A novel small-molecule PCSK9 inhibitor E28362 ameliorates hyperlipidemia and atherosclerosis

Abstract

Proprotein convertase subtilisin/kexin type 9 (PCSK9) binds to the epidermal growth factor precursor homologous domain A (EGF-A) of low-density lipoprotein receptor (LDLR) in the liver and triggers the degradation of LDLR via the lysosomal pathway, consequently leading to an elevation in plasma LDL-C levels. Inhibiting PCSK9 prolongs the lifespan of LDLR and maintains cholesterol homeostasis in the body. Thus, PCSK9 is an innovative pharmacological target for treating hypercholesterolemia and atherosclerosis. In this study, we discovered that E28362 was a novel small-molecule PCSK9 inhibitor by conducting a virtual screening of a library containing 40,000 compounds. E28362 (5, 10, 20 μM) dose-dependently increased the protein levels of LDLR in both total protein and the membrane fraction in both HepG2 and AML12 cells, and enhanced the uptake of DiI-LDL in AML12 cells. MTT assay showed that E28362 up to 80 μM had no obvious toxicity in HepG2, AML12, and HEK293a cells. The effects of E28362 on hyperlipidemia and atherosclerosis were evaluated in three different animal models. In high-fat diet-fed golden hamsters, administration of E28362 (6.7, 20, 60 mg·kg⁻¹·d⁻¹, i.g.) for 4 weeks significantly reduced plasma total cholesterol (TC), triglyceride (TG), low-density lipoprotein-cholesterol (LDL-C) and PCSK9 levels, and reduced liver TC and TG contents. In Western diet-fed ApoE^−/− mice (20, 60 mg·kg⁻¹·d⁻¹, i.g.) and human PCSK9 D374Y overexpression mice (60 mg·kg⁻¹·d⁻¹, i.g.), administration of E28362 for 12 weeks significantly decreased plasma LDL-C levels and the area of atherosclerotic lesions in en face aortas and aortic roots. Moreover, E28362 significantly increased the protein expression level of LDLR in the liver. We revealed that E28362 selectively bound to PCSK9 in HepG2 and AML12 cells, blocked the interaction between LDLR and PCSK9, and induced the degradation of PCSK9 through the ubiquitin-proteasome pathway, which finally resulted in increased LDLR protein levels. In conclusion, E28362 can block the interaction between PCSK9 and LDLR, induce the degradation of PCSK9, increase LDLR protein levels, and alleviate hyperlipidemia and atherosclerosis in three distinct animal models, suggesting that E28362 is a promising lead compound for the treatment of hyperlipidemia and atherosclerosis.

Introduction

Atherosclerotic cardiovascular disease (ASCVD) remains the leading cause of global human mortality [1, 2]. Dyslipidemia plays an important role in the development of ASCVD [3]. Increased levels of low-density lipoprotein cholesterol (LDL-C) contribute to the occurrence and development of atherosclerosis [1, 4–6], and lowering plasma LDL-C levels is beneficial to the management of hypercholesterolemia and reduces cardiovascular risks and the incidence of cardiovascular events [7, 8]. Currently, statins are recommended as the primary agents for lowering LDL-C levels and preventing ASCVD by inhibiting 3-hydroxy-3-methyl glutaryl coenzyme A reductase (HMGCR) and increasing the expression of low-density lipoprotein receptor (LDLR) [9, 10]. However, a number of patients are intolerant to statins, leading to an inability to attain the suggested LDL-C target levels [11–13]. Therefore, there is an urgent need to develop new therapeutic strategies to treatatherosclerosis.

Proprotein convertase subtilisin/kexin type 9 (PCSK9) is a serine protease that has been identified as a powerful drug target for the treatment of hypercholesterolemia and atherosclerosis [14, 15]. PCSK9 loss-of-function mutations have lower levels of LDL-C and a reduced incidence of cardiovascular diseases (CVDs) [16]. PCSK9 binds to the epidermal growth factor precursor homologous domain A (EGF-A) of LDLR in the liver and triggers the degradation of LDLR via the lysosomal pathway, consequently leading to an elevation in plasma LDL-C levels [17]. Inhibiting PCSK9 prolongs the lifespan of LDLR and maintains cholesterol homeostasis in the body. PCSK9 monoclonal antibodies (Alirocumab and Evolocumab), PCSK9 siRNA (Inclisiran), and a cyclic peptide small-molecule drug named MK-0616 have been demonstrated to efficiently reduce LDL-C levels [18–21], but are associated with high costs. Therefore, disrupting the protein-protein interaction (PPI) between PCSK9 and LDLR presents an efficient and promising avenue for identifying PCSK9 inhibitors [22–25], and the development of new low-cost small-molecule inhibitors disrupting the PCSK9-LDLR interaction is a novel strategy to inhibit PCSK9 and alleviate atherosclerosis.

In this study, we conducted a virtual screening of a library containing 40,000 compound structures to identify potential PCSK9 inhibitors which were subsequently validated using cell-based LDL uptake assays. Compound E28362 (6-(2-hydroxypropyl)-4-(p-tolyl)-3,4,6,7-tetrahydro-1H-pyrrolo[3,4-d] pyrimidine-2,5-dione) was discovered as a novel small-molecule PCSK9 inhibitor. E28362 effectively binds to PCSK9, thereby disrupting the PCSK9-LDLR interaction and facilitating the degradation of PCSK9 via the ubiquitin-proteasome pathway, ultimately leading to increased LDLR protein levels. Notably, our in vivo data demonstrate that E28362 significantly ameliorates hyperlipidemia and atherosclerosis progression through its inhibitory effect on PCSK9 in three distinct experimental animal models. Taken together, our findings suggest that E28362 is a potential lead compound for the treatment of hyperlipidemia and atherosclerosis.

Materials and methods

In silico screening of the chemical library

In this study, BIOVIA Discovery Studio 2021 (Dassault Systèmes, MA, USA) was used to perform the virtual screening process. The crystal structure of PCSK9 from the Protein Data Bank (PDB: 3BPS) was used to construct the screen model. Amino acids 367–381 within PCSK9, where the EGF-A domain of LDLR binds, were considered to be the target of the inhibitors, and Arg194, Asp238, Thr377, and Phe379 were selected as the key amino acids to define the binding pocket in virtual screening. After the removal of the EGF-A domain, the PCSK9 structure was remodeled by removing water molecules and supplementing hydrogen atoms. A maximum of 10 docking poses were calculated for each chemical. The library containing 40,000 compounds used in this screening was a part of ChemDiv’s stock-available discovery libraries kept in our laboratory. The compounds were all converted to a three-dimensional structure minimized by the module “Prepare Ligands” before the screen. The top seven chemicals with the highest LibDockScore values were selected for further evaluation.

Compound E28362

Compound E28362 used in the in vitro and in vivo experiments was synthesized by our laboratory. E28362 is yellow oily at room temperature, soluble to 3 mM in phosphate-buffered saline (PBS) and 1 M in DMSO. The ESI-MS, ¹H NMR, and ¹³C NMR data of E28362 are provided in Supplementary Fig. 1.

Cell culture

Human hepatocellular carcinoma (HepG2) cells and alpha mouse liver 12 (AML12) cells were purchased from ATCC (Rockville, MD, USA). Human embryonic kidney 293a (HEK293a) cells were preserved in our laboratory. All cells were cultured in Dulbecco’s modified Eagle’s medium (C11995500BT, Gibco, NY, USA) supplemented with 10% (v/v) fetal bovine serum (FBS, 10270-106, Gibco, NY, USA) in an incubator at 37 °C and 5% CO₂.

