TNF-α is a major cytokine implicated in rheumatoid arthritis (RA), and its expression is regulated at the transcriptional and posttranscriptional levels. However, the impact of changes in microRNA expression on posttranslational processes involved in TNF-α signaling networks is not well defined in RA. In this study, we evaluated the effect of miR-17, a member of the miR-17–92 cluster, on the TNF-α signaling pathway in human RA synovial fibroblasts (SFs). We demonstrated that miR-17 expression was significantly low in RA serum, SFs, and synovial tissues, as well as in the serum and joints of adjuvant-induced arthritis rats. RNA-sequencing analysis showed modulation of 664 genes by pre–miR-17 in human RA SFs. Ingenuity pathway analysis of RNA-sequencing data identified the ubiquitin proteasome system in the TNF-α signaling pathway as a primary target of miR-17. Western blot analysis confirmed the reduction in TRAF2, cIAP1, cIAP2, USP2, and PSMD13 expression by miR-17 in TNF-α–stimulated RA SFs. Immunoprecipitation assays showed that miR-17 restoration increased the K48-linked polyubiquitination of TRAF2, cIAP1, and cIAP2 in TNF-α–stimulated RA SFs. Thus, destabilization of TRAF2 by miR-17 reduced the ability of TRAF2 to associate with cIAP2, resulting in the downregulation of TNF-α–induced NF-κBp65, c-Jun, and STAT3 nuclear translocation and the production of IL-6, IL-8, MMP-1, and MMP-13 in human RA SFs. In conclusion, this study provides evidence for the role of miR-17 as a negative regulator of TNF-α signaling by modulating the protein ubiquitin processes in RA SFs.

MicroRNAs (miRNAs) are a highly conserved set of single-stranded noncoding RNAs (∼19–23 nt in length) that are important in many developmental and physiological processes (1, 2), and their aberrant expression correlated with inflammatory diseases, including rheumatoid arthritis (RA) (25). A key specificity determinant for miRNA target recognition is based on Watson–Crick pairing of the 5′-proximal seed region (nt 2–8) in the miRNA to the seed match site in the target mRNA, which is located primarily in the 3′UTR (6), with a small subset of miRNAs targeting the mRNA 5′UTR and/or the coding region (79). Although the exogenous delivery of different miRNAs was shown to regulate various target genes in specific cellular contexts, the predicted impact of changes in miRNA expression on cellular processes and cytokine signaling networks is difficult to predict. Recent studies showed significant changes in the expression of many genes by individual miRNA overexpression (1012); however, only a portion of differentially regulated genes were predicted direct targets, indicating that most of the changes in gene expression induced by miRNA transfections are indirect (12, 13).

Ubiquitination is a posttranslational modification process that plays a key role in various signal-transduction cascades by priming signaling proteins for degradation or stabilization through Lys-linked K48 or K63 ubiquitin chains, respectively (14). The ubiquitin proteasome system (UPS) consists of ubiquitin ligases (13) and proteomes that mediate posttranslational modifications in the cytokine or TLR signaling network. For instance, upon binding to TNF-α, TNFR1 recruits adapter proteins; the protein kinase RIP1; several ubiquitin E3 ligases, such as TRAF2, cIAP1, and cIAP2; and the deubiquitination (DUB) enzymes for downstream signaling (15). However, TNF-dependent recruitment of multiple ubiquitin ligases and DUB enzymes implies the importance of ubiquitination for regulating inflammation and cell death in this pathway. Being two different spectrums of biological processes, namely epigenetics and posttranslational, the influence of miRNAs on the ubiquitination of TNF-α signaling proteins is not well known in RA.

miR-17–92 is located in the locus of MIR17HG (miR-17–92 cluster host gene), also known as C13orf25 (chromosome 13 open reading frame 25). The miR-17–92 cluster transcript spans 800 nt and encodes six miRNAs that are transcribed from the same promoter (Supplemental Fig. 1A). These six miRNAs can be grouped into four families based on their seed regions: miR-17, miR-18, miR-19, and miR-92. The miR-17 and miR-19 families are composed of the pairs of miRNAs with identical seed regions: miR-19/miR-20a and miR-19a/miR-19b-1 (16). As oncomirs, these miRNAs are known to promote proliferation, inhibit apoptosis, and induce tumor angiogenesis (17, 18). Yet in some contexts, the miR-17 family was shown to negatively regulate cell proliferation (1921) and inhibit cell migration and invasion in cancer (22, 23). Recent studies showed that miR-20a from the same cluster regulates apoptosis signaling kinase (ASK)1, whereas miR-19a/b were shown to regulate IL-6 and matrix metalloproteinase (MMP)-3 expression in LPS-activated RA synovial fibroblasts (SFs) (24, 25). In contrast, TNF-α–induced miR-18a was reported to facilitate cartilage destruction and chronic inflammation in the joint through a positive feedback loop in NF-κB signaling, with a concomitant upregulation of MMPs and mediators of inflammation in RA SFs (26), suggesting differential effects of miRNAs from this cluster. In the current study, contrary to the other miRNAs in the same cluster, miR-17 expression was consistently low in RA synovial tissue (STs) and RA SFs, but not in osteoarthritis (OA) STs or OA SFs, making miR-17 more disease relevant. Thus, this study was undertaken to determine the role of miR-17 in RA pathogenesis.

The results from the current study showed that miR-17 is consistently low in the diseased serum, STs, and SFs, as well as in a rat adjuvant-induced arthritis (AIA) model of RA. To extend these findings, the current study was carried out to assess the effect of miR-17 overexpression on the posttranslational ubiquitination in TNF-α signaling. The results showed that miR-17 overexpression inhibited TRAF2 expression and its association with cIAP2, thereby suppressing TNF-α signaling pathways and downstream inflammatory proteins. This study provides novel insights into the role of miR-17 in downregulating TNF-α signaling by influencing the protein-ubiquitination pathway in RA SFs.

Rabbit polyclonal anti-TRAF2 (#sc-876), mouse monoclonal β-actin (#sc-47778), mouse monoclonal anti–USP-14 (#sc-100630), mouse monoclonal anti–MMP-1 (#sc-58377), rabbit polyclonal anti–MMP-13 (#sc-30073), rabbit polyclonal anti–lamin A/C (#sc-20681), and rabbit polyclonal anti–p-IκB-α (#sc-8404) Abs were purchased from Santa Cruz Biotech (Santa Cruz, CA). Rabbit monoclonal anti-cIAP1 (#7065), rabbit monoclonal anti-cIAP2 (#3031), rabbit anti-USP2 (#8036), rabbit monoclonal anti-RAD23A (#24555), anti-K63 polyubiquitin (#5621), anti-K48 polyubiquitin (#8081), anti–STAT-3 (#9132), anti–NF-κBp65 (#8242), anti–p-c-Jun (S73) (#9164), anti–p-p38 (T180/Y182) (#4511), anti–p-JNK (T183/Y185) (#9251), total JNK (#8690), total p-38 (#9258), and p–STAT-3 (S727) (#9134) Abs were purchased from Cell Signaling Technology (Beverly, MA). TRAF2 mouse monoclonal (#AM1895B) Ab for immunoprecipitation (IP) was purchased from Abgent (San Diego, CA), and anti-PSMD13 (#5937-1) Ab was purchased from Epitomics (Burlingame, CA). Total ASK1 (#ab131506) and p-ASK1 Thr838/845 Abs were purchased from Abcam (Cambridge, MA) and Cell Signaling Technology, respectively. Human Cytokine Array C5 (#AAH-CYT-5) was purchased from RayBiotech (Norcross, GA). SMARTpool ON-TARGET plus ASK1 small interfering RNA (siRNA) or negative control (NC) siRNA was purchased from GE Dharmacon (Lafayette, CO).

The procurement of deidentified human healthy/nondiseased (NL), OA, and RA tissue (ST) was obtained under a protocol approved by the Institutional Review Board (IRB#106628). Human SFs were derived from STs of patients diagnosed with OA or RA from autopsies/amputation who underwent total joint replacement surgery (mostly knee joints) or synovectomy. NL STs from nonarthritis individuals were obtained at the time of autopsy or amputation. ST from 18 RA patients (mean age ± SD: 74.2 ± 8.5 y), 12 OA patients (73.8 ± 12.7 y), and 8 NL subjects (60.5 ± 9.8 y) was used in the current study. The deidentified human NL, OA, and RA STs were obtained from the Cooperative Human Tissue Network (Columbus, OH) and the National Disease Research Interchange (Philadelphia, PA). Tissue specimens were washed by sterile PBS, minced, and processed as previously described (27). SFs were grown in RPMI 1640 containing 2 mM l-glutamine with 10% FBS, at 37°C, in a humidified atmosphere with 5% CO2. Cells were used between passages 5 and 10 for these studies. For some studies, RNA was prepared directly from ST from NL donors, OA and RA patients, or rat AIA or naive joints.

NL, OA, or RA SFs were plated in 60-mm dishes and used when >80% confluent. SFs were serum starved overnight and stimulated or not with TNF-α (20 ng/ml) for the indicated time, and cell lysates were prepared. Human STs and the joint homogenates from the AIA study were also ground to a fine powder in liquid nitrogen using a tissue pulverizer. Pulverized tissue was used to purify total RNA containing the miRNA fraction (miRNeasy kit; QIAGEN, Valencia, CA) to study miR-17, miR-18a, miR-19a, miR-19b, miR-20a, and miR-92 expression.

