METTL14-regulated PI3K/Akt signaling pathway via PTEN affects HDAC5-mediated epithelial–mesenchymal transition of renal tubular cells in diabetic kidney disease

Histone deacetylase 5 (HDAC5) belongs to class II HDAC subfamily and is reported to be increased in the kidneys of diabetic patients and animals. However, little is known about its function and the exact mechanism in diabetic kidney disease (DKD). Here, we found that HDAC5 was located in renal glomeruli and tubular cells, and significantly upregulated in diabetic mice and UUO mice, especially in renal tubular cells and interstitium. Knockdown of HDAC5 ameliorated high glucose-induced epithelial–mesenchymal transition (EMT) of HK2 cells, indicated in the increased E-cadherin and decreased α-SMA, via the downregulation of TGF-β1. Furthermore, HDAC5 expression was regulated by PI3K/Akt signaling pathway and inhibition of PI3K/Akt pathway by LY294002 treatment or Akt phosphorylation mutation reduced HDAC5 and TGF-β1 expression in vitro high glucose-cultured HK2 cells. Again, high glucose stimulation downregulated total m6A RNA methylation level of HK2 cells. Then, m6A demethylase inhibitor MA2 treatment decreased Akt phosphorylation, HDAC5, and TGF-β1 expression in high glucose-cultured HK2 cells. In addition, m6A modification-associated methylase METTL3 and METTL14 were decreased by high glucose at the levels of mRNA and protein. METTL14 not METTL3 overexpression led to PI3K/Akt pathway inactivation in high glucose-treated HK2 cells by enhancing PTEN, followed by HDAC5 and TGF-β1 expression downregulation. Finally, in vivo HDACs inhibitor TSA treatment alleviated extracellular matrix accumulation in kidneys of diabetic mice, accompanied with HDAC5, TGF-β1, and α-SMA expression downregulation. These above data suggest that METTL14-regulated PI3K/Akt signaling pathway via PTEN affected HDAC5-mediated EMT of renal tubular cells in diabetic kidney disease.


Introduction
Diabetic kidney disease (DKD) is a kind of severe chronic complication of diabetes mellitus (DM) and has been reported to have multiple morphologic changes including hypertrophy and proliferation of mesangial cells, epithelial-mesenchymal transition (EMT) of renal tubular cells, apoptosis of podocytes, and so on 1 . Among these, EMT of renal tubular cells is the most common reason to cause impaired renal function 2 . TGF-β1 pathway has been demonstrated to be the important pro-fibrosis factor in the pathogenesis of diabetic kidney disease 3 .
Accumulating evidence suggests that inhibition of histone deacetylase (HDAC) ameliorates diabetic kidney disease manifestations and phenotypes such as fibrosis, inflammation, cell death, and albuminuria 4,5 . A pan-HDAC inhibitor trichostatin A (TSA) was reported to suppress EMT, indicated in upregulation of E-cadherin and downregulation of collagen type I, in TGF-β1-treated human renal proximal tubular epithelial cells 6 . HDAC5 belongs to HDAC family that is found to regulate gene transcription by affecting histone acetylation and chromatin remodeling. Among four kinds of HDAC genes, HDAC5 belongs to class II HDAC subfamily, together with HDAC4, 6, 7, 9, and 10 7 . Recently, Wang and coworkers revealed the increased expression of HDAC2, HDAC4, and HDAC5 in the kidneys of STZ-diabetic rats and db/db mice, and increased expression of HDAC4 and HDAC5 in the kidneys of humans with diabetic kidney disease 8 . However, the effect and mechanism of HDAC5 on EMT of renal tubular cells of DN is still not known.
Phosphatidylinositol-3-kinase (PI3K), known as a lipid kinase, generates PIP3 that regulates the translocation of Akt to plasma membrane (PM) as a second messenger. Then, Akt is activated by phosphorylation modification of threonine 308 and serine 473 9 . PI3K/Akt signaling pathway plays a key regulatory role in the development of diabetic kidney disease. PI3K/Akt pathway is activated in renal tubular cells under the diabetic condition to regulate cell growth, EMT, and lipid metabolism 10,11 .
N6-methyladenosine (m6A) RNA methylation is an abundant and conservative RNA modification in mammals and is regulated by a series of "writer", "eraser", and "reader" proteins. METTL3-METTL14 complex catalyzes m6A modification of RNA as "writer" and FTO and ALKBH5 remove methyl from RNA as "eraser". "Reader" proteins are responsible for recognizing m6A methylated transcripts to regulate mRNA processing, translation, and degradation 12 . Deficiency of m6A RNA methylation is associated with many kinds of diseases, such as cancer and diabetes mellitus 13 .
In the present study, we first revealed that HDAC5 expression was increased in renal tubular cells of diabetic mice, fibrosis region of unilateral ureteral obstruction (UUO) mice, and in vitro high glucose-cultured human renal tubular cell line (HK2). Knockdown of HDAC5 inhibited high glucose-induced EMT in HK2 cells via TGF-β1 downregulation. Again, blocking PI3K/Akt pathway in high glucose-treated HK2 cells reduced HDAC5 expression, followed by TGF-β1 downregulation. Furthermore, m6A RNA methylation level was reduced in high glucose-stimulated HK2 cells. The m6A demethylase inhibitor MA2 suppressed Akt phosphorylation and HDAC5 expression. Then, m6A RNA methylation-related genes' detection revealed the downregulation of METTL3, METTL14, and FTO in high glucose-cultured HK2 cells. Overexpression of METTL14 not METTL3 reversed high glucose-activated Akt pathway and HDAC5 expression via PTEN. Finally, in vivo administration of TSA reduced HDAC5 expression and extracellular matrix (ECM) accumulation in the kidneys of diabetic mice.

