Study cases included those which matched to the “Diffuse glioma, MYB- or MYBL1-altered” family by the CNS methylation classifier. In this cohort of 14 cases (Table 1 and Supplementary Table 1), the median age was 11 years (range 1–64) and the male-to-female ratio was 1.8 (9 males and 5 females). All but one were supratentorial, and the remaining case was located in the pons. A final diagnosis of angiocentric glioma was reached in 6 cases, and the remaining 8 cases were diagnosed as diffuse astrocytoma, MYB- or MYBL1-altered. With respect to specific methylation class, the DKFZ Heidelberg classifier divides this family into 4 classes, labeled Angiocentric glioma, MYB/MYBL1-altered (AG_MYB), and Diffuse astrocytoma, MYB- or MYBL1-altered, subtypes B, C, and D. In our cohort, 4 cases matched to the Angiocentric glioma, MYB/MYBL1-altered class, and all 4 were deemed to represent angiocentric gliomas in the final integrated diagnosis (Table 1). Interestingly, 2 additional cases (Cases 1 and 7), deemed to represent angiocentric glioma and showing the characteristic perivascular arrangement of tumor cells as well as positive results of EMA stains (Supplementary Table 1), matched to 2 of the other classes in the family (LGG_MYB_D and LGG_MYB_B, respectively; Fig. 1a, b). By comparison, 2 cases of diffuse astrocytoma, MYB- or MYBL1-altered (Cases 4 and 11), in which perivascular accumulation of tumor cells was not appreciated and which matched to the methylation classes, LGG_MYB_C and LGG_MYB_D, respectively, are also shown in Fig. 1c and d. The remaining 8 cases of diffuse astrocytoma, MYB- or MYBL1-altered all matched to the classes, Diffuse astrocytoma, MYB- or MYBL1-altered, subtypes B, C or D.

Histopathology of MYB(L1)-altered gliomas. Top: 2 cases of Angiocentric glioma (a Case 1; b Case 7), matching to methylation classes, LGG_MYB_D and LGG_MYB_B, respectively. Bottom: 2 cases of Diffuse astrocytoma, MYB- or MYBL1-altered (c Case 4; d Case 11), matching to methylation classes, LGG_MYB_C and LGG_MYB_D, respectively

We then examined MYB and MYBL1 alterations at the sequencing level using the TruSeq RNA Exome assay. MYB::QKI alterations were detected in 3 cases, but only 2 of them were in-frame fusions (Cases 1 and 2; Table 1) and the third one (Case 5) turned out having a tail-to-tail configuration of the fusion partner genes (Fig. 2a). Another type of in-frame MYB fusion involving the PCDHG gene cluster was found in 2 cases (Cases 3 and 4). Case 6 had a MYB fusion with a known partner, MMP16, but it was also a tail-to-tail fusion which cannot express a functional fusion protein (Fig. 2b). MYB was found to be rearranged to an intergenic region in chr6:q26 in 2 cases (Cases 7 and 8; Table 1).

Tail-to-tail non-productive fusions in MYB/MYBL1-altered gliomas. Direction of each coding transcript is marked under the fusion diagram generated by Arriba. a MYB::QKI tail-to-tail rearrangement in Case 5. b MYB::MMP16 (identified as MMP16::MYB) tail-to-tail rearrangement in Case 6

No MYB(L1) fusion or rearrangement was detected in Case 9 by 3 different fusion callers including Arriba, although this case had been classified to the “Diffuse glioma, MYB- or MYBL1-altered” family with maximum confidence scores (Supplementary Table 1). When the manual review of the sequencing reads didn’t reveal anything, either, we took on an unconventional approach of examining the overall sequencing coverage. The RNA sequencing coverage track of Case 9 for the MYB region showed abundant reads up to exon 10, and then a significant reduction to the background noise level following exon 10 (Fig. 3). We interpreted that this pattern supports the presence of a genomic rearrangement downstream of exon 10. By comparison, Case 3 with a MYB in-frame fusion at exon 9, as well as Case 7 with a MYB truncation at exon 15, were also examined for the same MYB exons 7 to 15 region (Fig. 3). The sequencing coverage track of Case 3 showed a reduction in read counts after exon 9, agreeing with the expected pattern. Case 7 coverage track displayed the MYB sequencing reads persisting up to exon 15, also complying with our interpretation. These illustrations demonstrate another means of rearrangement interpretation, even in the absence of an event detectable by standard NGS workflows.

RNA sequencing coverage tracks of the MYB exons 7–15 region. Case 9 shows a reduction in read counts following exon 10, supporting a MYB rearrangement downstream of exon 10; Case 3 with a demonstrated in-frame MYB fusion at exon 9, shows a reduction in read counts following exon 9, representing a rearrangement downstream of exon 9; Case 7 with MYB truncation at exon 15, shows MYB read counts persistent up to exon 15

Cases 10 and 11 had MYBL1 rearrangements with the most common fusion partner, MMP16, however, both were out-of-frame fusions. Cases 12 and 13 had MYBL1 rearranged to intergenic regions in chr8:q21.3 and chr6:q26, respectively.

Overall, among the 9 cases with MYB alterations, 4 were expected to result in productive fusions, with the remaining 5 alterations not expected to result in a productive fusion. Among the 5 cases with MYBL1 alterations, only 1 was expected to result in a detectable in-frame MYBL1 fusion (Case 14; MYBL1::KHDRBS3). The cases whose rearrangements were re-evaluated as being unable to produce in-frame fusion proteins, were collectively deemed “MYB truncation/non-productive fusion” or “MYBL1 truncation/non-productive fusion” cases (Table 1).

We then turned our attention to the expression of relevant genes in these tumors. Whether in-frame fusion or truncation, most MYB-rearrangement cases showed high levels of MYB expression, when compared within an archival cohort of  > 1000 CNS tumor cases profiled in the course of clinical testing on the same RNA Exome platform (Fig. 4). All 4 "MYBL1 truncation/non-productive fusion" cases demonstrated high levels of MYBL1 expression, but the singular in-frame fusion case did not. Interestingly, we found QKI expression to be generally elevated in all MYB- or MYBL1-altered glioma cases, relative to the pan-CNS tumor cohort, and not limited to those cases with genomic rearrangement positions found near the QKI gene (Fig. 4), prompting a speculation that this KH-domain-containing RNA-binding protein may play a role in the physiology of this glioma group, beyond the frequent fusion partner to MYB. We did not find a similar correlation in the expression patterns of other fusion partners.

Expression levels of MYB, MYBL1 and QKI for each case in the cohort. Samples are sorted according to the type of fusion/rearrangement, and for comparison, the expression levels of 1005 pan-CNS tumor cases in the NCI clinical sequencing archives are also shown. The median level of each gene in each sample group is displayed as a line. TPM: transcripts per kilobase million