Dicer1 downregulation by multiple myeloma cells promotes the senescence and tumor-supporting capacity and decreases the differentiation potential of mesenchymal stem cells

Bone marrow mesenchymal stem cells (BMMSCs) facilitate the growth of multiple myeloma (MM) cells, but the underlying mechanisms remain unclear. This study demonstrates that the senescence of MM-MSCs significantly increased, as evidenced by a decrease in proliferation and increase in the number of cells positive for senescence-associated β-galactosidase activity. Senescent MM-MSCs displayed decreased differentiation potential and increased tumor-supporting capacity. Dicer1 knockdown in the MSCs of healthy controls promoted cellular senescence and tumor-supporting capacity, while decreasing the differentiation capacity. Dicer1 overexpression in MM-MSCs reversed the effects on differentiation and reduced cellular senescence. In addition, decreased expression of the microRNA-17 family was identified as a favorable element responsible for increasing senescence, with the expression of p21 increased in Dicer1 knockdown cells. Furthermore, we observed decreased expression of miR-93 and miR-20a in MM-MSCs, while upregulation of miR-93/miR-20a decreased cellular senescence, as evidenced by the increased p21 expression. Importantly, we found that myeloma cells could induce the senescence of MSCs from healthy controls, as observed from the decreased expression of Dicer1 and miR-93/miR-20a and increased expression of p21. Overall, MM cells downregulate Dicer1 in MSCs, which leads to senescence; in turn, senescent MSCs promote MM cell growth, which most likely contributes to disease progression.


Introduction
Multiple myeloma (MM) is a malignancy characterized by plasma cell proliferation, initially in the bone marrow microenvironment 1 . This tissue is a complicated network of extracellular matrix and various cells 2 . Undifferentiated and pluripotent bone marrow mesenchymal stem cells (BMMSCs) are the principle members of the bone marrow microenvironment. BMMSCs promote MM cell proliferation via different factors, leading to increased tumor supporting capacity and the development of drug resistance. Evidence has shown that MSCs isolated from MM (MM-MSCs) exhibit different behavior from that exhibited by the MSCs from healthy controls (HC-MSCs), including differential gene expression, impaired cytokine release, and decreased osteogenic differentiation potential [3][4][5][6] . In MM, the interaction between BMMSCs and MM cells plays a pivotal role in MM development. MM cells attach to BMMSCs and release cytokines that promote MM cell viability, migration, and invasive ability, ultimately enhancing MM progression 7 . This interaction usually leads to complicated changes, including dysregulation of mRNA and miRNA expression and activation or deactivation of signaling pathways, but the underlying mechanisms remain unclear. For example, the development of a senescence-like state in BMMSCs, and thereby a modulated secretory profile, has been reported to b Typical colonies at the third passage in six-well plates are exhibited. c MM-MSCs showed degressive cumulative CFU-F numbers at passages 1, 3, and 5 in comparison with HC-MSCs. d Doubling times of MSCs at passages 1, 3, and 5 were obviously higher in MM patients (n = 15) than of control group(n = 10). e Cell proliferation ability was determined by CCK-8 assay. After the culture period of 7 days, the count of living cells in accordance with the optical density from MM groups were growing more slowly than that from the control group (n = 4). f, g Cell cycle distribution of MSC was determined by flow cytometric analysis. We observed a reduction in S phase and an increase in G 0 /G 1 phase for MM-MSC (n = 10) compared to HC-MSCs(n = 5). h, i SA-β-gal was performed to identify the senescent MSCs. The amount of SA-β-galpositive cells was significantly elevated in MM-MSCs (n = 15) at each time point detection. j-l Expression of senescence-related gene p21 increased in MM-MSC (n = 46), as determined by RT-PCR and western blot. The results are expressed as means ± SD. Compared with HC-MSC, the significance was set as *p ≤ 0.05; **p ≤ 0.01 worsen osteogenic differentiation potential [8][9][10][11] . Senescence is a cellular state corresponding with a decrease in proliferation ability and variation in the release of proinflammatory cytokines and growth factors 12 . Senescent BMMSCs exhibit an increased senescence-associated βgalactosidase activity (SA-β-gal) and irregular cell morphology 13,14 . Generally, cell cycle of the senescent cells is often arrested at the G 0 /G 1 phase, in combination with the upregulation of different cell cycle inhibitors such as p21 [15][16][17] . Increasing body of report suggests that micro-RNAs (miRNAs) take part in the regulation of MSC senescence 18,19 . Furthermore, they play an important part in the self-renewal and differentiation of MSCs. As previously indicated, MM-MSCs show a different miRNA profile to that of HC-MSCs, but the roles of these deregulated miRNAs in MM-MSCs are not clearly known 20,21 . The loss of Dicer1, an RNAse III endonuclease essential for miRNA biogenesis, and the ablation of mature miRNA result in the inhibition of cell proliferation and the induction of a premature senescence phenotype [22][23][24] . Our previous study showed that Dicer1 expression is reduced in MSCs from myelodysplastic syndrome patients (MDS) and that this down-regulation promotes cellular senescence and decreases the stem cellsupporting capacity of these cells 17 . However, the effect of the Dicer1 gene on the pathogenesis of MM has not been studied.
In this study, we studied the senescent features of BMMSCs derived from MM patients and explored the biological function of Dicer1 in the senescence of MM-MSCs. Finally, we analyzed the decreased differentiation potential and the enhanced tumor-supporting capacity of MM-MSCs.