Methyl thiazolyl tetrazolium (MTT) assay

HepG2, AML12, and HEK293a cells in the logarithmic growth phase were seeded at 1 × 10⁴ cells per well in 96-well plates and cultured for 6 h. E28362 was diluted to 1.25, 5, 10, 20, 40, 60, and 80 μM. The cells were treated with E28362 for another 24 h. The blank control groups were set simultaneously. Then the supernatant was discarded, new medium (100 µL) containing MTT reagent (10 µL) (M1020, Solarbio, Beijing, China) was added, and cells were incubated at 37 °C and 5% CO₂ for another 4 h. Then the supernatant was discarded, formazan reagent (110 µL) was added, and cells were incubated at room temperature on a low-speed shaker for 10 min to fully dissolve the crystals. Then the optical density was determined at 490 nm using a multimode plate reader (EnVision 2105, PerkinElmer, Fremont, CA, USA).

DiI-LDL uptake assay

AML12 cells grown in a 96-well plate were treated with E28362 (5, 10 and 20 μM) or vehicle in DMEM with or without PCSK9 protein (4 μg/mL) for 18 h and then incubated with 2 μg/mL DiI-LDL (LDL labeled with 1,10-dioctadecyl-3,3,3’,3’-tetramethylindocarbocyanine perchlorate (DiI)), L3482, Invitrogen, CA, USA) for 4 h at 37 °C. After incubation, the cells were washed three times with PBS containing 0.5% bovine serum albumin (BSA), and the nuclei were stained with diamidino phenylindole (DAPI) (D1306, Invitrogen, CA, USA). Fluorescence images were captured by a living cell high-connotation image analysis system (Opetta CCS, PerkinElmer, MA, USA).

Western blot (WB)

Hepatic cells or liver samples were lysed in RIPA lysis buffer (C1053, APPLGEN, Beijing, China) containing 1× phosphorylase (P1260, Solarbio, Beijing, China) and protease inhibitor cocktail (P6730, Solarbio, Beijing, China). The protein concentrations were determined by the BCA method (Pierce™ BCA Protein Assay Kits, 23225, Thermo, Waltham, MA, USA). The protein samples were separated by 10% SDS-PAGE and then transferred onto 0.45-μm polyvinylidene fluoride membranes (ISEQ00010, Millipore, Bedford, MA, USA), which were blocked with 5% skimmed milk, incubated with primary antibodies overnight at 4 °C, washed, incubated with secondary antibodies for 1 h at room temperature, and washed again. The primary antibodies used were as follows: anti-PCSK9 (55206-1-AP (Proteintech, Rosemont, IL, USA) and AF3888 (R&D Systems, MN, USA)), anti-LDLR (10785-1-AP, Proteintech, Rosemont, IL, USA), anti-GAPDH (60004-1-AP, Proteintech, Rosemont, IL, USA), and anti-β-actin (20536-1-AP, Proteintech, Rosemont, IL, USA). A hypersensitive ECL chemiluminescence detection system (P90719, Millipore, Billerica, MA, USA) was used to detect the protein bands, and ImageJ software was used for analysis. GAPDH or β-actin was used for normalization.

Membrane LDLR immunofluorescence staining

AML12 cells seeded on a 96-well black/clear bottom plate were treated with E28362 (5, 10, or 20 μM), SBC-115076 (PCSK9 inhibitor, 10 μM), Pep2-8 (PCSK9 inhibitor, 10 μM) or vehicle for 18 h. The cells were fixed in 4% paraformaldehyde (PFA) (BL539A, Biosharp, Hefei, China) for 15 min, blocked with 3% BSA for 30 min, and incubated with anti-LDLR (10785-1-AP, Proteintech, Rosemont, IL, USA) for 4 h. The cells were then washed and incubated with Alexa Fluor 488-conjugated goat anti-rabbit IgG (H + L) (R37117, Thermo Fisher Scientific, Waltham, MA, USA) for 1 h. The cell membrane was stained with DiI (C1991S, Beyotime Biotechnology, Beijing, China), a fat-soluble dye that can combine rapidly with the phospholipid bilayer, for 15 min at 37 °C. Samples were imaged using a living cell high-connotation image analysis system (Opetta CCS).

Surface plasmon resonance

The binding affinity and kinetics between E28362 and human PCSK9 were further analyzed using OpenSPR (Wayen Biotechnologies, Inc., Shanghai, China). PCSK9 (purity > 90%) was purchased from ACRO Biosystems Group (PC9-H5223, ACRO, Beijing, China). PCSK9 was dissolved in sodium acetate solution with a pH value of 4.5 and immobilized on the surface of a dextran chip (Reichert, NY, USA). To determine the binding affinity of PCSK9 and E28362, solutions of E28362 at different concentrations (5–200 μM) in PBST-D (PBST with 0.5% DMSO, pH 7.4) were injected into the immobilized PCSK9 chambers. Detection was performed using a Reichert 2 SPR system in a PBST-D running buffer. The injection time of the compound was 30 s and the dissociation time was 150 s. The results were analyzed using TraceDrawer (Reichert, NY, USA).

Cellular thermal shift assay

Cellular thermal shift assay is a widely used method to determine the intracellular binding strength of a drug to its target protein [26]. HepG2 cells cultured in DMEM supplemented with 10% FBS were collected and lysed in RIPA (C1053, APPLGEN, Beijing, China) supplemented with 1× phosphorylase (P1260, Solarbio, Beijing, China) and protease inhibitor cocktail (P6730, Solarbio, Beijing, China). After lysis on ice for 20 min, the lysate was centrifuged at 20,000 × g for 20 min at 4 °C. DMSO or E28362 (20 μM) was added to an equal volume of lysate, followed by incubation at 4 °C for 60 min on a rotator. Subsequently, the samples were mixed gently and transferred into 10 PCR tubes (50 μL for each). Every tube was incubated in a PCR thermocycler for 3 min at the indicated temperature and then at 25 °C for 3 min. After centrifugation at 20,000 × g for 20 min at 4 °C, 10 μL of 5× protein loading buffer was added to 40 μL of centrifugated supernatant and samples were boiled at 100 °C for 10 min. PCSK9 and LDLR protein levels were determined by SDS-PAGE followed by WB. To determine the effects of different concentrations of E28362 on the protein stability of PCSK9, the lysate was incubated with gradient E28362 or an equal amount of DMSO at 4 °C for 1 h and incubated for 3 min at 57 °C and for 3 min at 25 °C. After centrifugation at 20,000 × g for 5 min, the supernatant was analyzed by SDS-PAGE and WB.

Bimolecular fluorescence complementation

The pBiFC-VN173 and pBiFC-VC155 plasmids were purchased from Addgene (#22010, #22011) [27]. The sequence encoding human PCSK9 lacking the 30-amino-acid signal peptide (31–692 aa) and a sequence encoding a FLAG-tag was cloned into pBiFC-VN173 (named pBiFC-VN173-flag PCSK9). A sequence encoding the human LDLR signal peptide and the first four amino acids of mature LDLR (1–25 aa) was cloned into pBiFC-VC155 before the C-terminal fragment of Venus [28], an improved version of yellow fluorescent proteins, and the sequence encoding LDLR lacking the signal peptide (22–861 aa) was cloned after Venus with an elastic linker (GGGGSGGGGS) and a C-terminal His-tag (named pBiFC-VC155-his LDLR).

