Defining tissue- and disease-associated macrophages using a transcriptome-based classification model

Abstract

Macrophages are heterogeneous multifunctional leukocytes which are regulated in a tissue-and disease-specific context. Many different studies have been published using in vitro macrophage models to study disease. Here, we aggregated public expression data to define consensus expression profiles for eight commonly-used in vitro macrophage models. Altogether, we observed well-known but also novel markers for different macrophage subtypes. Using these data we subsequently built the classifier macIDR, capable of distinguishing macrophage subsets with high accuracy (>0.95). This classifier was subsequently applied to transcriptional profiles of tissue-isolated and disease-associated macrophages to specifically define macrophage characteristics in vivo. Classification of these in vivo macrophages showed that alveolar macrophages displayed high resemblance to interleukin-10 activated macrophages, whereas macrophages from patients with chronic obstructive pulmonary disease patients displayed a drop in interferon-{gamma} signature. Adipose tissue-derived macrophages were classified as unstimulated macrophages, but resembled LPS-activated macrophages more in diabetic-obese patients. Finally, rheumatoid arthritic synovial macrophages showed characteristics of both interleukin-10 or interferon-{gamma} signatures. Altogether, our results suggest that macIDR is capable of identifying macrophage-specific changes as a result of tissue-and disease-specific stimuli and thereby can be used to better define and model populations of macrophages that contribute to disease.