Fam20C is a kinase that generates the majority of secreted phosphoproteins and regulates biomineralization. However, its potential roles in bone resorption and breast cancer bone metastasis are unknown. Here we show that Fam20C in the myeloid lineage suppresses osteoclastogenesis and bone resorption, during which, osteopontin (OPN) is the most abundant phosphoprotein secreted in a Fam20C-dependent manner. OPN phosphorylation by Fam20C decreased OPN secretion, and OPN neutralization reduced Fam20C deficiency–induced osteoclast differentiation and bone metastasis. In contrast, Fam20C in breast cancer cells promoted bone metastasis by facilitating the phosphorylation and secretion of BMP4, which in turn enhanced osteoclastogenesis. Mutation of the BMP4 phosphorylation site elevated BMP4 lysosomal degradation and reduced BMP4 secretion. In breast cancer cells, BMP4 depletion or treatment with a BMP4 signaling inhibitor diminished osteoclast differentiation and bone metastasis and abolished Fam20C-mediated regulation of these processes. Collectively, this study discovers distinct roles for Fam20C in myeloid cells and breast cancer cells and highlights OPN and BMP4 as potential therapeutic targets for breast cancer bone metastasis.
Osteoclastogenesis and bone metastasis are suppressed by myeloid-derived Fam20C, but enhanced by breast cancer–associated Fam20C, uncovering novel Fam20C functions and new therapeutic strategies via targeting Fam20C substrates OPN and BMP4.
Bone undergoes continuous remodeling through bone formation by osteoblasts and bone resorption by osteoclasts. Bone is also the predominant site of metastasis for almost all breast cancer subtypes. Bone metastasis occurs in 15% to 30% in patients with breast cancer, severely impeding patient survival and normal bone functions (1). During bone metastasis, osteoclasts and cancer cells form a vicious cycle (2–5), where cancer cells produce factors that directly or indirectly induce osteoclastogenesis, and in turn, osteoclast-mediated bone resorption releases and activates growth factors that facilitate cancer cell seeding and proliferation in bone.
Phosphorylation is an important posttranslational modification for protein functions and signal transduction. Family with sequence similarity 20 member C (Fam20C, also named as DMP4 or GEF-CK) is a newly identified kinase that is responsible for phosphorylating casein and biomineralization proteins, such as small integrin-binding ligand N-linked glycoproteins (SIBLINGs), at the S-x-E/pS (Ser-x-Glu/phospho-Ser) motif (6, 7). Phosphoproteome analysis shows that Fam20C generates the majority of secreted phosphoproteins in liver, breast, and osteoblast-like cells (8). These Fam20C substrates participate in a broad range of biological processes including cell adhesion and migration (8).
Fam20C is highly expressed in lactating mammary gland (9) and mineralized tissues (10, 11). In human, Fam20C is required for normal bone development and its mutation causes a lethal osteosclerotic disorder, Raine syndrome (11). Fam20C deficiency in bone-forming osteoblasts reduces mineralization and bone mass (12, 13). The engagement of Fam20C in lactating mammary tissue and biomineralization provides a possible link for Fam20C in breast cancer bone metastasis. However, little is known about the roles of Fam20C and secreted phosphoproteins during bone resorption by osteoclasts and breast cancer metastasis to bone. Here we report that Fam20C plays a dual role in myeloid cells and breast cancer cells to regulate the phosphorylation and secretion of substrate proteins, thereby controlling osteoclastogenesis, bone resorption, bone mass and breast cancer bone metastasis.
Materials and Methods
All reagents and key resources are provided in Supplementary Material.
To establish conditional Fam20C knockout mice in the hematopoietic or myeloid lineage, Fam20C flox mice (14) were bred with Vav1-iCre (VavCre) mice (15) or lysozyme-Cre (LysMCre) mice (16), respectively. Athymic nude mice were purchased from Charles River. All protocols for mouse experiments were approved by the Institutional Animal Care and Use Committee of the University of Texas Southwestern Medical Center (Dallas, TX). Mice were housed on a 12–12 light cycle and provided food and water ad libitum.
Plasmids and antibodies
Fam20C cDNA with C-terminal FLAG-tag or Flag-KDEL sequence and cDNA of BMP4, IGFBP7, MFGE8, and osteopontin (OPN) with C-terminal V5/His tag were cloned into pcDNA4.0 vector (6) for transient expression, and pTy-U6 vector for stable expression. Mutants of murine BMP4 (Addgene) were constructed via overlap extension PCR. Anti-Fam20C (ref. 17; Thermo Fisher Scientific, 50–173–4886), anti-V5 (Thermo Fisher Scientific, R96025), anti-osteopontin (Santa Cruz Biotechnology, sc-21742), and anti-β-actin (Sigma, A5441) antibodies were used in immunoblotting. Normal goat IgG (R&D Systems, AB-108-C) and anti-osteopontin (R&D Systems, AF808) were used in in vivo treatment.
