Funded Grants

Use of cerebrospinal fluid genetics for the diagnosis, subgrouping, and minimal residual disease monitoring in medulloblastoma (update March 2012)

Medulloblastoma is the most common malignant brain tumor in children, and despite recent improvements in treatment strategies, still approximately one third of children die from their disease. At least another third of patients can only be cured at high cost, namely severe neurological, mental, psychosocial, endocrine, cardiac, and renal deficits. In the past, diagnosis was solely derived from histopathological assessment of the tumor tissue. Recent evidence from genome-wide screening projects (including DNA copy-number analysis, gene expression profiling, and DNA-methylation analyses) clearly suggests that medulloblastoma is in fact a group of molecularly, epidemiologically, and prognostically distinct diseases comprising at least four major variants. Across all of these molecular variants, patients with microscopic or macroscopic dissemination of their tumor (throughout the CNS, occasionally even throughout the body) have a much worse probability of being cured.

One major reason why brain tumors as a group have replaced leukemias as the leading cause of cancer death in children, despite the latter being much more common, is that our knowledge of the genetic repertoire of leukemias is much more comprehensive. These scientific insights have been successfully exploited for molecular diagnostics, treatment monitoring, tailoring treatment intensity to disease risk, and molecular targeted therapies in selected cases for more than a decade. This body of knowledge is only now starting to grow for medulloblastoma, the deadliest brain tumor in children. A challenging decision on whether to regard patients as clinically high risk is the definition of “microscopic metastases”. To justify this diagnosis, “united cell structures” of malignant cells have to be present in the cerebrospinal fluid (CSF) of patients at the time of primary diagnosis, according to a widely used classification system proposed by Chang and colleagues in the late 1960s. It is anticipated, however, that almost all patients have some kind of CSF dissemination, and that medulloblastoma is in fact a disease that involves the entire CNS rather than only the cerebellum, where the primary tumor occurs. This may be illustrated by the fact that many patients who are considered non-metastatic at the time of diagnosis will subsequently develop a relapse at a distant site, e.g., the spine. Patients with recurrent disease typically have a fatal prognosis, and most will die within one year after tumor recurrence was diagnosed. As a result of this, a refined detection of CSF dissemination is urgently warranted. One very promising avenue to achieve this is by tracking and quantifying tumor-specific genetic aberrations in this fluid. Identifying tumor-specific mutations in DNA from the CSF could not only be used for the diagnosis of CSF dissemination, but also for molecular subgrouping, treatment monitoring (i.e., clearance of the CSF in the course of treatment), and surveillance of patients after completing treatment (i.e., detection of early microscopic relapse). The latter two aspects are again in analogy to a very successful concept that has been applied in leukemia patients for many years, namely “minimal residual disease monitoring”. In leukemia patients, this is achieved by detecting characteristic genetic aberrations of the leukemia cell clone in the bone marrow, down to a level of 1 tumor cell in 100,000 normal cells. This information is used all over the world to stratify patients into therapy groups with different treatment intensity, to monitor treatment response, and for early detection of leukemia recurrence. If it was possible to use CSF genetics for the same purpose in brain tumor patients, this would revolutionize current clinical practice.

Our proposed approach to test the feasibility of using CSF genetics in the diagnosis and treatment monitoring of medulloblastoma patients involves three important steps. In a first phase, we will investigate 30 medulloblastoma patients, from whom we have collected frozen tumor samples, CSF at primary diagnosis, and blood (germline control). By sequencing all coding exons in the tumor DNA using next-generation sequencing techniques and comparing these with the germline of the same patient, we will determine the repertoire of somatic (i.e., tumor specific) mutations in these 30 tumor samples. In a second step, we will quantify the abundance of tumor-specific mutations in the CSF of the respective patient down to a resolution of 1 copy of the mutant allele in 50 copies of the wildtype (germline) allele. Finally, in the third phase of the project we will be able to compare serial CSF samples from ten patients over the course of their disease to assess whether it is possible to quantify molecular CSF clearance after treatment.

In the first five months of funding for this project, we have made significant progress in optimizing DNA extraction, quantification and amplification procedures, and in demonstrating the feasibility of the proposed approach. Due to the decreasing costs of next-generation sequencing, we were in fact able to apply high-coverage next generation sequencing techniques to the CSF DNA in addition to the primary tumor and control. As such, the preliminary results were actually better than anticipated, and this has enabled us to significantly expand the scope of the proposal, by using whole-exome deep-sequencing of the CSF DNA, within the same budgetary requirements (as outlined in the accompanying documents).

On average, we were able to recover ~150ng of DNA from 1ml of CSF (range 80-950ng). To maximize the number of samples available, we used whole genome amplification (WGA) to boost the DNA quantity. This technique, however, is known to introduce a number of artifacts in the resulting DNA. To overcome this, we developed a novel multiplexing protocol to sequence independent amplifications from the same CSF DNA sample. We were thereby able to reduce the number of mutation calls from ~1000 per sample to only 10 high-confidence hits, exactly in line with the observed mutation rate when sequencing primary tumors. Some of these mutations were seen in a high fraction of the CSF DNA, despite being detected in only a minor fraction of the sequence data from the primary tumor. This may indicate that the cells being shed into the CSF derive from a distinct subclone of the main tumor bulk, in keeping with a recent report to which our group contributed. Further investigation of this phenomenon is now also a goal of the current proposal.

In summary, we strongly believe that genetic testing of the CSF of medulloblastoma patients has the potential to significantly influence clinical decision making in terms of diagnosis, assessment of therapy response, selection of therapeutic agents, and patient surveillance. Therefore, we think our proposal exactly serves the major goals of the JSM foundation call on brain cancer research, namely the generation of novel insights leading to increased cure rates and improved functional outcome for brain tumor patients in a truly translational sense.


  1. Jones DT*, Jäger N*, … , Eils R*, Pfister SM*, Lichter P*. Dissecting the genomic complexity underlying medulloblastoma. Nature. 2012 Aug 2;488(7409):100-5.
  2. Northcott PA, Shih DJ, … , Pfister SM, Marra MA*, Taylor MD*. Subgroup-specific structural variation across 1,000 medulloblastoma genomes. Nature. 2012 Aug 2;488(7409):49-56.
  3. Pugh TJ, … , Pfister SM, Roberts TM, Meyerson M, Pomeroy SL, Cho YJ. Medulloblastoma exome sequencing uncovers subtype-specific somatic mutations. Nature. 2012 Aug 2;488(7409):106-10.
  4. Schwartzentruber J*, Korshunov A*, Liu XY*, … , Majewski J*, Pfister SM*, Jabado N*. Driver mutations in histone H3.3 and chromatin remodelling genes in paediatric glioblastoma. Nature. 2012 Jan 29;482(7384):226-31.
  5. Rausch T*, Jones DT*, Zapatka M*, … , Lichter P*, Pfister SM*, Korbel JO*. Cell. 2012 Jan 20;148(1-2):59-71.
  6. Wu X, …, Taylor MD. Clonal selection drives genetic divergence of metastatic medulloblastoma. Nature. 482, 529-533 (2012).