RNF130 Regulates LDLR Availability and Plasma LDL Cholesterol Levels

Meet the First Author, see p 793

Editorial, see p 864

Elevated plasma low-density lipoprotein cholesterol (LDL-C) levels are a well-established risk factor for cardiovascular disease.¹ Plasma LDL particle concentration is regulated by the balance between synthesis and uptake from the circulation via the LDL receptors (LDLRs).² The liver expresses over 70% of whole body LDLR and is the major determinant of plasma LDL-C concentrations.^3,4 As such, regulation of LDLR is complex and under tight control by multiple pathways.

RNF130 (ring finger containing protein 130), also known as GOLIATH, is the mammalian homolog of the Drosophila protease-associated (PA) domain-containing E3 ligase Goliath (dGoliath). E3 ligases are part of the cellular ubiquitin system that not only controls the selective degradation of proteins by the 26S proteasome, but can also modify proteins and regulate their cellular localization.⁵ E3 ligases catalyze transfer of ubiquitin to the target protein, thereby conferring specificity and regulation to these processes. Most E3 ligases are cytosolic, presumably to increase their exposure to substrates, E2 conjugating enzymes, and ubiquitin, thus allowing for rapid modification and regulation of target proteins. Despite there being over 600 RING (really interesting new gene) E3 ligases,⁶ only a small number, that includes RNF130, contain a bona fide TMD (transmembrane domain).⁷ The function and molecular targets of RNF130 are unknown.^8,9

RNF130, like the other members of the PA domain-containing E3 ligase subfamily, exhibits a distinct domain architecture, consisting of a signal peptide, a PA domain, a transmembrane domain, and a RING domain.¹⁰ Ubiquitination events mediated by Drosophila Goliath family members have been shown to regulate specific cellular events such as endocytosis^7,11 and protein-protein interactions.^12,13 Given the highly conserved structural architecture between RNF130 and other PA-TM-RING proteins, we hypothesized that RNF130 may play a role in the endocytic recycling of membrane receptors.

In the present study, we identify RNF130 as a novel posttranslational regulator of hepatic LDLR and plasma LDL-C levels. Using a combination of molecular, biochemical, and in vivo metabolic analyses, we have determined that RNF130 ubiquitinates the LDLR, reduces plasma membrane LDLR localization and LDL uptake, and increases plasma LDL-C levels. We also demonstrate, using 3 different independent in vivo approaches, that loss of RNF130 in mice results in increased hepatic levels of LDLR and reduced plasma LDL-C levels. We propose that RNF130 allows for the modulation of LDLR endocytosis and recycling, and therefore, the regulation of available LDLR molecules at the cell surface.

Methods

Data Availability

The authors declare that all supporting data are available within the article and its Supplemental Materials.

Mice, Diets, and Treatments

Wild-type (WT) C57BL/6 mice were purchased from The Jackson Laboratory (JAX, No. 00664). Ldlr^−/− and Pcsk9^−/− mice on a C57BL/6 background were originally purchased from The Jackson Laboratory. Mice with a targeted disruption of Rnf130 were generated using embryonic stem (ES) cells from the European Conditional Mouse Mutagenesis program and the European Mouse Mutant Cell Repository with a knockout (KO)-first, conditional-ready cassette targeting exon 3 of Rnf130. To obtain germline transmission of the Rnf130 KO-first Tm1a allele (Rnf130^KO1-Tm1a) on a C57BL/6 background, Rnf130^KO1-Tm1a ES cells were injected into blastocysts and the resulting chimeras were bred with C57BL/6 mice. Germline transmission was confirmed by PCR genotyping for a minimum of 10 litters. All animals were maintained on normal rodent diet (5001; Ralston Purina Company) at UCLA on a 12 h/12 h light/dark cycle with unlimited access to food and water. All animals used in the study were male mice 8 to 10 weeks of age. As adenovirus and adeno-associated virus (AAV) transduction is known to be different between male and female mice, only male animals were used.¹⁴ Animals were randomly assigned to control or treatment groups. Human RNF130 was cloned as described below and shuttled into adenoviral expression vectors. Adenovirus particles were prepared using the AdEasy system (Agilent) and purified by cesium chloride gradient centrifugation. The virus was dialyzed for 48 hours and stored at −80 °C. Particles were quantified by serial dilution methods by detection of plaques in HEK293Ad (human embryonic kidney 293 adherent cells; Agilent) cells. Adenovirus (10⁹ plaque forming unit [PFU]) was delivered by tail-vein injection. Mice were treated with a single injection of either control adenovirus, or adenovirus expressing WT or mutant RNF130 on day 0, as indicated in the figure legend, and mice were euthanized 7 days later. The plasmids required for the manufacture of AAV, pAdDeltaF6 adeno helper plasmid (PL-F-PVADF6) and pAAV2/8 (PL-T-PV0007), were obtained from the University of Pennsylvania Vector Core. AAV was prepared by the triple-transfection method¹⁵ in HEK293T cells (ATCC [American Type Culture Collection], CRL-3216) and purified by cesium chloride gradient as previously described.¹⁶ Aliquots of concentrated virus were stored at −80 °C until injection. Viral titers were determined by quantitative PCR following DNAse digestion and normalized to a standard curve of known genome copies. AAV (5×10¹¹ genome copies) was delivered by intraperitoneal injection. A guide RNA (gRNA) sequence specific to Rnf130 was used to target mouse Rnf130 by CRISPR (clustered regularly interspaced short palindromic repeats)/Cas9 (5′gactgcacagtgatcgaagt). Gen. 2.5 16-mer ASO (antisense oligonucleotides) targeted to mouse Rnf130 (5′attctgttatcatgac) or control sequences were synthesized and purified by Ionis Pharmaceuticals (Carlsbad, CA). ASOs were delivered by intraperitoneal injection twice weekly at a dose of 25 mg/kg for 4 weeks (as described in the figure legends).

