Filling the gap: Genetic risk assessment in hypercholesterolemia using LDL-C and LPA genetic scores

Routine genetic testing in hypercholesterolemia patients reveals a causative monogenic variant in less than 50% of affected individuals. Incomplete genetic characterization is partly due to polygenic factors influencing low-density-lipoprotein-cholesterol (LDL-C). Additionally, functional variants in the LPA gene affect lipoprotein(a)-associated cholesterol concentrations but are difficult to determine due to the complex structure of the LPA gene. In this study we examined whether complementing standard sequencing with the analysis of genetic scores associated with LDL-C and Lp(a) concentrations improves the diagnostic output in hypercholesterolemia patients. 1.020 individuals including 252 clinically diagnosed hypercholesterolemia patients from the FH Register Austria were analyzed by massive-parallel-sequencing of candidate genes combined with array genotyping, identifying nine novel variants in LDLR . For each individual, validated genetic scores associated with elevated LDL-C and Lp(a) were calculated based on imputed genotypes. Integrating these scores especially the score for Lp(a) increased the proportion of individuals with a clearly defined disease etiology to 68.8% compared to 46.6% in standard genetic testing. The study highlights the major role of Lp(a) in disease etiology in clinically diagnosed hypercholesterolemia patients, of which parts are misclassified. Screening for monogenic causes of hypercholesterolemia and genetic scores for LDL-C and Lp(a) permits more precise diagnosis, allowing individualized treatment.

frequent functional variants that independently have small effect sizes. 4 A variable proportion of measured LDL-C derives from lipoprotein(a) [Lp(a)] that, contains apolipoprotein B-100 [apoB100] attached via disulfide bond to the hydrophilic, highly glycosylated apolipoprotein(a) [apo(a)]. 5 The remaining cases are attributed to unspecified genetic and exogenous factors. 6 The most common monogenic cause for FH are variants in the low-density-lipoprotein receptor (LDLR) gene (80%-90%). To date, the variant database ClinVar lists over 2000 disease-causing LDLR variants.
Homozygosity causes a very severe HC with clinical manifestation already in childhood. 5%-10% of FH patients carry one of a limited number of APOB gene variants that interfere with binding of LDL-C particles to its receptor (European ancestry mostly p.Arg3527Gln). 7 1%-3% have gain-of-function variants in the PCSK9 gene (proprotein convertase subtilisin/kexin Type 9) which impairs LDL receptor recycling. 8 The APOE gene variant p.Leu167del is rare (Spain, France, NW Europe) and has a less severe phenotype than LDLR-related FH; this variant leads to increased LDL concentrations due to reduced ApoEmediated LDL catabolism. 9 Biallelic variants in the genes LDLRAP1, ABCG5, ABCG8 and LIPA can manifest with phenotypes similar to classic FH. 10,11 Depending on severity and type of genetic variant, individuals with monogenic HC have a 3-to 10-fold risk of premature CAD. 12 Genetic variability at the LPA locus is the main determinant of human Lp(a) serum concentrations, which range from <0.1 to >200 mg/dL. The apo(a) protein contains multiple repeats of triple-looped 80-90 amino acid structures called kringles. Differences in the molecular weight of apo(a) from 300 to 800 kDa are due to a variable number of kringle repeats. There is an inverse relationship between the number of repeats and allele-associated Lp(a) concentrations; a kringle KIV-2 repeat number below 23 is used to distinguish small and large apo(a) isoforms. 13 Differences in Lp(a) level in same-size KIV-2 alleles are partly due to variants in the LPA gene. Due to its repetitive nature, sequencing of the LPA gene is difficult 14 and therefore not usually included in routine genetic investigations. As a surrogate for full LPA analysis, a weighted genetic score composed of 43 single-nucleotide variants in the LPA gene region has been developed allowing prediction of Lp(a) concentrations. 15 Genotyping array data from UK Biobank participants confirmed a high validity of the score particularly in White/European individuals in whom it explained approximately 60% of the Lp(a) variation. 16 Cardiovascular risk associated with HC appears to be similar for monogenic and polygenic causes, but elevated blood Lp(a) concentrations represent an additional independent risk factor. [17][18][19][20] Lp(a) has cholesterol-independent pro-inflammatory and pro-atherosclerotic properties that may partly be due to binding of oxidized phospholipids. About 20% of the European general population have a Lp(a) value of ≥50 mg/dL which increases the risk for CAD by up to 200%. At Lp(a) blood concentration of >180 mg/dL, lifetime risk resembles that of untreated heterozygous FH. 21 Elevated Lp(a) is a risk factor even at very low LDL-C concentrations, and the European Atherosclerosis Society recommends testing Lp(a) concentration at least once in adults. 18 However, Lp(a) measurement is challenging, 22 and there is substantial variability in the results provided by commercially available immunoassays. 23 Population screening of LDL-C concentrations is universally accepted as useful for the early diagnosis of an increased CAD risk.
Genetic investigations can assist in the identification of affected individuals for risk stratification and individualized therapy even before the onset of symptoms. The molecular diagnosis of FH clarifies the risk for relatives and enables family-based cascade screening. Polygenic scores for LDL-C and Lp(a) can potentially assist in clarifying the molecular background particularly in FH variant-negative individuals but are not included in routine genetic work-up of individuals with HC.
Considering that genetic factors are major determinants of lifelong CAD risk, genetic screening promises to be useful for predictive analyses in asymptomatic individuals. In the present study, we investigated the usefulness of combined genetic analysis of monogenic HC as well as LDL-C and Lp(a) genetic scores in HC cases from the Austrian FH Register.

