Gene expression profiles of 35 Jumanji domain-containing protein genes at the mRNA level were analyzed using the RNA-seq dataset obtained from the Cancer Genome Atlas Prostate Adenocarcinoma project (TCGA-PRAD). There were 499 patients with primary prostate cancers in this cohort, of which case-matched benign specimens from the prostate glands were obtained from 52 cases. The RNA-seq data in Fragments Per Kilobase per Million (FKPM) format were downloaded from the TCGA portal (https://portal.gdc.cancer.gov) and were then converted to Transcripts Per Million (TPM) reads format. Finally, the data were presented as log₂ [TPM + 1] for analysis. All analyses were conducted using the R-language package (version 3.6.3) on the bioinformatics platform (www.xiantao.love). RIOX2 gene expression profile was also assessed using the MSKCC dataset from 126 prostate cancers (16) and SU2C/PCF dataset from 429 patients (17) on the cBioportal platform (www.cbioportal.org).

The association of RIOX2 gene expression with patient survival outcomes was assessed using the Kaplan-Meier curve approach. Patients were divided into groups with high or low RIOX2 gene expression using the minimum p-value strategy (18). The Log-rank test was utilized to determine the significance of the hazard ratio (HR).

RIOX2 protein expression was assessed using an immunohistochemistry (IHC) approach on tissue microarray slides commercially obtained from Novus Biologicals, LLC (Avenue, CO). The tissue microarray contained 49 tissue sections derived from prostate cancer patients, of which there were nine case-matched pairs of normal and tumor sections. There were two cases of a Gleason score of 6 or 10, 15 cases of a Gleason score of 7 or 9, and 6 cases of a Gleason score of 8. The anti-RIOX2 polyclonal antibody was obtained from Sigma-Aldrich (St Louis, MO) and validated by the Human Protein Atlas project (HPA008080). The VECTASTAIN Elite kit (catalog #PK8200) from Vector Labs (Burlingame, CA) was used to visualize the immunosignal. A semi-quantitative approach was utilized to analyze the positive immunosignals, as described in our previous publication (19).

RIOX2 expression in LNCaP cells after androgen stimulation or deprivation was analyzed using the NCBI GEO datasets (GDS3358, GDS2728, and GDS3111). For androgen deprivation, LNCaP cells were cultured in 10% charcoal-stripped fetal bovine serum (cFBS) (testosterone < 0.03 ng/ml) for up to 48 weeks. RIOX2 gene expression was assessed using the Affymetrix Human Genome U133 Plus 2.0 arrays (Affymetrix, Santa Clara, CA), as described in a previous publication (20). For androgen stimulation, LNCaP cells were cultured in 8% cFBS for three days and then treated with dihydrotestosterone (DHT, 10 nM) with or without the AR antagonist Casodex (10 nM). RIOX2 gene expression was examined using the Affymetrix Human Genome U133AB platform (21, 22).

To evaluate the androgen effect on RIOX2 expression at the protein level, LNCaP cells were treated with synthetic androgen R1881 in 2% cFBS with or without anti-AR drugs Enzalutamide and Abiraterone. R1881 was obtained from ICN (Aurora, OH) and dissolved in absolute Ethanol. Enzalutamide and Abiraterone were purchased from Cayman Chemicals (Ann Arbor, MI) and dissolved in dimethyl sulfoxide. Total cellular proteins were subjected to SDS-PAGE electrophoresis, followed by anti-RIOX2 immunoblotting assay (HPA008080, Sigma-Aldrich Co., St Louis, MO), as described in our publication (14).

RNA-seq data were presented as a Log₂ [TPM + 1]) value. Data in each group were also shown with the MEAN plus/minus the SEM (standard error of the mean). ANOVA test was used for multiple group comparisons, and a Student t-test was conducted to determine the significance of the differences between the two groups. Data visualization was performed using the R packages (version 3.6.3) and GraphPad software (version 9.1.0).

