Open Access
Open Peer Review

This article has Open Peer Review reports available.

How does Open Peer Review work?

Allelic variants in vitamin D receptor gene are associated with adiposity measures in the central-European population

  • Julie Bienertová-Vašků1, 2Email authorView ORCID ID profile,
  • Filip Zlámal1,
  • Aneta Pohořalá1,
  • Ondřej Mikeš1,
  • Monika Goldbergová-Pávková2,
  • Jan Novák2,
  • Zbyněk Šplíchal2 and
  • Hynek Pikhart1, 3
Contributed equally
BMC Medical GeneticsBMC series – open, inclusive and trusted201718:90

https://doi.org/10.1186/s12881-017-0454-z

Received: 23 January 2017

Accepted: 14 August 2017

Published: 22 August 2017

Abstract

Background

There is an increasing body of evidence suggesting that vitamin D is involved in ethiopathogenesis of obesity and therefore the aim of the study was to investigate whether 5 selected SNPs in VDR (vitamin D receptor) gene are associated also with anthropometry in the obese and non-obese Central-European population.

Methods

A total of 882 Central European Caucasian individuals of Czech origin were recruited (n = 882, 232 M/650 F) and weight, height, BMI, lean body mass, fat mass, body fat, waist and hip circumference, waist–hip ratio (WHR) and skinfold thickness were measured. Univariate and multivariate models were constructed in order to investigate the relationship between anthropometry and VDR polymorphisms.

Results

In the univariate modeling, the CC genotype of FokI SNP was associated with reduced waist circumference (β = −3.48; 95%CI:-7.11;0.15; p = 0.060), sum of skin fold thickness (β = −6.53, 95% CI: -12.96;-0.11; p = 0.046) as well as total % of body fat (β = −3.14, 95% CI: -5.18;-1.09; p = 0.003) compared to TT genotype. The AC genotype of ApaI SNP was associated with reduced waist circumference compared to AA genotype (β = −4.37, 95% CI: -7.54;-1.20; p = 0.007). GG genotype of EcoRV SNP was associated with reduced sum of skin fold thickness compared to AA genotype (β = −7.77, 95% CI: -14.34;-1.21; p = 0.020). In the multivariate modelling, multiple significant associations of VDR with investigated traits were observed, too.

Conclusion

Our study suggests that genetic variability in the VDR region may be an important factor influencing anthropometric characteristics associated with obesity.

Keywords

VDR SNP Obesity Anthropometry Adiposity Gene Vitamin D Polymorphism Adiposity measure

Background

Obesity is an epidemic worldwide and represents a potential risk factor for many chronic health conditions including cardiometabolic diseases as well as cancer. The increase of obesity prevalence in some geographical regions is, despite an undoubtable genetic contribution, difficult to explain and is attributed mainly to the obesity-associated lifestyle factors, largely manageable by public health interventions. However, the exact proportion of genetic and/or environmental influences in the complex diseases is undoubtedly difficult to estimate.

Apart from its fundamental role in prevention of rickets, low vitamin D status has recently become a major public health problem due to its associations with several chronic metabolic diseases [1]. It is empirically known that obesity is associated with vitamin D status but the underlying pathophysiology has been uncertain and the role of vitamin D receptor variability in obesity is far from being understood. The possible positive associations of vitamin D status with obesity reported in the smaller studies have not been confirmed either in randomized controlled studies testing the effect of vitamin D supplementation on obesity risk [2, 3]. Therefore, the role of vitamin D in obesity prophylaxis and/or treatment remains to be further investigated.

Still, it seems that vitamin D effect on obesity risk is biologically plausible as in vitro experiments in rats have previously shown that administration of extensive doses of vitamin D2 may lead to an increase in energy expenditure due to uncoupling of oxidative phosphorylation in adipose tissues [4]. However, these data cannot be simply extrapolated to humans, as there are numerous differences in obesity pathophysiology between mice and humans.

Generally, vitamin D, that is derived from the diet or from 7-dehydrocholesterol, must be further activated to achieve full activity within the body [4]. The most active known metabolite of vitamin D is calcitriol, i.e. 1,25- dihydroxyvitamin D (1,25(OH)2D), and its effects on genome are mediated through a transcription factor activated by a ligand, vitamin D receptor (VDR) [5]. 1,25(OH)2D functions as a ligand by direct binding to the VDR locus, a key member of the steroid hormone receptor family, which in turn regulates the transcription of the numerous target genes.

The DNA polymorphisms that have been often reported for the VDR gene are: BsmI (rs1544410), ApaI (rs7975232), TaqI (rs731236), FokI (rs2228570) and a EcoRV site in the promoter region (rs4516035) [6]. In the past years, many studies reported these polymorphisms to be related to bone metabolism, cancer, type 1 diabetes and obesity [7]. However, the association with obesity was reported in some, but not all studies [8]. As vitamin D supplementation is commonly recommended to general population, it is important to know whether the beneficial effects are BMI dependent and more research to the role of vitamin D in body composition is therefore necessary.

