EP4 inhibition attenuates the development of diabetic and non-diabetic experimental kidney disease

The therapeutic targeting of prostanoid subtype receptors may slow the development of chronic kidney disease (CKD) through mechanisms that are distinct from those of upstream COX inhibition. Here, employing multiple experimental models of CKD, we studied the effects of inhibition of the EP4 receptor, one of four receptor subtypes for the prostanoid prostaglandin E2. In streptozotocin-diabetic endothelial nitric oxide synthase knockout mice, EP4 inhibition attenuated the development of albuminuria, whereas the COX inhibitor indomethacin did not. In Type 2 diabetic db/db mice, EP4 inhibition lowered albuminuria to a level comparable with that of the ACE inhibitor captopril. However, unlike captopril, EP4 inhibition had no effect on blood pressure or hyperfiltration although it did attenuate mesangial matrix accumulation. Indicating a glucose-independent mechanism of action, EP4 inhibition also attenuated proteinuria development and glomerular scarring in non-diabetic rats subjected to surgical renal mass ablation. Finally, in vitro, EP4 inhibition prevented transforming growth factor-ß1 induced dedifferentiation of glomerular podocytes. In rodent models of diabetic and non-diabetic CKD, EP4 inhibition attenuated renal injury through mechanisms that were distinct from either broadspectrum COX inhibition or “standard of care” renin angiotensin system blockade. EP4 inhibition may represent a viable repurposing opportunity for the treatment of CKD.


Results
The EP4 inhibitor ONO-AE3-208 attenuates albuminuria in streptozotocin-diabetic eNOS knockout mice. To determine whether inhibition of EP4 affects the development of experimental diabetic kidney disease, we initially performed studies in streptozotocin (STZ)-diabetic eNOS knockout (eNOS −/− ) mice that develop massive albuminuria very soon after the induction of diabetes 18 . Wildtype (C57BL/6) and eNOS −/− mice were made diabetic with STZ and were treated with ONO-AE3-208 (10 mg/kg/day) in drinking water beginning with the first intraperitoneal (i.p.) injection of STZ and continued for three weeks (Table 1). In comparison to their non-diabetic counterparts, body weight was lower in STZ-diabetic wildtype and eNOS −/− mice (Table 1). After three weeks of diabetes, kidney weight was increased in STZ-eNOS −/− mice but not in STZ-C57BL/6 mice, whereas kidney weight and kidney weight:body weight ratio were significantly lower in STZ-eNOS −/− mice treated with ONO-AE3-208 than in vehicle-treated STZ-eNOS −/− mice (Table 1). Two weeks after the first i.p. injection of STZ, urinary nephrin content was increased >10-fold in vehicle-treated STZ-eNOS −/− mice, whereas it was approximately 50% lower in ONO-AE3-208-treated mice (Fig. 1A). By three weeks, urinary albumin excretion rate (AER) had increased >40-fold in STZ-eNOS −/− mice compared to non-diabetic C57BL/6 mice, whereas AER was approximately 50% lower in ONO-AE3-208 treated STZ-eNOS −/− mice than vehicle-treated STZ-eNOS −/− mice (Fig. 1B). As expected, at this early stage of diabetes, glomerular volume was marginally albeit non-significantly increased in STZ-diabetic mice in comparison to their non-diabetic counterparts (Fig. 1C). Mesangial matrix index was also marginally increased in STZ-eNOS −/− mice ( Fig. 1D-I) and, although both glomerular volume and mesangial matrix index were numerically lower with ONO-AE3-208 treatment, changes in neither of these parameters achieved statistical significance.
