Circular and Fusion RNAs in Medulloblastoma Development

Abstract

Background. The cerebellar cancer medulloblastoma is the most common childhood cancer in the brain. Methods. RNA sequencing of 81 human biospecimens of medulloblastoma using pipelines to detect circular and fusion RNAs. Validation via PCR and Sanger sequencing. Results. 27, 56, 28 and 11 RNA circles were found to be uniquely up-regulated, while 149, 7, 20 and 15 uniquely down-regulated in the SHH, WNT, Group 3, and Group 4 medulloblastoma subtypes, respectively. Moreover, linear and circular fusion RNAs containing exons from distinct genes joined at canonical splice sites were also identified. These were generally expressed less than the circular RNAs, however the expression of both the linear and the circular fusions was comparable. Importantly, the expression of the fusions in medulloblastoma was also comparable to that of cerebellum. Conclusions. A significant number of fusions in tumor may be generated by mechanisms similar to the ones generating fusions in normal tissue. Some fusions could be rationalized by read-through transcription of two neighboring genes. However, for other fusions, e.g., a linear fusion with an exon from a downstream gene joined 5' to 3' with an exon from an upstream gene, more complicated splicing mechanisms, e.g., trans-splicing, have to be postulated.

Keywords: back-splicing; brain cancer; pediatric cancer; read-through transcription; trans-splicing.

Figure 1

Differentially expressed circular RNAs in human medulloblastoma and normal cerebellum. (A) Volcano plots of the RNA-seq data of the SHH, WNT, Group 3 and Group 4 medulloblastoma tumor and normal cerebellum samples. Cutoff thresholds are assessed with the DESeq2 method, where the Wald significance test is applied as |log2 fold change| > 1, padj < 0.05. Annotated are selected circular RNAs defined as the top 5 with highest normalized mean count across samples, the top 5 with lowest padj value, the top 5 with highest log2 fold change and the top 5 with lowest log2 fold change. (B) Venn diagrams depicting uniquely and commonly up- and down-regulated circular RNAs (circRNA) in SHH, WNT, Group 3 and Group 4 medulloblastoma subtypes compared to normal cerebellum. (C–E) Violin plots of the expression of uniquely up- or down-regulated circular RNAs (circRNA) in the SHH (C), WNT (D), Group 3 (E) and Group 4 (F) medulloblastomas that are also annotated as selected circular RNAs in (A).

Figure 1

Differentially expressed circular RNAs in human medulloblastoma and normal cerebellum. (A) Volcano plots of the RNA-seq data of the SHH, WNT, Group 3 and Group 4 medulloblastoma tumor and normal cerebellum samples. Cutoff thresholds are assessed with the DESeq2 method, where the Wald significance test is applied as |log2 fold change| > 1, padj < 0.05. Annotated are selected circular RNAs defined as the top 5 with highest normalized mean count across samples, the top 5 with lowest padj value, the top 5 with highest log2 fold change and the top 5 with lowest log2 fold change. (B) Venn diagrams depicting uniquely and commonly up- and down-regulated circular RNAs (circRNA) in SHH, WNT, Group 3 and Group 4 medulloblastoma subtypes compared to normal cerebellum. (C–E) Violin plots of the expression of uniquely up- or down-regulated circular RNAs (circRNA) in the SHH (C), WNT (D), Group 3 (E) and Group 4 (F) medulloblastomas that are also annotated as selected circular RNAs in (A).

Figure 2

Fusion transcripts in linear and circular RNAs from cerebellum and medulloblastoma. (A) Box plots of the expression of linear (mdata) and circular (cdata) fusion transcripts in cerebellum (CB) and medulloblastoma (MB). The y-axis shows the total counts per sample of all fusion transcripts (black rhombi indicate outlier samples). (B,C) Violin plots of the expression of the selected fusions (Table 1) in the linear (mdata) and circular (cdata) datasets from cerebellum (B) and medulloblastoma (C). Note that the selected TFG--ADGRG7 linear fusion in medulloblastoma is also found in the circular dataset. The lack of expression of a selected fusion in a dataset is not shown.

Figure 3

Electrophoresis and sequence of the NC0A2--TRAM1 and FBXO25--SEPTIN14 fusion PCR products. Agarose gel electrophoresis of PCR products generated from cDNA of the cerebellar samples 83 and 86 for the NC0A2--TRAM1 and FBXO25--SEPTIN14 fusions, respectively, using primer pairs specific for each fusion (Section 2 Materials and Methods). MW, molecular weight markers. The electropherograms of the sequenced PCR products, with the fusion junction indicated, is shown. Note the presence of an A nucleotide at position 12 of the SEPTIN14 sequence, indicated by an arrow, instead of the anticipated G nucleotide.

Figure 4

Schematic representation of molecular events leading to the observed ADAMTSL3--SH3GL3 circular fusion and the KANSL--ARL17A/B linear and circular fusions. (A) Read-through transcripts from the SH3GL3 gene extend into the downstream ADAMTSL3 gene, followed by back-splicing of exon 3 of ADAMTSL3 to exon 7 of SH3GL3. The resulting circle may or may not include additional ADAMTS3 or SH3GL3 exons, indicated by hatched lines on the diagram. (B) Transcripts from the KANSLI1 and ARL17A/B gene bring in proximity the exon 3 of KANSL1 to the exon 3 of ARL17A/B. The 3′ end of KANSLI1 exon 3 trans-splices to the 5′ end of ARL17A/B exon 3 producing the observed KANSLI1--ARL17A/B linear fusion. In cases where a back-splicing event also occurs, with the 3′ end of ARL17A/B exon 3 joining the 5′ end of KANSL1 exon 3, a circular RNA composed of these two exons is produced.