How myeloma promotes bone loss

Multiple myeloma can lead to bone loss by reducing the differentiation of mesenchymal stem cells (MSCs) into osteoblasts. Using a combination of single-cell RNA sequencing, in vitro coculture, and experiments with human myeloma cells and MSCs in mice, Liu et al. demonstrated how direct contact between myeloma cells and MSCs shifted the balance of MSC differentiation to favor adipogenesis over osteoblastogenesis. Integrin 4 on the surface of myeloma cells activated the adhesion molecule VCAM1 on MSCs, leading to protein kinase C 1 (PKC1)dependent repression of the E3 ubiquitin ligase MURF1 and subsequent stabilization of the adipocyte transcription factor PPAR2. These findings suggest a possible avenue for preventing or treating myeloma-induced bone loss in patients.

The suppression of bone formation is a hallmark of multiple myeloma. Myeloma cells inhibit osteoblastogenesis from mesenchymal stem cells (MSCs), which can also differentiate into adipocytes. We investigated myeloma-MSC interactions and the effects of such interactions on the differentiation of MSCs into adipocytes or osteoblasts using single-cell RNA sequencing, in vitro coculture, and subcutaneous injection of MSCs and myeloma cells into mice. Our results revealed that the 4 integrin subunit on myeloma cells stimulated vascular cell adhesion molecule1 (VCAM1) on MSCs, leading to the activation of protein kinase C 1 (PKC1) signaling and repression of the muscle ring-finger protein-1 (MURF1)mediated ubiquitylation of peroxisome proliferatoractivated receptor 2 (PPAR2). Stabilized PPAR2 proteins enhanced adipogenesis and consequently reduced osteoblastogenesis from MSCs, thus suppressing bone formation in vitro and in vivo. These findings reveal that suppressed bone formation is a direct consequence of myeloma-MSC contact that promotes the differentiation of MSCs into adipocytes at the expense of osteoblasts. Thus, this study provides a potential strategy for treating bone resorption in patients with myeloma by counteracting tumor-MSC interactions.

More than 80% of patients with multiple myeloma suffer from bone destruction, which greatly reduces their quality of life and has a severe negative impact on survival (1). New bone formation, which usually occurs at sites of previously resorbed bone, is strongly suppressed in patients with myeloma, and bone destruction rarely heals in these patients (2). Therefore, prevention of bone disease is a priority in myeloma treatment, and understanding the mechanisms by which myeloma cells disturb the bone marrow (BM) is fundamental to myeloma-associated bone diseases.

Osteoblasts originate from mesenchymal stem cells (MSCs) and are responsible for bone formation. It has been reported that myeloma cells inhibit MSC differentiation into mature osteoblasts (35). Osteoblasts and adipocytes arise from a common MSC-derived progenitor and exhibit lineage plasticity, which further complicates the relationship between these two cell types in myeloma cellinfiltrated BM (6). Traditionally, initiation of adipogenesis and osteogenesis has been widely regarded as mutually exclusive, and factors that inhibit osteoblastogenesis activate adipogenesis and vice versa (7). Previous studies have demonstrated that MSCs differentiate into either adipocytes or osteoblasts depending on the stimulator (8), and adipocytes transdifferentiate into osteoblasts in patients with several benign diseases (9). However, the underlying effects of myeloma cells on the activation of adipogenic transcriptional factors and the molecular mechanisms involved are still obscure.

Peroxisome proliferatoractivated receptor 2 (PPAR2) is a key transcription factor for the regulation of fatty acid storage and glucose metabolism (10), and it activates genes important for adipocyte differentiation and function (11). Previous findings have demonstrated that PPAR2 plays important roles not only in the activation of adipogenesis but also in the suppression of osteoblastogenesis (12, 13). In vitro coculture of MSCs from multiple myeloma patients with malignant plasma cell lines enhances adipocyte differentiation of the MSCs due to increased PPAR2 in the MSCs (14), suggesting that PPAR2 mediates myeloma-induced adipogenesis. However, the mechanism by which myeloma cells activate PPAR2 in MSCs, thereby causing MSCs to differentiate into adipocytes rather than osteoblasts, remains unclear.

In the present study, we demonstrated that myeloma cells enhanced the differentiation of human MSCs into adipocytes rather than osteoblasts by stabilizing PPAR2 protein through an integrin 4protein kinase C 1 (PKC1)muscle ring-finger protein-1 (MURF1) signaling pathway in MSCs. Our study thus provides a potential therapeutic strategy for myeloma-associated bone disease.

To determine whether myeloma cells affect MSC fate, we characterized the heterogeneity of human BMderived MSCs after exposure to myeloma cells. We cultured MSCs alone (controls) or cocultured them with myeloma cells in a 1:1 mixture of adipocyte:osteoblast (1:1 AD:OB) medium (Fig. 1A). An aliquot of cells was cultured for 48 hours and then subjected to single-cell RNA sequencing (scRNA-seq). We cultured another aliquot of cells for 2 weeks, removed the myeloma cells, and assessed the ability of the MSCs to differentiate into mature osteoblasts or adipocytes using Alizarin red-S, which stains calcium deposits, and Oil red O, which stains lipids (Fig. 1A). Trajectory analysis indicated the dynamic cellular transition processes of MSCs in vitro, in line with the in vivo MSC fates, reported by Wolock et al. (15). We observed a fate shift in MSC differentiation when MSCs were cocultured with myeloma cells (Fig. 1B). T-distributed stochastic neighbor embedding cluster analysis based on the entire transcriptome gene signature showed that both control and cocultured MSCs had specific transcriptome characteristics (Fig. 1C). After identification of genes with highly variable expression across the dataset, clusters were identified in each of the control and coculture groups (Fig. 1C). Enrichment analysis demonstrated that the adipokine signaling pathway and the mineral absorption pathway were among the 20 pathways most significantly changed in MSCs cocultured with myeloma cells (Fig. 1D). We identified clusters 0, 1, 6, and 8 in the MSCs cocultured with myeloma cells as being of adipogenic lineage because their expression of the specific markers of adipogenesis, the ADD1 and PPAR genes, were markedly higher than that of other clusters (Fig. 1E). These results demonstrated that myeloma cells at least partially increase MSC transformation into adipocytes.

