Oligodendroglioma with Neuronal Differentiation in Two Boxer Dogs

Abstract

Two cases of high-grade glioma comprising sheets of oligodendroglial cells multifocally disrupted by regions of remarkable neuronal differentiation are described. These tumours morphologically resemble ‘oligodendroglioma with ganglioglioma-like maturation’, a rare tumour of man, but appear to be phenotypically more aggressive. Neuronal markers (synaptophysin, neuron-specific enolase and βIII-tubulin) effectively highlight neuronal elements within these tumours and could potentially help to further investigate the prevalence and biological significance of neuronal differentiation in canine oligodendroglioma.

Oligodendroglioma, after meningioma, is the second most common primary intracranial tumour of dogs (Vandevelde et al., 2012). A distinct predilection for oligodendroglioma is seen in boxer dogs, Boston terriers and other closely-related breeds (Snyder et al., 2006; Vandevelde et al., 2012; Song et al., 2013; Dickinson et al., 2016; Koehler et al., 2018). Despite the relatively high incidence of canine oligodendroglioma and a strong interest in canine glioma as a model for human disease, morphological variants of this tumour entity are reported rarely (Kovi et al., 2013; Fadda et al., 2014; Rissi et al., 2015; Koehler et al., 2018). In this communication, we describe two cases of oligodendroglioma that exhibited extensive neuronal differentiation.

Case 1 was a 5-year-old neutered male boxer dog with an acute onset of generalized seizures that were treated with the oral extended release anticonvulsant levetiracetam (UCB Pharmaceuticals, Atlanta, Georgia, USA). Magnetic resonance imaging (MRI) revealed an intra-axial T1W-hypointense and T2W-hyperintense ring contrast-enhancing mass extending from the left olfactory lobe to the rostral aspect of the left thalamus. Gross total resection was performed and a sample of the surgically resected tumour was examined microscopically (Supplementary Figs. 1–3). Two populations of neoplastic cells were observed. The majority of the neoplastic cells were arranged in sheets and had the classic oligodendroglial ‘fried egg’ appearance with small round heterochromatic to moderately basophilic nuclei surrounded by clear perinuclear halos. A second, morphologically distinct neoplastic cell population with round eccentrically located nuclei, prominent nucleoli and a variable amount of eosinophilic cytoplasm tended to form multilayered ribbon-like structures. Both neoplastic cell populations were strongly immunoreactive for oligodendrocyte transcription factor 2 (olig2); however, synaptophysin immunoreactivity was observed only in the second population of neoplastic ‘neuronal’ cells (Supplementary Figs. 1–3 and Table 1). Mitoses in the oligodendroglial population averaged 1 per 10 high-power (×400) fields (HPFs), and no mitoses were observed in the neuronal population. Prominent glomeruloid microvascular proliferation and regions of necrosis indicated that the tumour was high grade.

Table 1.

Antibodies and immunohistochemical findings in two cases of canine oligodendroglioma with prominent neuronaldifferentiation

Antibody	Source	Dilution	Case 1		Case 2
			Biopsy	PM	Biopsy *	PM
Olig2 (RP)	Millipore, Burlington, Massachusetts, USA	1 in 1,000	4/4	4/2	4	4/2
SYN (MM)	Agilent, Santa Clara, California, USA	1 in 1,000	0/3	0/0†	ND	0/0†
GFAP (RP)	Agilent	1 in 1,000	1/1	0/1	ND	0/0
NSE (MM)	Agilent	1 in 200	1/1	2/4	ND	1/3
βIII (RP)	Abcam, Cambridge, UK	1 in 1,500	2/4	1/3	ND	1/4
MAP2 (CP)	Abcam	1 in 4,000	2/3	2/3	ND	2/2
CD34 (RP)	Abcam	1 in 500	ND	0/0	ND	ND
Nestin (RP)	Thermo Fisher Scientific, Waltham MA, USA	1 in 200	0/1	ND	ND	0/1
SOX2 (RP)	Novus Biologicals, Centennial, Colorado, USA	1 in 500	4/4	ND	ND	4/4

For immunohistochemistry (IHC), 4 μm formalin-fixed and paraffin wax-embedded sections of tissue were dewaxed and rehydrated, followed by antigen retrieval in a steamer prior to IHC performed with a Dako Autostainer (Dako, Carpinteria, California, USA). Detection ofbound primary antibody was achieved using the Dako EnVision + System-HRP kit, with DAB or AEC (for GFAP and NSE) as the chromogen.

