Characterization of Differentially Expressed miRNAs by CXCL12/SDF-1 in Human Bone Marrow Stromal Cells

Introduction

Recent developments in regenerative medicine seek to utilize adult stem cells as therapeutics to treat age-related musculoskeletal disease. Gene reprogramming via manipulating the cellular microenvironment is a promising and unique opportunity to direct adult stem cell fate. Bone marrow stromal cells (BMSCs) are mesenchymal lineage progenitors that give rise to multiple cell types, including osteoblasts, chondrocytes, adipocytes, and other connective tissue cells [1, 2]. BMSCs are thought to play an important role in the repair and maintenance of various musculoskeletal tissues. The migration and differentiation of adult stem cells, including BMSCs, is a highly regulated by interaction with the cellular milieu. Molecular factors within the milieu, such as growth factors and chemokines, play a vital role in the extracellular regulation of BMSCs [3].

Chemokines are a class of cytokines that induce cell chemotaxis to areas of inflammation. Stromal cell-derived factor 1 (SDF-1), also known as CXCL12, is a chemokine that plays a unique role in the migration of progenitor cell types [4]. SDF-1 mediates its effect primarily through interaction with CXC chemokine receptor-4 (CXCR4) [5]. SDF-1 is known to mediate cell migration, osteogenic differentiation, and cell survival of BMSCs [6,7,8,9,10,11,12]. Our group previously reported that SDF-1 accelerates osteogenic differentiation of BMSCs by regulating osteogenic factors (e.g. RUNX2, BMP2, osteocalcin, collagen alpha 1 type 1) [10]. We also reported that SDF-1 promotes BMSCs migration and osteogenic differentiation in vitro, as well as bone formation in vivo [8, 10]. Similarly, other groups found SDF-1 to be implicit during endochondral ossification and fracture healing, specifically the differentiation of BMSCs into hypertrophic chondrocytes [13,14,15]. Due to the clinical significance of these findings, understanding the molecular mechanisms by which SDF-1 exerts its effects on BMSCs proves highly valuable.

Given the ability of SDF-1 to influence various signaling pathways within BMSCs, we postulate that it likely mediates its control via epigenetic regulation [16]. Epigenetic regulation is a mechanism by which gene expression changes occur without changes in genetic makeup [17]. Epigenetic factors, including DNA methylation and miRNAs, are known to be implicit in the differentiation of BMSCs and musculoskeletal development [18,19,20,21,22]. MiRNAs are small non-coding RNAs that bind mRNA at the 3′-untranslated regions (3′-UTR). Upon binding, the miRNA prevents translation or promotes degradation of the mRNA, thus negatively regulating gene expression at the post-transcriptional level [23, 24]. Several reports suggest that miRNAs regulate almost all cellular events including cell proliferation, differentiation, and development [25,26,27,28]. We hypothesize that SDF-1 induces some of its effects via the regulation of miRNAs. In this study we investigate the SDF-1 dependent regulation of miRNAs within human BMSCs and their correlation with respect to survival, proliferation, migration, and differentiation.

Material and Methods

Results

MiRNA Differentially Expressed Following SDF-1 Treatment

miRNA microarray analysis was conducted to compare miRNA profiles of human BMSCs with and without SDF-1 treatment. We found 104 miRNAs to be differentially regulated (p < 0.05) with an absolute fold change of 1.5 or greater following SDF-1 treatment. Out of these 104 miRNAs, 49 miRNAs were downregulated, and 55 miRNAs were upregulated. Table 1 presents the top 50 most differentially expressed miRNAs. A heat map with hierarchical clustering visually depicts the unique profiles of miRNA from SDF-1 treated and control groups (Figure 1).

Table 1

Select (top 50) differentially expressed miRNAs in response to SDF-1 treatment in human BMSCs.

