Research ArticleCancer Metastasis

Thymidine phosphorylase exerts complex effects on bone resorption and formation in myeloma

See allHide authors and affiliations

Science Translational Medicine  24 Aug 2016:
Vol. 8, Issue 353, pp. 353ra113
DOI: 10.1126/scitranslmed.aad8949

Myeloma enzyme makes way for metastasis

Bone tissue is built up by osteoblasts and broken down by osteoclasts in a balanced remodeling process. In metastatic cancer, however, the balance is tipped, leading to the formation of cancerous growths in the bone. Attempts to prevent metastasis have not been successful in the clinic; thus, Liu and colleagues set out in search of a new pathway to target. The authors found that an enzyme produced by myeloma cells, called thymidine phosphorylase (TP), suppressed osteoblast activity (new bone formation) and enhanced osteoclast activity (bone resorption). Inhibiting TP reduced the incidence of myeloma-induced osteolytic bone lesions, suggesting a new target for translation to the clinic, especially because certain TP inhibitors are already approved for human use.

Abstract

Myelomatous bone disease is characterized by the development of lytic bone lesions and a concomitant reduction in bone formation, leading to chronic bone pain and fractures. To understand the underlying mechanism, we investigated the contribution of myeloma-expressed thymidine phosphorylase (TP) to bone lesions. In osteoblast progenitors, TP up-regulated the methylation of RUNX2 and osterix, leading to decreased bone formation. In osteoclast progenitors, TP up-regulated the methylation of IRF8 and thereby enhanced expression of NFATc1 (nuclear factor of activated T cells, cytoplasmic 1 protein), leading to increased bone resorption. TP reversibly catalyzes thymidine into thymine and 2-deoxy-d-ribose (2DDR). Myeloma-secreted 2DDR bound to integrin αVβ35β1 in the progenitors, activated PI3K (phosphoinositide 3-kinase)/Akt signaling, and increased DNMT3A (DNA methyltransferase 3A) expression, resulting in hypermethylation of RUNX2, osterix, and IRF8. This study elucidates an important mechanism for myeloma-induced bone lesions, suggesting that targeting TP may be a viable approach to healing resorbed bone in patients. Because TP overexpression is common in bone-metastatic tumors, our findings could have additional mechanistic implications.

INTRODUCTION

Bone is constantly being remodeled in a process where osteoclasts resorb bone, osteoblasts deposit type I collagen and other proteins in the resorbed lacunae, and, lastly, the collagen mineralizes to form bone (1, 2). The first partner in this “pas de deux” is the osteoclast, which arises from hematopoietic monocytic precursors and resorbs bone. The formation of osteoclasts requires the cytokine receptor activator of nuclear factor kB ligand (RANKL) and macrophage colony-stimulating factor (M-CSF) (3). RANKL enhances the expression of nuclear factor of activated T cells, cytoplasmic 1 protein (NFATc1), a transcriptional factor that up-regulates the expression of osteoclast differentiation–associated genes, such as tartrate-resistant acid phosphatase (TRAP), calcitonin receptor (CALCR), and cathepsin K (CTSK), whereas the transcriptional factor interferon regulatory factor 8 (IRF8) can suppress RANKL-induced NFATc1 expression (4). The second player in the remodeling cycle is the osteoblast, which is differentiated from mesenchymal stem cells (MSCs). This process requires the activation of core-binding factor α-1/runt-related transcription factor 2 (RUNX2) and osterix, which stimulate the expression of osteoblast differentiation–associated genes, such as bone γ-carboxyglutamic acid–containing protein (BGLAP), alkaline phosphatase (ALP), and collagen type I α1 (COL1A1).

This delicate balance between osteoclast-mediated bone resorption and osteoblast deposition of matrix is disrupted in certain types of malignancies, including multiple myeloma and solid tumors, such as breast and lung cancer, among others (5, 6). In multiple myeloma, tumor cells secrete RANKL or stimulate the release of RANKL by surrounding stromal cells, leading to enhanced osteoclast differentiation. Myeloma cells can also secrete dickkopf-related protein 1 (DKK1), which inhibits the Wnt/β-catenin signaling pathway and suppresses osteoblast differentiation. Attempts to target RANKL and DKK1 in myeloma therapeutically have achieved only modest success. For example, the anti-resorptive agent denosumab (a monoclonal antibody against RANKL) was examined in a phase 3 trial, but only moderately affected myeloma-induced lytic lesions were shown (7). BHQ880 (a monoclonal antibody against DKK1) is now in a phase 1/2 study, but its application in myeloma fails to restore new bone formation (8). Thus, bisphosphonates, which suppress osteoclast function, remain the mainstay in treatment of myeloma-induced bone disease. Unfortunately, bisphosphonates are less than fully effective and cause osteonecrosis of the jaw in 2 to 5% of treated patients (7). The goal of the current studies is to identify other factors produced by myeloma cells, which regulate both resorption and formation, and to determine whether this information can be used to prevent myeloma-induced bone disease.

Thymidine phosphorylase (TP), also called platelet-derived endothelial cell growth factor, is an enzyme that can reversibly catalyze the conversion of thymidine to thymine and 2-deoxy-d-ribose-1-phosphate (2DDR1P), which is further dephosphorylated into a smaller, more stable molecule, 2-deoxy-d-ribose (2DDR) (9). TP has been found in a wide range of normal tissues (10) and participates in wound healing and a variety of chronic inflammatory diseases (11). TP is highly expressed in many types of cancers, including lung and breast cancer, and plays an important role in angiogenesis and anti-apoptosis (12, 13). Clinically, elevated levels of TP are associated with cancer aggressiveness and poor prognosis (11), but TP has never been implicated in the regulation of bone resorption or formation. In previous studies, we demonstrated that p38 MAPK (mitogen-activated protein kinase) activity in myeloma cells induces osteolytic bone lesions (14, 15), and others demonstrated that this signaling regulates TP expression (16, 17). We therefore hypothesized that the TP gene plays a role in the pathogenesis of cancer-induced bone destruction in myeloma.

