Oxidative Stress Down-Regulates MiR-20b-5p, MiR-106a-5p and E2F1 Expression to Suppress the G1/S Transition of the Cell Cycle in Multipotent Stromal Cells

Oxidative stress has been linked to senescence and tumorigenesis via modulation of the cell cycle. Using a hydrogen peroxide (H2O2)-induced oxidative stress-induced premature senescence (OSIPS) model previously reported by our group, this study aimed to investigate the effects of oxidative stress on microRNA (miRNA) expression in relation to the G1-to-S-phase (G1/S) transition of the cell cycle and cell proliferation. On global miRNA analysis of the OSIPS cells, twelve significantly up- or down-regulated miRNAs were identified, the target genes of which are frequently associated with cancers. Four down-regulated miR-17 family miRNAs are predicted to target key pro- and anti-proliferative proteins of the p21/cyclin D-dependent kinase (CDK)/E2F1 pathway to modulate G1/S transition. Two miR-17 miRNAs, miR-20-5p and miR-106-5p, were confirmed to be rapidly and stably down-regulated under oxidative stress. While H2O2 treatment hampered G1/S transition and suppressed DNA synthesis, miR-20b-5p/miR-106a-5p over-expression rescued cells from growth arrest in promoting G1/S transition and DNA synthesis. Direct miR-20b-5p/miR-106a-5p regulation of p21, CCND1 and E2F1 was demonstrated by an inverse expression relationship in miRNA mimic-transfected cells. However, under oxidative stress, E2F1 expression was down-regulated, consistent with hampered G1/S transition and suppressed DNA synthesis and cell proliferation. To explain the observed E2F1 down-regulation under oxidative stress, a scheme is proposed which includes miR-20b-5p/miR-106a-5p-dependent regulation, miRNA-E2F1 autoregulatory feedback and E2F1 response to repair oxidative stress-induced DNA damages. The oxidative stress-modulated expression of miR-17 miRNAs and E2F1 may be used to develop strategies to retard or reverse MSC senescence in culture, or senescence in general.


Ivyspring International Publisher
Multipotent stromal cells (MSC) are able to self-renew and exhibit multilineage differentiation ability, thus, presenting great potentials in therapeutic applications in regenerative medicine and cell therapy [9,10]. However, proliferating MSC ages rapidly in culture, hampering therapeutic potentials [10,11]. The mechanism of such debilitation is not well understood. We have previously reported the establishment of oxidative stress-induced premature senescence (OSIPS) in Wharton's Jelly (WJ)-MSC by treatment with H2O2 [12]. H 2 O 2 treatment of cells causes accumulation of ROS to provoke oxidative stress, which, in turn, induces DNA single-and doublestrand breaks and telomere losses to accelerate senescence [13,14]. It has been also established that the onset of senescence involves hampered self-renewal properties through disruption of the G1-to-S-phase (G1/S) transition of the cell cycle [15,16].
One of the important regulators of gene expression is the microRNA (miRNA). MiRNAs are small single-stranded non-coding RNAs that regulate expression of target genes via degradation or translational repression of the targeted transcripts. Hence, miRNAs control a broad range of cellular functions in stem cells, including self-renewal, proliferation and regulation of G1/S transition of the cell cycle [17][18][19]. Several miRNAs have previously been reported to be differentially expressed in senescence cells [19][20][21][22][23]. MiRNAs have also been reported to regulate oxidative stress-induced cellular senescence (reviewed in [24]). However, details on pathways that depict interactions between miRNAs and cell-cycle factors in senescence cells remain to be fully elucidated.
The G1/S transition of the cell cycle is closely associated with senescence. Using the OSIPS MSC model, this work focused on further elucidating the biochemical and cellular effects of oxidative stress on miRNA expression in regulating the G1/S phase of the cell cycle. Differentially expressed miRNAs and the target transcripts in the H2O2-induced oxidativestressed MSC were identified, and selected miRNAtargeted genes in the p21/cyclin D-dependent kinase (CDK)/E2F1 pathway were verified by functional studies. The data show that down-regulated expression of miR-20b-5p and miR-106a-5p, both members of the miR-17 family, contribute to regulation of G1/S-phase transition of the cell cycle via promoting p21 expression and suppressing E2F1 in oxidativestressed multipotent stromal cells.

