2025 年 32 巻 1 号 p. 34-47
Aim: This study investigated the associations of the surface charge of low-density lipoprotein (LDL) with the serum LDL-cholesterol and atherosclerosis levels in a community-based Japanese population.
Methods: The study had a cross-sectional design and included 409 community residents aged 35-79 years who did not take medications for dyslipidemia. The potential electric charge of LDL and the zeta potential, which indicate the surface charge of LDL, were measured by laser Doppler microelectrophoresis. The correlations of the zeta potential of LDL (−mV) with the serum LDL-cholesterol levels (mg/dL), cardio-ankle vascular index (CAVI), and serum high-sensitivity C-reactive protein (hsCRP) levels (log-transformed values, mg/L) were examined using Pearson’s correlation coefficient (r). Linear regression models were constructed to examine these associations after adjusting for potential confounding factors.
Results: A total of 201 subjects with correctly stored samples were included in the primary analysis for zeta potential measurement. An inverse correlation was observed between the LDL zeta potential and the serum LDL-cholesterol levels (r=−0.20; p=0.004). This inverse association was observed after adjusting for sex, age, dietary cholesterol intake, smoking status, alcohol intake, body mass index, and the serum levels of the major classes of free fatty acids (standardized β=−6.94; p=0.005). However, the zeta potential of LDL showed almost no association with CAVI or the serum hsCRP levels. Similar patterns were observed in the 208 subjects with compromised samples as well as all the original 409 subjects.
Conclusion: A higher electronegative surface charge of LDL was associated with lower serum LDL-cholesterol levels in the general Japanese population.
Low-density lipoprotein (LDL) in the blood is a complex particle composed mainly of a single apolipoprotein B-100 molecule and lipids including cholesterol, triglycerides, and phospholipids1). LDL acts as a carrier in the circulation for cholesterol (free cholesterol and esterified cholesterol [i.e., cholesteryl ester]), mainly from the liver to peripheral tissues1). A high level of cholesterol within LDL is a major risk factor for cardiovascular diseases, especially coronary heart disease, due to the development of atherosclerosis2, 3). Therefore, the serum LDL-cholesterol level is a convenient measure for assessing future cardiovascular disease risk4, 5).
All lipoprotein particles have a net negative charge, with LDL having the lowest surface charge among all lipoproteins6). Several constituents and factors associated with lipoprotein particles are involved in regulating the surface charge of lipoprotein particles, mainly as a result of their own anionic charge and/or their conformation with other constituents within lipoprotein particles7-10). Interestingly, the surface charge may be associated with the cholesterol pool in lipoproteins, with evidence from experimental studies supporting this possibility11-14). However, evidence on this topic from epidemiological studies conducted in humans in natural settings is limited15).
We conducted an epidemiological study to investigate the association between the surface charge of LDL and the serum LDL-cholesterol levels in a community-based Japanese population. With respect to this topic, we also investigated the associations between the surface charge of LDL and the atherosclerosis levels using the cardio-ankle vascular index (CAVI)16) and serum high-sensitivity C-reactive protein (hsCRP) levels as variables17).
A cross-sectional study was conducted as part of the Dynamics of Lifestyle and Neighborhood Community on Health Study (DOSANCO Health Study), a community-based study involving 2100 study subjects of ≥ 3 years of age, carried out in the town of Suttu, Hokkaido, Japan18). In 2015, the response rate was 79.6% for all residents at home. In the present study, a detailed survey of Suttu residents aged 35-79 years was conducted to obtain evidence that may contribute to the prevention of metabolic disorders and atherosclerotic disease. Of the 1686 residents of 35-79 years of age, 1,379 (males, n=650; females, n=729) completed a self-administered questionnaire (response rate, 81.8%). Among these residents, 545 (males, n=245; females, n=300) underwent additional physical measurements and provided blood samples between August and November (response rate, 32.3%). Of the 545 subjects, 136 were deemed ineligible for inclusion for the following reasons: history of coronary heart disease and stroke (n=42); use of medications to control dyslipidemia (n=73); serum triglycerides ≥ 400 mg/dL and therefore ineligible for calculation of LDL-cholesterol using the Friedewald equation19) (n=6); missing data on the surface charge of LDL (n=7), CAVI (n=1), and the characteristics other than the surface charge of LDL, serum LDL-cholesterol, CAVI, and serum hsCRP (n=7). The remaining 409 subjects who were not taking anti-dyslipidemia medications (males, n=192; females, n=217) were considered eligible for the study and thus were included in the subsequent analyses.
The study protocol was approved by the Institutional Review Committee for Ethical Issues of the Faculty of Medicine (15-002, 16-007, and 17-008) and the Faculty of Health Sciences (16-10), Hokkaido University. Written informed consent was obtained from all subjects. The present study was approved by the Institutional Review Committee for Ethical Issues of the University of the Ryukyus (1643, 1644, and 1645).
Blood Sample Collection and Biochemical MeasurementsVenous blood samples were collected by cubital venipuncture after an overnight fast. Blood samples were then transported to a commercial laboratory (Daiichi Kishimoto Clinical Laboratories, Inc., Sapporo, Japan) for measurement of serum total cholesterol, high-density lipoprotein (HDL)-cholesterol, triglycerides, fasting plasma glucose, and glycated hemoglobin (HbA1c). The serum lipid levels were measured using standardized methods with an automatic analyzer (BioMajesty JCA-BM8060; Hitachi High-Tech Corporation, Tokyo, Japan and JEOL, Japan Electron Optics Laboratory, Ltd., Tokyo, Japan). The serum LDL-cholesterol levels were calculated using the Friedewald equation: total cholesterol level − HDL-cholesterol level − (triglyceride level/5)19).
