The Arg82Cys Polymorphism of the Protein Nepmucin Implies a Role in HDL Metabolism

Abstract Context Blood lipid levels are linked to the risk of cardiovascular disease and regulated by genetic factors. A low-frequency polymorphism Arg82Cys (rs72836561) in the membrane protein nepmucin, encoded by CD300LG, is associated with lower fasting concentration of high-density lipoprotein cholesterol (HDLc) and higher fasting triglycerides. However, whether the variant is linked to postprandial lipids and glycemic status remains elusive. Objective Here, we augment the genetic effect of Arg82Cys on fasting plasma concentrations of HDL subclasses, postprandial lipemia after a standardized high-fat meal, and glycemic status to further untangle its role in HDL metabolism. Methods We elucidated fasting associations with HDL subclasses in a population-based cohort study (Oxford BioBank, OBB), including 4522 healthy men and women. We investigated fasting and postprandial consequences on HDL metabolism in recall-by-genotype (RbG) studies (fasting: 20 carrier/20 noncarrier; postprandial: 7 carrier/17 noncarrier), and shed light on the synergistic interaction with glycemic status. Results A lower fasting plasma concentration of cholesterol in large HDL particles was found in healthy male carriers of the Cys82 polymorphism compared to noncarriers, both in the OBB (P = .004) and RbG studies (P = .005). In addition, the Cys82 polymorphism was associated with low fasting plasma concentrations of ApoA1 (P = .008) in the OBB cohort. On the contrary, we did not find differences in postprandial lipemia or 2-hour plasma glucose levels. Conclusion Taken together, our results indicate an association between the Arg82Cys variant and a lower concentration of HDL particles and HDLc, especially in larger HDL subclasses, suggesting a link between nepmucin and HDLc metabolism or maturation.

Dyslipidemia is a major risk factor for cardiovascular disease [1][2][3][4][5]. Higher blood levels of low-density lipoprotein cholesterol (LDLc) and triglycerides (TGs) are associated with an increased cardiovascular disease risk, whereas higher levels of high-density lipoprotein cholesterol (HDLc) are associated with a lower risk. The protective effect of HDLc has been primarily attributed to the beneficial role of HDLc in reverse cholesterol transport [6], whereby free cholesterol is transported via HDL particles from peripheral tissues to the liver, and cholesterol esters and TGs are exchanged between HDL and very low-density lipoprotein (VLDL), LDL, and other apolipoprotein B-containing particles [7,8]. HDL incorporates a range of particle sizes with different properties for transporting circulating cholesterol [9]. The main HDL categories based on diameter include very large (XL) and large (L) particles, also called HDL2 particles, and medium (M) and small (S) particles, also called HDL3 particles. Particle size may affect the cardioprotective potential of HDL, with several studies suggesting that the cardioprotective characteristics are harbored by the smaller HDL3 particles [10][11][12]. However, the literature remains controversial in this regard [13].
To shed light on the role of nepmucin in lipid homeostasis, we carried out population-based genetic association studies to investigate the nepmucin Arg82Cys polymorphism and fasting plasma HDL subclasses [27]. Furthermore, we performed recall-by-genotype (RbG) studies to detect additional subtle abnormalities in lipid metabolism that could not be detected in a fasting state alone, and studied synergistic interactions of Cys82 with glycemic status (study overview in Fig. 1).

Material and Methods
Study I-Nepmucin Arg82Cys Variant and Fasting Plasma High-density Lipoprotein Subclasses in 4522 British Men and Women HDL subclasses were measured by nuclear magnetic resonance (NMR)-spectroscopy in 4522 individuals from the Oxford BioBank (OBB) [28]. The OBB is a cohort of healthy, normoglycemic European ancestry men and women aged 30 to 50 years, randomly selected from the Oxfordshire area [29]. A high-throughput proton NMR platform containing approximately 230 metabolites has been performed on approximately 7100 OBB plasma samples [27,28]. The quantified HDL subclasses are defined as XL-HDL (average particle diameter 14.3 nm), L-HDL (12.1 nm), M-HDL (10.9 nm), and S-HDL (8.7 nm). The NMR platform has been used in multiple epidemiological and genetic studies [30,31], and the details of the method have been described elsewhere [27,32] Here, we performed a meta-analysis of the associations with fasting plasma HDL subclasses in 20 healthy male heterozygous carriers of the Cys82 variant and 20 matched noncarriers [33,34] and the baseline/fasting results from study III (7 heterozygous carriers of Cys82, and 17 noncarriers). The 20 Cys82 carriers and matched noncarriers were examined on 2 separate days as previously described [33,34], and biochemical profiling was performed the first examination day. The study protocol was approved by the Ethical Committee of Central Denmark Region (protocol No. 1-10-72-113-12) and undertaken in accordance with the principles of the Declaration of Helsinki II. The study participants gave their written informed consent before study participation.

