- HM
-
high MUFA
- HDL-C
-
HDL-cholesterol
- HS
-
high saturated fat
- LDL-C
-
LDL-cholesterol
- LF
-
low fat
- LPL
-
lipoprotein lipase
- P:S
-
PUFA:SFA ratio
- PPRE
-
peroxisome proliferator response element
- RISCK
-
Reading Imperial Surrey Cambridge King's
- sdLDL
-
small dense LDL
- SREBP
-
sterol regulatory element-binding protein
- TC
-
total cholesterol
- TZD
-
thiazolidinedione
The metabolic syndrome is defined by dyslipidaemia, glucose intolerance, hypertension and visceral obesity and is associated with an increase in the risk of type 2 diabetes and CVD( Reference Alberti, Zimmet and Shaw 1 ). Both environmental and genetic predisposition contribute to development. Among environmental factors, dietary habits (intake of fat, carbohydrate, alcohol and micronutrients) are of crucial importance. Low-fat (LF) diets reduce body weight( Reference Astrup, Grunwald and Melanson 2 ) and LF and high complex carbohydrate diets produce a significant reduction in total cholesterol (TC), LDL-cholesterol (LDL-C) and TAG( Reference Feldeisen and Tucker 3 , Reference Yu-Poth, Zhao and Etherton 4 ). Recently, the type of fat consumed, SFA, MUFA or PUFA, has received more attention. Atherogenic dyslipidaemia is characterised by increased TAG-rich lipoproteins, small LDL-C particles and reduced HDL-cholesterol (HDL-C)( Reference Grundy, Abate and Chandalia 5 ). Diets rich in SFA have an adverse effect( Reference Riccardi, Giacco and Rivellese 6 , Reference Vessby, Gustafsson and Tengblad 7 ), whereas consumption of a MUFA-rich diet at the expense of SFA promotes healthy blood lipid profiles, improves insulin sensitivity and regulates glucose levels( Reference Gillingham, Harris-Janz and Jones 8 ). Substitution of carbohydrate by MUFA generally decreases TAG. Effects of high MUFA (HM) intake on LDL-C are less well defined, with reports of a reduction( Reference Berglund, Lefevre and Ginsberg 9 , Reference Kris-Etherton, Pearson and Wan 10 ) or no effect( Reference Feldeisen and Tucker 3 ).
The lack of consistent outcomes in dietary intervention studies could reflect variation in genetic background. Understanding the nature of multiple gene–gene interaction and gene–environment interactions is pivotal in understanding the causes and progression of the metabolic syndrome and its management( Reference Roche, Phillips and Gibney 11 ). In population-based studies, the habitual dietary intake of fat is an important consideration in determining an association of any SNP with risk of metabolic syndrome.
PPAR-γ
PPAR-γ is a member of the nuclear hormone receptor superfamily( Reference Lehrke and Lazar 12 ), a transcription factor with extensive influence over expression of genes related to inflammation, adipose cell differentiation, atherosclerosis and metabolism( Reference Costa, Gallo and Letizia 13 ). The major natural ligands of PPAR-γ are PUFA, as well as prostanoids( Reference Xu, Lambert and Montana 14 ), which suggests a role in transducing nutritional to metabolic signals( Reference Semple, Chatterjee and O'Rahilly 15 ). Synthetic ligands include the thiazolidinediones (TZD)( Reference Lehrke and Lazar 12 ). On ligand-dependent activation, PPAR-γ heterodimerises with retinoid-X receptor-α and binds to a peroxisome proliferator response element (PPRE) in the promoter region of the target genes (Fig. 1).
Fig. 1. (Colour online) PPAR mechanism of action. The major natural ligands of PPAR-γ and PPAR-α are PUFA, as well as prostanoids. Upon ligand-dependent activation, both PPAR heterodimerise with retinoid-X receptor-α (RXR-α) and bind to a peroxisome proliferator response element (PPRE) in the promoter region of the target genes to initiate transcription.
Role in lipid homoeostasis
Expression of the LDL receptor gene is activated by sterol regulatory element-binding protein (SREBP)-2( Reference Hua, Yokoyama and Wu 16 ). Activated PPAR-γ up-regulates the insulin-induced gene INSIG1, the key regulator of SREBP activity( Reference Kast-Woelbern, Dana and Cesario 17 ). Reported effects of PPAR-γ agonist TZD are mainly increased HDL-C, increased size/decreased density of LDL-C particles and increased lipoprotein (a)( Reference Goldberg 18 ). PPAR-γ activation by troglitazone has been shown to reduce nuclear SREBP-2 and down-regulate LDL clearance from plasma by the liver LDL receptor( Reference Klopotek, Hirche and Eder 19 ) and troglitazone and rosiglitazone have been shown to increase plasma LDL-C concentrations( Reference Ovalle and Bell 20 ).
It is well known that n-3 fatty acids, ligands of PPAR-γ, decrease the plasma concentration of TAG( Reference Harris, Lu and Rambjor 21 ). PPAR-γ may mediate this effect through enhancement of synthesis, clearance or hydrolysis. Troglitazone has been shown to decrease SREBP-1 target genes fatty acid synthase (FASN) and glycerol-3-phosphate acyltransferase (GPAM) resulting in reduction of TAG synthesised from de novo-derived fatty acids, intracellular and secreted TAG concentrations( Reference Shah, Rader and Millar 22 ). Other PPAR-γ targets are fatty acid transport protein and CD36( Reference Tontonoz, Nagy and Alvarez 23 ), which facilitate the transport of fatty acids across cell membranes, and acyl-CoA synthetase, which facilities esterification to prevent their efflux( Reference Martin, Schoonjans and Lefebvre 24 ), so PPAR-γ also enhances clearance of TAG from plasma by this route. Lipoprotein lipase (LPL) is a rate-limiting determinant of plasma TAG hydrolysis and as the LPL gene is a target of PPAR-γ( Reference Schoonjans, Peinado-Onsurbe and Lefebvre 25 ) TAG could also be reduced by this mechanism. In summary, PPAR-γ activation by unsaturated fatty acids is expected to decrease TAG and possibly increase plasma LDL-C concentration.
