This study was approved by the Ethics Committee of the First Affiliated Hospital of Harbin Medical University (No. HMUIRB-2008-06), and the patients’ consent was obtained. All glioma tissue specimens were taken from surgically resected tissues of patients with glioma (n = 25), and non-tumor brain tissue (From patients undergoing epilepsy surgery) was used as a negative control group (n = 5). Tissue samples were stored in liquid nitrogen.

After screening out samples with defects in clinical information, we collected 546 glioma samples and five normal brain samples from The Cancer Genome Atlas (TCGA) (https://www.cancer.gov) and 749 glioma samples and 20 normal brain samples from the Chinese Glioma Genome Atlas (CGGA) (http://www.cgga.org.cn). The samples’ clinical information is shown in Supplementary Table S1, which mainly included glioma grade, 1p19q deletion status, IDH mutation status, age, and survival information.

The differences in TXNDC12 expression between normal brain tissue and gliomas were examined by Wilcoxon test. The relationship between TXNDC12 expression and the clinical characteristics of glioma patients was analyzed using the Kruskal test or Wilcoxon test. Survival and ROC curves were plotted using the Survival and ROC Curves package. The scoring of sample immune gene sets was completed using GSVA, Limma, and GSEABase software packages, and an immune heatmap was created using heatmap software. The CIBERSORT R script v1.03 calculated the infiltration of immune cells in glioma samples. The median TXNDC12 expression in samples was used as a threshold to divide samples into high and low expression groups, and the Vioplot software package was used to depict the differences in immune cell infiltration between the two groups.

According to the manufacturer’s instructions, the total RNA was extracted by TRIzol reagent (Carlsbad Invitrogen, California, United States). The total RNA concentration was detected using a NanoDrop2000 (Thermo Science^TM). RNA (1,000 ng) was reverse transcribed into cDNA using a qPCR RT Kit (TOYOBO, Japan). FastStart Universal 96 SYBR Green Master (ROX) (Roche, Germany) was used to detect the relative expression of TXNDC12 and housekeeping gene GAPDH by qRT-PCR. The primer sequences used in our procedure were as follows: TXNDC12, 5-GTC​CTG​CTG​ATT​GTG​AAA​ATG​GC-3 and 5-TGATCCATGTCGAG GGTCAAA-3; GAPDH, 5-AAT​CCC​ATC​ACC​ATC​TTC-3 and 5-AGGC TGTTGTCATACTTC-3′. A non-paired t-test was performed between the two groups, and p < 0.05 was considered statistically significant.

We selected 3 normal brain tissues, 3 WHO III gliomas, and 3 WHO IV glioma specimens from the pre-embedded wax blocks, and sliced them into 5 μm thick sections. The section tissue was deparaffinized with xylene, 3% hydrogen peroxide blocked the endogenous peroxidase activity, fetal bovine serum blocked at room temperature, and the primary antibody (4°C, overnight) and secondary antibody (2 h, room temperature) were incubated. Next, samples were visualized by using the diaminobenzidine (DAB) substrate kit. The primary antibody was TXNDC12 (A14403, ABclonal), and the secondary antibody was Goat Anti-Rabbit IgG (S0001, Affinity). Brown staining is a positive feature in cells.

The differential genes that were co-expressed with TXNDC12 in samples (cor > 0.5) were screened by cor.test function in the TCGA and CGGA databases, and the intersection was taken using a Venn diagram. The function of differential genes was analyzed using the R package (clusterProfiler, org.Hs.eg.db, enrichplot, ggplot2).

R × 64 3.5.2, strawberry-perl-5.30.2.1, and GraphPad Prism7 were used for plotting and statistics. p < 0.05 was considered statistically significant.

In samples from TCGA, TXNDC12 expression was significantly higher in glioma tissues compared to normal brain tissues (Figure 1A, p < 0.01). We further analyzed the relationship between TXNDC12 expression and glioma pathology grade. We found that TXNDC12 expression increased significantly as glioma grade increased, both in TCGA (Figure 1C, p < 0.01) and CGGA (Figure 1D, p < 0.01). Survival curve analysis showed that the prognosis of the TXNDC12 high expression group was poorer than that of the low expression group in both TCGA (Figure 1F, p < 0.01) or CGGA (Figure 1G, p < 0.01) (the cutoff value is the median). All the above results were verified by PCR of clinical glioma samples and the follow-up results of patients in our hospital (Figures 1B,E,H), p < 0.05).

