Influences of Vitamin D Levels and Vitamin D-Binding Protein Polymorphisms on Nonalcoholic Fatty Liver Disease Risk in a Chinese Population

Abstract

Introduction

Vitamin D-binding protein (VDBP) is correlated with nonalcoholic fatty liver disease (NAFLD) through the biological functions of regulating plasma vitamin D (VD) level and the inflammatory process.

Objective

This study aims to investigate the effects of VD level and VDBP gene polymorphisms on the risk of NAFLD in a Chinese population.

Methods

Plasma 25-hydroxyvitamin D₃ levels were measured and seven VDBP candidate genetic variants (rs222020, rs2282679, rs4588, rs1155563, rs7041, rs16847024, rs3733359) were genotyped among participants in this case-control study. The control group was frequency-matched to the NAFLD case group by age and gender. Correlation analysis and multiple linear regressions were used to screen determinants of 25-hydroxyvitamin D₃ levels. Multivariable unconditional logistic regression was performed to estimate odds ratio (OR) and 95% confidence interval (95% CI). The prediction capability of models containing independent factors was estimated by the area under the receiver operating characteristic curve and Hosmer-Lemeshow test.

Results

Age, body mass index, and triacylglycerol were independent factors influencing VD levels. Participants with low VD levels had significantly higher prevalence of NAFLD compared to subjects with normal VD levels (p < 0.001). A low VD level contributed to increased the risk of NAFLD, independent of metabolic factors known to affect VD levels (adjusted OR = 2.282, 95% CI = 1.422–3.661, p = 0.001). Logistic regression analysis showed that individuals carrying rs7041-G allele had a significantly decreased the risk of NAFLD occurrence compared to T allele (additive model: adjusted OR = 0.814, 95% CI = 0.713–0.929, p = 0.002; codominant model: adjusted OR = 0.623, 95% CI = 0.449–0.866, p = 0.005), after adjusting for age, gender, and overweight. Stratification by multiple metabolic disorders did not alter this relationship. Moreover, we developed a simple model including age, gender, metabolic disorders, and VDBP single nucleotide polymorphism (SNP) to assess NAFLD risk, an AUC of which being 0.817, significantly higher than the model not included VDBP SNP, with Hosmer-Lemeshow test fitting well (p = 0.182).

Conclusions

Low plasma VD levels may increase susceptibility to NAFLD, while rs7041-G allele in VDBP contributed to a decreased NAFLD risk among Chinese population. The VDBP variant significantly improved the capability for NAFLD risk assessment, which could be used for early screening and management of NAFLD.

Introduction

Nonalcoholic fatty liver disease (NAFLD) has been the most common chronic liver disease, affecting a quarter of the global population approximately [1]. Individuals with NAFLD are generally asymptomatic and have increased risk of liver-related, extrahepatic, and all-cause mortality [2], resulting in a huge health burden. Notably, NAFLD is shaped by multiple interactions between components of metabolic syndrome, environmental risk factors, and inherited susceptibility [3]. Numerous studies had proved that adipose tissue inflammation [4], gut microbiome [5], and genetic pathways [6, 7] were correlated with NAFLD. Considering the complex pathogenesis of this disease, it is essential to explore the underlying etiological mechanisms of NAFLD occurrence and progression.

Accumulating evidence indicated that vitamin D (VD) can not only play a classical role in skeletal homeostasis, but also act as a regulator in improving adipose tissue inflammation [8], liver fibrogenesis [9], hepatic aberrant fat accumulation [10], and insulin resistance [11]. It has been hypothesized that a low level of VD may be an environmental risk factor for NAFLD [12]. Previous studies indicated that low level of VD was an independent factor for the occurrence and severity of NAFLD [13, 14, 15]. However, recent studies reported contradictory findings that no significant association existed between plasma VD level and risk of NAFLD in different populations [16, 17]. The effect of low VD levels on NAFLD risk remains unclarified.

Serum VD originates from dietary intake and sun exposure, affected by environmental factors like season, geographic latitude, etc. However, only about one-quarter of interindividual difference in VD status was accounted for environmental factors [18, 19]. Classical family and twin studies identified that gene heritability contributed significantly to the variability of circulating VD concentration, with gene effect accounting for 22%–43% approximately [18, 19, 20]. Thus, variants of genes involved in the VD metabolic pathway are likely to affect VD status.

In large genome-wide association studies, it has been established that genetic polymorphisms of VD-binding protein (VDBP), a leading plasma carrier for VD and its metabolites in metabolic process, were correlated with serum VD concentrations [21, 22, 23]. Apart from transport role for VD, VDBP also has immune biologic functions like actin scavenging, macrophage activation, and complement C5a chemotaxis enhancement [24], and has been applied as a biomarker of liver fibrosis [25]. Furthermore, previous researches demonstrated the associations between single nucleotide polymorphisms (SNPs) in VDBP genes and liver diseases, such as hepatocellular carcinoma [26] and hepatitis C [27]. However, evidence regarding the joint effects of plasma VD and VDBP SNPs on NAFLD risk, a stress-related metabolic disease, is scant. Consequently, the aim of this study was to investigate the potential associations among plasma VD level, VDBP gene polymorphisms, and NAFLD risk in a Chinese population, attempting to explain the environmental and genetic etiology, and to further provide new ideas for the prevention and treatment of NAFLD.

