Inflammation perturbs hematopoiesis by remodeling specific compartments of the bone marrow niche

Abstract

Hematopoietic stem and progenitor cells (HSPC) are regulated by interactions with stromal cells in the bone marrow (BM) cavity, which can be segregated into two spatially defined central marrow (CM) and endosteal (Endo) compartments. However, the importance of this spatial compartmentalization for BM responses to complex conditions like inflammation remains largely unknown. Here, we extensively validate a combination of scRNA-seq profiling and matching flow cytometry isolation that reproducibly identifies 7 key CM and Endo populations and accurately surveys both niche locations. We demonstrate that inflammatory perturbations exert specific effects on different cellular compartments, with type I interferon responses causing leptin receptor-expressing mesenchymal stromal cells to abandon their normal stromal functions and instead adopt an inflammatory phenotype associated with overproduction of chemokines that modulate local monocyte dynamics in the surrounding microenvironment. Our results provide a comprehensive method for molecular and functional stromal characterization and highlight the importance of altered stomal cell activity in regulating hematopoietic responses to inflammatory challenges.

Figure 1. Stromal cells are arranged in two distinct spatial compartments.

A, Isolation strategy for stromal cells in the central marrow (CM) and endosteal (Endo) fractions. B, Uniform manifold and approximation projection (UMAP) of integrated 10X single cell RNA sequencing (scRNA-seq) stroma datasets derived from n = 3 independent experiments profiling CM (blue) and Endo (red) fractions. C, UMAP showing Louvain clusters with the numbers of analyzed cells, separation of mesenchymal and endothelial cells from hematopoietic contaminants, and cluster nomenclature. AEC: arterial endothelial cells, SEC: sinusoidal endothelial cell, MSC-S: Sca-1⁺ mesenchymal stromal cell (MSC), OPr: osteoblast progenitor, CPr: chondroblast progenitor, FPr: fibroblast progenitor, MSC-L: leptin receptor (LepR)⁺ MSC (adipo: adipodegnic, osteo: osteogenic), EryP: erythroid progenitor, MkP: megakaryocyte progenitor, HSPC: hematopoietic stem and progenitor cells, Pre-neut, pre-neutrophil. D, Feature plots showing expression of indicated genes in the 10X stroma dataset. E, Representative flow cytometry plots showing gating of key stromal populations in Endo and CM preparations (prep.) with flow/10X cluster correspondence. (m)MPr: (multipotent) mesenchymal progenitor. * denote contamination by EryP expressing low levels of CD31 marker. F, Projection of Smart-seq stromal cell transcriptomes onto the 10X stroma dataset.

Figure 2. Mesenchymal cells have two distinct origins in the BM niche.

A, UMAP of a 10X mesenchymal atlas created by integrating n = 3 published 10X scRNA-seq datasets of BM mesenchymal cells^,,, with major lineage branches indicated. Chrondo.: chondroblastic, Adipo.: adipocytic, Osteo.: osteoblastic, Fibro.: fibroblastic. B, Expression of gene signatures associated with Sca-1⁺ or LepR⁺ MSCs (left) and pseudotime calculated with Monocle3 using start nodes corresponding to the highest MSC gene scores (right). C, Projection of Smart-seq scRNA-seq transcriptomes onto the 10X mesenchymal atlas. D-E, GSEA of selectively enriched GO biological pathways in (D) MSC-L and (E) MSC-S. F, Representative confocal image of femur whole mount preparation (scale bar: 100 μm) showing cell masks for MSC-L (green) and MSC-S (purple). G, Bar charts showing frequency of MSC-L and MSC-S within central marrow (CM) and endosteal (Endo) bone regions as a proportion of total number of segmented cells. Data are means ± S.D. with points showing values for individual mice. P. values, derived from permutation tests (D,E) or Welch’s t tests (G).

Figure 3. Distinct mesenchymal progenitors generate bone and cartilage cells.