RA SFs or THP-1–differentiated macrophages were transfected with pre–miR-17 (Life Technologies, Carlsbad, CA) in six-well plates or 100- or 150-mm dishes. RA SFs were transfected with pre-miRNAs (100 nM) of miR-17 with NC–pre-miRNAs (Life Technologies) or anti–miR-17 (150 nM) with NC anti-miRNA using Lipofectamine RNAiMAX transfection reagent (Life Technologies) for 48 h and then stimulated or not with TNF-α (20 ng/ml) for 30 min or 24 h. Total RNA containing miRNA fraction or cell lysate were prepared after treatment. Protein expression was determined using Western immunoblotting. Transfection efficiency was confirmed by the significant upregulation of miR-17 expression using TaqMan assays (Life Technologies). THP-1 cells were differentiated into macrophages by treatment with PMA (300 ng/ml for 3 h) and then transfected with pre-miRNAs (100 nM) or NC using Lipofectamine RNAiMAX transfection reagent (Life Technologies) for 48 h, followed by treatment or not with TNF-α (20 ng/ml) for 30 min. RA SFs (n = 4) were also transfected with ASK1 siRNA or NC siRNA for 48 h using Lipofectamine RNAiMAX transfection reagent (Life Technologies) and then stimulated or not with TNF-α (20 ng/ml) for 24 h. RA SFs were pretreated with the selective inhibitor of ubiquitin-conjugating enzyme E1 (PYR41; 1 μM) for 2 h and transfected with pre–miR-17 or NC–pre-miRNA for 48 h, followed by 24 h of stimulation with TNF-α.

Total RNA was reverse transcribed using a SuperScript First Strand cDNA Synthesis Kit (Life Technologies), according to the manufacturer’s protocol. Expression of miR-17 (also known as miR-17-5p), miR-18a, miR-19a, miR-19b, miR-20a, and miR-92-1 was quantified using TaqMan microRNA Assays with U6snRNA as control (Life Technologies). Expression of ASK1 mRNA was quantified using SYBR Green quantitative real-time PCR, and GAPDH was used as control. Quantification of the relative expression was done using the ∆∆Ct method.

The Exiqon serum RNA purification protocol was followed for the total RNA–containing small RNA fraction using a miRCURY RNA Isolation Kit - Biofluids (Exiqon, Woburn, MA). Serum samples from healthy and RA donors, AIA rats, and naive controls were thawed on ice. A total of 200 μl of human serum or 50 μl of rat serum from each donor was transferred into a 1.5-ml Eppendorf tube and centrifuged at 3000 × g for 5 min at 4°C to remove debris. Serum was transferred into a new 1.5-ml Eppendorf tube, and 60 μl of lysis buffer was added containing 1.17 μl of carrier RNA (0.8 μg/μl) from bacteriophage MS2. Samples were incubated at room temperature for 3 min and subsequently mixed with 20 μl of protein precipitation solution. After centrifugation at 11,000 × g, the aqueous phase containing the RNA was carefully transferred into a new collection tube, and RNA was precipitated with isopropanol. The mixture was applied to an miRNA Mini spin column and washed several times, and RNA was eluted by the addition of 50 μl of RNase-free water. Extracted RNA was processed the same day for cDNA synthesis using the Universal cDNA synthesis kit II and UniSp6 RNA spike-in control primer. Quantitative real-time PCR was performed in 10-μl reactions containing 4 μl of 40× diluted reverse-transcription product, 5 μl of 2× SYBR Green Master Mix, and 1 μl of UniRT LNA PCR primers for miR-17. Reaction mixtures were incubated at 95°C for 10 min, followed by 40 cycles of 95°C for 10 s and 60°C for 1 min, followed by melting curve stage. miRNA-93-5p was used as reference control for sample analysis (28). A threshold cycle (Ct) was observed in exponential phases of the amplification, and quantification of the relative expression levels was determined by the ΔΔCt method.

Ingenuity Pathway Analysis (IPA) was used to interpret the differentially expressed genes in terms of an interaction network that might be altered as a result of RNA changes induced by miR-17 overexpression in RA SFs compared with NC. Genes with p < 0.05 (Student t test) were selected from mRNA Sequencing data, and a list of differentially regulated genes and their corresponding expression values were uploaded to the IPA application software (Ingenuity Systems, http://www.ingenuity.com). IPA was done with standard settings (Ingenuity knowledge case [gene only], direct and indirect, includes endogenous chemicals, consider only relationships where confidence = experimentally observed). TargetScan 7.0, PicTar, and miRanda algorithms were used to identify miR-17 binding sites in TRAF2, cIAP1 (BIRC2), cIAP2 (BIRC3), USP2, and PSMD13 mRNA.

Human RA SFs from two RA patients were transfected with pre–miR-17 or NC–pre-miRNA for 48 h, and total RNA was prepared using an miRNeasy kit (QIAGEN). Next, total RNA integrity was checked using an Agilent Technologies 2100 Bioanalyzer (Santa Clara, CA). A total of 10 ng of high-quality RNA was used to make cDNA for amplification with the Ion AmpliSeq Transcriptome Human Gene Expression Kit (Thermo Fisher Scientific, Grand Island, NY). The cDNA was subjected to 12 cycles of amplification with panel primers and barcoded with adapters, as recommended. Resulting sequencing libraries were quantified by quantitative PCR using SYBR FAST Master Mix (Kapa Biosystems, Wilmington, MA). Sets of eight libraries were balanced and pooled, and sequencing beads were produced on an Ion Chef. Sequencing was performed on an Ion P1 semiconductor sequencing chip using an Ion Proton System (Thermo Fisher Scientific). Data were collected, and primary analysis was performed using Torrent Suite software version 5.0.3. Reads were mapped to the panel, and expression values were determined. R Software version R-3.2.3 was used to generate heat map (29). Microarray data were submitted to the National Center for Biotechnology Information Gene Expression Omnibus under accession number GSE83930 (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?token=ahgtwusgflmxnah&acc=GSE83930).

A luciferase reporter construct encompassing the 3′UTR of ASK1 mRNA (NM_005923) with the binding sites for miR-17 and an empty control reporter construct was obtained from Genecopoeia (Rockville, MD). ASK1 3′UTR sequences were inserted downstream of the secreted Gaussia Luciferase reporter gene driven by the SV40 promoter. A secreted alkaline phosphatase reporter driven by a CMV promoter cloned into the same vector served as the internal control for normalization. RA SFs were cotransfected with ASK1 luciferase reporter construct (1 μg) with pre-miRNA (100 nM) of miR-17 or NC–pre-miRNA (100 nM) using Lipofectamine 2000 transfection reagent (all from Life Technologies). After 48 h, the conditioned media were collected, and luciferase activity was assayed using a Secrete-Pair Dual Luminescence Assay kit (Genecopoeia). Each experiment was repeated in three independent SF donors, and each assay was performed in triplicate.

To study the effect of miR-17 on TNF-α–induced NF-κBp65, p–c-Jun, and p–STAT-3 activation, RA SFs were transfected with pre–miR-17 or NC–pre-miRNA (100 nM) and then stimulated or not with TNF-α (20 ng/ml) for 30 min or 24 h. Upon termination, cells were washed with ice-cold PBS, collected by scraping, and centrifuged at 1500 × g for 5 min at 4°C. Nuclear fractions were prepared, and equal amounts of protein (15 μg) from nuclear fractions were evaluated by Western blotting to study the level of NF-κBp65, p–c-Jun, and p–STAT-3 expression.

SFs were lysed in RIPA lysis buffer with complete protease inhibitor mixture (Roche, Indianapolis, IN). Cell lysates/tissue homogenates/nuclear fractions were resolved on 10% SDS-PAGE gels and transferred to nitrocellulose membranes (Bio-Rad, Hercules, CA). Membranes were blocked with 5% nonfat dry milk powder/BSA in TBS containing 0.1% Tween-20 and probed with 1:1000 diluted polyclonal Abs or mAbs specific for TRAF2, cIAP1, cIAP2, USP2, RAD23A, USP14, PSMD13, K63, K48, ASK1, STAT3, p–c-Jun, p–IκB-α, NF-κBp65, p-p38, p-JNK, total p38, total JNK, MMP-1, MMP-13, or β-actin. Culture supernatants from miR-17–transfected and TNF-α–stimulated RA SFs were concentrated and used for MMP-1 and MMP-13 expression. Immunoreactive bands were visualized using HRP-linked secondary Abs and ECL (Bio-Rad Molecular Imager ChemiDoc XRS+). Images were analyzed using Gel Doc software. Each band was scanned using Image laboratory 5.1 software, and the expression values (pixels per band) are presented as mean ± SE.

Measurement of IL-6 and IL-8 in the culture supernatants of RA SFs was analyzed using ELISA, according to the manufacturer’s instructions (R&D Systems).

We used a commercial cytokine Ab-based array designed to detect 80 cytokines (RayBio Human Cytokine Array C5; RayBiotech). RA SFs were transfected with pre–miR-17 or NC–pre-miRNA for 48 h, followed by TNF-α stimulation for 24 h. Conditioned medium from four patients was pooled for each experimental condition and used in the assay. Experiments were performed essentially as recommended by the manufacturer. Briefly, array membranes were incubated for 1 h in 2 ml of blocking buffer, incubated overnight with 2 ml of the pooled (n = 4; 500 μl from each) culture supernatant, and washed. Next, a mixture of 80 biotinylated Abs diluted 1:250 was added to each membrane (1 ml/array membrane) and incubated for 5 h at 4°C overnight. Membranes were washed, sandwiched Ags were detected by ECL by incubating the membranes with 2 ml of a peroxidase-labeled streptavidin solution (diluted 1:1000), and signals were captured on x-ray films. Arrays were processed simultaneously for image acquisition. Each spot intensity was analyzed using GELDOC software, and the mean value was used. Normalized values were calculated using the formula: X (Ny) = X(y) * P1/P(y), where P1 is the average signal density of the positive control spot on the reference array, P(y) is an average signal density of the positive control spot on array “y,” X(y) is the signal density for a particular spot on array for sample “y,” and X (Ny) is the normalized value of a particular spot “X” on array for sample “y.”