In the present study, we investigated the relationship between anthropometric traits including BMI, a widespread measure for monitoring the prevalence of obesity at the population level, and selected genetic variants, namely BsmI, FokI, ApaI, EcoRV (GATA) and TaqI in the VDR gene, in the Central-European obese and lean population.

Methods

Subjects

A total of 882 Central European Caucasian individuals of Czech origin were recruited for the study in a mass media campaign targeting the South Moravia region of the Czech Republic (232 men and 650 women; mean BMI 31.9 (SD 7.4) kg/m2; median age 50.3 years; age range 18.8–79.7 years). Inclusion and exclusion criteria were reported elsewhere [7, 8]. Briefly, participation in the study was limited to individuals who were 1) not taking cholesterol-lowering medications; 2) were not currently on any lipid-lowering or weight-control diets; 3) were free from possible causes of secondary severe hypercholesterolemia (e.g., hyperthyroidism, pregnancy); and 4) were free from severe chronic illness (e.g., cancer, renal disease, heart failure). However, in contrast to the study by Ma et al., the subjects who worked night shifts were maintained in our study.

The study was conducted according to the guidelines set out in the Declaration of Helsinki and all procedures involving human subjects were approved by the Committee for Ethics of Medical Experiments on Human Subjects, Faculty of Medicine of Masaryk University (Brno, Czech Republic). Written informed consent was obtained from all subjects.

Data on personal or family history of obesity, birth weight, age at onset of obesity and its severity were obtained by a professional interviewer using a semi-structured questionnaire. A positive family history of obesity was estimated as having at least one obese relative with BMI ≥ 30 kg/m2 in the close family (siblings, parents and their siblings, or grandparents). Both the obese cases and the non-obese controls underwent the same examination focused on their anthropometric characteristics; the subjects of the study were also interviewed with respect to their current smoking status and its history, type of work, shift work and their nutritional habits using the 7-day food records on the grounds of which the total energy intake of the individuals was established.

Anthropometric characteristics

All phenotypic measurements were performed by experienced specialists and included weight, height, BMI, lean body mass, fat mass, body fat, waist and hip circumference, waist–hip ratio (WHR) and skinfold thickness, as described elsewhere [7]. Body composition was assessed by bioelectrical impedance analysis using the InBody 230 bioimpedance analyzer (Biospace CO Ltd., 518–10 Dogok 2-dong, Gangnam-gu, Seoul, Korea) with the subject in a standing position.

Genotyping

The following SNPs were genotyped: BsmI (rs1544410), ApaI (rs7975232), TaqI (rs731236), FokI (rs2228570) and a EcoRV (GATA) (rs4516035).

The selection of these particular SNPs was based on: (1) its frequency in the European Caucasian population; (2) its presumed functional or regulatory impact on feeding behaviour and its possible functional consequences and (3) a previously described association with obesity. DNA for analyses was extracted from 5 mL of the patients’ saliva collected after 3 h fasting. Genotyping of the polymorphism was performed as described previously using a standard PCR-based methodology with following restriction fragment length polymorphism analysis [9]. Restricted fragments were separated by electrophoresis on 2% agarose gels with ethidium bromide staining. To assess genotyping reliability, we performed double sampling in more than 20% of the samples and found no differences. We always used quality control, and negative controls were used to identify possible false positives.

All reactions were performed using the XP BIOER Cycler (BIOER Technology CO. Ltd., Japan), the overall genotypization success varied between 76.8% (TaqI SNP) and 89.0% (FokI SNP) (with 77.8% for ApaI, 88.8% for BsmI and 88.4% for EcoRV), missing genotypes were due to either consistent PCR dropout or depletion of template DNA.

Haplotype analysis

The haplotype frequencies were calculated using the method for reconstructing haplotypes from population data according to Stephens et al. [10] Bonferroni correction for multiple hypothesis testing was performed, where appropriate.

Statistics

Where applicable, it was first determined whether a variable was normally distributed using a battery of normality tests (e.g. Shapiro-Wilk test) and graphical tools; in cases of skewed variables, logarithmic transformation was performed and normal distribution was tested again. For descriptive purposes, mean values are presented using untransformed values. Results are expressed as mean values and standard deviations unless otherwise stated. ANOVA and Kruskal-Wallis test were used for multiple comparisons with subsequent Tukey-Kramer, respectively Benjamini-Hochberg p-value correction, where appropriate.

For the purpose of investigation of crude association between genotypes and obesity status, logistic regression was used.

Using the linear regression modelling, weight, height, lean body mass, total body fat, waist and hip circumference, WHR, BMI and total sum of skin folds were modelled as dependent variables, while the genotypes of the investigated SNPs (ApaI, BsmI, EcoRV, Fok-I, Taq-I), gender, smoking status, BMI (squared value) and age as independent variables. Firstly, the univariate models were constructed for the modelled dependent variable in relation to genotypes, separately for each polymorphism.