Study III-A Recall-by-Genotype Study: Lipid Meal Challenge in 24 Danish Men
We recruited 24 middle-aged healthy Danish men by genotype from the population-based Health2006 cohort [35] (n = 18) and among staff at the University of Copenhagen (n = 6). Of the 24 participants, 7 were heterozygous carriers of the nepmucin Cys82 variant and 17 were noncarriers ( Table 1). The carriers and noncarriers were in silico matched by age, body mass index (BMI), fasting glucose, and glycated hemoglobin A 1c concentrations. The genotypes of the study participants were not blinded to the examining personnel. The study was approved by the regional ethical committee of Copenhagen (protocol No. H-1-12012-136) and conducted in accordance with the principles of the Helsinki Declaration II. Written informed consent was obtained from all study participants.
The participants underwent a 6-hour standardized lipidrich meal challenge after a minimum of 8 hours overnight fasting. The meal consisted of a lipid-rich soup of water, chicken, leek, butter, and cream (80 g of saturated fat) in a volume of 675 mL. Total energy content was 4459 kJ (1065 kcal). The fat energy percentage was 66%, carbohydrate 16%, and protein 18%. A total of 1 g of acetaminophen was added in the meal to estimate gastric emptying time [36]. All participants consumed the entire meal within a 15-minute period. The participants were instructed to avoid excessive alcohol intake, smoking, and physical activity 48 hours before the examination. Height, weight, waist and hip circumferences were measured in light indoor clothes. Waist circumference was measured halfway between the rib cage and the superior iliac spine. Hip circumference was measured at the level of the major femoral trochanter. Blood samples were taken every 30 minutes during the first 2 hours and subsequently every hour during the last 4 hours of the meal challenge.

Study IV-Interaction Between the Arg82Cys Polymorphism and Glycemic Status on Fasting Plasma High-density Lipoprotein Subclasses
The interaction between the Arg82Cys polymorphism and glycemic status on fasting plasma lipid levels was tested in 1330 Danish men and women from the ADDITION-PRO study [37], of whom 831 were normoglycemic (2-hour plasma glucose < 7.8 mmol/L, fasting blood glucose [FBG] < 6.1 mmol/L) and 499 hyperglycemic (2-hour plasma glucose > 7.8 mmol/L, FBG > 6.1 mmol/L). Fasting plasma lipids were measured using a high-throughput serum NMR platform [27]. The study protocol was approved by the ethical committee of the Central Denmark Region (No. 20000183) and undertaken in accordance with the principles of the Declaration of Helsinki II. All participants gave written informed consent [38].

Biochemical Measurements in Studies II and III
Plasma glucose in studies II and III was measured by a glucoseoxidase method (Vitros 5600, Ortho Clinical Diagnostics). Serum insulin and C-peptide were measured by electrochemistry luminescence-immunoassay (Roche Diagnostics GmbH). Insulin resistance was estimated by the Homeostasis Model Assessment of Insulin Resistance (HOMA-IR) [39]. ApoB48 and apoB100 were measured in EDTA-plasma samples using the enzyme-linked immunosorbent assay method (ELISA; Shibayagi Co Ltd and Immuno-Biological Laboratories Co Ltd, respectively). Double standards were measured for each run. Intra-assay and inter-assay variation ranged from 3.5% to 5.6% and 2.8% to 8.6%, respectively. Plasma levels of TGs, HDLc, and total cholesterol (TC) were measured on a Vitros 5.

