Background

Malignant rhabdoid tumors (MRT) are highly aggressive embryonal tumors. They can arise at any anatomical location and their cell of origin is unknown. The most common site of origin is the central nervous system (CNS) where they are known as atypical teratoid/rhabdoid tumors (AT/RT) [1]. According to recent data from the German Childhood Cancer Registry, it is estimated that AT/RTs make up to 65% of all rhabdoid tumors, while 5–10% are originated in the kidneys and are called rhabdoid tumors of the kidney or RTK. The remaining 25% originate in extracranial/extrarenal locations. AT/RT is the most common malignant CNS tumor in patients under 6 months of age [2]. The current treatment protocol involves maximal surgical resection of the primary tumor, chemotherapy and radiation therapy depending on the age of the patient and location of the tumors. Maximal tumor resection is associated with increased overall survival rates [3]. Recent reports showed that high-dose chemotherapy including intrathecal chemotherapy has improved patients’ outcome to 70% survival in 2 years [4] and 45% survival in 6 years [5]. Other studies showed that radiation therapy in high-doses early in the course of the disease can also increase survival rates of children diagnosed with AT/RT [6, 7]. However these intensive adjuvant therapies have significant side effects including high risk of leukoencephalopathy or radiation necrosis [8]. Thus, there is an immediate need for a better understanding of the biological basis of AT/RT in order to develop novel and more effective targeted therapeutic approaches.

Genetically, the majority of AT/RT is characterized by inactivating mutations of the SMARCB1 gene, also known as hSNF5, BAF47 and INI1, while less than 5% exhibit mutations in the SMARCA4 gene, also known as BRG1 gene [9]. Both genes are part of the SWI/SNF chromatin remodeling complex. In fact, it has been demonstrated that more than 20% of human malignancies have a mutation in at least one subunit of this complex [10]. AT/RT most frequently arises sporadically, but there are also familial cases in the so called rhabdoid tumor predisposition syndrome (RTPS) [11].

Histologically, AT/RT may display a diverse combination of cellular types consisting of undifferentiated “small round blue cells”, mesenchymal and epithelial components. The classic rhabdoid phenotype, characterized by large cells with eccentrically placed nuclei, a prominent nucleolus, abundant eosinophilic cytoplasm and occasional pale cytoplasmic inclusion bodies may or may not be present [12]. Depending on the area of the tumor submitted to histologic examination it may be challenging to morphologically differentiate AT/RT from medulloblastoma (MB) and other embryonal brain tumors [13,14,15]. Therefore, establishing the diagnosis based solely on histopathological observations can be challenging. The immunohistochemical profiling of the tumor is also necessary for an accurate diagnosis. AT/RT frequently co-expresses mesenchymal and epithelial markers such as vimentin and epithelial membrane antigen (EMA) and does not express markers of skeletal muscle differentiation. Remarkably, about 90% of AT/RT demonstrate loss of expression of SAMRCB1 [16] by immunohistochemistry. This loss of SMARCB1 expression is currently the most reliable marker for AT/RT diagnosis. However, about 10% of the cases, which do not lose SMARCB1 expression, remain undiagnosed.

The aim of this study was to investigate new diagnostic markers and/or new genetic factors contributing to the pathophysiology of AT/RT. To achieve this goal, ee analyzed the gene expression profiles of frozen AT/RT samples. Because the AT/RT’s cell of origin is unknown, we chose to compare AT/RT to MB, which is the most common CNS pediatric embryonal tumor, but has a better outcome than AT/RT, with over 90% of cure rates for WNT group and 40–60% for group 3 [17, 18]. The result of this analysis revealed the neural cell adhesion molecule 1 (NCAM1) among the most significantly differentially expressed genes.

Methods

Tumor samples

Fresh frozen tumor tissues were provided by the Falk Brain tumor bank (Chicago, IL, USA) and by the Center for Childhood Cancer’s Biopathology Center (Columbus, OH, USA), which is a section of the Cooperative Human Tissue Network of The National Cancer Institute (Bethesda, MD, USA).

