COX-2 and c-kit expression in canine gliomas


J. M. Jankovsky
College of Veterinary Medicine
University of Tennessee
2407 River Drive Knoxville,
TN 37996-4542, USA


Gliomas are among the most common primary neural tumours of dogs. Cyclooxygenase-2 (COX-2) and c-kit overexpression are associated with increased aggressiveness of gliomas and decreased survival in human beings. COX-2 is the inducible form of cyclooxygenase, which catalyzes prostaglandin formation and may increase tumour proliferation and angiogenesis. C-kit is a tyrosine kinase receptor involved in normal cell physiology; c-kit is upregulated in some canine tumours. In this retrospective study, 20 canine gliomas were identified: 11 (55%) oligodendrogliomas, including 1 anaplastic variant; 1 (5%) oligoastrocytoma; and 8 (40%) astrocytomas, of which 2 were glioblastoma multiforme. None of the gliomas expressed COX-2. None of the gliomas were immunoreactive for c-kit, although all three high-grade tumours had intramural vascular expression. Consequently, COX-2 inhibitors would likely be ineffective against canine gliomas. C-kit inhibitors may have an anti-angiogenic effect in high-grade gliomas, but would likely be ineffective in low- and medium-grade tumours.


Gliomas are among the most common primary neural tumours of dogs, with boxers and Boston terriers at increased risk.1 The incidence of primary central nervous system (CNS) tumours is estimated at 14.5–30.0 per 100 000 dogs, with 20–36% of those being of glial origin.1,2 Classification of glial tumours as either astrocytoma or oligodendroglioma is based on histological characteristics and strong immunohistochemical detection of glial fibrillary acidic protein (GFAP) in astrocytomas, although the presence or absence of this immunomarker is not definitive. The veterinary World Health Organization (WHO) classification scheme further divides astrocytomas into low, medium and high grade [glioblastoma multiforme (GBM)]; oligodendrogliomas are classified as low or high grade (anaplastic).3 Mean age of dogs at diagnosis is 8.1 ± 3.1 years for oligodendrogliomas, 8.4 ± 4.3 years for GBM and 8.6 ± 3.3 years for astrocytomas, with most tumours located in the prosencephalon.1 Risk increases with age, but no significant gender predisposition has been noted.1,4,5 The prognosis is poor, with median survival for astrocytomas and oligodendrogliomas at 78 days and 73 days, respectively.6 Treatment generally consists of symptomatic therapy (corticosteroids and anticonvulsants), surgical debulking, radiotherapy and/or chemotherapy, usually with nitrosourea compounds.7,8

In human beings, glial tumours comprise 31% of CNS tumours, with GBM being the most frequent form; the incidence of human CNS tumours is 6.59 per 100 000 people.9,10 Gliomas are more common in Caucasians and males, and risk increases with age.11 Mean survival time for astrocytomas ranges from 0.4 year for GBM to 1.6 years for anaplastic variants, to over 10 years for low-grade tumours. For oligodendrogliomas, mean survival time ranges from 3.5 years for high grade to 11.6 years for low grade.12 Human and canine gliomas have similar morphology, imaging characteristics, histologic features and immunocytochemistry expression.5,13–15

Cyclooxygenase-2 (COX-2) is the inducible form of prostaglandin H synthase (cyclooxygenase), which catalyzes the conversion of arachidonic acid to prostaglandins. Overexpression of COX-2 has been shown to promote neoplastic cell proliferation and angiogenesis and decrease apoptosis.16 Strong COX-2 positivity has been noted in a variety of primary canine tumours, including meningiomas, mammary tumours and colorectal tumours.17–19 Several studies have found upregulation of COX-2 in human gliomas to be a strong negative prognostic factor, with increasing expression in higher-grade tumours.20–29 COX-2 inhibitors decrease cell proliferation and angiogenesis and increase tumour sensitivity to radiation in many tumour types, including human glial cell lines.16,25,30–34 If COX-2 is expressed in canine gliomas, similar treatment modalities may be effective.

C-kit (CD117) is a tyrosine kinase receptor involved in normal cell physiology, which is overexpressed in several canine neoplasms, including mast cell tumours and gastrointestinal stromal tumours.35,36 Some studies of human glial tumours associated increased c-kit positivity with higher grade and decreased survival.37–39 Preclinical and clinical studies have indicated that tyrosine kinase inhibitors, such as imatinib, show promise as targeted adjunct treatment for human gliomas.40 However, no similar data exists for canine gliomas.

