Background: Squamous cell carcinoma of the tonsil is one of the most frequent cancers of the oropharynx. The escalating rate of tonsil cancer during the last decades is associated with the increase of high risk-human papilloma virus (HR-HPV) infections. While the microbiome in oropharyngeal malignant diseases has been characterized to some extent, the microbial colonization of HR-HPV-associated tonsil cancer remains largely unknown.Results: Using 16S rRNA gene amplicon amplicon sequencing, we have characterized the microbiome of human palatine tonsil crypts in patients suffering from HR-HPV-associated tonsil cancer in comparison to an age-matched control cohort of sleep apnea patients. We found an increased abundance of the phyla Firmicutes and Actinobacteria in tumor patients, whereas the abundance of Spirochaetes and Synergistetes was significantly higher in the control cohort. Furthermore, the accumulation of several genera such as Veillonella, Streptococcus and Prevotella_7 in tonsillar crypts was associated with tonsil cancer. In contrast, Fusobacterium, Prevotella and Treponema_2 were enriched in sleep apnea patients. Machine learning-based bacterial species analysis indicated that a particular bacterial composition in tonsillar crypts is tumor-predictive. Species-specific PCR-based validation in extended patient cohorts confirmed that differential abundance of Filifactor alocis and Prevotella melaninogenica is a distinct trait of tonsil cancer.Conclusion: This study shows that tonsil cancer patients harbor a characteristic microbiome in the crypt environment that differs from the microbiome of sleep apnea patients on all phylogenetic levels. Moreover, our analysis indicates that profiling of microbial communities in distinct tonsillar niches provides microbiome-based avenues for the diagnosis of tonsil cancer.