PPARG gene Pro12Ala polymorphism
Four subtypes of PPAR-γ mRNA transcribed from different promoters give rise to two different PPAR-γ proteins( Reference Zieleniak, Wojcik and Wozniak 26 ). The PPAR-γ2 protein is exclusively expressed in adipose tissue( Reference Lehrke and Lazar 12 ). Since PPAR-γ regulates several genes in different tissues, variation in the PPARG gene is likely to be associated with an alteration of the expression levels of targets( Reference Costa, Gallo and Letizia 13 ). The most widely studied SNP is Pro12Ala in the PPAR-γ2 isoform, located in codon 12 of exon 3( Reference Tellechea, Aranguren and Perez 27 ). The frequency of the minor allele is 0·076 in Europeans( 28 ), lower in non-Caucasians( Reference Gouda, Sagoo and Harding 29 ).
Numerous studies have investigated association of Pro12Ala with the risk of obesity and diabetes. Results generally indicate a favourable effect of Ala12 carriage, but there are contrary findings. A meta-analysis of over 30 000 subjects, reported a significant association between Ala12 and the lowest risk of type 2 diabetes mellitus in overweight Caucasians( Reference Huguenin and Rosa 30 ). Ala12 associates with reduced risk of obesity in some studies( Reference Deeb, Fajas and Nemoto 31 ), but not others( Reference Altshuler, Hirschhorn and Klannemark 32 ). Contrary findings indicate association with increased risk of weight gain in obese patients( Reference Masud and Ye 33 ), and higher BMI, waist circumference and fat mass( Reference Robitaille, Despres and Perusse 34 , Reference Tonjes, Scholz and Loeffler 35 ). In a recent meta-analysis, Ala12 carriers had significantly increased TC and HDL-C and lower plasma TAG compared with Pro homozygotes( Reference Huang, Zhao and Zhao 36 ). Other studies have reported no association between Pro12Ala and TAG concentrations( Reference Mori, Ikegami and Kawaguchi 37 ) or plasma lipids( Reference Ringel, Engeli and Distler 38 , Reference Meshkani, Taghikhani and Larijani 39 ).
Pro12Ala and diet
An increase in PPAR-γ mRNA in adipose tissue of mice exposed to a high-fat diet( Reference Vidal-Puig, Jimenez-Linan and Lowell 40 ) suggested that dietary modulation might influence adipogenesis induced by PPAR-γ in response to raised plasma concentration of fatty acid ligands. PUFA affinities for PPAR-γ depend largely on their chain length and degree of saturation( Reference Xu, Lambert and Montana 14 ). Thus, the metabolic impact of this polymorphism is potentially dependent on gene interaction with different types of dietary fat. A direct effect was reported in functional studies, in which the PPAR-γ Ala variant had decreased binding affinity for the PPRE and thus reduced transactivation ability, both in TZD-induced adipogenesis and a luciferase reporter gene assay( Reference Deeb, Fajas and Nemoto 31 , Reference Masugi, Tamori and Mori 41 ).
The outcomes of previous studies on dietary interaction with Pro12Ala have been variable. Total fat intake was positively associated with increased BMI and waist circumference( Reference Robitaille, Despres and Perusse 34 ) and inversely correlated with plasma TC( Reference Memisoglu, Hu and Hankinson 42 ) in Pro12 homozygotes but not in Ala12 allele carriers. Memisoglu et al. ( Reference Memisoglu, Hu and Hankinson 42 ) found that intake of MUFA was inversely associated with BMI in Ala12 carriers, but not in Pro12 homozygotes. Thus, the responsiveness of Ala12 carriers to dietary fat only emerged when MUFA rather than total fat intake was analysed. Luan et al. ( Reference Luan, Browne and Harding 43 ) had previously shown greater sensitivity of Ala12 carriers to dietary PUFA in determination of BMI. Interaction between the PUFA:SFA (P:S) ratio and genotype in determining BMI was highly significant. As P:S increased, BMI decreased in Ala12 carriers but not in Pro12 homozygotes. Both findings( Reference Memisoglu, Hu and Hankinson 42 , Reference Luan, Browne and Harding 43 ) are compatible with unsaturated fatty acids acting as specific ligands for PPAR-γ( Reference Xu, Lambert and Montana 14 ) and lower transcriptional activity of the Ala variant reducing PPAR-γ-mediated adipogenesis( Reference Deeb, Fajas and Nemoto 31 ). The Ala12 variant appears to be a diet-dependent metabolic sensor, whose protective effect appears to depend on the amount and type of dietary fat.
Pro12Ala interaction with habitual dietary PUFA:SFA ratio in the Reading Imperial Surrey Cambridge King's study
The Reading Imperial Surrey Cambridge King's (RISCK) study is a parallel 2×2 factorial design compared with a control intervention, to investigate effects of dietary fat intake on variables characterising the metabolic syndrome( Reference Jebb, Lovegrove and Griffin 44 ). After a 4-week run-in on a high-SFA Western-type ‘reference diet’ (HS (high saturated fat)), subjects were randomised to continuation on the HS diet, a ‘HM’ diet in which SFA was reduced and replaced with MUFA and ‘LF diet’, in which SFA was reduced through replacement of total fat with carbohydrate. All participants followed prescribed diets for 24 weeks. A total of 549 subjects completed the RISCK study. Based on self-reported ethnicity, individuals of White, S. Asian, Black African and ‘other’ ancestry were distinguished. In view of the small sample size of the S. Asian and other ancestries and absence of the Pro12Ala SNP in Blacks, we chose to focus our genetic investigation on the White subjects only. Initially we were interested in the effect of P:S interaction with Pro12Ala genotype on plasma lipid concentrations. For this we utilised habitual intake at recruitment, as PUFA intake was constant in the interventions.
There was a significant interaction between dietary P:S ratio and genotype as a determinant of plasma concentrations of TC (P=0·02), LDL-C (P=0·002) and TAG (P=0·02) after adjustment for BMI, age and gender. When the P:S ratio was low (⩽0·33), mean plasma TC concentration in Ala12 carriers was significantly higher than in non-carriers (P=0·003). As P:S increased, the concentration of TC fell by 10%. The trend in reduction as the ratio increased from ⩽0·33 to >0·65 was significant (P=0·02). An even more significant difference was seen in LDL-C concentration between carriers and non-carriers in the lowest P:S quartile (P=0·0001). As P:S increased, the concentration fell by 19·5% in Ala12 carriers, but here the trend was NS (P>0·05). There were no significant differences in plasma TAG concentrations between Ala12 carriers and non-carriers in any P:S quartile. However, there was a significant trend in the reduction of plasma TAG in Ala12 carriers as the P:S ratio increased from 0·34 to >0·65, in which concentration fell by 50·0% (P=0·002). Plasma TAG concentrations stratified by genotype and P:S quartile are shown in Fig. 2.
