COX2 confers bone marrow stromal cells to promoting TNFα/TNFR1β-mediated myeloma cell growth and adhesion

Abstract

Purpose

Bone marrow stromal cells (BMSCs) have been implicated in multiple myeloma (MM) progression. However, the underlying mechanisms remain largely elusive. Therefore, we aimed to explore key factors in BMSCs that contribute to MM development.

Methods

RNA-sequencing was used to perform gene expression profiling in BMSCs. Enzyme-linked immunosorbent assays (ELISAs) were performed to determine the concentrations of PGE2 and TNFα in sera and conditioned media (CM). Western blotting, qRT-PCR and IHC were used to examine the expression of cyclooxygenase 2 (COX2) in BMSCs and to analyze the regulation of TNFα by COX2. Cell growth and adhesion assays were employed to explore the function of COX2 in vitro. A 5T33MMvt-KaLwRij mouse model was used to study the effects of COX2 inhibition in vivo.

Results

COX2 was found to be upregulated in MM patient-derived BMSCs and to play a critical role in BMSC-induced MM cell proliferation and adhesion. Administration of PGE2 to CM derived from BMSCs promoted MM cell proliferation and adhesion. Conversely, inhibition of COX2 in BMSCs greatly compromised BMSC-induced MM cell proliferation and adhesion. PCR array-based analysis of inflammatory cytokines indicated that COX2 upregulates the expression of TNFα. Subsequent rescue assays showed that an anti-TNFα monoclonal antibody could antagonize COX2-mediated MM cell proliferation and adhesion. Administration of NS398, a specific COX2 inhibitor, inhibited in vivo tumor growth and improved the survival of 5TMM mice.

Conclusions

Our results indicate that COX2 contributes to BMSC-induced MM proliferation and adhesion by increasing the secretion of PGE2 and TNFα. Targeting COX2 in BMSCs may serve as a potential therapeutic approach of treating MM.

Supplementary Information

The online version contains supplementary material available at 10.1007/s13402-021-00590-4.

Introduction

Multiple myeloma (MM) is a plasma cell malignancy characterized by the production of monoclonal proteins [1]. Both genetic alterations and the tumor microenvironment (TME) are essential for MM progression. Recurrent translocations involving the immunoglobulin heavy chain (IgH) locus are found in about 50% of MM patients, including Cyclin D1 t(11;14), FGFR3 t(4;14) and MAFB t(14;16) translocations [2, 3]. In addition, it has been reported that the bone marrow microenvironment plays a crucial role in supporting MM cell growth and survival, as well as the acquisition of therapy resistance [4–6].

The cellular composition of the bone marrow is complex and includes BMSCs, osteoblasts, osteoclasts, endothelial cells, immunocytes and adipocytes. Together, these components constitute a TME that favors the growth of MM cells [7]. BMSCs constitute a heterogeneous population of cells that reside within the bone marrow, which also contains their main progenitor-mesenchymal stem cells. Importantly, BMSC-mediated survival signaling or BMSC-mediated drug resistance have been identified as critical factors for MM pathogenesis [8]. According to previous studies, BMSCs exert their effects on MM cells mainly through two routes. Firstly, BMSCs have been shown to secrete a series of factors such as interleukin-6 (IL6) [9], insulin-like growth factor-1 (IGF-1) [10] and stromal-derived factor 1 alpha (SDF-1α) [11], which promote proliferation and resistance to conventional chemotherapeutic agents. Furthermore, BMSC-derived exosomes containing those factors support growth or confer drug resistance in MM cells [12]. Secondly, direct contact of MM cells with BMSCs may be essential for the acquisition of adhesion-mediated drug resistance [13]. However, there is still a lack of comprehensive information on BMSC-derived factors in MM progression. Thus, whether and how other factors in BMSCs contribute to the progression of MM requires further analysis.

COX2, a well-documented inflammation mediator, is often overexpressed in solid tumors and hematologic malignancies, including colorectal cancer, breast cancer and MM [14–17]. More recently, it has been reported that high expression of COX2 in cancer-associated fibroblasts (CAFs) contributes to cell migration and invasiveness in nasopharyngeal carcinoma [18]. Previous studies reported that COX2 may serve as a marker of poor prognosis in MM, and that silencing of COX2 may induce growth inhibition of MM cells [19]. As yet, however, stromal expression of COX2 has not been specifically evaluated in MM. Although COX2 is known to mediate inflammatory responses within the bone marrow microenvironment, it is still unclear whether COX2 affects interaction of BMSCs with MM cells to facilitate MM growth or drug resistance.

In this study, we found that COX2 is highly expressed in MM BMSCs using RNA-seq analysis. We subsequently determined the effects of COX2-dependent BMSCs on MM progression in vitro and in vivo. Our results indicate that COX2 may serve as a reliable biomarker for predicting the effect of BMSCs in MM and as novel therapeutic target.

Materials and methods

Cell culture, reagents and plasmids

Human MM cell lines MM1S and KMS11, and mouse MM cell line 5T33MMvt were cultured in RPMI-1640 supplemented with 10% FBS (Gibco, Grand Island, NY, USA), 100 units/ml penicillin and 100 μg/ml streptomycin (P/S). 293 T cells were purchased from the American Type Culture Collection (ATCC, Manassas, VA, USA) and grown in Dulbecco-modified minimum essential medium (DMEM) high glucose supplemented with 10% FBS and P/S. The human BM stromal cell line HS5 was maintained in DMEM low glucose supplemented with 10% FBS and P/S. All cells were cultured at 37 °C in a 5% CO₂ incubator.

Human BMSCs (hBMSCs) were derived from fresh bone marrow from healthy bone marrow transplantation donors (HD, n = 5) and newly diagnosed MM patients (n = 6). BMSC isolation was performed as described previously [20]. Briefly, BM mononuclear cells were collected via density gradient centrifugation with Ficoll–Hypaque and thereafter cultured in DMEM/F12 medium for about 3 weeks until the formation of a monolayer of cells morphologically similar to bone marrow fibroblast cells. Specimens were obtained with written informed consent from MM patients enrolled in the first and the third affiliated Xiangya Hospital of Central South University (CSU). The hBMSCs used in functional studies were cultured for less than five passages.

NS398, PGE2, AH6809 (EP2 antagonist) and AH23848 (EP4 antagonist) were purchased from Cayman Chemicals (MI, USA), and Bortezomib (BTZ) and FH535 were purchased from Selleck Chemicals (Houston, TX, USA). Human anti-COX2 antibody was purchased from Cell Signaling (Cell Signaling Technology, MA, USA). Human anti-αSMA and anti-CD138 antibodies were purchased from Abcam (Abcam, MA, USA). Anti-p-p65, anti-p65, anti-β-actin anti-GAPDH antibodies, and HRP-conjugated IgG secondary antibodies were purchased from Santa Cruz Biotechnology (Santa Cruz, CA, USA). Anti-EP2, anti-EP4, anti-TNFα, anti-CSF2, anti-TNFR1α and anti-TNFR1β antibodies were purchased from ABclonal (ABclonal Biotech Co, Hubei, China).

