Pleomorphic xanthoastrocytomas of adults: MRI features, molecular markers, and clinical outcomes

Fifty adult patients with pathologically-confirmed pleomorphic xanthoastrocytomas (PXAs) were retrospectively analyzed. Adult PXAs appeared as a single lesion in 47 patients and multiple lesions in 3 patients. Temporal lobe was the most common location (17/50). Twenty-two cases were superficial with obvious meningeal contact and 9 were closely adjacent to lateral ventricles. Three imaging patterns were differentiated, including a predominantly solid mass with or without cystic changes (n = 33), a predominantly cystic mass with an obvious mural nodule (n = 14), and a predominantly cystic mass with an uneven wall thickness (n = 3). The mean tumoral apparent diffusion coefficient (ADC) was 0.83 ± 0.17 × 10−3 mm2/s, and the mean ADC ratio was 1.02 ± 0.22. The V-raf murine sarcoma viral oncogenes homolog B1 (BRAF)V600E mutation was found in 12 of 29 patients. In 36 patients with isocitrate dehydrogenases 1 and 2 (IDH1/2) data, only one had IDH1 mutation and no patient had IDH2 mutation. Anaplastic features were common (24/50) and significantly associated with high rates of recurrence or progression (P < 0.001). In conclusion, this study expands our knowledge on the MRI features, molecular markers, and clinical outcomes of adult PXAs, to some extent different from pediatric PXAs.

(36%). Multiple lesions were found in 3 cases, and 1 case was concomitant with a spinal intramedullary lesion at the L2 level (Fig. 5). Twenty-two cases (44%) were superficial in the cerebral hemisphere and attached to the meninges (Fig. 1). The locations of 9 cases (18%) with supratentorial tumors were adjacent to and even invasion of the lateral ventricles (Figs 2 and 5). PXAs were also located in the lateral ventricles (n = 2), dorsal midbrain (n = 1), and thalamus (n = 1). Hemorrhage was observed in 9 cases (18%). Peritumoral edema was observed in 41 cases, 30 of which had moderate to severe (≥10 mm in diameter) peritumoral edema.
Three imaging patterns were differentiated: (1) a predominantly solid mass with or without cystic changes in 33 cases (Figs 2, 4 and 5); (2) a predominantly cystic mass with a mural nodule in 14 cases, with the mural nodule close to the meningeal side in 8 cases (Fig. 1); and (3) a predominantly cystic mass with an uneven wall thickness, without an obvious mural nodule in 3 cases (Fig. 3). In all 50 cases, the solid component and mural nodule appeared slightly hypo-or iso-intense on T1-weighted spin-echo imaging (T1WI), iso-or slightly hyper-intense on T2-weighted spin-echo imaging (T2WI), and iso-or slightly hyper-intense on T2-fluid-attenuated inversion recovery (FLAIR) sequence. The cystic component appeared homogeneously hypo-intense on T1WI, hyper-intense on T2WI, and slightly hypo-intense on T2 FLAIR. Marked contrast enhancement in the solid component and mural nodule was observed in 46 cases, and marked contrast enhancement in the cystic wall was observed in 8 cases.
The diffusion-weighed imaging (DWI) showed restricted diffusion in 11 cases (Figs 1C, 2C and 5D) and no evidence of restricted diffusion in 15 cases (Fig. 3C). Tumoral apparent diffusion coefficient (ADC) values were not calculated in 3 cystic cases without obvious mural nodules because there was not enough solid area to place the regions of interests (ROIs) without partial volume effects in the necrotic region. Therefore, tumoral ADC values were measured in only 23 patients. The mean tumoral ADC was 0.83 ± 0.17 × 10 −3 mm 2 /s (min-max: 0.58-1.42). The mean ADC ratio was 1.02 ± 0.22 (min-max: 0.68-1.66).
