Atypical Teratoid Rhabdoid Tumors (ATRTs): The British Columbia's Children's Hospital's Experience, 1986–2006

Abstract

As “atypical teratoid rhabdoid tumors” (ATRTs) may mimic “small round blue cell tumors” (SRBCT), we reexamined our ATRT experience focusing upon INI‐1 immunohistochemistry (IHC). All high‐grade pediatric brain tumors occurring from 1986–2006 at our institution underwent INI‐1 IHC. Clinicopathologic data from each INI‐1 immunonegative case were reviewed. Additional genetic, epigenetic and IHC analyses (including interrogation of INI‐1 and CLDN6) were performed on a subset of the INI‐1 immunonegative cases. Twelve INI‐1 IHC negative tumors were identified retrospectively, of which only two previously carried the diagnosis of ATRT. Overall, the clinicopathologic and genetic data supported the assertion that all 12 cases represented ATRT. Unexpectedly, three long‐term survivors (4.2, 7.0 and 8.5 years) were identified. As hypothesized, “teratoid” and “rhabdoid” histologic features were relatively infrequent despite gross total resections in some cases. Methylation specific polymer chain reaction (PCR) (MSP) revealed a uniform methylation pattern across all cases and gene promoters tested (ie, MGMT, HIC1, MLH3 and RASSF1); notably, all cases demonstrated unmethylated MGMT promoters. Our data demonstate that a primitive non‐rhabdoid histophenotype is common among ATRTs and highlights the diagnostic importance of INI‐1 IHC. Epigenetically, the MGMT promoter is usually unmethylated in ATRT, suggesting that potential temozolomide‐based chemotherapy may be of limited efficacy.

INTRODUCTION

Atypical teratoid rhabdoid tumor (ATRT) was originally identified by Rorke et al in 1996 as a rare and aggressive form of pediatric neoplasia (36). Subsequent studies have further delineated this central nervous system (CNS) entity (7). Based on recent data from the Central Brain Tumor Registry of the United States (CBTRUS), the incidence of ATRTs is estimated at 0.5/1 million (10). Children less than 2 years of age are typically affected, and these tumors are almost equally distributed between the supratentorial and infratentorial compartments. Most reports have suggested a slight male predominance (29). Despite multimodal treatment, including intensive chemotherapy, surgery and at times radiation, prognosis is very poor and most patients fail to survive 1 year after diagnosis 2, 11, 14, 15, 21, 34, 43.

The “teratoid” component of ATRTs is derived from histologic evidence of divergent differentiation along mesenchymal, neuroectodermal and epithelial lines 7, 29, 36. While the latter may be appreciated on routine stains, it is often far better highlighted via immunohistochemistry (IHC). Prior to the advent of INI‐1 IHC in 2004, the quintessential triad of immunopositivity in ATRT included vimentin, epithelial membrane antigen (EMA) and smooth muscle actin (SMA). The “rhabdoid” component of ATRT is often considered its defining histologic feature; classically, a large vesicular nucleus is associated with a voluminous belly of eosinophilic cytoplasm that may bear a smaller, round and intensely eosinophilic inclusion. Electron microscopic studies have proven the latter to represent a collection of intermediate filaments. Nonetheless, a component of primitive “small round blue cells” (SRBCs) may be seen in many ATRTs and justify their classification among the embryonal (grade IV) CNS neoplasms by the World Health Organization (WHO) (16).

The genetic abnormality in ATRT involves biallelic genetic alterations of the INI‐1/SMARCB1/hSNF5/BAF47 gene on 22q11.2 3, 21, 26, 31. INI‐1 encodes a protein that is a member of the ATP‐dependent switching/sucrose non‐fermentable (SWI/SNF) complex that is thought to be important in chromatin remodeling and cell cycle regulation 24, 25. Initial genetic studies suggested that approximately 75% of ATRTs were characterized by biallelic INI‐1 inactivation (3). As such, epigenetic mechanisms of INI‐1 inactivation were hypothesized to be involved in some cases of ATRT. However, Zhang et al (2002) were unable to detect INI‐1 promoter methylation in their cohort of 24 pediatric rhabdoid tumors (42). Epigenetic investigations of other genetic loci in ATRT have been infrequent, but in one study, Muhlisch et al (2006) demonstrated epigenetic repression of the tumor suppressor gene RASSF1A in four of six ATRTs using a combination of semiquantitative methylation specific polymer chain reaction (PCR) (MSP), bisulfite sequencing and real‐time polymer chain reaction (PCR) (RT‐PCR) (33). In a recent multiplatform genetic study by Jackson et al, nearly 100% of rhabdoid cases (including 36 ATRTs) demonstrated a variety of INI‐1 alterations that were typified by relatively large deletions, but also mutations (especially affecting exons 5 and 9) and loss of heterozygosity (LOH). Based upon their genome‐wide single nucleotide polymorphism array analyses, Jackson et al concluded that INI‐1 alterations per se were the sole pertinent genetic abnormality in ATRTs (26).

Identification of INI‐1 abnormalities in ATRT ultimately led to the development of INI‐1 IHC. As first illustrated by Judkins et al in 2004, the tumor cells of ATRT lose their normally ubiquitous nuclear positivity (27). Originally, the loss of INI‐1 staining was felt to be a sensitive and specific marker of ATRT, but recent studies have demonstrated similar findings in a number of tumors including epithelioid sarcoma, familial schwannomatosis, myoepithelial carcinoma, epithelioid malignant peripheral nerve sheath tumor, oligodendroglioma and cribiform neuroepithelial tumor 18, 22, 27, 35. However, concomitant mutation of INI‐1 has only been confirmed in a few of the foregoing entities. Despite this seeming lack in specificity, most neuropathologists would consider INI‐1 immunonegativity to be virtually diagnostic of ATRT in the context of a high‐grade, primitive, CNS tumor occurring in a very young child. Nonetheless, rare cases of histologically classic ATRT have been described that retain nuclear expression of INI‐1 (40). Genetic analyses by Hasselblatt et al on such an INI‐1 IHC retained case demonstrated mutation in another member of the SWI/SNF complex, SMARCA4/BRG1. This latter finding would support the notion that the SWI/SNF complex, not just INI‐1, plays a primary role in the generation of ATRT histophenotype (19).