Plasmids expressing wild-type K-Ras and N-Ras were cloned by inserting human K-Ras and N-Ras cDNAs into a cFlag-pcDNA4 vector. Next, pcDNA4 plasmids expressing mutant K-Ras/G12V and N-Ras/G12V (in which the glycines at positions 12 are replaced by valine) were constructed with specific primers using a Fast Mutagenesis System (TransGen Biotech, Beijing, China). To construct a lentiviral COX2 expression vector, COX2 cDNA was amplified by PCR with specific primers and inserted into a pLVX-IRES-ZsGreen1 vector. The primers used are listed in Additional file 1: Table S1.

Conditioned medium derived from BMSCs

BMSCs and HS5 cells (2 × 10⁵ cells/ml) were seeded into 60-mm culture dishes in DMEM/F12 or DMEM low glucose medium supplemented with 10% FBS and cultured overnight, and subsequently refreshed with 5 ml serum-free RPMI-1640 medium. After the culture supernatants were harvested 48 h later, cell debris was removed with centrifugation and the supernatants were stored at −80 °C until experimentation. To obtain conditioned medium (CM) of BMSCs with or without PGE2 stimulation, BMSCs derived from HDs, BMSCs from COX2 knockout mice (COX2^−/− BMSCs) and HS5-COX2-sh cells were treated with 10 μM PGE2 or PBS. To obtain CM of BMSCs treated with or without NS398, BMSCs derived from MM patients, BMSCs from COX2 heterozygous mice (COX2^+/− BMSCs) and HS5-Ctrl cells were cultured in serum-free RPMI-1640 with 20 μM NS398 or PBS. The CMs from these cultures were assayed for their capacity to stimulate proliferation, adhesion and drug resistance to BTZ of MM cells.

Patient samples

Six fresh bone marrow samples of MM patients were obtained from the first and the third affiliated Xiangya Hospital, CSU, Changsha, China. Serum samples of HD (n = 11) and newly diagnosed MM patients (n = 17) were obtained from the Chinese Academy of Medical Science & Peking Union Medical College, Tianjin, China. Serum samples were prepared for detection of PGE2 and TNFα using an ELISA Kit. Paraffin-embedded bone marrow tissues from HD (n = 31) and MM patients (n = 31) were collected from the Hunan Tumor Hospital (Changsha, China) and the second affiliated Xiangya Hospital, CSU. Quantitative immunofluorescence and immunocytochemistry were performed to detect COX2 expression in BMSCs. Written informed consent was obtained from all participants. All studies with human samples were approved by the Medical Ethics Committee of CSU and the Chinese Academy of Medical Sciences. Patient characteristics are summarized in Additional file 1: Table S2.

Quantitative immunofluorescence and immunocytochemistry (IHC)

For multiplexed quantitative immunofluorescence, α-SMA, COX2, CD138 and 4′,6-diamidino-2-phenylindole (DAPI) were simultaneously quantified on the same slide for every patient. Briefly, slides were deparaffinized and rehydrated before undergoing antigen retrieval using sodium citrate buffer (pH = 8). Next, the slides were incubated with 0.3% bovine serum albumin for 30 min to block non-specific antigens and probed with COX2 antibody overnight at 4 °C. After the slides were incubated with a HRP-conjugated secondary antibody, Opal Fluorophores were added according to manufacturer’s instructions (PerkinElmer, Waltham, MA, USA). The slides were subsequently stained for CD138 and α-SMA. Nuclei were tagged with DAPI (Life Technologies, Carlsbad, CA, USA). For IHC, the procedure was performed as previously reported [21, 22]. The processes before primary antibody incubation were performed the same as for multiplexed quantitative immunofluorescence. Next, the slides were incubated with HRP-conjugated secondary antibody and stained with 3,3′-diaminobenzidine tetrahydrochloride hydrate (DAB) for 10 min. Finally, cell nuclei were counterstained with hematoxylin. The stained sections were evaluated independently by two pathologists who were blinded to the clinical parameters.

Enzyme-linked immunosorbent assay (ELISA)

PGE2 and TNFα levels in the CM from BMSCs and patients’ serum were measured using prostaglandin E₂ ELISA kits (Cayman Chemicals, MI, USA) and TNFα ELISA kits (R&D Systems, Wiesbaden, Germany), respectively, according to the manufacturer’s instructions. Mouse IgG2b circulating levels were determined using specific ELISA kits (Bethyl Laboratories, Inc., Montgomery, TX, USA). Standard curves were processed in parallel for individual experiments to achieve precise quantification of sample concentrations.

Quantitative reverse transcriptase-PCR (qRT-PCR)

Total RNA was isolated from BMSCs and HS5 cells using Trizol reagent (Invitrogen-Life Technologies, CA, USA) as instructed by the manufacturer’s protocol. Next, complementary DNAs were synthesized from 2 μg total RNA using a SuperScript III First-Strand Synthesis SuperMix (Thermo Fisher Scientific, MA, USA) according to the manufacturer’s protocols. Amplification of specific targets was performed to determine mRNA levels using a Bio-Rad iCycler iQ Real-Time PCR Detection System (Bio-Rad, CA, USA). GAPDH or β-actin were used as loading controls. Relative mRNA levels were calculated as the value of 2ΔCt normalized to the respective control.

Western blot analysis

Western blotting was performed as previously described [23, 24]. BMSCs derived from HD and MM patients, and MM cells treated with CMs were lysed in RIPA buffer containing protease inhibitor and phosphatase inhibitor cocktails (Roche, Basel, Switzerland) on ice, after which protein levels were quantified using a BCA protein assay kit (Dingguo Biotech Co, Beijing, China) according to the manufacturer’s instructions. Total proteins were separated using 10%–12% sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and, after blotting onto membranes, analyzed using specific antibodies.

Immunofluorescence assay

Immunofluorescence was performed as previously described [25, 26]. Briefly, BMSCs were seeded onto glass coverslips and incubated overnight. Next, the cells were washed with PBS and fixed with 4% paraformaldehyde, permeabilized with 0.5% Triton X-100, and blocked with 4% bovine serum albumin. Subsequently, the samples were incubated with primary antibodies directed against α-SMA (1:100) overnight at 4 °C, followed by incubation with an Alexa fluor-594-conjugated secondary antibody. DAPI was used to stain the nuclei, and the samples were photographed using a fluorescence microscope.