Cell culture of Wharton's Jelly-derived MSC (WJ-MSC)
Human multipotent stromal cell lines (previously called mesenchymal stem cells in [12]) at passage 3 to 5, WJ0706, WJ0619 and WJ2000, derived from Wharton's Jelly (WJ) of the umbilical cord, were obtained with compliments from Cryocord Sdn Bhd, Selangor, Malaysia (http://www.cryocord.com.my), and has previously been described [12]. The cell lines were isolated and the MSC phenotype was confirmed at Cryocord as previously described [25]. WJ-MSC was cultured in complete Dulbecco's Modified Eagle Medium composed of nutrient mixture F12 (DMEM/F12) (Gibco, NY, USA) supplemented with 10% fetal bovine serum (Gibco) and 1% penicillin/ streptomycin. The medium was changed every 3 days and cells were re-plated at a density of 5,000 cells/cm 2 upon reaching 80% confluency. All cells were cultured at 37 o C in 5% carbon dioxide.

Hydrogen peroxide (H 2 O 2 ) treatment
H 2 O 2 treatments were performed on confluent cells (3.75 x 10 5 cells per T75 flask) as previously described [12]. In brief, cells at passage 6 to 8 were incubated in complete DMEM/F12 medium containing 200 µM H 2 O 2 for 2 h; cells were washed and further cultured in fresh complete DMEM/F12 medium without H 2 O 2 for 24 h for recovery before being harvested for subsequent assays. In the time-course experiments, the cells were treated with 200 µM H 2 O 2 for different durations, followed by cell recovery for 24 h before RNA preparation. In the miRNA stability test, the H 2 O 2 -treated cells were cultured in complete medium for different time points before being harvested for RNA preparation.

miRNA microarray analysis
For miRNA profiling analysis, the Gen. 7 version of miRCURY LNA™ microRNA Arrays (Exiqon, Vedbaek, Denmark), which included validated and T m -normalized LNA™-based capture probes, were used. The arrays included 1,896 of the 1,921 annotated human miRNAs in the miRBase 18.0. Included in the arrays were also 25 proprietary miRPlus™ human miRNAs not yet annotated in miRBase. The arrays were done in four replicate spots. The quality of the total RNA was verified by an Agilent 2100 Bioanalyzer profile. Only total RNA preparations with RNA Integrity (RIN) values > 7 were used in the array. Total RNA (450 ng) from both the experimental samples and the reference was labeled with Hy3™ or Hy5™ fluorescent label, respectively, using the miRCURY LNA™ microRNA Hi-Power Labeling Kit, Hy3™/ Hy5™ (Exiqon) following established procedure of the manufacturer. The Hy3™-labeled experimental and a Hy5™-labeled reference RNA samples were mixed pair-wise and hybridized to the miRCURY LNA™ microRNA Array using a Tecan HS4800™ hybridization station (Tecan, Grödig, Austria). The array slides were scanned using the Agilent G2565BA Microarray Scanner System (Agilent Technologies Inc., CA) and image analysis was carried out using the ImaGene® 9 microRNA Array Analysis Software (Exiqon). The quantified signals were backgroundcorrected (Normexp with offset value 10, see [26] and normalized using the global LOcally WEighted Scatterplot Smoothing (LOWESS) regression algorithm. For the present data set, a total of 1,469 probes were discarded by this filtering procedure, leaving 564 probes detectable above the background threshold for each sample. Unsupervised hierarchical clustering was performed using the complete-linkage method and the Euclidean distance measure.

RNA isolation, reverse transcription and quantitative RT-PCR
Total RNA was isolated from WJ-MSC using RNeasy Plus Mini Kit (Qiagen, CA, USA). For miRNA expression analysis, miRNA polyadenylation was first performed using 1 µg total RNA, followed by first-strand cDNA synthesis using the NCode™ miRNA first-strand cDNA synthesis and qRT-PCR kit (Invitrogen, CA, USA). Ten-fold diluted cDNA was used for quantitative RT-PCR using a SYBR select master mix kit (Applied Biosystems, TX, USA) in a Rotor-Gene Q (Qiagen) thermal cycler. The primers used for qRT-PCR analysis are shown in Supplementary Table S1. PCR was carried out under the following conditions: 50 °C for 2 min, 95 °C for 10 min followed by 40 cycles of 95 °C for 15 s and 60 °C for 1 min. Data was analyzed by using the comparative ΔΔCt method, and miRNA expression was normalized to the expression level of U6. All miRNA quantification experiments were performed in triplicates, and the data presented were derived from three independent experiments. miRNA target prediction, ontology and KEGG pathway analysis miRNA target transcripts were predicted using the web-based algorithms miRBase (www.mirbase. org), TargetScan Release 7.1 (http://www.targetscan. org/), DIANA-microT-CDs v5.0 (http://diana.imis. athena-innovation.gr) and miRTarBase (www.mirtar base.mbc.nctu.edu.tw). Ontology and KEGG pathway analysis were performed by using DAVID Bioinformatics Resources v6.8 [27,28].

