PABPC1 relevant bioinformatic profiling and prognostic value in gliomas

Gliomas represent the most prevalent and fatal brain tumor in adults [1–3]. According to the updated 2016 WHO classification system of CNS, with histology and molecular parameters, gliomas could be categorized as followed: lower grade gliomas (LGG) with isocitrate dehydrogenase (IDH) mutation with or without 1p/19q-codel, LGG with IDH-wildtype subtype, glioblastoma multiforme (GBM) with IDH mutation or not [4]. In order to improve diagnostic accuracy and treatment efficacy for the glioma patients, it is crucial to identify novel biomarkers and therapeutic targets.

RNA-binding proteins are versatile and can interact with many molecules. The poly adenosine tail of RNA can be bound by RNA poly(A) binding proteins (PABPs) specifically. PABPs are essential for initiation of translation and stabilization of mRNA [5–7]. In humans, PABPs include at least three functional proteins: PABP1, induced PABP (iPABP) and PABP3. Among them, PABP1 protein is encoded by the PABPC1 gene [8]. In many cancers, overexpression of PABPC1 promotes tumor progression and metastasis, such as esophageal cancer [9], gastric carcinoma [10], follicular thyroid cancer [11], bladder cancer [12] and so on. However, little is known about the genetic and epigenetic features of PABPC1 in gliomas.

To explore the clinical and molecular characteristics of PABPC1 in gliomas, we utilized the gene expression data from Chinese Glioma Genome Atlas (CGGA) dataset as training set and validated in The Cancer Genome Atlas (TCGA) dataset. We detected PABPC1 expression was downregulated in gliomas with higher malignance and high expression of PABPC1 in proneural subtype and showed a beneficial effect in patients with gliomas. This integrative study of PABPC1 might provide a potential therapeutic target for glioma patients in future.

Materials & methods

Samples collected in this study

In CGGA dataset (http://www.cgga.org.cn), we collected 301 whole grade glioma samples with mRNA microarray data as training set. In TCGA dataset, we downloaded mRNA sequencing data of 699 glioma samples online (https://cancergenome.nih.gov/) for validation. Thus, we obtained totally 1000 samples with mRNA expression data. Meanwhile, clinical follow-up and molecular information of these patients were collated.

IDH1/2 mutations

It was common to employ pyrosequencing technique to detect IDH mutation in CGGA [13], and IDH mutant status of TCGA patients downloaded online were mainly tested by pyrosequencing technique or whole exon sequencing technique.

TCGA subtype annotation

As for the TCGA subtype annotation, patients were divided into four subtypes (proneural–neural–classical–mesenchymal) using the published signatures and single-sample gene set enrichment analysis algorithm [14].

Immunohistochemical

Formalin-fixed, paraffin-embedded glioma specimens were cut (5 μm section), deparaffinized, rehydrated before antigen repair in citrate buffer. After blocking endogenous peroxidase activity with ethanol containing 3% hydrogen peroxidase, we incubated sections in rabbit anti-PABP1 antibody (number 4992, dilution:1:50, Cell Signaling Technology) over night at 4°C. Diaminobenzidine was used as chromogens and slides were counterstained with hematoxylin before mounting. Each stained slide was scored by two independent neuropathologists. All experiments were performed with negative and positive control to ensure the quality of staining. Staining was scored using a four-point scale from 0–3: 0 = no staining, 1 < 10% neoplastic cells positive staining, 2 = 10–30% neoplastic cells positive staining, 3 > 30% neoplastic cells positive staining. We selected specimens from 6 glioma patients (3 of LGG, 3 of GBM) for immunohistochemical analysis. Five fields at 400 × magnification were analyzed per specimen. The PABPC1 expression difference between two groups was tested by χ² test.

Gene ontology analysis

We filtered PABPC1 significantly correlated genes (Pearson |R| >0.5; p < 0.05) by the method of Pearson correlation analysis in TCGA dataset. Then PABPC1 positively associated genes were chosen to perform gene ontology (GO) analysis in DAVID (https://david.ncifcrf.gov/) [15].

Statistical analysis

R language (version 3.5.1, http://www.r-project.org) was the main data analysis and mapping tool. Student’s t-test was performed for PABPC1 expression difference. For the comparison between categorical variables, we applied a two-tailed χ² test. The time interval from diagnosis to death or the last follow-up was defined as overall survival (OS) time. Based on median PABPC1 expression, we divided patients into two groups, high- and low-PABPC1 groups. Kaplan–Meier estimate was used in survival analysis with two-sided log-rank test. There was significant difference when p-value ≤ 0.05.