Establishment of stable COX2 overexpression and knockdown cell lines

To obtain 293 T cells with COX2 overexpression, PLVX-COX2 constructs were co-transfected with packaging (psPAX2) and envelope (pMD2.G) vectors into 293 T cells. Next, GFP positive cells were isolated using a FACS Aria III cell sorter (BD Biosciences, CA, USA). Lentiviral COX2 shRNA expression vector transduction and cell selection were performed as previously described [25].

Cell viability assay

CM was collected as described previously. For MM cell treatment, CM was diluted with an equivalent volume of complete medium supplemented with 10% FBS. MM cells were cultured with the mixed medium in 96-well plates at a density of 5 × 10³ cells per well and incubated for 24 h and 48 h. To examine cell viability, viable cells were enumerated using trypan blue exclusion staining and counted manually using a hemocytometer at the indicated time points. Each experiment was repeated at least three times.

Cell cycle assay

Cell cycle progression of MM cells was quantified using propidium iodide staining (Sigma-Aldrich, Steinheim, Germany) in conjunction with a LSRFortessa X-20 flow cytometer (BD Biosciences, CA, USA), as described previously [27]. Briefly, cells subjected to different CM treatments were collected, washed twice with PBS and next fixed with cold 70% ethanol at −20 °C overnight. After washing with cold PBS, propidium iodide (25 μg/ml solution in PBS supplemented with RNase, 0.2% Triton-X 100) was applied for 30 min at 4 °C, after which the cell samples were analyzed by flow cytometry (gating strategy excluded cell doublets from analysis).

Cell apoptosis assay

Apoptosis of MM cells was quantified after treatment with the indicated CMs for 48 h. MM1S and KMS11 cells were exposed to 2 nM and 3 nM BTZ, respectively. Apoptosis was assessed using Annexin-V-PE/7aminoactinomycin-D staining as recommended by the manufacturer (BD-Pharmingen, CA, USA). Flow cytometric analysis was performed using a BD LSRFortessa X-20 cytofluorimetry system (Becton Dickinson-BD, CA, USA).

Cell adhesion assay

CMs from BMSCs or HS5 cells were used to examine the adhesive ability of MM cells to HS5 cells or fibronectin (FN), respectively. A co-culture system was used to detect direct adhesion of MM cells to BMSCs. Briefly, for CM pretreatment, KMS11 or MM1S cells were treated with CM for 48 h, after which the cells were harvested, resuspended and seeded at 1 × 10⁵ cells/well into 96-well plates in serum-free RPMI-1640 medium. Next, the plates were incubated with FN (50 μg/ml) overnight prior to seeding. For co-culture of HS5 or BMSCs and MM cells, HS5 cells or BMSCs (2 × 10⁴ cells/well) were seeded into 96-well plates and cultured overnight, followed by the addition of 1 × 10⁵ MM cells in 100 μl serum-free RPMI-1640 medium. After incubation for two hours, the nonadherent MM cells were removed. Adherent cells were subsequently stained with 0.2% crystal violet for 2 to 24 h at RT. Next, crystal violet was washed off with distilled water, and the plates were air-dried overnight at RT. The dye was then dissolved with 2% SDS for 2 h on a shaking platform, and ODs were measured at 570 nm using a microplate reader. The ODs from BMSCs cultured alone were considered as background absorption.

Mouse 5T33MM model

C57BL/KaLwRij mice were purchased from Harlan and housed in accordance with guidelines of the Ethical Committee for Animal Experiments of Cancer Research Institute, CSU (2017sydw0092). 5T33MMvt cells (3 × 10⁶ cells in 300 μl PBS) were injected into 12 weeks old C57BL/KaLwRij mice through the tail vein. After one week, mice were randomly divided into 4 groups, and subsequently treated with NS398 (n = 6, 5 mg/kg), BTZ (n = 5, 1 mg/kg), NS398 plus BTZ (n = 6, 5 mg/kg plus 1 mg/kg) and PBS (n = 5). In the second week, mice in the NS398 or NS398 plus BTZ groups were injected with NS398 (5 mg/kg) intraperitoneally for 5 consecutive days. After one-week treatment with NS398, BTZ (1 mg/kg) was injected intraperitoneally twice a week, whereas PBS was used as control. Peripheral blood was obtained every two weeks for IgG2b detection in plasma until the end of the experiment. After 8 weeks, two mice from the PBS group and one from the NS398 group exhibited signs of hindlimb weakness. Then, all animals were sacrificed. Dissected tibiae were employed to evaluate bone disease using a MicroCT scanner. Extramedullary myeloma tissues removed from the mice were subjected to CD138 staining.

Statistical analysis

Statistical analyses were performed using GraphPad Prism 5 software (GraphPad Prism Inc., CA, USA). The two-tailed t-test was utilized for comparison of two conditions. A p value < 0.05 was considered statistically significant and marked as p < 0.05*, p < 0.01** or p < 0.001***. Graphical results are presented as means with error bars (standard error of the mean, SEM).

Results

COX2 is up-regulated in BMSCs derived from MM patients

To confirm the characteristics of BMSCs derived from MM patients (MM-BMSCs, n = 6) or healthy donors (HD-BMSCs, n = 5), we assessed the expression of an established marker of fibroblasts named α-SMA [28] (Additional file 2: Fig. S1a). In addition, BMSCs are known to express the mesenchymal markers CD90, CD73 and CD105, and to lack expression of the hematopoietic markers CD31 and CD34 [29] (Additional file 2: Fig. S1b).

To compare global mRNA expression profiles between primary MM-BMSCs and HD-BMSCs, we performed RNA-sequencing using BMSCs from two healthy donors and two newly diagnosed MM patients. Gene set enrichment analysis (GSEA) revealed that the inflammatory response was significantly enriched in MM-BMSCs, implying that inflammation may be an important link between BMSCs and MM cells (Fig. 1a). Interestingly, PTGS2, also known as COX2, a key molecule in inflammatory responses, was the most upregulated pro-inflammatory factor in MM-BMSCs compared with HD-BMSCs (Fig. 1b).

Fig. 1.

