Magnetic resonance imaging surrogates of molecular subgroups in atypical teratoid/rhabdoid tumor

Abstract

Background

Recently, 3 molecular subgroups of atypical teratoid/rhabdoid tumor (ATRT) were identified, but little is known of their clinical and magnetic resonance imaging (MRI) characteristics.

Methods

A total of 43 patients with known molecular subgroup status (ATRT–sonic hedgehog [SHH], n = 17; ATRT-tyrosine [TYR], n = 16; ATRT–myelocytomatosis oncogene [MYC], n = 10) were retrieved from the EU-RHAB Registry and analyzed for clinical and MRI features.

Results

On MRI review, differences in preferential tumor location were confirmed, with ATRT-TYR being predominantly located infratentorially (P < 0.05). Peritumoral edema was more pronounced in ATRT-MYC compared with ATRT-SHH (P < 0.05) and ATRT-TYR (P < 0.05). Conversely, peripheral tumor cysts were found more frequently in ATRT-SHH (71%) and ATRT-TYR (94%) compared with ATRT-MYC (40%, P < 0.05). Contrast enhancement was absent in 29% of ATRT-SHH (0% of ATRT-TYR; 10% of ATRT-MYC; P < 0.05), and there was a trend toward strong contrast enhancement in ATRT-TYR and ATRT-MYC. We found the characteristic (bandlike) enhancement in 28% of ATRT as well as restricted diffusion in the majority of tumors. A midline/off-midline location in the posterior fossa was also not subgroup specific. Visible meningeal spread (M2) at diagnosis was rare throughout all subgroups.

Conclusion

These exploratory findings suggest that MRI features vary across the 3 molecular subgroups of ATRT. Within future prospective trials, MRI may aid diagnosis and treatment stratification.

Atypical teratoid/rhabdoid tumors (ATRTs) are rare, highly malignant embryonal central nervous system (CNS) neoplasms that comprise approximately 1%–2% of all pediatric brain tumors.¹ In patients less than 3 years of age, ATRTs account for up to 20% of brain tumor cases.² ATRTs occur at a median age of 18–22 months.³ Approximately 70% of all cases arise in children younger than 1 year of age, and over 90% of cases occur before 3 years of age.⁴ Rarely, ATRTs can be found in older children and in adults.^5,6 The sole recurrent genetic alterations are biallelic mutations of SMARCB1 (INI1, SNF5, BAF47) or (rarely) SMARCA4 (BRG1), both members of the SWItch/sucrose nonfermentable chromatin remodeling complex.^7–9 On the epigenetic level, ATRTs are surprisingly heterogeneous, and 3 molecular subgroups with characteristic gene expression patterns, DNA methylation profiles, and enhancer landscapes have been recently identified.^4,10

MR images of brain tumors that showed a combination of nonspecific singular features—including intratumoral hemorrhage, peripherally localized cysts, high cellularity seen as low T2 and apparent diffusion coefficient (ADC) signal, as well as a distinct bandlike “wavy” enhancement (present in 38% of cases) in an infant or young child—are suggestive of an ATRT. Furthermore, 15% of patients may display meningeal dissemination at diagnosis.¹¹ A rather rare phenomenon is bone involvement of ATRT.¹² The predominant location of an ATRT in the brain is (depending on age and/or molecular subgroup) either supratentorial (older patients and ATRT–myelocytomatosis oncogene [MYC] or infratentorial (younger patients and ATRT-tyrosine [TYR]). ATRT–sonic hedgehog (SHH) tumors can occur both supra- and infratentorially.⁴ The current analysis is the first to describe and compare MR imaging features of molecular subgroups in ATRT.