Olig2, oligodendrocyte transcription factor; SYN, synaptophysin; GFAP, glial fibrillary acidic protein; NSE, neuron-specific enolase; βIII, βIII-tubulin; MAP2, microtubule-associated protein 2; SOX2, sex-determining region-box 2; RP, rabbit polyclonal; MM, mouse monoclonal; CP, chicken polyclonal; PM, post mortem; ND, not done.

Immunohistochemical immunoreactivity scores: presented as oligodendroglial/neuronal cells; 0, none; 1, minimal; 2, mild; 3, moderate; 4. marked.

^*Biopsy comprised tumour-adjacent neuropil with a single infiltrating neoplastic cell population.

^†Immunoreactivity in surrounding normal neuropil (internal control) was minimal to absent.

The dog was enrolled in an immunotherapy trial using autologous tumour lysate vaccine with imiquimod as the adjuvant and novel checkpoint blockade (CD200) inhibitor as per translation study protocol (Zhengming et al., 2019). Progressive neurological signs (e.g. circling, ataxia, pacing and stupor) and MRI evidence of multifocal tumour recurrence were noted 3 months post resection. Oral temozolomide therapy (150 mg/m² for 5 days in 28-day cycles; Merck, Whitehouse Station, New Jersey, USA) was initiated, but 1 month later the owner requested humane destruction due to progressive neurological decline. At post-mortem examination, a variably demarcated, gelatinous, pale tan tumour filled the left lateral ventricle and multifocally effaced extensive regions of the right and left cerebral cortex, caudate nuclei, hippocampus, cerebellum and brainstem (Fig. 1).

Fig. 1.

Oligodendroglioma with neuronal differentiation in a boxer dog; case 1. Coronal section of formalin-fixed brain at the level of the lateral and third ventricles. A gelatinous mass multifocally replaces the cerebral cortex and fills the lateral ventricles. The arrowheads indicate areas of grossly-apparent tumour. Bar, 2 cm.

Microscopically, the post-mortem tumour specimen recapitulated the majority of the features described in the surgically resected sample. Broad sheets of oligodendroglial cells (Fig. 2 and Supplementary Figs. 4–9) were multifocally disrupted by ribbons of morphologically distinct neuronal cells, as described above. A novel feature of the post-mortem sample was the presence of several small well-demarcated islands of primitive neural tissue comprising clusters of the neuronal cells surrounded by a neuropil-like matrix (Fig. 3 and Supplementary Figs. 4–9). Compared with the surgically resected tumour, mitotic activity in the oligodendroglial cell population was brisk with mitoses averaging 10 per 10 HPFs; however, as before, no mitoses were seen in the neuronal regions. Prominent serpiginous necrosis with pseudopalisading of nuclei was noted in several regions of the post-mortem sample.

Fig. 2.

Oligodendroglioma with neuronal differentiation in a boxer dog; case 1. Sheets of oligodendroglial neoplastic cells with small round heterochromatic nuclei and perinuclear halos. Mitoses (arrowheads) are relatively frequent. HE. Bar, 100 μm.

Fig. 3.

Oligodendroglioma with neuronal differentiation in a boxer dog; case 1. Island of neuropil development. Clusters of pleomorphic neoplastic neuronal cells with large round, eccentrically located nuclei and abundant eosinophilic cytoplasm on a background of neuropil-like matrix. HE. Bar, 100 μm.

The second case was a 9-year and 9-month-old neutered female boxer dog with acute onset of generalized seizures initially controlled with phenobarbital, but levetiracetam had to be added to control breakthrough seizure activity. Pre-surgical workup revealed slightly elevated serum alkaline phosphatase concentration (consistent with phenobarbital administration) and signs of arrhythmogenic right ventricular ‘boxer dog’ cardiomyopathy. On MRI, an intraaxial T1W-hypointense, T2W-hyperintense non-contrast-enhancing tumour was identified in the rostral tip of the right temporal lobe and piriform cortex, which invaded the thalamus and midbrain. The tumour was surgically resected from the temporal lobe, but residual tumour was left in the thalamus and midbrain. Histopathological examination of the biopsy sample taken from the tumour margin showed neuropil infiltrated by neoplastic olig2-immunoreactive cells (Table 1).