COX2 expression is upregulated in MM-BMSCs compared with HD-BMSCs. (a). GSEA profiles showing significant enrichment of inflammatory response genes in MM-BMSCs compared with HD-BMSCs. (b). Fold change of five inflammatory related genes in MM-BMSCs compared with HD-BMSCs from two independent RNA-seq analyses. (c). Western blot analysis of COX2 expression in HD-BMSCs and MM-BMSCs. β-actin was used as loading control. The gray levels of the blots were quantified using Image J software. (d). Representative IHC images of COX2 expression in stroma of bone marrow specimens from 28 healthy donors (HDs) and 26 MM patients. Brown staining, COX2 positive; 0: negative, 1: weak, 2: strong; blue, cell nucleus. Red arrows point to COX2 positive stroma cells. Scale bar: 50 μm. (e). Quantification of COX2 expression in stroma of HD and MM patients. (f). PGE2 levels in CM from HD-BMSCs and MM-BMSCs measured by ELISA. (g). PGE2 levels in serum (n = 11 for HD, n = 17 for MM patients, respectively). (h). Correlation of PGE2 levels in serum with histological disease stage (DS). Data are represented as mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001. (Two-tailed Student’s t test for comparison of two groups, one-way ANOVA for more than two groups)

We next tested whether COX2 is up-regulated in BMSCs using qRT-PCR and Western blotting. Both COX2 mRNA and protein levels were found to be elevated in MM-BMSCs compared with HD-BMSCs (Additional file 2: Fig. S1c and Fig. 1c). In addition, multiplexed quantitative immunofluorescence analysis indicated a tendency of α-SMA and COX2 co-expression in MM patients (Additional file 2: Fig. S1d and Fig. S1e). We further examined COX2 expression in stromal cells by IHC staining and found that it was markedly increased in stromal cells of MM patients compared with those from healthy donors (HDs; p < 0.05, Fig. 1d, e). Moreover, COX2 levels were found to be significantly correlated with age and International Staging System (ISS) stage (p < 0.05, Table 1). Considering that PGE2 is the major product catalyzed by COX2, we next assessed PGE2 levels in CMs of MM-BMSCs as well as sera from MM patients using ELISA. Consistent with COX2 expression, we found that PGE2 secretion of MM-BMSCs was higher than that of HD-BMSCs (Fig. 1f). COX2 expression in BMSCs was significantly positively correlated with PGE2 levels in the CMs (Additional file 2: Fig. S1f). Furthermore, we confirmed simultaneously elevated PGE2 levels in serum samples from MM patients (413.9 pg/ml vs. 260.15 pg/ml, Fig. 1g). Additionally, we found that PGE2 levels were higher in MM patients with late Durie-Salmon (DS) stage III disease than in those with early stage (I and II) disease (Fig. 1h). Taken together, these results indicate that altered COX2/PGE2 levels in BMSCs may play a significant role in MM progression.

Table 1.

Correlation between COX2 expression in MM BMSCs and clinical characteristics

Clinical characteristics	COX2-negative (n = 10) (n/N × 100%)	COX2-positive (n = 16) (n/N × 100%)	P value
Sex			0.105
Male	7/10	5/16
Female	3/10	11/16
Age (years)			0.005**
< 60	4/10	15/16
≥ 60	6/10	1/16
ISS stage			0.034*
I	8/10	6/16
II-III	2/10	10/16
DS stage			1
I-II	3/10	4/16
III	7/10	12/16
Median calcium (mmol/l)	2.173	2.178	0.973
Hb level (g/L)	103.1	108.6	0.586
Plasma cells in BM (%)	20.9	30.4	0.458

Abbreviations: DS: Durie-Salmon. ISS: International Staging System. Hb: Hemoglobin. *Statistically significant, p < 0.05*, p < 0.01**. All MM patients were newly diagnosed

COX2 contributes to BMSC-mediated MM cell proliferation and adhesion

Consistent with earlier studies from other groups, we found that the proliferation of MM cells was accelerated in the presence of CM from MM-BMSCs compared with CM from HD-BMSCs (Additional file 2: Fig. S2a). To further examine whether COX2 was involved in BMSC-mediated proliferation of MM cells through PGE2, we assessed mRNA expression levels of PGE2-EP signaling pathway factors, including EP1, EP2, EP3 and EP4 in MM cells cultured with CM of MM BMSCs. We found that, based on the mRNA (Additional file 2: Fig. S2b) and protein (Additional file 2: Fig. S2c) levels, EP4 specifically mediated the aggressive phenotype induced by cultivation of MM cells in the presence of CM from MM BMSCs. To next explore whether MM BMSCs induce the aggressive phenotype through EP4, we evaluated MM cell growth by blocking EP4 with AH23848 (EP4 antagonist) when cultured with CM of MM BMSCs (Additional file 2: Fig. S2d). Notably, we found that EP4 antagonist treatment significantly blocked the effects of MM BMSCs on the growth and adhesion of MM cells (Fig. 2a). Therefore, we conclude that COX2 in MM-BMSCs promotes MM cell growth via PGE2.

Fig. 2.

COX2 is involved in BMSC-mediated MM cell proliferation and adhesion. (a). MM1S and KMS11 cells were incubated with CMs of HD-BMSCs or MM-BMSCs with or without EP4 antagonist AH23848, after which cell numbers were counted at 24 h and 48 h (n = 3). (b, c). Proliferation of MM1S and KMS11 cells after incubation with CMs from HD-BMSCs with PGE2 stimulation (b), or from MM-BMSCs with NS398 (0 or 20 μM) (c), (d, e). Cell cycle distributions of MM cells after incubation with CMs from HD-BMSCs with PGE2 stimulation (d), or from MM-BMSCs with NS398 for 48 h (e). (f, g). Adhesion of MM cells after incubation with CMs from HD-BMSCs with PGE2 stimulation (f) or MM-BMSCs with NS398 (g). Statistical analyses were performed using Student’s t test. *p < 0.05, **p < 0.01, ***p < 0.001

Because COX2 in MM-BMSCs facilitated MM cell growth through PGE2, PGE2 was subsequently used to mimic COX2 overexpression in HD-BMSCs. We found that the proliferation of MM1S and KMS11 cells was enhanced when incubated with CM of HD-BMSCs with PGE2 stimulation compared with CM of HD-BMSCs alone (Fig. 2b). In contrast, opposite trends were found in MM cells cultured with CM from MM-BMSCs with NS398, compared with CM from MM-BMSCs alone (Fig. 2c). To explore whether COX2 supports the growth of MM cells by promoting cell cycle progression, we performed flow cytometric analysis to quantify the distribution of cell cycle phase cells. We found that the S-phase population was markedly increased in MM cells cultured in CM from HD-BMSCs with PGE2 pretreatment compared with those cultured in CM from HD-BMSCs (34.833 ± 2.13% vs. 46.333 ± 1.65% for MM1S and 40.153 ± 2.749% vs. 52.989 ± 0.856% for KMS11, Fig. 2d). Conversely, we found that the number of cells in the S-phase population was decreased in MM1S and KMS11 cells cultured with CM of MM-BMSCs pretreated with NS398 (46.167 ± 0.736% vs. 32.8 ± 0.656% for MM1S and 56.605 ± 5.61% vs. 31.285 ± 0.901% for KMS11, Fig. 2e).