MicroRNA ID/Probeset ID	Fold-Change	p-Value
hsa-miR-654-5p	−2.5901	0.02547
hsa-miR-339-3p	−2.25636	0.02332
hsa-miR-1226-5p	−2.14647	0.00692
hsa-miR-23b-5p	−2.11548	0.02960
hsa-miR-330-3p	−2.07703	0.01226
hsa-miR-668-5p	−2.06023	0.00677
hsa-miR-3162-5p	−1.99216	0.02709
hsa-miR-3911	−1.94415	0.02799
hsa-miR-6831-5p	−1.8794	0.01496
hsa-miR-92b-5p	−1.87062	0.00466
hsa-mir-4673	−1.83306	0.02000
hsa-miR-493-3p	−1.83031	0.01104
hsa-miR-1233-5p	−1.81938	0.00178
hsa-miR-27a-5p	−1.78655	0.00742
hsa-miR-34c-3p	−1.77045	0.02972
hsa-miR-1275	−1.76887	0.00589
hsa-miR-1268b	−1.70929	0.01389
hsa-miR-4484	−1.69334	0.00523
hsa-miR-4640-5p	−1.67491	0.01385
hsa-miR-423-5p	−1.66952	0.00188
hsa-miR-23a-5p	−1.63963	0.00454
hsa-miR-134-5p	−1.6315	0.00105
hsa-miR-550a-3-5p	−1.58899	0.01087
hsa-miR-324-5p	−1.58876	0.01565
hsa-miR-6850-5p	−1.58559	0.00621
hsa-miR-376a-3p	1.76442	0.01016
hsa-miR-146b-5p	1.77157	0.01273
hsa-mir-548q	1.77686	0.02127
hsa-miR-940	1.8313	0.01507
hsa-miR-505-5p	1.88502	0.03045
hsa-miR-4688	1.88885	0.00777
hsa-miR-6769a-5p	1.92262	0.00585
hsa-miR-151b	2.00276	0.01503
hsa-miR-4284	2.02122	0.04587
hsa-miR-148a-3p	2.04888	0.01323
hsa-let-7f-5p	2.05026	0.00014
hsa-miR-7845-5p	2.12776	0.03009
hsa-miR-6891-5p	2.14821	0.00195
hsa-miR-493-5p	2.17137	0.00349
hsa-let-7d-3p	2.19535	0.04102
hsa-miR-503-5p	2.20447	0.01386
hsa-miR-3121-3p	2.27425	0.00499
hsa-miR-495-3p	2.36848	0.00776
hsa-miR-30b-5p	2.43776	0.03964
hsa-miR-376c-3p	2.47408	0.00670
hsa-miR-140-5p	2.48671	0.00463
hsa-miR-196a-5p	2.68803	0.02805
hsa-miR-1246	2.74544	0.01874
hsa-miR-4706	2.96761	0.01040
hsa-miR-24-2-5p	3.12261	0.00626

Figure 1

Hierarchical clustered heat map depicting miRNA expression in SDF-1 treated human BMSCs versus control (n = 3 each group). Differential expression pattern is observed between the miRNAs of the two groups.

Principal Component Analysis

Principle component analysis (PCA) was performed to distinguish the miRNA profile in SDF-1 treated versus control BMSCs. The PCA depicts unique clusters for treated samples versus control samples highlighting the clear distinction between these two groups (Figure 2).

Figure 2

Primary component analysis mappings of SDF-1 treated human BMSCs versus control. SDF-1 treatment and control group show a unique miRNA clustering pattern for each group. Differential expression pattern confirms that seen in miRNA heat map.

SDF-1 Versus Fracture Healing and Osteogenic Induced MiRNA Expression Profile

GEO dataset archived by Hadjiargyrou et al. (GEO accession GSE76197) was analyzed to determine the differential expression of miRNA in murine femur fracture model across 14 days, and results were compared to that of SDF-1. The analysis determined 12 common miRNAs were significantly upregulated in our study and fracture healing [Hadjiargyrou et al., 2016)] at one or more time points across the 14-day study. Conversely, two common miRNAs were significantly downregulated across both studies. These specific miRNAs are presented in Figure 3. Subsequent GEO search yielded several more datasets depicting differently regulated miRNAs during osteogenic differentiation of human mesenchymal stem/stromal cells (GSE159508, GSE134946, GSE72429, GSE115197). The analysis determined that several common miRNAs were differentially expressed across these studies and with SDF-1 treatment. These results are presented in Table 5.

Figure 3

Common miRNAs regulated in presence of SDF-1 treatment (at 72 hours) and during murine femur fracture healing (across multiple time points ranging from day 1 through day 14). Fracture healing data was retrieved from GEO dataset uploaded by Hadjiargyrou et al. (GEO accession GSE76197) in which murine bone fractures were generated and miRNA-enriched RNA was isolated from calluses at post-fracture days 1, 3, 5, 7, 11, and 14 and compared to intact bone (control) (n = 3). Microarray analysis revealed 306 and 374 up- and down-regulated miRNAs, respectively, of which 14 were similar to our study and presented in this figure with their respective fold changes. Significance in fold change (*) of GEO data was determined by GEO2R adjusted p-value < 0.05.

