Transcriptomic profiling of the myeloma bone-lining niche reveals BMP signalling inhibition to improve bone disease

Abstract

Multiple myeloma is an incurable, bone marrow-dwelling malignancy that disrupts bone homeostasis causing skeletal damage and pain. Mechanisms underlying myeloma-induced bone destruction are poorly understood and current therapies do not restore lost bone mass. Using transcriptomic profiling of isolated bone lining cell subtypes from a murine myeloma model, we find that bone morphogenetic protein (BMP) signalling is upregulated in stromal progenitor cells. BMP signalling has not previously been reported to be dysregulated in myeloma bone disease. Inhibition of BMP signalling in vivo using either a small molecule BMP receptor antagonist or a solubilized BMPR1a-FC receptor ligand trap prevents trabecular and cortical bone volume loss caused by myeloma, without increasing tumour burden. BMP inhibition directly reduces osteoclastogenesis, increases osteoblasts and bone formation, and suppresses bone marrow sclerostin levels. In summary we describe a novel role for the BMP pathway in myeloma-induced bone disease that can be therapeutically targeted.

Fig. 1

Transcriptome profiling of bone-lining niche components from myeloma-bearing mice identifies dysregulated BMP signalling. a Bone-lining cells disassociated from the bone surface of hindlimb long bones using enzyme digestion were FACS-sorted for RNA-Seq in 100 cell aliquots. Percentages shown are a proportion of the live cell gate. GFP+ cell abundance varied with tumour burden. The full gating strategy is shown in Supplementary Fig. 1. b Multidimensional Scaling (MDS) plot of the four cell types sorted, using leading log-fold changes in gene expression between each pair of samples. To confirm cell phenotypes, a heat map was generated using established marker gene expression for each cell type (colour scale; log RPKM values). c Heat map of stromal progenitor gene expression, showing genes differentially expressed with adj. p value <0.1 and abs(log 2FC) >1 (n = 82) (colour scale; centred and scaled log RPKM values). d Myeloma-bearing vs. healthy stromal progenitor GSEA enrichment plots for bone formation (FDR 0.019), bone loss (FDR 0.007) and bone remodelling (FDR 0.062) gene sets. e Overlap between limma-generated differentially expressed genes (adj. p value <0.1 and abs(log 2FC) >1) and GSEA-ranked top 100 genes (n = 18). Only Bmpr1b has established role in bone remodelling. GSEA of BMP signalling gene set showed enrichment in myeloma-bearing vs. healthy stromal progenitors (FDR 0.006). f The bone remodelling gene set (n = 151) supplemented with BMP family members (n = 44, 21%) and reanalysed: FDR decreased from 0.062 to 0.007. The gene set leading edge is shown, comprising 32% BMP pathway members, targets or interactors (marked with red arrows). Colour scale as per GSEA algorithm: the range of colours (red, pink, light blue, dark blue) shows the range of RPKM expression values (high, moderate, low, lowest). OP: osteoprogenitors; SP: stromal progenitors; EC: endothelial cells

Fig. 2

Inhibition of BMP signalling with LDN193189 reduces myeloma bone disease in vivo, with no effect on tumour burden. n = 10 per control group, 20 per myeloma group, total n = 60 for microCT-derived indices; n = 5 per control group, 10 per myeloma group, total n = 30 for other indices. Missing data values: see Methods. *P < 0.05, **p < 0.01, ***p < 0.001, NS, not significant, t test. Source data are provided as a Source Data file. Reconstructed tibial metaphyseal images were analysed using Bruker CTAn software to generate (a–d): a total BV/TV (bone volume/total volume). b Cortical volume in μm³. c Trabecular BV/TV. d Trabecular thickness in μm. e Reconstructed microCT images of tibial sagittal sections showing trabecular architecture. f Histomorphometric generated BV/TV (femoral metaphyseal sections). g Enumerated cortical lytic lesions on right medial tibial surface. h Effect of BMP6 on JJN-3, MM1-S (human) and 5TGM1 (mouse) myeloma cell line viability at 72 h (alamarBlue® assay), n = 4 replicates. Source data are provided as a Source Data file. i Dose effect of LDN193189 on BMP6-induced myeloma growth inhibition at 72 h. Viability was expressed relative to viability in control conditions, n = 3 replicates. Source data are provided as a Source Data file. j Effect of hBMP6, LDN193189 and contact co-culture with stromal line HS5 on proliferation of 5TGM1-GFP cells (GFP fluorescence intensity as proxy for growth), n = 4 replicates. k 5TGM1-GFP myeloma disease burden quantification: serum IgG2b level (mg/ml). l 5TGM1-GFP myeloma disease burden quantification: bone marrow GFP+ percentage. m Effect of LDN193189 on myeloma distribution between central marrow niche (flushed from marrow cavity) and bone-lining niche (collagenase-digested from cleaned bone fragment surfaces) (n = 8 per group). Data represent mean ± SD

Fig. 3

Inhibition of BMP signalling with BMPR1a-FC reduces myeloma bone disease in vivo, with no effect on tumour burden (5TGM1 model; n = 5 per control group, 10 per myeloma group, total n = 30. JJN-3 model; n = 5 per control group, 8 per myeloma group, total n = 26). *P < 0.05, **p < 0.01, ***p < 0.001, NS, not significant, t test. a Reconstructed microCT images of tibial sagittal sections showing trabecular architecture. Reconstructed tibial metaphyseal images analysed using Bruker CTAn software to generate b–d: b trabecular BV/TV (bone volume/total volume). c Trabecular thickness in μm. d Cortical volume in μm³. e Histomorphometric analysis of right femoral metaphysis sections using Osteomeasure software to generate BV/TV. f 5TGM1-GFP myeloma disease burden quantification using serum IgG2b level (mg/ml). g 5TGM1-GFP myeloma disease burden quantification using bone marrow GFP+ percentage. h Trabecular BV/TV (bone volume/total volume) in JJN-3 model. i Trabecular thickness in μm in JJN-3 model. j JJN-3 myeloma disease burden quantification using kappa free light chain quantification (mg/l). k JJN-3 bone marrow tumour burden was calculated by the sum of GFP+ and human BCMA+ cells as a proportion of murine CD45-negative population. Data represent mean ± SD