Senescent features of BMMSCs from patients with MM
HC-MSCs were characterized by spindle-like morphology, while MM-MSCs were larger and irregular (Fig. 1a). The clonogenic ability of the MSCs were identified by colony forming unit fibroblast (CFU-F). MM-MSCs showed degressive cumulative CFU-F numbers at passages 1, 3, and 5 in comparison with HC-MSCs (p < 0.05, Fig. 1b, c). In addition, the doubling times of MSCs at passages 1, 3, and 5 were obviously higher in MM patients than of control group (Fig. 1d). Cell proliferation ability was determined by CCK-8 assay. After the culture period of 7 days, the count of living cells in accordance with the optical density from MM groups were growing more slowly than that from the control group (Fig. 1e). To demonstrate whether the reductive MM-MSCs growth was caused by weak proliferation or intensive apoptosis, the distributed population of cells in each cell cycle phase was evaluated. No Sub-G 0 peak, in accordance with hypodiploid apoptotic cells, was existed in the BMMSC cultures (<1%). Nevertheless, the percentage of MM-MSCs in G 0 /G 1 phase was increased to 95.66 ± 1.26% (vs 87.78 ± 2.20%; **p < 0.01), while the percentage of MM-MSCs in S phase was reduced to 2.64 ± 0.87% (vs 10.77 ± 2.39%; **p < 0.05) when compared with that of HC-MSCs (Fig. 1f, g).
SA-β-gal was performed to identify the senescent MSCs. The number of SA-β-gal-positive cells was significantly elevated in MM-MSCs at each time point detection (Fig. 1h, i). From passage 3, the median percentage of senescent cells in HC-MSCs accounted for 4.4% of all HC-MSCs. At a cut off value of 4.4%, increased senescence was observed in 86.67% (26/30) of MM-MSCs. Next, we measured the expression of p21, which is involved in the regulation of cell senescence. Using RT-PCR and western blotting, we found an increase in p21 in MM-MSCs in comparison with the control group ( Fig. 1j-l). Moreover, the level of p21 expression increased in MM-MSCs (S-MM-MSCs, SA-β-gal-positive cells ≥4.4%) when compared with non-senescent MM-MSCs (NS-MM-MSCs, SA-β-gal-positive cells <4.4%).
In accordance with the above phenomena, primary MSCs (CD271 + ) from MM patients also exhibited raised cell senescence, which was displayed by an obviously increased amount of SA-β-gal positive cells and increased p21 expression level, in comparison with the healthy control group. The collective data indicate that the proliferation capacity decreased and the senescence increased in MSCs from MM patients.

