2.1 Reagents and cell lines

Normal human astrocytes (HA) were cultivated in an astrocyte medium (Zeye Biotechnology, Shanghai, China). The remaining glioma cell lines are from daily storage in our laboratory and have been regularly tested for STR identification and mycoplasma. These multiple glioma cell lines U-251, T98G, LN-229, A-172, and U-87MG were cultivated according to relevant literature reports.^9-11 JAK2/STAT3 signaling pathway inhibitor WP1066 (Selleck, S2796) was purchased from Selleck. For detailed information on antibodies, please see the Data S1.

2.2 Glioma samples

The normal brain tissues (NBT) were procured from individuals involved in automobile accidents who did not have glioma. Between June 2017 and January 2023, 112 glioma samples were acquired from Wuhan Union Hospital, and participant consent and ethics committee approval (UHCT-IEC-SOP-016-03-01) were required. Aside from collecting and managing clinical samples and data according to established guidelines, no patients underwent adjuvant, neoadjuvant, or radiotherapy before surgery. Tables S1 and S2 present more detailed information.

2.3 Plasmid construction and lentivirus

The pLVX-Puro-RNF122 overexpression plasmid was constructed, followed by the transfection of LN-229 and A-172 cells with the respective expression plasmid. Data S1 lists the specific procedure. Briefly, LN-229 and A-172 cells were transfected following the protocols after their growth to 85% confluence in six-well plates. A puromycin selection was performed on the cells after 48 h to remove uninfected cells. Based on the RNF122 sequences, siRNAs were designed, and the two siRNAs that had the most significant knock down impact were selected. For a more comprehensive examination, refer to the Data S1.

2.4 CCK-8 assay

According to the protocols, cell proliferation assays were carried out with CCK-8 (Cell Counting Kit-8, Dojindo, Tokyo, Japan). The cells were introduced onto 96-well plates using a fresh medium and subsequently exposed to CCK-8 solution. Cell proliferation was determined by measuring the absorption by a microplate reader, with more available details in the previous publication.^11-13

2.5 Western blotting (WB)

The tissues and cells underwent lysis and were quantified using RIPA buffer and bicinchoninic acid (BCA) (Servicebio, Wuhan, China). The lysates underwent separation through the utilization of sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE), subsequently proceeding to be moved onto a polyvinylidene fluoride (PVDF) membrane and further incubated with specific antibodies. Data S1 show detailed information on antibodies.

2.6 Quantitative real-time PCR (qRT-PCR)

Total RNA extraction, cDNA synthesis, and qRT-PCR were performed as mentioned earlier.^{10, 11, 14} The GAPDH was employed as a standard reference gene to determine the relative amounts of mRNA expression, with more available details in the Data S1.

2.7 Colony formation assay

A total of 400 reconstituted cells were cultivated in a full medium for 2 weeks, describing the detailed procedures in the Data S1.

2.8 Cignal finder cancer 10-pathway reporter array

A luciferase reporter was transfected into 96-well plates containing resuspended cells to detect common cancer pathways. A luciferase reaction was performed following the protocols after incubation, with more available details in the Data S1.

2.9 Immunohistochemistry (IHC) and Immunofluorescence (IF)

Our previous publications provide details on immunohistochemistry (IHC) and immunofluorescence (IF).^{15, 16} In brief, we incubated tissues with appropriate primary and secondary antibodies after dehydrating, paraffin embedding, and sectioning them.

