Dynamin-related protein 1 inhibition reduces hepatic PCSK9 secretion

Abstract

Open in new tabDownload slide

1. Introduction

Proteostasis is a physiologic process through which cells maintain protein homeostasis¹ and is achieved through biogenesis, folding, maturation, transport, secretion, and proteasomal and autophagosome-lysosomal protein degradation. Endoplasmic reticulum (ER) membrane remodeling through release and incorporation of trafficking vesicles initiates proteostasis processes.^2–4 Additionally, ER molecular chaperones participate in cholesterol metabolism-associated proteostasis. Glucose-regulated protein 94 (GRP94), an ER molecular chaperone, selectively binds proprotein convertase subtilisin/kexin type 9 (PCSK9).⁵ PCSK9 promotes degradation of a critical lipoprotein receptor, low-density lipoprotein receptor (LDLR), and in the absence of GRP94, LDLR is more sensitive to PCSK9-mediated degradation.⁵ PCSK9 strongly associates with hyperlipidemia,⁶ a cardiometabolic disease risk factor. PCSK9 small molecule inhibitor development has proven difficult,⁷ but, if achieved, it could provide a broader role for PCSK9 inhibition therapies.

PCSK9 is transported from ER via coat protein complex II (COPII) vesicles to Golgi, where autocatalytic cleavage and PCSK9 maturation occur before PCSK9 is secreted.² Mitochondrial reactive oxygen species increase PCSK9 secretion and associate with mitochondrial fission.^8,9 Trafficking vesicle-mediated remodeling of mitochondrial-ER membrane contact sites participates in mitochondrial fission.¹⁰ Dynamin-related protein 1 (DRP1) is a key driver of mitochondrial fission¹¹ and has been suggested as a therapeutic target in diseases associated with dysfunctional proteostasis.^12–14 The role of DRP1 in proteostasis is unclear, and whether DRP1 regulates PCSK9 secretion in unknown. Cytosolic DRP1 localizes to organelles, primarily mitochondria, but is also enriched in fractions containing secretory proteins and associates with ER tubules in liver.¹⁵ Given a putative association of DRP1 with hepatic secretory proteins, we examined the poorly defined functional roles of DRP1 in hepatic proteostasis.

2. Methods

2.1 Reagents

Major resources used in this study are included as Supplementary material online,Table S1. DMSO vehicle (final concentration 0.01% in cell culture) was purchased from Sigma Aldrich (St Louis, MO) and mitochondrial division inhibitor 1 (mdivi-1; >98% purity) from Sigma Aldrich. For cell culture, mdivi-1¹⁶ was incubated at a concentration of 50 μmol/L unless stated otherwise. Mdivi-1 in DMSO vehicle was added to cell culture media in a 50 mL tube, filling the tube at no more than 60% volume, vortexed hard for several minutes and then incubated at 37°C with repeated vortexing as necessary prior to adding to cells to ensure that the compound was fully solubilized. Pitavastatin and T0901317 were synthesized and provided by Kowa Company, Ltd. (Tokyo, Japan) and incubated at a dose of 1 μmol/L in all experiments. Chloroquine and MG132 were purchased from Sigma Aldrich and used at a concentration of 50 and 0.3 μmol/L, respectively.

2.2 Human subjects and tissues

The investigation conforms with the Declaration of Helsinki and does not contain any studies involving human subjects. Human cells were all commercially obtained (ATCC, Manassas, VA, Thermo Fisher, Waltham, MA, Research Blood Components, Watertown, MA; written informed consent was obtained by the vendors when applicable) and approved for use by the Partners Institutional Biosafety Committee (protocol #2017B000026).

2.3 Cell culture

Cells used in this study tested negative for mycoplasma using a MycoAlert mycoplasma detection kit (Lonza, Basel, Switzerland) according to manufacturer’s protocol. HepG2 cells were obtained from ATCC and cultured in EMEM with l-glutamine medium (ATCC), 1% penicillin–streptomycin from VWR International (Radnor, PA), with 10% foetal bovine serum (FBS) from Thermo Fisher, unless noted otherwise, in 12-well plates from Corning (Durham, NC). HepG2 cells were plated such that they were incubated with compounds at a density of approximately 60–70% to avoid clumping. Cell viability was assessed by Trypan Blue (VWR International) exclusion using live/dead cell counting on a Countess II LT automated cell counter (Thermo Fisher). Free cholesterol was extracted in 3:2 (hexane: isopropanol; 250 μL/well), dried down by vacuum centrifuge, and resuspended in 100 μL isopronal and measure using a Wako Free Cholesterol E kit (Thermo Fisher). Freshly isolated primary human hepatocytes plated at 100% confluency on 12-well plates were obtained from Thermo Fisher. Three commercially obtained (Thermo Fisher) cell donors were used to obtain experimental data: 43-year-old Caucasian female, 23-year-old Caucasian female, and a 43-year-old Caucasian male. Primary fresh hepatocytes were incubated immediately upon arrival and cultured in William’s E medium with hepatic maintenance supplement pack (Thermo Fisher), 1% penicillin–streptomycin, and 10% FBS. HEK293 cells (ATCC) were cultured in the same manner as HepG2 cells. For primary human macrophage cell culture studies, unpurified buffy coats were obtained from Research Blood Components, LLC (Watertown, MA) and isolated using an EasySep Human Monocyte Enrichment Kit without CD16 Depletion protocol (STEMCELL technologies, Cambridge, MA). Detailed cell culture methods, including HepG2 cell CRISPR gene editing and primary macrophage isolation and mdivi-1 incubation, are provided in the Supplementary material online.

2.4 Animal procedures

The investigation conforms with the Guide for the Care and Use of Laboratory Animals published by the US National Institutes of Health (2011). No procedures requiring anesthetic agents were used. Euthanasia was performed using carbon dioxide, according to AVMA Guidelines for the Euthanasia of Animals (2020) and approved by local animal welfare committees of Kyushu University and Beth Israel Deaconess Medical Center (protocols #010-2013, #010-2016). Detailed animal procedures are included in the Supplementary material online. Generation of Drp1LiKO mice was previously described.¹⁷

2.5 Histology

For tissue histology, frozen sections (7 μm thickness) were processed, with six equally distributed sections examined per sample. Plaques were stained using Oil Red O (Abcam) according to the manufacturer’s instructions. Necrotic plaque was assessed as the acellular areas of Oil Red O stained plaque. Alkaline phosphatase activity was stained with the Vector Red Alkaline Phosphatase Substrate Kit (Vector Laboratories). Calcification was imaged using von Kossa staining with 5% silver nitrate (American Master Tech Scientific, Lodi, CA). For HepG2 cell immunofluorescence, cells were grown on 0.1% gelatin (Sigma Aldrich)-coated Lab-Tek II chambered cover glass #1.5 borosilicate slides (Lab-Tek, Rochester, NY). Immunofluorescence slides were analyzed using a Nikon confocal microscope A1 (Nikon Instruments, Inc., Melville, NY). Images were processed with Elements 3.20 software (Nikon Instruments, Inc.). CellRox Green was used for oxidative stress assessment in a manner similar to another reported protocol.¹⁸ Detailed histology methods, including oxidative stress, mitochondrial, and autophagy assessments, are provided in the Supplementary material online.