Serum was separated from the remaining blood samples by centrifugation after blood coagulation. It was then stored at −80℃ for additional measurements. The serum hsCRP levels were measured using nephelometry at another commercial laboratory (SRL, Inc., Tokyo, Japan). The serum free fatty acid (FFA) levels and the zeta potential of LDL were measured in serum samples stored at our laboratory (Hokkaido University Faculty of Health Science). The serum FFA levels (µmol/L) were determined by liquid chromatography/tandem mass spectrometry (LC/MS)20). The serum samples were processed directly for derivatization using saponification21). The FFA subtypes were as follows: FFA4:0, FFA6:0, FFA12:0, FFA14:0, FFA16:0, FFA18:0, FFA18:1, FFA18:2, FFA18:3, FFA20:4, FFA20:5, and FFA22:6. The following subtypes were not detected and were therefore lacking in the dataset: FFA20:0, FFA22:0, FFA24:0, and FFA26:0. An FFA with a carbon number x and a double bond number y is described as FFAx:y (e.g., FFA18:2 for linoleic acid). FFAs were classified as saturated fatty acids (FAs) (FFA4:0, FFA6:0, FFA12:0, FFA14:0, FFA16:0, and FFA18:0), monounsaturated FAs (FFA18:1), n-6 polyunsaturated FAs (FFA18:2 and FFA20:4), and n-3 polyunsaturated FAs (FFA18:3, FFA20:5, and FFA22:6).
Lipoprotein SeparationThe lipoprotein fractions were isolated from the serum by sequential ultracentrifugation (Hitachi Himac CS120GXL; Eppendorf Himac Technologies Co., Ltd., Ibaragi, Japan) using an angle rotor (S140AT2) according to a modified protocol previously reported22). In brief, frozen serum samples were incubated at 37℃ for 3 min. The density of sera (240 µL) was adjusted to d=1.019 kg/L with potassium bromide solution (d=1.023, 720 µL), followed by centrifugation at 120,000 rpm for 70 min at 16℃. The upper 300 µL fraction containing chylomicrons, very low-density lipoprotein (VLDL), and intermediate-density lipoproteins (IDL) was removed thoroughly by pipetting. The remaining fraction was adjusted to d=1.063 kg/L with potassium bromide solution (d=1.160, 300 µL) and then centrifuged at 120,000 rpm for 110 min at 16℃. A 300 µL fraction of the upper layer containing LDL was isolated by pipetting and then stored at 4℃ until use. EDTA was added at a final concentration of 2 µM to prevent oxidation during storage. Before using the LDL, the EDTA was removed using an ultrafiltration device with a molecular cut-off of 100 kDa (Amicon Ultra 0.5 mL; Merck Millipore Ltd., Ireland), followed by addition of 1 mM phosphate with 10 mM NaCl buffer (pH 7.4) for the measurement of the zeta potential.
Measurement of the Zeta Potential of LDLThe zeta potential, specifically the potential electric charge of LDL (mV), indicates the surface charge of LDL and was measured at 25℃ in all the samples by applying 30 V between the electrodes using a Zetasizer Nano-ZS (Malvern Instruments, Malvern, UK). This measurement uses laser Doppler microelectrophoresis, which applies a direct electric field to the lipoprotein fractions. The zeta potential of LDL, expressed as a negative value, was calculated using software (Malvern Instruments, Malvern, UK). To avoid confusion in interpretation, we expressed the absolute values of the LDL zeta potential as alternative units (−mV) and used these values in the analyses.
Measurement of CAVIThe CAVI was measured in an air-conditioned room at 24-26℃ using a VaSera (VS-3000TN) CAVI instrument (Fukuda Denshi Co. Ltd., Tokyo, Japan). VaSera is equipped with both measurement and calculation systems and automatically displays the calculated CAVI values23). The cuffs were wrapped around the bilateral upper arms and ankles of the subjects while they were in the supine position. After a 5-minute rest, CAVI measurements were conducted with simultaneous measurement of the heart rate.
Assessments of Physical Indices and LifestyleThe body height and weight of the participants were measured, and the body mass index (BMI) was calculated as weight (kg)/height squared (m2). Other data collected using a self-administered questionnaire included age, sex, smoking habits, and medical history. The smoking status was assessed based on whether the subject had never smoked, was a former smoker, or was a current smoker. For this report, the subjects were classified as nonsmokers or current smokers. Dietary habits during the past month were assessed using a validated brief self-administered diet history questionnaire (BDHQ)24, 25). Briefly, this is a 10-page fixed-portion questionnaire that estimates the dietary intake of 58 food items. The food items and portion sizes in the questionnaire were primarily derived from a food list used in the National Health and Nutrition Survey in Japan and from several Japanese recipe books. Energy and nutrient intakes were calculated based on the weight of food intake obtained from the questionnaire and the Japanese Standard Tables of Food Composition. All subjects provided appropriate dietary data based on a total daily energy intake of 500-5000 kcal/day. Blood pressure was measured twice using an automatic manometer (HBP-T105S-N; Omron Colin Co., Ltd, Tokyo, Japan) after the subject rested for five min in the seated position. The means of the first and second readings were used for the analyses.
