Local recurrence continues to limit survival in medulloblastoma patients, largely related to the persistence of invasive cells at the site of tumour resection and leptomeningeal dissemination. Given the relative dearth of understanding of causative mechanisms behind the invasiveness of medulloblastomas, and a general lack of validated in vitro models with which to study them, our objectives were (1) to obtain quantitative data on the invasiveness of five distinct medulloblastoma cell lines within a 3-dimensional in vitro collagen-based model; and (2) to characterize some of the mechanisms behind invasion, specifically striving to identify proteolytic processes that occur as medulloblastoma cells disrupt and thereby invade the normal tissue surrounding them, and specific inhibitors of these proteolytic enzymes. Five different medulloblastoma cell lines (UW228-1, 2 and 3; Daoy, and Madsen) were implanted onto a 3-dimensional, type I collagen gel assay to assess tumour invasion distance and mean doubling time over 5 days. Proteolytic activity was assessed against collagen types I and IV by measuring the degradation of 3H-collagen I and IV to products soluble in 100% w/v trichloroacetic acid; and general (neutral) proteolytic activity evaluated by measuring the degradation of 3H-albumin. In other experiments, cells were pre-exposed to a variety of protease inhibitors, including inhibitors of metalloproteinases and cysteine, serine and aspartic proteases, and then plated to identify any inhibition of invasion. Inter-group differences in mean invasion distance were assessed by means of Student's t-tests for non-paired subjects, with P < 0.05 set as the threshold for statistical significance. For the inhibitor studies, an inhibition index, called the inhibitory concentration 50, IC-50, was calculated by performing a regression analysis for each inhibitor tested over a range of concentrations, for each cell line. Within hours of implantation, individual cells readily detached from the surface of the cell aggregates and invaded the collagen matrix, to distances of up to 1,200 mum and at rates of up to 300-mum per day; the UW228-1 cell line clearly was less invasive than the other four cell lines. Proteolytic activity was identified against collagen type I, but not against collagen type IV or albumin; but there was no apparent correlation between invasion distance and either cell doubling time or the amount of collagen type I proteolytic activity. Both metalloproteinase inhibitors suppressed tumour invasion, as did one of two cysteine protease inhibitors; but there was no tumour suppression with either serine or aspartic protease inhibition. MMP-1 and 2, and TIMP-1 and 2 all were detectable by Western blot analysis. Medulloblastoma cell invasiveness within the 3-dimensional model used here appears to depend upon a combination of metalloproteinase and cysteine protease activity, a finding that may suggest areas for potential future clinical investigation and therapy.