Protein extraction and western blotting analysis
Total protein lysates were prepared using the RIPA buffer (50 mM Tris•HCl pH 7.6, 150 mM NaCl, 1% NP-40, 0.5% sodium deoxycholate, protease inhibitor cocktail, 0.1% SDS) (Nacalai Tesque, Kyoto, Japan). Western blot analysis was performed as previously described [29] using the following rabbit primary antibodies: p21 (Cell Signaling, MA, USA), E2F1 (Abcam), CCND1 (Abcam) or GAPDH (Cell Signaling). The relative optical density values of the proteins bands were quantified by comparing the H2O2-treated cells relative to untreated cells, or miRNA mimic-transfected cells relative to the negative mimic control cells, after normalization to the GAPDH loading control using Image J software version 1.49v (National Institutes of Health, MD, USA).

Luciferase reporter assay
PCR products of full-length or segments of 3'-untranslated region (3'UTR) of the target genes harboring the predicted miR-20b-5p or miR-106a-5p binding sites were cloned into the dual luciferase (Firefly/Renilla) reporter vector pmirGLO using the restriction enzyme combinations SacI/XbaI or SalI/ XhoI [30]. The colorectal cancer cell line HCT-15 was co-transfected in 24-well plates using 1µl Lipofectamine 2000 (Invitrogen) with 200 ng empty pmirGLO or the pmirGLO/3'-UTR construct, and 10nM miRNA mimic or a validated nonspecific negative control mi-RNA mimic (Ambion, CA, USA) and the subsequent luciferase assay was performed as described [29].

Transfection of miRNA mimic and BrdU assays
To test the effects of miRNA over-expression, WJ0706 cells were transfected with 25 nM or 50 nM of miRNA mimics, or a validated nonspecific negative control miRNA mimic (Ambion) in 24-well plates using 1µl Lipofectamine2000 (Invitrogen) according to the manufacturers' protocols. Twenty-four hours post transfection, fresh medium was changed and the cells were harvested 48 h post-transfection for BrdU assays (Cell Signaling Technology) in triplicates in three independent biological experiments as described [29].

Cell cycle analysis
Twenty-four hours post-H 2 O 2 treatment and forty-eight hours post-miRNA transfection, the cells were harvested and were fixed overnight with 70% cold ethanol at -20 °C. The cells were then treated with 1mg/mL RNase A and stained with 10µg/mL propidium iodide in the dark for 30 min at 37 o C, followed by analysis on a FACS Canto-II analyzer (BD Biosciences, CA, USA). A total of 100,000 events were recorded for each sample. The experiments were done three times independently before quantification and statistical analysis.

Statistical analysis
Statistical analysis was performed in Microsoft Excel 2010 using the statistical formula to obtain the mean values, standard deviation and variance. Realtime RT-PCR results were reported as average of log (fold change) ± standard error. All data were analyzed by Student's t test using the negative controls as reference. Statistical significance was accepted at p<0.05.