Fig. 2. Mean TAG concentration with respect to quartiles of habitual dietary PUFA:SFA (P:S) ratio and PPARG Pro12Ala genotype in White subjects. The numbers of genotyped subjects with measurements in each quartile of P:S ratio ⩽0·33, 0·34–0·47, 0·48–0·65 and >0·65, were as follows: PP: 64, 76, 59, 49; PA+AA:16, 8, 19 and 17. The geometric mean concentrations of TAG are shown. Bars represent 95% CI. Dietary P:S ratio×genotype interaction determined by univariate analysis of covariance (ANCOVA) significantly influenced plasma TAG (P=0·02), after adjustment for BMI, gender and age. There was a significant trend in reduction of plasma TAG concentration between P:S ratio 0·34 to >0·65 (P=0·002) in Ala12 allele carriers.
As mentioned earlier, PPAR-γ activation by troglitazone has been shown to raise circulating LDL-C( Reference Klopotek, Hirche and Eder 19 ) and increased plasma concentration has been observed following TZD treatment( Reference Ovalle and Bell 20 ). As the PPAR-γ-Ala12 form has lower transactivational ability than the wild-type( Reference Deeb, Fajas and Nemoto 31 ), Ala12 allele carriers would be expected to show a fall in LDL-C concentration, as we observed in the higher P:S quartiles. However, at P:S <0·33, the concentration of PUFA ligand may not have been sufficient to activate LDL clearance in carriers of the low-activity isoform.
Plasma TAG concentration in Ala12 carriers fell consistently in the higher P:S quartiles. As mentioned earlier, PPAR-γ activity is expected to reduce plasma TAG( Reference Harris, Lu and Rambjor 21 ). Lindi et al. ( Reference Lindi, Schwab and Louheranta 45 ) found a significantly greater decrease in serum TAG concentration in Ala12 carriers than in Pro12 homozygotes in response to n-3 fatty acid supplementation, when the intake of SFA was below 10%, i.e. at high P:S intake. This is consistent with our finding of a fall in plasma TAG concentration in Ala12 carriers as P:S intake increased, but is inconsistent with reduced lipase activity associated with a less-active PPAR-γ-Ala isoform.
In order to determine whether gene interaction was related to decreased SFA, rather than increased PUFA, we utilised data from dietary interventions. As these did not differ in PUFA content, we were only able to investigate change in SFA. The HS and LF diets allowed comparison of high and low SFA, with constant MUFA and PUFA intake. As carriage of Ala12 was not significantly associated with change in either plasma LDL-C or TAG concentrations, the interaction does not appear to depend on a decrease in SFA.
PPAR-α
PPAR-α is a nuclear receptor mostly expressed in tissues with high levels of fatty acid oxidation, such as liver and muscle( Reference Auboeuf, Rieusset and Fajas 46 ) and regulates target genes involved in the transportation and oxidation of fatty acids( Reference Aoyama, Peters and Iritani 47 ). PPAR-α ligands can be both exogenous lipid-lowering drugs such as fibrate and fenofibrate and endogenous SFA and unsaturated fatty acid( Reference Yoon 48 , Reference Corton, Anderson and Stauber 49 ).
Like PPAR-γ, ligand-activated PPAR-α heterodimerises with retinoid-X receptor-α before binding to target gene promoters( Reference Zieleniak, Wojcik and Wozniak 26 ), which usually contain one or more PPRE( Reference Chu, Lin and Rao 50 ). In addition, PPAR-α transactivation is modulated by co-factors or co-repressors( Reference Yoon 48 ), which in the absence of a ligand inhibit its activity( Reference Pyper, Viswakarma and Yu 51 ). AMP-activated protein kinase activation increases expression of PPAR-α target genes in muscle( Reference Lee, Kim and Park 52 ). PPAR-α also appears to alter its own expression( Reference Pyper, Viswakarma and Yu 51 , Reference Pineda, Jamshidi and Flavell 53 ) and transcriptional activity is also regulated by phosphorylation, which stabilises its binding to the PPRE( Reference Yoon 48 ).
Role in lipid homoeostasis
In the liver, fibrate agonists of PPAR-α enhance fatty acid transport protein and acyl-CoA synthetase, which generate fatty acyl-CoA, carnitine palmitoyl transferase-1, essential for facilitating the entry of fatty acyl carnitine into mitochondria, and genes involved in mitochondrial β-oxidation( Reference Staels, Vu-Dac and Kosykh 54 ). Other genes involved in peroxisomal( Reference Coleman, Lewin and Van Horn 55 , Reference Jia, Qi and Zhang 56 ) and microsomal β-oxidation( Reference Reddy 57 ) are tightly regulated by PPAR-α. INSIG1, the key regulator of SREBP activity, is up-regulated by activation of PPAR-α in liver by clofibrate( Reference König, Koch and Spielmann 58 ), leading to reduction in expression of the SREBP-2 target LDL receptor gene LDLR ( Reference Hua, Yokoyama and Wu 16 ) and an increase in plasma LDL-C concentration. The hypotriglyceridaemic action of fibrates involves effects on LPL and apoC-III expression( Reference Lemieux, Salomon and Despres 59 , Reference Staels, Dallongeville and Auwerx 60 ) and on enzymes involved in TAG synthesis. PPAR-α induces LPL gene transcription( Reference Staels, Dallongeville and Auwerx 60 ) and represses expression of apoC-III, a natural inhibitor of LPL activity( Reference Lemieux, Salomon and Despres 59 ), which further enhances LPL-mediated catabolism of very low density lipoprotein (VLDL) production( Reference Staels, Dallongeville and Auwerx 60 ). Treatment with PPAR-α agonist WY14 643 reduces FASN and GPAM, resulting in reduced synthesis of TAG( Reference Shah, Rader and Millar 22 ).
