A better understanding of helper T cell differentiation using mathematical modelling and bioinformatics


Helper T cells play an important role in determining the outcome of an immune response. Helper T cells of distinct phenotypes induce different immune responses by producing distinct cytokines. The phenotype of a helper T cell and the production of the corresponding cytokine(s) is determined by so-called ‘master transcription factors’, and is subsequently fixed and maintained by epigenetic modification of the cytokine locus. We formulated a mechanistic mathematical model describing master transcription factor dynamics, and show that self-activation of the master transcription factor, together with the formation of hetero-dimeric complexes, explains the polarisation of helper T cells into multiple phenotypes. We extended our model to include cell division and the dynamics of cytokine expression and cytokine locus epigenetics. The extended model was used it to investigate the role of stochasticity in adopting and main- taining cytokine expression. We show that cell division opens up the cytokine locus by diluting CpG methylation. Surprisingly, stochasticity reduces the time needed by a cell to express cytokine. To investigate the changes that helper T cells undergo at the mRNA level, we applied microarray-assisted gene profiling to ex vivo skewed mouse splenocytes. Principle component analysis of our own data showed that the most important sources of variance in helper T cell differentiation are 1) activation, 2) time since activation, and 3) polarisation into Th phenotypes. To identify genes that are specific for a particular Th phenotype, we constructed a novel non-supervised scoring method called ‘Polar Score’, which finds the rare genes that are differ- entially regulated due to Th skewing, in a sea of genes up-regulated by T cell activation. We apply the Polar Score method to publicly available data, and identify candidate master regulators for the recently described Th9 phenotype. Additionally, we use our own data to identify genes that we expect to be related the Th1 and Th2 phenotypes. The specific cytokines produced by helper T lymphocytes and other cells de- termine the type of immune response that is raised. This suggests that cytokines measured in situ could be employed to reveal the type of immune response. By applying cluster analysis to immunoassay multiplex cytokine measurements of a cohort of controls and juvenile idiopathic arthritis (JIA), rheumatoid arthritis, osteo arthritis, and diabetes patients, we show that different autoimmune diseases can be recognised by their distinct cytokine patterns. Because several cytokines are co-expressed, suggesting that they are co-regulated, these results call for further analysis of the time courses of gene expression during helper T cell differentiation.

PhD thesis, Theoretical Biology & Bioinformatics, Utrecht University, NL