2.6 Omics

HepG2 cells were cultured in either light-labeled lysine or heavy-labeled lysine-containing stable isotope labeling by amino acids in cell culture (SILAC) media for about 2 weeks until heavy lysine was found to be >99% incorporated by mass spectrometry. Peptide samples were analyzed with a high-resolution/accuracy Q Exactive mass spectrometer fronted with a Nanospray FLEX ion source and coupled to an Easy-nLC1000 HPLC pump (Thermo Fisher Scientific). Mouse liver metabolite samples were analyzed using a Q Exactive coupled to Vanquish Horizon UHPLC system (Thermo Fisher Scientific). Detailed proteomics and metabolomics methods are included in the Supplementary material online.

2.7 RNA analysis

Human cells and mouse liver were homogenized with TRIzol reagent (Thermo Fisher) to isolate RNA. According to the manufacturer’s instructions, 2 μg total RNA was used to make 20 μL cDNA using the Quanta qScript cDNA Synthesis Kit (VWR International). mRNA levels were quantified on a 7900HT real-time PCR system (Thermo Fisher) using TaqMan qPCR with PerfeCTa FastMix II (VWR International). PCR probes are included in the Supplementary material online,Table S1. All mRNA levels were normalized to GAPDH using the ΔΔCT method. Cytoskeleton mRNA was assessed using the RT² Profiler™ PCR Array Human Cytoskeleton Regulators kit according to the manufacturer’s protocol (Qiagen Inc., Hilden, Germany).

2.8 Western blot and ELISA assays

Primary antibody information and detailed methods used for western blot and ELISA analysis are included in the Supplementary material online. For western analysis, protein levels were normalized to β-actin, tubulin, GAPDH, calnexin, and VDAC with the use of NIH ImageJ software for quantification. Membranes used in Main Figure western blots are included in the Supplementary material online.

2.9 Vesicle budding and DRP1 tethering assays

HepG2 cell microsomal membrane (ER) vesicle budding was assessed with an in vitro budding assay, using a NanoSight LM10 (Malvern Instruments, Malvern, UK) for nanoparticle tracking analysis to quantify budded vesicles. In vitro DRP1 tethering was performed using 100 μL in vitro synthesized vesicles incubated with 1 μg GFP-tagged DRP1 for 5 min and then imaged on a Nikon confocal microscope A1. Detailed vesicle budding and DRP1 tethering assay methods are included in the Supplementary material online.

2.10 Network analysis

2.11 Illustrations

Working model was made using a licensed Motifolio illustration tool kit (Motifolio Inc., Ellicott City, MD).

2.12 Statistical analysis

PRISM software (GraphPad, San Diego, CA) was used to analyze data with two-tailed unpaired Welch’s t-test or ANOVA with Tukey’s multiple comparisons test where appropriate. Normality and equal variance were tested using PRISM software. For box and whisker plots, the boxes extend from the 25th to 75th percentiles, middle line is plotted at the median, whiskers show min to max, and all individual data points are shown. SILAC mass spectrometry data analysis was performed using a one-sample t-test to assess how significant each protein’s ratio was from 1 (the ideal ratio if there is no effect of mdivi-1 incubation), the P-value was then adjusted using Benjamini and Hochberg method (FDR), and the results visualized using a volcano plot. Heat maps were made using liver metabolomics datasets with Qlucore software (Lund, Sweden).

3. Results

3.1 DRP1 inhibition reduced a hepatic secretome associated with cardiometabolic diseases

To examine DRP1 function in protein secretion, we quantitatively assessed a human hepatic secretome using SILAC mass spectrometry and inhibited DRP1 using mdivi-1, a small molecule DRP1 inhibitor. Mdivi-1 inhibits DRP1 GTPase activity without acting as a general dynamin family protein inhibitor,¹⁶ a finding that has been independently verified using human DRP1 immunopurified from human cells.²³ We validated that mdivi-1 inhibited HepG2 cell mitochondrial fission through assessing abundance of mitochondrial dynamics proteins, mitochondrial morphology, and mitochondrial function. Mdivi-1 incubation did not alter HepG2 cell total protein abundance of mitochondrial dynamics proteins including DRP1, mitochondrial fission factor (MFF), mitofusion 1 (MFN1), mitofusin 2 (MNF2), and optic atrophy mitochondrial dynamin-like GTPase (OPA1) (Fig. 1A). We also assessed HepG2 cell DRP1 phosphorylation levels that could regulate DRP1 activity, including on serine 616 that was not detected in our experimental conditions and on serine 637 that was not altered by mdivi-1 (Fig. 1B). Mdivi-1 reduced DRP1 mitochondrial localization and increased DRP1 cytosolic localization in HepG2 cells (Fig. 1C). Mdivi-1 increased HepG2 cell mitochondrial aspect ratio, assessed by confocal microscopy (Fig. 1D), demonstrating reduced mitochondrial fission. Additionally, mdivi-1 increased HepG2 cell intracellular calcium levels without altering protein kinase C activity (Fig. 1E), supporting increased mitochondrial calcium buffering following inhibition of mitochondrial fission rather than increasing cytosolic calcium that would be accessible to activate protein kinase C. We also assessed HepG2 cell mitochondrial membrane potential (Fig. 1F), a general marker of mitochondrial function, and HepG2 cell viability (Fig. 1G), which was unchanged by mdivi-1 incubation.

