Synthetic Site-Specific Antibody–Ligand Conjugates Promote Asialoglycoprotein Receptor-Mediated Degradation of Extracellular Human PCSK9

Abstract

Targeted degradation using cell-specific lysosome targeting receptors is emerging as a new therapeutic strategy for the elimination of disease-associated proteins. The liver-specific human asialoglycoprotein receptor (ASGPR) is a particularly attractive lysosome targeting receptor leveraged for targeted protein degradation (TPD). However, the efficiency of different glycan ligands for ASGPR-mediated lysosomal delivery remains to be further characterized. In this study, we applied a chemoenzymatic Fc glycan remodeling method to construct an array of site-specific antibody–ligand conjugates carrying natural bi- and tri-antennary N-glycans as well as synthetic tri-GalNAc ligands. Alirocumab, an anti-PCSK9 (proprotein convertase subtilisin/kexin type 9) antibody, and cetuximab (an anti-EGFR antibody) were chosen to demonstrate the ASGPR-mediated degradation of extracellular and membrane-associated proteins, respectively. It was found that the nature of the glycan ligands and the length of the spacer in the conjugates are critical for the receptor binding and the receptor-mediated degradation of PCSK9, which blocks low-density lipoprotein receptor (LDLR) function and adversely affects clearance of low-density lipoprotein cholesterol. Interestingly, the antibody–tri-GalNAc conjugates showed a clear hook effect for its binding to ASGPR, while antibody conjugates carrying the natural N-glycans did not. Both the antibody–tri-antennary N-glycan conjugate and the antibody–tri-GalNAc conjugate could significantly decrease extracellular PCSK9, as shown in the cell-based assays. However, the tri-GalNAc conjugate showed a clear hook effect in the receptor-mediated degradation of PCSK9, while the antibody conjugate carrying the natural N-glycans did not. The cetuximab–tri-GalNAc conjugates also showed a similar hook effect on degradation of the membrane-associated protein, epidermal growth factor receptor (EGFR). These results suggest that the two types of ligands may involve a distinct mode of interactions in the receptor binding and target-degradation processes. Interestingly, the alirocumab–tri-GalNAc conjugate was also found to upregulate LDLR levels in comparison with the antibody alone. This study showcases the potential of the targeted degradation strategy against PCSK9 for reducing low-density lipoprotein cholesterol, a risk factor for heart disease and stroke.

Graphical Abstract

INTRODUCTION

Protein degradation is an important mechanism in biological systems to maintain protein homeostasis and protein quality control.^1,2 In eukaryotes, there are several degradation compartments, including the proteasome, lysosome, and autophagosome.³ The degradative mechanisms of these compartments have been leveraged to perform targeted protein degradation and deplete pathogenic proteins or other proteins of interest (POI) with great success.^4–6 For example, Deshaies and co-workers have reported the first proteolysis targeting chimeras (PROTACs), consisting of a small molecule ligand for an E3 ubiquitin ligase and a ligand specific for the POI, as bifunctional molecules that can recruit ubiquitin ligase enzymes to target proteins resulting in their polyubiquitination and proteasomal degradation.⁴ PROTACs were shown to be effective degraders of intracellular target proteins, which encouraged the development of several other targeted protein degradation (TPD) strategies employing similar mechanisms of action such as specific and nongenetic inhibitor of apoptosis protein (IAP)-dependent protein erasers (SNIPERs), molecular glues, autophagy-targeting chimeras (AUTACs), and even bacteria PROTACs (BacPROTACs) for degradation of bacterial proteins.^6–9 The early development of sweeping antibodies, Abdegs and Seldegs, which make use of the FcRn receptor for degradation of extracellular proteins by protein-engineered pH-responsive antibodies or Fc-antigen fusions, has spurred the development of other TPD platforms for the degradation of extracellular proteins.^10–12 For example, Bertozzi and co-workers have developed lysosome-targeting chimeras (LYTACs) as a strategy for targeted degradation of extracellular proteins.⁵ This strategy has been expanded by other groups using different antibodies, novel heterobifunctional small molecules, and/or different receptors for targeted degradation of extracellular as well as membrane-associated proteins.^5,12–15

The LYTAC approach makes use of extracellular lysosome targeting receptors (LTRs) such as the asialoglycoprotein receptor (ASGPR) with specificity for galactose or N-acetylgalactosamine (GalNAc)-containing glycoproteins. Synthetic ligands for the LTRs are conjugated to monoclonal antibodies specific for the POI. In this way, the LYTAC can bridge the POI and the LTR to enable endocytosis and delivery of POI to the lysosome for degradation. To date, several LTRs, including CI-M6PR (cation-independent man-nose-6-phosphate receptor), ASGPR, integrin, and membrane-bound E3 ligase RNF43 have been successfully utilized for targeted protein degradation.^5,13,16–18 Despite the recent progress in ASGPR-dependent degradation strategies, much remains to be understood about the substrate scope of this receptor and what constitutes an effective glycan ligand for successful lysosomal delivery of the protein target. In fact, since the optimization of the high-affinity synthetic tri-GalNAc ligand for ASGPR, it has been used almost exclusively for the delivery of therapeutics or probes to the liver.¹⁹ Also, in most reports leveraging lectin-based LTRs for targeted degradation, synthetic glycan ligands are attached by random conjugation resulting in heterogeneous antibody conjugates that are difficult to characterize and optimize.^5,16,17 On the other hand, previous work from our lab showed that chemo-enzymatic antibody-glycan remodeling could be used to install a minimal M6P-containing structure site specifically on Asn297 of the antibody Fc and result in significant CI-M6PR mediated degradation of membrane-bound protein targets, demonstrating that this can be a viable approach to obtain structurally well-defined and homogeneous LYTACs.¹⁵