RA SFs were transfected with pre–miR-17 or NC–pre-miRNA in 150-mm dishes for 48 h, starved overnight, and stimulated with TNF-α for 30 min. Cells were washed two times in ice-cold 1× PBS, lysed in 500 μl of RIPA buffer as described earlier, and used for IP assays. Clear lysate was subjected to protein estimation using a DC Protein Assay (Bio-Rad). A total of 3 μg of Ab was used per milligram of whole-cell extract (1 mg) from each sample was subjected to IP using TRAF2 (Abgent), K63-ubiquitin (Cell Signaling Technologies), or K48-ubiqutin (Cell Signaling Technologies) Abs. A similar amount of nonspecific IgG control Ab (Flag M2; Sigma) was used as isotype control. Ab and whole-cell extract were incubated at 4°C on a rotor overnight, followed by incubation with protein G–Sepharose beads for 4 h to capture Ab and protein complex. Beads were subjected to three washes with RIPA wash buffer, followed by a final wash with 1× PBS. Protein beads complex was eluted by boiling in 2× SDS sample buffer and resolved on 4–15% Bio-Rad TGX gel using Western immunoblotting. For K63 and K48 IP assays, 120 μl of bead volume equivalent of Dynabeads (Invitrogen, Thermo Fisher Scientific) was subjected to cross-linking with 18 μl of K63 polyubiquitin Ab or 9 μl of K48 polyubiquitin Ab (Cell Signaling Technologies) using BS3 cross-linker (Thermo Fisher Scientific), as per the manufacturers’ instructions. Beads were divided equally into three samples for IP assay. Protein complex was eluted from cross-linked beads using 2× sample buffer for 15 min at 70°C.

Female Lewis rats (Harlan Laboratories, Indianapolis, IN) weighing 135–160 g were injected s.c. at the base of the tail with 300 μl (5 mg/ml) of lyophilized Mycobacterium butyricum (Difco Laboratories, Detroit, MI) in sterile mineral oil. The day of adjuvant injection was considered day 0. Body weight and ankle circumferences were measured on days 0, 8, and 18 in a blinded manner, as described previously (30). The study also included a naive (no-adjuvant) group for comparison. Naive and AIA rats were sacrificed on days 8 and 18. The circumferences of both hind ankles from each rat were averaged, and n is the number of rats used in each experimental group. Joints and serum samples were collected on day 18. The animal study was approved by the Institutional Animal Care and Use Committee of Washington State University.

The Student t test (unpaired two-tailed t test), followed by one-way ANOVA, was used to calculate statistical differences between different variables. Data shown are mean ± SE, unless stated otherwise. Comparisons were performed using the GraphPad Prism 6 software package, and p < 0.05 was considered significant.

miR-17–92 is a polycistronic miRNA cluster located in chromosome 13 (Supplemental Fig. 1A). To identify the basal expression levels of the miR-17–92-1 cluster in RA, we performed quantitative real-time PCR using RNA from SFs and STs of healthy (NL) donors and OA and RA patients. The expression levels of all six miRNAs (miR-17 [∼69%, p < 0.01], miR-18a [∼81%, p < 0.05], miR-19a [∼87%, p < 0.01], miR-20a [∼85%, p < 0.01], miR-19b [∼79%, p < 0.01], and miR-92-1 [∼80%, p < 0.05]) were significantly decreased in RA ST compared with NL ST (Supplemental Fig. 1B). Interestingly, we observed that miR-17 levels were consistently low in ST and SFs from RA donors compared with those from NL and OA donors (Fig. 1A, 1B). To further verify miR-17 expression, we used a rat AIA model in which inflammation starts around day 8 and peaks around day 18 (31). Transcriptional analysis of pulverized joint tissue from naive and AIA rats also showed a significant decrease in the expression of miR-17 on day 18 compared with the naive group (Fig. 1C, p < 0.05).

FIGURE 1.

Clinical significance of miR-17 expression levels in RA patients. Expression profile of miR-17 in ST (A) and in SFs (B) obtained from NL, OA, and RA donors. (C) Expression levels of miR-17 in AIA (day 18) model of human RA compared with the naive group. miRNA expression levels were evaluated using a TaqMan-based quantitative real-time PCR assay. Data are mean ± SEM for the indicated number of animals. (A–C) U6snRNA was used as a reference control. (D) Lower miR-17 levels in the serum of human RA patients. miR-93-5p was used as a reference control. (E) AIA rats showed an increase in ankle circumferences on day 18. (F) Lower miR-17 expression was associated with RA in these AIA rats. Expression of miR-17 was determined using quantitative real-time. miR-93-5p was used as a reference control. Data are mean ± SEM for the indicated number of serum donors or animals used per group. *p < 0.05, **p < 0.01.

FIGURE 1.

Clinical significance of miR-17 expression levels in RA patients. Expression profile of miR-17 in ST (A) and in SFs (B) obtained from NL, OA, and RA donors. (C) Expression levels of miR-17 in AIA (day 18) model of human RA compared with the naive group. miRNA expression levels were evaluated using a TaqMan-based quantitative real-time PCR assay. Data are mean ± SEM for the indicated number of animals. (A–C) U6snRNA was used as a reference control. (D) Lower miR-17 levels in the serum of human RA patients. miR-93-5p was used as a reference control. (E) AIA rats showed an increase in ankle circumferences on day 18. (F) Lower miR-17 expression was associated with RA in these AIA rats. Expression of miR-17 was determined using quantitative real-time. miR-93-5p was used as a reference control. Data are mean ± SEM for the indicated number of serum donors or animals used per group. *p < 0.05, **p < 0.01.

Close modal

Furthermore, we observed a significant decrease in miR-17 serum levels in RA patients (Fig. 1D, p < 0.05). To further validate these findings, we determined miR-17 levels in the serum of AIA and naive animals. AIA animals exhibited signs of severe arthritis on day 18, as reflected by a significant increase in ankle circumference (∼65%), compared with the naive group (Fig. 1E, p < 0.01). A lower serum miR-17 level, together with the increase in ankle circumferences, was observed in AIA rats on day 18 (Fig. 1F, p < 0.01).

To determine the role of miR-17 in RA, we used a gain-of-function model and performed RNA-sequencing analysis in RA SFs overexpressing miR-17. Among the panel of 20,803 genes, the expression of 15,067 genes, as shown in the representative heat map, was observed in RA SFs transfected with pre–miR-17 and NC–pre-miRNA (Supplemental Fig. 2A). A total of 664 significantly modulated genes (301 upregulated and 363 downregulated) was used for IPA. IPA predicted the protein ubiquitin pathway as a major canonical pathway affected by the differentially regulated genes (Supplemental Fig. 2B). Interestingly, IPA generated an interactome that showed connectivity among various ubiquitin ligases, NF-κB family proteins, AP-1/c-Jun, and 20S and 26S proteasome system (Supplemental Fig. 2C).

The heat map generated based on the list of identified ubiquitin ligases and other proteasome pathway proteins regulated by miR-17 overexpression showed some of the key ubiquitin and deubiquitin ligases in TNF-α signaling, such as TRAF2, cIAP2, USP2, PSMD13, and RAD23A (Fig. 2A, 2B). This suggests that miR-17 may be intimately involved in the regulation of UPS in TNF-α signaling. The IPA results confirmed that miR-17 overexpression may significantly influence canonical pathways with high pathological relevance, including the protein ubiquitin pathway, TNF-α related weak inducer of apoptosis, and TNFR1 signaling (Supplemental Fig. 3A). Functional network and canonical pathway analysis of 30 selected genes associated with TNF-α signaling and ubiquitination is shown in Fig. 3B and Supplemental Fig. 4.

FIGURE 2.

miR-17 regulates genes related to TNF-α signaling in RA SFs. (A) Heat map of the gene-expression data for selected genes related to TNF-α signaling and ubiquitination. Each row represents a single transcript, and each column represents a single sample. Red represents lower expression; green represents higher expression. (B) Fold change in selected genes from RNA-sequencing data (mean ± SD; n = 2). (C) Effect of miR-17 on the expression of selected genes (TRAF2, cIAP1, cIAP2, USP2, PSMD13, USP14, and RAD23A) was verified using quantitative real-time PCR. RA SFs (n = 4) were transfected with pre–miR-17 or NC–pre-miRNA for 48 h, and total RNA was prepared. GAPDH was used as a reference control. (D and E) Verification of RNA sequencing data for TRAF-2, cIAP1, cIAP2, USP2, USP14, PSMD13, and RAD23A (n = 4) by Western immunoblotting. RA SFs were transfected with pre–miR-17 or NC–pre-miRNA for 48 h, and cell lysates were prepared. Densitometry for Western blots for the indicated number of patients was performed, and values were normalized with β-actin. *p < 0.05, **p < 0.01.