Animals and groups
C57BL/6J mice, at 6-8 weeks of age, were purchased from Vital River Laboratory Animal Technology Company (Beijing, China) and were housed and handled according to the guidelines approved by Hebei Medical University Animal Care and Use Committee. PASS Sample Size Software was used to estimate sample size. Simple random sampling method was used for the group allocation of experimental mice and the detection was performed blindly. Twenty mice were randomly divided into three groups: normal mice group (N), diabetic mice group (DM), and diabetic mice administrated with TSA group (DM + TSA). Type 1 diabetic mice were made by intraperitoneally injecting streptozotocin (STZ, 150 mg/kg body weight). Three days later, mice with random blood glucose greater than 11.1 mmol/L were regarded as diabetic models. Mice were intraperitoneally injected with TSA at 0.5 mg/kg weight three times per week. After two months, all mice were sacrificed for the related detections. Also, twelve mice were randomly divided into two groups: sham group and unilateral ureteral obstruction (UUO) group. UUO mice were made according to the standard protocol. In detail, the animals were anesthetized with the injection of 10% chloral hydrate. A midline incision of abdomen was made and the left ureter was ligated at two points. Then the abdominal incision was stitched. The mice with the left ureter exposed but not ligated were regarded as sham mice. After two weeks, all animals were killed and the related detections were performed.

Small hairpin RNA plasmid construction
The shRNA plasmids aimed at HDAC5 and METTL14 were constructed and named as pGenesil-1-HDAC5 and pGenesil-1-METTL14. The target sequences of human HDAC5 (NM_001015053.1) and METTL14 (NM_020961.4) were, respectively, CTGTTATTAGCACCTTTAAGAA and GCTAATGT TGACATTGACT TA. The protocol was performed to construct plasmid as described previously 14 . The plasmid was identified with the methods of Sal I enzyme digestion and sequencing.

Plasmid transfection
Lipofectamine TM 3000 was used to perform cell transfection and the detailed protocol was as follows. A 2.5 µg plasmid and 5 µL P3000 were mixed with 125 µL serumfree DMEM medium, and then mixed with an additional 125 µL serum-free DMEM medium containing 5 µL Lipofectamine™ 3000. Five minutes later, the mixture was added into HK2 cells in 6-well plates. After 48 h, the cells were collected and the related detections were performed.

Protein extraction and western blot
Protein extraction was performed according to the method used in the previous research 15 . In detail, cells were washed with PBS and lysed for 30 min at 4°C with 2.5 mM Tris-HCl (pH 7.5), 100 mM NaCl, 0.1% Triton X-100, 1% NP-40, 30 mM sodium phosphate (pH 7.4), 1 mM sodium orthovanadate, 10 μg/mL leupeptin, and 10 μg/mL aprotinin. Subsequently, the homogenate was centrifuged at 12,000g for 30 min at 4°C and the supernatant containing protein was collected. Also, renal tissue protein was extracted following the same procedure. After protein quantitative assay, 30 μg total protein was separated on 10% SDS-PAGE gel and transferred onto PVDF membrane, followed by an incubation with 5% BSA at 37°C for 1 h. Next, blots were washed with TBST and incubated with primary antibodies overnight at 4°C. The next morning, blots were rinsed with TBST and incubated with horseradish peroxidase-conjugated secondary antibody for 1 h at room temperature. Bands were visualized using ECL detection reagents for blots incubation. For quantitative analysis, the integrated optical density (IOD) bands were evaluated with LabWorks software (UVP Laboratory Products, Upland, CA, USA), normalized by β-actin.