(A) System for coculturing of human MSCs with the human multiple myeloma cell (MM) line MM.1S in a 1:1 mixture of adipocyte (AD) and osteoblast (OB) medium. Cells were cocultured for 48 hours and then MSC-derived cells were subjected to single-cell RNA sequencing (scRNA-seq). As a control, scRNA-seq was also performed on MSCs cultured alone in 1:1 AD:OB medium. (B) The single-cell trajectory reconstructed by Monocle in the control (Ctrl) and coculture (Coculture) groups. Each point represents a cell, and colors indicate their respective group. n = 2 independent experiments. The trajectory constructed by Monocle is in black. (C) T-distributed stochastic neighbor embedding (t-SNE) plot depicting clusters of MSCs cultured alone (Ctrl) or cocultured with MM cells. The first two dimensions are shown. Each cluster represents individual cells with similar transcriptional profiles of MSCs or different MSC lineages, with total of 10 clusters from aggregated samples of two biologically independent experiments. (D) Enrichment analysis showing the 20 most significantly changed pathways in the MSCs cocultured with MM cells. Red indicates activated pathways, and green indicates repressed pathways. (E) Distributions of unique transcripts per cell and PPARG and CEBPB gene expression in all cell clusters. The red frame shows the highest expression among the clusters. TGF-, transforming growth factor.

The coculture of MSCs and myeloma cells resulted in lower Alizarin red-S staining and higher Oil red O staining in MSCs, indicating an increase in the generation of adipocytes, compared to culture of MSCs alone (Fig. 2A). We further labeled cocultured MSCs with antibodies recognizing the osteoblast marker osteocalcin or the adipocyte marker fatty acid binding protein 4 (FABP4) and analyzed them using flow cytometry. We observed that culturing MSCs in osteoblast medium increased the osteocalcin+ population and that coculturing MSCs with myeloma cells inhibited this increase. Also, culturing MSCs in adipocyte medium increased the FABP4+ population, and coculturing them with myeloma cells further increased it. When we cultured MSCs alone in the 1:1 AD:OB medium, both the osteocalcin+ and FABP4+ populations increased, whereas coculturing MSCs with myeloma cells reduced the osteocalcin+ population but increased the FABP4+ population (Fig. 2, B and C). We obtained similar effects on osteoblastogenesis (Fig. 2D) and adipogenesis (Fig. 2E) when we cocultured MSCs with six other myeloma cell lines or with CD138+ primary myeloma cells isolated from BM aspirates from five patients with myeloma, but not with plasma cells from healthy donors (Fig. 2, F and G). Real-time polymerase chain reaction (PCR) analysis further showed lower expression of the osteoblast differentiationassociated genes alkaline phosphatase (ALP), secreted phosphoprotein 1 (SPP1), collagen type I alpha 1 chain (COL1A1), and bone gamma-carboxyglutamate protein (BGLAP; Fig. 2H) and higher expression of the adipocyte differentiationassociated genes delta-like noncanonical Notch ligand 1 (DLK1), diacylglycerol O-acyltransferase 1 (DGAT1), FABP4, and fatty acid synthase (FASN; Fig. 2I) in MSCs cocultured with ARP-1 or MM.1S myeloma cells than in MSCs cultured alone. These results demonstrate that myeloma cells directed the differentiation of MSCs preferentially toward adipocytes than to osteoblasts.

(A) Representative images of Alizarin red-S and Oil red O staining (whole wells and enlarged views) of MSCs cultured alone or cocultured with ARP-1 or MM.1S myeloma cell lines in MSC medium, adipocyte (AD) medium, osteoblast (OB) medium, or mixed 1:1 AD:OB medium as indicated. n = 3 independent experiments. Scale bars, 5 mm (whole wells) and 20 m (enlargements). (B and C) Flow cytometric analysis showing the percentage of osteocalcin+ (B) and FABP4+ (C) cells in cultures of MSCs alone or in direct contact with ARP-1 cells in the indicated medium. Data are representative of three independent experiments with each sample analyzed in triplicate. (D and E) Quantification of Alizarin red-S (D) and Oil red O (E) staining of MSCs cultured alone (No MM) or cocultured with the six indicated myeloma cell lines. Combined data are from three biologically independent experiments. (F and G) Quantification of Alizarin red-S (F) and Oil red O (G) staining of MSCs cultured alone or cocultured with primary myeloma cells isolated from BM aspirates of five patients with myeloma (P1 to P5) or normal plasma cells from the BM of two healthy donors (PC1 and PC2). Combined data are from n = 3 experiments using the same donor source material. (H and I) Quantitative reverse transcription PCR showing the expression of the osteoblast differentiationassociated genes ALP, SPP1, COL1A1, and BGLAP (H) and the adipocyte differentiationassociated genes DLK1, DGAT1, FABP4, and FASN (I) in cells generated by coculture of MSCs with myeloma cells relative to expression of each gene in MSCs cultured alone. Combined data are from n = 3 independent experiments. All data are means SD. *P 0.05 and **P 0.01. P values were determined using one-way ANOVA with Tukeys multiple comparisons test.

We next investigated the mechanism of myeloma-induced shifting of MSCs from osteoblastogenesis to adipogenesis. We focused on PPAR2 because it is a key transcriptional factor for the activation of adipogenesis. scRNA-seq showed higher PPAR2 mRNA expression in MSCs cocultured with myeloma cells compared to MSCs cultured alone (Fig. 1E). Using the coculture system with MSCs and myeloma cells in a 1:1 mixture of adipocyte and osteoblast medium, we again observed the transformation of osteoblastogenesis into adipogenesis in MSCs cocultured with myeloma cells (Fig. 3A), as well as an increase in the abundance of PPAR2 in MSCs cultured with myeloma cells (Fig. 3B and fig. S1). To determine the importance of PPAR2 in MSC transformation, we added the PPAR2 antagonist G3335 to cocultures. G3335 inhibited the myeloma cellinduced increase in PPAR2 protein (Fig. 3B and fig. S1). Consistent with the Western blot results, G3335 treatment decreased Oil red O staining (Fig. 3C) and adipocyte gene expression (Fig. 3D) and increased Alizarin red-S staining (Fig. 3E) and osteoblast gene expression (Fig. 3F). These results suggest that PPAR2 mediated myeloma-induced MSC transformation into adipocytes.