Figure 1

DRP1 inhibition reduced HepG2 cell mitochondrial fission. (A) Western blot time course of HepG2 cell mitochondrial dynamics proteins, DRP1, MFF, MFN1, MFN2, and OPA1 with DMSO vehicle (0.01%) or mdivi-1 (50 μmol/L) incubation; n = 3 experimental replicates, analyzed by ANOVA. (B) HepG2 cells incubated as in A for 24 h, phosphorylated (p S637) DRP1 western blot; n = 3 experimental replicates. (C) Cytosolic and mitochondrial-enriched fractions DRP1 western blots (n = 5 experimental replicates) for HepG2 cells incubated as in B. (D) Mitotracker live HepG2 cell mitochondrial imaging and box and whisker plot of quantified mitochondrial aspect ratio; scale bars are 20 μm, n = 3 experimental replicates with 220 (DMSO) and 203 (mdivi-1) mitochondria assessed and analyzed by unpaired Welch’s t-test. (E) HepG2 cells incubated as in B, intracellular calcium (n = 5 experimental replicates) and protein kinase C activity (PKC) (n = 4 experimental replicates); analyzed by unpaired Welch’s t-test. (F) Mitochondrial membrane potential (n = 3 experimental replicates) and viability (G) for HepG2 cells incubated as in B, viability (n = 3 experimental replicates); analyzed by unparied Welch’s t-test. Error bars, mean ± SD, *P < 0.05, **P < 0.01, and ****P < 0.0001.

Open in new tabDownload slide

Using SILAC mass spectrometry, we identified that mdivi-1 decreased a total of 36 out of 217 (16.6%) HepG2 cell-secreted proteins with an FDR < 0.05 and average fold-change < 0.5 (Supplementary material online, Excel File 1, and Supplementary material online, Fig. 1). Supporting a role of DRP1 in cardiometabolic diseases, network analysis associated secreted proteins significantly altered by mdivi-1 with human disease gene modules for atherosclerosis, hypercholesterolemia, aortic stenosis, myocardial infarction, and non-alcoholic fatty liver disease (Fig. 2A). In addition, we performed pathway network analysis on secreted proteins significantly altered by mdivi-1 incubation, which highlighted several cardiometabolic disease-associated pathways, including glucose and lipid metabolism (Fig. 2B and Supplementary material online, Fig. 2 and Excel File 1). Given an association of oxidative stress with cardiometabolic disease and mitochondrial fission,¹⁴ we assessed oxidative stress in HepG2 cells. Mdivi-1 did not alter the low basal levels of oxidative stress in HepG2 cells but did reduce tert-butyl hydrogen peroxide-induced oxidative stress, assessed by an oxidative stress indicator, CellRox Green (Supplementary material online, Fig. 3).

Figure 2

DRP1 inhibition reduced a hepatocyte secretome associated with cardiometabolic diseases. (A) Diseases that are significantly associated (empirical P-value < 0.05) via their connections on the PPI network with the 92 differentially abundant [average fold-change (mdivi-1/DMSO vehicle) < 1, FDR < 0.05] HepG2 cell-secreted proteins identified using SILAC proteomics (Supplementary material online, Excel File 1). (B) Network representation of pathways showing enrichment as well as relationship with each other using the set of secreted proteins. Nodes represent pathways, and connections (edges) between nodes represent the shared proteins between the pathways that were captured in proteomics measurements. Green nodes indicate pathways that are also significantly enriched (FDR < 0.05) in the differentially abundant proteins with an FDR < 0.05, average fold-change (mdivi-1/DMSO vehicle) <0.5. Node size is proportional to the significance of enrichment [−log(q-value)]. Fully labeled network included as Supplementary material online, Figure 2; n = 6 experimental replicates.

Open in new tabDownload slide

Figure 3

DRP1 inhibition reduced hepatocyte PCSK9 secretion. (A) Secreted PCSK9 ELISA with 24-h mdivi-1 incubation; n = 3 experimental replicates. (B) Box and whisker and dot plots for HepG2 cell PCSK9 mRNA and secreted PCSK9 ELISA for incubation with DMSO vehicle (0.01%; mRNA n = 18, protein n = 29 experimental replicates; DMSO controls are the same samples in the individual graphs in panel B for relative comparison among the incubations), mdivi-1 (50 μmol/L; mRNA n = 17, protein n = 27), pitavastatin (1 μmol/L; mRNA n = 9, protein n = 18), pitavastatin +mdivi-1 (mRNA n = 9, protein n = 18), T0901317 (1 μmol/L; mRNA n = 9, protein n = 18), T0901317 +mdivi-1 (mRNA n = 9, protein n = 18), chloroquine (50 μmol/L; mRNA n = 8, protein n = 9), chloroquine +mdivi-1 (mRNA n = 9, protein n = 9), MG132 (0.3 μmol/L; mRNA n = 9, protein n = 9), or MG132 +mdivi-1 (mRNA n = 9, protein n = 9); analyzed by ANOVA. (C) HepG2 cells incubated as in B, total and pro/mature PCSK9 protein ratio; DMSO, mdivi-1 n = 6, choloroquine, choloroquine +mdivi-1, MG132, MG132 +mdivi-1 n = 3, analyzed by ANOVA. (D) HepG2 DRP1 CRISPR/Cas9 western blot and secreted PCSK9 ELISA; analyzed by unpaired Welch’s t-test, n = 3. (E) Box and whisker plots for primary hepatocyte secreted PCSK9 ELISA, analyzed by ANOVA; DMSO, mdivi-1 n = 9 experimental replicates from three donors, pitavastatin, pitavastatin +mdivi-1 n = 6. (F) HepG2 cells incubated as in B, SREBP-1c mRNA, analyzed by unpaired Welch’s t-test, n = 11. (G) HepG2 cells incubated as in B, box and whisker plots for LDLR protein and mRNA; analyzed by ANOVA, DMSO (mRNA n = 12, protein n = 7), mdivi-1 (mRNA n = 11, protein n = 8), pitavastatin (1 μmol/L; mRNA n = 9, protein n = 7), pitavastatin +mdivi-1 (mRNA n = 9, protein n = 7), T0901317 (1 μmol/L; mRNA n = 9, protein n = 8), or T0901317 +mdivi-1 (mRNA n = 9, protein n = 8). Error bars, mean ± SD; *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.

