Angiopoietin-like protein 3 governs LDL-cholesterol levels through endothelial lipase-dependent VLDL clearance

Angiopoietin-like protein (ANGPTL)3 regulates plasma lipids by inhibiting LPL and endothelial lipase (EL). ANGPTL3 inactivation lowers LDL-C independently of the classical LDLR-mediated pathway and represents a promising therapeutic approach for individuals with homozygous familial hypercholesterolemia due to LDLR mutations. Yet, how ANGPTL3 regulates LDL-C levels is unknown. Here, we demonstrate in hyperlipidemic humans and mice that ANGPTL3 controls VLDL catabolism upstream of LDL. Using kinetic, lipidomic, and biophysical studies, we show that ANGPTL3 inhibition reduces VLDL-lipid content and size, generating remnant particles that are efficiently removed from the circulation. This suggests that ANGPTL3 inhibition lowers LDL-C by limiting LDL particle production. Mechanistically, we discovered that EL is a key mediator of ANGPTL3’s novel pathway. Our experiments revealed that, although dispensable in the presence of LDLR, EL-mediated processing of VLDL becomes critical for LDLR-independent particle clearance. In the absence of EL and LDLR, ANGPTL3 inhibition perturbed VLDL catabolism, promoted accumulation of atypical remnants, and failed to reduce LDL-C. Taken together, we uncover ANGPTL3 at the helm of a novel EL-dependent pathway that lowers LDL-C in the absence of LDLR.


INTRODUCTION
Hypercholesterolemia is a major risk factor for atherosclerosis and coronary artery disease (CAD) (1).
Despite advances in drug development designed to lower LDL-cholesterol (LDL-C) levels, substantial risk of cardiovascular disease persists in patients with sub-optimal responses to current medical interventions (2). Treatment options and efficacy of existing therapeutics remain even more limited for individuals with genetic mutations in the LDL-Receptor (LDLR) pathway (FH: "Familial Hypercholesterolemia"). Homozygous FH patients exhibit severely elevated LDL-C levels and develop CAD prematurely (3). While additional lipid-lowering strategies are required to achieve sufficient LDL-lowering in the general population, LDLR-independent pathways for patients lacking functional LDLR represent a particularly urgent need for therapeutic intervention.
By contrast, the mechanism of how ANGPTL3 regulates LDL-C is unclear. Earlier studies revealed that ANGPTL3 blockade lowers LDL-C independently of LDLR (23,24), providing a new therapeutic avenue for the treatment of FH (25). Previous attempts to explore the mechanism of LDL-C reduction with ANGPTL3 inhibition raised questions regarding the contribution of hepatic VLDL production. In mice, ANGPTL3 inhibition was found to be associated with reduced VLDL-TG secretion but had no effect on VLDL-APOB production (24). By contrast, human ANGPTL3 LOF carriers were reported to have reduced VLDL-APOB production rates (6). Studies focused on lipoprotein clearance revealed that APOE or remnant receptors LDL Receptor Related Protein 1 (LRP1) or by guest, on  www.jlr.org Downloaded from 5 administered subcutaneously or intravenously to participants with varying degrees of dyslipidemia (28) (ClinicalTrials.gov Identifier: NCT01749878). Part of the results for Group A of this trial are described in this report.
Group A enrolled healthy men and women 18 to 65 years of age with a fasting triglyceride level of 150 to 450 mg per deciliter (1.7 to 5.1 mmol per liter) or an LDL cholesterol level of 100 mg per deciliter (2.6 mmol per liter) or greater.
Human genetic association analysis The analysis was performed in subjects of European ancestry from the UK Biobank study with available exome sequences. Exome sequencing and variant calling were performed at the Regeneron Genetics Center as previously described (29). Triglyceride levels were log10-transformed. All traits were residualized adjusting for age, age squared, sex, age-sex interaction, and 10 principal components of ancestry and subsequently transformed by rank-based inverse normal transformation. The association was tested using BOLT under linear mixed model (30). The effects in sd units and P-values were derived by testing the association against rank-based inverse normal transformed traits, while the effects in clinical units were derived by testing the association against residualized traits.

