ets1 associates with KMT5A to participate in high glucose-mediated EndMT via upregulation of PFN2 expression in diabetic nephropathy

Background Diabetic nephropathy (DN) is currently the leading cause of end-stage renal disease globally. The endothelial-to-mesenchymal transition (EndMT) of glomerular endothelial cells has been reported to play a crucial role in DN. As a specific form of epithelial-to-mesenchymal transition, EndMT and epithelial-to-mesenchymal transition may exhibit mutual modulators. Profilin 2 (PFN2) has been reported to participate in epithelial-to-mesenchymal transition. Moreover, ETS proto-oncogene 1 (ets1) and lysine methyltransferase 5A (KMT5A) have been reported to contribute to high glucose-mediated endothelial injury and epithelial-to-mesenchymal transition. In this study, we hypothesize ets1 associates with KMT5A to modulate PFN2 transcription, thus participating in high glucose-mediated EndMT in glomerular endothelial cells. Methods Immunohistochemistry (IHC) was performed to detect protein levels in the kidney tissues and/or aorta tissues of human subjects and rats. Western blot, qPCR and immunofluorescence were performed using human umbilical vein endothelial cells (HUVECs). Chromatin immunoprecipitation (ChIP) assays and dual luciferase assays were performed to assess transcriptional activity. The difference between the groups was compared by two-tailed unpaired t-tests or one-way ANOVAs. Results Our data indicated that vimentin, αSMA, S100A4 and PFN2 levels were increased, and CD31 levels were reduced in glomerular endothelial cells of DN patients and rats. Our cell experiments showed that high glucose induced EndMT by augmenting PFN2 expression in HUVECs. Moreover, high glucose increased ets1 expression. si-ets1 suppressed high glucose-induced PFN2 levels and EndMT. ets1 overexpression-mediated EndMT was reversed by si-PFN2. Furthermore, ets1 was determined to associate with KMT5A. High glucose attenuated KMT5A levels and histone H4 lysine 20 methylation (H4K20me1), one of the downstream targets of KMT5A. KMT5A upregulation suppressed high glucose-induced PFN2 levels and EndMT. sh-KMT5A-mediated EndMT was counteracted by si-PFN2. Furthermore, H4K20me1 and ets1 occupied the PFN2 promoter region. sh-KMT5A cooperated with ets1 overexpression to activate PFN2 promoter activity. Our in vivo study demonstrated that KMT5A was reduced, while ets1 was augmented, in glomerular endothelial cells of DN patients and rats. Conclusions The present study indicated that ets1 cooperated with KMT5A to transcribe PFN2, thus contributing to hyperglycemia-induced EndMT in the glomerular endothelial cells of DN patients and rats. Trial registration ChiCTR, ChiCTR2000029425. 2020/1/31, http://www.chictr.org.cn/showproj.aspx?proj=48548 Supplementary Information The online version contains supplementary material available at 10.1186/s10020-021-00339-7.


Background
Diabetic nephropathy (DN) has become one of the primary causes of mortality in diabetic patients and may result in end-stage renal disease (Packham et al. 2012;Tomino and Gohda 2015). DN is characterized by a gradual elevation of urine albumin, a rise in blood pressure, and a decline of the glomerular filtration rate (Boer et al. 2011;Ng et al. 2016). However, current treatment strategies can only delay and cannot prevent DN progression. Once DN develops to end-stage renal disease, the cost of treatment and cardiovascular-related mortality both increase (Alves and Lewis 2010;Xue et al. 2017). Therefore, the search for the potential mechanisms underlying DN is urgent. The endothelial-to-mesenchymal transition (EndMT) of glomerular endothelial cells has been reported to play a crucial role in the occurrence and progression of DN (Li et al. 2010;Kanasaki et al. 2014).
EndMT is an intricate cellular differentiation process whereby endothelial cells detach and migrate away from the endothelium, lose endothelial properties and acquire mesenchymal features. EndMT is defined as endothelial cell loss of the endothelial phenotype, including CD31, and acquisition of mesenchymal features, including vimentin (Liang et al. 2016). As EndMT is a specific form of epithelial-to-mesenchymal transition, it is possible that epithelial-to-mesenchymal transition modulators also play an important role in the modulation of EndMT (Saito 2013).
Profilins are well known as actin-binding proteins (Ding et al. 2012). Previous studies have indicated that profilin2 (PFN2) participates in epithelial-to-mesenchymal transition in cancer cells (Tang et al. 2015;Zhang et al. 2018). However, whether PFN2 participates in hyperglycemia-mediated EndMT in endothelial cells has not been reported. ets1, a member of the ETS family of transcription factors, plays roles in kidney development and glomerular integrity (Razzaque et al. 2001). Moreover, ets1 serves as a crucial transcription factor in the progression of DN (Geng et al. 2019;Liu et al. 2011). ets1 suppression was reported to attenuate angiotensin II-induced cardiac fibrosis via the inhibition of EndMT (Xu et al. 2019). EST1 may be a potential drug target for the treatment of diabetic nephropathy.
Lysine methyltransferase 5A (KMT5A) is the only known nucleosome-specific methyltransferase that modulates histone H4 lysine 20 through methylation (H4K20me1) (Beck et al. 2012). Our previous studies have shown that KMT5A downregulation participates in high glucose-mediated vascular endothelial injury (Chen et al. 2018(Chen et al. , 2020aQi et al. 2020;Wang et al. 2020;Shen et al. 2020). Moreover, KMT5A was reported to mediate epithelial-to-mesenchymal transition (Hou et al. 2016). However, whether KMT5A is involved in high glucoseinduced EndMT is still not well known. In the present study, we hypothesize that ets1 may associate with KMT5A to modulate PFN2 expression, thus performing a vital role in high glucose-mediated EndMT in glomerular endothelium.