Analysis of CRISPR-Based Gene Disruption

Liver genomic DNA was isolated using the Qiagen DNeasy kit. DNA (100 ng) was subjected to quantitative PCR using primers specific to the AAV-Control and AAV-gRNA vectors to assess the relative amount of virus present in the liver. A standard curve from plasmids used for virus production was used to quantify AAV genomes per microgram of DNA.

Off-target sites for Rnf130 gRNA were determined using the online bioinformatics tool, COSMID (CRISPR Off-Target Sites With Mismatches, Insertions, and/or Deletions) at https://crispr.bme.gatech.edu/ as previously described.¹⁶ For on-target editing, 50 μg cDNA was amplified using primers surrounding the gRNA target site. The resulting band was gel purified (Clontech) and sequenced by Sanger Sequencing. Editing efficiency was estimated using Synthego ICE (Inference of CRISPR Edits) https://www.biorxiv.org/content/early/2019/01/14/251082.

Plasma Analysis

High-density lipoprotein (HDL) and LDL/very low-density lipoprotein (VLDL) fractions were separated using a manganese chloride-heparin precipitation.¹⁷ Subsequent fractions and total cholesterol were measured using a colorimetric assay (Infinity Cholesterol Reagent; Thermo Scientific). Plasma cholesterol lipoprotein profiles were determined from individual mice using the modified column lipoprotein profile method as previously described.^18,19 Lipoprotein profiles are plotted as mean absorbance unit±SEM for all animals in each treatment group.

RNA Analysis

Liver tissue was homogenized and total RNA extracted using QIAzol reagent (Invitrogen Life Technologies). Five hundred nanograms total RNA was reverse transcribed using the High Capacity cDNA Reverse Transcriptase Kit (Applied Biosystems) and gene expression determined using a Lightcycler480 real-time quantitative PCR machine and SYBR-Green mastermix (Roche). Relative gene expression was determined using an efficiency corrected method, and efficiency was determined from a 3-log serial dilutions standard curve made from cDNA pooled from all samples. Results were normalized to 36b4 mRNA. Primer sequences are available upon request.

Protein Analysis

Liver tissue was homogenized and protein extracted using RIPA buffer with protease inhibitor cocktail mix (1 complete ethylenediaminetetraacetic acid (EDTA)-free protease inhibitor tablet [Roche], 25 μg/mL calpain inhibitor [Sigma], and 200 μmol/L polymethylsulfonyl fluoride [PMSF; Sigma]). Twenty-five to fifty micrograms protein was separated on an SDS-PAGE gel (BioRad) and transferred to a polyvinylidene fluoride membrane. Membranes were incubated overnight with antibodies to EGFR (epidermal growth factor receptor; 1:1000; Invitrogen), FLAG (1:1000; Sigma), RNF130 (1:1000; Novus Biologicals), GFP (green fluorescent protein; 1:1000; Santa Cruz Biotechnology), HA.11 (1:1000; Covance), LDLR (1:1000; Cayman), LRP1 (1:1000; Abcam), PDI (protein disulfide isomerase; 1:1000; Cell Signaling), and Transferrin Receptor (1:1000; Invitrogen). Proteins were detected with horseradish peroxidase (HRP)-conjugated secondary antibodies (1:10 000; GE Healthcare) and visualized using an AI600 Imager (GE Healthcare). Densitometry analysis was performed using ImageQuant (TL 8.1, GE Healthcare).

Plasmids and Expression Constructs

The pDEST47-hLDLR and K1/6/20RC29A mutant plasmids were a kind gift from Dr Peter Tontonoz.^20,21 The pcDNA3.1-(HA-Ubiquitin)₆ plasmid was a gift from Dr James Wohlschlegel (UCLA, United States). Human RNF130 was amplified (forward 5′ atgagctgcgcggggcgggcgggccctgccc; reverse 5′ gcttgaatgctaatgaggtagaatggttttga) from Hep3B (Hepatoma 3B; ATCC) cell cDNA using HiFi DNA polymerase (KAPA Biosciences). Human RNF130 was also generated with a 3×FLAG epitope engineered at amino acid position 28 after the signal peptide as a double-stranded DNA gBlock (IDT Technologies). Additionally, the C304A mutation was introduced into the RING domain of RNF130. Restriction digests and DNA sequencing were used to confirm all constructs used in this study. For adenoviral studies, full length human RNF130 was subcloned from pcDNA3.1 into the adenoviral vector.

gRNA Design

Staphylococcus aureus gRNAs targeting murine Rnf130 were designed by examining the coding sequence using SnapGene. Preliminary off-target prediction was completed using COSMID.²² The search terms used for the off-target search aimed to return the highest possible number of off-targets, using a NNGRR protospacer adjacent motif (PAM) instead of NNGRRT, and allowing 3 mismatches and 2 insertions or deletions within the gRNA and PAM sequence. Of the available gRNAs, we selected, cloned, and tested the gRNAs with the fewest predicted off-targets. The plasmid 1313 pAAV-U6-BbsI-MluI-gRNA-SA-HLP-SACas9-HA-OLLAS-spA (No. 109304; Addgene) was a gift from Dr William Lagor at Baylor College of Medicine.²³ This AAV vector backbone contains inverted terminal repeats surrounding a U6 promoter driving gRNA expression and a hybrid liver-specific promoterdriving expression of SaCas9 (Staphylococcus aureus Cas9). The gRNA was cloned into this vector by ligating annealed oligos matching the desired gRNA sequence into the BbsI cloning site behind the U6 promoter. The Control AAV-CRISPR used in experiments is the parent vector containing the BbsI cloning site. When the sequence including the BbsI site is queried using COSMID for the SaCas9 NNGRR PAM, no possible off-targets are returned.

Cell Culture and Transfection

HEK293T, Hep3B, and HepG2 (Hepatoma G2) cells were obtained from ATCC (CRL-3216, HB-8064, HB-8065). Cells were maintained in Dulbecco’s modified eagle medium (DMEM) containing 10% FBS. For transfections, cells were plated onto cell-culture-treated plates or coverslips (Corning) at 60% confluence on day 0. Cells were transfected with a total of 1 μg combined DNA plasmids using FuGENEHD (Promega) according to the manufacturer’s instructions. After 24 to 48 hours incubation, cells were collected for downstream processing (Western blot, immunoprecipitation) or fixed and mounted for imaging. Quantification of confocal images was performed using ImageJ.

