Skip main navigation
×

Do Genetic Modifiers of High-Density Lipoprotein Cholesterol and Triglyceride Levels Also Modify Their Response to a Lifestyle Intervention in the Setting of Obesity and Type-2 Diabetes Mellitus?

The Action for Health in Diabetes (Look AHEAD) Study
and for the Genetics Subgroup of the Action for Health in Diabetes (Look AHEAD) Study
Originally publishedhttps://doi.org/10.1161/CIRCGENETICS.113.000042Circulation: Cardiovascular Genetics. 2013;6:391–399

Abstract

Background—

High-density lipoprotein cholesterol (HDL-C) and triglycerides are cardiovascular risk factors susceptible to lifestyle behavior modification and genetics. We hypothesized that genetic variants identified by genome-wide association studies as associated with HDL-C or triglyceride levels modify 1-year treatment response to an intensive lifestyle intervention, relative to a usual care of diabetes mellitus support and education.

Methods and Results—

We evaluated 82 single-nucleotide polymorphisms, which represent 31 loci demonstrated by genome-wide association studies to be associated with HDL-C and triglycerides, in 3561 participants who consented for genetic studies and met eligibility criteria. Variants associated with higher baseline HDL-C levels, cholesterol ester transfer protein (CETP) rs3764261 and hepatic lipase (LIPC) rs8034802, were found to be associated with HDL-C increases with intensive lifestyle intervention (P=0.0038 and 0.013, respectively) and had nominally significant treatment interactions (P=0.047 and 0.046, respectively). The fatty acid desaturase-2 rs1535 variant, associated with low baseline HDL-C (P=0.017), was associated with HDL-C increases with intensive lifestyle intervention (0.0037) and had a nominal treatment interaction (P=0.035). Apolipoprotein B (rs693) and LIPC (rs8034802) single-nucleotide polymorphisms showed nominally significant associations with HDL-C and triglyceride changes with intensive lifestyle intervention and a treatment interaction (P<0.05). Phosphatidylglycerophosphate synthase-1 single-nucleotide polymorphisms (rs4082919) showed the most significant triglyceride treatment interaction in the full cohort (P=0.0009).

Conclusions—

This is the first study to identify genetic variants modifying lipid responses to a randomized lifestyle behavior intervention in overweight or obese individuals with diabetes mellitus. The effects of genetic factors on lipid changes may differ from the effects on baseline lipids and are modifiable by behavioral intervention.

Introduction

Significant changes in eating and physical activity behaviors that cause weight loss can improve insulin resistance and other biological markers relevant to cardiovascular disease risk in patients who are obese and have type-2 diabetes mellitus (T2D).1 Lifestyle modification is first-line therapy in efforts to raise high-density lipoprotein cholesterol (HDL-C) and to lower triglyceride levels.2 However, the response to behavioral intervention is inconsistent. HDL-C and triglycerides levels are heritable,3,4 as are lipid responses to overfeeding5 and exercise training,6 suggesting that genetic factors contribute to the lipid response to behavioral intervention. What is currently largely unknown is which common genetic factors influence or predict the HDL-C and triglyceride level response to behavior modification. Collectively, genome-wide association studies (GWASs) have identified single-nucleotide polymorphisms (SNPs) in ≥95 loci, the combined effects of which account for ≈10% to 12% of the total variance in HDL-C and triglyceride levels.7

Clinical Perspective on p 399

The Action for Health in Diabetes (Look AHEAD) study is a multicenter trial that randomly assigned participants with T2DM who were overweight or obese to an intensive lifestyle intervention (ILI), with the goal of producing 7% weight loss through calorie restriction and physical activity, or to diabetes mellitus support and education (DSE) with no weight loss or physical activity goals.8 Compared with large observational cohort studies, the Look AHEAD study has the unique strength of being able to analyze the effect of genetic factors on lipid trait change in response to 2 randomly assigned behavioral interventions. At baseline (year 0), the majority of Look AHEAD participants demonstrated decreased fitness9 and consumed a diet that exceeded recommended nutrient intake for total fat and saturated fat with reduced fiber content.10 After the first year (year 1), participants in the ILI arm lost significantly greater amounts of weight and showed greater improvement in fitness, waist circumference, and indices of diabetes mellitus control, including metabolic syndrome, diabetes mellitus medication use, hemoglobin A1-c, and fasting glucose, compared with participants in the DSE arm.11 Participants in both groups demonstrated improved HDL-C and triglyceride levels at year 1; however, the ILI group showed greater improvements compared with the DSE group.11 Changes in low-density lipoprotein cholesterol were similar in both groups.

Gene variants that modify the behavioral treatment response of HDL-C and triglycerides may have potential for directing personalized medical or behavioral treatment programs. With this overarching goal, we hypothesized that GWAS-identified SNPs associated with HDL-C and triglyceride levels are also associated with HDL-C and triglyceride responses to the behavioral therapy implemented in Look AHEAD. To test this hypothesis, we examined the interaction of SNP with treatment arm at 1 year and analyzed the HDL-C and triglyceride response to ILI and DSE between Look AHEAD carriers of the minor allele of HDL-C and triglyceride SNPs included on the ITMATBroad-CARE (IBC) array chip.