It has previously been reported that PGE2 from MM cells mediates adhesion of MM cells to FN [30], leading us to hypothesize that PGE2 released from BMSCs may modulate the adhesion of MM cells to BMSCs. To test this hypothesis, we co-cultured MM cells with HD-BMSCs or MM-BMSCs for 2 h and examined cell adhesion. We found that the adhesion capacity of MM1S cells to MM-BMSCs was enhanced (0.143 ± 0.013 for HD2, 0.253 ± 0.034 for MM2 and 0.219 ± 0.006 for MM3). Similar effects were observed in KMS11 cells (0.584 ± 0.055 for HD2, 1.003 ± 0.032 for MM2 and 1.271 ± 0.036 for MM3, Additional file 2: Fig. S2e). To further determine whether MM-BMSCs influence the adhesion of MM cells indirectly, we pretreated MM cells with CM from HD-BMSCs or MM-BMSCs for 48 h and subsequently co-cultured them with HS5 cells for 2 h. We found that the adhesion capacity of MM1S cells to HS5 was enhanced when the cells were cultured in CM of MM-BMSCs compared with that of HD-BMSCs (0.218 ± 0.045 for HD2, 0.464 ± 0.03 for MM2 and 0.469 ± 0.076 for MM3). Similar effects were observed in KMS11 cells (0.307 ± 0.038 for HD2, 0.671 ± 0.091 for MM2 and 0.82 ± 0.052 for MM3, Additional file 2: Fig. S2f).

To further assess the role of PGE2 in cell adhesion, we collected CM from HD-BMSCs with PGE2 stimulation prior to cultivation with MM cells. We found that the adhesion capacity of MM cells to HS5 was enhanced when cultured in CM with PGE2 stimulation (0.316 ± 0.006 vs. 0.616 ± 0.028 for MM1S and 0.857 ± 0.188 vs. 1.715 ± 0.167 for KMS11, Fig. 2f). In contrast, we found that the adhesion of MM cells was impaired in the presence of NS398 (1.052 ± 0.024 vs. 0.654 ± 0.037 for MM1S and 1.417 ± 0.033 vs. 1.042 ± 0.035 for KMS11, Fig. 2g). To next test whether MM-BMSCs enhance the adhesion of MM cells through PGE2, adhesion was tested in MM cells co-cultured with CM of MM-BMSCs and AH23848. We found that AH23848 attenuated the promoting effects of CM of MM-BMSCs on adhesion (Additional file 2: Fig. S2g). These results suggest that COX2/PGE2 in BMSCs enhances the proliferation and adhesion of MM cells.

COX2 knockdown impairs BMSC-induced MM cell proliferation and adhesion

To further explore the role of COX2 in MM progression, we knocked down its expression in HS5 cells using COX2 small hairpin RNAs (shRNAs). These reduced COX2 expression at both the mRNA (80%, Fig. 3a) and the protein (58%, Fig. 3b) level. Simultaneously, PGE2 was found to be impaired in CM collected from HS5-COX2-sh cells compared with HS5-Ctrl cells (1252 ± 82.8 vs. 500.5 ± 80.2, Additional file 2: Fig. S3a). CM from HS5-Ctrl and HS5-COX2-sh cells was subsequently used to culture MM cells, followed by proliferation and adhesion assessment. As expected, the proliferation of MM cells was suppressed after incubation with CM from HS5-COX2-sh cells, which was reversed by PGE2 addition (Fig. 3c). Consistently, CM from COX2-sh cells inhibited the proportion of S-phase MM cells (36.9 ± 2.3% vs. 24.9 ± 1% for MM1S and 44.5 ± 1.7% vs. 34.2 ± 0.513% for KMS11, Additional file 2: Fig. S3b), while PGE2 augmented the proportion of S-phase MM cells (32.5 ± 0.2% vs. 49.4 ± 1.5% for MM1S and 27.5 ± 2.711% vs. 65 ± 0.355% for KMS11, Additional file 2: Fig. S3b).

Fig. 3.

COX2 knockdown in BMSCs inhibits MM cell proliferation and adhesion. mRNA (a) and protein levels (b) of COX2 in HS5 COX2-sh cells examined by qRT-PCR and Western blotting, respectively. GAPDH and β-actin were used as internal controls for the qRT-PCR and Western blot assays, respectively. Proliferation of MM cells (c) and adhesion of MM cells to FN (d) analyzed after incubation with CMs from HS5-Ctrl or HS5-COX2-sh cells. Ctrl: CM from HS5-Ctrl cells, Ctrl + NS398: CM from HS5-Ctrl cells with NS398 treatment, HS5-COX2-sh: CM from HS5-COX2-sh cells, sh + PGE2: CM from HS5-COX2-sh cells with PGE2 stimulation. (e). COX2 protein expression in mouse COX2^+/− BMSCs and COX2^−/− BMSCs detected by Western blotting. (f). Proliferation of 5T33MMvt cells after incubation with CMs from COX2^−/− BMSCs or from COX2^+/− BMSCs. (g). Adhesion of 5T33MMvt cells after incubation with CMs from COX2^+/− BMSCs or COX2^−/− BMSCs. COX2^+/−  + NS398: COX2^+/− BMSCs treated with NS398, COX2^−/− + PGE2: COX2^−/− BMSCs treated with PGE2. Statistical analyses were performed using Student’s t test. *p < 0.05, **p < 0.01, ***p < 0.001

We next compared adhesion of MM cells treated with CM from HS5-Ctrl or HS5-COX2-sh cells. CM from HS5-COX2-sh cells significantly decreased the adhesion of MM cells to FN (0.449 ± 0.024 vs. 0.255 ± 0.009 for MM1S and 0.454 ± 0.022 vs. 0.302 ± 0.047 for KMS11, Fig. 3d). Expectedly, PGE2 pretreatment rescued the adhesion defects caused by COX2 knockdown (0.282 ± 0.01 vs. 0.463 ± 0.039 for MM1S and 0.23 ± 0.028 vs. 0.334 ± 0.012 for KMS11, Fig. 3d). Moreover, MM cells displayed a lower adhesion capacity to HS5-COX2-sh cells than to HS5-Ctrl cells (0.589 ± 0.078 vs. 0.259 ± 0.039 for MM1S and 0.396 ± 0.014 vs. 0.266 ± 0.008 for KMS11, Additional file 2: Fig. S3c).

To further validate the functional role of COX2 in BMSCs, we generated BMSCs from COX2 knockout mice (COX2^−/− BMSC). Because COX2^+/+ mice failed to breed offspring, we could not obtain COX2 wild-type BMCSs as controls. Instead, we prepared primary BMSCs from COX2^+/− mice (COX2^+/− BMSCs). Expectedly, COX2 protein was hardly detectable in COX2^−/− BMSCs (Fig. 3e). We subsequently examined whether CM from COX2^−/− BMSCs affects the growth and adhesion of 5T33MMvt cells. We found that treatment of CM from COX2^−/− BMSCs significantly impeded the growth of 5T33MMvt cells compared with COX2^+/− BMSCs, while PGE2 reversed the inhibitory effect of CM from COX2^−/− BMSCs (Fig. 3f). Furthermore, we found that incubation of 5T33MMvt cells with CM from COX2^−/− BMSCs attenuated 5T33MMvt cell adhesion compared with COX2^+/− BMSCs (0.321 ± 0.013 vs. 0.252 ± 0.001, Fig. 3g), while administration of PGE2 to COX2^−/− BMSC rescued the adhesion defective phenotype (0.252 ± 0.001 vs. 0.533 ± 0.007, Fig. 3g).