As INI‐1 immunonegativity does not appear to be restricted to ATRTs, some researchers have sought out alternative molecular/immunohistochemical markers. Using Affymetrix GeneChip® microarrays (Santa Clara, CA, USA) and IHC, Birks et al demonstrated that ATRTs overexpress claudin‐6 (CLDN6). As such, CLDN6 IHC was proposed as a novel “positive” marker of histologically classic (ie, with rhabdoid cells) ATRTs (4)

In emphasizing the utility of INI‐1 IHC, Judkins et al raised the possibility of INI‐1 IHC capturing a group of ATRTs with less classic morphology (27). Subsequently, Haberler et al (2006) reviewed their 40‐year experience with high‐grade pediatric CNS neoplasms (16). There were 289 neoplasms identified, rereviewed and stained with INI‐1 IHC. Twenty‐six INI‐1 negative tumors were documented. Only six of the 26 originally carried the diagnosis of ATRT. Upon histologic rereview and, without knowledge of the results of INI‐1 staining, a total of 18 of the 26 cases were considered to represent ATRT. Six medulloblastomas and two supratentorial primitive neuroectodermal tumors (sPNET) were thought to comprise the remaining eight cases. Given these results, Haberler et al suggested that a subset of ATRTs may demonstrate a paucity of rhabdoid cells due in part to suboptimal tumor sampling.

INI‐1 IHC was introduced at our institution in 2006. Prior to 2006, only two ATRTs were documented in our archives since 1986. In contrast, after the commencement of INI‐1 testing, three ATRTs were recognized in the following 3 years (ie, n = 5, 1986–2009). Based upon Haberler et al's data, we would expect that approximately 10% of our high‐grade pediatric CNS tumors represent ATRT. Alternatively, given recent CBTRUS data (incidence of ATRTs ∼0.5/1 million) and the greater Vancouver population (∼2 million), we would estimate an ATRT incidence of approximately one case/year (10). As such, we hypothesized that several ATRTs had gone unrecognized at our institution, perhaps in part because of the occurrence of “cryptic” cases (ie, ATRTs with a predominance of primitive cells and few teratoid features/rhabdoid cells).

METHODS

All high‐grade (ie, malignant) primary pediatric CNS tumors diagnosed from 1986 to 2006 at British Columbia's Children's Hospital (BCCH) were targeted for evaluation. With local research ethics board approval (#H09‐01657), tumors were identified with the aid of two BCCH databases; the first is maintained in the Division of Hematology, Oncology and Bone Marrow Transplantation and records all pediatric brain tumor cases, while the second resides in the Division of Anatomic Pathology. All potential cases identified through these databases were assigned a random three‐digit study number.

Patients from birth to 19 years of age at the time of original diagnosis were included for study. Low‐grade tumors, as well as cases with unavailable slides and blocks, were excluded from the study. In addition, upon initial pathologic review, those tumors whose histology clearly resided outside of the documented spectrum of ATRT (eg, classic glioblastoma), or those cases with inadequate viable tissue for IHC, were excluded from further analysis.

A representative formalin fixed and paraffin embedded (FFPE) block from each case was selected for INI‐1 IHC. Tissue was sectioned to a thickness of 4 µm. Purified murine Anti‐BAF47 Antibody (BD Transduction Laboratories, Mississauga, ON, Canada) was utilized at a dilution of 1:100, and samples were stained with the aid of the Ventana BenchMark XT automated stainer (Ventana Medical Systems Inc., Tucson, AZ, USA). Antigen retrieval was performed as per Ventana's protocol. Loss of the normally ubiquitous INI‐1 nuclear staining by light microscopy was only considered to represent a true negative if the appropriate internal positive control staining (ie, endothelial cells and lymphocytes) was present within the tumor. Additional external positive and negative control staining was performed as per routine protocol.

A detailed histologic analysis, using all of the original microscopic slides, was performed on the tumors that exhibited loss of INI‐1 staining. The hematoxylin and eosin (H&E) slides were assessed for the following features: rhabdoid cells, vesicular nuclei, vacuolated tumor cell cytoplasm, number of mitoses, necrosis, microvascular proliferation (MVP), tumor architecture (eg, sarcomatous, jumbled etc.), epithelial differentiation and SRBCs (ie, hyperchromatic enlarged nuclei with essentially no cytoplasm). With regard to the latter, while not required, the presence of “nuclear molding” was considered to be supportive evidence for the presence of SRBCs. The foregoing histologic features were assessed semiquantitatively (ie, 0: absent; 1+: rare; 2+: scattered; 3+: frequent) where appropriate. The results of the original IHC studies were also recorded.

Patient charts were reviewed. The following de‐identified clinical data were recorded: age at diagnosis, date of diagnosis, tumor location, presence of metastatic disease at diagnosis, extent of surgical resection, type of chemotherapy, radiation dose, time to relapse, survival in months from date of diagnosis and current clinical status. All data were entered into Microsoft Excel wherein the median event‐free and overall survivals were calculated. The results of cytogenetic analyses performed on the respective tumors were also recorded.