Subjects
Twenty diagnosed DN patients (type II diabetes) and twenty control participants (renal cancer patients with normal renal function) were consecutively recruited. This research abided by the Declaration of Helsinki and was authorized by the Ethics Committee of Huzhou Central Hospital. The license number of the present study is 20191209-01. Written informed consent was acquired from all the participants recruited. The Chronic Kidney Disease Epidemiology Collaboration equation (CKD-EPI) was used to calculate the glomerular filtration of human subjects. DN patients' staging protocol was defined according to the National Kidney Foundation's Kidney Disease Outcomes Quality Initiative classification scheme.

Rat model of DN
Male Sprague-Dawley rats weighing 180-200 g were selected for the present study. The animals were purchased from Shanghai SLAC Laboratories. This study abided by the Guide for the Care and Use of Laboratory Animals of Shanghai Medical College at Fudan University. Animals were housed in the care facilities as described previously Shen et al. 2020). Twenty rats were randomly assigned to the control group (con, n = 10) and the DN group (DN, n = 10). In the control group, the animals received a single intraperitoneal Table 1 Characteristics of participants and rats in control (con) and Diabetic Nephropathy (DN) group Data are presented as means ± SD. BMI (Body Mass Index), SBP (systolic blood pressure), DBP (diastolic blood pressure), HbA1c (glycated hemoglobin), FBG (fasting blood glucose), CREA (creatinine), ALB (albumin), CCr injection of citrate buffer (0.1 M, pH 4.5). In the DN group, after being raised with a high sugar-fat diet for 2 weeks, the animals were intraperitoneally injected once with streptozotocin (STZ, 50 mg/kg). Then, the animals were sent back to the care facilities for 6 weeks. Three days later, hyperglycemia in the DN group animals was confirmed by monitoring tail vein blood samples.

Collection of rat blood samples and urine specimens
One day before euthanasia, 24-h urine was collected, and albumin levels were detected after urine collection. Euthanasia of all the animals was carried out by intraperitoneal injection of thiopental sodium (40 mg/kg). Animal blood samples were gathered by cardiac puncture. After collection, the plasma and urine specimens were preserved at −80 °C until detection. Serum creatinine levels were detected with the use of a creatinine assay kit (Jiancheng Bio, Nanjing, China). Serum urea nitrogen (BUN) levels were detected with the use of a BUN assay kit (Jiancheng Bio, Nanjing, China). Twenty-four-hour urinary albumin levels were detected with the use of an enhanced BCA protein assay kit (Beyotime, Shanghai, China).

Hematoxylin-eosin (HE) staining, Masson staining and immunohistochemistry (IHC)
The tissues of human subjects and rats were wrapped in paraffin and then processed for HE, Masson staining and IHC. HE was used to detect histopathological abnormalities. Masson staining was used to determine the severity of fibrosis according to the kit instructions (Solarbrio, Beijing, China

RNA extraction and quantitative real-time PCR (qPCR)
Cell and tissue RNA was extracted by using TRIzol reagent (Invitrogen, Grand Island, NY, USA). Complementary DNA (cDNA) was synthesized with the use of PrimeScript RT reagent (TaKaRa). Then, we performed qPCR via Hieff UNICON ® qPCR TaqMan Probe Master Mix (Yeasen, Shanghai, China) on an ABI7500 Real-Time PCR system (Applied Biosystems). The primers employed are listed in Additional file 1: Table S1.

