Identification and in vitro characterization of two new PCSK9 Gain of Function variants found in patients with Familial Hypercholesterolemia

Familial hypercholesterolemia (FH) is an autosomal dominant disease caused by pathogenic variants in genes encoding for LDL receptor (LDLR), Apolipoprotein B and Proprotein convertase subtilisin/kexin type 9 (PCSK9). Among PCSK9 variants, only Gain-of- Function (GOF) variants lead to FH. Greater attention should be paid to the classification of variants as pathogenic. Two hundred sixty nine patients with a clinical suspect of FH were screened for variants in LDLR and the patients without pathogenic variants were screened for variants in PCSK9 and APOB. Functional characterization of PCSK9 variants was performed by assessment of protein secretion, of LDLR activity in presence of PCSK9 variant proteins as well as of the LDLR affinity of the PCSK9 variants. Among 81 patients without pathogenic variants in LDLR, 7 PCSK9 heterozygotes were found, 4 of whom were carriers of variants whose role in FH pathogenesis is still unknown. Functional characterization revealed that two variants (p.(Ser636Arg) and p.(Arg357Cys)) were GOF variants. In Conclusions, we demonstrated a GOF effect of 2 PCSK9 variants that can be considered as FH-causative variants. The study highlights the important role played by functional characterization in integrating diagnostic procedures when the pathogenicity of new variants has not been previously demonstrated.

to decreased LDLR levels on the cell membrane available for LDL uptake 3 . Another mechanism independent of endocytosis, indicating an intracellular action of PCSK9 in LDLR expression regulation, was observed 4 .
The discovery of PCSK9 as a player of LDL uptake opened new therapeutic avenues. In fact PCSK9 has become the target of several therapies administered in case of failure of traditional therapies or in the most severe cases 5 . Anti-PCSK9 monoclonal antibodies are the most promising 6,7 .
Two different types of pathogenic variants have been identified in this gene: 1. loss of function (LOF) variants producing a less functioning protein, causing an increase of LDLR amounts on the cell membrane and, consequently, hypocholesterolemia; 2. gain of function (GOF) variants producing a more functioning protein that degrades LDLR more efficiently, decreasing its levels and causing FH 8 .
An accurate evaluation of each variant identified during genetic screening is essential to define its pathogenic role. Although several bioinformatics tools are now available 9 , in silico predictions are not sufficiently effective to reliably assess the pathogenicity of variants and in particular of GOF variants 10 . Therefore, functional characterization by in vitro assays is the most effective and reliable method to evaluate the pathogenic role of gene variants and it is especially required for evaluating PCSK9 variants. Recent guidelines support this concept and suggest that, among different criteria, functional assays can provide strong evidence of pathogenicity 11 . Several methods have been proposed to functionally characterize FH causative variants [12][13][14] . We report herein the characterization of 4 rare variants in the PCSK9 gene following 3 different approaches.

Results
Genetic screening. The screening of LDLR gene revealed the presence of variants causative of FH in 188 patients out of 269. In the remaining 81 patients the genetic screening of PCSK9 revealed the presence of 7 rare missense variants at heterozygous status (2.6% of total patients): 3 variants previously identified in FH patients; 4 variants never associated to FH before. Table 1 reports data about the 7 PCSK9 variants together with the lipid profile of carriers. The PCSK9 variant c.1906A > C (p.(Ser636Arg)) was identified in a patient carrying also a very rare variant in the LDLR gene, the c.1336 C > G (p.(Leu446Val)). This LDLR variant has been found in ExAC with a MAF of 0.0008% and in EVS with a MAF of 0.0077% and was never reported as causative of FH. After performing in vitro assays to test the variant effects through the evaluation of the protein expression, the LDL binding and the LDL uptake, we concluded that the LDLR variant c.1336 C > G (p.(Leu446Val)) is not causative of FH ( Supplementary Fig. S3).