LYTAC strategies have been attempted for the degradation of several clinically relevant extracellular proteins, such as PDL-1 and HER-2.^{5,13,15,16,18} In the present study, we chose the proprotein convertase subtilisin/kexin type 9 (PCSK9) as a target to demonstrate the potential of LYTACs for the treatment of diseases like hypercholesterolemia. PCSK9 is a liver-secreted protease, the presence of which is known to reduce low-density lipoprotein receptor (LDLR) levels, resulting in higher levels of low-density lipoprotein cholesterol (LDL-C) and thus a higher risk of cardiovascular disease.²⁰ Extracellular PCSK9 binds LDLR receptor inducing receptor-mediated endocytosis, where PCSK9 prevents the recycling of LDLR to the cell surface and redirects the receptor to the lysosome for degradation (Scheme 1).^21,22 Several compelling genetic studies have established a clear association between PCSK9 activity and levels of circulating low-density lipoprotein (LDL-C), the main vehicle for cholesterol transport in the blood and a major risk factor for heart disease and stroke.^23–26 Specifically, patients with loss-of-function mutations in the PCSK9 gene were reported to have remarkably low LDL-C levels,^25,26 and PCSK9 gain-of-function mutants such as PCSK9 D374Y with 10-fold higher affinity for LDLR result in high LDL-C levels and hypercholesteremia.²⁴ Several therapies have been developed to block PCSK9 function, including PCSK9 small molecule inhibitors, siRNAs, and monoclonal antibodies, with monoclonal antibodies and siRNAs being the most advanced in the clinical pipeline.^27,28 However, paradoxically, dosing with anti-PCSK9 monoclonal antibodies results in elevated total serum PCSK9 levels in animal models and humans, which could be problematic after antibody clearance.^29–32Here, we sought to explore the LYTAC strategy for targeted degradation of extracellular PCSK9, which may enhance the efficacy of the anti-PCSK9 antibody for the treatment of hypercholesteremia (Figure 1). To test the hypothesis, we synthesized an array of site-specific antibody–ligand conjugates carrying natural bi- and triantennary N-glycans as well as the tri-GalNAc ligands using an Fc glycan remodeling method. A flow cytometry analysis demonstrated interesting and distinct profiles of binding of different antibody-glycan ligand conjugates to the cell-surface ASGPR. In addition, our cell-based assay results showed that the synthetic antibody-glycan ligand conjugates could significantly degrade the extracellular PCSK9. Interestingly, antibody-glycan conjugates carrying the tri-antennary N-glycan or the tri-GalNAc ligands showed differences in ASGPR receptor binding and in degradation of PCSK9, and a clear hook effect was observed for the antibody–tri-GalNAc conjugate in these processes but not for the antibody-natural N-glycan conjugate. During the preparation of this manuscript, Bagdanoff and co-workers have reported that heterobifunctional small molecules as well as antibody–tri-GalNAc ligands that are capable of binding to PCSK9 and the asialoglycoprotein receptor (ASGPR) could accelerate PCSK9 clearance in an animal model, demonstrating the in vivo efficacy of this LYTAC strategy.³³ Taken together, these studies demonstrate the potential of the LYTAC strategy for the accelerated clearance and degradation of pathogenic proteins in circulation such as PCSK9.

Scheme 1. Synthesis of ASGPR Ligands Used in the LYTAC Approach^a.

^a(A) Biantennary complex-type N-glycan-Asn is prepared by Pronase digestion of SGP, followed by sialidase treatment to remove terminal sialic acid. Compound 1 is reacted with NHS-PEG₅-DBCO to prepare G2-Asn-DBCO (compound 2). (B) Tri-antennary complex-type N-glycans are prepared from fetuin by Endo-F3 digestion, followed by sialidase treatment to obtain compound 3. Alternatively, Pronase digestion of fetuin followed by sialidase treatment gives the asialo-complex-type G3-Asn. Reaction with NHS-PEG₅-DBCO gives G3-Asn-DBCO (compound 5). (C) The tri-GalNAc-NH₂ (compound 6) starting material is synthesized according to Prakash et al.⁴⁵ and reacted with NHS-PEG₅-DBCO to give compound 7. DMC, 2-chloro-1,3-dimethylimmidizolinium chloride.

Figure 1.

Cholesterol regulation by the low-density lipoprotein receptor (LDLR) in human hepatocytes. (I) LDLR promotes the endocytic uptake of LDL-C to the endosome. (II) PCSK9 binds LDLR, promoting its endocytosis and redirects LDLR to the lysosome for degradation resulting in reduced LDLR levels and higher LDL-C levels. (III) PCSK9-specific monoclonal antibodies act as competitive inhibitors to disrupt PCSK9-LDLR interactions and rescue LDLR from PCSK9-mediated degradation. (IV) PCSK9-specific LYTACs leverage liver cell ASGPR to promote PCSK9 degradation, removing it from circulation. LDLR is rescued from degradation and can bind and clear LDL-C from the blood.

RESULTS AND DISCUSSION

Synthesis of High-Affinity Glycan Ligands for ASGPR.

We have previously described the use of a chemoenzymatic Fc glycan remodeling approach to install drugs and mannose-6-phosphate glycan ligands site-specifically on monoclonal antibodies.^15,34–37 For example, we have reported that M6P-modified LYTACs prepared through this chemoenzymatic method can degrade transmembrane proteins effectively through the CI-M6PR receptor.¹⁵ We sought to use this chemoenzymatic Fc glycan remodeling method to site-specifically install natural and synthetic glycan ligands of ASGPR to the Fc domain of an anti-PCSK9 antibody, which has not been explored before. Site-specific antibody conjugation is highly desirable since the method allows for tight control over location and valency of the attached cargoes.

While the synthetic tri-GalNAc ligand has proven successful in several liver-targeted therapies, identification of a natural glycan structure with equivalent functions would facilitate access to liver-targeted therapeutic glycoproteins and glycol-conjugates by way of existing genetic and chemoenzymatic glycoengineering strategies. Previous work from Lee and coworkers on rat hepatocytes has demonstrated that bi-, tri-, and tetra-antennary asialo-complex-type N-glycans with terminal β-galactose moieties are recognized by the rat asialoglycoprotein receptor.^38,39 The affinity of the biantennary complex-type N-glycan for rat ASGPR is relatively weak (ca. 50 μM), while the affinity of the tri- and tetra-antennary N-glycans is in the nanomolar range.^38,39 Such β-galactosides are also known ligands for the human asialoglycoprotein receptor (hASGPR); however, the optimal glycan ligands for hASGPR binding and subsequent internalization remain to be further evaluated.^40,41This consideration is particularly relevant to the design of liver-specific LYTACs leveraging the ASGPR receptor and warrants further investigation. To enable the chemoenzymatic synthesis of ASGPR-specific LYTACs and explore the substrate scope of the ASGPR receptor, several natural N-glycan structures were first prepared from the sialoglycopeptide (SGP) isolated from chicken egg yolks. Thus, the asialo-biantennary complex-type N-glycan-Asn (G2-Asn, 1) was prepared by protease digestion of SGP followed by sialidase digestion to remove terminal α(2,6)-sialic acid, as described in our previous reports.^42,43 The asialylated N-glycan (G2-Asn) was then reacted with 1.5 mol equiv NHS-PEG₅-DBCO to prepare G2-Asn-DBCO (2) and install a chemical handle at the Asn N-terminus for click conjugation (Scheme 1). The final product was purified by reversed-phase preparative HPLC to give 2. The tri-antennary N-glycan oxazoline (3), which serves as a substrate for antibody Fc glycan remodeling, was prepared from bovine fetuin following our previously reported procedure.⁴⁴ The DBCO-tagged tri-antennary N-glycan (5) was synthesized in several steps. First, protease digestion of bovine fetuin, followed by anion exchange and enzymatic desialylation, gave the Asn-linked tri-antennary N-glycan (4). Then, treatment of 4 with 1.5 mol equiv. of NHS-PEG₅-DBCO, followed by reversed-phase HPLC purification, afforded the G3-Asn-DBCO (5). After the synthesis of natural N-glycan ligands, the high-affinity tri-GalNAc ligand used by Bertozzi, Tang, and co-workers^16,17 was synthesized to enable comparison of synthetic and natural ASGPR ligands. Chemical synthesis of 6 was accomplished following the reported procedures.⁴⁵ Compound 6 was then reacted with 3 mol equiv. NHS-PEG₅-DBCO in DMSO and purified by reversed-phase HPLC to obtain pure tri-GalNAc-DBCO (7) (Scheme 1). The identity of compounds 2, 5, and 7 was confirmed by¹ H NMR and LC ESI-MS analysis.