ONO-AE3-208 reduces albuminuria and mesangial matrix accumulation in db/db mice. Although our experiments performed in STZ-eNOS −/− mice enabled us to detect an anti-albuminuric effect of ONO-AE3-208, we were cognizant of limitations of this experiment including the short duration of study (three weeks) and the confounding effects of eNOS absence. Thus, we decided to perform a second study, this time administering ONO-AE3-208 to Type 2 diabetic db/db mice, one of the most commonly studied mouse models of diabetes. We treated non-diabetic db/m mice and Type 2 diabetic db/db mice with ONO-AE3-208 or vehicle for eight weeks and, in this study, we compared the effects of ONO-AE3-208 with "standard of care" by treating an additional group of db/db mice with the ACE inhibitor, captopril (Table 2). There was no difference in blood glucose between db/db mice treated with vehicle, ONO-AE3-208 and captopril ( Table 2). As expected, systolic blood pressure (SBP) was lower in db/db mice treated with captopril than db/db mice treated with vehicle, whereas systemic pressure was not affected by ONO-AE3-208 (Table 2). Serum creatinine was decreased in vehicle-treated db/db mice in comparison to db/m mice, consistent with hyperfiltration (Fig. 3A). Captopril prevented the decrease in creatinine in db/db mice whereas ONO-AE3-208 had no effect (Fig. 3A) difference in serum creatinine, albuminuria was reduced with both ONO-AE3-208 and captopril ( Fig. 3B and C). Histologically, consistent with the marked lowering of blood pressure with ACE inhibition, the kidneys of  captopril-treated db/db mice showed pronounced juxtaglomerular hyperplasia, which was not observed in any of the other study groups (Fig. 3D-I). Further highlighting the difference in effect of EP4 inhibition and ACE inhibition, mesangial matrix accumulation was reduced with ONO-AE3-208 but not with captopril ( Fig. 3D-I).
ONO-AE3-208 attenuates kidney injury in subtotally nephrectomized rats. Whereas the db/db mouse model is widely studied as a model of diabetic kidney disease this model is also restricted in its applicability because db/db mice do not develop advanced glomerulosclerosis with glomerular filtration rate (GFR) decline as may be observed in patients. Indeed the renoprotective effects of ACE inhibition, now routinely used in diabetes management, were initially examined in a non-diabetic experimental model of raised intraglomerular pressure: the subtotally nephrectomized (SNx) rat 19 . Accordingly, to determine whether the reno-protective actions of ONO-AE3-208 extend to more advanced stages of nephropathy and to also determine whether the reno-protective effects of ONO-AE3-208 are glucose-dependent, we tested the effects of the EP4 inhibitor in SNx rats. Male Sprague Dawley rats underwent sham or SNx surgery and, one week later, were randomly allocated to receive either drinking water or ONO-AE3-208 in drinking water for seven weeks. In this series of experiments, we elected to compare the effects of ONO-AE3-208 when administered at two different doses: 1 mg/kg/ day and 10 mg/kg/day. SBP was increased and GFR was reduced in SNx rats in comparison to sham-operated  animals and neither dose of ONO-AE3-208 had an effect on either of these parameters (Table 3). In contrast, proteinuria (which was increased in SNx rats) was equivalently reduced with both 1 mg/kg/day and 10 mg/kg/ day of ONO-AE3-208 ( Fig. 4A-C). Because the anti-proteinuric effects were equivalent for the two doses of ONO-AE3-208, we combined the two treatment groups for histological analysis. Paralleling the changes in proteinuria, the magnitude of glomerular scarring, when assessed either on periodic acid-Schiff stained kidney sections ( Fig. 4D-H) or following immunostaining for collagen IV (Fig. 4I-M), was increased in vehicle-treated SNx rats and was attenuated with ONO-AE3-208.