These two BiFC plasmids (pBiFC-VN173-flag PCSK9 and pBiFC-VC155-his LDLR) were co-transfected into HEK293a cells for 6 h and then the medium was replaced with DMEM containing 10% FBS for 18 h. The cell culture medium was then changed to DMEM containing 10% FBS and Hoechst (a live cell nuclear dye) and the indicated concentration of E28362, followed by incubation for another 18 h. The cells were washed with PBS two times, and images were acquired by a living cell high-connotation image analysis system (Opetta CCS).

Protein stability and degradation analysis

Cycloheximide (CHX) is an antifungal antibiotic that can inhibit protein synthesis in eukaryotes by interfering with translocation during protein synthesis. MG-132 is an effective, reversible, and permeable inhibitor of the 26 S proteasome, which can inhibit proteasome chymotrypsin-like peptidase activity. HepG2 cells were treated with CHX (10 μg/mL) (66-81-9, Sigma-Aldrich, MO, USA) or MG132 (10 μM) (E2899, Selleck Chemicals, TX, USA) for 24 h and E28362 (10 μM) or an equal amount of DMSO for the indicated periods (0, 6, 9, 12, and 24 h). Total protein was extracted to analyze the effects on PCSK9 and LDLR protein degradation by WB.

For immunoprecipitation (IP), HepG2 cells were treated with (CHX, 10 μg/mL), MG132 (10 μM), and E28362 (10 μM) or DMSO for 18 h. The cells were collected and lysed in WB and IP cell lysis buffer (P0013, Beyotime, Shanghai, China) supplemented with protease inhibitors on ice for 20 min and centrifuged at 14,000 × g for 20 min at 4 °C. Then the cell lysates were incubated with anti-PCSK9 antibody for 24 h. The clarified lysates were incubated with Bioepitope^® protein A + G agarose IP beads (BD0048, Bioworld, Nanjing, China) at 4 °C for 2 h. After incubation, immune complexes were washed three times with wash buffer (PBST containing 100 μM PMSF, pH 7.4), and the bead-conjugated proteins were denatured in 1 × SDS loading buffer for 10 min at 100 °C. Samples were analyzed by SDS-PAGE and WB.

Animals

Thirty male golden hamsters about 7 weeks old were purchased from the Beijing Vital River Laboratory Animal Technology Co., Ltd. (Beijing, China). Seven-week-old male apolipoprotein E knockout (ApoE^−/−) mice and C57BL/6 J mice were purchased from the SpePharm (Beijing) Biotechnology Co., Ltd. (Beijing, China). All experimental procedures were approved (IMB–20210301D₃01, IMB–20210604D₃01, and IMB–20221013D₃01) and strictly followed the requirements of the Laboratory Animal Care and Use Committee of the Institute of Medicinal Biotechnology, Chinese Academy of Medical Sciences & Peking Union Medical College. All animals were housed at 23 °C with a 12/12-h light/dark cycle in a specific pathogen-free environment with free access to food and water. Investigators were blinded to the treatment conditions and all in vivo data analyses were performed by a blinded investigator.

Dyslipidemia golden hamsters treatment

After adaptive feeding for a week, thirty male golden hamsters were randomly divided into five groups (n = 6 golden hamsters per group), including the normal diet group (ND group), the high-fat diet (HFD) and solvent group (HFD group), the HFD and 6.7 mg/kg E28362 group (E28362-L group), the HFD and 20 mg/kg E28362 group (E28362-M group), and the HFD and 60 mg/kg E28362 group (E28362-H group). The ND group was fed a ND (SPF-F02-001, SpePharm, Beijing, China), and the other four groups were fed HFD (TP4H100 [20% fat, 40% fructose, 0.25% cholesterol], Trophic Animal Feed High-tech Co., Ltd., Nantong, China) for 2 weeks to induce hyperlipidemia, and the hamsters were then given E28362 (dissolved in 0.5% sodium carboxymethyl cellulose [CMC-Na]) or the same dosage of solvent (0.5% CMC-Na) once a day for another 4 weeks.

ApoE^−/− mice treatment

After adaptive feeding for a week, ApoE^−/− mice were randomly divided into three groups (n = 10 per group): the Western diet group (WD group), the E28362 low-dose group (20 mg/kg, E28362-L group), and the E28362 high-dose group (60 mg/kg, E28362-H group). C57BL/6 J mice fed a maintenance diet were used as the control group (n = 6). All ApoE^−/− mice were fed WD (TP26300 [21% fat, 0.2% cholesterol, 49.1% carbohydrate, 19.8% protein], Trophic Animal Feed High-tech Co., Ltd., Nantong, China) for 12 weeks. Meanwhile, the mice were intragastrically administered with E28362 or the same dosage of solvent (0.5% CMC-Na) for 12 weeks.

PCSK9 overexpression mice treatment

The AAV8-hPCSK9 D374Y overexpression and AAV8-Control viruses were constructed by OBiO Technology (Shanghai, China). After adaptive feeding for a week, thirty 6-week-old male mice were randomly divided into three groups (n = 10 per group), including the control, model, and E28362 (60 mg/kg, dissolved in 0.5% CMC-Na) groups. Mice in the model and E28362 treatment groups were injected once with AAV8-hPCSK9 D374Y overexpression virus (1 × 10¹¹ VG per mouse), and control mice were injected once with AAV8-Control virus (1 × 10¹¹ VG per mouse) via the tail vein. All mice were fed a WD (TP26300) for 12 weeks after one week of adaptation to tail vein injection. Meanwhile, the mice were intragastrically administered with E28362 or the same dosage of solvent (0.5% CMC-Na) once a day for 12 weeks.

Tissue harvest

Body weight was measured every week during the in vivo experiments. At the end of the experiments, the animals were fasted for 12 h and euthanized under isoflurane anesthesia. The livers were weighed and the liver index (liver index = liver mass / body mass × 100%) was calculated. Blood samples were collected into heparinized tubes and centrifuged at 3000 rpm for 20 min to prepare plasma, which was stored at −80 °C. Liver samples were immediately put into liquid nitrogen and stored at −80 °C until analysis of plasma lipid levels and protein extraction. The livers, hearts, and full-length aortas of mice were collected and fixed in 4% PFA to make sections.

Blood biochemistry analysis

The plasma total cholesterol (TC), triglyceride (TG), LDL-C, high-density lipoprotein cholesterol (HDL-C), alanine aminotransferase (ALT), aspartate transaminase (AST), creatinine (Cre), and urea levels were determined by an automatic blood biochemical analyzer (Hitachi 71800, Chiyoda, Japan) according to the manufacturer’s instructions.

PCSK9 ELISA

The PCSK9 content in cell supernatant was measured using a commercial ELISA kit (SBJ-R1123-96T, SBJ bio, Nanjing, China). The PCSK9 content in golden hamsters’ plasma was measured using a commercial ELISA kit (SBJ-M1191-96T, SBJ Bio), and PCSK9 levels in mouse plasma were measured using a commercial ELISA kit (ab215538, Abcam, Cambridge, MA, USA) according to the manufacturer’s instructions.