Here, we report an association of cancer-expressed TP with osteolytic bone lesions, as well as the ability of TP to orchestrate osteoclast-mediated resorption and decreased bone formation. Specifically, TP down-regulated the expression of IRF8 and thereby activated RANKL-induced NFATc1 expression, leading to an increase in osteoclastogenesis and bone resorption. We further observed that myeloma-expressed TP suppressed osteoblastogenesis and bone formation by down-regulating the expression of RUNX2 and osterix in human MSCs. Our findings not only elucidate a mechanism of cancer-induced suppression of osteoblast differentiation and activation of osteoclast differentiation and activity, but also implicate a potential therapeutic approach for cancer patients with osteolytic bone lesions by targeting TP.

RESULTS

Myeloma-expressed TP enhances lytic bone lesions

We confirmed the presence of myeloma (CD138+) cells in tissue array biopsies and found that TP+ cells in the bone marrow from 14 myeloma patients were significantly greater than those from 14 healthy donors (Fig. 1, A and B). TP was expressed in most of the bone marrow aspirates of primary myeloma cells (4 of 6 patients) and in most of the established human myeloma cell lines (4 of 6), but not in aspirates of plasma cells from normal subjects (Fig. 1C).

Fig. 1. TP is highly expressed in myeloma.

(A) Representative immunohistochemical images of bone marrow (BM) biopsies from tissue arrays from 14 healthy and 14 myeloma patients stained for CD138 and TP. Scale bar, 50 μm. MM, multiple myeloma. (B) Densitometry analysis of CD138+ cells or TP+ cells in (A). Data are box plots showing the distribution and median value of quantitative staining (n = 14). P values were determined by Student’s t test. (C) Western blot analysis of TP expression in normal plasma cells from four healthy donors, malignant plasma cells of six myeloma patients (Pt), and six established human myeloma cell lines. Primary plasma cells were isolated from the BM aspirates of healthy donors or myeloma patients. β-Actin served as loading control. Data are representative of triplicate blots.

To determine whether TP expression correlated with the frequency of bone lesions in myeloma patients, we performed two experiments using samples from tissue banks. In the first experiment, there was a strong positive correlation between TP gene expression in myeloma cells and bone lesion numbers in 52 patients (Fig. 2A); in the second experiment, there was a robust positive correlation between TP immunohistochemistry and bone lesion numbers (Fig. 2B). TP expression was higher in myeloma cells from patients with high bone lesion scores than in those from patients with low lesion numbers (Fig. 2C).

Fig. 2. Association of TP expression and lytic bone lesion in myeloma.

(A) Correlation coefficient between the mRNA levels of TP and numbers of bone lesion in myeloma patients (n = 52). P values were determined by Pearson coefficient. (B and C) BM biopsy samples from n = 13 patients in (A) were labeled with an anti-TP antibody. TP staining was analyzed using the Image-Pro Plus. (B) Correlation between TP staining in BM biopsies and the numbers of bone lesions in myeloma patients. P values were determined by Pearson coefficient. (C) Representative images of immunohistochemical staining showing TP expression in myeloma cells and CD138+ infiltrated myeloma cells within BM of the patient samples from (B) highlighted with red circles. Scale bar, 10 μm. (D to H) On the basis of the levels of TP expression in myeloma cells, patients’ myeloma cells were separated into high and low TP expression groups (TPhigh and TPlow; n = 5 patients’ BM aspirates per group). In addition, myeloma cells were injected into the bone chips of SCID-hu mice or SCID mouse femurs. Shown are representative x-ray images (D to G) and summarized data of the percentage of bone volumes versus total volumes (BV/TV) (H) of lytic lesion in the implanted human bone chips of SCID-hu mice injected with TPhigh and TPlow cells or in the femurs of SCID mice injected with myeloma cell lines ARP-1 [wild-type (WT), nontargeted shRNA (shCtrl), and TP shRNA (shTP)] and MM.1S [WT, control vector (Vec), and TP cDNA (TP)]. Data are averages ± SD (n = 5 mice per group, three replicate studies). P values were determined by Student’s t test.

We then isolated primary myeloma cells from the bone marrow aspirates of 10 newly diagnosed patients. On the basis of the TP expression in myeloma cells as detected by Western blot, the aspirates were separated into high and low TP expression groups: TPhigh and TPlow (n = 5 patients’ bone marrow aspirates per group) (fig. S1A). To determine the functional role of myeloma-expressed TP in lytic bone lesions, we injected TPhigh or TPlow cells into human bone chips that had been implanted into severe combined immunodeficient (SCID)–hu mice. Mice with no tumor cells (but with bone chips implanted) were controls. More lytic lesions and a lower percentage of bone volume versus total volume (BV/TV) were observed in the bone chips of mice injected with TPhigh cells than in those injected with TPlow cells (Fig. 2, D and H). Moreover, injecting wild-type ARP-1 cells, which have high TP expression, into mouse femurs caused more lytic lesions than wild-type MM.1S cells, which have low TP expression (Fig. 2, E and H, and fig. S1B). ARP-1 cells with TP knocked down by small hairpin RNAs (shRNAs; shTP ARP-1) were injected into mouse femurs and caused fewer lytic lesions than did ARP-1 cells with a nontargeted shRNA (shCtrl ARP-1) (Fig. 2, F and H, and fig. S1C). Conversely, MM.1S cells expressing TP complementary DNA (cDNA; TP MM.1S) caused more femur lesions than did wild-type MM.1S cells expressing a control vector (Vec MM.1S) (Fig. 2, G and H, and fig. S1D). In summary, cells with high TP expression induce the formation of lytic lesions in mouse and human bone.

To determine whether lytic lesions result from a change in the tumor burden within the marrow milieu, we measured the levels of myeloma-secreted M protein, a reflection of tumor burden, in the mouse sera. The modulation of TP expression in myeloma cells did not change the serum levels of M protein (fig. S1E). In addition, there were no differences in viability or apoptosis in control and modified ARP-1 or MM.1S cell lines (fig. S1, F and G), indicating that changes in TP expression did not affect myeloma cell growth or survival. Together, these results reveal that myeloma cells express TP and enhance lytic bone lesions in patients and mice with myeloma.