ONO-AE3-208 prevents podocyte dedifferentiation induced by transforming growth factor-ß1 (TGF-ß1). The difference in the renal effects of ONO-AE3-208 and captopril, despite both being accompanied by an equivalent reduction in albuminuria, suggested to us that the reno-protective effects of EP4 inhibition are not limited to actions on hemodynamic forces. Likewise, the observation that the reno-protective effect of ONO-AE3-208 was also apparent in non-diabetic SNx rats suggested to us that this effect was also not restricted to high glucose mediated events. We thus reasoned that the actions of ONO-AE3-208 may at least in part be mediated by preventing the deleterious effects of dysregulated growth factors or cytokines that are common to both diabetic and non-diabetic CKD and that these effects may occur in resident glomerular cells. We immunostained kidney sections from both mice and humans and we observed prominent expression of EP4 in glomerular podocytes of each species (Fig. 5A,B). By dual immunofluorescence, we observed an accumulation of the mesenchymal marker, α-smooth muscle actin (α-SMA) in the glomeruli of SNx rats, coinciding with segmental loss of the podocyte slit-pore protein, nephrin (Fig. 5C). Collectively, these findings led us to speculate that signaling through the EP4 receptor in podocytes may facilitate their dedifferentiation into more mesenchymal-like cells. Accordingly, in our final experiments we exposed differentiated immortalized mouse podocytes to the pro-fibrotic growth factor TGF-ß1, which is upregulated in both diabetic 20 and non-diabetic kidney disease 21 . TGF-ß1 caused the dedifferentiation of cultured podocytes as reflected by an upregulation in the expression of α-SMA, snail, slug, collagen I and collagen IV whereas this effect was negated by pre-treatment of podocytes with ONO-AE3-208 ( Fig. 5D-H).

Discussion
Despite recent advances 22 , new treatments are still urgently needed for people with CKD. For example, for people with diabetes, a diagnosis of CKD more than doubles healthcare costs 23,24 and even for individuals with Stage 3A kidney disease (GFR 45-59 ml/min) all-cause mortality is increased four-fold 25 . Here, in an effort to examine new treatment opportunities for CKD we explored the effect of pharmacological antagonism of the PGE 2 , EP4 receptor. Across three different experimental rodent models of diabetic and non-diabetic kidney disease we observed that EP4 inhibition attenuated albuminuria or proteinuria, independent risk markers of renal decline 26 and mortality 27 . The effects of EP4 antagonism were phenotypically distinct from those of either broadspectrum COX inhibition or ACE inhibition and were not limited to the diabetic state, attenuating glomerular scarring in both Type 2 diabetic db/db mice and non-diabetic rats with progressive proteinuric kidney disease induced by surgical renal mass ablation. In cultured podocytes, EP4 inhibition prevented podocyte dedifferentiation. EP4 is a typical seven transmembrane domain G-protein coupled receptor, which as a superfamily make up over 50% of all current drug targets 28 . Like the other EP subfamily member, EP2, EP4 is a G s α coupled receptor which activates adenylate cyclase to increase cyclic adenylyl monophosphate (cAMP) production on ligand-binding 8,9 . Its ligand, PGE 2 , is generated via three different PGE 2 synthases: microsomal prostaglandin E synthase (mPGES)-1, mPGES-2 and cytosolic PGES (cPGES). In rat mesangial cells, mPGES-1 was induced by TGF-ß1 29 and, more recently, mPGES-1-derived PGE 2 was shown to contribute to podocyte injury caused by adriamycin 30 . Although in the latter study, the EP receptor subtype responsible for promoting podocyte injury in response to adriamycin was not established, an earlier study employing podocyte-specific EP4 overexpressing or deficient mice pointed to a deleterious function of the EP4 subtype receptor in podocytes of mice subjected to renal mass ablation 13 . On the background of this collective body of evidence, we sought to examine the effects of EP4 inhibition on the development of glomerular disease in CKD. We elected to first examine the effects of EP4 inhibition in STZ-eNOS −/− mice. We selected this mouse model for two reasons. Firstly, we had earlier shown that glomerular injury in STZ-eNOS −/− mice predominantly occurs within podocytes 18 (which our histological survey revealed are the major glomerular cell-type that expresses EP4). Secondly, the magnitude and pace of albuminuria development in STZ-eNOS −/− mice occurring within two to three weeks of diabetes induction 18,31 , allows the relatively rapid in vivo screening of compounds for possible anti-albuminuric actions 31 . Whereas the EP4 inhibitor ONO-AE3-208 caused an approximate 50% reduction in urinary albumin excretion in STZ-eNOS −/− mice, the broadspectrum COX inhibitor indomethacin made no difference. This latter observation underscores the distinct effects of selective prostanoid receptor subtype targeting. It is possible, for example, that with COX inhibition, concurrent blockade of the synthesis of other prostanoids (e.g. prostaglandin I 2 31 ) or concurrent prevention of the activation of other prostanoid receptors (e.g. EP2 32 ) negates the apparently beneficial effects observed with selective PGE 2 -EP4 targeting.
