Circulating low-density LDL-c is a major risk factor for the development of cardiovascular diseases leading to atherosclerosis and myocardial complications. The LDLR is responsible for LDL-c uptake in many tissues, the most important one being in the liver, which clears ∼70% of LDL-c1. The major cholesterol-lowering drugs prescribed in clinics worldwide are HMG-CoA reductase inhibitors, known as statins. A number of outcome mega-epidemiological studies have shown that the absolute levels of LDL-c before or after statin treatment are directly proportional to the extent of nociceptive cardiovascular events, and thus, especially for patients at high risk, LDL-c levels <70 mg/dL (1.8 mM) are recommended2. However, many patients fail to achieve such levels even when high-intensity statin is combined with intestinal cholesterol-absorption inhibitors such as ezetimibe3.

The discovery of the 692-amino-acid (aa) protein PCSK9 (ref. 4), the ninth member of the proprotein convertase family of secretory proteases5, and its relationship to LDL-c6,7 revealed that PCSK9 directly regulates liver LDLR levels by dragging this receptor from the cell surface to acidic endosomes or lysosomes for degradation5,8. After its autocatalytic activation in the endoplasmic reticulum of hepatocytes, the prodomain of PCSK9 (amino acids 31–152) remains noncovalently but tightly bound to the mature catalytic domain (amino acids 152–692) (ref. 4). Thus PCSK9 is secreted as an inactive protease that is nonetheless able to bind to and reduce the levels of hepatic LDLR. Biochemical analyses9 and crystal structures of PCSK9 alone10 or in complex with a soluble form of the LDLR11 revealed that the EGFA domain of the LDLR directly binds the catalytic domain of PCSK9. However, this binding is not sufficient to effectively chaperone the LDLR to endosomes and lysosomes, and the C-terminal cysteine/histidine-rich domain (aa 453–692) is needed to direct the complex PCSK9–LDLR to degradation compartments12 by an as yet unknown mechanism13.

It became evident that blocking the binding of PCSK9 to the EGFA domain of LDLR on the cell surface (Fig. 1) might be an effective approach to inhibit the function of circulating PCSK9. Indeed, two very potent humanized monoclonal antibodies (evolocumab and alirocumab) can lower LDL-c levels by 50–60% when injected every two weeks14,15 and are now prescribed worldwide. Recent outcome data confirmed that such treatment reduced overall cardiovascular events by ∼20% (ref. 16). Another promising approach has been to inject RNAi-containing nanoparticles (inclisiran), resulting in substantial reduction of PCSK9 synthesis17. Phase 2 clinical trials demonstrated that a single injection may be enough to reduce LDL-c levels for at least six months18. Although other approaches to reduce PCSK9 levels have been proposed, no small-molecule inhibitor of PCSK9 function has been identified. The main problem has been the lack of a targetable pocket at the EGFA–PCSK9 interface, which is relatively flat and covers an ∼500-Å2 area19.

Figure 1: Inhibition of the PCSK9-induced LDLR degradation.
figure 1

Primary hepatocytes isolated from PCSK9 12-week-old male knockout mice7 were incubated with ∼5 μg/ml of human purified PCSK9 for 5 h at 37 °C in the absence (top) or presence (bottom) of 0.35 μM evolocumab, a PCSK9 mAb inhibitor. After the first 3 h of incubation, 5 μg/ml of the fluorescent DiI-LDL (red) was added to measure LDL internalization for the next 2 h. After the 5 h of incubation, the cells were fixed for analyses. The level of cell-surface LDLR (green) was measured by immunocytochemistry with an anti-mouse LDLR antibody. Nuclei are stained with DAPI (blue). Notice the increase in cell-surface LDLR immunolabeling (green) and DiI-LDL internalization (red) in the presence of the mAb. Scale bar, 25 μm.