Genetic testing to support the clinicopathologic diagnoses of ATRT was performed on a subset of the INI‐1 immunonegative cases. Fluorescence in situ hybridization (FISH): FISH analysis was carried out on FFPE tissue sections, as previously described (28). In brief, a dual‐color FISH probe set, manufactured by Vysis (Abbott Laboratories Inc., Des Plaines, IL, USA), was used to test for monosomy 22. DNA probes directed to 22q11.2 (TUPLE1/HIRA) and 22q13 (ARSA) were used. PCR and sequencing of INI‐1 exons: DNA extracted from FFPE tumor tissue was used for targeted PCR amplification of the entire INI‐1 coding region. Sequences for the exonic PCR primers are listed in the supplemental data section (Supporting Information Table S1) and are modified from Hulsebos et al (23). PCR was performed as previously described, and targeted bidirectional sequencing was done using standard Sanger sequencing (8). Analysis of DNA tracings was carried out using Mutation Surveyor version 3.2 (Softgenetics, State College, PA, USA). Multiplex ligation‐dependent probe amplification (MLPA): the SALSA P258 MLPA‐SMARCB1 test kit (MRC‐Holland, Amsterdam, The Netherlands) was used to determine copy number changes in INI‐1. MLPA was performed following the manufacturer's instructions. The reaction products were detected with an ABI‐3100 Genetic Analyzer (Applied Biosystems, Foster City, CA, USA). To size the PCR products and obtain peak areas the GeneMapper software (Applied Biosystems) was used. These data were exported into GeneMarker software (Softgenetics) for MLPA analysis.

CLDN6 gene expression microarray, quantitative RT‐PCR (QRT‐PCR) and IHC analyses: both CLDN6 gene expression microarray and QRT‐PCR analyses were performed as previously described (4). In brief, INI‐1 immunonegative samples were evaluated for gene expression using the Affymetrix U133Plus2 GeneChip microarrays. Surgical specimens that were snap frozen and stored in liquid nitrogen were utilized. RNA was extracted from the samples and then amplified and labeled. The quality of the RNA was verified and subsequently hybridized to the HG‐U133 Plus 2 GeneChips (Affymetrix). Data analysis was done using the R programming language (http://www.r‐project.org/). Microarray data from the processed tumor samples were background corrected and normalized using the gene chip robust multi‐array average (gcRMA) algorithm, resulting in log 2 expression values. Differential expression of genes was determined using limma, which employs an empirical Bayes approach to calculate a moderated T‐statistic. CLDN6 mRNA expression was assessed via QRT‐PCR using the Chromo4 Real Time Detector (Bio‐Rad, Hercules, CA, USA); the resulting log 2 values were analyzed and compared with the gene expression microarray for concordance. CLDN6 IHC was performed on FFPE tissue sections as previously described (4). After antigen retrieval, the subsequent steps were performed using the EnVision‐HRP kit (Dako, Glostrup, Denmark) on a Dako autostainer according to standard protocol. A primary rabbit polyclonal anti‐CLDN6 antibody (01‐8865; American Research Products, Belmont, MA, USA) was used at a dilution of 1:66. CLDN6 IHC was scored semiquantitatively as follows: “0,” no positive tumor cells; “1+,” focal positivity (<5%); “2+,” 5%–25% positive tumor cells; “3+,” >25% positive cells. Only membranous staining was considered to represent a true positive.

Epigenetic studies were performed on all INI‐1 immunonegative cases. Tissue samples and DNA stocks: de‐identified unstained slides cut from FFPE or snap frozen tissue was utilized for testing. Tumor tissue from unstained FFPE slides (taken from 10 sections cut at 4 µm) was deparaffinized in xylene followed by immersion in graded alcohols, followed by water, until rehydration. The tissue was incubated with proteinase K at 60°C overnight. Subsequently, genomic DNA was extracted using the Gentra PureGene Kit (Qiagen, Hilden, Germany) as per the manufacturer's protocol. DNA quantitation was performed using a Nanodrop ND‐1000 ultra‐violet (UV)‐Vis spectrophotometer (Nanodrop Technologies, Wilmington, DE, USA). Methylation analysis: we performed promoter methylation analyses of the following genes given their involvement in other forms of CNS neoplasia: MGMT, HIC1, MLH3 and RASSF1. Published protocols for MSP were utilized 9, 12, 13, 32, 44. A total of 500 ng of genomic DNA underwent bisulfite conversion using the EpiTect Bisulfite conversion kit (Qiagen) followed by MSP of the converted DNA using separate methylated‐ and unmethylated‐specific PCR reactions. Genomic DNA from Jurkat cell line, methylated excessively by Spiroplasma sp. strain MQ1 (SSSI) (New England Biolabs, Beverly, MA, USA), served as the positive control. Genomic DNA from a normal male donor (Promega, Madison, WI, USA) served as our negative control. The PCR products were separated electrophoretically in 1.5% agarose gels with ethidium bromide and visualized under UV illumination.

RESULTS

A total of 114 cases of high‐grade primary pediatric CNS neoplasia were diagnosed at BCCH from 1986−2006. In eight instances, the slides and blocks could not be located, and as such, these cases were excluded from further analyses. On initial histologic review, 12 cases were felt to be incompatible with the spectrum of pathology seen in ATRT; these cases were also excluded. Therefore, INI‐1 IHC was performed on 94 cases.

Clinical data (see Table 1)

Table 1.

