Genetics of Glioblastomas in Rare Anatomical Locations: Spinal Cord and Optic Nerve

Correspondence

Spinal cord diffuse astrocytomas, WHO grades II and III, are known not to harbor IDH1/2 mutations as do many of their supratentorial counterparts 4, but it is unknown whether glioblastomas (GBMs) of spinal cord or optic nerve share other mutations common to supratentorial GBMs, such as epidermal growth factor receptor (EGFR) or loss of phosphatase and tensin homolog (PTEN). In addition, both spinal cord and optic chiasm GBMs are midline in anatomic location and H3.F3A K27 mutation (histone K27M‐H3.3), a mutation originally identified in diffuse pontine gliomas in the pediatric population 2, 3, 6, has very recently been identified in adult GBMs (18 years of age and older) “enriched for tumors of infratentorial and midline localization,” although exact numbers of spinal cord/optic chiasm examples were not specified 9.

Rarity of GBMs in either optic chiasm or spinal cord (<1–3% of all GBMs) limits study size at any one institution and most are thus case reports or small cohorts, such as five spinal cord GBMs in the most recent series 7. We conducted a case search of departmental databases, 2006–present, and identified biopsies from seven spinal cord and two optic nerve GBMs from our institution (UCHSC). Two of 7 cord cases had occurred in pediatric patients, ages 8 and 17 years; the latter represented the only radiation‐induced example and occurred following therapy for standard risk medulloblastoma 10 years prior (see Table 1).

Table 1.

Glioblastomas of spinal cord and optic chiasm.

Case, age, surgery date	Location	ATRX/H3F3/IDH/P53	Genetics	Survival Time
#1 8 YO F, 2008 Biopsy, autopsy	Cervicomedullary to C2	ATRX‐Retained H3F3‐Negative IDH1‐Negative P53‐100%	Positive for amplification for EGFR; specimen failure for PTEN	Deceased‐48 h, autopsy performed
#2 38 YO M, 2009 Biopsy only	Cervicomedullary junction	ATRX‐Retained H3F3‐Positive IDH1‐Negative P53‐<5%	Negative for amplification of EGFR and PTEN loss	Deceased—date unknown
#3 69 YO M, 2010 Biopsy only	Upper cervical	ATRX‐Retained H3F3‐Negative IDH1‐Negative P53‐<5%	Negative for amplification of EGFR and PTEN loss	Deceased‐8 months, 15 days—no autopsy performed
#4 58 YO F, 2013 Biopsy only	Intramedullary C7‐T1	ATRX‐Retained H3F3‐Negative IDH1‐Negative P53‐<5%	Positive for EGFR amplification and PTEN loss	Alive‐1 yr, 4 months, 2 days to (last seen 11/26/14) Biopsy‐documented cerebellar metastasis/spread
#5 17 YO M, 2013 Radiation‐induced; (following treatment for standard risk medulloblastoma 10 years previous) Biopsy only	T12‐L2 region	ATRX‐Retained H3F3‐Negative IDH1‐Negative P53‐<5%	Negative for amplification of EGFR and PTEN loss	Deceased‐5 months, 4 days—no autopsy performed
#6 27 YO F, 2014 Biopsy only	T8‐12 region	ATRX‐Retained H3F3‐Not done IDH1‐Negative P53‐<5%	Negative for amplification of EGFR and PTEN loss	Alive 5 months, 21 days (last seen 6/10/15)—Now with cranial nerve palsies suggestive of spread up spinal column
#7 33 YO F, 2015 Biopsy only	Conus Medullaris	ATRX‐Lost H3F3‐Positive IDH1‐Negative P53‐40%	Negative for amplification of EGFR and PTEN loss	Status unknown; lost to  follow up
#8 56 YO M, 2015 Biopsy, autopsy	Optic nerve	ATRX‐Retained H3F3‐Negative IDH1‐Negative P53‐Not Done	Positive for amplification of EGFR and PTEN loss	Deceased, 26 days—autopsy performed
#9 45 YO M, 2006 Biopsy, autopsy	Suprachiasmatic region	ATRX‐Lost H3F3‐Negative IDH1‐Negative P53‐<5%	Positive for amplification of EGFR and PTEN loss	Deceased, 48 h—autopsy performed

Key: Exceptional cases bolded.

Abbreviations: IDH1 = isocitrate dehydrogenase 1, ATRX = alpha thalassemia/mental retardation syndrome X‐linked, EGFR = epidermal growth factor receptor, PTEN = phosphatase and tensin homolog.

Given the paucity of data on genetic features in GBMs from these rare sites, as noted by Karsy et al 5, tumors were assessed by immunohistochemistry for IDH1 R132H cytoplasmic expression (1:20 [manufacturer's recommended dilution], Histo Bio Tech/Dianova, Miami Beach, FL, USA), nuclear p53 expression (1:200, Dako Corp., Carpinteria, CA, USA), retention or loss of nuclear ATRX (1:200, alpha thalassemia/mental retardation syndrome X‐linked) immunostaining (using the recommended Sigma‐Aldrich antibody, St. Louis, MO, USA, and Ventana benchmark processing 10), and H3.F3A K27 (1:500, Millipore, Temecula, CA, USA). Scoring was subjective for IHC. Standard methodology fluorescence in situ hybridization was utilized for EGFR amplification and loss of PTEN/10 centromere.

