Identification of Activators of ERK5 Transcriptional Activity by High-Throughput Screening and the Role of Endothelial ERK5 in Vasoprotective Effects Induced by Statins and Antimalarial Agents

ERK5, an atypical mitogen activated protein kinase with transcriptional activity, has not only a kinase domain but also a transcriptional activity domain, the latter being regulated by an intramolecular interaction independent of its kinase activity (1). Using tamoxifen-inducible endothelial cell (EC)–specific ERK5 knockout (ERK5-EKO) mice, we have demonstrated that ERK5 possesses endothelial protective effects (1). We and others have shown that the laminar shear stress-mediated endothelial protection is due to ERK5 activation (2, 3), which inhibits leukocyte–endothelial interaction and adhesion molecule expression (1, 4–7). ERK5–Kruppel-like factor (KLF2) signaling is involved in the laminar flow–induced endothelial NO synthase (eNOS) expression and anti-inflammatory effects (3). These results collectively suggest that activating ERK5 is a novel approach to protecting ECs. To translate this idea into therapy, we performed high-throughput screening (HTS) of a large number of molecules to identify novel activators of ERK5 transcriptional activity. Of the most effective compounds screened were pitavastatin and quinacrine (QC; an old, but well-known, antimalarial drug).

Statins (or HMG-CoA reductase inhibitors) are principal therapeutic agents used in the treatment of hypercholesterolemia (8). However, they also exert cholesterol-independent effects such as improving endothelial function (9–12), reducing the progression of cardiac allograft vasculopathy (13), and improving 8-y survival of heart transplant recipients (13). Although some in vitro studies suggest that the cholesterol-independent statin effects may be achieved via Rho/ROCK (9), PI3K–Akt (10), ERK5–KLF2 (11, 12, 14), and KLF4 (15) signaling, molecular mechanisms through which statins improve vascular function and inhibit cardiac allograft vasculopathy remain unclear.

Due to their anti-inflammatory effects, antimalarial drugs quinacrine (QC), chloroquine (CQ), and hydrochloroquine (HCQ) are being used for the treatment of not only malaria but also for autoimmune diseases such as systemic lupus erythematosus and rheumatoid arthritis (16). Recently, the Hopkins Lupus Cohort and Lupus in Minorities: Nature Versus Nurture (LUMINA) nested case-control study showed that HCQ treatment was associated with a long-term protective effect on cardiovascular diseases (17), probably by inhibiting inflammation (16). The possible role of antimalarial agents as PG antagonists via inhibiting phospholipase A₂ was suggested (18), but recent publications challenged this concept (19–21). Therefore, the target molecule(s) of antimalarial agents in regulating inflammation remains unknown. More importantly, to our knowledge, there has never been a study on the role of antimalarial agents in cardiac allograft rejection.

The combination of HTS and studies on molecular mechanisms of ERK5 activators has allowed us to uncover the crucial role of ERK5, which is activated by pitavastatin and antimalarial drugs, in inhibiting endothelial inflammation and dysfunction. Furthermore, we found the crucial role of endothelial ERK5 in preventing cardiac allograft rejection. To our knowledge, this is the first report to define the role of endothelium during the course of cardiac allograft rejection and will suggest a new therapeutic application of known drugs including antimalaria drugs against endothelial inflammation and dysfunction.

Pitavastatin was kindly provided by Kowa Company (Tokyo, Japan). Rosuvastain was kindly supplied by AstraZeneca (London, U.K.). 6-Mercaptopurine monohydrate (6-MP) was purchased from Alfa Aesar (catalog number A12197; Johnson Mathey Company), and quinacrine dihydrochloride (QC) from Sigma-Aldrich (catalog number Q3251; St. Louis, MO). Chloroquine phosphate (CQ; CAS 50-63-5) and hydroxychloroquine sulfate (HCQ; CAS 747-36-4) were from Santa Cruz Biotechnology (Dallas, Texas). Anti–p-ERK5 (#3371; Thr²¹⁸/Tyr²²⁰) and anti-ERK5 (#3372) were from Cell Signaling Technology (Danvers, MA). Anti-tubulin (T-5168) was from Sigma-Aldrich, anti-MAPK/ERK kinase (MEK5; KAP-MA003E) was from StressGen (San Diego, CA), anti-CD45 (Sc-103012) was from Santa Cruz Biotechnology, and anti-CD3 (ab5690) was from Abcam (Cambridge, MA).