Material and Methods

Detailed methods are described in the Supplement. Of 4099 Look AHEAD participants whose DNA was collected, 353 subjects taking niacin or fibrates were excluded because of their effects on HDL-C and triglycerides, and 185 subjects were excluded because of genotyping failure yielding an effective sample size of 3561 participants. All participants included in this study signed informed consent for participation in the Look AHEAD trial and genetic analyses with Institutional Review Board approval at their local institution; this analysis was approved by the Tufts Medical Center and Miriam Hospital Institutional Review Boards. Genotyping was performed using the IBC chip. A search of published literature using the HUGE Navigator in April 2012 for HDL-C and triglycerides GWASs returned 89 SNPs associated with HDL-C alone, 48 SNPs associated with triglycerides alone, and 22 SNPs associated with both HDL-C and triglycerides; 82 SNPs were selected for this study that were either a GWAS SNP or a proxy for a GWAS SNP (Tables I and II in the online-only Data Supplement). HDL-C and triglyceride levels were measured after a >12-hour fast.11 Nontransformed HDL-C and log-transformed triglyceride levels were analyzed (Figure I in the online-only Data Supplement).

Baseline and year 1 measurements for each outcome of interest were modeled jointly, as bivariate normal variables with an unstructured covariance matrix. Three-way interaction models of individual SNP markers with measurement time (year 1 versus baseline) and study arm (ILI versus DSE) were estimated in Splus 8.212 using restricted maximum likelihood. An additive genetic model was used for all SNP markers, with genotype coded by the number of minor alleles (0/1/2 copies). Therefore, 4 distinct types of SNP effects were estimated, which can be interpreted as the effect of 1 additional copy of the corresponding minor allele on (1) baseline lipid levels within DSE (SNP main effect), (2) ILI-DSE differences in baseline lipid levels (SNP×treatment interaction), (3) 1-year change in lipids within DSE (SNP×time×treatment interaction), and (4) ILI-DSE differences in 1-year change in lipids (SNP×time×treatment interaction). Set type 2 parameters serve as a randomization check. No between-arm differences in baseline means were detected for any of the markers under consideration.

All our results are based on full 3-way hierarchical interaction models, with no additional model simplification. To aid with the interpretation of SNP×time×treatment interactions, we report marker effects on 1-year change separately for ILI and DSE. Main marker effects at baseline include all participants from both study arms.

Longitudinal outcomes were additionally adjusted for age, sex, hormonal replacement therapy at baseline, concurrent drug use (lipid medication, thiazolidindione medication, with pioglitazone and rosiglitazone effects modeled separately), study site, and the first 2 ancestry informative marker principal components (Methods in the online-only Data Supplement) as previously described.13,14 Other than study site, all of these covariates were fully interacted with time, treatment, and time by treatment interaction so as to allow for these covariate effects to vary across study arm and time point, in a manner similar to the SNP effects described earlier.

Using the method of Li and Ji15 for our experiment of 82 SNPs, we computed our multiplicity-adjusted threshold for significance of SNP main effects on baseline outcomes as P<0.0009, taking into account the effective number of uncorrelated markers being tested. However, because we expect SNP×treatment interactions to have smaller effect size than SNP main effects, we chose to report all such interaction findings that reached a P<0.05 threshold for nominal significance.16 The locus-wide threshold significance used in regional plots was P<2.9×10−6 using the approach described by Li and Ji.15

Results

Look AHEAD Genetic Study

Baseline characteristics of Look AHEAD study participants not taking fibrates or niacin for whom genotype data from the IBC array were available are shown in Table 1. Significant differences in year 1 lipid and thiazolidindione medication use were observed between participants in the DSE and ILI groups (Table 1),11 with statin use increasing in ILI to a lesser extent than in DSE. Look AHEAD Genetics Study participants in the ILI group showed highly significant year 1 differences in HDL-C and triglycerides as compared with participants in the DSE group (both P<0.0001; Table 1). However, they did not differ in terms of either low-density lipoprotein (P=0.48) or total cholesterol (P=0.26).

Table 1. Population Characteristics in Look AHEAD Genetic Subcohort

CharacteristicsTotal (n=3561)DSE (n=1797)ILI (n=1764)
Women, %57.957.558.3
Age, y (SD)59.0 (6.9)59.0 (6.8)59.1 (6.9)
Ethnicity, %
 Non-Hispanic Black605 (17.0)295 (16.4)310 (17.6)
 American Indian/Alaskan18 (0.5)8 (0.4)10 (0.6)
Native*
 Asian/Pacific Islander41 (1.1)19 (1.1)22 (1.2)
 Hispanic/Latino267 (7.5)136 (7.6)131 (7.4)
 Non-Hispanic White2555 (71.7)1298 (72.2)1257 (71.2)
 Other (multiple)75 (2.1)41 (2.3)34 (1.9)
Year 0 medication use, %
 Statin46.546.746.2
 Rosiglitazone15.015.714.2
 Pioglitazone12.512.712.2
 HRT (% women)58.258.657.9
Year 1 medication use, %
 Statin52.155.149.1
 Rosiglitazone14.816.912.6
 Pioglitazone12.314.410.1
BMI, kg/m2; mean (SD)
 Year 036.2 (6.06)36.16 (5.9)36.3 (6.22)
 Year 134.4 (6.24)35.77 (5.94)33.02 (6.24)
 Year 1−year 0 change−1.89 (2.72)−0.33 (2.83)−3.25 (2.61)
HDL-C, mg/dL; mean (SD)
 Year 043.73 (11.9)43.7 (11.8)43.8 (12.0)
 Year 146.1 (12.5)45.18 (11.96)47.1 (13.0)
 Year 1−year 0 change2.28 (7.03)1.3 (7.2)3.24 (6.83)
Triglyceride, mg/dL; median (IQR)
 Year 0151 (106 to 217)150 (107 to 214)152 (106 to 218)
 Year 1139 (94 to 191)142 (100 to 300)128 (90 to 180)
 Year 1−year 0 change−12 (−49.25 to 20)−5 (−40 to 26)−19 (−62 to 12)
Log triglyceride, mean (SD)
 Year 05.02 (0.54)5.02 (0.53)5.02 (0.54)
 Year 14.91 (0.53)4.96 (0.52)4.85 (0.53)
 Year 1−year 0 Change−0.12 (0.41)−0.05 (0.42)−0.18 (0.41)

