Single-cell mass cytometry on peripheral blood identifies immune cell subsets associated with primary biliary cholangitis

Abstract

The relationship between primary biliary cholangitis (PBC), a chronic cholestatic autoimmune liver disease, and the peripheral immune system remains to be fully understood. Herein, we performed the first mass cytometry (CyTOF)-based, immunophenotyping analysis of the peripheral immune system in PBC at single-cell resolution. CyTOF was performed on peripheral blood mononuclear cells (PBMCs) from PBC patients (n = 33) and age-/sex-matched healthy controls (n = 33) to obtain immune cell abundance and marker expression profiles. Hierarchical clustering methods were applied to identify immune cell types and subsets significantly associated with PBC. Subsets of gamma-delta T cells (CD3⁺TCRgd⁺), CD8⁺ T cells (CD3⁺CD8⁺CD161⁺PD1⁺), and memory B cells (CD3^-CD19⁺CD20⁺CD24⁺CD27⁺) were found to have lower abundance in PBC than in control. In contrast, higher abundance of subsets of monocytes and naïve B cells were observed in PBC compared to control. Furthermore, several naïve B cell (CD3^-CD19⁺CD20⁺CD24^-CD27^-) subsets were significantly higher in PBC patients with cirrhosis (indicative of late-stage disease) than in those without cirrhosis. Alternatively, subsets of memory B cells were lower in abundance in cirrhotic relative to non-cirrhotic PBC patients. Future immunophenotyping investigations could lead to better understanding of PBC pathogenesis and progression, and also to the discovery of novel biomarkers and treatment strategies.

Figure 1

Schematic overview of the current study’s analysis pipeline to investigate immune cell populations implicated in PBC using the Cytobank cloud-based platform (Cytobank, Inc.).

Figure 2

Qualitative analysis of PBMC immunophenotyping data using (A) viSNE and (B) FlowSOM reveals differences in immune cell lineages between PBC and control. (A) Mass cytometry samples from 33 PBC patients were concatenated into 100,000 randomly-sampled total events and mapped onto a t-SNE plot using viSNE (left). Analogously, samples from 33 age-/sex-matched controls were mapped using viSNE (right). Each point in the t-SNE plot represents a single event (e.g., cell) detected by the mass cytometer, and colors vary according to cell abundance density. Observed regional differences in cell densities correspond to differences in relative abundances of major immune cell lineages. (B) FlowSOM clusters cells into cell subsets based on their marker expression patterns, and generates a Minimum Spanning Tree (MST) of those clusters (left: PBC; right: control). Each node is characterized by a pie chart, whose diameter is proportional to the number of events, and whose colors indicate specific markers defined in the legend. The background colors group nodes into cell types that correspond to different major immune cell types (FlowSOM Metaclusters). Each link connects cell subsets of similar marker expression patterns. In the two MSTs of PBC and control, the nodes located in the same position correspond to the same cell subset.

Figure 3

FlowSOM metaclusters, which represent immune cell types, characterize differences in immune profiles between PBC and control. (A–D) FlowSOM metaclusters that varied in relative abundance, i.e., proportion, across study groups. Boxplots indicate relative abundances of Metacluster-3, Metacluster-4, Metacluster-16, and Metacluster-18, which corresponds to gamma-delta T cell (CD3⁺TCRgd⁺), CD8⁺ T cell (CD3⁺CD8⁺CD161⁺PD1⁺), memory B cell (CD3⁻CD19⁺CD27⁺CD38⁻), and naïve B cell (CD19⁺CD27⁻IgD⁺CCR7⁺) immune cell types, respectively. The ‘Cirrhosis’ and ‘Non-cirrhosis’ groups are subsets of the ‘PBC’ group. Numbers in parentheses indicate sample size of the group. Cirrhosis status was unavailable for one PBC patient. Horizontal bars indicate the Mann–Whitney U test performed on the respective group pairs (significance level: *0.01 ≤ p < 0.05; **0.001 ≤ p < 0.01; and ***p < 0.001). (E) Marker expression patterns for each of the nine FlowSOM clusters (derived from Metaclusters-3, -4, -16, and -18) found to be differentially abundant (i.e., fold-change ≥ 2.0 and p < 0.05) between PBC and control. Clusters from the same metacluster grouped according to their marker intensities. Color corresponds to marker intensity, which was scaled and centered for each marker.

Figure 4

CITRUS identifies differentially abundant immune cell subsets between PBC patients and age-/sex-matched controls. (A) CITRUS produces a radial hierarchical tree of cells subsets using an unsupervised clustering approach. Each node (i.e., cluster) represents a subset of cells, and each edge points from a parent node to child node(s). Only the highlighted nodes correspond to cell subsets of statistically significant differential abundance using a significance analysis of microarray (SAM) correlative association model (Benjamini–Hochberg adjusted p value < 0.05). (E) CITRUS map shows a significantly higher abundance of CD14⁺CD11c⁺CD66b⁺ cells, and lower abundances of CD19⁺CD24^hiCD27⁺ and CD3⁺CD8⁺CD161⁺ cells, in PBC compared to control. CITRUS maps overlaid with marker-specific intensities show their relative expression levels (across all nodes) in the proposed (B) CD19⁺CD24^hiCD27⁺; (C) CD14⁺CD11c⁺CD66b⁺; and (D) CD3⁺CD8⁺CD161⁺ cell subsets. The nodes circled in black indicate positive or high expression of a particular marker.