HEK293T cells, 4T1 cells, and MDA-MB-231 cells were purchased from the ATCC, certified to be free of Mycoplasma contamination. MDA-MB-231 subline (BoM-1833; ref. 18) was provided by Dr. Joan Massagué (Memorial Sloan Kettering Cancer Center, New York, NY). Py230 and Py8119 cells (19) were provided by Dr. Lesley Ellies (University of California, San Diego, La Jolla, CA). 4T1.2 cells (20) were provided by Drs. Robin Anderson (Peter MacCallum Cancer Centre, Melbourne, Victoria, Australia) and Yibin Kang (Princeton University, Princeton, NJ). Cells were tested as Mycoplasma-free using Mycoplasma PCR Detection Kit (Applied Biological Materials) and stored in liquid nitrogen. Cells were used for experiments within a month after thawing. Py230 and Py8119 cells were cultured in Ham F12K medium containing 5% FCS (Fetal Clone II), MITO Serum Extender and antibiotic–antimycotic; other cells were cultured in DMEM with 10% FBS and antibiotic–antimycotic, all at 37°C with 5% CO2. All reagents for cell culture were purchased from Thermo Fisher Scientific.
Preparation of stable cell lines
Single guide RNA (sgRNA) targeting human FAM20C in lentiCRISPR v2 vector (8) was cotransfected with psPAX and pMD2.G into HEK293T cells using FuGENE reagent (Thermo Fisher Scientific). To generate other stable cell lines, pTy-U6 or pLKO.1 plasmid was cotransfected with the expression plasmids for GAG-Pol-Rev, VSV-G, and pAdvantage into HEK293T cells using Lipofectamine 3000 (Thermo Fisher Scientific). Viruses were filtered and used for cell infection. Stably transfected cells were selected by puromycin.
Osteoclasts were differentiated as described previously (21, 22). Briefly, mouse bone marrow (BM) cells were filtered with a 40-μm cell strainer, and cultured with 40 ng/mL of mouse macrophage colony-stimulating factor (MCSF; R&D Systems) in α-MEM containing 10% FBS for 3 days (day 0 to 3), then with 100 ng/mL of mouse RANKL and 40 ng/mL of mouse MCSF (R&D Systems) for 3–4 days, in the presence or absence of 1 μmol/L rosiglitazone (Cayman Chemical). For OPN neutralization, 1 μg/mL anti-OPN antibody (R&D Systems) was added during the entire course of culture using normal goat IgG (R&D Systems) as control. Conditioned medium (CM; supplemented with 10% FBS) was added at 1:1 ratio to fresh culture medium. For BMP signaling inhibition, 500 nmol/L LDN-193189 (Selleckchem) was added. Osteoclast differentiation was quantified by the RNA expression of osteoclast marker genes using reverse-transcription quantitative qPCR on day 6, as well as TRAP staining of osteoclasts on day 7 when osteoclast formation starts in RANKL-treated group and is completed in RANKL/rosiglitazone-treated group. TRAP staining of osteoclasts was performed using the Leukocyte Acid Phosphatase Staining Kit (Sigma-Aldrich). Images of osteoclasts were taken with the NIS-Elements software (Nikon). Osteoclast number and size were quantified with ImageJ software.
Gene expression analyses
RNA was extracted with TRIzol reagent (Thermo Fisher Scientific) and reverse-transcribed using a High Capacity cDNA RT Kit (Fisher Scientific) and then analyzed using an Applied Biosystems 7700 real-time PCR instrument. All RNA expression was normalized by Rpl19 (mouse) or GAPDH (human). Proteins in CM were precipitated with trichloroacetic acid (Sigma-Aldrich) and washed with acetone (Thermo Fisher Scientific). Proteins in whole-cell extract or CM were analyzed with SDS-PAGE and immunoblotting. The intensity of protein bands was quantified with ImageJ software. For protein degradation analyses, cells were treated with 1 μmol/L MG132 (Sigma-Aldrich) or 100 μmol/L chloroquine (Thermo Fisher Scientific) for 24 hours before lysis. OPN levels in the culture medium were analyzed with an OPN ELISA kit (R&D Systems).
Mouse tibiae were isolated from 2-month-old male mice, fixed in 70% ethanol, and scanned for whole tibia (7-μm resolution), trabecular bone (3.5-μm resolution) and cortical bone (7-μm resolution) with micro-CT (μCT) using a Scanco μCT-35 instrument (SCANCO Medical) as described previously (23). Trabecular bone parameters were calculated using Scanco software to analyze the bone scans of the trabecular region directly distal to the proximal tibial growth plate. Mouse femurs were fixed in 4% paraformaldehyde for 24 hours, decalcified with 10% EDTA (pH 7.5) for 7–10 days, soaked in 30% sucrose in PBS for 24 hours, embedded in optimal cutting temperature compound (Thermo Fisher Scientific) and sectioned at 12–14 μm. TRAP staining of osteoclasts in femur sections was performed using the Leukocyte Acid Phosphatase staining kit. Images of bone sections were taken with the NIS-Elements software (Nikon). Histomorphometric analyses were conducted using the BIOQUANT Image Analysis software (Bioquant Version 14.1.6). Serum CTX-1 and P1NP were measured with a RatLaps EIA kit and a Rat/Mouse P1NP enzyme immunoassay kit (Immunodiagnostic Systems), respectively.
Mass spectrometry analyses of phosphopeptides
Phosphopeptide identification was performed as previously described (8) with minor modifications. Six 10-cm plates of BM cells from each genotype of three 2-month-old female littermate mice at a density of 2.8 × 107 cells/plate were cultured with 40 ng/mL of mouse MCSF for 3 days. Cells were cultured in serum-free medium (phenol red free) for 48 hours after being washed with PBS for four times and serum-free medium twice. CM was collected and spun down at 1,000 × g for 5 minutes to remove intact cells. The supernatant was centrifuged again at 10,000 × g for 20 minutes to remove debris. The resulting supernatant was added with 1 mmol/L Na3VO4, 1 mmol/L PMSF and 1x protease inhibitor cocktail (Sigma-Aldrich), and concentrated with Amicon Ultracel-3K units (EMD Millipore). Protein concentrations were measured with Pierce BCA protein assay (Thermo Fisher Scientific) and 600 μg of protein in each sample were subjected to the following procedures.