Myeloma-expressed TP enhances RANKL-induced osteoclastogenesis and bone resorption

We studied the ability of myeloma-expressed TP to regulate osteoclast differentiation in vitro using a standard osteoclast differentiation protocol (18), in which TRAP+ cell number and TRAP5b secretion were assessed in the presence of RANKL (10 ng/ml) and M-CSF (25 ng/ml), both of which are needed for osteoclast formation. The coculture of preosteoclasts (preOCs) with TPhigh patient myeloma cells or wild-type ARP-1 cells induced higher numbers of multinuclear TRAP+ cells (fig. S2A), more TRAP5b secretion (fig. S2B), and higher expression of the osteoclast genes TRAP, CALCR, and CTSK (fig. S2, C and D) than coculture with cells expressing low levels of TP (TPlow or wild-type MM.1S). Knockdown of TP expression in wild-type ARP-1 cells reduced osteoclast differentiation and activity (fig. S2, A, B, and E). In contrast, overexpression of TP in wild-type MM.1S cells enhanced the osteoclast differentiation and activity (fig. S2, A, B, and F).

To assess the role of myeloma-expressed TP in osteoclast-mediated bone resorption in vivo, we stained myeloma-bearing human bone chips or mouse femurs for TRAP and analyzed the percentage of bone surface eroded by osteoclasts (ES/BS) and the percentage of bone surface covered with osteoclasts (Oc. S/BS). The percentage of ES/BS (Fig. 3A) and Oc. S/BS was higher in the mice injected with myeloma cells that had high TP expression (TPhigh, wild-type ARP-1, shCtrl ARP-1, or TP MM.1S) than in those that had low TP expression (TPlow, shTP ARP-1, wild-type MM.1S, or Vec MM.1S) (Fig. 3, A and B). These results indicate that myeloma-expressed TP enhances osteoclastogenesis.

Fig. 3. Myeloma-expressed TP enhances osteoclast-mediated bone resorption and inhibits osteoblast-mediated bone formation in vivo.

The implanted human bone chips from SCID-hu mice injected with TPhigh and TPlow cells (n = 5 patients’ BM aspirates per group) or the femurs from SCID mice injected with myeloma cell lines ARP-1 [WT, nontargeted shRNA (shCtrl), and TP shRNA (shTP)] and MM.1S [WT, control vector (Vec), and TP cDNA (TP)] were fixed, TRAP- or toluidine blue–stained, and analyzed by BIOQUANT OSTEO software. (A to D) Percentage of bone surface eroded by osteoclasts (ES/BS) (A), percentage of bone surface covered with osteoclasts (Oc. S/BS) (B), percentage of osteoid surface (OS/BS) (C), and percentage of total bone surface lined with osteoblasts (Ob. S/BS) (D) in myeloma-bearing human bone chips or mouse femurs. (E and F) Bone formation rate (BFR/BS) was measured by calcein injection, and the undecalcified bone sections were imaged and analyzed. Shown are representative images or summarized data of bone formation in the femurs from SCID mice injected with myeloma cell lines ARP-1 (shCtrl and shTP) and MM.1S (Vec and TP). Scale bars, 20 μm. All data are averages ± SD (n = 5 mice per group, three replicate studies). All P values were determined by Student’s t test. y, year.

In preOCs, myeloma cells have been shown to enhance the expression of NFATc1, a transcriptional factor that plays a pivotal role in osteoclast gene expression and can be up-regulated by RANKL (19). We confirmed this effect of myeloma cells on NFATc1 expression in preOCs (fig. S3A) and further demonstrated that myeloma cells with high but not low TP down-regulated expression of the transcription factor IRF8 in preOCs (fig. S3A). To determine whether IRF8 mediates TP-induced osteoclast differentiation, its expression in preOCs was knocked down using IRF8 shRNA (shIRF8). When cocultured with wild-type ARP-1 cells, shIRF8-preOCs expressed higher levels of NFATc1 protein (fig. S3B) and secreted more TRAP5b (fig. S3C) than did control preOCs. Thus, myeloma-expressed TP abrogated the inhibitory effect of IRF8 on NFATc1 shown previously (3), leading to the promotion of osteoclast gene expression and osteoclastogenesis.

Myeloma-expressed TP inhibits osteoblastogenesis in vitro and bone formation in vivo

To determine whether myeloma-expressed TP regulates osteoblast differentiation in vitro, we cocultured the precursors of osteoblasts, MSCs, with patient myeloma cells and myeloma cell lines in osteoblast medium for 2 weeks. MSCs cultured alone in this medium served as a positive control. Mature osteoblasts that produced soluble ALP were positive for Alizarin red S staining (which is indicative of osteoblast-mediated bone formation activity) and expressed osteoblast differentiation–associated genes (fig. S4). In line with previous studies (20), the coculture of MSCs with myeloma cells inhibited osteoblast activity, but coculture with TPlow myeloma cells and low TP–expressing myeloma cell lines (wild-type MM.1S, Vec MM.1S, or shTP ARP-1) had comparatively more mature osteoblasts than those with high levels of TP (fig. S4).

To determine the role of myeloma-expressed TP in osteoblast-mediated bone formation in vivo, we collected myeloma-bearing human bone chips or mouse femurs, enumerated the osteoblasts localized on trabecular bone, and counted the percentage of osteoid surface (OS/BS) and of bone surface lined with osteoblasts (Ob. S/BS). The percentages of OS/BS (Fig. 3C) and Ob. S/BS (Fig. 3D) were lower in the mice injected with myeloma cells expressing high levels of TP (TPhigh, wild-type ARP-1, shCtrl ARP-1, or TP MM.1S) than in those injected with low-TP myeloma cells (TPlow, shTP ARP-1, wild-type MM.1S, or Vec MM.1S). In agreement with these data, bone formation rate was increased in mice injected with shTP ARP-1 cells (Fig. 3E) and decreased in mice with TP MM.1S cells (Fig. 3F).

