Twin Study: The Factors Affecting the Serum LDL-C and HDL-C Levels and an RNA-Seq Analysis in Mononuclear Cells in Monozygotic Twins

Abstract

Aim: A twin study is a valuable tool for elucidating the acquired factors against lifestyle diseases such as dyslipidemia, diabetes mellitus, and obesity. We aimed 1. to investigate the factors that affect low-density lipoprotein cholesterol (LDL-C) and high-density lipoprotein cholesterol (HDL-C) in monozygotic (MZ) twins, and 2. to identify genes which expression levels changed in pairs with large differences in LDL-C or HDL-C levels.

Methods: The registered database at the Center for Twin Research, Osaka University, containing 263 pairs of MZ twins, was analyzed. 1. The effects of smoking, exercise, nutritional factors, and anthropometric and biochemical parameters on LDL-C or HDL-C levels were examined in MZ twins. 2. RNA sequencing in the peripheral blood mononuclear cells of 59 pairs was analyzed for large differences of LDL-C or HDL-C groups.

Results: 1. The ΔLDL-C levels were significantly associated with an older age, the ΔTG levels, and ΔBMI. ΔHDL-C levels were associated with the ΔBMI, ΔTG, ΔTP, and ΔLDL-C levels. The HDL-C levels were affected by smoking and exercise habits. The intakes of cholesterol and saturated fatty acids were not associated with the LDL-C or HDL-C levels. 2. An RNA sequencing analysis revealed that the expression of genes related to the TLR4 and IFNG pathways was suppressed in accordance with the HDL-C levels in the larger ΔHDL-C group among the 59 pairs.

Conclusion: We identified the factors affecting the LDL-C or HDL-C levels in monozygotic twins. In addition, some types of inflammatory gene expression in peripheral blood mononuclear cells were suppressed in accordance with the HDL-C levels, thus suggesting the importance of weight management and exercise habits in addition to dietary instructions to control the LDL-C or HDL-C levels.

See editorial vol. 31: 1515-1516

Introduction

The serum low-density lipoprotein cholesterol (LDL-C) or high-density lipoprotein cholesterol (HDL-C) levels vary with environmental factors such as dietary habits and physical activity^{1, 2)}. Previously, it was reported that LDL-C increased by consuming diets containing cholesterol³⁾ or saturated fatty acids⁴⁾ while it decreased by consuming diets containing unsaturated fatty acids⁵⁾ and dietary fiber⁶⁾. It has also been reported that the HDL-C levels decreased by smoking⁷⁾. However, almost all clinical studies have been conducted in unrelated individuals with different genetic backgrounds, thus making it difficult to determine which factors affecting the lipid profile might be genetic or environmental. Because monozygotic (MZ) twins share almost the same genes, twin research is a valuable modality for determining the factors associated with various diseases.

The Osaka University Center for Twin Research was the first established center to hold systematic twin omics information and promote twin research in Japan based on the Osaka University Twin Registry. As of March 2022, 363 MZ and 73 dizygotic twin pairs were registered⁸⁾. Registered data included physical, epidemiological, and clinical laboratory data. Importantly, peripheral blood mononuclear cell (PBMC)-derived RNA sequencing (RNA-seq) data have also recently been registered, which is the first database of an RNA-seq analysis in MZ twins. Therefore, we aimed to investigate the environmental determinants of the serum LDL-C or HDL-C levels and to identify differentially expressed genes (DEGs) in MZ pairs with large differences in LDL-C or HDL-C levels.

Methods

i. Subjects

Japanese MZ pairs were recruited from a registry established by the Center for Twin Research at Osaka University⁸⁾. Among 363 pairs, we excluded MZ pairs who had a history and treatment history of thyroid disease and a treatment history of dyslipidemia and who were women aged 45 to 56 years who may be at the onset of menopause⁹⁾. Finally, we examined 263 pairs (Table 1) and 59 pairs with peripheral blood mononuclear cell (PBMC)-derived RNA sequencing (RNA-seq) data (Table 2). Blood samples were collected at 9:00 am after a 12-hour fast. All twins were nonalcoholic. Twins were examined on the same day. The zygosity of the twin pairs was confirmed by perfect matching of 15 short tandem repeat (STR) loci using the PowerPlex 16 System (Promega, WI, USA).

Table 1.Clinical characteristics of 263 MZ pairs

	All MZ twins	Female	Male
No. of samples, n	526 (263 pairs)	364 (182 pairs)	162 (81 pairs)
Age, y	44.5±20.2 (27-64)	39.9±17.8 (26-57)	54.9±21.4 (34-73)
LDL-C, mg/dL	110.1±28.6 (89.0-128.2)	107.2±28.0 (86.8-124.6)	116.6±28.9 (97.5-133.8)
HDL-C, mg/dL	64.1±15.0 (53.1-73.7)	67.7±14.3 (56.6-76.6)	55.9±13.4 (47.0-63.0)
TG, mg/dL	81.4±50.9 (49.3-96.6)	70.6±39.4 (44.6-81.7)	105.5±64.2 (61.3-130.2)
BMI, kg/m2	21.4±3.3 (19.0-23.3)	20.6±2.7 (18.7-22.0)	23.2±3.7 (20.7-25.5)

Age, LDL-C, HDL-C, TG, and BMI were presented by mean±SD and interquartile range.