Cell Surface Biotinylation

For cell surface biotinylation, HEK293T cells were transfected with plasmids encoding GFP-tagged LDLR, HA-tagged ubiquitin, and N-terminally FLAG-tagged WT or mutant (C304A) RNF130 as described above. After 36 hours, cells were washed in PBS++ (phosphate buffered saline with 0.02 mmol/L CaCl₂ and 0.1 5mmol/L MgCl₂) and then incubated for 30 minutes on ice with EZ-link SulfoNHS-SS Biotin (diluted in PBS++). The cells were washed in PBS++, and the reaction was quenched for 30 minutes at 4 °C in quenching buffer (PBS++ with 100 mmol/L glycine). Biotin-modified proteins were immunoprecipitated with NeutrAvidin streptavidin beads overnight at 4 °C. The following day, biotin-modified proteins were collected by centrifugation at 5000g for 5 minutes at 4 °C. Intracellular, unmodified proteins were collected from the supernatant of the 5000g spin. The streptavidin beads were washed 3× in PBS++ before proteins were removed from the beads by incubation at 42 °C for 20 minutes in Laemmli sample loading buffer supplemented with β-mercaptoethanol. Equal percent volume of individual fractions were subject to immunoblotting.

Immunoprecipitation

Cells were plated and transfected as described above. Total cell lysates were prepared in radioimmunoprecipitation assay (RIPA) buffer as described above. Lysates were cleared by centrifugation at 4 °C for 10 minutes at 10 000g. Protein concentration was determined using the bicinchonic acid (BCA) Assay (BioRad) with bovine serum albumin as a reference. To immunoprecipitate LDLR-GFP, equal amounts of protein of cleared lysate were incubated with anti-GFP polysera (1:1000) overnight prior to addition of protein-G agarose beads (Santa Cruz Biotechnology) for an additional 2 hours. Subsequently, beads were washed 3× with RIPA buffer supplemented with protease inhibitors. All incubations and washes were done at 4 °C with rotation. Proteins were eluted from the beads by incubating in Laemmli sample buffer at 70 °C for 30 minutes.

DiI-LDL Uptake Assay

Cells were plated on coverslips and transfected as above. After 24 hours, cells were treated with 50 μg/mL diI-LDL (3,3’-dioctadecylindocarbocyanine LDL) for 30 minutes at 4 °C. Subsequently, cells were placed at 37 °C for 4 hours prior to fixation. After 4 hours, cells were washed, fixed in 4% PFA for 15 minutes, and coverslips mounted onto slides before imaging using a 63× lens under immersion oil. Images were captured using a Zeiss AxioCam 506 camera, and postimage analysis and quantification was performed using ImageJ.

Imaging

Cells were plated and transfected as described above. At harvest, cells were washed, fixed in 4% PFA for 15 minutes, and then washed again before briefly being stored in PBS. Coverslips were mounted on to slides and imaged using a 63× lens under immersion oil. Images were captured using a Zeiss AxioCam 506 camera, and postimage analysis and quantification was performed using ImageJ. Images from 3 independent replicate experiments were analyzed and quantified. Representative images were chosen that best illustrated the mean/average of the quantification.

Statistical Analysis

Statistical analysis was performed using Prism Graphpad software (V8.0). All results are reported as mean±SEM, as stated in the figure legends. A P≤0.05 was considered significant, and statistical significance is shown as described in the figure legends. Gaussian distribution was determined using the Shapiro-Wilk normality test for sample size n>6. Data determined to be parametric were analyzed by unpaired 2-tailed Student t test to compare 2 independent groups or by 2-way ANOVA followed by Tukey HSD posthoc multiple comparison analysis for >2 groups. Data determined to be nonparametric were analyzed by Mann-Whitney U test for comparison of 2 groups and Kruskal-Wallis test with posthoc Dunn test for comparison of >2 groups. For sample size n<6, nonparametric tests (Mann-Whitney U test for comparison of 2 groups and Kruskal-Wallis with Dunn correction for comparison of >2 groups) were used. Outliers were determined using the robust regression and outlier removal (ROUT) outlier test with an false discovery rate (FDR) of 1%. Additionally, animals were removed from analyses due to attrition from in vivo studies because of excessive weight loss (>20% body weight) or noted mis-injects of adenovirus, AAV, or ASO.

Study Approval

All animal experiments were approved by the Office of Animal Research Oversight and the Institutional Animal Care and Use Committee at the University of California Los Angeles.

Results

RNF130 Expression Decreases Hepatic LDLR Levels

Members of the Rnf130 family of PA domain E3 ligases have previously been implicated in the regulation of endocytic recycling events.⁷ To determine whether RNF130 is also localized to endocytic compartments, we transfected HEK293 and HepG2 cells with plasmids encoding a GFP-tagged endosomal marker (Rab27-GFP) and an N-terminally FLAG-tagged human RNF130. Since the coding sequence for RNF130 contains a predicted signal peptide (amino acids 1–27) and addition of a tag on the carboxy-terminal end of the protein renders it nonfunctional (data not shown), we engineered a 3×FLAG-tag immediately after the signal peptide. Similar to other Rnf130 family members including Drosophila Godzilla^CG10277 and human RNF167, we observed that RNF130 was localized to punctate vesicular structures that colocalized with Rab27, a marker of endosomal compartments (Figure 1A), suggesting RNF130 may also play a role in regulating endocytic events.