Consecutively, multivariate models were constructed in relation to variables gender, age, BMI, BMI2, smoking status, comorbidity, type of work, shift work, total energy intake and a three-combination of genotype (ApaI-EcoRV-FokI; BsmI-EcoRV-FokI; TaqI-EcoRV-FokI). Type of work was determined as manual/non-manual work and sedentary/non-sedentary work. The multivariate models did not include all the genotypes of the 5 investigated polymorphisms, as the trio Apa-1-Bsm-1-Taq-1 is in strong linkage disequilibrium (LD) and there is a strong statistical dependency between these variables. If two of these SNPs or all the three of them were included into the model, this could significantly bias the confidence intervals in the model due to strong collinearity.

Missing data were imputed using Multiple imputation by chained equations method [11] and the regression models were built using imputed data. The total of 50 imputed datasets was used with the maximum of 20 iterations.

All analyses were performed using R, version 3.1.2 and STATA version 14. Consensual values of p < 0.05 were considered statistically significant.

Results

Baseline characteristics of the study cohort

The baseline description of the study cohort with respect to demographic and clinical parameters is presented in Table 1. The Hardy–Weinberg disequilibrium was assessed for all sub-cohorts, and none of the SNPs was significantly deviated from it. The relationship between the investigated polymorphisms and obesity status (based on the cut-off value of BMI 30.0 kg/m2) is given in Table 2. Significant differences between the obese and non-obese group were observed for TaqI, where the AA genotype was more frequent in non-obese individual (45.1% vs. 37.1%, p = 0.034).
Table 1

Basic anthropometric and demographic characteristics of the study subjects

Parameter

Unit

All

Females

Males

p-value*

   

Non-obese

Obese

Non-obese

Obese

 
  

882

281

369

90

142

 

Height

cm

167.8 ± 8.9

165.6 ± 6.8abc

163.6 ± 6.9ade

177.8 ± 7.3bd

176.9 ± 6.6ce

<0.001

Weight

kg

89.8 ± 22.2

68.9 ± 10.3abc

98.4 ± 16.9ade

82.2 ± 10.5bdf

113.8 ± 18.9cef

<0.001

BMI

kg/m2

31.9 ± 7.4

25.1 ± 3.4ab

36.7 ± 5.8ac

26.0 ± 2.6cd

36.4 ± 5.6bd

<0.001

Age

years

48.3 ± 14.2

43.0 ± 14.9ab

52.7 ± 12.1acd

44.3 ± 15.8ce

49.9 ± 12.8bde

<0.001

Skinfold thickness (tricipital)

mm

25.1 ± 8.0

20.6 ± 5.6abc

30.4 ± 5.9ade

16.4 ± 6.2bdf

23.6 ± 7.6cef

<0.001

Skinfold thickness (bicipital)

mm

18.4 ± 8.0

13.7 ± 5.1abc

23.3 ± 7.0ade

10.9 ± 5.7bdf

17.9 ± 7.3cef

<0.001

Skinfold thickness (supraspinal)

mm

22.8 ± 9.1

16.8 ± 5.9ab

27.5 ± 8.0acd

15.5 ± 6.0ce

25.4 ± 9.5bde

<0.001

Skinfold thickness (subscapular)

mm

25.7 ± 9.7

18.2 ± 6.3ab

30.7 ± 8.3ac

18.6 ± 6.6cd

29.8 ± 8.3bd

<0.001

Skinfold thickness (sum of all)

mm

91.5 ± 30.0

69.1 ± 18.9abc

111.6 ± 22.9ade

61.4 ± 20.3bdf

95.9 ± 25.2cef

<0.001

Body fat

%

37.1 ± 10.4

32.2 ± 7.3abc

45.8 ± 5.6ade

21.4 ± 6.0bdf

33.9 ± 6.5cef

<0.001

Body water

%

46.4 ± 6.9

49. 6 ± 5.0ab

40.8 ± 3.8acd

56.4 ± 4.6bce

48.4 ± 4.5de

<0.001

Waist circumference

cm

101.8 ± 18.6

82.4 ± 10.5abc

109.5 ± 12.9ade

93.7 ± 9.6bdf

119.4 ± 13.7cef

<0.001

Hip circumference

cm

113.3 ± 13.9

101.8 ± 7.4ab

122.4 ± 12.5acd

102.0 ± 5.5ce

117.3 ± 10.4bde

<0.001

Waist–hip ratio

---

0.90 ± 0.10

0.81 ± 0.08abc

0.90 ± 0.08ade

0.92 ± 0.07bdf

1.02 ± 0.06cef

<0.001

For a given variable, each pair of superscript letters (a, b, c, d, e, f) appearing in two diffefent groups (e.g. letter a in non-obese female and obese female group) mean statistically significant difference (p < 0.05) in levels of the variable between these groups in post-hoc test, * p-value for comparison between obese and non-obese individuals

Table 2

VDR polymorphisms and odds of being obese in the study cohort

Polymorphism

Obese

Non-obese

OR (95% CI)*

p for linear trend

BsmI

458

325

 

0.652

 CC (%)

178 (38.9)

127 (39.1)

