Pulmonary osteoclast-like cells in silica induced pulmonary fibrosis

Abstract

The pathophysiology of silicosis is poorly understood, limiting development of therapies for those who have been exposed to the respirable particle. We explored mechanisms of silica-induced pulmonary fibrosis in human lung samples collected from patients with occupational exposure to silica and in a longitudinal mouse model of silicosis using multiple modalities including whole-lung single-cell RNA sequencing and histological, biochemical, and physiologic assessments. In addition to pulmonary inflammation and fibrosis, intratracheal silica challenge induced osteoclast-like differentiation of alveolar macrophages and recruited monocytes, driven by induction of the osteoclastogenic cytokine, receptor activator of nuclear factor κΒ ligand (RANKL) in pulmonary lymphocytes, and alveolar type II cells. Anti-RANKL monoclonal antibody treatment suppressed silica-induced osteoclast-like differentiation in the lung and attenuated pulmonary fibrosis. We conclude that silica induces differentiation of pulmonary osteoclast-like cells leading to progressive lung injury, likely due to sustained elaboration of bone-resorbing proteases and hydrochloric acid. Interrupting osteoclast-like differentiation may therefore constitute a promising avenue for moderating lung damage in silicosis.

Fig. 1.. Silicosis features of progressive fibrosis, silicotic nodules, and infiltrating monocytes are recapitulated in the mouse model.

(A) H&E-stained human silicotic lung sections. Patient 1 pathologic features include extensive fibrosis (arrows, top) with lymphocyte predominant inflammation (*, top), collagenized silicotic nodules (*, middle), surrounding alveoli with septal inflammation (*, bottom), and intra-AM accumulations (arrows, bottom). Pathologic features in patient 2 include focal fibrosis with prominent dust macules and accumulations of pigment containing macrophages (arrows, top), rare small collagenized silicotic nodules (*, top and middle), and alveoli with lymphocyte predominant septal inflammation (*, bottom), and intra-AM aggregates (arrows, bottom). (B) H&E and (C) Masson’s trichrome reagent–stained lung sections collected from C57BL/6J silica-challenged mice (5 mg, i.t.) at d0 (pre-exposure), d28, and d56. Scale bars, 50 μm. (D) RT-qPCR of Tgfb1, Col1a1, Col3a1, and Fn1 from whole lung homogenates at indicated days after silica administration (5 mg i.t.) (N = 4 mice per group). (E) Hydroxyproline levels in whole lungs were quantified (N = 3 to 5 mice per group). (F) Pulmonary function tests at d56 postsilica challenge (5 mg, i.t.). The representative curves of pressure in the cylinder (Pcyl)/volume from the two groups and (G) compliance measurements were made using the forced oscillation method (see Materials and Methods) (N = 4 to 6 mice per group). Error bars show median with interquartile range. *P < 0.05 and **P < 0.01 by Mann-Whitney U test in two groups comparison and Kruskal-Wallis test followed by Dunn’s test in multiple comparisons.

Fig. 2.. snRNA-seq depicts the global molecular dynamics in silicosis.

snRNA-seq was performed on whole lung from i.t. silica-challenged mice at time 0, d7, d28, and d56. (A) Uniform Manifold Approximation Projection (UMAP) plot of all annotated cells, colored according to fine scale annotation. (B) Selected marker genes expression used for broad and fine-scale cell type annotation, colored by normalized expression, and size indicates percent cells of the cell type with detectable expression of the gene. (C) Cell state abundance relative to d0 shown over time. *P < 0.05, **P < 0.005, and ***P < 0.0005; (beta-binomial test, Benjamini-Hochberg correction). Gene set enrichment scores for inflammation (D) and fibrosis (E) shown over time summarized for all cells collected from each time point (left) and colored by individual cell state (right). T_regs, regulatory T cell; FC, fold change; PNECs, pulmonary neuroendocrine cells; BASCs, bronchioalveolar stem cells; ILCs, innate immune cells; pDCs, plasmacytoid dendritic cells; cDCs, classical dendritic cells; EC, endothelial cells; moDC, monocyte-derived dendritic cells.

Fig. 3.. Myeloid cells demonstrate remarkable heterogeneity and activation of osteoclast-related transcriptional programs in silicosis.

Analysis of myeloid populations identified in longitudinal snRNA-seq of whole lung from i.t. silica-challenged mice. (A) UMAP plot of myeloid cells. (B) Selected marker genes used for fine-scale cell annotation of myeloid cells. Size indicates percent of cells expressing the marker, and color indicates average expression value. (C) Relative abundance (normalized by library size factors) is shown for each myeloid cell state over time. (D) Differentially expressed genes (DEGs) relative to d0 were identified for AMs (false discovery rate). Osteoclast-related genes denoted in bold typeface. (E) UMAP of IMs colored according to fine-scale annotation (top) and pseudotime (bottom). (F) Heatmap of genes differentially expressed over pseudotime as IMs differentiate toward profibrotic phenotypes. Each row represents a gene, and color represents relative intensity of expression. Significant (g:Profiler, g:SCS) representative gene ontology (GO) terms shown on right. (G) Osteoclast differentiation enrichment scores for all myeloid cells plotted over time. Boxplots demonstrating cell state enrichment scores for (H) osteoclast differentiation, (I) osteoclast development, and (J) osteoclast signaling. (K) Select osteoclast genes plotted over time for myeloid cell states. *P < 0.05.

Fig. 4.. Single-cell analysis of coal miner lungs reveals a population of POLCs.