• Download figure
• Download PowerPoint

Using a panel of tissues isolated from C57BL6/J mice, we determined that Rnf130 is most highly expressed in liver, adipose, and brain, with ubiquitous expression in other tissues (Figure S1A). To understand the effects of modulating RNF130 expression on endocytosis in vivo, we generated adenovirus particles to overexpress human RNF130 in the livers of mice. WT mice were injected once on day 0 with either control adenovirus (Ad-Ctr) or adenovirus expressing human RNF130 (Ad-RNF130) and tissues harvested 7 days later (Figure 1B). To determine whether RNF130 grossly affected endocytosis, we first measured the levels of endocytosed receptors. To our surprise, overexpression of RNF130 (Figure 1C) significantly decreased levels of LDLR (Figure 1D and 1E), without changing the levels of other endocytosed receptors, EGFR, TfR (transferrin receptor), LRP-1 (LDLR-related protein 1), suggesting some specificity for the regulation of LDLR. Changes in LDLR protein (Figure 1D and 1E) were also independent of any change in Ldlr mRNA levels (Figure 1F), suggesting RNF130 regulates LDLR levels via a posttranscriptional mechanism.

Since hepatic LDLR is the major determinant of circulating plasma LDL-C levels, we also measured levels of plasma lipids. RNF130 overexpression resulted in significantly increased total (Figure 1G) and non-HDL (LDL/VLDL) plasma cholesterol levels (Figure 1H). Fast performance liquid chromatography (FPLC) analysis of plasma samples from individual mice demonstrated a significant increase in the LDL-C fraction (Figure 1I). In contrast, plasma triacylglycerol and HDL-C levels were either modestly decreased (Figure S1B) or unchanged (Figure S1C).

RNF130 Is a RING-Dependent E3 Ligase That Ubiquitinates LDLR and Redistributes LDLR Away From the Plasma Membrane

We next sought to determine the molecular mechanism underlying the regulation of LDL-C by RNF130. Like other members of the Rnf130 family, RNF130 has been reported to function as an E3 ubiquitin ligase.⁸ Therefore, we hypothesized that LDLR may be ubiquitinated by RNF130. To test this hypothesis, we transfected HEK293 cells with plasmids encoding GFP-tagged LDLR, HA-tagged ubiquitin, and, where indicated, an N-terminally FLAG-tagged human RNF130. As shown in Figure 2A, total LDLR protein was decreased in cells coexpressing RNF130 (Figure 2A; input) which was accompanied by greatly increased ubiquitination of LDLR (Figure 2A; ubiquitination).

• Download figure
• Download PowerPoint

LDLR has multiple highly conserved potential ubiquitination sites within its cytoplasmic tail (Lys⁸¹¹, Lys⁸¹⁶, Lys⁸³⁰, and Cys⁸³⁹). We repeated the experiments described above but using an LDLR construct where these 4 residues were mutated to arginine or alanine, respectively. Combined mutation of all 4 residues (K811/816/830R/C839A) prevented both ubiquitination of LDLR (Figure 2B; ubiquitination) and the loss of LDLR-GFP protein (Figure 2B; input). Consistent with our in vivo observations (Figure 1D), overexpression of RNF130 had no effect on levels of additional endocytosed receptors EGFR, TfR, and the related lipoprotein receptor LRP-1 (Figure S2A), further supporting that this ubiquitination event is specific to LDLR and not a gross change in receptors known to be internalized via clathrin-mediated endocytosis.

Having determined that RNF130 requires specific residues in the cytoplasmic tail of LDLR to be intact for a ubiquitination event, we next turned to assessing whether the E3 ubiquitin ligase function of RNF130 was essential for this observation. The catalytic domain required for ubiquitination activity of RING E3 ligases is the RING domain, and specific mutations in this domain have been shown to impair ubiquitination of target proteins.^6,24 RNF130 containing a single-point mutation in the RING domain (Cys³⁰⁴ →Ala³⁰⁴) failed to ubiquitinate LDLR (Figure 2C; ubiquitination) or lower LDLR protein levels (Figure 2C; input). Further, the levels of FLAG-tagged RNF130 were elevated when the RING domain was mutated, compared with WT RNF130 (Figure 2C; input) with no significant difference in RNF130 mRNA expression (Figure S2B), suggesting that RNF130 can also catalyze its own ubiquitination and degradation. Such self-ubiquitination is in agreement with previously published reports of RNF130^8,9 and other RING E3 ligases.

A significant percent of LDLR protein is normally expressed at the cell surface which is important for its function in the endocytic pathway that internalizes LDL-C particles.²⁵ To determine whether RNF130 affects LDLR abundance at the cell surface, we transfected HEK293 cells with plasmids encoding GFP-tagged LDLR, HA-tagged ubiquitin, and N-terminally FLAG-tagged WT or mutant (C304A) RNF130. Transfected cells were exposed to biotin to label cell surface proteins before quenching the reaction to prevent subsequent modification of intracellular proteins released after cell lysis. As expected, WT FLAG-RNF130 but not mutant FLAG-RNF130^C304A, reduced GFP-tagged LDLR protein levels in whole cell lysates (Figure 2D; immunoblot:GFP lanes 2–4). WT RNF130 decreased the abundance of GFP-tagged LDLR at the cell surface (Figure 2D; immunoblot:GFP lane 7 versus 6). In contrast, levels of GFP-LDLR were not reduced in cells following coexpression of mutant RNF130^C304A (Figure 2D; immunoblot:GFP lane 8 versus 6), consistent with the inability of mutant RNF130^C304A to degrade LDLR. Analysis of intracellular LDLR levels indicated that there was no change in intracellular LDLR levels following coexpression of WT RNF130 (Figure 2D; immunoblot:GFP lane 11 versus 10). Together, these data suggest that WT RNF130 has proportionally greater effects on LDLR abundance at the cell surface.