HUVECs were cultured as described previously (22). Bovine aortic endothelial cells were cultured in M199 medium (catalog number 11150-059; Invitrogen, Grand Island. NY) supplemented with 10% FBS Clone III (catalog number SH 30109.03; HyClone, Logan, UT), 1% NEM-amino acid (catalog number 11130-051; Invitrogen), and 1% Antibiotic-Antimycotic (catalog number 30-004Cl; Cellgro; Mediatech, Manassas, VA). Human pulmonary aortic endothelial cells were cultured in M200 medium (Cascade Biologics, catalog number M-200-500; Life Technologies, Grand Island, NY) supplemented with 2% FBS (catalog number S11050; Atlanta Biologicals, Lawrenceville, GA) and low serum growth supplement (catalog number S-003-10; Cascade Biologicals). The HeLa cell line stably expressing ERK5 was cultured in DMEM containing 10% FBS (catalog number S11050; Atlanta Biologicals) and 100 μg/ml G418 (catalog number 345810; Calbiochem, San Diego, CA). All cells were maintained at 37°C in the humidified atmosphere of 5% CO₂ and 95% air.

We generated a stable HeLa cell line coexpressing pG5-Luc, which contains five Gal4 binding sites upstream of a minimal TATA box and pBIND-ERK5 fused to Gal4 (pBIND-ERK5). Using these reporter cells, the MicroSource SPECTRUM Collection (http://www.msdiscovery.com/spectrum.html), which consists of ∼2000 structurally diverse compounds selected by medicinal chemists and biologists for testing a wide range of biological activities, was screened at the University of Rochester HTS Core facility. Included in this collection are the following three classes of compounds: 1) known drugs that have reached clinical trial stages within the United States have been assigned the United States Approved Name or U.S. Pharmacopeia (USP) status and are found in the USP Dictionary of United States Approved Name and International Drug Names (2005, USP); 2) known drugs from Europe and Asia; and 3) natural compounds with already known pharmacologic effects. A major advantage of screening this group of compounds is that most of them have known biological activities that can be investigated in more detail if a specific hit is found.

Stable cells plated on a 384-well plate (2500 cells/well) were treated with test compounds at the concentration of 5 μM for 18 h. The level of luciferase activity was assayed using a Luciferase kit (Promega, Madison, WI), and a series of positive and negative control compounds were used as references.

Cells were transfected with pG5-Luc and Gal4-ERK5 constructs using Opti-MEM (Invitrogen) containing Plus-Lipofectamine mixture. After transfection, cells were treated with various concentrations of pitavastatin, CQ, and HCQ for 24 h. Cells were then harvested, and the level of luciferase activity was assayed using a dual-luciferase reporter system (Promega) and measured using a TD-20/20 Luminometer (Turner Designs, Sunnyvale, CA). Transfections were performed in triplicates, and each experiment was repeated at least three times. Because Gal4-ERK5 also contains the Renilla luciferase gene, the expression and transfection efficiencies can be normalized by Renilla luciferase activity. We verified our HTS data by detecting both luciferase and Renilla activity, which also controlled for toxic effects of the compounds.

Cells were transfected with the pG5-luc vector and various pBIND (Gal4) and pACT (VP16) plasmids (Promega), as indicated, for 4 h. Cells were then washed, and complete M200 medium supplemented with low serum growth supplement and 2% FBS was added. In CQ-treated samples, cells were then incubated 5 more h prior to CQ treatment. The pG5-luc vector contains five Gal4 binding sites upstream of a minimal TATA box, which in turn, is upstream of the firefly luciferase gene. pBIND containing Gal4 was fused with SMRT1, and pACT containing VP16 was fused with peroxisome proliferator activated receptor γ (PPARγ). Cells were harvested 24 h after transfection unless otherwise indicated, and luciferase activity was assayed as described previously (1).

Total RNA from cells was isolated using an RNeasy Plus Mini Kit (catalog number 74136; Qiagen, Valencia, CA) according to the manufacturer’s instruction. Total RNA from cardiac allografts harvested 5 d after transplantation was isolated using an RNeasy Fibrous Tissue Mini Kit (catalog number 74704; Qiagen). At the time of harvest, cardiac allografts were stored in RNAlater solution (catalog number AM7020; Ambion, Grand Island, NY) until use. To isolate total RNA, stored hearts were taken out of the storage solution, minced using a razor blade, and transferred to a 1.5-ml tube with 600 μl buffer RLT Plus (2-ME added). The lysate was loaded onto a QIAshredder spin column (Qiagen) placed in a 2-ml collection tube, centrifuged for 2 min at 15,000 rpm, and purification was continued according to the manufacturer’s instructions. First-strand cDNA was synthesized, and target cDNA levels were quantified by real-time RT-PCR using a MyiQ2 Optics Module (Bio-Rad) as described previously (22).