Statistical AnalysisWe encountered a problem in the storage of the frozen blood samples during the mid-course of the measurements of LDL zeta potential due to an approximately 2-day electrical outage without electrical backup caused by the Hokkaido Eastern Iburi Earthquake on September 6, 2018. Nearly half of the data on the LDL zeta potential were collected from samples that had been stored correctly (n=201), whereas the remaining data were collected from samples compromised by the electrical outage (n=208). Therefore, we considered this issue during the data analysis. We first performed a series of data analyses on the 201 correctly stored samples and then carried out similar analyses on the 208 samples affected by the electrical outage. Finally, we analyzed all of the original 409 samples collected in the study.
The relationship between the LDL zeta potential (−mV) and the serum LDL-cholesterol levels (mg/dL) was examined by calculating the Pearson’s correlation coefficient (r). Linear regression models were constructed to investigate this association, after adjusting for potential confounding factors. The following covariates were included step-by-step as confounders in the regression models: age (years, as a continuous variable) and sex (male or female) (model 1); dietary cholesterol intake (log-transformed values, mg/day, as a continuous variable) (model 2); smoking status (non-smoker or current smoker), alcohol intake (none, 0.1-19.9, or ≥ 20 g/day, using two dummy variables with none as the reference), and BMI (kg/m2, as a continuous variable) (model 3); serum saturated FA (log-transformed values, µmol/L, as a continuous variable), monounsaturated FA (log-transformed values, µmol/L, as a continuous variable), n-6 polyunsaturated FA (log-transformed values, µmol/L, as a continuous variable), and n-3 polyunsaturated FA levels (log-transformed values, µmol/L, as a continuous variable) (model 4). Because the original data on the dietary cholesterol intake and the serum levels of major FFA classes showed skewed distributions, the values were expressed as transformed natural logarithms. For models of all 409 subjects, further adjustment was made for the sample status (correct or compromised storage).
The relationships of the LDL zeta potential (−mV) with CAVI and the serum hsCRP levels (log-transformed values, mg/L) were examined by calculating Pearson’s correlation coefficient (r). Linear regression models were also used to investigate these associations following adjustment for the following potential confounding factors: age, sex (model 1), smoking status, alcohol intake, BMI, hypertension status, diabetes status, heart rate (per minute, as a continuous variable) only for CAVI (model 2), and the serum LDL-cholesterol levels (mg/dL, as a continuous variable) (model 3). For hsCRP, the analyses were performed after excluding subjects with marked inflammation, defined as serum hsCRP ＞10.0 mg/L, which was largely attributable to sources other than atherosclerotic lesions (e.g., acute infection, chronic severe inflammatory disease)17) (n=4 for correctly stored samples and n=2 for compromised samples).
Similar analyses were performed in the subjects stratified by demographic, lifestyle, and clinical characteristics including sex, age (35-54 and 55-79 years), smoking status (non-smoker and current smoker), alcohol intake (none and ≥ 0.1 g/day), BMI (≤ 24.9 and ≥ 25 kg/m2)26), comorbidity status including hypertension and diabetes (absence of both and presence of either/both), and the serum LDL-cholesterol levels (≤ 139 and ≥ 140 mg/dL)4). Hypertension was defined as systolic blood pressure ≥ 140 mmHg, diastolic blood pressure ≥ 90 mmHg27) and/or taking medications to control hypertension, whereas diabetes was defined as fasting plasma glucose ≥ 126 mg/dL and/or HbA1c ≥ 6.5%28) and/or taking medications to control diabetes. The aim of these analyses was to determine whether the associations of the LDL zeta potential with the serum LDL-cholesterol levels, CAVI, and the serum hsCRP levels were similar across the different demographic, lifestyle, and clinical characteristics groups.
Finally, we investigated the relationships between the LDL zeta potential (−mV) and the fasting serum levels of the major FFA classes and total FFA, which we considered to be measurable substances that may have regulated the surface charge of LDL12, 29, 30). These relationships were examined by calculating Pearson’s correlation coefficient (r), while linear regression models were used to investigate these associations after adjusting for age and sex.
The analyses were performed using the Stata 17 software program (StataCorp LP, College Station, TX, USA). All probability values were two-tailed and the significance level was set at p＜0.05.
Table 1 summarizes the characteristics of the study population stratified according to the blood sample status. The mean age (standard deviation [SD]) of the original 409 study subjects was 56.8 (12.4) years, with females accounting for 53.1% of these subjects. The mean (SD) LDL zeta potential, serum LDL-cholesterol levels, and CAVI of the 409 subjects were 11.1 (1.8) −mV, 124.7 (30.8) mg/dL, and 8.0 (1.2), respectively. The median (interquartile range) serum hsCRP levels of the 403 subjects without marked inflammation was 0.44 (0.22, 0.88) mg/dL.