We analyzed the expression profiles of 35 Jumanji domain-containing protein genes in prostate cancers, and the list of genes was derived from two review articles (1, 2). We conducted two types of comparisons, case-matched pair of 52 benign-malignant tissues and a group cohort (benign vs. malignant) using the TCGA-PRAD dataset. Our results showed that nine genes were significantly upregulated, and three genes were significantly downregulated in both types of comparisons ( Figures 1A, B ). In contrast, the other 23 genes showed no significant or inconsistent differences between the two comparisons ( Figures S1A-D ). ROC analysis revealed that except for KDM5C (AUC = 0.597), all other genes exhibited a potent AUC value (> 0.6) in distinguishing malignant tissue from the benign compartment ( Figure 1C ). Specifically, the gene HR, a lysine demethylase and nuclear receptor corepressor, showed an AUC value of 0.814, representing a potential diagnostic marker for prostate cancer.

We then analyzed these altered genes’ association with disease-specific survival outcomes in prostate cancer patients. With five clinicopathological factors, COX regression analysis revealed that only RIOX2 upregulation and patient serum PSA level was significantly associated with disease-specific survival in the univariate test ( Table S1 ). After eliminating other genes, RIOX2 upregulation showed a significant association with disease-specific survival in univariate and multivariate tests ( Table 1 ). These data suggest that RIOX2 is an independent prognostic factor in prostate cancer, like serum PSA level.

To verify RIOX2 upregulation at the protein level, we conducted an immunohistochemical assay using a prostate cancer tissue array containing 40 cases. Of these tissue sections, there were nine case-matched benign-malignant tissue pairs. As shown in Figure 2 , RIOX2 protein immunosignals were significantly increased in malignant tissues compared to their benign compartments. Semi-quantitative analysis of the immunosignals revealed a drastic difference in pairwise ( Figure 2E ) and group cohort comparisons ( Figure 2F ). These data indicated that RIOX2 protein expression was also upregulated in prostate cancers.

Genomic alterations in cancer cells include mutations, copy number variations (CNV), and gene fusions. These alterations indicate tumor heterogeneity and homogeneity and are important factors associated with gene expression and cancer progression. We then examined the genetic alterations of the RIOX2 gene using the TCGA-PRAD RNA-seq dataset. Among 491 cases, only three (0.6%) showed genetic deep deletion or amplification. There were 10 cases with shallow deletion (2%) and 45 with genetic gain (9.16%). Two patients showed point mutations; one was A344V, and the other had two simultaneous point mutations, R197H & E447D. Consistently, RIOX2 mRNA levels were significantly increased along with the genetic gain and amplification ( Figure 3A ). Similar results were also observed from the MSKCC dataset (16); RIOX2 mRNA levels were significantly higher in prostate cancers with genetic gain ( Figure 3B ). In addition, RIOX2 mRNA levels were strongly correlated with the fraction value of genomic alterations in prostate cancers. Patients with ETS fusion, ERG-ACGH fusion, and ERG-GEX fusion all showed an increased level of RIOX2 mRNA ( Figures 3C–F ).

Since the low incidence (< 10%) of RIOX2 genetic gain/amplification was unlikely a driving cause of RIOX2 upregulation in prostate cancers, we turned to epigenetic alterations. We examined the promoter methylation levels within the RIOX2 gene locus. Our results showed that RIOX2 promoter methylation was significantly lower in cancer tissues compared to benign tissues ( Figure 3G ). A very strong reverse correlation was noticed between RIOX2 promoter methylation and mRNA levels in cancer tissues ( Figure 3H ). Further analysis revealed that the methylation intensity around the -1127bp region upstream of the transcription start site (TSS) was significantly correlated with RIOX2 mRNA levels ( Figure 3I ), indicating a potential mechanism of epigenetic modification.