Materials and Methods

Study Participants and Design

This study was a case-control study that recruited participants who underwent a physical examination in a community hospital during July to September 2018 in Nanjing (Jiangsu, China). The diagnostic criterion for NAFLD case was accorded with the “Guideline of prevention and treatment for nonalcoholic fatty liver disease: a 2018 update” [28]: (1) evidence of diffuse hepatic steatosis determined by imaging or histology; (2) in the absence of secondary causes of hepatic fat accumulation, such as heavy alcohol consumption, hepatitis C, or medication use. NAFLD controls were frequency matched to cases by 6-year age bands and gender, aiming to enable the matching factors have similar constituent ratios between the case and control groups. We excluded subjects with (1) history of viral hepatitis, autoimmune hepatitis, or other liver diseases; (2) history of excessive alcohol consumption (≥30 g/day for males and ≥20 g/day for females); (3) presence of acute or chronic gastrointestinal diseases, severe multisystem illness, or malignancy; (4) history of liver transplant within the past year or complications of advanced liver disease; (5) history of mental disorders.

A sample size calculation was conducted at the beginning of the research. We presumed the mutation frequency of VDBP gene was 20% in general population referred to the previous studies, gene-induced odds ratio (OR) was 1.5, the significant level α exceeded 5% and the power of test 1-β reached 80%. After calculated by NCSS-PASS (version 15.0, Dawson edition; Kaysville, UT, USA), the minimum sample size of NAFLD cases was estimated to be 534 and was fulfilled in this study. The study protocol was permitted by the Institutional Ethics Review Committee of Nanjing Medical University (Nanjing, China) and all participants signed informed consents.

Data Collection

The demographic and clinical information of all the participants were collected by structured questionnaires and electronic medical record system. Anthropometric indices, including weight, height, waist circumference (WC), systolic blood pressure (SBP), and diastolic blood pressure (DBP), of the subjects were measured. Body mass index (BMI) was calculated as weight divided by height squared. The definition of overweight was BMI ≥24 kg/m², in line with the Chinese standard [29]. Moreover, each participant supplied an approximate 5-mL venous blood sample for laboratory testing after overnight fasting, including assessments of triacylglycerol (TG), total cholesterol (TC), and fasting plasma glucose (FPG). All the laboratory data were measured by an automatic biochemical analyzer (Mindray BC-860, China). We defined metabolic disorders based on “Guideline of prevention and treatment for nonalcoholic fatty liver disease: a 2018 update” [28], including abdominal obesity (WC >90 cm in males and >85 cm in females), hypertension (SBP/DBP ≥130/85 mm Hg), hypertriglyceridemia (serum triglyceride ≥1.7 mmol/L), and hyperglycemia (FPG ≥5.6 mmol/L). In addition, the abdominal ultrasonography was performed by experienced technicians using a Logiq E9 ultrasound system (General Electric [GE] Healthcare, Milwaukee, WI, USA) for imaging diagnosis of NAFLD. The authenticity and reliability of the collected data were ensured by quality control program throughout the entire investigation process.

Plasma 25(OH)D₃ Level Determination

Plasma 25-hydroxyvitamin D3 (25(OH)D₃) is the main circulating form of VD, commonly used as the indicator to assess serum VD status [30]. In a subgroup of 647 subjects (330 NAFLD cases and 317 controls) selected at random, plasma 25(OH)D₃ levels were measured by enzyme-linked immunosorbent assay (ELISA) (human 25-dihydroxy vitamin D₃ (25(OH)D₃) ELISA Kit; Jin Yibai Biological Technology Co., Ltd.; Nanjing, China). We defined a low VD level (25(OH)D₃ <20 ng/mL) and a normal VD level (25(OH)D₃ ≥20 ng/mL) according to the Endocrine Society clinical practice guideline in the USA [31].

SNPs Selection and Genotyping Assays

Flows within the selection of candidate SNPs were as follows: (1) the information of VDBP gene polymorphisms for China Han Beijing population were screened from the 1000 Genomes Project resources (http://www.1000genomes.org/); (2) then the gene polymorphisms were inputted into the Haploview software (version 4.2; Broad Institute, Cambridge, MA, USA), with minimum minor allele frequency and p value of Hardy-Weinberg equilibrium (HWE) test set to 0.05; (3) the genotypes and minimum minor allele frequencies of the SNPs in the Chinese population were obtained from NCBI dbSNP database (http://www.ncbi.nlm.nih.gov/SNP); (4) we reviewed literatures related to the selected SNPs and metabolic stress disorders (obesity, type-2 diabetes mellitus, metabolic syndrome, inflammatory diseases, etc.), combining with the potential biological function of SNPs from the RegulomeDB online database (http://www.regulomedb.org/) and the UCSC Genome Bioinformatics website (http://genome.ucsc.edu/) to finalize disease-related SNPs. Finally, 7 candidate SNPs of VDBP gene were obtained, including rs222020 (T>C), rs2282679 (A>C), rs4588 (C>A), rs1155563 (T>C), rs7041 (T>G), rs16847024 (C>T), and rs3733359 (C>T).