| Overall (n = 409) | Blood sample status | p value for difference | ||
|---|---|---|---|---|
| Correctly stored (n = 201) | Compromised storage (n = 208) | |||
| Age, years | 56.8 (12.4) | 56.6 (13.3) | 57.0 (11.5) | 0.72 |
| Females | 53.1% (217) | 65.2% (131) | 41.4% (86) | ＜0.001 |
| Zeta potential of LDL, −mV | 11.1 (1.8) | 10.9 (1.6) | 11.4 (1.9) | 0.002 |
| Dietary cholesterol intake, mg/day | 342.6 (238.6, 478.5) | 380.8 (248.8, 486.2) | 321.2 (223.1, 463.3) | 0.02 |
| Current smoker | 56.2% (230) | 52.7% (106) | 59.6% (124) | 0.16 |
| Alcohol intake, g/day | 1.2 (0, 18.2) | 1.1 (0, 14.4) | 2.3 (0, 27.1) | 0.11 |
| Body mass index, kg/m2 | 23.6 (3.6) | 23.5 (3.8) | 23.7 (3.5) | 0.66 |
| Serum saturated FA, μmol/L | 205.6 (154.9, 295.4) | 202.8 (154.7, 299.1) | 209.8 (155.1, 293.0) | 0.81 |
| Serum monounsaturated FA, μmol/L | 193.8 (134.5, 275.2) | 194.3 (134.5, 282.5) | 192.4 (134.1, 268.2) | 0.93 |
| Serum n-6 polyunsaturated FA, μmol/L | 89.0 (63.6, 122.2) | 90.5 (61.7, 122.2) | 87.2 (64.8, 122.5) | 0.93 |
| Serum n-3 polyunsaturated FA, μmol/L | 39.2 (27.7, 56.6) | 39.2 (27.3, 55.5) | 39.3 (27.9, 57.7) | 0.65 |
| Serum total free FA, μmol/L | 535.2 (382.1, 739.0) | 532.4 (376.6, 740.4) | 538.7 (383.7, 734.7) | 0.90 |
| Hypertension | 39.1% (160) | 38.3% (77) | 39.9% (83) | 0.74 |
| Diabetes | 11.5% (47) | 9.5% (19) | 13.5% (28) | 0.20 |
| Heart rate, per min | 63.5 (9.0) | 63.5 (8.6) | 63.5 (9.3) | 0.97 |
| Serum LDL-cholesterol, mg/dL | 124.6 (30.8) | 124.2 (32.9) | 125.0 (28.7) | 0.78 |
| Cardio-ankle vascular index | 8.0 (1.2) | 7.9 (1.2) | 8.1 (1.2) | 0.15 |
| Serum hsCRP, mg/L | 0.45 (0.22, 0.93) | 0.47 (0.21, 0.98) | 0.44 (0.23, 0.89) | 0.90 |
| Serum hsCRP, mg/L＊ | 0.44 (0.22, 0.88) | 0.45 (0.21, 0.88) | 0.43 (0.23, 0.87) | 0.96 |
Abbreviations: FA, fatty acid; hsCRP, high-sensitivity C-reactive protein; LDL, low-density lipoprotein.
Blood sample status represents whether the samples were stored correctly until measurement of the zeta potential of LDL.
Hypertension was defined as systolic blood pressure ≥ 140 mmHg, diastolic blood pressure ≥ 90 mmHg [reference 27] and/or taking medications to control hypertension.
Diabetes was defined as fasting plasma glucose ≥ 126 mg/dL and/or HbA1c ≥ 6.5% [reference 28] and/or taking medications to control diabetes.
＊Subjects with marked inflammation, defined as serum hsCRP ＞10.0 mg/L [reference 17], were excluded (n = 4 for correctly stored samples and n = 2 for compromised samples).
Data are expressed as mean (standard deviation), median (interquartile range) or % (number) of subjects.
The unpaired t-test, Mann-Whitney U test, or Chi-square test were used to compare each characteristic in each blood sample group.
The mean age (SD) of the 201 study subjects with correctly stored samples was 56.6 (13.3) years, with females accounting for 65.2% of these subjects. The mean (SD) LDL zeta potential, the serum LDL-cholesterol levels, and CAVI of the 201 subjects were 10.9 (1.6) −mV, 124.2 (32.9) mg/d L, and 7.9 (1.2), respectively). The median (interquartile range) of the serum hsCRP levels of the 197 subjects without marked inflammation was 0.45 (0.21, 0.88) mg/dL. The LDL zeta potential in the 201 subjects with correctly stored samples was slightly, but significantly, lower by 0.5 −mV than that measured in the 208 subjects with compromised samples. The majority of other characteristics were broadly similar across blood sample statuses, except for the prevalence of a female sex and the dietary cholesterol intake.
LDL Zeta Potential and the Serum LDL-Cholesterol LevelsFig.1 shows a scatter plot used to examine the relationship between LDL zeta potential and serum LDL-cholesterol levels. There was a slight but significant inverse association between LDL zeta potential (−mV) and the serum LDL-cholesterol levels in the 201 subjects with correctly stored samples (r=0.20; p=0.004). The association of interest remained significantly inverse even after adjusting for major potential confounding factors, including sex, age, dietary cholesterol intake, and the serum levels of major FFA classes (fully adjusted standardized β=−6.94; 95% confidence interval, −11.78 to −2.11; p=0.005) (Table 2).
Pearson’s correlation coefficients (r) for each group (a, b, and c) were −0.20 (p=0.004), −0.27 (p＜0.001), and −0.23 (p＜0.001), respectively. Abbreviations: LDL, low-density lipoprotein.
| Standardized β coefficient (95% confidence interval) | p value | |
|---|---|---|
| Age and sex-adjusted, model 1 | −7.41 (−12.39 to −2.43) | 0.004 |
| Multivariate-adjusted, model 2 | −7.67 (−12.66 to −2.69) | 0.003 |
| Multivariate-adjusted, model 3 | −6.63 (−11.55 to −1.71) | 0.008 |
| Multivariate-adjusted, model 4 | −6.94 (−11.78 to −2.11) | 0.005 |
Abbreviations: FA, fatty acid; LDL, low-density lipoprotein.
Four different linear regression models were used to examine the standardized β coefficient: model 1, adjusted for age and sex; model 2, adjusted for the same covariates in model 1, in addition to dietary cholesterol intake; model 3, adjusted for the same covariates in model 2, in addition to smoking status, alcohol intake, and body mass index; model 4 adjusted for the same covariates in model 3, in addition to serum saturated FA, monounsaturated FA, n-6 polyunsaturated FA, and n-3 polyunsaturated FA levels. Log-transformed values were used for dietary cholesterol intake and serum levels of major free FA classes.
