Genetic spectrum of familial hypercholesterolemia and correlations with clinical expression: Implications for diagnosis improvement

Abstract Familial hypercholesterolemia (FH) is the most common genetic disease caused by variants in LDLR, APOB, PCSK9 genes; it is characterized by high levels of LDL‐cholesterol and premature cardiovascular disease. We aim to perform a retrospective analysis of a genetically screened population (528 unrelated patients—342 adults and 186 children) to evaluate the biochemical and clinical correlations with the different genetic statuses. Genetic screening was performed by traditional sequencing and some patients were re‐analyzed by next‐generation‐sequencing. Pathogenic variants, mainly missense in the LDLR gene, were identified in 402/528 patients (76.1%), including 4 homozygotes, 17 compound heterozygotes and 1 double heterozygotes. A gradual increase of LDL‐cholesterol was observed from patients without pathogenic variants to patients with a defective variant, to patients with a null variant and to patients with two variants. Six variants accounted for 51% of patients; a large variability of LDL‐cholesterol was observed among patients carrying the same variant. The frequency of pathogenic variants gradually increased from unlikely FH to definite FH, according to the Dutch Lipid Clinic Network criteria. Genetic diagnosis can help prognostic evaluation of FH patients, discriminating between the different genetic statuses or variant types. Clinical suspicion of FH should be considered even if few symptoms are present or if LDL‐cholesterol is only mildly increased.

In addition to HeFH also a homozygous FH (HoFH) form has been described; this latter is caused by the presence of pathogenic variants on two alleles of candidate genes (homozygosis, compound heterozygosis or double heterozygosis) and recent data suggest a prevalence of $1:300.000. 7 The HoFH form is more severe than Previous reports about large Italian FH population have been published. The most recent one analyzed the genetic architecture of FH in patients collected by lipid clinics across the whole nation but did not performed a genotype-phenotype correlation. 9 This aspect was investigated by another Italian study revealing differences between LDLR variants leading to receptor defective or receptor negative proteins. 10 However, these studies did not perform a separate analysis and a comparison between adults and children.
The detection of a pathogenic variant in one of candidate genes is a cardiovascular risk factor independently from the LDL-c levels. 11 The variant identified in a patient can be searched in the patient relatives (cascade screening) improving the identification and early treatment of additional FH patients. 12 However, in about 20%-30% of patients with a clinical suspicion of FH no pathogenic variants are identified 2 ; a few such patients could be carriers of pathogenic variants which caused other rare diseases whose phenotype overlaps with FH such as Sitosterolaemia (MIM: 210250 and 618 666) or Lysosomal Acid Lipase Deficiency (MIM: 278000) (FH phenocopies). 4,13 A polygenic base has been hypothesized considering the accumulation of common small-effect LDL-c-raising alleles, 14 although the diagnostic power of the calculated risk score are not useful for FH diagnosis. 15 Here we report the genetic spectrum emerging from the retrospective analysis of an Italian population genetically screened in the last 11 years highlighting the complexity of FH genetics. We also ana-

| Clinical assessment and biochemical evaluation
Patients were questioned about personal and family history of hypercholesterolemia and cardiovascular diseases; presence of tendon xanthomas, corneal arcus and carotid plaque were also verified. Tendon xanthomas were considered present if at the inspection and palpation of Achilles tendons, tendon at the dorsum of hands, elbows and knees a diffuse enlargement or nodules are present. Other collected data are: smoking habits, presence of other diseases (such as hypertension, diabetes, thyroid dysfunction). Body mass index (BMI) was calculated as weight (kg)/height 2 (m 2 ).
Total cholesterol, HDL-cholesterol (HDL-c) and triglycerides reported in this study have been evaluated in absence of lipidlowering therapy and were measured by standard enzymatic methods, whereas LDL-c was calculated by the Friedewald formula. In case of patients on therapy at the first observation, the pre-therapy LDL-c was calculated according to a formula previously. 16 The non-HDLcholesterol (non-HDL-c) and the LDL/HDL ratio were also calculated.
Clinical diagnostic criteria for FH diagnosis were applied: Dutch Lipid Clinic Network (DLCN) Score was calculated for adult patients and interpreted according to Nordestgaard et al. 2 ; Simon Broome diagnostic criteria were considered for all patients. 8 Because some clinical and familial information were difficult to retrieve for several patients, 43 patients were classified as "Unlikely FH" and 267 as "No FH" according to DLCN and Simon Broome, respectively.