UMAP of annotated cell types colored by (A) fine annotation, (B) case-control status, and (C) patient-reported tobacco use. (D) Cells were broadly and finely annotated using canonical marker genes. (E) UMAP of myeloid cells. (F) Osteoclast-related gene expression for all myeloid cells, colored by normalized average expression and sized according to percent cells expressing the marker. (G) Osteoclast development and (H) osteoclast signaling enrichment scores for each cell grouped by annotation. (I) Normalized cell counts of POLCs per subject. POLC, pulmonary osteoclast like cells; SMC, smooth muscle cells; CWP, coal workers pneumoconiosis.

Fig. 5.. Prominent osteoclast-like phenotypes and molecular markers are evident in BAL and in whole lung of silica-challenged mice.

(A) TRAP and (B) CTSK staining of human silicotic lung tissue (see figs. S7 to S9). Arrows indicate TRAP and CTSK positive macrophage adjacent to fibrotic regions. (C) Mouse lung tissue was stained for TRAP (top) and CTSK (bottom) at d0, d28, and d56. (D) BAL MNGCs collected at d28 stained with Giemsa, TRAP, and CTSK (see fig. S10). (E) TRAP5b, quantified by ELISA, in BALF from saline- or silica-treated mice collected d28 postchallenge (N = 4 mice per group). N.D., not detected. (F) CTSK activity measured by radiant efficiency of i.t. Cat K 680 FAST delivered d6 postsilica challenge (N = 4 mice per group) (see fig. S11). (G) RT-qPCR for osteoclast genes in BAL cells (N = 4 mice per group) (see fig. S12). (H) Scheme for ex vivo osteoclast experiments. (I) BAL cells collected at d14 post i.t. challenge with saline (I-1, I-3, and I-5) or silica (I-2, I-4, and I-6). Representative images chosen from cultures supplemented with RANKL (25 ng/ml; see fig. S13) After d6, cells cultured on plastic were stained for TRAP activity (I-1 and I-2; yellow arrows). Cells cultured on bone used phalloidin staining to visualize actin rings (I-3 and I-4; yellow arrows), and resorbed bone area was visualized by peroxidase-conjugated wheat germ agglutinin/horse radish peroxidase staining after removing cells (I-5 and I-6; yellow arrows). Scale bars, 200 μm (I-1 and I-2) and 50 μm (I-3, I-4, I-5, and I-6). Number of actin ring (J), actin ring size (K), total actin ring area (L), and bone resorption area (M) were quantified. Error bars show median with interquartile range. *P < 0.05 and **P < 0.01 by Mann-Whitney U test or Kruskal-Wallis test followed by Dunn’s test. Scale bars are 20 μm unless noted.

Fig. 6.. Osteoclast differentiation and activation are RANKL dependent.

Analysis of BAL cytokines and cell populations in silica-challenged mice (5 mg, i.t.). TNF-α (A), IL-1β (B), IL-6 (C), M-CSF (D), IL-4 (E), RANKL (F), and OPG (G) concentrations, measured by ELISA in BALF at indicated time points (N = 4 mice per group). RANKL gene expression, measured by RT-qPCR, in BAL (H) and whole lung homogenates (I) (N = 4 mice per group). RANKL protein concentration, measured by ELISA, in BAL (J) and AM (K) lysates from silica-treated mice on d14 postchallenge (N = 3 to 4 mice per group). Number of RANKL-positive and IFN-γ–negative cells in (L) B cells, (M) CD4⁺ T cells, (N) CD8⁺ T cells, and (O) NK cells from whole lungs collected d14 postchallenge (N = 4 mice per group) (see fig. S16). (P) RANKL in AT2 cell lysates from silica-treated mice on d14 postchallenge, measured by ELISA (N = 3 mice per group). Error bars show median with interquartile range. *P < 0.05, **P < 0.01, and ***P < 0.001 by Mann-Whitney U test in two groups comparison and Kruskal-Wallis test followed by Dunn’s test in multiple comparisons.

Fig. 7.. Inhibiting the RANKL-dependent differentiation of POLCs ameliorates pulmonary fibrosis in the silicosis mouse model.

(A) Study design for anti-RANKL mAb studies. Mice treated with RANKL mAb or Ctrl IgG (0.25 mg per mouse, i.p., three times per week) were challenged with silica (5 mg) and euthanized d28 later. (B) RT-qPCR for osteoclast-related genes: Acp5 (TRAP), Ctsk, Mmp9, and Atp6v0d2 in BAL cells (N = 7 to 9 mice per group) (see fig. S18). (C) Lung sections were stained for TRAP enzymatic activity (top) or with anti-CTSK (bottom) (N = 6 mice per group) (see fig. S19). Scale bars, 20 μm. (D) TRAP5b in BALF, measured by ELISA (N = 7 to 9 mice per group). Paraffin-embedded lung sections were stained with (E) H&E and (F) Masson’s trichrome reagent (N = 6 mice per group). Scale bars, 20 μm. (G) Expression of fibrosis-related genes quantified in right lungs using RT-qPCR (N = 7 to 9 mice per group). (H) Hydroxyproline levels in left lungs were quantified (N = 7 to 9 mice per group). Pulmonary function tests performed included static compliance (I) and P-V curve analysis (J) (see fig. S20) (N = 4 to 6 mice per group). Error bars show median with interquartile range. *P < 0.05, **P < 0.01, and ***P < 0.001 by Mann-Whitney U test in two groups comparison and Kruskal-Wallis test followed by Dunn’s test in multiple comparisons.