Genomic DNA was extracted from EDTA anticoagulated blood samples by magnetic bead method (blood genomic extraction kit; Pangu Genome Nanotechnology Co., Ltd.; Nanjing, China). Ultraviolet spectrophotometer (UV-2700220V CH) was employed to test the concentration and purity of DNA. TaqMan allelic discrimination assay was used to genotype the seven selected SNPs on the ABI 7900HT Sequence Detection System (Applied Biosystems, Foster City, CA, USA; catalog numbers: C____948853_20, C__26407519_10, C___8278879_10, C___8278782_20, C___3133594_30, C__33133975_10, C__25652813_40). All laboratory assistants were trained uniformly and blind to the clinical data of the participants. We chose 10% samples randomly to repeat the experiment and achieved 100% consistency rate for quality control. The success rate of genotyping all candidate SNPs was over 95%.

Covariates

The following covariates were included in the logistic regressions. Demographic covariates were gender and age. Clinical variables included BMI and TG. Overweight was categorized by BMI ≥24 kg/m². Age and gender were adjusted in all models to control both the selection bias introduced by the matching variables and the original confounding effects [32, 33].

Statistical Analysis

Normally distributed continuous variables were expressed as mean ± standard deviation and those non-normally distributed was median (interquartile range). Categorical variables were represented by number and percentage. Distributions of baseline data among NAFLD cases and controls groups were compared by χ² test, independent-samples t test, or Mann-Whitney U test, if applicable.

Pearson correlation, point-biserial correlation, and Spearman correlation were used for correlation analyses. Multiple linear regression was adopted to screen determinants of VD levels. The unconditional logistic regression analysis was used to estimate OR and 95% confidence interval (95% CI) on account of the frequency-matching in our study [34]. HWE of VDBP SNPs in the controls group was inspected by the goodness-of-fit χ² test. We used codominant model (mutant homozygous type vs. wild homozygous type; heterozygous type vs. wild homozygous type), dominate model ([mutant homozygous type + heterozygous type] vs. wild homozygous type), recessive model (mutant homozygous type vs. [heterozygous type + wild homozygous type]), and additive model (mutant homozygous type vs. heterozygous type vs. wild homozygous type) in the analyses of single SNP. The heterogeneity between subgroups in stratified analysis was examined using the Q test.

To screen predictors for NAFLD risk assessment, logistic regression model (forward: likelihood ratio) was conducted. Linearity between logits of dependent variable and the continuous independent predictors in the model was assessed by Box-Tidwell Test [35]. Collinearity diagnostics (tolerance and variance inflation factor) were carried out by linear regression analysis. The area under the receiver operating characteristic curve (AUROC) and Hosmer-Lemeshow test were performed to assess the predictive power of the model. The AUROCs of the models were compared by DeLong test [36]. p value <0.05 in two-tailed test was considered to be statistically significant. Bonferroni correction was applied to multiple comparisons among genotypes, with p value adjusted to 7.14 × 10⁻³ (0.05/7) [37]. SPSS (version 23.0, SPSS Inc., Chicago, IL, USA) and MedCalc (Version 19.1, MedCalc Inc, Ostend, Belgium) software were utilized for all statistical analyses in our research.

Results

Description of the Sample Populations

According to the results of abdominal ultrasonography and the 2018 update Guideline, a total of 1,137 NAFLD cases and 1,886 NAFLD controls were recruited into our study. The demographic and clinical characteristics of the 2 groups were shown in Table 1. Distributions of age and gender were all comparable among the 2 groups (all p > 0.05). Other than these two characteristics, significant differences were detected in all other indicators, including BMI, WC, SBP, DBP, TG, TC, FPG, 25(OH)D₃, and in rates of overweight, abdominal obesity, hypertension, hypertriglyceridemia, hyperglycemia (all p < 0.001).

Table 1.

Baseline characteristics between groups of NAFLD cases and controls

Variables	Controls (n = 1,886)	NAFLD cases (n = 1,137)	p value
Age, years	39.85±9.71	40.35±8.14	0.124b
Age n (%)
≤40 years	958 (50.8)	597 (52.5)	0.362a
>40 years	928 (49.2)	540 (47.5)
Gender
Male	1,567 (83.1)	967 (85.0)	0.156a
Female	319 (16.9)	170 (15.0)
BMI, kg/m2	22.70±2.50	25.47±2.55	<0.001b
Overweight, n (%)
No	1,282 (71.7)	309 (28.3)	<0.001a
Yes	506 (28.3)	783 (71.7)
WC, cm	81.40±7.96	88.95±7.98	<0.001b
Abdominal obesity, n (%)
No	1,559 (89.6)	610 (57.1)	<0.001a
Yes	181 (10.4)	458 (42.9)
SBP, mm Hg	124.74±13.69	130.51±15.29	<0.001b
DBP, mm Hg	74.85±9.42	79.62±10.95	<0.001b
Hypertension, n (%)
No	1,151 (64.4)	518 (47.5)	<0.001a
Yes	635 (35.6)	573 (52.5)
TC, mmol/L	4.51±0.81	4.77±0.90	<0.001b
TG, mmol/L*	1.03 (0.78, 1.34)	1.66 (1.19, 2.29)	<0.001c
Hypertriglyceridemia, n (%)
No	1,543 (86.5)	582 (51.3)	<0.001a
Yes	240 (13.5)	553 (48.7)
FPG, mmol/L*	2.60 (2.31, 4.75)	3.98 (2.37, 4.92)	<0.001c
Hyperglycemia, n (%)
No	1,716 (96.2)	1,043 (91.9)	<0.001a
Yes	67 (3.8)	92 (8.1)
25(OH)D3, ng/mL*d	20.00 (16.84, 26.64)	15.77 (11.93, 20.69)	<0.001c

