Massive Dissemination From Spinal Cord Gangliogliomas Negative for BRAF V600E

Abstract

Objectives

Metastatic leptomeningeal spread from spinal cord gangliogliomas (GGs) is exceedingly rare.

Methods

Two adult women, aged 27 and 51 years, died of massive disseminations of cervicothoracic GGs 4 and 6 years, respectively, after initial diagnoses; full autopsies were performed. BRAF status was assessed by VE1 immunohistochemistry (IHC), Sanger sequencing, and single-nucleotide base extension assay (SNaPshot).

Results

The 27-year-old underwent two biopsies, chemotherapy, radiation, and ventriculoperitoneal shunt placement; she developed craniospinal and peritoneal dissemination. Autopsy confirmed shunt-mediated peritoneal metastases, microscopic bone marrow involvement, and profuse spinal and supratentorial leptomeningeal and parenchymal spread. The 51-year-old underwent two resections, radiation, and chemotherapy and developed pancytopenia with biopsy-proven bony metastases 15 months before death. Autopsy demonstrated leptomeningeal, subpial, and subependymal metastases. The tumors in both primary and metastatic sites were BRAF negative by VE1 IHC and two different mutational analyses. This compared with negative BRAF results for an additional four nonmetastatic adult nonsupratentorial GGs and in our study.

Conclusions

We document two rare cases of massively metastatic spinal cord GGs in adult patients who were negative for BRAF V600E mutations via multiple methods.

Gangliogliomas (GGs) are low-grade glial-neuronal neoplasms that occur most commonly in the temporal lobe of pediatric patients.¹ They account for 6% (adult) and 27% (pediatric) of all intramedullary spinal cord neoplasms; conversely, approximately 3% of all GGs are primary to the spinal cord.^2,3 Primary spinal cord GGs usually follow a benign clinical course, with a 5-year progression-free survival rate of 67%, although aggressive behavior has been reported.^4–10 Transformation to a higher grade tumor may occur more frequently in adults.^11–13 Intracerebral, leptomeningeal, and intraventricular spread from primary spinal cord GG is exceedingly rare.⁸

BRAF c.1799T>A (p.V600E) mutations occur in 18% to 57% of GGs, although the exact rate of the mutation is unknown for primary spinal cord GGs, due to their rarity.^3,14 The immunohistochemistry (IHC) detection of the mutant BRAF protein with the VE1 monoclonal antibody has facilitated more rapid testing, and a high rate of concordance with Sanger sequencing (60/62; 97%) was shown in a recent report.¹ BRAF assessment by VE1 monoclonal antibody reactivity recently has been associated with a shortened recurrence-free survival within pediatric GGs, but these data are not known for adult GGs or specifically for spinal cord examples.¹⁵

Although dissemination from spinal cord GGs is too rare to accrue large numbers of cases, we took advantage of available BRAF status testing to assess both primary and metastatic tumor deposits from our two patients for status. The extensive autopsy information provides insights as to the extent of metastatic spread possible from spinal cord GGs, and the BRAF status information adds to the limited literature on BRAF mutational status in nonsupratentorial GGs.

Case Reports

Patient 1

This 27-year-old woman, who died in 2012, sought treatment in 2007 for left-hand paresthesias. Magnetic resonance imaging (MRI) scan demonstrated a 3.5-cm-long intramedullary mass extending from C4 to C7. Biopsy samples demonstrated a tumor composed exclusively of monotonous small round cells with scant wispy cytoplasm, embedded in an abundant mucinous matrix Image 1A. The tumor was devoid of calcification, microvascular proliferation, necrosis, neoplastic ganglion cells, or ependymal canals. The copious mucin, glomeruloid vasculature, vascular hyalinization, and subtle radial perivascular arrangements raised the consideration of ependymoma or pilocytic astrocytoma more than that of diffuse astrocytoma. Scattered mitotic figures and an MIB-1 labeling index of 8% to 9% (predilute; Ventana Medical Systems, Tucson, AZ) were indicative of anaplastic change, particularly if the tumor could be established as being astrocytic in origin. MIB-1 was assessed manually on a 1,000-cell count, using an ocular scored grid and focusing on the highest labeled area within the tumor. Glial fibrillary acidic protein (GFAP, 1:100; DAKO, Carpinteria, CA) IHC was focally positive only in areas of tumor-surrounding vessels, and synaptophysin (predilute; Ventana Medical Systems) IHC was negative. The diagnosis of glioma, or possible ependymoma, was rendered. Six months later, symptoms worsened, and an MRI scan showed enlargement of the tumor, and further resection was performed.