Table 5

List of common miRNAs differently expressed after SDF-1 treatment and after various osteogenic differentiation protocols (retrieved via GEO database analysis). GSE159508 presents human periodontal ligament stem cells (hPDLSCs) cultured in osteogenic medium for 14 days compared to hPDLSCs cultured in normal medium (10% FBS in DMEM) for 14 days. GSE134946 and GSE115197 presents human BMSCs cultured in osteogenic medium for 7 days compared to non-induced BMSCs from day 0. GSE72429 presents human synovial membrane MSCs and adipose derived stem cells (ADSCs) osteogenically differentiated compared to undifferentiated controls. Differential expression is reported as miRNA fold change after treatment compared to respective control. All reported fold changes are statistically significant based upon GEO2R adjusted p-value.

MicroRNA ID	SDF-1 Treatment (n = 3)	GEO Accession ID
		GSE159508 (n = 3)	GSE134946 (n = 3)	GSE72429* (n = 3)	GSE72429** (n = 2)	GSE115197 (n = 2)
miR-654-5p	−2.59	-	−2.95	-	-	-
miR-339-3p	−2.26	-	−3.03	-	−2.10	-
miR-493-3p	−1.83	-	−3.95	-	-	-
miR-27a-5p	−1.79	−1.26	-	-	-	-
miR-146b-5p	1.77	2.44	-	-	9.25	0.01
mir-548q	1.78	1.39	-	7.29	-	-
miR-940	1.83	1.59	1.32	0.86	-	-
miR-505-5p	1.89	0.70	-	-	-	-
miR-4284	2.02	-	-	1.14	-	-
miR-148a-3p	2.05	1.06	-	-	-	-
let-7d-3p	2.20	1.28	-	-	-	-
miR-24-2-5p	3.12	-	-	0.25	-	-

1. *

   Synovial membrane MSC samples, osteogenic differentiation versus undifferentiated control

2. **

   Adipose derived stem cell samples, osteogenic differentiation versus adipogenic differentiation (control)

SDF-1 Versus BMP2 Induced MiRNA Expression Profile

GEO dataset archived by Bae et al. (GEO accession GSE37036) was analyzed to determine the miRNA expression profile with BMP2 treatment, and results were compared to that of SDF-1. The analysis revealed several common miRNAs were differentially expressed in both studies. Specifically, let-7f-5p, miR-140-5p, miR-196a-5p, and miR-24-2-5p were upregulated in both SDF-1 and BMP2 treatment, as presented in Figure 4. miR-330-3p was downregulated with the treatment of either growth factor. This suggests that both BMP2 and SDF-1 induce osteogenic differentiation by regulating the same downstream osteogenic miRNAs. Furthermore, SDF-1 is known to have potential BMP2 induced bone formation, a phenomenon that could be explained in part by shared downstream osteogenic miRNA.

Figure 4

Common miRNAs regulated in the presence of SDF-1 treatment (at 72 hours) and BMP2 treatment (at 72 hours). BMP2 data was determined from GEO database analysis of dataset by Bae et al. (GEO accession GSE37036) in which C2C12 cells were treated with BMP2 (300 ng/mL) or vehicle (control) for 72 hours (n = 2). Afterwards miRNA was isolated and Exiqon miRNA microarray was performed yielding a total of 34 differentially expressed miRNAs, of which five common miRNAs showed similar trend in the presence of SDF-1 and BMP2 treatment. Reported p-values are adjusted values calculated by GEO2R analyzer that indicate significance in the fold change (differential expression) for BMP2 treatment versus control.

Discussion

Adult bone marrow stromal cells are the primary source of stem cells in the field of regenerative medicine. As mentioned previously, the extracellular milieu plays a paramount role in controlling the migration and differentiation of BMSCs. Manipulating this environment offers a unique opportunity to control cell fate and achieve enhanced therapeutic results. Based on previous findings from our group and others, extracellular SDF-1 is beneficial for osteogenesis and fracture healing [8,9,10, 13,14,15]. We hypothesized that extracellular SDF-1 mediates its effects on BMSCs via miRNA-dependent gene regulation.