Overweight, BMI >24 kg/m²; abdominal obesity, WC >90 cm in males and >85 cm in females; hypertension, SBP/DBP ≥130/85 mm Hg; hypertriglyceridemia, serum triglyceride ≥1.7 mmol/L; hyperglycemia, FPG ≥5.6 mmol/L. NAFLD, nonalcoholic fatty liver disease; BMI, body mass index; WC, waist circumference; SBP, systolic blood pressure; DBP, diastolic blood pressure; TC, total cholesterol; TG, triacylglycerol; FPG, fasting plasma glucose. Results in bold type indicate statistically significant.

^*Non-normally distributed variables.

^aχ² test.

^bIndependent-samples t test.

^cMann-Whitney U test.

^DSerum 25(OH)D₃ levels were measured in 330 NAFLD cases and 317 controls.

Determinants of Plasma 25(OH)D₃ Levels

Plasma 25(OH)D₃ were Log10 transformed into normal distribution. Pearson correlation indicated that age (r = −0.148, p < 0.001), BMI (r = −0.179, p < 0.001), WC (r = −0.161, p < 0.001), DBP (r = −0.080, p = 0.048), TC (r = −0.105, p = 0.008), TG (r = −0.201, p < 0.001), and FPG (r = −0.143, p < 0.001) were negatively correlated with 25(OH)D₃ levels. No correlation was observed between the seven selected SNPs and VD levels (all p > 0.05). Multivariate linear regression analysis showed that age (β = −0.114, p = 0.030), BMI (β = −0.136, p = 0.038), and TG (β = −0.136, p = 0.007) were independently associated with 25(OH)D₃ levels (Table 2).

Table 2.

The association of clinical parameters and genetic polymorphisms with VD levels

Variables	Correlation analysis		Multiple linear regression
	r/rs	p value	standardized β	p value
Age, years	−0.148	0.001a	−0.114	0.030
Gender	−0.051	0.193b	−	−
BMI, kg/m2	−0.179	<0.001a	−0.136	0.038
WC, cm	−0.161	<0.001a	−0.020	0.765
SBP, mm Hg	−0.064	0.113a	−	−
DBP, mm Hg	−0.080	0.048a	0.057	0.208
TC, mmol/L	−0.105	0.008a	−0.002	0.955
TG, mmol/L*	−0.201	<0.001a	−0.136	0.007
FPG, mmol/L*	−0.143	<0.001a	−0.032	0.541
rs222020	<0.001	0.994c	−	−
rs2282679	−0.044	0.264c	−	−
rs4588	−0.064	0.107c	−	−
rs1155563	−0.062	0.114c	−	−
rs7041	0.056	0.159c	−	−
rsl 6847024	0.031	0.436c	−	−
rs3733359	0.004	0.921c	−	−

were Log10 transformed for normality, and the seven SNPs were coded in additive models. BMI, body mass index; WC, waist circumference; SBP, systolic blood pressure; DBP, diastolic blood pressure; TC, total cholesterol; TG, triacylglycerol; FPG, fasting plasma glucose. Results in bold type indicate statistically significant.

^*Non-normally distributed variables were Log10 transformed for normality.

^aPearson correlation.

^bPoint-biserial correlation.

^cSpearman correlation.

Associations between 25(OH)D₃ Levels and NAFLD

As shown in Figure 1, plasma 25(OH)₃ levels were categorized into 2 groups: low VD levels (<20 ng/mL) and normal VD levels (≥20 ng/mL), and prevalence of low VD levels in our study was 61.36% (397/647). Participants with low VD levels had significantly higher prevalence of NAFLD compared to subjects with normal levels (p < 0.001).

Fig. 1.

Percentage of NAFLD cases and controls according to plasma 25(OH)D₃ levels. NAFLD, nonalcoholic fatty liver disease; 25(OH)D₃, 25-hydroxyvitamin D₃. χ² test between the 2 groups: χ² = 31.070, p < 0.001.

Logistic regression analysis (VD ≥20 ng/mL as reference) showed that the odds of NAFLD were increased for low VD levels (crude OR = 2.501, 95% CI = 1.806–3.463, p < 0.001). After adjustment for age and gender, low VD levels were associated with higher the risk of NAFLD than normal VD levels (adjusted OR = 2.624, 95% CI = 1.882–3.659, p < 0.001). When further adjusting for BMI and TG, the odds of NAFLD remained significantly elevated for low VD levels (adjusted OR = 2.282, 95% CI = 1.422–3.661, p = 0.001) (Table 3).