Methods

We retrieved 43 patients with ATRT from the EU-RHAB Registry. Molecular subgrouping (ATRT-SHH, ATRT-TYR, ATRT-MYC) had been performed using Infinium HumanMethylation450 BeadChip (P.J., M.K.), and initial MRI studies at diagnosis were obtained from the Neuroradiological Reference Center of the multicenter German Hirntumor (HIT) trials. MRI scans had been acquired with MR scanners of different manufacturers at 1.0–3.0 tesla field strength. A retrospective review of all MRI scans was performed by 2 experienced radiologists (M.W-M. and J.N.) in consensus. We analyzed and compared patient data (sex, age at diagnosis) and MR morphology in order to search for differences between the 3 molecular subgroups. For image reading, we applied standardized criteria of our Reference Center: localization (supratentorial, infratentorial, supra/infratentorial, midline/off-midline, further specification of exact neuro-anatomical origin) and volume of tumors (approximation formula: a × b × c × 0.5; in mL, categorized in 6 groups), T2 signal, presence of peripheral cysts and peritumoral edema (maximum width in centimeters in axial plane), signs of intratumoral hemorrhage, bone involvement, restriction in diffusion-weighted imaging (DWI; signal in DWI/ADC compared with normal brain), and enhancement pattern after gadolinium administration (no/mild/strong enhancement, and presence of characteristic “bandlike” pattern as previously described¹¹). Another criterion was evidence of synchronous lesions and meningeal dissemination (when visible in contrast-enhanced T1-weighted images, referring to M2 stage). We further tested whether differences of ATRT imaging features were location dependent (supra/infratentorial), or indeed related to molecular subgroups. Our dataset did not meet all criteria (<20% of cells with an expected frequency of n < 5) for log-linear regression and subsequent multivariate analysis. Thus, analyses of variance (ANOVAs; for age, edema) and chi-square tests were performed as appropriate. A P-value < 0.05 was regarded as significant. Statistical analyses were computed using SPSS for Windows (v21). All procedures were approved by the local and central ethics committee of the HIT studies and performed in accordance with the Helsinki Declaration of the World Medical Association.

Results

The mean age at diagnosis of the 18 females and 25 males harboring ATRT was 2.4 ± 3.8 years. Molecular subgroups comprised 17 ATRT-SHH, 16 ATRT-TYR, and 10 ATRT-MYC. Sex distribution and age at diagnosis did not differ between molecular subgroups (ANOVA, P = 0.37). Tumor location, however, was significantly different between the 3 molecular subgroups: ATRT-TYR (13/16 or 81% infratentorial), ATRT-MYC (7/10 or 70% supratentorial), and ATRT-SHH (balanced: 6/17 or 35% infratentorial, 6/17 or 35% supratentorial, 5/17 or 29% both supra- and infratentorial; rounded values) (P = 0.002, as illustrated in Fig.1 and Supplementary Figure S1; MR images in Supplementary Figure S2). Regarding tumor location in the midline versus off-midline, we found an almost equal distribution in our cohort (23/43 or 53% midline, 20/43 or 47% off-midline) and no significant differences between molecular subgroups (P = 0.85; Supplementary Figure S3A). All (n = 5) of the ATRT-SHH tumors with both supra- and infratentorial location were midline tumors. There was also no significant difference between subgroups when supra- or infratentorial location was analyzed separately (infratentorial P = 0.37, supratentorial P = 0.30). Regarding exact neuro-anatomical origin of the tumors and side/hemisphere of occurrence, we did not find significant differences between ATRT subgroups (P = 0.09 and P = 0.11, respectively). Among all ATRTs in our cohort, there was only one tumor (ATRT-MYC) of spinal location (Supplementary Figure S3E). Tumor volume was not significantly different between molecular subgroups in ATRT (P = 0.43), with volumes of >10 mL in the majority of ATRTs (37/43 tumors). Beyond location, Fig. 1 further illustrates tumor size according to molecular ATRT subgroups. When further analyzed for large (≥10 mL) versus small (<10 mL) tumors, there was also no significant difference (P = 0.82).

Analyzing peritumoral edema, ANOVA revealed significant differences between molecular ATRT subgroups (P = 0.002). Post-hoc Scheffé tests showed significantly more edema in ATRT-MYC (mean = 1.85 cm; Supplementary Figure S2E) compared with ATRT-SHH (mean = 0.29 cm; P = 0.005) and ATRT-TYR (mean = 0.14 cm; P = 0.002). Differences between ATRT-SHH and ATRT-TYR were not significantly different (P = 0.924). The single case with spinal location showed abundant edema of 7 cm, not allowing for further analysis. When analyzed exclusively for location (supratentorial, infratentorial, supra/infratentorial, spinal; n = 43), differences were highly significant (ANOVA, P = 0.001), with supratentorial ATRT (mean = 1.1 cm) showing more edema compared with infratentorial (mean = 0.06 cm) or supra/infratentorial (no edema) tumors (Supplementary Figure S2). A dichotomy analysis for supra- or infratentorial location (cases with supra/infratentorial, spinal location excluded; n = 37) revealed the same significance (ANOVA, P = 0.001), with more edema in supratentorial ATRT compared with infratentorial tumors. Within the groups of supratentorial tumors, we found no significant differences between subgroups (P = 0.77). There was a significant difference in distribution of edema in infratentorial ATRT between molecular subgroups (P = 0.03), with no edema in infratentorial ATRT-SHH (0/6), little edema in 1/12 ATRT-MYC, and “medium”-grade edema in 1/1 ATRT-TYR.

