A 2-year longitudinal PBMC gene expression profiling of MS patients receiving interferon-beta 1b predicts therapy outcome
Objective: A systematic time-dependent gene expression analysis from Interferon beta 1b (IFNbeta-1b) treated Multiple sclerosis (MS) patients delivers sets of genes that provide a rational basis for predicting the success in modulation of the disease course.
Background: Microarrays are facilitating medical research by documenting detailed responses of cells and tissues to both disease and the intended and unintended effects of drug treatments.
Method: Two independent cohorts of RRMS patients were classified by stringent clinical criteria as IFNbeta-1b responders or non-responders after 2 years follow-up. Using Affymetrix HGU133AB chips, thus interrogating approximately 45.000 human RNA transcripts, we performed a full genome study monitoring 25 treated MS patients at 5 points. Samples were taken before first treatment (t0), after 48h (t1), 4 weeks (t2), 1 year (t3) and 2 years (t4). Transcripts passing the detection threshold of the Affymetrix statistical algorithms (MAS 5.0) were used for analysis. Permutation analysis provided with an estimate of data robustness. In order to distinguish subgroups of treated patients based on generated RNA-profiles and clinical parameters, supervised learning algorithms (k-means, PCA) applying the bootstrap/LOOCV method have been used on 18 individuals (training set) and for validation of prognostic model performance, data from remaining seven MS patients (test set) receiving same therapy were tested for model accuracy. Absolute expression values and ratios were used as discriminating variables.
Results: Various decision trees containing sets of in total 16 genes perform reasonably well with a good outcome at 48h, 1 month or 1 year into therapy. Clinical parameter (EDSS, relapse rate, EP) correlate differently well with attained molecular RNA profiles.
Relevance: Considering the gene expression variation in humans, attaining reliable surrogate biomarkers that predict the outcome of a disease modifying treatment is extremely demanding. The value of selected genes is high in presented cohort, but will need further confirmation in (inter)national meta-analyses.