BMI indicates body mass index; DSE, diabetes mellitus support and education; HDL-C, high-density lipoprotein cholesterol; ILI, intensive lifestyle intervention; IQR, interquartile range; and Look AHEAD, Action for Health in Diabetes.

*The number of American Indian participants included in this study is proportionally less than the parent Look AHEAD trial because of limitations in genetic consent.

SNPs Associated With Baseline Lipid Traits and Treatment Response

SNPs that show an interaction with response to behavioral treatment for HDL-C and log-transformed triglyceride levels below a level of nominal significance (SNP×treatment P<0.05) are shown in Tables 2 and 3, respectively, while SNPs that show an association with baseline levels averaged across the 2 study arms below a significance threshold corrected for multiple hypothesis testing specific to this analysis (P<0.0009) are shown in Tables IV and V in the online-only Data Supplement. Out of 82 SNPs selected for analysis based on previous HDL-C or triglyceride GWASs, we identified 20 SNPs associated with baseline HDL-C levels and 12 SNPs that demonstrated evidence of behavioral treatment effect modification; cholesterol ester transfer protein (CETP) rs3764261 showed both a significant baseline HDL-C association and a nominal treatment interaction. For triglyceride levels, we identified 27 SNPs associated with baseline levels and 6 SNPs that demonstrated evidence of behavioral treatment effect modification. SNPs in lipoprotein lipase (LPL) and phospholipid transfer protein (PLTP) were associated with both baseline HDL-C and triglycerides. Polymorphisms in CETP and lecithin-cholesterol acyltransferase (LCAT) were found to be only associated with baseline HDL-C, whereas SNPs in angiopoietin-like protein 3 (ANGPTL3), glucokinase regulatory protein (GCKR), propionyl-coenzyme A carboxylase β chain (PCCB), tribbles-like protein 1 (TRIB1), FADS-2, and cartilage intermediate layer protein 2 (CILP2) were only associated with baseline triglyceride levels. The direction of effect of significant minor allele association with baseline HDL-C and triglyceride levels in Look AHEAD agreed with the published GWAS association. The replication of many GWAS HDL-C and triglyceride SNP associations with baseline Look AHEAD measures indicates that these SNP associations are unlikely to be weakened significantly by T2D and obesity.

Table 2. Genetic Predictors of Year 1 Change in HDL-C

Baseline HDL1-Year HDL Change
ILI and DSEILIDSE
GeneSNPGroupnMajor AlleleMinor Allele/GWAS EffectMAFβSEP ValueβSEP ValueβSEP ValueSNP×Tx Interaction (P Value)
APOBrs693*All3559GA↓42.500.220.260.4089−0.510.270.05970.270.260.29950.0385
NHW2544GA↓47.860.170.280.5474−0.560.300.06060.290.300.33990.0458
GCKRrs1260326All3559GA35.210.180.270.5008−0.500.280.07290.450.270.09370.0141
NHW2544GA40.410.610.280.0289−0.560.310.06830.350.290.22600.0311
GCKRrs780094All3560GA35.390.110.270.6769−0.460.270.09520.390.270.14610.0272
NHW2545GA39.800.510.280.0684−0.520.310.09160.380.290.19480.0345
GCKRrs780093All3526GA34.910.190.270.4715−0.480.280.07990.420.270.11810.0190
NHW2524GA39.850.530.280.0605−0.540.300.07670.380.290.19590.0299
FADS1rs174548All3557GC↓29.10−0.460.290.10860.720.290.0132−0.170.280.55660.0295
NHW2542GC↓27.95−0.660.310.03400.890.330.0073−0.110.330.74510.0326
FADS-2rs1535All3559AG31.33−0.670.280.01740.820.280.0037−0.020.280.94790.0352
NHW2544AG32.19−0.640.300.03160.790.320.0120−0.020.310.94620.0664
ZNF259rs12286037All3561GA9.03−0.260.450.5631−0.770.440.08190.500.460.28080.0472
NHW2546GA↓6.34−0.390.570.5000−0.270.630.66970.240.590.68640.5570
LIPCrs1800588*All3558GA↑27.970.860.300.00410.870.300.00370.040.300.89970.0488
NHW2543GA↑20.741.030.340.00261.040.360.0044−0.220.360.54130.0141
LIPCrs8034802*All3555TA↑31.050.630.280.02250.700.280.0131−0.090.280.75110.0463
NHW2542TA↑27.450.650.310.03450.880.330.0081−0.020.320.95710.0522
LIPCrs588136*All3558AG↑26.410.800.300.00780.500.300.1033−0.150.300.62490.1324
NHW2543AG↑20.781.090.340.00130.940.370.0106−0.230.350.52100.0223
CETPrs3764261All3559CA↑31.482.790.272.5×10−240.810.280.00380.030.270.90340.0471
NHW2546CA↑31.632.700.301.5×10−190.680.330.03660.000.310.99520.1326
LILRA3rs103294All3561GA↑18.350.350.330.28670.230.330.4887−0.700.330.03400.0472
NHW2546GA↑18.870.570.360.10730.170.380.6522−0.380.380.30520.2976