PPARA gene Leu162Val polymorphism
The human PPAR-α gene PPARA gene contains fifteen coding SNP. The active isoform PPARA1 encodes the entire region, whereas PPARA2 is truncated( Reference Golembesky, Gammon and North 61 ). The most widely studied SNP Leu162Val is located in codon 162 of exon 5( Reference Tai, Demissie and Cupples 62 ). The frequency of the minor allele (Val162) is 0·042 in Europeans( 28 ). Many studies have examined association of PPARA Leu162Val with plasma lipid profiles, with conflicting results. Associations of Val162 with higher( Reference Robitaille, Brouillette and Houde 63 – Reference Shin, Kanaya and Krauss 66 ) and lower( Reference Nielsen, Hansen and Echwald 67 ) concentrations of plasma TAG have been found. Val162 has been associated with higher levels of LDL-C( Reference Tai, Demissie and Cupples 62 , Reference Sparsø, Hussain and Andersen 64 ) and with higher( Reference Flavell, Pineda Torra and Jamshidi 68 ) and lower( Reference Uthurralt, Gordish-Dressman and Bradbury 65 , Reference Tai, Collins and Robins 69 ) concentrations of HDL-C. Higher concentrations of apoA-1( Reference Flavell, Pineda Torra and Jamshidi 68 ), apoC-III( Reference Tai, Demissie and Cupples 62 , Reference Shin, Kanaya and Krauss 66 ) and apoB( Reference Vohl, Lepage and Gaudet 70 ) have been found in Val162 carriers. However, several studies have found no associations with lipid profile, BMI, body fat composition or insulin sensitivity( Reference Silbernagel, Stefan and Hoffmann 71 , Reference Skogsberg, Kannisto and Cassel 72 ). Only one other investigation has examined PPARA Leu162Val and PPARG Pro12Ala interaction in determination of plasma lipid concentrations, which found no effect in obese subjects( Reference Aberle, Hopfer and Beil 73 ).
Reports of the relative activities of the Leu162 and Val162 PPAR-α isoforms in vitro have been contradictory, possibly owing to dependence on ligand concentration. Sapone et al. ( Reference Sapone, Peters and Sakai 74 ) found Val162 allele had greater activity than Leu162 at high, but lower activity at low ligand concentration. Flavell et al. ( Reference Flavell, Pineda Torra and Jamshidi 68 ) originally found Val162 showed greater transactivation in a reporter construct. However, recently Rudkowska et al. found transcription to be higher in Leu162 than Val162 constructs containing the LPL PPRE, after n-3 fatty acid transactivation( Reference Rudkowska, Caron-Dorval and Verreault 75 ). They also found an inverse correlation between LPL activities and plasma TAG levels in Leu162 homozygotes but not in Val162 carriers( Reference Rudkowska, Verreault and Barbier 76 ), suggesting that Val162 has lower transactivational ability than Leu162 under physiological conditions.
Leu162Val and diet
Reports of PPARA Leu162Val interaction with fatty acid intake in determination of plasma lipids are inconsistent, including no interaction with PUFA( Reference Tai, Corella and Demissie 77 ), Val162 allele association with higher TC, LDL-C and apoA1 after a high-PUFA diet( Reference Paradis, Fontaine-Bisson and Bossé 78 ) and higher TAG and apoCII after low PUFA intake( Reference Tai, Corella and Demissie 77 ). In the latter, when PUFA intake was less than 4%, Val162 carriers had higher plasma TAG compared with Leu162 homozygotes, but when PUFA intake was more than 8%, Val162 allele carriers had lower plasma TAG. In Leu162 homozygotes, waist circumference increased with a higher intake of dietary fat, but no significant interaction was found in determining TC, LDL-C, HDL-C or apoB concentrations( Reference Robitaille, Brouillette and Houde 63 ). Only one other study has examined PPARG Pro12Ala and PPARA Leu162Val after dietary intervention. After a 2·5-year low-energy diet, in non-diabetic obese women there were significant favourable changes in lipid profile, but no significant interactive effects on anthropometric or biochemical characteristics at baseline or at the follow-up( Reference Aldhoon, Zamrazilova and Aldhoon Hainerová 79 ).
PPARG Pro12Ala andPPARA Leu162Val interaction in the Reading Imperial Surrey Cambridge King's study: effect of MUFA
We hypothesised that carriage of PPARG Pro12Ala and PPARA Leu162Val allelic combinations might influence concentration of plasma lipids according to the availability of dietary unsaturated fatty acid ligands. At baseline, after a 4-week run-in on the HS diet, carriage of the PPARG Ala12 allele was associated with a modest increase in plasma TC (n 415; P=0·05), LDL-C (P=0·04) and apoB (P=0·03) after adjustment for BMI, age, gender and ethnicity. Although SFA are relatively poor stimulators of PPAR-γ activity( Reference Xu, Lambert and Montana 14 ) these outcomes are likely to reflect lower transactivation of target genes by the PPAR-γ-Ala form( Reference Deeb, Fajas and Nemoto 31 ). The PPARA Leu162Val genotype was not associated with concentrations of plasma lipids at baseline, but PPARA Val162Leu×PPARG Pro12Ala genotype interaction influenced TC (P=0·04) concentration after adjustment for covariates.
After HM and LF diets, plasma TC, LDL-C and apoB concentrations were reduced (P<0·001), but surprisingly there was no change in TAG concentration( Reference Jebb, Lovegrove and Griffin 44 ). Independent associations of PPARG Pro12Ala or PPARA Leu162Val genotypes with changes in concentrations of plasma lipids with respect to baseline were NS after randomisation to diets. However, there was significant interaction between the two genotypes as determinants of plasma LDL-C concentration, (P=0·003) and small dense LDL (sdLDL) as a proportion of LDL (P=0·012) after adjustment for change in BMI, age, gender and ethnicity. Carriage of both variant alleles was associated with a greater reduction in LDL-C and proportion as sdLDL after HM diet than after LF diet. PUFA is a stronger activator of PPAR than MUFA( Reference Xu, Lambert and Montana 14 ), but was constant in both interventions. As PPAR variant carriage affected plasma lipids only after the HM diet, the effects may depend on HM concentration.
Fig. 3 shows the follow-up concentrations of plasma LDL-C and sdLDL as a proportion of LDL after the HM and LF diets above the baseline, with respect to PPARG Pro12Ala and PPARA Leu162Val genotype combinations. The results of gene×gene interaction were highly significant for these data. Our ANOVA model used the variability of the whole dataset to measure the background variation, and produced evidence of a significant effect of gene–gene interaction on LDL-C and proportion as sdLDL. The significance should nevertheless be treated with caution and confirmation awaits replication in a larger sample.