Open in new tabDownload slide

3.2 DRP1 inhibition reduced PCSK9 secretion in human hepatocytes

Using SILAC mass spectrometry, we identified a 74% reduction in secreted PCSK9, which was the top secreted protein reduced by mdivi-1 incubation in HepG2 cells (Supplementary material online, Fig. 1 and Excel File 1). Given the importance of PCSK9 in cardiometabolic diseases and as it was the top reduced protein, we focused our study on DRP1 regulation of PCSK9 secretion. We validated that mdivi-1 reduced PCSK9 secretion in HepG2 cells by ELISA in a concentration-dependent manner (Fig. 3A). PCSK9 mRNA and protein were reduced by mdivi-1 in HepG2 cells (Fig. 3B,C) and assessed with an antibody validated with CRISPR/Cas9 PCSK9-deficient and PCSK9-overexpression cell lysate (Supplementary material online, Fig. 4A). To validate specificity of mdivi-1, we deleted DRP1 by CRISPR/Cas9 in HepG2 cells, which reduced PCSK9 secretion (−66%) (Fig. 3D). Mdivi-1 similarly reduced PCSK9 secretion in primary human hepatocytes (Fig. 3E), demonstrating that mdivi-1 is a novel inhibitor of PCSK9 secretion.

Figure 4

DRP1 participated in HepG2 cell proteostasis. (A) Phosphorylated (P S349) p62, p62, LC3I, and LC3II western blots from HepG2 cells incubated for 24 h with vehicle (0.01% DMSO), mdivi-1 (50 μmol/L), chloroquine (50 μmol/L), or MG132 (0.3 μmol/L); n = 3 experimental replicates, except DMSO and mdivi-1 for p62 where n = 6; analyzed by ANOVA. (B) LC3, p62, DAPI, and LAMP1 representative confocal images for HepG2 cells incubated as in A; n = 3 experimental replicates, scale bars, 10 μm. (C) HepG2 cells incubated as in A, cytosolic, mitochondrial, and microsomal fractionated GRP94, p62, and MFF; n = 3 experimental replicates, analyzed by unpaired Welch’s t-test. (D) Representative electron microscopy images showing ER morphology (lines surrrounding mitochondria) in control CRISPR and DRP1 knockout CRISPR-Cas9cells (n = 3 experimental replicates). Scale bars, 500 nm. Error bars, mean ± SD; *P < 0.05, ***P < 0.001; A < 0.05, B < 0.01, C < 0.001, D < 0.0001 vs. DMSO; E < 0.05, F < 0.01, G < 0.001, H < 0.0001 vs. mdivi-1; I < 0.01, J < 0.001, K < 0.0001 vs. chloroquine; L < 0.05, M < 0.01, N < 0.001, O < 0.0001 vs. chloroquine +mdivi-1; P < 0.05, Q < 0.0001 vs. MG132.

Open in new tabDownload slide

3.3 DRP1 inhibition increased PCSK9 proteasomal degradation

In addition to transcriptional PCSK9 inhibition, proteostasis regulation mediated through DRP1 inhibition may also reduce PCSK9 secretion. To assess DRP1 regulation of PCSK9 secretion via a proteostasis mechanism, we co-incubated HepG2 cells with mdivi-1 and an autophagosome-lysosome fusion and degradation inhibitor, chloroquine, or with a proteasome inhibitor, MG132. Mdivi-1 suppressed PCSK9 secretion with and without chloroquine or MG132 co-incubation in HepG2 cells (Fig. 3B). Chloroquine incubation increased cellular PCSK9 protein that was reduced by mdivi-1 co-incubation (Fig. 3C). Co-incubation of MG132 with mdivi-1 resulted in significant accumulation of an ER-localized immature form of PCSK9, pro-PCSK9 compared to mature PCSK9 (Fig. 3C). These data support mdivi-1 partially reduced PCSK9 secretion via increasing PCSK9 proteasomal degradation in the early secretory pathway.

3.4 Mdivi-1 reduced SREBP-1c without altering statin induction of LDLR expression

Sterol regulatory element-binding protein (SREBP) transcription factors increase PCSK9 expression;⁷ therefore, we assessed a role of SREBP in mdivi-1 regulation of PCSK9. To examine if DRP1 inhibition reduced PCSK9 via inhibiting SREBP, we tested mdivi-1 co-incubation with two established SREBP regulators that induce PCSK9: pitavastatin that leads to SREBP-2 activation,²⁴ and a liver X receptor agonist, T0901317 that activates SREBP-1.²⁵ HepG2 cell PCSK9 secretion was decreased when mdivi-1 was incubated alone or combined with pitavastatin (−68%) or T0901317 (−64%) (Fig. 3B). Supporting combinatory therapy potential, mdivi-1 attenuated PCSK9 secretion induction by pitavastatin in primary human hepatocytes (Fig. 3E). Mdivi-1 reduced mRNA levels of an SREBP-1 isoform, SREBP-1c in HepG2 cells (Fig. 3F), but not SREBP-1a or SREBP-2 mRNA levels (Supplementary material online, Fig. 2C). Mdivi-1 did not alter protein abundance of SREBP-2, an SREBP-2-regulated gene, 3-hydroxy-3-methylglutaryl-coenzyme A reductase (HMGCR), or cholesterol efflux genes ATP-binding cassette subfamily A member 1 and ATP-binding cassette subfamily G member 1 (Supplementary material online, Fig. 5A,B). This data mechanistically explains the PCSK9 transcriptional inhibition we observed in HepG2 cells likely occurred via a reduction in SREBP-1c.

Figure 5

DRP1 tethered trafficking vesicles at ER. (A) Representative HepG2 cell DRP1 immunogold electron microscopy. Arrows, ER exit sites DRP1 (black dots); arrow heads, cytosolic vesicles DRP1; color box zoomed-in insets, DRP1 tethering trafficking vesicles at ER; scale bars, 100 nm. (B) Confocal analysis of human DRP1 protein tethering unilaminar vesicles; white arrows indicate examples of tethered vesicles; n = 3 experimental replicates. Representative confocal Z-stack image of GFP-tagged DRP1 tethering vesicles, width 97.70 μm, height 97.7 μm, and depth 38.40 μm. Image also included as Supplementary material online, Video 1. (C) Working model. DRP1 inhibition suppresses remodeling of select ER microdomains and select protein trafficking out of ER, including PCSK9, leading to increased PCSK9 degradation by trafficking from ER to proteasomes via ER chaperone GRP94 to maintain ER homeostasis. Additionally, DRP1 inhibition transcriptionally suppresses PCSK9 mRNA in specific conditions via SREBP regulation, such as in HepG2 cells by reducing SREBP-1c.