FIGURE 2.

miR-17 regulates genes related to TNF-α signaling in RA SFs. (A) Heat map of the gene-expression data for selected genes related to TNF-α signaling and ubiquitination. Each row represents a single transcript, and each column represents a single sample. Red represents lower expression; green represents higher expression. (B) Fold change in selected genes from RNA-sequencing data (mean ± SD; n = 2). (C) Effect of miR-17 on the expression of selected genes (TRAF2, cIAP1, cIAP2, USP2, PSMD13, USP14, and RAD23A) was verified using quantitative real-time PCR. RA SFs (n = 4) were transfected with pre–miR-17 or NC–pre-miRNA for 48 h, and total RNA was prepared. GAPDH was used as a reference control. (D and E) Verification of RNA sequencing data for TRAF-2, cIAP1, cIAP2, USP2, USP14, PSMD13, and RAD23A (n = 4) by Western immunoblotting. RA SFs were transfected with pre–miR-17 or NC–pre-miRNA for 48 h, and cell lysates were prepared. Densitometry for Western blots for the indicated number of patients was performed, and values were normalized with β-actin. *p < 0.05, **p < 0.01.

Close modal
FIGURE 3.

miR-17 regulates ubiquitination and DUB processes to inhibit TNF-α signaling in RA SFs. (A and B) Exogenous delivery of miR-17 and miR-20a in RA SFs (n = 4) showing a significant increase in the expression of both miRNAs after 48 h of transfection. (C) Effect of miR-17 on TRAF2, cIAP1, cIAP2, USP2, USP-14, PSMD13, and RAD23A in TNF-α–stimulated RA SFs. RA SFs were transfected with pre–miR-17 for 48 h, followed by stimulation with TNF-α for 24 h. Cell lysates (n = 4) were prepared and assayed for their expression by Western immunoblotting. Densitometry for the Western blots for the indicated number of patients was performed, and values were normalized with β-actin. (D) RA SFs were transfected with pre–miR-17/pre–miR-20a/NC–pre-miRNA for 48 h, followed by TNF-α stimulation for 30 min. Total cell lysates (n = 3) were used to determine the effect on global expression of K63-linked (upper panel) and K48-linked (lower panel) polyubiquitination. (E and F) Densitometry analysis for K63 and K48 band intensity values normalized to β-actin. RA SFs (n = 3) were transfected with pre–miR-17/pre–miR-20a/NC–pre-miRNA for 48 h, followed by 30 min of TNF-α stimulation, immunoprecipitated for K63 polyubiquitin (G) or K48 polyubiquitin (H), and probed for the expression of TRAF2, cIAP1, and cIAP2. *p < 0.05, **p < 0.01.

FIGURE 3.

miR-17 regulates ubiquitination and DUB processes to inhibit TNF-α signaling in RA SFs. (A and B) Exogenous delivery of miR-17 and miR-20a in RA SFs (n = 4) showing a significant increase in the expression of both miRNAs after 48 h of transfection. (C) Effect of miR-17 on TRAF2, cIAP1, cIAP2, USP2, USP-14, PSMD13, and RAD23A in TNF-α–stimulated RA SFs. RA SFs were transfected with pre–miR-17 for 48 h, followed by stimulation with TNF-α for 24 h. Cell lysates (n = 4) were prepared and assayed for their expression by Western immunoblotting. Densitometry for the Western blots for the indicated number of patients was performed, and values were normalized with β-actin. (D) RA SFs were transfected with pre–miR-17/pre–miR-20a/NC–pre-miRNA for 48 h, followed by TNF-α stimulation for 30 min. Total cell lysates (n = 3) were used to determine the effect on global expression of K63-linked (upper panel) and K48-linked (lower panel) polyubiquitination. (E and F) Densitometry analysis for K63 and K48 band intensity values normalized to β-actin. RA SFs (n = 3) were transfected with pre–miR-17/pre–miR-20a/NC–pre-miRNA for 48 h, followed by 30 min of TNF-α stimulation, immunoprecipitated for K63 polyubiquitin (G) or K48 polyubiquitin (H), and probed for the expression of TRAF2, cIAP1, and cIAP2. *p < 0.05, **p < 0.01.

Close modal

RNA-sequencing results were confirmed by quantitative real-time PCR and Western blot analysis in RA SFs with miR-17 overexpression. Our results showed a significant decrease in the basal expression levels of TRAF2, cIAP1, cIAP2, USP2, and PSMD13 proteins and no change in the constitutive expression levels of RAD23A and USP14 (Fig. 2C–E).

To determine the impact of miR-17 on ubiquitination mechanisms essential for TNF-α signaling, RA SFs were transiently transfected with miR-17 mimic or an NC and stimulated with TNF-α for 24 h (Fig. 3). For comparison, we also included miR-20a, another miR-17 family member shown to have similar function (22, 32). Transfection with pre–miR-17 or miR-20a for 48 h resulted in a significant increase in the expression of both miRNAs compared with NC-transfected samples (Fig. 3A, 3B). Consistent with our sequencing results, Western blot analysis showed that TRAF2, cIAP1, cIAP2, PSMD13, and USP2 expression was significantly reduced in TNF-α–stimulated RA SFs transfected with pre–miR-17 compared with the NC (Fig. 3C). However, the expression of RAD23A and USP14 was not altered by miR-17 overexpression in TNF-α–stimulated RA SFs (Fig. 3C). We used bioinformatics algorithms (TargetScan 7.0, PicTar, and miRanda) to determine whether miR-17 regulates the expression of TRAF2, cIAP1, cIAP2, USP2, and PSMD13 via the sequence spanning the 3′UTR of their mRNA. No direct binding site of miR-17 in the 3′UTR of TRAF2, cIAP1, cIAP2, PSMD13, or USP2 mRNA was observed (data not shown), which indicates that miR-17 may impact cellular processes in RA SFs to regulate their gene expression.

To investigate the mechanism of TRAF2 degradation by miR-17 in response to TNF-α stimulation, we analyzed K63- and K48-linked ubiquitination patterns in miR-17–transfected RA SFs. Surprisingly, our results showed that miR-17 and miR-20a increased global K63 ubiquitination in TNF-α–stimulated RA SFs (Fig. 3D, 3E), which suggests that miR-17 and miR-20a may enhance total K63 ubiquitination processes to stabilize certain TNF-α signaling proteins. A modest decrease in global K48-linked ubiquitination was also observed in TNF-α–stimulated RA SFs; however, this was not statistically significant (Fig. 3F).

To further examine the effect of miR-17 on K63 and K48 ubiquitination specific to TRAF2, cIAP1, and cIAP2 proteins, TNF-α–stimulated RA SFs, with or without control or miR-17 overexpression, were immunoprecipitated with K63- and K48-linked ubiquitinated proteins from cell lysates and probed for TRAF2, cIAP1, and cIAP2 (Fig. 3G, 3H); however, we did not observe significant changes in K63-mediated ubiquitination of TRAF2, cIAP1, and cIAP2 (Fig. 3G). In contrast, Western blotting results with K48-linked immunoprecipitated proteins showed that miR-17 and miR-20a enhanced K48-mediated ubiquitination of TRAF2, cIAP1, and cIAP2 compared with the NC in TNF-α–stimulated RA SFs (Fig. 3H). These results suggest that miR-17 induces the K48-mediated ubiquitination of the key TNF-α signaling proteins TRAF2, cIAP1, and cIAP2 that may influence their stability and, thereby, inhibit downstream signaling events in RA SFs.

To further understand the impact of miR-17 on the early events in TNF-α signaling, we transfected RA SFs with NC–pre-miRNA, pre–miR-17, or pre–miR-20a, followed by TNF-α stimulation for 30 min. Cell lysates were immunoprecipitated with TRAF2 Ab and analyzed for its association with the signaling partners cIAP1, cIAP2, and RAD23A in RA SFs. We found that miR-17 preferentially inhibited the association of cIAP2 with TRAF2 in TNF-α–stimulated RA SFs but elicited no effect on RAD23A (Fig. 4A).

FIGURE 4.

miR-17 interferes with the association of TRAF2-cIAP2 in TNF-α–stimulated RA SFs. (A) RA SFs (n = 4) were transfected with pre–miR-17/pre–miR-20a/NC–pre-miRNA for 48 h, followed by TNF-α-stimulation for 30 min. Cell lysates were immunoprecipitated with TRAF2 or IgG and probed for changes in the expression of cIAP1, cIAP2, and RAD23A. Densitometric analysis of cIAP1 and cIAP2 band intensity. (B) RA SFs were transfected with miR-17 for 48 h, followed by stimulation with TNF-α for 30 min. Cell lysates (n = 4) were assayed for the expression of TRAF2, cIAP1, cIAP2, RAD23A, and β-actin by Western immunoblotting. (C) THP-1–differentiated macrophages (n = 3) were transfected with pre–miR-17/pre–miR-20a/NC–pre-miRNA for 48 h, followed by TNF-α stimulation for 30 min. Total cell lysates were immunoprecipitated with TRAF2 or IgG and probed for changes in the expression of cIAP1 and cIAP2. Densitometric analysis of cIAP1 and cIAP2 band intensity is shown. (D) THP-1–differentiated macrophages were transfected with miR-17 for 48 h, followed by stimulation with TNF-α for 30 min. Cell lysates (n = 3) were assayed for the expression of TRAF2, cIAP1, cIAP2, and β-actin by Western immunoblotting. *p < 0.05, **p < 0.01.

