have no competing interests.

Authors’ Contributions

Di Hu, Kaisheng Yuan, Xiaocong Mo and Yin Li conceived and designed the experiments; Di Hu, Kaisheng Yuan, Xiaocong Mo and Juyu Luo conducted the research; Di Hu, Yin Li , Kaisheng Yuan and Hang Liu contributed materials and analysis tools; Di Hu, Xiaocong Mo, Juyu Luo, Kaisheng Yuan and Cheng Huang analyzed the results; Di Hu, Kaisheng Yuan and Xiaocong Mo wrote the paper. All authors reviewed the manuscript. Di Hu, Kaisheng Yuan and Xiaocong Mo contributed equally to this work. Yin Li is the corresponding author.

Funding Statement

This study was supported by the National Natural Science Foundation of China (No. 81900160).

ETHICAL APPROVAL STATEMENT

The study was approved by the Ethics Committee of the First Affiliated Hospital of Jinan University, China. GEO database belong to public database. The patients involved in the database have obtained ethical approval. Users can download relevant data for free for research and publish relevant articles.

Acknowledgments

We acknowledge GEO database for providing their platforms and contributors for uploading their meaningful datasets.

[1] T. Tsujino, K. Komura, T. Inamoto, and H. Azuma, "CRISPR Screen Contributes to Novel Target Discovery in Prostate Cancer," Int J Mol Sci, vol. 22, no. 23, 2021.

[2] S. V. Carlsson, and A. J. Vickers, "Screening for Prostate Cancer," Med Clin North Am, vol. 104, no. 6, pp. 1051-1062, 2020.

[3] A. Cimadamore, R. Mazzucchelli, A. Lopez-Beltran, F. Massari, M. Santoni, M. Scarpelli, L. Cheng, and R. Montironi, "Prostate Cancer in 2021: Novelties in Prognostic and Therapeutic Biomarker Evaluation," Cancers (Basel), vol. 13, no. 14, 2021.

[4] K. W. Fisher, R. Montironi, A. Lopez Beltran, H. Moch, L. Wang, M. Scarpelli, S. R. Williamson, M. O. Koch, and L. Cheng, "Molecular foundations for personalized therapy in prostate cancer," Curr Drug Targets, vol. 16, no. 2, pp. 103-14, 2015.

[5] T. L. Lotan, F. L. Carvalho, S. B. Peskoe, J. L. Hicks, J. Good, H. Fedor, E. Humphreys, M. Han, E. A. Platz, J. A. Squire, A. M. De Marzo, and D. M. Berman, "PTEN loss is associated with upgrading of prostate cancer from biopsy to radical prostatectomy," Mod Pathol, vol. 28, no. 1, pp. 128-137, 2015.

[6] C. G. Picanço-Albuquerque, C. L. Morais, F. L. Carvalho, S. B. Peskoe, J. L. Hicks, O. Ludkovski, T. Vidotto, H. Fedor, E. Humphreys, M. Han, E. A. Platz, A. M. De Marzo, D. M. Berman, T. L. Lotan, and J. A. Squire, "In prostate cancer needle biopsies, detections of PTEN loss by fluorescence in situ hybridization (FISH) and by immunohistochemistry (IHC) are concordant and show consistent association with upgrading," Virchows Arch, vol. 468, no. 5, pp. 607-17, 2016.

[7] G. Chakraborty, J. Armenia, Y. Z. Mazzu, S. Nandakumar, K. H. Stopsack, M. O. Atiq, K. Komura, L. Jehane, R. Hirani, K. Chadalavada, Y. Yoshikawa, N. A. Khan, Y. Chen, W. Abida, L. A. Mucci, G. M. Lee, G. J. Nanjangud, and P. W. Kantoff, "Significance of BRCA2 and RB1 Co-loss in Aggressive Prostate Cancer Progression," Clin Cancer Res, vol. 26, no. 8, pp. 2047-2064, 2020.