Open in new tabDownload slide

In HepG2 cells, mdivi-1 increased LDLR mRNA levels leading to a corresponding increase in LDLR protein: mdivi-1 incubation alone (+5.8-fold), in combination with pitavastatin (+3-fold) or T0901317 (+9-fold) (Fig. 3G). In agreement with increased LDLR abundance, 24-h incubation with mdivi-1 increased free cholesterol in HepG2 cells (Supplementary material online, Fig. 5C). As the mdivi-1-mediated increase in LDLR protein was too large to be explained only by reduced PCSK9 degradation of LDLR, we assessed LDLR transcriptional regulators. In addition to SREBP-2 that was not increased by mdivi-1, LDLR is transcriptionally induced by a cell-cycle regulator, c-Jun.²⁶ Activation of c-Jun incurs in part by c-Jun N-terminal kinase that also phosphorylates MFN2, promoting MFN2 degradation and reducing mitochondrial fusion, thus linking c-Jun to mitochondrial dynamics.⁹ Mdivi-1 induced c-Jun in HepG2 cells (Supplementary material online, Fig. 4C). Changes in c-Jun mirrored those in LDLR in HepG2 cells, which explains increased LDLR transcription by mdivi-1 in HepG2 cells as likely being mediated via c-Jun.

3.5 DRP1 inhibition reduced autophagic flux in human liver cells

As DRP1 inhibition increased PCSK9 proteasomal degradation, we assessed whether this phenomenon was related to impaired autophagic flux. To examine autophagic flux, we incubated HepG2 cells with mdivi-1 and chloroquine and then assessed protein levels of the autophagy-related proteins, p62, and light chain 3 (LC3). Mdivi-1 increased p62 and serine 349 phosphorylated p62 protein, which were further increased by an autophagy inhibitor, chloroquine (Fig. 4A). Co-incubation with mdivi-1 did not significantly alter increased p62 levels observed with chloroquine incubation alone (Fig. 4A). Co-incubation with mdivi-1 increased LC3 to a lesser extent than chloroquine alone (Fig. 4A), demonstrating mdivi-1 reduced, but did not fully inhibit autophagic flux in HepG2 cells. To assess whether increased proteasomal degradation with mdivi-1 incubation was tied to reduce autophagic flux, we co-incubated mdivi-1 with MG132. Supporting mdivi-1 association with increased proteasomal degradation, co-incubation with mdivi-1 and MG132 increased p62 and LC3 compared to mdivi-1 or MG132 incubation alone, assessed by western blot (Fig. 4A). Immunofluorescence confirmed these results for LC3 and p62, along with demonstrating altered proteostasis when mdivi-1 was combined with MG132, a condition in which perinuclear lysosomal-associated membrane protein 1 and p62-positive aggregates formed (Fig. 4B). Using cell fractionation, we observed p62 induction by mdivi-1 in HepG2 cell cytosolic-enriched fractions, but not mitochondrial or ER-enriched microsomal fractions (Fig. 4C). No differences were observed in protein abundances in HepG2 cell cytosolic-, mitochondrial-, and microsomal-enriched fractions for autophagosome formation proteins, including autophagy-related protein 9, autophagy-related protein 14, phosphatidylinositol 3-kinase, regulatory subunit 4, or class III phosphoinositide 3-kinase (Supplementary material online, Fig. 6A).

Figure 6

Drp1 hepatocyte deficiency reduced PCSK9 secretion in mice. (A) Flox/flox and Drp1LiKO liver tissue DRP1, OPA1, MFF, LDLR, PCSK9, and HMGCR western blots; n = 5 mice/group. (B) Circulating total cholesterol, lipoprotein profiles and (C) circulating PCSK9 from non-fasted standard laboratory diet-fed flox/flox and Drp1LiKO mice; n = 5 mice/group. (D) Flox/flox and Drp1LiKO liver p62 and LC3; n = 5 mice/group. (E) Flox/flox and Drp1LiKO liver tissue amino acid reverse phase (RP+ and RP−, positive and negative ion mode, respectively) and Intrada amino acids (Intrada AA+ and Intrada AA−) metabolomics heat map, scale bar indicates ratio change; n = 5 mice/group. Error bars, mean ± SD; data analyzed by unpaired Welch’s t-test, *P < 0.05, **P < 0.01, and ***P < 0.001.

Open in new tabDownload slide

3.6 DRP1 promotes ER remodelling by tethering trafficking vesicles to ER

We assessed if DRP1 regulates hepatic proteostasis via altering cytoskeleton or DRP1 ER localization. Mdivi-1 incubation did not significantly alter mRNA levels of 79 cytoskeleton-related genes in HepG2 cells and primary human hepatocytes (Supplementary material online, Fig. 6B). Mouse liver ER is swollen and dispersed in DRP1 hepatocyte-deficient mice (Drp1LiKO) on a high-fat diet, but closer to normal ER morphology in Drp1LiKO mice on a normal laboratory diet.¹⁷ To examine ER morphology in human liver cells, we performed electron microscopy using HepG2 cells in which DRP1 was deleted by CRISPR/Cas9. DRP1-deficient cells had no apparent gross ER morphologic changes, such as swelling or dispersion compared to control cells, and ER appeared as tubular networks alongside mitochondria in control and DRP1-deficient cells, assessed by electron microscopy (Fig. 4D).

To examine localization of DRP1 in HepG2 cells, we used electron microscopy with DRP1 immunogold-labeling with a knockout validated antibody. We observed DRP1 at mitochondria, cytosol/cytosolic vesicles, and ER, notably ER exit sites, but largely absent from Golgi (Fig. 5A). In addition to clustering at ER exit sites and trafficking vesicles, DRP1 highly aggregated at vesicle membrane tethered to ER exit sites (Fig. 5A). ER remodeling involves release and incorporation of trafficking vesicles. Therefore, we sought to determine if DRP1 remodels select ER micro-domains via a functional action on ER-associated vesicles. We first assessed a possibility that DRP1 may play a fission role at ER using a HepG2 cell COPII vesicle budding assay; however, mdivi-1 did not inhibit COPII vesicle budding (Supplementary material online, Fig. 6C). Evidence also suggested that DRP1 functions as a membrane tether protein leading to membrane remodeling.^27,28 Vesicle remodeling at mitochondrial-ER-associated membranes and ER exit sites is involved in inducing autophagy and mitochondrial fission;^3,10 although, a role of DRP1 in this function has not been assessed. To validate DRP1 membrane tethering, we immunopurified green fluorescence protein-tagged human DRP1 and incubated it with unilaminar vesicles, which increased vesicle tethering at DRP1-enriched areas on tethered membrane contact sites (Fig. 5B and Supplementary material online, Video 1). These data support DRP1 transiently localized to ER exit sites and trafficking membrane vesicles where it may serve in part to tether membrane promoting remodeling of select ER micro-domains that participates in mitochondrial fission.