Fast protein liquid chromatography (FPLC) assay

FPLC was performed as previously reported [29]. Briefly, a total of 350 μL pooled plasma from each group was injected into a Superose 6 Increase 10/300 GL column (GE Healthcare), followed by elution using elution buffer (8.75 g/L NaCl, 5.68 g/L Na₂HPO₄, 0.02% [v/v] 0.5 M EDTA) at a flow rate of 0.3 mL/min using ÄKTA explorer (GE Healthcare). The eluent was collected into about 45 fractions (500 μL each fraction) in tubes. The cholesterol amount in each tube was determined using a total cholesterol test kit (7020, Beijing Leadman Biochemistry Co., Ltd., Beijing, China).

Lesion assessment

Atherosclerotic lesion size was analyzed as previously described [30]. Briefly, the aortas fixed in 4% PFA were washed with PBS and then the adventitial fat was removed. The en face aortas were split longitudinally, and Oil red O (ORO) (MA0120-3, Meilun, Dalian, China) staining and microscopic analysis were conducted using a Leica S8 APO microscope (Wetzlar, Germany) and a digital camera (Sony) [30]. ImageJ was used to assess the ratio of atherosclerotic lesion area to total aortic area (plaque area / total area × 100%).

Formalin-fixed hearts were embedded in an OCT embedding compound (Tissue-Tek, Torrance, CA, USA) and the aortic roots were cut into 8-μm sections. The frozen cross-sections of aortic root tissues were permeabilized with 60% isopropanol and then stained with ORO working solution for 30 min. After washing three times with water, sections were incubated with hematoxylin to stain nuclei. Images were acquired using a Leica microscope (DM2500, Leica, Wetzlar, Germany). The lesions on the stained aortic root cross-sections were analyzed by ImageJ, and the ratio of atherosclerotic lesion area to total aortic area (plaque area / total area × 100%) was calculated.

The Verhöeff-van Gieson (VVG) staining experiments were performed as described in the supplementary methods. The necrotic core areas on the stained aortic root cross-sections were analyzed by ImageJ. The immunofluorence staining for CD68 in the aortic root frozen sections was performed as described in the supplementary methods. The images were captured using a slice scanner (Pannoramic MIDI, 3DHISTECH Ltd., Budapest, Hungary) and then analyzed by ImageJ software.

Histological analysis of liver

Formalin-fixed livers were embedded in paraffin and then cut into 8-μm sections. The liver sections were then stained with hematoxylin and eosin (H&E). Images were captured using a slice scanner to analyze the structure (Pannoramic MIDI).

Formalin-fixed livers were embedded in OCT and cut into 8-μm sections for ORO staining as previously described [30]. The frozen sections were stained with ORO working solution and hematoxylin as described for aortic root tissues. Images were acquired using a slice scanner (Pannoramic MIDI) to analyze intracellular lipid droplets.

Liver lipid content assay

Briefly, 25 mg of liver tissue was homogenized in 1 mL lysis buffer by a GeneReady UltraCool sample preparation system (BSH-C2, Hangzhou LifeReal Biotechnology Co., Ltd., Hangzhou, China). TC and TG contents in the livers were quantified using commercial kits (E1015, E1013, Applygen Technologies, Inc, Beijing, China) and normalized to the weight of the livers.

Statistical analyses

All data are presented as mean ± SEM. For in vitro studies, at least three independent experiments were performed. The two-tailed unpaired Student t-test was used for comparisons between the two groups. One-way analysis of variance (ANOVA) with Dunnett’s multiple comparison test was used to compare three or more groups. GraphPad Prism 8 was used to analyze the data. P < 0.05 was considered significant.

Results

E28362 is a PCSK9 inhibitor and increases the uptake of LDL in AML12 cells

To obtain small-molecule PCSK9 inhibitors, a binding pocket was set on the surface of PCSK9 (PDB: 3BPS) where it interacts with the EGF-A domain of LDLR, and a virtual screening was performed to screen PCSK9 inhibitors from a library containing 40,000 compound structures. Seven hits were selected from the compound library. Among these seven hits, E28362 (6-(2-hydroxypropyl)-4-(p-tolyl)-3,4,6,7-tetrahydro-1H-pyrrolo[3,4-d]pyrimidine-2,5-dione, C₁₆H₁₉O₃N₃, MW: 301.3 Da) (Fig. 1a) may form intermolecular hydrogen bonds with Arg194, Thr377, and Phe379 of PCSK9, and there was a π-π interaction between the tolyl dihydropyrimidine ketone of E28362 and the phenyl group of Phe379 of PCSK9 (Fig. 1b–d).

Fig. 1. Identification of novel PCSK9 inhibitors using virtual screening assays.

a The structure of E28362 (C₁₆H₁₉O₃N₃). b The E28362 binding pocket of PCSK9. c The key amino acids of PCSK9 interacting with E28362. d The binding mode of E28362 was predicted with induced-fit docking. e The cytotoxic effects of E28362 on AML12 cells after incubation with E28362 (0–80 μM) for 24 h. f, g E28362 increased the uptake of DiI-LDL in AML12 cells. AML12 cells were treated with DMSO or E28362 (5, 10 or 20 μM) in DMEM with or without human PCSK9 protein (4 μg/mL) for 18 h, and then the medium was replaced with DMEM containing 2 μg/mL DiI-LDL for another 4 h. Nuclei were stained with Hoechst. Representative images are shown (f). Scale bars: 100 μm. The average fluorescence intensity of DiI-LDL (g) in AML12 cells was calculated. Values are presented as mean ± SEM. Statistical significance was calculated using the unpaired two-tailed Student’s t-test. ^∗∗∗P < 0.001 vs E28362 (0 μM) + DiI-LDL (2 μg/mL) group; ^{^}P < 0.05, ^{^^}P < 0.01 vs E28362 (0 µM) + PCSK9 (4 μg/mL) + DiI-LDL (2 μg/mL) group; ^###P < 0.001 vs E28362 (0 μM) group; ^$$$P < 0.001.

The function of LDLR is to re-uptake LDL in the liver [31]. By analyzing the effects of the screened compounds on LDL uptake, we found that E28362 (5, 10, 20 μM) could significantly increase the intake of DiI-LDL in AML12 cells, and the increased DiI-LDL uptake effect of E28362 was blunted when PCSK9 protein was co-incubated with E28362 (Fig. 1f, g). In addition, MTT assay showed that E28362 from 0–80 μM had no obvious toxicity in AML12 cells (Fig. 1e).

E28362 increases the LDLR protein level in HepG2 and AML12 cells

PCSK9 can bind to LDLR and mediate its degradation, and the main function of PCSK9 inhibitors is to increase the LDLR protein level. Therefore, we examined the LDLR protein levels in hepatocytes after E28362 treatment. As shown in Fig. 2a, b, E28362 at 5, 10, and 20 μM significantly increased the total protein level of LDLR in both HepG2 and AML12 cells. E28362 treated for 9 h, 12 h, and 24 h could also increase the total protein level of LDLR in HepG2 cells (Fig. 2c).

Fig. 2. E28362 increases LDLR protein expression in hepatocytes.