There was also an inverse association between the LDL zeta potential and the serum LDL-cholesterol levels in 208 subjects with compromised samples and a total of 409 subjects (Fig.1 and Supplementary Table 1).
| Standardized β coefficient (95% confidence interval) | p value | |
|---|---|---|
| 208 subjects with compromised samples | ||
| Age and sex-adjusted, model 1 | −6.17 (−9.81 to −2.52) | 0.001 |
| Multivariate-adjusted, model 2 | −6.13 (−9.79 to −2.48) | 0.001 |
| Multivariate-adjusted, model 3 | −6.53 (−10.30 to −2.76) | 0.001 |
| Multivariate-adjusted, model 4 | −5.95 (−9.73 to −2.17) | 0.002 |
| Total 409 subjects | ||
| Age and sex-adjusted, model 1 | −6.75 (−9.72 to −3.78) | ＜0.001 |
| Multivariate-adjusted, model 2 | −6.76 (−9.73 to −3.79) | ＜0.001 |
| Multivariate-adjusted, model 3 | −6.58 (−9.55 to −3.61) | ＜0.001 |
| Multivariate-adjusted, model 4 | −6.45 (−9.40 to −3.51) | ＜0.001 |
Abbreviations: FA, fatty acid; LDL, low-density lipoprotein.
Four different linear regression models were used to examine the standardized β coefficient: model 1, adjusted for age and sex; model 2, adjusted for the same covariates in model 1, in addition to dietary cholesterol intake; model 3, adjusted for the same covariates in model 2, in addition to smoking status, alcohol intake, and body mass index; model 4, adjusted for the same covariates in model 3, in addition to serum saturated FA, monounsaturated FA, n-6 polyunsaturated FA, and n-3 polyunsaturated FA levels. For models of all 409 subjects, further adjustment was made for sample status. Log-transformed values were used for dietary cholesterol intake and serum levels of major free FA classes.
When the analyses were performed on data stratified by sex, age, smoking status, alcohol intake, BMI, comorbidity status, and the serum LDL-cholesterol levels, the inverse association observed in the 201 subjects with correctly stored samples was broadly similar across all strata (Table 3). The patterns were also applicable to all strata of the 208 subjects with compromised samples (data not shown) and all strata of the 409 subjects (Supplementary Table 2).
| Standardized β coefficient (95% confidence interval) | p value | |
|---|---|---|
| Sex | ||
| Males (n= 70), multivariate-adjusted | −11.11 (−22.09 to −0.14) | 0.047 |
| Females (n= 131), multivariate-adjusted | −7.65 (−13.32 to −1.97) | 0.009 |
| Age | ||
| 35-54 years (n= 86), multivariate-adjusted | −8.41 (−15.58 to −1.25) | 0.02 |
| 55-79 years (n= 115), multivariate-adjusted | −6.36 (−13.00 to 0.28) | 0.06 |
| Smoking status | ||
| Non-smoker (n= 95), multivariate-adjusted | −6.34 (−12.93 to 0.25) | 0.06 |
| Current smoker (n= 106), multivariate-adjusted | −10.22 (−18.45 to −2.00) | 0.02 |
| Alcohol intake | ||
| None (n= 86), multivariate-adjusted | −4.90 (−13.2 to 3.37) | 0.24 |
| ≥ 0.1 g/day (n= 115), multivariate-adjusted | −9.93 (−16.19 to −3.68) | 0.002 |
| Body mass index | ||
| ≤ 24.9 kg/m2 (n= 138), multivariate-adjusted | −2.69 (−8.33 to 2.95) | 0.35 |
| ≥ 25 kg/m2 (n= 63), multivariate-adjusted | −16.11 (−26.76 to −5.45) | 0.004 |
| Comorbidity status (hypertension and diabetes) | ||
| Neither (n= 115), multivariate-adjusted | −6.84 (−13.52 to −0.16) | 0.045 |
| Either or both (n= 86), multivariate-adjusted | −6.05 (−13.34 to −1.23) | 0.10 |
| Serum LDL-cholesterol levels | ||
| ≤ 139 mg/dL (n= 138), multivariate-adjusted | −3.44 (−7.34 to 0.45) | 0.08 |
| ≥ 140 mg/dL (n= 63), multivariate-adjusted | −5.24 (−12.76 to 2.28) | 0.17 |
Abbreviations: FA, fatty acid; LDL, low-density lipoprotein.
Hypertension was defined as systolic blood pressure ≥ 140 mmHg, diastolic blood pressure ≥ 90 mmHg [reference 27] and/or taking medications to control hypertension.
Diabetes was defined as fasting plasma glucose ≥ 126 mg/dL and/or HbA1c ≥ 6.5% [reference 28] and/or taking medications to control diabetes.
