PPARα activation directly upregulates thrombomodulin in the diabetic retina

Two large clinical studies showed that fenofibrate, a commonly used peroxisome proliferator-activated receptor α (PPARα) agonist, has protective effects against diabetic retinopathy. However, the underlying mechanism has not been clarified. We performed genome-wide analyses of gene expression and PPARα binding sites in vascular endothelial cells treated with the selective PPARα modulator pemafibrate and identified 221 target genes of PPARα including THBD, which encodes thrombomodulin (TM). ChIP-qPCR and luciferase reporter analyses showed that PPARα directly regulated THBD expression via binding to the promoter. In the rat diabetic retina, treatment with pemafibrate inhibited the expression of inflammatory molecules such as VCAM-1 and MCP1, and these effects were attenuated by intravitreal injection of small interfering RNA targeted to THBD. Furthermore, pemafibrate treatment inhibited diabetes-induced vascular leukostasis and leakage through the upregulation of THBD. Our results indicate that PPARα activation inhibits inflammatory and vasopermeable responses in the diabetic retina through the upregulation of TM.

www.nature.com/scientificreports/ neovascularization, vasoregression, and vascular hyperpermeability 9,[13][14][15] . Another recent study has also found that important factors in DR such as VEGF and TNFα are downregulated by PPARα activation in the retina of diabetic animal models 16 . However, the mechanisms by which PPARα activation exerts protective effects against DR are not fully understood. Fenofibrate, which has a long history of clinical use as a lipid-lowering drug, has poor PPAR subtype selectivity 17 . Fenofibrate treatment sometimes results in elevation of the transaminase, homocysteine, and creatine levels in patients. On the other hand, pemafibrate, a novel selective PPARα modulator, has greater PPARα activation potency and higher subtype selectivity than fenofibrate [18][19][20] . Therefore, pemafibrate may reduce inflammation and angiogenesis more effectively than fenofibrate in DR patients.
Thrombomodulin (TM) is a transmembrane protein expressed on the surface of endothelial cells and is encoded by the THBD gene 21 . TM converts thrombin to the anticoagulant form to reduce blood coagulation and inhibits inflammation in blood vessels [22][23][24][25][26] . Recently, recombinant TM has been developed and can potentially be used to treat patients with inflammatory and thrombotic diseases 27,28 .
In the current study, we performed genome-wide analyses of gene expression and PPARα binding sites in vascular endothelial cells treated with pemafibrate and found that PPARα directly regulates the expression of THBD in endothelial cells. Furthermore, PPARα activation inhibited retinal inflammation through the upregulation of TM in a rat model of DR. Thus, upregulation of TM by PPARα activation can be a potential therapeutic strategy against DR.