There were also significant differences (P = 0.01) regarding the occurrence of peripheral cysts (present in 31/43 or 72% of all cases). Whereas only 4/10 ATRT-MYC (40%) showed cysts in the tumor periphery, those cysts were found in 12/17 (71%) ATRT-SHH and in up to 15/16 (94%) ATRT-TYR (Fig. 2). With respect to tumor location (supratentorial P = 0.22, infratentorial P = 0.21), there were no significant differences in distribution of peripheral cysts between subgroups, demonstrating that this finding is not primarily location dependent.

When analyzed for either presence or absence of contrast enhancement, ATRT-SHH did not show any enhancement in 5/17 cases (29%; Fig. 3C, D), whereas all ATRT-TYR and 9/10 (90%) ATRT-MYC enhanced. These findings were statistically significant (P = 0.047). There was a trend (P = 0.05) regarding gadolinium enhancement intensity toward strong enhancement in ATRT-TYR (7/16 or 44%; Fig. 3A) and ATRT-MYC (7/10 or 70%; Fig. 3B), whereas only 3/17 (18%) ATRT-SHH showed strong contrast enhancement. We found a bandlike enhancement (previously reported as characteristic for ATRT) in 12/43 patients (28%; Fig. 3A, B). This pattern was present throughout all molecular subgroups (31% of ATRT-TYR, 29% of ATRT-SHH, 20% of ATRT-MYC) at similar rates. There was no significant difference with respect to enhancing percentage of solid tumor between ATRT subgroups (P = 0.32).

The results in the categories of T2 signal (P = 0.10), ADC (P = 0.82), and intratumoral hemorrhage (P = 0.26) were as well not significantly different in the overall analysis.

When calculated for supra- and infratentorial tumors separately, we found significant differences regarding ADC (P = 0.03) with 1/2 infratentorial ATRT-MYC and 1/5 supratentorial ATRT-SHH showing high ADC values compared with normal brain, whereas in all other tumors ADC was either low (diffusion restriction) or not provided. In the remaining categories, we did not find significant differences with respect to supra- or infratentorial tumor location.

Regarding tumor dissemination, we found a total of 6/43 tumors (14%) with macroscopic (detectable by MRI, corresponding to M2 status) meningeal spread; however, we found no significant differences between molecular subgroups (ATRT-TYR, n = 19%; ATRT-MYC, 20%; ATRT-SHH, 6%) for this category. Furthermore, there was only a single tumor (ATRT-TYR) with a synchronous lesion (Supplementary Figure S3D) and no tumor with bone involvement in our cohort, not allowing for statistical analysis in these categories. Table 1 summarizes the most relevant MRI features of ATRT molecular subgroups. Supplementary Table S1 displays a detailed summary of the analyzed data, with P-values listed for each category.

Discussion

Several studies described MR imaging characteristics of ATRT without molecular stratification.^11,13–15 We found significant differences between molecular ATRT subgroups in the categories of tumor location, peritumoral edema, peripheral cysts, diffusion restriction, and contrast enhancement.

On a molecular level, recent data demonstrate that ATRT is not a homogeneous disease, and 3 distinct molecular subgroups with preferred locations in the brain have been identified.^4,10 In this study, we confirm the findings of Johann et al that ATRT-TYR predominantly occur infratentorially, whereas ATRT-MYC tumors were preferentially of supratentorial location, and ATRT-SHH tumors can be found supra- and infratentorially in a balanced distribution.⁴ However, we found some ATRT-TYR tumors that were located supratentorially, as well as ATRT-MYC tumors with an infratentorial location. Thus, supra- or infratentorial location by itself is not an exclusive parameter to differentiate ATRT subgroups, which is in line with previous findings.⁴ Case-based atypical locations have been described for ATRT (eg, with tumors occurring with relation to peripheral nerves or originating from the cavernous sinus^16–18); furthermore, bone involvement and synchronous tumors have been reported.^12,19,20 It has been suggested that an off-midline location may be characteristic for infratentorial ATRT,^13,14,21 in comparison to other tumor entities. However, we found 50% of midline tumors (cerebellar vermis, fourth ventricle) in infratentorial ATRT-TYR, which agrees with our previous study showing 5/11 infratentorial tumors arising from midline structures (cerebellar vermis, pons).¹¹ An off-midline location thus seems not specific for ATRT in the posterior fossa and may not be of help as a single criterion to differentiate ATRT from medulloblastoma. Differences in anatomical location (supra/infratentorial) and remarkable epigenetic differences between ATRT subgroups may nonetheless point to a different cellular origin. However, no specific cells of origin have been described for molecular subgroups in ATRT to date. Our analysis of exact neuro-anatomical location of ATRT did not result in significant differences between molecular subgroups. Here, it is of note that preferably infratentorial tumors (ATRT-TYR and ATRT-SHH) could not always be assigned to their potential origin due to large tumor size. We found a single ATRT-MYC with spinal location, and only very few tumors with relation to the pineal gland and thalamus/basal ganglia. There were, however, no tumors that originated from midbrain structures/pons.