Table 3.

The logistic regression analysis of plasma 25(OH)D₃ levels for NAFLD risk

25(OH)D3 levels	OR	95% CI	p value
Model 1
≥20 ng/mL	1.00 (ref)	−	−
<20 ng/mL	2.501	1.806–3.463	<0.001
Per 1 ng/mL increase	0.984	0.975–0.993	0.001
Model 2
≥20 ng/mL	1.00 (ref)	−	−
<20 ng/mL	2.624	1.882−3.659	<0.001
Per 1 ng/mL increase	0.983	0.974−0.992	<0.001
Model 3
≥20 ng/mL	1.00 (ref)	−	−
<20 ng/mL	2.282	1.422−3.661	0.001
Per 1 ng/mL increase	0.990	0.980–1.001	0.073

Model 1: unadjusted. Model 2: adjusted for age, gender. Model 3: further adjusted for BMI and TG. 25(OH)D₃, 25-hydroxyvitamin D₃; OR, odds ratio; CI, confidence interval. Results in bold type indicate statistically significant.

Moreover, we further assessed the association between VD concentration as a continuous variable and NAFLD risk. After adjusted by age and gender, NAFLD risk decreased 1.7% for each 1 ng/mL increase in 25(OH)D₃ (adjusted OR = 0.983, 95% CI = 0.974–0.992, p < 0.001). The reduction of NAFLD risk with increasing VD became nonsignificant after adjustment for BMI and TG (adjusted OR = 0.990, 95% CI = 0.980–1.001, p = 0.073) (Table 3).

Associations between VDBP SNPs and NAFLD

The genotype distributions of the seven SNPs in VDBP gene among groups of controls and NAFLD cases were shown in Table 4. The genotype frequencies of all the SNPs in the controls group were in HWE (all p > 0.05). It is widely proved that obesity or overweight is the most meaningful risk factor for NAFLD [38]. Therefore, we added overweight and age, gender as adjustment factors to the logistic regression analyses. The additive genetic model showed that rs7041-G significantly decreased the risk of NAFLD (adjusted OR = 0.814, 95% CI = 0.713–0.929, p = 0.002). Moreover, the codominant model indicated that the rs7041-GG contributed to a reduced the risk of NAFLD in comparison with the TT wild genotype (adjusted OR = 0.623, 95% CI = 0.449–0.866, p = 0.005). After Bonferroni correction, the association between VDBP variants rs7041 and NAFLD risk remained statistically significant.

Table 4.

Associations between VDBP SNPs and the occurrence of NAFLD

Genotype	Controls n (%)	NAFLD cases n (%)	p valuea	OR (95% CI)b	p valuea
rs222020			0.842
TT	749 (40.0)	461 (41.1)		−
TC	879 (47.0)	519 (46.2)		0.922 (0.771–1.104)	0.376
CC	243 (13.0)	143 (12.7)		0.995 (0.760–1.302)	0.968
Dominant				0.937 (0.790–1.111)	0.455
Recessive				1.038 (0.807–1.336)	0.769
Additive				0.975 (0.861–1.104)	0.689
rs2282679			0.174
AA	856 (46.2)	523 (46.2)		−
AC	833 (44.9)	510 (45.0)		1.058 (0.888–1.260)	0.530
CC	165 (8.9)	100 (8.8)		1.076 (0.790–1.466)	0.641
Dominant				1.061 (0.897–1.254)	0.492
Recessive				1.047 (0.778–1.408)	0.762
Additive				1.046 (0.918–1.192)	0.503
rs4588			0.585
CC	889 (47.5)	525 (46.5)		−
CA	817 (43.6)	495 (43.9)		1.068 (0.896–1.272)	0.465
AA	167 (8.9)	108 (9.6)		1.217 (0.901–1.644)	0.200
Dominant				1.092 (0.924–1.291)	0.302
Recessive				1.179 (0.884–1.573)	0.262
Additive				1.089 (0.957–1.239)	0.195
rs1155563			0.790
TT	644 (34.6)	384 (34.2)		−
TC	889 (47.8)	545 (48.5)		1.013 (0.840–1.221)	0.893
CC	327 (17.6)	194 (17.3)		1.016 (0.794–1.299)	0.902
Dominant				1.014 (0.850–1.209)	0.881
Recessive				1.008 (0.808–1.257)	0.944
Additive				1.009 (0.895–1.137)	0.889
rs7041			0.981
TT	959 (51.0)	613 (54.0)		−
TG	766 (40.7)	446 (39.3)		0.844 (0.708–1.005)	0.057
GG	157 (8.3)	77 (6.8)		0.623 (0.449–0.866)	0.005
Dominant				0.804 (0.680–0.951)	0.011
Recessive				0.672 (0.488–0.924)	0.015
Additive				0.814 (0.713–0.929)	0.002
rs16847024			0.983
CC	1,404 (75.1)	846 (74.9)		−
CT	431 (23.1)	263 (23.3)		1.067 (0.875–1.301)	0.523
TT	34 (1.8)	21 (1.9)		1.124 (0.597–2.118)	0.717
Dominant				1.071 (0.883–1.299)	0.488
Recessive				1.107 (0.589–2.082)	0.752
Additive				1.065 (0.895–1.267)	0.477
rs3733359			0.846
CC	792 (42.3)	485 (42.9)		−
CT	859 (45.9)	517 (45.7)		0.976 (0.817–1.165)	0.787
TT	220 (11.8)	129 (11.4)		0.948 (0.717–1.252)	0.705
Dominant				0.970 (0.820–1.148)	0.725
Recessive				0.960 (0.738–1.248)	0.759
Additive				0.974 (0.859–1.104)	0.683