Global expression analysis identifies deregulated miRNAs associated with signaling processes, cell cycle regulation and tumorigenesis under oxidative stress
To identify oxidative stress-associated miRNAs in the OSIPS cells, global miRNA expression profiles of three WJ-MSC lines, WJ0706, WJ0619 and WJ2000, and the respective H 2 O 2 -treated cells, were established by microarray analysis. Only twenty-eight upregulated and twenty-two down-regulated miRNAs were identified in the H 2 O 2 -treated cells relative to the untreated cells (Supplementary Table S2). A heatmap was constructed using the fifty deregulated miRNAs for unsupervised hierarchical clustering analysis: the data were clustered into the control (untreated) and H2O2-treated groups as anticipated ( Figure 1A). Out of the fifty deregulated miRNAs, only seven upregulated miRNAs answered to the criteria of log 2 (fold change)>1.0 and p<0.05, and five downregulated miRNAs fulfilled the criteria of log 2 (fold change) <-1.0 and p<0.05 (Table 1); the twelve miRNAs were further subjected to gene ontology and KEGG pathway analyses. Amongst the top 10 affected biological and molecular processes (Supplementary  Table S3), most notable are 258 genes expressed in processes associated with gene regulation (gene count n=85, 32.9%), kinase activity and signaling (n=42, 16.3%) and cell cycle regulation (n=5, 1.9%) ( Table 2). In particular, the targeted processes of posttranscriptional gene silencing by RNA (GO:0035194) and miRNA-mediated inhibition of translation (GO: 0035278) are consistent with the post-transcriptional regulation properties of miRNAs. KEGG analysis also predicts involvement of signaling pathways (hsa04012 & hsa04070, n=13) and the cell cycle process (hsa04110) ( Figure 1B, asterisked) in both the up-and down-regulated miRNA groups. It is noteworthy that twenty-eight (45.9%) of the sixty-one predicted genes in the KEGG pathways of the up-regulated miRNAs, and forty-eight (73.8%) of the sixty-five downregulated miRNAs are associated with pathogenesis pathways of various cancers ( Figure 1B).  Data is extracted from the biological and molecular processes in the gene ontology analysis of the significantly and differentially expressed miRNAs in OSIPS cells as presented in Supplementary Table S3. Combining the up-and down-regulated biological and molecular processes, there are 258 genes; the percentages shown in brackets are percentages of the total gene counts.  Supplementary Table S2 for a full list). Color code for relative miRNA expression levels: red and green represent expression levels below and above the reference channel, respectively. Normalized log2(fold change) ratio values were used for the analysis. The down-regulated miR-17 family miRNAs that were further investigated in this work are boxed (see also Table 1). (B) Top 10 KEGG pathways of the differentially expressed miRNAs. The number of predicted genes in each cancer-associated pathway is shown on the right end of the bar, and the percentage of the total number of the predicted cancer genes is shown in the box. Asterisks indicate pathways that are associated with signaling or the cell cycle (see also Supplementary Table S3). In the up-regulated miRNA group (Table 1), miR-146a-5p and miR-146b-5p have previously been shown to be deregulated in expression in late-passage fibroblast cells [31]. The remaining five up-regulated miRNAs were first reported in deep sequencing of small RNA transcriptomes of several types of cancer [32][33][34][35][36]. In the down-regulated group, miR-16-5p is a tumor suppressor miRNA reportedly involved in regulating the cell cycle and apoptosis [37], consistent with the gene ontology data above ( Figure 1B & Table  2). The remaining down-regulated miRNAs all belong to the miR-17 family and are highly homologous in the sequences, and share the same seed sequence (Figure 2A) [38]. Down-regulated expression of the four miR-17 family members was confirmed in real-time RT-PCR analysis in the three WJ-MSC cell lines after H2O2 treatment ( Figure 2B). Of the four miR-17 miRNAs, the chromosome 13-associated miR-17-5p and miR-20a-5p have previously been studied in relation to differentiation, tumorigenesis and the G1-to-S-phase transition of the cell cycle [30,[39][40]. However, the chromosome X-linked miR-20b-5p and miR-106a-5p have not previously been investigated, and were the focus of the present work in relation to oxidative stress-induced deregulation of the G1/S-phase transition.
To first investigate the expression kinetics of miR-20b-5p and miR-106a-5p under oxidative stress, WJ0706 cells were treated with H2O2 for various periods of time. Data showed that H 2 O 2 treatment for 0.5 h was sufficient to rapidly down-regulate expression of both miRNAs, and the suppressed miRNA levels were maintained for up to 2 h ( Figure  2C). The data, thus, established a direct and rapid responsiveness of miR-20b-5p/miR-106a-5p to H2O2-induced oxidative stress. It was further shown that after 2 h H 2 O 2 treatment, the suppressed miRNA expression levels were maintained for up to 48 h in the cell recovery period in fresh medium without H 2 O 2 ( Figure 2D), indicating irreversible long-term miRNA down-regulation in the oxidative-stressed cells.