Fig. 4

Inhibition of BMP signalling reduces osteoclast number in vivo and in vitro. a Osteoclasts (indicated with arrows) from myeloma-bearing mice: vehicle vs. LDN193189 treated and vehicle vs. BMPR1a-FC treated (TRAP and haematoxylin-stained). Scale bar = 100 μm. b, c Histomorphometric analysis of TRAP-stained osteoclast bone surface density (N.Oc.B.Pm) and osteoclast surface/ trabecular surface ratio (OcS/BS) in healthy and myeloma-bearing mice. n = 5 per control group, 10 per myeloma group and total n = 30. Missing data values: see Methods. b Vehicle vs. LDN193189 treated. c Vehicle vs. BMPR1a-FC treated. d Serum TRAP-5b concentration, indicating global osteoclast burden, in vehicle vs. LDN193189-treated and vehicle vs. BMPR1a-FC treated healthy and myeloma-bearing mice. e RNA-Seq-generated in vivo Tnfsf11:Tnfrsf11b (RANKL to OPG) expression in osteoblasts. Ratio in healthy vs. myeloma-bearing mice and vehicle vs. LDN193189-treated myeloma-bearing mice. f Pooled human CD14+ PBMC osteoclast differentiation, with treatments as indicated commenced 24 h after seeding, and TRAP-stained at day 10. Images were taken at ×70 magnification. g Quantified osteoclasts (≥3 nuclei) when CD14+ PBMCs treated as f were enumerated. Results from same experimental replicate are paired to account for inter-replicate variation (n = 5 replicates, paired t test). *P < 0.05, **p < 0.01, ***p < 0.001, NS, not significant, t test. Data represent mean ± SD

Fig. 5

BMP inhibition promotes osteoblast differentiation in myeloma in vivo via down-regulation of Wnt inhibitors. a Mineralisation of murine pre-osteoblast cell line 2T3 in the presence of LDN193189 or BMPR1a-FC. b–e Histomorphometric analysis of haematoxylin- and TRAP-stained femur sections from healthy and myeloma-bearing mice: n = 5 per control group, 10 per myeloma group, total n = 30. Missing data points: see Methods. b Bone-lining cell density (N.Ob/T.Ar, per mm³) in mice treated with vehicle vs. LDN193189 and vehicle vs. BMPR1a-FC. c Mineral apposition rate (μm per day) and bone formation rate (μm³/μm² per day) in healthy and myeloma-bearing mice treated with vehicle vs. LDN193189. d Images of bone-lining cell layer in healthy mice treated with vehicle, LDN193189 or BMPR1a-FC. e Bone-lining cell layer thickness (μm) in healthy mice treated with vehicle vs. BMPR1a-FC and vehicle vs. LDN193189. f Images of bone-lining cell layer in healthy mice treated with vehicle vs. BMPR1a-Fc, stained for osterix and vitamin D receptor (VDR). Blue = Hoescht, green = osterix, red = VDR, yellow/orange = expression of osterix and VDR. Scale bar = 100 μm. g Gene sets enriched (FDR <0.25, n > 25 genes) on GSEA of stromal progenitors from LDN193189 vs. vehicle-treated myeloma-bearing mice. h Effect of siRNA double knockdown of Acvr1 and Bmpr1a (rat osteosarcoma line UMR-106.01) on Acvr1, Bmpr1a, Smad6, Dkk1 and Sost (sclerostin) gene expression (n = 4 replicates). Effect of incubation with LDN193189 (100 nM, 24 h) on expression of same genes was performed for comparison (n = 6 replicates). i Expression of Polr2a (control) and BMP-response genes Smad7, Sma6, Id1, Id3, Atoh8, and Dkk1 by myeloma patients’ bone marrow stromal cells after 24 h incubation with LDN193189 (100 nM) (n = 6 samples) or BMPR1a-FC (10 μg/ml) (n = 4 samples). Sost not detectable. j Bone marrow plasma sclerostin level in vehicle vs. BMPR1a-FC-treated and vehicle vs. LDN193189-treated myeloma-bearing mice. *P < 0.05, **p < 0.01, ***p < 0.001, NS, not significant. b–e, j t Test, h one-way ANOVA with Dunnett’s multiple comparison test and i paired ratio t test. Data represent mean ± SD

Fig. 6

Mechanistic model of the involvement of BMP signalling inhibition in the restoration of bone mass in multiple myeloma. a BMP signalling in adult bone homeostasis couples osteoblast and osteoclast development. By controlling expression of Wnt inhibitors, BMP signalling also prevents excess matrix deposition. Wnt inhibitors are negative regulators of osteoblast development and bone-lining cell reactivation. b Myeloma overrides the balance between osteoclast and osteoblast leading to excess bone resorption without repair. High Wnt inhibitor levels, induced by myeloma-induced osteocyte toxicity and apoptosis, are critical to this uncoupling by preventing osteoblast development and bone-lining cell activation. c Inhibition of BMP signalling reverses the excess Wnt inhibitor levels and osteoclast recruitment, thereby allowing bone-lining cells to reactivate to matrix-secreting osteoblasts to restore damage. Figure produced using Servier Medical Art (http://www.servier.com, licensed under a creative commons attribution 3.0 unported licence)