H3 K27M Mutation in Gangliogliomas can be Associated with Poor Prognosis

Spinal cord 7, 15, 16 and brainstem 1, 2, 12, 14 diffuse midline gliomas have been shown to have H3F3A (K27M) mutations, which are paralleled by presence of strong, diffuse nuclear immunoreactivity using a high fidelity immunohistochemistry (IHC) stain 17. Although a World Health Organization (WHO) grade of IV has been assigned to diffuse midline glioma, H3 K27M‐mutant 8, it is becoming apparent that rarely low grade tumors, including pilocytic astrocytoma, may also show this mutation 9. However, in the latter case, despite the long patient survival, the tumor eventually progressed and caused demise 9. This suggests that the presence of mutation may portend adverse prognosis in any tumor type in which it occurs.

H3 K27M mutation appears to be very rare in pilocytic astrocytomas in general. K27M was reported to be negative/absent in 15 of 15 midline pilocytic astrocytomas (by IHC) 16, 40 of 40 pilocytic astrocytomas (by IHC, from unspecified anatomical locations) 17, and 15 of 15 pilocytic and 5 of 5 pilomyxoid astrocytomas (by pyrosequencing, from unspecified anatomical locations) 6. However, one high grade glioma positive for K27M IHC has been reported that contained a prominent pilomyxoid component 16.

Gangliogliomas (GGs) are another tumor type that usually lack H3 K27M mutation. Gielen et al in a large survey of 163 pediatric tumors of all types showed that the majority of tumor types from unspecified anatomical locations lack mutation (by pyrosequencing), including 10 of 10 GGs 6. Venneti et al demonstrated that 17 of 17 pediatric and 2 of 2 adult GGs were negative for mutation (IHC and sequencing), all WHO grade I 17. Anatomical location was not specified, however, and thus it is unclear how many midline examples were included.

Nevertheless, similar to the situation with pilocytic astrocytoma, exceptional reports of GGs positive for H3 K27M IHC are starting to appear. These include two spinal cord anaplastic GGs 6, one malignant glioma with epithelioid and rhabdoid features as well as ganglionic differentiation 16, one thalamic GG initially grade I and then recurrent as anaplastic grade III 7 years later 11, and one cerebellar GG, initially WHO grade I followed by rapid progression and anaplastic GG 2 months later 11. Similar to the above‐cited pilocytic astrocytoma with mutation 9, the two patients with progression suffered demise at 8 years, 2 months and 33 months after diagnosis 11.

Given our previous experience with a cohort of pediatric patients with brainstem GGs on whom we had testing for BRAF V600E mutational status and clinical follow‐up 3, as well as with two adult patients with spinal cord GGs who succumbed and had massive metastatic tumor dissemination at autopsy 13, we re‐interrogated these cases for the H3K27M mutation, using the high‐fidelity antibody (Millipore, Temecula, CA) 17. We also assessed three additional unpublished midline pediatric GGs with autopsy documentation of metastatic disease, as well as several surviving adult patients with midline GGs. Given the known difficulty of diagnosis of GG on small biopsies, especially from brainstem and spinal cord locations, and the inability to utilize IDH1 R132H immunohistochemistry to distinguish diffuse infiltrating gliomas with entrapped non‐neoplastic ganglion cells from true GGs in these sites 5, 10, only definitive GG examples were utilized for study. Previously published cases had been well‐documented histologically 3, 13.

Details of patient age, tumor anatomical location, and results for H3 K27M IHC, BRAF V600E mutation via Sanger sequencing, BRAF VE1 IHC and ATRX IHC are documented in the Table 1. Four of the five autopsied cases had disseminated metastatic disease, including the two previously published adult cases 13. The other three autopsied cases in our study were all from pediatric patients and had not been previously published.

Table 1.