Senescent MM-MSCs exhibited decreased differentiation
On account of that cell dysfunction is relevant to cell senescence, we identified the ability of senescent MM-MSC and HC-MSCs to differentiate and to promote tumor cell proliferation. The osteoblastic and adipogenic differentiation capabilities of MSCs were assessed by immunohistochemical method and associated genes expression analysis. In comparison with HC-MSCs and NS-MM-MSCs, senescent MM-MSC showed significantly reduced osteogenic differentiation potential, which is indicated by the results of mineralization analysis and activated ALP evaluation (Fig. 2a-c). In accordance with the immunohistochemical staining analysis, the mRNA expressions of RUNX2 and ALP, which play crucial role in the process of osteoblastic differentiation, were notably decreased in senescent MM-MSCs undergoing osteoblastic differentiation (Fig. 2d, e). The ability of adipogenic differentiation was identified by Oil Red-O staining, and it was increased significantly in HC-MSCs than that in senescent MM-MSCs (**p < 0.01; Fig. 2f, g). The differences of the adipogenic differentiation capability between HC-MSCs and senescent MM-MSCs were also revealed in the expressions of FABP4 and C/EBPa mRNAs (Fig. 2h, i).
In addition, there exists a significant difference in these two group of MM patients. The patients with high β-gal staining MM-MSCs (S-MM) showed different disease progression compared to those with low β-gal staining MM-MSCs(NS-MM). Progression free-survival (PFS) was calculated from the date of the diagnosis to the date of progression. Patients with less than 4.4% SA-β-galpositive cells at diagnosis, had significantly longer PFS (mean of 29 v 11 months; p = 0.014; Supplementary Figure S1) than patients with more than 4.4% SA-β-galpositive cells.

Low expression of Dicer1 was exhibited in MM-MSCs, and MM cells induced senescence in HC-MSCs
The attenuation of Dicer1 resulted in rapid cell senescence in a variety of cells. We examined the level of Dicer1 in in vitro expanded and primary (CD271 + ) MSCs derived from MM patients and healthy controls. Results from RT-PCR and western blot exhibited that the expression of Dicer1 was declined in MM-MSCs ( Fig. 3a-d).
To determine whether myeloma cells could affect MSCs, HC-MSCs were co-cultured with myeloma cell lines (such as NCI-H929, OPM-2, and KMS-12-BM) or with primary CD138 + tumors cells from three MM patients for 24 h and the expressions of Dicer1, p21, miR-93 and miR-20a was determined by RT-PCR. We used trans-well inserts with 1-μm pores. MM cell lines or MM primary cells were cultured in the upper chamber of the inserts. We observed that myeloma cells could significantly decrease the expression of Dicer1, p21, miR-93 and miR-20a in HC-MSCs (Fig. 3e, f). The cells in the lower chamber were stained by SA-β-gal assay and results confirm that the cells were undergoing senescence. A few of staining were identified in the control MSC, while SAβ-gal activity was clearly obvious in the co-cultured MSC (data not shown).