Transcription factors such as RUNX2 and osterix promote osteoblast differentiation by up-regulating the expression of osteoblast differentiation–associated genes (21). In line with previous studies (20), we observed that the coculture of MSCs with myeloma cells reduced the expression of RUNX2 and osterix (fig. S5). Moreover, the expression of these transcription factors was significantly lower when they were cocultured with myeloma cells with high levels of TP than when they were cocultured with lower-TP myeloma cells (fig. S5A). Moreover, myeloma-expressed TP modulated the binding activity of RUNX2 and osterix to the promoter of osteoblast genes BGLAP and COL1A1 (fig. S5, B and C), a finding consistent with the effect of TP on bone formation.

TP down-regulates the expression of RUNX2, osterix, and IRF8 through hypermethylation of CpG islands

DNA methylation of CpG dinucleotides is a key epigenetic modification that influences tissue- and context-specific gene expression (22) and is generally associated with gene silencing (23). To determine whether myeloma-expressed TP regulates the methylation of CpG islands (CGI) in RUNX2, osterix, and IRF8, we designed MSP [methylation-specific polymerase chain reaction (PCR)] and BSP (bisulfite sequencing PCR) primers targeting their CpG-rich regions (Fig. 4A and table S2). Methylation of RUNX2 and osterix in MSCs and IRF8 in preOCs was higher in the coculture of myeloma cells with high TP expression (wild-type ARP-1, shCtrl ARP-1, and TP MM.1S) than in the coculture of those with low TP expression (shTP ARP-1, wild-type MM.1S, and Vec MM.1S) (Fig. 4B). BSP analysis confirmed these results (Fig. 4, C to E), and similar methylation data were obtained from healthy individual and myeloma patient samples expressing different levels of TP (Fig. 4F).

Fig. 4. TP inhibits the expression of RUNX2, osterix, and IRF8 through hypermethylation of their CGIs.

(A) Schematic diagrams of CpG-rich test regions on the promoter of RUNX2 or osterix in human MSCs and on the promoter of IRF8 in human preOCs. The arrow indicates the translation-initiating ATG site. The CpG-rich test region is marked with a horizontal bar. TSS, transcription start site. (B to E) MSCs or preOCs were cocultured with myeloma cells ARP-1 [WT, nontargeted shRNA (shCtrl), and TP shRNA (shTP)] and MM.1S [WT, control vector (Vec), and TP cDNA (TP)] in their respective medium for 7 days. After cultures, bisulfite-treated genomic DNA was subjected to MSP or BSP analysis. (B) DNA gel electrophoresis shows the unmethylated (U) and methylated (M) PCR products from MSP analysis. Sequencing results from BSP analysis shows percentage of methylation in the promoter of RUNX2 or osterix in MSCs and in the promoter of IRF8 in preOCs cocultured with myeloma cell lines ARP-1 and MM.1S (C), shCtrl or shTP ARP-1 cells (D), or Vec or TP MM.1S cells (E). Cultured MSCs or preOCs without myeloma (No MM) served as a control. Data in (C) to (E) are individual cultures with averages ± SD (n = 5) of three experiments. P values were determined by Student’s t test. (F) Summary of BSP analysis showing percentage of methylation in the promoter of RUNX2 or osterix in MSCs and in the promoter of IRF8 in preOCs, of healthy donors and TPhigh or TPlow patients. Data are individual patients’ samples with averages ± SD (n = 5) of three experiments. P values were determined by Student’s t test.

DNA methyltransferases (DNMTs) are important for the methylation of gene promoters. We observed higher gene expression and enzyme activity of DNMT3A, but not DNMT1 or DNMT3B, in MSCs and preOCs cocultured with high-TP myeloma cells (wild-type ARP-1, shCtrl ARP-1, and TP MM.1S) than in those cocultured with low-TP cells (shTP ARP-1, wild-type MM.1S, and Vec MM.1S) (fig. S6, A and B). The clinical relevance was further studied by correlating TP expression in myeloma cells from patients or number of bone lesions in these patients with DNMT3A expression in MSCs and preOCs from the same patients. We found positive correlations between TP expression and number of bone lesions with DNMT3A in MSCs or preOCs (fig. S6C). Collectively, these data demonstrate that myeloma-expressed TP up-regulates DNMT3A expression and activity in MSCs and preOCs, consequently enhancing methylation of CGIs in the promoters of RUNX2, osterix, and IRF8 genes, suppressing osteoblast differentiation, and increasing osteoclast activity. These signaling changes shift the balance of bone remodeling toward a net loss of bone tissue.

TP up-regulates DNMT3A expression through 2DDR-mediated signaling

TP degrades thymidine into thymine and 2DDR1P; the unstable 2DDR1P is further dephosphorylated to 2DDR (11). As expected, in vitro myeloma cells with high levels of TP secreted more 2DDR than did low-TP cells (fig. S7A). Similarly, higher levels of 2DDR were observed in the serum of mice bearing high-TP myeloma cells (TP MM.1S and wild-type ARP-1) than that of mice bearing low-TP cells (wild-type MM.1S and shTP ARP-1) (fig. S7B).

To test whether 2DDR mediates TP-induced suppression of osteoblastogenesis, we cultured MSCs in osteoblast medium with incremental doses of 2DDR for 2 weeks. The addition of 2DDR down-regulated ALP production, Alizarin red S staining, and the expression of osteoblast genes in a dose-dependent manner (fig. S7C). Furthermore, the addition of 2DDR decreased the expression of RUNX2 and osterix (Fig. 5A) and up-regulated the expression and activity of DNMT3A in MSCs (Fig. 5B). The culture of MSCs in medium with 2DDR induced the hypermethylation of CGIs on the RUNX2 and osterix genes (Fig. 5C) and inhibited ALP production and Alizarin red S staining (Fig. 5D). Knockdown of DNMT3A in MSCs partially abrogated the effects of 2DDR (Fig. 5, C and D).