3.7 Mdivi-1 reduced ER localized MFF and increased cytosolic GRP94

We assessed how DRP1 may transiently associate with ER in hepatocytes via receptor-mediated interaction by examining mdivi-1 effects on a DRP1 receptor, MFF,²⁹ and a DRP1-binding protein syntaxin 17 (STX17).³⁰ We found that mdivi-1 significantly reduced ER-localized (microsomal fraction) MFF without altering mitochondrial MFF protein abundance (Fig. 4C) or STX17 cellular localization. This supports a likely novel mechanism of mdivi-1 in inhibiting DRP1-mediated mitochondrial fission via reducing a transient ER pool of DRP1 by lowering ER MFF abundance; although, we cannot exclude an effect of mdivi-1 on DRP1 via its interaction with STX17 based solely on our localization data. We found that mdivi-1 increased cytosolic levels of a PCSK9-binding ER chaperone protein, GRP94 (Fig. 4C). As proteasomes are in the cytosolic fraction, this data supports that DRP1 inhibition likely increased PCSK9 trafficking from ER to proteasomes via GRP94 to maintain ER proteostasis. We present a working model for DRP1 regulation of PCSK9 secretion (Fig. 5C). DRP1 transiently translocates from cytosol to ER via association with its receptor, MFF on ER where it may assist in tethering of trafficking vesicles that drive hepatic remodeling of select ER microdomains, ER remodeling-associated mitochondrial fission, and autophagy initiation. To maintain ER homeostasis, PCSK9 is instead transported from ER to proteasomes via GRP94 where it is degraded. DRP1 mitochondrial fission may also regulate hepatic proteostasis by reducing mitochondrial calcium buffering. Additionally, DRP1 inhibition regulates PCSK9 transcription via reduced SREBP-1c in some conditions.

3.8 PCSK9 secretion is reduced in hepatocyte Drp1-deficient mice

To validate DRP1 inhibition-reduced PCSK9 secretion in vivo, we utilized Drp1LiKO mice that are deficient in hepatocyte DRP1.¹⁷ We first verified that DRP1 was absent in hepatocytes of Drp1LiKO mice by western blot (Fig. 6A). We then assessed additional mitochondrial dynamics proteins and found reductions in hepatic MFN1 and MFN2, MFF, and the long form of OPA1 that promotes mitochondrial fusion, but not in the short form of OPA1 that facilitates mitochondrial fission (Fig. 6A and Supplementary material online, Fig. 7A). Body weight, glucose, triglycerides, plasma aspartate aminotransferase, and alanine aminotransferase did not significantly change in 1-month-old, non-fasted, normal laboratory diet-fed Drp1LiKO mice (Supplementary material online, Fig. 4B,C,D,E). Total cholesterol (−21%) and LDL cholesterol concentrations modestly, but significantly decreased in Drp1LiKO mice (Fig. 6B). HDL cholesterol was non-significantly decreased in Drp1LiKO mice with lower levels observed in four of five mice examined that likely contributed to the overall reduction in total cholesterol (Fig. 6B). Providing in vivo validation that inhibiting DRP1-reduced PCSK9 secretion, plasma PCSK9 concentrations were reduced 78.5% in Drp1LiKO mice compared to flox/flox controls (Fig. 6C). Liver SREBP1 and SREBP2 protein and mRNA levels were unchanged in Drp1LiKO mice on a normal laboratory diet (Supplementary material online, Fig. 7F). Supporting altered PCSK9 proteostasis, liver tissue mRNA, and protein levels for PCSK9 and an SREBP target gene, HMGCR did not significantly change, whereas liver LDLR protein but not mRNA increased in Drp1LiKO mouse livers (Fig. 6A and Supplementary material online, Fig. 7G).

Figure 7

Mdivi-1 reduced calcified atherosclerotic plaque in diabetic Apoe^−/− mice. (A) Representative images for Oil Red O staining (red color) with quantification of total plaque and necrosis plaque area, (B) TNAP activity (red color) and von Kossa stain (black color, calcification indicated with black arrow), and (C) MAC3 immunohistochemistry (brown color) in aortic roots of Apoe^−/−; STZ mice intraperitoneally injected twice weekly for 15 weeks with either DMSO vehicle or mdivi-1 (50 mg/kg); n = 8 mice/group, with six equally distributed aortic root cryosections assessed per mouse, scale bars, 200 μm. Data analyzed by unpaired Welch’s t-test, (D) plasma IL-6 concentrations and (E) aortic root plaque macrophage (F4/80, green color) mitochondria (TOMM20, red color, dash-line boxes indicated zoomed in areas showing mitochondria only) in 15-week-mdivi-1 administered Apoe^−/−; STZ mice [n = 8 mice/group with 69 (DMSO) and 81 (mdivi-1) mitochondria analyzed]; scale bars, 20 μm, data analyzed by unpaired Welch’s t-test. (F) Human peripheral blood mononuclear cell-derived macrophages pre-incubated with DMSO (0.01%) or mdivi-1 (25 μmol/L) and then incubated with lipopolysaccharide, IL-6, and TNF-α-secreted protein and mRNA (n = 5 donors); analyzed by ANOVA. Error bars are mean ± SD, *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.

Open in new tabDownload slide

3.9 Drp1 deficiency increased liver p62

To assess whether proteostasis alterations observed with DRP1 inhibition in human liver cells also occurred in vivo, we measured p62 protein in Drp1LiKO mice livers. Drp1LiKO mice had increased liver p62 protein without significantly altered LC3 protein abundance (Fig. 6D). A major regulator of autophagy is mTOR, which inhibits autophagy and is in turn regulated by ER stress and amino acid availability.^25,31 ER stress marker protein abundance for binding immunoglobulin protein and C/EBP homologous protein in human liver cells incubated with mdivi-1 and in livers of 1-month-old, normal laboratory diet-fed Drp1LiKO mice were unchanged (Supplementary material online, Fig. 8A). To further exclude mTOR pathway as a regulator of proteostasis with DRP1 inhibition, we performed Drp1LiKO liver amino acid metabolomics. A modest but significant increase in phenylalanine was observed by reverse phase and Intrada amino acid assessment in Drp1LiKO livers, whereas no significant changes were consistently observed with both detection methods for alanine, aspartic acid, glutamic acid, glutamine, glycine, histidine, isoleucine, leucine, lysine, methionine, proline, serine, threonine, tryptophan, tyrosine, and valine amino acids (Fig. 6E and Supplementary material online, Excel File 2). Additionally, we did not observe changes in total or phosphorylated levels of mTOR target proteins p70S6 kinase and Unc-51 like autophagy activating kinase 1 in in Drp1LiKO livers (Supplementary material online, Fig. 8B).