Knockdown of Dicer1 induced senescence and inhibited differentiation of HC-MSCs
We then constructed a recombinant lentivirus to decrease the expression of Dicer1 to further study the biological effect To demonstrate whether the cells were growing senescent, Dicer1-KD MSCs or control groups were identified . e, f HC-MSC (n = 4) were co-cultured with HMCLs: NCI-H929, OPM-2, and KMS-12-BM (e), or with primary CD138 + MM tumors cells from three MM patients (f) for 24 h. Relative mRNA expression levels of p21, Dicer1, miR-93, and miR-20a in MSC after co-cultured with Myeloma cells. The results are expressed as means ± SD. Compared with controls, the significance was set as *p ≤ 0.05; **p ≤ 0.01 once they reached confluence. On the 14th day, Dicer1-KD MSCs stopped division and exhibited a large, flattened shape. Cells were collected and replanted in fresh medium at higher or lower densities and the proliferation cannot be promoted. To identify that Dicer1-KD MSCs were becoming senescence, cells were then stained by SA-β-gal assay. Light staining was determined in the control and negative groups, while SA-β-gal activity was quite clear in the Dicer1-KD MSCs (Fig. 4e). Exceeding 40% of the Dicer1-KD MSCs were growing senescence after 7 days, in comparison with less than 14% of the control and negative groups (Fig. 4f). The above results reveal that down-regulation of Dicer1 is effective for MSC senescence induction. Furthermore, Dicer1 KD caused an increasing proportion of cells in the G 1 phase and a decrease of those in the S phase without inducing apoptosis (Fig. 4h).
We measured the biological role of Dicer1-KD on the capability of MSCs to differentiate into osteogenic and adipogenic cells. Dicer1-KD MSCs were treated with osteogenic induction and exhibited decreased mineralization when stained with Alizarin red (Fig. 4i). Furthermore, they showed a reduction in ALP activity compared with negative MSCs and control cells (**p < 0.01) (Fig. 4j, k). In the process of osteogenic induction, mRNA levels of ALP and RUNX2 were elevated significantly in the control MSCs. On the contrary, the expressions of these mRNAs increased mildly in Dicer1-KD MSCs (Fig. 4l, m).
We also assessed the ability of MSCs differentiating into an adipogenic lineage treated with knockdown of Dicer1. On Day 21, Oil red-O-stained cultures revealed that excessive 91% of control MSCs could differentiate into adipocytes, exhibiting plenty of cellular lipid droplets. On the contrary, the cellular lipid amount in Dicer1-KD MSCs was rare, consisting of small droplets (Supplementary Figure S2A). In the adipogenic inductive process, the expressions of C/EBPa and FABP4 were significantly  (Fig. 5a). The results of the Edu assay showed that Dicer1-KD MSCs and MM-MSC promoted myeloma cell proliferation, and that the proportion of S phase cells was increased compared with that in the HC-MSC or MSC-Negative groups (Fig. 5a, b).
Relapse and chemo resistance are the main features of the clinical course of MM, and it is well-known that the bone marrow microenvironment protects MM cells against chemotherapy. Here we examined whether MM-MSCs could protect MM cells against chemotherapyinduced apoptosis. Bortezomib is a clinically available proteasome inhibitor that is currently among the most potent chemotherapeutic drugs used in the treatment of MM. When the human MM cell line NCI-H929 or MMderived CD138 + cells were cultured in complete medium with 5 nM Bortezomib for 48 h, approximately 33.4% apoptotic cells (early and late) were observed. However, in the presence of MSC-Dicer1-shRNA and MM-MSC, the proportion of apoptotic cells significantly decreased to 9.8% and 10.1%, respectively (Fig. 5c-e).
MSCs interact with MM cells, releasing cytokines which promote the viability of MM cells. Variational chemokine levels could cause the promoted tumor supporting effect of MM-MSCs. The expression of cytokines was then  In the following experiment, we performed MM-MSC transfected with an adenovirus carrying the Dicer1 gene (AD-Dicer1) or green fluorescent protein (AD-GFP). The mRNA and protein expressions of Dicer1 was significantly elevated in AD-Dicer1 MSCs (Fig. 6a-d) levels. It was shown that MM-MSCs transfected with AD-Dicer1 proliferate more quickly than either MM-MSCs or MM-MSCs transfected with adenovirus carrying green fluorescent protein (AD-GFP) (Fig. 6h). There were no evident SA-β-gal-positive cells in MM-MSCs with Dicer1 overexpression (Fig. 6e, f). Furthermore, overexpression of Dicer1 gave rise to an increasing number of cells in the S phase and decreased in the G1 phase (Fig. 6g). Furthermore, AD-Dicer1 MSCs showed increasing osteoblastic (Fig. 6i-m) and adipogenic differentiation ability (Supplementary Figure S4) in comparison with MSCs which was treated with GFP vectors transfection.

MiR-17 family members participated in Dicer1 KD-induced senescence
To identify whether lower expression of Dicer1 in MSC could show effect on miRNA biogenesis and take part in cell senescence, we detected a number of specific miRNAs which possibly target senescence-associated molecules, such as p21. miRBase, PicTar and Targetscans were performed to screen targets. miR-20a expression was significantly decreased in Dicer1-KD MSCs in comparison with both control MSCs and the negative group (Fig. 7a,  b). Decreased expressions of miR-93 and miR-20a was also appeared in MM-MSCs when compared with HC-MSCs. To demonstrate that p21 is a target of miR-93 and miR-20a in MM-MSCs, studies involving miR-93/miR-20a up-regulation were employed. MM-MSCs were transfected with miR-93 and miR-20a harboring lentiviruses, and p21 levels after 48 h transfection was also detected. In three analyzed samples of MM-MSCs, the overexpression led to decreased p21 expression (Fig. 7d) compared with that in cells transfected with scrambled control lentivirus.