(A) Representative images of Oil red O or Alizarin red-S staining of MSCs cultured alone or cocultured with ARP-1 or MM.1S myeloma cells in 1:1 OB:AD medium and treated with the PPAR2 antagonist G3335 as indicated. Scale bar, 5 mm. (B) Representative Western blot for PPAR2 in cells treated as in (A). Quantitation is presented in fig. S1. Actin is a loading control. (C to F) Quantitative analysis of Oil red O staining (C), adipocyte differentiationassociated gene expression (D), Alizarin red-S staining (E), and osteoblast differentiationassociated gene expression (F) in cells treated as in (A). Data are means SD from n = 3 independent experiments. *P 0.05 and **P 0.01. P values were determined using Students t test for paired samples (D and F) and one-way ANOVA with Tukeys multiple comparisons test (C and E).

To determine whether myeloma cells distort MSC transformation through myeloma-secreted soluble factors or cell-to-cell contact, we cocultured MSCs with ARP-1 or MM.1S myeloma cells in 1:1 AD:OB medium either together or separated by transwell inserts. We observed that the transwell coculture had a slight effect on increased Oil red O staining, whereas cell-to-cell contact coculture in the mixed medium produced much more significant boost of this staining, suggesting that direct interaction between MSCs and myeloma cells was needed for enhancing adipogenesis from MSCs (Fig. 4A). When we added supernatants collected from 24-hour cultures of ARP-1 or MM.1S cells to MSC cultures, we obtained results similar to those for the transwell coculture (Fig. 4A), reaffirming the importance of direct contact of MSCs with myeloma cells.

(A) Oil red O staining in MSCs cultured alone (No MM) or cocultured with ARP-1 or MM.1S myeloma cells in 1:1 AD:OB medium directly (cell-cell) or separated by transwell inserts (Trans) or in myeloma cell culture media (sup). Staining was quantified relative to the No MM condition. Representative data are from three independent experiments. (B to D) Relative Oil red O staining (B) and the relative expression of the indicated osteoblast (C) and adipocyte (D) marker genes in MSCs cultured alone (No MM) or cocultured with ARP-1 or MM.1S cells with or without neutralizing antibodies against integrin subunits 4, 5, V, or L. Combined data are from three independent experiments. (E) Western blot showing integrin 4 and integrin 1 in ARP-1 and MM.1S cells expressing shRNA targeting integrin 4 (4 KD) or nontargeted control shRNAs (NT Ctrl). Actin is a loading control. (Blot is a representative of three independent experiments, and blot quantitation data are presented in fig. S2C. (F to J) PPAR2 protein (F), Alizarin red-S staining (G), Oil red O staining (H), osteoblast marker gene expression (I), and adipocyte marker gene expression (J) in MSCs cultured alone or cocultured with ARP-1 or MM.1S cells expressing NT Ctrl or 4 KD shRNA. Blots in (E) and (F) are representative of three independent experiments, and blot quantitation is presented in fig. S2 (A and D). Data in (G) to (J) are means SD from n = 3 independent experiments using MSCs derived from BM aspirates of three healthy donors. Data are **P 0.01. P values were determined using one-way ANOVA with Tukeys multiple comparisons test.

To identify the specific molecules involved in adipocyte differentiation, we tested the effect of blocking antibodies against various integrins, which are highly expressed in myeloma cells, in cocultures of MSCs with ARP-1 or MM.1S cells in 1:1 AD:OB medium. The addition of an antibody against integrin 4but not antibodies against integrins 5, V, or L or a control immunoglobulin G (IgG)markedly reduced Oil red O staining in cocultures with both myeloma cell lines (Fig. 4B). The addition of the antibody recognizing integrin 4 to cocultures of MSCs and ARP-1 cells in the mixed medium also increased osteoblast gene expression (Fig. 4C) and decreased adipocyte gene expression (Fig. 4D) substantially more than did the addition of the control IgG. To determine whether integrin 4 affected PPAR2 production in MSCs, we infected ARP-1 and MM.1S cells with a lentivirus carrying short hairpin RNAs (shRNAs) targeting integrin 4 (fig. S2A). Integrin 4 knockdown (4 KD) reduced integrin 4 production without changing the cell viability or proliferation, whereas integrin 1 remained unchanged in ARP-1 and MM.1S cells (Fig. 4E and fig. S2, A to C). We also cocultured MSCs with control or 4 KD myeloma cells in the mixed medium. Western blot analysis demonstrated that 4 KD in myeloma cells reduced PPAR2 protein production in MSCs more than did myeloma cells expressing a nontargeting control shRNA (Fig. 4F and fig. S2D). In addition, coculture of MSCs with 4 KD myeloma cells induced higher Alizarin red-S staining (Fig. 4G) and osteoblast gene expression (Fig. 4H) but lower Oil red O staining (Fig. 4I) and adipocyte gene expression (Fig. 4J) compared to MSCs cocultured with myeloma cells expressing the control shRNA.

Because vascular cell adhesion molecule1 (VCAM1) is a major ligand of integrin 4, we investigated whether it mediated myeloma-induced MSC transformation by adding a blocking antibody against VCAM1 or control IgG to MSC and myeloma cell cocultures. Addition of the antibody, but not IgG, increased Alizarin red-S staining (Fig. 5A) and osteoblast gene expression (Fig. 5B) but decreased Oil red O staining (Fig. 5C) and adipocyte gene expression (Fig. 5D) in MSCs. To determine whether binding of integrin 4 to VCAM1 induced an increase in PPAR2, we constructed MSCs with reduced expression of VCAM1 using a lentivirus carrying VCAM1 shRNAs (VCAM1 KD) (Fig. 5E and fig. S3A) and cocultured myeloma cells with control or VCAM1 KD MSCs. Western blot analysis showed that cocultured VCAM1 KD MSCs had reduced PPAR2 protein production compared to cocultured MSCs expressing nontargeting control shRNA (Fig. 5F and fig. S3B). We also found that VCAM1 KD in MSCs considerably abrogated myeloma-induced suppression of osteoblastogenesis and activation of adipogenesis, because Oil red O staining and adipocyte gene expression decreased significantly (Fig. 5, G and H), whereas Alizarin red-S staining and osteoblast gene expression both increased (Fig. 5, I and J).