Materials and methods

Pathology records from 2001 to 2011 at the University of Tennessee College of Veterinary Medicine (UTCVM) and Tifton Veterinary Diagnostic and Investigational Laboratory (TVDIL), University of Georgia, yielded 26 canine glial tumour samples, of which 20 met the inclusion criteria of this study. Cases were included if they were definitively identified as gliomas and if sufficient tissue was available for immunohistochemical processing.

All slides were reviewed by two board-certified pathologists (K. M. N., S. J. N.) to determine the type of glioma (oligodendroglioma, astrocytoma and mixed) based on histology. Additionally, immunohistochemistry (IHC) for GFAP was performed to aid confirmation of cell of origin. A representative 5 µm section was cut from a formalin-fixed, paraffin-embedded tissue block for each case and placed on a charged glass slide. Slides were deparaffinized with xylene and rehydrated in an ethanol series. All slides were stained for GFAP, COX-2 and c-kit. COX-2 and c-kit slides were heated in pH 6.0 citrate buffer (Dako, Carpinteria, CA, USA) for 25 min at 95 °C for antigen retrieval and then cooled for 20 min (antigen retrieval was not needed for GFAP slides). Processing was performed on an autostainer (Dako, S3400). All slides were treated with 3% H2O2 for 5 min, followed by nonserum protein block (Dako) for 5 min. Primary antibodies included monoclonal GFAP (Invitrogen, Carlsbad, CA, USA; 1:50 for 30 min), polyclonal COX-2 (Cayman, Ann Arbor, MI, USA; 1:1000 for 30 min) and polyclonal c-kit (Dako; 1:500 for 60 min). A labelled polymer system (Envision + System HRP Anti-rabbit, Dako) was applied to all slides. Sections were stained with DAB+ chromogen (Dako) for 10 min and counterstained with haematoxylin. Positive controls included normal canine kidney and cerebral neurons for COX-2, canine mast cell tumour and Purkinje cells for c-kit and normal canine brain for GFAP. Negative controls substituted universal negative control + rabbit serum (Dako) for the primary antibodies.


The 20 canine glial tumours were separated into subtypes: 11 oligodendrogliomas (Fig. 1A) including 1 anaplastic type, 8 astrocytomas (Fig. 1B) including 2 GBM (Fig. 1C) and 1 oligoastrocytoma, based on histologic features and GFAP expression (Fig. 1D–F). Tissues were obtained from necropsy (n = 17) or surgical biopsy (n = 3), with 17 cases arising from the brain and 3 from the spinal cord. Fifteen gliomas were located in the prosencephalon, one oligodendroglioma was found in the mesencephalon, one astrocytoma was in the rhombencephalon, and three gliomas (one each oligodendroglioma, oligoastrocytoma and GBM) were in the spinal cord. The mean age of all dogs was 7.4 years (range 3–11 years), with a mean age of 7.0 years for dogs with oligodendrogliomas and 7.9 years for astrocytomas. The gender ratio was 4:1, with 16 males (nine castrated) and 4 spayed females. Breeds represented were 12 boxers, 3 Boston terriers, 2 mixed breeds and 1 each French bulldog, mastiff and Siberian husky. A summary of these findings is presented in Table 1.

Figure 1.

Canine brain. (A) Oligodendroglioma [haematoxylin and eosin (HE), ×200 magnification]. (B) Astrocytoma (HE, ×200). (C) GBM with endothelial proliferation (HE, ×200). (D) Oligodendroglioma IHC for GFAP; note negative expression (×400). (E) Astrocytoma IHC with strong GFAP expression (×400). (F) GBM IHC with heterogeneous GFAP expression (×400).

Table 1.  COX-2 and c-kit immunoreactivity in 20 canine gliomas
BreedSexAge (years)Tumour typeLocationCOX-2C-kit
  1. FS, female spayed; GBM, glioblastoma multiforme; M, male; MC, male castrated; NA, not available; V, vascular positivity.

Boston terrierM9AstrocytomaProsencephalon
Mixed breedMC7AstrocytomaProsencephalon
Siberian huskyFS11AstrocytomaRhombencephalon
Boston terrierMC11AstrocytomaProsencephalon
BoxerM4GBMSpinal cordV
BoxerMC7Anaplastic oligodendrogliomaProsencephalonV
French bulldogMCNAOligodendrogliomaProsencephalon
BoxerM3OligodendrogliomaSpinal cord
Mixed breedMNAOligodendrogliomaProsencephalon
Boston terrierM8OligodendrogliomaProsencephalon
BoxerM6OligoastrocytomaSpinal cord

No significant COX-2 expression was found (Table 1), although there were rare strongly immunoreactive cells in one GBM (Fig. 2A). Strong COX-2 expression was present in macula densa of control kidney sections (Fig. 2B), and occasional weak to moderate COX-2 expression was noted in the neurons (Fig. 2C) and choroid plexus of the normal brains used as control.