Antibodies and antisense oligonucleotides
The fully human anti-ANGPTL3 (evinacumab, originally REGN1500) (11) monoclonal antibody was derived using Regeneron's VelocImmune technology platform (31,32). An isotype-matched antibody with irrelevant specificity was used as control (REGN1945). Antibodies were diluted in saline and administered to mice (10 mg/kg for single dose, or 25mg/kg for multi-dose) by subcutaneous injection. Evinacumab binds ANGPTL3 from multiple species, including mouse and human, with comparable affinity (11). Single antibody administration studies To establish a baseline for serum chemistry parameters, serum samples were collected 7 days prior to antibody administration in the non-fasted state (fed ad libitum). On study day 0, mice were sorted into treatment groups based on their serum LDL-C levels. Mice were administered single s.c. injections of REGN1500 or isotype control antibodies at the indicated doses. Subsequent serum samples were collected from non-fasted mice during the duration of the study and analyzed for serum chemistry parameters, FPLC or HPLC analyses, and APOB levels. Mice were euthanized 1 week after injection and livers were collected and frozen for subsequent lipid content and/or enzymatic measurements.

Multiple antibody administration studies
Non-fasted baseline serum chemistry was determined on chow diet, and mice were sorted into treatment groups based on their LDL-C levels. Mice were then injected with REGN1500 or isotype control antibody (25mg/kg s.c.) once a week for 3 weeks. Body weights were measured weekly. Mice were euthanized 1 week after the last injection and livers were collected and frozen for subsequent lipid content and/or enzymatic measurements, or collected in RNAlater (ThermoFisher) for transcriptome studies. In some studies, mice were placed on high fat diet (Research Diets, D12492; 60% fat by calories) for 2 weeks, serum chemistry was measured again, and mice were then sorted into treatment groups.

Lipid analysis
Circulating TG, total cholesterol (TC), LDL-C, HDL-C, NEFA, alanine aminotransferase (ALT), and aspartate aminotransferase (AST) levels were determined in serum using an ADVIA® Chemistry XPT blood chemistry analyzer (Bayer). Non HDL-C levels were calculated by subtracting HDL-C from TC values. Phospholipids were measured using Phospholipids C assay (Wako Diagnostics). Hepatic TBARS were measured using TBARS Parameter Assay Kit (R&D) according to manufacturer's instructions. Liver lipid levels were evaluated as previously described (11 (1) Lipids analyzed in the methanol extract platform included fatty acids, oxidized fatty acids, bile acids, and lysoglycerophospholipids: Proteins were precipitated from 300μL of the defrosted lipoprotein samples by adding 3 volumes of methanol in 1.5 mL microtubes at room temperature. The methanol used for extraction was spiked with metabolites not detected in unspiked lipoprotein extracts. After brief vortex mixing the samples were incubated 1 hour at -20 C. 1000μL of the supernatants were collected after centrifugation at 18,000 x g for 5 minutes, dried and reconstituted in 70μL methanol before being transferred to vials for UHPLC-MS analysis.
(2) The chloroform/methanol extract platform provided coverage over glycerolipids, cholesteryl esters, sphingolipids and glycerophospholipids: Three volumes of chloroform / methanol (2:1) were added to 400μL of defrosted lipoproteins after a brief vortex mixing. The chloroform / methanol used for extraction was spiked with metabolites not detected in unspiked lipoprotein extracts. After brief vortex mixing the samples were incubated 1 hour at -20 C.
After centrifugation at 18,000 x g for 5 minutes, 800μL of the organic phase was collected and the solvent removed.
Then, the extracts were dried and reconstituted in 100μL acetonitrile / isopropanol (1:1), centrifuged (18,000 x g for 5 minutes), and transferred to vials for UHPLC-MS analysis. Specific chromatographic separation conditions and mass spectrometric detection conditions for each platform have been reported (35).

Quality controls
Different types of quality control (QC) samples were used to assess the data quality.

Data pre-processing and normalization
Following metabolite identification, all data were processed using the TargetLynx application manager for MassLynx 4.1 software (Waters Corp., Milford, USA). A set of predefined retention time, mass-to-charge ratio pairs, Rt-m/z, corresponding to metabolites included in the analysis was fed into the program. Associated extracted ion chromatograms (mass tolerance window = 0.05 Da) was then peak-detected and noise-reduced in both the LC and MS domains such that only true metabolite related features were processed by the software. A list of chromatographic peak areas was then generated for each sample injection. An approximated linear detection range was defined for each identified metabolite, assuming similar detector response levels for all metabolites belonging to a given chemical class represented by a single standard compound (35). Metabolites for which more than 30% of data points were found outside their corresponding linear detection range were excluded from statistical analyses.
Normalization factors were calculated for each metabolite by dividing their intensities in each sample by the recorded intensity of an appropriate internal standard in that same sample, following the procedure described before (36). Reads were decoded based on their barcodes and read quality was evaluated with FastQC (www.bioinformatics.babraham.ac.uk/projects/fastqc/). Reads were mapped to the mouse genome (NCBI GRCm38)

Lipidomics data analysis
using ArrayStudio® software (OmicSoft®, Cary, NC) allowing two mismatches. Reads mapped to the exons of a gene were summed at the gene level. Differentially expressed genes were identified by DESeq2 (37) package and significantly perturbed genes were defined with fold changes no less than 1.5 in either up or down direction and with p-values of at least 0.01.