The following primers were used: human GAPDH forward: 5′-GGT GGT CTC CTC TGA CTT CAA CA-3′ and reverse: 5′-GTT GCT GTA GCC AAA TTC GTT GT-3′; human KLF2 forward: 5′-GCA CGC ACA CAG GTG AGA AG-3′ and reverse: 5′-ACC AGT CAC AGT TTG GGA GGG-3′; human eNOS forward: 5′-GTG GCT GTC TGC ATG GAC CT-3′ and reverse: 5′-CCA CGA TGG TGA CTT TGG CT-3′; human KLF4-forward: 5′-ACC AGG CAC TAC CGT AAA CAC A-3′ and reverse: 5′-GGT CCG ACC TGG AAA ATG CT-3′; human VCAM1 forward: 5′-GGT GGG ACA CAA ATA AGG GTT TTG G-3′ and reverse: 5′-CTT GCA ATT CTT TTA CAG CCT GCC-3′; mouse IL-1 forward: 5′-CAA CCA ACA AGT GAT ATT CTC CAT G-3′ and reverse: 5′-GAT CCA CAC TCT CCA GCT GCA-3′; mouse IL-6 forward: 5′-CAG AAT TGC CAT CGT ACA ACT CTT TTC TCA-3′ and reverse: 5′-AAG TGC ATC ATC GTT GTT CAT ACA-3′; mouse IL-10 forward: 5′-GGT TGC CAA GCC TTA TCG GA-3′ and reverse: 5′-ACC TGC TCC ACT GCC TTG CT-3′; mouse IFN-γ forward: 5′-TCA AGT GGC ATA GAT GTG GAA GAA-3′ and reverse: 5′-TGG CTC TGC AGG ATT TTC ATG-3′; mouse TNF-α forward: 5′-CAT CTT CTC AAA ATT CGA GTG ACA A-3′ and reverse: 5′-TGG GAG TAG ACA AGG TAC AAC CC-3′; mouse β-actin forward: 5′-AGA GGG AAA TCG TGC GTG AC-3′ and reverse: 5′-CAA TAG TGA TGA CCT GGC CGT-3′ (23); mouse MCP1 forward: 5′-CCA CTC ACC TGC TGC TAC TCA T-3′ and reverse: 5′-TGG TGA TCC TCT TGT AGC TCT CC-3′ (24).

Cells were transfected with a reporter gene encoding KLF2 and eNOS promoter as described previously (22, 25). Transfected cells were then treated with various concentrations of pitavastatin for 24 h. KLF2 and eNOS promoter luciferase activity was assayed using a dual-luciferase reporter assay system. Experiments were repeated three times.

In vitro ERK5 kinase activity was measured as we described previously (22). For each reaction, 1 μg purified GST–myocyte enhancer factor-2C (MEF2C)–GST fusion protein on beads and 0.5 μg GST-ERK5 fusion protein were added to 30 μl reaction buffer (10 mM MgCl₂, 10 mM MnCl₂, and 25 mM HEPES [pH 7.5]). Different doses of pitavastatin were added to the mixture, and reactions were allowed to take place for 30 min at 30°C in the dark with vigorous shaking. Cold ATP (final concentration 1 mM) and 1 μCi γ-[³²P]-ATP were added to the reaction mixture, which was then gently mixed and further incubated as before for 30 min. Finally, 6× sample buffer was added to terminate the reaction. The samples were boiled, electrophoresed on SDS-polyacrylamide gels, and transferred to polyvinylidene difluoride membranes. The film exposed to the membrane was developed. MEF2C phosphorylation was quantified using Fujifilm Image Gauge software (Version 4.0). Amounts of GST-MEF2C fusion protein used in each sample were visualized by Poncreau S staining. Amounts of rGST-ERK5 fusion protein used in each sample were detected by Western blotting using anti-ERK5 Ab.

Western blotting was performed as described previously (22).