Chemoenzymatic Synthesis of the Antibody–Ligand Conjugates Carrying Natural and Synthetic ASGPR Glycan Ligands.

With compounds 2, 3, 5, and 7 in hand, we next examined their site-specific conjugation to alirocumab (8a), an FDA-approved monoclonal antibody specific for PCSK9, using the enzymatic Fc glycan remodeling method. Thus, alirocumab (antibody 8a) was deglycosylated with immobilized wild-type Endo-S2 from Streptococcus pyogenes to give the Fucα1,6GlcNAc-alirocumab intermediate (9a), which serves as an acceptor for the subsequent transglycosylation reactions (Scheme 2). A natural tri-antennary N-glycan structure was then site-specifically introduced by incubation of 9a with Endo-F3 D165A mutant in the presence of G3-oxazoline (compound 3) following previously reported procedures to give G3F-alirocumab (antibody 10a), with triantennary N-glycan on the Fc domain.⁴⁴ Alternatively, antibody 9a could be incubated with azido-modified sugar oxazoline donors to install a click handle on the monoclonal antibody. This was done by two different methods to enable comparison of the effects of ligand-antibody distance on hASGPR binding/ activity. First, antibody 9a was incubated with Endo-S2 D184M in the presence of di-N₃-S2G2-ox to install the azido-modified sialo-complex-type N-glycan on the Fc, giving antibody 11a, as previously reported by our lab.³⁴ Separately, antibody 8a was transformed in one-pot to antibody 12a by incubation with the di-N₃-Man-GN-ox disaccharide in the presence of WT Endo-S2, giving an antibody intermediate with azide groups attached to the shorter disaccharide (Scheme 2).³⁵ With intermediates 11a and 12a in hand, SPAAC (strain-promoted azide-alkyne cycloaddition) click chemistry was performed to install high-affinity ASGPR ligands 2, 5, and 7 on the nonreducing ends of the Fc glycans. Three molar equivalents (per azide) of the DBCO-modified natural N-glycan ligands 2 and 5 were reacted with intermediate 11a in a phosphate buffer to afford novel antibody glycoclusters 13a and 14a in just 1 h, as determined by LC-ESI-MS. Alternatively, three molar equivalents of the synthetic tri-GalNAc-DBCO ligand were reacted with antibody intermediates 11a and 12a in phosphate buffer containing 10% DMSO. The reactions were monitored by LC-ESI-MS and found to be complete within 6–8 h to afford antibody conjugates 15a and 16a.

Scheme 2. Chemoenzymatic Synthesis of Alirocumab and Cetuximab Site-Specific LYTACs^a.

^a(A) Native antibodies are deglycosylated with immobilized WT Endo-S2 to give the GNF-antibody acceptor, which can be transformed to antibody variants 10–12 by transglycosylation with various mutant endoglycosidase enzymes and sugar-oxazolines. (B) Click conjugation of ASGPR high-affinity ligands by reaction of DBCO-tagged glycans with azido-modified antibody intermediates.

To demonstrate the feasibility of the conjugation method, we used cetuximab as another model antibody to show the Fc site-specific conjugation, even when additional N-glycans are present in the Fab domains. Cetuximab is an epidermal growth factor receptor (EGFR)-targeting monoclonal antibody used for the treatment of colorectal and head and neck cancers.⁴⁶ The cetuximab–ligand conjugates would provide an opportunity to test the targeted degradation of membrane-associated protein (EGFR) by the antibody conjugates. Thus, commercially available cetuximab (antibody 8b) was deglycosylated using immobilized WT Endo-S2 to selectively remove the Fc N-glycans and give the GNF-cetuximab acceptor (antibody 9b, Scheme 2). Importantly, Endo-S2 is specific for deglycosylation of Fc N-glycans and has very low activity on N-glycans of the Fab domains.⁴⁷ Chemoenzymatic glycan remodeling was then done as described for the alirocumab glycoforms to prepare antibodies 10b–11b, the di-azido-disaccharide-modified antibody was not prepared. Lastly, DBCO-modified ligands 2, 5, and 7 were conjugated to antibody intermediate 11b by click chemistry to give conjugates 13b–15b (Scheme 2).

Binding of LYTAC Glycan Ligands to ASGPR on the Liver Cell Surface.