Fig. 3. Interaction between PPARG Pro12Ala and PPARA Leu162Val genotype groups after dietary treatments influences plasma LDL-cholesterol (LDL-C) concentration and small dense LDL (sdLDL) as proportion of LDL. Interaction between PPARG Pro12Ala and PPARA Leu162Val genotypes was a significant determinant of change in plasma concentrations of (a) LDL-C (P=0·003) and (b) sdLDL as proportion of LDL (P=0·012) after high MUFA (HM) and low fat (LF) diets, after adjustment for baseline values, change in BMI, age, gender and ethnicity using three-way ANOVA. PP represents subjects homozygous for the PPARG Pro12 allele and PA+AA carriers of the Ala12 allele. LL represents subjects homozygous for the PPARA Leu162 allele and LV+VV carriers of the Val162 allele. Mean follow-up concentrations of LDL-C (mmol/l) and sdLDL as proportion of LDL (%) adjusted for baseline values after 24 weeks on HM or LF diets are shown. Bars indicate 95% CI. The figure is based on subjects with genotypes for both SNP and measurements of plasma lipids after HM and LF diets. The numbers of subjects in each genotype group LL/PP, LL/PP+PA, LV+VV/PP and LV+VV/PP+PA were as follows: HM diet: 121, 24, 17, 4; LF diet: 126, 34, 9 and 4.
As explained earlier, PPAR-γ activation by troglitazone reduces nuclear SREBP-2 and down-regulates LDL clearance from plasma by SREBP-2 target, the liver LDL receptor( Reference Klopotek, Hirche and Eder 19 ). Expression of the LDL receptor is also reduced by clofibrate( Reference König, Koch and Spielmann 58 ). Activation of PPAR-α and PPAR-γ would thus impair LDL receptor expression, down-regulate LDL clearance from plasma and increase circulating LDL-C concentration, as found in response to TZD( Reference Ovalle and Bell 20 ), but LDL apoB-100 levels generally decrease in response to fibrates( Reference Shah, Rader and Millar 22 ). PPAR-γ-Ala12 and PPAR-α-Val162 forms have lower transactivational ability than the wild types( Reference Deeb, Fajas and Nemoto 31 , Reference Rudkowska, Verreault and Barbier 76 ). Hence, carriers of PPARG Ala12 and PPARA Val162 would express higher LDL receptor activity, leading to maximum clearance and the largest fall in LDL-C concentration, as we observed. All the other genotype combinations showed smaller reductions in LDL-C after the HM diet. As Ala12 was associated with higher TC concentration and interaction with Val162 yielded higher LDL-C after the HS diet, the lower LDL-C in carriers of both variants after the HM diet appears to be a response to increased availability of MUFA.
One of the most consistent effects of TZD is to increase the mean LDL particle size and/or reduce LDL density( Reference Goldberg, Kendall and Deeg 80 ). Were PPAR-γ to be implicated directly, carriage of the lower activity PPAR-γ-Ala form would be expected to associate with a higher proportion of small LDL particles. This was found to be the case by Hamada et al. ( Reference Hamada, Kotani and Tsuzaki 81 ), where PPARG Ala12 carriers had a significantly higher proportion of sdLDL fractions four to seven independent of lipid concentration. As mentioned previously, high-fat intake is associated with an increase in large LDL and decrease in sdLDL( Reference Krauss and Dreon 82 ). Bouchard-Mercier et al. ( Reference Bouchard-Mercier, Godin and Lamarche 83 ) found no significant change in LDL peak particle diameter in PPARG Pro12 homozygotes or Ala12 carriers after high SFA intake, but a significant increase in LDL peak particle diameter in Ala12 carriers after high intake of PUFA, which unlike SFA are PPAR-γ activators( Reference Xu, Lambert and Montana 14 ). They found that high SFA intake associated with larger LDL particle size in PPARA Leu162 homozygotes, but with a higher proportion of sdLDL in Val162 carriers. Fibrate ligands of PPAR-α can reduce production of VLDL( Reference Shah, Rader and Millar 22 ) and lower sdLDL( Reference Caslake, Packard and Gaw 84 , Reference Berneis and Krauss 85 ), and so in carriers of the less-active PPAR-α-Val form, activation by dietary ligands could result in a shift to a higher proportion of sdLDL. We found no significant change in the proportion of sdLDL in carriers of PPARG Ala12 or PPARA Val162 on switching from the HS diet at baseline to the HM or LF diets, but a significant reduction in the proportion of sdLDL in carriers of both PPARA Val162 and PPARG Ala12 alleles after the HM diet. This cannot be explained by reduced activity of both variants, because as indicated above, this would be expected to lead to a higher proportion of sdLDL.
Adiponectin
Adiponectin is a 244-amino-acid plasma protein secreted exclusively by adipocytes. It is an insulin sensitising adipokine with anti-atherogenic, anti-diabetic and anti-inflammatory functions( Reference Oh, Lee and Rhee 86 ). Plasma adiponectin concentration is negatively correlated with human obesity, hypertension, insulin resistance and increased plasma TAG concentrations( Reference Isobe, Saitoh and Takagi 87 , Reference Brochu-Gaudreau, Rehfeldt and Blouin 88 ). In the circulation, adiponectin is present in three oligomeric complexes, with different biological functions, acting though distinct signalling pathways. The basic trimer is the low-molecular-weight isoform( Reference Tsao, Tomas and Murrey 89 ). The hexametric isoform is formed through the association of two homotrimers( Reference Wang, Lam and Yau 90 ). High-molecular-weight adiponectin is the biologically active form.
Effect of gender, age and ethnicity
The sexual dimorphism of adiponectin is well known; males have significantly lower plasma concentrations than females( Reference Marques-Vidal, Bochud and Paccaud 91 ). The gender differences have been attributed primarily to the inhibitory effect of testosterone on adiponectin production established in vitro ( Reference Nishizawa, Shimomura and Kishida 92 ). Adiponectin concentrations generally increase with age( Reference Adamczak, Rzepka and Chudek 93 ), mainly explained by changes in sex hormones( Reference Isobe, Saitoh and Takagi 87 ). As insulin sensitivity declines with age, this may reflect development of resistance, or survival in those with higher concentrations. Cohen et al. ( Reference Cohen, Gammon and Signorello 94 ) reported significantly lower concentrations in Black than in White individuals.
