Inference of an in situ epidermal intracellular signaling cascade

Abstract

The stratified architecture of the epidermis makes it an ideal system in which to investigate intracellular signaling pathways within the context of a native tissue. We have applied quantitative imaging protocols to investigate the expression of 13 total-proteins and 4 phosphorylated-proteins across human epidermis. Plasma membrane, nuclear and/or cytoplasmic protein expression levels were measured along the gradient of keratinocyte differentiation. Dynamic Bayesian network techniques were used to infer conditional dependencies between the expression levels of target molecules and construct an associated cascade topology. The resulting networks were compared against a canonical network to investigate the extent to which known biochemical interactions could be recapitulated in situ. Biochemical evidence from the literature supported the majority (71-86%) of inferred network edges, however, overall coverage of the canonical network was relatively low (12-31%). Identified edges may represent key signaling pathway interactions which occur during keratinocyte differentiation. Inferred networks were ranked by model likelihood given the data and the top five were used to construct a consensus network. Several edges were present within this consensus network yet absent from the canonical network. These edges may represent putative interactions which occur in human epidermis.