While various studies have shown the pervasive role of miRNAs regulating cellular events, no study has previously identified SDF-1 regulation of miRNAs in human BMSCs. Our identification of these miRNAs provides an important link in understanding the mechanism by which SDF-1 exerts its osteogenic effect on BMSCs. Herein we identified a list of miRNAs that were differentially expressed in BMSCs following SDF-1 treatment. Several of these miRNAs are also regulated in osteogenic differentiation and fracture healing. Comparing our findings with the GEO dataset by Hadjiargyrou et al. (in which they analyzed miRNA expression in murine femur fracture) 14 common miRNAs were expressed similarly across both studies (Figure 3). Furthermore, we identified several common miRNAs regulated in the presence of SDF-1 and during osteogenic differentiation of mesenchymal stem/stromal cells (Table 5). We speculate that SDF-1 partially enhances fracture healing and osteogenic differentiation through the regulation of these miRNAs.

Several groups have demonstrated SDF-1 potentiates BMP2 induced bone formation [10, 30, 31]. BMP2 is known to enhance the proliferation and osteogenic differentiation of BMSCs [32,33,34]. Comparing our findings with the GEO dataset by Bae et al. (in which miRNA expression was analyzed in cells treated with BMP2) several common miRNAs were differentially expressed across both studies (Figure 4). These findings suggest SDF-1 and BMP2 could act synergistically via shared downstream osteogenic miRNA. Additionally, our results show SDF-1 downregulates miR-654-5p, a miRNA known to directly target BMP2 within BMSCs (verified by reporter assay) [35]. MiR-654-5p is also observed to be persistently decreased in patients during osteoblast differentiation [36]. Taken together, this is consistent with our prior finding that SDF-1 increases BMP2 mRNA levels in vitro [10]. BMP2 is approved by the FDA for lumbar fusion surgery and is used off-label in many other applications, despite concerns of pro-oncogenic effects [37]. Our findings support the use of lower BMP2 dosage in conjunction with synergistic molecules to achieve clinical outcomes with decreased oncogenic risk.

Existing literature indicates that several differentially expressed miRNAs by SDF1 play an important role in osteogenic differentiation of BMSCs. MiR-339 has been found to inhibit the osteogenic differentiation of BMSC via the direct targeting of DLX5, a known factor in osteogenic differentiation [38]. We speculate that SDF-1 likely enhances BMSC osteogenesis via the downregulation of miR-339-3p. MiR-503-5p has been found to promote bone formation by targeting the negative osteogenic regulator SMURF1 within BMSCs [39]. Our results suggest SDF-1 decreases SMURF1 by upregulating miR-503-3p. Let-7f-5p has been found to promote cell survival in various cell lines by targeting caspase-3 and caspase-9 while increasing levels of anti-apoptotic factors Bcl-2 and Bcl-xL [40, 41]. Taken with our results, this indicates an increase in cell survival of BMSCs mediated by SDF-1, consistent with previous reports [9]. We also identified several novel miRNAs with no prior known functions. Bioinformatic analysis revealed a considerable number of these miRNAs are involved in stem cell homing and commitment to musculoskeletal lineage (Table 2).

To better understand the collective effect of differentially expressed miRNA, KEGG pathway annotation and GO enrichment analysis was conducted. KEGG annotation identified several cellular pathways potentially affected, with fatty acid biosynthesis, thyroid hormone signaling, and mucin-type O-glycan biosynthesis pathways most relevant. Fatty acids and their metabolites have been well linked to stem cell proliferation and differentiation [42]. Both saturated and unsaturated fatty acids have differential effects on BMSC survival [43]. Furthermore, polyunsaturated fatty acids have been found to modulate proliferation, migration, and differentiation of various tissue-specific sources of MSCs [44,45,46]. GO analysis identified several cellular processes (TLR signaling, mitotic cell cycle, post-translation protein modification, extracellular matrix disassembly, apoptotic signaling pathway) that are vital in BMSC survival, proliferation, migration, and differentiation.

SDF-1 signaling has been recognized for its chemotactic and osteogenic function in BMSCs. Our findings further solidify its expanded role in BMSC biology. We have elucidated several miRNAs that are upregulated and downregulated in response to SDF-1. Among these miRNAs, we correlated some of their functions with roles identified in existing literature, and we characterized novel miRNAs concerning their potential targets in BMSCs (Table 2). We also determined key signaling pathways that are collectively influenced by the differentially expressed miRNAs. Limitations to our study do exist. Our study only utilized BMSCs in tissue culture and thus needs to be validated in vivo. Moreover, our study only utilized one concentration and one time period for SDF-1 treatment. It is valuable to understand the dose and time-dependent effects of SDF-1 in vitro and in vivo models. Overall, our study indicated that SDF-1 induces some of its osteogenic and chemotactic effects by regulating miRNAs in human BMSCs.