Statistical analysis
Statistical and graphical data analyses were performed using Microsoft Excel and Prism 7 (Graphpad Software, Inc.).
Data are expressed as mean ± standard error of the mean. Mean values were compared using unpaired two-tailed ttests, one-way or two-way analysis of variance (ANOVA) as implemented in the Graphpad Prism 7.0 software (Graphpad Software, Inc.). In box and whisker plots, the middle line is plotted at the median, the upper and lower hinges correspond to the first and third quartiles, and the upper and lower whiskers display the full range of variation (min. to max.). Grubbs' test was used to determine and remove significant outliers.

EL is necessary for the LDLR-independent LDL-C lowering effect of ANGPTL3 inhibition.
To investigate the LDLR-independent mechanism of LDL-C reduction, we employed monoclonal ANGPTL3 antibody (evinacumab/REGN1500) (11) to inactivate circulating ANGPTL3 in Ldlr-deficient mice. Consistent with previous findings, ANGPTL3 antibody lowered plasma levels of TG and cholesterol independently of LDLR (24) (supplemental Fig. S1A, B). Additional analysis revealed that ANGPTL3 inhibition also lowered serum phospholipid levels ( Fig. 1A), indicative of de-repression of EL, whose phospholipase activity is responsible for HDL-C reduction upon ANGPTL3 inhibition (11). FPLC analysis provided insights into the composition of individual lipoproteins and showed that beyond HDL, ANGPTL3 inhibition also led to reduction in VLDL-and LDL-phospholipids (Fig. 1A, supplemental Fig. S1B). In order to understand if the reduction in phospholipids is driven by their transfer from VLDL to HDL as a consequence of LPL-hydrolysis (38), we evaluated the activity of the responsible enzyme, phospholipid transfer protein (PLTP) (39). We found that evinacumab lowered PLTP activity, making it unlikely to explain the reduction in VLDL-phospholipids (Fig. 1B). Instead, the data suggested that ANGPTL3 inhibition promotes phospholipid hydrolysis on VLDL. While LPL has low phospholipase activity in vitro, both EL and hepatic lipase (HL) are the predominant phospholipases in vivo (reviewed by ref. 20). However, we have previously shown that HL is unaffected by ANGPTL3 (11,40). Therefore, we hypothesized that ANGPTL3 inhibition may enable EL to impact APOB-containing lipoproteins and contribute to the LDLR-independent LDL-C lowering effect.
To test our hypothesis, we employed EL-deficient mice (Lipg -/-), which on chow diet exhibit elevated HDL-C and circulating phospholipids (19) (see HPLC in Fig. 1C, supplemental Fig. S2A). While Lipg -/mice also showed a trend towards increased LDL-C levels (supplemental Fig. S2A), these effects were more difficult to assess, as mice typically have very low LDL-C on chow diet. We therefore subjected Lipg -/mice to high fat diet (HFD, 60% kcal from fat) to elevate baseline LDL-C levels before administering evinacumab. Curiously, ANGPTL3 inhibition reduced LDL-C and total cholesterol to similar extent in WT and Lipg -/mice ( Fig. 1D), suggesting either that EL is entirely dispensable for LDL-C reduction upon ANGPTL3 inhibition, or the availability of LDLR allows for clearance of atherogenic particles in Lipg -/mice and hence masks the role of EL.
To answer this question, we generated mice deficient in both EL and LDLR (Lipg  (41). In addition to further elevated HDL-C, the increase in VLDL-C and LDL-C was particularly notable (Fig. 1C). Lipg -/-Ldlr -/mice also displayed increased APOB levels (present as single copy per VLDL and LDL particle) in comparison to Ldlr -/-(supplemental Fig. S2A). Further analysis revealed that the APOB content of LDL was similar and APOB content of VLDL was increased by ~20% in Lipg -/-Ldlr -/compared to Ldlr -/mice (Fig. 1C). The greatly raised cholesterol content in Lipg -/-Ldlr -/mice ( Fig 1C, supplemental Fig. S2A) contrasted with the minor elevation in particle numbers (judged by APOB content), suggesting that APOB-lipoproteins carried more lipids per particle when EL and LDLR were absent.
Next, we tested the effect of ANGPTL3 inhibition on lipid levels in these mice on chow diet. Administration of evinacumab led to TG reduction in both Ldlr -/and Lipg -/-Ldlr -/mice ( Fig.1E). However, in contrast to EL single ablation, double EL/LDLR knockout rendered evinacumab ineffective in reducing LDL-C, non HDL-C and phospholipids (Fig. 1E). The HPLC analysis further revealed the differential effect of ANGPTL3 inhibition on all cholesterol fractions between Ldlr -/and Lipg -/-Ldlr -/mice, demonstrating EL is necessary for evinacumab-driven, LDLR-independent LDL-C reduction (supplemental Fig. S2B). Taken together, our studies identified EL as key mediator of a novel LDL-C lowering pathway operating in the absence of LDLR.
To evaluate if EL contributes to LDL-C metabolism in humans, we queried the UK Biobank for genetic associations of known LIPG variants with serum lipids (up to 180,374 individuals, supplemental Table S1). The LIPG Asn396Ser variant (MAF=0.0134), which severely blunts EL enzymatic activity, was associated with increased HDL-C (effect: +4.25 mg/dL, P=1.7x10 -101 ), as previously reported (42  Conversely, HDL-TG reduction appeared to be driven by LPL de-repression, as TGs were similarly reduced in Ldlr -/and Lipg -/-Ldlr -/mice (supplemental Fig. S3C). Although HDL was not our main focus, the data supporting a requirement for EL in HDL-phospholipid, but not TG-turnover following ANGPTL3 inhibition helped to validate our approach.