Table 2.Clinical characteristics of 59 MZ pairs with PBMC-derived RNA-seq data

	All MZ twins	Female	Male
No. of samples, n	118 (59 pairs)	76 (38 pairs)	42 (21 pairs)
Age, y	57.1±20.5 (40-74)	48.1±19.3 (30-64)	73.5±10.1 (69-82)
LDL-C, mg/dL	116.5±27.4(96.4-135.4)	116.9±28.12(95.5-131.8)	115.9±26.2(97.6-136.9)
HDL-C, mg/dL	64.2±17.3 (52.0-74.9)	68.7±16.2 (55.2-81.3)	56.0±16.4 (45.4-63.0)
TG, mg/dL	92.9±54.5(54.2-116.6)	83.1±48.9(50.9-103.5)	110.6±60.0(64.3-151.7)
BMI, kg/m2	21.9±3.1 (19.8-24.2)	21.1±2.6 (19.3-22.8)	23.4±3.3 (21.6-25.3)

Age, LDL-C, HDL-C, TG, and BMI were presented by mean±SD and interquartile range.

ii. Laboratory Tests

Thirteen parameters, including HbA1c, total protein (TP), uric acid (UA), creatinine (Cre), urea nitrogen (UN), glucose (Glu), triglyceride (TG), LDL-C, HDL-C, aspartate aminotransferase (AST), alkaline phosphatase (ALP), gamma-glutamyl transpeptidase (GGT), and creatine kinase (CK) were measured by Beckman Coulter, Inc., according to the method of the IFCC project¹⁰⁾.

iii. An RNA-Seq Analysis and Quality Control

Whole blood samples were frozen and stored in Tempus Blood RNA Tube (Thermo Fisher Scientific, CA, USA) after RNA stabilization. RNA was extracted from the stored samples using the Tempus Spin RNA Isolation Kit (Thermo Fisher Scientific, CA, USA), and ribosomal and globin RNA were depleted. Stranded total RNA libraries were prepared according to the manufacturer’s protocol for low-throughput experiments. The gene expression levels were measured using the Illumina NovaSeq 6000 system (Illumina Inc., CA, USA).

The fastq files containing the sequenced reads for each sample were downloaded, and quality control of the sequenced raw reads was analyzed using the FastQC program (v0.11.7). The overall read quality, total bases, total reads, GC (%), and basic statistics were calculated. To reduce any bias in the analysis, artifacts, such as low-quality reads, adaptor sequences, contaminant DNA, and PCR duplicates, were removed. Trimmed reads were mapped to the reference genome (GRCh38) using HISAT2 (v2.1.0)¹¹⁾, which is known to handle splice read mapping through the Browtie2 aligner (v2.3.4.1)¹²⁾. Transcripts were assembled using StringTie (v2.1.3b)¹³⁾ with aligned reads. The expression profiles were represented as normalization values, which were calculated based on the transcript length and depth of coverage. We used Transcripts Per Kilobase Million (TPM)¹⁴⁾ as a normalized value for the gene expression levels.

iv. Investigation of Smoking, Exercise, and Drinking Habits

Smoking, exercise, and drinking habits were assessed using Breslow’s Seven Healthy Habits questionnaire¹⁵⁾. Smoking habit options were 1) never smoker 2) daily smoker 3) former smoker, exercise habit options were 1) frequent exerciser 2) occasional exerciser 3) infrequent exerciser, and drinking habit options were 1) non-drinker 2) occasional drinker 3) daily drinker. To investigate the association between smoking habits and serum LDL-C or HDL-C levels, we examined 15 MZ pairs in which one of the options was 1) never smoker or 3) former smoker and 2) daily smoker using the Wilcoxon signed-rank test. In a similar analysis of exercise habits, we examined 23 MZ pairs in which one of the options was 3) infrequent exerciser and the other was 1) frequent exerciser. In the analysis of drinking habits, we examined 59 MZ pairs in which one of the options was 1) non-drinker and the other was 2) occasional drinker or 3) daily drinker.

v. Calculation of ΔLDL-C, ΔHDL-C, and other Parameters, and Determination of Environmental Factors

To investigate the environmental factors associated with the serum LDL-C or HDL-C levels in MZ twins, we used the differences in LDL-C (ΔLDL-C) and HDL-C (ΔHDL-C) levels. ΔLDL-C was calculated from the difference between high and low LDL-C levels in each pair, and ΔHDL-C levels were calculated in the same manner (Fig.1). Differences in the other parameters were calculated in the same direction as ΔLDL-C or ΔHDL-C levels. In stepwise linear regression analyses, the objective variables were log₁₀(ΔLDL-C+3) or log₁₀(ΔHDL-C+1.5) and the explanatory variables were age, sex, ΔTP, ΔUA, ΔCre, ΔUN, ΔGlu, ΔHbA1c, ΔTG, ΔLDL-C or ΔHDL-C, ΔAST, ΔALP, ΔGGT, ΔCK, ΔSBP, and ΔBMI. Stepwise linear regression analyses were performed using minimized Akaike’s Information Criterion (AIC). Nutrient intake was obtained from the brief-type self-administered diet history questionnaire (BDHQ)^{16, 17)}, and the differences were calculated in the same way. Correlations between the ΔLDL-C, ΔHDL-C levels, and the nutrient intake were confirmed using the Spearman’s rank correlation.

Fig.1.