Take rs222020 as example: codominant model (TC vs. TT; CC vs. TT), dominant model (TC + CC vs. TT), recessive model (CC vs. TC + TT), and additive model (CC vs. TC vs. TT). NAFLD, nonalcoholic fatty liver disease; OR, odds ratio; CI, confidence interval. Results in bold type indicate statistically significant.

^aχ² test in HWE for controls.

^bBinary logistic regression analysis adjusted for age, gender, and overweight, and p value was adjusted to 0.00714 (0.05/7) by Bonferroni correction.

Further stratification analysis of rs7041 on the susceptibility to NAFLD was performed according to gender, age, overweight, abdominal obesity, hypertension, hypertriglyceridemia, and hyperglycemia. We employed the additive genetic model of rs7041 to calculate the OR and 95% CI within each subgroup after adjusting for age, gender, and overweight. As shown in Table 5, significant decreased risk of incident NAFLD in rs7041-G remain existed in the stratification of age and overweight (all p < 0.05), along with male (adjusted OR = 0.803, 95% CI = 0.696–0.926, p = 0.003) and participants with nonabdominal obesity (adjusted OR = 0.793, 95% CI = 0.676–0.929, p = 0.004), nonhypertension (adjusted OR = 0.795, 95% CI = 0.663–0.952, p = 0.013), nonhypertriglyceridemia (adjusted OR = 0.826, 95% CI = 0.697–0.978, p = 0.026), and nonhyperglycemia (adjusted OR = 0.820, 95% CI = 0.712–0.943, p = 0.005). Furthermore, we assessed the heterogeneity of the subgroups, but no significant difference was found (all p > 0.05).

Table 5.

Stratified analysis of the association between VDBP-rs7041 genotypes and NAFLD risk

Variables	Controls, n (TT/TG/GG)	NAFLD cases, n (TT/TG/GG)	OR (95% CI)a	p valuea	p valueb
Gender
Male	803/628/132	528/371/67	0.803 (0.696–0.926)	0.003	0.856
Female	156/138/25	85/75/10	0.834 (0.577–1.207)	0.336
Age
≤40 years	491/392/73	318/239/39	0.805 (0.665–0.974)	0.026	0.978
>40 years	468/374/84	295/207/38	0.808 (0.669–0.977)	0.027
Overweight
No	658/523/99	176/117/15	0.796 (0.648–0.979)	0.031	0.782
Yes	242/213/51	413/313/57	0.827 (0.695–0.985)	0.033
Abdominal obesity
No	790/635/132	342/228/40	0.793 (0.676–0.929)	0.004	0.406
Yes	88/80/13	236/189/32	0.915 (0.693–1.210)	0.535
Hypertension
No	578/476/95	272/216/29	0.795 (0.663–0.952)	0.013	0.604
Yes	322/258/55	317/213/43	0.854 (0.701–1.040)	0.116
Hypertriglyceridemia
No	778/635/127	309/238/35	0.826 (0.697–0.978)	0.026	0.774
Yes	126/93/20	302/208/42	0.789 (0.607–1.027)	0.078
Hyperglycemia
No	871/701/140	552/424/66	0.820 (0.712–0.943)	0.005	0.704
Yes	33/27/7	59/22/11	0.733 (0.418–1.285)	0.278

NAFLD, nonalcoholic fatty liver disease; OR, odds ratio; CI, confidence interval. Overweight, BMI >24 kg/m²; abdominal obesity, WC >90 cm in males and >85 cm in females; hypertension, SBP/DBP ≥130/85 mm Hg; hypertriglyceridemia, serum triglyceride ≥1.7 mmol/L; hyperglycemia, FPG ≥5.6 mmol/L. Results in bold type indicate statistically significant.

^aAdjusted for age, gender, and overweight in additive model (except stratification variable itself).

^bp value for heterogeneity test between subgroups.

Variables Significantly Related to NAFLD

Binary logistic regression analysis (forward: likelihood ratio) was conducted to further determine the independent factors affecting the risk of NAFLD. We incorporated gender, age, overweight, abdominal obesity, hypertension, hypertriglyceridemia, hyperglycemia, and rs7041 (additive model) into analysis. Age was included as a categorical variable for that the assumption of linearity was not met for its continuous form. Results showed that female (OR = 1.770, 95% CI = 1.360–2.303, p < 0.001), overweight (OR = 3.902, 95% CI = 3.171–4.802, p < 0.001), abdominal obesity (OR = 2.694, 95% CI = 2.104–3.451, p < 0.001), hypertension (OR = 1.518, 95% CI = 1.254–1.837, p < 0.001), and hypertriglyceridemia (OR = 4.601, 95% CI = 3.726–5.680, p < 0.001) were statistically significant promoters for NAFLD initiation. Oppositely, age >40 years (OR = 0.649, 95% CI = 0.530–0.794, p < 0.001) and rs7041-G (OR = 0.820, 95% CI = 0.708–0.951, p = 0.009) independently impeded the occurrence of NAFLD (Table 6).