CM samples were mixed with an equal volume of 10% SDS in 100 mmol/L of triethylammonium bicarbonate (TEAB) buffer. Tris (2-carboxyethyl) phosphine (TCEP; 10 mmol/L) was added and the samples were incubated at 56°C for 30 minutes, followed by the addition of 20 mmol/L iodoacetamide and incubation at room temperature for 30 minutes in the dark. Solutions were acidified with 12% phosphoric acid, followed by the addition of 6 volumes of S-Trap binding buffer (90% methanol, 10% 1M TEAB), and then loaded onto an S-Trap column (Protifi). The columns were washed three times with binding buffer followed by the addition of trypsin (1:10) in 50 mmol/L TEAB and overnight incubation at 37°C. Peptides were recovered by eluting the S-Trap columns with 50 mmol/L TEAB, 0.2% formic acid (FA), and 50% ACN in 0.2% FA, sequentially. The eluate containing the peptides was dried and then cleaned using a 96 well Waters Oasis HLB solid-phase extraction plate.
Peptide samples were reconstituted with the binding buffer from the Thermo High-Select TiO2 Phosphopeptide Enrichment kit. The pH of the samples was verified to be less than 3 and samples were then loaded to TiO2 Spin Tips. The flow-through was collected for secondary enrichment with High-Select Fe-NTA Phosphopeptide enrichment columns (Thermo). The phosphopeptides collected from each enrichment step were combined, dried, and reconstituted in 2% (v/v) ACN, 0.1% trifluoroacetic acid in water. The unphosphorylated peptide fractions from each step were prepared the same way for analysis as well.
All samples were injected onto an Orbitrap Fusion Lumos mass spectrometer coupled to an Ultimate 3000 RSLC-Nano liquid chromatography system. Samples were injected onto an EasySpray column (Thermo) and eluted with a gradient from 0% to 28% buffer B. Buffer A contained 2% (v/v) ACN and 0.1% formic acid in water, and buffer B contained 80% (v/v) ACN, 10% (v/v) trifluoroethanol, and 0.1% formic acid in water. Mass spectrometer was operated in positive ion mode with a source voltage of 1.5 kV and an ion transfer tube temperature of 275°C. MS scans were acquired at a 120,000 resolution in the Orbitrap and up to 10 MS/MS spectra were obtained in the ion trap for each full spectrum acquired using higher energy collisional dissociation (HCD) for ions with charges 2–7. Dynamic exclusion was set for 25 seconds after an ion was selected for fragmentation.
Raw MS data files were analyzed using Proteome Discoverer v2.4 SP1 (Thermo Fisher Scientific), with peptide identification performed using Sequest HT searching against the mouse protein database from UniProt. Fragment and precursor tolerances of 10 ppm and 0.6 Da were specified, and three missed cleavages were allowed. Phosphorylation of Ser, Thr, and Tyr was set as variable modifications for the phosphopeptide fractions. The false-discovery rate (FDR) cutoff was 1% for all peptides. The raw data has been deposited to the Mass Spectrometry Interactive Virtual Environment (MassIVE) under the accession number MSV000086418.
Phosphoproteins identified from mass spectrometry were analyzed as reported (8). Briefly, they were analyzed with signal peptide predictors phobius (24) and SignalP 5.0 (25) to evaluate the potential for secretion. Protein sequences were also submitted to the transmembrane helix predictors phobius (24) and topcons (26) to obtain potential membrane topologies. The identified phosphorylation sites were mapped to membrane topologies, and the proteins with phosphorylation sites located at the predicted extracellular/lumen surface were considered as phosphorylated transmembrane proteins. The secreted or transmembrane proteins predicted by either method were further filtered with annotated subcellular localizations as secretion or transmembrane. The resulted candidates were determined as phosphoproteins in the secretory pathway.
The mRNA expression data and clinical data were obtained from Molecular Taxonomy of Breast Cancer International Consortium (METABRIC; refs. 27, 28). Patients with breast cancer were divided into two groups by the median expression of OPN. Patients with breast cancer with triple-negative status of ER, PR, and HER2 were divided into two groups by the median expression of Fam20C or its targets. The OS curves of these patients were analyzed in Prism software.
Analyses of primary tumor growth
Female nude mice (6–8 week old) were injected with 5 × 105 cells in 50 μL PBS with 50% Matrigel (Thermo Fisher Scientific) into the fourth mammary fat pad. Tumor size was measured every 3 days using electronic calipers. Tumor volume was calculated as (length×width2)/2.
Bone metastasis analyses
Under the guidance of a VisualSonics Vevo 770 small-animal ultrasound device, luciferase-labeled cancer cells were injected into the left cardiac ventricle of 6- to 8-week-old female mice so that they could bypass the lung and efficiently migrate to the bone (29). Bone metastases were detected and quantified weekly after injection by bioluminescent imaging (BLI) using a Caliper Xenogen IVIS Spectrum instrument. Mice with massive BLI signals (>1010) on chest only were excluded as failed intracardiac injections. Luciferase-labeled Py8119 cells were injected into C57BL/6J mice at 2.5×104 cells per mouse in 100 μL PBS. Luciferase-labeled human MDA231-BoM-1833 cells were injected into nude mice at 1×105 cells per mouse. For in vivo antibody treatment, 20 μg/mouse of anti-OPN neutralizing antibody or normal goat IgG (R&D Systems) was intraperitoneal injected every three days starting from one day before intracardiac injection of cancer cells. For in vivo inhibitor treatment, mice were treated with DMSO or 3 mg/kg of LDN-193189 by intraperitoneal injection daily starting on the day of tumor cell injection.