Definitions of ΔLDL-C and ΔHDL-C and the methods for determining the environmental factors affecting the serum LDL-C or HDL-C levels

vi. Classification of LD Groups and Analyses Using PBMC-Derived RNA-Seq Data

To identify the DEGs associated with serum LDL-C or HDL-C levels in 59 MZ pairs with PBMC-derived RNA-seq data, the top 20 pairs of ΔLDL-C or ΔHDL-C distributions were designated as the Large Difference (LD) group, LDL-C_LD and HDL-C_LD groups, respectively. DEGs were genes with P＜0.05 and were identified using the DESeq2 package in R software. Fold changes in the LDL-C_LD group were calculated by log2 transformation by dividing the expression of high LDL-C levels by the expression of low LDL-C levels in each pair, and fold changes in the HDL-C_LD group were calculated in the same manner. Enrichment analyses were performed using STRING (https://string-db.org/) and IPA (QIAGEN), and heatmaps were plotted using the R software program. In each heatmap, 59 pairs were aligned on the horizontal axis in the order of ΔLDL-C or ΔHDL-C levels, and the related genes were listed on the vertical axis. Each cell indicates the fold-change in the genes. In addition, immune response-related genes and TLR4- or IFNG-regulated genes were analyzed using Spearman’s correlation with HDL-C adjusted for smoking habits, exercise habits, BMI, TG, TP, and LDL-C. Within-pair differences in adjusted HDL-C (Δadjusted HDL-C) were also determined with the differences in those genes in the same direction as Δadjusted HDL-C. Genes with │ρ│≧0.3 were designated as positively or negatively correlated genes with adjusted HDL-C or Δadjusted HDL-C, respectively.

vii. Statistical Analyses

The Wilcoxon signed-rank test, linear regression analyses, and Spearman’s rank correlation were performed using the JMP Pro software program 17.0.0 (SAS Institute Japan Co., Ltd., Tokyo, Japan).

Results

Distribution and Difference in the Serum LDL-C and HDL-C Levels in MZ Pairs

First, we examined the distribution of the serum lipid profiles and the differences in the serum lipid profiles. In this study, 15.4% of the subjects (81 subjects) had LDL-C levels greater than 140 mg/dL (Fig.2A), 2.9% (15 subjects) had HDL-C levels less than 40 mg/dL (Fig.2B), and 8.6% (45 subjects) had TG levels greater than 150 mg/dL (Fig.2D). Interestingly, 17.5% of the pairs (46 pairs) had ΔLDL-C levels greater than 30 mg/dL (Fig.2E), and 16.7% (44 pairs) had ΔHDL-C levels greater than 15 mg/dL (Fig.2F) in MZ pairs that shared almost the same genes. Of the 46 pairs with ΔLDL-C levels greater than 30 mg/dL and 44 pairs with ΔHDL-C levels greater than 15 mg/dL, 11 pairs belonged to both groups. Regarding TG, the distribution of ΔTG levels was smaller than expected. These results suggest that environmental factors had caused variations in the serum LDL-C and HDL-C levels in MZ twins.

Fig.2. Distributions of the serum lipid profiles and Δ serum lipid profiles

Distribution of serum (A) LDL-C, (B) HDL-C, (C) TC, and (D) TG in 526 MZ twins. The orange bars indicate subjects with LDL-C greater than 140 mg/dL, HDL-C less than 40 mg/dL, or TG greater than 150 mg/dL. The distribution of absolute values of (E) ΔLDL-C, (F) ΔHDL-C, (G) ΔTC, and (H) ΔTG in 263 MZ pairs. Orange bars indicate MZ pairs with ΔLDL-C greater than 30 mg/dL or ΔHDL-C greater than 15 mg/dL

Smoking and Exercise Habits Affected the Serum HDL-C Levels in MZ Twins

Previous studies have reported that the serum LDL-C or HDL-C levels are influenced by smoking, exercise, and drinking habits^{1, 2, 7, 18, 19)}. Therefore, we investigated the association between these habits and the LDL-C or HDL-C levels in MZ pairs using the Wilcoxon signed-rank test. There were 15 pairs of a never/former smoker and a smoker, 23 pairs of an infrequent exerciser and a frequent exerciser, and 53 pairs of a non-drinker and a drinker. LDL-C levels were not significantly associated with any of these habits (Fig.3A, B, C). Interestingly, the HDL-C level significantly decreased in smokers (Fig.3D). Furthermore, the HDL-C levels were significantly elevated in frequent exercisers (Fig.3E). Drinking habits were not associated with the HDL-C levels (Fig.3F). These results suggest that smoking and exercise habits affect the serum HDL-C levels in MZ twins.

Fig.3. Elucidation of the associations between the serum LDL-C, HDL-C levels and lifestyle habits by the Wilcoxon signed-rank test

The serum LDL-C levels of MZ pairs with different (A) smoking (15 pairs), (B) exercise (23 pairs), or (C) drinking habits (53 pairs). Serum HDL-C levels of MZ pairs with different (D) smoking (15 pairs), (E) exercise (23 pairs), or (F) drinking habits (53 pairs).