[8] A. Flammiger, L. Weisbach, H. Huland, P. Tennstedt, R. Simon, S. Minner, C. Bokemeyer, G. Sauter, T. Schlomm, and M. Trepel, "High tissue density of FOXP3+ T cells is associated with clinical outcome in prostate cancer," Eur J Cancer, vol. 49, no. 6, pp. 1273-9, 2013.

[9] D. Bansal, M. A. Reimers, E. M. Knoche, and R. K. Pachynski, "Immunotherapy and Immunotherapy Combinations in Metastatic Castration-Resistant Prostate Cancer," Cancers (Basel), vol. 13, no. 2, 2021.

[10] B. Song, S. H. Park, J. C. Zhao, K. W. Fong, S. Li, Y. Lee, Y. A. Yang, S. Sridhar, X. Lu, S. A. Abdulkadir, R. L. Vessella, C. Morrissey, T. M. Kuzel, W. Catalona, X. Yang, and J. Yu, "Targeting FOXA1-mediated repression of TGF-β signaling suppresses castration-resistant prostate cancer progression," J Clin Invest, vol. 129, no. 2, pp. 569-582, 2019.

[11] X. Rui, S. Shao, L. Wang, and J. Leng, "Identification of recurrence marker associated with immune infiltration in prostate cancer with radical resection and build prognostic nomogram," BMC Cancer, vol. 19, no. 1, pp. 1179, 2019.

[12] B. Liu, Y. Xie, and Z. Wu, "Identification of Candidate Genes and Pathways in Nonsegmental Vitiligo Using Integrated Bioinformatics Methods," Dermatology, vol. 237, no. 3, pp. 464-472, 2021.

[13] K. Yuan, R. Zeng, P. Deng, A. Zhang, H. Liu, N. Wang, Y. Tang, Z. Yin, and H. Liu, "Identification and Verification of Immune-Related Genes Prognostic Signature Based on ssGSEA for Adrenocortical Carcinoma (ACC)," Int J Gen Med, vol. 15, pp. 1471-1483, 2022.

[14] C. Ren, M. Li, W. Du, J. Lü, Y. Zheng, H. Xu, and R. Quan, "Comprehensive Bioinformatics Analysis Reveals Hub Genes and Inflammation State of Rheumatoid Arthritis," Biomed Res Int, vol. 2020, pp. 6943103, 2020.

[15] A. Zito, M. Lualdi, P. Granata, D. Cocciadiferro, A. Novelli, T. Alberio, R. Casalone, and M. Fasano, "Gene Set Enrichment Analysis of Interaction Networks Weighted by Node Centrality," Front Genet, vol. 12, pp. 577623, 2021.

[16] M. Zhang, K. Zhu, H. Pu, Z. Wang, H. Zhao, J. Zhang, and Y. Wang, "An Immune-Related Signature Predicts Survival in Patients With Lung Adenocarcinoma," Front Oncol, vol. 9, pp. 1314, 2019.

[17] H. Jiang, J. Gu, J. Du, X. Qi, C. Qian, and B. Fei, "A 21‑gene Support Vector Machine classifier and a 10‑gene risk score system constructed for patients with gastric cancer," Mol Med Rep, vol. 21, no. 1, pp. 347-359, 2020.

[18] Q. Cheng, X. Chen, H. Wu, and Y. Du, "Three hematologic/immune system-specific expressed genes are considered as the potential biomarkers for the diagnosis of early rheumatoid arthritis through bioinformatics analysis," J Transl Med, vol. 19, no. 1, pp. 18, 2021.

[19] Y. Jiao, Y. Li, P. Jiang, W. Han, and Y. Liu, "PGM5: a novel diagnostic and prognostic biomarker for liver cancer," PeerJ, vol. 7, pp. e7070, 2019.

