2.1 Clinical characteristics

The clinical characteristics of the 64 hypercholesterolemic patients who received either a statin (n = 25) or a PCSK9i (n = 39) are provided in Table 1. No significant differences between the two groups were found, except for the fact that PCSK9i users were older (p < 0.001), and had higher triglycerides (TG) levels (p < 0.05) and a higher body mass index (BMI) (p < 0.01). Although chemerin levels were not related to sex, LDLR mutation, or history of smoking (Table 2), they were elevated in patients who had experienced a myocardial infarction (p < 0.05) and in those with hypertension (p < 0.001). Circulating chemerin correlated positively with TG (R = 0.37, p < 0.01; Table 1), hsCRP (R = 0.42, p < 0.001), and thromobocyte levels (R = 0.44, p < 0.05), and negatively with HDL-C and apolipoprotein A-I (ApoA-I) levels (R = −0.32 and −0.27, respectively, p < 0.05 for both).

Table S1 provides the clinical characteristics of 23 hypercholesterolemic patients that received a PCSK9i (alirocumab n = 12, evolocumab, n = 11) on top of a statin (fluvastatin n = 5, simvastatin n = 6, atorvastatin n = 6, and rosuvastatin n = 6). Their age and BMI were comparable to those receiving PCSK9i only.

2.2 Effects of statins and PCSK9i on lipid levels and inflammatory markers

Statins and PCSK9i administration resulted in the expected effects on lipids (Figure 1A–D), that is, decreases in TG, total cholesterol (TC), and LDL-C, and an increase in HDL-C (p < 0.01 for all). Yet, only statin treatment reduced chemerin (p < 0.0001; Figure 1E) and hsCRP (p < 0.001; Figure 1F). Additionally, the effect of PCSK9i on TG (p < 0.05), TC (p < 0.001), LDL-C (p < 0.0001), and HDL-C (p < 0.01) was larger than that of statins. The reductions in LDL-C and chemerin were unrelated to sampling time (R = 0.14, p = 0.24; R = −0.16, p = 0.17, respectively). Furthermore, all above described correlations with circulating chemerin remained unaltered after treatment (Table 1), except the negative correlation between chemerin and HDL-C which disappeared after statin treatment. Table S2 shows that the statin effect on chemerin is seen with fluvastatin, simvastatin, and atorvastatin, but not rosuvastatin.

In patients that received PCSK9i treatment on top of a statin, further decreases in TG (p < 0.05), TC (p < 0.0001), apolipoApoB (p < 0.0001), and LDL-C (p < 0.0001) were observed, while HDL-C (p < 0.01) increased (Table S1 and Figure S1). Yet, there was no change in either chemerin or hsCRP. Dual treatment did not affect the positive correlation between chemerin and TG, while the correlation between chemerin and HDL-C now became positive (R = 0.45, p < 0.05).

In summary, both lipid-lowering therapies exerted their expected effects on lipids, but only statins additionally lowered chemerin and hsCRP.

2.3 Chemerin distribution across lipoprotein subfractions

At baseline, in both treatment groups more than 75% of chemerin was in the free form, that is, unbound to lipoproteins (Figure 2). Among the lipoproteins fractions, chemerin occurred predominantly in the HDL₃ and HDL₂ fractions. Treatment did not alter this distribution, and reductions in absolute amounts occurred only in the free (p < 0.001) and HDL₃ (p < 0.01) fractions after statin treatment, while a similar tendency was observed for the HDL₂ fraction (p = NS). Taken together, these data indicate that the statin-induced reduction in chemerin is a general phenomenon, that is, it is not limited to a specific fraction.

2.4 Identification of chemerin–ApoA-I interaction in hypercholesterolemia plasma

The chemerin occurrence in the smaller HDL fractions might imply that chemerin interacts with ApoA-I.^{18, 19} To investigate such interaction, streptavidin-labeled anti-human chemerin was added to a hypercholesterolemic patient plasma pool. Chemerin could be detected in the elution fluid of the streptavidin beads, and this was also true for ApoA-I (Figure 3A,B). Moreover, when supplementing the plasma pool with human chemerin-His protein, His tag, chemerin and ApoA-I were subsequently detected in the elution fluid of His-tag beads (Figure 3C–E). In summary, these data support physical chemerin–ApoA-I interaction in hypercholesterolemic plasma.