Tamoxifen (4-hydroxytamoxifen [4-OHT])–inducible EC-specific ERK5 KO mice (VeCad-CreER-ERK5/flox/flox or inducible ERK5-EKO mice) were generated as we described previously (22). To induce deletion of the endothelial Erk5 gene, ERK5-EKO mice were injected with peanut oil with or without 2 mg 4-OHT for 5 consecutive d. Sustained reduction of endothelial ERK5 was confirmed by Western blotting using mouse endothelial cells isolated from ERK5-EKO mice 2 wk after 4-OHT injection as described previously (22). At age 8–10 wk, mice (C57BL/6) were made diabetic by i.p. injection of a single dose of freshly prepared streptozotocin (STZ) solution (150 mg/kg body weight in citrate saline [pH 4.5]) using a 26.5-gauge needle. Diabetic status was confirmed by blood glucose measurements. Animals were maintained according to the protocol approved by the University of Rochester Institutional Animal Care and Use Committee.

Mice were anesthetized, and blood was collected from the abdominal artery. Plasma was prepared, and total cholesterol, high-density lipoprotein (HDL), and non-HDL (low-density lipoprotein [LDL] and very LDL) concentrations were measured enzymatically using commercially available kits (catalog number ab65390; Abcam) according to the manufacturer’s instructions. Plasma glucose was measured using blood glucose meter (Advocate, model TD-4223F; Diabetic Supply of Suncoast, Baton Rouge, LA).

Leukocyte rolling on the endothelium, leukocyte rolling flux (number of rolling leukocytes passing a perpendicular line placed across the target vessel in 1 min), and vascular reactivity were measured and analyzed as described previously (22).

To determine the effect of ERK5 activation by statin on allograft rejection and immune responses, heterotopic cardiac transplantation was performed. Two weeks before transplantation, VeCad-CreER-ERK5/flox/flox mice (inducible ERK5-EKO) and VeCad-CreER–wild-type (WT) mice (being used as a control), both with C57BL/6 genetic background, were injected with 4-OHT. BALB/c male mice (8–10 wk old) were used as recipients. Cardiac transplantation was performed as described previously (26). Briefly, donor’s ascending aorta and pulmonary artery were anastomosed to recipient’s abdominal aorta and inferior vena cava, respectively. All anastomoses were done in an end-to-side fashion. Following cardiac transplantation, manual palpation was performed twice a day to gauge allograft rejection, and cessation of beating was interpreted as rejection. This step was done in a blind fashion.

To assess the progress of graft rejection and immune responses, the transplanted heart was harvested 5 d after the surgery. The recipient mice were anesthetized with i.p. injection of ketamine (50 mg/kg) and xylazine (10 mg/kg) in saline, sacrificed, and the grafted hearts harvested. The vascular anastomoses were cut off and discarded, and the heart was divided into two segments: the apex segment was stored in RNAlater solution (catalog number AM7020; Ambion, Grand Island, NY) for RNA isolation, and the remaining heart was fixed in acidic methanol (methanol/acetic acid/distilled deionized H₂O = 6:3:1), embedded in paraffin, sectioned (5 μm), and stained by H&E or specific Abs.

Parenchymal rejection (PR) was assessed in allografts 5 d after transplantation using the modified grading system of the Working Formulation of the International Society for Heart and Lung Transplantation and was scored on a scale of 0–4 (0, no rejection; 1, focal mononuclear cell infiltrates; 2, focal mononuclear infiltrates with necrosis; 3, multifocal infiltrates with necrosis; and 4, widespread infiltrates with hemorrhage and/or vasculitis). Occlusion (graft arterial disease [GAD]) was scored on a scale of 0–4 on the extent of luminal occlusion averaged over 10 arteries (0, <10%; 1, 10–25%; 2, 25–50%; 3, 50–75%; and 4, >75% occlusion of the vascular lumen) as described previously (27).

Data are expressed as mean ± SD. Nonparametric methods were used to analyze differences between independent groups. Comparisons between two groups for graft survival were performed by using Kaplan–Meier and log-rank tests. Comparisons among more than two groups were analyzed by Kruskal–Wallis test, followed by Bonferroni multiple-comparison test. Statistic analysis was performed using GraphPad Prism software for Macintosh version 5.00 (GraphPad Software, San Diego, CA). The sample size, the time intervals, and where the correlation data come from are listed in each figure and figure legend. A p value < 0.05 was accepted as being statistically significant and indicated by one asterisk (*), and a value <0.01 is indicated by two asterisks (**).

