BackgroundTherapeutic targeting of tumour necrosis factor (TNF)-α is highly effective in ankylosing spondylitis (AS) patients. However, since one-third of anti-TNF-treated AS patients do not show an adequate clinical response there is an urgent need for new biomarkers that would aid clinicians in their decision-making to select appropriate therapeutic options. Thus, the aim of this explorative study was to identify cell-based biomarkers in peripheral blood that could be used for a pre-treatment stratification of AS patients.MethodsA high-dimensional, multi-parametric flow cytometric approach was applied to identify baseline predictors in 31 AS patients before treatment with the TNF blockers adalimumab (TNF-neutralisation) and etanercept (soluble TNF receptor).ResultsAs the major result, the frequencies of natural killer (NK) cells, and in particular CD8-positive (CD8+) NK cell subsets, were most predictive for therapeutic outcome in AS patients. While an inverse correlation between classical CD56+/CD16+ NK cells and reduction of disease activity was observed, the CD8+ NK cell subset behaved in the opposite direction. At baseline, responders showed significantly increased frequencies of CD8+ NK cells compared with non-responders.ConclusionsThis is the first study demonstrating that the composition of the NK cell compartment has predictive power for prediction of therapeutic outcome for anti-TNF-α blockers, and we identified CD8+ NK cells as a potential new player in the TNF-α-driven chronic inflammatory immune response of AS.Electronic supplementary materialThe online version of this article (10.1186/s13075-018-1692-y) contains supplementary material, which is available to authorized users.