MiR-199a-3p/5p participated in TGF-β and EGF induced EMT by targeting DUSP5/MAP3K11 in pterygium

Background Recently, it has been reported that miRNA is involved in pterygium, however the exact underlying mechanism in pterygium is unrevealed and require further investigation. Methods The differential expression of miRNA in pterygium was profiled using microarray and validated with quantitative real-time polymerase chain reaction (qRT-PCR). Human conjunctival epithelial cells (HCEs) were cultured and treated with transforming growth factor β (TGF-β) and epidermal growth factor (EGF) and transfected with miR-199a-3p/5p mimic and inhibitor. Markers of epithelial-mesenchymal transition (EMT) in HCEs were detected using western blot and immunohistochemistry. Cell migration ability was determined using wound healing and transwell assay, while apoptosis was determined by flow cytometry. The target genes of miR-199a were confirmed by the dual-luciferase reporter assay. Results TGF-β and EGF could induced EMT in HCEs and increase miR-199a-3p/5p but suppress target genes, DUSP5 and MAP3K11. With the occurrence of EMT, cell migration ability was enhanced, and apoptosis was impeded. Promoting miR-199a-3p/5p expression could induce EMT in HCEs without TGF-β and EGF, while suppressing miR-199a-3p/5p could inhibit EMT in TGF-β and EGF induced HCEs. In a word, TGF-β and EGF induced EMT could be regulated with miR-199a-3p/5p-DUSP5/MAP3K11 axes. The validated results in tissues showed that, compared with control conjunctival tissues, miR-199a-3p/5p were more overexpressed in pterygium, while DUSP5/MAP3K11 were lower expressed. In addition, bioinformatics analysis indicated the miR-199a-3p/5p-DUSP5/MAP3K11 was belong to MAPK signalling pathway. Conclusions TGF-β and EGF induce EMT of HCEs through miR-199a-3p/5p-DUSP5/MAP3K11 axes, which explains the pathogenesis of EMT in pterygium and may provide new targets for pterygium prevention and therapy.


Background
Pterygium is a common ocular surface disease with a triangular-shaped lesion growing in the limbal conjunctiva and progressively invading toward cornea, affecting nearly 200 million people globally and the prevalence can even be as high as 22% in some countries [1][2][3][4]. Ultraviolet light (UVB) exposure induced chronic irritation of eyes is widely considered as the dominant risk factor of pterygium. Excessive UVB can lead to the inactivation of p53, and inactivated p53 promoted the process of epithelial-mesenchymal transition (EMT), contributing to the pathogenesis of pterygium [5][6][7]. Nowadays, conjunctival auto-transplantation is the most common treatment for pterygium, but the recurrence rates range from 2 to 39% [8][9][10].
The process of EMT, manifested in epithelial cells losing the polarity and cell-cell adhesion, and obtaining the migratory ability to become mesenchymal cells, has been verified as one of the most significant characteristic of pterygium [11][12][13][14]. Transforming growth factor β (TGF-β and epidermal growth factor (EGF) were found to be two fatal inducement in the process of EMT in various diseases [14][15][16][17]. TGF-β and EGF are overexpressed in pterygium, and it has been reported that both of them can be activated by UVB [18,19]. TGF-β signalling stimulates fibroblasts migration, proliferation and myofibroblasts differentiation, which play a significant role as profibrotic agents in pterygium [20,21]. While EGF promotes excessive keratinization and the overexpressed receptor of EGF (EGFR) leaded to increased proliferation in pterygium epithelial cells [22,23]. So, we planned to induce EMT in human conjunctival epithelial cells (HCEs) with TGF-β and EGF, which imitates the EMT initiation in pterygium as a cell model.
On another hand, current studies indicated miRNAs were involved in pterygium [2,6,[24][25][26][27], but based on a are a relatively small number of samples [2,[26][27][28]. MiRNAs are a large class of small non-coding RNAs, which can complementarily or partially complementarily bind to the 3′-untranslated regions (3′-UTR) of target mRNAs, leading to degradation or translation repression [29,30]. Based on results of miRNA microarray in the present study, miR-199a-3p/5p showed a significantly higher expression level in pterygium that in conjunctiva. MiR-199a-3p/5p was reported to suppress EMT in various cancer, such as testicular germ cell tumor, non-small cell lung cancer, head and neck cancer, and so on [31][32][33]. But they were also reported to promote the EMT process in hepatic fibrosis and idiopathic pulmonary fibrosis (IPF) [15,34,35]. These contradictory results suggest that miR-199a-3p/5p might have different functions in different types of EMT in various diseases. Pterygium is essentially characterized as a kind of fibrosis disease, since after EMT, the conjunctival epithelial cells obtain motility and grow toward the cornea, rather than inter-tissue invasion like tumor.
So, we hypothesized that miR-199a-3p/5p might promote EMT in fibrotic pterygium. Our objective was to explore the role of miR-199a-3p/5p in EMT of pterygium based on the established pterygium EMT cell model using HCEs with TGF-β and EGF, and eventually to find new perspectives on treatments for pterygium occurrence and development.