Bioinformatics analysis
In the present study, inbio-discover website was used to explore the potential proteins that interact with ets1. Then, these proteins were retrieved and loaded into R.studio for gene ontology (GO) mapping and annotation (p < 0.01 was the cut-off criterion).
For endogenous IP, cell protein extracts were incubated with specific primary antibodies and 50 μl protein A/G Dynabeads (Thermo Fisher, USA) at 4 °C for 12 h. Then, 8 μl of input, IgG and IP extracts were processed for Western blotting.

Immunofluorescence (IF) staining
HUVECs were plated onto glass slides with the corresponding treatment. After fixation with 4% paraformaldehyde, HUVECs were permeabilized with 0.3% Triton X-100 for 5 min. Then, the cells were blocked with 1% bovine serum albumin at room temperature for 1 h. Then, the cells were incubated with specific primary antibodies at 4 °C for 12 h. 4,6-Diamidino-2-phenylindole (DAPI) was employed to stain the nucleus.

Chromatin immunoprecipitation (ChIP) assay
ChIP assays were performed with the use of a Simple ChIP Plus Sonication Chromatin IP Kit (Cell Signaling Technology, MA) as described previously Shen et al. 2020). In brief, HUVECs (1 × 10 7 ) were fixed with 1% formaldehyde at room temperature for 10 min. Then, the reaction process was terminated by glycine. Chromatin was sheared with the use of a Microson XL ultrasonic cell disruptor XL (Misonix). Then, 10 μl of ultrasonic solution was employed as input. The rest of the ultrasonic solution was cultured with specific antibodies against ets1 (Proteintech, Wuhan, China) and H4K20me1 (Abcam, Cambridge, UK), as well as a negative control IgG at 4 °C for 12 h. After protein G magnetic beads absorbed the immunoprecipitants, the DNA-protein interaction was terminated by incubation at 65 °C for 2 h. After purification, the enriched DNA sequences were determined by PCR. The PFN2 oligonucleotide primers were as follows: forward, 5′-TGG GGG ACC CCT AAT TCC AT-3′, and reverse, 5'-GGA CCC AGG TCT GGG AAA GA-3′.

Dual-luciferase assay
The activity of the PFN2 promoter was assessed by a Promega Dual-luciferase Assay Kit (Madison, WI, USA) as described previously ). The PFN2 promoter was connected to the pGL3-basic vector, thus establishing the pGL3-PFN2 construct. HUVECs were transfected with the pGL3-PFN2 plasmid. The promoter activity of PFN2 was evaluated with the corresponding treatments.

Statistical analysis
In the present study, the sample sizes were calculated by evaluation of the effect of hyperglycemia or high glucose on PFN2 mRNA expression, which was detected in our pilot in vivo or in vitro experiments. Statistical significance may be acquired with a sample size of 10 in in vivo experiments and 5 in in vitro experiments. The results are expressed as the mean ± standard deviation. The difference between the groups was compared by two-tailed unpaired t-tests or one-way ANOVAs with GraphPad Prism Version 7.0 (GraphPad Software, San Diego, CA). P < 0.05 was considered significant.

Fig. 2
High-glucose treatment mediated EndMT via the upregulation of PFN2 expression in HUVECs. A Results from the Western blot analysis of CD31, vimentin, αSMA, S100A4 and PFN2 levels in HUVECs with the corresponding treatment. B Compared with the control group, the mRNA expression of CD31 was decreased in hyperglycemic HUVECs. C Compared with the control group, the mRNA expression of vimentin was increased in hyperglycemic HUVECs. D Compared with the control group, the mRNA expression of αSMA was increased in hyperglycemic HUVECs. E Compared with the control group, the mRNA expression of S100A4 was increased in hyperglycemic HUVECs. F Compared with the control group, the mRNA expression of PFN2 was increased in hyperglycemic HUVECs. G Results from the Western blot analysis of PFN2, CD31, vimentin, αSMA and S100A4 levels in HUVECs with the corresponding treatment. H The effects of si-PFN2 were confirmed by qPCR. I Compared with high-glucose treatment, si-PFN2 increased CD31 mRNA expression in hyperglycemic HUVECs. J Compared with high-glucose treatment, si-PFN2 decreased vimentin mRNA expression in hyperglycemic HUVECs. K Compared with high-glucose treatment, si-PFN2 decreased αSMA mRNA expression in hyperglycemic HUVECs. L Compared with high-glucose treatment, si-PFN2 decreased S100A4 mRNA expression in hyperglycemic HUVECs. (* p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001, n = 5/group)