To perform a cell-based assay for ERK5 transcriptional activity, we generated a HeLa cell line stably expressing pG5-Luc and pBIND-ERK5 vectors. The pG5-luc vector contains five Gal4 binding sites upstream of a minimal TATA box, which, in turn, is upstream of the firefly luciferase gene. The pBIND-ERK5 vector contains Gal4 fused with ERK5. When ERK5 transcription activity is activated, Gal4 of pBIND-ERK5 bind the Gal4 binding site of pG5-luc, which results in increased firefly luciferase expression in cells. For ERK5 activator screening, cells were treated with compounds in the MicroSource SPECTRUM Collection for 18 h and luciferase activity determined. A scoring system was generated based on the intensity of luciferase activity ranging from 4000 (highest) to 0 (lowest) in an arbitrary unit, and statistical assessment of this assay gave a Z’ factor of 0.57, which is suitable for HTS. From the screening, three compounds that had scores >500 were selected and reassayed in duplicates. Among the three compounds identified as novel ERK5 activators, two were pitavastatin and 6-MP (an activator of Nur77, which is an ERK5 downstream target) (12, 28), and the third was QC (Fig. 1A). Pitavastatin, 6-MP, and QC were then confirmed to increase ERK5 transcriptional activity in a dose-dependent manner using the same HeLa cell line (Fig. 1B–D).

To examine whether pitavastatin, 6-MP, and QC could also activate ERK5 transcriptional activity in ECs, we used human pulmonary aortic endothelial cells and bovine aortic endothelial cells. Although 6-MP increased ERK5 transcriptional activity in the genetically engineered HeLa cells, we could not detect any significant activation in ECs (Supplemental Fig. 1A, 1C). In contrast, both pitavastatin and QC increased ERK5 transcriptional activity in ECs we tested (Supplemental Fig. 1B–D).

Among the 2000 compounds screened, several different statins were present. Interestingly, however, not all increased ERK5 transcriptional activity. Rosuvastatin showed no activation (score <50), whereas simvastatin marginally increased ERK5 transcriptional activity (score 220). Pitavastatin, simvastatin, and atorvastatin increased ERK5 transcriptional activity in HUVECs (Supplemental Fig. 1E, 1G). Because KLF2 promoter activity is positively affected by certain statins only (12), it is likely that the similar statin selectivity exists for ERK5 activation. We also determined the effect of CQ and HCQ on ERK5 transcriptional activity in HUVECs. As shown in Fig. 1E, they dose-dependently increased ERK5 transcriptional activity. It was reported that as much as 100 μM CQ or HCQ was necessary in in vitro experiments to generate an intracellular drug level similar to that in patients receiving 400 mg/d HCQ for antirheumatic therapy (29). Therefore, the 5-μM dose we used should be regarded as highly selective for the drug effect. To further characterize pitavastatin-induced ERK5 activation, we examined if this statin was able to induce ERK5-TEY motif phosphorylation in HUVECs and found that it could do so at a level similar to that achieved by steady laminar flow (Fig. 2A). Because KLF2 and eNOS are regulated by ERK5 in ECs (2, 3), we then examined the effect of pitavastatin on KLF2 and eNOS promoter activity in HUVECs and found that both were dose-dependently increased by pitavastatin treatment (Fig. 2B, 2C). ERK5 small interfering RNA (siRNA) was then used to determine the role of ERK5 in pitavastatin-induced eNOS and KLF2/4 mRNA expression in HUVECs. Indeed, ERK5 depletion inhibited the pitavastatin effect on eNOS and KLF2/4 mRNA expression (Fig. 2D–G).

To determine the role of MEK5, an upstream kinase of ERK5 (Fig. 3A), in the action of pitavastatin, we depleted MEK5 by siRNA in HUVECs. In these cells, we also expressed constitutively active form of MAPK kinase kinase 3 (CA-MEKK3), which is an upstream kinase of MEK5, and detected ERK5 transcriptional activity. As shown in Fig. 3B and 3C, MEK5 depletion abolished CA-MEKK3–mediated ERK5 transcriptional activity. In contrast, we detected no inhibition by MEK5 siRNA on pitavastatin-induced ERK5 transcriptional activity (Fig. 3D), suggesting that pitavastatin exerts its effect directly on ERK5. This possibility was tested using a cell-free system, in which rERK5 kinase activity was assayed using GST-MEF2C as a substrate in the presence of pitavastatin. Pitavastatin dose-dependently activated ERK5 kinase activity with an IC₅₀ of ≈0.43 μM (Fig. 3E). We then examined the involvement of MEK5 in pitavastatin-induced KLF2/4 and eNOS expression in ECs and found that MEK5 depletion had no effect on the basal KLF2 induction (Fig. 3F) and only ≈40–50% inhibition in pitavastatin-induced KLF4 and eNOS induction (Fig. 3G, 3H). Of note, because we observed complete inhibition of steady laminar flow-induced eNOS mRNA expression by MEK5 siRNA (Fig. 3I), the level of MEK5 depletion by siRNA was biologically effective.