Results

Patients

We selected 301 glioma patients of whole grade from CGGA as discovery cohort and 699 glioma patients of whole grades from TCGA for validation in our study (Table 1). Median age of patients in CGGA was 42 (from 13 to 70) and 47 (from 14 to 89) in TCGA. Male accounted for 60% in CGGA and 52% in TCGA. CGGA contained 122 grade II, 51 WHO grade III and 128 grade IV gliomas (GBMs). TCGA dataset included 223 grade II, 245 grade III and 168 grade IV gliomas. Molecular subtype of two datasets were shown in the table as well. We also collected data on genetic alteration. In CGGA, 135 (45%) tumors were detected with IDH mutation and 43 (14%) tumors harbored 1p/19q codeletion. In TCGA, 441(63%) tumors exhibited IDH mutation and 172 (25%) showed 1p19q codeletion. Meanwhile, MGMT promoter methylation status of tumors was also available in two datasets. According to the updated 2016 WHO classification system of CNS [4], glioma patients in two databases were finally classified based on histologic and molecular data (Supplementary Table 1).

PABPC1 expression was downregulated in gliomas with higher malignance

In order to investigate the distribution characteristic of PABPC1 expression, we compared the expression of PABPC1 in gliomas of different grades, histopathology and IDH mutation status. In CGGA dataset, PABPC1 showed the lowest expression in glioblastoma (GBM; Figure 1A). The result was validated in TCGA (Figure 1C). Next, we divided patients into IDH-mutant groups and IDH-wildtype and found that PABPC1 was poorly expressed in IDH-wildtype groups compared with IDH-mutant groups across different grades (Figure 1B & D). We also divided patients into groups according to 2016 WHO classification and found that PABPC1 was poorly expressed in malignant molecular phenotypes (IDH-wildtype or 1p/19q-intact, Supplementary Figure 1A& B). To further clarify the relationship between PABPC1 and glioma grade at the protein expression level, we selected specimens from six glioma patients (3 of LGG, 3 of GBM, 5 fields at 400 × magnification were analyzed per specimen) for immunohistochemical analysis. Two representative stained sections were showed in Supplementary Figure 1C&D. Consistent with the results above, the expression of PABPC1 was lower in GBM (p = 0.0028l χ² test). Thus, both in RNA and protein expression level, PABPC1 expression was downregulated in gliomas with higher malignance.

PABPC1 was a potential marker for proneural molecular subtype

To depict the correlation between PABPC1 and molecular subtype, we further explored PABPC1 expression in different TCGA molecular subtype. In Figure 2A & B, our result showed that PABPC1 was highly expressed in proneural subtype in both datasets. We were inspired that PABPC1 might be a potential biomarker for proneural subtype. Hence, we drew the receiver operating characteristic curve for PABPC1 expression and proneural subtype. In CGGA, the area under the receiver operating characteristic curve was 0.786, and at the optimal cutoff point (0.268), the sensitivity and specificity were 65.1 and 82.8%, respectively (Figure 2C). In TCGA, the area under the receiver operating characteristic curve was 0.875, and at the optimal cutoff point (13.281), the sensitivity and specificity were 84.4 and 77.1%, respectively (Figure 2D). It revealed that PABPC1 was a potential marker for proneural subtype.

PABPC1 showed significant prognostic power

To explore the prognostic value of PABPC1, we performed the Kaplan–Meier survival analysis. The patients were divided into two groups, high and low PAPBC1 groups, based on the median expression level. It showed that glioma patients of high PAPBC1 expression group had a significantly longer OS than PABPC1 low expression group in training dataset (Figure 3A; p = 0.0022). Next, we explored the prognostic value of PABPC1 in different grade gliomas and found that the OS difference was significant in WHO grade III (p = 0.029; Figure 3C) and GBM (p = 0.045; Figure 3D), except for grade II (p = 0.43; Figure 3B).

The 2016 WHO guideline classified gliomas into five subtypes based on histopathology and IDH status and 1p19q codeletion [4]. Since five glioma subtypes had different characteristics and prognosis [16], we further evaluated the prognostic value of PABPC1 in five subgroups of patients. For LGG, survival time of high PABPC1 group was obviously longer than that of low PABPC1 group in LGG-IDHmut-codel (p = 0.041; Figure 3E), LGG-IDHmut-noncodel (p = 0.028; Figure 3F) and LGG-IDH-wild (p = 0.041; Figure 3G). There were similar differences in prognosis between two groups for GBM-IDHwt (p = 0.046; Figure 3H) and GBM-IDHmut (p = 0.046; Figure 3I).