We analyzed the relationship between TXNDC12 and the prognosis of patients with glioma by establishing a Cox regression model. Univariate analysis revealed that TXNDC12 is a significant risk factor for glioma in TCGA-RNA-seq (HR = 4.182, 95% CI = 3.185–5.490, p < 0.001) (Figure 2A) and CGGA-RNA-seq (HR = 2.186, 95%CI = 1.943–2.460, p < 0.001) (Figure 2D). We found that TXNDC12 is an independent risk factor for glioma in CGGA-RNA-seq by multivariate analysis (HR = 1.371, 95%CI = 1.188–1.584, p < 0.001) (Figure 2E). However, the result was not statistically significant in TCGA-RNA-seq (HR = 1.359, 95% CI = 0.915–2.018, p = 0.128) (Figure 2B). Furthermore, we found that TXNDC12 was an prognostic factor for 3-year and 5-year survival in glioma patients both in TCGA-RNA-seq (AUC_3year = 0.787, AUC_5year = 0.755) (Figure 2C) and CGGA-RNA-seq (AUC_3year = 0.748, AUC_5year = 0.774) (Figure 2F) by ROC curves.

To test the validity and authenticity of the analysis results from TCGA and CGGA databases applicable to actual clinical patients, we collected follow-up data of glioma patients in our hospital (Table 1) for validation. The results showed that expression of TXNDC12 (HR = 1.604, 95%CI = 1.178–2.182, p < 0.005) and WHO grade of glioma (HR = 2.059, 95% CI = 1.158–3.662, p < 0.05) were risk factors for glioma patients. However, patient age, gender, KPS (Karnofsky Performance Scale) score, and tumor location were not statistically significant.

1p19q has important prognostic implications for glioma patients as a molecular marker for pathological classification. We have demonstrated that high expression of TXNDC12 in glioma predicts adverse outcomes in patients. To explore whether there is any relationship between TXNDC12 and 1p19q status, we found that the expression of TXNDC12 was significantly lower in the 1p19q Codel (1p19q co-deletion) group compared to the 1p19q Non-codel (1p19q non-codeletion) group, both in TCGA (Figure 3A, p < 0.005) and CGGA (Figure 3E, p < 0.005). The expression of TXNDC12 in the 1p19q codel subgroup increased with glioma WHO grade (Figure 3F, p < 0.005) in CGGA. However, this difference was not statistically significant in TCGA (Figure 3B, p = 0.073). In the 1p19q Non-codel subgroup, TXNDC12 expression significantly increased with the rise of the grade of gliomas in both TCGA (Figure 3C, p < 0.005) and CGGA (Figure 3G, p < 0.005). In the 1p19q Non-codel subtype, TCGA (Figure 3D, p < 0.005) and CGGA (Figure 3H, p < 0.005), patients in the TXNDC12 high expression group had a worse prognosis than those in the TXNDC12 low expression group. In the 1p19q Codeletion subtype, CGGA (Figure 3H, p < 0.005),patients in the TXNDC12 high expression group had a worse prognosis than those in the TXNDC12 low expression group, while this change was not statistically significant in the TCGA database.

Similar to 1p19q, IDH is another glioma typing biomarker widely used clinically. We found that the expression of TXNDC12 in the IDH Wildtype group was significantly higher than that in the IDH Mutant group in TCGA (Figure 4A, p < 0.005) and CGGA (Figure 4E, p < 0.005). In the IDH Mutant subgroups, TCGA (Figure 4B, p < 0.005) and CGGA (Figure 4F, p < 0.005), the expression of TXNDC12 increased significantly with the rise of glioma grade, and the same results were observed in the IDH Wildtype subgroup (Figure 4C, p < 0.005 and Figure 4G, p < 0.005). In the IDH Wildtype subtype, TCGA (Figure 4D, p < 0.005) and CGGA (Figure 4H, p < 0.005), patients in the TXNDC12 high expression group had a worse prognosis than those in the TXNDC12 low expression group. In the IDH Mutant subtype, CGGA (Figure 4H, p < 0.005), patients in the TXNDC12 high expression group had a worse prognosis than those in the TXNDC12 low expression group. However, in the TCGA, the prognosis of MT-high (IDH Mutant combined with TXNDC12 high expression) is better than that of MT-low (IDH Mutant combined with TXNDC12 low expression).