Real-time PCR
Total RNA was extracted from HK2 cells using Trizol reagent in accordance with standard procedure. The cDNA was synthesized using PrimeScript TM RT reagent Kit with gDNA Eraser. Then SYBR Ⓡ Premix Ex Taq TM II (Tli RNaseH Plus) kit was used for real-time PCR detection in accordance with the protocol as follows: 95°C for 5 min, 95°C for 10 s, 55°C for 30 s, 72°C for 10 s, 40 cycles. Likewise, melting curves were made to confirm the amplification specificity. Primer sequences were as given in Table 1 and 18S was used to normalize the relative expression as an internal reference. The results were analyzed using the 2 −ΔΔCT method and shown as relative quantity (gene/18S).
Immunofluorescence HK2 cells seeded on the cover slips in 6-well plate were fixed with 4% paraformaldehyde, permeabilized with 0.3% Triton X-100, and blocked with goat serum, in turn. Then, cells were incubated with primary antibodies overnight at 4°C. After washing with PBS, cells were incubated with DyLight 594-conjugated secondary antibody at 37°C for 2 h. After washing with PBS and counterstaing with DAPI, cells were observed and a photograph was taken under an inverted fluorescence microscope.

Immunohistochemistry
HDAC5 expression in the kidneys of diabetic mice and UUO mice was detected by the method of immunohistochemistry. In detail, sections were boiled in pressure cooker for antigen retrieval, after deparaffinization and hydration. Afterwards, sections were treated using 0.3% H 2 O 2 at room temperature for 10 min to quench endogenous peroxidase. Next, goat serum was used to incubate sections for 30 min at 37°C to block non-specific staining. Then, sections were incubated with primary antibody overnight at 4°C. After rinsing with PBS, sections were incubated with biotin-conjugated secondary antibody at 37°C for 30 min and HRP-conjugated streptavidin at 37°C for 30 min, in turn. Finally, sections were incubated with 3,3′-diaminobenzidine tetrahydrochloride (DAB) to show the positive staining. PBS was used to replace primary antibody for negative control.

Masson trichrome staining
Masson trichrome staining was applied to determine the ECM accumulation and fibrosis in the kidneys of diabetic mice and UUO mice, according to the protocol as follows. Renal tissue sections were deparaffinized, rehydrated, and immersed in Weigert iron haematoxylin solution for 10 min. Subsequently, rinsed with water, sections were immersed in Ponceau acid fuchsin solution for 10 min. Then sections were treated with phosphomolybdicphosphotungstic acid for 10 min for differentiation. After immersed in aniline blue solution for 10 min, the sections were differentiated by treatment with 1% acetic acid for 1 min, followed by dehydration, clearing, and mounting.

Total N6-methyladenosine RNA methylation detection
Total N6-methyladenosine RNA methylation was detected according to the instruction of EpiQuik TM m6A RNA Methylation Quantification Kit. RNA extraction from HK2 cells was performed using Trizol reagent. First, the 80 μL BS (binding solution) was added to each well. Then RNA samples, negative control, and positive control were added into the designated wells and placed for 90 min at 37°C. After washing with WB (washing buffer) for three times, 50 μL diluted CA (capture antibody) was added to each well and incubated for 60 min at room temperature. After rinsing with WB, 50 μL diluted DA (detection antibody) and 50 μL diluted ES (enhancer solution) were added into each well and incubated for 30 min at room temperature, in turn. Then, the 100 μL DS (developer solution) was used to incubate each well for 10 min to produce color change, followed by the supplement of 100 μL SS (stop solution) to stop the reaction. Finally, the absorbance was read on a microplate reader at 450 nm. The amount of m6A RNA of samples was calculated according to the standard curve. Then the percentage of m6A RNA methylation was the ratio of the amount of m6A RNA and the amount of input total RNA.

Statistical analysis
All data were presented as mean ± SD from at least three independent experiments. GraphPad Prism 6 software was used for statistical analyses. For experiments with two groups, statistical analyses were performed using Student's t-test. For more than two groups, statistical analyses were performed using one-way analysis of variance (ANOVA), followed by the Bonferroni post hoc test to determine the differences within and between groups. The Kruskal-Wallis test was used for the comparisons of data with non-normal distribution or heterogeneity of variance. P value <0.05 was considered as significant.

