Article Text
PDF
Abstract
We present a detailed analysis of myeloid cell populations found in the kidneys of lupus nephritis (LN) patients, based on the single-cell RNA-sequencing (scRNA-seq) data collected as part of the Accelerating Medicines Partnership (AMP) in RA/SLE consortium. Overall, 23,819 cells isolated from 156 LN patients and 30 healthy donors passed QC. Clustering of these cells (figure 1A) identified populations of CD14 and CD16 monocytes, two subsets of tissue-resident macrophages and several types of dendritic cells (DCs). In addition, we found several transcriptionally-distinct subsets of differentiated macrophages, that were missing from healthy donors (figure 1B- C). The ratio between the frequency of these macrophage subsets and that of infiltrating monocytes positively correlated with the Activity Index (AI) (figure 1D).
To infer the origins of the observed disease-specific macrophages, we compared them to several published scRNA-seq datasets of blood and kidney samples, and performed in addition trajectory analysis. Our results suggested that these subsets likely originate from both infiltrating monocytes and tissue-resident macrophages (figure 2A). Furthermore, our analysis indicated that the differentiation into disease-specific macrophages mostly takes place within the kidney. To identify putative extracellular signals driving the differentiation of infiltrating CD16 monocytes into disease-related activation states, we performed in vitro experiments in which CD16 monocytes were stimulated with a wide array of cytokines and molecules suggested to play role in SLE pathology, such as immune complexes (ICs) and various types of cellular debris. We measured transcriptional changes associated with each in vitro condition, and utilized the generated data to identify enriched signatures in the AMP scRNA-seq data, using gene set enrichment analysis (GSEA). This analysis suggested that apoptotic cells likely promote differentiation of CD16 monocytes into a phagocytic state (cluster 11; figure 2B). In contrast, ICs containing TLR7 ligands, as well as IFNγ, were found to be plausible drivers of differentiation into an activation state that was characterized by high production of several proinflammatory cytokines and chemokines (‘high producers’ – clusters 12 and 13; figure 2C-D). Of note, our analysis suggested that through chemokine production, these ‘high producers’ may play a central role in recruiting and retaining the phagocytic macrophage subsets.
The frequency of a single population of disease-specific macrophages positively correlated with both the AI and the Chronicity Index (CI; figure 3A-B). This population (cluster 17) was characterized by the upregulation of a set of genes associated with lipid metabolism. While previous studies have reported the presence of a similar macrophage subset in other tissues, this has not yet been demonstrated in kidneys. Furthermore, our analysis identified subclusters within this population, associated with different specific pathways, that were separately correlated with the AI and CI. In particular, we found a proinflammatory signature in these cells that was negatively correlated with the AI and positively correlated with the CI (figure 3C-D).
A systemic differential expression analysis showed that several myeloid subsets modulated their gene expression in a manner correlated with the AI, compared to healthy donors; a particular clear response was observed in CD16 monocytes (cluster 2), in both proliferative/mixed and pure membranous LN (figure 4A-B). GSEA suggested that these changes were driven, at least in part, by type I and type II IFN, IL-6 and TNFβ. A conjoint analysis of changes in subset frequencies and of the differentially expressed (DE) genes in these populations and in glomerular endothelial cells pointed to the concurrent upregulation of molecules that may promote fibrosis, and in particular fibronectin in CD16 monocytes and integrins capable of binding it in glomerular endothelial cells (figure 4D-F); furthermore, several of the phagocytic macrophages derived from CD16 monocytes upregulated a set of genes regulating the extracellular matrix (figure 4G). Of note, these observations were found in LN patients that had a 0 glomerular CI (defined as the sum of glomerular subscores of the CI), suggesting that these molecular events precede fibrosis and may promote it. An increase in glomerular CI was associated with significant changes in gene expression, in particular in cDC2 and pDCs (clusters 4 & 15), as well as 2 specific populations of phagocytic macrophages (clusters 14 and 15; figure 4C); these changes included a decrease in the interferon response. We verified the reproducibility of these signatures in an independent set of LN patients.
Taken together, our results shed light on the mechanisms of kidney inflammation in LN, and provide a detailed view of the different subsets and activation states of myeloid cells found in LN kidneys and the putative relations between them, as well as the extracellular signals giving rise to these states.
(A) 24 subsets of myeloid cells, identified through clustering of the kidney scRNA-seq data of AMP RA/SLE. (B) The relative frequencies of myeloid cell clusters (fraction out of all myeloid cells) in LN patients. (C) The relative frequencies of myeloid cell clusters (fraction out of all myeloid cells) in healthy donors. (D) The ratio between differentiated macrophages (clusters 5, 7, 8, 10, 14, 16, 17 and 18) and monocytes (clusters 1, 2, 3 and 6), and its relation with the NIH Activity Index. Each point corresponds to a single LN patient. Only patients with at least 20 myeloid cells were considered in this analysis. Reported is the Spearman correlation and its associated p-value.
(A) The results of trajectory analysis using the PAGA software package. Lines represent putative transitions between cell clusters; only transitions found in our analysis to be highly plausible are shown. (B-D) The results of gene set enrichment analysis (GSEA) of the AMP SLE kidney scRNA-seq data, based on experiments involving the in vitro stimulation of CD16 monocytes with various signals. Each panel shows the enrichment score of genes found to be upregulated by a specific condition; gene ranking is based on comparison of the phagocytic cluster 11 (B) or ‘high producers’ cluster 12 (C-D) with the cluster corresponding to CD16 monocytes entering the kidney from blood (cluster 2). (B) Stimulation by apoptotic cells. (C) Stimulation by immune complexes containing R848, a TLR7 ligand. (D) Stimulation by IFNγ.
(A-B) Association of cell neighborhood frequencies with the Activity Index (A) and Chronicity Index (B), as computed using Covarying Neighborhood Analysis (CNA). Only correlations passing an FDR threshold of 0.1 are shown. Positive correlations are denoted by red, negative correlations by blue. Cluster 17 is circled in green. (C-D) The relation between the relative frequency of subcluster 1, within cluster 17, and the Activity Index (C) and Chronicity Index (D). Reported are the values of the Spearman correlation and their associated p-values.
(A-C) The number of DE genes as a function of available cells. Outlier subsets with an exceptionally high number of DE genes are highlighted. Lines are provided as visual guides only. (A) The comparison of patients with proliferative/mixed LN and high AI to healthy donors. (B) The comparison of patients with pure membranous LN and high AI to healthy donors. (C) Comparing patients with high CI to patients with low CI (proliferative/mixed). (D) The expression of fibronectin (FN1) in myeloid cells. (E-F) The upregulation of integrins in glomerular endothelial cells, comparing patients with proliferative/mixed LN and high AI with healthy donors. (G) The average scaled expression of 238 regulators of the ECM; elevated expression is observed in clusters 7, 8, 10, 16, 17 and 23.
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.