Genotyping
The rs72836561 (Arg82Cys) polymorphism of CD300LG was genotyped on the Illumina HumanExome Beadchip 12v1_A and the genotypes were called using GenCall, Genotyping module (version 1.9.4), or GenomeStudio software (version 2011.1, Illumina). Samples were excluded if they showed a low call rate, abnormal mean heterozygosity, high singleton count, non-European ancestry, sex discrepancy, or duplicate discordance. Genetic variants were excluded if they showed a low call rate, deviation from Hardy-Weinberg equilibrium, duplication, chromosome or allele mismatch, GenTrain score less than 0.6, cluster separation score less than 0.4, or a deviation in manual cluster checks. Missing genotypes were subsequently re-called using zCall and a second round of quality control was performed to exclude poor quality samples and variants.

Statistical Analysis
Analyses were performed using R (versions 2.13.2 and 3.4.3) and GraphPad Prism (version 8.3.0). In study I, the association between Arg82Cys and blood lipids was examined by linear regression using an additive genetic model, adjusting for age, sex, and BMI. Meta-analyses of aggregate data of the association of the Arg82Cys variant with fasting HDL subclasses (studies II and III) were performed using inverse variance-weighted meta-analysis from linear regression analyses adjusted for age and BMI, using the R package meta and the function forest.meta, where heterogeneity was assessed by I 2 and the P value for Cochran Q test [41]. If P was greater than .10 in the Cochran Q test, we applied a fixed-effects model. If P was less than .10, a random-effects model was employed. In studies II and III, a P value equal to or below .05 was defined as statistically significant. In study III, the associations between Arg82Cys polymorphism and postprandial plasma lipid levels were tested by linear regression analyses of the incremental area under the curve (AUC) and by unpaired t tests between genotype groups for specific time intervals. Incremental AUC was calculated using the trapezoidal method [42,43]. In study IV, the interaction between the Arg82Cys variant and glycemic status (hyperglycemic vs normoglycemic) on lipid levels was tested by incorporating an interaction term in linear regression models. The results in studies I and IV were corrected for multiple testing by Bonferroni correction (P CORRECTED = .05/17 = .003).

Association of Arg82Cys Variant With High-density Lipoprotein Subclasses in 4522 British Men and Women
To elucidate the association of Arg82Cys with lipids in HDL subclasses in a fasting state, we examined associations with HDL, HDLc, HDL cholesterol esters, and ApoA1 concentrations measured by NMR spectroscopy in 4522 men and women from the OBB cohort (study I). The results are given as effect-per-allele from the additive genetic model. We found that the Cys82 variant was associated with 0.037 mg/L lower ApoA1 levels (P = .008), 0.04 nm lower HDL mean diameter (P = .003), 0.04 µmol/L lower HDL2 cholesterol (P = .004), and 0.04 µM lower HDL3 cholesterol (P = .02) per allele (Table 2). In analyses stratified by HDL particle size, we found that the We found no association between Cys82 carriers and concentration of S-HDL particles, suggesting that the Cys82 variant specifically reduces the levels of larger HDL particles (Fig. 2). We did not find an association with HOMA-IR, ApoB, nor LDL cholesterol (Supplementary  (Tables 3 and 4), we performed a meta-analysis to test whether there are differences between Cys82 carriers and noncarriers in relation to fasting TGs, TC, LDLc, HDLc, or HDL subclasses measured with ultracentrifugation. We found that HDL2c was 0.3 mmol/L lower (P = .005) and the HDL2c/HDL3c ratio was 0.5 units lower (P = .0004) in Cys82 carriers than in noncarriers (Table 4). Furthermore, the  plasma HDL3 TG was 0.01 mmol/L higher in Cys82 carriers than in noncarriers (P = .0003), whereas there was no difference in plasma HDL2 TG content (P = .1). We also found that the HDL2/HDL3 TG ratio was 1.25 units lower in Cys82 carriers than in noncarriers (P = .002).