Occurrence of EndMT and increase in PFN2 expression in DN patients and rats
The characteristics of the control and DN participants are shown in Table 1. The DN patients recruited in the present study were in DN stage 2 (DN2), DN stage 3 (DN3) or DN stage 4 (DN4) according to the eGFR. HE staining of the participants is shown in Fig. 1A. Masson staining resulted in more blue staining in the renal biopsy specimens of DN patients, which represented collagen deposition and extensive interstitial fibrosis (Fig. 1A). EndMT in glomerular endothelial cells has been reported to play a crucial role in DN and fibrosis (Li et al. 2010;Kanasaki et al. 2014). Therefore, we detected CD31, vimentin, αSMA and S100A4 in renal biopsy specimens of DN patients and control participants. Compared with control participants, CD31 expression was decreased, while vimentin, αSMA and S100A4 expression was increased in glomerular endothelial cells of DN patients (Fig. 1B). More importantly, our data indicated that vimentin, αSMA, and S100A4 gradually increased, while CD31 gradually decreased in glomeruli when the DN stage progressed (Fig. 1B). PFN2 was reported to be involved in epithelial-to-mesenchymal transition (Tang et al. 2015;Zhang et al. 2018). Next, we detected PFN2 expression and found that PFN2 expression in glomerular endothelial cells was higher in DN patients than in control participants (Fig. 1B). Moreover, PFN2 was found to gradually increase in glomeruli when the DN stage progressed (Fig. 1B). These data indicated that as DN progresses, so does vascular injury.
The characteristics of the control and DN rats are listed in Table 1. Accordingly, HE staining and Masson trichrome staining indicated renal injury and interstitial fibrosis in the glomerular endothelial cells of DN rats (Additional file 2: Figure S1A). Moreover, the levels of PFN2, vimentin, αSMA and S100A4 in the kidney tissues (Additional file 2: Figure S1B) of DN rats increased compared with the levels in the control group. The levels of CD31 in the kidney tissues (Additional file 2: Figure S1B) of DN rats were lower than the levels of CD31 in the kidney tissues of the control group. Furthermore, the protein and mRNA levels of PFN2, vimentin, αSMA and S100A4 in aortic tissues (Additional file 3: Figure S2A, C-F) of DN rats were higher than the protein and mRNA levels of PFN2, vimentin, αSMA and S100A4 in aortic tissues of the control group. The protein and mRNA levels of CD31 in aortic tissues (Additional file 3: Figure S2A, B) of DN rats were lower than the protein and mRNA levels of CD31 in aortic tissues of the control group. The blood glucose evolution of rats in this study is provided in Additional file 3: Figure S2G. These data may indicate that EndMT and PFN2 participate in the occurrence and progression of DN.

High-glucose treatment induced EndMT via an increase in PFN2 levels in HUVECs
To explore whether PFN2 modulated EndMT in the glomerular endothelial cells of DN patients and rats, HUVECs were employed in the present study. To determine whether a high glucose level participated in the induction of EndMT and PFN2 expression in HUVECs, cells were cultured in DMEM with 5 mM glucose (con, 5 mM, 6 days) or DMEM with 25 mM glucose (HG, 25 mM, 6 days). Our data showed that high-glucose treatment inhibited CD31 expression while increasing vimentin, αSMA and S100A4 expression in HUVECs ( Fig. 2A-E). These data indicated that high-glucose treatment mediated EndMT in HUVECs. Mannitol treatment did not affect CD31, vimentin, αSMA or S100A4 expression ( Fig. 2A-E). Previous research indicated that PFN2 participates in epithelialto-mesenchymal transition (Tang et al. 2015;Zhang et al. 2018). Therefore, PFN2 expression was detected in HUVECs. The results indicated that high-glucose treatment augmented PFN2 protein ( Fig. 2A) and mRNA expression (Fig. 2F) in HUVECs. To further illustrate whether PFN2 plays an important role in (See figure on next page.) Fig. 3 ets1 participated in high glucose-induced EndMT by augmenting PFN2 expression in HUVECs. A Results from the Western blot analysis of ets1, PFN2, CD31, vimentin, αSMA and S100A4 in the HUVECs with the corresponding treatment. B The effects of si-ets1 were confirmed by qPCR. C Compared with high-glucose treatment, si-ets1 decreased PFN2 mRNA expression in hyperglycemic HUVECs. D Compared with high-glucose treatment, si-ets1 increased CD31 mRNA expression in hyperglycemic HUVECs. E Compared with high-glucose treatment, si-ets1 decreased vimentin mRNA expression in hyperglycemic HUVECs. F Compared with high-glucose treatment, si-ets1 decreased αSMA mRNA expression in hyperglycemic HUVECs. G Compared with high-glucose treatment, si-ets1 decreased S100A4 mRNA expression in hyperglycemic HUVECs. H Results from the Western blot analysis of ets1, PFN2, CD31, vimentin, αSMA and S100A4 in the HUVECs with the corresponding treatment. I The effects of ets1 overexpression were confirmed by qPCR. J The effects of si-PFN2 were confirmed by qPCR. K ets1 overexpression inhibited CD31 mRNA expression, which was reversed by si-PFN2 treatment. L ets1 overexpression increased vimentin mRNA expression, which was reversed by si-PFN2 treatment. M ets1 overexpression increased αSMA mRNA expression, which was reversed by si-PFN2 treatment. N ets1 overexpression increased S100A4 mRNA expression, which was reversed by si-PFN2 treatment. (* p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001, n = 5/group) Lu et al. Mol Med (2021)  high glucose-mediated EndMT, two independent siR-NAs against PFN2 were used in the present study. The effects of si-PFN2 were verified by Western blotting (Fig. 2G) and qPCR (Fig. 2H). Our data demonstrated that si-PFN2 counteracted the high glucose-mediated decrease in CD31 expression and the increase in vimentin, αSMA and S100A4 expression (Fig. 2G, I-L). These data indicated that PFN2 positively modulated EndMT in hyperglycemic HUVECs.