1.00

 CT (%)

213 (46.5)

157 (48.3)

0.97 (0.71;1.32)

 TT (%)

67 (14.6)

41 (12.6)

1.17 (0.74;1.83)

FokI

460

325

 

0.055

 TT (%)

75 (16.3)

73 (22.5)

1.00

 TC (%)

228 (49.6)

153 (47.1)

0.94 (0.68;1.30)

 CC (%)

157 (34.1)

99 (30.4)

0.65 (0.43;0.98)

ApaI

407

279

 

0.195

 AA (%)

112 (27.5)

67 (24.0)

1.00

 AC (%)

206 (50.6)

141 (50.5)

0.87 (0.60;1.27)

 CC (%)

89 (21.9)

71 (25.5)

0.75 (0.49;1.16)

TaqI

402

275

 

0.034

 AA (%)

149 (37.1)

124 (45.1)

1.00

 AG (%)

191 (47.5)

118 (42.9)

1.35 (0.97;1.88)

 GG (%)

62 (15.4)

33 (12.0)

1.56 (0.96;2.54)

EcoRV

455

325

 

0.668

 AA (%)

131 (28.8)

91 (28.0)

1.00

 AG (%)

235 (51.6)

166 (51.1)

0.98 (0.70;1.37)

 GG (%)

89 (19.6)

68 (20.9)

0.90 (0.60;1.38)

The given values are outputs of the univariable logistic regression model, * 95% confidence interval

Association of investigated VDR polymorphisms with the anthropometric measures

Briefly, we observed multiple associations of the investigated polymorphisms with anthropometric parameters as shown in Fig. 1.
Fig. 1

Involvement of vitamin D variability in obesity pathogenesis in relation to the results of the univariate analysis

In the univariable analysis (Table 3), FokI was borderly associated with waist circumference, and CC genotype was associated with reduced waist circumference compared to T allele carriers (β = −3.48, 95% CI: -7.11;0.15, p = 0.060). Moreover, CC genotype was also associated with reduction of the sum of skin fold thickness compared to TT genotype (β = −6.53, 95% CI: -12.96;-0.11, p = 0.046) as well as reduction of total % of body fat (β = −3.14, 95% CI: -5.18;-1.09, p = 0.003). The AC heterozygote genotype of ApaI was associated with reduced waist circumference compared to AA genotype (β = −4.37, 95% CI: -7.54;-1.20, p = 0.007). The TT genotype of BsmI was borderly associated with increased waist circumference compared to CC genotype (β = 4.05, 95% CI: -0.01;8.12, p = 0.051). GG genotype of EcoRV was associated with reduced sum of skin fold thickness compared to AA genotype (β = −7.77, 95% CI: -14.34;-1.21, p = 0.020). The GG genotype of TaqI was associated with increased waist circumference compared to AA genotype (β = 4.32, 95% CI: 0.30;8.34, p = 0.035).
Table 3

Univariable linear regression coefficients of studied polymorphisms with waist circumference, sum of skinfold thicknesses and total body fat

SNP

Genotypes

Waist circumference [cm]

Sum of skin fold thickness [cm]

Total body fat [%]

β estimate

(95% CI)

p

β estimate

(95% CI)

p

β estimate

(95% CI)

P

ApaI

AA

(ref)

(ref)

(ref)

AC

−4.37

(−7.54;-1.20)

0.007

−2.93

(−8.57;2.71)

0.308

−0.54

(−2.34;1.26)

0.555

CC

−3.12

(−6.92;0.67)

0.106

−1.85

(−8.62;4.92)

0.591

0.18

(−1.93;2.29)

0.869

BsmI

CC

(ref)

(ref)

(ref)

CT

−0.28

(−3.04;2.49)

0.845

−0.54

(−5.59;4.51)

0.834

−0.31

(−1.85;1.23)

0.689

TT

4.05

(−0.01;8.12)

0.051

−1.72

(−8.35;5.72)

0.714

−1.43

(−3.64;0.78)

0.203

EcoRV

AA

(ref)

(ref)

(ref)

AG

0.75

(−2.24;3.73)

0.624

−2.01

(−7.45;3.42)

0.467

−0.88

(−2.57;0.81)

0.305

GG

−1.32

(−5.05;2.40)

0.486

−7.77

(−14.34;-1.21)

0.020

−0.91

(−2.99;1.16)

0.389

FokI

TT

(ref)

(ref)

(ref)

TC

−1.94

(−4.79;0.90)

0.181

−4.13

(−9.11;0.86)

0.104

−1.52

(−3.14;0.10)

0.065

CC

−3.48

(−7.11;0.15)

0.060

−6.53

(−12.96;-0.11)

0.046

−3.14

(−5.18;-1.09)

0.003

TaqI

AA

(ref)

(ref)

(ref)

AG

0.65

(−2.15;3.46)

0.648

3.09

(−2.01;8.18)

0.234

0.33

(−1.24;1.91)

0.678

GG

4.32

(0.30;8.34)

0.035

3.37

(−3.69;10.43)