Fig. 5. 2DDR inhibits osteoblast differentiation and activates osteoclast differentiation by up-regulating DNMT3A expression.

Human MSCs or preOCs were cultured in medium without (0) or with 0.5, 1, or 2 mM 2DDR for 48 hours. In some studies, MSCs or preOCs carried with nontargeted shRNAs (shCtrl) or DNMT3A shRNAs (shDNMT3A) were cultured with phosphate-buffered saline (PBS) or 1 mM 2DDR. (A to D) Expression of RUNX2 and osterix (A), DNMT3A mRNA expression and activity (B), methylation of CGIs in the promoter regions of RUNX2 and osterix (C), and ALP activity and Alizarin red S staining (D) in MSC-derived cells after 2DDR treatment. OD490, optical density at 490 nm. (E to H) Expression of IRF8 (E), DNMT3A mRNA expression and activity (F), methylation of CGIs in the promoter region of IRF8 (G), and the number of multiple nuclear (≥3) TRAP+ cells and secretion of TRAP5b (H) in preOC-derived cells after 2DDR treatment. mRNA expression was normalized to cells without 2DDR (set at 1). The levels of β-actin served as loading controls. Data are averages ± SD (n = 3). P values were determined by Student’s t test. Each experiment was repeated three times.

To examine the role of 2DDR in the activation of osteoclastogenesis, we cultured preOCs with RANKL (10 ng/ml) and M-CSF (25 ng/ml) without or with escalating doses of 2DDR for 1 week. 2DDR enhanced multinuclear TRAP+ cell numbers, TRAP5b secretion, and the expression of the osteoclast genes TRAP, CALCR, and CTSK in a dose-dependent manner (fig. S7D). 2DDR also reduced IRF8 expression (Fig. 5E), up-regulated DNMT3A expression (Fig. 5F), and induced hypermethylation of CGIs in IRF8 (Fig. 5G). Knocking down DNMT3A in preOCs reversed 2DDR-mediated IRF8 hypermethylation (Fig. 5G) and increased osteoclast formation (Fig. 5H).

We next investigated the signaling pathways by which 2DDR may regulate DNMT3A expression. 2DDR is known to bind integrins αVβ3 and α5β1 (11). We confirmed in the literature that all MSCs in this study express both integrins (24) and that preOCs express αVβ3 (25). The addition of an antibody against either the α5 subunit of α5β1 or the αV subunit of αVβ3, but not control immunoglobulin G, to the coculture of MSCs with 2DDR enhanced osteoblast formation; the addition of both anti-α5 and anti-αV antibodies had a synergistic effect (fig. S8A). Moreover, application of both antibodies inhibited 2DDR-stimulated DNMT3A expression in MSCs (fig. S8B), indicating that 2DDR regulates osteoblast differentiation and DNMT3A expression via αVβ3 and/or α5β1.

Furthermore, we examined the extracellular signal–regulated kinase (ERK) and phosphoinositide 3-kinase (PI3K)/Akt signaling pathways downstream of α5β1 and αVβ3. The addition of 2DDR to the culture of MSCs up-regulated the phosphorylated levels of Akt and the integrin-mediated downstream molecules focal adhesion kinase (FAK), paxillin, and p130Cas but did not change phosphorylated levels of ERK or nonphosphorylated levels of these molecules (fig. S8C). Akt phosphorylation induced by 2DDR was abrogated with antibodies against both α5 and αV (fig. S8D). An Akt inhibitor, LY294002, blocked 2DDR-induced DNMT3A expression in MSCs (fig. S8E).

Addition of anti-αV antibodies to the culture of preOCs reduced 2DDR-induced TRAP5b secretion and DNMT3A expression (fig. S8, F and G). Addition of 2DDR up-regulated the levels of phosphorylated Akt, FAK, paxillin, and p130Cas in cultured preOCs (fig. S8H), and blocking αV with the antibodies (fig. S8I) reduced the phosphorylation of Akt. Similar to the results in MSCs, administration of LY294002 significantly reduced 2DDR-induced DNMT3A expression in preOCs (fig. S8J).

To validate the antibody blocking study, we knocked down αV and/or α5 expression in MSCs using small interfering RNAs (siRNAs) (fig. S9A). Adding 2DDR to cultures of siαV or siα5 MSCs reduced osteoblast formation, DNMT3A expression, and Akt phosphorylation in MSCs (fig. S9, B to D). The siRNAs against Akt1/2 blocked 2DDR-induced DNMT3A expression in MSCs (fig. S9, E and F). Moreover, knockdown of αV expression in preOCs reduced 2DDR-induced TRAP5b secretion and DNMT3A expression (fig. S9, G to I), and knockdown of Akt1/2 (fig. S9J) significantly reduced 2DDR-induced DNMT3A mRNA (fig. S9, K and L). These results suggest that TP regulates 2DDR secretion from myeloma cells, and 2DDR enhances DNA methylation of RUNX2 and osterix in MSCs and IRF8 in preOCs through the αVβ35β1-PI3K/Akt-DNMT3A signaling pathway.

Inhibiting TP reduces myeloma-induced osteolytic bone lesions

Toward a therapeutic approach, we asked whether inhibiting TP can prevent myeloma-induced osteolytic bone lesions. For this purpose, ARP-1 cells, which express high levels of TP, were directly injected into the femurs of SCID mice. Mouse serum was collected to measure circulating M protein levels for monitoring tumor burden. When myeloma was established, mice were treated with vehicle control or TP inhibitors 7-deazaxanthine (7DX) or tipiracil hydrochloride (TPI). ARP-1 cells caused osteolytic bone lesions, increased osteoclastogenesis, and reduced bone volume and osteoblastogenesis (Fig. 6, A to D). TPI or 7DX treatment significantly reduced ARP-1–induced bone lesions (Fig. 6, A to D). Treatment with 7DX or TPI significantly reduced Dnmt3a expression in mouse MSCs or preOCs (Fig. 6E) and 2DDR levels in the serum of ARP-1 tumor-bearing mice (Fig. 6F).