(A to D) Alizarin red-S staining (A), Oil red O staining (B), and real-time PCR analysis of the expression of osteoblast (C) and adipocyte (D) marker genes in MSCs cultured alone (No MM) or cocultured with ARP-1 or MM.1S myeloma cells in the presence of a neutralizing antibody against VCAM1 or IgG (control). Data are from n = 3 independent experiments. (E) Western blotting analysis showing VCAM1 in the MSCs infected with a lentivirus carrying nontargeted control shRNAs (NT Ctrl-MSCs) or human VCAM1 shRNAs (VCAM1 KD-MSCs). Actin is a loading control. Blot is a representative of three independent experiments, and blot quantitation is presented in fig. S3A. (F to J) PPAR2 protein (F), adipocyte gene expression (G), Oil red O staining (H), Alizarin red-S staining (I), and osteoblast gene expression (J) in MSCs expressing NT Ctrl or VCAM1 shRNAs cocultured with ARP-1 or MM.1S cells in 1:1 OB:AD medium. Blot in (F) is a representative of three independent experiments, and blot quantitation is presented in fig. S3B. Data are means SD from n = 3 independent experiments. *P 0.05 and **P 0.01. P values were determined using one-way ANOVA with Tukeys multiple comparisons test except in (G) and (J), where Students t test for paired samples were used.

Because VCAM1 stimulates intracellular signaling that results in the activation of protein kinase C (PKC), we examined PKC activation in cocultures. Coculture of myeloma cells and MSCs enhanced the phosphorylation of PKC1 but did not affect phosphorylation of the PKC isoforms PKC, PKC, or PKC/ or the abundance of total PKC and reduced the phosphorylation of PKC and PKC (Fig. 6, A and B). Addition of the PKC inhibitor Go6976 to the cocultures markedly reduced PKC1 phosphorylation and PPAR2 protein in MSC cells cocultured with ARP-1 or MM.1S cells (Fig. 6C and fig. S4). Functionally, treatment of cocultures with Go6976 reduced Oil red O staining and increased Alizarin red-S staining (Fig. 6, D to F). Together, these results demonstrate that myeloma cells activated PPAR2 in MSCs and induced MSC differentiation into adipocytes rather than osteoblasts through the integrin 4-VCAM1-PKC1 pathway.

(A) Western blotting for all phosphorylated PKCs (p-PKC pan), the indicated phosphorylated PKC isoforms, and total PKC in MSCs cultured alone or cocultured with ARP-1 or MM.1S myeloma cells. The abundances of total PKC served as protein loading controls. (B) Quantification of the phosphorylation of PKC isoforms in MSCs cocultured with myeloma cells in (A) relative to the MSC-only control. The cutoff values are fold change more than twofold or less than 0.5-fold. (C) Western blotting for phosphorylated PKC1, total PKC, and PPAR2 in MSCs cocultured with ARP-1 or MM.1S cells in the presence of the PKC inhibitor Go6976 or DMSO (control). Actin is a loading control. Blot is a representative of three independent experiments, and blot quantitation is presented in fig. S4. (D) Representative images of Oil red O staining and Alizarin red-S staining of MSCs cultured alone or cocultured with ARP-1 or MM.1S myeloma cells in the presence of the PKC inhibitor Go6976 or DMSO (control). Scale bar, 5 mm. (E and F) Quantification of Oil red O staining (E) and Alizarin red-S staining (F), in cells treated as in (D). Data are means SD from n = 3 independent experiments. *P 0.05 and **P 0.01. P values were determined using one-way ANOVA with Tukeys multiple comparisons test.

Because a key mechanism of regulation of PPAR2 is its ubiquitylation-dependent proteasome-mediated degradation (16), we added the proteasome inhibitor MG132 to cultures of MSCs. We found that treatment with MG132 increased the presence of PPAR2 protein in MSCs in a time- and dose-dependent manner (Fig. 7A and fig. S5A). MG132 treatment causes the accumulation of ubiquitylated PPAR2 in MSCs, and coculturing these cells with myeloma cells reduced PPAR2 ubiquitylation (Fig. 7B and fig. S5B). However, the addition of a neutralizing antibody against VCAM1 to the cocultures restored ubiquitylation of PPAR2 (Fig. 7C and fig. S5C). These results suggested that myeloma cells activate PPAR2 in MSCs through inhibition of its ubiquitylation.

(A) Western blotting analysis for PPAR2 in MSCs cultured in 1:1 OB:AD medium and treated with the proteasome inhibitor MG132 for the indicated amounts of time. Actin is a loading control. (B) Immunoblotting (IB) for ubiquitin in PPAR2 immunoprecipitates (IP) from MSCs cultured alone or cocultured with ARP-1 or MM.1S myeloma cells in the presence of MG132. (C) Western blotting for ubiquitin in PPAR2 immunoprecipitates from MSCs cocultured with ARP-1 or MM.1S cells in the presence of MG132 and an antibody against VCAM1 or IgG (control). (D) Expression of the E3 ligaseencoding genes USP7, MURF1, MKRN1, CRBN, CRL4B, and TRIM23 in MSCs cocultured with myeloma cells relative to the expression in MSCs cultured alone (No MM). Data are means SD from n = 3 independent experiments. **P 0.01. P values were determined using one-way ANOVA with Tukeys multiple comparisons test. (E) Western blotting for USP7, MURF1, and MKRN1 in MSCs cultured alone or cocultured with myeloma cells. (F) Western blotting for MURF1 in MSCs cocultured with ARP-1 or MM.1S myeloma cells and treated with Go6976 or DMSO (control) as indicated. (G) Immunoblotting for MURF1 or PPAR2 in PPAR2 or MURF1 immunoprecipitates, respectively, from MSCs. IgG immunoprecipitates and whole-cell lysate (input) were used as controls. (H) Immunoblotting for ubiquitin in PPAR2 immunoprecipitates from MSCs expressing nontarget control (NT Ctrl) or MURF1 shRNAs in the presence of MG132. Each blot is representative of n = 3 independent experiments, and blot quantitation is presented in fig. S5.