The total LDLR protein level was measured by WB (a–c) and the LDLR protein level in the membrane (d, e) was determined by immunofluorescence staining. a, b HepG2 (a) and AML12 (b) cells were treated with DMSO or E28362 at 0, 5, 10, and 20 μM for 18 h. c HepG2 cells were treated with E28362 (10 μM) for 0, 3, 6, 9, 12, 24 h. d, e HepG2 cells were treated with E28362 (0, 5, 10, and 20 μM), SBC-115076 (10 μM), or Pep2-8 (10 μM) for 18 h and then incubated with Alexa Fluor 488-conjugated goat anti-rabbit IgG (H + L). The blank group cells were only treated with DMSO. d Representative immunoblot and images are shown. Green, membrane LDLR proteins stained with Alexa Fluor 488-conjugated goat anti-rabbit IgG; red, membranes stained with DiI; blue, nuclei stained with DAPI. Scale bars: 100 μm. e The average fluorescence intensity of LDLR was calculated by ImageJ. f The cytotoxic effects of E28362 (0–80 μM) on HepG2 cells after incubation for 24 h. Values are presented as means ± SEM. Statistical significance was calculated using an unpaired 2-tailed Student’s t-test. ^*P < 0.05, ^**P < 0.01, ^***P < 0.001 vs E28362 (0 μM).

We further detected the LDLR protein in the membrane of AML12 cells after treatment with E28362 by immunofluorescence. The results showed that the E28362 treatment significantly increased the LDLR protein level in the membrane of AML12 cells (Fig. 2d, e), and the increased membrane LDLR protein effect of E28362 (10 μM) was equivalent to the small molecular PCSK9 inhibitor SBC-115076 (10 μM) and PCSK9 peptide inhibitor Pep2-8 (10 μM) (Fig. 2d, e). MTT assay showed that E28362 in 0–80 μM had no obvious toxicity in AML12 cells (Fig. 2f). Taken together, these results suggest that E28362 can increase the LDLR protein level and the DiI-LDL uptake in hepatocytes.

E28362 improves dyslipidemia and decreases hepatic lipid accumulation in HFD-induced golden hamsters

To investigate the effects of E28362 on hyperlipidemia in vivo, we first established an HFD-induced hyperlipidemia model in golden hamsters, whose lipid metabolism resembles that of humans more closely. Golden hamsters were fed HFD for 2 weeks to induce hyperlipidemia, and then E28362 was administered for 4 weeks (Fig. 3a). There was no significant difference in body weight (Fig. 3b and Supplementary Fig. 2a), liver weight (Supplementary Fig. 2b), or liver index (Supplementary Fig. 2c) between the E28362 groups and the HFD group at the endpoint. Compared with the ND group, the TC, TG, and LDL-C levels were significantly higher in the HFD group (Fig. 3c–e). Compared with the HFD group, plasma TC (by 21.50%, 23.22%, and 28.97%) (Fig. 3c), TG (by 53.51%, 52.59%, and 47.10%) (Fig. 3d), and LDL-C (by 16.97%, 43.64%, and 23.05%) levels (Fig. 3e) were significantly decreased in the E28362-L, E28362-M, and E28362-H groups, respectively. FPLC experiment also demonstrated that E28362-L and E28362-H treated groups visually reduced the level of LDL-C in golden hamster plasma when compared with HFD group (Fig. 3f). In addition, E28362-H group hamsters had lower plasma ALT levels compared with the HFD group (Fig. 3g). There were no visible differences in AST (Fig. 3h), Cre (Fig. 3i), and urea (Fig. 3j) levels between the E28362 treated groups and the HFD group. These data indicate that E28362 could improve dyslipidemia in golden hamsters without exerting toxic effects.

Fig. 3. E28362 improves the plasma profile and alleviates hepatic lipid accumulation in golden hamsters.

Male golden hamsters were intragastrically administered with vehicle or E28362 (6.7, 20, or 60 mg/kg per day) for 4 weeks (n = 6 per group). a Schematic diagram of the experimental design and treatment. b The body weight of golden hamsters at the end of the experiment. c–e Plasma TC (c), TG (d), and LDL-C (e) levels. f Plasma was pooled per group and the distribution of cholesterol over the individual lipoproteins was analyzed after separation by fast protein liquid chromatography (FPLC). g–j Plasma ALT (g), AST (h), Cre (i), and urea (j) levels. k Representative pictures of macroscopic appearance (Scale bars: 1 cm), hematoxylin and eosin (H&E) staining (Scale bars: 50 μm), and Oil red O (ORO) staining (Scale bars: 50 μm). l, m Hepatic TC and TG contents. n, o The hepatic protein expression of LDLR was determined by WB. Representative images are shown. Protein levels were normalized to β-actin. p Plasma PCSK9 content was measured by a commercial ELISA kit. Values are presented as means ± SEM. n = 6 per group. Blue, red, green, purple and orange-yellow colors represented ND, HFD, E28362-L, E28362-M, and E28362-H, respectively. Statistical significance was calculated with one-way ANOVA. ^#P < 0.05, ^###P < 0.001 vs ND group; ^*P < 0.05, ^**P < 0.01, ^***P < 0.001 vs HFD group.

Then the effects of E28362 on the liver were investigated. As shown in Fig. 3k, based on the macroscopic appearance and morphology of the livers, hamsters in the HFD group developed fatty liver, while livers in the E28362 groups were in a healthier state. H&E staining revealed obvious hepatocyte damage in the HFD group (Fig. 3k); E28362 significantly reduced the HFD-induced hepatic steatosis and ballooning in livers (Fig. 3k). ORO staining revealed obvious lipid accumulation in the HFD group (Fig. 3k); E28362 significantly reduced hepatic lipid droplet accumulation when compared with HFD group (Fig. 3k). In addition, liver TC (Fig. 3l) contents were also significantly decreased in the E28362 groups compared with the HFD group; liver TG (Fig. 3m) contents were also significantly decreased in the E28362-M group compared with the HFD group. These data indicate that E28362 improved hepatocyte damage and reduced lipid droplet accumulation in the liver.

Furthermore, the LDLR protein levels in livers were all significantly increased in the three E28362-treated groups compared with the HFD group (Fig. 3n, o), which is consistent with the in vitro results. Interestingly, plasma PCSK9 protein levels were also significantly lower in the E28362-H group compared with the HFD group (Fig. 3p) suggesting that E28362 might affect the secretion of PCSK9. In summary, E28362 reduced the plasma PCSK9 level and increased the liver LDLR protein level, thus playing a lipid-lowering role in hyperlipidemia golden hamsters.

E28362 alleviates atherogenesis in WD-induced ApoE^−/− mice

After confirming the beneficial effects of E28362 on dyslipidemia in the golden hamster model, we investigated whether E28362 has anti-atherosclerotic effects in ApoE^−/− mice. Atherosclerosis-prone ApoE^−/− mice were fed WD and simultaneously administered with vehicle (WD group), 30 mg/kg E28362 (E28362-L group), or 60 mg/kg E28362 (E28362-H group) for 12 weeks (Supplementary Fig. 3a). As shown in Supplementary Fig. 3b, there was no significant difference in body weight between the E28362 groups and the WD group at the start and endpoint.

As illustrated in Fig. 4a, plasma TC levels of mice in the HFD group were significantly higher than those in the ND group (Fig. 4a); E28362 had no significant effects on plasma TC and TG levels compared with the WD group (Fig. 4a, b). Notably, FPLC analysis results showed that the very-low-density lipoprotein–cholesterol (VLDL-C) and LDL-C levels in the plasma in both the E28362-L and E28362-H groups were decreased compared with those in the WD group (Fig. 4c). In addition, the levels of AST (Supplementary Fig. 3c), ALT (Supplementary Fig. 3d), Cre (Supplementary Fig. 3e), and urea (Supplementary Fig. 3f) were not visibly different between the E28362 groups and the WD group.