A-C, Assessment of osteoblastic vs. chondrogenic differentiation from Endo mesenchymal populations: (A) diffusion map created with PHATE from Smart-seq scRNA-seq datasets showing lineage pathways inferred by Slingshot (left) and Louvain clustering (right); (B) GSEA of selectively enriched GO biological pathways (NES: normalized enrichment score, P.adj.: adjusted P values); and (C) violin plot showing expression of Cd24a and percentage of cells in which expression was detectable in cluster 1 (cl.1) and cluster 2 (cl.2) cells. Lines show median values. D, Representative flow cytometry plots (left) and quantification (right) of the proportion of cells expressing CD24 in the CD51⁺ Endo fraction. FMO: fluorescence minus one. E, Expression of indicated genes measured by qRT-PCR in flow cytometry-isolated stroma populations. Results are normalized to expression of Actb. F, Representative images (left) and quantification (right) of chondrocyte differentiation from Endo CD51⁺/CD24^- OPr and CD51⁺/CD24⁺ CPr assessed by toluidine blue staining of colonies after 21 days. Results are gray scale intensity values per area. Data are means ± S.D. with points showing values for individual mice (D,E) or MSC colonies (F). P. values, derived from permutation tests (B), Student’s t tests (F), or one-way ANOVA with Tukey post-hoc test (E).

Figure 4. Inflammatory challenge remodels distinct niche compartments.

A-C, Niche remodeling following interferon-mediated inflammatory challenge: (A) inflammation model produced by repeated injection of polyinosinic:polycytidylic acid (pIC) every second day (q48h) for up to 13 days; (B) representative flow cytometry plots showing increased expression of Sca-1 in inflammatory MSC-Ls (iMSC-L) in the CM of 3 day-pIC-injected mice; and (C) quantification of CM and Endo populations at indicated times following repeated pIC injections. Data are means ± S.D. with n = 3-5 mice per time point. D-F, 10X scRNA-seq profiling of CM and Endo fractions isolated from control (n = 1, 2 male mice) and 3 day-pIC-injected (n = 2, 2 male mice each) mice: (D) UMAP of merged datasets with number of analyzed cells and hematopoietic contaminants (Hem.) shown in black; (E) frequency of identified CM and Endo cell types; and (F) gene score for Hallmark interferon (IFN)-α geneset in indicated cell types. G, Expression of indicated interferon genes in Smart-seq mesenchymal populations. Data are violin plots of Log10 normalized (norm.) Smart-seq counts with median lines. H, GFP conversion upon pIC treatment in Mx1-Cre:Rosa26-mT/mG lineage tracing mice with schematic of experimental design (top left), representative flow cytometry plots (right), and quantification (bottom left) of the proportion of cells converting from tdTomato to GFP in the indicated stromal population, with or without pIC treatment. Data are means ± S.D. with points showing values for individual mice. P. values, derived from two-way ANOVA with Sidak’s post hoc test (C), one-way ANOVA with Tukey’s post hoc test (H), or Wilcoxon rank sum test (G).

Figure 5. Reduced contributions to stromal niche integrity from iMSC-Ls.

A-D, Smart-seq scRNA-seq profiling of MSC-Ls isolated from CM of young untreated control mice (Ctrl, n = 3 biological replicates, 2-3 male mice pooled in each replicate) and iMSC-Ls isolated from CM of 24-month-old mice (Old, n = 1 male mouse) and 3 day-pIC-injected young mice (pIC1/2, n = 2 biological replicates, 1 male & 1 female mouse pooled in each replicate): (A) UMAP of merged datasets with number of individual (i)MSC-L cell analyzed; (B) expression of genes encoding Sca-1 constituents (data are violin plots of Log10 normalized (norm.) Smart-seq counts with median lines); (C) volcano plot showing differentially expressed genes (colored points are genes with log₂ fold change >|1| and adjusted p value <10^-5); and (D) GSEA of selectively enriched GO biological pathways (NES: normalized enrichment score, P.adj.: adjusted P values) in Smart-seq (i)MSC-L dataset. E, Representative images (left) and quantification (right) of CFU-F obtained from CM (i)MSC-Ls isolated from 3 days (3d) PBS (Ctrl) or pIC-treated mice (1,000 cells/35-mm well cultured for 8 days in 5% O₂/iROCK conditions). F, Quantification of CFU-F obtained from Endo MSC-S/(m)MPr isolated from 3d Ctrl or pIC-treated mice (300 cells/35-mm well cultured for 11 days). G-H, Representative micro-computed tomography images from (G) central femurs and (H) quantification of trabecular bone connectivity density and bone volume/total volume (BV/TV) in mice injected with PBS (Ctrl) or pIC for 13 days (13d). I, Effect of 13d pIC treatment on MSC-L response to bone drilling injury with experimental scheme (left) and representative confocal image of tibia whole mount preparation (scale bar: 100 μm) showing bone repair by MSC-L derived osteoblasts (yellow) at the drill site (white arrowheads). Data are means ± S.D. with points showing values for individual mice. P. values, derived from Wilcoxon rank sum test (B), permutation tests (D), or Student’s t test (E,F,H).