Subsequent larger biopsy specimens revealed hypercellularity and an MIB-1 labeling index of 14%. IDH-1 (1:40; HistoBioTec, Miami Beach, FL) was negative. IHC was again equivocal to negative for synaptophysin or neurofilament protein (clone 2F11, predilute; Ventana Medical Systems). A minute focus of tissue containing larger sized neurons could not confidently be interpreted as neoplastic vs normal anterior horn cells because of the paucity of the ganglion cells in H&E-stained sections, and the near-normal synaptophysin IHC pattern in this site did not fully meet the criteria as described by the World Health Organization (WHO) or the series of GGs by various authors.^16–18 Electron microscopy (EM) failed to identify ependymal features but showed possible neuronal differentiation. However, given the absence of definitive large neoplastic ganglion cells on light microscopy, a diagnosis of WHO grade III probable anaplastic astrocytoma was rendered.

Soon thereafter, diplopia, headache, and thoracic pain prompted additional MRI studies revealing leptomeningeal spread throughout the cervical, thoracic, and lumbar regions. Chemotherapy and radiation were begun in April 2009. Hydrocephalus developed in June 2010, necessitating ventriculoperitoneal (VP) shunt placement. By March 2011, an MRI showed an intramedullary upper cervical spinal cord tumor, as well as bulky, nodular, intracranial leptomeningeal and thoracic and lumbar spine metastases with enhancement of the nerve roots. Additional radiation therapy was administered to the posterior fossa and upper cervical spinal cord.

In April 2012, the patient was admitted to the hospital with ascites and symptoms associated with lumbosacral nerve involvement. MRI scans proximate to death showed an extensive intraparenchymal and leptomeningeal tumor in the cervical cord region Image 1B, supratentorial spread with leptomeningeal enhancement in the pontine cistern Image 1C, and massive lumbosacral leptomeningeal tumor spread. Computed tomographic studies of the abdomen and pelvis verified massive ascites but no liver, renal, or cardiac masses. She was treated with prednisone and discharged to home hospice in April 2012.

Autopsy examination revealed 3.5 L of ascites, corresponding to the premortem imaging. There was extensive gross tumor spread via the VP shunt to the omentum, pericardium, diaphragm Image 1D, and serosal surfaces of most abdominal organs, including serosa of the intestine and especially to the pelvic region near the terminus of the VP shunt that included near-complete obliteration of both fallopian tubes. Microscopically, all of these tumor deposits consisted solely of a monotonous population of small tumor cells (Image 1D), embedded in an abundant Alcian blue–positive matrix; cells were exclusively S100 immunoreactive, similar to her premortem spinal cord biopsy specimens. Microscopically, the bone marrow was focally positive for tumor histologically similar to the small cell population (Image 1D).

Neuropathologic examination showed massive parenchymal involvement of the cervical spinal cord by GG, thick leptomeningeal tumor deposits with adhesion of the dura overlying the dorsum of the cord, and significant expansion of spinal nerve roots by GG tumor infiltrates Image 1E. Meninges at the base of the brain were opacified, with gelatinous tumor encasement of the optic nerves and chiasm as well as all cranial nerves. Obstruction of the foramina of Luschka and Magendie correlated with obstructive hydrocephalus and ventricular dilatation. There were also extensive tumor deposits throughout the chiasmatic cistern (Image 1E).