G1/S transition and DNA synthesis
In stem cells, the G1/S transition of the cell cycle is accelerated [15,[41][42] while it is retarded in senescence cells [43][44][45]. The effects of H 2 O 2 -induced oxidative stress on the cell cycle were evaluated in flow cytometric (FACS) analysis. Representative FACS profiles of the three cell lines analyzed are shown in Figure 3A; combined data of three independent experiments of the three cell lines tested are shown in Figure 3B. The results showed that the fractions of G1-phase cells significantly increased on H2O2 treatment, with concurrent declines of the S1 and G2 cell populations, indicating hampered G1/S transition [46]. To obtain supporting evidence, DNA synthesis rates of H 2 O 2 -treated cells were determined in BrdU assays ( Figure 3C). The data, indeed, showed that on H 2 O 2 treatment, DNA synthesis rates were significantly suppressed, supporting that the G1-phase cells were blocked, or delayed, from entry into the S phase.
To elucidate the association between the miR-20b-5p and miR-106a-5p and cell proliferation, transient over-expression of the miRNA mimics in H2O2-treated and late replicative senescence (RS) WJ0706 cells was performed, followed by cell counts in cell growth analysis. Since the two miRNAs shares an identical seed sequences and a high degree of sequence homology (18/21) (Figure 2A), the interchangeability of the two miRNAs was also tested.
On 2 h H2O2 treatment, followed by further culturing in H 2 O 2 -free medium, the negative miRNA mimic-transfected control cells entered growth arrest on day 2 and up to day 10. On the other hand, growth arrest of the miR-20b-5p and/or miR-106a-5p mimic-transfected cells ended when the cells entered the log phase of growth on day 6 to day 8 ( Figure 4A). The growth delay was probably due to delays for the exogenous miRNAs to exert their gene regulatory and the subsequent biochemical actions. The late-passage RS cells did not seem to have a clear growth-arrest phase as shown in the early-passage cells on H2O2 treatment; however, the miRNA mimic-transfected RS cells grew more rapidly than the negative miRNA mimic-transfected control cells ( Figure 4B). Furthermore, the miR-20b-5p and miR-106a-5p mimics, when used individually at 50 nM, or in combination at a combined final concentration of 50 nM, displayed similar growth patterns. The results indicate that the two miRNAs are interchangeable in exerting biological functions, echoing our previous hypothesis that miRNA redundancy provides a fail-proof scheme in miRNA regulation of crucial cellular events [47].  The biological role of miR-20b-5p and miR-106a-5p in modulating cell proliferation, the cell cycle and DNA synthesis was next investigated in MTT, flow cytometry and BrdU assays, respectively ( Figures  4C-4E). In MTT analysis, WJ0706 cells not under oxidative stress maintained a steady growth rate both in the negative miRNA mimic-transfected control cells and in the miRNA mimic-transfected cells; however, miR-20b-5p and/or miR-106a-5p over-expression consistently resulted in higher numbers of viable cells and, thus, higher cell proliferation rates ( Figure 4C). Interchangeability of the two miRNAs was again observed. Likewise, flow cytometric analysis of the miRNA mimic-transfected cells showed that either one of the miRNA mimics resulted in a significant decrease in the G1-phase cell population with concurrent increases in the S-and G2-phase cells when compared with transfection of a negative control miRNA mimic; furthermore, a significant G1 cell reduction was observed on the combined used of both the miRNA mimics ( Figure 4D). The data support that miR-20b-5p and miR-106a-5p play a role in modulating the G1/S transition. The finding was further supported by data of BrdU analysis, which showed significantly enhanced DNA synthesis rates on single or co-transfection of the two miRNAs ( Figure 4E).
Taken together, evidences presented in this (Figures 3 & 4) and the preceding section ( Figure 2) indicate that H2O2-induced oxidative stress leads to down-regulated expression of miR-20b-5p and miR-106a-5p, and suppresses G1/S transition and DNA synthesis. On the other hand, the miR-20b-5p and miR-106a-5p over-expression enhances cell growth and proliferation, the G1/S transition and DNA synthesis. Since miRNAs are negative regulators, the H2O2 and miRNA data are consistent.