The LYTACs carrying natural N-glycans and the synthetic tri-GalNAc cluster moiety were then evaluated for their ability to bind hASGPR on the liver cell surface by flow cytometry. To eliminate the contribution of antibody-antigen interactions associated with the anti-EGFR antibody (cetuximab), alirocumab-based LYTACs were used in the binding studies and in a culturing condition, where HepG2 cells secrete significantly reduced levels of PCSK9 (Figure S31). Initially, an anti-human phycoerythrin (PE)-conjugated secondary antibody was used for detection of each LYTAC on the cell surface; however, the secondary antibody was unable to bind the antibody sugar conjugates 10a and 13a–16a. Alternatively, the glycoengineered antibodies and their conjugates were biotinylated chemically using an NHS-biotin activated ester and binding-detected using a streptavidin-PE conjugate. Asialofetuin, a known binder of the asialoglycoprotein receptor, was also biotinylated and used as a positive control. Importantly, the number of biotin units conjugated to each glycoprotein is comparable so that cell-associated fluorescence intensity reflects the relative binding level of each conjugate. HepG2 cells were seeded in 96-well plates and incubated overnight in serum-free media. Then, cells were incubated with serial dilutions of each biotinylated LYTAC and binding-detected by flow cytometry. As expected, little to no binding was observed for the native alirocumab antibody, which contains mainly the mono-galactosylated and nongalactosylated biantennary complex-type N-glycans. As a positive control, the asialofetuin-bearing galactosylated bi- and tri-antennary complex-type N-glycans showed robust binding to the cell surface (Figure 2A). This result confirms that the binding of the antibody-glycan conjugates to cell-surface ASGPR is dependent on the presence of terminal galactose/GalNAc in the conjugates.

Figure 2.

Recognition of glycan ligands on the biotinylated LYTACs by ASGPR on HepG2 cell surface. (A) Mean fluorescence intensity (MFI) measurements of HepG2 cells incubated with serial dilutions of each LYTAC. (B) Ratio of maximum fluorescence to EC50 for each LYTAC; EC50′s were obtained by nonlinear regression of each hyperbolic segment of each binding curve.

Interestingly, LYTAC 10a with the tri-antennary N-glycan being directly attached to Asn297 in the antibody demonstrated only marginally better binding than the native antibody (8a), suggesting that the N-glycans on this position might not be accessible to ASGPR recognition. Similarly, weak binding was observed by LYTAC 16a containing the high-affinity tri-GalNAc ligand attached to the short disaccharide core attached to the Fc glycosylation site. On the other hand, LYTAC 15a containing the same tri-GalNAc ligand attached by a click reaction to the more extended sialo-complex-type N-glycan at the Fc domain was the strongest binder to cell-surface ASGPR, indicating that proper spacing from the bulky Fc domain is critical for effective recognition of glycan ligands attached to antibodies. However, a decrease in the binding signal was observed at concentrations higher than 400 nM, suggesting that excess soluble 15a acts as a competitive inhibitor of the 15a/ASGPR complex, a phenomenon known as the prozone or hook effect.⁴⁸ The hook effect is an immunologic phenomenon, whereby the effectiveness of antibodies to form immune complexes can be impaired when concentrations of an antibody or an antigen are very high. It has been observed for some receptor–ligand and/or antibody-antigen interactions.^49,50 Lastly, LYTAC 14a containing natural triantennary G3 N-glycans attached by a click reaction showed ASGPR binding comparable to the antibody–tri-GalNAc conjugate (15a) and was the strongest binding ligand above 400 nM concentration. In contrast to the antibody–tri-GalNAc conjugate (15a), the antibody conjugate carrying the natural tri-antennary N-glycans (14a) did not show a hook effect in binding to ASGPR and demonstrated steadily increasing cell-associated fluorescence even to the highest concentration of 1.2 μM. Some cell-surface binding was observed by the antibody conjugate (13a) carrying biantennary N-glycans. However, the binding profile was erratic and was not included for simplicity. As described in previous work, the affinity of the biantennary complex-type N-glycan for rat ASGPR is relatively weak, in the range of ca. 50 μM, while the affinity of the tri-antennary N-glycans for rat ASGPR is estimated in the range of nM concentrations.^38,39 Our binding affinity data with the human ASGPR appear to be consistent with the previously reported results. However, it remains to be further investigated why the antibody conjugates carrying the natural N-glycans do not show a hook effect while the antibody conjugates carrying the synthetic tri-GalNAc cluster moiety demonstrate a clear hook effect in the receptor binding.

To assess the relative binding affinity of each glycoprotein more quantitatively, the EC₅₀ was determined for each glycoprotein displaying significant binding above background levels (i.e., LYTACs 14a and 15a, and asialofetuin). This was done by nonlinear regression of the hyperbolic segments of the binding curves before a decrease in fluorescence intensity due to the hook effect. These values were then normalized to the maximum fluorescence intensity achieved by saturated binding of each biotinylated glycoprotein and plotted side by side for comparison (Figure 2B). Indeed, LYTAC 15a is the best binder to cell-surface ASGPR with the highest F_max/EC₅₀ ratio (3.8). LYTAC 14a was the next best binder (F_max/EC₅₀ = 2.9) with comparable binding to the cell-surface ASGPR receptor. Lastly, the biotinylated asialofetuin bound only weakly with an F_max/EC₅₀ ratio of 1.2, indicating that the hASGPR can recognize tri-antennary N-glycans attached by long flexible linkers better than tri-antennary N-glycans in the context of highly glycosylated glycoproteins. These results provide important insights into glycan preference and selectivity for the human asialoglycoprotein receptor.

Targeted Protein Degradation of PCSK9 and Effects on Low-Density Lipoprotein Receptor (LDLR).

Having confirmed that interactions of LYTACs 10a and 13a–16a with ASGPR on the liver cell surface depend on the presence of terminal galactose or N-acetylgalactosamine, alirocumab-based LYTACs were then investigated for their ability to degrade extracellular PCSK9. Degradation of PCSK9 was assessed by western blot of the cell culture media. Figure 3 demonstrates the western blot data for HepG2 cells treated with varying concentrations of each antibody or antibody conjugate in the presence of PCSK9 D374Y, the PCSK9 D374Y mutant alone, or serum-deficient media (control) over 48 h. No significant degradation was observed by commercial antibody 8a, consistent with previous reports on this antibody (Figure 3A).⁵¹ Treatment with the antibody conjugates (10a and 13a) carrying the weak affinity biantennary N-glycan ligand or the high-affinity tri-antennary complex-type N-glycan ligand attached to the antibody with a short linker, 10a and 13a, respectively, also failed to degrade extracellular PCSK9. These results are consistent with the weak binding affinity of 10a and 13a to cell-surface ASGPR, as demonstrated by the flow cytometry analysis (Figure 2). This result suggests that clustered biantennary N-glycans and tri-antennary N-glycans attached directly to the bulky Fc domain are ineffective ligands for lysosomal targeting of extracellular soluble proteins (Figure 3B). However, LYTAC 14a, in which the tri-antennary N-glycan was attached on the extended N-glycan core, demonstrated ~30% degradation of extracellular PCSK9 at 200 nM concentration, indicating that the tri-antennary G3 N-glycan can be an effective ligand for targeted degradation when connected by a long linker to the N-glycan on Asn297.