Pterygium samples and control conjunctiva tissue samples
A total of 263 conjunctiva tissues were analysed in this study, including 234 pterygium samples and 29 control conjunctiva tissues. All pterygium samples were obtained from patients who diagnosed with primary pterygium undergoing a surgical resection and control conjunctiva tissues were collected from patients who underwent cataract surgery from December 2014 to October 2019 at Zhongnan Hospital of Wuhan University. Samples were stored at -80 °C immediately after surgery until used.
This study was approved by the Medical Ethical Committee of Zhongnan Hospital of Wuhan University and followed the tenets of the Declaration of Helsinki and its later amendments. Informed consents were obtained from all patients before the study was carried out.

MiRNA microarray assay
Total RNA was isolated from frozen tissue using miRNeasy Mini Kit (Qiagen, Hilden, Germany) according to the manufacturer's instruction before evaluated by the NanoDrop2000 spectrophotometer (Thermo Scientific, MA, USA). Total RNA was purified and labelled using FlashTag ™ Biotin HSR RNA Labelling Kit (P/N 901,911, Affymetrix) according to the manufacturer's instructions to obtain biotin labelled miRNA. Array hybridization and wash were performed by GeneChip ® Hybridization, Wash and Stain Kit (P/N900720, Affymetrix Santa Clara CA, USA) and GeneChip Eukaryotic Hybridization Control Kit (P/N 900,454, Affymetrix Santa Clara CA, USA) in Hybridization Oven 645 (P/N 00-0331 (220 V), Affymetrix Santa Clara CA, USA) and Fluidics Station 450 (P/N 00-0079, Affymetrix Santa Clara CA, USA) according to the manufacturer's instructions. Arrays were scanned by GeneChip ® Scanner 7G (Affymetrix, Santa Clara, CA, USA) using Command Console Software 3.2 (Affymetrix, Santa Clara, CA, USA) with default settings. Raw data was normalized by RMA and DABG algorithm, Expression Console (Affymetrix, Santa Clara, CA, USA).

Rtranscription and quantitative real-time PCR
MiRNA and cDNA were synthesized from 500 ng total RNA by RevetAid RT Reverse Transcription Kit (Thermo Scientific, MA, USA) with specific miRNAs stem-loop RT primers (Table 1) and oligo d(T) 18 , using reverse transcription and quantitative real-time polymerase chain reaction (qRT-PCR). The miRNAs or mRNA expression levels were detected using Bio-Rad CFX96 ™ real-time PCR detection system (Bio-Rad, CA, USA) and each sample was detected in duplicate. Each assay consisted of 1 × SYBR Green quantitative real-time PCR master mix (Bio-Rad, CA, USA), 0.5 μM of forward and reverse primers (Table 1), and 1 μl cDNA template in a total volume of 20 μl. Non-template control was used as negative control, miRNAs reactions were normalized to U6, mRNA was normalized to GAPDH and the relative expression levels were calculated using the 2 −△Cq method.

Cell culture
The human conjunctiva epithelial cell line (HConEpic, HCE) was purchased from BeNa Culture Collection (Beijing, China), which was primary culture cells (Additional file 1). HCEs were supplemented with High glucose Dulbecco's modified Eagle's medium (H-DMEM, HyClone) containing 10% fetal bovine serum (Gibco) and 1uM penicillin-streptomycin (Gibco). Cells were grown at a humidified atmosphere of 5% CO 2 at 37 °C. After 24 h incubation in growth medium, TGF-β and EGF were added to obstain the optimal concentrations (10 nM & 20 nM). The medium was changed every other day and TGF-β and EGF were reintroduced to maintain the concentrations, and the cells that harvested after 7 days were used for subsequent experiments.