A linear regression model was used to examine the standardized β coefficient adjusted for age, sex, dietary cholesterol intake, smoking status, alcohol intake, body mass index, and serum saturated FA, monounsaturated FA, n-6 polyunsaturated FA, and n-3 polyunsaturated FA levels. Log- transformed values were used for dietary cholesterol intake and serum levels of major free FA classes.
| Standardized β coefficient (95% confidence interval) | p value | |
|---|---|---|
| Sex | ||
| Males (n= 192), multivariate-adjusted | −6.65 (−10.94 to −2.37) | 0.003 |
| Females (n= 217), multivariate-adjusted | −8.17 (−12.54 to −3.79) | ＜0.0001 |
| Age | ||
| 35-54 years (n= 166), multivariate-adjusted | −8.23 (−12.83 to −3.62) | 0.001 |
| 55-79 years (n= 243), multivariate-adjusted | −5.85 (−9.74 to −1.97) | 0.003 |
| Smoking status | ||
| Non-smoker (n= 179), multivariate-adjusted | −5.61 (−10.64 to −0.58) | 0.03 |
| Current smoker (n= 230), multivariate-adjusted | −7.79 (−11.67 to −3.91) | ＜0.0001 |
| Alcohol intake | ||
| None (n= 171), multivariate-adjusted | −6.96 (−11.05 to −2.86) | 0.001 |
| ≥ 0.1 g/day (n= 238), multivariate-adjusted | −6.26 (−10.58 to −1.95) | 0.005 |
| Body mass index | ||
| ≤ 24.9 kg/m2 (n= 275), multivariate-adjusted | −4.76 (−8.12 to −1.40) | 0.006 |
| ≥ 25 kg/m2 (n= 134), multivariate-adjusted | −10.30 (−16.34 to −4.25) | 0.001 |
| Comorbidity status (hypertension and diabetes) | ||
| Neither (n= 230), multivariate-adjusted | −5.77 (−10.16 to −1.39) | 0.01 |
| Either or both (n= 179), multivariate-adjusted | −6.94 (−11.00 to −2.89) | 0.001 |
| Serum LDL-cholesterol levels | ||
| ≤ 139 mg/dL (n= 288), multivariate-adjusted | −3.45 (−5.66 to −1.25) | 0.002 |
| ≥ 140 mg/dL (n= 121), multivariate-adjusted | −4.37 (−9.20 to 0.46) | 0.08 |
Abbreviations: FA, fatty acid; LDL, low-density lipoprotein.
Hypertension was defined as systolic blood pressure ≥ 140 mmHg, diastolic blood pressure ≥ 90 mmHg [reference 27] and/or taking medications to control hypertension.
Diabetes was defined as fasting plasma glucose ≥ 126 mg/dL and/or HbA1c ≥ 6.5% [reference 28] and/or taking medications to control diabetes.
A linear regression model was used to examine the standardized β coefficient adjusted for age, sex, dietary cholesterol intake, smoking status, alcohol intake, body mass index, serum saturated FA, monounsaturated FA, n-6 polyunsaturated FA, and n-3 polyunsaturated FA levels, and sample status. Log-transformed values were used for dietary cholesterol intake and serum levels of major free FA classes.
Fig.2 shows a scatter plot used to examine the relationship between LDL zeta potential and CAVI. There was almost no association between the LDL zeta potential (−mV) and CAVI in 201 subjects with correctly stored samples (r=−0.04; p=0.60). The association of interest remained absent even after adjusting for major potential confounding factors, including sex, age, traditional cardiovascular risk factors, heart rate, and the serum LDL-cholesterol levels. Although the value of the standardized β coefficient was tiny, it was statistically significant (fully adjusted standardized β=−0.15; 95% confidence interval, −0.27 to −0.04; p=0.01) (Table 4).
Pearson’s correlation coefficients (r) for each group (a, b, and c) were −0.04 (p=0.60), −0.02 (p=0.74), and −0.02 (p=0.72), respectively. Abbreviations: LDL, low-density lipoprotein.
| Standardized β coefficient (95% confidence interval) | p value | |
|---|---|---|
| Age and sex-adjusted, model 1 | −0.10 (−0.22 to 0.02) | 0.12 |
| Multivariate-adjusted, model 2 | −0.13 (−0.25 to −0.02) | 0.03 |
| Multivariate-adjusted, model 3 | −0.15 (−0.27 to −0.04) | 0.01 |
Abbreviations: LDL, low-density lipoprotein.
Three different linear regression models were used to examine the standardized β coefficient: model 1, adjusted for age and sex; model 2, adjusted for the same covariates in model 1, in addition to smoking status, alcohol intake, body mass index, hypertension status, diabetes status, and heart rate; model 3 adjusted for the same covariates in model 2, in addition to serum LDL-cholesterol levels.
Fig.3 shows a scatter plot used to examine the relationship between the LDL zeta potential and the serum hsCRP levels. There was almost no association between the LDL zeta potential (−mV) and the hsCRP levels (log-transformed values, mg/L) in 197 subjects with correctly stored samples after excluding those with marked inflammation (r=0.02; p=0.78). The association of interest remained absent even after adjusting for major potential confounding factors, including sex, age, traditional cardiovascular risk factors, and the serum LDL-cholesterol levels (fully adjusted standardized β=−0.004; 95% confidence interval, −0.16 to 0.16; p=0.96) (Table 5).
Pearson’s correlation coefficients (r) for each group (a, b, and c) were −0.02 (p=0.78), −0.06 (p=0.42), and −0.02 (p=0.69), respectively. Abbreviations: hsCRP, high-sensitivity C-reactive protein; LDL, low-density lipoprotein. ＊Subjects with marked inflammation, defined as serum hsCRP ＞10.0 mg/L [reference 17], were excluded.
| Standardized β coefficient (95% confidence interval) | p value | |
|---|---|---|
| Age and sex-adjusted, model 1 | 0.003 (−0.16 to 0.17) | 0.97 |
| Multivariate-adjusted, model 2 | −0.0008 (−0.16 to 0.16) | 0.99 |
| Multivariate-adjusted, model 3 | −0.004 (−0.16 to 0.16) | 0.96 |
Abbreviations: hsCRP, high-sensitivity C-reactive protein; LDL, low-density lipoprotein.