Figure 3.

PCSK9 degradation assessed by western blot of HepG2 media after treatment with each LYTAC. HepG2 cells treated with PCSK9 D374Y and serial dilutions of (A) native antibody 8a, (B) LYTAC 13a or 10a, (C) LYTAC 14a or 15a, and (D) LYTAC 16a for 48 h. Media of cells treated with PCSK9 D374Y only and OptiMEM media were included as positive and negative controls of each blot, respectively.

Gratifyingly, more pronounced degradation was observed by LYTAC 15a with a synthetic tri-GalNAc ligand attached to each arm of the biantennary N-glycan core. Approximately 60% of extracellular PCSK9 D374Y mutant was degraded at a 50 nM concentration of 15a, corresponding to an equimolar concentration of the LYTAC relative to PCSK9. The hook effect was also observed for this LYTAC, with reduced PCSK9 degradation observed at higher concentrations, e.g., 33% degradation was observed at 100 nM of 15a, and no apparent degradation was shown at 200 nM. While the mechanism of this phenomenon needs further clarification, one possibility is that the competitive binding of adjacent tri-GalNAc clusters of the antibody–ligand conjugates results in fewer productive endocytosis events by the ASGPR-LYTAC-PCSK9 ternary complex. Notably, this effect was not observed for LYTAC 14a bearing natural tri-antennary G3 N-glycans (Figure 3C) and is reflective of the same patterns observed in the cell-surface binding analysis. LYTAC 16a, containing the same tri-GalNAc ligand attached to a short disaccharide, was less effective than LYTAC 15a and degraded only 34% of extracellular PCSK9. The lower degradation induced by LYTAC 16a as opposed to LYTAC 15a containing the same tri-GalNAc ligand is likely due to the lower accessibility of the tri-GalNAc ligands attached to the bulky IgG Fc domain by a shorter glycan “linker”. Interestingly, the hook effect is also visible for LYTAC 16a, which induces decreased PCSK9 degradation at higher concentrations. As demonstrated by the western blot data in Figure 3, this effect is unique to LYTACs containing the synthetic tri-GalNAc ligand. Furthermore, preliminary time-course experiments revealed that incubation of HepG2 cells with LYTAC 15a for 48 h resulted in maximal degradation of ~65% extracellular PCSK9 (Figure S32), demonstrating similar degradation kinetics as other antibody-based degraders reported by Bertozzi, Tang, and co-workers.^16,17 Thus, significant lysosomal degradation of the clinically relevant PCSK9 protein was achieved by LYTACs 14a–16a. Notably, the results indicate that natural tri-antennary N-glycan ligands can be leveraged for targeted protein degradation without the hook effect.

To determine the effects of PCSK9 degradation on low-density lipoprotein receptor (LDLR) levels, western blots were performed on the lysates of treated cells. Specifically, LDLR levels were compared in cells treated with commercial antibody 8a and LYTACs 14a and 15a to determine whether targeted degradation of PCSK9 was a viable means of increasing cellular LDLR. As expected, treatment of cells with exogenous PCSK9 reduced the levels of LDLR (approximately 20%) relative to the untreated control bands. Interestingly, a lower-molecular-weight double band was also observed for the PCSK9-treated cells (Figure A–B). This phenomenon might be attributed to partial LDLR proteolysis through PCSK9-mediated internalization and subsequent degradation, as previously reported in some cases.^52,53 LDLR levels were rescued from PCSK9-induced degradation by treatment with commercial antibody 8a and up to 17% increase in LDLR protein was detected at the highest antibody concentration tested (Figure 4A). By comparison, treatment with LYTAC 14a did not result in a significant increase in LDLR relative to the control sample. However, treatment with the tri-GalNAc-modified LYTAC (15a) resulted in a 40% increase in total LDLR at 50 nM concentration (Figure 4B). Importantly, significant levels of PCSK9 degradation were detected for LYTAC 15a at this concentration, suggesting that PCSK9 degradation is responsible for the rescue of LDLR. These results indicate that PCSK9 degradation may be a relevant therapeutic avenue for the upregulation of the LDL receptor, which is responsible for the uptake and clearance of circulating LDL-C. Further experiments are needed to determine if the upregulation in total LDLR observed after treatment with PCSK9-targeting LYTAC 15a results in significantly increased LDL-C clearance in vivo. During the preparation of this manuscript, Bagdanoff and co-workers have demonstrated that a tri-GalNAc-modified PCSK9 antibody, prepared using the ThioBridge method for antibody conjugation, and other small heterobifunctional molecules were able to degrade approximately 60–70% PCSK9 in mouse serum, as evaluated in mouse models.³³ Our cell-based analysis data are consistent with their in vivo data. The above-cited study provides the first in vivo example of the clinical relevance for PCSK9 targeted degradation mediated by a LYTAC technology. In our study, we used the PCSK9 gain-of-function mutant D374Y, a more potent driver of hypercholesteremia than WT PCSK9, to demonstrate the targeted protein degradation. In addition, the Fc glycosylation site-specific bioconjugation described here provides an alternative approach to producing site-specific antibody-glycan ligand conjugates for targeted protein degradation.^54,55

Figure 4.

LDLR levels assessed by western blot of HepG2 cell lysates after treatment with each LYTAC. HepG2 cells treated with PCSK9 D374Y and serial dilutions of (A) native antibody 8a and (B) LYTAC 14a or 15a for 48 h. Media of cells treated with PCSK9 D374Y only and OptiMEM media were included as positive and negative controls of each blot, respectively.

Targeted Protein Degradation of Epidermal Growth Factor Receptor (EGFR).