HDAC5 expression was increased in renal tubular cells of diabetic mice and high glucose-treated HK2 cells
We first detected the expression of HDAC5 in the kidneys of diabetic mice by the methods of western blot and immunohistochemistry. As seen in Fig. 1A, HDAC5 was increased by 2.80 times in the kidneys of diabetic mice compared with those of normal mice (P < 0.05). Furthermore, immunohistochemistry revealed that HDAC5 expressed in renal glomeruli, renal tubular cells, and renal interstitium. Especially, HDAC5 expression was enhanced in renal tubular cells of diabetic mice versus normal mice (Fig. 1B). Again, we explored the effect of high glucose known as the main feature of diabetes mellitus on HDAC5 expression in vitro-cultured human renal tubular cells (HK2 cells) and the results showed that HDAC5 was increased by 41.01% in HK2 cells treated with high glucose (40 mmol/L glucose) for 48 h compared to those treated with normal glucose (10 mmol/L glucose) (Fig. 1C). Similarly, immunofluorescence also confirmed the overexpression of HDAC5 in high glucose medium-cultured HK2 cells, indicated in green fluorescence in the cytoplasm of cells (Fig. 1D). In addition, HDAC5 mRNA was found to be upregulated by 31.15 times in high glucose-stimulated HK2 cells versus normal glucosetreated cells using real-time PCR technology (Fig. 1E).
Interestingly, in unilateral ureteral obstruction (UUO) mice known as the classical renal fibrosis model, we found the prominently increased expression of HDAC5 in renal tubular cells located in fibrotic interstitium region, in comparison with those cells located in normal region (Fig. 1F, G). The above data suggested that HDAC5 expression was upregulated in renal tubular cells of diabetic kidney disease, which might be involved in epithelial-mesenchymal transition (EMT) of renal tubular cells and renal interstitial fibrosis.

Knockdown of HDAC5 ameliorated high glucose-induced EMT of HK2 cells
To further elucidate the function of HDAC5 in renal tubular cells of diabetes mellitus, we knocked down HDAC5 in HK2 cells treated with high glucose using shRNA plasmid targeted at HDAC5 (pGenesil-1-HDAC5). It could be seen in Fig. 2A that transfection efficiency was more than 80% and HDAC5 was reduced by 53.24% in pGenesil-1-HDAC5transfected HK2 cells compared with pGenesil-1-transfected HK2 cells (P < 0.05) (Fig. 2B). Then as seen in Fig. 2C, epithelial marker E-cadherin and mesenchymal marker α-SMA were regulated by HDAC5 knockdown in HK2 cells treated with high glucose. Statistical analysis revealed that E-cadherin was increased by 39.06% and α-SMA was decreased by 46.75% in high glucose-treated HK2 cell transfected with pGenesil-1-HDAC5 compared with those transfected with pGenesil-1 (P < 0.05). In line, immunofluorescence staining of α-SMA also presented that the high glucose-cultured HK2 cells successfully transfected with pGenesil-1 (indicated in GFP expression) showed the same α-SMA expression as those failed to be transfected with pGenesil-1 (no GFP expression). Differently, HK2 cells successfully transfected with pGenesil-1-HDAC5 plasmid showed evident green fluorescence (GFP expression), and the reduced α-SMA expression (red fluorescence) (Fig. 2D). Furthermore, cell migration experiment showed that 24 h-high glucose treatment enhanced cell migration ability of HK2 cells, which was prevented with pGenesil-1-HDAC5 transfection. In detail, the rest area was decreased by 43.20% in high glucose-treated HK2 cells versus normal glucose-cultured HK2 cells (P < 0.05). While pGenesil-1-HDAC5 transfection increased the rest area by 1.78 times compared with pGenesil-1 transfection in HK2 cells (P < 0.05) (Fig. 2E). Moreover, we detected ECM protein expression and found that high glucose-induced increased fibronectin, collagen 1, and collagen 3 mRNA. In detail, fibronectin, collagen 1, and collagen 3 mRNA were, respectively, increased by 4.53, 5.77, and 6.77 times (P < 0.05) (Fig. 2F). Again, high glucose-increased fibronectin, collagen 1, and collagen 3 mRNA were effectively avoided with HDAC5 knockdown (Fig. 2G). Therefore, these data suggested that HDAC5 played a key role in regulating EMT of renal tubular cells and ECM accumulation.  TGF-β1 pathway was involved in HDAC5-regulated EMT of HK2 cells TGF-β1 pathway was the important signaling pathway to regulate EMT of renal tubular cells under the condition of hyperglycemia. We speculated that TGF-β1 pathway might be involved in HDAC5-reversed EMT of HK2, as the downstream targets of HDAC5. The results of western blot are illustrated in Fig. 3A that HDAC5 knockdown decreased TGF-β1 expression in both normal glucose-cultured HK2 cells and high glucose-cultured HK2 cells, especially in normal glucose-cultured cells. Statistical analysis confirmed TGF-β1 was, respectively, decreased by 54.36% and 24.35% in normal and high glucose-cultured HK2 cells transfected with pGenesil-1-HDAC5 versus pGenesil-1 (P < 0.05). In line with the results of western blot, immunofluorescence detection also showed that in pGenesil-1-transfected HK2 cells, there was no difference in TGF-β1 expression between transfected cells (with GFP expression) and untransfected cells (without GFP expression) (Fig. 3B). On the contrary, in pGenesil-1-HDAC5-transfected HK2 cells, the transfected cells indicated in GFP expression showed a decreased TGF-β1 expression compared with untransfected cells (Fig. 3C). Furthermore, exogenous TGF-β1 addition reversed the effect of HDAC5 knockdown on α-SMA in high glucose-treated HK2 cells. Statistical analysis revealed that α-SMA was, respectively, increased by 1.43 times and 1.72 times with 24 h-TGF-β1 treatment and 48 h-TGF-β1 treatment versus only pGenesil-1-HDAC5-transfected cells (P < 0.05) (Fig. 3D). Therefore, these above findings advised that HDAC5 knockdown ameliorated high glucose-induced EMT of renal tubular cells via TGF-β1 pathways.