Downloaded from
Intriguingly, the effects of EL ablation on LDL lipid composition and amount were as striking as those seen with HDL (Fig. 2C, first column). Despite equivalent LDL particle numbers between Ldlr -/and Lipg -/-Ldlr -/mice (based on APOB content of LDL fraction, Fig. 1C), LDL from Lipg -/-Ldlr -/mice comprised increased contents of ChoE and glycerophospholipids. Several TG species were also significantly enriched, albeit to a lesser extent (Fig.   2C, first column). Conversely, ANGPTL3 inhibition in Ldlr -/mice reversed this pattern and yielded significant reductions in >85% of individual phospholipid and TG species (Fig. 2C, second column; supplemental Table S3). EL played a critical role in mediating these effects: whereas TG-hydrolysis was similarly effective in Ldlr -/and Lipg -/- Endothelial lipase promotes VLDL catabolism and APOB particle reduction.
To determine if the altered lipid content of LDL was caused by changes in its metabolic precursor, we examined the lipid composition of VLDL. The reduction in VLDL-lipid content was greater than that seen for LDL with ANGPTL3 inhibition (compare 2 middle columns of  Table S4).
On the other hand, dissecting the role of EL de-repression in VLDL-processing and cholesterol reduction turned out to be more complex than we initially envisioned. Unlike for LDL and HDL, evinacumab diminished VLDL-cholesterol to similar extents in Ldlr -/and Lipg -/-Ldlr -/mice, and appeared dispensable for evinacumab-driven VLDL-phospholipid reduction (Fig. 3A, 2 middle columns; Fig. 3B). While at first glance, these data suggested little contribution of EL to VLDL metabolism, closer inspection revealed that ANGPTL3 inhibition lowered VLDLphospholipids by ~2x more in LDLR-vs. EL/LDLR-deficient mice. Phosphatidylcholines were particularly affected, evident from reduced VLDL-PC species in Ldlr -/relative to Lipg -/-Ldlr -/mice (Fig. 3C, supplemental Table S4).

Downloaded from
Given that VLDL is the precursor for LDL, the lipidomic data raised the question of how the extensively processed VLDL (TG low PL low ChoE low ) could give rise to PL/ChoE-enriched LDL (TG low PL High ChoE high ) when EL is missing (Figs. 2C, 2E and 3A). To reconcile these results, we evaluated how ANGPTL3 inhibition affects the relative distribution of VLDL and LDL particles by using APOB levels as a proxy readout. It is noteworthy that APOB and LDL-C levels were similarly influenced by EL (Figs. 1E, 3D): while ANGPTL3 inhibition lowered the total number of APOB-containing particles in Ldlr -/mice (Fig. 3D), it had no effect on APOB levels in Lipg -/-Ldlr -/mice, mimicking the effect on LDL. To determine particle distribution upon ANGPTL3 inhibition, we performed APOB western blots on FPLC-separated serum from chow-fed Ldlr -/and Lipg -/-Ldlr -/mice. In Ldlr -/mice, evinacumab reduced APOB in VLDL and LDL fractions by similar amounts, suggestive of particle clearance (Fig.   3E). In Lipg -/-Ldlr -/mice, ANGPTL3 inhibition lowered APOB in VLDL to the same extent as in Ldlr -/mice.
However, it increased APOB in LDL fractions, indicative of particle accumulation. Hence, whereas total APOB was largely unchanged in the absence of EL, evinacumab triggered APOB-containing lipoprotein re-distribution (Figs. 3D, E). The increased conversion of VLDL to LDL, followed by LDL accumulation suggested generation of atypical remnants and defects in APOB-containing particle clearance when EL was missing. Indeed, the resulting particles represent the PL/ChoE-rich LDL that we identified by mass spectrometry (Fig. 2E).
Taken together, our data suggest that ANGPTL3 inhibition elicits APOB-containing lipoprotein catabolism, and promotes EL-driven modifications to facilitate LDLR-independent particle uptake. Without EL/LDLR, partially processed LDL accumulates and clearance is perturbed (Fig. 3E). That said, the data do not allow us to unequivocally determine which particles undergo clearance. ANGPTL3 inhibition, via de-repression of EL, could either drive VLDL remnant uptake, removing LDL precursor, or it could directly promote LDL clearance to lower LDL-C.