To determine the involvement of MEK5 kinase activation on pitavastatin-induced ERK5 TEY motif phosphorylation, we transfected ECs with MEK5 siRNA. As shown in Supplemental Fig. 2A, we found that the deletion of MEK5 inhibited this phosphorylation, suggesting that MEK5 is involved in ERK5 TEF motif phosphorylation. In contrast, MEK5 depletion did not inhibit ERK5 transcriptional activity (Fig. 3D, Supplemental Fig. 2B). These data are consistent with our earlier suggestion (2) that the regulation of ERK5 transcriptional activity can be independent of ERK5 kinase activity or ERK5 TEF motif phosphorylation.

Of the three antimalarial drugs (CQ, HCQ, and CQ) found as ERK5 activators, CQ and HCQ but not QC are commonly used to treat systemic lupus erythematosus and rheumatoid arthritis (16), as they reduce the severity of renal dysfunction through their immune-modulatory, anti-inflammatory, and antithrombotic effects. Thus, we investigated effects of CQ and HCQ on ERK5 signaling. Although pitavastatin increased KLF2/4 expression (Fig. 4A), and CQ increased ERK5 phosphorylation (Fig. 4B), neither CQ nor HCQ was able to upregulate KLF2/4 expression (Fig. 4A). However, they inhibited TNF-α–induced VCAM-1 expression (Fig. 4C), suggesting that both CQ and HCQ had an anti-inflammatory effect via KLF2/4-independent mechanisms. Interestingly, their inhibitory effect on TNF-α–induced VCAM-1 mRNA expression was abolished by the depletion of ERK5 (Fig. 4C) but not MEK5 (Fig. 4D). These data show a crucial role of ERK5 but not KLF2/4 or MEK5 in certain anti-inflammatory effects of these antimalarial agents on ECs.

ERK5 is not only functional as a kinase or a coactivator, but also can release a corepressor, SMRT, from PPARs by direct binding with PPARs, as we have reported previously (1). It has been reported that PPARs can inhibit NF-κB transcriptional activity by releasing a corepressor such as Bcl6, SMRT1, and NCoR1 (30–33). Therefore, we examined whether CQ could release SMRT1 from PPARγ as we had previously shown by steady laminar flow (1). We found that CQ treatment indeed disrupted binding between PPARγ and SMRT (Fig. 4E). Interestingly, the deletion of ERK5 not only inhibited this disruption, but also accelerated binding between PPARγ and SMRT1 after CQ treatment (Fig. 4E, 4F). These results can provide a mechanistic insight into how CQ can inhibit TNF-induced NF-κB activation without increasing KLF2 expression and also provides a new functional mechanism of ERK5 in regulating inflammation.

Leukocyte rolling and adhesion are characteristics of vascular inflammation, and vascular reactivity to acetylcholine (ACh) is an indicator of normal vascular function. Because increased leukocyte rolling (34) and decreased ACh-induced vasodilation (35) have been reported in diabetes mellitus (DM) (22), we used this model to study vasoprotective effects of pitavastatin. In an STZ-treated mouse diabetic model, leukocyte rolling was increased and ACh-induced vasodilation was dampened compared with those of the vehicle-treated mice (Fig. 5A–D, Supplemental Videos 1–3). When DM mice were treated with pitavastatin, we observed a clear recovery from the DM-induced vascular inflammation and dysfunction (Fig. 5A–D, Supplemental Video 3). Unlike the effect of lowering blood glucose levels by insulin on leukocyte rolling/adhesion (22), pitavastatin affected neither blood glucose nor total cholesterol levels (Fig. 5E), demonstrating that pitavastatin’s action on leukocyte rolling/adhesion is independent of both blood glucose and cholesterol-lowering effects.

To test the role of pitavastatin-mediated endothelial ERK5 activation in these vasoprotective effects, we assayed leukocyte rolling and ACh-induced vasodilation in inducible ERK5-EKO mice. Leukocyte rolling was increased and ACh-induced vasodilation was decreased in 4-OHT–treated ERK5-EKO, but not in VE-Cad–CreER/WT mice (Fig. 5F–J, Supplemental Videos 4–6). Pitavastatin and the depletion of endothelial ERK5 affected neither total cholesterol nor blood glucose levels (Fig. 5G). Interestingly, pitavastatin had no effect on leukocyte rolling and vessel reactivity in ERK5-EKO mice (Fig. 5F, 5H–J, Supplemental Videos 6, 7). These results show a critical role for endothelial ERK5 in mediating the vasoprotective effects of pitavastatin.