Clinical and IHC features of gangliogliomas

PT #	Age, gender	PT # in original papers*	Histology	Location initially/plus later sites tested	Recurrent	Overall length of survival/follow‐up in years	Died of disease	BRAF exon 15 sequencing	BRAF VE1 IHC	ATRX IHC	H3 K27M IHC
1	14 yr. F	2	GG	BS	N	6.13	N	WT	Neg	Retained	Neg
2	4 yr. F	3	GG	CC/BS	Y	7.52	N	V600E	VE1 +	Retained	Neg
2	4 yr. F	3	GG (second excision)	CC/BS		7.52		ND	ND	Retained	Neg
3	12 yr. F	4	GG	BS	Y	6.52	N	V600E	VE1 +	Retained	Neg
4	13 yr. F	7	GG	CB/BS/CC	N	13.3	N	WT	Neg	Retained	Neg
5	5 yr. M	12	GG	CB/BS	Y	2.24	Y	WT	Neg	Retained	Neg
6	6‐day‐old F	13	GG	BS	N	1.83	N	V600‐W604 deletion	ND	Retained	Neg
7	6 yr. F	–	GG metastatic autopsied	CC	Y	1.61	Y	WT†	ND	Retained	Pos
7	6 yr. F	–	Biopsy of metastasis, upgraded to anaplastic	Intraventricular metastasis	Y	1.61	Y	ND	ND	Retained	Pos
7	6 yr. F	–	Upgraded to anaplastic	Autopsy: metastasis within CNS	–	1.61	–	ND	ND	Retained	Pos
8	13 yr. M	–	GG not metastatic autopsied	BS	Y	0.79	Y	ND	Neg	Indeterminate	Pos
9	16 yr. F	–	GG metastatic autopsied	BS	Y	1.92	Y	V600E	Pos	Retained	Neg
9	16 yr. F	–	No upgrade at autopsy	Autopsy: metastasis within CNS	–	1.84	–	ND	ND	Retained	Neg
10	62 yr. M	–	GG	CC	N	0.25	N **	ND	Neg	Retained	Neg
11	53 yr. F	–	GG	T	N	3.05	N ***	WT	ND	Retained	Neg
12	20 yr. M		GG	CB	unknown	unknown	N	WT	ND	ND	Neg
13	47 yr. F	2	GG metastatic autopsied	CT	Y	6.34	Y	WT	Neg	Retained	Neg
14	24 yr. F	1	GG metastatic autopsied	CC	Y	3.85	–	WT	Neg	Retained	Neg
14	24 yr. F	1	Anaplastic	Autopsy specimen: systemic metastasis	–	3.85	–	ND	ND	Retained	Neg

Key: F, female; M, male; GG, ganglioglioma at time of surgical biopsy or resection; CC, cervical cord; BS, brainstem; CB, cerebellum; T, thoracic spinal cord; CT, cervicothoracic cord; follow‐up from first biopsy/resection to demise or survival as of 9/28/16; N, no; Y, yes; IHC, immunohistochemistry; WT, wildtype; ND, not done; *, Reference 3 for patients 1–6; Reference 16 for patients 11, 12, patients 7–10 not previously published; **, follow‐up <1 year; ***, stable on 3‐year follow‐up. ^†FoundationOne screen of 315 candidate cancer genes including BRAF.

Patient 7 was a child who presented with an extensive cervical and upper thoracic cord tumor (Figure 1a) that intraoperatively manifested as a bulging, hyperemic intrinsic mass (Figure 1b). Extensive resection was diagnosed as GG, WHO grade I, based on the presence of numerous neoplastic ganglion cells (Figure 1b), binucleate tumor ganglion cells (Figure 1c), multifocal calcifications (Figure 1d, arrowheads), near‐absence of mitotic activity and low MIB‐1 cell cycle labeling, as performed on three different blocks. (See also Supporting Information File 1 for more images of this case, including prominent calcifications.) The patient had local tumor progression and, despite chemoradiation, developed multifocal metastases to bilateral frontal horns, septum pellucidum and left occipital horn 11 months after initial diagnosis. Biopsy from one of her ventricular metastases now showed anaplastic transformation of the GG. Death occurred 9 months later, at which time autopsy documented involvement of C6‐T2, as well as the extensive cerebral intraventricular metastatic deposits. Both metastatic cerebral deposits and recurrent spinal cord disease showed anaplastic transformation of the GG, with palisading necrosis. Initial spinal cord resection specimen, biopsy sample from the intraventricular metastasis, and an autopsy sample from her central nervous system metastasis were all positive for H3 K27M (see Table 1) (Figure 1d, initial biopsy illustrated, arrowheads highlighting multifocal calcifications). Testing of tissue specimens taken at three different time points, all positive, indicates neither gain nor loss of mutation over the course of disease. While double labeling with immunostains for H3 K27M and neuronal markers was not performed, it was felt that both neoplastic ganglion cells and neoplastic glial cells were positive for the mutation.

