Abstract
Obesity is a metabolic disease that promotes the development of a number of other pathologies. Despite its high disease burden, the underlying pathophysiology of obesity is poorly understood. Emerging research has indicated that adipocytes transfer their mitochondria to macrophages in white adipose tissue as a mechanism of cell-to-cell communication and that this process is impaired in obesity. However, the diversity of intercellular mitochondria transfer axes that occurs in adipose and its regulation in obesity are not known. Here, we utilized 31-color spectral flow cytometry of adipocyte-specific mitochondria reporter (MitoFat) mice to comprehensively analyze intercellular mitochondria transfer from adipocytes to other cell types in white, beige, and brown adipose tissues. Employing manifold machine learning, we generated reference clusters of cells in 5-month (young) and 20-month-old (aged) MitoFat mice fed a normal chow diet (low fat diet). Using the reference clusters and manifold, we then mapped differences in immune cell populations using nearest neighbor search approximations in MitoFat mice fed normal chow, high-fat diet (HFD), high-fat diet with low palmitate (LP-HFD). The degree of mitochondria transfer from adipocytes to each of the various cell clusters was determined for each tissue and for each condition. We observed that adipocytes transfer their mitochondria to a wide range of immune cell populations, most notably macrophages. Although aged mice develop obesity, surprisingly they do not exhibit decreased mitochondria transfer from adipocytes to macrophages in vivo in white, beige, or brown adipose tissue. In contrast, young mice fed a HFD highly enriched in palmitate exhibit obesity and markedly reduced mitochondria transfer from adipocytes to macrophages. The decrease in mitochondria transfer was largely ameliorated by the replacement of palmitate with medium chain fatty acids, suggesting a potential direct dietary mechanism in the alteration of mitochondria transfer. Overall, the 31-color quantification increased granularity, allowing us to quantify differences in immune populations and mitochondria transfer by tissue, age, and diet. Similar machine-learning approaches could be used to investigate both basic biological and clinical questions by effectively reducing dimensions, mitigating batch effect, and enabling comparisons across different tissues, timepoints, or conditions.
Prognostic Significance of Immune Cell Populations Identified by Machine Learning in Colorectal Cancer Using Routine Hematoxylin and Eosin–Stained Sections