To demonstrate the versatility of this platform in generating site-specific LYTACs targeting various proteins of interest, cetuximab-based LYTACs were chemoenzymatically engineered for the targeted degradation of membrane-bound EGFR (Figure 5A). HepG2 cells were treated with various concentrations of commercially available cetuximab 8b and LYTACs 10b and 13b–15b for 48 h in complete growth media. Total EGFR levels were then evaluated by western blot on the lysates of treated cells. As expected, little to no degradation was observed by commercial cetuximab (8b). LYTACs bearing the tri-antennary N-glycans attached directly to the Fc domain (LYTAC 10b) or the antibody conjugate carrying the biantennary N-glycans (LYTAC 13b) failed to degrade cell-surface EGFR (Figure 5B). However, LYTAC 14b bearing the tri-antennary complex-type N-glycans attached to the extended natural N-glycan core by click chemistry induced significant degradation of EGFR (up to 30%) at concentrations as low as 25 nM. Tri-GalNAc-modified LYTAC 15b, on the other hand, degraded as much as 45% EGFR at a concentration of 50 nM, with lower degradation values at higher LYTAC concentrations (Figure 5C). This result is comparable to degradation values of EGFR achieved with site-specifically modified M6P-based LYTACs, as well as the degradation level achieved by AbTAC bispecific antibodies leveraging E3 ligase RNF43.^15,56 To demonstrate that the observed degradation is ASGPR-dependent, HeLa cells (EGFR⁺ ASGPR⁻) that do not express ASGPR were exposed to commercial cetuximab 8b and LYTACs 14b and 15b. Indeed, no EGFR degradation was observed after any of these treatments, suggesting that EGFR degradation induced by LYTACs 14b and 15b was ASGPR-dependent (Figure 5D). Time-course degradation experiments were also performed for EGFR LYTACs, and maximal degradation was found after 48 h, in accordance with reports by Bertozzi, Tang, and co-workers (Figure S34).^16,17

Figure 5.

(A) Scheme of EGFR degradation by cetuximab-based LYTACs containing different ASGPR glycan ligands. EGFR levels assessed by western blot of HepG2 cell lysates after treatment with each LYTAC. HepG2 cells treated with serial dilutions of (B) native antibody 8b, or LYTACs 10b, or 13b or (C) LYTACs 14b or 15b for 48 h. (D) EGFR⁺ASGPR⁻ HeLa cells treated with native antibody 8b or LYTACs 14b and 15b for 48 h. Lysates of cells treated with complete growth media were included as negative controls for each blot.

These results provide evidence that natural tri-antennary asialo-complex-type N-glycans (G3) can be effective ligands for the lysosome-targeted degradation of membrane-bound and secreted extracellular proteins when attached with a sufficiently long spacer to monoclonal antibodies. The antibody conjugates displaying the synthetic tri-GalNAc cluster ligand were also found to be effective degraders of both secreted and membrane-bound target proteins, but the higher concentrations of the tri-GalNAc modified antibody conjugates were found to result in less degradation of the targeted proteins, due to the hook effect. It is well known that ASGPR interactions with Gal/GalNAc residues are weakened in the acidic pH environment of the endosome, resulting in dissociation of the calcium ion required for glycan binding, leading to glycanreceptor dissociation and receptor recycling. Several studies have identified three key residues in the carbohydrate recognition domain of the ASGPR that are protonated at reduced pH, including His²⁵⁶ which makes hydrophobic contacts with the methyl group of the GalNAc N-acetyl group and is responsible for the strong selectivity for GalNAc over galactose.^57–59 We propose that this extra hydrophobic contact may disfavor protonation of His²⁵⁶ and make the ASGPR-GalNAc interaction more resistant to the acid-induced dissociation. This would allow for recycling of the ASGPR-Tri-GalNAc-LYTAC complex to the cell surface to capture and deliver more target protein to the lysosome compared to the LYTACs carrying the galactose-terminated tri-antennary complex-type N-glycans.

The hook effect was observed in the targeted degradation of both PCSK9 and EGFR. However, this effect was found to be unique to antibody conjugates containing the synthetic tri-GalNAc cluster ligand and was not observed for LYTACs displaying the natural tri-antennary complex-type N-glycans. Furthermore, in contrast to chemoenzymatically engineered M6P-based LYTACs,¹⁵ natural tri-antennary N-glycan structures transferred enzymatically to the antibody Fc domain are not able to promote ASGPR-dependent degradation of target proteins. This finding highlights the distinct structural requirements of CI-M6PR and ASGPR-dependent LYTACs for the degradation of target proteins. For example, the CIM6PR receptor is composed of a long stalk of 15 receptor subunits, several of which contain mannose-6-phosphate binding domains that could accommodate binding to small M6P-containing glycans on antibodies while still allowing the antibody to form a ternary complex with the protein of interest.⁶⁰ The asialoglycoprotein receptor, on the other hand, consists of two distinct subunits (ASGR1 and ASGR2) that traverse the membrane and bind Gal and GalNAc-containing oligosaccharides only when displayed in a specific geometry.^39,61,62 The ASGPR glycan ligands conjugated to the antibody, therefore, must be in a proper configuration/orientation so that each Gal/GalNAc arm can engage a respective ASGPR subunit to achieve high-affinity binding.

CONCLUSIONS

A facile synthesis of site-specific antibody–ligand conjugates carrying different ligands for asialoglycoprotein receptor (ASGPR), including natural N-glycans and the synthetic tri-GalNAc cluster, was achieved by a chemoenzymatic Fc glycan remodeling method. Alirocumab and cetuximab, two therapeutic monoclonal antibodies that are against circulating PCSK9 and membrane-associated EGFR, respectively, were chosen for evaluating the ASGPR-mediated degradation of the targeted proteins. It was found that both the nature of the glycan ligands and the length of the spacer for the conjugation are critical for receptor binding and the receptor-mediated degradation of PCSK9 and EGFR. Interestingly, the antibody–tri-GalNAc cluster conjugates showed a clear hook effect for their binding to hASGPR and the degradation of PCSK9, i.e., a high concentration of the antibody–tri-GalNAc conjugates results in decreased affinity for the receptor and reduced efficacy for PCSK9 degradation. However, the antibody conjugates carrying the natural tri-antennary complex-type N-glycans showed significant affinity for the cell-surface receptor and dose-dependent degradation of PCSK9 without the adverse hook effect. This new finding provides important insight into the selection of appropriate glycan ligands for ASGPR-mediated targeted protein degradation. Future work should be directed to a more detailed mechanistic analysis of an expanded panel of natural and synthetic glycan ligands for cell-surface receptor binding and cell-based protein degradation to identify the best ligands for ASGPR-mediated protein degradation. In addition, in vivo experiments will be needed to evaluate the clinical potential of the site-specific antibody-glycan conjugates for targeted degradation of PCSK9, which offers a promising avenue for the treatment of high cholesterol.

MATERIALS AND METHODS

Materials.