PI3K/Akt pathway mediated high glucose-induced HDAC5 upregulation in HK2 cells
Considering that PI3K/Akt pathway is the pivotal pathway to regulate renal tubular cell function in diabetes mellitus, we further explored the potential influence of PI3K/Akt pathway on high glucose-induced HDAC5 upregulation in HK2 cells. Firstly, we verified the quick activation of Akt pathway in HK2 cell stimulated with high glucose. Phospho-Akt (Ser 473) and phospho-Akt (Thr 308) were, respectively, increased by 1.20 times and 1.23 times with high glucose stimulation for 1 h versus normal glucose treatment in HK2 cells (P < 0.05) (Fig. 4A). Subsequently, inhibition of Akt pathway with LY294002 effectively prevented high glucosecaused HDAC5 upregulation. Compared with DMSO control treatment, LY294002 treatment decreased HDAC5 expression by 50.43% in high glucose-cultured HK2 cells (P < 0.05) (Fig. 4B). Also, the results of immunofluorescence were similar to the results of western blot. LY294002 inhibited high glucose-induced HDAC5 overexpression in HK2 cells (Fig. 4C). In turn, downstream target of HDAC5, TGF-β1 was also downregulated with the treatment of LY294002 in high glucose-cultured HK2 cells. In detail, TGF-β1 was decreased by 38.25% with LY294002 treatment (P < 0.05) (Fig. 4D).

N6-methyladenosine demethylase inhibitor MA2 ameliorated high glucose-induced Akt pathway activation and HDAC5 expression in HK2 cells
N6-methyladenosine (m6A) RNA methylation is the most common RNA modification to regulate cell signaling conduction and gene expression. We first detected the level of total m6A mRNA methylation in high glucosetreated HK2 cells. The results were shown in Fig. 5A that high glucose reduced the level of total m6A RNA methylation in HK2 cells, which was prevented with 50 μmol/L MA2 treatment known as FTO (m6A RNA demethylase) inhibitor for 12 h. In detail, total m6A RNA methylation was, respectively, decreased by 28.95% and 64.25% with high glucose treatment for 24 h and 48 h versus normal glucose treatment (P < 0.05). Also, total m6A RNA methylation was enhanced by 1.79 times with MA2 treatment in 48 h-high glucose-cultured HK2 cells (P < 0.05). In addition, to further elucidate whether m6A RNA methylation was associated with Akt pathway activity and HDAC5 expression, we used MA2 to treat high glucose-cultured HK2 cells. Then Akt phosphorylation was significantly decreased with the treatment of MA2, followed by HDAC5 downregulation. Statistical analysis revealed that phospho-Akt (Ser 473) and phospho-Akt (Thr 308) were, respectively, decreased by 27.94% and 43.97% in high glucose-cultured HK2 cells with the treatment of MA2 compared to DMSO treatment (P < 0.05) (Fig. 5B). Similarly, HDAC5 expression was also reduced by 50.70% with MA2 treatment (P < 0.05) (Fig. 5C). Also, immunofluorescence detection showed the similar results to western blot that MA2 stimulation decreased HDAC5 protein expression (Fig. 5D). Again, TGF-β1 and EMT marker α-SMA were also decreased by the treatment of MA2 in high glucose-cultured HK2 cells. Statistical analysis revealed a 20.04% decrease of TGF-β1 and a 31.19% decrease of α-SMA in MA2-treated HK2 cells versus DMSO-treated cells (P < 0.05) (Fig. 5E). The above data suggested that N6methyladenosine mRNA methylation level might mediate high glucose-induced Akt pathway activation, HDAC5 upregulation and EMT of renal tubular cells.