Because cell adhesion-mediated drug resistance is considered to be one of the mechanisms for relapse [13], we next measured apoptosis of MM cells treated with CMs of HS5-COX2-sh cells and HS5-Ctrl cells in the presence or absence of FN. We found that MM cells were more resistant to BTZ when exposed to FN-coated plates than to uncoated plates (51.7 ± 0.186 vs. 41.9 ± 0.239, Additional file 2: Fig. S3d). Consistently, we found that FN-mediated drug resistance was rescued by NS398 (71 ± 0.722 vs. 67.3 ± 0.5, p > 0.05, Additional file 2: Fig. S3d). These data suggest that COX2 in BMSCs confers BTZ resistance via both contact-dependent and contact-independent (i.e., PGE2-mediated) mechanisms.

COX2 upregulates TNFα expression in BMSCs

PGE2 has been reported to act as an activator of Wnt/β-catenin signaling, which promotes aggressive phenotypes of tumor cells. To verify whether MM-BMSCs contribute to proliferation and adhesion via Wnt/β-catenin signaling, MM cells were treated with combined CMs of HD-BMSCs with PGE2 and FH535 (Additional file 2: Fig. S4a). Interestingly, we found that FH535 attenuated the promoting effects of PGE2 on MM cell proliferation (Additional file 2: Fig. S4b). However, no repressive effect of FH535 on MM cell adhesion was observed (Additional file 2: Fig. S4c). We speculated that MM-BMSCs may enhance cell adhesion through other mechanisms for enriched inflammatory responses in MM-BMSCs. To this end, we first examined the expression of 15 inflammatory cytokines using a Real-time PCR assay. We found that most inflammatory cytokines were upregulated in MM-BMSCs compared with HD-BMSCs except IL1β (Fig. 4a). To next investigate whether COX2 regulates the expression of those cytokines, we assessed their expression in HS5-COX2-sh cells. Among them, CSF2, IL6, IL1β and TNFα were significantly reduced in these cells (Fig. 4b). Because IL1β was down-regulated in MM-BMSCs and IL6 has been widely studied in BMSCs, CSF2 and TNFα were selected for further analysis. Unexpectedly, we found that the CSF2 protein level exhibited no difference in HS5-COX2-sh cells compared with HS5-Ctrl cells, while the expression of TNFα was reduced in HS5-COX2-sh cells (Additional file 2: Fig. S4d). TNFα is a signaling cytokine of the NF-κB pathway, while COX2 acts as a target gene of NF-κB signaling [31]. As yet, however, it is unknown whether TNFα is regulated by COX2.

Fig. 4.

COX2 upregulates TNFα expression in BMSCs. (a). Heat map of the expression of 15 inflammatory-related genes in HD-BMSCs and MM-BMSCs detected by qRT-PCR. The genes are ordered from highest to lowest expression by fold change in MM-BMSCs relative to HD-BMSCs. (b). Relative mRNA levels of the above genes analyzed in HS5-COX2-sh cells by qRT-PCR. (c). Relative protein levels of COX2 and TNFα in HD-BMSCs and MM-BMSCs (left) and 293 T cells with or without exogenous COX2 overexpression (right) detected by Western blotting. (d). Relative protein levels of COX2 and TNFα in HS5 cells with NS398 treatment (left), HS5-COX2-sh cells (middle), and COX2^−/− BMSCs (right) detected by Western blotting. (e). TNFα level in CM from MM-BMSCs and HD-BMSCs detected by ELISA. (f). TNFα level in serum from HD and MM patients detected by ELISA. (g). Correlations between PGE2 and TNFα levels in sera of MM patients analyzed by Pearson’s test. (h). Relative mRNA levels of TNFα and PGE2 downstream receptors in HD-BMSCs cultured with PGE2 analyzed by qRT-PCR. (i). TNFα, EP2 and EP4 protein levels in HD-BMSCs cultured with PGE2 and AH6809 or AH23848 detected by Western blotting. Data are presented as mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001 (Two-tailed Student’s t test)

Thus, we next investigated whether TNFα is indeed regulated by COX2. Upregulation of TNFα was found in MM-BMSCs and 293 T cells after exogenous COX2 overexpression (Fig. 4c). Conversely, TNFα was downregulated in HS5 COX2 knockdown cells and COX2^−/− BMSCs (Fig. 4d). Because TNFα is a secreted protein, we assessed TNFα in CM from BMSCs as well as in serum samples from healthy donors (HD) and MM patients. Consistent with the increased expression of TNFα, the secretion of TNFα was elevated in CM from MM-BMSCs compared with HD-BMSCs (1.269 ± 0.233 for HD2, 4.907 ± 0.211 for MM2 and 6.234 ± 0.536 for MM3, Fig. 4e). In contrast, we found that TNFα secretion was reduced in HS5 COX2 knockdown cells compared with HS5-Ctrl cells (Additional file 2: Fig. S4e). Moreover, TNFα levels in sera from MM patients were significantly higher than in those from healthy donors (28 pg/ml vs. 8.32 pg/ml, Fig. 4f). Interestingly, the PGE2 levels positively correlated with TNFα levels in the serum samples (n = 14, r = 0.724, Fig. 4g). However, high TNF-α levels in MM patients were not correlated with disease stage (DS) (Additional file 2: Fig. S4f). Furthermore, TNFα levels in the sera were significantly correlated with sex, ISS stage and the percentage of plasma cells in MM patients (Additional file 1: Table S3).

To further delineate whether COX2 promotes TNFα secretion through PGE2, we examined the mRNA expression levels of the TNFα and PGE2 downstream receptors in HD-BMSCs supplemented with PGE2. Upregulation of TNFα, EP2 and EP4 was observed in HD-BMSCs treated with PGE2 (Fig. 4h). We next examined TNFα expression of HD-BMSCs treated with PGE2 and AH6809 (EP2 antagonist) or AH23848 (EP4 antagonist). Interestingly, we found that EP2 rather than EP4 was responsible for PGE2-mediated TNFα expression (Fig. 4i). These results indicate that COX2 promotes TNFα secretion through PGE2.