The dog was enrolled in the immunotherapy trial described above and was monitored by the referring veterinary neurologist. Follow-up imaging was aborted due to cardiac issues under anaesthesia. Seven months post resection the dog began to show neurological signs, which progressed within 1 month to stupor and collapse. The owner requested humane destruction and the referring neurologist shipped the formalin-fixed brain for pathological examination. An approximately 3 × 2.5 × 2.5 cm soft friable tan mass with multifocal dark red–brown discolouration replaced much of the right temporal lobe and midbrain.

Similar to case 1, sheets of neoplastic cells with a classical oligodendroglial appearance were multifocally disrupted by ribbon-like structures lined by multiple layers of cells with abundant eosinophilic and strongly βIII-tubulin-immunoreactive cytoplasm (Table 1 and Fig. 4). These ribbons of neuronal differentiation were surrounded by prominent glomeruloid microvascular proliferation, regionally extensive necrosis and multiple vessels occluded by organizing fibrin thrombi. Unlike case 1, no islands of primitive neural tissue were observed. Mitoses in the oligodendroglial population were frequent (20 per 10 HPFs) and often exhibited atypical morphology. No mitoses were observed in the neuronal cells.

Fig. 4.

Oligodendroglioma with neuronal differentiation in a boxer dog; case 2. Multilayered ribbons of strongly βIII-tubulin-immunoreactive cells are present in regions of tumour exhibiting marked glomeruloid microvascular proliferation (asterisks). IHC. Bar, 100 μm.

In this report, we describe two cases of high-grade oligodendroglioma with prominent neuronal differentiation. We do not believe that the neuronal differentiation seen in these cases was related to the experimental immunotherapy received by the dogs, because neuronal components in case 1 were identified in the treatment-naïve biopsy sample. A prior report in the veterinary literature documenting ‘neurocytic’ differentiation in an oligodendroglioma in a dog (Rissi et al., 2015) described a single cluster of morphologically distinct synaptophysin immunoreactive cells within an intraventricular oligodendroglioma. By comparison, the degree of neuronal differentiation in the current cases was far more extensive than previously described. This is especially true in case 1, where islands of tissue resembling immature brain were sharply demarcated from the surrounding sheets of oligodendroglial cells. These areas of ‘brain-like’ development were similar to foci that have been described in ‘oligodendroglioma with ganglioglioma-like maturation’, a rare human brain tumour (Perry et al., 2010). Traditionally, oligodendrogliomas are believed to originate from an oligodendroglial precursor cell; however, the occurrence of neuronal differentiation in cases of human oligodendroglioma has led to novel tumour ontogeny studies in genetically modified mice (Colin et al., 2006; Zong et al., 2015). The apparent existence of neuronal differentiation in canine oligodendroglioma makes it possible that these tumour ontogeny studies are also relevant to canine glioma biology.

In human medicine, although rarely seen, areas of neuronal differentiation, especially ganglioglioma-like maturation within an oligodendroglioma, can pose a significant diagnostic challenge, especially when only small surgically resected sections are available for examination (Wharton et al., 1998; Perry et al., 2002, 2010; Hirose et al., 2013). In order to provide appropriate patient care, it is vital to differentiate generally benign tumours such as neurocytoma, gangliocytoma and ganglioglioma from oligodendroglioma. In veterinary medicine, brain tumour biopsy is becoming more frequent, meaning that veterinary pathologists will be faced with similar diagnostic challenges, albeit with fewer molecular tools to aid in making an accurate tumour diagnosis.

The majority of human oligodendrogliomas exhibit chromosomal aberrations, usually 1p, 19q co-deletion (Perry et al., 2010; Louis et al., 2016). These codeletions have been shown to exist in both cell populations in oligodendroglial tumours exhibiting neuronal differentiation (Perry et al., 2010; Hirose et al., 2013). Although syntenic deletion events have been described in canine oligodendroglioma, the prevalence of these deletions in canine glioma has not been determined, and rapid diagnostic tests to identify these chromosomal aberrations are not available (Dickinson et al., 2016). In the absence of molecular tests, cell morphology, immunohistochemistry, clinical suspicion and the diagnostic experience of the pathologist continue to be important for accurate diagnosis.