| Study cohort/FH Register Austria
The study cohort included 252 patients with HC from the FH Register Austria. This register exists since 2015 and includes a wide ranch of demographic, biochemical, lifestyle and genetic data (Supplementary   Tables S1-S3). Patients were assessed using the Dutch Lipid Clinic Network Score (DLCN), which is based on family and patient history of premature CAD, untreated LDL-C concentrations (>155, >190, >250 and >325 mg/dL) and presence of xanthomata, xantelasmata, or arci cornealis prior to the age of 45 years. The clinical diagnosis was defined as possible (score 3-5), probable (score 6-8) or definite (score > 8) FH. As part of the diagnostic work-up, samples were genetically analyzed with a genotyping array, the Global Screening Array (Illumina GSA) and complementary massive parallel sequencing to investigate (likely) pathogenic variants 24 in hypercholesterolemia genes LDLR, APOB, PCSK9, LDLRAP1, ABCG5, ABCG8 and LIPA. Seven hundred and sixty eight fully pseudonymized adult Europeans, without known FH diagnosis, were used as population controls and analyzed by GSA genotyping (for details see Supplementary Materials S1).

| Biochemical characteristics of lipids and lipoproteins
All HC patients were clinically assessed with the DLCN in the University Hospital for Internal Medicine I at the Medical University of Innsbruck. Laboratory investigations included measuring blood concentrations of LDL-C, HDL cholesterol, total cholesterol, triglycerides and Lp(a). Baseline LDL-C levels were defined as those prior to cholesterol-lowering therapy. When native lipid levels were missing, LDL-C levels were therapy-adjusted as previously described. 25 For Lp(a) we used the first therapy-naïve value for analysis. LDL-C was considered elevated at ≥190 mg/dL, and the Lp(a) cut-off was set at ≥50 mg/dL as previously described. 17 2.3 | Genotype imputation, quality control, and genetic scores After genotyping with standard methods 26 quality control was performed before imputation. Principal component analysis using LASER Server 27 was done to estimate the genetic ancestry of all 1.020 individuals and matched to population samples. Filtered genotypes were imputed with the Michigan Imputation Server 28 using the Haplotype Reference Consortium panel (HRC r1.1 2016). 29 The imputation quality was r 2 > 0.6. We applied the published validated polygenic score for LDL-C (LDL GS) (http://www.pgscatalog.org/, PGS Catalog ID PGS000115) and the locus-based LPA genetic score (LPA GS) (PGS Catalog ID PGS000667) 15,30-32 (for details see Supplementary Materials S1). Statistical significance was set to p < 0.05.