We detected peritumoral edema of varying degrees in 30% of our cases, compared with 52% in our previous study.¹¹ Other authors found opposing results, from no significant edema¹³ to 100% edema.¹⁵ For infratentorial ATRT, statistical analysis revealed significance between molecular subgroups (P = 0.03); however, only 2/21 infratentorial tumors showed edema, thus this result should be interpreted with caution. The discussion of the highly significant finding that edema occurred more often in supratentorial ATRT compared with infratentorial tumors seems more complex. On one hand we found that ATRT-MYC show significantly more edema compared with ATRT-SHH and ATRT-TYR, with mean edema width in ATRT-MYC > ATRT-SHH > ATRT-TYR. The distinct anatomical location of molecular ATRT subgroups (with ATRT-MYC preferably in supratentorial location, ATRT-TYR in infratentorial location, and ATRT-SHH in infra- and/or supratentorial location) may, however, indicate that location is an important factor with respect to peritumoral edema, considering that supratentorial tumors have more space to grow and tend to be clinically symptomatic at later stages (with more edema). Thus, location seems to be related to molecular subgroups, whereas edema seems to be related to location. Distribution and width of edema may be interpreted in line with tumor location, in order to differentiate molecular ATRT subgroups in MRI.

An imaging parameter that has been described as typical for ATRT is the evidence of cysts in the periphery of the tumor, separating solid tumor components from the adjacent brain.^11,13–15 In contrast to these studies (“eccentric” cysts in 20%–55% of cases), we found peripheral cysts in up to 72% among all ATRT cases, with higher frequency in infratentorial tumors (81% vs 63%). This is somewhat in contrast to our previous study (without molecular subgrouping), where peripheral cysts were seen in 13 tumors (9 supratentorial and only 4 infratentorial).¹¹ Interestingly, the frequency of peripheral cysts differed significantly between molecular subgroups, ranging from only 40% in ATRT-MYC to 71% in ATRT-SHH and even 94% in the ATRT-TYR group. We further demonstrate that these differences are not dependent on tumor location (supra/infratentorial), thus molecular subgroup may significantly influence the occurrence of peripheral cysts. We cannot explain this finding on a molecular basis, but consider it to be important regarding patient stratification in ATRT subgroups by MRI. However, a multivariate analysis was not possible on our dataset, and thus our results may need confirmation on a larger cohort.

Diffusion restriction as shown by low ADC (compared with normal brain) can be regarded as typical for most ATRTs (and throughout all molecular subgroups in our study), which has been demonstrated before.^{11,17,21–25} Our significant finding that ADC may appear high in very few ATRT-MYC and ATRT-SHH tumors may not be regarded as helpful for clinical use, due to low sample size and lack of ADC data in 23% of cases in our study (DWI/ADC not provided by referring centers). Both low to intermediate signal in T2 and restricted diffusion reflect the high cellularity of solid tumors. As a prediction of molecular subgroups by T2 and ADC seems impossible (although we did not find high ADC values in any ATRT-TYR), these 2 parameters are rather able to help differentiate ATRT from other tumor entities with low cellularity, such as low-grade glioma.

