Single-cell RNA sequencing identifies macrophage transcriptional heterogeneities in granulomatous diseases

To the Editor:

Both sarcoidosis and chronic beryllium disease (CBD) are granulomatous diseases with overlapping clinical features, however, their clinical courses differ. CBD may have slow clinical progression and rarely resolves while sarcoidosis has variable progression and may resolve depending on the stage of the disease [1]. These commonalities and differences imply differences in antigen(s), and potentially clearance and persistence. Alveolar macrophages play an important role in this process driving innate [2] and adaptive immune responses [3]. Furthermore, subpopulations of macrophages may promote granulomatous inflammation and resolution of lung injury [4], key aspects of sarcoidosis and CBD. Using single-cell RNA sequencing (scRNA-seq), we explored the common and unique pathways between progressive (SarcP) and remitting (SarcR) sarcoidosis as well as CBD and sarcoidosis, focusing on macrophages and macrophages subpopulations (recruited vs. resident).

Cases were enrolled from the National Jewish Health (NJH) Granuloma Clinic and healthy controls were used from a recently published study, also recruited at NJH [5]. They underwent bronchoscopy with bronchoalveolar lavage (BAL) for clinical or research purposes. We used scRNA-seq to analyze BAL cells from SarcP (N=2), SarcR (N=2), and CBD (N=3) compared to beryllium sensitized non-diseased subjects (BeS; N=2) or healthy controls (Cont; N=4). We choose BeS as a control group to explore the underlying biological differences in subjects who are at risk to develop CBD/granulomatous disease and also healthy controls. Established criteria were used to diagnose SarcP and SarcR [6] and BeS and CBD [7]. Subjects were on no treatment for at least 3-6 months prior to collection of the BAL except for one CBD patient who received 5-aminosalicylic acid (5-ASA) anti-inflammatory therapy intermittently. We used the R package Seurat [8] to project cells unto Uniform Manifold Approximation and Projection (UMAP) 2D space, identify cell clusters, and detect differentially expressed (DE) genes. Cluster identities were assigned based on the expression of key marker genes. With a specific interest in macrophage subpopulations, we used FCN1 as a marker for recruited macrophages and FABP4 for resident macrophages [5]. We observed very few M2-like macrophages in the recruited population-based on the expression of SPP1 [9]. Downstream pathway and network analyses were performed using MetaCore from Clarivate Analytics. DiffBind analysis of bulk Assay for Transposase-Accessible Chromatin (ATAC-seq) data was used to assess chromatin accessibility using the same subjects’ BAL specimens (total BAL cell) [10]. Following standard quality checks, ATAC sequence reads were restricted to autosomes and subsetted to a nucleosome-free fraction (reads with insert sizes <100bp). Peaks (read pile-ups) were called using Genrich’s ATAC-Seq mode. Non-overlapping consensus peaks between samples were determined using DiffBind. Occupancy analysis (presence/absence of peaks between sample groups) and affinity analysis (differential number of reads within peaks between sample groups were also performed using DiffBind.

After quality control using mitochondrial reads, RNA features, and RNA counts, one BeS patient was excluded due to poor sample quality. A total of 40,760 cells were used for the final analysis. The cell identities are shown in Figure 1A. No major difference was observed regarding cell composition among groups. We focused our analysis on macrophages (excluding cell clusters going through the cell cycle). We first compared disease groups to healthy controls. Using False Discovery Rate (FDR) adjusted p<0.05 and absolute log fold change (log FC)>0.25, we identified 191, 242, and 234 DE genes within all macrophages for the comparisons of SarcP vs. Con, SarcR vs. Con, and CBD vs. Con, respectively, with 82 genes common to all three comparisons (Figure 1B); 47 out of the 82 genes were common to all comparisons in regards to recruited macrophages. Network analysis (Figure 1D) of the 82 overlapping genes identified three key hubs centered on AP-1 complexes (FOS/FOSB/FOSL2/JUNB/JUND), KLF4, and UBC. Pathway analysis of the 82 overlapping genes identified significant pathways (FDR-adjusted p<0.05) related to the immune system, including Immune Response Antigen Presentation by MHC class II. Within this pathway, sarcoidosis and CBD share common sub-pathways related to RHOB (upregulated), RHOA (upregulated), and MARCH1 (downregulated). In addition, immune response HSP60 and HSP70/TLR Signaling Pathway is also shared by sarcoidosis and CBD. Genes related to heat shock proteins such as UBC (Ubiquitin) and HSP70 family including HSPA1A, HSPA1B, and HSPA8 are shared by sarcoidosis and CBD (all upregulated). There were also unique pathways specific to sarcoidosis phenotypes and CBD. For example, the MANR receptor was downregulated in SarcR and CBD. The expression of CDC42, a GTPase related to micropinocytosis, was higher in SarcR. MHC Class II beta chain was increased in expression in SarcP vs SarcR. Taken together, these pathways ultimately regulate CD4+ T cells and stimulate the immunogenic response. We next compared disease groups to BeS. For SarcP vs. BeS, SarcR vs. BeS, and CBD vs. BeS, we identified 186, 202, and 141 DE genes within all macrophages, respectively, with 53 genes common to all three comparisons (Figure 1C). Network analysis of the 53 overlapping genes identified a key hub centered on AP-1 complexes, also found in comparison to the healthy controls. Pathway analysis of the 53 overlapping genes identified similar pathways related to the immune system such as Immune Response Antigen Presentation by MHC class II. Focusing specifically on recruited macrophages, there were 34 genes common to all three comparisons. Overall, we observed dysregulation of similar pathways when comparing disease groups to healthy controls and to BeS. Focusing specifically on epigenetic regulation of sarcoidosis progression, occupancy analysis of ATAC-seq data revealed differences at 12,333 loci that have open chromatin in SarcP but not SarcR with only 134 in SarcR but not SarcP; 12,333 loci were highly enriched (adjusted p<1x10⁻²⁰) in immune processes relevant to sarcoidosis (e.g., leukocyte activation, immune effector process, myeloid cell activation involved in immune response, leukocyte degranulation, and others). Focusing on affinity analysis, we observed statistically significant differences (Benjamini-Hochberg FDR adjusted p<0.05) in chromatin accessibility at 3,512 loci, with enrichment for similar gene categories. Among promoters with open chromatin in SarcP but not SarcR was the promoter of HLA-DRB5. TXN, CXCL2, and CXCL3 were downregulated in sarcoidosis and CBD with much lower expression in SarcP vs. SarcR.