We analyzed the association of RIOX2 expression with patient clinicopathological features. Our results showed that higher levels of RIOX2 expression were seen in patients with late-stage (T4 vs. T2/3, Figure 4A ), higher Gleason scores (8-10 vs. 6-7, Figure 4B ), and tumors with TP53 mutation ( Figure 4C ). Deceased patients due to prostate cancer also showed a significantly higher level of RIOX2 mRNA levels compared to alive patients ( Figure 4D ). Kaplan-Myer curve analysis also showed that RIOX2 upregulation was significantly associated with disease-specific survival outcomes ( Figure 4E ). However, the association of RIOX2 expression with progression-free interval was at the borderline of significance (Log-rank p = 0.054, Figure 4F ). However, after stratification of patients into subgroups, RIOX2 expression was significantly associated with the progression-free interval in patients with PSA level below 4 ng/ml ( Figure 4G ), with residue tumors after surgery ( Figure 4H ), without lymph node invasion ( Figure 4I ), or with higher Gleason scores ( Figure 4J ). ROC analysis indicated that RIOX2 expression is a robust prognostic factor (AUC = 0.88 - 0.971) of disease-specific survival for 5-10 years ( Figure 4K ).

The clinical obstacle in managing metastatic prostate cancers is the castration-resistant progression after androgen deprivation therapy. Due to the vast implication of androgen receptor antagonists for castration-resistant prostate cancers (CRPC), neuroendocrinal progression of prostate cancer (NEPC) emerged as the worst stage of prostate cancer without means to cure (23, 24). We analyzed RIOX2 expression profiles in CRPC and NEPC specimens using the SU2C/PCF dataset (17). Our results showed that RIOX2 expression was significantly correlated with AR score ( Figure 5A ), an AR signal activity index (17). Meanwhile, RIOX2 expression was also strongly correlated with the expression levels of the AR-V7 splicing variant ( Figure 5B ), a potent indicator of treatment resistance in CRPC patients (25). In contrast, a robust reverse correlation was observed between RIOX2 expression and NEPC score ( Figure 5C ), an index of NEPC progression (17). Consistently, RIOX2 mRNA levels were significantly lower in NEPC tumors than in CRPC tumors ( Figure 5D ).

To determine if RIOX2 expression was directly regulated by the androgen/AR signal pathway, we analyzed RIOX2 expression in prostate cancer LNCaP cells after androgen stimulation. As shown in Figures 5E–G , androgen stimulation up to 16-24 h had no significantly enhancing effect on RIOX2 expression at the mRNA ( Figures 5E, F ) and protein levels ( Figure 5G ) in LNCaP cells. The AR antagonists, Casodex (Bicalutamide), Enzalutamide, and Abiraterone, did not significantly impact RIOX2 expression. Consistently, RIOX2 mRNA expression was observed in all prostate cancer cell lines with or without AR expression ( Supplemental Figure S2 ). These data indicated that RIOX2 expression might not be a direct target gene of the AR signal pathway.

RIOX2 gene was initially reported as a c-Myc downstream target involved in cell proliferation (26), and c-Myc is a proven oncogene in prostate cancer (27). Therefore, we examined the correlation of RIOX2 expression with the MYC family and its related genes. Consistent with a previous report (26), a strong positive correlation was found between RIOX2 expression and c-Myc ( Figure 6A ), L-Myc ( Figure 6B ), and Myc binding proteins ( Figures 6C, D ), while only a moderate correlation was noticed between RIOX2 expression and Myc associated factor-X ( Figure 6E ). In contrast, RIOX2 expression was not correlated with N-Myc ( Figure 6F ) and MYCBP-associated protein ( Figure 6G ). Interestingly, among these Myc family members, only c-Myc was significantly upregulated in prostate cancers compared to benign compartments ( Figures 6J, K ), suggesting that c-Myc is potentially a major regulator of RIOX2 expression in prostate cancer.

Recent reports showed that RIOX2 expression in human cancers was also modulated by zinc finger proteins ZNF143 (28) and IKZF1 (29) in liver cancers. We conducted a Spearman correlation analysis and found that RIOX2 expression was strongly correlated with ZNF143 expression ( Figure 6H ) while moderately with IKZF1 expression ( Figure 6I ) in prostate cancers. However, both ZNF143 and IKZF1 expression had no significant differences between malignant and benign tissues in prostate cancers ( Figures 6J, K ). These data suggest that ZNF143 or IKZF1 is unlikely an upstream modulator for RIOX2 upregulation in prostate cancer.