Association of Nepmucin Arg82Cys Polymorphism With Postprandial Lipemia
To test whether the Arg82Cys genotype affects postprandial fluctuation of blood lipids in a time-dependent manner, we examined differences between 7 heterozygous Cys82 variant carriers and 17 noncarriers in relation to AUCs for HDLc, TGs, TC, apoB48, and apoB100 (Fig. 3A) during a lipid-rich meal challenge (study III). The baseline characteristics were similar between the carriers and noncarriers of the Cys82 variant (Table 1). We found no statistically significant differences in the AUCs or lipid levels at specific time points between Cys82 carriers and noncarriers during the meal challenge (Fig. 3B-3F and Table 5).

Interaction of Arg82Cys Variant With Glycemic Status With Serum Lipid Levels
Previous studies have suggested an association of the Cys82 variant with abnormal glucose metabolism [34,45,46], which could be mechanistically linked to the association between Arg82Cys and lipid levels. To test whether glycemic status modifies the association between Arg82Cys and HDL subclasses, we tested the interaction with glycemic status among 1330 participants in the Danish ADDITION-PRO cohort (study IV). We stratified the participants based on 2-hour plasma glucose and FBG concentration into a normoglycemic (normal GT) group (n = 831) and a hyperglycemic (abnormal GT) group (n = 499). No statistically significant interaction between the Cys82 variant and glycemic status was found (P > .05 for interaction; data not shown).

Discussion
Nepmucin is a type-I membrane protein expressed in vascular endothelial cells that shows high expression in placental tissue, skeletal muscle, and adipose tissue [23,33,47]. The Arg82Cys nepmucin polymorphism was originally identified in an exome-sequencing study for association with lower fasting plasma HDL cholesterol and higher fasting plasma TG concentration [14], and the association has been replicated in subsequent fasting [17,18,48,49], and (for HDL) nonfasting studies [50] (Supplementary Table S2 [44]). Here, we studied association between the Cys82 polymorphism and fasting plasma concentrations of HDL and its subclasses to draw a more detailed image of its role in lipid metabolism. Furthermore, we assessed postprandial changes after a highfat stimulus, to detect additional subtle abnormalities in lipid metabolism that could not be detected in a fasting state, and evaluated potential synergistic effects of Cys82 with glucose metabolism on fasting lipid levels.
Our analyses in the OBB cohort, applying an additive genetic model in 4522 British individuals to HDL subclasses measured by NMR spectroscopy, showed that Cys82 is associated with a smaller HDL diameter and lower cholesterol concentration in M, L, and XL-HDL particles. The associations were strongest in the L-HDL subclass. In a meta-analysis of fasting lipids from 2 RbG studies, including 27 carriers of the Cys82 polymorphism and 37 noncarriers, we observed lower cholesterol levels in plasma HDL2 particles and higher TG levels in plasma HDL3 particles of Cys82 carriers. Our Meta-analysis of the association of nepmucin Arg82Cys with fasting plasma lipid levels in studies II and III. A total of 37 noncarriers (study II: 20, study III: 17) and 27 carriers (study II: 20, study III: 7) were analyzed. No statistically significant heterogeneity between studies II and III was found. Abbreviations: HDL, high-density lipoprotein; LDL, low-density lipoprotein. a Log-transformed effects and SE.  results are concordant with a published GWAS of HDL subclasses measured by NMR spectroscopy in up to 24 925 individuals [17] that showed the Cys82 variant was associated with lower concentrations of L and M-HDL particles [17] (see Supplementary Table S2 [ 44]). Interestingly, smaller HDL diameter has been associated with adverse cardiometabolic outcomes [12,51,52]. In line with this, the Cys82 polymorphism reached nominal significance in the latest GWAS for cardiovascular outcomes (β = 0.057; P = 2.2E-3) [53] (see Supplementary Table S2 [ 44]) and is linked to increased atherosclerosis (P = 4.18E-6), and peripheral artery disease in FINNGEN (P = 2.73E-5/unpublished data, URL: https:// r5.finngen.fi/variant/17-43848758-C-T). Furthermore, we found the Arg82 polymorphism to be linked to decreased ApoA1 level-a protein component of HDL involved in lipid metabolism (see Table 2), which has previously been linked to cardioprotective properties [11]. In our study of 24 Danish men who were challenged with an oral lipid load, we did not see a difference in postprandial lipemia between Cys82 carriers and noncarriers. While the study sample size was limited, the RbG design enhanced statistical power by including a relatively large proportion of carriers with the rare allele. We estimated to have 90% statistical power to exclude an allele-dependent reduction in postprandial HDL plasma levels exceeding 0.17 mmol/L in a test of unpaired means (α = 0.05). This may suggest that the link between Arg82Cys and lipemia is subjected to a fasting state and effects may be obscured in a postprandial state.
A previous RbG study among 42 healthy male carriers and 20 noncarriers of the Cys82 variant showed an  association with lower CD300LG messenger RNA expression in muscle and white adipose tissue, as well as with higher intramyocellular lipid content and forearm glucose uptake [34]. Overall, the findings suggested a role for nepmucin in the regulation of glucose and lipid homeostasis. We therefore hypothesized that the association of Arg82Cys with HDL subclasses might be modified by glycemic status. However, we found no epidemiological evidence of such an interaction. These findings are supported by a previous study from 2013, in which Arg82Cys was not associated with glycemic traits (fasting glucose, 2-hour oral glucose tolerance test, glycated hemoglobin A 1c ) in a linear model [14]. Taken together, the results suggest that nepmucin is involved in lipid transport and lipoprotein maturation. As nepmucin has been shown to adhere to several polar lipids that are known to be present in HDL particles [9,10], we speculate that nepmucin could affect the maturation of HDL molecules from the small (lipid-poor) HDL to the very large spherical HDL particle (eg, via lecithin cholesterol acyltransferase, cholesterol ester transfer protein, and hepatic lipase) [7,54]. Also, the low levels of ApoA1 in carriers of Arg82Cys may indicate decreased uptake of cholesterol and other lipids in HDL (and thus the general decrease in HDL content) or it may simply reflect smaller HDL particles [55]. Because ApoA1 is a cofactor for lecithin cholesterol acyltransferase and thus is involved in the formation and reverse transport of cholesterol esters, specific binding assays between nepmucin and HDL are needed to confirm the mechanistic basis for the link between nepmucin and HDL metabolism. Considering previous studies that suggested a reduction of HDL2 and a shift toward smaller HDL subclasses is linked to obesity [56], it may also be of interest to focus on an obese population in future studies.