2.5 Statins and PCSK9i counteract the inhibitory effect of chemerin on the cholesterol efflux ability of HDL

Chemerin–ApoA-I interaction could be involved in the HDL-induced cholesterol efflux response. We tested this hypothesis in differentiated THP-1 cells. After 24 h of incubation, chemerin was below detection limit in the medium of differentiated THP-1 cells, but easily detectable when the HDL₃ fraction had been added to the medium (Figure S2A). HDL₃ fraction addition decreased the key proteins ATP Binding Cassette subfamily A member 1 (ABCA1) and ATP Binding Cassette subfamily G member 1 (ABCG1) (Figure S2B), indicating inhibition of cholesterol transport. Moreover, adding the HDL₃ fraction increased CMKLR1 but not chemerin receptor C-C motif chemokine receptor-like 2 (CCRL2; Figure S2B). The anti-human chemerin antibody enhanced cholesterol efflux during incubation with the HDL₃ fraction (Figure 4A), and this was not altered by α-NETA, pravastatin, or alirocumab. Alirocumab even increased CMKLR1 under these conditions (Figure S2B). To investigate the effect of chemerin on cholesterol transport in differentiated THP-1 cells, chemerin-9, the active isoform of chemerin, was used. Chemerin-9 inhibited cholesterol efflux in the presence of both human serum and human ApoA-I (serving as the cholesterol acceptor) when compared with vehicle (Figure 4B). The CMKLR1 antagonist α-NETA, pravastatin, atorvastatin, rosuvastatin, alirocumab, and evolocumab prevented this effect in the presence of serum, while only α-NETA and pravastatin prevented it in the presence of ApoA-I (Figures 4B and S3). Western blot analysis of cell lysates (Figure 4C) revealed that all antagonists, either with or without chemerin-9, upregulated HMGCR, ABCA1, and ABCG1, while a similar tendency was shown for liver X receptors α and β (LXRα and LXRβ). In addition, alirocumab and pravastatin, both alone and in the presence of chemerin-9, upregulated LDLR. Taken together, these data indicate that chemerin inhibits cholesterol efflux via CMKLR1, and that both statins and PCSK9i prevent this effect, potentially by interfering with key proteins involved in cholesterol transport (ABCA1 and ABCG1).

2.6 Statins, not PCSK9i, suppresses chemerin release from hepatic cells

Given the hepatic origin of circulating chemerin,²² we used HepG2 cells to verify whether statins or PCSK9i directly affect chemerin synthesis. As expected, HepG2 cells release chemerin in a time-dependent manner (Figure S4A). Chemerin could not be detected in fetal bovine serum (FBS)-containing medium in the absence of cells (Figure S4A), excluding that its release represented chemerin in FBS. Pravastatin, but not alirocumab, concentration-dependently suppressed the release of chemerin into the medium (Figure 5A), and the degree of suppression (around 15−20%) mimicked the in vivo suppression of circulating chemerin by statins (Figure 1E). Neither drug affected cellular chemerin (Figure 5D), and although both drugs tended to decrease chemerin gene expression (Figure 5B), this was not significant. Only pravastatin upregulated LDLR gene expression (by 436%, p < 0.0001; Figure 5C), while both pravastatin (p < 0.01) and alirocumab (p < 0.05) at most modestly upregulated LDLR at the protein level (Figure 5D). This implies that the increase in LDLR gene expression after pravastatin did not result in a parallel increase in LDLR protein levels. No change in the unfolded protein response-related proteins glucose-regulated protein 78, CCAAT-enhancer-binding protein homologous protein, protein kinase R-like endoplasmic reticulum kinase, X-box binding protein 1, activating transcription factor 6, and eukaryotic translation initiation factor 2 subunit 1 was observed (Figure S4B,C). This argues against an alteration in the accumulation of unfolded or misfolded proteins as an explanation of the effect of pravastatin.

Since the transcription factor sterol-regulatory-element-binding protein 2 (SREBP2) is responsible for both LDLR and chemerin transcription, one further possibility is that upregulating LDLR transcription after statin exposure, diminishes concomitant chemerin transcription. In agreement with this concept, knockdown of LDLR during incubation with pravastatin even increased chemerin release, while additional SREBP2 knockdown prevented this effect (Figure 5E). LDLR knockdown, SREBP2 knockdown, or both during pravastatin or alirocumab exposure did not affect chemerin mRNA levels (Figure 5F), although they did reduce LDLR mRNA levels during pravastatin exposure (Figure 5G). As expected, SREBP2 knockdown diminished the SREBP2 mRNA levels during both pravastatin and alirocumab exposure (Figure 5H). Cell lysate data revealed that pravastatin during LDLR knockdown upregulated SREBP2 and that additional SREBP2 knockdown prevented this (Figure S4D). No such observations were made during incubation with alirocumab, although LDLR knockdown during alirocumab incubation did downregulate the LDLR protein level to the same degree as during pravastatin incubation.