Age, sex, ancestry principal components, and study site statistically adjusted. All analyses quantify the effect of additional copies of the minor allele in an additive model. Baseline P<0.0009 and interaction and 1-year P<0.05 shown in bold. Arrows indicate the direction of the HDL-C effect published in genome-wide association studies (GWASs; Tables I and III in the online-only Data Supplement). ALL indicates all racial and ethnic groups; DSE, diabetes mellitus support and education; ILI, intensive lifestyle intervention; MAF, minor allele frequency; and SNP, single-nucleotide polymorphisms.

*Both high-density lipoprotein cholesterol (HDL-C) and triglyceride SNP×Tx interaction association.

Table 3. Genetic Predictors of Year 1 Change in Triglycerides (Logarithmic Scale)

Baseline log(TRIG)1-Year log(TRIG) Change
ILI and DSEILIDSE
GeneSNPGroupnMajor AlleleMinor Allele/GWAS EffectMAFβSEP ValueβSEP ValueβSEP ValueSNP×Tx Interaction (P Value)
APOBrs693*ALL3559GA42.2−0.010.010.54150.040.020.01110.000.020.91500.0833
NHW2544GA47.86−0.010.010.40780.050.020.0065−0.020.020.30740.0083
AFF1rs3775214ALL3559AG↓34.390.000.010.91020.010.020.3673−0.040.020.01410.0177
NHW2545AG↓38.53−0.010.010.55020.020.020.2725−0.040.020.02650.0192
LIPCrs1800588*ALL3558GA↑27.970.010.010.6773−0.020.020.23360.020.020.36420.1377
NHW2543GA↑20.740.010.020.5658−0.030.020.11410.030.020.13060.0288
LIPCrs8034802*ALL3555TA↑31.050.010.010.6744−0.030.020.08170.020.020.27450.0447
NHW2545TA↑27.450.010.020.5854−0.040.020.04060.030.020.06800.0061
LIPCrs588136*ALL3558AG↑26.410.000.010.8066−0.020.020.26870.040.020.01060.0099
NHW2543AG↑20.780.010.020.7661−0.040.020.08480.050.020.02380.0050
PGS1rs4082919ALL3561CA44.72−0.030.010.0219−0.030.020.06190.040.020.00480.0009
NHW2546CA47.68−0.020.010.1520−0.030.020.06070.030.020.13790.0176

Age, sex, ancestry principal components, and study site statistically adjusted. All analyses quantify the effect of additional copies of the minor allele in an additive model. Interaction and 1-year P<0.05 shown in bold. Arrows indicate the direction of the triglyceride effect published in genome-wide association studies (GWASs; Table I and III in the online-only Data Supplement). ALL indicates all racial and ethnic groups; DSE, diabetes mellitus support and education; ILI, intensive lifestyle intervention; SNP, single-nucleotide polymorphisms; and TRIG, triglyceride.

*Both high-density lipoprotein cholesterol (HDL-C) and Triglyceride SNP×Tx Interaction association.

In agreement with previous studies,17,18 numerous LPL SNPs were associated with both baseline HDL-C and triglyceride levels in Look AHEAD. LPL rs17410962 was most strongly associated with baseline HDL-C (β±SE=2.41±0.36 mg/dL; P=3.6×10−11) and log(triglycerides) (β±SE=−0.12±0.02; P=1.4×10−10). In the original scale of the data, this implies an 11% reduction in baseline triglyceride levels per copy of the minor allele (β=0.89; 95% confidence interval, 0.85–0.92). PLTP SNP rs6065904 was also found to be significantly associated with baseline HDL-C and triglyceride levels. However, LPL and PLTP minor allele carriers demonstrated year 1 changes in lipid traits that were in the same direction in the DSE and ILI treatment groups with a nonsignificant SNP×treatment interaction P value. The lack of a treatment response association for LPL and PLTP SNPs indicated that genetic variants significantly associated with baseline HDL-C and triglyceride levels do not necessarily predict behavioral treatment response.

Genetic Predictors of Differential Lipid Level Response to Behavioral Therapy

Three hepatic lipase (LIPC) SNPs demonstrated evidence of behavioral treatment effect modification at year 1 for both HDL-C and triglyceride changes (Tables 2 and 3), in either the full Look AHEAD cohort or just in non-Hispanic white participants. LIPC rs8034802 minor allele carriers showed a greater increase with ILI compared with DSE for HDL-C (ILI per allele change±SE=0.70±0.28 [P=0.013] versus DSE per allele change±SE=−0.09±0.28 [P=0.75]; nominal SNP×treatment interaction P=0.046) and a greater decrease in log(triglycerides) (ILI per allele change±SE=−0.03±0.02 [P=0.082] versus DSE per allele change±SE=0.02±0.02 [P=0.27]; SNP×treatment interaction P=0.045). The direction of treatment effect was opposite for HDL-C and triglycerides. LIPC is known to biologically modify triglycerides and HDL-C, and here we demonstrate that LIPC variants are associated with the triglyceride and HDL-C response to a lifestyle intervention designed to reduce obesity and increase physical fitness.