Table 6.

Variables significantly related to NAFLD in logistic regression analysis

Variables	Unstandardized β	OR	95% CI	p value
Gender (female vs. male)	0.571	1.770	1.360–2.303	<0.001
Age (>40 vs. ≤40 years)	−0.432	0.649	0.530–0.794	<0.001
Overweight (yes vs. no)	1.361	3.902	3.171–4.802	<0.001
Abdominal obesity (yes vs. no)	0.991	2.694	2.104–3.451	<0.001
Hypertension (yes vs. no)	0.417	1.518	1.254–1.837	<0.001
Hypertriglyceridemia (yes vs. no)	1.526	4.601	3.726–5.680	<0.001
rs7041 (GG vs. TG vs. TT)	−0.198	0.820	0.708–0.951	0.009
Constant	−2.292	−	−	−
Hosmer-Lemeshow test: χ2 = 11.355, *p = 0.182

Overweight, BMI >24 kg/m²; abdominal obesity, WC >90 cm in males and >85 cm in females; hypertension, SBP/ DBP ≥130/85 mm Hg; hypertriglyceridemia, serum triglyceride ≥1.7 mmol/L. Results in bold type indicate statistically significant.

^*p value for Hosmer-Lemeshow test.

We further united the determined independent factors into models for risk assessment of NAFLD, and the equations of models can be expressed as Model 1 = (−2.390) + 0.566 × gender + (−0.434) × age + 1.348 × overweight + 0.995 × abdominal obesity + 0.416 × hypertension + 1.531 × hypertriglyceridemia; Model 2 = (−2.292) + 0.571 × gender + (−0.432) × age + 1.361 × overweight + 0.991 × abdominal obesity + 0.417 × hypertension + 1.526 × hypertriglyceridemia + (−0.198) × rs7041. The numeric coding of all categorical variables was presented in online supplementary Table 1 (see www.karger.com/doi/10.1159/000522193 for all online suppl. material). Collinearity diagnostics showed no evidence of collinearity among the independent factors in the models (tolerance >0.1; variance inflation factor <10). As shown in Figure 2, the AUROCs of model 1 and model 2 were 0.813 (95% CI = 0.798–0.827) and 0.817 (95% CI = 0.802–0.831), respectively. The DeLong test between the two models showed that discrimination of model 2 was significantly higher than model 1 (p = 0.0075). Additionally, the Hosmer-Lemeshow goodness-of-fit test indicated model 2 fitted the observed data well (p = 0.182).

Fig. 2.

The ROC curves of the established models for assessing NAFLD risk.

Discussion

In this study, the associations of plasma VD levels and VDBP genetic polymorphisms (rs222020, rs2282679, rs4588, rs1155563, rs7041, rs16847024, rs3733359) with the risk of NAFLD were investigated in a Chinese Han population for the first time. We found that a low level of 25(OH)D₃ and VDBP-rs7041-G allele were all significantly associated with the risk of NAFLD occurrence.

There was a high prevalence of low VD levels among subjects in East China of our study, consistent with the high prevalence in other regions [39, 40]. Age, BMI, and TG were proved to be independent factors influencing VD levels in our study, partially explained by the hypothesis that fat-soluble VD may be sequestered in adipose tissue and thus decrease serum VD levels [41]. Accumulating studies reported that plasma levels of VD have profound influences on the susceptibility to NAFLD [42]. As observed in our study, NAFLD occurrence was more prevalent among subjects with low levels of VD than with normal VD levels, and low VD levels increased the risk of NAFLD. The effect of VD could be explained by the anti-fibrosis, anti-inflammatory, and insulin-regulating characteristics [9, 11, 43]. However, the susceptibility to NAFLD varies considerably among subjects with similar VD levels. Consequently, the correlations of VD level and VDBP with NAFLD risk were elucidated in our study, considering that NAFLD is influenced by a combination of hereditary and environmental factors.

VDBP, also known as the group-specific component (Gc-globulin), is a member of the albuminoid superfamily [44]. VDBP is mostly synthesized by the hepatocytes and reported to be expressed in the liver, fat cells, and neutrophils [45]. The physiological functions of VDBP are diverse. Specifically, VDBP is the primary plasma transporter for VD and its metabolites, and may promote Mets, type-2 diabetes, and NAFLD occurrence by affecting plasma levels of VD [46]; VDBP is a transport of fatty acids and may be related with lipid metabolism [47]; VDBP plays a role in inflammatory process by activating macrophage and C5a-mediated chemotaxis [48]. All the functions suggested that VDBP may have the potential effects on the pathogenesis of NAFLD.

