Background: Glioblastoma multiforme (GBM) is the most common primary brain cancer, with a dire 5-year survival rate of less than 10%. Despite comprehensive initial treatment, recurrences occur in nine out of ten patients. Key challenges in GBM therapy include the lack of targeted treatment options and diagnostic markers for recurrent tumors.

Methods: This study includes a cohort of 55 GBM patients, examining tissue samples from primary tumors and five matching recurrences. Together with each tissue specimen, matching serum samples were collected before and after surgery. All samples were measured using mass-spectrometry-based proteomics in data-independent acquisition mode.

Results: In tissue, we identified 6300 proteins on average per sample. Hierarchical clustering of primary tumors revealed four distinct subgroups, independent of previous pathological annotations. The subgroups included:

A neuronal cluster marked by elevated mature neuron markers

An innate immunity cluster with increased protease expression

A mixed cluster with high translational activity

A stem-cell cluster with increased stem-cell marker expression

The primary clustering drivers were neurodevelopmental and inflammatory processes, with semi-specific peptide levels indicating proteolytic activity intensifying alongside inflammation levels. This pattern was expanded and confirmed through an extra analysis including proteins of lower coverage. Patients in the neuronal cluster showed significantly longer survival compared to those in the stem-cell cluster. In patient-matched differential expression analysis, recurrent tumors exhibited significant changes in protein expression highlighting the proteomic plasticity of recurrent GBM. Serum proteome analysis, utilizing a depletion-based protocol, yielded 730 protein identifications. It revealed stable, patient-specific proteome compositions despite increased inflammation markers post-surgery. This finding was consistent with an independent control cohort of serum from patients undergoing surgery for meningioma (MNG). Several proteins decreased in GBM serum after tumor removal, with some remaining unchanged in MNG serum. Additionally, circulating proteolytic products matched to tissue proteolytic activity, and one fragment of proteolysis-activated receptor 2 (PAR2) consistently decreased after the removal of tumors assigned to the innate immunity cluster.

Conclusion: We identified distinct tumor subgroups in a large GBM cohort together with circulating marker candidates. This may enable recurrence risk prediction and personalized therapies in the future.