Genes Selectively Associated With HDL-C Response

Of SNPs associated with baseline HDL-C or triglycerides, only 1 SNP strongly associated with baseline trait levels was also associated with a significant behavioral treatment response. CETP rs3764261 was strongly associated with baseline HDL-C (P=2.5×10−24) and nominally associated with HDL-C change in response to ILI (P=0.0038) and showed a nominal treatment interaction at year 1 (SNP×treatment interaction P=0.047) in the full Look AHEAD cohort. The HDL-C increase in minor allele carriers of rs3764261 in the ILI group was lower in NHW participants compared with all Look AHEAD participants, and the SNP×treatment interaction no longer reached significance (P=0.13).

To illustrate the genotypic effect of CETP rs3764261 on HDL-C change in the entire Look AHEAD cohort, we calculated the expected HDL-C treatment response of 60-year-old men and women in the absence of lipid medication, pioglitazone, or rosiglitazone, based on our longitudinal statistical model. In response to ILI in both men and women, CETP rs3764261 minor allele carriers showed a greater increase in HDL-C than noncarriers (0.81 mg/dL per minor allele copy; 95% confidence interval, 0.26–1.36), with no significant difference by minor allele status observed in the DSE group (Figure 1; Table 2). Stratification of the treatment effects by sex and genotypic group revealed a highly significant HDL-C response to ILI compared with DSE within each stratum (all P<0.002), with the exception of female homozygous carriers of the major allele (CC; Figure 1).

Figure 1.

Figure 1. High-density lipoprotein cholesterol (HDL-C) behavioral treatment response by cholesterol ester transfer protein (CETP) rs3764261 genotype. Graph shows model predictions from the Action for Health in Diabetes (Look AHEAD) cohort for the mean and 95% confidence interval HDL-C difference (year 1−year 0) by CETP rs3764261 genotype as predicted for 60-year-old men and women not receiving a lipid-lowering medication, pioglitazone or rosiglitazone randomized to intensive lifestyle intervention (ILI; black dots) and diabetes mellitus support and education (DSE; shaded dots). The HDL-C difference in women was adjusted for the presence of hormone replacement therapy. The presence of ≥1 minor alleles in men and women was significantly associated with increased HDL-C change in the setting of ILI (P=0.0038 for both) but not for DSE (P=not significant for both). The HDL-C difference was significantly greater in the participants given ILI compared with DSE (**P<0.01; ***P<0.001), except for women CC carriers.

By comparison, the minor allele at zinc finger 259 (ZNF259) rs12286037, which GWAS predicted to be associated with a reduced HDL-C, showed a nominally significant treatment interaction (SNP×treatment interaction P=0.047) with a trend toward a lower HDL-C in response to behavioral intervention (P=0.082). APOB rs693 showed a nominally significant treatment response (ILI per allele change±SE=−0.51±0.27 versus DSE per allele change±SE=+0.27±0.26; SNP×treatment interaction P=0.0385) with the overall ILI treatment response in the same direction as predicted by GWAS. Finally, LIPC rs8034802 minor allele carriers showed a higher baseline HDL-C and a greater increase in HDL-C in response to ILI and not DSE. These findings indicate that ILI may strengthen the genetic association by promoting HDL-C change in the same direction as at baseline.

In response to behavioral treatment, FADS-2 rs1535 minor allele carriers demonstrated a significantly positive HDL-C response and nominal treatment interaction (ILI per allele change±SE=+0.82±0.28, P=0.0037 versus DSE per allele change±SE=−0.02±0.28, P=0.95; SNP×treatment interaction P=0.035). SNPs in GCKR selected for analysis based on their association with triglycerides were found to have a nominally significant SNP×treatment interaction for HDL-C including GCKR-P446L (rs1260326), which in ILI showed a per allele change±SE=−0.50±0.28 versus DSE per allele change±SE=+0.45±0.27 and SNP×treatment interaction P=0.014.

Genes Selectively Associated With Triglyceride Response Alone

No SNP associated with baseline log(triglycerides) also showed an SNP×Tx interaction for triglyceride response to behavioral intervention. A single SNP in AF4/FMR2 family member 1 (AFF1), APOB, phosphatidylglycerophosphate synthase-1 (PGS1), and 3 LIPC SNPs showed a nominally significant triglyceride behavioral treatment interaction (Table 3). The strongest association with log(triglycerides) change was found with PGS1 rs4082919 (ILI per allele change±SE=−0.03±0.02 versus DSE per allele change±SE=+0.04±0.02; SNP×treatment interaction P<0.0009). Converting to the original measurement scale, this corresponds to a 3% reduction in triglyceride change within ILI per copy of the minor allele (β=0.97; 95% confidence interval, 0.93–1.01) versus a 4% increase within DSE (β=1.04; 95% confidence interval, 1.00–1.08). None of the SNPs showed a significant change in HDL-C in response to ILI.