Fig. 4. E28362 inhibits the development of atherosclerosis in ApoE^−/− mice.

Male ApoE^−/− mice were intragastrically administered with vehicle or E28362 (30 or 60 mg/kg per day) for 12 weeks. a, b Plasma TC (a) and TG (b) levels were measured (n = 6 for ND group; n = 10 for WD, E28362-L, and E28362-H group). c Plasma was pooled per group and the distribution of cholesterol over the individual lipoproteins was analyzed after separation by FPLC (n = 6 for ND group; n = 10 for WD, E28362-L, and E28362-H group). d The plasma PCSK9 content was measured by a commercial ELISA kit (n = 6 for ND group; n = 10 for WD, E28362-L, and E28362-H group). e Representative gross lesions of the aortic arch. f, g Representative ORO-stained images of the full length of the aorta (f) and quantification of the percentage of the plaque area vs. the area of the full-length aorta (g) (n = 10 per group). h, i Representative image of the ORO-stained sections from the aortic root (h) and quantification of aortic root lesion areas (n = 5-6 for per group) (i). j, k Hepatic TC and TG contents (n = 6 for ND group; n = 10 for WD, E28362-L, and E28362-H group). l–n The hepatic protein expression of PCSK9 and LDLR was determined by WB (n = 6 for ND group; n = 10 for WD, E28362-L, and E28362-H group). Representative images are shown. Protein levels were normalized to GAPDH. Values are presented as means ± SEM. For all data, blue, red, green and purple lines showed the fractions of ND, WD, E28362-L, and E28362-H, respectively. Statistical significance was calculated with one-way ANOVA. ^#P < 0.05, ^###P < 0.001 vs ND group; ^*P < 0.05, ^**P < 0.01, ^***P < 0.001 vs WD group.

Notably, E28362 treatment resulted in smaller plaque areas in both the en face aortas and the aortic root compared to the WD group (Fig. 4e–i). ORO staining results showed that the plaque area in en face aortas was significantly reduced in both the E28362-L (3.89% ± 0.46% vs 10.42% ± 0.44%, P < 0.001) and E28362-H (2.42% ± 0.28% vs 10.42% ± 0.44%, P < 0.001) groups compared with the WD group (Fig. 4f, g). The plaque area and amount of atherosclerotic plaque quantified from both ORO-stained aortic root cross-sections were also obviously reduced in E28362-L group (50.60% ± 11.57%, P < 0.001) and E28362-H group (63.55% ± 6.43%, P < 0.01) compared with WD group (Fig. 4h, i). CD68 staining in atherosclerotic plaques in aortic root cross-sections showed that there were less macrophages infiltration in E28362 treated groups than the WD group (Supplementary Fig. 3h). VVG staining results showed that E28362 treatments significantly reduced the necrotic core area in comparison with the WD group (Supplementary Fig. 3i), which is consistent with the results of atherosclerotic plaques in the aortic root (Fig. 4h). In addition, the TC content in the liver was significantly decreased after treatment with 60 mg/kg E28362 (Fig. 4j), while the TG content in the liver was significantly increased in E28362 groups compared with the WD group (Fig. 4k).

To investigate the potential mechanisms by which E28362 prevents atherosclerosis and reduces plasma VLDL-C and LDL-C levels, we measured the plasma levels of PCSK9 and the protein expression levels of LDLR and PCSK9. As shown in Fig. 4d, there was no significant difference in the plasma PCSK9 levels between E28362 and WD group. Interestingly, WB results showed that E28362 significantly increased hepatic LDLR protein level and decreased PCSK9 protein level (Fig. 4l–n). Taken together, these data demonstrate that E28362 could retard atherosclerosis development in ApoE^−/− mice.

E28362 binds to PCSK9 and blocks the interaction between PCSK9 and LDLR

We then aimed to analyze how E28362 exerts anti-atherosclerotic effects in vitro. Since E28362 was identified by a virtual screening method to aim to disrupt the PPI between PCSK9 and the EGF-A domain of LDLR, we investigated whether E28362 can bind to PCSK9 by surface plasmon resonance, isothermal titration calorimetry, and cellular thermal shift assays, respectively. As shown in Fig. 5a, our surface plasmon resonance assay results showed that E28362 could bind to PCSK9 at concentrations of 5 to 200 μM with a K_D value of 14.8 μM, indicating that E28362 and PCSK9 have a moderately strong bond. Isothermal titration calorimetry assay results showed that E28362 could bind to human PCSK9 protein (Supplementary Fig. 4, 5) with a dissociation constant K_D value of 6.84 ± 4.23 μM (Supplementary Fig. 5). Cellular thermal shift assay is another technique to investigate the binding ability of compounds and proteins that resembles the normal physiological state of cells more closely [26]. The basic principle of cellular thermal shift assays is that high temperatures cause protein denaturation and precipitation; however, proteins bound to a molecule can be more stable and withstand higher temperatures. Here, we specifically investigated the ability of E28362 to bind to PCSK9 by changing the temperature or the compound concentrations. As shown in Fig. 5b, when the temperature was fixed, more PCSK9 protein was retained as the concentration of E28362 was increased, and when the concentration of E28362 was fixed, PCSK9 was degraded more slowly as the increasing temperature (Fig. 5c). LDLR did not exhibit the similar behaviors (Fig. 5b, c). These results indicate that E28362 exclusively binds to PCSK9 rather than LDLR.

Fig. 5. E28362 binds to PCSK9 and blocks the interaction between PCSK9 and LDLR.

a Surface plasmon resonance sensorgram for the interaction between E28362 and human PCSK9 protein. The dissociation constant (K_D) value is shown. b The amount of PCSK9 rather than LDLR retained increased with the increase of E28362 concentration in the cellular thermal shift assay. Protein band intensities were quantified by ImageJ. c E28362 improved the thermal stability of PCSK9 rather than LDLR in the cellular thermal shift assay. Protein band intensities were quantified by ImageJ. Blot intensities were normalized to the intensity obtained for the 4 °C sample. Values are presented as means ± SEM. Statistical significance was calculated using an unpaired 2-tailed Student’s t-test. b ^*P < 0.05 vs E28362 (0 μM). c ^∗P < 0.05, E28362 (20 μM) at 59 °C vs DMSO at 59 °C. d Schematic diagram of the PCSK9-LDLR BiFC model. e HEK293a cells were co-transfected with or without pBiFC-VC155-his LDLR and pBiFC-VN173-flag PCSK9 plasmids, and then the transfected cells were treated with DMSO or E28362 (2.5, 5, 10, and 20 μM) for 18 h. The average fluorescence intensity of Venus was calculated. f The cytotoxic effects of E28362 (0–80 μM) on HEK293a cells after incubation for 24 h. e, f Values are presented as means ± SEM. Statistical significance was calculated using an unpaired 2-tailed Student’s t-test. ^∗∗∗P < 0.0001 vs E28362 (0 μM).

