Cells are constantly communicating with each other to keep tissues healthy and functioning properly. Much of what we know about cell to cell communication focuses on proteins, such as hormones or signaling molecules that bind to receptors on neighboring cells. However, cells also communicate using small molecules called metabolites. These metabolite driven signals are harder to detect and study.

Researchers at Boston Children’s Hospital have developed a new computational tool called MEBOCOST to better understand this type of communication. Their goal was to identify metabolite mediated cell cell communication, also called mCCC, using single cell RNA sequencing data.

Single cell RNA sequencing allows researchers to measure which genes are active in individual cells. From these gene activity patterns, scientists can infer what enzymes and metabolic pathways are working inside each cell. MEBOCOST combines this gene expression information with a method called metabolic flux balance analysis, which estimates how metabolites flow through cellular pathways. By integrating these approaches, the algorithm predicts which cells are likely producing certain metabolites and which neighboring cells are equipped to respond to them.

MEBOCOST detects metabolite-sensor communications between cells with scRNA-seq

(A) Cartoon to show that MEBOCOST predicts metabolite-mediated cell-cell communications, in which sender cells secret metabolites and receiver cells receive metabolites through three types of sensor proteins, including cell surface transporter, cell surface receptor, and nuclear receptor. (B) Cartoons to show the seven steps in MEBOCOST to identify cell-cell metabolite-sensor communications. [1] The RNA expression data and cell type annotation obtained from scRNA-seq were taken as the input data to MEBOCOST. [2] A knowledgebase of enzymes (2.1) and sensor proteins (2.2) for metabolites is incorporated into the MEBOCOST algorithm. [3] The RNA expression values of the metabolite enzymes were extracted from scRNA-seq data. [4] The RNA expression values of sensor proteins for metabolites were extracted from the scRNA-seq data. [5] Calculate the enzyme-sensor co-expression score between two cell groups by taking the product of the mean of enzyme expression values in the sender cells and the mean of sensor expression values in the receiver cells. [6] Shuffling single cell labels to generate a statistical null distribution to calculate the P-value for the enzyme-sensor co-expression score. [7] Incorporating metabolite efflux rates to indicate secretion in sender cells and influx rates to indicate uptake activity in receiver cells, respectively, for combining with the enzyme-sensor co-expression score to identify communication events.

The team tested MEBOCOST using simulated datasets, spatial transcriptomics data, CRISPR screening results, and clinical patient samples. Across these different types of data, the tool consistently detected biologically meaningful metabolite based communication events.

When the researchers applied MEBOCOST to single cell RNA sequencing datasets from human white adipose tissue, they found that macrophages were a major source of metabolic communication changes in obesity. This suggests that immune cells in fat tissue may influence how other cells behave through metabolite signals.

They also analyzed brown adipose tissue in mice. The algorithm not only rediscovered known communication pathways, but also identified new ones. One example involved glutamine mediated signaling from endothelial cells to adipocytes. Experimental validation confirmed that this communication pathway plays a role in adipocyte differentiation, which is the process by which precursor cells mature into fat cells.

Overall, MEBOCOST provides researchers with a powerful way to map metabolite driven communication between cells in different tissues and disease settings. By expanding the focus beyond proteins to include metabolites, this work opens new opportunities to understand how cellular communities function in health and disease.

Availability – MEBOCOST is freely available at https://github.com/kaifuchenlab/MEBOCOST.