Microscopically, a small blue/mucin-rich tumor involved the leptomeninges at all spinal cord levels and the cord parenchyma throughout the cervical cord; the lower thoracic and lumbosacral cord, however, were spared of parenchymal involvement. Parenchymal cord tumor deposits formed nodular, discrete, subpial foci within the cervical cord.

However, unlike premortem biopsy specimens or metastatic deposits, the cervical spinal cord immediately rostral and caudal to the original biopsy sites contained much smaller, additional discrete, separate foci of tumor, which were now diagnostic for GG. These were located within discrete foci in the dorsal column (Image 1E) composed of large ganglion cell–rich areas predominantly contralateral to sites of the original biopsy specimens and small cell/mucin–rich areas of tumor. The cervical cord demonstrated demarcated parenchymal tumor (Image 1E), leptomeningeal thickening and dural adhesion especially dorsally, and expansion of spinal nerve roots by tumor (Image 1E). There was also tumor within the ventricular walls (Image 1E), which was composed of irregularly shaped and occasionally binucleate ganglion cells, establishing their neoplastic nature Image 1F. The tumor was recognized only at autopsy to be anaplastic GG, not anaplastic astrocytoma.

The leptomeningeal and subependymal tumor deposits in the cerebrum, brainstem, and cerebellum were all exclusively composed of small cell/mucin–rich tumor (Image 1D). Individual small tumor cells infiltrating cranial nerves were also identified.

Patient 2

This 51-year-old woman, who died in 2008, had first sought treatment for weakness and numbness of the lower extremities in 2002. An MRI and subsequent surgical excision led to the diagnosis of cervicothoracic spinal cord GG, WHO grade I. Most of the biopsy specimen was composed of a nodular mass containing numerous, closely juxtaposed, large dysmorphic ganglion cells Image 2A. This component was sharply demarcated from a much smaller volume of tumor composed of small cells embedded in mucin. The tumor ganglion cells were immunoreactive for chromogranin (1:50; BioCare, Concord, CA) and surrounded by perisomal synaptophysin IHC immunoreactivity that further highlighted the extremely large numbers of these neoplastic ganglion cells. The discrete small cell component was minimally immunoreactive for synaptophysin but strongly immunoreactive for S100 protein (predilute; Ventana Medical Systems) and focally for GFAP. Clinically progressive symptoms prompted further tumor resection in 2004; the tumor was histologically similar and again diagnosed as ganglioglioma, WHO grade I. She received radiation and chemotherapy; pancytopenia prompted a premortem bone marrow biopsy in August 2007. The bone marrow demonstrated cords of small S100-positive cells embedded within a sclerotic desmoplastic background Image 2B. An aspirate smear showed small cells with uniform nuclear features and focal rosette-like aggregation (Image 2B, inset). Metastatic GG was diagnosed. She then developed stage II breast cancer and underwent a left mastectomy with lymph node dissection in 2007 at an outside hospital. By report, two lymph nodes were positive, and she was treated with tamoxifen.

By 2008, she had progressive lower extremity weakness and paresthesias, and an MRI showed multifocal spinal cord–enhancing parenchymal tumor, with cerebrospinal fluid (CSF) seeding Image 2C. Unusual T1 low-intensity lesions were seen in the basis pontis, with one area noted in a subpial location Image 2D. In November 2008, she died of aspiration pneumonia.

Autopsy examination revealed widely disseminated metastatic tumor involving lungs, lymph nodes, liver, adrenals, kidneys, spleen, pancreas, heart, multifocal bone sites, and small bowel. The right breast was extensively sampled, but no tumor was identified. Microscopically, a highly malignant tumor composed of glands with cribriform architecture suggested breast carcinoma, but unlike typical mammary carcinoma, it showed strong and diffuse cytoplasmic immunoreactivity for S100 protein, focal immunoreactivity for synaptophysin, and negative immunostaining for CK7 (1:300; DAKO) or BRST-2 (predilute; Ventana Medical Systems) on multiple sections. Postmortem EM examination of two different lung and liver metastases revealed desmosomes and proved an epithelial origin for the tumor, negating any consideration that these systemic metastases were from the GG. No features of melanoma or GG were seen on EM. The diagnosis was neuroendocrine carcinoma, likely of breast origin, a highly aggressive mammary carcinoma.¹⁹ Unfortunately, the original slides and blocks from her premortem left mastectomy performed at the outside hospital were not available for reexamination. No typical breast adenocarcinoma could be identified anywhere in the autopsy. Two small hemorrhagic metastases of this neoplasm were also identified in the cerebrum and cerebellum.