Conclusion
Taken together, our results indicate an association between the Arg82Cys variant and a lower concentration of HDL particles and HDLc, especially in larger HDL subclasses, and lower ApoA1 level, suggesting a link between nepmucin and HDLc metabolism.
Foundation, and the strategic and infrastructural research funding from the University of Oulu, Finland, as well as by the British Heart Foundation, the Wellcome Trust (Nos. 090532 and 098381) and the Medical Research Council, UK. The present project was funded by the Lundbeck Foundation through the LuCamp project (www.lucamp.org) and, furthermore, by the Danish Diabetes Academy. N.T.K., C.C., J.S., and U.K. were funded by the Lundbeck Foundation. C.C. was funded by the Novo Nordisk Foundation. O.P. was founder and director of the LuCamp project (www.lucamp. org). The Oxford BioBank (www.oxfordbiobank.org.uk) and Oxford Bioresource are funded by the NIHR Oxford Biomedical Research Centre (BRC). The views expressed are those of the authors and not necessarily those of the NIHR or the Department of Health and Social care.

Disclosures
T.H. and O.P. own stock in Novo Nordisk Inc. The other authors have nothing to disclose.

Data Availability
Some data generated or analyzed during this study are included in this published article or in the data repositories listed in "References" [44]. Supplementary Table S1 contains associations of Arg82Cys with HOMA-IR, APOB, and LDL in study I. Supplementary Table S2 contains associations of Arg82Cys from previously published GWAS. Otherwise, restrictions apply to the availability of some data generated or analyzed during this study to preserve patient confidentiality or because they were used under license. The corresponding author will on request detail the restrictions and any conditions under which access to some of the data may be provided.