The patients bearing PCSK9 rare variants were also screened for APOB variants in order to exclude the presence of additional variants causing FH and no rare variants were found. For the 4 PCSK9 rare variants never associated to FH, we performed the in silico predictions of pathogenicity (Supplementary Table S1) and we characterize the function of variant protein by several approaches.  Fig. 1, secretion of all the variants was similar than wt PCSK9 secretion. Relative values of PCSK9 determined by ELISA to total cellular protein showed no differences among expression of the different PCSK9 variants compared to wt PCSK9. The secreted protein in the transfected cells expressed as ng PCSK9/total PCSK9 were: wt 16 (FITC-LDL) uptake by the cells was measured as described in materials and methods. As shown in Fig. 2A, LDL uptake was significantly reduced upon expression of p.(Arg357Cys) and p.(Ser636Arg) PCSK9 variants (≈25%), compared to wt PCSK9. The p.(Asp374Tyr) variant (used as a GOF control) was the most potent in reducing LDL uptake in those cells (≈50%) while p.(Pro331Ala) and p.(His643Arg) variants showed similar activities than wt PCSK9 ( Fig. 2A). This approach investigates both intracellular and extracellular effects of PCSK9 variants. In order to deeply investigate the action mechanism of PCSK9 variants and to test only their extracellular action, the LDLR assay was performed incubating cells in a culture medium and adding the recombinant purified PCSK9 variants exogenously. For that purpose, HEK293 and HepG2 cells (to test the PCSK9 behaviours in a hepatic cell line), Fig. 2B and C, respectively, were treated with 5 μg/mL PCSK9 variants. Then, cells were incubated with 20 μg/mL FITC-LDL to determine the extent of LDL uptake. As shown in Fig

p.(Arg357Cys) and p.(Ser636Arg) PCSK9 variants show higher affinity for LDLR than wt PCSK9.
Next, we tested binding affinities of wt PCSK9, p.(Pro331Ala), p.(Arg357Cys), p.(Ser636Arg) and p.(His643Arg) PCSK9 variants for LDLR at both pH 7.2 and pH 5.5 using a solid phase binding immunoassay. Figure 3 shows the LDLR-PCSK9 binding curves obtained with purified LDLR ectodomain incubated at a fixed concentration with serial dilutions of each PCSK9 variant. The EC 50 values calculated from these binding curves are shown in Table 2, indicating that at pH 7.2 the EC 50 for wt LDLR is 112.2 nM, very similar to the previously reported values 15 . Affinity values of p.(Pro331Ala) variant to LDLR were similar to wt PCSK9, whereas for the p.(His643Arg) variant, only the value at pH 7.2 is similar to the wt, being its EC 50 at pH 5.5 higher than the wt (44.2 nM vs. 23.2 nM, respectively) ( Table 2).
However, EC 50 values for wt LDLR of p.(Arg357Cys) and p.(Ser636Arg) variants were 50.9 and 47.4 nM, respectively, thus showing a higher affinity for LDLR and confirming the GOF activity. In addition and as internal control for method validation, EC 50 value of p.(Asp374Tyr) GOF was determined, and as expected, this variant shows a higher affinity to LDLR compared to wt PCSK9 (19.3 nM vs. 112.2 nM, respectively) ( Table 2). EC 50 values for wt LDLR of the different variants was also determined at acid pH (5.5) but no differences among them were found, being the affinities of p.(Arg357Cys) and p.(Ser636Arg) variants very similar to p.(Asp374Tyr) GOF variant, and higher than those of wt, p.(Pro331Ala), and p.(His643Arg) PCSK9 variants ( Table 2).

Discussion
Variants in the PCSK9 gene are responsible for about 1% of FH cases, whereas the LDLR variants account for most of cases 1 . The action mechanism of PCSK9 is at the basis of the dual effect of its variants, LOF variants causing hypocholesterolemia and GOF variants causing FH. The rare variants in the PCSK9 gene should be correctly evaluated before claiming their role as LOF or GOF. According to recent guidelines, a strict pathogenic classification is needed to correctly define the role of variants identified during re-sequencing studies. Functional studies are qualified as "strong" criteria for assessing variant pathogenicity 11 .
In this study, 7 rare variants in the PCSK9 gene (2.6% of total examined population) were identified, 4 of which had never been previously described in FH patients. The prevalence of FH-causing variants in PCSK9 was 1.8% when considering only the GOF variants, i.e. the 3 previously identified variants [16][17][18][19][20][21] together with the 2 variants identified in this study. The prevalence of PCSK9 variants causative of FH varies among different countries being very low in the Dutch and English populations (0.1-2%) 22 and higher in the French population, reaching 5% in the study of Abifadel et al. 16 . However, in some studies only the most frequent PCSK9 variant p.(Asp374Tyr) was searched 23 or only rare variants already annotated as pathogenic are reported 24 , partially justifying the very low frequency of PCSK9 rare variants.