Clinical data on INI‐1 negative cases (n = 12)

Case #	Age, year/sex (M/F)	Initial diagnosis	Site	Metastases at DX*	SX	RTX (grays)	Chemotherapy	Relapse site	Current status	Length of survival (years)
174	0.3F	MBL	CBL	M0	STR	No	VP16/VCR/CPM/CDDP 5x (CCG 9921)	Local	DOD	0.47
163	0.5F	MBL	CBL	M0T2	GTR	No	Intensive + AUTO (HS III REG D)	Local	DOD	1.10
206	0.8M	ATRT	Brainstem	M0	PR	No	None	Local	DOD	0.16
166	1.0F	MBL	CBL	M0	?	No	Intensive + AUTO (HS III REG D)	Local	DOD	0.87
200	1.3M	MBL	CBL	M0T4	STR	No	Intensive + AUTO (HS III REG D)	N/A	Alive	4.2
147	1.4F	MBL	CBL	M0T3a	GTR	No	Intensive + AUTO (COG 99703)	N/A	Alive	8.5
112	1.8F	SPNET	Frontal	M0	GTR	No	Intensive + AUTO (HS II)	Ventricles	DOD	1.41
179	3.1F	MBL	CBL	M0	GTR	Craniospinal, (54/23.4)	VCR/CDDP/CCNU 8x	Brain and spine	DOD	1.73
207	3.8M	ATRT	CBL	M3T4	STR	At relapse (45.5)	VICE 2x, then Docetaxel (CCG0962)	Brain and spine	DOD	2.11
209	0.9F	SPNET	Pineal	M0	PR	No	VP16/VCR/CPM/CDDP 5x (CCG 9921 A)	N/A	DOD	0.44
115	1.5F	MBL	CBL	?	?	No	None	N/A	DNOD†	0.76
212	9.0F	Gliosarcoma	Parietal	M0	GTR	Focal (59)	VCR, PRED, CCNU 8x (CCG 945)	N/A	Alive	7

^* Chang staging.

^† Died of cardiac arrest.

ATRT = atypical teratoid rhabdoid tumor; AUTO = autologous stem cell transplant; CBL = cerebellum; CPM = cyclophoshamide; CDDP = cisplatin; CCNU = lomustine; CCG = Children's Cancer Group; COG = Children's Oncology Group; DOD = dead of disease; DNOD = dead not of disease; DX = diagnosis; F = female; GTR = gross total resection; HS = Head Start; M = male; MBL = medulloblastoma; PR = partial resection; PRED = prednisone; REG = regimen; RTX = radiotherapy; STR = subtotal resection; SPNET = supratentorial primitive neuroectodermal tumor; SX = surgery; VCR = vincristine; “VICE” = vincristine, ifosfamide, carboplatin and etoposide, VP16, etoposide; ? = unknown.

Twelve cases, including the two previously diagnosed cases of ATRT, were found to be INI‐1 immunonegative (ie, 12/94 = 12.7%). Nine of the 12 were female. Age at diagnosis ranged from 0.3 to 9 years; 75% were <2 years old. Aside from the two previously known cases of ATRT, initial diagnoses included seven medulloblastomas, two supratentorial primitive neuroectodermal tumor (sPNET, frontal lobe and pineal gland, respectively) and one gliosarcoma (parietal lobe); accordingly, the majority of these INI‐1 immunonegative cases were infratentorial (9 of 12). Metastatic status at diagnosis was known in 11 of 12 cases; only 1 of 11 presented with metastatic disease (case #207). Partial, subtotal or gross total surgical resections were performed on 10 of 12 cases with the extent of resection being judged from the operative reports. Three of 12 patients underwent radiation therapy and 10 of 12 received varied forms of chemotherapy (see Table 1 for details). Tumor relapse was noted in seven of the 10 patients where such data were recorded; relapse was local in five, diffuse (ie, brain and spinal cord) in two and ventricular in one (see below). The vast majority of subjects were dead of disease (DOD) within 1 year; the median event‐free and overall survivals were both ≤1 year (0.76 and 1 year, respectively). At the time of last follow‐up, nine were deceased [only one “due not to disease,” (DNOD)] and, remarkably, three long‐term survivors were identified (#200: 4.2 years, #212: 7.0 years and #147: 8.5 years). No clear correlation was observed between long‐term survivorship and the particular type of therapy received. However, among the five patients that received myeloablative chemotherapy, two were long‐term survivors (#147 and #200).

Pathology (see Table 2)

Table 2.

Pathologic data on INI‐1 negative cases (n = 12)

Case #	Additional patterns*	Cytoplasmic vacuoles	Vesicular nuclei	Rhabdoid cells	SRBCs	Necrosis	MVP	Mitoses
174	SARC +	2+	1+	1+	1+	Yes	No	2–3+
163	SARC +	1+	1+	0	1+	Yes	No	3+
206	SARC + and papillae +	3+	2+	2+	1+	Yes	Yes	3+
166	None	3+	2+	1–2+	1+	Yes	Yes	3+
200	None	1+	1–2+	0	1+	Yes	No	1+
147	None	2+	1+	0	2+	Yes	No	2+
112	SARC +	1+	1+	0	3+	Yes	No	1+
179	None	0	1–2+	1+	1+	Yes	No	3+
207	Chordoid‐like +	1+	2+	2+	1+	Yes	No	2+
209	None	2+	2+	0–1+	0	Yes	No	3+
115	SARC ++	1+	1–2+	1+	1+	Yes	No	3+
212	SARC +++	0–1+	1+	0	1+	Yes	Yes	3+

The frequency/relative abundance of a given histologic feature was semiquantitatively determined (ie, 0: absent; 1+: rare; 2+: scattered; 3+: frequent).

^* All tumors displayed a predominantly jumbled/disorganized pattern (except case #212) in addition to the patterns listed.

MVP = microvascular proliferation; SARC = sarcomatous; SRBCs = small round blue cells.