＊Subjects with marked inflammation, defined as serum hsCRP ＞10.0 mg/L [reference 17], were excluded.
Three different linear regression models were used to examine the standardized β coefficient: model 1, adjusted for age and sex; model 2, adjusted for the same covariates in model 1, in addition to smoking status, alcohol intake, body mass index, hypertension status, and diabetes status; model 3 adjusted for the same covariates in model 2, in addition to serum LDL-cholesterol levels. Log-transformed values were used for the serum hsCRP levels.
There were also almost no associations between the LDL zeta potential and the CAVI or serum hsCRP levels in 208 (or 206) subjects with compromised samples and 409 (or 403) subjects after excluding those with marked inflammation (data not shown).
When the analyses were performed on the data stratified by demographic, lifestyle, and clinical characteristics, the null associations between the LDL zeta potential and CAVI or the serum hsCRP levels observed in 201 (or 197) subjects with correctly stored samples were broadly similar across all strata for these two groups (data not shown). The patterns were also applicable to all strata of the 208 (or 206) subjects with compromised samples and all strata for the 409 (or 403) subjects (data not shown).
LDL Zeta Potential and the Serum FFA LevelsThere were almost no associations between the fasting serum levels of the major FFA classes and total FFA (log-transformed values, µmol/L) and LDL zeta potential (−mV) in the 201 subjects with correctly stored samples (all |r|＜0.04; p ≥ 0.05) (Table 6).
| Standardized β coefficient (95% confidence interval) | p value | |
|---|---|---|
| Serum saturated FA, | ||
| age and sex-adjusted | 0.05 (−0.17 to 0.27) | 0.66 |
| Serum monounsaturated FA, | ||
| age and sex-adjusted | 0.008 (−0.21 to 0.23) | 0.94 |
| Serum n-6 polyunsaturated FA, | ||
| age and sex-adjusted | 0.02 (−0.20 to 0.24) | 0.84 |
| Serum n-3 polyunsaturated FA, | ||
| age and sex-adjusted | 0.06 (−0.17 to 0.29) | 0.60 |
| Serum total free FA, | ||
| age and sex-adjusted | 0.03 (−0.19 to 0.25) | 0.80 |
Abbreviations: FA, fatty acid; LDL, low-density lipoprotein.
A linear regression model was used to examine the standardized β coefficient adjusted for age and sex.
Similar patterns were also observed in 208 subjects with compromised samples and the whole population of 409 subjects (data not shown).
This cross-sectional study was conducted to investigate the association between the zeta potential of LDL and the serum LDL-cholesterol levels, CAVI, and serum hsCRP levels in a community-based Japanese population not taking dyslipidemia medications. Study subjects with a higher electronegative zeta potential of LDL had significantly lower serum LDL-cholesterol levels after adjusting for major potential confounding factors, including sex, age, dietary cholesterol intake, and serum levels of major FFA classes. This inverse association was present even within the serum LDL-cholesterol levels that were considered clinically normal. However, the study showed that the LDL zeta potential did not correlate with CAVI or the serum hsCRP levels, both of which are indicators of atherosclerosis16, 17). One important issue that should be acknowledged is that the LDL zeta potential represents the average level of surface charge of all LDL, not the percentage of the most electronegative LDL subfraction (L5) with atherogenic properties, which is distinct from that of the other LDL subfractions9, 10).
Epidemiological studies relevant to our study were limited to both the serum LDL-cholesterol and atherosclerosis levels. Védie et al.15) conducted a cross-sectional study on 104 male patients with asymptomatic hypercholesterolemia and used agarose electrophoresis to measure the surface charge of LDL based on the relative electrophoretic mobility (i.e., the LDL-to-albumin migration distance ratio). This previous study on subjects with borderline, but not overt, dyslipidemia showed an inverse association between the surface charge of LDL and the serum LDL-cholesterol levels15). This finding is consistent with the findings of the current study. To the best of our knowledge, our study is the first to show an inverse association between the surface charge of LDL and the serum LDL-cholesterol levels in the general population, using an epidemiological approach, after adjusting for several major confounding factors. In addition, the association observed was consistent across several demographic, lifestyle, and clinical characteristics. Therefore, these results provide new insights into the relevant academic fields.
The surface charge of LDL is dependent on several constituents, including proteins and lipids7, 8). Cholesteryl ester is the dominant type of cholesterol in LDL but has little influence on its surface charge7), whereas in HDL, it contributes to the increased electronegativity of the surface charge6). Other constituents or factors associated with LDL that affect the surface charge may therefore regulate the amount of cholesterol pooled within LDL in circulation. However, our epidemiological study could not identify the factors that determine the surface charge of LDL or clarify the causal mechanism of the association between LDL zeta potential and the serum LDL-cholesterol levels. Therefore, we can only speculate on a possible causal mechanism based on evidence from relevant experimental studies.
Several chemical and compositional changes that generate electronegative LDL (L5)9, 10) are key factors that should be considered when speculating on the cause of the inverse association between the surface charge of LDL and the serum LDL-cholesterol levels. Briefly, the composition of phospholipids is altered mainly by lipolysis of phosphatidylcholine into lysophosphatidylcholine and non-esterified FAs31, 32). Increased levels of non-esterified FAs are associated with increased electronegativity of the LDL surface charge12, 29, 30). Lysine residues of apolipoprotein B-100 are modified with increased electronegativity of the LDL surface charge33), resulting in misfolded structures of apolipoprotein B-100, which have a decreased content of α-helices and an increased content of β-sheets34, 35). These changes in both phospholipids and apolipoprotein B-100 in LDL lead to changes in its conformation (i.e., decreased size and increased density of LDL particles) and also increase the electronegativity of LDL6, 9, 10, 36). Electronegative LDL has impaired binding to the LDL receptor, probably because of the modified lysine residues of apolipoprotein B-100 33, 37), and may therefore be prone to remain circulating in the blood. When focusing on the final stage of electronegative LDL, our results do not support the possibility that a prolonged presence of modified LDL electronegatively in the blood may contribute to increased serum levels of LDL-cholesterol.