2.10 Bioinformatics analysis

The Cancer Genome Atlas Program (TCGA) (https://cancergenome.nih.gov/) was accessed to download transcript data of glioma samples and clinical information while accessing the Genotype-Tissue Expression (GTEx, https://gtexportal.org/home/) to download the transcript data of NBT.¹⁷ Batch effect is a common problem in statistical analysis, especially in biological and medical research. Different experimental batches may cause non-biological variation between data. The purpose of dealing with batch effect is to reduce or eliminate this variation so as to more accurately interpret the experimental results. The following are some common methods used by our team to deal with batch effect. (1) Normalization and standardization: During the data processing stage, normalization or standardization is used to reduce the differences between batches. For example, Z-score standardization can be used to transform the data of each batch to a scale with the same mean and variance. (2) Batch effect correction method: (2.1) Linear regression: Use a linear regression model to remove the effect of batch as a predictor variable from the response variable. (2.2) ComBat method: This is a widely used Bayesian method suitable for adjusting for systematic differences between batches and is often used in gene expression data analysis. (2.3) Empirical Bayes method: Use information from the overall data to estimate and adjust the effects of each batch. (3) Mixed models: Using mixed effects models, batch can be included as a random effect in the model, thus accounting for the variation between batches. (4) Data integration techniques: When integrating multiple datasets, methods such as principal component analysis (PCA) and multidimensional scaling (MDS) are used to visualize and correct for batch effects. Proper handling of batch effects is a key step to ensure accurate and reliable statistical analysis results. In different situations, it may be necessary to combine one or more of the above methods.

2.11 Xenograft model

From Beijing Vital River Animal Technology Co. Ltd. (Beijing, China), 30 nude mice (6–8 weeks, female) were purchased. Tongji Medical College's Animal Experiments Committee approved all animal studies. A humane endpoint was established in our study following AAALAC guidelines. Following the random categorization into distinct cohorts, a cohort of female BALB/c nude mice aged 6–8 weeks underwent anesthesia with three mice in each cohort. Subsequently, a suspension of 7 μL LN-229 and A-172 cells (6 × 10⁵ cells) (with indicated treatment) was administered intracranially in the mouse employing a stereotaxic apparatus. Tumor size estimation was conducted utilizing the formula V = (D × d²)/2, where D and d refer to the longest and shortest diameter, respectively.

2.12 Statistical analysis

First, we use normality examination to ensure that the inspection data follows a normal distribution (Gaussian distribution). The statistical analyses were performed using SPSS 22.0 software (SPSS, Inc., Chicago, IL) and GraphPad Prism (version 8.0; GraphPad Inc., La Jolla, CA, USA), reporting the data as mean ± SD. Moreover, the statistical methods employed for evaluating the significance of data include the utilization of the unpaired or paired Student's t-test for two cohorts and the two-way ANOVA, then using Dunnett's test for involving more than two cohorts. Also, the chi-square test, Pearson's correlation, and one-way analysis of variance were conducted. Cox regression analysis and Log-rank test were utilized to determine survival variation and hazard ratio. p < 0.05 showed statistically significant. If the data do not follow a normal distribution, it may be necessary to perform a data transformation (such as a logarithmic transformation) or use nonparametric statistical methods.

3.1 RNF122 expression was increased in glioma and negatively correlated to prognosis

The candidate genes contributing to tumor progression were identified using high-throughput sequencing (NBT vs. LGG; LGG vs. HGG; NBT: NBT; LGG: low-grade glioma; HGG: high-grade glioma). The top 16 upregulated genes were altered more than fivefold in glioma compared to NBT (Figure 1A). RNF122 was significantly overexpressed among them. WB and qRT-PCR analysis revealed that RNF122 was significantly overexpressed in glioma and increased with tumor grade compared to NBT, especially in HGG (Figure 1B,C). The typical IHC staining micrographs in Figure 1D indicate that RNF122 staining intensity increased significantly with tumor grade in gliomas more than in NBT. Furthermore, the IHC staining score and malignancy grade were positively correlated (Figure 1E). Kaplan–Meier analysis showed a poorer prognosis for patients with RNF122 overexpression (Figure 1F,G). Data analysis from the TCGA tumor database also demonstrated that RNF122 was overexpressed in gliomas and inversely related to patient outcomes (Figure S1A).

Herein, we conducted a receiver operating characteristic curve (ROC) analysis to compare the predictive value of RNF122-based and WHO grade-based, as well as the combination of both, in assessing the pathological and clinical outcomes of RNF122. The results indicate that the combination model exhibited superior predictive ability for clinical outcomes compared to the WHO grade-based model alone (Figure S1B,C). Furthermore, the examined association between RNF122 mRNA levels and clinicopathological characteristics in 112 glioma patients (Table S1) revealed that the RNF122 mRNA expression level had a significant correlation to the tumor size (p < 0.001), KPS (p < 0.001), and WHO stage (p < 0.001). Based on univariate and multivariate Cox regression analyses, RNF122 mRNA expression level was correlated to WHO stage; hence, it was found to be an independent predictor of unfavorable survival outcomes in glioma patients (Table S2), indicating that RNF122 might be a potential glioma biomarker.