Novel SNPs Associated With Differential Lipid Trait Response to Behavioral Therapy

Next, we asked whether alternate SNPs within CETP, LPL, LIPC, BUD13-APOA1 Region, FADS1/2/3, GCKR, and LCAT-DPEP2 that regulate HDL-C and triglyceride were more strongly associated with differential lipid trait response compared with the GWAS SNPs. Regional plots showing associations for baseline HDL-C, year 1 change with DSE and ILI, and differential change are shown for CETP in Figure 2A–2D. Interestingly, we observed SNPs nominally associated with HDL-C change within ILI and differential ILI-DSE response (P<0.05) in the 5′-region of CETP, which is the same region in which SNPs were highly associated with baseline HDL-C levels. We identified SNPs in LIPC, BUD13-APO complex, and FADS1/2/3, but not LPL or LCAT-DPEP2, which were significantly associated with differential change for each locus (Figure IIA–IIF in the online-only Data Supplement). However, none of these locus-wide SNPs showed an association markedly stronger than that of the GWAS SNPs.

Figure 2.

Figure 2. CETP locus single-nucleotide polymorphisms (SNP) association with high-density lipoprotein cholesterol (HDL-C) response to diabetes mellitus support and education (DSE) and intensive lifestyle intervention (ILI). CETP regional plot including all SNPs available on the IBC assay in Action for Health in Diabetes (Look AHEAD) with their genomic location and −log(P value) for trait association. SNPs identified by genome-wide association studies to be associated with baseline HDL-C (circles) are indicated by name with their P value for each comparison; all other SNPs are represented as diamonds. The colors represent the strength of the LD between the selected top SNP and each marker. The top SNP always have the darkest red color. r2 values calculated with respect to the SNP marker showing the most significant associations with baseline values of the trait of interest.

Discussion

To our knowledge, this is the largest study to analyze the interaction of genetic factors with a randomly assigned behavioral intervention on lipid trait change in the setting of established T2D. The role of genetic factors in polygenic dyslipidemia has been well defined, and examples of genetic modification of lipid behavioral treatment response have begun to emerge.19,20 Here, we present findings from Look AHEAD, taking advantage of the unique strength of the randomized trial study design in which the 2 standardized behavioral interventions (ILI and DSE) were randomly assigned, with ILI producing greater improvements in HDL-C and triglyceride levels relative to DSE at 1 year of follow-up.11,21 Although we replicate the association of many SNPs with baseline HDL-C and triglyceride levels, including several SNPs achieving a genomic level of significance, interestingly, only 1 SNP, CETP rs3764261, was strongly associated with baseline HDL-C and predicted behavioral treatment response. CETP rs3764261 showed a minor allele dose HDL-C difference at year 1 with ILI but not with DSE. Strikingly, women in the full Look AHEAD cohort who carried both major alleles (CC) did not have a significant HDL-C response to ILI, suggesting HDL-C resistance to behavioral treatment (Figure 1). Similarly, minor alleles within GCKR, APOB, and ZNF259 predicted resistance to HDL-C improvement.

The Women’s Genome Health Study,16 a prospective cohort study of healthy women, previously demonstrated significant effect modification for CETP SNP (rs1532624) with physical activity. Here, we demonstrate that rs3764261, which is modestly correlated with rs1532624 (r2=0.59), minor allele carriers had a greater increase in HDL-C in response to ILI with a significant SNP×treatment interaction (interaction P=0.047). By comparison, Sarzynski et al22 found that rs3764261 did not modify the HDL-C response after bariatric surgery, which does not have a fitness intervention. We were unable to replicate findings from the Women’s Genome Health Study for LPL (rs10096633, which we studied with rs17410962; r2=0.96) and LIPG (rs4939883, which we studied with rs2156552; r2=0.95). Differences between the Women’s Genome Health Study and Look AHEAD participants may explain our results. Look AHEAD includes both men and women, all participants have T2D, and a median body mass index of ≈36 kg/m2, whereas Women’s Genome Health Study participants have a low incidence of T2D and were not overweight (median body mass index, 24.1–25.7 kg/m2). Look AHEAD participants also received a randomized, controlled behavioral intervention in the ILI arm, whereas Women’s Genome Health Study observed genotype associations in the setting of usual self-reported physical activity. Collectively, our findings suggest genetic variation in CETP can modify HDL-C response to lifestyle intervention.

Three LIPC variants were associated with both HDL-C and to a lesser degree with triglyceride response treatment interaction in the entire Look AHEAD cohort or the NHW subset including the LIPC-514(C/T) polymorphism (rs1800588), which has been associated with LPL expression,23 activity,24 and particularly in the setting of a low-fat diet.25,26 We found that LIPC-514(C/T) polymorphism (rs1800588) minor allele carriers showed a significantly greater HDL-C increase in response to ILI and not in response to DSE.

We identified significant HDL-C treatment interactions with SNPs previously associated with diet and metabolic factors. The Diabetes Prevention Program demonstrated an interaction of GCKR-P446L (rs1260326) with a lifestyle intervention on triglyceride levels.19 In Look AHEAD, GCKR-P446L modified the behavioral treatment response of HDL-C but not triglyceride, although it selectively was associated with baseline triglyceride levels. GCKR-P446L has been shown to be associated with reduced GCKR expression, lower glucose–stimulated GCK inhibitory activity,27 and to interact with plasma N-3 polyunsaturated fat levels modulating fasting insulin levels and inflammatory markers.28GCKR rs780094 also interacts with dietary whole grain intake on fasting insulin levels29 and modifies the HDL-C response to behavioral treatment. The FADS1/2/3 locus SNPs described here, which have an association with differential HDL-C response to behavioral intervention, have previously been associated with low-density lipoprotein response to dietary polyunsaturated fatty acid.30APOB rs693, which has important effects on low-density lipoprotein cholesterol,31 was found here to be associated with the HDL-C behavioral treatment response. It is worth emphasizing that weight loss in Look AHEAD participants, particularly in the ILI group, correlated with meal replacement consumption32 and cessation of binge eating practices.33 We speculate that improved diet in ILI participants contributes to the differential genetic associations of GCKR, FADS1/2/3, and APOB SNPs with year 1 HDL-C changes.

