Tumour tissue from the pons of 66 DIPG patients was screened for K27M mutation in histone H3 as previously described [
10]. 42/66 (64 %) were found to be mutated for K27M-H3.3, with an additional eight patients with K27M-H3.1 mutations (12 %). Mutant allele frequencies from deep sequencing suggest histone mutations are heterozygously present in all tumour cells. Of the 14 biopsy tested for histone mutations, 7 (50 %) were mutated for K27M-H3.3 and none had histone H3.1 mutations. Of the 52 autopsy cases tested for histone H3 mutations, 35 (67 %) were mutated for K27M-H3.3 and 8 (15 %) were mutated for K27M-H3.1. We previously performed survival analysis in a subset of these DIPG patients based on histone mutation status [
10]. In this larger series, the observation that patients with K27M mutation in either histone H3.1 or H3.3 have worse overall survival compared to patients with no histone mutations remains significant (Fig.
2b, Log rank,
p = 0.0026). On multivariate analysis (Cox regression), which included histone mutational status, age of diagnosis, histological grade and sex, only histone mutation status was determined to be a significant predictor of overall survival with a hazard ratio of 2.8 (95 % confidence intervals, 1.35–5.78,
p = 0.006, Table
2). DIPG patients who had wild-type H3 were diagnosed at a significantly younger age than patients with K27M-H3 (wild-type 4.84 years ± 4.13 vs. K27M-H3 7.36 years ± 3.31,
p = 0.018). There was also correlation between histologic grade and mutational status. Eighty-eight percent of GBMs harboured a K27M mutation in H3 versus 60 % of AAs and 71 % of LGA. K27M-H3.3 was found at a statistically higher ratio in GBM patients (78 %) than patients with AA histology (33 %) (
p = 0.0016). The association of anaplastic astrocytoma histology and K27M-H3.1 mutations approached significance (
p = 0.058). All K27M mutant tumours were glial (tumours with PNET histology were wild-type), but represented a spectrum of WHO grades (II–IV). There was no correlation between K27M-mutation status and leptomeningeal spread. Some patients were wild-type for H3.3 yet exhibited features of high-grade astrocytic tumours such as pseudopalisading necrosis, vascular endothelial proliferation and mitoses (Fig.
1d). Interestingly, despite typical high-grade glioma features, patients who were wild-type for H3 with GBM histology survived significantly longer than patients mutated for K27M with GBM histology (mean survival of 1.99 years vs. 0.96 years;
p = 0.026). Conversely, some DIPG patients exhibited low-grade tumours with low proliferative index, yet were mutated for K27M-H3.3 (Fig.
1b) and exhibited a clinical course typical of high-grade gliomas. DIPG with grade II and III histology carrying the mutation do just as poorly as GBM. The overall survival of K27M-H3.3 tumours with grade II and III histology was 0.82 ± 0.47 years compared to 0.91 ± 0.77 years for K27M-H3.3 grade IV tumours (
p > 0.05). Patients with WHO grade II diffuse astrocytomas, but wild-type for H3 survived longer than K27M-H3.3 mutant WHO grade II astrocytomas. Histologically, these cases all had similar features of diffusely infiltrating astrocytic tumour with rare to no mitotic figures, low MIB1, diffuse GFAP immunopositivity and rare to no P53 immunopositivity. Most DIPGs were positive for GFAP, with more diffuse staining present in the grade II lesions and no GFAP immunopositivity present in DIPG with PNET histology.
Table 2
Multivariate Cox regression analysis
K27M-H3 | 2.793 | 0.006 | 1.350 | 5.777 |
Histology | 0.808 | 0.222 | 0.574 | 1.138 |
Age Dx | 1.016 | 0.693 | 0.940 | 1.097 |
Sex | 1.267 | 0.412 | 0.720 | 2.231 |