To investigate the mechanism by which myeloma cells inhibited PPAR2 ubiquitylation, we examined the E3 ubiquitin ligases known to induce ubiquitylation of PPARs (17). Among the tested ligases, we found that MURF1 mRNA (Fig. 7D) and MURF1 protein (Fig. 7E and fig. S5D) were reduced in MSCs cocultured with myeloma cells. Addition of the PKC inhibitor Go6976 to the cocultures increased MURF1 protein in MSCs (Fig. 7F and fig. S5E), indicating that myeloma cells inhibited MURF1 production in MSCs through the PKC signaling pathway. Because the effects of MURF1 on PPAR2 ubiquitylation are unclear, we examined the interaction of these two proteins in MSCs. Co-immunoprecipitation of PPAR2 from MSCs demonstrated an interaction between MURF1 and PPAR2 (Fig. 7G), and knockdown of MURF1 in MSCs reduced the ubiquitylation of PPAR2 (Fig. 7H and fig. S5, F and G). These results demonstrate that myeloma cells activated PPAR2 in MSCs by reducing MURF1-mediated ubiquitylation of PPAR2.

To test the influence of myeloma cells on MSC differentiation in vivo, we established an extramedullary bone formation model in mice. Matrigel containing MSCs and Matrigel containing MSCs plus -irradiated ARP-1 cells were subcutaneously implanted into the right and left flanks of nonobese diabetic/severe combined immunodeficiency/interleukin-2rnull mice, respectively (Fig. 8A). Each sample also included human endothelial colony-forming cells (ECFCs) to stimulate blood vessel formation in the implant. In line with results of a previous study (18), we observed lower bone density in the extramedullary bones that formed in the left flanks, which were implanted with MSCs plus irradiated myeloma cells, compared to the extramedullary bones that formed on the right side, which were implanted with MSCs alone (Fig. 8A). Furthermore, we examined subcutaneous tissues on both sides of mice using histologic or immunohistochemical staining with antibodies against the mature osteoblast marker osteocalcin, the adipocyte marker perilipin, the myeloma marker CD138, and human MURF1. We observed lower numbers of new bones and osteocalcin+ osteoblasts and higher numbers of perilipin+ adipocytes in tissues on the sides of mice implanted with both MSCs and myeloma cells, reduction of MURF1 abundance in tissues on the sides of mice implanted with MSCs alone, and CD138+ cells only in tissues on the sides of mice implanted with myeloma cells (Fig. 8B).

(A) Representative images of subcutaneous tissues and bone density in mice implanted with human MSCs plus ECFCs in the right flank and MSCs plus ECFCs mixed with ARP-1 myeloma cells in the left flank. The arrows indicate bone formation in subcutaneous tissue, and the bars indicate bone density. (B) Representative hematoxylin and eosin (H&E) and immunohistochemical staining for the osteoblast marker osteocalcin, the adipocyte marker perilipin, the myeloma cell marker CD138+, and MURF1 of the subcutaneous tissues from (A). Scale bar, 20 m. Data represent n = 3 independent experiments with five mice each. (C) Expression of MURF1 in MSCs from BM aspirates from 12 patients with myeloma and 12 age-matched healthy donors relative to expression in a randomly selected sample from healthy donor. Data are from n = 3 experiments using the same donor source material. *P 0.05. P values were determined using Students t test. (D and E) Western blotting for MURF1 and PPAR2 (D) and Alizarin red-S and Oil red O staining (E) in MSCs from BM aspirates from three healthy donors and three patients with myeloma. Blots and images are representative of three experiments using the same donor materials, and blot quantitation is presented in fig. S6. Scale bars, 5 mm (whole wells) and 100 m (enlargements). (F) Quantitation of Alizarin red-S and Oil red O staining in the cultures of MSCs from BM aspirates from healthy donors and patients with myeloma in (C). Data are from n = 3 experiments using the same donor source material. P values were determined using Students t test. OD490, optical density at 490 nm.

We also isolated MSCs from the BM of 12 healthy human donors and 12 age-matched patients with myeloma and found markedly lower MURF1 mRNA expression in patient-derived MSCs compared to healthy donor MSCs (Fig. 8C). Western blotting validated the negative correlation between MURF1 and PPAR2 at the protein level in MSCs isolated from 3 of 12 samples in both groups (Fig. 8D and fig. S6). When we cultured these primary MSCs in 1:1 AD:OB medium, we found lower Alizarin red-S staining and higher Oil red O staining in cultures of patient-derived MSCs than in cultures of healthy donor MSCs (Fig. 8, E and F). These findings demonstrate that myeloma cells reduced MURF1 in MSCs and skewed MSC differentiation to favor adipogenesis, resulting in the suppression of osteoblast-mediated new bone formation in myeloma-bearing mice and in cells from patients with myeloma.

Using scRNA-seq, an in vitro coculture system, and mouse models, we demonstrated that myeloma cells shift the differentiation of MSCs into adipocytes rather than osteoblasts. Mechanistic studies revealed that integrin 4 on myeloma cells bound to VCAM1 on MSCs and inhibited ubiquitylation of PPAR2 through PKC-MURF1 signaling. The resulting increase in PPAR2 enhanced adipogenesis and suppressed osteoblastogenesis from MSCs. Thus, our study elucidates a previously unknown mechanism underlying myeloma-induced suppression of osteoblast-mediated bone formation and provides a potential approach for treating bone resorption in patients with myeloma.