Inflammation plays a major role in graft rejection. Because pitavastatin activates ERK5 and reduces inflammatory responses, we examined if this statin could have a protective effect against acute cardiac allograft rejection. We performed a total allomismatched heterotopic cardiac transplantation, in which hearts from either VE-Cad–CreER/WT or ERK5-EKO mice (C57BL/6 background) were transplanted to the abdominal aorta of BALB/c recipient mice. Pitavastatin treatment significantly prolonged the survival of VE-Cad–CreER/WT allografts (Fig. 6A, left panel) but was unable to prolong the survival of ERK5-EKO allografts (Fig. 6A, middle and right panels).

Because cytokines and MCP-1 are involved in allograft rejection (36, 37), we investigated pitavastatin’s effects on their expression in harvested allografts. In VE-Cad–CreER/WT allografts after 5 d of transplantation, pitavastatin treatment increased IL-10 expression while decreasing the expression of TNF-α, IFN-γ, MCP-1, and IL-1. IL-6 mRNA expression had a tendency to be also reduced by pitavastatin treatment in VE-Cad–CreER/WT allografts after 5 d of transplantation, but the difference did not reach statistical significance. However, these changes were not found in the ERK5-EKO allografts (Fig. 6B). Taken together, these data suggest that pitavastatin’s effect on allograft survival is mediated by endothelial ERK5 that regulates expression of inflammatory molecules.

Five days after transplantation, vehicle-treated VE-Cad–CreER/WT allografts showed severe coronary arteriolitis with perivascular edema and myocardial necrosis. Inflammatory cells and thrombi were present in the vessel lumen of vehicle-treated VE-Cad–CreER/WT allografts. Although these lesions were morphologically different from chronic GAD lesions, we examined PR and GAD using the GAD scoring system and quantified the extent of early luminal occlusions as reported previously (27, 36). PR and GAD scores were significantly lower in pitavastatin-treated VE-Cad–CreER/WT allografts than in vehicle treated VE-Cad–CreER/WT allografts, but there was no difference between pitavastatin- and vehicle-treated ERK5-EKO allografts (Fig. 7A, 7B). However, we found similar luminal occlusions in both VE-Cad–CreER/WT and ERK5-EKO allografts (Fig. 7), suggesting that pitavastatin prevents vascular occlusion due to accumulated inflammatory cells and thrombosis in allografts and that this graft survival effect is mediated by endothelial ERK5.

Because cardiac allograft rejection is associated with accumulation of inflammatory cells and increased expression of proinflammatory molecules (36), we examined CD45- and CD3-positive cell infiltration in allografts. Such cells were detected on 5 d after transplantation in vehicle-treated allografts (Fig. 8A, 8B). The number of CD45-positive, but not CD3-positive cells, was decreased in pitavastatin-treated VE-Cad–CreER/WT allografts. However, there was no difference in the extent of infiltration of CD45-positive cells between vehicle- and pitavastatin-treated ERK5-EKO allografts (Fig. 8A, 8B).

We identified certain statins and antimalarial agents as potent ERK5 activators by HTS, which are shown to induce anti-inflammatory effects via activating endothelial ERK5. Pitavastatin is one of the statins identified as an ERK5 activator. Interestingly, depletion by siRNA of ERK5 but not MEK5 inhibited pitavastatin-mediated KLF2 induction, strongly suggesting that there is a direct interaction between ERK5 and pitavastatin. Indeed, our in vitro kinase assay showed activation of rERK5 by pitavastatin in a cell-free system. Importantly, our study shows that pitavastatin reduced the extent of vessel occlusion by inflammatory cells and thrombi in allografts, but this beneficial effect was completely abolished by ERK5 depletion in ECs. In a clinical study (38), the contribution of coronary endothelial dysfunction after heart transplantation has been suggested, but the mechanism for this remains largely unclear. To our knowledge, this is the first report that defines the crucial role of endothelial ERK5 in the process of cardiac allograft rejection, especially under statins treatment. Although the anti-inflammatory role of ERK5 (22) and statins (12) was reported, the role of ERK5 in statin-induced protection against endothelial dysfunction and cardiac allograft rejection has not been explored in vivo. In this study, we have established using both in vitro and in vivo approaches that via endothelial ERK5, certain statins and antimalaria drugs provide protection against endothelial dysfunction. Our comprehensive approaches to detect small molecules by HTS and determine not only their molecular mechanism in vitro but also pathological consequence in vivo have established the critical role of ERK5 in endothelial protection and identified endothelial ERK5 as a novel drug target for heart transplantation. Lastly, we have also established a novel screening assay for detecting ERK5 activators, which provides a technical contribution for further search of compounds that possibly prevent endothelial dysfunction and subsequent cardiac allograft rejection.