Plasmids construction and transient transfection
MiR-199a-3p and miR-199a-5p mimics and inhibitors were purchased from RiboBio Co. Ltd (Ribo, China) and Table 1 The list of primer sequences short hairpin RNAs (shRNAs) knocking down DUSP5 and MAP3K11 were purchased from Shanghai Genechem Co. Ltd (Genechem, China) ( Table 2). To overexpress DUSP5 and MAP3K11, a full-length coding sequence (CDS) was amplified, and the EcoRI-HF and XhoI sites (NEB, USA) were used to insert the CDS product into pCMV-myc vector (Invitrogen, NY, USA). Before transfection, 2*10 5 cells were seeded into each well of 6-well plates. After 24 h incubation in growth medium without penicillin-streptomycin, the cells were transiently transfected with miRNA mimics and inhibitors using riboFECT ™ CP (Ribo, China), or with plasmids using Lipofectamine 2000 (Invitrogen, Carlsbad, CA, USA). The cells were incubated for an additional 24-48 h for the following experiments.

Wound healing and transwell assay
HCEs dealt with TGF-β and EGF for 7d, were seeded equivalently into 6-well culture plates and then treated with transient transfection for 24 h before scratching or resuspending. A wound was scratched onto the monolayer with a sterile 20ul tip (Axygen, Union City, CA, USA). Images of HCEs migrating into the wound were captured at time points of 0, 24 and 48 h by an inverted microscope. The migration assay was performed using upper chambers of Transwell insert (0.8um pore size, Corning Incorporated, Costar, USA) with 4*10 5 cells in serum-free medium for 48 h. After migration, cells passed through the coated membrane to the lower surface, where cell were fixed with 4% paraformaldehyde and stained with 0.1% crystal violet. The cell was counted under a microscope.

Cell apoptosis determined by flow cytometry
The flow cytometry was performed to analyse the apoptosis of HCEs. Cells were re-suspended with Annexin V-FITC and propidiumiodide (PI) successively according to the manufacture's protocol (BestBio, China) at the concentration of 10 6 cells/ml. Cell apoptosis was analysed by flow cytometry (FACSCanto II; BD Bioscience, Franklin Lake, NJ).

Western blot analysis
After cells were seeded in 6-well plates for 48-72 h, cells were washed with PBS and harvested in RIPA with phosphorylase inhibitor and phenylmethanesulfonyl fluoride (PMSF). Equal amounts of the supernatant were loaded per lane and resolved by SDS-polyacrylamide electrophoresis. Then, proteins were transferred onto polyvinylidene fluoride (PVDF) membrane (Millipore, USA) and blocked by 5% BSA. Primary antibodies should be probed overnight at 4 °C, including rabbit anti-MAP3K11, anti-DUSP5, anti-Vimentin, anti-E-Cadherin, anti-N-Cadherin antibodies and mouse anti-GAPDH (abClone) antibody. Membranes were washed in Tris buffered saline tween (TBS-T) and incubated with horseradish peroxidase-conjugated anti-rabbit or anti-mouse secondary antibodies. Membranes were washed in TBS-T and then exposed using the electrochemical luminescence (ECL) system. Protein loading was normalized by GAPDH.

MiRNAs target gene prediction and GO/pathway analysis
Differentially expressed miRNAs were subjected to target gene prediction analysis using TargetScan, miRDB, mirTarBase, miRanda, Targetminer and miR-NAorg. The predicted results of target genes were shown by Venn diagram. GO network maps and term enrichment analysis were performed by using plugin of Cytoscape: ClueGO and CluePedia, with terms

Statistical analysis
All data were analysed by GraphPad-Prism8.0 (Graph Pad, CA, USA) in independent t-test, Mann-Whitney U test or Pearson Correlation. p < 0.05 (two-tailed) was considered statistically significant.

Expression profiling showed that miR-199a was upregulated in pterygium
To obtain miRNAs expression profiling in pterygium, miRNA microarray analysis was performed in 3 pairs of pterygium and control conjunctiva tissues. In the screened 2578 miRNAs, there were 1362 upregulated and 1216 downregulated miRNAs in pterygium, compared with control conjunctiva (Fig. 1a). Using p < 0.05 and |log- map for visualization (Fig. 1c). Among them, 30 miRNAs were upregulated, and 10 miRNAs were downregulated. Furthermore, some of the differentially expressed miRNAs were verified in a small group of samples as the explorer category, including 55 pterygium samples and 12 control conjunctiva tissues. Coincident with the results of miRNA microarray, the qRT-PCR verified the differential expressed miRNAs except for miR-675-5p, which presented no significant decrease in the 55 pterygium samples (p > 0.05) (Fig. 1d). Based on the essential role of miR-199 family in the EMT process, and the primary exploration results of microarray, miR-199a-3p and miR-199a-5p were selected for further exploration.