Figure 1.

A. Uniform manifold approximation and projection (UMAP) analysis of cell identifies of all samples demonstrates the subpopulations of macrophages including recruited macrophages and resident macrophages. Cluster markers for main cell populations were identified in Seurat, and additional markers were used to distinguish recruited (FCN1) from resident (FABP4) macrophages. 1B. Analysis of the 82 DE genes (using healthy control) in common to CBD and sarcoidosis in all macrophages. Heatmap of log fold change for each disease comparison is shown. 1C. Analysis of the 53 DE genes (using BeS control) in common to CBD and sarcoidosis in all macrophages. Heatmap of log fold change for each disease comparison is shown. *Genes that are also DE between progressive sarcoidosis (Sarc P) and remitting sarcoidosis (Sarc R). 1D. Network of the overlap DE genes (using healthy control) between sarcoidosis and CBD. Green means stimulate, red means inhibit, and grey means uncertain.

In our study, we identified different macrophage subpopulations using predefined gene markers. A novel network centered on AP-1 complexes was common to sarcoidosis and CBD, in addition to the MHC class II-related pathway within macrophages, using both healthy controls and BeS as comparator groups. We also found several unique pathways for each disease/phenotype that may explain differences in the disease course. To our knowledge, this study is among the first to highlight shared pathways of these granulomatous diseases using single-cell technologies and recognize unique molecular pathways that underpin sarcoidosis progression. The observation of HLA-DRB5 upregulation in SarcP vs. SarcR is similar to previous findings in progressive sarcoidosis using bulk RNA-seq [11]. In another shared network (centered on AP-1 complexes) with lower recruited macrophage FOSB expression, AP-1 might alter the granulomatous inflammatory response, as its activation in alveolar macrophages is critical in promoting inflammation during acute lung injury [12]. Another important set of genes in this network are heat shock proteins (Ubiquitin and HSP70 family). Two studies found that polymorphisms within the HSP70-Hom gene were associated with sarcoidosis [13]. Antibodies to HSP70 or UBC were also present in a subset of patients with sarcoidosis [14]. We also identified a novel key hub gene, KLF4 as upregulated in sarcoidosis and CBD. KLF4 expression is induced in M2 macrophage [15] and in mediastinal lymph nodes of sarcoidosis patients, where the shift towards M2 macrophage subset has been observed [16]. These previous studies and our results imply that M2 macrophages may drive granulomatous inflammation and immunopathogenesis. JUNB (part of AP-1 complex), HLA-DPA1, and PLAUR were DE in both sarcoidosis and CBD and also found in our bulk peripheral blood mononuclear cells transcriptome study [17], suggesting that these genes may play important roles in the immune response outside of the lung.

We focused this first single-cell transcriptome study in granulomatous lung disease on macrophages, however, our future work will use single-cell technologies to characterize T cell transcriptional changes, given the importance of macrophage T-cell interactions and our previous publication on epigenetic and transcriptional changes in T-cell immune pathways in CBD and sarcoidosis [6]. The main limitations of our study are the small sample size and the fact that we were not able to include all varieties of sarcoidosis phenotypes. Despite these limitations, this study offers investigators cell-specific transcriptional changes (genes and networks/pathways) to consider mechanisms of granulomatous disease and as potential drug targets.