Chemicals, reagents, and solvents were purchased from Sigma-Aldrich or TCI America and used as received unless otherwise specified. Monoclonal antibodies cetuximab and alirocumab were purchased from RefDrug and buffer-exchanged to PBS before use. Preparative HPLC was performed with Waters 1525 Binary HPLC pump coupled with 2489 UV/Vis Detector under UV 214 and 280 nm with a Waters Symmetry C18 column (7 μm, 19 mm × 300 mm) using water containing 0.1% trifluoracetic acid as phase A and MeCN containing 0.1% trifluoracetic acid as phase B. Semi-preparative HPLC was performed on the same instrument with an Agilent Eclipse XDB-C18 column (5 μm, 9.4 mm × 250 mm) using water containing 0.1% trifluoroacetic acid as phase A and MeCN containing 0.1% trifluoroacetic acid as phase B.

General Method for Antibody Deglycosylation Using Immobilized WT Endo-S2.

Wild-type Endo-S2 enzyme was immobilized following previously reported procedures, and the final concentration of the immobilized enzyme was determined to be 1.2 mg/mL by BCA assay. To a solution of monoclonal antibody in PBS (10 mg/mL) was added immobilized WT Endo-S2 to a final concentration of 0.5% w/w (50 μg), and the mixture was mixed end over end at room temperature overnight. The next day, the mixture was centrifuged to pellet the enzyme, and deglycosylation was confirmed by LC-ESI-MS. Once complete, the supernatant was filtered through a 0.2 μm cellulose syringe filter and buffer-exchanged to 100 mM Tris pH 7 buffer.

ESI-MS: calcd for GNF-Ali (9a), M = 146,675 Da; found (m/z), deconvolution of the ESI-MS, M = 146,677 Da. Calcd for GNF-Ali Fc fragment (9a), M = 24,102 Da; found (m/z), deconvolution of the ESI-MS, M = 24,102 Da.

ESI-MS: calcd for GNF-Cet (9b), M = 150,009 Da; found (m/z), deconvolution of the ESI-MS, M = 150,006 Da. Calcd for GNF-Cet Fc fragment (9b), M = 24,123 Da; found (m/z), deconvolution of the ESI-MS, M = 24,134 Da.

Protein A Purification of Monoclonal Antibodies and Their Conjugates.

Antibody reaction mixtures were diluted to 5 mL in PBS buffer pH 7.4 and loaded onto a pre-equilibrated HiTrap Protein A HP 1 mL column (GE Healthcare) by a syringe pump. The column was washed thoroughly with 20 mL of PBS pH 7.4 at 1 mL/min flow rate using an AKTA Pure FPLC system (Cytiva) and then eluted with 30 mM sodium citrate pH 2.6 buffer in one step. Samples were collected as 1 mL fractions in tubes containing 100 μL of 1 M Tris pH 7. Antibody-containing fractions were identified by absorbance at 280 nm and immediately buffer-exchanged to the appropriate buffer using Amicon Ultra 10 kDa MWCO centrifugal filters (Millipore). Antibody samples were collected, and final concentrations were determined by a NanoDrop.

General Method for Transglycosylation of Deglycosylated Antibodies with G3-Oxazoline.

Transglycosylation reactions with Endo-F3 D165A mutant were done following a previously reported procedure with some modifications.⁴⁴ To a solution of deglycosylated GNF-antibody in Tris buffer (15 mg/mL) was added 40 mol equiv of G3-oxazoline as a 100 mg/mL stock solution in water. Then, Endo-F3 D165A mutant was added to a final concentration of 0.2 mg/mL (14 μg) in a volume of ~70 μL and incubated at 37 °C for 20 min. After 20 min, the reaction was found to be complete for the transglycosylation of alirocumab, and the mixture was purified by protein A column, as described above, to give antibody 10a.

For transglycosylation of cetuximab, the reaction was only 70% complete after 20 min, as judged by LC-ESI-MS. Another 30 mol equiv of G3-oxazoline was added, and after 40 min, the reaction was complete and purified by protein A column to obtain antibody 10b. LC-ESI-MS data for antibodies 10a and 10b can be found in the Supporting Information.

Transglycosylation of Deglycosylated Antibodies with di-N₃-S2G2-Oxazoline.

Transglycosylation of deglycosylated alirocumab and cetuximab with di-N₃-S2G2-oxazoline was done following previously reported procedures to prepare antibodies 11a and 11b.³⁴

ESI-MS: calcd for di-N₃-S2G2-Ali (11a), M = 151,835 Da; found (m/z), deconvolution of the ESI-MS, M = 151,835 Da. Calcd for di-N₃-S2G2-Ali Fc fragment (11a), M = 26,681 Da; found (m/z), deconvolution of the ESI-MS, M = 26,681 Da.

ESI-MS: calcd for di-N₃-S2G2-Cet (11b), M = 155,162 Da; found (m/z), deconvolution of the ESI-MS, M = 155,161 Da. Calcd for di-N₃-S2G2-Cet Fc fragment (11b), M = 26,712 Da; found (m/z), deconvolution of the ESI-MS, M = 26,714 Da.

Transglycosylation of Deglycosylated Antibodies with di-N₃-Mannose-β (1,4)-GlcNAc-Oxazoline.

Transglycosylation of deglycosylated alirocumab with di-N₃-mannose-β (1,4)-GlcNAcoxazoline was done following previously reported procedures to prepare the antibody 12a.³⁵

ESI-MS: calcd for di-N₃-Man-GlcNAc-Ali (12a), M = 148,034 Da; found (m/z), deconvolution of the ESI-MS, M = 148,038 Da. Calcd for di-N₃-Man-GlcNAc-Ali Fc fragment (12a), M = 24,781 Da; found (m/z), deconvolution of the ESI-MS, M = 24,782 Da.

General Method for Click Conjugation of Azido-Modified Antibodies with DBCO-Tagged Glycan Ligands.

To a solution of azido-modified antibody in PBS (2 mg/mL) was added 12 mol equiv of DBCO-tagged glycan ligand (3 mol equiv per azide). The mixture was inverted end over end at room temperature, and the reaction progress was monitored by LC-ESI-MS. After 1 h, click conjugation of N-glycan-Asn-DBCO ligands was found to be complete with four units of the N-glycan ligand (bi- or tri-antennary) attached to the antibody Fc. Antibody conjugates were purified by spin filtration over a 10 kDa MWCO spin filter (Millipore). For conjugation of the synthetic Tri-GalNAc-DBCO ligand, longer reaction times of approximately 6–8 h were required for reaction completion under the same conditions. Tri-GalNAc-modified antibody LYTACs were also purified by spin filtration. LC-ESI-MS data can be found in the Supporting Information.

Cell Lines and Reagents.