miR-20b-5p/miR-106a-5p and oxidative stress modulate the p21/CDK/E2F pathway
In normal cells, enhanced cell proliferation may be due to more rapid G1/S-phase transition via concerted regulation of pro-and anti-proliferative factors of the p21/CDK/E2F pathway [11,17]. Database interrogation and bioinformatics analysis have predicted that the miR-20a-5p and miR-106a-5p target multiple components of the p21/CDK/E2F1 pathway, which modulates the G1/S transition of the cell cycle ( Figure 5), as has been forecast by KEGG pathway analysis above ( Figure 1B & Table 2). In this work, further functional analysis was done on the p21 protein, the downstream cyclin D1 and D2 (CCND1/2) and E2F1.
MiR-20b-5p/miR-106a-5p targeting of transcripts of p21, CCND1, CCND2 and E2F1 was first verified by luciferase assays (Figures 6A & 6B). The putative miRNA target sites, which are common for both miRNAs (Figure 6A), was cloned into the dual luciferase (Firefly/Renilla) reporter vector pmirGLO. In the presence of a miR-20b-5p mimic, the p21 and E2F1, but not CCDN1/2, constructs showed reduced luciferase activities, whereas the miR-106a-5p mimic down-regulated luciferase expression of all four proteins ( Figure 6B), signifying miRNA targeting of the transcripts. MiRNA targeting was further demonstrated by an inverse relationship in the miRNA levels and levels of the targeted transcripts and proteins in the miRNA mimic-transfected cells. Over-expression of miR-20b-5p or miR-106a-5p significantly down-regulated the p21, CCND1 and E2F1 mRNA levels ( Figure 6C, left panel). Likewise, the proteins levels were also down-regulated ( Figure  6C, right panel) supporting miRNA targeting.
Under H2O2-induced oxidative stress, results showed elevated p21 and CCND1 mRNA and protein levels in the different MSC cell lines tested ( Figure  6D). However, the E2F1 mRNA levels were all significantly low in all the three cell lines tested ( Figure 6D, left panel); however, the E2F1 protein level was lower than the control only in WJ2000 cells, but no significant changes were observed in the other two cell lines on H2O2 treatment ( Figure 6D, right panel). Taken together, the data show that miR-20b-5p and miR-106a-5p target and down-regulate expression of p21, CCND1 and E2F1, and that oxidative stress exerts the reverse effects in up-regulating p21 and CCND1 expression, but not E2F1. Both the miRNA and oxidative stress effects may be linked to their effects on suppressed cell proliferation and DNA synthesis and hampered G1/S transition described above.

Discussion
Oxidative stress-deregulated miR-17 family miRNAs are frequently associated with cancers and senescence In this study, twelve miRNAs that are significantly deregulated in oxidative stress-induced senescence cells were identified (Table 1), many of which are predicted to target genes associated with various cancers ( Figure 1B; Supplementary Table S3). MiR-146a has previously been shown to be up-regulated by p53-binding protein-1 to regulate NF-kB activities associated with metastasis of xenografted breast cancer cells in nude mice [48]. Besides the two miR-146 family miRNAs, the other five up-regulated miRNAs were first reported in deep sequencing of small RNA transcriptomes of malignant B cells, melanoma and breast cancers [32-36], and have not been fully annotated and investigated. It is, however, noteworthy that the gene encoding miR-4732-5p is located in close proximity to the ERBB2 receptor gene, which is frequently amplified and over-expressed in breast cancer cells [34]; miR-4732-5p may be collaterally amplified in cancer cells. In the down-regulated miRNA group ( including this study, on the deregulated expression of the miR-17 family miRNAs are consistent with the notion that oxidative stress, and possibly cellular senescence, share some common features with the tumorigenesis process [3-4, 54,55].  In the up-regulated miRNA group, deregulated expression of miR-146a has previously been shown in long-term culturing or late-passage replicative senescence human fibroblast and endothelial cells [31,56]. Besides possible roles in tumorigenesis, the chromosome 13-associated miR-17 and miR-20a have also previously been investigated in replicative aging models, including MSC [57]. The chromosome Xassociated miR-20b-5p and miR-106a-5p were shown in this work to target both pro-and anti-proliferative proteins (Figure 4), echoing previous reports [30,[38][39][40]. In stress-induced premature senescence fibroblasts, miR-17 suppresses p21 expression [58], as was also observed for miR-20b-5p and miR-106a-5p, in this work ( Figure 6C).

