Expression of FOXP2 in the local cohort of glioblastoma patients.
The IHC analysis, using a specific FOXP2 antibody, showed that the protein was mainly located in the nucleus and the cytoplasm. The mean percentage of positivity in hot-spot areas was 28.33% (SD=32.29), with an adequate correlation between the results of the two independent observers (Spearman`s rho= 0.897; p<0.001).
Prognostic evaluation of FOXP2 expression in the local cohort of glioblastoma patients.
The univariate Cox regression analysis showed that higher FOXP2 expression was associated with a worse prognosis in both progression free survival (PFS) (HR=1.711, 95% C.I. [1.040 – 2.814]; p=0.034) and overall survival (OS) (HR=1.809, 95% C.I. [1.129 – 2.900]; p=.014). The univariate Cox regression analysis also included other variables that had previously been associated with the prognosis of glioblastoma (tables 1 and 2). The association of FOXP2 IHC positivity with worse PFS and OS was still significant when this variable was included in a multivariate Cox regression analysis with those variables that showed statistical significance in the univariate Cox regression analysis (p<0.1) (tables 1 and 2).
Comparison between high and low FOXP2 expression groupsin the local cohort of glioblastoma patients.
Using the median (p50=14.0) of FOXP2 % positivity expression, a comparative analysis between patients with low FOXP2 expression (<=p50) and high FOXP2 expression (>p50) was performed. This cutoff was selected because it was considered the best way to show prognostic differences (figure 1). No clinical, radiological or molecular differences were identified between both groups (table 3). However, according to the Cox regression results, patients with high FOXP2 expression presented a worse PFS (176.1 vs. 339.9 days; p=0.024) and a worse OS (381.9 vs. 624.9 days; p=0.042) (table 3, figure 1).
Expression of FOXP2 in the TCGA cohort of glioblastoma patients.
The mean expression of FOXP2 RNA was 25.75 RPKM (SD=52.02) (from RNA seq V2 data). The mean expression of FOXP2 in normal brain had been established in previous assays at around 3.715 RPKM +/-0.605 (data extracted from https://www.ncbi.nlm.nih.gov/gene/). Bearing this figure in mind, one hundred and two glioblastoma patients (70.83%) presented an over-expression of FOXP2 RNA.
Prognostic evaluation of FOXP2 expression in the TCGA cohort of glioblastoma patients.
The univariate Cox regression analysis did not show any effect of FOXP2 mRNA expression on OS (HR=1.142, 95% C.I. [0.787 – 1.659], p=0.484) (table 4). Using the median of FOXP2 expression as cutoff, the Log-Rank test and the Kaplan-Meier curves were performed to test differences in OS. Similarly to the Cox regression results, no significant differences were identified in the median survival between low and high FOXP2 expression (419.0 vs. 343.0 days; Log-Rank; p=0.426) (figure 2). PFS was not evaluated here, because data from TCGA was not accurate enough. Other clinical and molecular data were evaluated, but only the age showed an association with OS (table 4).
Comparison between high and low FOXP2 expression groupsin the TCGA cohort of glioblastoma patients.
A comparison between patients with high and low FOXP2 RNA expression was performed using the median expression of FOXP2 as cutoff (p50=7.1668). Differences were identified in the molecular classification distribution, where the “high RNA FOXP2” expression group presented a higher percentage of patients classified in the neural and proneural groups (table 5). These groups presented higher FOXP2 RNA levels than the classical and mesenchymal subtypes, but this difference was only statistically significant for the comparisons neural vs. mesenchymal and proneural vs. mesenchymal (ANOVA; p=0.024 and p=0.030, respectively) (supplementary figure 2).
Regarding the broad molecular information available in the TCGA patients, additional comparisons between low and high FOXP2 expression groups were performed. These comparisons were focused on DNA copy-number variation (CNV) and DNA mutations. Firstly, the frequency of chromosomal gains or losses was analyzed in both FOXP2 RNA expression groups. It should be noted that twelve patients (19.7%) from the high FOXP2 RNA expression group showed losses in 15q. This genomic loss was significantly different to those in the low FOXP2 RNA expression group (uncorrected p value=0.038), but did not reach significance with corrected FDR values (FDR>0.1) (supplementary table 7). No other statistical differences in the frequency of chromosomal alterations between both groups were identified.
Secondly, DNA CNV analysis was performed. Analysis of focal amplifications and deletions in the top-10 glioblastoma cancer-related genes with focal CNVs showed a larger number of amplifications of CDK4 (24.7% vs. 9.9%; p=0.027; FDR>0.1) in the high FOXP2 RNA expression group (supplementary table 8). CDK4 is located in chromosome 12 (cytoband: 12q14.1) and it encodes a protein member of the Ser/The protein kinase family. This kinase is responsible for the phosphorylation of retinoblastoma gene product (Rb). On the other hand, more patients in the low FOXP2 RNA expression group showed a notable deletion of CDKN2B (59.2% vs. 42.5%; p=0.048; FDR>0.1) (supplementary table 8). The protein encoded by this gene is also involved in cell growth and in the control of the cell cycle.
Finally, the mutational burden of the selected TCGA patients was analyzed. The most common driver mutations among the selected genes in the TCGA cohort of patients affected PTEN, TP53 and EGFR (supplementary table 5). Comparisons between the incidence of mutations in the two FOXP2 RNA groups are reported in supplementary table 9. No differences in the incidence of PTEN and TP53 mutations in low and high FOXP2 RNA expression groups were identified, but more patients with a missense mutation in EGFR gene were identified in the low FOXP2 RNA expression group (38.0% vs. 16.4%; p=0.042). Only one patient presented a driver mutation in ATRX. PCLO gene showed a higher incidence of passenger mutations in the low FOXP2 RNA expression group (15.5% vs. 2.8%; p=0.016; FDR=0.16). This gene encodes a protein of the presynaptic cytoskeletal matrix and is involved in establishing active synaptic zones and in synaptic vesicle trafficking. Its role in glioma pathogenesis has not been fully studied.
Expression and prognostic evaluation of hsa-miR-181a-2-3p and hsa-miR-20a-3p: two FOXP2-targeted miRNAs.
Regarding the results described above and bearing in mind the intense regulation of FOXP2 expression by different miRNAs, post-hoc analysis using the expression of some miRNAs were performed. Two miRNAs that targeted FOXP2 were selected (see Methods and supplementary table 4): hsa-miR-181a-2-3p and hsa-miR-20a-3p.
The OS of patients with a low expression of hsa-miR-181a-2-3p was influenced by FOXP2 expression. Those with a high FOXP2 expression have a median OS of 239.0 days (95% C.I. [49.5 –428.5]) and patients with low FOXP2 expression have a median OS of 419.0 days (95% C.I. [308.7–529.2]). This difference was statistically significant (p=0.028) (figure 3a). On the other hand, patients with a high expression level of has-miR-181a-2-3p did not show differences in OS between low and high expression groups of FOXP2 (333.0 vs. 395.0 days; Log-Rank test; p=0.604) (figure 3a).
The OS of patients with a low expression of hsa-miR-20a-3p was also influenced by FOXP2 expression. Those with a high FOXP2 expression have a median OS of 239.0 days (95% C.I. [62.2 – 415.8]) and patients with low FOXP2 expression have a median OS of 448.0 days (95% C.I. [368.7–527.3]). This difference was statistically significant (p=0.020) (figure 3b). On the other hand, patients with a high expression level of has-miR-20a-3p did not show differences in OS between low and high expression groups of FOXP2 (333.0 vs. 454.0 days; Log-Rank test; p=0.511) (figure 3b).