[20] C. Zhang, J. H. Zheng, Z. H. Lin, H. Y. Lv, Z. M. Ye, Y. P. Chen, and X. Y. Zhang, "Profiles of immune cell infiltration and immune-related genes in the tumor microenvironment of osteosarcoma," Aging (Albany NY), vol. 12, no. 4, pp. 3486-3501, 2020.

[21] H. Zhang, R. Liu, L. Sun, W. Guo, X. Ji, and X. Hu, "Comprehensive Analysis of Gene Expression Changes and Validation in Hepatocellular Carcinoma," Onco Targets Ther, vol. 14, pp. 1021-1031, 2021.

[22] J. Gao, L. Shi, J. Gu, D. Zhang, W. Wang, X. Zhu, and J. Liu, "Difference of immune cell infiltration between stable and unstable carotid artery atherosclerosis," J Cell Mol Med, vol. 25, no. 23, pp. 10973-10979, 2021.

[23] C. Ma, H. Luo, J. Cao, C. Gao, X. Fa, and G. Wang, "Independent prognostic implications of RRM2 in lung adenocarcinoma," J Cancer, vol. 11, no. 23, pp. 7009-7022, 2020.

[24] Y. Fang, S. Huang, L. Han, S. Wang, and B. Xiong, "Comprehensive Analysis of Peritoneal Metastasis Sequencing Data to Identify LINC00924 as a Prognostic Biomarker in Gastric Cancer," Cancer Manag Res, vol. 13, pp. 5599-5611, 2021.

[25] I. Tsaur, M. P. Brandt, E. Juengel, C. Manceau, and G. Ploussard, "Immunotherapy in prostate cancer: new horizon of hurdles and hopes," World J Urol, vol. 39, no. 5, pp. 1387-1403, 2021.

[26] Z. Liu, H. Li, and S. Pan, "Discovery and Validation of Key Biomarkers Based on Immune Infiltrates in Alzheimer's Disease," Front Genet, vol. 12, pp. 658323, 2021.

[27] D. Zeng, R. Zhou, Y. Yu, Y. Luo, J. Zhang, H. Sun, J. Bin, Y. Liao, J. Rao, Y. Zhang, and W. Liao, "Gene expression profiles for a prognostic immunoscore in gastric cancer," Br J Surg, vol. 105, no. 10, pp. 1338-1348, 2018.

[28] Y. Xiong, K. Wang, H. Zhou, L. Peng, W. You, and Z. Fu, "Profiles of immune infiltration in colorectal cancer and their clinical significant: A gene expression-based study," Cancer Med, vol. 7, no. 9, pp. 4496-4508, 2018.

[29] H. R. Ali, L. Chlon, P. D. Pharoah, F. Markowetz, and C. Caldas, "Patterns of Immune Infiltration in Breast Cancer and Their Clinical Implications: A Gene-Expression-Based Retrospective Study," PLoS Med, vol. 13, no. 12, pp. e1002194, 2016.

[30] W. Zhang, W. Chai, Z. Zhu, and X. Li, "Aldehyde oxidase 1 promoted the occurrence and development of colorectal cancer by up-regulation of expression of CD133," Int Immunopharmacol, vol. 85, pp. 106618, 2020.

[31] W. Li, M. Middha, M. Bicak, D. D. Sjoberg, E. Vertosick, A. Dahlin, C. Häggström, G. Hallmans, A. C. Rönn, P. Stattin, O. Melander, D. Ulmert, H. Lilja, and R. J. Klein, "Genome-wide Scan Identifies Role for AOX1 in Prostate Cancer Survival," Eur Urol, vol. 74, no. 6, pp. 710-719, 2018.

[32] X. Shi, J. Wang, S. Dai, L. Qin, J. Zhou, and Y. Chen, "Apolipoprotein C1 (APOC1): A Novel Diagnostic and Prognostic Biomarker for Cervical Cancer," Onco Targets Ther, vol. 13, pp. 12881-12891, 2020.