Effect on insulin sensitivity
Adiponectin acts on two receptors AdipoR1 in skeletal muscle and AdipoR2, more abundant in the liver( Reference Kadowaki and Yamauchi 95 ) (Fig. 4). In the liver, adiponectin activates PPAR-α and AMP-activated protein kinase, a key energy sensor that maintains cellular energy homoeostasis, via AdipoR2. Activation of AMP-activated protein kinase down-regulates enzymes involved in gluconeogenesis, phosphoenolpyruvate carboxykinase and glucose-6-phosphatase. It also increases the inhibitory phosphorylation of acetyl coenzyme A carboxylase, promoting fatty acid oxidation, and inhibits the action of genes such as SREBP, required for fatty acid synthesis. The activation of PPAR-α by AMP-activated protein kinase decreases TAG in the liver by stimulating fatty acid oxidation( Reference Kadowaki and Yamauchi 95 ). In muscle, adiponectin acts via AdipoR1 to stimulate fatty acid oxidation and glucose utilisation. AdipoR1 targets genes such as CD36, involved in fatty acid transport, acyl-CoA oxidase, involved in fatty acid oxidation and uncoupling protein-2, involved in energy dissipation as heat( Reference Kadowaki and Yamauchi 95 ). Therefore, adiponectin increases fatty acid oxidation in liver and muscle, leading to reduced adipose tissue mass, a fall in pro-inflammatory cytokines and promotion of insulin signalling.
Fig. 4. Adiponectin mechanism of action. Adiponectin activates AMP kinase (AMPK) and PPAR-α in liver and skeletal muscle. In muscle, globular and full-length adiponectin activate AMPK, stimulating inhibitory phosphorylation of acetyl-CoA carboxylase (ACC), promoting fatty-acid oxidation, and GLUT4 translocation promoting glucose uptake. Activation of PPAR-α also leads to stimulation of fatty-acid oxidation and decreased TAG. In the liver, full-length adiponectin activates AMPK, thereby reducing enzymes involved in gluconeogenesis, also increasing phosphorylation of ACC and stimulating fatty-acid oxidation. Activation of PPAR-α decreases TAG as in muscle. All actions increase insulin sensitivity. PEPCK, phosphoenolpyruvate carboxykinase; G6Pase, glucose-6-phosphatase. (Adapted from Kadowaki and Yamauchi( Reference Kadowaki and Yamauchi 95 )).
Effect on plasma lipid profile
Adiponectin is correlated negatively with plasma TAG( Reference Hotta, Funahashi and Arita 96 ) and positively with HDL-C concentration( Reference Matsubara, Maruoka and Katayose 97 ). The mechanism may relate to insulin resistance. Insulin is a well-known stimulator of adipose tissue LPL activity( Reference Nellemann, Gormsen and Christiansen 98 ), which catalyses the rate-limiting step in the hydrolysis of the TAG component in circulating VLDL and chylomicrons( Reference Schittmayer and Birner-Gruenberger 99 ). Adiponectin promotes mitochondrial fatty acid oxidation and reduction in circulating fatty acids, which in turn promotes LPL activity( Reference Ganguly, Schram and Fang 100 ). Adiponectin also activates PPAR-α, which up-regulates expression of apo-proteins A-I and A-II, promoting hepatic HDL-C secretion( Reference Mohamadkhani, Sayemiri and Ghanbari 101 ).
ADIPOQ gene polymorphisms
Fifty-three SNP have been identified at the adiponectin gene ADIPOQ locus( Reference Heid, Wagner and Gohlke 102 ). There are many, often conflicting, reports of SNP associations with circulating adiponectin concentrations( Reference Yang and Chuang 103 ) and various metabolic syndrome traits. In an earlier study of SNP at the ADIPOQ locus −11391 G/A, −10066 G/A, −7734 A/C and +276 G/T in this laboratory, we found −10066G, −11391A, −7734A and +276T were significantly associated with higher serum adiponectin concentration in two large cohorts( Reference Kyriakou, Collins and Spencer-Jones 104 ). Association of elevated adiponectin with the −11391 A-allele has been reported widely( Reference Heid, Wagner and Gohlke 102 ,105–Reference Bouatia-Naji, Meyre and Lobbens 108 ), although one group found lower adiponectin in G-allele carriers( Reference Petrone, Zavarella and Caiazzo 109 ). Associations between +276G and lower adiponectin concentrations have also been reported in Spanish( Reference Gonzalez-Sanchez, Zabena and Martinez-Larrad 110 ), European( Reference Mackevics, Heid and Wagner 111 ), Korean( Reference Jang, Lee and Chae 112 ) and Japanese( Reference Hara, Boutin and Mori 113 ) subjects. The −10066G allele has also been associated with higher adiponectin concentration elsewhere( Reference Woo, Dolan and Deka 107 ).
Association between ADIPOQ gene variants and metabolic syndrome risk factors has been established in many studies. +276G carriage predisposed to higher CVD risk in Koreans( Reference Jang, Lee and Chae 112 ). In Italians +276T was a risk allele in one study( Reference Filippi, Sentinelli and Trischitta 114 ) and protective in another( Reference Chiodini, Specchia and Gori 115 ). In Spanish subjects +276G was associated with impaired glucose tolerance( Reference Gonzalez-Sanchez, Zabena and Martinez-Larrad 110 ) and higher homoeostatic model assessment of insulin resistance in Korean( Reference Jang, Lee and Chae 112 ), Italian( Reference Menzaghi, Ercolino and Di Paola 116 ) and Japanese( Reference Hara, Boutin and Mori 113 ) subjects. +276T has been associated with lower( Reference Menzaghi, Ercolino and Di Paola 116 , Reference Menzaghi, Ercolino and Salvemini 117 ) and higher( Reference Melistas, Mantzoros and Kontogianni 118 ) homoeostatic model assessment of insulin resistance. Higher LDL-C and lower HDL-C levels have been found in +276T allele carriers( Reference Berthier, Houde and Cote 119 ) and +276G was associated with higher concentration of TAG in Koreans( Reference Jang, Lee and Chae 112 ). The +276G allele was associated with higher BMI in Italians( Reference Menzaghi, Ercolino and Di Paola 116 ) but with lower BMI in Swedish and African Americans( Reference Ukkola, Ravussin and Jacobson 120 , Reference Beebe-Dimmer, Zuhlke and Ray 121 ). Higher waist:hip ratio was found in carriers of the −11391A allele( Reference Dolley, Bertrais and Frochot 122 ) and there are reports of increased risk of insulin resistance and type 2 diabetes associated with the −11391 G/G genotype( Reference Vasseur, Helbecque and Dina 105 , Reference Petrone, Zavarella and Caiazzo 109 ).