Neuropathologic gross examination showed near-holocord parenchymal involvement by GG, with significant distortion and blurring of spinal cord landmarks, expansion of the cord diameter, and gross sparing only of the lumbosacral cord regions. Numerous small discrete periventricular/subependymal tan tumor nodules studded the third, fourth, and lateral ventricles. Although leptomeningeal opacification was not appreciated grossly, there were focal subpial discrete gelatinous masses in the basis pontis parenchyma Image 2E and the cerebral cortex. These were histologically composed of small cells embedded in a mucin-rich stroma (Image 2E), similar to the first patient’s small cell tumor metastases.

All spinal cord levels except the most caudal were occupied by GG. There were several tumor nodules composed of small cell/mucin–rich cells, while other tumor nodules were exclusively composed of large dysmorphic ganglion cells. Both of the GG components were present at virtually all spinal cord levels and nearly equal in amount but with little admixture.

The same striking histologic dichotomy was seen in tumor metastases in non–spinal cord central nervous system sites. Supra- and infratentorial periventricular tumor deposits contained exclusively large dysmorphic ganglion cells, while parenchymal tumor metastases, including basis pontis and cerebral cortex tumor metastases, were composed exclusively of small cell/mucin–rich, S100-immunopositive tumor cells; MIB-1 was less than 5%. Small tumor cells permeated nearby leptomeninges, suggesting that parenchymal supra- and infratentorial subpial deposits were CSF disseminated.

BRAF Assessment of Primary and Metastatic Tumor Deposits in Patients 1 and 2

One primary spinal cord and one metastatic tumor deposit from each patient were assessed for BRAF mutations by Sanger sequencing as well as for mutant BRAF protein via IHC VE1 antibody. Care was taken to test both the large ganglion cell–rich and the small blue/mucin-rich portions of each GG. The metastatic spinal cord tumor assessed from patient 1 consisted solely of the small cell/mucin–rich element devoid of neoplastic large ganglion cells, while the spinal cord sample from patient 2 was almost exclusively composed of the latter. Metastatic tumor deposit assessed from patient 1 was composed of the pure small cell/mucin–rich element and taken from a systemic omental metastasis; from patient 2, a similar-appearing metastatic small cell/mucin–rich tumor metastasis from the cerebral hemisphere was assessed.

Since previously published studies on BRAF mutational testing have included a paucity of adult nonsupratentorial GGs,³ we also tested four adult non–cerebral hemispheric GGs (two nonmetastatic spinal cord GGs and two nonmetastatic posterior fossa GGs) for both BRAF mutational and IHC studies. Additional comparison cases consisted of a large cohort of 22 pediatric GGs (12 brainstem and 10 nonbrainstem) assessed by Sanger sequencing and VE1 IHC.

Methods for BRAF mutational analyses as assessed by Sanger sequencing have been previously published in detail by our group.^20,21 Single-nucleotide base extension (SNaPshot) assessment for BRAF c.1799T mutation, an assay with higher analytic sensitivity compared with Sanger sequencing, was able to be performed on the subset of cases with sufficient DNA with a methodology identical to previously published methods.^22,23 Immunostaining for BRAF VE1 antibody used a methodology similar to previously published methods.¹

Neither the two metastatic GGs in this study nor the additional four adult non–cerebral hemispheric GGs showed the BRAF mutation by Sanger sequencing. The metastatic deposits, the initial biopsy specimens from patient 1, and the comparison spinal cord GGs (microdissected for optimal selection of the large neoplastic ganglion cells) were able to be further assessed by SNaPshot and were confirmed as negative; other samples could not be further assessed by SNaPshot due to the limited quantity of available extracted nucleic acid. In contrast, V600E-mutated BRAF was identified by Sanger sequencing in eight (67%) of 12 pediatric brainstem GGs and five (50%) of 10 nonbrainstem pediatric GGs. Thus, regardless of which cell population was detected or targeted, no BRAF mutation was found in primary or metastatic spinal cord GGs in this study.