In previous studies, we have proved the significance of TXNDC12 alone, TXNDC12 combined with IDH, and TXNDC12 combined with 1p19q on glioma patient prognosis. However, we know that these three molecular changes may occur in the same patient in actual clinical cases. Reliant on any single molecule to analyze glioma patients’ pathological condition and prognosis may be relatively weak and confusing. Therefore, we combined IDH, 1p19q, and TXNDC12 to explore their combined effects on patient prognosis with glioma. In Figure 5A (TCGA, p < 0.005) and Figure 5E (CGGA, p < 0.005), we found that TXNDC12 had the lowest expression in the MT-Codel (IDH Mutation combining 1p19q Codeletion) group, while the MT-Non-codel (IDH Mutant combining 1p19q Non-codeletion) was moderate, and the WT-Non-codel (IDH Wildtype combining 1p19q Non-codeletion) had the highest expression. Survival analysis showed that the MT-Codel subgroup had the longest survival time, followed by the MT-Non-codel group. The WT-Non-codel group had the shortest survival time (Figure 5B, p < 0.005; and Figure 5F, CGGA, p < 0.005). To further investigate the effect of the difference in TXNDC12 expression among the three subtypes on glioma prognosis, we divided the samples of each subtype into TXNDC12 high-expression and low-expression groups according to the median TXNDC12 expression in the samples. That is, the above three groups were divided into six subgroups (Figure 5C, TCGA; and Figure 5G, CGGA), and survival curves were drawn for Figure 5D (TCGA) and Figure 5H (CGGA). The results showed that TXNDC12 high expression group had shorter survival within the same subtype compared to the low expression group in CGGA (Figure 5D). The WT-Non-codel subtype also showed the same results in TCGA, but the difference in survival curves was not statistically significant between MT-codel and MT-Non-codel subtypes (Figure 5H).

We know that there is an immunosuppressive phenomenon in glioma lesions, which affects the efficacy of many therapeutic measures against glioma. To investigate the relationship between TXNDC12 and the local immune microenvironment glioma, we analyzed data from TCGA (Figure 6A) and CGGA (Figure 6B) databases. The results showed that the clinical features of gliomas in the TXNDC12 high expression group were mainly associated with IDH wild-type, 1p19q non-codeletion, age >41 years, higher WHO grade, reduced tumor cell purity, increased immune cells, and extracellular matrix in the tumor tissue. Further analysis of the infiltrating immune cell components revealed that infiltration of immune cells in glioma tissue is inconsistent across populations and mast cells decreased in the TXNDC12 high expression group (Figures 6C,D).

We found that the combination of IDH and 1p19q status has an absolute predictive value for the prognosis of glioma patients. The expression of TXNDC12 is significantly increased at the mRNA level in gliomas; however, it is unclear whether the expression of the protein encoded by the final target TXNDC12 gene also has corresponding changes. We took normal brain tissue, WHO III, and WHO IV glioma tissues for HE staining (Figures 7A–C) and performed TXNDC12 immunohistochemical analysis (Figures 7D–F,). HE staining showed that WHO III and WHO IV gliomas were dense, with large, and dark stained nuclei and irregular shape (Figures 7B,C). Immunohistochemistry showed an increase in TXNDC12-positive cells in WHO III and WHO IV glioma tissues compared to normal brain tissues (Figure 7E,F).

Co-expression analysis yielded 1301 TXNDC12-related genes in TCGA and 4,374 in CGGA and intersected these genes to obtain 398 genes, as shown in Figure 8A. The GO enrichment analysis of the biological processes (Figure 8B), cellular components (Figure 8C), and molecular functions (Figure 8D) involved in these genes revealed that the functions of these genes were mainly focused on RNA synthesis, antigen presentation and processing, neutrophil-mediated immune response, and glycosylation. The KEGG analysis (Figures 8E,F) shows that major pathways are enriched in glycosyl transfer reaction and DNA transcription.