ADIPOQ polymorphisms and diet
Inconsistent associations between the ADIPOQ variants and serum adiponectin, BMI and insulin resistance suggest that environmental influences may be influential. A few studies have explored the relationship between dietary factors and adiponectin concentrations or gene–nutrient interactions involving SNP at the ADIPOQ locus. In the largest study to date, in American Whites, −11391 A-allele carriers in the highest fiftieth percentile of MUFA intake had lower BMI and risk of obesity compared with G/G homozygotes( Reference Warodomwichit, Shen and Arnett 123 ). In another study, after switching from an SFA- to MUFA-rich diet, −11377 C/C homozygotes were significantly less insulin resistant compared with G-allele carriers( Reference Pérez-Martínez, López-Miranda and Cruz-Teno 124 ). In a recent study, an interaction between ADIPOQ −11377 C/G genotype with SFA, but not MUFA or PUFA, significantly affected homoeostatic model assessment of insulin resistance, but there were no significant effects on serum adiponectin concentration( Reference Ferguson, Phillips and Tierney 125 ).
ADIPOQ and PPAR-γ
One potential pathway for dietary interaction with ADIPOQ is via activation of PPAR-γ( Reference Kim, Jang and Paik 126 ). PPAR-γ agonists such as TZD have been clearly shown to increase serum adiponectin concentrations in both human subjects and rodents( Reference Iwaki, Matsuda and Maeda 127 ). PUFA have been reported to increase plasma adiponectin concentrations and may up-regulate ADIPOQ by acting as ligands of PPAR-γ( Reference Iwaki, Matsuda and Maeda 127 ). Both natural and artificial ligands of PPAR-γ enhance the expression of adiponectin mRNA in adipose tissue and dramatically increase plasma concentration of adiponectin( Reference Maeda, Takahashi and Funahashi 128 ). The mechanism involves a functional PPRE and a responsive element of liver receptor homolog-1 in the ADIPOQ promoter( Reference Iwaki, Matsuda and Maeda 127 ).
ADIPOQ −10066 G/A in the Reading Imperial Surrey Cambridge King's study: effect of MUFA
Diets low in carbohydrate( Reference Pischon, Girman and Rifai 129 ) and high in unsaturated fat increase adiponectin( Reference Esposito, Pontillo and Di Palo 130 ). We hypothesised that variants in ADIPOQ could interact with dietary intake of unsaturated fat and age to influence serum adiponectin in the absence of significant change in BMI, in RISCK study participants.
After 4-week run-in on the HS diet, there were significant differences between males and females in fasting glucose and TAG (higher in males), HDL-C, adiponectin, insulin sensitivity and percentage body fat (lower in males). Adiponectin positively correlated with age (β=0·217, P<0·001) and negatively with BMI (β=−0·161, P<0·001) in agreement with previous reports( Reference Isobe, Saitoh and Takagi 87 , Reference Adamczak, Rzepka and Chudek 93 ). Adiponectin was significantly higher in White Europeans than in S. Asians (P=0·001) and Black Africans (P=0·001) as reported previously( Reference Cohen, Gammon and Signorello 94 ) and higher in females (mean 11·1 (sd 6·2) μg/ml) than males (mean 8·5 (sd 4·1) μg/ml) (P<0·001), as is well known( Reference Marques-Vidal, Bochud and Paccaud 91 ). However, there were no significant interactions between gender×age (P=0·697), gender×BMI (P=0·139) or gender×ethnicity (P=0·15) in determination of serum adiponectin concentration.
Surprisingly, replacement of SFA by isoenergetic MUFA or carbohydrate diets for 24 weeks did not significantly improve adiponectin concentration. Previously, we reported no significant effect on insulin sensitivity following this dietary regimen( Reference Jebb, Lovegrove and Griffin 44 ). Small changes in adiponectin concentration after dietary intervention may not have been sufficient to affect insulin sensitivity, or the intervention period may not have been long enough to produce an effect. This is consistent with other reports( Reference Yannakoulia, Yiannakouris and Bluher 131 – Reference Peake, Kriketos and Denyer 133 ). Long-term effects were seen only after a 10-year Mediterranean diet in diabetic women( Reference Mantzoros, Williams and Manson 134 ). These data suggest that adiponectin concentrations are unlikely to be affected by relatively short-term dietary changes, but reflect intakes over longer time periods( Reference Pischon, Girman and Rifai 129 ).
We hypothesised that stratification by genotype might uncover influential interaction between diet and ADIPOQ variants in determination of serum adiponectin concentration following dietary intervention. Our genetic investigations were based on the White subjects. We investigated four SNP which we previously showed to have the strongest replicated associations with serum adiponectin( Reference Kyriakou, Collins and Spencer-Jones 104 ): −11391 G/A is located in the promoter region −10066 G/A and −7734 A/C are both located in intron 1 and +276 G/T is in intron 2.
After the 4-week run-in on HS diet, +276T was associated with higher (n 340; P=0·006) and −10066A with lower serum adiponectin concentration (n 360; P=0·03) after adjustment for covariates, in agreement with previous reports( Reference Kyriakou, Collins and Spencer-Jones 104 , Reference Woo, Dolan and Deka 107 ).There were no significant differences in the change in serum adiponectin concentration after HM or LF diets, with the exception of −10066 G/A. After the HM diet GG subjects showed a 3·8% increase (95% CI −0·1, 7·7) and GA+AA subjects a 2·6% decrease (95% CI −5·6, 0·4) in serum adiponectin (P=0·006 for difference, after adjustment for change in BMI, age and gender). However, gene×diet interaction in determination of serum adiponectin was NS (P=0·12) after adjustments( Reference AlSaleh, O'Dell and Frost 135 ).