| Ethics declaration
The data used in the study were generated in the standard diagnostic process with full informed consent from the investigated individuals.

| Demographic data
Of the 252 clinically suspected FH patients, 18 were excluded because they either carried variants of uncertain significance or homozygous LDLR variants, register data was missing, or they were family members from cascade screening that were FH variantnegative. In total, 234 patient samples were included in the analysis.

| Novel variants in LDLR, APOB and LDLRAP1
We identified nine novel variants in the LDLR, APOB and LDLRAP1 genes not previously reported in any disease-specific database: two LDLR splice-donor-site variants, one LDLR cryptic splice-acceptor variant, one LDLR frameshift variant, one LDLR missense variant, two large LDLR deletions, an APOB missense variant, and a homozygous LDLRAP1 missense variant. Applying ACMG guidelines, these variants were classified as variants of uncertain significance, likely pathogenic or pathogenic. All variants were submitted to ClinVar (Supplementary Table S2). and PCSK9 (n = 1). We also identified one individual with a homozygous pathogenic LDLRAP1 variant that phenotypically mimics heterozygous FH and was therefore included in the analysis (Table S3). One     17 Our results show that the available LPA genetic score can be reliably used to assess the probability of increased Lp(a) concentrations.

| Monogenic hypercholesterolemia is primarily cause by LDLR variants
A low LPA genetic score in an individual with HC suggests that Lp(a) cholesterol did not contribute substantially to overall cholesterol concentrations. To investigate this further, and to adjust for the cholesterol fraction of the Lp(a) particle, 34 we subtracted one third of the Lp(a) concentrations (a conservative estimate) from total LDL-C in the groups of HC patients that are variant-negative with an unclear disease origin, all FH variant-positives, and FH variant-negatives with a high LPA genetic score. The difference between total Lp(a)-adjusted LDL-C and total LDL-C was only significant ( p = 0.002) in FH variantnegative individuals with high LPA genetic scores (see Figure 4). This calculation showed that the high LDL-C that is used for HC diagnosis is strongly influenced by the high Lp(a) levels in LPA genetic scorepositive individuals.

| Differentiating genetic risk groups for HC
In the studied HC cohort, 46.6% of individuals carried a monogenic vari-

| Genetic risk in the general population
Up to one in 220 individuals of the general population is estimated to be a carrier of a FH-causing variant. 2 When applying the genetic score cut-offs to the population controls, 24.3% carried a high genetic score compared to 31.2% with an elevated genetic score in our HC cohort.
In the population control, a small number of 21 individuals (2.7%) had an elevated LDL score and a substantial fraction of 166 individuals (21.6%) had an elevated LPA score; out of these, three individuals (0.4%) had both genetic scores above the 95th percentile (see