FIGURE 4.

miR-17 interferes with the association of TRAF2-cIAP2 in TNF-α–stimulated RA SFs. (A) RA SFs (n = 4) were transfected with pre–miR-17/pre–miR-20a/NC–pre-miRNA for 48 h, followed by TNF-α-stimulation for 30 min. Cell lysates were immunoprecipitated with TRAF2 or IgG and probed for changes in the expression of cIAP1, cIAP2, and RAD23A. Densitometric analysis of cIAP1 and cIAP2 band intensity. (B) RA SFs were transfected with miR-17 for 48 h, followed by stimulation with TNF-α for 30 min. Cell lysates (n = 4) were assayed for the expression of TRAF2, cIAP1, cIAP2, RAD23A, and β-actin by Western immunoblotting. (C) THP-1–differentiated macrophages (n = 3) were transfected with pre–miR-17/pre–miR-20a/NC–pre-miRNA for 48 h, followed by TNF-α stimulation for 30 min. Total cell lysates were immunoprecipitated with TRAF2 or IgG and probed for changes in the expression of cIAP1 and cIAP2. Densitometric analysis of cIAP1 and cIAP2 band intensity is shown. (D) THP-1–differentiated macrophages were transfected with miR-17 for 48 h, followed by stimulation with TNF-α for 30 min. Cell lysates (n = 3) were assayed for the expression of TRAF2, cIAP1, cIAP2, and β-actin by Western immunoblotting. *p < 0.05, **p < 0.01.

Close modal

To investigate whether this effect of miR-17 is specific to RA SFs, THP-1–differentiated macrophages were transfected with NC–pre-miRNA, pre–miR-17, or pre–miR-20a, followed by TNF-α stimulation for 30 min. Cell lysates were immunoprecipitated with TRAF2 and probed for the same panel of proteins shown in Fig. 4A and 4B. In agreement with the findings in RA SFs, we found a decrease in the association of TRAF2 with cIAP2 in macrophages (Fig. 4C). Similar to RA SFs, no early impact of miR-17 overexpression was observed on the expression of these proteins, as seen in the analyzed inputs (Fig. 4B, 4D). To study the impact of miR-17 on the ubiquitination pathways, a chemical inhibitor of ubiquitin-conjugating enzyme E1 (PYR41) with no effect on E2 or E3 ubiquitin ligases was used. RA SFs were pretreated with PYR41 for 2 h and then transfected with miR-17 (Supplemental Fig. 2D, 2E). Our results showed that inhibiting the ubiquitin E1 ligase had no effect on TNF-α–induced IL-6 and IL-8 production, whereas miR-17 decreased production in RA SFs (Supplemental Fig. 2D, 2E). These findings suggest that miR-17 might interfere with the UPS to impact the stability and efficiency of ubiquitin E3 ligase TRAF2 to associate with cIAP1/cIAP2 complex and participate in TNF-α signaling.

RA SFs were transfected with pre–miR-17 and then stimulated with TNF-α for 30 min or 24 h to assess the regulation of signaling proteins and soluble proteins, respectively. We observed that miR-17 overexpression led to a decrease in p-p38 and p-JNK expression in TNF-α–stimulated RA SFs (Fig. 5A, 5B). ASK1 (a serine-threonine kinase) regulates downstream p38 and JNK pathways and was shown to play a critical role in RA pathogenesis via TNF-α signaling (33, 34). Activation of ASK1 is tightly regulated by the phosphorylation of threonine residue (Thr838 and Thr845 of human and mouse ASK1, respectively) (33, 34). A higher expression of total ASK1 and ASK1 Thr838 was found in RA SFs compared with NL SFs (Fig. 6A–C). Interestingly, we found that miR-17 binds to the ASK1 3′UTR to regulate its expression in RA SFs, which may downregulate p-p38 and p-JNK (Fig. 6D–F). Interestingly, ASK1 knockdown also inhibited TNF-α–induced IL-6 and IL-8 production in RA SFs (Fig. 6G). miR-17 was shown to target the STAT3 3′UTR and regulate STAT3 expression, thereby leading to a loss of suppressive function in myeloid-derived suppressor cells (35, 36). Consistent with the previous findings, an overexpression of miR-17 reduced the nuclear translocation of p-STAT3 in TNF-α–stimulated RA SFs (Fig. 5A, 5B). Furthermore, we found that miR-17 also inhibited TNF-α–induced IκB-α phosphorylation compared with NC (Fig. 5A, 5B). In accordance with these findings, our results showed that miR-17 moderately reduced TNF-α–induced activation and nuclear translocation of transcription factors NF-κBp65 and p–c-Jun in RA SFs (Fig. 5A, 5B).

FIGURE 5.

Impact of miR-17 on MAPK/NF-κB/STAT3 pathways and downstream mediators in TNF-α–stimulated RA SFs. (A and B) Effect of miR-17 on p38, JNK, and IκB-α phosphorylation and nuclear translocation of NF-κBp65, p–c-Jun, and p-STAT3 in TNF-α–stimulated RA SFs. RA SFs (n = 3) were transfected with pre–miR-17 for 48 h and then stimulated or not with TNF-α for 30 min. Total cell lysates were probed for p38, JNK, and IκB-α, and nuclear fractions of NF-κBp65, p–c-Jun, and p-STAT3 were prepared and assayed using Western immunoblotting. (B) Densitometric analysis of Western blots (n = 3) is shown. Band intensities were normalized to their total respective forms, β-actin, or lamin A/C (for nuclear fraction). (C) Human cytokine Ab array (Ray Biotech) was used to measure the secretion of 80 cytokines in the pooled conditioned medium (n = 4) from miR-17 or NC-miRNA overexpressed and TNF-α (20 ng/ml) 24 h stimulated RA SFs. (D and E) RA SFs (n = 4) were transfected with pre–miR-17 for 48 h and then stimulated with TNF-α for 24 h to determine IL-6 and IL-8 production using quantitative ELISA. Data are mean ± SEM for the indicated number of SF donors. (F and G) RA SFs (n = 4) were transfected with anti–miR-17 for 48 h and stimulated with TNF-α for 24 h to determine IL-6 and IL-8 production using quantitative ELISA. Data are mean ± SEM for the indicated number of SF donors. (H and I) RA SFs were transfected with pre–miR-17 or NC for 48 h, followed by TNF-α stimulation for 24 h. Conditioned medium (n = 3) was concentrated using Amicon Ultra Centrifugal filters (Millipore) and analyzed for MMP-1 and MMP-13 protein expression using Western immunoblotting. Densitometric analysis of MMP-1 and MMP-13 (n = 3) is shown. *p < 0.05, **p < 0.01.

FIGURE 5.

Impact of miR-17 on MAPK/NF-κB/STAT3 pathways and downstream mediators in TNF-α–stimulated RA SFs. (A and B) Effect of miR-17 on p38, JNK, and IκB-α phosphorylation and nuclear translocation of NF-κBp65, p–c-Jun, and p-STAT3 in TNF-α–stimulated RA SFs. RA SFs (n = 3) were transfected with pre–miR-17 for 48 h and then stimulated or not with TNF-α for 30 min. Total cell lysates were probed for p38, JNK, and IκB-α, and nuclear fractions of NF-κBp65, p–c-Jun, and p-STAT3 were prepared and assayed using Western immunoblotting. (B) Densitometric analysis of Western blots (n = 3) is shown. Band intensities were normalized to their total respective forms, β-actin, or lamin A/C (for nuclear fraction). (C) Human cytokine Ab array (Ray Biotech) was used to measure the secretion of 80 cytokines in the pooled conditioned medium (n = 4) from miR-17 or NC-miRNA overexpressed and TNF-α (20 ng/ml) 24 h stimulated RA SFs. (D and E) RA SFs (n = 4) were transfected with pre–miR-17 for 48 h and then stimulated with TNF-α for 24 h to determine IL-6 and IL-8 production using quantitative ELISA. Data are mean ± SEM for the indicated number of SF donors. (F and G) RA SFs (n = 4) were transfected with anti–miR-17 for 48 h and stimulated with TNF-α for 24 h to determine IL-6 and IL-8 production using quantitative ELISA. Data are mean ± SEM for the indicated number of SF donors. (H and I) RA SFs were transfected with pre–miR-17 or NC for 48 h, followed by TNF-α stimulation for 24 h. Conditioned medium (n = 3) was concentrated using Amicon Ultra Centrifugal filters (Millipore) and analyzed for MMP-1 and MMP-13 protein expression using Western immunoblotting. Densitometric analysis of MMP-1 and MMP-13 (n = 3) is shown. *p < 0.05, **p < 0.01.

Close modal
FIGURE 6.

miR-17 may directly target the ASK1 3′UTR to regulate its expression in RA SFs. (AC) Expression of total ASK1 and p-ASK Thr838 protein was evaluated in NL, OA, and RA SFs by Western immunoblotting. (D) Bioinformatics analysis of the predicted binding sites for miR-17 in the ASK1 mRNA 3′UTR and their cross-species conservation. (E) Inhibitory effect of miR-17 on ASK1 3′UTR luciferase reporter activity in RA SFs. ASK1 reporter vectors were transfected in RA SFs (n = 3) with pre–miR-17 or NC–pre-miRNA. The experiment was performed in triplicate. (F) The effect of pre–miR-17 transfection on ASK1 mRNA and protein expression at the basal level and in TNF-α–stimulated RA SFs (n = 3) was assessed using quantitative real-time PCR and Western immunoblotting, respectively. GAPDH or β-actin was used as a loading control. (G) The effect of ASK-1 siRNA on TNF-α–induced production of IL-6 and IL-8 in RA SFs was analyzed using ELISA. RA SFs (n = 4) were transfected with ASK1 siRNA for 48 h and stimulated with TNF-α for 24 h. Conditioned medium was analyzed for IL-6 and IL-8 production. Data are mean ± SEM for the indicated number of patients using different donors. (H) Schematic representation of the effect of miR-17 on TNF-α signaling proteins in RA SFs. Our data suggest that miR-17 modulation of the UPS is an important mechanism that inhibits SF–mediated inflammation and tissue destruction in RA. *p < 0.05, **p < 0.01.