The anti-albuminuric effect that we observed in STZ-eNOS −/− mice was subsequently replicated in Type 2 diabetic db/db mice that were treated with ONO-AE3-208 for eight weeks and, in this longer duration study, was accompanied by attenuated mesangial matrix accumulation. One of the principal mechanisms by which glomerular injury occurs in diabetes is through glomerular hypertension 33 . Indeed, the reno-protective effects of ACE inhibitors are considered to be, at least in part, mediated by a diminution of intraglomerular pressure caused by preferential dilatation of the efferent arteriole 34 . In contrast, EP4 appears to play an important role in sustaining vasodilatation of the preglomerular afferent arteriole 35 . In the present study, serum creatinine was lower in db/ db mice than non-diabetic db/m mice suggestive of renal hyperfiltration (although GFR was not directly determined in these animals). This reduction in creatinine was normalized by the ACE inhibitor captopril suggestive of a reduction in intraglomerular pressure. However, it was unaltered by ONO-AE3-208. Thus, the effects of ONO-AE3-208 on albuminuria and matrix deposition in db/db mice are unlikely to be a consequence of renal hemodynamic actions. GFR was similarly unaffected by EP4 inhibition in SNx rats, where it was reduced in comparison to control animals and in contrast to the diabetic mouse study. Despite this lack of effect on GFR with EP4 inhibition, proteinuria and glomerular injury were both significantly reduced with ONO-AE3-208 in SNx rats. This disparity is not surprising and is reminiscent of the actions of ACE inhibition in this model. For instance, in the original report published over 30 years ago, treatment of subtotally nephrectomized Munich-Wistar rats with the ACE inhibitor enalapril reduced proteinuria and glomerular scarring without affecting single nephron GFR 19 . As stated at the outset, in the opinion of the investigators it is unlikely that the reno-protective actions of EP4 inhibition that were observed during the present series of experiments can be attributed to effects on a single pathway in a single renal cell-type. However, the collective observations suggest that the effects may be, at least partly, mediated by direct actions on resident glomerular cells. Consistent with this supposition, in cultured podocytes EP4 inhibition prevented TGF-ß1-induced podocyte dedifferentiation. Whereas EP4 is recognized as a G s α-coupled receptor, the intracellular signaling pathways that it initiates are more complex than this. For example, EP4 is also coupled with G i α 36 and this relationship may explain the comparatively lower ability of EP4 signaling to raise cAMP levels compared to EP2 9 . Through increasing cAMP levels, EP4 can activate both protein kinase A (PKA) and exchange protein directly activated by cAMP (Epac) 9 . Activation of the receptor can also induce intracellular signaling through a number of cAMP-independent second messengers, including phosphatidylinositol 3-kinase (PI3K) 9 , extracellular signal-regulated kinase (ERK) 9 , and p38 mitogen-activated protein kinase (p38 MAPK) 37 . Thus, there are several potential, and by no means mutually exclusive, points of convergence for TGF-ß1-and EP4-regulated pathways. In prostate cancer cells, for example, EP4 inhibition prevented TGF-ß signaling through the PI3K/Akt pathway 38 , whereas PI3K/Akt 39, 40 , ERK 40 and p38 MAPK 41 have each been implicated in TGF-ß1-induced epithelial mesenchymal transition.
The present report has weaknesses. Firstly, as emphasized above, the precise means by which EP4 inhibition prevented kidney or podocyte injury in the experiments herein described remains uncertain. Secondly, whilst the findings are consistent with some reports (e.g. refs 11, 13 and 15) they are at variance with others 14 . However, a strength of the study is the evaluation of the actions of one EP4 inhibitor in multiple different models of kidney disease. In the mouse and rat models of diabetic and non-diabetic kidney disease studied and with the particular small molecule (ONO-AE3-208) administered for the dose and duration described, EP4 inhibition consistently resulted in a preservation of glomerular permselectivity and a prevention of renal injury. Existing therapies for CKD, e.g. RAS-blockade 42 and sodium-glucose cotransporter 2 inhibition 22 , likely confer renoprotection through actions at both the cell and organ level and it seems reasonable to presume that the effects of EP4 inhibition are similarly broad. Amongst these broad effects, is the preservation of podocyte differentiation when challenged by the pro-fibrotic cytokine, TGF-ß1.