Suppressed differentiation of osteoblasts is well known to be a key reason for bone loss and skeleton-related events in patients with myeloma (19). The molecules and pathways involved in myeloma-induced suppression of osteoblastogenesis include the Wnt signaling inhibitor Dickkopf-related protein 1 (DKK-1) (2, 20). However, antibody-mediated blocking of DKK-1 function cannot restore new bone formation completely or heal myeloma-induced resorbed bone, suggesting that additional factors expressed by myeloma cells critically affect bone formation. In the present study, we demonstrated that the 4 subunit of integrin, which is highly abundant in myeloma cells, promoted MSC differentiation into adipocytes, demonstrating that adhesion moleculesbut not soluble factorsproduced by myeloma cells primarily mediated the shift from osteoblastogenesis to adipogenesis. Integrin 41, also known as very late antigen-4, is a cell surface heterodimer present on malignant cells in patients with many types of cancer, including myeloma (21). It is a key adhesion molecule that acts as a receptor for the extracellular matrix protein fibronectin and the cellular receptor VCAM1. Interaction between integrin 41 and VCAM1 can activate mature osteoclast formation in patients with bone-metastatic breast cancer (22). In patients with multiple myeloma, this interaction promotes the secretion of interleukin-7 by tumor cells, which inhibits the expression of RUNX-2, which encodes a transcription factor that is essential for osteoblast differentiation, and RUNX-2 transcriptional regulatory activity in MSCs (23). This interaction also increases DKK-1 secretion by myeloma cells. Adding to these known mechanisms, we revealed that binding of myeloma cell integrin 41 to VCAM1 on the MSC surface activated the PKC signaling pathway. We also identified activation of PKC1, suppression of the downstream mediator MURF1, and the fundamental roles of such signaling pathways in the promotion of the MSC-derived adipocyte lineage. PKCs are also reportedly associated with Jagged-Notch signaling pathways, and they can regulate the transition of embryonic stem cells differentiating into postmitotic neurons (24). Some immunomodulatory drugs, such as lenalidomide, may affect osteoblast differentiation through this pathway (25), indicating the important role of Jagged-Notch in osteoblast differentiation from MSCs. We may further investigate their impacts and mechanisms on myeloma-induced the shift of MSC fates in our next studies.

BM adipocytes are recognized as important regulators of bone remodeling rather than just being inert filler cells (26, 27). Normal BM adipocytes have been shown to be reprogrammed by myeloma cells and gain the ability to resorb bone in myeloma patients in remission (13). Focusing on the determination of MSC fate in this study, we investigated the molecular mechanism underlying the shift from osteoblastogenesis to adipogenesis induced by myeloma cells. Lineage-tracing experiments have revealed that adipocytes can also originate from osterix-positive cells and are closely related to osteoblasts (28). Chan et al. (29) reported that BM adipocytes were derived from a progenitor cell that was also the progenitor for osteoblasts. In addition, Gao et al. (30) reported plasticity between BM adipocytes and osteoblasts and potential transdifferentiation and transformation between these two identities after initiating differentiation. Despite this new knowledge about the balance between osteoblastogenesis and adipogenesis, how myeloma cells regulate this balance and transformation of MSCs is still unclear.

scRNA-seq can identify subpopulations using the transcriptome to avoid the complicated isolation procedures after cell-cell contact culture (15). We found that MSCs could be naturally divided into two populations by transcriptomic data, and at least one cluster of MSCs cocultured with myeloma cells highly expressed adipocyte marker genes. Coculture of myeloma cells pushed MSC differentiation toward adipocytes rather than osteoblasts, resulting in the suppression of bone formation in the in vivo extramedullary bone assay. Because MSCs are pluripotent stem cells capable of differentiation into other cell types, such as chondrocytes and skeletal muscle cells (31), whether myeloma cells affect MSC transformation into these cell types instead of osteoblasts remains unclear. It is possible that the observed differentiation from MSC to adipocyte in the presence of myeloma cells might have been rather the result of a differentiation of MSCs into osteoblasts followed by a transdifferentiation from osteoblast into adipocyte. Further investigation is needed to address this possibility.

Like other transcription factors and coregulators, PPAR2 can undergo posttranslational modifications, such as phosphorylation, acetylation, and SUMOylation (32). Researchers have identified the key enzymes and target amino acid sites involved in these modifications, but modification of PPAR2 by ubiquitylation, especially that induced by myeloma cells, is still unclear. Many E3 ligases, such as MURF1 and makorin ring finger protein 1 (MKRN1), are reported to be regulators of ubiquitylation of PPAR proteins (17, 3335), whereas investigators have identified only polyubiquitylation at Lys184 and Lys185 (K184/185) mediated by MKRN1 (16). In the present study, we demonstrated that the E3 ligase MURF1 contributed to PPAR2 ubiquitylation, and inhibition of MURF1 by myeloma cells reduced PPAR2 ubiquitylation, leading to enhanced protein stability in MSCs. MURF1 contains a canonical N-terminal RING-containing E3 ligase that is required for its ubiquitin ligase activity (36). Others have reported dysregulation of MURF1 in experimental models of fasting, diabetes, cancer, denervation, and immobilization (37). However, none have reported the substrate proteins, such as PPAR2, that are targeted for proteasomal degradation by MURF1 in patients with myeloma bone disease. Although the amino acids in PPAR2 that MURF1 targets remain to be identified, we demonstrated that the reduced MURF1 production in MSCs induced by myeloma cells was critical for the inhibition of PPAR2 ubiquitylation and thus stabilization of the PPAR2 protein. Other posttranslational modifications may also regulate PPAR2 protein, especially SUMOylation, which was not addressed in the current study. For example, the transcriptional activity of PPAR2 can be inhibited by SUMOylation at Lys107 to regulate insulin sensitivity (38), and growth differentiation factor 11 promotes osteoblastogenesis through enhancement of PPAR2 SUMOylation (39). A logical next step could be the investigation of the role of SUMOylation in myeloma-induced MSC transformation and how it interplays with the mechanisms described here.

In summary, our results shed light on the cross-talk between myeloma cells and MSCs and the impact of this interaction on the determination of the MSC-derived adipocyte lineage and the suppression of osteoblastogenesis from MSCs. Myeloma cell integrin 4 promoted phosphorylation of PKC1 through VCAM1, and the activated PKC1 reduced the production of MURF1 in MSCs, leading to reduced PPAR2 ubiquitylation. Therefore, counteracting 4-VCAM1-MURF1mediated adipogenesis from MSCs may be a promising strategy to heal myeloma-induced bone resorption.

Myeloma cell lines ARP-1 and ARK were provided by University of Arkansas for Medical Sciences (Little Rock, AR, USA), and others were purchased from American Type Culture Collection. Primary myeloma cells or normal plasma cells were isolated from the BM aspirates of patients with myeloma or healthy donors using antibody-coated magnetic beads against CD138, respectively (Miltenyi Biotec Inc.) (40). The cells were maintained in RPMI 1640 medium with 10% fetal bovine serum (FBS). MSCs from BM of healthy donors or patients with myeloma were maintained and augmented in Dulbeccos modified Eagles medium (DMEM) with 10% FBS (13). Information of healthy donors and patients were listed in table S1. The study was approved by the Institutional Review Board at The University of Texas MD Anderson Cancer Center.

