Finding the Perfect Partner for Medulloblastoma Prognostication

Abstract

The field of cancer prognostication is littered with broken relationships. New predictors that begin with such promise, in a blush of honeymoon excitement, frequently fade and fail the test of time. Medulloblastoma—the most common malignant pediatric brain tumor—is no exception. The relationship between survival and dozens of molecular markers for medulloblastoma has been reported by more than 200 articles in the last 30 years, yet not one of those markers is used in the clinic today. A more sophisticated approach to selecting the perfect partner for robust prognostication in medulloblastoma is well overdue. In the article that accompanies this editorial, Remke et al report the potential prognostic significance of a poorly understood gene, follistatin-like 5 (FSTL5), among patients with medulloblastoma. The authors used a novel platform, the 4 44K Agilent Whole Human Genome Oligo Microarray (Agilent Technologies, Bumoblingen, Germany), to segregate the transcriptomes of 64 medulloblastomas, including 24 adult cases, into distinct subgroups. (This article is the first, to my knowledge, to report an analysis of medulloblastoma using this platform.) This process identified the well-established molecular subgroups of medulloblastoma that have emerged from Affymetrixbased (Affymetrix, Santa Clara, CA) studies: the wingless (WNT) subtype that confers an excellent outcome; the sonic hedgehog and D subtypes that are associated with intermediate survival; and the highly aggressive C or MYC-subtype, so called because of its high level of expression of MYC. It is important to note that although the origins and underlying biology of the WNT and sonic hedgehog subtypes are well understood, little is known about Cand D-subtype tumors, which include the majority—and most aggressive forms—of medulloblastoma. Therefore, the authors looked for genes that might drive these less understood tumors. FSTL5 caught their attention. Because this gene was expressed most highly in C-subtype medulloblastoma, it was significantly associated with a poor clinical outcome across the entire cohort of 64 tumors. Additionally, through published gene expression profiles and analysis of their own data, the authors noted that FSTL5 expression among D-subtype medulloblastomas—a clinically quite heterogeneous subgroup—ranged from low to the highest observed in C-subtype tumors. This led the authors to propose that high FSTL5 expression might identify the D-subtype tumors with the worst prognosis. To test this, the authors turned to an immunohistochemical analysis of two tissue microarrays that contained 235 medulloblastomas. FSTL5 protein expression was assessed using the B01P antibody (Abnova, Taipei City, Taiwan). Tumors were scored in a semiquantitative manner (positive v negative). Features associated with high-risk disease (for example, a higher incidence of large-cell/anaplastic histology) and metastatic disease were observed among 84 tumors (36%) that were judged to be FSTL5 immunopositive (described by the authors as displaying “clear evidence of cytoplasmic and/or nuclear staining”). FSTL5 immunopositivity also seemed to occur more frequently among tumors that displayed a C-subtype rather than a WNT-subtype immunoprofile, and, in keeping with the authors’ transcriptomic studies, was significantly associated with a diminished progression-free and overall survival. This survival significance held true independently of other disease features such as tumor subtype (C and D) and patient age (pediatric and adult), and, in a Brier score analysis, reduced the error in predicting treatment failure among patients with Cand D-subtype medulloblastoma. In light of their data, the authors propose FSTL5 as a promising biomarker in medulloblastoma, with particular value for identifying “high-risk group D patients, as well as low-risk group C patients so they can be properly stratified in the setting of a clinical trial.” A robust biomarker that could better guide the treatment of poorly understood Cand D-subtype medulloblastoma would truly be an advance for the field. But why should FSTL5 expression be any more successful than the numerous other potential markers of medulloblastoma prognosis that have gone by the wayside? One reason is the rigor with which Remke et al have performed their analysis. As a leading group in the study of medulloblastoma, they know the disease well. Previous potential prognostic markers of medulloblastoma, such as ERBB2 and TRKC, were first described in the pregenome era when medulloblastoma was regarded as a single disease. Transcriptome profiling and mouse modeling have since shown that medulloblastoma likely comprises a series of different diseases, each with discrete cells of origin and driving mutations. By setting the prognostic significance of FSTL5 in the context of these disease subtypes, the authors provide a much more rigorous and comprehensive assessment of the potential prognostic value of this variable. An additional notable finding of the study is that FSTL5 protein expression can be detected in formalin-fixed, paraffin-embedded material using immunohistochemistry. This technique is standard within diagnostic pathology laboratories, additionally increasing the probability that this marker may ultimately prove useful for guiding routine clinical decisions. JOURNAL OF CLINICAL ONCOLOGY E D I T O R I A L S VOLUME 29 NUMBER 29 OCTOBER 1