Fig. 6. Administration of TP inhibitor in myeloma-bearing mice reduces bone lesions and osteoclastogenesis and enhances osteoblastogenesis.

ARP-1 cells were injected into the femurs of SCID mice. Mice without myeloma cells served as controls (No MM). After 3 weeks, mice were treated with PBS as vehicle controls or TP inhibitor 7DX (200 μg/kg) or TPI (300 μg/kg). After treatment, mice were scanned for radiography, and mouse femurs were subjected to toluidine blue staining or TRAP staining. (A) Representative x-ray images of mouse femurs. (B to D) Percentage of bone volume to total volume (BV/TV) (B), percentage of bone surface eroded by osteoclasts (ES/BS) and of bone surface covered with osteoclasts (Oc. S/BS) (C), and percentage of osteoid surface (OS/BS) and of bone surface lined with osteoblasts (Ob. S/BS) (D). Data are averages ± SD (n = 5 mice per group, three replicate studies). (E) Dnmt3a mRNA expression in murine MSCs and preOCs isolated from BM aspirates of ARP-1–bearing mice. Data are averages relative to no MM–bearing mice (No MM) treated with vehicle (set at 1) ± SD (n = 5 mice/group, three replicate studies). (F) 2DDR levels in the serum of ARP-1–bearing mice. Data are averages relative to that in no MM–bearing mice (No MM) treated with vehicle (set at 1) ± SD (n = 5 mice per group, three replicate studies). All P values were determined by Student’s t test. (G) Depiction of signaling pathways involved in the myeloma TP-mediated suppression of osteoblastogenesis and activation of osteoclastogenesis. OB, osteoblast.

Using two additional human myeloma cell lines, RPMI 8226 (high TP expression) and U266 (low TP expression), we observed similar effects of TP inhibitors on myeloma-induced bone lesions both in vitro and in vivo (Fig. 7). The coculture of RPMI 8226 cells secreting more 2DDR (Fig. 7A) increased osteoclast (Fig. 7B) and suppressed osteoblast (Fig. 7C) differentiation and activity, compared to those in the coculture of U266 cells (secreting less 2DDR). Treatment of RPMI 8226–bearing mice with TPI or 7DX enhanced bone volume and osteoblastogenesis and reduced osteoclastogenesis (Fig. 7, D to F).

Fig. 7. TP expressed by myeloma cells regulates osteoblast and osteoclast differentiation in vitro and in vivo.

(A) Relative levels of 2DDR in human myeloma cell lines RPMI 8226 or U266 were measured after 48 hours of culture. Data are relative to RPMI 8226. Data are averages ± SD (n = 3) of three experiments. (B) PreOCs were cultured alone or cocultured with RPMI 8226 or U266 cells in medium without or with RANKL (10 ng/ml) for 1 week. PreOCs alone without or with RANKL (10 ng/ml) served as controls. Numbers of multinuclear (≥3) TRAP+ cells and the relative expression of osteoclast differentiation–associated genes TRAP, CALCR, and CTSK were measured. mRNA expression was normalized to cells without myeloma (No MM, set to 1). Data are averages ± SD (n = 3) of three experiments. (C) MSCs were cocultured with RPMI 8226 or U266 in osteoblast medium for 2 weeks and then stained with Alizarin red S. The relative expression of osteoblast differentiation–associated genes BGLAP, ALP, and COL1A1 was determined in attached cells. mRNA expression was normalized to cells without myeloma (No MM, set to 1). Data are averages ± SD (n = 3) of five experiments. (D to F) RPMI 8226 cells (5 × 105 cells per mouse) were injected into the femurs of SCID mice. Mice without myeloma cell injection served as controls (No MM). After 3 weeks, mice were intraperitoneally injected with PBS as vehicle control or the TP inhibitors 7DX (200 μg/kg) or TPI (300 μg/kg) three times per week for 2 weeks. After treatment, mice were scanned for radiography, and mouse femurs were subjected to toluidine blue staining or TRAP staining. We calculated the percentage of bone volume to total volume (BV/TV) (D), the percentage of bone surface eroded by osteoclasts (ES/BS) and of bone surface covered with osteoclasts (Oc. S/BS) (E), and the percentage of osteoid surface (OS/BS) and of bone surface lined with osteoblasts (Ob. S/BS) (F). Data are averages ± SD (n = 5 mice per group, three replicate studies). All P values were determined by Student’s t test.

DISCUSSION

Our study reveals an important biological function of TP in the pathogenesis of myeloma-associated osteolytic bone lesions (Fig. 6G) and indicates that counteracting TP activity may be effective for prevention or treatment of osteolytic bone lesions in myeloma patients. We found that TP reversibly catalyzes conversion of thymidine into thymine and 2DDR. Myeloma-secreted 2DDR binds to integrins αVβ3 and α5β1 in osteoblast progenitors, activates PI3K/Akt signaling, and increases DNMT3A expression and methylation of RUNX2 and osterix, leading to decreased osteoblastogenesis. The secreted 2DDR also binds to integrin αVβ3 in osteoclast progenitors, activates PI3K/Akt signaling, and increases DNMT3A expression and methylation of IRF8, leading to increased NFATc1 expression and osteoclastogenesis. The net effect of TP is to suppress osteoblast-mediated bone formation and activate osteoclast-mediated bone resorption, the hallmarks of myeloma-induced bone disease.

DNA methylation regulation has been shown to be one of the most important regulatory mechanisms of gene expression (26). For instance, CpG methylation is catalyzed by a family of DNMTs (27), and DNMT3A and DNMT3B specifically are able to modify unmethylated DNAs. However, the involvement of DNA methylation in osteoblast and osteoclast differentiation is unclear. Our results demonstrated that myeloma-expressed TP up-regulated DNMT3A, but not DNMT1 or DNMT3B, expression in MSCs via 2DDR, the metabolic product of thymidine, leading to highly methylated CGIs in RUNX2 and osterix gene promoters. Hypermethylation caused down-regulation of these two genes, resulting in decreased osteoblast-mediated bone formation, a hallmark of myeloma. In addition, myeloma-expressed TP enhanced osteoclast differentiation. TP/2DDR down-regulated IRF8 expression in preOCs through DNA hypermethylation by DNMT3A, thereby enhancing RANKL-induced NFATc1 expression, demonstrating a critical role of DNA methylation in myeloma-induced activation of osteoclastogenesis. These results point to a new epigenetic mechanism underlying myeloma-induced bone disease.

