The T2‐FLAIR mismatch sign as an imaging biomarker for oligodendrogliomas in dogs

Abstract Background In humans, the T2‐weighted (T2W)—fluid‐attenuated inversion recovery (FLAIR) mismatch sign (T2FMM) is a specific imaging biomarker for the isocitrate dehydrogenase 1 (IDH1)‐mutated, 1p/19q non‐codeleted low‐grade astrocytomas (LGA). The T2FMM is characterized by a homogeneous hyperintense T2W signal and a hypointense signal with a hyperintense peripheral rim on FLAIR sequences. In gliomas in dogs, the T2FMM has not been described. Hypotheses/Objectives In dogs with focal intra‐axial brain lesions, T2FMM will discriminate gliomas from other lesions. The T2FMM will be associated with the LGA phenotype and presence of microcysts on histopathology. Interobserver agreement for T2FMM magnetic resonance imaging (MRI) features will be high. Animals One hundred eighty‐six dogs with histopathologically diagnosed focal intra‐axial lesions on brain MRI including oligodendrogliomas (n = 90), astrocytomas (n = 47), undefined gliomas (n = 9), cerebrovascular accidents (n = 33), and inflammatory lesions (n = 7). Methods Two blinded raters evaluated the 186 MRI studies and identified cases with the T2FMM. Histopathologic and immunohistochemical slides of T2FMM cases were evaluated for morphologic features and IDH1‐mutations and compared to cases without the T2FMM. Gene expression analyses were performed on a subset of oligodendrogliomas (n = 10) with and without T2FMM. Results The T2FMM was identified in 14/186 (8%) of MRI studies, and all dogs with T2FMM had oligodendrogliomas (n = 12 low‐grade [LGO], n = 2 high‐grade [HGO]; P < .001). Microcystic change was significantly associated with the T2FMM (P < .00001). In oligodendrogliomas with T2FMM, IDH1‐mutations or specific differentially expressed genes were not identified. Conclusion and Clinical Importance The T2FMM can be readily identified on routinely obtained MRI sequences. It is a specific biomarker for oligodendroglioma in dogs, and was significantly associated with non‐enhancing LGO.


| INTRODUCTION
In dogs, gliomas account for 35% of primary brain tumors, and feature remarkable neuroimaging, clinical, and phenotypic similarities to gliomas in humans, making them an excellent model for studies in humans. [1][2][3][4] Characterizing glioma genetic drivers in dogs and identifying molecular determinants common to humans and dogs only recently has been investigated. 4 Glioma classification in humans is based on 2021 World Health Organization criteria, including genotypic, molecular and phenotypic tumor features, describing entities in which the presence of isocitrate dehydrogenase 1/2 (IDH1/2) mutations and 1p/19q codeletion status are considered important predictors of positive prognosis and longer overall survival than are found in patients carrying the wild-type IDH1/2 gene. [5][6][7] Genomic features of gliomas in humans also have been associated with certain magnetic resonance imaging (MRI) biomarkers such as the T2-weighted (T2W)-fluid-attenuated inversion recovery (FLAIR) mismatch (T2FMM) sign, which has high to perfect specificity and moderate to low sensitivity to predict IDH1-mutant, 1p/19q non-codeleted low-grade astrocytomas (LGA). [8][9][10][11][12][13] The T2FMM is characterized by a homogeneous hyperintense signal on T2W and a hypointense signal with a hyperintense peripheral rim on FLAIR. 8,13 Although histopathology is the gold standard to diagnose gliomas, antemortem diagnosis by biopsy rarely is performed in veterinary medicine because of practical limitations, such as availability of centers with experience performing brain biopsy or surgery in dogs with gliomas. 1,[12][13][14][15][16] Hence, the antemortem diagnosis of gliomas is often presumptive based on MRI features. 1,2,16,17 A recent classification scheme for gliomas in dogs recognizes tumor types as oligodendrogliomas, astrocytomas, or undefined gliomas, and classifies them as low-grade or high-grade according to their morphologic features. 3 Although the qualitative MRI features of gliomas in dogs have been widely studied, no reliable imaging features distinguish among tumor types or grades, but tumor contrast enhancement occurs more frequently in high-grade than low-grade gliomas. 1,2,18,19 A lack of biological specificity in MRI signals has hindered identification of MRI biomarkers for glioma in dogs, which share imaging characteristics with other intracranial diseases such as abscesses, granulomas, meningoencephalitis of unknown etiology, and cerebrovascular accidents (CVA). 1,2,20,21 The T2FMM and its possible association with tumor phenotype have not been investigated previously in dogs.
The purpose of our study was to evaluate the presence of T2FMM in dogs with clinical and MRI evidence of focal intra-axial brain diseases that can mimic gliomas. We hypothesized that the presence of the T2FMM would: (a) discriminate gliomas from other intraaxial brain lesions; (b) be predictive of LGA phenotype; and (c) be associated with microcysts on histopathology of the mismatched tumor regions. Additionally, we investigated whether the T2FMM sign was associated with a particular genotype in a subset of dogs with gliomas.