[33] W. P. Su, L. N. Sun, S. L. Yang, H. Zhao, T. Y. Zeng, W. Z. Wu, and D. Wang, "Apolipoprotein C1 promotes prostate cancer cell proliferation in vitro," J Biochem Mol Toxicol, vol. 32, no. 7, pp. e22158, 2018.

[34] R. Cartoni, M. W. Norsworthy, F. Bei, C. Wang, S. Li, Y. Zhang, C. V. Gabel, T. L. Schwarz, and Z. He, "The Mammalian-Specific Protein Armcx1 Regulates Mitochondrial Transport during Axon Regeneration," Neuron, vol. 92, no. 6, pp. 1294-1307, 2016.

[35] S. Jauhiainen, J. P. Laakkonen, K. Ketola, P. I. Toivanen, T. Nieminen, T. Ninchoji, A. L. Levonen, M. U. Kaikkonen, and S. Ylä-Herttuala, "Axon Guidance-Related Factor FLRT3 Regulates VEGF-Signaling and Endothelial Cell Function," Front Physiol, vol. 10, pp. 224, 2019.

[36] B. V. Chakravarthi, S. S. Pathi, M. T. Goswami, M. Cieślik, H. Zheng, S. Nallasivam, S. R. Arekapudi, X. Jing, J. Siddiqui, J. Athanikar, S. L. Carskadon, R. J. Lonigro, L. P. Kunju, A. M. Chinnaiyan, N. Palanisamy, and S. Varambally, "The miR-124-prolyl hydroxylase P4HA1-MMP1 axis plays a critical role in prostate cancer progression," Oncotarget, vol. 5, no. 16, pp. 6654-69, 2014.

[37] Y. Lv, Y. Jin, Y. Zhou, J. Jin, Z. Ma, and Z. Ren, "Deep sequencing of transcriptome profiling of GSTM2 knock-down in swine testis cells," Sci Rep, vol. 6, pp. 38254, 2016.

[38] A. Albawardi, J. Livingstone, S. Almarzooqi, N. Palanisamy, K. E. Houlahan, A. A. A. Awwad, R. A. Abdelsalam, P. C. Boutros, and T. A. Bismar, "Copy Number Profiles of Prostate Cancer in Men of Middle Eastern Ancestry," Cancers (Basel), vol. 13, no. 10, 2021.

[39] P. G. Patel, T. Wessel, A. Kawashima, J. B. A. Okello, T. Jamaspishvili, K. P. Guérard, L. Lee, A. Y. Lee, N. E. How, D. Dion, E. Scarlata, C. L. Jackson, S. Boursalie, T. Sack, R. Dunn, M. Moussa, K. Mackie, A. Ellis, E. Marra, J. Chin, K. Siddiqui, K. Hetou, L. A. Pickard, V. Arthur-Hayward, G. Bauman, S. Chevalier, F. Brimo, P. C. Boutros, D. J. Lapointe Ph, J. M. S. Bartlett, R. J. Gooding, and D. M. Berman, "A three-gene DNA methylation biomarker accurately classifies early stage prostate cancer," Prostate, vol. 79, no. 14, pp. 1705-1714, 2019.

[40] J. C. Angulo, J. I. Lopez, J. F. Dorado, M. Sanchez-Chapado, B. Colas, and S. Ropero, "A DNA Hypermethylation Profile Independently Predicts Biochemical Recurrence Following Radical Prostatectomy," Urol Int, vol. 97, no. 1, pp. 16-25, 2016.

[41] N. Ashour, J. C. Angulo, G. Andrés, R. Alelú, A. González-Corpas, M. V. Toledo, J. M. Rodríguez-Barbero, J. I. López, M. Sánchez-Chapado, and S. Ropero, "A DNA hypermethylation profile reveals new potential biomarkers for prostate cancer diagnosis and prognosis," Prostate, vol. 74, no. 12, pp. 1171-82, 2014.