LIPC, AFF1, and PGS1 SNPs that showed nominally significant treatment interactions for triglycerides all showed significant associations with triglyceride change in the DSE group and not in the ILI group. These examples suggest that genetic variants may have an effect on triglyceride change observable only in more sedentary people with a stable weight. We interpret this finding to indicate that the influence of genetic factors on triglyceride response may not be capable of modifying the response to an intensive behavioral intervention designed to achieve weight loss and improved fitness. PGS1 rs4082919 showed the strongest SNP×treatment interaction for triglyceride response, which is surprising because that SNP was included in our analysis based on its association with HDL-C and not triglycerides.7 By comparison, GCKR SNPs included in the analysis because of their association with triglyceride levels7 showed an SNP×treatment interaction for HDL-C and not triglycerides. These examples indicate that SNPs associated with baseline traits may modify the treatment response of a related biomarker.

We did not identify any novel SNPs included in the genes under study that affect HDL-C or triglyceride response to behavioral intervention. We cannot exclude the possibility of there being genomic loci that modify HDL-C and triglyceride behavioral treatment response somewhere in the genome. Future genome-wide association analysis will allow the determination of whether common gene variants modify behavioral treatment response.

We recognize several limitations of our study. First, the size of Look AHEAD is smaller than many GWASs, and, therefore, our ability to replicate SNP associations with baseline lipid levels or to distinguish their effect on behavioral treatment response was limited. Second, we acknowledge that our findings of genotype-treatment response interaction cannot be tested in a replication cohort because there is currently no cohort available with similar enrollment criteria subject to a randomized behavioral treatment. Furthermore, we report nominally significant interaction P<0.05 that does not surpass our experiment-wide significance threshold used for testing main SNP effects on baseline outcomes. SNP×treatment interactions are likely to have smaller effect sizes than SNP main effects, suggesting that P<0.05, but above the threshold corrected for multiple testing, may still be informative. However, we acknowledge that because of a lack of replication and the nominal significance of our interaction P values, our findings must be considered as hypothesis generating. Third, although SNPs studied here were selected on the basis of their previous association with HDL-C and triglycerides by well-powered GWASs, we cannot exclude the possibility that there are other important gene variants that influence HDL-C and triglyceride response to behavioral intervention that were not studied. Furthermore, we cannot exclude the possibility that SNP associations demonstrated here may have been influenced by unappreciated genetic structure within the population. We sought to control for population stratification through the inclusion of the first 2 principal components derived from ancestry informative markers as covariates in our models; an approach that for cardiovascular disease risk factors yields information close to self-reported race,34 while offering the ability to classify race and ethnicity for every available Look AHEAD participant. In addition, we find that association results for Look AHEAD participants who self-reported their race and ethnicity as NHW were similar to the results from the entire cohort. We acknowledge that our results in Look AHEAD may be most reflective of NHWs and may not be generalizable to all racial and ethnic groups. Finally, we acknowledge that although our findings from Look AHEAD may apply to a growing population of individuals with T2D, our findings may not be generalizable to a nondiabetic population.

Overall, our findings demonstrate evidence that select variants contained within genes that have undisputed roles in lipid traits seem to modify HDL-C and triglyceride response to lifestyle intervention. However, the effect of genetic factors on lipid responses to behavioral intervention may not be consistently predicted by their baseline associations and are susceptible to modification. These results provide new insight into the biology of HDL-C and triglycerides and demonstrate the genetic effects on the response to an environmental change. It should be noted that all of the individual genetic effects were smaller than the aggregate HDL-C and triglyceride responses to ILI. Therefore, we conclude that although there is evidence of significant effect modification, individual genetic factors do not prevent favorable HDL-C and triglyceride response to behavioral treatment. Collectively, we conclude that an intensive behavioral intervention for T2D and obesity is effective despite the presence of genetic factors that resist HDL-C and triglyceride responses.

Footnotes

The online-only Data Supplement is available at http://circgenetics.ahajournals.org/lookup/suppl/doi:10.1161/CIRCGENETICS.113.000042/-/DC1.