Human MSCs were generated from BM mononuclear cells from fetal bones of healthy human donors, characterized using flow cytometry, and labeled with antibodies against MSC markers (CD44, CD90, and CD166) (41). Mature adipocytes were generated from MSCs using an adipocyte medium, which was formulated of DMEM medium with 10% FBS, 1 M dexamethasone, 0.2 mM indomethacin, insulin (10 g/ml), and 0.5 mM 3-isobutyl-l-methylxanthine (41). Mature adipocytes were fixed with 4% paraformaldehyde, stained with Oil red O for 1 hour, and observed under a light microscope. Mature osteoblasts were generated from MSCs using an osteoblast medium, which was formulated of alpha MEM medium with 10% FBS, 100 nM dexamethasone, 10 mM -glycerophosphate, and 0.05 mM l-ascorbic acid 2-phosphate (42). The bone-forming activity of osteoblasts was determined using Alizarin red-S staining (43, 44). Human MSCs were cultured alone or cocultured with myeloma cells at a ratio of 5:1 in MSC medium, osteoblast medium, adipocyte medium, or 1:1 mixed of osteoblast and adipocyte medium with or without inhibitors (G3335 or Go6976) or neutralizing antibodies for 2 weeks. Addition of dimethyl sulfoxide (DMSO) served as vehicle control for inhibitor-treatment experiments, and addition of IgG served as control for antibody-neutralizing experiments. In the transwell nondirect contact model, adipocytes were seeded onto the bottom of culture wells and cocultured with the myeloma cells on the insert. In direct contact coculture system, MSCs were seeded together with the myeloma cells in the culture wells to allow direct cell-cell contact. Supernatants collected from 24-hour cultures of myeloma cells were added to the MSCs in mixed osteoblast and adipocyte medium at a ratio of 1:5. In the experiments with primary cells, MSCs were cultured in the mixed medium for a week (45) and then cocultured with primary myeloma cells isolated from BM aspirates from patients with myeloma or normal plasma cells from BM of healthy donors for another week. Medium, inhibitors, and antibodies were refreshed every 3 days. After culture, the myeloma cells were removed, and the residual cells were stained with Alizarin red-S to assess osteoblast differentiation and with Oil red O to assess adipocyte differentiation. Culture of MSCs alone served as a control.

Single-cell preparation, complementary DNA (cDNA) library synthesis, RNA sequencing, and data analysis were performed by Gene Denovo Inc. Briefly, 1 106 MSCs were plated for 6 hours, 5 106 myeloma cells were added to the MSCs directly, and the cells were cocultured in mixed culture media for 48 hours; control MSC cells were cultured alone at the same media and then mixed with myeloma cells at the same ratio just before preparation for analysis. After removal of dead cells, the cells in these groups were counted using a Countess II Automated Cell Counter, and the concentration was adjusted to 1000 cells/l. The single-cell suspensions were bar-coded labeled and reverse-transcribed into scRNA-seq library using the Chromium Single Cell 3 GEM, Library and Gel Bead Kit (10X Genomics). The cDNA libraries from two independent experiments were sequenced on the Illumina HiSeq X-Ten platform, and data were pooled for the analysis. Myeloma cells were excluded using CD138 markers. The raw scRNA-seq data were aligned, filtered, and normalized using Cell Ranger (10X Genomics) software (tables S2 to S6). The cell subpopulation was grouped by graph-based clustering based on the gene expression profile of each cells in Seurat (tables S7 and S8). Subsequent data analysis including standardization, cell subpopulation, difference of gene expression, and marker gene screening were achieved by Seurat software.

MSCs were cultured alone or cocultured with myeloma cells with or without G3335 or neutralizing antibodies for 48 hours. In some experiments, MG132 was added to the cultures 6 hours before the cell collection. Addition of DMSO served as vehicle control for inhibitor experiment; addition of IgG served as neutralizing antibody control.

Quantitative real-time PCR was performed as described (46). The primers are listed in table S9. For Western blotting, cells were lysed with 1 lysis buffer (Cell Signaling Technology), subjected to 4 to 20% gradient gel electrophoresis, transferred to, and immunoblotted with antibodies against integrin 4 (R&D Systems), integrin 1, VCAM1, PKC, MURF1, and phosphorylated isoforms of PKC along with p-PKC-pan (Cell Signaling Technology) and PPAR2 (Santa Cruz Biotechnology). The membrane was stripped and reprobed with an antibody against -actin to ensure equal protein loading, and last, signals were detected using peroxidase-conjugated secondary antibody followed by enhanced chemiluminescence system (Millipore) in the MiniChem system (Saizhi Biotech), and quantitative analysis of blots were performed using the Fiji-based ImageJ software (version 1.51n, National Institutes of Health, Bethesda, MA, USA).

Viral particles were produced by human embryonic kidney 293T cells transfected with PMD2G and PSPAX2 packaging plasmids (Addgene) together with lentivirus-expressing shRNA vectors targeting 4, MURF1, or VCAM1 (Sigma-Aldrich). Nontargeted shRNA control (Sigma-Aldrich) was used as control. Sequences for knocking down specific genes are the following: 4, 5-CCGGGCTCCGTGTTATCAAGATTATCTCGAGATAATCTTGATAACACGGAGCTTTTT-3; VCAM1, 5-CCGGGGAATTAATTATCCAAGTTACCTCGAGGTAACTTGGATAATTAATTCCTTTTTTG-3; MURF1, 5-CCGGGAAGAGGAAGAGTCCACAGAACTCGAGTTCTGTGGACTCTTCCTCTTCTTTTTG-3 or 5-CCGGGTATAATAATGCCTGGTCATTCTCGAGAATGACCAGGCATTATTATACTTTTTG-3. Supernatants carrying the viral particles were harvested 48 hours later and concentrated to a 100 volume using polyethylene glycol 8000 (Sigma-Aldrich). MSCs (1 106 cells) were seeded 6 hours before the infection. Concentrated viral particles were added to MSCs or myeloma cells, respectively, in the presence of polybrene (8 g/ml) for 12 hours. The medium was then changed, and cells were cultured for another 48 hours until further management.