Akt pathway agonist insulin prevented MA2-reduced HDAC5 expression in high glucose-treated HK2
In order to determine the direct regulation of m6A RNA modification on Akt pathway, as well as HDAC5 expression in high glucose-treated HK2, we pretreated HK2 cells with insulin (activator of PI3K/Akt pathway) and revealed that MA2-reduced HDAC5 expression was effectively  reversed. Statistical analysis confirmed that the phospho-Akt (Ser 473) and phospho-Akt (Thr 308) were, respectively, enhanced by 1.60 times and 1.43 times in insulin plus MA2-treated HK2 cells compared with only MA2treated cells (Fig. 5F). Similarly, HDAC5 expression was also increased by 1.45 times in insulin plus MA2-treated HK2 cells compared with only MA2-treated cells (P < 0.05) (Fig. 5G). The results of immunofluorescence were similar to those of western blot. Insulin treatment significantly increased HDAC5 expression, indicated in red Fig. 5 The effect of MA2 and insulin on phospho-Akt (Ser 473), phospho-Akt (Thr 308), Akt, HDAC5, TGF-β1, and α-SMA in HK2 cells. A Total m6A methylation level detection in HK2 cells treated with high glucose. n = 3, *means P < 0.05 versus N group. B Western blot and statistical analyses of phospho-Akt (Ser 473), phospho-Akt (Thr 308), and Akt in high glucose-cultured HK2 cells treated with DMSO or MA2. n = 3, *means P < 0.05 versus DMSO group. C Western blot and statistical analyses of HDAC5 in high glucose-cultured HK2 cells treated with DMSO or MA2. n = 3, *means P < 0.05 versus DMSO group. D Immunofluorescence of HDAC5 in high glucose-cultured HK2 cells treated with DMSO or MA2. Bar: 25 μm. E Western blot and statistical analyses of TGF-β1 and α-SMA in high glucose-cultured HK2 cells treated with DMSO or MA2. n = 3, * means P < 0.05 versus DMSO group. F Western blot and statistical analyses of phospho-Akt (Ser 473) and phospho-Akt (Thr 308) in HK2 cells. n = 3, *means P < 0.05 versus DMSO group, # means P < 0.05 versus MA2 group. G Western blot and statistical analyses of HDAC5 in HK2 cells. n = 3, *means P < 0.05 versus DMSO group, # means P < 0.05 versus MA2 group. H Immunofluorescence of HDAC5 expression in high glucose-cultured HK2 cells treated with MA2 or MA2 plus insulin. White arrows showed the positive expression of HDAC5. Bar: 25 μm.