ANGPTL3 inhibition promotes VLDL processing and clearance.
To evaluate the relevance of EL-modifications on LDLR-independent lipoprotein uptake, we performed plasma clearance kinetic studies in Ldlr -/mice. VLDL and LDL were isolated from Ldlr -/or Lipg -/-Ldlr -/donor mice, radiolabeled with [ 3 H]-cholesteryl ether (CE), and injected into Ldlr -/recipients to monitor their rates of disappearance from the circulation (Fig. 4A, left panel). Notably, clearance of VLDL was significantly delayed when donor mice lacked EL. In contrast, the absence of EL did not affect LDL kinetics (Fig. 4A, right panel).

Downloaded from
Next, we assessed how ANGPTL3 influences LDLR-independent lipoprotein clearance. Neither LDL
Again, no differences were seen for LDL clearance in Apoe -/mice. Thus, the mouse data suggested ANGPTL3 inhibition predominantly modifies VLDL catabolism.
To understand if ANGPTL3 inhibition elicited similar effects on VLDL remodeling in humans, we performed lipoprotein analyses using samples of participants of the evinacumab clinical trial (28) (Phase 1 single ascending dose study). Hyperlipidemic human individuals (LDL-C≥100mg/dL, 150≤TG≤450mg/dL) received a single ascending dose of evinacumab or placebo by either subcutaneous or intravenous administration to evaluate the optimal dosing regimen ( Fig. 4C). At the indicated timepoints, blood was collected and subjected to NMR analysis to determine lipoprotein characteristics. The magnitude and duration of lipid reductions were dose-proportional, with evinacumab significantly reducing VLDL/chylomicron particles by 77% at the highest dose tested. The effect on LDL particle reduction was more modest (29%) and similar in magnitude to HDL (33%, Fig. 4C, supplemental Fig. S4E).
Furthermore, NMR provided important insights into changes in lipoprotein size, which was not readily apparent from the mouse studies. Specifically, evinacumab reduced VLDL/chylomicron, but not LDL or HDL particle size in human individuals (Fig. 4D). Thus, the clinical data suggested that evinacumab does not alter LDL remodeling directly and provided further evidence that ANGPTL3 inhibition primarily affects VLDL.
Taken together, our studies emphasized the importance of ANGPTL3/EL in governing VLDL remodeling in hyperlipidemic humans and mice. Whereas VLDL remnants could be generated without EL (Fig. 3), its VLDL by guest, on July 25, 2020 www.jlr.org Downloaded from phospholipid-modifications appeared to be necessary for LDLR-independent particle removal at the juncture between VLDL remnants and their processing to become LDL.

ANGPTL3 has no impact on hepatic lipid homeostasis.
The data above suggested that ANGPTL3 governs LDL-C levels by controlling vascular lipolysis and VLDL clearance. Prior studies have proposed additional, yet conflicting roles for ANGPTL3 in regulating hepatic lipid metabolism. In mice, ANGPTL3 inhibition has been shown to lower liver lipid content (10), and to reduce hepatic VLDL-TG secretion without impacting VLDL-APOB production (24). By contrast, decreased VLDL-APOB production rates were reported in human individuals with ANGPTL3 LOF variants (6). To further evaluate how ANGPTL3 regulates hepatic VLDL production to influence plasma LDL-C, we investigated VLDL assembly in LDLR-deficient mice in greater detail.
Liver transcriptome analysis of Ldlr -/mice revealed no impact of ANGPTL3 inhibition on hepatic mRNA expression, including genes related to APOB-lipidation or lipases EL (Lipg) and HL (Lipc) (supplemental Fig. S5A, supplemental Table S5). These findings were consistent with the data reported in Angptl3-deficient mice (24).
Furthermore, post-translational mechanisms of VLDL assembly (46) were unperturbed by evinacumab: neither hepatic microsomal triglyceride transfer protein (MTP) nor intracellular phospholipid transfer protein (PLTP) activities were affected, mirroring our findings in Angptl3 -/mice (supplemental Fig. S5B). Lipid peroxidation, which at elevated levels is associated with pre-secretory VLDL degradation (46), was also unchanged (supplemental Fig.   S5B). Evinacumab had no impact on liver lipids in Ldlr -/mice, even after multi-dose treatment, and no changes in liver fat were seen in Angptl3 -/mice (supplemental Fig. S5C) (40). Collectively, our data show ANGPTL3 has no apparent effect on hepatic VLDL assembly, although we cannot rule out secondary effects as a result of increased lipolysis and return of lipid substrates to the liver.

