BMP4, a strong better prognosis predictor, has a subtype preference and cell development association in gliomas

Whole genome mRNA expression microarray data and clinical information of 220 glioma samples of all grades from Chinese Glioma Genome Atlas (CGGA) database [8] (http://​www.​cgga.​org.​cn) were obtained as discovery set and 202 glioma expression files from the cancer genome atlas (TCGA) database [9] (http://​cancergenome.​nih.​gov), the Repository for Molecular Brain Neoplasia Data (REMBRANDT, https://​caintegrator.​nci.​nih.​gov/​rembrandt/​) and GSE16011 data [10] (http://​www.​ncbi.​nlm.​nih.​gov/​geo/​query/​acc.​cgi?​acc=​GSE16011) were obtained as validation sets.

After Pearson correlation analysis, gene ontology analysis of the positively correlated genes (r>0.4, p<0.05) were analyzed by DAVID (http://​david.​abcc.​ncifcrf.​gov/​home.​jsp).

Gene set variation analysis with BMP4 expression was analyzed by GSVA package [11] of R [12]. Gene list was obtained from GSVAdata package [13].

Briefly, Immunoperoxidase staining for BMP4 (Abcam, ab39973) were performed following the standard protocol recommended by the manufacturer. Each slide stained for and BMP4 was individually reviewed and scored by two independent observers. Discrepancies in scoring between the two observers were resolved by additional review of the specimens and discussion between the reviewers until a consensus was achieved. Approximately 15-20 fields at 400× magnification were analyzed per specimen. The proportion of positively stained tumor cells was graded as follows: 0, no positive tumor cells; 1, <5% positive tumor cells; 2, 5-20% positive tumor cells; and 3, >20% positive tumor cells. The intensity of staining was recorded on a scale of 0 (no staining), 1 (weak staining, light yellow), 2 (moderate staining, yellowish brown) and 3 (strong staining, brown). The staining index was calculated as follows: staining index = staining intensity × tumor cell staining grade. High BMP4 expression was defined as a staining index score ≥4, while low expression was defined as a staining index <4.

For molecular subtype annotation of the 4 datasets, we applied prediction analysis of microarrays (PAM) as previously reported [8]. Quantitative results were shown as mean ± standard deviation. The difference of BMP expression was tested by the Student t-test in microarray data and by chi-square test in IHC results. Overall survival time (OS) was calculated from the date of diagnosis until death or the last follow-up. The survival curve of patients with high or low expressed BMP4 was calculated with the Kaplan-Meier method and the difference was analyzed using the two-sided log-rank test. A p-value < 0.05 was considered statistically significant. All the data analysis was performed in GraphPad Prism and R.

Written informed consent was obtained from the patient for publication of this report and any accompanying images. The study was performed with the approval of Ethics Committee of Capital Medical University and was in compliance with the Helsinki Declaration.

Of the 220 glioma patients in the training set, there were 97 grade II gliomas (astrocytomas, oligodendrogliomas, and oligoastrocytomas), 34 grade III gliomas (anaplastic astrocytomas, anaplastic oligodendrogliomas, and anaplastic oligoastrocytomas) and 89 grade IV gliomas (GBMs). Clinical information (age, gender, preoperational Karnofsky Performance Scale (KPS) score, and treatment) were obtained from medical records of CGGA database, which were listed in Table 1.

After screening of differently expressed genes in the discovery dataset, we found that BMP4 was in the list of genes significantly differently expressed between LGGs and HGGs. Both in CGGA and other 3 validation datasets, the expression of BMP4 in LGG was higher than that in HGG (Figure 1).

We further validated the protein expression level of BMP4 in an independent group of 77 glioma patients by IHC (Figure 2). Similar to the findings above, BMP4 showed a higher expression status in LGGs than that in HGGs (p<0.05, chi-square test). Thus, BMP4 expression showed a LGG preference both in mRNA and protein level.

We confirmed the prognosis of the 220 patients, and got 216 patients for further prognosis analysis (the prognosis of 2 patients were not available and the OS of another 2 patients were too short which might due to other complications other than glioma). As is shown in Figure 3, both anaplastic glioma (Figure 3B) and GBM (Figure 3C) patients with high or low expression of BMP4 had considerable different prognosis. But there was only a marginal p value in LGGs (Figure 3A). The results were similar in the validation set (Figure 3D-I). Therefore, BMP4 was a better prognostic marker in anaplastic gliomas and glioblastomas. The high or low expression was defined as higher or lower than the median individual in each grade.

As BMP4 showed association with LGGs and better prognosis, we screened its expression in different molecular subtypes of gliomas. As previously reported, we annotated the 4 datasets by TCGA and CGGA classification systems by PAM [8, 9]. BMP4 showed a Proneural and G1 subtype preference. Patients with IDH1 gene mutation also showed higher expression of BMP4 than those with wild-type IDH1 gene (Figure 4). The expression difference was shown in Figure 5.

After Pearson correlation analysis of the data from CGGA, the significantly positively correlated genes (Pink marked genes in Figure 4) were used for GO analysis. The top 20 GO terms listed in Table 2 showed that BMP4 was associated with cell development, differentiation and biogenesis.

As GO analysis showed that BMP4 had a tight association with cell development, differentiation and biogenesis, especially astrocytes, we performed GSVA with BMP4 expression (Figure 6). Genes up- and down-regulated in astrocytes went positively and negatively with BMP4 expression. Meanwhile, it has been reported that Proneural gliomas are characterized by oligodentrocytic genes which is in concordance with the present study [9]. Oligodentrocytic genes were accumulated in patients with higher BMP4 expression, which was preferentially expressed in Proneural gliomas.