METTL14-regulated Akt pathway and HDAC5 expression in high glucose-treated HK2 cells
Considering that m6A RNA modification was regulated by methylase and demethylase, we further explored which enzyme of m6A RNA modification was involved in high glucose-induced Akt pathway activation and HDAC5 upregulation in HK2 cells. Then we first detected FTO, METTL3, and METTL14 mRNA expression by the method of real-time PCR in high glucose-stimulated cells. As illustrated in Fig. 6A, high glucose treatment significantly decreased FTO, METTL3, and METTL14 mRNA expression. In detail, FTO, METTL3, and METTL14 mRNA were, respectively, decreased by 56.68%, 61.71%, and 52.69% with high glucose treatment in HK2 cells compared with normal glucose treatment (P < 0.05). Similarly, western blot results also showed that high glucose reduced FTO, METTL3, and METTL14 at the level of protein in HK2 cells. Statistical analysis revealed that FTO, METTL3, and METTL14 protein were, respectively, decreased by 20.36%, 14.72%, and 12.80% with high glucose treatment versus normal glucose treatment (P < 0.05) (Fig. 6B).
PTEN mediated MA2 and METTL14-inhibited Akt pathway activation in high glucose-cultured HK2 cells Considering that m6A RNA modification affected RNA degradation or protein translation, and not directly affected protein phosphorylation, we speculated that MA2 and METTL14 might regulate Akt signaling pathway by affecting the upstream regulator of Akt. In our previous study, PTEN was revealed to be the crucial regulator of Akt pathway in renal tubular cells of diabetic kidney disease. Here, we first detected the effect of MA2 on PTEN expression in high glucose-cultured HK2 cells and the results showed that MA2 significantly increased PTEN expression at the level of protein. Statistical analysis revealed that PTEN protein was increased by 1.32 times by the treatment of MA2 in HK2 cells (P < 0.05) (Fig. 6F). Similarly, the results of real-time PCR showed that PTEN mRNA level was also increased by 1.46 times in MA2-treated HK2 cells (P < 0.05) (Fig. 6G). Then, in normal glucose-cultured HK2 cells, METTL14 shRNA plasmid transfection caused a 21.81% decrease of PTEN expression, similar to the effect of high glucose on PTEN protein. Accordingly, phospho-Akt (Ser 473) and phospho-Akt (Thr 308) were, respectively, enhanced by 1.23 times and 1.56 times with pGenesil-1-METTL14 transfection versus pGenesil-1 transfection (P < 0.05). Subsequently, HDAC5 protein expression was also increased (Fig. 6H). Furthermore, we overexpressed METTL3 and METTL14 in high glucose-cultured HK2 cells and found that pcDNA3.1-METTL14 not pcDNA3.1-METTL3 significantly increased PTEN expression by 1.38 times versus pcDNA3.1 (P < 0.05). There was no evident difference in PTEN expression between pcDNA3.1-transfected HK2 cells and pcDNA3.1-METTL3-transfected cells (Fig. 6I). Therefore, it was suggested that PTEN upregulation was involved in MA2 and METTL14-caused Akt pathway inactivation and HDAC5 downregulation in HK2 cells.
HDACi TSA treatment decreased HDAC5 expression, EMT, and ECM accumulation in renal tubular cells of diabetic mice Again, we further explored the in vivo effect of HDACs inhibitor on HDAC5 expression, EMT and ECM deposit in the kidneys of diabetic mice. It could be seen in Fig. 7A that HDAC5 was effectively decreased in the renal tubular cells of trichostatin A (TSA)-administrated diabetic mice. In line, overexpression of TGF-β1 in the kidneys of diabetic mice was significantly inhibited with TSA treatment (Fig. 7B). Furthermore, immunohistochemistry results also revealed that EMT marker α-SMA was markedly enhanced in the renal tubular cells of diabetic mice compared with those of normal mice, which was avoided with the administration of TSA (Fig. 7C). Then Masson staining showed that the cumulative ECM in renal interstitium of diabetic mice was effectively prevented in the kidneys of TSA-treated diabetic mice (Fig. 7D). Therefore, these findings suggested that HDACi was the promising drug to ameliorate EMT of renal tubular cells and ECM deposit of diabetic kidney disease.