| Monogenic HC
The studied cohort was extracted from the database of the Austrian  Figure 1B). This confirms that the CAD risk caused by elevated Lp(a) concentrations follows a distinct pathomechanism from LDL-C. 36 Current risk models for cardiovascular disease primarily rely on clinical data, but evidence that genome-wide variation contributes to this risk is widely investigated. 3,15,30,31,[37][38][39][40] In accordance with the literature, 41  to the Spanish group they found that the 6-SNV score explained only about 3% of the baseline LDL-C variance. 42,43 In a Canadian study, a high LDL-C score based on 28 SNVs (cut-off ≥80th percentile) was associated with only a relatively small, non-significant increase in the risk of premature cardiovascular events in monogenic variant-negative HC individuals, but the highest cardiovascular risk was observed in individuals with both, a monogenic FH and a high LDL-C score. 38 The same group developed the 223-SNV score used in our study and set a higher cut off at the 95th percentile to define polygenic HC; using this approach, both monogenic and polygenic hypercholesterolemia were associated with an increased risk for cardiovascular events compared to the general population. In their data set (n = 455 191) the maximum variance of LDL-C levels explained by the 223-SNV genetic score was 10.19%. 31 Therefore, we only included individuals with the very highest LDL scores above the 95th percentile (n = 11) into our group of polygenic HC.
Our results suggest that both a LDL-C polygenic score and a LPA single-gene score may explain elevated high LDL-C in a variable fash- ion. An LDL score above the 95th percentile can serve as a surrogate marker for polygenic HC in monogenic variant-negative individuals.
Notably, we did not identify any individuals with LDL genetic scores below the 5th percentile in monogenic variant-negative HC individuals, which might be indicative of a protective effect (Figure 2). While both high LDL-C and Lp(a) put patients at high cardiovascular risk, their disease etiologies follow distinct pathogenetic mechanisms, 19 and patients with high Lp(a) are frequently misdiagnosed as phenotypic FH. 20 Lp(a) is primarily determined by LPA gene variability, 44 which is difficult to assess in a routine genetic diagnostic setting but may be estimated using the LPA score as a surrogate marker. In our cohort, we confirmed that the LPA genetic score is a reliable predictor of high biochemical Lp(a) (Figure 3). Furthermore, our results confirm that the cholesterol fraction of Lp(a) contributes significantly to high LDL-C concentrations particularly in the group with high LPA genetic scores ( Figure 4). We show that the higher the LPA genetic score the higher is the cholesterol that needs to be attributed to Lp(a). With a positive predictive value of 90%, the LPA genetic score can be used as a valuable surrogate for biochemical Lp(a) measurements in large genetic screening efforts and therefore complement monogenic testing within a single intervention. We also detected HC individuals with a combination of monogenic and multi-allelic predisposition (  47,48 When there is a combined risk of monogenic and polygenic origin, an individualized combination of treatment will be most beneficial. 49

| Study limitations
Overall, our results are robust and in coherence with earlier studies.
However, there are some limitations mostly due to cohort size. Our samples were retrieved from the FH Register Austria, involving a subgroup of patients from Austrian and therefore focus on a Middle-European population. To generalize our findings, further investigation with different ethnicities is needed. Part of the single nucleotide polymorphisms used to compute the genetic scores were not on the GSA and therefore needed an imputation step. However, we only used variants with a high imputation quality r 2 > 0.6. For large-scale screening, the GSA could be customized to include all known variants in the FH genes and all genetic variants included in the genetic scores. This way, the more expensive and data intense sequencing could be avoided.

| Conclusion
By implementing genetic scores associated with high LDL-C and Lp(a) levels, the presented genetic screening approach increases the number of HC patients in which a distinct disease etiology was identified from 46.6% to 68.8% (Table 1). The high proportion of individuals with elevated Lp(a) can be predicted using the very effective LPA genetic score with a PPV of 90%. Furthermore, the results show the clinical relevance of Lp(a) cholesterol in LDL-C measurement, and the incorrect classification of part of HC patients. In our study, 31.2% of individuals in the non-monogenic HC cohort but also 24.3% of the population controls were carriers of multi-allelic risks for hypercholesterolemia. As opportunistic and genetic newborn screening are implemented, the early detection of CAD risk by a routine array method can be easily integrated to provide risk estimates for one of the most common and actionable genetic conditions. However, before developing concepts for implementing genetic scores in HC clinics, it is necessary to ensure that potential benefits are robust, reproducible, equitable, and cost-effective.

ACKNOWLEDGMENTS
The authors thank the participants and individuals involved in the Austrian FH Register. We would like to thank the following individuals for their constructive input to the research and the manuscript: Christoph Binder, Stefan Coassin, and Johannes Schwaninger. This work has been supported by grants from the Austrian Heart Foundation and Novartis/Austria.

CONFLICT OF INTEREST STATEMENT
The authors declare no conflict of interest.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.