In the field of dermatology, microarray-based studies have led to the discovery of complex gene expression patterns underlying the activation of specific cells infiltrating in skin lesions of chronic inflammatory conditions 1 like psoriasis and atopic dermatitis. These diseases are accompanied by characteristic changes in the cellular composition of the skin, which are usually determined based on flow cytometry or immunohistochemistry. However, these techniques are often constrained by the availability of cell type-specific biomarkers and are typically restricted to identification of a small subset of cell types. Recently, robust deconvolution methodologies 2 like CIBERSORT 3 have been implemented for fast and accurate computation of cell composition in bulk tissue based on gene expression data from biopsies. In this study, we apply CIBERSORT to inflammatory skin conditions and derive the phenotype-specific relative fractions of cell subsets associated with healthy and diseased skin. This deconvolution method requires two inputs, (i) the expression data from the sample of interest and (ii) a reference matrix with the gene expression signatures from isolated cell types. CIBERSORT was originally developed for deconvolution of blood cell subsets and the study of immune cell infiltration in cancers. Thus, the published signature matrix accounted only for leukocytes and did not include any skin-specific cell types. Here, those cell subsets were first added to the signature matrix based on publicly available expression data from basal and suprabasal epidermal cells. The result was a skin-specific signature matrix, which is comprised of twenty-six different cell phenotypes. Analysis of these cell type-specific signatures showed that it is possible to reliably distinguish leukocyte subsets, basal and suprabasal keratinocytes. Then, raw microarray data from ten studies was gathered from Gene Expression Omnibus. Pooled data comprised skin biopsies from healthy, lesional and non-lesional tissue of patients with psoriasis or atopic dermatitis. The deconvolution analysis was implemented for each dataset and the relative fraction of each cell subset was derived. The deconvolution analysis confirmed the characteristic changes in cellular infiltrates known to be involved in skin lesions of psoriasis and atopic dermatitis. Further, it shed light on the key differences between the lesional and non-lesional phenotypes of these diseases, as well as on the effect of phototherapy on skin cell type composition. To evaluate the robustness of the method, we compared the relative cell counts computed for each dataset. This assessment showed consistency in the numbers across datasets. Overall, our results suggest that computer-based cellular profiling of inflammatory skin conditions is a strong, fast and effective alternative to current experimental approaches, which only account for a fraction of the cell types found in the skin. The use of this tool in dermatology is, however, currently limited by the completeness and accuracy of the signature matrix. In the future other cell types could be included to strengthen the approach. The application of deconvolution analysis to dermatological diseases is of great interest for tracing the effect of diverse therapeutic approaches. References 1. Nomura, I. et al. Distinct patterns of gene expression in the skin lesions of atopic dermatitis and psoriasis. J. Allergy Clin. Immunol. 112, 1195–1202 (2003). 2. Lenz, M., Schuldt, B. M., Müller, F.-J. & Schuppert, A. PhysioSpace: Relating Gene Expression Experiments from Heterogeneous Sources Using Shared Physiological Processes. PLoS ONE 8, e77627 (2013). 3. Newman, A. m. ( 1, 2 ) et al. Robust enumeration of cell subsets from tissue expression profiles. Nat. Methods 12, 453–457 (2015).
|Status||Gepubliceerd - 2017|
|Evenement||ECCB 2017 - 16th European Conference on Computational Biology - Prague, Tsjechië|
Duur: 21 jul 2017 → 25 jul 2017
|Congres||ECCB 2017 - 16th European Conference on Computational Biology|
|Periode||21/07/17 → 25/07/17|