ΔLDL-C was Associated with Aging, ΔTG, and ΔBMI and ΔHDL-C was Associated with ΔBMI, ΔTG, ΔTP, and ΔLDL-C in MZ Pairs

Next, we investigated the anthropometric and clinical laboratory parameters associated with the LDL-C or HDL-C levels using stepwise linear regression analyses for ΔLDL-C or ΔHDL-C levels. ΔLDL-C levels were significantly associated with age, ΔTG, and ΔBMI (Table 3). In the same analysis, ΔHDL-C levels were significantly associated with ΔBMI, ΔTG, ΔTP, and ΔLDL-C levels (Table 4). These results suggest that the TG levels and weight gain were associated with changes in LDL-C and HDL-C levels in the MZ pairs.

Table 3.Linear regression analysis for ΔLDL-C

Explanatory variable	β (95%CI)	p value
Age	0.003 (0.001, 0.005)	＜0.001
ΔTG	0.001 (0.0002, 0.002)	0.011
ΔBMI	0.023 (0.004, 0.041)	0.013
ΔCre	0.389 (-0.029, 0.807)	0.069
ΔSBP	-0.002 (-0.004, 0.0006)	0.143

Table 4.Linear regression analysis for ΔHDL-C

Explanatory variable	β (95%CI)	p value
ΔBMI	-0.041 (-0.059, -0.023)	＜0.001
ΔTG	-0.002 (-0.003, -0.001)	＜0.001
ΔTP	0.157 (0.061, 0.262)	0.003
ΔLDL-C	0.003 (0.001, 0.004)	0.004
ΔALP	-0.0004 (-0.001, 0.0002)	0.120

Previously, it was reported that some nutrients were associated with the serum LDL-C or HDL-C levels; therefore, we investigated the nutrient intake data obtained from the BDHQ method^{16, 17)} using Spearman’s rank correlation. Neither LDL-C nor HDL-C was associated with any nutrient intake, such as cholesterol, saturated fatty acids, unsaturated fatty acids, or dietary fiber intake (Supplemental Table 1 and 2).

Supplemental Table 1.Spearman’s rank correlation between ΔLDL-C and Δ nutrient intakes

ΔLDL-C vs	Spearman’s rank correlation coefficient	p value
Δenergy	0.04	0.47
Δprotein	0.06	0.30
Δfat	0.07	0.23
Δsaturated fatty acids	0.04	0.51
Δmono-unsaturated fatty acids	0.08	0.22
Δpoly-unsaturated fatty acids	0.09	0.13
Δcarbohydrate	0.01	0.90
Δash	0.05	0.44
Δretinol	0.07	0.25
Δβ carotene	0.03	0.62
Δvitamin A	0.04	0.51
Δvitamin D	0.01	0.93
Δvitamin E	0.06	0.33
Δvitamin K	0.08	0.22
Δvitamin B1	0.06	0.32
Δvitamin B2	0.04	0.54
Δniacin	0.06	0.36
Δvitamin B6	0.02	0.70
Δvitamin B12	0.02	0.80
Δfolic acid	0.04	0.52
Δpantothenic acid	0.05	0.39
Δvitamin C	0.05	0.45
Δcholesterol	0.01	0.84
Δfiber	0.04	0.52
Δsalt	0.05	0.39
Δsugar	0.06	0.32
Δalcohol	-0.01	0.85
Δdaidzein	0.06	0.31
Δgenistein	0.06	0.32

Supplemental Table 2.Spearman’s rank correlation between ΔHDL-C and Δ nutrient intakes

ΔHDL-C vs	Spearman’s rank correlation coefficients	p value
Δenergy	0.06	0.30
Δprotein	0.01	0.82
Δfat	-0.04	0.51
Δsaturated fatty acids	-0.01	0.93
Δmono-unsaturated fatty acids	-0.05	0.43
Δpoly-unsaturated fatty acids	-0.05	0.39
Δcarbohydrate	0.07	0.22
Δash	-0.01	0.90
Δretinol	0.02	0.73
Δβ carotene	-0.03	0.63
Δvitamin A	-0.04	0.57
Δvitamin D	0.06	0.32
Δvitamin E	-0.04	0.54
Δvitamin K	-0.04	0.47
Δvitamin B1	0.01	0.93
Δvitamin B2	-0.03	0.64
Δniacin	0.05	0.41
Δvitamin B6	0.02	0.79
Δvitamin B12	0.06	0.34
Δfolic acid	-0.07	0.29
Δpantothenic acid	0.01	0.87
Δvitamin C	-0.02	0.69
Δcholesterol	-0.03	0.60
Δfiber	-0.03	0.69
Δsalt	-0.01	0.92
Δsugar	-0.01	0.87
Δalcohol	0.12	0.05
Δdaidzein	＜-0.01	0.99
Δgenistein	＜-0.01	1.00

RNA-Seq Analyses of PBMCs in the LDL_LD Group

To explore the influence of serum lipids on leukocyte gene expression, RNA-seq analyses of PBMCs were performed using 59 MZ pairs. The top 20 pairs (1/3) of ΔLDL-C or ΔHDL-C distributions were designated as the LDL-C_LD and HDL-C_LD groups, respectively. In the LDL-C_LD group, ΔLDL-C levels ranged from 28.9 to 81.8 mg/dL (Fig.4A, B). In the LDL-C_LD group, 512 genes were upregulated and 488 genes were downregulated (Fig.4C). A GO analysis revealed that cellular protein metabolic process-related genes were upregulated, and cellular nitrogen compound metabolic process-related genes were downregulated (Fig.4D, Supplemental Tables 3 and 4). Furthermore, an upstream regulator analysis was performed using IPA and plotted as heat maps (Fig.4E, F). In each heatmap, 59 pairs were aligned on the horizontal axis in the order of ΔLDL-C levels, and the related genes were listed on the vertical axis. Each cell indicates the fold-change in the genes. PTEN-regulated genes were upregulated in the pairs framed in black, which were in the LDL-C_LD group. Similarly, MYC-regulated genes were downregulated in the LDL-C_LD group.