The EMT cell model was established using HCEs induced by TGF-β and EGF
We attempted to establish the pterygium EMT-cell model in HCEs, by using TGF-β and EGF stimulation. After cultured in the inducement of 10 nM TGF-β and 20 nM EGF for 7 days, the HCEs grew in a long spindle shape, showing a mesenchymal cell phenotype (Fig. 2a). The characteristic of EMT-HCEs was identified by EMT markers detected by qRT-PCR, western blot and immunofluorescence. Compared with control HCEs, the expression level of N-cadherin and Vimentin increased significantly, while E-cadherin decreased (Fig. 2b-d). Then, the wound healing and transwell assays were executed and found that the migration ability of induced cells was also improved after EMT activation (Fig. 2e, f ). Furthermore, decreased cell apoptosis was observed by flow cytometry with the Annexin V-FITC/PI reagent (Fig. 2g).
To further investigate that whether the EMT-HCEs model could represent the EMT process in pterygium, we compared expressions of EMT markers between primary culture pterygium cells and HCEs and found the EMT markers were significantly lower in the latter (Additional file 2).

DUSP5 and MAP3K11 were the targets of miR-199a-3p and miR-199a-5p
The predicted target genes DUSP5 and MAP3K11 were first selected from multiple miRNA related databases, then filtrated with Venn diagram (Fig. 5a and b). The predicted binding between the 3′-untranslated region (3′-UTR) of DUSP5/MAP3K11 and miR-199a-3p/5p were confirmed by dual-luciferase reporter gene assay  Fig. 5c and d). Luciferase activity of DUSP5-WT 3′UTR could be inhibited by miR-199a-3p mimic, but DUSP5-MUT 3′UTR could not (Fig. 5e). There were 2 predicted binding sites between the 3′UTR of MAP3K11 and miR-199a-5p, and luciferase activity of MAP3K11-WT (wild type) 3′UTR was inhibited by miR-199a-5p mimic. With mutation at either of the two binding sites (MAP3K11-MUT1 and MAP3K11-MUT2), luciferase activity could still be inhibited by mimic, but the inhibition rate was approximately 50% while with mutations at both binding sites (MAP3K11-MUT1 + 2) it was nearly 100% (Fig. 5f ). The negative correlation between miR-199a-3p/5p and their target genes in both mRNA and protein levels were observed in tissues and cell models to further verify that miR-199a-3p/5p downregulated the expression of DUSP5 and MAP3K11 (Fig. 5g-j).

DUSP5/MAP3K11 inhibition could promote EMT in HCEs
It was shown that DUSP5-shRNA2# and MAP3K11-shRNA1# had the best effects on DUSP5/MAP3K11 knockdown, and both of the overexpression vectors demonstrate significantly upregulations of the two gene ( Fig. 6a-d). Some HCEs were treated with TGF-β and EGF for 36 h to extract RNA, while the others were treated for 48 h to extract protein or use for wound healing and transwell assay. (Fig. 6e and f ). Knockdown of either DUSP5 or MAP3K11 promoted the HCEs migration, and when either of the two genes was overexpressed cell motility was inhibited ( Fig. 6g and h).

Suppression on DUSP5/MAP3K11 induced by TGF-β and EGF could be alleviated by decreased miR-199a-3p/5p expression
Along with the prolongation of TGF-β and EGF induction, the expressions of miR-199a-3p and miR-199a-5p were increased, while the expressions of DUSP5 and MAP3K11 were decreased. Moreover, compared with HCE, the expression levels of DUSP5 and MAP3K11 of induced HCEs were significantly reduced (Fig. 7a-e). However, inhibiting the expression of miR-199a-3p/5p in EMT-HCEs resulted in increased expression of corresponding target genes (Fig. 7f-i). That is to say, the inhibition on DUSP5 and MAP3K11 expressions induced by TGF-β and EGF was weakened by decreasing the expression of miR-199a-3p or miR-199a-5p. So, we speculated that miR-199a-3p and miR-199a-5p might target DUSP5 and MAP3K11 to function in EMT of pterygium.