Cell proliferation and migration assay
Cells with a luciferase reporter were seeded into 96-well plates at a density of 2,000 cells per well, and cell proliferation was measured by testing luciferase activity on days 1, 3, 5, 7, and 9. Cancer cell migration was quantified using corning transwell plates (8-μm pore size, Thermo Fisher Scientific). Briefly, 5 × 105 luciferase-labeled cells in 100 μL serum-free medium were seeded in the top chamber and 600 μL medium containing 10% FBS was placed in the bottom chamber. After 24 hours, the relative number of cells that migrated to the bottom chamber was quantified by measuring luciferase activity.
Quantification and statistical analysis
We performed all statistical analyses in GraphPad Prism 7 software with data from three or more biologically independent replicates using two-sided Student t test or two-way ANOVA with post hoc Tukey multiple comparisons or Sidak multiple comparisons. Two-sided log-rank test was used to analyze the OS of patients. Western blot analyses were repeated at least twice. Block randomization was used for animal/sample allocation. Statistical parameters and tests are reported in the Figures and corresponding Figure Legends. Results are represented as mean ± SEM. The P values are *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001; NS, nonsignificant (P > 0.05).
Fam20C deletion in the hematopoietic or myeloid lineage enhances osteoclast differentiation and causes bone loss
Bone resorbing osteoclasts are differentiated from the myeloid lineage, which is derived from hematopoietic stem cells. To study the role of Fam20C during osteoclastogenesis, we generated mice with Fam20C conditional knockout in the hematopoietic stem cells by breeding Fam20C flox/flox mice with Vav1-iCre mice (Fam20C-VavCre). Osteoclast differentiation was examined by ex vivo culture of primary BM cells with RANKL induction and rosiglitazone enhancement (21). Complete deletion of Fam20C during osteoclastogenesis was achieved in Fam20C-VacCre mice (Supplementary Fig. S1A). qRT-PCR showed that Fam20C-VavCre cultures expressed higher levels of osteoclast markers than wild-type (WT) cells, such as tartrate-resistant acid phosphatase (TRAP), Cathepsin K (CTSK), Carbonic anhydrase 2 (CAR2), calcitonin receptor (Calcr), and Osteoclast-associated immunoglobulin-like receptor (Oscar; Fig. 1A). Fam20C-VavCre cultures also formed larger osteoclasts shown as multinucleated cells stained positive for the osteoclast marker TRAP (Fig. 1B; Supplementary Fig. S1B). Bone histomorphometric analyses showed increased osteoclast surface and number in the femurs of Fam20C-VavCre mice (Fig. 1C). Serum carboxy-terminal telopeptides of type I collagen (CTX-1), a bone resorption marker, was also higher in Fam20C-VavCre mice compared with WT littermate control mice (Fig. 1D). Serum procollagen type 1 N-terminal propeptide (P1NP), a bone formation marker, was similar in Fam20C-VavCre mice and WT mice (Supplementary Fig. S1C). These results indicate increased osteoclast differentiation and bone resorption in Fam20C-VavCre mice.
To further delineate the role of Fam20C in the myeloid lineage, we generated myeloid-specific Fam20C knockout mice by breeding Fam20C flox/flox mice with lysozyme-Cre mice (Fam20C-LysMCre). Fam20C-LysMCre osteoclast differentiation cultures had approximately 90% Fam20C deletion (Supplementary Fig. S1A), leading to higher expression of osteoclast markers such as TRAP, CTSK, CAR2, Calcr, and Oscar (Fig. 2A), and larger osteoclasts (Fig. 2B; Supplementary Fig. S1B). Histomorphometry showed that osteoclast surface was increased in Fam20C-LysMCre mice compared with littermate controls (Fig. 2C). However, serum CTX-1 was unaltered (Fig. 2D), which might be due to an incomplete Fam20C deletion by LysMCre than VavCre (Supplementary Fig. S1A). Serum P1NP levels were also similar (Supplementary Fig. S1C). Bone architecture of mouse tibia was analyzed by μCT imaging. Fam20C-LysMCre mice had less trabecular volume, trabecular number, connectivity density, bone mineral density (BMD) and bone surface, as well as more trabecular separation, structure model index (SMI), and trabecular porosity than WT mice (Fig. 2E and F). Osteoclast and macrophage are both differentiated from the myeloid lineage, our gene expression analyses showed that macrophage polarization was not affected by Fam20C-VavCre or Fam20C-LysMCre (Supplementary Fig. S1D and S1E). Together, these data indicate that Fam20C in the hematopoietic/myeloid lineages suppresses osteoclast differentiation.