0.348

−0.18

(−2.47;2.11)

0.876

95%CI 95% confidence interval

The results of the multivariate analysis for the waist circumference, sum of skin fold thickness and total body fat are summarized in Table 4. Briefly, multiple effects of the investigated SNPs on anthropometric characteristics of the study subjects were observed in the multivariate analysis based on genotype three-combinations ApaI-EcoRV-FokI; BsmI-EcoRV-FokI; TaqI-EcoRV-FokI in the model.
Table 4

Multivariable models of waist circumference, sum of skin fold thickness and total body fat adjusted for sex, age, BMI, BMI2, total energy intake, smoking, type of work, shift work, comorbidities

 

SNP

Genotypes

Waist circumference [cm]

Sum of skin fold thickness [cm]

Total body fat [%]

β estimate

p

β estimate

p

β estimate

p

Var. 1

ApaI

AA

(ref)

(ref)

(ref)

AC

−1.68

(−2.88;-0.48)

0.006

−0.45

(−4.19;3.28)

0.811

0.35

(−0.27;0.96)

0.270

CC

−0.69

(−2.09;0.71)

0.333

0.08

(−4.70;4.86)

0.974

0.56

(−0.16;1.28)

0.130

EcoRV

AA

(ref)

(ref)

(ref)

AG

0.66

(−0.40;1.72)

0.220

−1.12

(−4.76;2.52)

0.545

−0.26

(−0.84;0.33)

0.389

GG

0.73

(−0.65;2.11)

0.301

−5.21

(−9.62;-0.80)

0.021

−0.51

(−1.26;0.23)

0.176

FokI

TT

(ref)

(ref)

(ref)

TC

0.19

(−0.84;1.21)

0.720

−0.66

(−3.81;2.49)

0.680

0.01

(−0.58;0.60)

0.974

CC

0.01

(−1.28;1.30)

0.987

0.79

(−3.60;5.18)

0.723

−0.21

(−0.95;0.53)

0.584

Var. 2

BsmI

CC

(ref)

(ref)

(ref)

CT

−0.50

(−1.50;0.49)

0.319

−0.42

(−3.81;2.96)

0.805

−0.02

(−0.56;0.53)

0.957

TT

0.70

(−0.82;2.23)

0.364

−1.77

(−6.68;3.15)

0.478

−1.17

(−1.95;-0.38)

0.004

EcoRV

AA

(ref)

(ref)

(ref)

AG

0.70

(−0.36;1.76)

0.196

−1.10

(−4.75;2.56)

0.554

−0.25

(−0.84;0.33)

0.392

GG

0.67

(−0.72;2.07)

0.342

−5.17

(−9.59;-0.75)

0.022

−0.47

(−1.21;0.28)

0.218

FokI

TT

(ref)

(ref)

(ref)

TC

0.20

(−0.84;1.23)

0.710

−0.73

(−3.86;2.39)

0.645

−0.04

(−0.63;0.54)

0.883

CC

−0.03

(−1.32;1.25)

0.959

0.71

(−3.67;5.08)

0.750

−0.26

(−0.99;0.48)

0.492

Var. 3

TaqI

AA

(ref)

(ref)

(ref)

AG

−1.08

(−2.14;-0.02)

0.046

0.46

(−3.12;4.03)

0.801

−0.30

(−0.86;0.26)

0.292

GG

−0.35

(−1.86;1.16)

0.649

−1.07

(−5.76;3.61)

0.651

−0.87

(−1.67;-0.07)

0.033

EcoRV

AA

(ref)

(ref)

(ref)

AG

0.65

(−0.41;1.71)

0.229

−1.07

(−4.69;2.55)

0.559

−0.25

(−0.84;0.33)

0.398

GG

0.67

(−0.73;2.07)

0.344

−5.14

(−9.54;-0.74)

0.022

−0.51

(−1.25;0.24)

0.180

FokI

TT

(ref)

(ref)

(ref)

TC

0.18

(−0.85;1.21)

0.735

−0.75

(−3.88;2.38)

0.638

−0.03

(−0.61;0.56)

0.931

CC

0.00

(−1.29;1.29)

0.999

0.67

(−3.74;5.09)

0.763

−0.20

(−0.94;0.53)

0.588

Var: combinations of three genotypes in the given model: Var 1: ApaI-EcoRV-FokI; Var 2: BsmI-EcoRV-FokI; Var 3: TaqI-EcoRV-FokI in the model

The results of haplotype analysis are reported in Table 5, using the obesity status as a dependent variable (Table 5). Briefly, no significant association of any of identified haplotypes with obesity status was observed. Similarly, no significant association of haplotypes was observed when analysing other anthropometric variables (% body fat, skin fold thickness, waist circumference) (data not shown).
Table 5

VDR haplotype association with obesity

haplotype

Haplotype frequency in controls

Haplotype frequency in obese cases

OR (95% CI)

p-value

CCTGA

0.136

0.136

1

-

CCCAA

0.088

0.052

0.60 (0.29–1.24)