Histopathological and molecular findings. According to the World Health Organization (WHO) classification, 26 (52%) and 24 (48%) cases were histopathologically diagnosed with Grade II PXA and Grade III anaplastic PXA, respectively. Grossly, some tumors were superficial in the cerebral hemisphere and attached to the  (Figs 1E and 2E). Immunohistochemically, the tumor cells were positive for glial fibrillary acid proteins and S-100 proteins. V-raf murine sarcoma viral oncogenes homolog B1 (BRAF) V600E status was evaluated in 29 patients, of which 9 cases were Grade II PXA and 20 cases were Grade III anaplastic PXAs. We found that BRAF V600E mutation was observed in 12 patients (Fig. 1F) and non-V600E mutation in 17 patients (Fig. 2F). We also found that 4 patients (20%) of anaplastic PXAs demonstrated the mutation. The isocitrate dehydrogenases 1 and 2 (IDH1/2) mutation status were evaluated in 36 patients, and we found that IDH1 was mutated in only 1 patient, and not mutated in other 35 patients (Fig. 4D). IDH2 was not mutated in all 36 patients (Fig. 4E). (Table 3)   Grade III anaplastic PXA, were successfully followed up (range: 3 months to 4.5 years). Out of the 30 patients, 14 cases had evidence of recurrence or progression (including 2 cases with Grade II PXA and 12 cases with Grade III anaplastic PXA), and one case eventually died. The rates of recurrence or progression were higher in patients diagnosed with Grade III anaplastic PXA (92%) compared to those diagnosed with Grade II PXA (12%), with statistical significance (P < 0.001).

Discussion
In 1979, Kepes et al. 10 described for the first time a series of 12 young patients with a distinctive form of astrocytoma was thought to arise from subpial astrocytes due to the presence of "basal lamina". The astrocytoma was subsequently named as a PXA or Kepes tumor. The typical histopathological features of PXAs include marked cellular pleomorphism with spindle cells, in addition to multi-nucleated giant cells with bizarre nuclei and lipid-filled xanthomatous cells. On immunohistochemical analysis, all tumors are positive for glial fibrillary acid proteins and S-100 proteins 1,2 . In a later study of immunohistochemistry and electron microscope, Giannini et al. 11 revealed the expression of mixed glioneuronal markers in PXAs. Based on the literature, it has been suggested that PXAs arise from less well-differentiated precursor cells with multipotent cell fate potential. According to the WHO classification of CNS neoplasms in 2007 12 , PXA is a distinctive subtype of astrocytoma and is classified as WHO Grade II. The outcomes in cases of PXAs are relatively favorable, with 5-and 10-year overall survival rates of 75-81% and 67-70%, respectively, after total resection 2,13 . Although PXA is generally considered as an indolent neoplasm, it is associated with a higher frequency of recurrence, malignant transformation, and death when compared with other astrocytic tumors of good prognosis, such as pilocytic astrocytomas, indicating the more aggressive nature of PXAs. Lesions with significant mitotic activity (i.e., five or more mitoses per 10 HPF) regardless of the necrosis are designated as PXAs with anaplastic features, as proposed by Giannini et al. 2 . It has been reported that 21% of PXA cases in children and adolescents had anaplastic features or malignant transformation at initial presentation 6 . A study dedicated solely to pediatric PXAs found that 27% of cases had anaplastic features 4 . This entity has now been acknowledged and included in the 2016 update of the WHO Classification of Tumors of the CNS as Grade III anaplastic PXA 14 . In our study, the ratio of anaplastic PXAs was about 48%, which was higher than the ratio of anaplastic PXAs in children and adolescents from previous studies, indicating that adult PXAs may be more likely to have anaplastic features. In reviewing the literature, PXAs with anaplastic features have been shown to predict the risk of recurrence 2,15 . A previous study 6 has also shown the recurrence rates to be higher in patients diagnosed with anaplastic PXAs compared to those diagnosed with Grade II PXAs. Similar results were found in the present study. PXAs occur primarily in children and young adults, with no bias between genders 1 . PXAs have also been described in infants and elderly patients, with variable prognoses 1,2,13 . The median age of the adult patients in the present series was 36 years old, with two patients older than 65 years. There was a male predominance in the present series, inconsistent with the literature 1 . The typical clinical presentation in cases of PXAs includes a long history of epilepsy, especially in young patients, most commonly in the second decade of life 16 . However, in our adult sample, headaches, which are the atypical clinical presentation, were more common than seizures.
PXAs tend to be superficial and attached to the meninges, with a predilection for the temporal lobe, followed by the parietal, frontal, and occipital lobes 2, 13 . Other rare locations are the cerebellum 17 , spinal cord 18 , hypothalamus 19 , pineal region 20 , sella turcica 21 , and retina 22 . In the present series, 44% of cases were superficial in the cerebral hemisphere and attached to the meninges. In these cases, the tumors may have arisen from subpial astrocytes due to the presence of basal lamina, which is a characteristic feature of these astrocytes. In addition, 9 cases of supratentorial PXAs in the present series were adjacent to and even invasion of the lateral ventricles. There are few reports of such an association in the literature, and its significance needs to be confirmed in future studies. Besides, in the present series, PXAs were also observed to be located in the lateral ventricles, dorsal midbrain, and thalamus.