In addition to signaling for degradation in the proteasome, protein ubiquitination can also serve as an internalization signal for specific proteins at the plasma membrane.^7,26 As effects of RNF130 on LDLR protein levels were greater at the cell surface, we hypothesized that the ubiquitination of LDLR by RNF130 may result in redistribution of LDLR away from the cell surface, in addition to possibly increasing degradation of LDLR. Immunocytochemical staining showed that coexpression of RNF130 dramatically decreased the levels of LDLR at the cell surface (Figure 2E; middle). Although there was still LDLR present in intracellular compartments (Figure 2E; middle), quantification of mean fluorescent intensity demonstrated a significant reduction in GFP fluorescence indicating that total levels of GFP-tagged LDLR were also decreased with coexpression of RNF130 (Figure 2F). In contrast to WT RNF130, cells expressing mutant RNF130^C304A continued to express high levels of LDLR both at the cell surface and in intracellular compartments (Figure 2E; right panel and Figure 2F). Additionally, levels and localization of mutant LDLR-GFP (MUT-LDLR-GFP) were unchanged in cells expressing WT RNF130 (Figure S2C and S2D).

The data of Figure 2D through 2F demonstrate that RNF130-mediated ubiquitination of LDLR results in the redistribution of LDLR away from the plasma membrane. We next used immunocytochemical staining to ask what intracellular compartments LDLR is redistributed to. These studies demonstrated that LDLR is redistributed to intracellular vesicles that colocalize with NPC-1 (Niemann Pick C Type-1), a marker of late endosomes and lysosomes (Figure S2E), consistent with lysosomal degradation of LDLR. We also probed the degradative mechanism that occurs using inhibitors of proteasomal degradation. Consistent with previous findings that LDLR degradation occurs in the lysosome, the proteasome inhibitor MG132 (Cbz-leu-leu-leucinal) stabilized RNF130 protein but did not stabilize LDLR (Figure S2F). Finally, we asked whether RNF130-mediated ubiquitination of LDLR affects uptake of LDL particles. We cotransfected HepG2 cells with LDLR-GFP and either control plasmid or plasmid expressing WT RNF130. Following transfection, cells were incubated in the presence of diI-LDL particles to determine LDL uptake. These studies demonstrated that LDL uptake was significantly reduced in cells expressing RNF130, likely as a consequence of RNF130-mediated LDLR ubiquitination and redistribution away from the plasma membrane (Figure S2G and S2H).

E3-Ligase Activity of RNF130 Is Required for the Regulation of Hepatic LDLR and Plasma LDL-C

To assess whether the E3 ubiquitin ligase function of RNF130 was required for its ability to increase plasma LDL-C in vivo, we treated WT mice with either control or adenovirus overexpressing FLAG-tagged WT RNF130 or RNF130 containing the C304A point mutation in the RING domain (Figure 3A). Consistent with previous studies (Figure 1), FLAG-RNF130 overexpression in WT mice (Figure S3A and S3B) increased total and LDL cholesterol levels (Figure 3B and 3C), and this was associated with a decrease in LDLR protein (Figure S3A and S3C). In contrast, expression of mutant RNF130^C304A (Figure S3A and S3B) failed to alter plasma cholesterol levels (Figure 3B and 3C) or LDLR protein (Figure S3A and S3C). Collectively, these experiments together with the data of Figure 2, demonstrate that RNF130 regulates LDL-C through its function as a ubiquitin ligase, and that the consequence of this ubiquitination is likely a combination of redistribution of LDLR from the plasma membrane and degradation of LDLR protein.

• Download figure
• Download PowerPoint

To determine if the RNF130-mediated increase in LDL-C is dependent on the presence of hepatic LDLR, we overexpressed RNF130 in LDLR knockout (Ldlr^−/−) mice (Figure 3D). Overexpression of RNF130 (Figure S3D) in Ldlr^−/− mice did not change plasma total cholesterol (Figure 3E and 3F), LDL/VLDL-C (Figure 3F, Figure S3E), HDL-C, or triacylglycerol levels (Figure S3F and S3G), demonstrating that LDLR is required for the RNF130-dependent increase in LDL-C levels in vivo.

Efficacy of RNF130-Mediated Regulation of Plasma Cholesterol Levels Is dependent on Hepatic LDLR Abundance

Based on these data, we hypothesized that overexpression of RNF130 may have more potent effects on plasma LDL-C in mouse models that exhibit elevated endogenous levels of hepatic LDLR protein compared with WT mice.

We overexpressed RNF130 in Pcsk9 KO (proprotein convertase subtilin/kexin type 9; Pcsk9^–/–) mice, a model of increased hepatic LDL receptors.^27,28 PCSK9 normally binds to the LDLR/LDL-C complex as it is being internalized and signals for the LDLR to be degraded in the lysosome, instead of being recycled back to the plasma membrane.²⁹ Consequently, compared with Pcsk9^+/+ mice, Pcsk9^−/− mice have elevated levels of hepatic LDLR protein and significantly reduced levels of plasma LDL-C.²⁸ Overexpression of RNF130 in Pcsk9^−/− mice (Figure 4A and 4B) resulted in increased plasma total cholesterol (Figure 4C and 4E), LDL/VLDL-C (Figure 4D and 4E), and HDL-C levels (Figure S4A), with no change in plasma triacylglycerol levels (Figure S4B). Consistent with data from WT mice (Figure 1E and 1F), in the absence of any change in Ldlr mRNA (Figure 4F), RNF130 overexpression resulted in significantly decreased LDLR protein levels (Figure 4G).

• Download figure
• Download PowerPoint

Together these data show that overexpression of RNF130 in WT and Pcsk9^−/− mice increased LDL/VLDL-C levels by 36% (Figure 1H and 1I) and 70% (Figure 4D and 4E), respectively, suggesting that the efficacy of RNF130 for increasing LDL/VLDL-C is enhanced when basal hepatic LDLR levels are elevated. Together these data demonstrate that overexpression of RNF130 increases plasma LDL-C by a process dependent on both the expression and availability of hepatic LDLR.