Discussion
MM cells are believed to have a reciprocal interaction with the surrounding tumor microenvironment, which consists of endothelial cells, osteoclasts, fibroblasts, macrophages, and other cells. These cells play a crucial part in support of the proliferation, survival, chemo-resistance, and migration of MM cells 5 . A lot of studies showed that the role of the BM microenvironment in cytokines and growth factors secretion that could promote MM pathological mechanism and therefore take part in disease progression. After co-culture with MM cells, normal MSCs showed a phenotype as similar with that of MM-MSCs 25 , demonstrating that MM cells induce this phenotypic change, either by immediate contact or by soluble elements 3,26 . The current results indicated a new mechanism by which BMMSCs act as an oncogenic factor in MM.
In the present study, we first confirmed that MM-MSCs were more likely to be senescent than HC-MSCs. Although MSCs from all MM patients were expanded successfully in vitro, MM-MSCs showed a reduced expansion potential when compared to HC-MSCs. These observations are in line with the findings of several previously published studies, but conflict with other research in which MM-MSCs had normal proliferative characteristics. The typical features of cellular senescence include enlarged morphology, decreased cellular proliferation, and increased SA-β-gal expression. Exactly as expected, most of the cultured MM-MSCs were larger, flatter, irregular and expressed observably higher p21 (the senescence-related molecule). It is clear that the amount of SA-β-gal positive cells was significantly higher in MM-MSCs than HC-MSCs. Our findings are in agreement with previous studies, but these studies did not detail the mechanisms and functional changes associated with senescence in the pathogenesis of MM. The characteristics of senescent MSCs include changes in cellular functions. We demonstrated that senescent MM-MSCs displayed decreasing differentiation capability and increased tumor supporting ability. The depressed osteogenic and adipogenic potentials in senescent MM-MSCs have been verified in vitro by quantitative and continuous detection. In particular, some cell surface molecules and subsequent downstream signaling pathways take part in the regulation of MM-related bone destroying process, in which the balance of bone resorption and formation are no longer acts as a result of the increasing activity of osteoclasts, but rather the osteoblast activity is decreased, leading to an uncoupled or severely imbalanced bone remodeling process. In addition, our previous study showed that MM patients exhibit an impairment of osteogenic differentiation of BMMSCs as compared with healthy controls 27   i Typical images after Alizarin Red S staining on day 21 of osteogenic differentiation. j Relative calcium production (OD 572 nm) by AD-Dicer1 MSC, was significantly higher after 21 days of differentiation as compared with controls. k The ALP activity of AD-Dicer1 MSC was significantly increased after 3 days osteogenic differentiation. l, m Relative RUNX2 and ALP mRNA expression levels. The results are expressed as means ± SD. The average of three replicates is displayed. Compared with controls, the significance was set as *p ≤ 0.05, ** p ≤ 0.01 proportions of S phase myeloma cells. In accordance with our findings, some studies showed that MM-MSCs displayed tumor-supporting capability compared with healthy controls. However, some of the previous reports demonstrated that MSCs possibly have inhibition capability. For example, Ramasamy et al. reported that MSCs act directly on tumor cells, suppressing proliferation and inducing apoptosis in tumor cells 28 . Whether MSCs promote or inhibit myeloma cell growth remains controversial. The diversities in these findings are probably due to the different origin of MSCs, cellular populations and experimental protocols in the respective studies.