Written informed parental consents were obtained prior to sample collection. The study was approved by the Institutional review board of Ann and Robert H. Lurie Children’s Hospital of Chicago (IRB#2009-13778).

Gene expression (GE) profile

Total RNA was isolated using Trizol Reagent (Invitrogen, Carlsbad, CA, USA) according to the manufacturer’s instructions. The RNA concentration and quality were assessed by NanoDrop (ThermoFisher Scientific, Waltham, MA, USA) and Bioanalzyer 2100 (Agilent technologies Inc., Santa Clara, CA, USA) respectively. GE profiling was performed using Illumina HT-12 BeadChip whole-genome expression arrays (Illumina, San Diego, CA, USA). Each RNA sample was amplified using the Ambion Illumina RNA amplification kit with biotin UTP labeling (Enzo Biochem, New York, NY, USA). In vitro transcription was completed in order to synthesize biotin-labeled cDNA using T7 RNA polymerase. The cDNA was then column-purified and checked for size and yield using the Bio-Rad Experion system (Bio-Rad Laboratories, Hercules, CA, USA). A total of 1.5 μg of cDNA was hybridized to each array using standard Illumina protocols. Slides were scanned on an Illumina Beadstation and analyzed using BeadStudio (Illumina, San Diego, CA, USA). Data were normalized using the quantile normalization procedure from the bioconductor package, affy (www.bioconductor.org). Cluster and TreeView (www.eisenlab.org) were used for data clustering and visualization.

Quantitative real-time PCR (qRT-PCR)

A total of 1 μg of RNA was used to make cDNA using the high capacity RNA-to-cDNA Kit (ThermoFisher Scientific, Carlsbad, CA, USA), and the expression of the selected gene was validated by real-time PCR using the TaqMan gene expression assay for NCAM1 (Hs00941830_m1) (ThermoFisher Scientific, Carlsbad, CA, USA). Reactions were performed in triplicates with adequate positive and negative controls. The normalized GE levels were calculated by the ΔΔCt method using the housekeeping gene GAPDH (Hs02758991_g1) as reference [19], and a pool of all samples as calibrator.

Immunohistochemistry (IHC) on Tissue microarray (TMA)

To assess the expression of genes of interest by immunohistochemistry, two TMAs were analyzed. The TMAs were designed as follows: (1) a rhabdoid tumor array with 35 unique cases in duplicates with control cores consisting of cerebellum (three cases), kidney (three cases), and tonsil (three cases) and a (2) MB array with 40 unique cases in duplicates and cores of cerebellum as control. Cores were placed randomly across the blocks.

TMA sections (5 μm thick) were stained using standard immunohistochemical procedures with the following antibodies; polyclonal NCAM1 antibody (1:100, Millipore, USA) and polyclonal hSNF5 antibody (1:200 Novus Biologicals, USA) Interpretation of the TMA slides was performed in a blinded fashion.

Results

GE profile showed significant downregulation of NCAM1 in AT/RT

GE profiling was performed in 14 AT/RTs, 6 MBs and four normal brain tissues including three fetal and one adult. The analysis revealed 1,002 genes significantly differentially expressed with a fold change higher than 2 or lower than −2 and a p-value ≤−0.05 when comparing AT/RT and MB (Fig. 1a). Among the most differentially expressed genes, NCAM1 was significantly downregulated in AT/RT when compared to MB and normal brain tissue (P = 0.0046 and 0.0001 respectively, one-way ANOVA) (Fig. 1b). Data was validated by qRT-PCR in 5 AT/RTs, 5 MBs and three normal brain tissues. The results confirmed the GE findings that expression of NCAM1 in AT/RT is significantly lower than in MB and normal brain tissue (P = 0.0321 and 0.0047 respectively, one-way ANOVA) (Fig. 1c).