Activation of PPAR-γ by unsaturated fatty acids increases with chain length and degree of unsaturation( Reference Sanderson, de Groot and Hooiveld 136 ). The switch from SFA to MUFA could lead to increased expression of the ADIPOQ gene and serum adiponectin concentration through increased availability of PPAR-γ-activating ligands. The PPRE lies in a 1·3 kb linkage disequilibrium block( Reference Heid, Wagner and Gohlke 102 ). If the −10066A-allele was in linkage disequilibrium with a variant in the PPRE reducing affinity for the receptor, this could account for higher serum adiponectin in response to MUFA in GG homozygotes and the lower concentration in A-allele carriers.
We were interested to discover whether the strong relationship between adiponectin concentration and age seen at baseline was modified by diet. There was no significant interaction between either genotype or diet in determining adiponectin concentration. We then looked at whether age×genotype interaction was influential after dietary intervention. Fig. 5 compares the effect of HM and LF diets on % change in serum adiponectin concentration in White −10066 GG homozygotes and A-allele carriers. In GG homozygotes over 40 years of age, adiponectin concentration increased progressively after the HM diet and decreased after the LF diet. The difference in % change in serum adiponectin between GG subjects on HM and LF diets in the oldest 61–70-year age group was significant (P=0·003). In A-allele carriers there was little change in serum adiponectin concentration compared with baseline with increasing age, after HM or LF diet. Interaction between gene×age×diet in determination of change in serum adiponectin concentration approached significance after adjustment for gender and change in BMI (n 303; P=0·07). However, interaction between gene×age×diet×gender was NS after adjustment for change in BMI( Reference AlSaleh, O'Dell and Frost 135 ).
Fig. 5. Effect of high-MUFA (HM) and low-fat (LF) diets on adiponectin concentration with respect to −10066 G/A genotype and age in White subjects. % changes (95% CI) in geometric mean adiponectin concentration adjusted for change in BMI and gender are shown in each age group, after subjects consumed HM (n 151) or LF (n 152) diets. (a) −10066 GG subjects (n 111) and (b) −10066 GA+AA subjects (n 192). The number of genotyped subjects in each age group 35–40, 41–50, 51–60 and 61–70 years were as follows: HM diet: GG 6, 18, 13, 20; GA+AA 13, 22, 30 and 29; LF diet: GG 9, 8, 21, 16; GA+AA 13, 27, 36 and 22. Interaction between gene×age×diet in determination of change in serum adiponectin concentration found by analysis of covariance (ANCOVA) was NS after adjustment for change in BMI (n 303; P=0·07). *Denotes significant difference in % change in serum adiponectin between GG subjects on HM and LF diets (P=0·003). (From AlSaleh A et al. ( Reference AlSaleh, O'Dell and Frost 135 ).)
Serum adiponectin might be expected to be lower in GG subjects after the LF diet, in which carbohydrates replace PPAR-γ-activating fatty acids, than after the HM diet. In A-allele carriers, substitution of carbohydrate for MUFA would have little effect if reduced affinity of the PPRE, rather than ligand activation were to be the rate-limiting step. This would be compatible with other reports of lower serum adiponectin after high-carbohydrate( Reference Pischon, Girman and Rifai 129 ) and higher serum adiponectin with a diet rich in MUFA( Reference Yeung, Appel and Miller 137 ). If aging is associated with the development of adiponectin resistance, the change in adiponectin concentrations may reflect a capability of responding by increasing production after HM, but not LF, diets.
Conclusion
The strength of the RISCK study lies in its design as a randomised, tightly controlled feeding trial with high adherence and retention rates and diets with practical relevance to the general population. Analysis of White subjects showed that at the lowest PUFA:SFA intake, carriage of the less active PPAR-γ Ala12 isoform associated with higher plasma TC and LDL-C. The significant trends in the reduction of plasma TC and TAG in Ala12 carriers as the P:S ratio increased suggests that these subjects might be advised to maintain a high PUFA:SFA intake ratio to reduce plasma concentrations of atherogenic lipids. sdLDL particles are recognised as an important risk factor for CVD and numerous dietary elements have a significant impact on several characteristics of the LDL size phenotype. Significant predictive value of individual disease risk or responses to diet could potentially be gained by combining genotype information from the PPARA Leu162Val and PPARG Pro12Ala loci. The switch from SFA to MUFA could lead to increased expression of the ADIPOQ gene and serum adiponectin concentration through increased availability of PPAR-γ-activating ligands. In White ADIPOQ −10066 GG homozygotes, increase in adiponectin with age suggests that a HM diet may help to increase adiponectin concentrations with advancing years.
Limitations to these SNP association studies include relatively small sample sizes, and multiple testing remains a controversial issue in interpretation. Replication in other cohorts is the most reliable method to distinguish true from false-positive associations. Substantiated effects of common SNP in modifying the outcome of dietary intervention studies in larger samples should help in the identification of individuals at risk of complex disease who would benefit from personalized dietary recommendations.
Acknowledgements
Research based on the RISCK study was supported by the UK Food Standards Agency (project NO2031), registered at clinicaltrials.gov as ISRCTN29111298. We acknowledge the contributions of the principal investigators and their staff at the four other centres: Bruce Griffin (University of Surrey), Gary Frost (Imperial College London), Julie Lovegrove (University of Reading) and Susan Jebb (Cambridge). Foods were supplied by Unilever Food and Health Research Institute (Unilever R&D, Vlaardingen, The Netherlands), Cereal Partners UK (Welwyn Garden City, Hertfordshire, UK), Grampian (Banff, UK), Weetabix Ltd (Kettering, UK) and Sainsbury's Supermarkets Ltd (London, UK). None of these providers had any role in the design and implementation of the study or analysis and interpretation of the data. A. A. gratefully acknowledges a PhD studentship from the Saudi Arabian Ministry of Higher Education. T. A. B. S. has acted as a consultant to Seven Seas and is a member of the Scientific Advisory Committee for the Global Dairy Platform and external scientific review committee of the Malaysian Palm Oil Board, and chairs Cadbury's Global Nutrition Advisory Panel. All research groups received products from a range of food companies gratis for research purposes, including Archer Daniel Mills, Croda, Matthews Foods, Nestle, PepsiCo, Jordan, GSK, and Unilever. A. A. and S. D. O. declare no conflicts of interest. A. A. drafted this paper and S. D. O. and T. A. B. S. discussed and modified it.