| Genetic screening
Genetic screening included the sequence analysis of all LDLR exons together with the intron-exon junctions as previously reported. 17 If no pathogenic variants were detected, large rearrangements in LDLR were searched by Multiplex Ligation-dependent Probe Amplification.
When no pathogenic variants in LDLR were detected, the screening of PCSK9 was performed by sequence analysis of all exons together with the intron-exon junctions, while APOB sequence analysis was limited to the region coding for the LDLR binding region (a portion of exon 26 and the whole exon 29 with the intron-exon junctions as described in Rubba et al. 18 ). Finally, the screening was extended to all LDLRAP1 exons together with the intron-exon junctions, if no pathogenic variants in LDLR, APOB and PCSK9 genes were found. For homozygotes/ compound heterozygotes, the variant presence was ascertained in both parents, allowing to confirm that the two variants were present on the two different alleles. This was not performed for the double heterozygote, but as the two variants are present in two different genes on two different chromosomes, it was unnecessary to perform further analyses to establish the patient genotype.  Missense, deletion/insertion without frameshift and promoter variants in the LDLR gene were considered defective variants, whereas nonsense, splicing, deletion/insertion leading to frameshift and large rearrangements were defined null variants as reported in ACMG guidelines. 19 Missense variants in APOB and PCSK9 genes were considered defective because the protein alteration did not lead to a complete loss of LDL-LDLR binding and uptake.   A report of the frequencies of the different genetic statuses is reported in Figure 1, whereas a plot of all different variants found at heterozygous status is reported in Figure S1. As a pathogenicity criterion is based on the previous identification of variants in additional FH patients, we report the VUS in Table S3 in order to facilitate future studies on FH genetics.

| Statistical analysis
We further sequenced a subgroup of 49 patients (38 adults and 11 children) with different genetic status: 29 without pathogenic variants, 17 HeFH for variants in LDLR, 1 HeFH for a single variant in LDLRAP1 gene, 2 compound heterozygotes for LDLR variants and 1 double heterozygote for variants in LDLR and PCSK9 genes. All previously identified variants were confirmed and no additional rare variants (pathogenic, VUS or benign) were identified in these patients and consequently no changes in the previous genetic diagnosis were present.

| Genotype-phenotype correlation analysis
We compared the untreated LDL-c levels of patients with different genetic statuses, further distinguishing HeFH in patients with a defective or a null variant. The violin plot reported in Figure 2 shows the gradual increase of LDL-c observed from patients without pathogenic variants to patients with a defective variant, to patients with a null variant and to HoFH patients. This plot also highlights the great variability of LDL-c levels in each group. As these groups contains both pediatric and adult patients, we repeated the analysis further dividing patients for age ( Figure S3). By this analysis, we observed no differences of LDL-c between children carrying a defective or a null variant or between HeFH and HoFH. Only the differences between patients without a pathogenic variant and the other three groups were statistically significant.
Analyzing only the heterozygous patients with a missense variant, we verified that patients with the variant in LDLR gene (n = 223) showed a worse phenotype than patients with a variant in

| Analysis of DLCN and Simon Broome criteria
We calculated the DLCN score for adult patients and performed the clinical diagnosis accordingly. As reported in Figure 5A,     On the other hand, the extended panels, including genes leading to FH phenocopies, would identify a high number of rare variants in other lipid-associated genes. Unfortunately, to date the pathogenicity evidences of variants in FH-phenocopy genes are too few to easily establish their role in FH development. 38 An additional confounding factor in FH patient identification is