In order to test the effect of PCSK9 rare variants on LDLR, functional characterization was performed by assessing protein secretion, LDLR activity in the presence of PCSK9 variant proteins as well as LDLR affinity of PCSK9 variants. Although the amounts of secreted PCSK9 appeared to be similar among the studied variants and the wt protein, we observed different behaviours of the LDLR regulation: 2 variants were GOF and 2 variants did As for the 3 variants previously described in other studies on FH patients, the p.(Asp35Tyr) variant was identified in a French patient by Abifadel et al. who reported the variant as being responsible for a novel Tyr-sulfation site creation, which may enhance the intracellular activity of PCSK9 16 . The p.(Ser465Leu)was previously reported by our group as a variant associated with an extremely variable FH phenotype 17 . A subsequent study reported the presence of this variant in 10 patients from the Netherlands with a mild hypercholesterolemia 18 . The variant p.(Arg469Trp) was firstly identified in a double heterozygote patient also carrying a null variant in the LDLR gene 19 . The variant was also described by Kotowski et al. in a multi-ethnic population from the United States: black carrier subjects showed a variable lipid profile ranging from low to high LDL cholesterol, whereas the only white carrier showed increased LDL levels 21 . In the next generation sequencing databases ExAC and gnomAD, the variant p.(Arg469Trp) was reported at higher frequency in African populations respect to other populations, whereas in the EVS and 1 kG the variant was exclusively found in African derived population. The presence of other in linkage variants might be responsible for the different phenotypes observed in the two ethnic groups. The functional characterization performed by Geschwindner et al. 20 with the PCSK9 variant purified from insect cells and incubated on HepG2 cells revealed a very mild LOF effect, although this approach might be influenced by the use of a non-human protein-production system.
In conclusion, in our study 4 PCSK9 variants with an unknown effect on FH pathogenesis were identified. After extensive functional evaluation, we demonstrated a GOF effect of the p.(Arg357Cys) and p.(Ser636Arg) variants that could be considered as FH causative variants.

Methods
Patients and genetic screening. Based on biochemical and clinical features, 269 patients were suspected of FH and enrolled in this study. Simon Broome criteria were followed for patient inclusion with the exception of  Table 2. EC 50 values for the binding of PCSK9 variants to LDLR, as determined by solid-phase immunoassay at pH 7.2 and pH 5.5. * Data are reported as mean ± standard deviation.
paediatric patients that were also included if LDL cholesterol levels were higher than 90 th percentile 25 and a clear hypercholesterolemia was found in a parent.
The study was performed according to the current version of the Helsinki Declaration and was approved by the Ethical Committee of the Università degli Studi di Napoli Federico II (Number 157/13, September 9, 2013). Informed consent was obtained for each patient. All patients underwent the genetic screening of the LDLR gene by amplification of promoter, exons and exon-intron junctions followed by direct sequencing as previously described 26 . If no pathogenic variants were detected, Multiplex Ligation-dependent Probe Amplification (MLPA) was performed as previously reported 26 to search for large rearrangements in the LDLR gene. Molecular analysis of the PCSK9 gene was performed in patients without pathogenic variants in the LDLR gene. PCSK9 screening included the amplification and direct sequencing of promoter, exons and exon-intron junctions 27 . APOB screening included the analysis of the region coding for the LDLR binding region, i.e. a portion of the exon 26 (c.9670-c.11916, p.Lys3181-p.Asn3929) and the exon 29 27 .
Identified variants were checked against pathogenic variants databases: Leiden Open Variation Database (LOVD) and Human Gene Mutation Database (HGMD). Variants not present in these databases or never reported as causative of FH in literature were searched in next generation sequencing databases: Exome Aggregation Consortium (ExAC), Genome Aggregation Database (gnomAD -it consists in ExAC data integrated with additional exome and genome data); Exome Variant Server (EVS) and 1000 genomes. Rare variants are defined as variants with a Minor Allele Frequency (MAF) less than 1%.