To investigate whether E28362 binding to PCSK9 affects its interaction with LDLR, we constructed a cell-based BiFC model to examine whether E28362 can block the PCSK9/LDLR interaction. PCSK9 and LDLR were expressed in fusion with two different parts (truncated at the 155th and 177th amino acids) of the Venus fluorescent protein (Fig. 5d). The interaction of PCSK9 and LDLR will bring the two parts of the Venus protein together to restore the fluorescence signal (Fig. 5e). If the compound can block the PCSK9/LDLR interaction, the fluorescence signal will be attenuated. The results showed significant fluorescence in the vehicle-treated PCSK9-LDLR group compared with the negative control group (Fig. 5e). E28362 at 2.5, 5, 10, and 20 μM could significantly attenuate the fluorescence signals compared with E28362 at 0 μM (Fig. 5e). MTT assay showed that E28362 from 0–80 μM had no obvious toxicity in HEK293a cells (Fig. 5f). These results indicate that E28362 can block the interaction between LDLR and PCSK9 after binding to PCSK9.

E28362 induces PCSK9 degradation through the ubiquitin-proteasome pathway after binding to PCSK9 in hepatic cells

As E28362 reduces PCSK9 levels in vivo (Fig. 4l) and robustly increases LDLR protein levels in vitro (Fig. 2a–e) and in vivo (Figs. 3n, 4l), we examined the PCSK9 protein expression levels in HepG2 and AML12 cells after treatment with E28362. As shown in Fig. 6a, E28362 at 10 μM and 20 μM visibly decreased PCSK9 protein levels in both HepG2 and AML12 cells. E28362 (10 μM) treated for 3 h or more significantly reduced the level of PCSK9 protein in HepG2 cells (Fig. 6b). We then examined whether E28362 treatment affects the PCSK9 levels secreted into the media. The ELISA results showed that E28362 significantly decreased the amount of PCSK9 secreted by HepG2 cells (Fig. 6c).

Fig. 6. E28362 induces PCSK9 degradation through the ubiquitin-proteasome pathway.

a E28362 decreased the PCSK9 protein level in HepG2 and AML12 cells. HepG2 or AML12 cells were treated with E28362 (0, 5, 10, and 20 μM) for 18 h. ^*P < 0.05, ^**P < 0.01 vs E28362 (0 μM). b E28362 (10 μM) decreased the PCSK9 protein level when treated for the indicated times in HepG2 cells. ^*P < 0.05, ^**P < 0.01, ^***P < 0.001 vs E28362 treated duration (0 h). c E28362 decreased the secretion of PCSK9 protein into the media in HepG2 cells. ^**P < 0.01, ^***P < 0.001 vs E28362 (0 μM) (d, e) HepG2 cells were treated with or without DSMO, cycloheximide (CHX, 10 μg/mL), MG132 (10 μM), and E28362 (10 μM) for the indicated periods. ^∗P < 0.05, ^∗∗P < 0.01 vs HepG2 treated with DMSO for the indicated time when CHX was present. f HepG2 cells were treated with (CHX, 10 μg/mL), MG132 (10 μM), and E28362 (10 μM) or DMSO for 18 h and lysed, and then the cell lysates were incubated with anti-PCSK9 for 24 h. After that, the mixtures were incubated with A + G agarose immunoprecipitation (IP) beads for 2 h. Samples were then separated and proteins were detected by WB. For all data, values are presented as means ± SEM; statistical significance was calculated using an unpaired 2-tailed Student’s t-test.

Inspired by studies related to the degradation of target proteins caused by small-molecule compounds [32], we investigated whether E28362 can affect the protein degradation of PCSK9 after binding. CHX is the most commonly used reagent for inhibiting protein synthesis in laboratory settings. We used CHX to block protein synthesis in HepG2 cells (Supplementary Fig. 6) and then treated the cells with E28362 for different time periods to check whether the protein level would be affected (Fig. 6d, e). The experimental results indicate that E28362 reduced the level of PCSK9 protein and correspondingly increased the level of LDLR protein (Fig. 6d). This result suggests that E28362 affects PCSK9 during the post-translation process.

Since the ubiquitin-proteasome pathway is an important pathway for protein degradation [33], we analyzed whether E28362 mediates the degradation of PCSK9 through the ubiquitin-proteasome pathway. MG132 inhibits the activity of the 26 S proteasome by covalently binding to the active site of the β subunit and effectively blocks the proteolytic activity of the 26 S proteasome complex, which can inhibit the activity of different types of proteases such as serine protease and calpain [34]. Our data showed that the E28362-induced degradation of PCSK9 was visibly blocked when proteasome activity was inhibited by MG132 (Fig. 6d, e), indicating that it induces PCSK9 degradation through the ubiquitin-proteasome pathway. An IP assay was then performed to confirm the effect of E28362-mediated ubiquitination on PCSK9 protein levels. After treating cells with MG132 and E28362 or DMSO, the cell lysates were immunoprecipitated with an anti-PCSK9 antibody, followed by WB. E28362 promoted the ubiquitination of PCSK9 compared with DMSO (Fig. 6f). Taken together, these results illustrate that E28362 could bind to PCSK9 to induce PCSK9 degradation through the ubiquitin-proteasome pathway and thus downregulate PCSK9 and decrease its secretion, which finally leads to the presence of more LDLR to take up LDL.

E28362 alleviates the development of atherosclerosis in PCSK9 D374Y overexpression mice

To further confirm that the effect of E28362 in vivo comes from the inhibition of PCSK9, we constructed an atherosclerosis model by overexpressing PCSK9 in C57BL/6 J mice (Supplementary Fig. 7a) and analyzed whether the change in PCSK9 levels is responsible for the pathological state. As shown in Supplementary Fig. 7b, there was no significant difference in body weights between the E28362 groups and the model group at the start and end point. As shown in Fig. 7a–c, mice overexpressing PCSK9 had an obvious increase in TC, TG, and LDL-C levels compared with the control group. E28362 significantly decreased the plasma TC (Fig. 7a), TG (Fig. 7b), and LDL-C (Fig. 7c) levels compared with the model group. FPLC analysis also demonstrated that E28362 significantly decreased the plasma cholesterol levels in VLDL and LDL particles in the separated fractions (Fig. 7d). Importantly, E28362 significantly reduced the en face aortic and aortic arch plaque areas in PCSK9 overexpression mice compared with the model group (Fig. 7i–m). In addition, the TC and TG contents in the liver (Fig. 7n, o) and the plasma levels of AST, ALT, Cre, and urea (Fig. 7e–h) were not affected by the E28362 treatment when compared with the model group.

Fig. 7. E28362 decreased PCSK9 levels and alleviated the development of atherosclerosis in PCSK9 overexpression mice.