In a total allomismatched cardiac transplantation model, pitavastatin significantly prolonged allograft survival and reduced the expression of inflammatory gene expression while increasing anti-inflammatory IL-10 gene expression. A decrease in CD45-positive cells was observed in pitavastatin-treated VE-Cad–CreER/WT allografts, but such a decrease was not found in pitavastatin-treated ERK5-EKO allografts. In contrast to a previous report (39), infiltration of CD3-positive cells was not affected by pitavastatin. Because ERK5 is not essential for T cell development and survival (40), the lack of pitavastatin’s effects on T cell infiltration may be reasonable.

Antimalarial therapy for patients with autoimmune diseases has been an effective treatment since 1955, as this class of drugs improves the survival and remission rates with a good safety profile (16, 17, 41). It is worth noting that antimalarial agents are recommended for patients with lupus nephritis as they reduce the severity of renal dysfunction through their immune-modulatory, anti-inflammatory, and antithrombotic effects, which subsequently prevent endothelial dysfunction and reduce renal inflammation (16). In our study, we found that the anti-inflammatory effect of CQ and HCQ was mediated by an ERK5-dependent but MEK5- or KLF2/4-independent mechanism. To our knowledge, this is the first report defining endothelial ERK5 as a primary action site and proposing a novel application of ERK5 activators such as antimalarial agents in cardiac allograft transplantation. Because ERK5 also increases the expression of other anti-inflammatory genes and the activity of transcription factors including PPARs and Nur77 (1, 42, 43), these molecules may be also involved in the CQ- and HCQ-induced anti-inflammatory effects, which need further investigation. Taken together, our results have revealed a novel role for ERK5 as a direct target of certain statins and antimalarial agents, and as such, they prevent endothelial inflammation and dysfunction and suppress acute rejection of cardiac allografts.

As shown in Supplemental Fig. 2C, we could not find ERK1/2 activation by pitavastatin in ECs. However, it has been reported that pitavastatin can inhibit (not activate) lysophosphatidylcholine-induced ERK1/2 activation in vascular smooth muscle cells (44). Because we did not observe inhibition of the basal level of ERK1/2 activity by pitavastatin, ERK1/2 may not be a major player in regulating NF-κB activation in ECs.

Although we detected a significant increase in ERK5 TEY motif phosphorylation by pitavastatin and CQ, we could not detect visible band-shift of ERK5 by pitavastatin and CQ compared with steady laminar flow stimulation. As we reported previously, ERK5 band shift is not only caused by phosphorylation, but also by other posttranslational modification such as SUMOylation and ubiquitination and possibly by other molecules' interactions (2). As we showed in the study, we found that both pitavastatin and CQ increased ERK5 transactivation in an MEK5-independent manner. In addition, ERK5 has multiple phosphorylation sites besides the TEY motif (22, 45). These data suggest that both pitavastatin and CQ increase TEY motif phosphorylation but may not induce phosphorylation of other sites or other molecular association with ERK5, including MEK5 (1). Therefore, it is possible that the cause of band shift induced by CQ and pitavastatin is different from that induced by steady laminar flow (1).

When we overexpressed ERK5 WT and ERK5 TEF motif phosphorylation site mutant (ERK5-AEF), ERK5-AEF inhibited pitavastatin and CQ-induced ERK5 transcriptional activity (Supplemental Fig. 2D, 2E). The deletion of MEK5 inhibited pitavastatin-induced ERK5 TEY motif phosphorylation, but not ERK5 transcriptional activity (Supplemental Fig. 2A, 2B). Taken together, although the induction of ERK5 TEY motif phosphorylation by pitavastatin is not necessary for the pitavastatin-induced increase of ERK5 transcriptional activity, the basal level of ERK5 TEY motif phosphorylation is necessary for full activation of ERK5 transcriptional activity induced by pitavastatin.

We thank Drs. Cathy Tournier and Xin Wang for providing ERK5^flox/flox mice, Carolyn J. Giancursio, Hannah Cushman, Shane A. Cieri for technical support, and Yingqian Xu for preparation of HUVECs.

The authors have no financial conflicts of interest.