FIGURE 6.

miR-17 may directly target the ASK1 3′UTR to regulate its expression in RA SFs. (AC) Expression of total ASK1 and p-ASK Thr838 protein was evaluated in NL, OA, and RA SFs by Western immunoblotting. (D) Bioinformatics analysis of the predicted binding sites for miR-17 in the ASK1 mRNA 3′UTR and their cross-species conservation. (E) Inhibitory effect of miR-17 on ASK1 3′UTR luciferase reporter activity in RA SFs. ASK1 reporter vectors were transfected in RA SFs (n = 3) with pre–miR-17 or NC–pre-miRNA. The experiment was performed in triplicate. (F) The effect of pre–miR-17 transfection on ASK1 mRNA and protein expression at the basal level and in TNF-α–stimulated RA SFs (n = 3) was assessed using quantitative real-time PCR and Western immunoblotting, respectively. GAPDH or β-actin was used as a loading control. (G) The effect of ASK-1 siRNA on TNF-α–induced production of IL-6 and IL-8 in RA SFs was analyzed using ELISA. RA SFs (n = 4) were transfected with ASK1 siRNA for 48 h and stimulated with TNF-α for 24 h. Conditioned medium was analyzed for IL-6 and IL-8 production. Data are mean ± SEM for the indicated number of patients using different donors. (H) Schematic representation of the effect of miR-17 on TNF-α signaling proteins in RA SFs. Our data suggest that miR-17 modulation of the UPS is an important mechanism that inhibits SF–mediated inflammation and tissue destruction in RA. *p < 0.05, **p < 0.01.

Close modal

To understand the impact of TNF-α signaling inhibition by miR-17, we determined the effect of miR-17 overexpression on the production of 80 cytokines, chemokines, and growth factors using cytokine array. Densitometry analysis of the array membrane results obtained from the conditioned media of RA SFs treated with NC–pre-miRNA or miR-17 in the presence of TNF-α showed a marked inhibition of several proinflammatory cytokines and chemokines known to contribute to RA pathogenesis, including G-CSF (∼24%), GM-CSF (∼19%), GRO (∼17%), GRO-α (∼31%), CCL23 (∼21%), IL-3 (∼16%), IL-6 (∼15%), IL-7 (∼26%), IL-8 (∼19%), MCP-1 (∼15%), M-CSF (∼25%), MDC (∼24%), MIG (∼24%), MIP-1β (∼25%), and TNF-β (∼22%) (Fig. 5C). To confirm this array result, we used the same conditioned media to quantitate IL-6 and IL-8 production by ELISA. Consistent with the array results, miR-17 overexpression inhibited TNF-α–induced IL-6 and IL-8 production in RA SFs (Fig. 5D, 5E). In contrast, the inhibition of endogenous miR-17 expression showed a significant increase in IL-6 and IL-8 production in TNF-α–stimulated RA SFs (Fig. 5F, 5G). Extending these findings, our results also showed that overexpression of miR-17 inhibited TNF-α–induced MMP-1 and MMP-13 production in RA SFs (Fig. 5H, 5I), suggesting a protective role for miR-17 in cartilage degradation.

In the current study, we identified the role of miR-17 in modulating posttranslational protein ubiquitination pathways to downregulate TNF-α signaling events in RA (Fig. 6H). We found that miR-17 expression is very low in RA, and its restoration in vitro may induce K48-linked ubiquitination of TRAF2 and cIAP2 and may reduce their association, which results in a reduction in TNF-α–induced inflammatory mediators in RA SFs. Importantly, this study validated the IPA data that suggest miR-17 may influence posttranslational modifications independent of its 3′UTR binding. These findings provide an opportunity for further understanding of the impact of miRNA on cellular and posttranslational mechanisms, such as the protein ubiquitination pathway, that are important in the cytokine signaling networks for RA pathogenesis.

TRAF2, an E3 ubiquitin ligase, is a critical upstream component in the TNF-α signaling that activates MAPK and NF-κB pathways through the recruitment of cIAP1 and cIAP2 to the TNFR signaling complex (37, 38). TRAF2 is tightly regulated by posttranslational modifications, such as autophosphorylation, ubiquitination, or DUB (39). Among these, ubiquitination plays an important role in diverse cellular events and signaling; however, this mechanism in proinflammatory cytokine signaling networks has not been well studied in RA SFs (40, 41). Our results showing that miR-17 reduced K48-linked ubiquitination in RA SFs further confirms the IPA findings of the influence on protein ubiquitin pathways. The UPS is indispensable for TNF-α signal transduction and balances the cellular expression and activity of proteins by ubiquitination and degradation (42). In contrast, DUB removes ubiquitin conjugates by activating ATP hydrolysis, thereby rescuing proteins from degradation (43). Among the different DUB enzymes (USP14, USP2, and PSMD13) modulated by miR-17 in RNA-sequencing data, our results in RA SF lysates confirmed that miR-17 suppressed the expression of PSMD13 and USP2 in TNF-α–stimulated RA SFs. These results suggest that miR-17 affects the stability of TRAF2 and/or cIAP2 proteins, in part, by regulating DUB activity of PSMD13, an important proteome component, and USP2, a ubiquitin-specific protease important in TNF-α–induced RA SFs. USP2 was shown to target multiple substrates for protein stability (e.g., p53, cyclin D1) (44), and its mRNA may be modulated by TNF-α in a cell-specific manner (45). Further investigation is needed to understand the mechanism involved in downregulation of USP2 by miR-17. In agreement with our results, the miR-17–92 cluster was shown to target E3 ubiquitin ligases, thereby affecting PTEN subcellular localization through monoubiquitination of limb-innervating lateral motor neurons (46).

Our findings from IP assays showed that miR-17 and miR-20a enhanced K48-mediated ubiquitination of TRAF2, cIAP1, and cIAP2, suggesting that these miRNAs induce K48-mediated proteasomal degradation of key TNF-α signaling proteins TRAF2, cIAP1, and cIAP2, inhibiting their expression and downstream signaling. Further validation studies in THP-1–activated macrophages also confirmed that the proteasomal degradation of TRAF2 and cIAP2 by miR-17 further reduced their association and, possibly, downstream signaling events.

In classical TNF-α signaling, TNFR2 forms a complex with TRAF2 and cIAP1/2 to further recruit and activate RIPK1, which further activates TAK1 kinase activity, leading to activation of the NF-κB and MAPK pathways (15). As a result of the lack of TRAF2/cIAP2 association and efficient signaling activation by miR-17, we observed inhibition of the activation of TNF-α–induced p-p38, p-JNK, and p–IκB-α and a consequent reduction in the nuclear translocation of p–c-Jun and NF-κBp65. The inhibition of p–c-Jun and NF-κB transcription factors also caused a significant reduction in TNF-α–induced MMP-1 and MMP-13 production in RA SFs. We showed recently that a similar reduction in TRAF6 and TAK1 association by epigallocatechin-3-gallate inhibited IL-1β signaling and downstream mediators of inflammation in RA SFs (47). It is worth noting that MMP-1 and MMP-13 have no 3′UTR binding regions for miR-17, which further suggests that this might be an indirect effect of miR-17 on tissue remodeling mediated in RA. Our results showed that miR-17 targets the ASK1 3′UTR to regulate ASK1 expression, which may result in a reduction in p38 and JNK phosphorylation in RA SFs. A decrease in TNF-α–induced IL-6 and IL-8 production was observed in response to miR-17 overexpression, as well as after ASK1 knockdown in RA SFs. Importantly, in addition to directly targeting ASK1, miR-17 may influence the association of TRAF2 and cIAP2 in TNF-α–stimulated RA SFs. We hypothesize that miR-17 may regulate TNF-α signaling at multiple steps in RA, and its influence on TRAF2 and cIAP2 may be the earliest and most prominent in the TNF-α signaling pathway. In addition, recent studies showed that miR-17 targets the STAT3 3′UTR to regulate its expression in myeloid-derived suppressor cells and in other cell types (35, 36), suggesting that STAT3 regulation may be a direct downstream effect of miR-17 through 3′UTR binding.

The miR-17–92 cluster was shown to influence acquired and innate immune responses (48); however, individual members of this cluster showed different biological activities. This is evident from our results whereby miR-17 mediated the suppression of TRAF2/cIAP2 association. Furthermore, miR-17 was shown to promote osteogenic differentiation in the inflammatory microenvironment, and miR-20a promoted osteogenesis of human myeloid-derived suppressor cells via BMP signaling (49, 50). In contrast, miR-18a enhanced MMP-1, IL-6, and IL-8 expression in TNF-α–stimulated RA SFs (26). Another study showed the suppression of miR-19a and miR-19b in RA SFs by TLR ligands and identified TLR2 as a direct target of the miR-19 family (25). These findings suggest that individual miRNAs in a cluster might have completely different influences, depending on the level of expression and the cytokine signaling pathways involved that contribute to RA pathogenesis.