In general, the histology from all cases was remarkably similar. By definition, all cases chosen for analysis were high grade (ie, malignant), and histologic rereview was confirmatory. These cases were generally typified by a combination of high mitotic rates and necrosis. MVP was identified in a subset (3 of 12). Somewhat contrary to the literature, SRBCs (as strictly defined) were relatively infrequent in any given case (10 of 12 cases had “0” to “1+”); nonetheless, nearly all cases (11 of 12) demonstrated at least some SRBCs (Figure 1A). While rhabdoid cells were seen in the majority of cases (7 of 12), they were either rare (n = 4) or absent (n = 5) in 75% of cases (9 of 12) (Figure 1B). In contrast to a rhabdoid cell‐rich picture, cytoplasmic vacuoles were seen in the tumor cells of most cases (10 of 12) and were prominent in five (Figure 1A, left, C and D). In addition, vesicular tumor nuclei were also typical (12 of 12), but generally speaking, the tumor nuclei were not large and did not display the prominent large nucleoli of classic “rhabdoid cells” (Figure 1A–C). Architecturally, a jumbled pattern was most common (11 of 12) (Figure 1D). A subset of cases additionally exhibited a somewhat spindled cytology and sarcomatous architecture either focally (n = 5) or diffusely (n = 1) (Figure 1E). Papillae and chordoid‐like morphologies were each focally appreciated in one case. Notably, outright regions of neuroglial or epithelial differentiation (ie, glands or epithelial surfaces) were not appreciated. Rare perivascular pseudorosettes (akin to ependymoma) were seen in one case.

Figure 1.

Representative FFPE microscopic findings. A. Focal SRBCs (right‐middle of the figure). Tumor cells with vacuolated cytoplasm are seen in the left of the image (H&E, ×400). B. Rhabdoid cells (arrows) (H&E, ×400). C. Tumor cells with vacuolated cytoplasm and vesicular nuclei (H&E, ×400). D. Jumbled/disorganized pattern. Even at this low power, one can appreciate the diffuse distribution of vacuolated tumor cells and a cluster of SRBCs (arrow). No signs of divergent differentiation are appreciated (H&E, ×100). E. A fascicle of tumor cells, interpreted as sarcomatous/mesenchymal differentiation, cuts across the figure and is highlighted by arrows (H&E, ×200). F. Loss of INI‐1 staining in tumor cells, with retention in non‐neoplastic endothelial cells, via IHC (×200).

Original IHC investigations varied in scope. With the exception of vimentin (five of five diffusely positive), positive staining patterns of the tumors were usually focal and patchy: glial fibrillary acidic protein (GFAP), five of 11; EMA, three of four; SMA, two of five; synaptophysin, three of nine. Desmin staining was negative in all five cases tested. In keeping with the high‐grade histology on routine staining, the Ki67 proliferative index was qualitatively high in all three cases tested.

Genetic analyses and CLDN6 IHC (see Table 3)

Table 3.

Genetic data and CLDN6 IHC on n = 12 INI‐1 negative cases

Case #	Karyotype: monosomy 22	Loss 22q FISH	INI‐1 exon sequencing: exon; nucleotide change; amino acid effect	MLPA	CLDN6 IHC	CLDN6 mRNA expression (log 2), microarray	CLDN6 mRNA expression (log 2), QRT‐PCR
174	Normal	Yes	5; CGA to TGA; p.R201X	N/A	2+	N/A	N/A
163	Normal	No	#1: 9; c.1145 het_delC; frame shift	N/A	1+	N/A	N/A
			#2: 6; c.665_666het_delTC; frame shift
206	N/A	Yes	N/A	HOMO E1‐9 del	2+	8.1	11.3
166	Normal	Yes	5; c.[501‐1G>A]; splice	Normal	2+	N/A	N/A
200	Normal	No	N/A	N/A	0	N/A	0.2
147	Yes	Yes	N/A	HOMO E1‐9 del	0	7.3	20.6
112	Abnormal*	No§	No mutation found	N/A	0	N/A	N/A
179	Yes	Yes	7; TCA to TGA; p.S284X	N/A	2+	N/A	N/A
207	Normal	F	No mutation found	N/A	2+	N/A	N/A
209	Normal	No	N/A	N/A	1+	N/A	N/A
115	Yes†	F	No mutation found	N/A	1+	N/A	N/A
212	Yes‡	Yes	No mutation found	N/A	0	N/A	1.4

For CLDN6 IHC scoring details please see the main text.

^* 46, XX, t(22;22)(q11or12;q13).

^† 45, XX, t(9;19)(q22;q13.3), −22.

^‡ 46, XX der (1) t(1;1)(p36;q23), del (6)(p21p24), −22.

^§ Loss 22q11.2 only.

N/A = testing not performed; F = failed test; HOMO = homozygous; IHC = immunohistochemistry; FISH = fluorescence in situ hybridization; MLPA = multiplex ligation‐dependent probe amplification; QRT‐PCR = quantitative real‐time polymerase chain reaction.

Cytogenetic analysis was originally undertaken on 11 of the 12 cases. In six cases, karyotyping was normal. Monosomy for chromosome 22 was detected in four cases. Additionally, one case exhibited a t(22;22) (q11 or 12; q13) with a breakpoint involving the INI‐1 locus. Translocations not involving chromosome 22, but affecting other chromosomes, were detected in two cases (#115 and #212); notably, both of the latter also harbored monosomy 22.

FISH analysis yielded results in 10 of 12 cases. Six cases demonstrated evidence of 22q loss, while one case demonstrated loss at 22q11.2 only (ie, solitary loss of the TUPLE1/HIRA probe signal in #112). Three cases did not exhibit any involvement of 22q by FISH. Of note, FISH testing identified two cases with loss of 22q that were normal by karyotype.