Although there are some imaging reports of PXAs 3-5 , reports dedicated solely to adult imaging studies are rare, and DWI characteristics of adult PXAs have not been previously described. The present study described the MRI metrics of this kind of rare tumor. Consistent with the histopathological descriptor "pleomorphic", the imaging features of PXAs in the present study were varied, leading to some difficulties in the preoperative diagnosis. Previous reports [3][4][5] indicated that PXAs were typically predominantly cystic, with a contrast-enhancing mural nodule close to the meningeal side. However, in our study, only 28% of PXAs presented as a predominantly cystic mass with a mural nodule; 66% of PXAs appeared as a predominantly solid mass with or without cystic changes, whereas others appeared as a predominantly cystic mass with uneven wall thicknesses and without obvious mural nodules. In all 50 cases, the solid component and mural nodule appeared slightly hypo-or iso-intense on T1WI, iso-or slightly hyper-intense on T2WI, and iso-or slightly hyper-intense on T2 FLAIR. The cystic component appeared homogeneously hypo-intense on T1WI, hyper-intense on T2WI, and slightly hypo-intense on T2 FLAIR. The solid components and mural nodules mainly showed marked contrast enhancement, and the cystic wall showed enhancement or no enhancement. Moreover, our findings are similar to those in previously published reports in terms of the frequency of peritumoral edema, with marked peritumoral edema being found in 60% of PXAs. Hemorrhage in PXAs has been described as an uncommon finding, but it was present in 18% of tumors in our series. And a study of PXAs in childhood also reported 33% of the patients had hemorrhage 4 .
Most MRI findings of PXAs are based on single lesions, while reports of multiple lesions are extremely rare. In 2013, Montano et al. 23 reported a primary anaplastic PXA patient with multiple brain lesions located in the region of the left parietotemporal junction, left thalamus, and left hippocampus. They concluded that in this case the PXA was best classified as a multicentric tumor and proposed that multicentricity should be considered as a marker of an anaplastic phenotype in PXA. However, the findings of the present study do not entirely support this proposal. Multiple brain lesions were present not only in the anaplastic PXAs in the current cohort, but also in the Grade II PXA. Due to the similar radiological features of these lesions in different brain regions, we suggest that the lesions in these three cases should be better classified as either disseminated or multicentric. The Grade II PXA case with multiple lesions was concomitant with a spinal intramedullary lesion at the L2 level, and no further lesions were observed along the rest of the spine, suggesting that the lesion possibly represented intramedullary metastasis, although such metastasis is very rare. In the absence of a biopsy or autopsy examination of the spinal lesion, an entity other than a PXA cannot be excluded. The spinal spread of a secondary PXA with anaplastic features is relatively common, occurring in approximately 1/3 of patients 24 . There are very few reports of spinal tumors in patients with primary PXAs without anaplastic features. However, Grade II PXAs also have the biological characteristic of distant spread 25 . The detection of an unresectable spinal lesion underlines the importance of expanding radiological follow-up in PXA patients to imaging of the spine.
DWI has become an important tool in the preoperative characterization of brain tumors. Humphries et al. 26 found an inverse relationship between lesional cellularity and the ADC, with a higher degree of malignancy associated with more intensive tumor spread, and restriction of the free diffusibility of water molecules determining a low ADC in vivo. Moore et al. 4 have reported the DWI characteristics of pediatric PXAs. They found that the mean ADC of the tumors was 0.91 ± 0.22 × 10 −3 mm 2 /s and the mean ADC ratio was 1.14 ± 0.26. These values were lower than those of pilocytic astrocytomas and supratentorial gangliogliomas. This feature is recognized with a similar frequency in the adult PXAs in the present series. In the 23 tumors where the ADC values could be reliably measured, the mean ADC of the tumors was 0.83 ± 0.17 × 10 −3 mm 2 /s, and the mean ADC ratio was 1.02 ± 0.22. It is important to recognize that the relatively low ADC values and ADC ratios of PXAs in both children and adults are not unusual, which is unlike other low-grade neoplasms. Recently, the BRAF V600E mutation has been known as a common finding in certain CNS tumors, most commonly in PXAs (nearly 50-60%) 27,28 . Amongst CNS tumors, the presence of BRAF V600E may be helpful in distinguishing PXAs from diffusely infiltrating gliomas that always lack BRAF V600E mutation. Conversely, IDH1/2 mutations, which are frequently present in diffusely infiltrating gliomas, rarely occur in PXAs 29 29 has also reported that one case of PXAs was positive for the IDH1 mutation. Identification of BRAF V600E mutations in PXAs could be helpful in deciding the proper therapy. Indeed, for these patients, the use of targeted therapies, such as vemurafenib, was discussed in previous case reports 8,9 . This novel therapeutic approach must be validated, and may ultimately represent a new opportunity for these patients.