ets1 participated in high glucose-induced EndMT by augmenting PFN2 expression in HUVECs
Studies have indicated that ets1 plays a crucial role in the progression of DN (Geng et al. 2019;Liu et al. 2011). Accordingly, high glucose upregulated ets1 protein and mRNA levels in HUVECs (Additional file 4: Figure S3A, B).
To further confirm the effect of ets1 on high glucose-mediated PFN2 expression and EndMT, both loss-of-function and gain-of-function approaches were used in the present study. The effect of si-ets1 was confirmed in this study (Fig. 3A, B). si-ets1 reversed high glucose-induced PFN2 expression (Fig. 3A, C). Moreover, si-ets1 counteracted the high glucose-mediated reduction in CD31 (Fig. 3A, D) and augmented vimentin, αSMA and S100A4 (Fig. 3A, E-G) levels in HUVECs. Moreover, the effect of ets1 overexpression was similar to the effect of high-glucose treatment ( Fig. 3H-N). To determine whether the effects of ets1 overexpression were acquired by augmenting PFN2 levels, we silenced PFN2 in ets1-upregulated HUVECs. Our data found that PFN2 silencing counteracted ets1 overexpression-mediated reduction in CD31 (Fig. 3H, K) and augmented vimentin, αSMA and S100A4 (Fig. 3H, L-N) expression in HUVECs. These data demonstrated that upregulated ets1 augmented PFN2 levels in hyperglycemic HUVECs, thus participating in high glucose-mediated EndMT in HUVECs.

ets1 associated with KMT5A
To uncover the potential regulatory mechanism by which ets1 regulates PFN2 levels as well as EndMT in hyperglycemic HUVECs, bioinformatics was employed to predict the proteins that associate with ets1 in the present study. Many proteins that associated with ets1 are shown in Table 2, including KMT5A (https:// inbio-disco ver. intom ics. com/ map. html# search). The results of GO pathway enrichment analysis showed that the top ten enriched GO terms in biological processes (BP), molecular functions (MF) and cellular components (CC) were arranged in Fig. 4A,    including histone methylation which was highlighted by the red boxes. Our previous studies demonstrated that KMT5A downregulation participates in high glucosemediated cell injury in HUVECs (Chen et al. 2018(Chen et al. , 2020aQi et al. 2020;Wang et al. 2020;Shen et al. 2020). Moreover, KMT5A also mediates the occurrence of epithelial-tomesenchymal transition (Hou et al. 2016). The association between ets1 and KMT5A in HUVECs was confirmed by CoIP experiments (Fig. 4B). IF experiments revealed that ets1 and KMT5A were co-positioned in the nucleus of hyperglycemic HUVECs (Fig. 4C). Furthermore, our data indicated that high-glucose treatment inhibited KMT5A levels (Fig. 4D, E) in HUVECs. Accordingly, the expression of H4K20me1, a downstream target of KMT5A, was inhibited in the HG group (Fig. 4D).