A proposed scheme on miR-20b-5p/miR-106a-5p-dependent and -independent regulation of E2F1 expression and G1/S transition under oxidative stress
To investigate the role of miRNA in oxidative-stressed cells in this work, miR-20a-5p and miR-106a-5p were first identified to be down-regulated in H 2 O 2 -treated cells (Figures 1 & 2). H 2 O 2 treatment was further shown to suppress G1/S transition and DNA synthesis (Figure 3) whereas the opposite effects were shown when miR-20b-5p and miR-106a-5p were over-expressed (Figure 4). Since miRNAs are negative regulators, the H 2 O 2 and miRNA data are consistent. MiR-20a-5p and miR-106a-5p targeting of p21, CCND1 and E2F1 was predicted by bioinformatics analysis (Figure 5), and was confirmed in luciferase assays ( Figures 6A & 6B) and in correlation with down-regulation of the three targets on miRNA over-expression ( Figure 6C). On H2O2 treatment, p21 and CCND1 levels were up-regulated ( Figure 6D). However, H 2 O 2 treatment resulted in suppressed E2F1 mRNA levels in all the three MSC cell lines tested, and suppressed E2F1 protein level only in WJ2000 cells, but the E2F1 levels were not significantly altered in the other two cell lines ( Figure 6D).
Based on the data presented in this and a previous work [12], and also drawing literature on the analysis of other miR-17 members, a scheme is proposed here to explain the E2F1 response and in relation to miR-20a-5p/miR-106a-5p expression under oxidative stress. It is proposed that under oxidative stress, four separate but inter-related miR-20a-5p/miR-106a-5p-dependent and -independent events collectively modulate E2F1 expression and G1/S transition to affect cell proliferation and oxidative stressed-induced premature senescence (Figure 7). In event (I), oxidative stress downregulates expression of miR-20b-5p and miR-106a-5p, which, in turn, up-regulates p21 expression. In event (II), the two miRNAs independently and individually target and regulate CCND1/2 and CDK4/6. Despite the fact that p21 is an inhibitor of CCND1/2-CDK4/6, CCND1/2 is still up-regulated, and the E2F1 levels are suppressed. To explain the discrepancies, event (III) is proposed based on reports of an auto-regulatory feedback loop between E2F factors and other miR-17 family members, viz. miR-17-5p and miR-20a, and also other miRNAs [59][60][61][62]. MiR-20b-5p/miR-106a-5p, being members of the same miR-17 family, and E2F1 may also be under auto-regulatory feedback control under oxidative stress in that the miRNAs inhibit E2F1 expression, and, on the other hand, E2F1 binds to the promoter of the miR-17 miRNA clusters to block miRNA expression. Event (IV) is miRNA-independent in which oxidative stress triggers DNA damages, which activate the DNA damage response to recruit E2F1 and CCND1 to the sites of the DNA breaks to promote DNA repair [63][64], and to initiate temporary cell cycle arrest under oxidative stress [65][66]. Hence, the E2F1 levels in the H2O2-treated cells are proposed to be the outcome of the cumulative effects of the interactions and balances between the four miRNA-dependent and -independent events, and are subjected to further modulation by multiple factors, including the physiological state of the cell and the relative spatial-and temporal-dependent abundance of specific factors. Besides affecting the G1/S transition of the cell cycle reported here, H2O2-induced oxidative stress may also exert influences on MSC differentiation (reviewed in [67]) and other cellular process via miRNA regulation. These are important issues worthy of further investigations.

Conclusions
Global miRNA profiling of MSC under oxidative stress and KEGG pathway analysis in this work lend further support that oxidative stress share numerous affected target genes with the tumorigenesis process. This work also shows that oxidative stress down-regulates expression of the miR-17 family miRNAs, and that the miR-17 family members, miR-20b-5p and miR-106a-5p, inhibit expression of E2F1 via targeting p21 and CCND1 to suppress the G1/S-phase transition of the cell cycle, possibly resulting in premature senescence. Understanding the role of the miR-17 family miRNAs under oxidative stress may lead to the development of strategies to retard or reverse senescence in MSC culture, and to facilitate regenerative medicine.