A key determinant for miRNA target recognition is the seed match site in the target mRNA, which is located primarily in the 3′UTR (6). In the current study, in silico prediction analysis showed no direct binding site for miR-17 in the 3′UTRs of TRAF2, cIAP1, cIAP2, PSMD13, and USP2 mRNAs, indicating that miR-17 may not target the respective mRNA 3′UTRs of these genes to regulate their expression in RA SFs. A recently described model to examine indirect effect of miRNA transfection postulates that the transfected miRNAs may compensate endogenous miRNAs for the available RNA-induced silencing complex and, consequently, alter the regulation of their downstream targets (51). Recent studies showed that, among all of the genes that were differentially expressed with the miRNA modulation, <20% were predicted miRNA targets (12), suggesting that the underlying coordinated changes in the overall patterns of gene expression involve the modulation of centralized hub genes (52). These regulatory genes (i.e., hub genes) have the potential to control a group of downstream genes to force differential gene-expression outcomes (53). Our finding is an important step toward understanding the impact of miRNAs on the essential cellular processes and TNF-α signaling network in RA SFs, a cell type that has not been targeted for therapeutic approaches. Further studies are required to extend these in vitro findings in animal models of human RA to understand the synovial and systemic benefit of therapeutically restoring miR-17 expression levels in RA.

We thank Karen Porter and Sadiq Umar for help with the AIA animal study.

This work was supported by National Institutes of Health Grant AR063104 (to S.A.), an Arthritis Foundation Innovative Research Grant (to S.A.), and start-up funds from Washington State University.

The sequences presented in this article have been submitted to the National Center for Biotechnology Information Gene Expression Omnibus under accession number GSE83930.

The online version of this article contains supplemental material.

Abbreviations used in this article:

AIA

adjuvant-induced arthritis

ASK

apoptosis signaling kinase

DUB

deubiquitination

IP

immunoprecipitation

IPA

Ingenuity Pathway Analysis

miRNA

microRNA

MMP

matrix metalloproteinase

NC

negative control

NL

healthy/nondiseased

OA

osteoarthritis

RA

rheumatoid arthritis

SF

synovial fibroblast

siRNA

small interfering RNA

ST

synovial tissue

UPS

ubiquitin proteasome system

UTR

untranslated region.