Fig.4. RNA-seq analyses in the LDL-C_LD group

(A) The distribution of ΔLDL-C in 59 MZ pairs and the top 20 pairs (1/3 of the distribution) were designated as the LDL-C_LD group used in RNA-seq analyses (orange). (B) Characteristics of the LDL-C_LD group. (C) A volcano plot showing differentially expressed genes (DEGs) in those with higher LDL-C compared to those with lower LDL-C in MZ pairs. The upregulated genes (red), and the downregulated genes (blue) with P＜0.05. (D) A GO analysis using upregulated DEGs (upper, red) and downregulated DEGs (lower, blue). Statistical analyses were carried out using FDR correction. A default FDR ＜0.05 was considered statistically significant. (E) Heatmaps of PTEN-regulated genes and (F) MYC-regulated genes. In each heatmap, 59 pairs were aligned on the horizontal axis in order of ΔLDL-C and related genes were listed on the vertical axis. Each cell indicated the fold change of the genes.

Supplemental Table 3.LDL-C_LD group upregulated genes GO Process TOP 10 terms

GO term (term ID)	Observed count	-log10FDR
Cellular protein modification process (GO:0006464)	112	5.46
Cellular protein metabolic process (GO:0044267)	127	5.46
Macromolecule modification (GO:0043412)	113	4.69
Protein metabolic process (GO:0019538)	133	3.80
Ubiquitin-dependent protein catabolic process (GO:0006511)	32	3.55
Protein catabolic process (GO:0030163)	37	3.51
Organonitrogen compound metabolic process (GO:1901564)	154	3.51
Proteolysis involved in cellular protein catabolic process (GO:0051603)	33	3.22
Cellular protein catabolic process (GO:0044257)	34	3.21
Protein modification by small protein conjugation or removal (GO:0070647)	46	3.17

Supplemental Table 4.LDL-C_LD group downregulated genes GO Process TOP 10 terms

GO term (term ID)	Observed count	-log10FDR
Cellular nitrogen compound metabolic process (GO:0034641)	164	21.08
Cellular aromatic compound metabolic process (GO:0006725)	145	18.12
Nucleobase-containing compound metabolic process (GO:0006139)	137	17.77
Translation (GO:0006412)	48	17.63
Heterocycle metabolic process (GO:0046483)	142	17.63
RNA catabolic process (GO:0006401)	40	17.20
Organic cyclic compound metabolic process (GO:1901360)	149	17.20
RNA metabolic process (GO:0016070)	99	17.07
Gene expression (GO:0010467)	115	17.00
Peptide metabolic process (GO:0006518)	54	16.95

RNA-Seq Analyses of PBMCs in the HDL_LD Group

Similarly, in the HDL-C_LD group, ΔHDL-C ranged from 11.3 to 37.2 mg/dL (Fig.5A, B). In the HDL-C_LD group, 645 genes were upregulated and 478 genes were downregulated (Fig.5C). A GO analysis revealed that cell cycle-related genes were upregulated, and immune response-related genes were downregulated (Fig.5D, Supplemental Table 5 and 6). As in the LDL-C_LD group, an upstream regulator analysis was performed. TLR4-regulated and IFNG-regulated genes were clearly downregulated in the HDL-C_LD group (Fig.5E, F), thus suggesting that subjects with lower HDL-C levels had a higher expression of inflammatory genes than subjects with higher HDL-C levels. To further investigate the association between “HDL-C levels” and immune-related genes, we additionally performed Spearman’s rank correlation of immune-related genes, which were identified in “Immune response” in a GO Process analysis (Fig.5D), “TLR4-regulated genes” (Fig.5E), and “IFNG-regulated genes” (Fig.5F), by applying HDL-C after adjusting for smoking habits, exercise habits, BMI, TG, TP, and LDL-C levels in the HDL-C_LD group. Importantly, 36 genes, including IL2RG, CD44, and STAT3, were negatively correlated with the adjusted HDL-C levels (Supplemental Table 7). Furthermore, we performed the same analysis with “within-pair differences in adjusted HDL-C” (Δadjusted HDL-C) and differences in immune-related genes; 23 genes were negatively correlated with Δadjusted HDL-C (Supplemental Table 8). Through both analyses, five genes, TNFAIP3, NLRC5, PRDM1, ACSS1, and CDC25B, were identified. Even after adjusting for smoking habits, exercise habits, BMI, TG, TP, and LDL-C levels, we confirmed that inflammatory gene expression in PBMC was more suppressed in the subject with higher HDL-C levels in each pair.

Fig.5. RNA-seq analyses in the HDL-C_LD group

(A) The distribution of ΔHDL-C in 59 MZ pairs and the top 20 pairs (1/3 of the distribution) were designated as the HDL-C_LD group used in RNA-seq analyses (orange). (B) Characteristics of the HDL-C_LD group. (C) A volcano plot showing differentially expressed genes (DEGs) in those with higher HDL-C compared to those with lower HDL-C in MZ pairs. The upregulated genes (red), and the downregulated genes (blue) with P＜0.05. (D) A GO analysis using upregulated DEGs (upper, red) and downregulated DEGs (lower, blue). Statistical analyses were carried out using FDR correction. A default FDR ＜0.05 was considered to be statistically significant. (E) Heatmaps of TLR4-regulated genes and (F) IFNG-regulated genes. In each heatmap, 59 pairs were aligned on the horizontal axis in order of ΔHDL-C and related genes were listed on the vertical axis. Each cell indicated the fold change of the genes.