| Study design and animals
Ours was a retrospective study of dogs with histologically confirmed, focal intra-axial forebrain lesions associated with clinical signs of intracranial disease that were referred to our neurology practice between 2004 and 2022 for biopsy of their brain lesions or for treatment of presumed intracranial glioma. 16 An institutional clinical trials database was searched for cases that met the following inclusion criteria: clinical signs of intracranial disease, presence of a focal intra-axial forebrain lesion on brain MRI examination, and histopathologic confirmation of the lesion by stereotactic brain biopsy, surgical brain biopsy, or necropsy. 16,21 Because dogs were referred from multiple institutions, MRI scans were performed on 0.2T, 1.0T, 1.5T or 3T units, and MRI sequences obtained were not standardized. At a minimum, qualitative evaluations of conventional MRI images included T1-weighted (T1W) and T1W contrast-enhanced images in at least 2 planes, T2-weighted (T2W) images in 2 planes, and FLAIR sequences. Proton density (PD), T2W gradient echo images (T2*GRE), and diffusion weighted imaging with anisotropic diffusion coefficient (DWI/ADC) sequences were reviewed when available. Images were reviewed independently by 2 raters including a board-certified neurologist with more than 25 years of experience (JHR) and a veterinarian with 2 years of experience (JGM). The MR images were evaluated for the presence or absence of the core features of the T2FMM sign including a complete or nearly complete homogeneously hyperintense lesion signal on T2W sequences, and a complete or nearly complete relatively hypointense central signal except for a hyperintense peripheral rim on FLAIR. 8,13 To mitigate possible variability associated with qualitative reviewer interpretations of "complete" or "near-complete" FLAIR signal suppression, we also required that ≥80% of the high T2W signal volume was FLAIR null (F null ) to be considered a T2FMM positive case, as calculated using the equation: where the F null V is the lesion volume that suppressed on FLAIR sequences and TTV is the total T2W lesion volume. 22 The F null V and TTV were determined using commercial image analysis software (Osirix MD, v11.0.4, Pixmeo, Switzerland) from transverse sequences using manually defined regions of interest (ROIs) from contiguous slices representing the hypointense (null) portion of the lesion on FLAIR sequences and the homogenously hyperintense portion of the lesion on T2W images, respectively. 22 The F null V and TTV then were calculated using the software's ROI volume function. Additionally, T1W signal intensity (hypo-, iso-, or hyperintense) and presence (present or absent) of lesion contrast enhancement were assessed.

| Histopathologic and immunohistochemical analyses
Histopathologic diagnoses for all lesions were recoded from the surgi- used. An IDH1-mutant genotyped grade IV astrocytoma from a human was used as a positive control, and the primary antibody was omitted from negative controls.

| T2FMM histopathologic and immunohistochemical analyses
Significant correlations between the presence of microcysts and myxoid matrix lakes were found in oligodendrogliomas presenting with the T2FMM sign (Table 3). No other grade agnostic histopathological features of oligodendrogliomas were significantly associated with the T2FMM sign.

| Immunohistochemical and molecular analysis
Immunoreactivity to the IDH1-mutation was not observed in any the oligodendrogliomas with (n = 0/14) or without (n = 0/15) T A B L E 1 Prevalence, diagnoses, and MRI features of the T2-FLAIR mismatch sign in 186 dogs with focal, intra-axial brain lesions.