Conclusions
In summary, this study expands our knowledge on the MRI features, molecular markers, and clinical outcomes of adult PXAs, to some extent different from pediatric PXAs. The adult PXAs appear as predominantly solid masses with single lesions in the overwhelming majority of cases, and multiple lesions in few cases. Relatively lower ADC values and ADC ratios of the solid components may distinguish these tumors from other low-grade neoplasms. Anaplastic features are common at initial presentation, conferring more unfavorable clinical outcomes. The evaluation of BRAF V600E and IDH1/2 mutation status may be helpful in distinguishing PXAs from diffusely infiltrating gliomas, and useful for the therapeutic implications of PXAs. MRI protocol and evaluation. All 50 patients underwent MRI scans before surgery, using a 3 T MRI scanner of a variety of manufactures including Siemens, GE, and Philips. The sequences of the plain scan included axial and sagittal T1WI, axial T2WI, and axial T2 FLAIR sequence. DWI, performed on a 3 T Siemens MRI scanner at the same section position as the axial T1WI using a spin echo single-shot echo-planar sequence with the b values of 0 and 1000 s/mm 2 , was available in 26 patients. The ADC maps were generated with a monoexponential model on a voxel-by-voxel basis for all imaging planes. After the patient received intravenous administration of a standard dose of Gd-DTPA (0.1 mmol/kg; Bayer Healthcare Co., Ltd., Guangzhou, China), axial, sagittal, and coronal T1WI sequences were acquired. Imaging studies were reviewed by two neuroradiologists with 27 and 9 years of experience in neuroimaging. Tumor size, number, location, hemorrhage, degree of peritumoral edema, imaging pattern, signal, and enhancement characteristics were evaluated on MRI images. All the DWI data were transferred to a Syngo workstation for analysis. The solid components of the tumors were sampled by placing up to three nonoverlapping ROIs on the ADC map, avoiding cystic or necrotic parts. If there were multiple brain lesions, the largest one that reflected diffusion-weighted findings for most PXA lesions was selected as the target lesion to increase the accuracy of the measurement. The mean tumoral ADC value was calculated. Further, three ROIs with the same dimensions were placed on the contralesional thalamus because the thalamus maintains normal signal intensity on DWI, even in marked hydrocephalus cases, to serve as a reliable internal control 32 . Then, to assess the objective difference in the ADC value between the tumor and the contralesional brain tissues, the ADC ratio was calculated as the mean ADC of the tumor divided by the mean ADC of the corresponding contralesional thalamus. During the delineation, the two neuroradiologists were blinded to all the clinical information and delineated the characteristics of the tumors from each patient, independently. And they worked together on MRI images of each patient via a consensus reading. BRAF V600E and IDH1/2 mutations analysis. Five 10-µm sections from each selected FFPE block were subjected to genomic DNA extraction. DNA was isolated using E.Z.N.A. ® FFPE DNA kit (D3399-02; Omega Bio-tek, US) according to the manufacturer's instructions. Next, the genomic DNA was amplified by polymerase chain reaction (PCR) and the PCR products were purified. The purified products were sequenced and then analyzed using Chromas Lite software. Compared with the reference sequence, the BRAF V600E mutation and IDH1/2 alterations in the hotspot codons R132H and R172K were obtained. Statistical analysis. Statistical analysis were performed using SPSS version 21.0 (SPSS, Inc, USA). Data were described as number (percentage), median (range), and mean ± SD. Comparison of the rates of recurrence or progression between groups was evaluated using the Fisher's exact probability test. The P value <0.05 was considered as statistically significant.