HepG2 and HeLa cells were purchased from ATCC and cultured in T75 flasks at 37 °C and with 5% CO₂. HepG2 and HeLa cells were cultured in EMEM medium containing 10% FBS and 1% penicillin/streptomycin antibiotics. For serum depletion of HepG2 cells, OptiMEM serum-free medium (Gibco) was used with no other additives. Antibodies used in western blotting experiments are listed in Supporting Information Table S1.

Flow Cytometry Binding Assay of Biotinylated LYTACs to Cell-Surface ASGPR.

HepG2 cells in a T75 flask were serum-starved in OptiMEM media for 12–16 h and then detached by incubation with trypsin (0.25%)-EDTA (0.02%) solution (Quality Biological) at 37 °C. Then, the cells were resuspended in wash/stain buffer (DPBS, 2% FBS, 0.1% NaN₃) before counting. HepG2 cells were then seeded to 3 × 10⁵ cells per well of a 96-well V-bottom plate in 200 μL volumes, and the plate was spun at 1,400 rpm for 5 min to pellet the cells. The wash/stain buffer was removed, and 100 μL of Fc block solution (Human BD Fc Block, BD Cat. # 564220) was added to the wells. The plate was incubated in the dark for 10 min at room temperature and spun again at 1400 rpm for 5 min. Test antibody and LYTAC dilutions were then prepared in wash/stain buffer in a fresh 96-well V-bottom plate and then added in 100 μL volumes to the cells. Wash/stain buffer only was used as a negative control and added to the designated control wells. The plate was incubated for 30–45 min in the dark at room temperature, and 100 μL of wash buffer was then added to each well before pelleting the cells again. The supernatant was removed, and a diluted solution of streptavidin-PE conjugate (BD Cat. # 349023) was added in 100 μL volumes to the test and control wells. As a positive control, PE-conjugated mouse anti-ASGPR1 antibody (BD Cat. # 563655) was added at 25 μg per test to the control wells. The plate was incubated for 30 min in the dark before pelleting the cells. The supernatant was removed, and each well was washed with 200 μL of wash/stain buffer. The cells were pelleted and washed again before resuspending the cells in 50 μL 1% paraformaldehyde to fix the cells. The samples were analyzed for fluorescence intensity using a Fortessa/LSR instrument under the high-throughput option for the plate reader. Fluorescence data were analyzed with Flowjo Ver.9.9.6 (Flowjo, Ashland, OR).

Epidermal Growth Factor Receptor (EGFR) Protein Degradation Assay.

HepG2 and Hep3B cells were seeded at 100,000 cells/well in a 24-well plate one day prior to the degradation assay. The following day, the media was discarded, and adherent cells were treated with variable concentrations of each cetuximab LYTAC as a solution in EMEM complete growth medium (0.5 mL). Cells were incubated with cetuximab LYTACs for 48 h, at which point, the media was discarded and cells were washed twice with 0.5 mL of PBS. Finally, 100 μL of Pierce RIPA lysis buffer containing cOmplete Mini protease inhibitor (Sigma) and phosphatase inhibitor cocktail (Cell Signaling Technologies) was added, and the cells were lysed for 30 min on ice. Cells were scraped, transferred to 1.5 mL centrifuge tubes, and centrifuged at 13,500 rpm for 15 min to pellet cell debris. The supernatant was collected inro new microcentrifuge tubes and assayed for total protein concentration by Pierce BCA assay (Thermo). Equal amounts of cell lysate were loaded and separated on an 8–16% SDS-PAGE gel and imaged using a Bio-Rad Gel Doc EZ imager. Lysate protein was then transferred to a nitrocellulose blot and blocked with 5% nonfat dairy milk (ChemCruz) for 1 h at room temperature. Nitrocellulose blots were washed with TBST (Tris-buffered saline containing 0.1% Tween-20) as the washing buffer and then incubated with primary antibodies overnight at 4 °C (rabbit anti-EGFR 1:1000 dilution and rabbit anti-β-actin 1:5000 dilution). Blots were washed three times in TBST and then incubated with secondary antibody (goat anti-rabbit IgG [H + L] HRP-conjugate 1:2500 dilution) for 1 h at room temperature before washing four more times with TBST buffer. Finally, ECL Plus chemifluorescent reagent (Thermo) was added to the blots and incubated for 5 min before performing fluorescent imaging using a Molecular Devices Storm 860 imager with excitation wavelength at 450 nm.

PCSK9 Protein Degradation Assay.

HepG2 cells were seeded at 100,000 cells/well in a 24-well plate one day prior to the degradation assay. The following day, the media was discarded, and wells were washed once with 0.5 mL PBS before 0.5 mL of serum-deficient OptiMEM media was added to each well. HepG2 cells were then serum-deprived for 12–16 h. Meanwhile, human PCSK9 D374Y gain-of-function mutant (ACROBiosystems) was prepared as a 10 μg/mL stock solution in OptiMEM and, separately, alirocumab LYTACs were diluted to twofold higher than the desired concentration for the assay in OptiMEM. Equal volumes (150 μL) of the PCSK9 D374Y stock solution and LYTAC solution were then mixed and incubated for 30 min at room temperature. Importantly, the PCSK9 solution was incubated with varying concentrations of each LYTAC to test for a dose–response. A positive control consisted of the PCSK9 D374Y stock solution mixed with OptiMEM containing no LYTAC, and OptiMEM alone was used as a negative control. All experimental and control samples were added to the cells in the 24-well plate as 250 μL aliquots, and cells were incubated for 48 h at 37 °C. Then, the media in each well was collected into separate 1.5 mL microcentrifuge tubes and labeled, and cells were washed twice with 0.5 mL of PBS. Finally, 100 μL of RIPA lysis buffer containing cOmplete Mini protease inhibitor cocktail (Sigma) and phosphatase inhibitor cocktail (Cell Signaling Technologies) was added, and cells were lysed for 30 min on ice. Cell lysate samples were then collected, as described above. PCSK9 levels in the media were tested by western blot using a polyclonal anti-PCSK9 primary antibody (rabbit anti-PCSK9 1:1000), where the total protein content was normalized by adding the same volume of media to each well of an 8–16% SDS-PAGE gel. Protein was then transferred to a nitrocellulose blot and processed as described above for the EGFR degradation assay. Levels of low-density lipoprotein receptor (LDLR) were also tested by western blot using an anti-LDLR primary antibody (rabbit anti-LDLR 1:2000 dilution) on the HepG2 cell lysate and normalized to β-actin as a loading control (rabbit anti-β-actin 1:5000 dilution), following the general procedure outlined above for the EGFR degradation assay.