Noninvasive CNS lymphoma risk stratification and brain cancer classification by circulating tumor DNA profiling

Abstract: Lymphomas of the central nervous system are highly heterogeneous and pose several challenges for clinicians and patients. The initial diagnosis of CNSL is significantly hampered by the sensitivity and specificity of imaging-based methods and requires in most cases an invasive brain biopsy for histopathological assessment, which introduces neurological morbidity and can be inconclusive in certain cases. Moreover, existing methods fail to identify patients with high risk or low risk for treatment failure and are unable to predict the clinical outcome accurately. Therefore, there is an unmet clinical need for more reliable and less invasive biomarkers in patients with CNSL. In the current study, we addressed these fundamental limitations and examined the potential of liquid biopsy from blood and CSF in patients with CNSL. We applied CAPPSeq and PhasED-seq – two targeted NGS-based methods for ultrasensitive genotyping and monitoring – to 67 tumor, 115 plasma, and 46 CSF samples from patients with either PCNSL or iSCNSL. At diagnosis, ctDNA is readily detectable in plasma and CSF samples, and ctDNA levels are significantly different in plasma and CSF. Furthermore, ctDNA concentrations in plasma and CSF were associated with certain radiographic parameters: We showed that ctDNA levels correlate with tumor volume in plasma, whereas CSF ctDNA concentrations are associated with tumor localization near the ventricles. When compared to DLBCL, we showed that plasma ctDNA levels of CNSL patients are substantially lower than in plasma of DLBCL patients. However, ctDNA concentrations in CSF were not affected, suggesting that factors such as the BBB might limit the amount of ctDNA in the blood. In order to stratify patients before and during treatment, we correlated plasma ctDNA with the survival of patients and revealed that assessment of ctDNA before therapy was a strong predictive biomarker for the survival of patients. Moreover, when combined with the tumor volume, we could define three distinct groups with a low, intermediate, and high risk of lymphoma progression or death. This outcome prediction was supplemented and improved by sampling blood from patients during curative-intent treatment. We found that ctDNA assessment during treatment was also predictive for the survival of patients and that this could help obtain real-time information for individualized outcome prediction and potential treatment adjustments. Further, we developed a machine learning classification algorithm for distinguishing CNSL from other brain cancers or metastases based on the ctDNA mutational profile. Independent validation showed correct classification in a large portion of patients from CSF, indicating that noninvasive diagnosis from CSF might be a viable alternative to routinely performed invasive surgical procedures.

Overall, this study highlights the potential of non-invasive ctDNA assessment for early outcome prediction, MRD monitoring, and disease classification in CNSL patients. Moreover, these results emphasize the possibility of biopsy-free diagnosis and the ability to guide treatment decisions and optimizations in the future

Angaben zum Objekt

• Standort: Deutsche Nationalbibliothek Frankfurt am Main
• Umfang: Online-Ressource
• Sprache: Englisch
• Anmerkungen: Universität Freiburg, Dissertation, 2022

Beteiligte, Orts- und Zeitangaben

• Ereignis: Veröffentlichung
• (wo): Freiburg
• (wer): Universität
• (wann): 2022
• Urheber: Mutter, Jurik
• Beteiligte Personen und Organisationen: Pahl, Heike L. ; Frew, Ian ; Börries, Melanie ; Universitätsklinikum Freiburg. Klinik für Innere Medizin I ; Albert-Ludwigs-Universität Freiburg. Fakultät für Biologie

Weitere Informationen

• DOI: 10.6094/UNIFR/226720
• URN: urn:nbn:de:bsz:25-freidok-2267204
• Rechteinformation: Open Access; Der Zugriff auf das Objekt ist unbeschränkt möglich.
• Letzte Aktualisierung: 25.03.2025, 13:57 MEZ