BMSC-secreted TNFα promotes MM cell proliferation and adhesion by activating the NF-κB pathway

TNFα has been reported to serve as a survival factor for MM cells [32, 33]. However, whether and how COX2 induction of TNFα in BMSCs may affect MM progression is still unclear. To investigate whether COX2 promotes MM cell proliferation and adhesion through TNFα, we first examined the proliferation of KMS11 and MM1S cells after TNFα treatment or after incubation with antibody directed to TNFα. As expected, TNFα antagonized the repressive effect of CM with COX2 inhibition on MM cell growth, while the antibody directed to TNFα reversed the enhanced effect of CM with high PGE2 expression (Fig. 5a). We further assessed the impact of TNFα on 5T33MMvt MM cell survival after the cells were exposed to CM from COX2^−/− BMSCs. Consistent with the results of the human MM cells, TNFα attenuated the repressive effect of CM from COX2^−/− BMSCs on murine MM cell growth (Fig. S5a). Thus, we conclude that COX2 in BMSCs contributes to MM cell proliferation through TNFα.

Fig. 5.

COX2-induced TNFα promotes MM cell proliferation and adhesion via the NF-κB pathway. (a). Proliferation of MM1S and KMS11 cells after incubation with CMs from HS5 cells with different treatments. Ctrl + NS398 + TNFα: CM from HS5-Ctrl with NS398 treatment supplemented with TNFα, COX2-sh + PGE2 + αTNFα: CM from HS5-COX2-sh cells with PGE2 stimulation supplemented with αTNFα. (b). Adhesion of MM cells to FN detected after incubation with CMs from HS5 cells. (c). Detection of Annexin V-positive apoptotic cells after incubation with CMs in the presence of 2 nM BTZ pre-incubated with or without FN. (d). Detection of TNFR1β and p-p65 levels in MM cells after incubation with CMs from HS5 cells for 48 h. (e). Ras protein levels in KMS11 cells after exogenous overexpression of mutant Ras (K-Ras/G12V or N-Ras/G12V) with or without PGE2 detected by Western blotting. Proliferation (f) and adhesion (g) of KMS11 cells expressing mutant Ras cultured with CM of HD-BMSCs in the presence or absence of PGE2. *p < 0.05, **p < 0.01, ***p < 0.001; NS, not significant

Next we investigated whether the adhesion of MM cells is affected by TNFα. As expected, adhesion of both human MM cells and mouse 5T33MMvt MM cells to FN was augmented by TNFα, while attenuated by the antibody directed to TNFα (Fig. 5b and Fig. S5b). Since COX2 is known to be involved in cell adhesion-mediated drug resistance, also known as CAM-DR (Additional file 2: Fig. S3d), we next asked whether TNFα plays a role in MM drug resistance through the regulation of cell adhesion. We found that TNFα antagonized the repressive effects of NS398 on resistance to BTZ, whereas the antibody directed to TNFα enhanced MM cell sensitivity to BTZ (Fig. 5c and Additional file 2: Fig. S5c). These data strongly suggest that COX2/PGE2-TNFα, secreted by BMSCs, is responsible for promoting MM cell proliferation, adhesion and drug resistance.

Both extracellular domains of TNFR1α and TNFR1β are able to bind to the same TNF ligand, TNFα [34], therefore leading us to assess the expression of TNFR1α and TNFR1β in MM cells pretreated with CM. First, we examined the levels of TNFR1α and TNFR1β in KMS11 and MM1S cells. We found that TNFR1β, but not TNFR1α, was expressed in MM cells. In addition, we found that treatment with CM from MM-BMSCs pre-incubated with NS398 significantly suppressed TNFR1β protein expression in KMS11 cells, while TNFα restored this expression (Additional file 2: Fig. S5d). Therefore, TNFR1β, but not TNFR1α, likely mediates the TNFα signaling pathway in MM cells. Moreover, MM cells treated with CM from HS5 pre-incubated with NS398 exhibited decreased TNFR1β and NF-κB activity compared with the control, whereas TNFα restored TNFR1β and NF-κB activity in both KMS11 and MM1S cells (Fig. 5d). Our results may be explained by a compensatory cross-talk between canonical and non-canonical TNFα pathways [35].

Since previous work has shown that mutant Ras in MM cells participates in enhanced adhesion [30], we next checked the presence of Ras mutations in MM1S and KMS11 cells. We found that KMS11 cells have both wild-type K-Ras and N-Ras, while MM1S cells have wild-type N-Ras but mutant K-Ras (the glycine at position 12 is replaced with alanine). Therefore, we assume that MM-BMSCs promote the proliferation of MM cells with both wild-type and mutant Ras. To further verify whether PGE2 is enhanced via the Ras/MAPK pathway, MM cells expressing mutant K-Ras/G12V and N-Ras/G12V (in which the glycines at positions 12 are replaced with valine) were subjected to proliferation and adhesion assays (Fig. 5e). As expected, the proliferation and adhesion was enhanced in KMS11 cells with exogenously overexpressed mutant Ras, but PGE2 could not further enhance the aggressive phenotype (Fig. 5f and 5g). Because the amount of mutant Ras was increased by 200–400 fold after exogenous overexpression, we propose that the effect of PGE2 may be restricted by overactivation of mutant Ras. Taken together, these data suggest that PGE2 and/or TNFα secreted by BMSCs contribute to MM cell proliferation, adhesion and drug resistance via TNFR1β-mediated NF-κB activation.

Inhibition of COX2 suppresses MM cell proliferation in vivo

To explore the in vivo role of COX2 in MM, we investigated whether COX2 plays a functional role in 5T33MMvt-KaLwRij mice. Briefly, 5T33MMvt-KaLwRij mice were inoculated intravenously with 3 × 10⁶ 5T33MMvt cells, and after 2 weeks the mice were randomly divided into four groups (6 mice each group) and next treated with NS398 (5 mg/kg), BTZ (1 mg/kg), NS398 plus BTZ (5 mg/kg plus 1 mg/kg) and PBS, respectively (Fig. 6a). Compared to the PBS group, the MM mice treated with NS398 or BTZ showed a lower morbidity (Fig. 6b) and an extended survival (Fig. 6c). Surprisingly, the inhibitory effect of NS398 on MM progression was even more significant than that of BTZ. Moreover, IgG2b was markedly reduced by BTZ or NS398 plus BTZ at week 5 and 7 (p = 0.019 and p = 0.028 at week 5 and p = 0.018 and p = 0.038 at week 7, Fig. 6d), though there was only a minimal difference in the first three weeks (p > 0.05 for each case). Since bone destruction is an indicator to evaluate MM progression, we evaluated the occurrence of bone defects in the respective groups. Intriguingly, we found that the bone defects of tibiae were significantly alleviated after treatment with NS398 compared with PBS (Fig. 6e). Collectively, our results provide evidence that COX2 inhibition can improve the survival of MM mice.

Fig. 6.