Figure 6. iMSC-Ls modulate local monocytes by chemokine production.

A-C, Effect of 3 days (3d) pIC treatment on MSC-L chemokine production: (A) expression of Ccl5 and Cxcl9 genes in Smart-seq (i)MSC-L dataset (data are violin plots of Log10 normalized (norm.) Smart-seq counts with median lines); (B) production of CCL5 and CXCL9 in BM fluids of 3d PBS (Ctrl) or pIC-treated mice; and (C) measurement of CCL5 in supernatant from 24 hours (24h) cultured (i)MSC-Ls with experimental scheme on the left. D-E, 10X scRNA-seq profiling of BM cells isolated from control (n = 1, 2 female mice) and 3d pIC-treated (n = 1, 1 male & 1 female mouse) mice: (D) UMAP of merged datasets with number of analyzed cells; and (E) frequency of identified BM cell types. DC: dendritic cells. F, Chord plots showing predicted senders and receivers for CCL5-CCR1 interactions between indicated mature BM and stromal cell types. G, Monocyte counts in peripheral blood (PB) and BM of 3d Ctrl or pIC-treated mice. H, Representative confocal images of BM sections (left) and quantification (right) of monocytes (scale bar: 50 μm). I, Changes in monocyte migration with scheme of the transwell assays (left) using 10⁶ c-Kit-depleted Ifnar^-/- BM cells in the upper chamber and BM fluid from 3d Ctrl or pIC-treated mice in the bottom chamber, and quantification (right) of the number of monocytes and neutrophils migrating into lower chamber after 2 hours (2h). Data are means ± S.D. with points showing values for individual mice. P. values, derived from Wilcoxon rank sum test (A), or Student’s t test (B,C,G,H,I).

Figure 7. iMSC-L emergence and functions are dependent on type I interferons.

A, Top 3 HOMER identified transcription factor (TF) motifs enriched in genes differentially expressed in iMSC-Ls. B, Expression of Irf7 and Stat1 genes in Smart-seq (i)MSC-L dataset. Data are violin plots of Log10 normalized (norm.) Smart-seq counts with median lines. C, Mouse lines (left), representative flow cytometry plots (middle), and quantification (right) of iMSC-Ls and monocytes in 3 days (3d) PBS (Ctrl) or pIC-treated wild type (WT) or Ifnar^-/- mice. D, Mouse lines (left), representative flow cytometry plots (middle) and quantification (right) of iMSC-Ls and monocytes in 3 days (3d) PBS (Ctrl) or pIC-treated WT or MSC-L-specific Ifnar-deleted (Lepr-Cre⁺:Ifnar^f/f or Ifnar^ΔMSC-L) mice. E, Ccl5 and Irf7 expression measured by qRT-PCR in the indicated flow cytometry-isolated MSC-Ls. Results are normalized to expression of Actb. F, Kinetics of iMSC-L resolution with line plots showing evolution in the numbers of iMSC-Ls (left graph), and frequency of BM monocytes as well as CCL5 BM fluid concentration (right graph) following an initial 3d pIC treatment (n = 4-5 mice per time point). Data are means ± S.D. with points showing values for individual mice. P. values, derived from Wilcoxon rank sum test (B), or one-way ANOVA with Tukey’s post-hoc test (C,D,E,F).