3.2 RNF122 overexpression (oeRNF122) promotes glioblastoma cell growth

The first step was to measure RNF122 protein in Normal HA and five glioma cell lines (U-251, T98G, LN-229, A-172, and U-87MG) by WB. According to the results, RNF122 protein levels are greater in glioma cell lines than in HA (Figure S1D). Therefore, we selected two cell lines, LN-229 and A-172, with moderate expression of RNF122 protein for subsequent research. Next, overexpressed RNF122 in LN-229 and A-172 cells were verified using WB (Figure S1E). Based on functional tests, oeRNF122 advanced GBM cell proliferation, migration, and invasion (Figure 2A–C).

Furthermore, in vivo animal experiments again showed that oeRNF122 promoted tumor growth. Ki-67 IF staining was utilized to assess the proliferative tumor index, indicating that proliferation in the oeRNF122 cohorts was more elevated than in the control cohorts (Figure 2D). Besides, the flow cytometric analysis revealed that the progression of the cell cycle of oeRNF122 cells was enhanced in contrast to cells transfected with the control vector (Figure S2A). The above data all suggested that oeRNF122 enhances tumor growth by promoting the cell cycle progression.

3.3 RNF122 knockdown suppresses glioblastoma cell growth

RNF122 knocked down efficiency in LN-229 and A-172 cells, which were validated by WB (Figure S1E). The si-RNF122#2 has the highest knockdown efficiency and will be used as a follow-up study. Functional tests indicated that RNF122 knocking down curbed GBM cell proliferation, migration, and invasion (Figure 3A–C), which was verified further by the in vivo animal experiments. Performing Ki-67 IF staining to assess the proliferative tumor index showed that proliferation in the RNF122 knockdown cohorts was reduced more than in the normal cohorts (Figure 3D). Additionally, it was observed through flow cytometric analysis that at the G1 phase, the cell cycle progression of cells transfected with si-RNF122 was arrested in contrast to cells transfected with si-NC (Figure S2B). These data suggest that RNF122 knockdown suppresses tumor growth by inducing S-phase cell cycle arrest.

3.4 RNF122 activates the JAK2/STAT3/c-Myc signaling

Based on the Cignal finder cancer 10-pathway reporter array used for screening the potentially involved signaling pathways in this process, the JAK/STAT signaling axis was significantly suppressed by RNF122 knocking down in LN-229 and A-172 cells, in contrast to the other signaling axis (Figure S3A). Moreover, GSEA revealed that RNF122 was significantly correlated to JAK/STAT signaling (Figure S3B). Additionally, performing RNF122 knocking down and overexpressing in LN-229 and A-172 cells, respectively, for validating the RNF122 ability to promote tumor growth further through JAK/STAT pathway activation. WB was employed to assess the expression level of JAK1, p-JAK1, JAK2, p-JAK2, STAT1, p-STAT1, STAT2, p-STAT2, STAT3, p-STAT3, and c-Myc. The outcomes indicated that only p-JAK2, p-STAT3, and c-Myc were downregulated by RNF122 knocking down while promoted by RNF122 overexpression (Figure 4A,B). In addition, we employed qRT-PCR to assess the RNF122, JAK2, and STAT3 expression levels in tumor samples and performed correlation analysis. The outcomes indicated a significant positive connection between the RNF122 expression levels and the JAK2 and STAT3 (Figure S3C). Interestingly, our analysis of TCGA data also revealed a positive connection between RNF122 and JAK2, STAT3, and c-Myc (Figure S4A). The above data all show that RNF122 activates the JAK2/STAT3/c-Myc pathway.