Fig. 1
figure 1

GE of AT/RT, MB and normal brain tissue. a Hierarchical clustering of GE profiling showed 1,002 genes significantly differentially expressed with ≥2 or ≤−2 fold-change, and p-value <0.05. b Analysis of GE data showed that NCAM1 was significantly downregulated in AT/RT when compared to MB and normal brain tissues (**P < 0.01, ****P < 0.0001, one-way ANOVA). c Validation of GE data by qRT-PCR. NCAM1 expression level was significantly lower in AT/RT when compared to MB and normal brain tissues (*P < 0.05, **P < 0.01, one-way ANOVA)

Full size image

IHC showed loss of NCAM1 expression in AT/RT

Hematoxylin and eosin staining (H&E) of TMA revealed AT/RT with various amounts of rhabdoid cells within undifferentiated small round blue cells. MB cores showed typical small round blue cell tumors composed of sheets of undifferentiated cells with minimal cytoplasm, hyperchromatic and anaplastic nuclei. High mitotic activity was observed in both AT/RT and MB cores. IHC for SMARCB1 showed loss of expression in tumor cells of all AT/RT samples with retention of expression in endothelial cells, fibroblasts and inflammatory cells within the tumors. NCAM1 expression was observed in cell membranes and extracellular space of MB and controls but not in AT/RT (Fig. 2a). It was observed that 67.3 and 10.5% of the samples were immunohistochemically negative for NCAM1 in the AT/RT and the MB set of tumors, respectively (P < 0.01, Fisher exact test) (Fig. 2b).

Fig. 2
figure 2

Histopathological and immunohistochemical comparison between AT/RT and MB in an independent set of tumors using TMAs. a H&E staining highlights the morphological characteristics of AT/RT and MB. While MB is predominantly represented by undifferentiated small blue round tumor cells, AT/RT displayed of rhabdoid phenotype (inserts: 40X). IHC for SMARCB1 showed loss of expression in AT/RT while expression is retained in MB. IHC for NCAM1 showed loss of expression in AT/RT with high expression levels at the cell membrane and extracellular spaces in MB. b Left: representative image of a TMA stained with NCAM1 antibody. Middle: the positive/negative ratio for NCAM1 in AT/RT samples. Right: the positive/negative ratio for NCAM1 in MB samples

Full size image

Discussion

Since the cell of origin of AT/RT is unknown [20], we chose to compare the GE profile of AT/RT with MB as both entities are malignant, embryonal brain tumors that arise in young children and display overlapping histological features. It is important to note however, that MB commonly responds to current available modalities of therapy while AT/RT does not. This fact is reflected in 5-year overall survival rates of 73.0% for MB and 2-year overall survival of 35.5% for AT/RT [21].

Here we report loss of NCAM1 expression in AT/RT. While we understand that the small number of samples included in this study represents a limitation on data interpretation, the downregulation of NCAM1 detected by microarray GE profiling, was verified by qRT-PCR and validated at the protein level, by IHC in an independent set of tumor samples. This validation in multiple levels strengthens the significance of our data. Furthermore, our group previously reported NCAM1 to be sharply downregulated in rhabdoid tumors from the kidney (RTK) [22], what also reassures the significance of our results.

NCAM1, also known as CD56 is part of the immunoglobulin superfamily and is expressed on the surface of neural cells [23] and certain cells of the immune system [24]. NCAM1 is classically known as a cell surface molecule and cell adhesion in neuronal cells. However it has many other known functions including exon guidance and repair, migration, apoptosis and cell proliferation [25, 26]. Its multiple functions depend mainly on alternative splicing, glycosylation/polysialylation status and pattern of expression at different developmental stages. Alternative splicing generate multiple isoforms of NCAM1 which play major roles in prognosis of malignancies [26]. NCAM1 glycosylation and polysialylation have also been associated with cancer [27]. Although the literature in the field is still controversial there is a tendency to correlate low NCAM1 expression to a poorer prognosis. As an example, NCAM1 immnostaining was found to be inversely correlated with the histological grade in gliomas [28]; its levels of expression were reported to be markedly reduced by the malignant transformation in thyroid follicular carcinoma and papillary carcinoma [29], and with worst prognosis in colon cancer [30].

Conclusion

Our data show that loss of NCAM1 expression might function as a potential new diagnostic marker for AT/RT, when considered in association with SMARCB1 (INI1) or SMARCA4 (BRG1) inactivation. Further studies are needed to explore the functional significance of NCAM1 loss of expression in the biology of AT/RT.