MLPA analysis was performed on three of 12 cases. Two of the three cases yielded abnormal results, in particular homozygous deletion of exons 1 through 9. Both of these latter cases also demonstrated chromosome 22 abnormalities by either karyotype and/or FISH.

INI‐1 exon sequencing identified mutations in four of eight cases tested. Five mutations (two in one case) were identified in total. All mutations were considered to have deleterious effects on protein expression. Two of the mutations encoded premature stop codons, two led to shifts in the reading frame and one resulted in a splice site mutation. Three of the four cases demonstrated either monosomy 22 by karyotype and/or loss of 22q by FISH; only one mutated case (#163) demonstrated a normal karyotype and no FISH abnormality (ie, genetic abnormalities were only detected via INI‐1 exon sequencing).

CLDN6: CLDN6 IHC was performed on all 12 INI‐1 immunonegative cases and eight were positive. Interestingly, three of the four immunonegative cases were associated with long‐term survival. Six cases underwent CLDN6 mRNA gene expression microarray analysis, but good quality RNA was obtained in only two cases; both cases demonstrated high gene expression levels of CLDN6. Four of six cases contained enough viable RNA for gene expression via QRT‐PCR; of these four, two expressed high levels of CLDN6 mRNA, while two cases exhibited low levels of CLDN6. Overall, the gene expression data were consistent with the protein levels, as measured by IHC, in all but one case (#147) wherein CLDN6 gene expression was high both by microarray and QRT‐PCR yet IHC was negative.

Methylation assays (see Table 4)

Table 4.

Epigenetics. MSP results on INI‐1 negative cases (n = 12)

Case #	MGMT	HIC1	RASSF1A	MLH3
174	M	M	M	U
163	U	M	M	U
206	F	M	M	U
166	U	M	M	U
200*	U	M	M	U
147*	U	M	M	U
112	U	M	M	U
179	U	M	M	F
207	F	M	F	F
209	F	M	M	U
115	U	M	F	F
212*	U	M	M	M

^* Long‐term survivor.

U = unmethylated; M = methylated; F = test failed.

MSP data generally revealed similar findings across all cases (ie, there was no pattern distinguishing the long‐term ATRT survivors). Promoter MGMT analysis was informative in nine cases and all were unmethylated (Figure 2). HIC1 promoter analysis revealed hypermethylation in all 12 cases. Similarly, RASSF1A testing revealed promoter hypermethylation in 10 cases with readable results. MLH3 analysis revealed an unmethylated promoter in eight of nine readable cases; case #212 was the solitary case demonstrating a hypermethylated promoter.

Figure 2.

MGMT MSP on the 12 INI‐1 immunonegative tumors. A majority of the cases exhibited no methylation of the MGMT promoter, except case #174 that demonstrated promoter hypermethylation. M = methylated allele‐specific MSP assay; U = unmethylated allele‐specific MSP assay; ME = hypermethylated DNA control; UM = unmethylated DNA control.

Long‐term survivors and clinicopathologic correlation

The clinicopathologic features of the three long‐term survivors were generally indistinguishable from the remaining nine cases. Importantly, two of the three long‐term survivors had additional genetic/IHC characteristics that further supported their redesignation as ATRTs; only case #200, a 1.3‐year‐old male originally diagnosed as a medulloblastoma, failed to display 22q11.2 or CLDN6 alterations. Nonetheless, this case demonstrated other features that were in keeping with ATRT (ie, young age and compatible histology, including INI‐1 immunonegativity).

Among the cohort of 12 INI‐1 immunonegative cases, #212 was unique in several respects. This 9‐year‐old female originally carried the diagnosis of “gliosarcoma.” Both her older age and the initial diagnosis were unusual for most confirmed cases of ATRT. Not only did she receive adjuvant radiotherapy (only one other patient received such treatment), but she was the only subject in our series treated with CCG 945‐type chemotherapy [ie, prednisone, lomustine (CCNU) and vincristine (PCV)]. Moreover, her tumor displayed a unique karyotype; in addition to monosomy 22, there was a reciprocal chromosome 1 translocation and a 6p deletion. Light microscopy uniquely revealed a diffuse fascicular architecture and rich intercellular reticulin deposition (Figure 3A–C). GFAP stained numerous and generally solitary large eosinophilic cells that presumably were originally considered to be the “glial” part of the tumor (Figure 3E). Upon initial rereview, some of these latter cells exhibited morphologies suspicious for rhabdoid cells (Figure 3B); however, after INI‐1 IHC, the latter cells retained nuclear positivity suggesting a reactive origin (ie, atypical gliosis) (Figure 3F).

Figure 3.

Histology from the long‐term surviving case #212 that was originally diagnosed as “gliosarcoma.” A. Diffuse fascicular architecture (H&E, ×200). B. Scattered large tumor cells mimic the appearance of rhabdoid cells. In passing, dense eosinophilic strands are seen between tumor cells and are consistent with reticulin (H&E, ×400). C. Diffuse and dense intercellular reticulin deposition (Gordon and Sweet's reticulin, ×200). D. The Ki67 proliferative index is markedly elevated, in keeping with a high‐grade neoplasm (×100). E. GFAP IHC highlights the large eosinophilic cells that are seen on routine staining (×100). F. INI‐1 IHC reveals nuclear negativity in the tumor cells with little cytoplasm (confirming the diagnosis of ATRT), while the cells with abundant cytoplasm (ie, rhabdoid mimics, arrow) retain nuclear positivity.