3.5 RNF122 influences glioblastoma growth through JAK/STAT signaling activation

By not only treating LN-229 and A-172 cells with JAK/STAT inhibitor WP1066 (6 μM, 48 h) to further determine the RNF122 ability to promote cell proliferation through JAK/STAT activation but also by investigating RNF122 overexpression effect on proliferation, it was revealed that consistent with the earlier conclusion, RNF122-OE induced p-JAK2, p-STAT3, and c-Myc, but not JAK2 and STAT3. Nevertheless, this effect was weakened in treating LN-229 and A-172 cells with JAK/STAT inhibitor WP1066 (Figure 5A). Additionally, depending on the CCK-8, colony formation, Transwell migration/invasion, and in vivo animal experiments, WP1066 led to a significant reverse of the RNF122 effect on enhancing LN-229 and A-172 cell proliferation, migration, and invasion (Figures 5B,C and 6A), which was parallel to the in vitro study which verified that WP1066 treatment significantly decreased tumor growth in mice with GBM intracranial orthotopic transplantation more than in the respective cohorts. The reduced Ki-67 expression in xenograft tumors of mice treated with WP1066 implies that JAK/STAT inhibitor WP1066 could counteract the RNF122 influence on enhancing cell proliferation (Figure 6B). Collectively, RNF122 might enhance the progression of GBM by JAK/STAT signaling activation.

3.6 RNF122 promotes STAT3-mediated transcription of c-Myc

Transcription factors demonstrate a propensity to bind to specific DNA sequences to regulate the expression of genes. By employing the JASPAR database,¹⁸ the observation of STAT3 binding to the promoter of c-Myc was determined. To demonstrate the transcriptional regulation of c-Myc by STAT3 and the potential enhancement of this modulation by RNF122, luciferase vectors containing either the wild-type or mutated c-Myc promoters were generated and subsequently transfected into LN-229/A-172 cells (Figure 7A). The luciferase assay provided evidence that the upregulation of STAT3 led to the stimulation of WT c-Myc promoter activity, as indicated by an increase in luciferase activity. Conversely, the overexpression did not impact mut c-Myc promoter activity. Furthermore, RNF122 augmented the luciferase activity induced by STAT3 (Figure 7B). Moreover, the ChIP assays provided evidence of STAT3's binding to the c-Myc promoter, with RNF122 being identified as a facilitator of this interaction (Figure 7C). The upregulation of STAT3 led to a notable increase in both c-Myc mRNA and protein levels, subsequently facilitated by RNF122 (Figure 7D).

3.7 The involvement of c-Myc in glioblastoma cell proliferation, migration, and invasion is facilitated by RNF122

The c-Myc proteins have significant importance in the gene expression regulation of cell cycle progression, cell growth, and apoptosis, thereby being indispensable for maintaining normal cellular functionality.^{11, 19, 20} Mutations or overexpression of Myc genes has been implicated in the deregulation of cell growth and are known to have a significant function in cancer pathogenesis. The Myc gene is frequently observed to be overexpressed across diverse cancer types.^21-23 In LN-229 and A-172 cells, short hairpin RNAs (siRNAs) targeting c-Myc were utilized to investigate further its role in RNF122-mediated GBM cell proliferation, migration, and invasion. As si-c-Myc#1 with the highest knockdown efficiency, it will be used as a follow-up study (Figure S4B). Our previous data showed that c-Myc expression was suppressed by RNF122 knocking down while promoted by RNF122 overexpression (Figure 4A,B). Besides, to verify whether RNF122 regulates c-Myc expression further through JAK/STAT pathway activation. We overexpressed RNF122 added JAK/STAT signaling pathway inhibitors WP1066 to treat the cells and employed WB to ascertain the c-Myc expression level. The outcomes indicated that RNF122 overexpression might increase the c-Myc expression level, while WP1066 could weaken the ability of overexpression of RNF122 to upregulate c-Myc protein (Figure 5A). Besides, c-Myc silencing attenuated the promoting impacts of RNF122 overexpression on LN-229 and A-172 cell proliferation (Figure 8A,B), migration, and invasion (Figure 8C), which was parallel to the in vitro study, where c-Myc silencing attenuated the promoting influences of RNF122 overexpression on LN-229 and A-172 cell growth in mice with GBM intracranial orthotopic transplantation tumor model more than in the corresponding cohorts. The reduced Ki-67 expression in xenograft tumors of c-Myc silencing cohorts also implied that c-Myc silencing could counteract the RNF122 influence on promoting cell growth (Figure 9). Collectively, RNF122 enhances tumor development through JAK2/STAT3/c-Myc signaling axis activation.