Supplemental Table 5.HDL-C_LD group upregulated genes GO Process TOP 10 terms

GO term (term ID)	Observed count	-log10FDR
Cell cycle (GO:0007049)	84	6.92
Cell cycle process (GO:0022402)	67	6.21
Mitotic cell cycle process (GO:1903047)	49	5.72
Mitotic cell cycle (GO:0000278)	51	5.09
Organelle organization (GO:0006996)	151	4.20
Cell division (GO:0051301)	38	3.70
Cotranslational protein targeting to membrane (GO:0006613)	16	3.64
Regulation of cell cycle (GO:0051726)	68	3.40
Protein localization to endoplasmic reticulum (GO:0070972)	18	3.25
Chromosome segregation (GO:0007059)	25	3.08

Supplemental Table 6.HDL-C_LD group downregulated genes GO Process TOP 10 terms

GO term (term ID)	Observed count	-log10FDR
Immune response (GO:0006955)	80	9.25
Immune system process (GO:0002376)	106	9.25
Immune effector process (GO:0002252)	56	7.82
Innate immune response (GO:0045087)	44	6.56
Interspecies interaction between organisms (GO:0044419)	81	6.54
Cellular nitrogen compound metabolic (GO:0034641)	116	5.47
Nitrogen compound metabolic process (GO:0006807)	196	5.47
Response to other organism (GO:0051707)	59	5.22
Response to biotic stimulus (GO:0009607)	60	4.68
Defense response to other organism (GO:0098542)	47	4.55

Supplemental Table 7.Spearman’s rank correlation between adjusted HDL-C and gene expressions

Gene	Spearman’s rank correlation coefficient	p value
TNFAIP3	-0.515	0.001
GBF1	-0.488	0.001
TAPBP	-0.472	0.002
PRKD2	-0.466	0.002
DIAPH1	-0.457	0.003
IL2RG	-0.452	0.003
NLRC5	-0.445	0.004
CD44	-0.442	0.004
CNN2	-0.432	0.005
HUWE1	-0.432	0.005
PRDM1	-0.430	0.006
RNF19B	-0.418	0.007
HSPA5	-0.414	0.008
SF3A1	-0.412	0.008
TRIM28	-0.382	0.015
PKM	-0.373	0.018
STAT3	-0.373	0.018
PML	-0.371	0.018
DYNC1H1	-0.371	0.018
ADA2	-0.364	0.021
HSP90AA1	-0.360	0.023
CD74	-0.359	0.023
NCSTN	-0.358	0.024
PRKCB	-0.355	0.026
ANKHD1	-0.352	0.026
CASP2	-0.351	0.026
TRIM52	-0.345	0.029
SPTAN1	-0.344	0.030
PDXK	-0.337	0.034
ACSS1	-0.330	0.038
PRPF8	-0.328	0.039
SKAP1	-0.328	0.039
CDC25B	-0.326	0.040
CCND2	-0.322	0.043
PLCG2	-0.3148	0.048

Supplemental Table 8.Spearman’s rank correlation between Δadjusted HDL-C and Δgene expressions

Gene	Spearman’s rank correlation coefficient	p value
ΔSLAMF7	-0.570	0.009
ΔZCCHC3	-0.478	0.033
ΔACSS1	-0.462	0.041
ΔXBP1	-0.457	0.043
ΔTUBB	-0.448	0.048
ΔPRDM1	-0.430	0.058
ΔGBP4	-0.424	0.062
ΔCERS5	-0.415	0.069
ΔCDC25B	-0.387	0.092
ΔGM2A	-0.377	0.101
ΔPSMD2	-0.371	0.107
ΔMR1	-0.364	0.115
ΔGBP3	-0.347	0.133
ΔRFX5	-0.343	0.139
ΔSEC61A1	-0.343	0.139
ΔLDHA	-0.341	0.141
ΔBTN3A3	-0.320	0.169
ΔNFKB1	-0.317	0.173
ΔTNFAIP3	-0.314	0.177
ΔOAS2	-0.313	0.179
ΔLCP1	-0.310	0.184
ΔHLA-DQA1	-0.305	0.191
ΔNLRC5	-0.301	0.198

Discussion

Although previous studies have demonstrated that smoking, exercise habits, and some nutrient factors may affect the LDL-C or HDL-C levels, few studies have so far been performed in twin individuals^1-7). Here, we confirmed that a smoker had lower HDL-C levels than a never smoker or a former smoker in each of the 15 pairs. In addition, a frequent exerciser had higher HDL-C levels than an infrequent exerciser in each of the 23 pairs. Regarding anthropometric and clinical laboratory parameters, pairs with large differences in the LDL-C levels were significantly associated with age (older pairs), ΔTG, and ΔBMI. Similarly, pairs with a large difference in the HDL-C levels showed significant associations with ΔBMI and ΔTG, thus indicating that body weight management is crucial for maintaining favorable lipid profiles. Regarding nutrient factors, we could not detect any significant associations in this study. We speculate that this may be associated with the method of collecting nutritional data and the characterization of this study itself. In the current study, nutrient intake data were obtained using the BDHQ method, which is characterized by a one-month recording of the frequency and content of food consumption^{16, 17)}. While this method is convenient for assessing dietary habits in a large-scale study owing to its simplicity and low cost, there may be some possibilities of systematic and reporting errors due to the structure of the questionnaire. Therefore, we speculate that we could have had more samples to detect previously reported nutrients. Moreover, we may not be able to detect associations if the twins have similar food preferences.