3.10 Mdivi-1 reduced advanced calcified plaque in aortic roots of diabetic Apoe^−/− mice

Mdivi-1 reduced early atherosclerotic plaque formation in Apoe^−/− mice with streptozotocin (STZ)-induced diabetes.⁸ PCSK9 has been suggested to not be a regulator of plasma cholesterol in Apoe^−/− mice.² These studies suggest that mdivi-1 may provide additional mechanistic benefits not involving PCSK9 and lipid metabolism in Apoe^−/− mice. Whether small molecule DRP1 inhibition suppresses advanced atherosclerotic plaque formation including an associated and critical cardiovascular pathology, vascular calcification is unknown. Therefore, we assessed the effects of mdivi-1 on advanced atherosclerotic plaque in diabetic Apoe^−/− mice. We performed an initial assessment of mdivi-1 pharmacokinetics when administered via intraperitoneal injection to Apoe^−/− mice (50 mg/kg, a dose reflecting the 50 μmol/L incubation used with human liver cells). Maximum plasma mdivi-1 concentration was observed 1 h after injection and was almost negligible by 6 h (Supplemental material online, Fig. 9A).

To assess the effects of DRP1 inhibition on advanced atherosclerotic plaque, 10-week-old Apoe^−/−; STZ mice were fed a normal laboratory diet for 15 weeks, during which time mice were interperitoneally injected with mdivi-1 twice per week to reflect a prior study assessing DRP1 in early plaque formation.⁸ Advanced atherosclerotic plaque area (−26.5%) and necrotic plaque area (−48%) were decreased in aortic roots of twice weekly mdivi-1-injected Apoe^−/−; STZ mice (Fig. 7A). Additionally, aortic root calcification, a marker of advanced plaque, was reduced in mdivi-1-administered mice and assessed by tissue non-specific alkaline phosphatase activity and von Kossa staining (Fig. 7B). Body weight, plasma glucose, total cholesterol, and triglycerides were not significantly changed by twice weekly mdivi-1 injections in Apoe^−/−; STZ mice (Supplementary material online, Fig. 9B,C). We further assessed the effects of DRP1 inhibition on oxidative stress and inflammation. Mdivi-1 reduced aortic root plaque oxidative stress, assessed by CellRox Green (Supplemental material online, Fig. 2). Macrophage accumulation in aortic roots was decreased relative to total atherosclerotic plaque area, as was an inflammatory cytokine, interleukin-6 (IL-6) in plasma of mdivi-1-administered Apoe^−/−; STZ mice (Fig. 7C,D). Mdivi-1 increased aortic root plaque macrophage mitochondrial aspect ratio (Fig. 7E), supporting reduced mitochondrial fission.

In agreement with our mouse data, mdivi-1 reduced mRNA and secreted protein levels of inflammatory cytokines, IL-6, and tumor necrosis factor-α in primary human macrophages derived from peripheral blood mononuclear cells and incubated with lipopolysaccharide (Fig. 7F), a condition that stimulates mitochondrial fission and inflammation.³² These data support that mdivi-1 likely inhibited advanced atherosclerotic plaque formation in aortic roots of diabetic Apoe^−/− mice and mechanistically worked via reducing oxidative stress and inflammation.

4. Discussion

Our data demonstrate a novel finding that DRP1 inhibition promotes hepatic proteasomal degradation of PCSK9 in the early secretory pathway. DRP1 inhibition did not reduce PCSK9 secretion by blocking global ER-to-Golgi trafficking, but instead acted selectively. We found that DRP1 inhibition increased GRP94 localization in cytosol where proteasomes are located. GRP94 is a selective ER chaperone that binds PCSK9, preventing PCSK9 interaction with LDLR at ER,⁵ supporting mdivi-1 likely induced a proteasomal GRP94-PCSK9 degradation pathway.

Additionally, we observed SREBP-mediated transcriptional regulation in specific experimental conditions including mdivi-1 suppression of PCSK9 mRNA levels in HepG2 cells via reduced SREBP-1c. While SREBP-1c regulates PCSK9 in some conditions, SREBP-2 has been suggested to be the major regulator of PCSK9.³³ Whether DRP1 inhibition also modulates SREBP-2 activity in different conditions than we examined requires further study.

We propose a novel tethering function of DRP1 beyond its established fission role, with DRP1 mediating distinct ER-microdomain remodeling that likely contributes to ER constriction of mitochondria driving mitochondrial fission. ER contains various micro-domains, including mitochondrial-associated membrane that is enriched in mitochondrial dynamics proteins (DRP1, MFF, MFN) and ER chaperones.^34–36 Proteomics profiling of mitochondria-associated membrane did not identify SREBP-2 in mitochondrial-associated membrane,^35,36 suggesting that SREBP-2 may reside in different ER microdomains. Deficiency in a COPII protein, SEC24A reduced PCSK9 trafficking out of ER without disrupting SREBP-2 activity,² further supporting different ER micro-domains may be utilized for PCSK9 and SREBP-2 trafficking.

Mdivi-1 may inhibit DRP1 functions by multiple condition-dependent mechanisms, including regulation of DRP1 GTPase activity, total protein abundance, phosphorylation, and localization. Our data support a role of transient DRP1 interacting with ER-localized MFF in regulating hepatocyte remodeling of select ER microdomains, which is disrupted by mdivi-1 incubation. Other DRP1-binding proteins, including STX17, may play a role in DRP1 function at ER. In mouse embryonic fibroblasts, DRP1 regulates ER tubulation independent of DRP1 GTPase activity and MFF,³⁷ but whether this is through DRP1 interaction with STX17 is unknown. As DRP1 tethering does not require DRP1 GTPase activity,²⁸ it is possibile that mdivi-1 could partially regulate PCSK9 abundance independent of DRP1 GTPase activity, but this remains to be assessed.

Hepatocyte DRP1-deficiency reduced the total abundance of other mitochondrial dynamics proteins (MFF, MFN1, MF2, OPA1), which was not observed in HepG2 cells incubated with mdivi-1. While the reason for this difference is unclear, it may involve mdivi-1 altering DRP1 activity via its localization that does not induce a strong compensatory response.