Results
Genome-wide analysis of PPARα-targeted genes in vascular endothelial cells. To investigate the mechanism of the protective effects against DR by PPARα activation, we performed DNA microarray analysis and ChIP-seq of PPARα in HUVECs treated with pemafibrate. Microarray analysis showed that pemafibrate treatment for 24 h upregulated 1,062 genes (> 1.5-fold, Fig. 1a, pemafibrate-induced genes) and downregulated 477 genes (> 1.5-fold) compared with DMSO treatment (control). Pemafibrate-induced genes included known direct targets of PPARα such as PDK4 18 . The top 50 upregulated and downregulated genes are listed in Tables 1 and 2, respectively.
We also performed ChIP-seq of retinoid X receptor α (RXRα), which is a heterodimer partner of PPARα, and found RXRα binding near the TSS of THBD regardless of pemafibrate treatment (Fig. 1c). Because TM is reported to inhibit inflammation in blood vessels 23,26 , we hypothesize that the upregulation of TM by PPARα activation could inhibit the inflammatory response in the diabetic retina.
PPARα directly upregulates THBD expression. A combination of DNA microarray and ChIP-seq analyses of HUVECs identified THBD as one of the target genes of PPARα. To confirm this, we performed immunoblot analysis and showed that pemafibrate treatment upregulated TM protein expression in HUVECs as well as HRMECs (Fig. 2a,b). Q-PCR analysis revealed that the upregulation of THBD by pemafibrate was blunted when PPARα was knocked down by small interfering RNA (siRNA) targeted to PPARα in HUVECs and HRMECs (Fig. 2c,d). Thus, pemafibrate-mediated induction of THBD is dependent on PPARα in HUVECs and HRMECs. To determine whether PPARα directly regulates THBD expression, we examined the physical and functional interactions of PPARα with THBD in ChIP-qPCR and luciferase reporter analysis, respectively. ChIP-qPCR analysis confirmed PPARα binding on the promoter region of THBD in HUVECs treated with pemafibrate (Fig. 2e).
PPARα preferentially binds DNA as a heterodimer to the PPAR-responsive element (PPRE), which is composed of two nuclear receptor consensus half-sites of AGG TCA organized as a direct repeat (direct repeat 1 [DR1]) 31 . PPARα forms heterodimers with RXRα, and the heterodimers bind the PPRE located in PPARαregulated genes. Scanning of the promoter sequences of human THBD identified two putative DR1 motifs (-189 bp to 177 bp and -1,135 bp to -1,123 bp from the TSS) (Fig. 2f). The former motif is conserved in mouse and rat Thbd genes, while the latter is not (Fig. 2g). Accordingly, we generated a reporter plasmid containing the promoter of THBD (~ 1.3 kb) and performed luciferase reporter analysis (Fig. 2h). Because pemafibrate affected the control reporter activities (i.e., renilla luciferase and β-gal) (data not shown), we used fenofibric acid as a PPARα agonist in the luciferase reporter assay. HUVECs were transfected with the reporter plasmid together with expression plasmids of PPARα and RXRα and then treated with fenofibric acid to determine PPARα-mediated transactivation. Treatment with fenofibric acid increased the activity of the luciferase reporter containing the THBD promoter by threefold. Deletion of two putative DR1 motifs blunted the transactivation by fenofibric acid. Mutation in the DR1 motif at -1,135 bp from the TSS alone had no effect on fenofibric acid-mediated transactivation, while mutation in the DR1 motif at -189 bp from the TSS or mutations in both DR1 motifs abolished transactivation. These results indicate that PPARα directly upregulates THBD expression in a ligand-dependent manner by binding to the DR1 motif at -189 bp from the TSS in HUVECs. www.nature.com/scientificreports/ Pemafibrate inhibits TM-dependent retinal inflammation in diabetic rats. Cell culture experiments showed that PPARα directly upregulated TM, which has antiinflammatory effects. We hypothesized that PPARα activation by pemafibrate could inhibit inflammation through upregulation of TM in the diabetic retina. To examine this, we first administered oral pemafibrate to rats via their feed and showed that pemafibrate (10 mg/kg or 30 mg/kg) significantly increased mRNA expression of PPARα target genes such as Pdk4 and Thbd in the retina (Fig. 3a,b). Next, we set up a TM knockdown system in the rat retina by intravitreal injection of siRNA. Intravitreal injection of 500 pmol of siRNA targeted to Thbd for 14 days successfully decreased the protein expression of TM (Fig. 3c). In the STZ-induced diabetic rat model, the protein levels of inflammatory molecules such as ICAM, MCP1, and VCAM-1 in the retina were elevated compared with the retinas of nondiabetic rats (Fig. 3d), as previously reported 32,33 . Oral intake of pemafibrate in STZ-induced diabetic rats markedly inhibited the elevation of inflammatory molecules, indicating that PPARα activation inhibits inflammatory responses in the retinas of diabetic rats. Furthermore, knockdown of TM in diabetic rats attenuated the pemafibrate-mediated inhibition of elevation of inflammatory molecules. These results indicate that the inhibition of PPARα activation by pemafibrate inhibits inflammation through the upregulation of TM in diabetic rat retinas.

Pemafibrate inhibits TM-dependent retinal vascular leukostasis and leakage. Vascular leuko-
stasis and leakage increase cumulatively with the progression of DR 34 . To determine whether PPARα activation inhibits vascular leukostasis and leakage in the retina, we examined them in diabetic rats treated with pemafibrate. FITC-labeled adherent leukocytes were not observed in control rats, while they were clearly observed in the retinal vasculature of untreated diabetic rats (Fig. 4a). Pemafibrate treatment significantly reduced the number of leukocytes in the rat retinas (p < 0.05) (Fig. 4a,b), indicating that it inhibits vascular leukostasis in diabetic rats. TM knockdown by siRNA attenuated the pemafibrate-mediated reduction of leukocytes in the retinal vasculature of diabetic rats (p < 0.05) (Fig. 4a,b), while control siRNA treatment did not, indicating that PPARα activation by pemafibrate inhibits vascular leukostasis through TM in the retina of diabetic rats. Next, we examined whether pemafibrate treatment reduces retinal vascular leakage in diabetic rats. The fluorescence intensity of FITC-dextran in the retinal vessels was significantly increased in diabetic rats compared with control rats (p < 0.05), indicating increased vascular leakage in the former (Fig. 5a,b). Pemafibrate treatment significantly reduced the florescence intensity of FITC-dextran in diabetic rats (p < 0.05) (Fig. 5a,b), indicating that PPARα activation by pemafibrate inhibits vascular leakage. In addition, TM knockdown by siRNA attenuated pemafibrate-mediated inhibition of vascular leakage in diabetic rats (p < 0.05) (Fig. 5a,b). These results revealed that PPARα activation by pemafibrate inhibits retinal vascular leukostasis and leakage via TM in diabetic rats. Table 3. List of overlapped genes.