We also performed the survival analysis in TCGA dataset and obtained similar results though no statistical significance was detected in LGG-IDH-wild and GBM-IDHmut (Supplementary Figure 2). Therefore, PABPC1 was a positively prognosis predictor in gliomas, and higher expression of the PABPC1 indicated excellent survival.

Functional annotation of PABPC1

To better understand the underlying biological function of PABPC1, we filtered genes that were significantly correlated with PABPC1 in TCGA dataset (correlation |R| >0.5; p < 0.05) utilizing Pearson correlation analysis. As a result, 187 positively correlated genes and 116 negatively correlated genes met the criteria, respectively (Supplementary Table 2). We chose the positive genes (187 genes) for GO Analysis and detected enrichment in biological function of translation (False Discovery Rate [FDR] = 1.48E-23, Benjamini = 6.45E-24), cytoplasmic translation (FDR = 9.91E-05, Benjamini = 2.16E-05), formation of translation preinitiation complex (FDR = 0.001539, Benjamini = 2.24E-04) (Figure 4A & B & Supplementary Figure 3), which was consistent with previous research [17–20].

Discussion

Gliomas are associated with high disability and mortality rate, which seriously endanger human health and quality of life. Standards surgical resection with radiochemotherapy fail to improve survival of glioma patients substantially [2,21]. Hence, new diagnostic and treatment strategies are urgently needed. New biomarkers are currently being developed for identifying cancer risk, determining prognosis, detecting recurrence and predicting responses to specific drugs [22,23]. It is in urgent need to identify more biomarkers and establish new molecular classification systems for further precision medicine.

In glioma patients, many prognosis biomarkers have been reported, such as IDH mutation, MGMTp methylation, 1p/19q codeletion, EGFR and so on [24]. And PD-L1 [25], TIM-3 [26], DCTD [27], pHH3 [28] and FGFR3 [29] have been demonstrated as biomarker in our previous study.

We have noticed that control of mRNA stability and translation plays a crucial role in various biology processes, such as cell growth and neuronal plasticity [30,31]. mRNA-binding proteins (mRBPs) play a key role in maintaining mRNA stability and controlling transport and translation [32]. Hundreds of such mRBPs have been discovered and studied over the years [33]. Previous studies both in vivo and in vitro have confirmed that PABPC1 control cell proliferation [34–37], promotes cell migration and spreading [38], which is closely related to the malignant phenotype of cancer.

At present, there are few reports on the characteristics and significance of PABPC1 in gliomas. We enrolled totally 1000 glioma patients with transcriptome data from CGGA and TCGA dataset in this study. We found that PABPC1 was upregulated in beneficial pathological type such as low grade, IDH1/2 mutation and 1p/19q co-deletion. Moreover, PABPC1 was preferentially expressed in proneural subtype. These results indicated that PABPC1 expression was related to beneficial biological process. These beneficial biological processes might contribute to tumor stability and therapy sensitivity.

GO analysis of positively related genes gave us a real insight into the biological functions of PABPC1. We found that PABPC1 was significantly related to translation. Translation refers to the cellular metabolic process of protein formation, using mRNA to form specific polypeptide chain. Changes in translation can seriously affect gene expression and dysregulation of translation have been documented in several disorders, including cancer [39,40]. Meanwhile, the relationship between translation and many well-established cancer genes (e.g., PIK3CA, MYC, PTEN) has been deeply revealed [39,41]. For example, MYC hyperactivation increased translation of m7GpppN cap (eIF4F)-dependent mRNAs, resulting in chromosomal instability and cancer development [42]. Therefore, the role of translation function in cancer cannot be underestimated and the mechanism of PABPC1-regulating translation needs deeper investigation in gliomas.

For prognosis analysis, survival curve showed that high-PABPC1 expression predicted significantly better OS in glioma either in whole grade gliomas, grade II, grade III or GBM. These results indicated that PABPC1 was a beneficial molecular prognosis marker with potential important therapeutic implications in gliomas. Due to the important role of PABPC1 in translation and gene stabilization, the mechanism of the downregulation of this gene in glioma patients with poor prognosis needs to be clarified.

Nowadays, traditional cytotoxic agents no longer meet the requirements of modern precision medicine in glioma and an amount of selective, mechanism-based agents come forth. Our understanding of the gliomagenesis has improved, resulting in new treatment options like targeted therapy. This study will provide new insights for targeted therapy of gliomas.

Conclusion

We first summarized the molecular features and potential clinical value of PABPC1 in gliomas. PABPC1 is a favorable biomarker for gliomas and predicts better survival for patients of glioma. Our finding improves the understanding of PABPC1 as a novel therapeutic target for gliomas.