Diagnosis of oligodendroglioma on post-mortem specimens is generally considered straightforward, since large sections of tumour are available for examination, and the histomorphology of oligodendroglioma is well-defined. Classically, the neoplastic cells in oligodendroglioma have a ‘fried egg’ appearance and form broad sheets subdivided by delicate branching blood vessels resembling ‘chicken wire’ (Vandevelde et al., 2012; Louis et al., 2016; Koehler et al., 2018). However, the characteristic fried egg appearance used to identify oligodendroglial cells is an artefact due to delayed fixation, which is absent in frozen and promptly fixed specimens (Louis et al., 2016). Without this artefact, differentiating oligodendroglioma from other tumours such as diffuse astrocytoma can be difficult (Vandevelde et al., 2012; Louis et al., 2016; Spitzbarth et al., 2017; Koehler et al., 2018).

For veterinary pathologists examining tumours with ambiguous morphology, immunohistochemistry remains the primary tool to differentiate oligodendroglioma from astrocytoma, other gliomas and round cell tumours. Oligodendrogliomas generally exhibit diffuse and intense labelling with olig2 and, when well-preserved, can display CNPase immunoreactivity (Koehler et al., 2018). By comparison, astrocytomas are usually less olig2 immunoreactive and do not exhibit CNPase immunoreactivity (Koehler et al., 2018). Glial fibrillary acidic protein (GFAP) is a useful astrocytic marker in human cases of astrocytoma; however, labelling of canine tumours is unreliable (Spitzbarth et al., 2017; Koehler et al., 2018).

The neuronal portions of the tumours in the current report exhibited mild to moderate olig2 immunoreactivity and were not GFAP immunoreactive (Table 1 and Supplementary Figs. 4–9). To further characterize these morphologically distinct neoplastic cells, we performed immunohistochemistry with several antibodies targeting neural differentiation markers, including: synaptophysin, neuron-specific enolase (NSE), MAP-2, βIII-tubulin (Table 1 and Supplementary Figs. 4–9), and the stem cell markers nestin, sox2 and CD34. These markers have been used routinely by human pathologists and were validated on normal canine control tissues in our laboratory (Table 1). We found synaptophysin, NSE and βIII-tubulin immunoreactivity to be specific to the neuronal neoplastic population within the biopsy specimen. However, synaptophysin immunoreactivity, as demonstrated in the current cases, is highly sensitive to prolonged fixation and appears to only provide diagnostic value in well-preserved specimens. The MAP-2 antibody exhibited a high level of background, labelled tumour-associated macrophages, and failed to differentiate the two neoplastic cell populations. The sox2 antibody intensely and diffusely labelled both groups of neoplastic cells, while the nestin and CD34 antibodies exhibited little to no immunoreactivity. Based on these findings, we concluded that synaptophysin, NSE and βIII-tubulin are useful markers for identifying neuronal differentiation in canine glioma and applied them to a subsample of 20 archived formalin-fixed, paraffin wax-embedded specimens of oligodendroglioma. We identified neuronal elements in 11 cases of canine oligodendroglioma, but could not determine whether these neuronal elements represented a neoplastic subpopulation or were simply entrapped neurons.

The two current cases exhibited morphological features that appear to be unique to canine glioma when compared with the human disease (Koehler et al., 2018). For instance, although the canine tumours spread throughout the grey matter of the cerebrum and invaded the ventricles, infiltration of neoplastic cells into the white matter was limited. Additionally, as is common in canine oligodendroglioma, these tumours had wide expanses of necrosis and prominent glomeruloid microvascular proliferation consistent with a highly aggressive tumour phenotype. These features are less common and, when present, less pronounced in human oligodendroglioma. The advanced disease state at the time of detection and the common practice of humane destruction in companion animals make direct comparison between human and canine glioma challenging. However, in our experience canine oligodendroglioma behaves more aggressively than its human counterpart, which may point to a more primitive tumour precursor cell with a greater degree of pluripotency.

The clinical and biological significance of neuronal differentiation in canine oligodendroglioma are, as yet, unclear. The same is true of human oligodendroglioma with ganglioglioma-like maturation, which has primarily been described as a potential diagnostic challenge for pathologists examining surgical biopsy samples (Perry et al., 2010). Canine brain tumour biopsies and advanced cancer therapies are becoming more common, and veterinary pathologists are being asked to make increasingly challenging diagnoses while laying the groundwork for ongoing efforts to characterize molecular signatures of canine brain tumours.