Although case #147 initially carried a clinical suspicion of neurofibromatosis 2 (NF2) (in light of a suspected unilateral acoustic neuroma by imaging), upon chart review, no other clinical features of NF2 were observed in this subject (including a negative family history). No other unique features were noted in case #147. Likewise, case #200 was also quite typical for ATRT with exception of long‐term survival.

DISCUSSION

Since Rorke et al's initial description of 53 patients in 1996, much has been learned about ATRT (36). As is obvious from their designation, the histologic “teratoid” and particularly the “rhabdoid” features of ATRT were held up as diagnostic pillars in the early literature. Over time, this histologic characterization has evolved and essentially led to the redefinition of ATRT 16, 19, 27. Key to the latter has been the identification of the primary genetic basis of ATRT. As first reported by Versteege et al in 1998, biallelic inactivation of the INI‐1 locus characterizes almost all ATRTs (39). Subsequently, an immunohistochemical assay was developed for the INI‐1 gene product in 2004, and today, INI‐1 IHC has become standard in the histologic workup of ATRT (27). The latter has greatly facilitated diagnosis by highlighting the presence of a subgroup of ATRTs that have a paucity of rhabdoid cells. According to Haberler et al, these ATRTs were essentially indistinguishable from other embryonal CNS neoplasms (ie, sPNET and medulloblastoma) because of tissue under sampling (16).

Retrospective clinicopathologic analysis of each of the 12 BCCH INI‐1 immunonegative cases supports their true designation as ATRT. The vast majority of cases affected children who were ≤2 years old, and despite multimodal treatment, all but three died within approximately 2 years. All of the cases displayed histology falling within the reported spectrum for ATRT. In fact, the most common histophenotype was one that lacked conspicuous teratoid and rhabdoid features and in turn mimicked what many would consider either sPNET or medulloblastoma; however, the prominence of primitive non‐rhabdoid cells with vacuolated cytoplasm among our INI‐1 immunonegative cases was a relatively distinguishing feature. Our supplementary genetic and immunohistochemical data further supported the position that these 12 cases represented ATRT. All but three (#200, #207 and #209) of the 12 cases demonstrated genetic evidence of INI‐1 alteration by at least one testing modality; notably, only one of these three “genetically negative” cases was negative by CLDN6 IHC.

Our study offers additional data that bolsters the general histologic redefinition of ATRT. Prior to INI‐1 IHC, only two ATRTs were diagnosed at our institution from 1986 to 2006. With the aid of INI‐1 IHC, 10 previously unrecognized ATRTs were uncovered. INI‐1 IHC was clearly essential to the identification of these latter cases as rhabdoid cells were “rare to absent” in 9 of 10. In contrast to Haberler et al, our contention is that the relative paucity of rhabdoid cells in these ATRTs was not because of under sampling as four gross totally resected cases exhibited either no (n = 3) or rare (n = 1) rhabdoid cells (16). This data suggests that rhabdoid cells are not essential for the diagnosis of ATRT. Recently, Sunol et al offered a similar perspective given that only three of their 11 ATRTs displayed rhabdoid cells (37). The cellular morphology in our cases was more often typified by primitive small to medium vesicular nuclei that were often associated with vacuolated cytoplasm. Prominent cytoplasmic vacuolation in ATRTs was also remarked upon in the series presented by Burger et al and Wharton et al, yielding an overall “spongy appearance” and “clear cell” foci, respectively 7, 41. Moreover, despite being included in the embryonal category, SRBCs were not, strictly speaking, numerous in our cases, a finding also similar to Burger et al (7). Architecturally, “divergent differentiation” (ie, “teratoid”) was not conspicuous in most of our cases on routine staining. Rather, a disorganized or “jumbled” pattern usually predominated. Divergent differentiation was largely limited to focal regions of sarcomatoid architecture in a subset of our cases. In particular, outright areas of neuroglial or epithelial differentiation were not seen. Therefore, as our cases demonstrate, many ATRTs may not be particularly “teratoid” or “rhabdoid” microscopically (ie, “cryptic”), and as such, the moniker of ATRT may be misleading in a significant number of cases. Accordingly, it behooves neuropathologists to be aware of such cases because ATRTs are generally considered to be more aggressive than medulloblastoma and, accordingly, demand more intensive treatment regimens (31). We contend that enhanced recognition of “cryptic” ATRTs will result in more accurate epidemiologic data and a more refined our conception of this entity (see below).

With the exception of the CLDN6 immunonegative results, our three long‐term survivors did not exhibit clinicopathologic features that facilitated their detection from the remaining INI‐1 immunonegative cases. Nonetheless, some unusual features were particularly appreciated in the long‐term survivors. Case #212 was particularly unique given the patient's older age (9 years) and her histology that mimicked gliosarcoma. The latter is considered highly unusual, as most ATRTs are often misdiagnosed as medulloblastoma, sPNET or choroid plexus carcinoma 6, 16. Case #212 also displayed had a unique karyotype among the cohort of 12 INI‐1 immunonegative cases that featured t(1;1)(p36;q23) and a 6p21‐24 deletion, as well as the typical monosomy 22 (21). Epigenetically, case #212 was unique in exhibiting hypermethylation of the MLH3 promoter.