Integrins are important for bone remodeling (28). For example, integrins α5β1 and αVβ3, which are expressed in MSCs and mature osteoblasts, have been shown to promote osteoblast differentiation through FAK-ERK1/2 signaling and maintain osteoblast survival through activation of PI3K/Akt signaling (24). In addition, αVβ3 expressed in preOCs and mature osteoclasts has been implicated in the regulation of osteoclast attachment (28). The αVβ3 is also essential for osteoclast-mediated matrix degradation, an effect that occurs through collaboration of c-Fms with the ERK/c-Fos signaling pathway. Hotchkiss et al. suggested that activation of α5β1 and αVβ3 and downstream signaling pathways by TP and 2DDR affects endothelial cell migration (29), leading us to examine the role of integrins in TP-mediated bone remodeling. We found that TP/2DDR enhances osteoclast differentiation by interaction with preOC-expressed αVβ3 and inhibits osteoblast differentiation by interacting with MSC-expressed α5β1 and αVβ3. The interaction of 2DDR with these integrins activated the PI3K/Akt signaling pathway, but not the ERK1/2 signaling pathway, leading to DNMT3A expression. Collectively, these results demonstrate that αVβ35β1-mediated signaling plays an important role in bone remodeling through the effects of TP on regulation of osteoclastogenesis and osteoblastogenesis.

There are limitations to this study. Bone remodeling is a complex event in which multiple factors, regulating both formation and resorption, collaborate to maintain bone stability. In myeloma, several other factors, including RANKL, DKK1, macrophage inflammatory protein–1α, and stromal cell–derived factor–1, are known to be dysregulated, and it is not clear how TP may interact with these factors. In addition, although our results with TP inhibitors in animal models show promising results, questions remain about their efficacy in human disease. Because these models operate in a murine bone marrow microenvironment, which could be somewhat different from human marrow, it is not clear whether TP inhibitors will be effective in the human marrow environment. Despite these concerns, we think that it is reasonable to extend these studies to humans because TP inhibitors are now U.S. Food and Drug Administration (FDA)–approved for other purposes. In 2015, a TP inhibitor used in this study, TPI, combined with a nucleoside metabolic inhibitor (trifluridine) was approved by the FDA for treatment of patients with metastatic colorectal cancer (30). In addition, several DNMT inhibitors are now in clinical trials for cancer treatment: azacitidine is in phase 3 trials for myelodysplastic syndromes and acute myeloid leukemia (AML), and decitabine is in phase 3 trials for AML and chronic myelogenous leukemia (31).

The therapeutic effect of these drugs in myeloma or myeloma-induced bone disease remains unknown, but our data in cell lines, patient cells, and rodent models make a compelling case for a role of TP in the genesis of myeloma-induced bone disease and, thus, encourage repurposing of these inhibitors. An in-depth analysis of the therapeutic efficacy of TP inhibitors in the SCID-hu mouse model, an excellent model for myeloma translational research because it provides a human-like marrow microenvironment, would be a logical next step. This model could facilitate the translation of these inhibitors into human studies of myeloma bone disease. Because TP is often expressed by other malignancies including breast, prostate, and lung cancer, these findings may also have broader implications for the genesis of bone metastasis caused by these and other tumors.

MATERIALS AND METHODS

Study design

We hypothesized that TP induces bone destruction based on an observation of high TP expression in tumor cells (12, 13). In a myeloma setting, this study was designed to evaluate the relationship between TP expression and cancer-associated bone lesions. It encompassed three main objectives: to determine the role of TP in myeloma-induced bone lesions, to elucidate the mechanism of TP-induced bone lesions, and to validate the mechanisms in vivo and in vitro using mouse models and patient samples, respectively. In the first objective, we separated all tested primary myeloma cells and human myeloma cell lines into high-TP– and low-TP–expressing cells, injected these cells into mice, and assessed osteoclast-mediated bone resorption and osteoblast-mediated bone formation by radiography and bone histomorphometry. We also knocked down or overexpressed TP in myeloma cells to assess whether modulating TP expression affects bone formation/resorption. TRAP and Alizarin red S staining in the coculture of myeloma cells with MSCs or preOCs determined the importance of TP to bone cell differentiation and activity. In the second aim, MSP and BSP were performed to analyze DNA methylation in the promoters of RUNX2 and osterix in MSCs or in the promoter of IRF8 in preOCs. We also examined DNMT3A and 2DDR expression and the integrin-PI3K/Akt signaling pathway in MSCs or preOCs. In the third objective, we evaluated the correlations among TP expression, DNMT3A expression, and bone lesions using samples from randomly selected myeloma patients. In addition, we confirmed the mechanism using the mouse models. Myeloma-bearing mice were randomly selected for the treatment with two TP inhibitors.

All patient samples were obtained from the Myeloma Tissue Bank of the University of Texas MD Anderson Cancer Center. Bone lesions in humans were characterized by a radiologist who was blinded to the severity of clinical bone disease. The number of samples required to achieve a correlation coefficient (r) ≥ 0.7, a power of 80%, and the level of significance at 5% was determined to be at least seven samples. In mouse studies, the sample size, the composition of replicates, and the intermediate end point were based on previous knowledge of the mouse models (14). The sample size was initially estimated using power analysis with our previous knowledge on the bone histomorphometric analysis in myeloma-bearing mice. We needed five mice per group to ensure a power of 80% to detect the changes in bone volume between the different TP expression groups with two-sided type I error rate controlled at the 0.05 level.

The final end point before sacrifice was in accordance with the Institutional Animal Care and Use Committee policies and was predefined. All data were included in the analysis, and the criteria for interpretation were established prospectively. Experiments were performed three to five times (as indicated in the figure legends). Animal results were verified by repetition over a 3-year period.