The mechanism for DRP1 regulation of PCSK9 trafficking and ER proteostasis may be two-fold involving tethering functions of trafficking vesicles at ER and mitochondrial calcium buffering. We found increased mitochondrial calcium buffering with DRP1 inhibition. In agreement, calcium builds up in mitochondria in Drp1-deficient macrophages leading to less lysosomal fusion with phagosomes, reducing efferocytosis.³⁸ The role of DRP1 in promoting efferocytosis should be considered for atherosclerotic therapeutic development; however, conditions with elevated mitochondrial stress, such as cardiometabolic disease, support preclinical promise for DRP1-targeting therapies.^14,17,39,40 Macrophage Drp1-deficient mice have reduced femoral artery intimal thickening after wire injury,³⁹ a model with elevated mitochondrial stress and inflammation. Activated cardiac DRP1 contributes to heart dysfunction in high-fat diet-induced obese and diabetic mice and high-fat and high-cholesterol fed rhesus monkeys.^40,Drp1LiKO mice also have improved glucose tolerance and reduced obesity on high-fat diet.¹⁷

Whether mdivi-1 or similar small molecules can reduce PCSK9 in vivo remains to be demonstrated and requires substantial pharmacokinetics and pharmacodynamics testing beyond the scope of this initial mechanistic study. Future studies could also assess DRP1LiKO on PCSK9-depenent mechanisms in Apoe*3Leiden.CETP atherosclerotic mice,⁴¹ including with induced diabetes that promotes vascular calcification.⁴² In diabetic Apoe^−/− mice, in which atherosclerosis is not regulated via PCSK9, mdivi-1 suppressed oxidative stress, inflammation, and formation of calcified atherosclerotic plaque.

As such, our present study also supports non-PCSK9 mechanistic benefits of DRP1 inhibition in cardiometabolic disease. We previously demonstrated DRP1 inhibition suppressed cardiovascular cell calcification in vitro, including by suppressing oxidative stress, tissue non-specific alkaline phosphatase activity, and collagen secretion.¹⁴ Mitochondrial calcium buffering could also play a role in calcification formation via annexin-mediated hydroxyapatite formation;⁴³ whether DRP1 regulates annexin-mediated cardiovascular calcification is unknown.

In contrast to our results in Drp1LiKO mice, a different hepatocyte-specific Drp1 deletion, targeting exons 3–5 of Drp1, increased serum lipid concentrations that were corrected by compound deficiency in Drp1 and Opa1.^44,Drp1LiKO mice have an in-frame deletion of exon 2 of Drp1.¹⁷ While both Drp1 tissue-specific models are driven by an albumin promoter targeting hepatocytes and have elevated hepatic p62, we found that Drp1LiKO mice had a corresponding reduction in a long form of OPA1 that promotes mitochondrial fusion, which does not occur in mice targeting Drp1 exons 3–5.⁴⁴

We are not aware of reported circulating PCSK9 concentrations in any patients with DRP1 mutations, and serum cholesterol measurements are typically not reported in these paediatric patients. However, in line with our present study in which serum cholesterol was modestly reduced in Drp1LiKO mice, some humans with no wild-type DRP1 have been reported with serum lipids within a normal range.⁴⁵ DRP1 inhibition in our experimental conditions was similar to PCSK9 heterozygosity in mice. Drp1LiKO mice had a 78.5% reduction in circulating PCSK9, a result similar to a 70% reduction in circulating PCSK9 and 30% increase in LDLR in heterozygous PCSK9-deficient mice.⁴⁶ Complete PCSK9 loss-of-function mutations in humans exhibit strong reductions (>80%) in LDL cholesterol, whereas heterozygous patients exhibit mild (15–28%) reductions in LDL cholesterol.³³ Notably, heterozygous PCSK9 loss-of-function still confers reduced (46–88%) cardiovascular disease risk.³³

Serum lipids in Drp1LiKO mice on a high-cholesterol diet have not been reported; however, Drp1LiKO mice on a high-fat diet have nearly half the level of serum cholesterol as controls as well as reduced very-low density lipoprotein secretion.¹⁷ Mdivi-1 reduced the secretion of several apolipoproteins in HepG2 cells, supporting an involvement of DRP1 in lipoprotein secretion. We observed a modest but significant reduction in LDL cholesterol and a non-significant decrease in HDL; with HDL reduced in four of the five Drp1LiKO mice, we examined on a normal laboratory diet that likely contributed to our observed significant reduction in total cholesterol.

Humans heterozygous for DRP1 deficiency exhibit no abnormal clinical pathology.⁴⁵ A small number of humans without any wild-type DRP1 survive post-natally with mutations in DRP1, including dominant negative and complete loss-of-protein mutations that result in a wide range of pathologies from normal early development to distinct neuronal development issues.⁴⁵ How additional mitochondrial dynamics proteins respond to various changes in DRP1 are a possible mechanism for observed effects in different experimental models and might explain highly heterogeneous pathology in DRP1-deficient humans.

Supplementary material

Supplementary material is available at Cardiovascular Research online.

Data availability

The data underlying this article are available in the article and in its online Supplementary Material.

Authors’ contributions

M.A.R. conceived the study and wrote the manuscript. E.A. contributed to the concept and manuscript writing. E.A. and M.A. provided overall supervision and funding for the study. M.N. and L.W. provided Drp1LiKO and flox/flox mouse tissues, blood glucose levels, and body weight. A.D. analyzed mouse cholesterol by FPLC. M.A.R., J.D.H., T.O., C.G., S.A.S., A.H., F.S., H.H., M.C.W., and A.K.M. collected and analyzed all other data. All authors contributed to manuscript editing.

Acknowledgements

We thank L.H.L. for statistical analysis assistance and I.Y. for pharmacokinetics assistance.

Funding

This work was supported by research grants from Kowa Company, Ltd. to M.A.; National Institutes of Health grants (grant numbers R01HL136431, R01HL147095, and R01HL141917) to E.A.; the Japanese Society for the Promotion of Science KAKENHI and the Medical Research Encouragement Prize of the Japan Medical Association to M.N.

Conflict of interests: M.A.R., M.A., and E.A. are inventors on a related patent [WO/2018/052891]. H.H. and T.O. are Kowa Company, Ltd employees, and worked at Brigham and Women’s Hospital when this work was conducted.

References