1
Bartel
D. P.
2004
.
MicroRNAs: genomics, biogenesis, mechanism, and function.
Cell
116
:
281
297
.
2
Akhtar
N.
,
Haqqi
T. M.
.
2013
.
MicroRNAs: tiny regulators with the great potential for gene regulation.
In
microRNAs in Toxicology and Medicine.
Sahu
S. C.
, ed.
Wiley-Blackwell
,
New York
, p.
287
300
.
3
Murata
K.
,
Yoshitomi
H.
,
Tanida
S.
,
Ishikawa
M.
,
Nishitani
K.
,
Ito
H.
,
Nakamura
T.
.
2010
.
Plasma and synovial fluid microRNAs as potential biomarkers of rheumatoid arthritis and osteoarthritis.
Arthritis Res. Ther.
12
:
R86
.
4
Salehi
E.
,
Eftekhari
R.
,
Oraei
M.
,
Gharib
A.
,
Bidad
K.
.
2015
.
MicroRNAs in rheumatoid arthritis.
Clin. Rheumatol.
34
:
615
628
.
5
Mogilyansky
E.
,
Rigoutsos
I.
.
2013
.
The miR-17/92 cluster: a comprehensive update on its genomics, genetics, functions and increasingly important and numerous roles in health and disease.
Cell Death Differ.
20
:
1603
1614
.
6
Lewis
B. P.
,
Burge
C. B.
,
Bartel
D. P.
.
2005
.
Conserved seed pairing, often flanked by adenosines, indicates that thousands of human genes are microRNA targets.
Cell
120
:
15
20
.
7
Akhtar
N.
,
Makki
M. S.
,
Haqqi
T. M.
.
2015
.
MicroRNA-602 and microRNA-608 regulate sonic hedgehog expression via target sites in the coding region in human chondrocytes.
Arthritis Rheumatol.
67
:
423
434
.
8
Lee
I.
,
Ajay
S. S.
,
Yook
J. I.
,
Kim
H. S.
,
Hong
S. H.
,
Kim
N. H.
,
Dhanasekaran
S. M.
,
Chinnaiyan
A. M.
,
Athey
B. D.
.
2009
.
New class of microRNA targets containing simultaneous 5′-UTR and 3′-UTR interaction sites.
Genome Res.
19
:
1175
1183
.
9
Brümmer
A.
,
Hausser
J.
.
2014
.
MicroRNA binding sites in the coding region of mRNAs: extending the repertoire of post-transcriptional gene regulation.
BioEssays
36
:
617
626
.
10
Baek
D.
,
Villén
J.
,
Shin
C.
,
Camargo
F. D.
,
Gygi
S. P.
,
Bartel
D. P.
.
2008
.
The impact of microRNAs on protein output.
Nature
455
:
64
71
.
11
Selbach
M.
,
Schwanhäusser
B.
,
Thierfelder
N.
,
Fang
Z.
,
Khanin
R.
,
Rajewsky
N.
.
2008
.
Widespread changes in protein synthesis induced by microRNAs.
Nature
455
:
58
63
.
12
Shahab
S. W.
,
Matyunina
L. V.
,
Hill
C. G.
,
Wang
L.
,
Mezencev
R.
,
Walker
L. D.
,
McDonald
J. F.
.
2012
.
The effects of microRNA transfections on global patterns of gene expression in ovarian cancer cells are functionally coordinated.
BMC Med. Genomics
5
:
33
.
13
Lim
L. P.
,
Lau
N. C.
,
Garrett-Engele
P.
,
Grimson
A.
,
Schelter
J. M.
,
Castle
J.
,
Bartel
D. P.
,
Linsley
P. S.
,
Johnson
J. M.
.
2005
.
Microarray analysis shows that some microRNAs downregulate large numbers of target mRNAs.
Nature
433
:
769
773
.
14
Tseng
P. H.
,
Matsuzawa
A.
,
Zhang
W.
,
Mino
T.
,
Vignali
D. A.
,
Karin
M.
.
2010
.
Different modes of ubiquitination of the adaptor TRAF3 selectively activate the expression of type I interferons and proinflammatory cytokines.
Nat. Immunol.
11
:
70
75
.
15
Gyrd-Hansen
M.
,
Meier
P.
.
2010
.
IAPs: from caspase inhibitors to modulators of NF-kappaB, inflammation and cancer.
Nat. Rev. Cancer
10
:
561
574
.
16
Mendell
J. T.
2008
.
miRiad roles for the miR-17-92 cluster in development and disease.
Cell
133
:
217
222
.
17
Olive
V.
,
Jiang
I.
,
He
L.
.
2010
.
mir-17-92, a cluster of miRNAs in the midst of the cancer network.
Int. J. Biochem. Cell Biol.
42
:
1348
1354
.
18
Grillari
J.
,
Hackl
M.
,
Grillari-Voglauer
R.
.
2010
.
miR-17-92 cluster: ups and downs in cancer and aging.
Biogerontology
11
:
501
506
.
19
Poliseno
L.
,
Pitto
L.
,
Simili
M.
,
Mariani
L.
,
Riccardi
L.
,
Ciucci
A.
,
Rizzo
M.
,
Evangelista
M.
,
Mercatanti
A.
,
Pandolfi
P. P.
,
Rainaldi
G.
.
2008
.
The proto-oncogene LRF is under post-transcriptional control of MiR-20a: implications for senescence.
PLoS One
3
:
e2542
.
20
Yu
Z.
,
Wang
C.
,
Wang
M.
,
Li
Z.
,
Casimiro
M. C.
,
Liu
M.
,
Wu
K.
,
Whittle
J.
,
Ju
X.
,
Hyslop
T.
, et al
.
2008
.
A cyclin D1/microRNA 17/20 regulatory feedback loop in control of breast cancer cell proliferation.
J. Cell Biol.
182
:
509
517
.
21
Hossain
A.
,
Kuo
M. T.
,
Saunders
G. F.
.
2006
.
Mir-17-5p regulates breast cancer cell proliferation by inhibiting translation of AIB1 mRNA.
Mol. Cell. Biol.
26
:
8191
8201
.
22
Chang
C. C.
,
Yang
Y. J.
,
Li
Y. J.
,
Chen
S. T.
,
Lin
B. R.
,
Wu
T. S.
,
Lin
S. K.
,
Kuo
M. Y.
,
Tan
C. T.
.
2013
.
MicroRNA-17/20a functions to inhibit cell migration and can be used a prognostic marker in oral squamous cell carcinoma.
Oral Oncol.
49
:
923
931
.
23
Lin
Y. H.
,
Liao
C. J.
,
Huang
Y. H.
,
Wu
M. H.
,
Chi
H. C.
,
Wu
S. M.
,
Chen
C. Y.
,
Tseng
Y. H.
,
Tsai
C. Y.
,
Chung
I. H.
, et al
.
2013
.
Thyroid hormone receptor represses miR-17 expression to enhance tumor metastasis in human hepatoma cells.
Oncogene
32
:
4509
4518
.
24
Philippe
L.
,
Alsaleh
G.
,
Pichot
A.
,
Ostermann
E.
,
Zuber
G.
,
Frisch
B.
,
Sibilia
J.
,
Pfeffer
S.
,
Bahram
S.
,
Wachsmann
D.
,
Georgel
P.
.
2013
.
MiR-20a regulates ASK1 expression and TLR4-dependent cytokine release in rheumatoid fibroblast-like synoviocytes.
Ann. Rheum. Dis.
72
:
1071
1079
.
25
Philippe
L.
,
Alsaleh
G.
,
Suffert
G.
,
Meyer
A.
,
Georgel
P.
,
Sibilia
J.
,
Wachsmann
D.
,
Pfeffer
S.
.
2012
.
TLR2 expression is regulated by microRNA miR-19 in rheumatoid fibroblast-like synoviocytes.
J. Immunol.
188
:
454
461
.
26
Trenkmann
M.
,
Brock
M.
,
Gay
R. E.
,
Michel
B. A.
,
Gay
S.
,
Huber
L. C.
.
2013
.
Tumor necrosis factor α-induced microRNA-18a activates rheumatoid arthritis synovial fibroblasts through a feedback loop in NF-κB signaling.
Arthritis Rheum.
65
:
916
927
.
27
Ahmed
S.
,
Pakozdi
A.
,
Koch
A. E.
.
2006
.
Regulation of interleukin-1beta-induced chemokine production and matrix metalloproteinase 2 activation by epigallocatechin-3-gallate in rheumatoid arthritis synovial fibroblasts.
Arthritis Rheum.
54
:
2393
2401
.
28
Mestdagh
P.
,
Van Vlierberghe
P.
,
De Weer
A.
,
Muth
D.
,
Westermann
F.
,
Speleman
F.
,
Vandesompele
J.
.
2009
.
A novel and universal method for microRNA RT-qPCR data normalization.
Genome Biol.
10
:
R64
.
29
Khomtchouk
B. B.
,
Van Booven
D. J.
,
Wahlestedt
C.
.
2014
.
HeatmapGenerator: high performance RNAseq and microarray visualization software suite to examine differential gene expression levels using an R and C++ hybrid computational pipeline.
Source Code Biol. Med.
9
:
30
.
30
Ahmed
S.
,
Marotte
H.
,
Kwan
K.
,
Ruth
J. H.
,
Campbell
P. L.
,
Rabquer
B. J.
,
Pakozdi
A.
,
Koch
A. E.
.
2008
.
Epigallocatechin-3-gallate inhibits IL-6 synthesis and suppresses transsignaling by enhancing soluble gp130 production.
Proc. Natl. Acad. Sci. USA
105
:
14692
14697
.
31
Joe
B.
,
Wilder
R. L.
.
1999
.
Animal models of rheumatoid arthritis.
Mol. Med. Today
5
:
367
369
.
32
Yu
Z.
,
Willmarth
N. E.
,
Zhou
J.
,
Katiyar
S.
,
Wang
M.
,
Liu
Y.
,
McCue
P. A.
,
Quong
A. A.
,
Lisanti
M. P.
,
Pestell
R. G.
.
2010
.
microRNA 17/20 inhibits cellular invasion and tumor metastasis in breast cancer by heterotypic signaling.
Proc. Natl. Acad. Sci. USA
107
:
8231
8236
.
33
Hayakawa
T.
,
Matsuzawa
A.
,
Noguchi
T.
,
Takeda
K.
,
Ichijo
H.
.
2006
.
The ASK1-MAP kinase pathways in immune and stress responses.
Microbes Infect.
8
:
1098
1107
.
34
Matsuzawa
A.
,
Saegusa
K.
,
Noguchi
T.
,
Sadamitsu
C.
,
Nishitoh
H.
,
Nagai
S.
,
Koyasu
S.
,
Matsumoto
K.
,
Takeda
K.
,
Ichijo
H.
.
2005
.
ROS-dependent activation of the TRAF6-ASK1-p38 pathway is selectively required for TLR4-mediated innate immunity.
Nat. Immunol.
6
:
587
592
.
35
Zhang
M.
,
Liu
Q.
,
Mi
S.
,
Liang
X.
,
Zhang
Z.
,
Su
X.
,
Liu
J.
,
Chen
Y.
,
Wang
M.
,
Zhang
Y.
, et al
.
2011
.
Both miR-17-5p and miR-20a alleviate suppressive potential of myeloid-derived suppressor cells by modulating STAT3 expression.
J. Immunol.
186
:
4716
4724
.
36
Kumar
R.
,
Sahu
S. K.
,
Kumar
M.
,
Jana
K.
,
Gupta
P.
,
Gupta
U. D.
,
Kundu
M.
,
Basu
J.
.
2015
.
MicroRNA 17-5p regulates autophagy in Mycobacterium tuberculosis-infected macrophages by targeting Mcl-1 and STAT3.
Cell Microbiol.
18
:
679
691
.
37
Rothe
M.
,
Pan
M. G.
,
Henzel
W. J.
,
Ayres
T. M.
,
Goeddel
D. V.
.
1995
.
The TNFR2-TRAF signaling complex contains two novel proteins related to baculoviral inhibitor of apoptosis proteins.
Cell
83
:
1243
1252
.
38
Liu
Z. G.
,
Hsu
H.
,
Goeddel
D. V.
,
Karin
M.
.
1996
.
Dissection of TNF receptor 1 effector functions: JNK activation is not linked to apoptosis while NF-kappaB activation prevents cell death.
Cell
87
:
565
576
.
39
Brown
K. D.
,
Hostager
B. S.
,
Bishop
G. A.
.
2002
.
Regulation of TRAF2 signaling by self-induced degradation.
J. Biol. Chem.
277
:
19433
19438
.
40
Chen
Z. J.
2012
.
Ubiquitination in signaling to and activation of IKK.
Immunol. Rev.
246
:
95
106
.
41
van Wijk
S. J.
,
Timmers
H. T.
.
2010
.
The family of ubiquitin-conjugating enzymes (E2s): deciding between life and death of proteins.
FASEB J.
24
:
981
993
.
42
Wullaert
A.
,
Heyninck
K.
,
Janssens
S.
,
Beyaert
R.
.
2006
.
Ubiquitin: tool and target for intracellular NF-kappaB inhibitors.
Trends Immunol.
27
:
533
540
.
43
Phillips
A. H.
,
Zhang
Y.
,
Cunningham
C. N.
,
Zhou
L.
,
Forrest
W. F.
,
Liu
P. S.
,
Steffek
M.
,
Lee
J.
,
Tam
C.
,
Helgason
E.
, et al
.
2013
.
Conformational dynamics control ubiquitin-deubiquitinase interactions and influence in vivo signaling.
Proc. Natl. Acad. Sci. USA
110
:
11379
11384
.
44
Davis
M. I.
,
Simeonov
A.
.
2015
.
Ubiquitin-specific proteases as druggable targets.
Drug Target Rev.
2
:
60
64
.
45
Haimerl
F.
,
Erhardt
A.
,
Sass
G.
,
Tiegs
G.
.
2009
.
Down-regulation of the de-ubiquitinating enzyme ubiquitin-specific protease 2 contributes to tumor necrosis factor-alpha-induced hepatocyte survival.
J. Biol. Chem.
284
:
495
504
.
46
Tung
Y. T.
,
Lu
Y. L.
,
Peng
K. C.
,
Yen
Y. P.
,
Chang
M.
,
Li
J.
,
Jung
H.
,
Thams
S.
,
Huang
Y. P.
,
Hung
J. H.
,
Chen
J. A.
.
2015
.
Mir-17∼92 governs motor neuron subtype survival by mediating nuclear PTEN.
Cell Rep.
11
:
1305
1318
.
47
Singh
A. K.
,
Umar
S.
,
Riegsecker
S.
,
Chourasia
M.
,
Ahmed
S.
.
2016
.
Regulation of transforming growth factor β-activated kinase activation by epigallocatechin-3-gallate in rheumatoid arthritis synovial fibroblasts: suppression of K(63)-linked autoubiquitination of tumor necrosis factor receptor-associated factor 6.
Arthritis Rheumatol.
68
:
347
358
.
48
Tsitsiou
E.
,
Lindsay
M. A.
.
2009
.
microRNAs and the immune response.
Curr. Opin. Pharmacol.
9
:
514
520
.
49
Liu
Y.
,
Liu
W.
,
Hu
C.
,
Xue
Z.
,
Wang
G.
,
Ding
B.
,
Luo
H.
,
Tang
L.
,
Kong
X.
,
Chen
X.
, et al
.
2011
.
MiR-17 modulates osteogenic differentiation through a coherent feed-forward loop in mesenchymal stem cells isolated from periodontal ligaments of patients with periodontitis.
Stem Cells
29
:
1804
1816
.
50
Zhang
J. F.
,
Fu
W. M.
,
He
M. L.
,
Xie
W. D.
,
Lv
Q.
,
Wan
G.
,
Li
G.
,
Wang
H.
,
Lu
G.
,
Hu
X.
, et al
.
2011
.
MiRNA-20a promotes osteogenic differentiation of human mesenchymal stem cells by co-regulating BMP signaling.
RNA Biol.
8
:
829
838
.
51
Khan
A. A.
,
Betel
D.
,
Miller
M. L.
,
Sander
C.
,
Leslie
C. S.
,
Marks
D. S.
.
2009
.
Transfection of small RNAs globally perturbs gene regulation by endogenous microRNAs.
Nat. Biotechnol.
27
:
549
555
.
52
Shalgi
R.
,
Lieber
D.
,
Oren
M.
,
Pilpel
Y.
.
2007
.
Global and local architecture of the mammalian microRNA-transcription factor regulatory network.
PLOS Comput. Biol.
3
:
e131
.
53
Seo
C. H.
,
Kim
J. R.
,
Kim
M. S.
,
Cho
K. H.
.
2009
.
Hub genes with positive feedbacks function as master switches in developmental gene regulatory networks.
Bioinformatics
25
:
1898
1904
.

The authors have no financial conflicts of interest.

Supplementary data