Figure 1.

Pediatric gangliogliomas at autopsy. a–d. From patient 7—It shows the cervical spinal cord location (a, magnetic resonance imaging T1 with contrast, intraoperative photograph), WHO grade I features with numerous neoplastic ganglion cells (b, arrowheads) and basophilic microcalcifications (hematoxylin and eosin [H&E]), binucleated, neurofilament‐immunoreactive (phosphorylated NFP) neoplastic neurons (c), and H3 K27M immunoreactivity (d, with numerous interspersed microcalcifications archetypal for GG at arrowheads). e–g. From patient 8—It illustrates the patient's pontine GG (e, magnetic resonance imaging T2), tumoral neoplastic neurons (arrowheads) and microcalcifications (d, H&E), and H3 K27M immunoreactivity (e, with possible neoplastic ganglion cell at arrowhead). h–j. From patient 9—It shows that despite massive dissemination at autopsy to basilar leptomeninges (f, arrowhead), with encasement of spinal cord (g, whole mount, H&E), the tumor is nevertheless negative for H3 K27M by immunohistochemistry (j).

Patient 8 was a child who presented with a pontine mass (Figure 1e) that was diagnosed on stereotactic biopsy as a GG with numerous neoplastic ganglion cells (Figure 1f, arrowhead) and scattered basophilic calcifications (Figure 1f). The tumor showed no mitotic activity, only rare cells with MIB‐1 cell cycle labeling, and was diagnosed as GG, WHO grade I (see also Supporting Information File 2 for more images of this case). Both neoplastic ganglion cells (Figure 1e, arrowhead) and the neoplastic glial cell component were H3 K27M positive (Figure 1g), similar to a case of Solomon et al 16, although confidence was somewhat reduced by the small volume of tissue remaining in the tissue block for H3 K27M. This patient also developed clinical progression, eventually entered hospice case, and succumbed 9½ months after diagnosis. Autopsy showed tumor confined to parenchyma of pons (initial location) and adjacent medulla, but no cerebrospinal or distant parenchymal metastatic disease. Tumor at autopsy had also upgraded to anaplastic GG with palisading necrosis and microvascular proliferation. Tumors from patients 7 and 8 did not show loss of nuclear ATRX (4).

Patient 9 was a child who also presented with a pontine mass that was found to be GG, WHO grade I at biopsy. Her tumor progressed and she succumbed 23 months later. At autopsy, she was found to have bulky leptomeningeal metastases obscuring the brainstem (Figure 1h) and encasing spinal cord (Figure 1i), despite being H3 K27M immunonegative (Figure 1j). This case was positive for BRAF V600E mutation by Sanger sequencing, as were several of our original pediatric brainstem GGs without clinical progression 3. We also found these to be negative for H3 K27M immunostaining (see Table 1). None of the tested pediatric brainstem GGs from our original study 3 was positive for H3 K27M (see Table 1).

Neither of the two adult previously‐published, widely‐metastatic autopsied spinal cord gangliogliomas 13 showed H3 K27M immunopositivity in their surgical specimen (see Table 1). The negative H3 K27M status persisted throughout the disease course in the patient with systemic organ dissemination at autopsy, in that both the surgical specimen and the autopsy metastatic specimen were negative. None of the other adult, non‐metastatic midline GGs that we tested were positive.