Cells were harvested and lysed using NP-40 lysis buffer supplemented with complete protease inhibitors, and the supernatant was precleaned with protein G beads (Thermo Fisher Scientific) and incubated with a mouse antibody against MURF1 (Santa Cruz Biotechnology) or monoclonal rabbit antibody against PPAR2 antibody (Santa Cruz Biotechnology) at 4C overnight with protein A/G agarose beads (Thermo Fisher Scientific). The next day, the pellet was washed four times with lysis buffer and then subjected to Western blot analysis using the antibodies against PPAR2 or MURF1. IgG was used as a control and total cell lysates (5%) were used as input controls.

For a ubiquitylation assay, diluted lysates were incubated with an antibody against PPAR2 at 4C overnight after precleaning with protein G beads (Thermo Fisher Scientific). Protein G beads were added to the washed lysate/antibody mixture at 4C for 4 hours. The resin was washed and applied to Western blot analysis using an antibody against ubiquitin.

MSCs were cultured alone or cocultured with myeloma cells for 2 weeks. Abundance of FABP4 and osteocalcin was assessed by immunofluorescence using fluorescein isothiocyanate or allophycocyanin-conjugated antibodies (BD Biosciences). After staining, cells were resuspended in phosphate-buffered saline with 1% FBS and analyzed using a BD LSR Fortessa flow cytometer.

The animal experiments in the present study were approved by the MD Anderson Institutional Animal Care and Use Committee. In vivo extramedullary bone formation in nonobese diabetic/severe combined immunodeficiency/interleukin-2rnull mice was established and examined (18). Briefly, MSCs alone or a mixture of human MSCs (1.5 106) and human ECFCs (1.5 106) in 0.2 ml of Matrigel (Corning Inc.) was subcutaneously injected into the right flanks of mice. This mixture and an additional 2 105 -irradiated (5000 centigrays) myeloma cells were injected into the left flanks of the mice. At 8 weeks after implantation, subcutaneous tissues were established, and the mice were intraperitoneally injected with OsteoSense 750 to assess new bone formation in those tissues. The subcutaneous tissues were collected after the mice were sacrificed and subjected to hematoxylin and eosin or immunohistochemical staining of cells labeled with an antibody against osteocalcin (a marker of mature osteoblasts), an antibody against perilipin (a marker of mature adipocytes), or an antibody against CD138 (a marker of myeloma cells).

The subcutaneous tissues were extracted from the mice and then formalin-fixed and paraffin-embedded. Tissue sections were deparaffinized with xylene and rehydrated to water through a graded alcohol series. Endogenous peroxidase activity was quenched with 3% hydrogen peroxide. The presence of CD138 (R&D Systems), osteocalcin, perilipin, and MURF1 (Abcam) in tissues was detected using specific antibodies. Signals were detected using secondary biotinylated antibodies and streptavidin/horseradish peroxidase. Chromagen 3,3-diaminobenzidine/H2O2 (Dako) was used, and slides were counterstained with hematoxylin. All slides were observed under a light microscope, and images were captured using a SPOT RT camera (Diagnostic Instruments).

Experimental values were expressed as means SD unless indicated otherwise. Statistical significance was analyzed using the GraphPad Prism v7.0 with two-tailed unpaired Students t tests for comparison of two groups and one-way analysis of variance (ANOVA) with Tukeys multiple comparisons test for comparison of more than two groups. P values less than 0.05 were considered statistically significant. All results were reproduced in at least three independent experiments.

stke.sciencemag.org/cgi/content/full/13/633/eaay8203/DC1

Fig. S1. G3335 inhibits PPAR2 accumulation in MSCs cocultured with myeloma cells.

Fig. S2. 4 KD in myeloma cells.

Fig. S3. VCAM1 knockdown in MSCs.

Fig. S4. PKC inhibition reduces PKC1 phosphorylation and PPAR2 abundance in MSCs cocultured with myeloma cells.

Fig. S5. Coculture with myeloma cells reduces ubiquitylation of PPAR2 in MSCs.

Fig. S6. MSCs from patients with myeloma show decreased MURF1 and increased PPAR2.

Table S1. Characteristics of patients with myeloma and healthy donors.

Table S2. Read quality control of the samples for scRNA-seq.

Table S3. Mapping quality control of aligned scRNA-seq data.

Table S4. Basic information of the aggregated samples for scRNA-seq before and after normalization.

Table S5. Information of each sample after aggregation.

Table S6. Cell quality control showing the cell numbers before and after the filtration.

Table S7. Number of cells in each subpopulation.

Table S8. Number of cells in each subpopulation of control and cocultured samples.

Table S9. Primers used in the quantitative reverse transcription PCR analysis.

Data file S1. scRNA-seq data from control sample.

Data file S2. scRNA-seq data from coculture sample.

Acknowledgments: We thank M. J. Li from Department of Genetics, Tianjin Medical University for the evaluation of our statistical analysis. Funding: This work was supported by R01 grants from NCI (CA190863 and CA193362 to J.Y.) and by the Research Scholar Grant from the American Cancer Society (127337-RSG-15-069-01-TBG to J.Y.). It was also supported by NIH/NCI (Core Labs at UT MD Anderson Cancer Center, P30CA016672) for the Small Animal Imaging and Research Histopathology Facilities. Author contributions: Z.L. and J.Y. designed all experiments and wrote the manuscript. H.L., Z.L., and J.H performed all experiments and statistical analysis. P.L. provided and interpreted patient samples. Q.T. provided critical suggestions. Conflict of interests: The authors declare that they have no competing interests. Data and materials availability: All of the data needed to evaluate the conclusions in the paper are provided in the main text or the Supplementary Materials. Stable cell lines carrying targeted shRNA are available with a materials transfer agreement between Houston Methodist Research Institute and the requesting institution.

View original post here:
Myeloma cells shift osteoblastogenesis to adipogenesis by inhibiting the ubiquitin ligase MURF1 in mesenchymal stem cells - Science