Given the known poor prognosis of these tumors 7, not surprisingly, the few surviving patients in our cohort had been diagnosed the most recently (see Table 1). Spinal cord GBMs originated in cervicomedullary and upper cervical regions in 3 patients, and C7‐T1, T12‐L2 and T8‐T12, lumbar region in one each, underscoring the diversity of anatomical levels of cord involvement seen with these tumors (Figure 1A). Tumors were expansile, with ill‐defined borders (Figure 1A,B), and variably hemorrhagic (Figure 1B). One spinal cord GBM patient with survival >1 year developed anterograde, biopsy‐proven spread to the cerebellum (case 4). Another by neuroimaging was felt to have cranial nerve palsies suggestive of spread from the spinal cord primary (case 6). Nuclear pleomorphism (Figure 1C), immunoreactivity for glial fibrillary acidic protein (Figure 1D), and either necrosis or microvascular proliferation were identified on biopsies, as is typical of supratentorial counterparts.

Figure 1.

A. MRI illustrating the most uncommon anatomical location for spinal cord GBMs, that is, the lumbar region; note the expansile, nondiscrete features. Case 7 illustrated. B. At autopsy, this pediatric patient with a cervicomedullary spinal cord GBM was found to have extensive hemorrhagic features. Case 1 illustrated. C. GBMs manifested nuclear pleomorphism and hypercellularity; mitoses and vascular proliferation were present elsewhere. Case 7 illustrated. Hematoxylin and eosin, 400×. D. The lumbar GBM had no features to suggest derivation from a myxopapillary ependymoma or lower grade glioma precursor; strong GFAP fibrillary cytoplasm was present. Case 7 illustrated. 400×. E. Two of 9 cases manifested nuclear loss of ATRX; note the endothelial cell retention of the IHC (upper left), demonstrating tissue fidelity. Case 7 illustrated. 400×. F. Histone H3.F3 K27M IHC showed positive nuclear expression, paralleling mutation, in only 2/8 assessable cases. Case 7 illustrated. 400×. G. Most cases manifested no Histone H3.F3 K27M IHC expression; IHC assessment has shown good correlation with mutational status. 400×. H. Optic chiasm GBM at autopsy arising in the right optic nerve (arrowhead), with lesser involvement of the left optic nerve (arrow). Note the massive acute hemorrhage into the hypothalamus with rupture of lamina terminalis. The latter was responsible for demise 48 h after biopsy. Case 9 illustrated. I. Optic chiasm GBM demonstrating pleomorphic glial cells infiltrating nerve, albeit with preservation of fibrovascular septae. Case 8 illustrated. H&E. 100×.

None of the 7 spinal cord GBMs showed IDH1‐immunoreactivity, paralleling the work of Ellezam et al in grades II and III diffuse gliomas from spinal cord sites 4. Most supratentorial GBMs show no ATRX mutation 10 and most spinal cord GBMs in our cohort also had no ATRX mutation, as assessed by IHC retention of nuclear ATRX. The single exception was case 7 (Figure 1E) (see Table 1).

Immunostaining for H3.F3A K27M status was performed at University of California‐San Francisco; this immunostain has been shown to be sensitive and specific for the K27 mutation 11. H3.F3 K27 IHC showed nuclear immunopositivity in 2 spinal cord GBMs (Figure 1E), for comparison). Nuclear p53 expression was strong (100%, 40%) in two‐seventh cases (cases 1, 7, respectively). Two spinal cord GBMs manifested EGFR amplification (cases 1, 4), while the other 5 spinal cord cases were negative for both EGFR amplification and PTEN loss. Genetic results showed no correlation with spinal cord tumor level.

Both optic chiasmal GBMs (Figure 1H,I) were epicentered on the chiasm at presentation; tumors originating in hypothalamus but extending into chiasm terminally had been excluded from study. Neither patient with an optic chiasm GBM (cases 8, 9) showed parenchymal or cerebrospinal fluid dissemination at autopsy, nor did one autopsied spinal cord GBM patient (case 1), but the interval to demise after diagnosis was short in all 3 of these patients (48 h–26 days). Both optic chiasm tumors occurred in adult males and both had amplification of EGFR and loss of PTEN. One manifested nuclear ATRX loss (case 9), but neither showed K27M immunoreactivity/mutation. Thus, GBMs from optic chiasm appear to share EGFR amplification and loss of PTEN with supratentorial primary GBM counterparts. In contrast, only the minority (two‐seventh) of spinal cord cases had one or both of these (EGFR, PTEN) mutations.

Strong conclusions are impossible with the small case numbers. Our results in spinal cord and optic chiasm GBMs do extend those of others who have shown that adult gliomas, predominantly high grade, from anatomical midline sites such as brainstem 8 and thalamus 1 differ from their supratentorial hemispheric counterparts as well, including the presence of more frequent H3.F3A K27M mutations. As in our study, however, in neither of the studies from brainstem 8 or thalamus 1 was the presence of K27 mutation found in the majority of tested cases. Specifically, Aihara et al identified K27M mutation by Sanger sequencing in 10/20 tested thalamic gliomas from young adults 1 and Reyes‐Botero et al in two‐eighth diffuse intrinsic brainstem gliomas in adults 8. Using actual sequencing, the latter group did show one case with an additional HISTIH3B mutation, a mutation which was not assessed in our cohort. The lack of sequencing is an acknowledged limitation of our study. Nevertheless, GBMs from optic nerve and spinal cord appear to be a “mixed bag” genetically.

Note added in revision

A study of 36 spinal cord gliomas of various grades and types (18 GBMs) by Gessi et al. has recently appeared online and also demonstrates H3.F3A K27M mutations in adults and children (52%, 54%) (Acta Neuropathol 2015;130:435–437).