Partial Knockout of Rnf130 in Mice Results in Increased Hepatic LDLR and Reduced Plasma Cholesterol

Thus far, we have shown that RNF130 gain-of-function results in reduced hepatic LDLR and increased plasma LDL-C levels. To perform the reverse loss-of-function (LoF), we first generated mice with a targeted disruption of Rnf130 (Figure S5A; Tm1a allele). Germline transmission was confirmed by PCR genotyping (Figure S5B) for a minimum of 10 litters.³⁰ These knockout-first, conditional-ready mice have the potential to generate global knockout animals (Figure S5A; Tm1b allele) by crossing with mice expressing Cre recombinase or conditional-ready animals with a floxed allele (Figure S5A; Tm1c allele) by crossing with mice expressing Flp recombinase. We attempted to generate mice with both Tm1b and Tm1c alleles (Figure S5A). However, we were unable to obtain any mice with the correct genotype. Upon sequencing of the ES cells used to generate the Tm1a mice (Figure S5A), we discovered that the downstream loxP site (Figure S5A; solid red triangle) required for these recombination events was not present in any of the clones analyzed, thus explaining the lack of Tm1b and Tm1c mice. We, therefore, proceeded to investigate the loss of Rnf130 function in vivo using our knockout-first Tm1a allele mice (Figure S5A and S5B).

When heterozygous Rnf130^+/– Tm1a mice were crossed with one another, pups were obtained at the ratio of 1 Rnf130^+/+: 2 Rnf130^+/–: 0.6 Rnf130^−/− (n=35 litters, 204 pups total, mean litter size 5.8 pups per litter) which differs from the expected Mendelian ratio of 1:2:1 (Figure 5A), suggesting that homozygous deletion of Rnf130 is not well tolerated. We next investigated the phenotype of the Rnf130^−/− mice that survived birth. Surviving Rnf130^−/− Tm1a pups were able to nurse, feed, and mature into adults, and the body weight of adult mice at 10 weeks of age was not significantly different from either heterozygous (Rnf130^+/–) or WT (Rnf130^+/+) mice (Figure S5C). Gene expression analysis of livers from Rnf130^+/+, Rnf130^+/–, and Rnf130^−/− mice demonstrated that Rnf130^−/− mice had 30% residual Rnf130 mRNA expression (Figure 5B), suggesting that surviving hypomorphic mice have escaped entire loss of Rnf130. Expression analysis of predicted Rnf130 transcripts and the knockout cassette indicate that a truncated mRNA is not the reason for the 30% residual Rnf130 expression (Figure S5D through S5F). Furthermore, Rnf130^+/– mice only experienced a 30% knockdown of mRNA expression, significantly lower than the 50% that would be expected of heterozygous animals (Figure 5B). We also asked what happened to Rnf130 expression in other tissues in which Rnf130 is also highly expressed, such as the brain (Figure S1A). Gene expression analysis of Rnf130 (Figure S5G), its predicted transcripts (Figure S5H and S5I), and the knockout cassette (Figure S5J) also demonstrated that Rnf130^−/− mice had 30% residual Rnf130 mRNA expression in the brain.

• Download figure
• Download PowerPoint

Given the observed partial knockdown of Rnf130 mRNA (Figure 5 and Figure S5C through S5J), we next wanted to determine what happened to RNF130 protein expression in these mice. We tested all commercially available antibodies to mouse RNF130 and found that none were specific for endogenous mouse RNF130 (Figure S5K through S5O).

Importantly, plasma lipid analysis in Rnf130^−/− mice showed a significant reduction in circulating plasma total cholesterol (Figure 5C) and these changes appeared to be present in the LDL fraction (Figure 5D). Consistent with such a decrease in plasma LDL-C, there was a concomitant increase in LDLR protein expression in Rnf130^−/− mice compared with their WT littermates (Figure 5E).

Liver-Specific Disruption of Rnf130 Reduces Plasma Cholesterol Levels

Given that hepatic LDLR levels have a major role in regulating clearance of LDL-C from the circulation, we next sought to determine whether acute reductions in RNF130 in the livers of adult mice reduced plasma LDL-C levels. We used an all-in-one AAV-CRISPR vector expressing a small gRNA targeting exon 5 of Rnf130, and SaCas9 (Figure S6A). We targeted exon 5 because exon 5 encodes the RING domain, which is critical for RNF130 function (Figures 2 and 3). The CRISPR system was packaged into AAV 2/8 which has a high tropism for liver,^31,32 and we increased liver-specificity further by expressing SaCas9 under the control of a liver-specific promoter (Figure S6A). C57BL/6 mice were assigned into groups and blood samples drawn to ensure equivalent levels of plasma cholesterol at baseline (Figure S6B).

Mice were injected once on day 0 with either control AAV-CRISPR or Rnf130 AAV-CRISPR (Figure 6A) at a dose of 5×10¹¹ genome copies per animal. After 2 weeks, there was no significant difference in body weight between animals treated with either control AAV-CRISPR or Rnf130 AAV-CRISPR (Figure S6C). Hepatic Rnf130 mRNA expression was decreased by 70% in mice treated with Rnf130 AAV-CRISPR (Figure 6B). To ensure comparable viral load had been achieved in both groups, we performed PCR to amplify within SaCas9 (Figure S6D) and measured viral genome copies by quantitative PCR (Figure S6E). These data confirm that viral loads were similar in both groups. To confirm the decreases seen in Rnf130 mRNA were the consequence of CRISPR-mediated gene disruption, we sequenced samples from control- and Rnf130-AAV-CRISPR treated animals and performed editing analyses using Synthego ICE. As expected, we detected INDELs in Rnf130 exon 5 in mice treated with Rnf130 AAV-CRISPR (Figure S6F). These analyses also show increased values for discordant sequences in these mice when compared with control treated animals (Figure S6G), suggesting some INDELs are affecting an area larger than the 8bp measurement window. Together, these data indicate that a proportion of the remaining mRNA is disrupted, and that the observed decrease in Rnf130 mRNA levels are the consequence of CRISPR gene editing. We also wanted to determine effects on RNF130 protein levels following AAV-CRISPR-mediated disruption of Rnf130 mRNA. As in Figure S5, we repeated our protein analysis using all commercially available antibodies in the livers of mice treated with either control- or Rnf130-AAV-CRISPR and found that none were specific for endogenous mouse RNF130 (Figure S6H through S6L).