Next, for the first time, we performed RNA-seq on PBMCs from 59 MZ pairs. Interestingly, we found that the upregulation of PTEN-regulated genes was observed in the LDL-C_LD group. It has been reported that PTEN is a negative regulator of insulin signaling²⁰⁾, thus suggesting that high LDL-C levels was associated with enhancement of insulin resistance. In addition, in the HDL-C_LD group, TLR4-regulated and IFNG-regulated inflammatory genes were widely suppressed in one subject with lower HDL-C levels compared to the other with higher HDL-C levels. HDL can efflux cholesterol. In a clinical study by Khera AV, et al., it was reported that the cholesterol efflux capacity of HDL from macrophages had a strong inverse association with atherosclerosis²¹⁾. Furthermore, it has been reported that patients with Tangier disease, which is characterized by the absence or extremely low HDL-C levels, display premature systemic atherosclerosis^{22, 23)}. An impairment in the cholesterol efflux leads to cellular cholesterol accumulation and inflammatory response^{24, 25)}. Although the HDL-C levels were not as low as in those with Tangier disease, our findings suggest that lower HDL-C levels were associated with higher inflammatory gene expression in two subjects sharing the same genetic background.

This study is associated with several limitations. First, we did not perform analyses by sex because the sample size was insufficient. Although we excluded MZ pairs from women aged 45-56 years, who may have been at the onset of menopause, there is a possibility that we failed to detect the influence of female hormones. Second, RNA-seq was performed on 59 of 263 pairs. The limitation of the number of RNA-seq analyses may have prevented us from performing a proper analysis. Finally, since the analysis was performed only in Japanese, our results may not be universally applicable to other ethnic groups.

Conclusions

The serum HDL-C levels were significantly associated with smoking and exercise, and the LDL-C and HDL-C levels were associated with the BMI and serum TG levels in monozygotic twins. In addition, some types of inflammatory gene expression in peripheral blood mononuclear cells were suppressed in accordance with the HDL-C levels, thus suggesting the importance of weight management in addition to dietary instructions to control LDL-C or HDL-C levels.

Acknowledgements

The authors are thankful to Beckman Coulter, Inc. (Tokyo, JAPAN) for their collaborative work and Kumiko Furukawa for their technical assistance.

The members of the Osaka Twin Research Group: Norio Sakai, Masanori Takahashi, Teiji Nishio, Kei Kamide, Shinji Kihara, Hiroko Watanabe, Mikio Watanabe, and Dousatsu Sakata, Center for Twin Research, Osaka University Graduate School of Medicine.

Conflicts of Interest

We all declare no competing interests.

Funding

This research was supported by a grant from the Japan Agency for Medical Research and Development (AMED), grant number J230705536, and the Japan Society for the Promotion of Science (JSPS) KAKENHI grants, grant numbers 19H04048, 21K08323, 22H02967, 22K08670, 23K07968 from the Japan Society for the Promotion of Science.