Fam20C phosphorylates OPN and retards its secretion from osteoclast precursors
To identify the substrates that mediate the effects of Fam20C on osteoclastogenesis, we performed label-free phosphoprotein quantification by phosphopeptide enrichment and mass spectrometry with CM from WT and Fam20C-VavCre osteoclast precursors (3-day culture of BM cells with MCSF). We identified 136 Fam20C-dependent phosphoproteins that displayed more than 2-fold reduction of phosphorylation in Fam20C-VavCre osteoclast precursors (Supplementary Fig. S2A and S2B) or were only present in WT osteoclast precursors (Supplementary Table S1). We further restricted our analyses to phosphoproteins in the secretory pathway that are either secreted proteins or transmembrane proteins containing signal peptide or luminal/extracellular domain. We identified 32 Fam20C-dependent phosphopeptides representing 9 phosphoproteins that consisted of 83.5% abundance of all phosphoproteins in the secretory pathway (Fig. 3A; Supplementary Table S1). Among these phosphoproteins, OPN was the most abundant in WT with multiple phosphorylation sites, with at least 28-fold higher level than other Fam20C targets (Fig. 3A).
OPN is a Fam20C substrate (8, 30), and stimulates osteoclast differentiation and bone resorption (31, 32). Mass spectrometry analysis before phosphopeptide enrichment showed that total OPN protein level in WT CM was 2.8-fold lower than that in Fam20C-VavCre CM (Supplementary Table S1). ELISA assays confirmed that the amount of secreted OPN in the CM from Fam20C-LysMCre and Fam20C-VavCre osteoclast precursors were 135% and 271% of that in the WT controls, respectively (Fig. 3B). Similarly, in differentiating osteoclasts after RANKL stimulation, the levels of secreted OPN from Fam20C-LysMCre and Fam20C-VavCre cultures were also higher than their respective WT controls, and this difference in OPN secretion was diminished as osteoclast matured (Fig. 3C). In contrast, OPN mRNA levels were similar in Fam20C-LysMCre and Fam20C-VavCre compared with their respective WT controls (Supplementary Fig. S2C). Western blot analysis confirmed the depletion of Fam20C protein secretion and OPN phosphorylation in Fam20C-LysMCre and Fam20C-VavCre cells (Fig. 3D; Supplementary Fig. S2D). Western blot also showed that although the levels of secreted OPN was 2.6-fold and 8.9-fold higher in Fam20C-LysMCre and Fam20C-VavCre than their WT controls, respectively, total OPN levels in the whole-cell lysate were similar (Fig. 3D). These findings suggest that OPN phosphorylation by Fam20C impairs OPN secretion from osteoclast precursors without affecting OPN expression.
High OPN level facilitates not only bone resorption (32), but also cancer development and bone metastasis (33). Elevated plasma OPN level is found in breast cancer patients with bone metastases compared to healthy volunteers, and is correlated with shorter patient survival (34). We analyzed mRNA expression data from the METABRIC (27, 28). Patients with breast cancer of all types with high OPN expression have about 21-month shorter OS than those with low OPN expression (Supplementary Fig. S2E). Our qRT-PCR assays also revealed that the mRNA expression levels of OPN in three bone metastasis–prone breast cancer cell lines, MDA-231-BoM-1833 cells (1833 cells), 4T1.2 cells, and Py8119 cells, were significantly higher than those in their parental cell lines (Supplementary Fig. S2F). This suggests that the increased OPN secretion from Fam20C-deficient osteoclast precursors may also exacerbate breast cancer bone metastasis through promoting both breast cancer cell growth and osteoclast differentiation.
The proosteoclastogenic effects of myeloid Fam20C deficiency can be blocked by an OPN-neutralizing antibody
To test whether the elevated OPN secretion from Fam20C-deficient myeloid cells functionally mediates Fam20C regulation of osteoclast differentiation, we utilized an OPN-neutralizing antibody. Upon OPN neutralization, the higher expression of osteoclast differentiation markers in Fam20C-LysMCre and Fam20C-VavCre cultures was reduced to comparable levels as that in WT cultures (Fig. 4A and B), supporting OPN as a key Fam20C functional target during the regulation of osteoclast differentiation.
Myeloid Fam20C suppresses breast cancer bone metastasis
To determine the effects of Fam20C deficiency in the myeloid lineage on bone metastasis of breast cancer cells in vivo, we used our previously described experimental mouse models (18, 23) in which luciferase-labeled Py8119 mouse cancer cells were transplanted via ultrasound-guided intracardiac injection into WT mice and Fam20C-VavCre mice or Fam20C-LysMCre mice. The development of bone metastases from Py8119 were exacerbated in both Fam20C-VavCre mice and Fam20C-LysMCre mice compared with their respective WT control mice (Fig. 4C and D; Supplementary Fig. S2G). Moreover, treatment with an OPN-neutralizing antibody showed a trend, although not significant, of rescuing the elevated bone metastasis in Fam20C-VavCre mice (Supplementary Fig. S2G). Together, these findings indicate that Fam20C in the myeloid lineage suppresses breast cancer bone metastasis by elevating the secretion of OPN.
Fam20C in breast cancer cells promotes primary tumor growth and bone metastasis
To examine the functions of Fam20C in breast cancer cells, we generated Fam20C knockout (KO) 1833 cells with CRISPR-Cas9 (Supplementary Fig. S3A and S3B). The growth rate of 1833-Fam20C-KO cells in vitro was much slower than 1833 control cells (Fig. 5A). After implantation into the fourth mammary fat pad of nude mice, the tumor growth rate of 1833-Fam20C-KO cells in vivo was also significant slower than 1833 control cells (Fig. 5B and C; Supplementary Fig. S3C). These results suggest that Fam20C accelerates primary mammary tumor growth and may represent a new therapeutic target for breast cancer.