Cell lines and primary cells

Primary myeloma cells were isolated from the bone marrow aspirates of newly diagnosed myeloma patients using anti-CD138 antibody–coated magnetic beads (Miltenyi Biotec Inc.). The cells were maintained in RPMI 1640 medium with 10% fetal bovine serum. Normal plasma cells were isolated from the peripheral blood of healthy donors, as previously described (18). MSCs and monocytes from myeloma patients were isolated from bone marrow aspirates and cultured as previously described (14). This study was approved by the Institutional Review Board of the University of Texas MD Anderson Cancer Center.

DNMT3A activity analysis

Nuclear extracts were isolated using the EpiQuik Nuclear Extraction Kit (EpiGentek), and 3 μl of nuclear extracts from cells was added to each reaction well according to the manufacturer’s protocol. DNMT3A activity was measured using the EpiQuik DNA Methyltransferase Activity/Inhibition Assay Kit (EpiGentek) as described previously (32).

Measurement of 2DDR levels

The relative levels of 2DDR in culture medium and mouse serum were measured as described previously (33). Briefly, the samples were degraded in 1.0 M HCl at 80°C, and the absorbance at 261 and 277 nm was taken. The concentration was determined on the basis of the calibration curve.

Mouse models

C.B-17 SCID mice purchased from Harlan Laboratories were maintained in American Association of Laboratory Animal Care–accredited facilities. The mouse studies were approved by the Institutional Animal Care and Use Committee of the University of Texas MD Anderson Cancer Center. Cultured myeloma cells (5 × 105 cells per mouse) were injected into the femurs of 6- to 8-week-old SCID mice (34). SCID-hu hosts were established as reported previously (14). Human fetal bone chips were implanted subcutaneously into the right flanks of the mice. Primary myeloma cells (1 × 106 cell per mouse) were injected into the implanted human bone chips to establish myeloma. In some experiments, PBS or 7DX (200 μg/kg) or TPI (300 μg/kg) was injected into the peritoneum of mice three times a week for 2 weeks, beginning 3 weeks after myeloma cell injection.

Monitoring tumor burden and bone lesions in mice

Sera were collected from the mice weekly and were tested for myeloma-secreted M proteins or light chains by enzyme-linked immunosorbent assay or for 2DDR levels by spectrophotometric analysis. To measure the size of lytic bone lesions, radiographs were scanned with a Faxitron cabinet X-ray system.

Bone histomorphometry

SCID mouse femurs or human bone chips of SCID-hu mice were fixed in 10% neutral-buffered formalin and were decalcified, and the bone sections were stained with toluidine blue or TRAP following standard protocols. To assess dynamic histomorphometric indices, mice were given two injections of calcein (20 mg/kg; Sigma-Aldrich) at 3 and 6 days before dissection. The paraffin-fixed femurs were embedded and sectioned. Both analyses were carried out using BIOQUANT OSTEO software (BIOQUANT Image Analysis Co.). Mouse MSCs and monocytes were isolated and cultured as previously described (35, 36).

Statistical analysis

In directly comparing two sets of quantitative data, statistical significance was determined with SPSS software (version 10.0) using unpaired Student’s t tests. P < 0.05 was considered statistically significant. All results were reproduced in at least three independent experiments.

SUPPLEMENTARY MATERIALS

www.sciencetranslationalmedicine.org/cgi/content/full/8/353/353ra113/DC1

Methods

Fig. S1. Modulation of TP expression does not affect the growth and survival of myeloma cells.

Fig. S2. Myeloma-expressed TP enhances RANKL-mediated osteoclast differentiation and activity in vitro.

Fig. S3. Myeloma-expressed TP enhances NFATc1 expression and activity through inhibition of IRF8.

Fig. S4. Myeloma-expressed TP inhibits osteoblast differentiation and activity in vitro.

Fig. S5. TP inhibits the expression and activities of RUNX2 and osterix in vitro.

Fig. S6. Myeloma-expressed TP enhances DNMT3A levels in MSCs and preOCs.

Fig. S7. Myeloma cells with high TP expression secrete more 2DDR, which affects osteoclast and osteoblast differentiation in vitro.

Fig. S8. 2DDR up-regulates DNMT3A through α5β1Vβ3-PI3K/Akt signaling pathways.

Fig. S9. Knockdown of integrins or Akt1/2 abrogates the effects of 2DDR on DNMT3A expression.

Table S1. Primers for reverse transcription PCR and quantitative PCR.

Table S2. Primers for chromatin immunoprecipitation PCR.

Table S3. Primers for MSP and BSP.

REFERENCES AND NOTES

Acknowledgments: We thank the staff at MD Anderson’s Small Animal Imaging Facility for their assistance with x-ray imaging. Funding: This work was supported by the National Cancer Institute R01s (CA190863 and CA193362), the National Cancer Institute UTMDACC SPORE in Multiple Myeloma Career Development Award (CDP-060315) and Developmental Research Program (DRP-00013585), the American Cancer Society Research scholar grant (127337-RSG-15-069-01-TBG), the University of Texas MD Anderson Cancer Center Institutional Research Grants Basic Research, the Leukemia Research Foundation, the American Society of Hematology, and the National Natural Science Foundation of China (grant no. 81470356). Author contributions: H.L. and J.Y. designed all experiments and wrote the manuscript; H.L., Z.L., J.D., and J. He performed all experiments and statistical analyses; P.L. and J.J.S. provided the myeloma patients’ samples; R.E.D., R.F.G., J. Hou, and B.A. provided critical suggestions; and M.W.S. and N.N. provided the injection technology and osteodata analysis. All authors reviewed the final manuscript. Competing interests: The authors declare that they have no competing financial interests. Data and materials availability: Myeloma cell lines ARP-1, CAG, and ARK were provided by the University of Arkansas for Medical Sciences. Modified cell lines can be obtained from the corresponding authors through a material transfer agreement.
View Abstract

Navigate This Article