Multiple remnant receptors could contribute to VLDL clearance upon ANGPTL3 inhibition.
Given our new insights into the role of EL in APOE-and LDLR-independent VLDL clearance, we postulated that unrestrained EL may promote VLDL uptake via SR-B1, akin to its action on HDL (22,47). To address this, we used antisense oligonucleotides (ASOs) for in vivo knockdown of chow-fed SR-B1/Scarb1 in Ldlr -/mice (34). SR-by guest, on July 25, 2020 www.jlr.org Downloaded from B1 ASO yielded ~90% Scarb1 knockdown in liver and significantly elevated serum cholesterol (supplemental Fig.   S6A-C). Despite a 3.5-fold increase in non HDL-C over baseline, evinacumab potently lowered cholesterol in APOBlipoproteins and hence did not require SR-B1 (supplemental Fig. S6A-C). Of note, ANGPTL3 inhibition had no effect on HDL-C in Scarb1 ASO-treated mice, consistent with SR-B1's established role as HDL-receptor (supplemental Fig. S6A-C) (22).
Given the multitude of remnant receptors known to compensate for each other (47,50,51), we hypothesized that a combination of receptors could mediate VLDL remnant clearance upon ANGPTL3 inhibition. We therefore generated quadruple lipoprotein receptor-deficient mice by administering ASOs targeting SR-B1 alone or in combination with LRP1 ASO (33,34) to Sdc1 -/-Ldlr -/mice (Fig. 5B). Knockdown of additional receptors yielded a gene dosage-dependent increase in serum cholesterol: targeting SR-B1 strikingly elevated cholesterol (834mg/dL), while combined SR-B1 and LRP1 ASO treatment further accentuated these effects in Sdc1 -/-Ldlr -/mice on chow diet (1639mg/dL vs. 252mg/dL for Ctrl ASO). However, evinacumab still potently reduced serum cholesterol, even in mice with reduced or absent expression of four major VLDL remnant receptors (Fig. 5B, C).
Given the widespread expression of Lrp1 and Scarb1, incomplete knockdown in the liver (Fig. 5D) and in extrahepatic tissues (data not shown) could provide an explanation for this effect. On the other hand, additional remnant receptors (50,52,53) could further compensate and mediate evinacumab-driven VLDL uptake. The potential for redundancy was already apparent from our studies in Lipg (single) KO mice with functional LDLR, in which evinacumab lowered LDL-C likely through LDLR (Fig. 1D). That said, we cannot completely rule out non-receptor mediated clearance mechanisms, or other contributing factors, e.g. potential effects of EL on hepatic VLDL secretion.
While intriguing, exploring these options is beyond the scope of the present investigation. Most importantly, the by guest, on July 25, 2020 www.jlr.org Downloaded from 20 notion that ANGPTL3 inhibition appears to engage multiple pathways underscores the efficiency and versatility of this pathway, and provides an explanation for its potent reduction of LDL-C.