In recent years, miRNAs, a pivotal group of moderating factors, have been deemed to be prominent regulators of cellular senescence 18,19 . It is reported that Dicer1 is an RNAse III endonuclease responsible for microRNA biogenesis 22 . Furthermore, our preliminary findings showed that the mRNA and protein levels of Dicer1 in MSCs isolated from MDS patients was significantly decreased compared with MSCs from health controls 17 . On the basis of these findings, we evaluated Dicer1 expression in MM-MSCs. It was demonstrated that the expression of Dicer1 decreased in primary and expanded MM-MSCs. In the current study, we investigated the effects of a lower expression of Dicer1 in MM-MSCs by generating a stable knockdown model. Our study is mainly concerned with the biological function of Dicer1 in cell senescence, differentiation and tumor support of Dicer1-KD MSC expedited cellar senescence, which manifested as flattened cell morphology, increased SA-β-gal positive incidence and a decreased percentage of cells in S phase. Experiments involving Dicer1 overexpression further verified these results. Upregulation of Dicer1 expression in MM-MSC reversed cellular senescence and promoted differentiation. Under-expression of Dicer1 interferes with the biogenesis of miRNA, which is relevant to cellular senescence in endothelial cells. However, the role of miRNA in the senescence of BMMSCs remains ambiguous. Senescent cells generally exhibit increased expression of cell cycle inhibiting factor such as p53/p21, which could be adjusted by miRNA [29][30][31] . In our current study, decreased expression of the miR-17 family (miR-106a/b, miR-17-5p, miR-20a/b, and miR-93) was determined to be a critical factor responsible for elevated p21 expression in Dicer1-KD MSC. Actually, miRNAs in the miR-17 family have been identified as regulators of cell cycle through targeting p21 in many other studies [32][33][34] . It was demonstrated that a lower of expression of the miR-17 family could give rise to the Dicer1-KD-induced cell senescence in MSCs in our present and previous studies. Lately, substantial miRNA expression was reported to be downregulated in MM-MSCs, demonstrating that reduced miRNA expression probably be associated with the pathogenic mechanism of MM. Our result showed that expression of miR-93 and miR-20a remarkably decreased in MM-MSCs. Moreover, overexpression of miR-93/ miR-20a reversed cellular senescence in MM-MSCs through targeting p21. It was confirmed that miRNA showed a critical effect on increasing cell senescence of MM-MSCs.
Dicer1-KD MSCs displayed diminished osteogenic differentiation potential, characterized by the reduction in the expression of osteoblastic maker genes, calcium deposits and ALP activity, which was consistent with previous studies. Furthermore, upregulation of Dicer1 increased the impaired osteoblastic differentiation capability of MM-MSCs. Recently, many studies demonstrated that osteolytic bone lesions of MM patients resulted from the abnormal osteoblastic function in patients 6,[35][36][37] . Our study reported that the reduction of miRNA processing induced by Dicer1-KD was effective to immensely decrease the differentiation ability of MSCs, which indicated that exact level of mature miRNAs is important for MSC osteoblast differentiation. Most importantly, it provides a new perspective for the study of bone lesions in MM.
The senescent MSCs not only exhibited an impaired differentiation potential but also an enhancement in their support of tumor growth. So far, knowledge concerning the effect of MSC senescence in tumor growth support is restricted. In our results, in addition to the classic features of cell senescence, Dicer1-KD MSCs exhibited a significant increase in their capacity to support myeloma cells. Furthermore, BMMSCs derived from patients with MM were found to deliver the cytokine IL-6, which is known to promote tumor formation and progression. The tumor microenvironment-derived GDF15 was also able to significantly increase cell survival in primary MM cells.