KMT5A suppression participated in high glucose-mediated EndMT by augmenting PFN2 levels in HUVECs
To determine whether KMT5A modulates PFN2 levels and EndMT in hyperglycemic HUVECs, both loss-offunction and gain-of-function approaches were employed in the present study. The effect of KMT5A overexpression was verified in the present study (Fig. 5A, B). The results demonstrated that KMT5A overexpression reversed high glucose-induced PFN2 (Fig. 5A, C). Moreover, KMT5A overexpression counteracted the high glucose-mediated reduction in CD31 (Fig. 5A, D) and augmented vimentin, αSMA and S100A4 (Fig. 3A, E-G) levels in HUVECs. The effect of sh-KMT5A was verified in the present study (Additional file 5: Figure S4A, B). The effect of sh-KMT5A was similar to the effect of high-glucose treatment (Additional file 5: Figure S4A, C-G). To determine whether the effects of sh-KMT5A were acquired via the augmentation of PFN2 expression, PFN2 was silenced in HUVECs in which KMT5A was downregulated. Our data indicated that PFN2 silencing counteracted the KMT5A downregulation-mediated EndMT in HUVECs (Fig. 5H-N). These data indicated that KMT5A downregulation augmented PFN2 levels, thus participating in high glucose-mediated EndMT in HUVECs.

ets1 cooperated with KMT5A to regulate PFN2 transcriptional activity in HUVECs
To explore whether PFN2 is directly transcribed by ets1 and KMT5A, the genome-wide distribution of ets1 and H4K20me1 was detected in the present study. ChIP assays demonstrated that ets1 and H4K20me1 both occupied the PFN2 promoter region (Fig. 6A). The potential ets1 binding site is exhibited in Fig. 6B. Moreover, KMT5A overexpression and si-ets1 both decreased PFN2 promoter activity (Fig. 6C). sh-KMT5A and ets1 overexpression both augmented PFN2 promoter activity (Fig. 6C). Moreover, sh-KMT5A acted in concert with ets1 overexpression to augment PFN2 promoter activity in HUVECs (Fig. 6C). However, mutant KMT5A R259G had no effect on PFN2 promoter activity (Fig. 6C). These data indicated that ets1 cooperated with KMT5A to modulate PFN2 promoter activity in hyperglycemic HUVECs. Furthermore, KMT5A overexpression inhibited PFN2 protein and mRNA levels, while mutant KMT5A R259G did not affect PFN2 levels ( Fig. 6D-F). Our data indicated that KMT5A-mediated H4K20me1 was necessary to modulate the PFN2 transcript in HUVECs. Furthermore, ets1 overexpression decreased KMT5A expression (Fig. 6G-I). Accordingly, sh-KMT5A augmented ets1 expression in HUVECs (Fig. 6J-L). Our study showed that ets1 and KMT5A inhibited each other in HUVECs.

High glucose-mediated increases in ets1 and decreases in KMT5A were verified in DN participants and rats
To explore whether the protein and/or mRNA expression of ets1 and KMT5A in DN participants and rats was consistent with the protein and/or mRNA expression of ets1 and KMT5A of the in vitro study, ets1 and KMT5A levels were assessed in kidney tissues and/or aortic tissues of DN patients and rats. The results indicated that ets1 increased and KMT5A decreased in glomerular endothelial cells of DN patients (Fig. 7A) and rats (Additional file 6: Figure S5A). More importantly, our data indicated that ets1 levels gradually increased, while KMT5A gradually decreased in glomeruli of DN patients when the (See figure on next page.) Fig. 5 KMT5A suppression participated in high glucose-mediated EndMT by augmenting PFN2 expression in HUVECs. A Results from the Western blot analysis of KMT5A, PFN2, CD31, vimentin, αSMA and S100A4 in the HUVECs with the corresponding treatment. B The effects of KMT5A overexpression were verified by qPCR. C The high glucose-mediated increase in PFN2 mRNA expression was inhibited by KMT5A overexpression. D The high glucose-mediated decrease in CD31 mRNA expression was counteracted by KMT5A overexpression. E The high glucose-mediated increase in vimentin mRNA expression was inhibited by KMT5A overexpression. F The high glucose-mediated increase in αSMA mRNA expression was inhibited by KMT5A overexpression. G The high glucose-mediated increase in S100A4 mRNA expression was inhibited by KMT5A overexpression. H Results from the Western blot analysis of KMT5A, PFN2, CD31, vimentin, αSMA and S100A4 in the HUVECs with the corresponding treatment. I si-PFN2 did not affect KMT5A mRNA expression. J The effect of si-PFN2 was verified by qPCR. K si-PFN2 upregulated CD31 mRNA expression in sh-KMT5A-treated HUVECs. L si-PFN2 decreased vimentin mRNA expression in sh-KMT5A-treated HUVECs. M si-PFN2 decreased αSMA mRNA expression in sh-KMT5A-treated HUVECs. N si-PFN2 decreased S100A4 mRNA expression in sh-KMT5A-treated HUVECs. (* p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001, n = 5/group) Lu et al. Mol Med (2021)  DN stage progressed (Fig. 7A). Accordingly, the expression of ets1 increased, while the expression of KMT5A decreased in aortic tissues of DN rats (Additional file 6: Figure S5B-D). In conclusion, our study showed that ets1 and KMT5A cooperated to regulate PFN2 transcription, thus mediating EndMT in DN patients and rats (Fig. 7B).