VE1 immunoreactivity was not identified in large dysmorphic neoplastic ganglion cells Image 2F or in the small cell/mucin–rich component in either patient 1 or 2 or in the four adult nonmetastatic, nonsupratentorial GG cases. In contrast, strong VE1 immunostaining was found in eight (67%) of 12 pediatric brainstem GGs (representative example in Image 2F) and five (50%) of 10 nonbrainstem pediatric GGs, with perfect correlation in the controls between VE1 IHC and positive BRAF mutational status in the control group.

Discussion

Three interesting clinicopathologic points emerge from these cases. First, both of our spinal cord GGs had two discrete, separate components to their tumors: cohesive groups of large dysmorphic ganglion cells and small monotonous, S100-positive cells embedded in abundant Alcian blue–positive mucinous matrix. In patient 1, the small cell/mucin–rich component was more abundant in the spinal cord parenchyma and was solely responsible for the extensive spinal cord/supratentorial/infratentorial leptomeningeal deposits and serosal metastases in the peritoneal cavity, the latter in relationship to the VP shunt, as well as microscopic bone marrow involvement. Large dysmorphic ganglion cells were extremely focal and found only at autopsy. This component is histologically necessary for a confident diagnosis of GG but had not been present in two premortem biopsy specimens, probably due to sampling artifact. In patient 2, the small cell/mucin–rich component produced symptomatic bone marrow metastases, and the two dichotomous components of the GGs were shown to have metastasized to separate intracranial sites at autopsy.

Second, there has been debate if the large neoplastic ganglion cell component in GGs is truly neoplastic or if only the small cell, often glial, portion can be so considered; this was recently addressed by Koelsche et al.¹ The fact that the CSF metastatic intraventricular deposits in patient 2 were composed solely of the large dysmorphic ganglion cells strongly suggests that, indeed, these are a true neoplastic component of GG.

Third, neither the four adult nonsupratentorial, nonmetastatic GG cases nor the two metastatic adult spinal cord GGs in our study showed BRAF mutation or VE1 immunoreactivity, even with the more sensitive method of the single-nucleotide base extension assay (SNaPshot). This compares with the higher percentage of BRAF mutation and VE1 immunoreactivity seen in more than half of the 22 pediatric patients in our study.

The two adult metastatic vs four nonmetastatic cases are clearly too few to draw any conclusions about relationships between BRAF mutational status and tumor behavior, but our two metastatic adult cases and four nonmetastatic adult GGs add to the numbers in the literature of spinal, nonsupratentorial GGs examined. In the largest survey study to date of 1,320 tumors, five cerebellar, three brainstem, and two spinal cord GGs were available for BRAF mutational analysis by Sanger sequencing, with two of 10 positive.³ Restricting molecular studies of neuroglial tumors by specific anatomical location may be important to further elucidate the biology of these tumors. An example of this is the low rate of IDH-1 immunostaining in spinal cord diffuse astrocytomas compared with their supratentorial counterparts.²⁴ A corollary to this is that, although IDH-1 IHC can assist in the differential diagnosis of supratentorial GGs from diffuse infiltrating gliomas that encompass normal neurons,²⁵ this is not possible in spinal cord GGs, given the IDH-1–negative status of most spinal cord astrocytomas. Similarly, it appears that BRAF mutation and the finding of negative VE1 immunoreactivity will not likely be diagnostically helpful in the spinal cord in distinguishing astrocytomas from GGs. Although BRAF assessment by VE1 monoclonal antibody reactivity recently has been associated with a shortened recurrence-free survival within pediatric GGs,¹² BRAF status may not relate to prognosis in spinal cord GGs. Clearly, larger numbers of cases will have to be accrued and tested before definitive conclusions, however, can be drawn.