COX2 inhibition suppresses MM cell proliferation in vivo. (a). Scheme of the group distribution for the in vivo study. Morbidity (b) and survival curves (c) of 5T33MMvt-KaLwRij mice from different experimental groups. Survival curves were analyzed using Mantel-Cox log-rank test. (d). IgG2b concentrations in sera of the different experimental groups determined by ELISA. Data are presented as mean ± SD. (e). microCT images of tibiae from different experimental groups. (f). Model of our hypothesis in this work

Discussion

In this study, we report a functional role of BMSC-derived COX2 in MM development. We show that a high expression of COX2 in BMSCs promotes the growth and enhances the adhesion of MM cells. Subsequent analyses indicated that PGE2 and TNFα are released from BMSCs, facilitating the growth and adhesion of MM cells and inducing NF-κB activation via TNFR1β-mediated signaling. Using a 5T33MMvt-KaLwRij mouse model we found that the administration of NS398 attenuated the progression and improved the survival of MM mice. Thus, COX2 may be a promising target for anti-myeloma therapy.

BMSCs are well known to facilitate the growth of both solid tumors and hematologic malignancies [36–39]. Previous studies have shown that interaction with BM stroma may activate Notch signaling in MM and other malignant lymphoid cells, being an important cause of drug-resistance [8]. Additionally IL1β, a BMSC paracrine factor, has been reported to promote the development of prostate cancer via AR repression and p62 induction [36]. Indeed, LPS-treated primary cultured murine and human BMSCs promote the adhesion and proliferation of melanoma cells, whereas the COX2 inhibitor celecoxib could abolish this process [37]. As yet, however, little is known about the differences in gene expression signatures between HD-BMSCs and MM-BMSCs. In this study, we performed RNA-seq analysis of BMSCs from MM patients and found that COX2 is upregulated in MM-BMSCs compared with HD-BMSCs. Previous work has shown that COX2 is regulated by several pro-inflammatory factors such as IL6, TNFα and IL1β. In addition, several studies have reported that stromal-derived IL6 may promote proliferation and induce drug resistance through paracrine actions in neuroblastoma and MM [38, 39]. We found through qPCR-array analysis that IL6 and TNFα are upregulated in MM-BMSCs. Therefore, we suspect that COX2 may be induced by excessive IL6 during the transformation of MM BMSCs from normal BMSCs. This possibility needs further investigation.

Previous studies have reported that direct interactions of tumor cells with surrounding stromal cells may influence tumor cell proliferation and metastasis [40]. Increased adherence of MM cells to stromal cells has been reported to support their proliferation and drug resistance, and to cause osteolytic lesions and angiogenesis [41, 42]. Here, we found that COX2-derived PGE2 not only promotes MM cell proliferation, but also enhances MM cell adherence to BMSCs or FN. In fact, high PGE2 levels enhance interactions of MM cells with BMSCs or FN, while COX2 inhibitors impair MM cell adhesion via specific actions on COX2 rather than on N-Ras or K-Ras [30]. Accumulating evidence shows that increased adhesion of MM cells to BMSCs contributes to drug resistance. Interestingly, Azab et al. reported that hypoxia induces MM cell aggressiveness via upregulation of molecules related to epithelial-mesenchymal transition (EMT), and this observation has been partly attributed to decreased adhesion of MM cells to the bone marrow stroma under hypoxic conditions [43]. Collectively, adhesion of MM cells to BMSCs probably induces two independent processes, i.e., metastasis and drug resistance. Previous studies have shown that high COX2 expression correlates with a poor prognosis of MM patients. We initially speculated that MM cells may be the source that releases PGE2. Our results showed, however, that BMSCs rather than tumor cells are the main source of PGE2, and that PGE2 in serum of MM patients was significantly correlated with disease stage (DS). Therefore, COX2 inhibition may reduce the risk of drug resistance and tumor recurrence, at least partly by inhibiting the secretion of PGE2 from BMSCs.

In this study we show that COX2 induces PGE2 and/or TNFα secretion in BMSCs and, subsequently, upregulates TNFRII levels in MM cells. As a result, nuclear factor-kappaB (NF-κB) transcription is increased. TNFα reportedly blocks the apoptosis of tumor cells [44], binds to either TNFRI or TNFRII and recruits several proteins to form a complex to stimulate downstream signaling. TNFRI binds to the adaptor molecule Fas-associated death domain, thereby stimulating the caspase cascade and, consequently, cell death. However, TNFRII-mediated RelB:p50/NF-κB activation has been reported to mediate pro-survival signaling in MM cells [45]. Previous studies have shown that NF-κB plays a critical role in the pathogenesis of MM and that activation of the NF-κB pathway contributes to the growth and metastasis of MM cells [46]. Intriguingly, our results indicate that phosphorylation of p65 is elevated after exposure of MM cells to CM containing TNFα. This effect may be due to cross-talk between classical and non-classical NF-κB pathways. However, the exact mechanism by which TNFRII activates p65 remains to be determined.

In addition to the aforementioned results, several groups have shown that COX inhibitors can significantly inhibit osteoclast and osteoblast differentiation in vitro [47]. Moreover, Ono et al. reported that COX2 plays an important role in the osteolysis of bone metastasis in mammary carcinoma through the production of PGE2 [48]. In our study, we found that COX2 inhibition significantly reduced bone defects in a MM mouse model. It will be of great interest to investigate whether MM patients with bone defects can similarly benefit from COX2 inhibitor treatment.

In the present study, we validated a critical role of COX2 in BMSCs in the regulation of MM development, and unveiled a novel mechanism by which COX2 contributes to MM progression. In addition, our results raise the possibility of applying COX2 inhibitors, such as celecoxib and aspirin, for treating MM.

Abbreviations

MM

multiple myeloma

BMSCs

bone marrow stromal cells

HD-BMSCs

bone marrow stromal cells derived from healthy donors

MM-BMSCs

bone marrow stromal cells derived from MM patients

CM

Conditioned medium

COX2

Cyclooxygenase 2

ELISA

Enzyme-linked immunosorbent assay

IHC

Immunohistochemistry

PGE2

Prostaglandin E 2

GFP

Green fluorescent protein

TNFα

Tumor necrosis factor alpha

TNFR1β

Tumor necrosis factor receptor 1β

NF-κB

Nuclear factor kappa B

CAM-DR

Cell adhesion mediated drug resistance

Authors’ contributions

Wen Zhou designed this study. Chunmei Kuang and Yinghong Zhu performed the experiments and contributed to data acquisition. Wen Zhou, Jiliang Xia and Chunmei Kuang wrote and revised the manuscript. Guizhu Liu provided COX2^−/− mice. YingHong Zhu, Jingyu Zhang and Xuan Wu performed the in vivo study. Yangjian Ou and Lu Xie analyzed the RNA sequencing data. Mu Hao, Lugui Qiu, Xin Li and Songqing Fan provided patient samples. Wenxia Zhang provided HD samples. Yongjun Guan, Hang Xu, Jiabin Liu, Jiaojiao Guo, Xiangling Feng and Guancheng Li interpreted the data and revised the manuscript. All authors reviewed and approved the final version of the manuscript.

Declaration

Conflict of interest

All authors declare no conflict of interest.