0.166

CTTGA

0.017

0.003

0.24 (0.03–1.77)

0.163

CCTAA

0.116

0.113

0.87 (0.42–1.77)

0.699

ATCAA

0.011

0.009

0.54 (0.15–1.92)

0.341

ACCAA

0.026

0.025

1.13 (0.36–3.52)

0.834

CCCGA

0.127

0.128

1.00 (0.52–1.92)

0.994

ATCAG

0.058

0.056

0.98 (0.45–2.15)

0.960

ACCGA

0.035

0.029

0.70 (0.25–1.96)

0.492

ACTAA

0.046

0.042

0.83 (0.36–1.96)

0.678

ATTAG

0.089

0.096

1.01 (0.55–1.86)

0.977

ATCGG

0.070

0.086

1.18 (0.64–2.18)

0.598

ACTGA

0.024

0.049

1.89 (0.6–5.94)

0.274

ATTGG

0.081

0.102

1.20 (0.59–2.47)

0.613

Benjamini-Hochberg p-value correction for multiple hypothesis testing was performed, where appropriate

Discussion

In this study, we identified multiple associations of selected SNPs in VDR region with anthropometry. More specifically, all the investigated SNPs except for EcoRV (i.e. BsmI, TaqI, FokI, ApaI) were associated with waist circumference. EcoRV and FokI were associated with sum of skin fold thicknesses and FokI was also associated with total % of body fat.

Several studies have observed associations of VDR SNPs with anthropometric traits and yielded contradictory results. In a study by Ferrarezi, BsmI, ApaI and FokI genotypes were significantly associated with height in pubertal, but not prepubertal children, while the homozygous carriers of the minor allele of BsmI were by 0.65 z-scores (approx. equal to 4 cm) higher than the homozygous carriers of the major allele (p = 0.0006) [12]. However, these reports contradict a large analysis from 2008 that reports only negligible influence of FokI on height in large samples from Australia (N = 3906) and the Netherlands (N = 1689) [13]. On the other hand, in a more recent, large Tromsø study on 9471 subjects, significant association of BsmI and ApaI with height was observed [14].

Several studies attempted to link variability in VDR region with obesity and adiposity traits on smaller population samples. In a Saudi cohort, the BsmI minor allele (T allele) was significantly more frequent in obese individuals as well as TaqI (G allele) [15]. In a small Greek cohort, TaqI was significantly associated with obesity, (OR: 2.07, 95% CI: 1.123;3.816, p = 0.019), contributing to an elevated BMI of 3 kg/m2 per risk allele [16]. Study on 1773 healthy female adults recruited from western New York tested associations of 14 SNPs in VDR with the following 3 phenotypic measures of adiposity (body mass index (kg/m2), waist circumference (cm), and abdominal height (cm)), and significant association of another SNP within the VDR locus, the rs3782905, was observed. The mean waist circumference for women with the minor homozygote genotype of rs3782905 was 4.4 cm larger than for women with the common homozygous genotype [17]. In a Polish cohort of 351 postmenopausal women, no association of BsmI with BMI, total fat volume and visceral fat (as determined by total body dual-energy X-ray absorptiometry) was observed [18]. The observed differences in adiposity measures between our presented cohort and this closely geographically related Polish cohort can be explained by different methodology of measurement (X-ray absorptiometry in Polish study vs. bioimpedance in present study) or by different age structure of the cohorts, as the Polish cohort included only postmenopausal women, while our cohort included both men and women, and the premenopausal women were included, too.

Importantly, all of the studied polymorphisms were reported to have functional consequences. A cluster of three polymorphisms BsmI, ApaI, and TaqI at the 3′ end of the VDR gene are in near complete linkage disequilibrium among Caucasians. It is known that these polymorphisms do not influence protein sequence; however, there is indirect evidence that these polymorphisms could lead to altered transactivation [19]. Tajouri, et al. [20] and Partridge, et al. [21] investigated the functional effects of the FokI polymorphism, which has been demonstrated to alter the transcription initiation start site resulting in a shorter protein (F genotype), with demonstrated increased receptor activity. In a French cohort of adolescent girls, the EcoRV was associated with vitamin D status and calcium levels and it can be also suggested that due to its location in the promoter of the VDR gene, this SNP may have an important functional effect on expression of the gene [22]. In a novel study by Khan et al., the G allele of the rs4328262 in VDR region was associated with increased VAT volume (β = 45.7; P < 0.001), while the A allele of another SNP (rs11574070) was nominally associated with body fat percentage (β = 0.96; P = 0.002) in the African American population [23]. In this study, none of the VDR SNPs analyzed showed any link with WC or BMI, which is contradictory to our observations. As the study by Khan et al. investigated the African-American population, while our study looked on the white European population, it could be suggested that the effects of VDR on body composition could by co-dependent on the ethnicity of the subjects.