Abbreviations

LDL-C; low-density lipoprotein cholesterol

HDL-C; high-density lipoprotein cholesterol

TC; total cholesterol

TG; triglyceride

RNA-seq; RNA sequencing

References

• 1)  Kelley GA, Kelley KS and Tran ZV: Aerobic exercise and lipids and lipoproteins in women: a meta-analysis of randomized controlled trials. J Womens Health (Larchmt), 2004; 13: 1148-1164
• 2)  Kelley GA and Kelley KS: Aerobic exercise and lipids and lipoproteins in men: a meta-analysis of randomized controlled trials. J Mens Health Gend, 2006; 3: 61-70
• 3)  Zhong VW, Van Horn L, Cornelis MC, Wilkins JT, Ning H, Carnethon MR, Greenland P, Mentz RJ, Tucker KL, Zhao L, Norwood AF, Lloyd-Jones DM and Allen NB: Associations of Dietary Cholesterol or Egg Consumption With Incident Cardiovascular Disease and Mortality. JAMA, 2019; 321: 1081-1095
• 4)  Nakamura Y, Okuda N, Turin TC, Fujiyoshi A, Okamura T, Hayakawa T, Yoshita K, Miura K, Ueshima H and Group NDR: Fatty acids intakes and serum lipid profiles: NIPPON DATA90 and the national nutrition monitoring. J Epidemiol, 2010; 20 Suppl 3: S544-548
• 5)  Vafeiadou K, Weech M, Altowaijri H, Todd S, Yaqoob P, Jackson KG and Lovegrove JA: Replacement of saturated with unsaturated fats had no impact on vascular function but beneficial effects on lipid biomarkers, E-selectin, and blood pressure: results from the randomized, controlled Dietary Intervention and VAScular function (DIVAS) study. Am J Clin Nutr, 2015; 102: 40-48
• 6)  Brown L, Rosner B, Willett WW and Sacks FM: Cholesterol-lowering effects of dietary fiber: a meta-analysis. Am J Clin Nutr, 1999; 69: 30-42
• 7)  Gepner AD, Piper ME, Johnson HM, Fiore MC, Baker TB and Stein JH: Effects of smoking and smoking cessation on lipids and lipoproteins: outcomes from a randomized clinical trial. Am Heart J, 2011; 161: 145-151
• 8)  Honda C, Watanabe M, Tomizawa R, Osaka Twin Research G and Sakai N: Update on Osaka University Twin Registry: An Overview of Multidisciplinary Research Resources and Biobank at Osaka University Center for Twin Research. Twin Res Hum Genet, 2019; 22: 597-601
• 9)  Tamada T and Iwasaki H: Age at natural menopause in Japanese women. Nihon Sanka Fujinka Gakkai Zasshi, 1995; 47: 947
• 10)  Ichihara K: Statistical considerations for harmonization of the global multicenter study on reference values. Clin Chim Acta, 2014; 432: 108-118
• 11)  Kim D, Paggi JM, Park C, Bennett C and Salzberg SL: Graph-based genome alignment and genotyping with HISAT2 and HISAT-genotype. Nat Biotechnol, 2019; 37: 907-915
• 12)  Langmead B and Salzberg SL: Fast gapped-read alignment with Bowtie 2. Nat Methods, 2012; 9: 357-359
• 13)  Pertea M, Pertea GM, Antonescu CM, Chang TC, Mendell JT and Salzberg SL: StringTie enables improved reconstruction of a transcriptome from RNA-seq reads. Nat Biotechnol, 2015; 33: 290-295
• 14)  Li B, Ruotti V, Stewart RM, Thomson JA and Dewey CN: RNA-Seq gene expression estimation with read mapping uncertainty. Bioinformatics, 2010; 26: 493-500
• 15)  Belloc NB and Breslow L: Relationship of physical health status and health practices. Prev Med, 1972; 1: 409-421
• 16)  Sasaki S, Yanagibori R and Amano K: Self-administered diet history questionnaire developed for health education: a relative validation of the test-version by comparison with 3-day diet record in women. J Epidemiol, 1998; 8: 203-215
• 17)  Otsuka R, Tamakoshi K, Yatsuya H, Murata C, Sekiya A, Wada K, Zhang HM, Matsushita K, Sugiura K, Takefuji S, OuYang P, Nagasawa N, Kondo T, Sasaki S and Toyoshima H: Eating fast leads to obesity: findings based on self-administered questionnaires among middle-aged Japanese men and women. J Epidemiol, 2006; 16: 117-124
• 18)  Nakamura M, Yamamoto Y, Imaoka W, Kuroshima T, Toragai R, Ito Y, Kanda E, E JS and Ai M: Relationships between Smoking Status, Cardiovascular Risk Factors, and Lipoproteins in a Large Japanese Population. J Atheroscler Thromb, 2021; 28: 942-953
• 19)  Thornton J, Symes C and Heaton K: Moderate alcohol intake reduces bile cholesterol saturation and raises HDL cholesterol. Lancet, 1983; 2: 819-822
• 20)  Nakashima N, Sharma PM, Imamura T, Bookstein R and Olefsky JM: The tumor suppressor PTEN negatively regulates insulin signaling in 3T3-L1 adipocytes. J Biol Chem, 2000; 275: 12889-12895
• 21)  Khera AV, Cuchel M, de la Llera-Moya M, Rodrigues A, Burke MF, Jafri K, French BC, Phillips JA, Mucksavage ML, Wilensky RL, Mohler ER, Rothblat GH and Rader DJ: Cholesterol efflux capacity, high-density lipoprotein function, and atherosclerosis. N Engl J Med, 2011; 364: 127-135
• 22)  Koseki M, Yamashita S, Ogura M, Ishigaki Y, Ono K, Tsukamoto K, Hori M, Matsuki K, Yokoyama S and Harada-Shiba M: Current Diagnosis and Management of Tangier Disease. J Atheroscler Thromb, 2021; 28: 802-810
• 23)  Muratsu J, Koseki M, Masuda D, Yasuga Y, Tomoyama S, Ataka K, Yagi Y, Nakagawa A, Hamada H, Fujita S, Hattori H, Ohama T, Nishida M, Hiraoka H, Matsuzawa Y and Yamashita S: Accelerated Atherogenicity in Tangier Disease. J Atheroscler Thromb, 2018; 25: 1076-1085
• 24)  Li Y, Schwabe RF, DeVries-Seimon T, Yao PM, Gerbod-Giannone MC, Tall AR, Davis RJ, Flavell R, Brenner DA and Tabas I: Free cholesterol-loaded macrophages are an abundant source of tumor necrosis factor-alpha and interleukin-6: model of NF-kappaB- and map kinase-dependent inflammation in advanced atherosclerosis. J Biol Chem, 2005; 280: 21763-21772
• 25)  Koseki M, Hirano K, Masuda D, Ikegami C, Tanaka M, Ota A, Sandoval JC, Nakagawa-Toyama Y, Sato SB, Kobayashi T, Shimada Y, Ohno-Iwashita Y, Matsuura F, Shimomura I and Yamashita S: Increased lipid rafts and accelerated lipopolysaccharide-induced tumor necrosis factor-alpha secretion in Abca1-deficient macrophages. J Lipid Res, 2007; 48: 299-306