Conclusions
In this study, we explored the feature and molecular mechanism of MSCs senescence in MM patients. Dicer1 plays a crucial role in the senescent process of MM-MSCs by regulating miRNA. Senescent MM-MSCs have a decreased differentiation potential and an increased tumor-supporting capacity. Therefore, much attention should be paid to the therapeutic exploitation of MSCs in bone lesion diseases and tumor progression.

Patients
A total of forty-six patients with newly diagnosed MM (n = 46) were enrolled between January 2015 and June 2017 in this study (Table 1). All experiments were approved by the Ethics Committee of Shanghai Jiao tong University Affiliated Sixth People's Hospital (Shanghai, China) and written informed consent was obtained from all subjects. Patients were diagnosed according to the criteria recently defined by the International Myeloma Working Group 38 , and were staged according to the criteria of ISS 39 . For the patients enrolled, they were all uniformly treated with bortezomib and dexamethasone. The median follow-up was 24 months (rang 9-38 months). The bone marrow was extracted for experiments before the treatment start. Eighteen healthy controls with the median age of 66 years (age range, 46-79 years) were analyzed in our study and they were matched for gender and age.

Cell culture
Bone marrow MSCs from patients and controls were isolated, cultured and harvested. The procedure was detailed in our previous studies. Human MM cell lines NCI-H929, OPM-2, and KMS-12-BM were cultured in RPMI 1640 medium (Gibco, Grand Island, NY, USA), supplemented with 10% fetal bovine serum (Gibco), 100 units/mL penicillin, and 100 mg/mL streptomycin.

Cell proliferation
The proliferation of BMMSCs was detected by using CCK-8 (Dojindo, Japan) and by measuring the cell doubling time from P1 to P5. To evaluate the clonogenic potential of BMMSCs, colony-forming unit fibroblast (CFU-F) at P1, P3, and P5 were counted (colonies were stained with Giemsa).

Co-culture of BMMSCs with myeloma cells and 5-Ethynyl-2′-deoxyuridine (EdU) incorporation assay
The human multiple myeloma cell lines NCI-H929 were purchased from the American Type Culture Collection (ATCC, Manassas, VA, USA). The cells were cultured in RPMI 1640 (Gibco) medium supplemented with 10% fetal bovine serum (Gibco),100 units/mL penicillin, and 100 mg/mL streptomycin in a humidified atmosphere with 5% CO 2 at 37°C. NCI-H929 were co-cultured with MSCs (HC-MSCs or MM-MSC) for 24 h. We used a trans well inserts with 1-μm pores. MM cell lines or MM primary cells were cultured in the upper chamber of the inserts. (For cell cycle and proliferation analysis, cells were exposed to 10 μM EdU (Invitrogen) in culture medium for 45 min. EdU was detected with Alexa Fluor 647-azide with the use of a Click-iT EdU Flow Cytometry Assay Kit (Invitrogen). Flow cytometry was performed with a FACS Calibur flow cytometer (Becton Dickinson, SanJose, CA, USA) and Cell Quest software.

SA-β-gal assay
BMMSCs were plated into 6 well culture-plates and maintained until 80% confluence. The cells were incubated with SA-β-gal staining solution (Beyotime, China) according to the manufacturer's instructions. Under inverted microscope, the number of SA-β-gal positive cells (blue cells) were calculated as senescent cells.

Cell cycle analysis
Cells were fixed by precooled ethanol for at least 24 h and stained with the PI/RNase Staining Solution kit (Invitrogen, Carlsbad, CA, USA) according to the manufacturer's directions. The cell cycle analysis was evaluated by flow cytometry and MultiCycle software.

Cell apoptosis analysis
Apoptosis of MM cells was induced by Bortezomib (Selleck, Shanghai, China). The quantification of apoptotic cells was performed by using the AnnexinV-FITC Apoptosis Detection Kit (Invitrogen, Carlsbad, CA, USA) according to the manufacturer's instructions. The analyses were performed on the FACScan flow cytometer.

Real-time PCR
Total RNA was isolated using the RNeasy Mini Kit (Qiagen, Germany) according to the manufacturer's instructions. For mRNA detection, reverse transcription of RNA was performed using the ReverTra Ace qPCR RT Kit (TOYOBO, Osaka, Japan), and real-time PCR (RT-PCR) was carried out using SYBR® Premix Ex Taq™ II (Tli RNaseH Plus) (Takara, Kusatsu, Shiga, Japan). The primers are listed in Supplementary Table. The process was detailed in our previous study 40 .

Dicer1 shRNA transductions
The HC-MSCs were transfected with shDicer1-eGFP vectors (shRNA group) and negative vector. And detailed information about cell transductions is provided in our previous articles 17 .

Cell transductions of adenoviral vectors for expression of Dicer1
The process of the construction of adenoviral vectors for expression of Dicer1 was detailed previously 17,41 . The MM-MSCs were planted in 6-well plates and cultivated for 24 h. The cells were transfected with 20 μl adenoviral vectors and cultured for 12 h and then the medium was changed. The cells were cultured for another 36 h and harvested. The transfection was performed according to the manufacturer's protocol.

Statistical analysis
The data were presented as mean ± SD. All statistical analyses were performed using the SPSS 17.0 System. Comparison of mRNA levels between healthy controls and different MM subtypes was using a Student's t test, one-way analysis of variance (ANOVA) was used to assess multiple pairwise comparisons p < 0.05 was considered statistically significant. Estimation of PFS was performed using the method of Kaplan and Meier.