Discussion
In the current study, we identified THBD encoding TM as one of the direct target genes of PPARα in vascular endothelial cells. PPARα activation by pemafibrate in a rat model of DR upregulated THBD and inhibited retinal inflammation and vascular leukostasis and leakage. To the best of our knowledge, this is the first report showing the novel molecular mechanism by which PPARα activation protects against DR. PPARα is a ligand-activated transcription factor, and its target genes are involved in fatty acid metabolism in tissues with high oxidative rates such as the liver, heart, skeletal muscle, and kidney 8,9 . PPARα is also expressed in other tissues and cells including the intestine, vascular endothelium, and immune cells 15 . Large clinical trials and several reports revealed that PPARα activation by an agonist has protective effects against DR in type 2 diabetes patients 10,11 , although the mechanisms of action of the PPARα agonist were unclear. Because DR is characterized by progressive loss of vascular cells and infiltration of inflammatory cells, we specifically focused on vascular endothelial cells and performed genome-wide analyses of PPARα binding sites and gene expression using ChIP-seq analysis and DNA microarray. Our ChIP-seq analysis identified 6,017 genomic PPARα binding sites, which were assigned to 4,186 proximal genes. Microarray analysis revealed 1,062 genes with expression levels upregulated by pemafibrate treatment in HUVECs.
The majority of PPARα binding sites identified by ChIP-seq were located at intragenic regions or 5′ proximal regions (< 10 kb from the TSS) in HUVECs, consistent with previous ChIP-seq analysis in human hepatoma cells 35 . These results are in line with several studies showing that transcription factor binding sites are generally distributed around the TSS 36,37 . Among 4,186 genes annotated as bound by PPARα, 221 including THBD were upregulated by pemafibrate treatment, indicating that these genes are direct PPARα targets in HUVECs. However, pemafibrate treatment did not affect the expression of the majority of genes annotated as bound by PPARα. Some PPARα binding sites may participate in transcriptional regulation of distal genes or in other mechanisms such as noncoding RNA and DNA methylation. It is also possible that our ChIP-seq analysis contained false-positive signals due to the nonspecific binding of antibody. Therefore, the functional relevance of PPARα binding sites needs to be elucidated further.
VEGF has been shown to be one of predominant factors regulating pathological conditions including chronic inflammation and resulting abnormal vasopermeable and angiogenic responses in DR. Our microarray analysis identified 1,062 upregulated and 477 downregulated genes in HUVECs treated with pemafibrate. VEGF mRNA was not altered in the present microarray analysis. Since it is possible that PPARα has transcriptional and nontranscriptional roles in protection against DR, more detailed protein analyses are needed to confirm the direct effects of PPARα activation on VEGF expression. In contrast, the inhibitory effects of PPARα on VEGF expression were clearly demonstrated in in vivo settings. Chen et al. reported the inhibitory effects of fenofibrate on HIF-1 and VEGF expression in the whole retina in a type 1 diabetes rat model and mouse oxygen-induced retinal angiogenesis model 38 . Tomita et al. found that VEGF is suppressed by PPARα activation, which inhibits HIF activity through serum FGF 12 induced and secreted in the liver 39 . Another study showed that PPARα ligands may suppress angiogenesis indirectly by inhibiting tumor cell production of VEGF and FGF2 and by increasing thrombospondin-1 40 . Although we need further detailed protein expression analyses in in vitro studies, these data may indicate that PPARα activation alone does not substantially affect VEGF expression, although it exerts obvious inhibitory effects through cross-reaction with other cytokines and organs. PPARα effects on VEGFinduced intracellular signaling have not been reported and remain to be elucidated.
Among 221 direct PPARα target genes, we focused on THBD that encodes TM. TM is an integral membrane protein expressed on the surface of endothelial cells and exerts antiinflammatory effects via several mechanisms [22][23][24][25][26] . TM inhibits inflammation through the activation of protease-activated receptor-1 (PAR-1) in the form of activated protein C (APC) 23,24 . Moreover, TM is a critical cofactor for thrombin-mediated activation of the thrombin-activatable fibrinolysis inhibitor (TAFI) 21 . The proinflammatory mediators are inactivated by TAFI 25 . TM, via thrombin-mediated activation of protein C and TAFI, provides protection against inflammation. Thrombin is an important inflammatory factor in retinal vascular diseases including DR 41 . It was reported that thrombin and prothrombin were increased in the vitreous of patients with proliferative DR compared with nondiabetic individuals 42 . Additionally, it was found that the lectin-like domain of TM sequesters inflammatory factors such as high-mobility group-B1 (HMGB-1) protein and lipopolysaccharide 43,44 . HMGB-1 promotes the process of cell apoptosis by activating the transcription factor nuclear factor-κB after binding to its receptors 45,46 . For these reasons, we hypothesized that PPARα activation inhibited the retinal vascular damage caused by retinal inflammation and apoptosis via the regulation of TM.
In the transcriptional regulation of THBD, our studies revealed that PPARα binds to the TSS upstream of THBD to activate gene expression. Furthermore, our luciferase reporter assay showed that the DR1 motif at 189 bp from the TSS is responsible for transactivation of the THBD promoter by PPARα. Previous studies demonstrated that other transcriptional factors (e.g., Kruppel-like factor 2 and RXRs) also bind to this region to transactivate the THBD promoter 47,48 , suggesting the importance of this region for the activity of the THBD promoter in vascular endothelial cells.
We used the rat model of DR to show that pemafibrate treatment prevents the upregulation of inflammatory molecules as well as vascular leukostasis and leakage. It was reported that CCL2, VCAM-1, and ICAM-1 are key inflammatory molecules in the diabetic retina which lead to leukocyte adhesion and vascular leakage 32,33 . Inflammatory mediators such as thrombin and HMGB-1 induce these inflammatory molecules 43,45,46 , while TM has inhibitory effects on inflammatory molecules. Therefore, we performed TM knockdown experiments in diabetic rats to determine whether the antiinflammatory effect of pemafibrate is mediated via the action of TM.
In vivo experiments demonstrated that the antiinflammatory effect of pemafibrate was canceled by intravitreal injection of Thbd siRNA. These results suggest that the upregulation of TM is one mechanism of the protective effects of PPARα modulators (e.g., pemafibrate) against DR and that a modulator upregulating TM could be a   novel therapeutic option for DR treatment. Further studies will be required to elucidate other mechanisms by which PPARα activation protects against DR.