Hence, all of the mentioned polymorphisms have proved functional impact. In our study, we observed association of this SNP with all the three investigated adiposity phenotypes, i.e. waist circumference, sum of skin fold thickness and total % of body fat which contradicts the results of the study by Ochs-Balkom et al. [17]. However, the study by Osch-Balkom included only postmenopausal women recruited in frame of a case-control design for breast cancer, while the presented study included male and female population including premenopausal women.

Strenghts and limitations of the study

The major strength of the study is the relatively large sample size, standardized phenotypic measurements performed by the trained specialists and 7-day food records obtained from the individuals. Also, our study included males as well as females and was not focused only on premenopausal females. On the other hand, the major limitation may be a monocentric character of study that included only white Europeans. Moreover, phenotypic data on circulating levels of 1,25 dihydroxyvitamin D and/or 25-hydroxivitamin D are not available from our study individuals and hence presumptions on possible influence of investigated VDR SNPs on vitamin D levels are only indirect and based on literature.

Conclusion

In conclusion, data from this study confirm that the genetic variability in the VDR region may be an important factor influencing anthropometric parameters associated with obesity, i.e. waist circumference, sum of skin fold thickness and total % of body fat in the Central-European population.

Abbreviations

KMO measurement: 

Kaiser-Meyer-Olkin measurement

VDR: 

Vitamin D Receptor

Declarations

Acknowledgements

The authors are very grateful to Andrea Stejskalova and Dana Ambrozkova for excellent technical support. Moreover, the authors would like to thank David Konecny for valuable comments and assistance in language correction during preparation of the manuscript.

Funding

This research was financially supported by Masaryk University (CETOCOEN PLUS (CZ.02.1.01/0.0/0.0/15_003/0000469)) and the RECETOX Research Infrastructure (LM2015051 project of the Ministry of Education, Youth and Sports of the Czech Republic).

Availability of data and materials

The datasets used and/or analysed during the current study available from the corresponding author on reasonable request.

Authors’ contributions

All authors confirmed they have contributed to this paper. JBV was responsible for the design and interpretation of all study data and drafted the manuscript. FZ was responsible for conducting the data analysis and interpretation of results. OM, AP, JN, ZŠ and LA were responsible for the measurement/evaluation of the anthropometric data and all contributed to manuscript preparation. MGP performed the analysis of haplotypes. HP supervised the analyses and critically revised the manuscript. All authors contributed to the intellectual content and approved the final version of the manuscript.

Consent for publication

Not applicable.

Ethics approval and consent to participate

The study was conducted according to the guidelines set out in the Declaration of Helsinki and all procedures involving human subjects were approved by the Committee for Ethics of Medical Experiments on Human Subjects, Faculty of Medicine of Masaryk University (Brno, Czech Republic). Written informed consent was obtained from all subjects.

Competing interests

On behalf of all authors, the corresponding author states that there is no conflict of interest.

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.

Authors’ Affiliations

(1)
Research Centre for Toxic Compounds in the Environment, Faculty of Sciences, Masaryk University
(2)
Department of Pathological Physiology, Faculty of Medicine, Masaryk University
(3)
Research Department of Epidemiology and Public Health, University College London