| DISCUSSION
Although our initial hypothesis associating the T2FMM sign with the LGA phenotype in dogs was rejected, our results indicate that the T2FMM is a highly specific indicator of oligodendrogliomas in dogs, particularly non-enhancing LGO phenotypes. Evaluation of the T2FMM sign is simple and it can be easily identified by evaluators with a range of experience using routinely obtained conventional MRI sequences. We evaluated the performance of the T2FMM sign as a biomarker in a large population of dogs with different intra-axial brain diseases, in contrast with prior studies that considered selected groups of dogs with only gliomas or oligodendrogliomas. 2,18,19 Therefore, our results are more clinically relevant because they are based on a more representative sample of dogs with intracranial diseases that can mimic gliomas seen in clinical practice.
F I G U R E 1 Magnetic resonance imaging features of canine low-grade (top three rows) and high-grade (bottom two rows) oligodendrogliomas in which the T2FMM sign was present and absent. The T2FMM sign is characterized by a homogeneous hyperintense T2W signal and a hypointense signal with a hyperintense peripheral rim on FLAIR sequences. T2FMM, T2-fluid-attenuated inversion recovery mismatch sign.
T A B L E 2 Interobserver intraclass correlations (ICC) of MRI features defining the T2-FLAIR mismatch sign. Previous studies have indicated that no MRI features reliably predict glioma phenotypes or grades in populations that included dogs with astrocytomas and oligodendrogliomas, although the presence of contrast enhancement has been associated with high-grade tumors. 1,2,15,18 Although MRI features compatible with the T2FMM sign previously were included in a study of oligodendrogliomas in dogs, ours is the first study that specifically identified a significant relationship between an MRI biomarker and the oligodendroglial phenotype in dogs. 18 In adult humans, the T2FMM sign is a highly accurate biomarker for IDH1-mutated, 1p/19q non-codeleted astrocytomas, with specificity and sensitivity as high as 100% and 31%, respectively. [8][9][10][11][12] We found a similar high specificity and low sensitivity for the T2FMM to detect oligodendrogliomas in dogs. It is important to note that the T2FMM sign has been reported in other tumor types in humans, such as IDH1-mutated, 1p/19q co-deleted oligodendrogliomas and IDH1 wild-type glioblastomas. 11,13 Although dogs remain a valuable animal model for comparative studies of glioma, our study adds additional evidence of important differences in glioma biology between dogs and humans, because we did not identify associations between T2FMM and astrocytomas in dogs or the presence of IDH1 mutations.
One possible reason for our discordant T2FMM findings relative to humans may be that gliomas in dogs have been shown to be more closely related to gliomas in children at the genetic and epigenetic levels, and considerable pathophysiological differences exist between glioma presentations in human adults and children. 4 We did not identify an association of the T2FMM with IDH1 mutations in dogs, but this finding was not unexpected given the rarity in which IDH1 mutations have been identified in dogs with gliomas. 1,4 Overexpression of PDGFRA has been shown to be a common driver of tumorigenesis in humans and dogs with gliomas, and to be present more frequently in high-grade gliomas. 4,23-28 T A B L E 3 Grade agnostic histological features of canine oligodendrogliomas with and without the T2-FLAIR mismatch sign. F I G U R E 2 Volcano plot showing cutoffs of false discover rate-adjusted P-values and fold change for all genes (gray circles) and differentially expressed (red circles) between canine oligodendrogliomas with and without the T2FMM. T2FMM, T2-fluid-attenuated inversion recovery mismatch sign.
In addition, CCL2 is a known progression-inducing factor in humans with gliomas and has been reported to be significantly upregulated in HGA. 23 In our study, no specific genotype was associated with oligodendrogliomas exhibiting T2FMM, but PDGFRA and CCL2 were significantly upregulated among oligodendrogliomas without the sign, which likely reflects sampling bias of inclusion of more HGO in the group without the T2FMM sign. We also did not assess for 1p/19q codeletions, because these have not yet been identified in dogs with gliomas. 4 The small sample sizes included in the molecular analyses could have precluded us from identifying particular genotypes associated with the T2FMM.
Our results highlight that the primary value of the T2FMM in dogs is its specificity for oligodendroglial tumors, rather than its sensitivity. By using more rigid criteria for the T2FMM sign in dogs, we hoped to avoid difficulties that have been encountered in studies in humans that applied purely qualitative definitions of T2FMM to cases, which have subsequently resulted in reductions in the specificity of this biomarker. 13 Although inter-rater agreements for T2FMM criteria were good to excellent in our study, the primary discordance between reviewers arose from analysis of the FLAIR null tumor burden in cases imaged on low-field magnets. Magnetic field strength also can influence interpretation of T2FMM, because higher magnetic field strengths will result in more FLAIR signal attenuation, making the FLAIR null signal more obvious and thus F null volumetric ROI easier to contour. 29 Conversely, stronger magnetic fields also can influence interpretation of the T2W lesion signal, with glioma T2W signal being less homogenous on 3T than on 1.5T scanners. 13 False positive cases of the T2FMM sign have been reported in children with gliomas, including pilomyxoid astrocytoma and K27M mutant midline gliomas, as well as non-neoplastic diseases of the brain. [30][31][32] In these tumors, T2FMM occurs because of the presence of microcystic changes in the tumor regions of mismatch, which is consistent with our findings because microcysts were present in all T2FMM cases reported here. 12 The biological basis for the T2FMM appearance of microcysts is currently unknown, because microcysts also have been found in IDH1-mutated, 1p/19q non-codeleted astrocytomas and in oligodendrogliomas in dogs that do not present with the T2FMM sign. 12 We also identified a significant association between the presence of myxoid matrix lakes and dogs with the T2FMM sign, a histologic feature that has not been correlated with this specific sign in humans. Myxoid tissue is variably present in oligodendrogliomas in dogs and in other non-neural soft tissue neoplasms, and consists of a mucopolysaccharide matrix of sulfated and nonsulfated glycosaminoglycans with high water content. 32,33 Myxoid lakes are also a rare microscopic finding in other rare types of neuroepithelial tumors, such as dysembryoplastic neuroepithelial tumor (DNET), a benign neuroglial tumor with a heterogeneous cellular composition, containing myxoid tissue. [34][35][36] In humans with DNET an MRI feature very similar to the T2FMM sign is found, known as the "bright rim sign," which is described as a complete or incomplete hyperintense well-defined rim around the DNET on FLAIR sequence. 34 Because DNETs also commonly have cystic structures and abundant myxoid matrix content, which make them appear hyperintense on T2W and hypointense on T1W, they can mimic the T2FMM sign. 34

| CONCLUSION
Although the presence of the T2FMM sign is uncommon in dogs with gliomas, it is an easily identifiable and specific imaging biomarker for oligodendrogliomas in dogs. Although we did not identify false positive cases of T2FMM in our study, in humans false positive cases occurring in other types of neuroepithelial tumors and non-neoplastic diseases of the brain are reported.