In summary, in rodent models of diabetic and non-diabetic CKD, EP4 inhibition attenuated renal injury through mechanisms that were distinct from either broadspectrum COX inhibition or RAS-blockade. EP4 inhibitors have reached clinical trial for other indications 43 and may offer a viable repurposing opportunity for the treatment of CKD.

Methods
Animal studies. Streptozotocin (STZ)-diabetic eNOS −/− mice. Male C57BL/6 and eNOS −/− (C57BL/6 genetic background) mice (The Jackson Laboratory, Bar Harbor, ME) were studied at eight weeks of age. Mice received a daily i.p. injection of STZ (55 mg/kg in 0.1 M citrate buffer, pH 4.5) or citrate buffer alone after a 4 hour fast for five consecutive days. Animals received ONO-AE3-208 (Medchemexpress, Monmouth Junction, NJ) at a dose of 10 mg/kg/day in drinking water 17 or drinking water alone, for three weeks beginning on the day of the first injection of STZ. In a previous report, ONO-AE3-208 administered orally to mice as a 10 mg/kg bolus achieved a peak plasma concentration of 677 ng/ml after 0.25 hours with 18% bioavailability 17 . Urine nephrin content (Exocell, Philadelphia, PA) and urine albumin excretion (Assaypro, St. Charles, MO) were determined by ELISA after housing mice in individual metabolic cages for 24 hours. Blood glucose was determined by OneTouch UltraMini (LifeScan Canada Ltd., Burnaby, British Columbia, Canada). To determine the effect of broadspectrum COX inhibition, male control and STZ-diabetic/6 and eNOS −/− mice were treated with either indomethacin (4 mg/kg/day in drinking water 44 , Cayman Chemical, Ann Arbor, MI) or drinking water alone beginning with the first i.p. injection of STZ and continued for two weeks (n = 10/group).
db/db mice. Male db/m and db/db mice on a BKS background (The Jackson Laboratory) aged eight weeks were randomly allocated to receive either ONO-AE3-208 (10 mg/kg/day in drinking water) or drinking water alone for eight weeks. An additional group of db/db mice were treated contemporaneously with captopril (Sigma-Aldrich, Oakville, Ontario, Canada) at a dose of 20 mg/kg/day in drinking water 18 . Blood glucose and urine albumin excretion were determined as already described. SBP was determined using a CODA non-invasive blood pressure system (Kent Scientific, Torrington, CT) 18 . Serum creatinine was determined by HPLC (Vanderbilt University, Nashville, TN). For silver staining, urine volumes containing 0.5 µg creatinine (Creatinine Companion, Exocell, Philadelphia, PA) were solubilized in sample buffer (ThermoFisher Scientific, Rockford, IL) and separated by SDS-PAGE before staining with a ProteoSilver Stain kit (Sigma Aldrich).
Subtotally nephrectomized rats. Male Sprague Dawley rats (Charles River, Montreal, Quebec) aged eight weeks underwent sham or subtotal nephrectomy surgery as previously described 45 . Briefly, for subtotal nephrectomy surgeries, under isoflurane anesthesia, the right kidney was removed via subcapsular nephrectomy and infarction of two thirds of the left kidney was achieved by selective ligation of two out of three of the branches of the left renal artery. Sham surgery involved laparotomy and manipulation of both kidneys prior to wound closure. One week later, rats were randomized to receive ONO-AE3-208 (1 mg/kg/day or 10 mg/kg/day) in drinking water or drinking water alone and they were followed for a further seven weeks. SBP was determined by tail cuff plethysmography (Powerlab, ADInstruments, Colorado Springs, CO) as previously described 46 . GFR was determined by single shot FITC inulin clearance with repeated sampling via the tail vein as previously described 45 . Urine protein excretion was determined using the benzethonium chloride method after 24 hour metabolic caging and urine creatinine was determined by autoanalyzer (Advia 1650, Siemens Medical Solutions Diagnostics, Tarrytown, NY). Mouse and rat kidney tissue was immersion-fixed in 10% neutral buffered formalin before routine processing and sectioning. All experimental procedures adhered to the guidelines of the Canadian Council on Animal Care and were approved by the St. Michael's Hospital Animal Care Committee, Toronto, Ontario, Canada. Glomerular volume. Glomerular volume (Gv) was calculated on 4 μm periodic acid Schiff-stained kidney sections using the formula: GV = (ß/k)(GA) 3/2 where ß = 1.38 pertains to the sphere and k = 1.10 is the distribution coefficient 47 .
