As an initial step, a comparative supervised analysis of gene-expression profiles generated for a large cohort of various pediatric and adult brain tumors was performed. When comparing the 13 ETMR samples with all other brain tumors, we identified a specific gene signature that could readily distinguish ETMR samples from all other brain tumors. Among the 50 most differentially expressed genes, 18 (LIN28A
, and HES5
) were highly upregulated in all ETMR samples. However, only one gene, LIN28A
, was highly upregulated in ETMR but not in any other brain tumor (Fig. ). Nine ETMR samples showed particularly high LIN28A expression (>1,000). For all other brain tumors, only a subset of AT/RTs and a few other tumors showed low LIN28A
mRNA levels, but otherwise, it was not expressed at all in other pediatric and adult brain tumors. Normal CNS and non-CNS tissues also did not express LIN28A
. These data strongly suggest that LIN28A
expression could be exploited as a diagnostic biomarker to specifically detect ETMR. Interestingly, Picard et al. [20
] recently analysed gene-expression profiles of 51 CNS PNETs and identified 3 molecular subgroups, one of which was distinguished by high expression of LIN28A
, frequent amplification of 19q13.42, young age, and dismal prognosis. Our data now suggest that all tumors in this subgroup fit well to the diagnosis of ETMR.
Fig. 1 LIN28A expression is highly upregulated only in ETMR cases when compared to other pediatric and adult brain tumors (n = 1404) and normal CNS and non-CNS tissues (n = 353). The numbers in parentheses indicate the number (more ...) LIN28A
and its homolog LIN28B
encode proteins that bind small RNA and function as negative regulators of the let
family of miRNAs, which may act as tumor suppressor miRNAs [24
]. LIN28A is a conserved cytoplasmic protein, but may be imported to the nucleus where it regulates the translation and stability of mRNA. In addition, LIN28A has been implicated in stem cell pluripotency and metabolism, is expressed widely in early embryogenesis, and is downregulated upon differentiation [7
]. Some recent studies suggest that LIN28A
function as oncogenes promoting tumor transformation and progression, with high expression associated with unfavorable clinical course in cancers of the ovary, colon, esophagus, and sympathetic nervous system (neuroblastoma) [7
]. A potential mechanism for the oncogenic potency of LIN28A and LIN28B might be mediated through let
repression and consecutive upregulation or stabilization of let
targets, such as MYC(N)
]. Of note, we indeed observed strikingly tight associations between expression levels of LIN28A
and those of MYCN
, and HMGA2
in the 13 ETMRs (Fig. ). Moreover, recent data of Molenaar et al. [15
] showed that Lin28B
overexpression in the mouse sympathetic adrenergic lineage induced development of neuroblastomas marked by low let
miRNA and high MYCN expression. Interestingly, LIN28B
is also highly expressed in all 13 ETMRs, but its diagnostic utility is limited, given that it is also found in other brain tumors, including medulloblastoma.
Expression of LIN28A strongly and significantly correlates with expression of HMGA2 (a), MYCN (b), CDK6 (c), and NRAS in ETMR (d)
To test whether LIN28A expression could be used as a sensitive and specific diagnostic marker for ETMR, we performed IHC screening of a large cohort of 816 malignant pediatric brain tumors (Table ). Strong and diffuse LIN28A cytoplasmic immunostaining was found in all 37 histologically classic ETMR (Fig. ). Thirty-six of these tumors (97 %) also harbored high-level amplification of the 19q13.42 locus, which was not present in any other tumors analyzed (Table ). LIN28A positivity was found to be more prominent and intense in multilayered rosettes and poorly differentiated small cell tumor areas, whereas only single collections of positive cells were observed within neuropil-like tumor parts. Analysis of nine ETMR samples obtained from corresponding tumor recurrences showed that the intensity of LIN28A immunoexpression and the number of stained cells were significantly higher in recurrent lesions in comparison to their primaries (Fig. ). In addition, we had the opportunity to compare the results of LIN28A mRNA and protein expression for six ETMR samples (4 tumors with expression gene levels >1,000, and 2 samples with lower LIN28A expression). All these tumors showed clear LIN28A positivity. Nevertheless, the two ETMR samples with lower LIN28A mRNA expression levels contained extensive areas of LIN28A immunonegative neuropil, whereas the four tumors with high LIN28A mRNA expression were composed predominantly of LIN28A immunopositive, histologically primitive small cell areas.
Fig. 3 LIN28A immunohistochemistry in pediatric malignant CNS tumors. Microscopic appearance of ETMR composed of clusters of multilayered rosettes embedded in abundant neuropil (initially diagnosed as ETANTR) (a). Intense LIN28A expression in poorly differentiated (more ...)
Fig. 4 LIN28A expression patterns in primary and recurrent ETMR. Primary ETMR with single multilayered rosettes and abundant neuropil (a). LIN28A expression is limited with rosettes and small cell collections (b). The ETMR relapse from the same patient is composed (more ...)
Among the other tumor entities, focal cytoplasmic immunoexpression of LIN28A was only found in single cell collections of six AT/RT cases (12 %). Each of these six tumors had obvious losses of nuclear INI1 protein expression, in line with their initial histopathological diagnosis (Fig. ). In contrast, all other pediatric CNS malignant tumors, including 44 additional AT/RT samples, 41 CNS PNETs, 334 medulloblastomas, 223 anaplastic ependymomas, and 131 pediatric glioblastomas were completely negative for LIN28A immunostaining (Fig. ). Finally, we performed survival analysis for all these patients based on LIN28A reactivity (Fig. ). Tumors with diffuse and wide-spread immunoexpression of LIN28A (considered as “positive” samples and representing all ETMR cases) showed an extremely poor overall survival (5-year overall survival estimate of 0 %) in comparison to other malignant pediatric brain tumors with absence or rare focal positivity of LIN28A (5-year survival 68 %, log-rank p < 0.001).
Fig. 5 Kaplan-Meier overall survival analysis for LIN28A positive malignant brain tumors versus LIN28A negative malignant brain tumors in children. Overall 5-year survival was 0 % for LIN28A positive versus 68 % for LIN28A negative tumors (log-rank, (more ...)
Therefore, this study represents a report of LIN28A as a highly specific and sensitive diagnostic marker for the distinct pathological verification of ETMR and, correspondingly, as a new important diagnostic tool to be routinely applied in all malignant childhood brain tumors where ETMR may be considered a possible diagnosis. Variability of LIN28A
gene expression in ETMR samples correlated closely with the intratumoral heterogeneity observed immunohistochemically. Focal positivity of LIN28A in a small subset of AT/RTs is not considered problematic when taking the simultaneous loss of INI1 expression into account, with the latter representing a fairly specific molecular marker for AT/RT. On the other hand, LIN28A immunoexpression has also been detected recently in CNS germ-cell tumors with diffuse staining patterns found in germinomas, embryonal carcinomas and yolk sac tumors but being only focal in immature teratomas [2
]. Nevertheless, these tumors usually affect older children (apart from the some immature teratomas of infancy) than those at risk for ETMRs, are typically not situated in those locations commonly involved by ETMRs, and do not enter the differential diagnosis on histopathologic and immunohistochemical grounds.
Our current data also confirmed that ETMR is a unique clinico-pathologic entity from a molecular perspective, exhibiting 19q13.42 amplification and high levels of LIN28A expression accompanied by particularly aggressive clinical behavior. It will now be important to understand the biological consequence of overexpression of such universal and direct miRNA translation regulators as LIN28A and LIN28B in ETMR, taking into account the prototypic amplification of the oncogenic miRNA cluster at 19q13.42 in these tumors.