LDLR-ectodomain production and purification. The LDLR construct encoding the N-terminal extracellular ectodomain (1-789 amino acids) plus c-myc and His tags was purified by affinity chromatography from cells transfected with the pcDNA3.1-EC-LDLR-His plasmid, kindly provided by Prof. Leren 28 . Briefly, HEK293 cells at 70-80% confluency were transfected with the plasmid by calcium phosphate method for 24-48 h and selected in successive passages by geneticin (G-418 sulphate, Gibco, Invitrogen). For EC-LDLR expression and purification, the growing medium of positively transfected cells was changed to Opti-MEM (Invitrogen) without geneticin and maintained under these conditions for other three days. Then the medium was harvested, supplemented with protease inhibitors (complete EDTA-free, Roche) and the LDLR ectodomain was affinity purified using one-step nickel affinity chromatography. For protein long-term maintenance, the buffer was changed to storage buffer (50 mM Tris-HCl, 50 mM NaCl, 10% glycerol, and 0.01% Brij-35, pH 7.5) 29  PCSK9 purification from stably transfected HEK293 cells. HEK293 cells grown to subconfluence were transfected with the different PCSK9 plasmids and subcultured with geneticin G418 sulphate (Gibco) according to the manufacturer's instructions to obtain the stably transfected cells. For PCSK9 purification, stably transfected HEK293 cells were grown at 80% confluence in DMEM medium, then, culture medium was replaced by Opti-MEM (Invitrogen) without geneticin and cells were maintained under these conditions for 48 h. Finally, the medium was harvested and PCSK9 was purified using one-step nickel affinity chromatography. Purified PCSK9 variants were stored at −80 °C in 50 mM Tris-HCl buffer supplemented with 150 mM NaCl and 10% glycerol, pH 8.0.
Lipoprotein labelling with FITC. LDL was labelled with FITC as previously described 31 . Briefly, 10 µL of FITC (2 mg/mL) were added to 1 mL LDL (1 mg/mL) in 0.1 M NaHCO 3 , pH 9.0, was mixed for 2 h by slow rocking at room temperature. The unreacted dye was removed by gel filtration on a sephadex G-25 column equilibrated with PBS EDTA-free buffer. All fractions were assayed for protein content using bovine serum albumin as standard (Pierce BCA protein assay, Pierce). Quantification of LDL uptake by flow cytometry. 48 h after transfection with the plasmids containing the different PCSK9 variants, HEK293 cells were incubated for 4 h, at 37 °C with 20 µg/mL FITC-LDL and lipoprotein uptake was determined as previously described 31 . In addition, LDL uptake was determined using purified PCSK9 variants in HepG2 cells and HEK293 cells. Briefly, 2 µg/mL of each purified PCSK9 variant was added to the cell culture medium and 2 h post-addition, 20 µg/mL FITC-LDL was added to the medium and LDL uptake was determined 4 h after lipoprotein addition. In both experimental approaches, after incubation with FITC-LDL, cells were washed twice in PBS-1%BSA, fixed on 4% formaldehyde for 10 min and washed again twice with PBS-1%BSA. The amount of internalized LDL was determined as described before by adding Trypan blue solution (Sigma-Aldrich, Steinheim, Germany) to a final concentration of 0.2% .
Fluorescence intensities were measured in a FACSCalibur ™ (BD Bioscience, NJ, USA) flow cytometer as previously described 31 . For each sample, fluorescence of 10,000 events was acquired for data analysis. All measurements have been performed at least in triplicate.
Solid-phase immunoassay for PCSK9-LDLR ectodomain binding. Purified LDLR ectodomain diluted in working buffer (10 mM Tris-HCl, pH 7.5, 50 mM NaCl, 2 mM CaCl 2 ) was coated at a fixed concentration onto 96-well microtiter plates by incubation ON at 4 °C. Plates were then blocked and incubated with a serial dilution of each sample diluted in working buffer during 2 hours at RT, and then washed thoroughly with working buffer supplemented with 0.1% (w/v) Tween 20 (Sigma-Aldrich, MO, USA). For ligand detection, the antibodies (mouse monoclonal anti-DDK, clone OTI4C5, Origene, USA; and peroxidase-conjugated horse anti-mouse, Cell Signaling, USA) were diluted in working buffer supplemented with 5% (w/v) BSA, applied directly to the plate and incubated for 1 hour at RT, with an extensive washing between both incubations. After a final wash, antibody binding was determined using 50 μL per well of 2,2´-Azino-bis (3-ethylbenzothiazoline-6-sulfonic acid) substrate solution (Sigma-Aldrich, MO, USA) and measuring colour change at 405 nm. The time course for colour development was essentially linear and measurements were taken 30-60 min after the addition of substrate. For data processing, all absorbance values were corrected for unspecific binding, relativized to maximum and EC 50 values were extracted from curves after fitting the data to 5-parameter logistic (5-PL) equation (SigmaPlot 13.0, Systat Software Inc., CA, USA).