INI‐1 IHC was instrumental in the identification of three cases of long‐term survival. Despite an abysmal median overall and event‐free survival in our cases as a whole (≤1 year each, respectively), three patients were alive and disease‐free at 4.2, 7.0 and 8.5 years, respectively, following multimodal therapy. Interestingly, each of these three patients underwent a somewhat unique combination of therapy. While two of these subjects received gross total resections, a subtotal resection was achieved in the third. Each of these three long‐term survivors underwent a different form of chemotherapy and only one received radiotherapy. Seemingly, our data would suggest that gross total resection and/or radiotherapy are not mandatory for long‐term survival in ATRT. Our data imply that designing a successful treatment approach in ATRT is challenging and that while multimodal therapy in ATRT is likely beneficial, other factors (eg, underlying tumor biology) may independently influence the ultimate prognosis for a given patient [eg, bone morphogenic pathway (BMP) pathway gene expression, see below](5). Without the aid of INI‐1 IHC, these three long‐term ATRT survivors would not have been detected, and the important clinicopathologic correlations arising from which could not have been made.

The pathogenesis of long‐term survivorship in rare cases of ATRT is unclear. However, our data raise the possibility that CLDN6 molecularly influences prognosis in ATRT. In accordance with Birks et al, we performed CLDN6 IHC and gene expression analysis on our INI‐1 immunonegative tumors. The fact that the majority of our cases illustrated upregulation of CLDN6 by IHC (8 of 12, 75%) is in keeping with the data of Birks et al and supports our contention that the 12 INI‐1 immunonegative tumors are in fact ATRTs (4). Notably, in four of the five cases with both IHC and gene expression data (the latter was obtained either via gene expression microarrays and/or QRT‐PCR), CLDN6 expression was concordant. The solitary non‐concordant case (#147) displayed CLDN6 immunonegativity yet high gene expression levels by both microarray and QRT‐PCR. The latter could be explained by regionally restricted gene/protein expression, support for which emanates from Birks et al who emphasized the patchy nature of CLDN6 IHC positivity (4). An unexpected finding among our cases was the association between decreased CLDN6 activity and long‐term survivorship in ATRT. Of the four CLDN6 immunonegative tumors, three were associated with long‐term survival. As such, this data suggest that CLDN6 may represent a prognostic marker in ATRT.

The CLDN6 data presented herein must be considered preliminary given the limited size of our study and the recent contradictory results reported by Antonelli et al claiming that CLDN6 was an insensitive and nonspecific marker of ATRT (1). Although our study and Birks et al utilized the same CLDN6 antibody as Antonelli et al, differences in staining techniques and immunohistochemical interpretation may account for Antonelli et al's discrepant results 1, 4. In particular, Antonelli et al's use of a more dilute primary CLDN6 antibody (1:100) vs. our study, and Birks et al (1:66) may account for the insensitivity of staining among their ATRT cases (17/59, 29%). Moreover, the nonspecificity of CLDN6 IHC for ATRT as demonstrated by Antonelli et al may in part be attributed to their inclusion of cytoplasmic staining (in addition to membranous) as truly positive. As claudins are a critical component of tight junctions that are found in cell membranes, it seems most apt that only membranous staining should be considered as meaningful and that cytoplasmic staining is nonspecific (4).

Our epigenetic analyses did not distinguish the three long‐term ATRT survivors from the remaining nine cases incurring short survival. Nonetheless, a consistent pattern of promoter methylation was seen across all cases for the individual assays performed (with the exception of MLH3 and case #212 previously.). In particular, our cases of ATRT revealed promoters that were MGMT and MLH3 unmethylated, while HIC1 and RASSF1A were methylated. This epigenetic uniformity across cases would support our assertion that all cases carry a common diagnosis: ATRT. Our data support Muhlisch et al's work that demonstrated RASSF1A methylation in the majority of their ATRTs (four of six) (33). Additionally, our results provide novel baseline data with respect to epigenetic alterations in ATRT. In regard to MGMT, the literature suggests that promoter hypermethylation is a marker of improved prognosis among glioblastomas that are treated with temozolomide (TMZ) (20). The MGMT promoter was unmethylated across our ATRTs; as such, and in contrast to the glioblastoma literature, our data would suggest that potential TMZ‐based chemotherapeutic regimens in ATRT might be of limited efficacy. In support of this assertion, Hashizume et al found that ATRT xenografts that highly expressed MGMT (ie, unmethylated promoters) displayed resistance to TMZ treatment in an orthotopic tumor transplantation model (17).

Most patients with ATRT appear to benefit from a more intensive treatment regimen than medulloblastoma. A variety of such treatments have been proposed for ATRT and have included radiation, myeloablative chemotherapy with stem cell rescue, intrathecal chemotherapy and even rhabdomyosarcoma‐like chemotherapy 11, 14, 34, 38, 43. However, these intensive and potentially life‐sparing treatment options come at the cost of harmful long‐term sequelae (eg, cognitive, motor and visual impairment, epilepsy, etc.) (30). As such, it would be ideal to direct the most intensive regimens toward those ATRT patients with the poorest prognoses. Although preliminary, CLDN6 holds promise as a useful prognostic and predictive tool in the treatment armamentarium of ATRT. Besides CLDN6, other potential ATRT risk stratifiers have been recently uncovered. Based upon gene expression microarrays performed on 18 cases, multiple BMP genes (including BMP4) were upregulated in those ATRTs demonstrating short survival (5).

In summary, we have demonstrated the somewhat common occurrence of ATRTs that display a paucity of “teratoid” and “rhabdoid” features. INI‐1 IHC retrospectively identified 12 cases of putative ATRT (10 previously unrecognized), and subsequent IHC and genetic data supported this recategorization. Unexpectedly, INI‐1 IHC highlighted a subset of long‐term survivors (n = 3) that were united by CLDN6 immunonegativity, raising the potential prognostic utility of such testing. Although our epigenetic data did not reveal a common profile among the long‐term survivors, it did suggest that TMZ‐based chemotherapeutic regimens might be less efficacious in ATRT than in glioblastoma.

Supporting information

Table S1. Methylation Specific PCR (MSP) primers.

Supporting info item