Following gadolinium administration, we observed variable degrees of contrast enhancement in the majority of tumors, consistent with the literature.^{13–15,21,25} However, 6/43 tumors (14%) did not enhance, a finding that was reported before in up to 16% of ATRT.^11,23 Interestingly, the majority of the non-enhancing tumors (5 of 6) were ATRT-SHH (29% of ATRT-SHH subgroup), whereas none of ATRT-TYR and only 10% of ATRT-MYC were non-enhancing. In contrast, ATRT-TYR and ATRT-MYC displayed strong enhancement more frequently, compared with ATRT-SHH. Although not present in the majority of ATRT, a distinct and unusual pattern of a “wavy” bandlike enhancement surrounding a central hypointensity has been described in 38% of cases in our previous study.¹¹ We found this typical enhancement pattern in 28% of all ATRT, which basically matches our previous findings. Interestingly, this frequency was consistently found throughout all molecular ATRT subgroups (31% of ATRT-TYR, 29% of ATRT-SHH, 20% of ATRT-MYC).

Limitations of this study originate in its retrospective design and lack of a validation cohort. Furthermore, MRI studies provided for central neuroradiological review were obtained with scanners from different manufacturers and with different sequence parameters. The multicenter approach within the HIT trials, however, provides us with a comparatively large cohort of rare tumors such as ATRT, including molecular characterization/subtyping. Advanced imaging techniques such as MR spectroscopy or MR perfusion, which harbor the potential for a more accurate assessment of the tumors’ metabolic and proliferative activity, were not generally provided by the referring centers. To date, only very few case-based reports exist applying these techniques in ATRT.^25–27 Future studies may further investigate anatomical, functional, and molecular imaging of molecular subgroups during treatment and surveillance/relapse of ATRT in a prospective fashion.

With this study on a cohort of 43 patients, we first provide detailed information on MR imaging characteristics and neuro-anatomical location of molecular ATRT subgroups. We found a morphological spectrum showing tumors of high cellularity in T2 and DWI/ADC, and we confirm that location (supra/infratentorial) can be a valuable parameter to assign tumors to one of the molecular subgroups. It is of note that ATRT-MYC show more edema compared with ATRT-SHH and ATRT-TYR, but edema formation may be related to (supratentorial) location rather than to molecular subgroups. In contrast to previous studies, we show that an off-midline location is not specific for infratentorial ATRT. The findings of peripheral cysts and/or a bandlike “wavy” enhancement, if present, are important characteristics to differentiate ATRT from other tumor entities in children <3 years of age, and could be observed throughout all molecular subgroups. Interestingly, peripheral cysts occurred most frequently in ATRT-TYR. A strong contrast enhancement points at ATRT-TYR or ATRT-MYC subgroups, whereas enhancement was absent in almost one-third of ATRT-SHH tumors. Intratumor hemorrhage is a common, yet unspecific MRI feature in ATRT, and meningeal dissemination can be absent initially. In summary, a differentiation of ATRT molecular subgroups by means of MRI alone seems difficult at present and needs further evaluation.

Recent research demonstrated that ATRT subgroups are associated with distinct genotypic, chromatin, and functional landscapes that correlate with cellular responses to various signaling and epigenetic pathway inhibitors; compounds specifically targeting these pathways or agents that alter the epigenetic state of the cell are currently being evaluated.^8,10 The knowledge of MRI surrogates for molecular ATRT subgroups, provided by this study, might contribute to risk assessment and individual patient stratification within the frameworks of radiogenomics.

Funding

This work was supported by the German Childhood Cancer Foundation. M.H. is supported by DFG (HA 3060/8-1) and the IZKF Münster (Ha3/019/15). EU-RHAB is supported by grants to M.C.F. by the “Deutsche Kinderkrebsstiftung” DKKS 2014.04, the parents organization Lichtblicke, Augsburg and the parents organization Horizont, Weseke, Germany.

Acknowledgments

We thank P. Neumayer and I. Lechner for expert assistance in data acquisition and managing.

Conflict of interest statement

None declared.

Authorship statement

This work evolved from different institutions within the framework of the multicenter HIT studies. We state that all authors meet the ICMJE criteria of authorship:

- Substantial contributions to the conception or design of the work; or the acquisition, analysis, or interpretation of data for the work; AND

- Drafting the work or revising it critically for important intellectual content; AND

  Final approval of the version to be published;

MRI reader 1, writing of manuscript (J.N.), EU-RHAB Registry and management of patient data (K.N., M.C.F.); statistical analysis (A.H.); interpretation of data (radiology and pathology), writing of manuscript (L.A.V.); molecular analysis (neuropathology), interpretation of data (neuropathology) (M.H.); interpretation of patient and molecular data (pediatric oncology) (P.D.J., M.K.); interpretation of patient and molecular data (pediatric oncology), EU-RHAB Registry (M.C.F.), MRI reader 2, interpretation of MRI data (neuroradiology) (M.W-M.).

References

Author notes