Given the important roles of Fam20C in breast and bone, we examined whether Fam20C in breast cancer cells affects skeletal metastasis. We found that Fam20C expression was significantly higher in bone-met-prone sublines (both 1833 cells and Py8119 cells) compared with their parental cells (Supplementary Fig. S3D). Moreover, 1833-Fam20C-KO cells exhibited a significantly reduced transwell migration capability in vitro compared with the control cells (Fig. 5D).
Our in vivo analyses reveal that bone metastasis of intracardiacally injected 1833-Fam20C-KO cells in nude mice was significantly impaired compared with 1833 control cells (Fig. 5E; Supplementary Fig. S3E). Zoledronate (ZOL) is a member of the bisphosphonate family of osteoclast inhibitor drugs that is often used in clinic to treat osteoporosis and cancer bone metastasis. Treatment with ZOL impeded the development of bone metastases and largely impaired the effects of cancer cell Fam20C KO on bone metastasis (Fig. 5E), suggesting that osteoclast may be a key player in the regulation of bone metastasis by cancer cell Fam20C.
As a complementary approach to CRISPR KO, we also generated Fam20C knockdown (KD) 1833 cells (1833-shFam20C) with Fam20C shRNA (Supplementary Fig. S3F). Similarly, Fam20C KD in breast cancer cells also significantly reduced the development of bone metastases compared with nontargeting shRNA control after intracardiac injection (Fig. 5F; Supplementary Fig. S3G). Collectively, these results suggest that Fam20C in cancer cells promotes skeletal metastasis via multiple mechanisms by directly enhancing cancer cell proliferation and mobility as well as indirectly activating osteoclasts in the bone metastatic niche.
Because Fam20C-LysMCre or Fam20C-VavCre mice are in the immune-competent C57BL/6 background, mouse Py8119 cells, but not human 1833 cells, are compatible for studying bone metastasis in these mice. However, we failed in generating Fam20C-KD Py8119 cells probably due to low Fam20C levels in Py8119 cells (Supplementary Fig. S3F). On the other hand, we did not observe significant difference of bone metastases in nude mice intracardiacally injected with MDA-MB-231 cells stably overexpressing Fam20C-WT or enzymatic inactive D478A mutant (Supplementary Fig. S3H), which is possibly due to endogenous WT Fam20C present in D478A-mutant stable cells, although Fam20C endogenous level is lower in 231 cells than in 1833 cells (Supplementary Fig. S3D).
Fam20C and its substrate BMP4 enhance breast cancer cell–induced osteoclast differentiation
Considering that Fam20C regulates cell–cell communications by modulating protein phosphorylation and secretion, we tested whether Fam20C in breast cancer cells affected the ability of cancer cell CM to stimulate osteoclastogenesis from mouse BM precursors (Fig. 6A). Compared with CM from 1833 control cells, CM from 1833-Fam20C-KO cells showed a significantly lower ability to induce osteoclast differentiation (Fig. 6B; Supplementary Fig. S4A) and the expression of osteoclast markers such as TRAP and CTSK (Fig. 6C; Supplementary Fig. S4B). Consistent with this observation, osteoclast differentiation was enhanced by CM from 1833 cells overexpressing Fam20C WT, but not its catalytic inactive mutant D478A (DA; Fig. 6D). To examine whether these effects resulted from secreted Fam20C, we engineered a Fam20C construct with a KDEL sequence at the C-terminus to prevent Fam20C secretion but maintain its capability of phosphorylating secreted proteins (8). CM from 1833 cells overexpressing Fam20C-WT-KDEL was still able to increase osteoclast differentiation compared with Fam20C-DA-KDEL and vector control (Fig. 6D), indicating that secreted Fam20C substrates in CM, rather than secreted Fam20C, are required to promote osteoclast differentiation.
Previous phosphoproteome analysis of CM from breast cancer cell line MDA-MB-231 has identified several secreted Fam20C substrates with important roles in bone development, including bone morphogenetic protein 4 (BMP4), insulin-like growth factor-binding protein 7 (IGFBP7), and milk fat globule-EGF factor 8 (MFGE8; ref. 8). We established 1833 cells that stably overexpress Fam20C, BMP4, IGFBP7, MFGE8, OPN, or a GFP control (Supplementary Fig. S4C). Elevated OPN levels in CM enhanced osteoclast differentiation (Supplementary Fig. S4D), confirming the stimulatory effect of OPN on osteoclastogenesis. However, Fam20C-dependent OPN phosphorylation is not found in CM of MDA-MB-231 cells (8). And even if Fam20C KO also leads to loss of OPN phosphorylation and increased OPN secretion in cancer cells, it cannot explain the reduced osteoclast differentiation by CM from 1833-Fam20C-KO cells. Thus, distinct from osteoclast precursors, Fam20C in breast cancer cells may regulate osteoclastogenesis via a different mechanism that is independent of OPN.
Among these stable cells overexpressing Fam20C or its substrates (Supplementary Fig. S4C), CM from 1833-BMP4 cells had the strongest effect on elevating osteoclast differentiation (Fig. 6E; Supplementary Fig. S4D). Our analysis of METABRIC data revealed that the OS of patients with triple-negative breast cancer (TNBC) was not correlated with the expression levels of Fam20C, IGFBP7, or MFGE8 (Supplementary Fig. S4E). Interestingly, patients with TNBC with low BMP4 expression levels in tumor had a much better OS (median survival, 195.4 months) than those with high BMP4 levels (median survival, 108.5 months; Fig. 6F). BMP4 expression levels in BM cells and cultured osteoclasts were much lower than that in breast cancer cells (Supplementary Fig. S4F), suggesting that BMP4 in the bone metastatic niche is primarily produced by breast cancer cells. Therefore, TNBC cells producing higher BMP4 may enhance osteoclast differentiation, and consequently trigger the development of bone metastasis and shorten patient survival time.