[42] S. Li, L. Wang, S. Sun, and Q. Wu, "Hepsin: a multifunctional transmembrane serine protease in pathobiology," Febs j, vol. 288, no. 18, pp. 5252-5264, 2021.

[43] X. Ma, J. Guo, K. Liu, L. Chen, D. Liu, S. Dong, J. Xia, Q. Long, Y. Yue, P. Zhao, F. Hu, Z. Xiao, X. Pan, K. Xiao, Z. Cheng, Z. Ke, Z. S. Chen, and C. Zou, "Identification of a distinct luminal subgroup diagnosing and stratifying early stage prostate cancer by tissue-based single-cell RNA sequencing," Mol Cancer, vol. 19, no. 1, pp. 147, 2020.

[44] F. Petitprez, N. Fossati, Y. Vano, M. Freschi, E. Becht, R. Lucianò, J. Calderaro, T. Guédet, L. Lacroix, P. M. V. Rancoita, F. Montorsi, W. H. Fridman, C. Sautès-Fridman, A. Briganti, C. Doglioni, and M. Bellone, "PD-L1 Expression and CD8(+) T-cell Infiltrate are Associated with Clinical Progression in Patients with Node-positive Prostate Cancer," Eur Urol Focus, vol. 5, no. 2, pp. 192-196, 2019.

[45] E. Zhang, F. Dai, Y. Mao, W. He, F. Liu, W. Ma, and Y. Qiao, "Differences of the immune cell landscape between normal and tumor tissue in human prostate," Clin Transl Oncol, vol. 22, no. 3, pp. 344-350, 2020.

[46] S. Malekghasemi, J. Majidi, B. Baradaran, and L. Aghebati-Maleki, "Prostate cancer cells modulate the differentiation of THP-1 cells in response to etoposide and TLR agonists treatments," Cell Biol Int, vol. 44, no. 10, pp. 2031-2041, 2020.

[47] M. D. Jelic, A. D. Mandic, S. M. Maricic, and B. U. Srdjenovic, "Oxidative stress and its role in cancer," J Cancer Res Ther, vol. 17, no. 1, pp. 22-28, 2021.

[48] S. Chikara, L. D. Nagaprashantha, J. Singhal, D. Horne, S. Awasthi, and S. S. Singhal, "Oxidative stress and dietary phytochemicals: Role in cancer chemoprevention and treatment," Cancer Lett, vol. 413, pp. 122-134, 2018.

[49] L. Xiong, Y. Feng, W. Hu, J. Tan, S. Li, and H. Wang, "Expression of AOX1 Predicts Prognosis of Clear Cell Renal Cell Carcinoma," Front Genet, vol. 12, pp. 683173, 2021.

[50] J. Wu, Y. Wei, T. Li, L. Lin, Z. Yang, and L. Ye, "DNA Methylation-Mediated Lowly Expressed AOX1 Promotes Cell Migration and Invasion of Prostate Cancer," Urol Int, pp. 1-9, 2022.

[51] K. Yuan, R. Zeng, Peng. Deng, "Identification and validation of immune-related biomarkers based on machine learning in patients with prostate cancer, 21 April 2022, PREPRINT (Version 1) available at Research Square.

Table 1 Information for selected GEO datasets

GEO accession	country	Platform	Samples		Category
GEO accession	country	Platform	PCa	Normal	Category
GSE8511	USA	GPL1708	24	16	Train group
GSE14206	Italy	GPL887	53	14	Train group
GSE46602	Denmark	GPL570	36	14	Train group
GSE55945	USA	GPL570	13	8	Train group
GSE69223	Germany	GPL570	15	15	Test group
GSE71016	USA	GPL16699	48	47	Test group

Download PDF

Version 2

posted

You are reading this latest preprint version

Identification and validation of hub genes based on machine learning in prostate cancer and AOX1 is an oxidative stress-related biomarker

Status:

Version 2

Abstract

Figures

Introduction

Materials And Methods

Results

Discussion

Conclusions

Declarations

References

Tables

Status:

Version 2