Finally, in a concentration-dependent manner, both atorvastatin and rosuvastatin, but not evolocumab, were also capable of decreasing chemerin release into the medium (Figure 5I).

In summary, these data imply that statins, but not PCSK9i, diminish chemerin secretion from HepG2 cells, and that this involves LDLR upregulation.

4.1 Study population

Patients (n = 87) diagnosed with hypercholesterolemia and attending the outpatient clinic at Erasmus Medical Center in Rotterdam, the Netherlands, between October 2011 and October 2019 were included in this study. A total of 64 of patients received either a statin, or, in case of statin intolerance (defined as statin-associated myalgia for at least three different statins, in accordance with the criteria described by the European Atherosclerosis Society/European Society of Cardiology consensus),³⁸ a PCSK9i. A total of 23 patients received statin treatment combined with a PCSK9i. All hypercholesterolemic patients receiving a PCSK9i met the Dutch criteria for treatment with such a drug, that is, having an LDL-C level > 2.5 mmol/L without cardiovascular disease events or an LDL-C level > 1.8 mmol/L with a history of cardiovascular disease events.³⁸ Figure S5 provides the flowchart for patient inclusion from our existing open cohort study.³⁹ Hypercholesterolemia diagnosis was based on a documented LDL receptor mutation or clinical hypercholesterolemia criteria (Dutch Lipid Clinic Network score ≥6).⁴⁰ The category “normal blood cholesterol” was defined as having a TC < 5 mmol/L and LDL-C < 3 mmol/L, and not being on blood lipid-lowering treatment.^{38, 40} In agreement with standard clinical dose recommendations, statin treatment involved simvastatin, fluvastatin, rosuvastatin, or atorvastatin (40, 80, 20, and 20 mg per day, respectively), while PCSK9i treatment involved alirocumab or evolocumab (140 and 150 mg every 2 weeks, respectively).² A follow-up visit occurred between 30 and 250 days after initiation of the therapy. Venous blood samples were collected in EDTA tubes after an overnight fast both at the start of therapy and during follow-up. Blood was centrifuged at 2500×g, and the resulting plasma was stored at −80°C until analysis. Anthropometric data were obtained from medical records. Participants’ medical history, family history, and lifestyle information were assessed via a questionnaire. Hypertension was defined as systolic blood pressure > 140 mmHg and/or diastolic blood pressure > 90 mmHg, or the use of antihypertensive therapy. Mean arterial pressure was calculated using the formula (2 × diastolic pressure + systolic pressure)/3. Participants were categorized as smokers or never smokers based on their self-reported smoking status.

4.2 Lipoprotein subfraction isolation

Plasma lipoprotein subfractions were separated using the density-gradient UC method described previously.⁴¹ The density gradient was prepared by adding KBr (0.35 g/mL) to plasma, resulting in a density of 1.26 g/mL, of which 1 mL was placed in an ultracentrifuge tube. On top of this layer, KBr solutions with densities of 1.21, 1.10, 1.063, 1.04, and 1.02 g/mL, each comprising 1.9 mL, were sequentially added, followed by 1 mL of water. The samples were then ultracentrifuged at 221,633xg for 18 h at 4°C using a SW41 rotor in an L-40 Beckman ultracentrifuge (Beckman Instruments, USA). The resulting density gradient was fractionated, starting from the bottom of the tube, into 250 µL fractions. The fractions with densities of 1.26–1.21 g/mL were designated as free protein without lipoprotein binding level (Free). Fractions with densities of 1.125−1.21 and 1.062−1.125 g/mL were identified as HDL₃ and HDL₂, respectively. Additionally, fractions with densities ranging from 1.038−1.063 and 1.019−1.038 g/mL were characterized as small-dense LDL (dLDL) and buoyant LDL (bLDL), respectively. Fractions with a density ≤ 0.971 g/mL represented VLDL.⁴²

4.3 Cell culture

HepG2 and THP-1 cells were obtained from the American Type Culture Collection (ATCC, Manassas, VA, USA). HepG2 cells were cultured in complete medium with Eagle's Minimum Essential Medium (ATCC) and 10% FBS (GE Healthcare, Eindhoven, the Netherlands) in T-75 flasks (Costar, Corning, NY, USA) at 37°C with 5% CO₂. The medium was refreshed every 2−3 days and the cells were subcultured at 80−85% confluence. THP-1 cells were maintained in complete medium composed of RPMI-1640 (ATCC) supplemented with 10% FBS and 0.05 mmol/L 2-mercaptoethanol (Gibco, Grand Island, NY, USA) in T-75 flasks (Costar) at 37°C with 5% CO₂. The use of the initial thawed vial was limited to 10 passages to control for genetic drift in the experiments.