Methods
Chemical reagents. Pemafibrate and fenofibric acid were kindly provided by Kowa Co., Ltd. (Nagoya, Japan).  ChIP-qPCR analysis. ChIP samples were analyzed by quantitative PCR using the gene-specific primers listed in Supplementary Table S2. ChIP signals were divided by no-antibody signals (input DNA) and presented as fold enrichment.  Luciferase reporter assay. The luciferase reporter assay was performed with the Beta-Glo assay system (Promega, Madison, WI, USA). HUVECs seeded on 24-well plates at 50% confluence in Opti-MEM (Life Technologies/Thermo Fisher Scientific) were transfected with 0.2 µg of reporter plasmid. HUVECs were incubated for 24 h in normal growth medium and then treated with 100 μM of fenofibric acid or vehicle (DMSO). Luciferase assays were performed 48 h after fenofibric acid treatment. Luciferase activity was normalized to the β-gal activity level.
Type 1 diabetes rat model. Type 1 diabetes was induced in male Wistar rats by an intraperitoneal injection of STZ (60 mg/kg). Briefly, 6 to 8 weeks after STZ injection, the rats were fed pemafibrate (10 mg/kg) for 2 weeks. Intravitreal injection of control or Thbd siRNA was performed 6 and 7 weeks after STZ injections. The rats were euthanized 8 weeks after STZ injection. Rats in which blood glucose levels were greater than 300 mg/ dl at tissue harvesting were defined as diabetic and used for the experiments.
Retinal vascular leukostasis ssay. The retinal vascular leukostasis assay was performed as described previously 38 . Briefly, rats were deeply anesthetized, and PBS was injected into the left ventricle. The anesthetized rats were perfused with PBS to remove nonadherent leukocytes in vessels. After injection of PBS, FITCconjugated concanavalin-A (40 μg/ml) (Vector Laboratories, Burlingame, CA, USA) was injected into the left ventricle. The retinas were surgically isolated and then flat mounted. FITC-labeled adherent leukocytes in the vasculature were counted under a fluorescence microscope by an operator masked to treatment allocation.

Measurement of vascular leakage in the retina.
Vascular leakage in the rat retinas was investigated using fluorescein angiography, as previously described 52 . Briefly, rats were deeply anesthetized, and FITC-dextran (Sigma-Aldrich, St. Louis, MO, USA) was injected into the left ventricle. After 5 min, the retinas were flat mounted and observed under a fluorescence microscope. Vascular leakage was quantitatively analyzed using Image J software by determining florescence intensities of FITC-dextran in the retina vessels.
Study approval. All animal protocols in this study were performed in accordance with the institutional guidelines and were approved by the St. Marianna University Graduate School of Medicine Institutional Animal Care and Use Committee.

Data availability
All data generated or analyzed during this study are included in the published article (and its supplementary files).