DISCUSSION
Epidemiological studies over the past decades have placed increasing emphasis on the importance of LDL-C reduction in the prevention of CAD (2,54). Despite the availability of a number of lipid-lowering therapies, treatment options remain limited for patients with deficiencies in the LDLR pathway. ANGPTL3 inhibition has emerged as a promising strategy for these patients, as it is associated with a dramatic reduction of both cholesterol and TGs, and lowers LDL-C independently of LDLR (9,11,24,25). While the pathways contributing to TG and HDL-C reduction with ANGPTL3 inactivation have been well characterized (11,14,17), much less is known about the mechanism responsible for its LDL-C lowering effect. Here, we reveal an under-appreciated role for EL in APOBcontaining lipoprotein metabolism. Importantly, we demonstrate that ANGPTL3's regulation of EL is not limited to its effect on HDL-C, but is required for the modulation of LDL-C levels when LDLR is absent. Mechanistically, we show that EL de-repression by evinacumab leads to extensive remodeling of VLDL, resulting in formation of lipiddepleted remnant particles, which accelerates their clearance from the circulation. This, in turn, leads to depletion of the LDL precursor pool and reduces LDL-C levels (Fig. 6). Such an EL-dependent alternative pathway likely functions unnoticed alongside LDLR during homeostasis, but becomes more relevant when LDLR is dysfunctional.
While a prior study using EL overexpression hinted at the potential role of EL in the catabolism of APOBcontaining lipoproteins (55), Lipg -/-Ldlr -/mice proved instrumental in uncovering new concepts under physiologically relevant conditions. By employing lipidomic and kinetic studies, we revealed a role for EL in determining APOB-containing particle composition and clearance upon ANGPTL3 inhibition. We found that ELmodifications on VLDL were necessary for LDLR-independent particle removal at the juncture between VLDL remnants and their further processing to become LDL. In the absence of both EL and LDLR, ANGPTL3 inhibition led to generation of partially processed remnants, which blunted their clearance and led to accumulation of TGdepleted, cholesterol-and phospholipid-rich LDL.
Such a scenario can be explained by EL-dependent VLDL processing in the circulation, but is more difficult to reconcile with reduced hepatic VLDL output. In fact, ANGPTL3 inhibition did not affect the hepatic transcriptome, by guest, on July 25, 2020 www.jlr.org Downloaded from VLDL assembly and liver lipid content (supplemental Figure S5). In this regard, our findings differ from previous studies reporting lower VLDL-TG secretion (24) and reduced liver fat (10) upon ANGPTL3 inhibition. Prior evaluation of VLDL-TG secretion depended on lipolysis inhibitor Triton WR-1339 (24). We speculate that the observed VLDL-TG reduction may be due to rapid unrestrained LPL activity towards nascent VLDL upon ANGPTL3 inhibition, and/or incomplete LPL inhibition by Triton WR-1339. If ANGPTL3 inhibition indeed led to secretion of smaller VLDL with less TG-content, our data from EL-deficient mice (where lipolysis led to accumulation of TGdepleted remnants) suggest that TG-reduction on its own may have little impact on the clearance of these particles.
That said, it is still feasible that the combination of altered VLDL-TG secretion and EL-driven catabolism of modified particles ultimately promotes VLDL remnant clearance and thus LDL-C reduction. Subsequent studies in Lpl conditional KO mice, or even Lipg/Lpl double KO mice should resolve these discrepancies, as the absence of lipolysis will allow to better determine the lipid composition of nascent VLDL upon ANGPTL3 blockade.
Regarding liver fat, ASO-mediated Angptl3 inhibition was found to reduce liver TG content (10), whereas evinacumab had no such effects (supplemental Fig. S5C). While this key difference may be rooted in the mechanism of pharmacological inhibition (modulating hepatic gene expression vs. protein blockade in the circulation), it is noteworthy that Angptl3 -/mice have similar liver TG content as WT mice (40). Moreover, in contrast to our current and prior studies (24), Musunuru and colleagues reported decreased VLDL-APOB production rates in human individuals with ANGPTL3 LOF variants (6). However, these findings were limited to 2 homozygous and 3 heterozygous ANGPTL3 LOF carriers. A recent NMR study on a larger number of ANGPTL3 LOF carriers found that among APOB-lipoproteins, the greatest cholesterol-lowering effect was in the VLDL remnant fraction (56) Nonetheless, evinacumab's action on VLDL remnants represents a mechanism that is distinct from existing lipid-lowering treatments, including statins, PCSK9-, MTP-and APOB-inhibitors (54). Consistent with this notion, a recent publication revealed that triple combination treatments of statins and antibodies targeting PCSK9 and ANGPTL3 yielded additive effects on cholesterol and atherosclerosis reductions in mice (59). ANGPTL3 inhibition even operates differently than APOC3 or ANGPTL4 blockade, which also activate LPL and promote TG hydrolysis, however have only minor effects on cholesterol reduction (60)(61)(62). These findings illustrate that increased rates of hydrolysis are not invariably linked to greater hepatic particle clearance. While individuals with heterozygous loss of APOC3 exhibit increased conversion of VLDL to LDL, no concomitant increase in VLDL remnant clearance kinetics was seen (63). We observed similar effects in EL/LDLR double KO mice, in which the absence of EL impaired VLDL remnant clearance typically seen with ANGPTL3 inhibition and promoted its conversion to LDL. Indeed, the key distinction between APOC3 and ANGPTL3 appears to be their differential ability to modulate EL: unlike for ANGPTL3, APOC3 has not been reported to affect EL activity. Thus, ANGPTL3 deficiency represents a unique scenario where de-repression of both LPL and EL drives VLDL remodeling and its LDLR-independent clearance, thereby restricting LDL production.