4.4 Plasma pool pull-down assay

A 3-mL hypercholesterolemia plasma pool was generated as described above. To conduct the pull-down assay, 300 µL of this plasma pool was incubated for 30 min at room temperature with either 1 µg anti-human chemerin labeled with streptavidin (R&D Systems, Abingdon, UK) or 5 µg recombinant human chemerin-His protein (Thermo Fisher Scientific). Subsequently, 50 µL of prewashed Dynabeads of streptavidin (Thermo Fisher Scientific) was added to the plasma pool with anti-human chemerin labeled with streptavidin, and 50 µL of prewashed Dynabeads of His-tag (Thermo Fisher Scientific) was added to the plasma pool with human chemerin-His protein. The same amount of Dynabeads was also added to blank plasma. Plasmas were incubated overnight at 4°C while being rotated. Next, the Dynabeads were removed with a magnet, and washed four times with wash buffer (50 mmol/L sodium phosphate, pH 8.0 containing 300 mmol/L NaCl and 0.01% Tween-20). Next, 50 µL elution buffer (identical to the wash buffer and additionally containing 300 mmol/L imidazole) was added to elute the pulled-down proteins, which were collected for further western blotting. Western blotting was performed in parallel in the wash buffer and plasma remaining after the Dynabeads removal.

4.5 Biochemical measurements

TG, TC, LDL-C, and HDL-C were determined using standard clinical chemistry techniques at the Clinical Chemistry Department of the Erasmus MC. Chemerin in plasma samples, lipoprotein subfractions, and cell culture medium was assessed by a commercial enzyme-linked immunosorbent assay (ELISA) according to the manufacturer's guidelines (R&D Systems nr. DY2324). The detection limit was 31.2 pg/mL. Plasma hsCRP levels were also quantified by commercial kit, according to the manufacturer's instructions (R&D Systems nr. DY1707). The detection limit was 15.6 pg/mL.

4.6 Western blotting and real-time PCR

Protein levels from cell lysates were quantified using the BCA assay kit (Thermo Fisher Scientific). For immunoblotting analysis, cell lysates containing an equal amount of protein (10–25 µg), 10-µL plasma pool samples, and 15 µL of wash buffer and pulled-down samples were resolved through SDS-PAGE and subsequently probed using primary antibodies as detailed in Table S3. HRP-conjugated goat anti-mouse or goat anti-rabbit antibodies (Bio-Rad Laboratories, Hercules, CA, USA) were then added, followed by detection with ECL (Bio-Rad Laboratories).

Total RNA from the cell lysates was extracted using a commercial kit according to the manufacturer's instructions (Zymo Research, Orange, CA, USA), and reverse transcribed using PrimeScript RT Master Mix for real-time PCR (Takara Shuzo, Kyoto, Japan). Real-time PCR was performed using SYBR®Premix Ex TaqTM II kit (Qiagen, Venlo, The Netherlands) according to the manufacturer's instructions as described before.⁴³ Primer sequences are listed in Table S4. Expression levels were determined using a standard curve, and human 36B4 was used as an internal control to normalize gene expression levels.

4.7 Statistical analysis

Statistical analysis was conducted using GraphPad Prism 9.5.0 (GraphPad Software, Inc., San Diego, CA, USA). Data of plasma are presented as mean ± standard deviation for normal distributions and median with range for non-normal distributions. Data of cell experiments are presented as mean ± SEM. The normality of the data distribution was evaluated using the Kolmogorov-Smirnov test. For the clinical data, variables were compared using either a paired t-test or Wilcoxon test. Categorical variables were compared using the χ²-test. To examine the association between chemerin and other variables, Spearman's correlation analysis was performed, after correction for sex, age, and BMI. The levels of chemerin in lipoprotein subfractions was determined using two-way ANOVA and Sidak's multiple comparison test. One-way ANOVA was used to compare parametric data, while Kruskal–Wallis was used to compare nonparametric data for western blotting of cell lysates, qPCR data of HepG2 cells, ELISA's of the medium, and the percent of cholesterol efflux. p < 0.05 was considered statistically significant.