References

  1. Bouillon R, Carmeliet G, Lieben L, Watanabe M, Perino A, Auwerx J, et al. Vitamin D and energy homeostasis—of mice and men. Nat Rev Endocrinol. 2014;10:79–87.View ArticlePubMedGoogle Scholar
  2. Jamka M, Woźniewicz M, Jeszka J, Mardas M, Bogdański P, Stelmach-Mardas M. The effect of vitamin D supplementation on insulin and glucose metabolism in overweight and obese individuals: systematic review with meta-analysis. Sci Rep. 2015;5:16142.View ArticlePubMedPubMed CentralGoogle Scholar
  3. Chandler PD, Wang L, Zhang X, Sesso HD, Moorthy MV, Obi O, et al. Effect of vitamin D supplementation alone or with calcium on adiposity measures: a systematic review and meta-analysis of randomized controlled trials. Nutr Rev. 2015;73:577–93.View ArticlePubMedPubMed CentralGoogle Scholar
  4. Dusso AS, Brown AJ, Slatopolsky E. Vitamin D. Am J Physiol Renal Physiol. 2005;289:F8–28.View ArticlePubMedGoogle Scholar
  5. Dusso AS. Update on the biologic role of the vitamin D endocrine system. Curr Vasc Pharmacol. 2014;12:272–7.View ArticlePubMedGoogle Scholar
  6. Egan JB, Thompson PA, Ashbeck EL, Conti DV, Duggan D, Hibler E, et al. Genetic polymorphisms in vitamin D receptor VDR/RXRA influence the likelihood of colon adenoma recurrence. Cancer Res. 2010;70:1496–504.View ArticlePubMedPubMed CentralGoogle Scholar
  7. Bienertová-Vasků J, Bienert P, Forejt M, Tomandl J, Brázdová Z, Vasků A. Genotype x nutrient association of common polymorphisms in obesity-related genes with food preferences and time structure of energy intake. Br J Nutr. 2010;103:352–9.View ArticlePubMedGoogle Scholar
  8. Ma Y, Bertone ER, Stanek EJ, Reed GW, Hebert JR, Cohen NL, et al. Association between eating patterns and obesity in a free-living US adult population. Am J Epidemiol. 2003;158:85–92.View ArticlePubMedGoogle Scholar
  9. Bienertová-Vašků J, Drábová K, Zlámal F, Tomandl J, Kýr M, Šplíchal Z, et al. Pre-treatment VD levels and VDR receptors as potential predictors of occurrence and overall survival in paediatric patients with solid tumours-a single institution pilot study. Tumour Biol. 2016;37(7):9209-19. doi:10.1007/s13277-016-4820-9.
  10. Stephens M, Smith NJ, Donnelly P. A new statistical method for haplotype reconstruction from population data. Am J Hum Genet. 2001;68:978–89.View ArticlePubMedPubMed CentralGoogle Scholar
  11. Multiple Imputation by Chained Equations (MICE): Implementation in Stata | Royston | Journal of Statistical Software [Internet]. [cited 2016 Mar 16] Available from: https://www.jstatsoft.org/article/view/v045i04
  12. Ferrarezi DAF, Bellili-Muñoz N, Nicolau C, Cheurfa N, Guazzelli IC, Frazzatto E, et al. Allelic variations in the vitamin D receptor gene, insulin secretion and parents’ heights are independently associated with height in obese children and adolescents. Metabolism. 2012;61:1413–21.View ArticlePubMedGoogle Scholar
  13. Macgregor S, Hottenga J-J, Lind PA, Suchiman HED, Willemsen G, Slagboom PE, et al. Vitamin D receptor gene polymorphisms have negligible effect on human height. Twin res. Hum. Genet. Off. J. Int. Soc. Twin Stud. 2008;11:488–94.Google Scholar
  14. Jorde R, Svartberg J, Joakimsen RM, Grimnes G. Associations between polymorphisms related to calcium metabolism and human height: the Tromsø study. Ann Hum Genet. 2012;76:200–10.View ArticlePubMedGoogle Scholar
  15. Al-Daghri NM, Guerini FR, Al-Attas OS, Alokail MS, Alkharfy KM, Draz HM, et al. Vitamin D receptor gene polymorphisms are associated with obesity and inflammosome activity. PLoS One. 2014;9:e102141.View ArticlePubMedPubMed CentralGoogle Scholar
  16. Vasilopoulos Y, Sarafidou T, Kotsa K, Papadimitriou M, Goutzelas Y, Stamatis C, et al. VDR TaqI is associated with obesity in the Greek population. Gene. 2013;512:237–9.View ArticlePubMedGoogle Scholar
  17. Ochs-Balcom HM, Chennamaneni R, Millen AE, Shields PG, Marian C, Trevisan M, et al. Vitamin D receptor gene polymorphisms are associated with adiposity phenotypes. Am J Clin Nutr. 2011;93:5–10.View ArticlePubMedGoogle Scholar
  18. Tworowska-Bardzińska U, Lwow F, Kubicka E, Łaczmański Ł, Jedzrzejuk D, Dunajska K, et al. The vitamin D receptor gene BsmI polymorphism is not associated with anthropometric and biochemical parameters describing metabolic syndrome in postmenopausal women. Gynecol. Endocrinol Off J Int Soc Gynecol Endocrinol. 2008;24:514–8.View ArticleGoogle Scholar
  19. Durrin LK, Haile RW, Ingles SA, Coetzee GA. Vitamin D receptor 3′-untranslated region polymorphisms: lack of effect on mRNA stability. Biochim Biophys Acta. 1999;1453:311–20.View ArticlePubMedGoogle Scholar
  20. Tajouri L, Ovcaric M, Curtain R, Johnson MP, Griffiths LR, Csurhes P, et al. Variation in the vitamin D receptor gene is associated with multiple sclerosis in an Australian population. J Neurogenet. 2005;19:25–38.View ArticlePubMedGoogle Scholar
  21. Partridge JM, Weatherby SJM, Woolmore JA, Highland DJ, Fryer AA, Mann CLA, et al. Susceptibility and outcome in MS: associations with polymorphisms in pigmentation-related genes. Neurology. 2004;62:2323–5.View ArticlePubMedGoogle Scholar
  22. Jehan F, Voloc A, Esterle L, Walrant-Debray O, Nguyen T-M, Garabedian M. Growth, calcium status and vitamin D receptor (VDR) promoter genotype in European children with normal or low calcium intake. J Steroid Biochem Mol Biol. 2010;121:117–20.View ArticlePubMedGoogle Scholar
  23. Khan RJ, Riestra P, Gebreab SY, Wilson JG, Gaye A, Xu R, et al. Vitamin D receptor gene polymorphisms are associated with abdominal visceral adipose tissue volume and serum Adipokine concentrations but not with body mass index or waist circumference in African Americans: the Jackson heart study. J Nutr. 2016;146:1476–82.View ArticlePubMedPubMed CentralGoogle Scholar

Copyright

© The Author(s). 2017