BMP4 phosphorylation by Fam20C enhances BMP4 stability and breast cancer cell–induced osteoclast differentiation
Fam20C phosphorylates BMP4 at Serine 91 in MDA-MB-231 cells (8). We next generated BMP4 phosphomimetic mutant S91D (Serine to Aspartic acid) and nonphosphorylatable mutant S91A (serine to alanine). The ability of CM from 1833-BMP4 cells to induce osteoclast differentiation was abolished by Fam20C KO or BMP4-S91A mutation, but augmented by BMP4-S91D mutation (Fig. 6G). Reexpressing Fam20C-WT, but not Fam20C-DA, with BMP4 in 1833-Fam20C-KO cells significantly increased cancer cell CM-induced expression of osteoclast maker TRAP to a similar level induced by BMP4-S91D (Fig. 6H). These data suggest that the phosphorylation of BMP4 by Fam20C in breast cancer cells enhances cancer cell–induced osteoclast differentiation.
To dissect how Fam20C regulates BMP4 functions, we examined whether BMP4 phosphorylation affects its protein stability and secretion. BMP4 protein levels (both full-length and mature forms, both intracellular and secreted forms), but not mRNA levels, were greatly reduced in 1833-Fam20C-KO cells compared with 1833 control cells (Fig. 6I; Supplementary Fig. S5A). BMP4 protein levels were further decreased by BMP4-S91A mutation in both 1833 and 1833-Fam20C-KO cells compared with BMP4-WT (Fig. 6I). Proteasome and lysosome are the two major systems for secreted protein degradation (35). Basal BMP4 protein levels were elevated by both the lysosome inhibitor chloroquine (Fig. 6J) and the proteasome inhibitor MG132 (Supplementary Fig. S5B). However, the reduced protein stability caused by BMP4-S91A mutation was only rescued by chloroquine, not MG132 (Fig. 6J; Supplementary Fig. S5B). Moreover, the reduced BMP4-S91A level in another breast cancer cell line Py8119 was rescued to a similar level as WT BMP4 by the treatment of chloroquine (Supplementary Fig. S5C). Together, these results indicate that Fam20C-mediated BMP4 phosphorylation at S91 prevents the lysosomal degradation of BMP4 protein.
Inhibition of BMP4 signaling blocks bone metastasis and Fam20C regulation in cancer cells
To further confirm the roles of Fam20C substrate BMP4 in breast cancer bone metastasis, we established BMP4 KD stable cells with BMP4 shRNA (Supplementary Fig. S5D). Comparing with CM from 1833-shNT control cells, CM from 1833-shBMP4 significantly reduced osteoclast size (Fig. 7A; Supplementary Fig. S5E) and the expression of osteoclast marker (Fig. 7B). The growth rate of 1833-shBMP4 in vitro was only slightly slower than that of 1833-shNT cells (Supplementary Fig. S5F). After intracardiac injection into nude mice, 1833-shBMP4 cells developed much less bone metastases than 1833-shNT cells (Fig. 7C; Supplementary Fig. S5G). Serum CTX-1 was significantly lower in mice injected with 1833-shBMP4 cells compared with 1833-shNT cells (Fig. 7D). These results indicate that BMP4 deletion in breast cancer cells diminishes their ability to elevate osteoclast differentiation and bone resorption in the bone environment, thus blocks skeletal metastasis. Stably overexpressing BMP4-WT or its mutants S91D and S91A in 1833 cells did not significantly affect bone metastasis in nude mice after intracardiac injection (Supplementary Fig. S5H), which is probably due to the presence of endogenous WT-BMP4 (Supplementary Fig. S5D).
LDN-193189 (LDN), a selective inhibitor of BMP receptors ALK2 and ALK3, can efficiently suppress BMP4-induced Smad1/5/8 phosphorylation (36). LDN treatment in vitro significantly attenuated the ability of cancer cell CM to induce the expression of osteoclast markers TRAP and CTSK during osteoclast differentiation (Fig. 7E). LDN treatment in vivo also reduced bone metastasis of intracardiacally injected 1833 cells, but not 1833-Fam20C-KO–negative control cells (Fig. 7F; Supplementary Fig. S5I). These results indicate that BMP4 phosphorylation by Fam20C and BMP4 signaling are required for the bone metastasis development of Fam20C-expressing breast cancer, and reveal a Fam20C-BMP4 signaling cascade within the cancer–osteoclast crosstalk in the bone metastatic niche.
During breast cancer development, many Fam20C substrates such as IGFBPs and BMP4 modulate tumor growth (37, 38). During osteolytic metastasis, the vicious cycle between cancer cells and osteoclasts is also well established (5). The nutrient-rich osseous environment and the abundant growth factors released upon bone resorption support the homing, survival and proliferation of breast cancer cells in bone. Meanwhile, breast cancer cells produce factors to induce osteoclast differentiation and activation. A number of proteins have been reported to be secreted by breast cancer cells to facilitate bone metastasis, such as parathyroid hormone–related protein (PTHrP), MCSF, IL11, VEGF (39). Nonetheless, how osteoclasts and breast cancer cells control the secretion of these factors is not well understood. Our findings reveal diverse roles of Fam20C in controlling the secretion of different proteins from osteoclasts and breast cancer cells (Fig. 7G). In osteoclast precursors, Fam20C phosphorylates OPN and prevents OPN secretion, leading to decreased osteoclast differentiation and breast cancer bone metastasis. But in breast cancer cells, Fam20C phosphorylates BMP4 and enhances BMP4 stability and secretion, resulting in elevated osteoclast differentiation and bone metastasis. Our findings also uncover new therapeutic strategies for breast cancer bone metastasis by treatment with OPN neutralizing antibody and BMP4 signaling inhibitor.