• Download figure
• Download PowerPoint

Notably, AAV-CRISPR-mediated decreases in Rnf130 mRNA expression were accompanied by significantly decreased plasma cholesterol levels (Figure 6C) that was mainly attributable to decreases in the LDL fraction (Figure S6M). Consistent with previous data in surviving Rnf130^−/− mice (Figure 5), we show increased LDLR protein levels in mice with reduced hepatic Rnf130 expression (Figure 6D and 6E). Taken together, our data show that acute disruption of hepatic Rnf130 in adult mice results in increased hepatic LDLR protein levels and decreased plasma LDL-C levels.

Silencing Endogenous Rnf130 Increases LDLR Protein Levels and Decreases Plasma LDL-C Levels

The studies described above demonstrate that reducing hepatic Rnf130 mRNA expression by at least 70% either by gene targeting (Figure 5) or CRISPR/Cas9 (Figure 6) is sufficient to increase hepatic LDLR protein levels and lower plasma LDL-C levels in mice. Human and mouse RNF130 protein are highly conserved and share 98% sequence identity suggesting potential conserved function between species. The RNF130 locus was previously linked with plasma LDL-C levels in a small population (8090 individuals) of African American patients.³³ To explore whether there were any variants associated with RNF130 in larger data sets, we surveyed multiple summary-level, publicly available association databases, such as UK Biobank and the Global Lipid Genetics Consortium.^34,35 We identified multiple additional variants in the RNF130 locus that were associated with plasma cholesterol (Figure S7A). We also asked whether RNF130 expression was associated with coronary artery disease and cardiometabolic disease outcomes using the STARNET study (Stockholm-Tartu Atherosclerosis Reverse Networks Engineering Task).³⁶RNF130 was expressed in a liver gene expression coregulatory network module that is significantly associated with cardiometabolic outcomes (Figure S7B). We also found that hepatic RNF130 expression was significantly elevated in individuals with coronary artery disease compared with healthy individuals (P=7.2×10⁻¹⁴). Lastly, we compared these linkage results from human studies with results from a diverse panel of 120 inbred mouse strains known as the Hybrid Mouse Diversity Panel. In this panel of mice, we identified significant correlation between hepatic mRNA expression of Rnf130 and plasma lipid traits including total (Figure S7C) and LDL (Figure S7D) cholesterol levels.

To assess the translational potential of our gain-of-function and LoF studies, we used an ASO (Ionis Pharmaceuticals) to silence Rnf130 in vivo. Mice were treated twice weekly with either control ASO or RNF130 ASO at a dose of 25 mg/kg for 4 weeks (Figure 7A). In contrast to other genetic approaches to disrupt Rnf130 expression (Figures 5 and 6), treatment with Rnf130 ASO for 4 weeks reduced hepatic Rnf130 mRNA levels by >90% (Figure 7B). As we observed in Figure S5 and S6, none of the commercially available antibodies were specific for endogenous mouse RNF130 protein (Figure S7E through S7I).

• Download figure
• Download PowerPoint

Rnf130 silencing caused significant decreases in total (Figure 7C), LDL (Figure 7D and 7E), and HDL (Figure 7E, Figure S7J) cholesterol, with no significant change in plasma triacylglycerol, or liver enzymes ALT and AST (Figure S7K through S7M). These decreases in plasma LDL-C concentrations were accompanied by significant increases in hepatic LDLR protein levels (Figure 7F). These data demonstrate that silencing Rnf130 in adult mice using an ASO compound results in increased LDLR protein levels and lower circulating plasma LDL-C levels.

Next, to determine whether these effects on LDL-C were dependent on the availability of LDLR, we repeated our ASO silencing experiment in Ldlr^−/− mice (Figure 7G). As observed in WT mice treated with Rnf130 ASO (Figure 7B), Rnf130 mRNA expression in ASO treated Ldlr^−/− mice was decreased by nearly 90% (Figure 7H). However, silencing Rnf130 in Ldlr^−/− mice had no effect on plasma total and LDL-C concentrations (Figure 7I and 7J), plasma HDL-C and triacylglycerol (Figure S7N and S7O), or plasma ALT and AST (Figure S7P and S7Q), confirming that the effects on plasma cholesterol observed with silencing Rnf130 require expression of LDLR.

Discussion

RNF130 is a member of the RING domain family of E3 ubiquitin ligases, which promote the transfer of ubiquitin from ubiquitin-conjugating enzymes (E2s) to lysine residues in target proteins.⁶ Furthermore, RNF130 is also part of the unique PA-TM-RING family of RING E3 ubiquitin ligases that contain a transmembrane domain. Previous studies of other PA-TM-RING ligase family members RNF167 and RNF13 have shown that these ligases are localized to both the plasma membrane and endosomes.⁷ It was further shown that this localization dictates target specificity and facilitates endocytic recycling of target proteins. In the present study, we show that RNF130 localizes to endosomes and ubiquitinates the LDLR and redistributes LDLR protein away from the plasma membrane. This ubiquitination of LDLR was dependent on both the catalytic RING domain of RNF130 and specific residues in the cytoplasmic tail of LDLR. We use mice with gain-of function of RNF130, together with our in vitro studies to demonstrate that RNF130 ubiquitinates and redistributes LDLR away from the plasma membrane resulting in decreased levels of cell surface LDLR. We hypothesize that as a result, LDLR-mediated clearance of LDL from the plasma is impaired during RNF130 overexpression leading to elevated plasma LDL-C levels. Using 3 independent approaches to reduce Rnf130 expression in vivo, we further provide data demonstrating that RNF130 LoF leads to increased LDLR abundance and reduced plasma LDL-C levels.