Discussion
The main finding of this study was that high glucose participated in EndMT via the augmentation of PFN2 transcription, thus mediating endothelial cell injury and the occurrence of DN. Moreover, high-glucose treatment decreased KMT5A levels and augmented ets1 levels. Furthermore, ets1 and H4K20me1 both occupied the PFN2 promoter region. Mechanistic studies indicated that KMT5A associates with ets1 to regulate PFN2 transcriptional activity, thus mediating EndMT in hyperglycemic HUVECs. Epithelial-to-mesenchymal transition is a complicated form of cell phenotype reprogramming and plays a crucial role in organ injury (Duan et al. 2012). The epithelial-to-mesenchymal transition of renal tubular epithelial cells plays an important role in renal interstitial myofibroblasts, thus participating in the progression of renal fibrosis (Stasi et al. 2017). Recently, EndMT of glomerular endothelial cells was proven to participate in the occurrence and progression of DN (Li et al. 2010;Kanasaki et al. 2014). In the present study, the results indicated that CD31 was reduced, while vimentin, αSMA and S100A4 were augmented in the glomerular endothelial cells of DN participants and rats (Fig. 1, Additional file 2: Figure S1). Moreover, HE staining and Masson trichrome staining indicated renal injury and interstitial fibrosis in the glomeruli of DN participants and rats (Fig. 1, Additional file 2: Figure S1). These data were in agreement with a recent study that indicated EndMT in glomeruli aggravates DN (Peng et al. 2016). To determine whether EndMT in DN patients and rats was caused by hyperglycemia, we used high glucose-treated HUVECs as an in vitro model. The protein and mRNA levels of CD31, vimentin, αSMA and S100A4 confirmed that high glucose participated in the occurrence of EndMT, which was similar to one previous study (Yu et al. 2017). We deduced that epithelial-to-mesenchymal transition and EndMT may have shared modulators (Saito 2013). PFN2 has been reported to participate in epithelial-tomesenchymal transition (Tang et al. 2015;Zhang et al. 2018). In the present study, PFN2 levels were found to be increased in DN patients and rats (Fig. 1). Moreover, high-glucose treatment induced PFN2 expression ( Fig. 2A, F) and EndMT ( Fig. 2A-E) in HUVECs. The inhibition of PFN2 expression reversed high glucosemediated EndMT (Fig. 2G-L). Our data indicated that PFN2 was involved in high glucose-mediated EndMT in HUVECs. In diabetic nephropathy or high-glucose conditions, several other types of cells undergo damage, including podocytes and tubular epithelial cells. Previous studies have demonstrated that crosstalk between tubular epithelial cells and glomerular endothelial cells plays an important role in diabetic nephropathy . Exosomes from high glucose-treated glomerular endothelial cells have been reported to trigger epithelial-mesenchymal transition and dysfunction in podocytes (Wu et al. 2017). These studies indicated that endothelial cells, podocytes, and tubular epithelial cells interact with each other, and all participate in the progression of diabetic nephropathy. ets1, a member of the ETS family of transcription factors, plays a crucial role in kidney development and glomerular integrity (Razzaque et al. 2001). The expression of ets1 has been reported to increase in both glomeruli and interstitium during the progression of rat crescentic glomerulonephritis (Naito et al. 2000). Moreover, ets1 serves as a vital transcription factor in the progression of DN (Geng et al. 2019). ets1 upregulation was reported to participate in glomerulosclerosis and renal interstitial fibrosis in experimental DN (Liu et al. 2011). Similarly, ets1 suppression protects against angiotensin II-mediated cardiac fibrosis via inhibition of EndMT (Xu et al. 2019). The mechanism by which ets1 modulates EndMT involves the transactivation of Twist1 promoter activity (Millien et al. 2018) and the activation of the ZEB1 promoter (Sinh et al. 2017;Dave et al. 2011). The present study showed that high glucose augmented ets1 expression in HUVECs (Additional file 4: Figure S3A, B). Moreover, si-ets1 reversed high glucosemediated PFN2 upregulation and EndMT (Fig. 3A-G). Furthermore, ets1 overexpression increased PFN2 expression and mediated EndMT, which was counteracted by si-PFN2 (Fig. 3H-N). These data indicated that ets1 was involved in high glucose-mediated EndMT in hyperglycemic HUVECs via the augmentation of PFN2 expression.