Male C57BL/6 J mice were injected with PCSK9 D374Y overexpression virus or the corresponding control virus through the tail vein, and then the mice were fed a WD and intragastrically administered with vehicle or E28362 (60 mg/kg per day) for 12 weeks. a–c Plasma TC (a), TG (b), and LDL-C (c) levels were measured (n = 10 per group). d Plasma was pooled per group and the distribution of cholesterol over the individual lipoproteins was analyzed after separation by FPLC (n = 10 per group). e–h Plasma ALT (e), AST (f), Cre (g), and urea (h) levels were measured (n = 10 per group). i Representative gross lesions of the aortic arch. j Representative ORO-stained full length of the aorta. k Percentage of the plaque area vs. the area of the full-length aorta (n = 10 per group). l Representative image of ORO-stained sections from the aortic root sections. Scale bars: 100 μm. m Aortic root lesion areas (n = 10 per group). n, o Hepatic TC and TG contents (n = 10 per group). p–r WB analysis was conducted to determine the hepatic protein expression levels of PCSK9 and LDLR (n = 10 per group). Representative images are shown and the relative protein level was calculated by ImageJ. s The plasma PCSK9 content was measured by a commercial ELISA kit (n = 10 per group). For all data, blue, red and green lines showed the fractions of control, model, and E28362, respectively. Values are presented as means ± SEM. ^#P < 0.05; ^###P < 0.001 vs Control group; ^*P < 0.05, ^**P < 0.01, ^***P < 0.001 vs Model group. a–c, e, h, n–s Statistical significance was calculated with one-way ANOVA; k, m Statistical significance was calculated using an unpaired 2-tailed Student’s t-test. t Schematic diagram shows the hypolipidemic and anti-atherosclerotic mechanism of E28362. Small-molecule compound E28362 binds to PCSK9, blocks the protein-protein interaction (PPI) between PCSK9 and LDLR, and induces the degradation of PCSK9, which leads to increased LDLR protein level, decreased LDL-C level, and finally exerts beneficial effects on hyperlipidemia and atherosclerosis.

Furthermore, the levels of PCSK9 and LDLR in the liver and the plasma levels of PCSK9 were determined by WB and ELISA. WB results showed that the PCSK9 level was significantly increased in the model and E28362 groups compared with the control group, indicating that PCSK9 was successfully expressed (Fig. 7p). Compared with the vehicle-treated model group, E28362 visibly decreased the PCSK9 protein level and increased the LDLR protein level (Fig. 7p–r). ELISA assay results showed that the plasma level of PCSK9 in the model group was significantly higher than that of the control group; E28362 visibly decreased the plasma PCSK9 level (Fig. 7s). These results prove that E28362 can effectively alleviate the development of atherosclerosis by inhibiting PCSK9.

Discussion

PCSK9 has been identified as an effective drug target for the treatment of hypercholesterolemia and atherosclerosis. The primary hindrance to the widespread adoption of PCSK9 inhibitors like monoclonal antibodies and siRNA is the cost of application, which makes these inhibitors not recommended for primary prevention in most familial hypercholesterolemia cases [35]. In addition, PCSK9 inhibitors like monoclonal antibodies and siRNA are all administered by injection. Therefore, many efforts have been put into the development of oral small-molecule PCSK9 inhibitors aiming to develop convenient alternatives and improve patient compliance.

In this study, we demonstrated that the small-molecule compound E28362 is a novel PCSK9 inhibitor that has potent effects against hyperlipidemia and atherosclerosis. E28362 notably reduced plasma VLDL-C (Figs. 4c and 7d) in two atherosclerosis mice models and LDL-C levels in three distinct animal models (Figs. 3f, 4c, and 7d). E28362 notably reduced plasma VLDL-C (Figs. 4c and 7d) in two atherosclerosis mice models and LDL-C levels in three distinct animal models (Figs. 3f, 4c, and 7d). Notably, E28362 visibly decreased the plaque area of en face aortas (Figs. 4f and 7j) and aortic root plaques (Figs. 4h and 7l) in both ApoE^−/− and hPCSK9 D374Y overexpression atherosclerosis models, underscoring its remarkable anti-atherosclerotic efficacy. E28362 binds to PCSK9 (Fig. 5a–i, Supplementary Fig. 5), disrupts the interaction between PCSK9 and LDLR (Fig. 5k, l), and enhances PCSK9 degradation via the ubiquitin-proteasome pathway (Fig. 6d, e), thereby preserving more LDLR proteins in hepatocytes for heightened LDL-C absorption into the liver. E28362 significantly decreased plasma PCSK9 levels (Figs. 3q and 7s) in both golden hamsters and human PCSK9 D374Y overexpression mice, suggesting that it can overcome the problem of increased plasma PCSK9 levels induced by monoclonal anti-PCSK9 antibodies [36, 37]. Therefore, our data underline that the anti-hyperlipidemic and anti-atherosclerotic effects of E28362 are mainly achieved by inhibition of PCSK9 and upregulation of LDLR to lower LDL-C levels (Fig. 7t).

PCSK9 might also regulate plasma TG-rich lipoprotein levels [38]. Monoclonal anti-PCSK9 antibodies were reported to reduce plasma TG levels in clinical trials [39–41]. In this study, we found that E28362 as a small-molecule PCSK9 inhibitor also significantly decreases plasma TG levels (Figs. 3d and 7b) in both golden hamsters and human PCSK9 D374Y overexpression mice. The plasma TG-lowering effect of E28362 might be another reason for the improvement of dyslipidemia and atherosclerosis.

It was noticed that E28362 treatment improved fatty liver in golden hamsters (Fig. 3k). Some studies showed that the application of PCSK9 inhibitors (alirocumab and evolocumab) [42, 43] could improve fatty liver diseases, which might be related to the eneutrophil-to-hepatocyte communication [42]. However, it has been reported that PCSK9 interacts with apolipoprotein B-100 (APOB-100) and prevents its degradation, and PCSK9 also promotes the secretion of VLDL-APOB100 from the liver [38, 44]; mice lacking Pcsk9 and carriers of the R46L loss-of-function variant exhibit significantly increased accumulation of TGs in the liver, which might be related to the increased VLDL receptor and scavenger receptor CD36 cell-surface levels [38, 45, 46]. Therefore, the reason why E28362 increased liver TG contents in ApoE^−/− mice (Fig. 4k) and the liver-specific overexpressed hPCSK9 D374Y mice have notably reduced liver TG levels (Fig. 7o) might be caused by the change of PCSK9 contents. However, the function of PCSK9 on the liver TGs and the mechanism of E28362 on fatty liver still need to be explored.

At present, some small-molecule PCSK9 inhibitors [47–52] have been found. The mechanism studies showed that 7030B-C5 [51] inhibited PCSK9 transcription and translation; NYX-PCSK9i [49], SBC-115076 (Patent: WO 2014150326 A1), and Compound 13 [24] disrupted the PCSK9-LDLR interaction. Our data demonstrate that E28362 as a novel small-molecule PCSK9 inhibitor can bind to PCSK9, block the interaction between PCSK9 and LDLR, and induce the degradation of PCSK9 (Fig. 7t). The increased LDLR effect of E28362 is equivalent to the small molecular PCSK9-LDLR interaction inhibitor SBC-115076 and PCSK9 peptide inhibitor Pep-2 (Fig. 2d, e). Though PCSK9 proteins having relatively flat binding pockets [53], small-molecule PCSK9 inhibitors including CB_36 [25], compound 13 [24], and E28362 were successfully found through computer virtual screening, which indicates that this relatively low-cost method is feasible to discover PCSK9-LDLR interaction inhibitors.

In conclusion, E28362 is a novel PCSK9 inhibitor that can bind to PCSK9, block its interaction with LDLR, and induce the degradation of PCSK9, exerting beneficial effects on hyperlipidemia and atherosclerosis (Fig. 7t), and it might be a potential lead compound for the treatment of hyperlipidemia and atherosclerosis.

Author contributions

WZW conducted the study and wrote the manuscript. SYS, YNX, and CL designed and supervised the project. CL, MHC, and TMX synthesized and analyzed compound E28362. WZW, JQL, LJL, LW, and YXW did the in vitro studies and analyzed the in vitro data. WZW, YYZ, RS, YNL, XHJ, YHZ, SWL, and YX performed in vivo experiments and analyzed the in vivo data.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41401-024-01305-9.