Figure 2.

Canine. (A) GBM with only rare positive COX-2 expression (×400 magnification). (B) Strong cytoplasmic COX-2 immunoreactivity of the macula densa (×400). (C) Moderate to strong COX-2 positivity of cortical neurons (×200). (D) IHC for c-kit in a GBM; neoplastic cells lack expression, but positivity of proliferative endothelium noted (×400). (E) Strong c-kit positivity of mast cell tumour (×400). (F) Cytoplasmic expression of c-kit in cerebellar Purkinje cells (×200).

Similarly, there was no significant c-kit expression in any of the tumours (Table 1). Rare, widely scattered immunoreactive neoplastic cells were present in one astrocytoma. Weak intramural and endothelial immunoreactivity was seen in vessel walls of all three high-grade gliomas (Fig. 2D, Table 1). Positivity for c-kit was also noted in a canine mast cell tumour (Fig. 2E) and in scattered neurons within nonaffected brain, particularly Purkinje cells (Fig. 2F).


Gliomas are among the most common CNS tumours in dogs. The mean ages in this study for canine oligodendrogliomas and astrocytomas were 7 and 7.9 years, respectively, which is similar to published averages of 8.1 ± 3.1 years and 8.6 ± 3.3 years.1 This study also agreed with published literature on approximately equal numbers of oligodendrogliomas and astrocytomas, the most common location being the prosencephalon, and representative breeds being overwhelmingly brachycephalics, particularly boxers and Boston terriers.1 Males were overrepresented with a ratio of 4:1, whereas previous studies had more equal representation. The reason for this is unknown. In human gliomas, the male:female ratio is 1.3:1 overall, with very similar rates for GBM but ratios approaching 4:1 for oligodendrogliomas.11

The lack of COX-2 immunoreactivity indicates that COX-2 inhibitors are unlikely to have COX-dependent effects for the treatment of canine gliomas. This is in contrast to human studies, which found COX-2 positivity ranging from 18 to 100%, especially in high-grade tumours.20,21,23–29 Lack of COX-2 expression may be because of species differences, fewer high-grade tumours, sample size or other factors such as age of tissue. The antibody used in this study detected COX-2 both in the control canine tissue and in scattered cortical neurons of canine brain, showing its efficacy in canine tissue. Additionally, there was reactivity of the choroid plexus, which had been noted in a study of canine meningiomas.19 Increased time between sample collection and fixation could lead to degradation of the target epitopes, yet COX-2 expression was consistently negative in all formalin-fixed gliomas, whether collected from necropsy or surgical biopsy. Age of the samples is not believed to be a factor, because a lack of COX-2 positivity in tumour cells was consistent throughout. However, it is possible that frozen sections would be more sensitive for the detection of low levels of COX-2, as less degradation of the epitopes would be expected to occur. Other methods to quantitate the level of COX-2 expression could be used, including Western blot and measurement of prostaglandin E2 levels.

Negative c-kit expression suggests that c-kit inhibitors are likewise ineffective for canine gliomas. However, our detection of c-kit in vascular endothelium suggests that c-kit inhibitors may provide an anti-angiogenic effect in high-grade gliomas. Studies of human gliomas found c-kit immunoreactivity ranging from 15.6 to 75% of neoplastic cells, with GBMs expressing significantly more c-kit than lower-grade tumours.37–39,41 Additionally, endothelial c-kit positivity has been detected in high-grade human gliomas, particularly GBMs, and only rarely in lower-grade gliomas.38,42 The finding of endothelial c-kit in high-grade gliomas in this study, as well as in the positive control tissue and Purkinje cells, implies that the antibody worked properly.

In conclusion, COX-2 inhibitors may still provide antineoplastic benefits for canine gliomas through COX-independent mechanisms despite the lack of COX-2 expression in this study. Studies of human gliomas and cell lines have shown that some selective COX-2 inhibitors can induce apoptosis, decrease angiogenesis and inhibit cell cycle progression independently of COX-2 expression by the target tissues.16 Given the vascular immunoreactivity seen in the high-grade tumours, c-kit inhibition may be beneficial by decreasing angiogenesis. Clinical studies may prove helpful in determining the value of these medications for dogs with gliomas.


We thank David Durtschi and the UTCVM Histology Lab for assistance with IHC, Misty Bailey for technical editing and Anik Vasington for graphical support. The University of Tennessee Center of Excellence in Livestock Diseases and Human Health funded J. M. J.'s employment.

Conflict of interest

The authors of this manuscript have no conflicts of interest to declare.