Being a highly polymorphic protein, the Gc-globulin is encoded by the VDBP gene. The human VDBP gene is 35 kb in length with 13 exons and 12 introns, localized at the long arm of chromosome 4 (4q12-q13) [44, 49]. More than 120 variants constitute the VDBP gene, the most common variants of which are rs7041 and rs4588 [50, 51]. The binding between VDBP and VD appears to be influenced by genetic variations in Gc gene, with rs7041 proved to have a greater affinity and a higher transport efficiency for VD metabolites [52]. Moreover, several studies found that VDBP common genetic polymorphisms regulated the levels of VD [53], as well as the response to VD supplementation therapy [54].

The common variants of VDBP (rs7041, rs4588, rs2282679, rs222020, rs1155563) have been reported to be associated with a low level of VD [55, 56, 57, 58], and data regarding effects of rs16847024 and rs3733359 on VD level are limited. However, recent studies found no connection of rs7041, rs4588, and rs2282679 with low levels of VD [59, 60]. In the present study, no significant relationship between the seven VDBP SNPs above and plasma VD levels was observed either after adjustment for age, gender, and metabolic factors. Such inconsistent results may be explained by backgrounds of sun exposure, season, dietary intake, and skin color [61].

VDBP-rs7041 is a widely investigated functional SNP, and the relationship between rs7041 and VD metabolism-related disease has been explored [62]. Although research on the association between rs7041 and NAFLD is absent, this genetic variant has been investigated in hepatocellular carcinoma [26] and hepatitis C virus infection [27]. Our study found that VDBP-rs7041-G significantly decreased the risk of NAFLD occurrence. Similarly, previous studies indicated that the mutant G allele of rs7041 possibly be the protector against the risk of metabolic syndrome [59] and cardiovascular disease [63]. Further studies on a larger sample are warranted to confirm the effects of VDBP variants on VD levels and NAFLD risk.

Obesity, hyperglycemia, dyslipidemia, and hypertension are all widely recognized metabolic factors increasing the risk of NAFLD [64, 65]. In this context, stratified analysis was performed to evaluate the impact of gender, age, overweight, abdominal obesity, hypertension, hypertriglyceridemia, and hyperglycemia on the associations between VDBP variants and NAFLD. The protective effect of rs7041-G remained across all the stratified variables with no significant heterogeneity found, suggesting that these traditional metabolic risk factors for NAFLD did not change such effect.

To further identify determinants of NAFLD risk, binary logistic regression screened gender, age, overweight, abdominal obesity, hypertension, hypertriglyceridemia, and rs7041 as independent factors, which were accorded with the findings of previous studies [65, 66]. Based on independent factors above, we developed simple models for risk assessment of NAFLD. The discrimination and goodness of fit of the model 2 with the inclusion of rs7041 were better than the model 1 without rs7041. Differences between the two models represented the significant contribution of this genetic variant to NAFLD risk assessment. Of note, the elevated prediction capability of rs7041 was slight, possibly caused by the weak effect of single locus, which prompts us to take polygenic loci into the risk assessment of NAFLD in future studies.

There are several limitations in our study. First, all the participants were enrolled from the same community, possibly leading to selection bias in the study. We matched NAFLD cases and controls by age and gender and took potential confounding factors into adjustment or as stratification in analyses in order to minimize such bias. Second, the measurements of plasma VD levels of all selected subjects were conducted in the same season, resulting in failure in determining the effects of different seasons on VD. Third, adipose mass indicators like body fat mass or surrogate markers were not included in the study. We took BMI, WC, TC, and TG into analysis to partially reflect the impacts of adipose mass on VD levels. Fourth, information on supplements, sun exposure, and dietary intake of VD was not collected, which may lead to residual confusion in our study. Finally, the association between VDBP-rs7041 and the susceptibility to NAFLD determined in our study need to be verified functionally in a larger sample.

Conclusion

In conclusion, our findings revealed that a low level of plasma VD and the mutant rs7041-G allele in VDBP all had significant effects on the risk of NAFLD occurrence in a Chinese population, which may contribute to elucidate NAFLD environmental and genetic etiology. On the basis of the findings, we developed an excellent prediction model comprised of VDBP variant and metabolic disorders to assess and identify patients with high risk of NAFLD early.

Statement of Ethics

The study was approved by Nanjing Medical University (Ethical Review No. 2018-596). Written informed consent was obtained from all subjects for being included in the study.

Conflict of Interest Statement

The authors declare that they have no conflicts of interest.

Funding Sources

This work was supported in part by the Natural Science Foundation of Jiangsu Province (grant no. BK20181369), the Six Talent Peak Project in Jiangsu Province (grant No. 2019-WSN-049), and Priority Academic Program Development of Jiangsu Higher Education Institutions (grant No. PAPD [2018] 87).

Author Contributions

J.W. designed and organized the study; M.W., M.W., R.Z., L.Z., Y.D., and Z.T. contributed to the planning, designing, and analyses of the experiments; C.S., H.W., W.Z., and Y.C. did data collection and quality control. M.W. and J.W. wrote and critical revised the manuscript. All authors read and approved the final manuscript.

Data Availability Statement

The data sets generated and analyzed during this study are not publicly available due to the legal and ethical grounds, which can be accessed from the corresponding author on a reasonable request.

Supplementary Material

Supplementary data