In regard to clearance, our studies have unmasked multiple layers of redundancy. ANGPTL3 inhibition effectively lowered LDL-C in Lipg -/-(single KO) mice, likely due to the presence of LDLR and its ability to bind VLDL remnants (1,3). Without LDLR, EL activity becomes critical, yet clearance of EL-modified VLDL may involve multiple redundant receptors (47)(48)(49)(50)(51). Deletion of LRP1 and SDC1 were previously shown to be dispensable for evinacumab's LDL-C lowering effect (24). While in the current study, our combinatorial approach to blunt hepatic LDLR/SDC1/LRP1/SR-B1 activities did not abolish LDL-C reduction upon ANGPTL3 inhibition, it is feasible that their extrahepatic expression may contribute to evinacumab's effect as well. Future tissue uptake studies with radiolabeled VLDL will help to pinpoint the potential clearance receptor(s). Coupling these studies with VLDL, dually labeled with triglyceride and cholesteryl (or retinyl) ester tracers will reveal further insights into the relative kinetics of TG-lipolysis versus remnant uptake.
How then does EL activation lead to VLDL clearance? Given that EL functions as a phospholipase, its impact on cholesterol is thought to be secondary to lipoprotein remodeling (20). We surmise that the reduction of surface phospholipids may expose additional APOB domains which facilitate remnant receptor interactions and lipoprotein by guest, on July 25, 2020 www.jlr.org Downloaded from 23 uptake. Conversely, the lack of EL may generate phospholipid-enriched remnants of discoid or irregular shape and size, which potentially prevents efficient interaction with hepatocytes. EL could also change VLDL apoprotein content and alter plasma clearance rates independently of APOE (24) in this way. Electron microscopy and proteomic approaches will shed light on the role of EL on APOB lipoprotein structure and composition in future studies. Besides EL, hepatic lipase (HL) is also known to hydrolyze phospholipids (20,64), yet prior studies have shown that ANGPTL3 does not regulate HL activity (11,40). In vitro data suggest that even LPL has low phospholipase activity (64)(65)(66), however, vascular lipases seem to have broader substrate specificity in vitro, as EL was also shown to exhibit appreciable, albeit weak TG-lipase activity in assays with purified enzyme (64). Our in vivo studies show that LPL activation had no substantial effect on phospholipids and particle clearance in EL/LDLR double KO mice ( Figure   1E). Thus, neither HL nor LPL appear to be able to compensate for EL de-repression when LDLR is missing, potentially due to competition between high levels of HDL and APOB-lipoproteins for lipolytic enzymes. Finally, it should be noted that although EL's enzymatic activity led to substantial remodeling of APOB-lipoproteins, its nonenzymatic 'bridging' function may also facilitate remnant clearance through uptake receptors (49).
In summary, we have uncovered a novel mechanism that lowers LDL-C by ANGPTL3 inactivation. Through de-repression of EL, ANGPTL3 inhibition promotes VLDL processing and clearance. This reduces LDL production and bypasses the requirement for LDLR (see schematic in Fig. 6). While EL appears to be dispensable for LDL-C regulation in normal individuals, its effect on VLDL remodeling and clearance becomes critical without LDLR.
Together, these studies advance our understanding of lipoprotein metabolism and provide important insights into the mechanism by which ANGPTL3 inhibition lowers LDL-C in patients lacking functional LDLR.

DATA AVAILABILITY
All data associated with this study are present in the paper or the supplemental Materials, and all figures except Fig.   6 (schematic) have associated raw data. Any materials that can be shared will be released via a material transfer agreement. The accession number for the sequencing data reported in this paper is NCBI GEO: GSE143726. 24 We thank Y.Xin and Q.Su for help with RNA-sequencing data analysis, and A.    administered ANGPTL3 mAb or control mAb (Phase 1 single ascending dose clinical study) (28), and serum was collected at the indicated timepoints. Mean ± s.e.m. are shown for each timepoint (n = 18 in placebo; 10 in 5mg/kg IV; 9 in 10mg/kg IV; 11 in 20mg/kg IV; 12 in 150mg SC and 9 in 250mg SC groups). Human VLDL/chylomicron and LDL concentration (C), and particle size analysis by NMR (D). P-values from two-way ANOVA with Tukey correction posttest: *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001 relative to placebo (color of asterisks indicates relevant group); # P<0.0001 regarding all ANGPTL3 mAb groups relative to placebo; & P<0.0001 regarding all IVadministered ANGPTL3 mAb groups relative to placebo.

APOB-containing lipoprotein turnover in the circulation is governed by ANGPTL3
Liver Liver