Raine syndrome is an autosomal-recessive disorder caused by FAM20C mutation or deletion. This disease shows symptoms including disorders of bone architectures, especially osteosclerosis. But R408W mutation of FAM20C is reported to induce osteomalacia (40). Global inactivation of Fam20C by Sox2-Cre in mice leads to osteomalacia and increased bone porosity (41). These previous findings demonstrate that Fam20C deficiency leads to complicated bone disorder phenotypes involving multiple bone cell types. Osteoblast-specific Fam20C deletion in mice reduces bone formation (12, 13). However, the roles of Fam20C in osteoclasts are not clear. In this study, we examined the roles of Fam20C in osteoclasts using two mouse strains that carry specific Fam20C deletion in different osteoclast differentiation stages. Both strains showed elevated osteoclastogenesis. We further demonstrated that Fam20C deletion leads to decreased OPN phosphorylation and increased OPN secretion. OPN can stimulate cell signaling in osteoclast precursors, thus promoting osteoclast differentiation and bone resorption (31, 32). Patients with breast cancer with metastases to bone have elevated plasma OPN levels, and elevated plasma OPN is associated with reduced patient survival (34). Here we show that treatment with an OPN-neutralizing antibody can impede osteoclast differentiation and breast cancer bone metastasis. In addition, breast cancer cells also express RANKL to stimulate osteoclast differentiation (42). We found that RANKL decreased the expression of Fam20C in osteoclasts (Supplementary Fig. S1A), which could further enhance OPN secretion and osteoclast differentiation. OPN phosphorylation in bone marrow cells likely occur in the lumen of secretory pathway before secreting to the extracellular space (8). Nonsecretable Fam20C-KDEL in breast cancer cells had a similar osteoclast-stimulating effect as secretable Fam20C-WT (Fig. 6D), and Fam20C in breast cancer cells promoted osteoclast differentiation (Fig. 6B and C). Thus Fam20C in breast cancer cells is unlikely to have a similar regulation of suppressing osteoclast OPN secretion as Fam20C in BM cells.
Intriguingly, our study uncovers the opposite effects of Fam20C deficiency in breast cancer cells and osteoclasts during bone metastasis. If specific uptake by cancer cells but not by myeloid cells can be achieved, Fam20C inhibitors may suppress primary breast tumor growth and bone metastasis. Similarly, if specific uptake by myeloid cells but not by cancer cells can be achieved, Fam20C activators may impede bone loss and breast cancer skeletal metastasis. However, if uptake occurs in both cancer cells and myeloid cells, Fam20C inhibitors may reduce tumor growth but risk increasing bone metastasis. Our study suggests that the pro-bone-met effects of Fam20C inhibitors in myeloid cells may be prevented by cotreatment with OPN-neutralizing antibodies or antiresorptive drugs such as bisphosphonates. However, anti-OPN treatment in vivo is challenging due to the fast OPN turnover rate and relatively high plasma OPN concentration in human (43). New OPN-neutralizing antibodies with higher affinity need to be developed.
The Fam20C substrate BMP4 can drive the differentiation of both osteoblast and osteoclast (44). BMP signaling has dual roles in cancer development as both tumor promoter and suppressors (38). Despite this, an increasing number of BMP signaling inhibitory molecules have been developed with great advancements in their selectivity and safety. These molecules, such as Dalantercept, are under clinical trials to test their potency in treating cancers (45). Phosphorylated BMP4 has been detected in the CM from breast cancer cell lines (8, 46), but not in the plasma of patients with breast cancer or healthy human plasma/serum (46–48). It is likely that the phosphorylated BMP4 mainly functions in a paracrine manner in the local bone metastatic niche. Improper cleavage of BMP4 causes rapid lysosomal degradation (49), indicating that BMP4 lysosomal degradation is an important regulatory mechanism for producing mature BMP4 for secretion. Our study supports a key role of BMP4 in fueling the vicious cycle between breast cancer cells and osteoclasts, highlighting BMP4 signaling inhibitors as effective therapeutic strategy to impede bone metastasis and an indirect approach to suppress Fam20C's functions. Moreover, our findings reveal a novel regulation of BMP4 phosphorylation, stability and secretion by the Fam20C kinase in breast cancer cells, through which Fam20C exerts indirect effects to stimulate osteoclastogenesis and the bone metastatic niche.
No disclosures were reported.
H. Zuo: Conceptualization, data curation, formal analysis, validation, investigation, methodology, writing–original draft, writing–review and editing. D. Yang: Data curation. Y. Wan: Conceptualization, supervision, funding acquisition, writing–review and editing.
The authors thank Drs. Vincent S. Tagliabracci and Brenden C. Park for providing constructs, Fam20C flox mice, and assistance in establishing CRISPR stable cells; Dr. Orson Moe for sharing the Bioquant software; Dr. Qiwen Yang for assistance with tumor analyses. This work was supported by Charles Pak Family Breast Cancer-Bone Initiative Grant (to Y. Wan).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.