We agree with Joyon et al who suggested that “ganglioglioma grade I with H3.3 K27M mutation could be prone to a malignant transformation” 11 and extend their observations to suggest that such tumors may additionally develop metastatic tumor deposits. We expand numbers of midline GGs tested for H3 K27M in this study and document that the majority of midline GGs are negative, agreeing with previous literature on GGs from all anatomical locations 6, 17. Conversely, however, we note that absence of H3 K27M mutation in a GG does not automatically guarantee good prognosis, since several of our GG patients with demise showed dissemination at autopsy and were H3 K27M negative.

Despite the small numbers, we conclude that H3 K27M IHC testing is warranted for all midline GGs. Even if detection is rare in midline GGs, H3 K27M IHC when positive can provide prognostic information beyond grade. Our two GGs with H3 K27M immunopositivity, coupled with the previously published cases of Hochart et al 9 and Joyon et al 11, suggest that an adverse prognosis can be associated with this mutation, even when it occurs in midline tumors other than diffuse midline gliomas 8, such as midline ganglioglioma or pilocytic astrocytoma.

Finally, the calcifications, numerous cytologically atypical and irregularly‐placed neoplastic ganglion cells, and relatively sharp demarcation in our two pediatric H3 K27M‐positive cases (one spinal cord, one pontine) suggested to us that our two immunopositive GGs, when combined with the other well‐documented examples in the literature 11, were not misdiagnosed diffuse midline gliomas with entrapped normal ganglion cells.

However, we acknowledge that an alternate interpretation is possible, namely metaplastic neuronal change within a diffuse glioma, H3 K27M mutant. The latter was the explanation favored by the authors for a mutant case with numerous neoplastic ganglion cells reported by Solomon et al 16. In that case, other regions of that same tumor manifested epithelioid/rhabdoid features, suggesting metaplastic change had occurred in only a subpopulation. These authors made the point, however, that, in their series, “all tumors had at least some classic astrocytic morphology with ovoid to elongate nuclei containing coarse chromatin”; in other words conventional infiltrating glioma. In our two H3 K27M tumors, no areas of classic astrocytic morphology/infiltrating glioma were identified and the heavy calcifications would be unusual for diffuse/infiltrating glioma.

In addition, they found that in mutant tumors “a wide morphologic spectrum was encountered including gliomas with giant cells, epithelioid and rhabdoid cells, primitive neuroectodermal tumor (PNET)‐like foci, neuropil‐like islands, pilomyxoid features, ependymal‐like areas, sarcomatous transformation, ganglionic differentiation and pleomorphic xanthoastrocytoma (PXA)‐like areas” 16. These findings reinforce our conclusions from this study—namely, that H3 K27M mutational status cannot not be predicted based on morphological features of the tumor, only by midline location.

Indeed, we felt the two mutant GGs could not be distinguished microscopically from other GGs without mutation in our series. We further provide Aperio imaging files from the slides of the two H3 K27M‐mutant GG cases (hosted by the publisher on their website, with free viewing software, available at: http://www.leicabiosystems.com/digital-pathology/digital-pathology-management/imagescope/ so that the reader can view these two cases and form their own impressions.

Supporting information

Additional supporting information may be found in the online version of this article at the publisher's web‐site.

Figure S1. Patient 7 had a ganglioglioma (GG) immunopositive for H3 K27M; tumor was heavily calcified (a) and contained numerous neoplastic ganglion cells (b, c) which showed strong immunoreactivity for synaptophysin in the neoplastic ganglion cells (d), focal immunoreactivity for chromogranin (e), and diffuse nuclear immunostaining for H3 K27M (f).

Figure S2. Patient 8 had a GG immunopositive for H3 K27M; tumor showed microcalcifications and numerous neoplastic ganglion cells, seen best in lower half of (a), strong immunoreactivity for synaptophysin in the neoplastic ganglion cells (b), glial fibrillary acidic protein (GFAP) immunostaining in the glioma component of the GG (c), very low cell cycling rate on MIB‐1 immunostaining (d), and diffuse nuclear immunostaining for H3 K27< (e, f).

Supplemental Aperio scans. Scanned slides from the two patients with gangliogliomas immunoreactive for H3 K27M, patients 7 and 8, to show classic GG histological features.

Supporting Information

Supporting Information

Supporting Information

Supporting Information

Supporting Information