Mesangial matrix index and glomerulosclerosis index. The magnitude of mesangial matrix deposition in the diabetic mouse studies or glomerulosclerosis in the SNx rat study was determined on periodic acid-Schiff stained kidney sections using a semi-quantitative scoring method on approximately 50 glomerular profiles per kidney section as previously described 31,45 .
Human studies. Formalin Immunohistochemistry. Immunohistochemistry was performed with antibodies in the following concentrations: collagen IV 1:500 (EMD Millipore, Darmstadt, Germany); EP4 1:100 (Cayman Chemical). Incubation with phosphate buffered saline in place of the primary antibody served as the negative control. After incubation with the appropriate horseradish peroxidase conjugated secondary antibody, sections were labeled with Liquid Diaminobenzidine and Substrate Chromogen (Dako North America Inc., Carpinteria, CA) before counterstaining in Mayer's hematoxylin. For quantitation of glomerular collagen IV in rat kidney sections, slides were scanned with the Aperio ScanScope System (Aperio Technologies Inc., Vista, CA). The proportional glomerular area positively immunostaining for collagen IV was determined in 30 randomly selected glomerular profiles from each kidney section using ImageScope (Aperio Technologies Inc.). Cultured mouse podocytes. Differentiated immortalized mouse podocytes were cultured as previously described 48 . Cells were serum starved for 16 hours before incubation with ONO-AE3-208 (1 µM 49 ) or vehicle (DMSO) for 1 hour before supplementation of the medium with 10 ng/ml recombinant TGF-ß1 (BioRad Laboratories Inc., Hercules, CA) for 48 hours. In a previous study of EP4 overexpressing human colonic epithelial cells, incubation with 1 µM ONO-AE3-208 resulted in a >90% reduction in PGE 2 -induced luciferase activity, whereas the response in EP2 overexpressing cells was unaffected 49 . RNA was isolated from cell extracts using TRIzol Reagent (ThermoFisher Scientific). SuperScript III Reverse Transcriptase (ThermoFisher Scientific) was used to reverse transcribe cDNA from 1 µg RNA. Primers were designed and validated using Primer-BLAST (http://www.ncbi.nlm.nih.gov/tools/primer-blast/) and were synthesized by Integrated DNA Technologies (Coralville, IA). Primer sequences were as follows: α-SMA forward CAGGGAGTAATGGTTGGAAT, reverse TCTCAAACATAATCTGGGTCA; snail forward GCCGGAAGCCCAACTATAGCGA, reverse TTCAGAGCGCCCAGGCTGAGGTACT; slug forward TGTTGCAGTGAGGGCAAGAA, reverse GACCCTGGTTGCTTCAAGGA; collagen I forward CAGTCGATTCACCTACAGCACG, reverse GGGATGGAGGGAGTTTACACG; collagen IV forward AGGACTTCCCGGCTTGAATG, reverse TCCCCCTTATCTCCTTGGGG; RPL13a forward GCTCTCAAGGTTGTTCGGCTGA, reverse AGATCTGCTTCTTCTTCCGATA. Gene expression was determined using SYBR green (Wisent Inc, St. Bruno, Quebec, Canada) on a ViiA7 Real-Time PCR System (ThermoFisher Scientific). Experiments were performed in triplicate and data analyses were performed using the Applied Biosystems Comparative C T method.