Discussion
In the present study, we first revealed the increased expression of HDAC5 in renal tubular cells and interstitium of diabetic mice and UUO mice, especially UUO mice. Also, HDAC5 overexpression promoted EMT of renal tubular cells, indicated in E-cadherin downregulation and α-SMA upregulation. Similarly, Choi et al. found that piceatannol suppressed renal fibrosis, shown as decreased ECM protein, reduced connective tissue growth factor (CTGF) and α-SMA in UUO kidneys, accompanied with reduced HDAC4 and HDAC5 protein expression 16 . As well known, EMT was the important characteristic of tumor to regulate invasion and metastasis, and HDAC5 was also reported to mediate EMT of tumor cells. In non-small cell lung cancer, the loss of miR-589-5p led to HDAC5 expression upregulation, regulating cell cycle and EMT-related genes 17 . As well, knockdown of HDAC5 by siRNA technology in glioma cells suppressed the doxorubicin-induced EMT 18 . Recently, Jaquva et al. revealed that HDAC5 overexpression in multiple urothelial carcinoma (UC) cell lines decreased cell proliferation, however in VM-Cub-1 cell line HDAC5 overexpression induced a dramatic EMT change, with the weak effect of HDAC4 on cell phenotypes 19 . Taken together it is suggested that HDAC5 is a promising target to ameliorate diabetic kidney disease via regulating EMT. TGF-β1 is well known as pro-fibrosis factor and previously reported to increase in renal tubular cells of diabetes mellitus 20 . Our findings revealed that TGF-β1 was involved in HDAC5-regulated EMT in renal tubular cells. As so far, there was no report on the direct relationship between HDAC5 and TGF-β1, whereas HDACs inhibitors (HDACi) or other HDACs members were proven to affect TGF-β1 expression in fibrosis models including renal fibrosis. Suberoylanilide hydroxamic acid (SAHA) alleviated live fibrosis by suppressing TGF-β1 pathway in the fibrotic rats 21 . In line, Yang et al. revealed that FK228 known as a selective inhibitor of class I HDACs significantly suppressed renal interstitial fibrosis via Smad and non-Smad pathways in a murine model of UUO 22 . Additionally, the mesothelial cells (MCs) isolated from effluent of peritoneal dialysis (PD) patients were treated with MS-275, a HDAC1-3 inhibitor, and downregulated mesenchymal markers including TGF-β1 to promote EMT reversal. Further genetic silencing experiments confirmed that HDAC1 played a major role in EMT reversal 23 . Combined with these findings, our experiments suggested that TGF-β1 was the important downstream target of HDAC5-regulated EMT in diabetic kidney disease. Here, we also found knockdown of HDAC5 in normal glucose-cultured HK2 cells led to a more significant downregulation of TGF-β1 than high glucosecultured cells. Considering that hyperglycemia had been proven to activate TGF-β1 pathway in renal tubular cells of diabetic kidney disease 24 , we speculated that hyperglycemia environment compromised the inhibitory effect of HDAC5 knockdown on TGF-β1 in HK2 cells.
Cell signaling pathway mediated multiple functions of cells including proliferation, migration, apoptosis, autophagy, and metabolism. PI3K/Akt pathway has been proven to be activated in renal tubular cells of diabetes mellitus 25 . Here, we also revealed that PI3K/Akt pathway regulated HDAC5 expression and inhibition of PI3K/Akt pathway ameliorated high glucose-induced EMT of renal tubular cells by the downregulation of HDAC5. Similar to our findings, in L6 myocytes decreased Akt phosphorylation resulted from low pH (pH 7.0) stimulation caused the reduced nuclear content of HDAC5 26 . Also, Pietruczuk et al. found that two inhibitors of PI3K/ Akt pathway, wortmannin and SC66, evidently attenuated the phosphorylation of HDAC5 and nuclear content in IGF-1-induced vascular smooth muscle cells (VSMCs) 27 . Summarily, PI3K/Akt pathway is the upstream signaling pathway to regulate HDAC5 expression in renal tubular cells of diabetic kidney disease.
Inspired by Liu et al.'s finding that m6A mRNA methylation regulated Akt activity in endometrial cancer cells 12 , we further explored the potential role of m6A mRNA methylation level in high glucose-induced Akt pathway activation and HDAC5 expression in renal tubular cells. Similar to the previous studies that m6A RNA methylation was decreased in peripheral blood RNA from type 2 diabetes mellitus (T2DM) patients compared with control group 28 , we also confirmed high glucose-induced m6A RNA methylation level downregulation in vitro-cultured HK2 cells. Subsequently, m6A demethylase inhibitor MA2 or overexpression of METTL14 inhibited Akt pathway activation followed by HDAC5 downregulation in high glucose-cultured HK2 cells. Fig. 7 The effect of TSA on HDAC5, TGF-β1, α-SMA, and ECM in the kidney of diabetic mice (n = 6). A Immunohistochemistry of HDAC5 in the kidney of normal mice, diabetic mice and diabetic mice treated with TSA. Black arrows showed the positive signal of HDAC5. B Immunohistochemistry of TGF-β1 in the kidney of normal mice, diabetic mice and diabetic mice treated with TSA. Black arrows showed the positive signal of TGF-β1. C Immunohistochemistry of α-SMA in the kidney of normal mice, diabetic mice and diabetic mice treated with TSA. Black arrows showed the positive signal of α-SMA. D Masson staining of the kidney of normal mice, diabetic mice and diabetic mice treated with TSA. Red arrows showed ECM accumulation. E Model of the effect of m6A RNA methylation-regulated PTEN/PI3K/Akt signaling pathway on HDAC5-mediated epithelial-mesenchymal transition in high glucose-treated renal tubular cells.
Recently, the regulation of m6A modification on Akt pathway was also revealed in gastric cancer cells. Zhang et al. confirmed that in gastric cancer cell lines HGC-27 and MGC803, m6A modification suppression via METTL14 knockdown promoted cell proliferation and invasion by activating Wnt and PI3K-Akt signaling. On the contrary, m6A elevation via FTO knockdown reversed these phenotypes 29 . Therefore, high glucose-caused m6A RNA methylation level downregulation activated Akt pathway in renal tubular cells, leading to HDAC5 expression abnormality.
PTEN was known as the negative regulator of Akt pathway and proven to be regulated by MA2 and METTL14 in HK2 cells in this study. PTEN mediated MA2 and METTL14 overexpression-caused Akt signaling pathway inactivation in HK2 cells. Similarly, the relationship between PTEN and m6A modifications was also reported in the human acute myeloid leukemia MOLM-13 cell line. Single-nucleotide-resolution mapping of mA coupled with ribosome profiling revealed that mA modification promoted the PTEN mRNA translation 30 . Also, Wang et al. found that in renal clear cell carcinoma METTL14 mRNA was likely to regulate PTEN mRNA via changing m6A RNA modification level 31 . Therefore, it was suggested that PTEN might be the target of m6A RNA modification in HK2 cells affecting Akt pathway.