Our previous studies showed that KMT5A participates in high glucose-mediated overexpression of endothelial adhesion molecule expression (Chen et al. 2018;Shen et al. 2020), proinflammatory enzyme and proinflammatory cytokine production (Qi et al. 2020), antioxidant imbalance ) and NOD-like receptor pyrin domain 3 inflammasome activation  in endothelial cells, thus mediating vascular endothelial injury (Chen et al. 2018(Chen et al. , 2020aQi et al. 2020;Wang et al. 2020;Shen et al. 2020). However, the Fig. 7 High glucose-mediated increases in ets1 and decreases in KMT5A were verified in DN patients and rats. A Immunostaining of ets1 and KMT5A in renal biopsy specimens of DN patients and control participants. Magnification: ×20. Scale bar: 20 μM. (* p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001, n = 20 for the control group, n = 7 for the DN2 group, n = 7 for the DN3 group, n = 6 for the DN4 group). B Schematic representation of the working model regulation of EndMT by KMT5A has not been reported. Indeed, KMT5A has been reported to be involved in epithelial-to-mesenchymal transition via the modulation of TWIST (Yang et al. 2012) and cooperation with zinc finger E-box-binding homeobox 1 (Hou et al. 2016). As a specific form of epithelial-to-mesenchymal transition, it is possible that EndMT and epithelial-to-mesenchymal transition share mutual modulators (Saito 2013). In the present study, KMT5A overexpression counteracted high glucose-mediated PFN2 upregulation (Fig. 5A, C) and EndMT (Fig. 3A, D-G) in HUVECs. Moreover, H4K20me1, a direct target of KMT5A, occupied the PFN2 promoter region (Fig. 6A). Furthermore, si-PFN2 counteracted sh-KMT5A-mediated EndMT (Fig. 5H-N). These data indicated that high glucose-mediated KMT5A downregulation participated in EndMT in hyperglycemic HUVECs by augmenting PFN2 expression. KMT5A is the only known H4K20me1 regulator, but H4K20me1 is not the only KMT5A target (Beck et al. 2012). In our further research, one way to assess the EndMT-specific effects of KMT5A downregulation in hyperglycemic HUVECs would be to employ gene expression microarrays and explore the expression status of EndMT-specific genes.
The transcriptional activity of ets1 has been reported to be regulated by epigenetic modifications Sobral et al. 2020). The present study represented an association between KMT5A and ets1 (Fig. 4B, C). Moreover, ets1 and H4K20me1 both occupied the PFN2 promoter region (Fig. 6A). Furthermore, sh-KMT5A augmented the promoting action of ets1 on PFN2 promoter activity (Fig. 6C). The results showed that KMT5A cooperated with ets1 to adjust PFN2 transcription, thus participating in EndMT in high glucose-treated HUVECs (Fig. 7C). Furthermore, KMT5A upregulation reduced PFN2 levels (Fig. 6D, F). However, KMT5A R259G (one KMT5A mutant) had no effect on PFN2 transcription (Fig. 6D, F). Our study showed that KMT5A-induced H4K20me1 participated in the regulation of PFN2 transcription.
This study has some limitations. First, whether KMT5A associates with ets1 in a direct or indirect manner is worthy of further study. Second, HUVECs were used to establish an in vitro model, and other primary endothelial cells should be used to verify our results in the present study. Third, the mechanism of mutual suppression between KMT5A and ets1 deserves further research. Fourth, the potential mechanism by which PFN2 enhances EndMT in hyperglycemic HUVECs was not well investigated in this study. Previous studies have indicated that vasodilator-stimulated phosphoprotein (VASP) (Reinhard et al. 1995) serves as a partner of PFN2, and VASP has been reported to participate in epithelial-mesenchymal transition . Whether PFN2 participates in EndMT via association with VASP or other partners deserves further research. Fifth, the number of cases in each DN group is limited, so the present study needs more cases to be further confirmed.

Conclusions
Our data indicated that KMT5A levels were reduced, ets1 and PFN2 levels were upregulated, and EndMT was induced in DN patients and rats. The present study also demonstrated that high glucose induced EndMT by augmenting PFN2 levels in hyperglycemic HUVECs. Moreover, high glucose levels suppressed KMT5A levels and increased ets1 levels in vascular endothelial cells. Furthermore, KMT5A cooperated with ets1 to modulate PFN2 levels, thus participating in high glucose-mediated EndMT in hyperglycemic HUVECs.