Figure 1.
Workflow of this study to determine the functional impact of 70 rare missense variants on LDLR protein activities and improve rare variant association testing for plasma LDL-C and the risk for early-onset MI.
Variants were identified through whole-exome sequencing of 3,235 individuals from the Italian Study of Early-onset Myocardial Infarction (ATVB) cohort.
Table 1.
Association of a burden of rare variants in the low-density lipoprotein receptor (LDLR) gene with plasma low-density lipoprotein cholesterol (LDL-C) levels and the risk for early-onset myocardial infarction.
Figure 2.
Systematic functional profiling of low-density lipoprotein receptor (LDLR) alleles.
(A) LDLR missense variants were functionally characterized by monitoring cellular uptake of fluorescently-labeled LDL (DiI-LDL; red) into cells (see Methods). Shown are automatically acquired images of HeLa-Kyoto cells transiently expressing siRNA-resistant full-length human wildtype LDLR linked to EGFP (LDLR’-GFP), empty GFP-control plasmid, or two FH mutants known to inhibit transport (p.G549D; FH class-2) or endocytosis (p.Y828C; FH class-4) of the LDLR protein. Arrows denote GFP-positive cells. Note the localization of FH mutants to different subcellular compartments. Bars = 15μm. (B) Graphs depict relative signal intensities of total DiI-LDL in endosome-like subcellular compartments (total LDL signal; y-axis, in arbitrary units) plotted against total cellular GFP expression (x-axis, in arbitrary units) for wildtype LDLR (LDLR’-GFP, upper panel) and indicated FH mutants. Each graph depicts results from a single experimental replica upon either overexpression of the respective cDNA-construct (left graphs) or complementation settings (i.e., siRNA knockdown of endogenous LDLR followed by reconstitution with indicated LDLR-GFP constructs; right graphs). Each dot represents one individual cell. Dashed vertical bars separate cells classified as GFP-negative (left from bar) from cells defined as GFP-positive. Dashed horizontal lines in complementation setting indicate mean total LDL signal in control siRNA-treated cells expressing endogenous LDLR. Cells where total LDL signal fell above this threshold (indicating over-compensation by LDLR’-GFP expression) were not respected for quantifications (see Methods). (C) LDLR activity was measured with five phenotypic parameters: total LDL signal in endosome-like compartments, LDL concentration, number (seg. number) and area (seg. area) of subcellular DiI-positive endosome-like structures, and cellular GFP-expression. The heatmap represents means from all experimental replicas per variant under the overexpression setting. Red reflects reduced, blue increased signal relative to wildtype LDLR’-GFP. Phenotypes meeting statistical criteria as described in Methods are framed in orange. Bar graph depicts total LDL signal ±SD normalized to wildtype LDLR’-GFP.
Figure 3.
Cell-based functional profiling distinguishes disruptive from non-disruptive rare missense variants in the low-density lipoprotein receptor (LDLR) gene as identified through exome sequencing of 3,325 individuals.
(A) Rare LDLR missense variants from exome sequencing of the ATVB cohort were individually introduced into LDLR’-GFP, transiently expressed in HeLa-cells, and the impact on cellular uptake of fluorescently-labeled LDL (DiI-LDL) was quantified for the indicated four parameters and GFP-expression. Shown are means from 3–4 independent experiments per variant relative to wild-type LDLR’-GFP. Phenotypes (red, reducing; blue, increasing) meeting statistical criteria (p<0.05; deviation >1) are framed in orange. Variants that significantly reduced LDL-uptake in ≥3 DiI-LDL parameters, including total LDL signal, were classified as “disruptive-missense” (for details, see Methods). OSD, O-linked sugars domain (exon15); TMD, transmembrane domain (exons16–17); CT, carboxy-terminus (exon18). (B) Comparison of mean total DiI-LDL signal intensities within endosome-like intracellular compartments (“total LDL signal”) between the overexpression setting (ΔLDLR’-GFP only) and a complementation setting (siRNA against endogenous LDLR together with ΔLDLR’-GFP) relative to wildtype LDLR’-GFP (wt, black circle) and GFP control (GFP, open circle). Variants classified as “non-disruptive” failed to reach significance in any parameter under neither experimental setting. (C) For 14 variants classified as disruptive-missense, impact upon overexpression on cellular levels of free cholesterol (FC) was determined. Shown are means±s.d. relative to wildtype LDLR’-GFP from 2–4 independent experiments. (D) Determination of subcellular localization of LDLR’-GFP disruptive-missense variants. Shown are maximal projections of confocal stacks of representative cells transiently transfected with indicated mutants. Bar graphs reflect ΔLDLR’-GFP levels on Western Blots (shown in S5 Fig.; means from 2 experiments) of endoplasmic reticulum (ER; 120kDa) relative to post-ER (160kDa) form of the LDLR protein relative to total wildtype LDLR as determined by ratiometric measurements. For each mutant, contribution of ER- relative to total LDLR’-GFP protein are indicated in percent.
Figure 4.
Functions and distribution of LDLR rare missense alleles identified through exome sequencing of 3,235 individuals.
(A) Plasma LDL-C (in mg/dl) in LDLR missense allele carriers (dots) from the ATVB cohort according to functional category (for classification, see Methods). LoF, loss-of-function. Means are indicated by horizontal bars. p-value was determined by 2-sided, 2-tailed Student’s t-test. (B, C) Individual LDLR missense variants identified through exome sequencing of indicated number of individuals are depicted according to genomic position starting at the 5’end (top). The numbers next to each variant represent the number of times the respective variant was observed in cases and controls, respectively, with regard to plasma LDL-C levels (b) and early-onset myocardial infarction (MI; c). Colors in circles represent indicated functional classes as determined either by an overlap of four bioinformatic prediction tools (PolyPhen-2, SIFT, MutationAssessor and MutationTaster; see Methods) (“prediction”) or cell-based experimental studies of LDL-uptake. Variants in bold have been observed in both, cases and controls. (D) Power calculations for the number of sequenced individuals needed to reach exome-wide significance (p<2.5×10-6, reflected by power = 1) for association with MI-risk when the indicated classes of rare LDLR alleles are taken into account. For details, see Methods.
Figure 5.
Impact of individual LDLR missense variants on cellular LDL-uptake correlates with single-variant association results for plasma LDL-C in ~40,000 individuals.
(A, B) Comparison of results from cell-based functional profiling of LDL-uptake (overexpression and complementation) and single variant association tests for 16 LDLR missense variants represented on the exome-chip and identified by genotyping 39,186 individuals from multiple cohorts. Shown are means from 3–4 independent experiments per variant relative to wild-type LDLR’-GFP. Phenotypes (red, reducing; blue, increasing) meeting statistical criteria as described in Methods are framed in orange. Effect size (beta) is provided as the difference of means of plasma LDL-C levels between variant carriers and non-carriers (in mg/dl).