Image 1.

A, Light microscopic appearance of initial spinal cord biopsy specimen of patient 1 showing relatively monotonous, small cells embedded in abundant mucin (H&E; ×200). B, Sagittal magnetic resonance imaging (MRI), with gadolinium enhancement, from patient 1 shows parenchymal cervical and leptomeningeal tumor spread (arrows). C, Axial MRI, with gadolinium, from patient 1 verifies leptomeningeal metastatic spread, with linear leptomeningeal enhancement surrounding the basis pontis (arrows). D, Autopsy from patient 1 shows (top) widespread peritoneal metastatic ganglioglioma tumor spread with white patches of tumor (arrows) throughout the abdominal serosal surfaces, histologically (bottom) composed of small cells embedded in rich mucin (H&E; ×400). E, Cervical cord (top), whole-mount section, from patient 1 shows discrete small blue cell/mucin–rich tumor nodules (black arrows), as well as a separate nodule containing large dysmorphic ganglion cells in the dorsal column (outlined by arrowheads); note massive spinal nerve expansion by tumor (red arrow) (H&E). Optic nerve region (bottom) from patient 1, on whole-mount section, shows small cell/mucin–rich leptomeningeal tumor at the base of brain (arrow left) as well as periventricular/subependymal tumor in the third ventricle (arrow right) (H&E. F, Cervical cord tumor from patient 1, taken from the region outlined by arrowheads in E (top), contained scattered, dispersed dysmorphic and binucleate neoplastic ganglion cells of varying sizes and shapes; these were identified only at autopsy and then allowed a correct diagnosis of ganglioglioma (GG), not anaplastic astrocytoma, now that the classic GG component was present (H&E; ×400).

Image 2.

A, Initial spinal cord biopsy specimen from the easily diagnosed ganglioglioma, World Health Organization grade I, in patient 2 shows the discrete nature and sharp demarcation (arrows) between the eosinophilic areas containing numerous large dysmorphic ganglion cells (inset) and the small cell/mucin–rich tumor in this patient (H&E; ×100, inset ×600). B, Spinal cord metastases were diagnosed in a bone marrow biopsy specimen 15 months before death in patient 2; note the cords of small tumor cells embedded in fibrotic, trichrome-rich, and desmoplastic matrix (Masson’s trichrome stain; ×200). Inset: aspirate smear revealed the small tumor cells to have cytologically bland, uniform nuclei and wispy cytoplasm, with focal rosette-like formation (H&E; ×600). C, Sagittal magnetic resonance imaging (MRI), with gadolinium, of patient 2 spinal cord prior to death shows widespread multifocal spinal cord involvement by tumor (arrow). D, Sagittal MRI, with gadolinium, revealed low-intensity signals in the basis pontis in patient 2 (arrow); etiology was unclear premortem, but postmortem, this corresponded to parenchymal tumor metastases of the small cell/mucin–rich tumor component. E, Autopsy section of the pons (top) shows white parenchymal tumor near the pial surface; compare with the premortem neuroimaging of this patient in D. Whole mount (H&E). Parenchymal tumor metastases (bottom) of the small cell/mucin–rich tumor component (H&E; ×400). F, VE1 immunohistochemistry from the original spinal cord ganglioglioma biopsy specimen (top). This case was also negative for BRAF mutation by Sanger sequencing. Immunostaining for BRAF VE1 with light hematoxylin counterstain (×600). VE1 immunohistochemistry in a comparison control non–spinal cord pediatric ganglioglioma (bottom) is strongly immunopositive in both large dysmorphic ganglion cells and the background neuropil of the tumor. This case was also positive for BRAF mutation by Sanger sequencing. Immunostaining for BRAF VE1 with light hematoxylin counterstain (×400).