Full size image

In this issue, Zhang et al.20 have fortuitously identified a targetable pocket in PCSK9 that is very close to the EGFA–PCSK9 interaction surface in the X-ray crystal structure of a PCSK9–mAb33 complex. This cryptic groove attracts the N-terminal 10-aa P′-helix peptide (aa 153–162; SIPWNLERIT) of the catalytic domain, which sprang out upon autocatalytic cleavage at Gln152 of the prodomain of PCSK9, keeping it >20 Å away from the catalytic pocket. The groove-bound P′ helix is predominant in the natural state of the zymogen-processed mature PCSK9. In the P′-helix-bound state, the interaction with the EGFA domain of the LDLR is stabilized. However, owing to major steric constraints, the P′ helix peptide can be displaced from the groove in the presence of mAb33, as evidenced by its accessibility to cleavage by proteases such as Factor XIa at Arg160, thereby suggesting that the P′-helix segment is intrinsically labile.

In order to define the critical residues in the P′-helix peptide, Zhang et al. used a phage-display strategy in which multiple variants of the P′ helix were fused via a glycine-serine-glycine (GSG) linker to a variant of the 13-residue LDLR-binding peptide Pep2-8V2 (ref. 21). Following affinity maturation, nine sequences robustly bound PCSK9, seven of which contained the WNLxRI motif in the native PCSK9 P′-helix peptide. This led to the synthesis of a potent 30-residue bivalent PCSK9 antagonist to LDLR binding in vitro (IC50 ∼23 nM), consisting of a 14-residue disulfide-linked cyclic P′-helix-binding peptide C–terminally fused to the LDLR-binding peptide, Pep2-8–GSG–CRLPWNLQRIGLPC. This peptide completely antagonized the effects of extracellular PCSK9 on induction of surface LDLR degradation in human hepatic HepG2 cells, with a 20-fold superior activity (5 μM) compared to Pep2-8 alone (100 μM).

This finding prompted an intense engineering effort to obtain a shorter antagonist lacking the Pep2-8 anchoring sequence, taking into account that the groove interaction of the native P′ helix is facilitated by its being tethered to the catalytic domain. These efforts ultimately produced a modestly active (>75 μM for surface LDLR restoration by 70% in HepG2 cells), first-generation 16-residue linear peptide antagonist of PCSK9 activity MESFPGWNLV(homoR)IGLLR. The N-terminal FPG motif together with the Trp7 of the core peptide form a β-turn in which Pro5 projects into the LDLR-binding site, thereby causing a steric clash with the EGFA domain that underlines the antagonistic activity of this peptide.

The identification of a critical P′ helix in PCSK9 rationalizes the reported loss of function of the only known P′ helix p.N157K natural mutation22, which should disrupt the WNLxRI motif. It would be informative to perform site-directed mutagenesis to systematically validate critical residues in the P′-helix sequence of PCSK9. Interestingly, of the nine known mammalian proprotein convertases5, only PCSK9 has a tripeptide SIP preceding the WNLxRI motif, whereas all other mature human PCSKs have between 21 (PC1) and 13 (PC4) N-terminal residues preceding their common WY(L/M) tripeptide motif. As noted by Zhang et al.20, this longer N-terminal stretch may hinder the intrinsic lability of the presumed P′ helix in other proprotein convertases.

PCSK9 was reported to be cleaved by furin at Arg218, resulting in a drastically less active circulating form of PCSK9 lacking residues 153–218 (PCSK9-Δ218)23. This processing by furin would release the P′ helix, probably contributing to the loss of function of PCSK9-Δ218. Furthermore, it was reported that the mAb LY3015014 targeting the N-terminal stretch of PCSK9 (residues 160–181) inhibits function but still allows furin cleavage, resulting in antibodies that are longer lasting than those prescribed in clinics, which disrupt the LDLR–PCSK9 interaction but prevent furin cleavage24. This mAb is expected to cover part of the P′-helix segment and abolish its PCSK9–EGFA stabilization potential.

The future goal of identifying nonpeptide orally active inhibitors of PCSK9 is challenging, but theoretically achievable, as reported for a number of peptidomimetic drugs targeting proteases, cancer, pain and viral infections. The present findings provide a framework for in vivo functional validations and pharmacodynamic analyses of MESFPGWNLV(homoR)IGLLR, or an optimized stable derivative, in model animals and ultimately in humans. Additional functional grooves in other parts of the PCSK9 protein have yet to be reported. Nonetheless, the existence of a targetable groove in the catalytic subunit provided the first glimpse and hope for the not-too-distant future isolation of small molecule inhibitors of PCSK9.