Notably, there is evidence that FA anions increase the surface charge of LDL, which then suppresses the activity of both cholesteryl ester transfer protein (CETP) and lipid transfer inhibitor protein (LTIP)12). CETP is a plasma protein that mediates the reciprocal transfer of lipids between lipoproteins, especially the exchange of cholesteryl esters in HDL and triglycerides in VLDL38). LTIP mainly inhibits the lipid exchange reaction between LDL, which provides cholesteryl ester, and VLDL, which provides triglycerides13). Suppressed activity of these two proteins ultimately leads to reduced storage of cholesteryl esters in LDL13, 14). Consequently, the increased electronegativity of the surface charge of LDL may decrease the serum LDL-cholesterol levels. Our data could not clarify whether non-esterified FAs within LDL contributed to the increase in electronegativity of the surface charge of LDL. However, our results indicated that FFAs in circulation during the fasting state were not involved in determining the surface charge of LDL. Similarly, there is evidence that phosphatidylinositol, a minor phospholipid in lipoproteins39), has a greater effect on increasing the electronegativity of the surface charge than other constituents (e.g., phosphatidylcholine, phosphatidic acid, and phosphatidylserine)40). The incubation of human plasma with phosphatidylinositol has been shown to decrease the intensity of the stained bands of cholesteryl ester and triglycerides in LDL, whereas the staining intensity for HDL increases, resulting in an almost constant staining intensity for all lipoproteins11). A similar incubation was also reported to reduce the amount of cholesteryl ester in both VLDL and LDL, and to increase the cholesteryl ester content in HDL11). Further studies are therefore warranted to clarify the underlying mechanisms of the relationship between the zeta potential of LDL and the serum LDL-cholesterol levels observed in our study.
Electronegative LDL interacts with the lectin-like oxidized LDL receptor 1 (LOX-1) in endothelial cells and induces an atherogenic response41). Rodríguez et al.42) used the zeta potential of LDL to investigate its association with carotid intima-media thickness (cIMT) in Spanish individuals with and without systemic lupus erythematosus (SLE). In 78 subjects with SLE, a higher electronegative zeta potential of LDL was shown to be associated with an increase in cIMT, with a correlation coefficient (r) of approximately 0.3 42). In contrast, in 32 subjects without SLE, no significant correlation was observed between the LDL zeta potential and cIMT (details not described)42). Our results were not in line with the concept of electronegative LDL with atherogenic properties9, 10), although the CAVI indicator we used was originally regarded as an index of arterial stiffness rather than an index of atherosclerosis23). Moreover, our results showed no association between the LDL zeta potential and the serum hsCRP level, which is a marker of the inflammatory response that contributes to atherosclerosis17). Therefore, there is a need for further large-scale studies using the same and different indices of atherosclerosis.
The present study had several original points and strengths, including the enrolment of community residents in the study population, the larger size of the study population in comparison to a previous study in this area15), and consideration of potential confounding factors, especially dietary cholesterol intake and serum levels of major FFA classes. However, the present study is also associated with several limitations. First, the cross-sectional design could not confirm a forward perspective association, especially considering that a lower surface charge of LDL may cause higher serum LDL-cholesterol levels. Second, we did not collect data on the serum levels of LDL particles and therefore were unable to account for this factor in the analyses. Evidence suggests that there is no association between the surface charge of LDL and the serum apolipoprotein B levels15), which may be closely correlated with the serum levels of LDL particles. Therefore, the serum levels of LDL particles may have had a minor confounding effect on our results. Third, we were only able to collect reliable data on the LDL zeta potential from approximately half of the study population due to a problem with the storage of the frozen blood samples. This problem may have caused an average systematic measurement error of 0.5 −mV in the measurement of the LDL zeta potential of the remaining samples. However, the patterns for the associations of interest were broadly similar, irrespective of sample status and other characteristics. Therefore, our study provides robust findings, at least in the direction of the association of interest. Finally, our study population included residents from a single northern, rural coastal community in Japan. Lifestyle factors in this community may have affected the dietary cholesterol intake and the serum FFA levels in the study population, and such factors can also influence the cholesterol metabolism. Therefore, caution should be exercised when generalizing the results of this study.
This study demonstrated that a higher electronegative surface charge of LDL was associated with lower serum LDL-cholesterol levels in the general Japanese population. Our results suggest that a lower surface charge of LDL stimulates cholesterol storage in LDL. Larger-scale longitudinal studies are needed to confirm a definitive causal relationship between these two parameters. However, our data showed that the surface charge of LDL was not correlated with either CAVI or the serum hsCRP levels.
We express special gratitude to all study participants, Suttu municipal government officers, Suttu clinic staff members, and other individuals who were involved in the DOSANCO Health Study.
This study was supported by grants from the Integration Research for Agriculture and Interdisciplinary Fields (14538261) and the Japan Society for the Promotion of Science, KAKENHI (26670322, 16K15353, and 17H04120).
The authors declare no conflicts of interest.