Earlier studies demonstrated that IDOL (inducible degrader of the LDLR), another RING E3 ubiquitin ligase, also ubiquitinates LDLR.²¹ Importantly, there are a number of contrasting features between RNF130 and IDOL. RNF130, unlike IDOL, contains a transmembrane domain, restricting its cellular localization. IDOL, however, lacks a transmembrane domain, is cytoplasmic, and contains a FERM binding domain critical for its interaction with LDLR family members. A recent study incorporating a CRISPR screen in human SV589j cells sought to identify regulators of LDL-C movement between the lysosome and endoplasmic reticulum.³⁷ Of note, this latter study identified disruption of both RNF130 and IDOL as upregulating cell surface levels of LDLR, consistent with RNF130 functioning to regulate LDLR protein levels at the cell surface. In contrast to RNF130, deletion of IDOL in mice had no effect on hepatic LDLR protein or plasma LDL-C levels, although such deletion did have effects on LDLR expression in peripheral tissues.³⁸ Based on these findings and data presented in the current report, we postulate that RNF130 functions by ubiquitinating LDLR at the plasma membrane to signal for receptor internalization, and by ubiquitinating endosomal LDLR to prevent recycling to the cell surface. These differences suggest that RNF130 and IDOL may have important, contrasting roles in the regulation of hepatic LDLR and plasma LDL-C.

PCSK9 is an important posttranslational regulator of LDLR levels that functions to reduce expression of LDLR protein.^28,39 Treatment of hyperlipidemic patients with monoclonal antibodies that target and inactivate PCSK9 results in increased hepatic LDLR expression and significantly reduced plasma LDL-C.^40,41 Our finding that overexpression of RNF130 is able to decrease LDLR levels (and concomitantly increase plasma LDL-C levels) in Pcsk9^−/− mice suggests that RNF130’s effects on LDLR protein are independent of PCSK9.

PCSK9 is known to have a complex reciprocal regulatory relationship with LDLR.^27,42 Studies have shown that PCSK9 is most effective in targeting the LDLR when the LDLR is bound to LDL particles.^43,44 Alternative means of posttranslationally increasing the availability of unbound LDLR, such as decreased expression of RNF130, are likely to further promote/enhance clearance of LDL-C and are, therefore, attractive additional therapeutic targets. ASO silencing of Rnf130 resulted in a 50% decrease in LDL-C, a change comparable to inhibiting PCSK9,^40,41 further supporting that RNF130 is a significant regulator of LDLR and LDL-C levels.

Whole body disruption of Rnf130 expression was incomplete and surviving pups exhibited only a 70% decrease in Rnf130 mRNA, suggesting that some residual expression of Rnf130 is necessary for survival. This is in agreement with publicly available human data, in which LoF mutations of the RNF130 gene are significantly lower than the expected odds. Only 9 LoF mutations in RNF130 have been described in the gnomAD database, which is 46% of the expected rate of LoF mutation.⁴⁵ Additionally, currently no homozygotes have been identified for these mutations, suggesting that loss of RNF130 is not well tolerated. While a 70% reduction in Rnf130 mRNA expression in Rnf130^−/− mice was sufficient to significantly decrease LDL-C by 45%, the same was not true in heterozygous mice. Rnf130^+/– mice had only a 30% reduction in Rnf130 expression, significantly less than the expected 50% reduction, but no significant change in plasma LDL-C. Finally, ASO silencing of Rnf130, which induced >90% reduction in hepatic Rnf130 mRNA levels, elicited the largest decrease in plasma LDL-C levels. The increase in hepatic LDLR and decrease in plasma LDL-C after treatment with ASO is comparable to a recently reported gain-of-function mutation in LDLR,⁴⁶ which caused a 74% decrease in LDL-C in humans. Taken together, these data suggest that RNF130 regulates cholesterol levels in a dose-dependent manner.

Most complex traits, such as coronary artery disease and plasma LDL-C, are determined by a number of genetic contributors that each have a modest effect making it difficult to identify genetic variants that increase disease risk, thus hindering the discovery of new drug targets. In fact, genome-wide association study loci only collectively explain ≈15% of the genetic variance in coronary artery disease, suggesting that there are meaningful associations that lie beneath the current genome-wide significance thresholds. Additionally, while genome-wide association studies have been a useful tool for identifying regulatory loci, in most cases the molecular link or the specific gene explaining the association between the predicted loci and the disease remains unknown. Our data demonstrating that RNF130 regulates plasma LDL-C and hepatic LDLR availability point to the importance of understanding the molecular mechanisms underlying subgenome wide significant loci.

In conclusion, our data highlight the complex nature of the posttranslational regulation of LDLR. We have identified RNF130 as a novel posttranslational regulator of hepatic LDLR abundance and activity, that may have important implications for targeting the LDLR pathway to lower plasma LDL-C levels. We further provide preclinical evidence that the RNF130-LDLR pathway could be targeted to enhance LDL clearance from the circulation.

Article Information

Acknowledgments

The authors thank Ionis Pharmaceuticals for providing antisense oligonucleotide compounds. The authors especially thank Dr. Peter Edwards for critical discussion and reading of the manuscript. The authors thank all members of the Tarling-Vallim (past and present) and Tontonoz labs at UCLA for advice, discussion, and sharing reagents.

Author Contributions

E.J. Tarling and T.Q. de Aguiar Vallim conceived and oversaw the project. B.L. Clifford, K.E. Jarrett, J. Cheng, A. Cheng, M. Seldin, P. Morand, T.Q. de Aguiar Vallim, and E.J. Tarling performed experiments. K.E. Jarrett designed and purified adeno-associated virus (AAV-CRISPR). B.L. Clifford and K.E. Jarrett designed and performed AAV-CRISPR experiments. J. Cheng and A. Cheng assisted with animal experiments. P. Morand assisted with Western blotting. M. Seldin performed functional genomic analyses. A. Baldan performed FPLC lipoprotein profiles. R. Lee provided antisense oligonucleotide compounds. Data analysis and statistical analyses were performed by B.L. Clifford and E.J. Tarling. Figures were generated by B.L. Clifford and E.J. Tarling. The article was written by B.L. Clifford and E.J. Tarling. All authors revised and approved the final article.

Supplemental Materials

Figures S1–S7

Footnotes