MiR-199a-3p/5p potentiated migration in EMT of HCEs induced by TGF-β and EGF by targeting DUSP5/MAP3K11
Transwell assays were then used to determine the effect of miR-199a-3p/5p-DUSP5/MAP3K11 axes regulation on TGF-β and EGF-induced migration of HCEs. Treated with 10 nM TGF-β and 20 nM EGF for 7 days, the migration ability of HCEs was increased, while the effects were inhibited when either of DUSP5 or MAP3K11 was overexpressed. The migration ability of HCEs was enhanced when miR-199a-3p or miR-199a-5p was overexpressed by mimics but decreased when either of the two miR-NAs was knockdown by inhibitors ( Fig. 8a and b). The inhibition of miR-199a-3p or miR-199a-5p resulted in the enhancement of DUSP5 or MAP3K11-mediated inhibition on EMT-HCEs migration, while conversely promoting miR-199a-3p or miR-199a-5p could further induce migration of EMT-HCEs. The change trend of EMT markers was consistent with that of migration ability. Overexpression of DUSP5 or MAP3K11 rescued the effects of TGF-β and EGF-induced EMT, while miR-199a-3p/5p could further promote the induction by regulating the expression of DUSP5/MAP3K11 respectively ( Fig. 8c and d). In short, the promotion of EMT by miR-199a-3p/5p was achieved by regulating target genes DUSP5/MAP3K11.

Downstream pathway prediction by bioinformatics
In order to explore the possible downstream pathway of miR-199a-3p and miR-199a-5p in greater depth in the occurrence and development of EMT in pterygium, we subjected the potential targets of the two miRNAs to pathway enrichment analysis. The network showed that MiR-199a-3p and miR-199a-5p target DUSP5 and MAP3K11 respectively. a, b Venn diagram of miR-199a-3p and miR-199a-5p target genes prediction results from 4 databases. c, d Putative miR-199a-3p and miR-199a-5p binding sites in the 3′UTR of DUSP5 and MAP3K11, and the designed mutation sites. e, f A luciferase reporter plasmid containing wild type or mutant type were co-transfected with relevant mimic, mimic NC, inhibitor and inhibitor NC into HCEs. Luciferase activity was normalized to Renilla luciferase activity. Of note, MAP3K11 3′UTR had 2 binding sites withmiR-199a-5p, so 3 kinds of mutants were construct, including mutant at the first site, mutant at the second site and mutant at both sites (**p < 0.01, ***p < 0.001, ****p < 0.0001). g-j After HCEs were transfected with miR-199a-3p/miR-199a-5p mimic, mimic NC, miR-199a-3p/ miR-199a-5p inhibitor and inhibitor NC, the variations of DUSP5 and MAP3K11 expression were analyzed with qRT-PCR and western blot (*p < 0.05, **p < 0.01, ***p < 0.001) miR-199a-3p and miR-199a-5p might participate in the development of pterygium by affecting MAPK signalling pathway, TGF-β signalling pathway, PI3K-Akt signalling pathway, focal adhesion and others (Fig. 10a). From the pathway enrichment network, both of the DUSP5 and MAP3K11 were related with MAPK signalling pathway and we speculated that miR-199a might be involved in the MAPK pathway by targeting DUSP5 and MAP3K11, to promote EMT in HCE. Consistent with other researches, both of miR-199a-3p and miR-199a-5p had the positive effect on EMT processes and induced cell migration through MAPK signalling pathway, such as miR-199a-3p-DUSP5-ERK and miR-199a-5p-MAP3K11-JNK-p53 pathway. The pathway network was established based on our wet-lab experiments and dry-lab analysis (Fig. 10b).

Discussion
MiR-199a-3p-DUSP5 and miR-199a-5p-MAP3K11 axes regulated EMT in pterygium were characterized in our study. In accordance with previous researches, both activated TGF-β and EGF stimulated the process of migration, proliferation and EMT in many different diseases [36][37][38], our study proved the feasibility of building EMT cell model using TGF-β and EGF. Furthermore, it presented that miR-199a-3p and miR-199a-5p participated in the process of EMT induced by TGF-β and EGF, by targeting DUSP5 and MAP3K11 in pterygium. First, we got the miRNA expression profile of pterygium using the Affymetrix' miRNA 4.0 microarray and validated it in pterygium samples, where we first observed the significant increase of miR-199a in pterygium. Although there were several researches, using microarray to screen the differentially expressed miR-NAs in pterygium [2,[26][27][28], the Affymetrix' miRNA 4.0 microarray provided more complete measurements of miRNA, compared to GeneChip ® miRNA 2.0 and 3.0, based on reports in Sanger miRBase 20.0 database. For instance, a larger number of miRNAs reads, reached to 6659, were discovered in our study compared to previous pterygium microarray researches, such as Silin Chen's (GSE21346), only 1380 reads.
MiR-199 family, consisting of miR-199a/b-3p and miR-199a/b-5p, is encoded within the Dynamin (DNM) genes and exhibit high conservation across species [39]. MiR-199a-3p/5p are found to inhibit EMT in various cancers, while they also can activate EMT in fibrosis [15,34,35]. The reason that why miR-199a-3p/5p have different expression level in EMT lies in the intrinsic mechanism of different types of EMT [12].
Second, miR-199 family functioning in pterygium is consist of miR-199a/b-3p and miR-199a/b-5p, and encoded within the Dynamin (DNM) genes, which exhibits high conservation across species [39]. MiR-199a-3p/5p are found to inhibit EMT in various cancers, while they also can activate EMT in fibrosis diseases [15,34,35]. The reason that why miR-199a-3p/5p have contradiction functions in EMT of pterygium compared with cancers, probably due to the essential fibrosis characteristic of pterygium [12].
Third, there are mounting evidences suggesting that miR-199 family acts as a fatal effector of TGF-β signalling in many diseases, regulating multiple disordered processes, including cell proliferation, apoptosis, migration, invasion and EMT [15,34,40]. The most typical biomarker change in EMT was the conversation from E-cadherin to N-cadherin, accompanied by the increase of cell migration ability and the decrease of cell apoptosis [41]. We found that as increasing time of TGF-β and EGF treatment, expressions of both miR-199a-3p and miR-199a-5p were increased gradually. The EMT characteristics including migration etc. of induced HCEs could be suppressed by inhibiting the expression of either miR-199a-3p or miR-199a-5p. It disclosed that miR-199a-3p/5p possibly acted as TGF-β effector in pterygium.
Forth, Antoon JW et al. recently found that inhibition of p38 mitogen-activated protein kinase (MAPK) signalling pathway even could reverse EMT [42], which is previously considered as a key process of inducting and maintaining the inflammation in various disease [43]. Accordingly, our study found that both DUSP5 and MAP3K11 implicated in MAPK signalling pathway, might be a potential therapeutic agent targeted specifically to reverse EMT in pterygium. Consistently, knockdown DUSP5 or MAP3K11 promoted EMT in the present study. However, upregulated DUSP5 inhibited the process of EMT in gastric cancer and hepatocellular carcinoma through MAPK pathway, and the cells showed a reduced migration ability and increased apoptosis [44,45]. Furthermore, combining with existing literature, our pathway bioinformatics analysis also speculated that The shRNAs which had the best knockdown efficiency were selected with qRT-PCR. c, d The effect of pCMV-myc-DUSP5, pCMV-myc-MAP3K11, DUSP5-shRNA2# and MAP3K11-shRNA#1. HCEs were transfected with pCMV-myc and scr as control. e Cell migration was determined by the wound healing assay (**p < 0.01). f Cell migration was determined by the transwell assay (**p < 0.01). g, h The changes in EMT markers, after HCEs were transfected with pCMV-myc-DUSP5, pCMV-myc-MAP3K11, pCMV-myc, DUSP5-shRNA2#, MAP3K11-shRNA#1 and shRNA control. (*p < 0.05, **p < 0.01, ***p < 0.001) (See figure on next page.)  . 7 The effects of TGF-β and EGF on expressions of DUSP5 and MAP3K11 could be regulated by miR-199a-3p and miR-199a-5p. a-d HCEs were induced with TGF-β (10 nM) and EGF (20 nM) for 7 days. The expressions of miR-199a-3p, miR-199a-5p, DUSP5 and MAP3K11 were detected by qRT-PCR analysis (*p < 0.05, **p < 0.01). e The protein expression of DUSP5 and MAP3K11 were analyzed with western blot (*p < 0.05, **p < 0.01). f, g.