Correspondence to Gordon S. Huggins, MD, MCRI Center for Translational Genomics, Tufts Medical Center and Tufts University School of Medicine, 800 Washington St, Boston, MA 02111. E-mail

References

  • 1. Anderson JW, Kendall CW, Jenkins DJ. Importance of weight management in type 2 diabetes: review with meta-analysis of clinical studies.J Am Coll Nutr. 2003; 22:331–339.CrossrefMedlineGoogle Scholar
  • 2. Miller M, Stone NJ, Ballantyne C, Bittner V, Criqui MH, Ginsberg HN, et al.; American Heart Association Clinical Lipidology, Thrombosis, and Prevention Committee of the Council on Nutrition, Physical Activity, and Metabolism; Council on Arteriosclerosis, Thrombosis and Vascular Biology; Council on Cardiovascular Nursing; Council on the Kidney in Cardiovascular Disease. Triglycerides and cardiovascular disease: a scientific statement from the American Heart Association.Circulation. 2011; 123:2292–2333.LinkGoogle Scholar
  • 3. Pérusse L, Rice T, Després JP, Bergeron J, Province MA, Gagnon J, et al.. Familial resemblance of plasma lipids, lipoproteins and postheparin lipoprotein and hepatic lipases in the HERITAGE Family Study.Arterioscler Thromb Vasc Biol. 1997; 17:3263–3269.CrossrefMedlineGoogle Scholar
  • 4. Pilia G, Chen WM, Scuteri A, Orrú M, Albai G, Dei M, et al.. Heritability of cardiovascular and personality traits in 6,148 Sardinians.PLoS Genet. 2006; 2:e132.CrossrefMedlineGoogle Scholar
  • 5. Bouchard C, Tremblay A, Despres JP, Poehlman ET, Theriault G, Nadeau A, et al.. Sensitivity to overfeeding: the Quebec experiment with identical twins.Prog Food Nutr Sci. 1988; 12:45–72.MedlineGoogle Scholar
  • 6. Després JP, Moorjani S, Tremblay A, Poehlman ET, Lupien PJ, Nadeau A, et al.. Heredity and changes in plasma lipids and lipoproteins after short-term exercise training in men.Arteriosclerosis. 1988; 8:402–409.CrossrefMedlineGoogle Scholar
  • 7. Teslovich TM, Musunuru K, Smith AV, Edmondson AC, Stylianou IM, Koseki M, et al.. Biological, clinical and population relevance of 95 loci for blood lipids.Nature. 2010; 466:707–713.CrossrefMedlineGoogle Scholar
  • 8. Bray G, Gregg E, Haffner S, Pi-Sunyer XF, WagenKnecht LE, Walkup M, et al.. Baseline characteristics of the randomised cohort from the Look AHEAD (Action for Health in Diabetes) study.Diab Vasc Dis Res. 2006; 3:202–215.CrossrefMedlineGoogle Scholar
  • 9. Ribisl PM, Lang W, Jaramillo SA, Jakicic JM, Stewart KJ, Bahnson J, et al.; Look AHEAD Research Group. Exercise capacity and cardiovascular/metabolic characteristics of overweight and obese individuals with type 2 diabetes: the Look AHEAD clinical trial.Diabetes Care. 2007; 30:2679–2684.CrossrefMedlineGoogle Scholar
  • 10. Vitolins MZ, Anderson AM, Delahanty L, Raynor H, Miller GD, Mobley C, et al.; Look AHEAD Research Group. Action for Health in Diabetes (Look AHEAD) trial: baseline evaluation of selected nutrients and food group intake.J Am Diet Assoc. 2009; 109:1367–1375.CrossrefMedlineGoogle Scholar
  • 11. Pi-Sunyer X, Blackburn G, Brancati FL, Bray GA, Bright R, Clark JM, et al.. Reduction in weight and cardiovascular disease risk factors in individuals with type 2 diabetes: one-year results of the look AHEAD trial.Diabetes Care. 2007; 30:1374–1383.CrossrefMedlineGoogle Scholar
  • 12. TIBCO Software I.TIBCO Spotfire SPLUS 8.2 for Solaris/Linux User’s Guide. Seattle, WA: TIBCO Software, Inc; 2010.Google Scholar
  • 13. McCaffery JM, Papandonatos GD, Peter I, Huggins GS, Raynor HA, Delahanty LM, et al.; Genetic Subgroup of Look AHEAD; Look AHEAD Research Group. Obesity susceptibility loci and dietary intake in the Look AHEAD trial.Am J Clin Nutr. 2012; 95:1477–1486.CrossrefMedlineGoogle Scholar
  • 14. McCaffery JM, Papandonatos GD, Huggins GS, Peter I, Kahn S, Knowler WC, et al.. FTO predicts weight regain in the Look AHEAD clinical trial.Int J Obes. April 3, 2013. 10.1038/ijo.2013.54. http://www.nature.com/ijo/journal/vaop/ncurrent/full/ijo201354a.html. Accessed April 3, 2013.CrossrefGoogle Scholar
  • 15. Li J, Ji L. Adjusting multiple testing in multilocus analyses using the eigenvalues of a correlation matrix.Heredity (Edinb). 2005; 95:221–227.CrossrefMedlineGoogle Scholar
  • 16. Ahmad T, Chasman DI, Buring JE, Lee IM, Ridker PM, Everett BM. Physical activity modifies the effect of LPL, LIPC, and CETP polymorphisms on HDL-C levels and the risk of myocardial infarction in women of European ancestry.Circ Cardiovasc Genet. 2011; 4:74–80.LinkGoogle Scholar
  • 17. Kathiresan S, Melander O, Guiducci C, Surti A, Burtt NP, Rieder MJ, et al.. Six new loci associated with blood low-density lipoprotein cholesterol, high-density lipoprotein cholesterol or triglycerides in humans.Nat Genet. 2008; 40:189–197.CrossrefMedlineGoogle Scholar
  • 18. Willer CJ, Sanna S, Jackson AU, Scuteri A, Bonnycastle LL, Clarke R, et al.. Newly identified loci that influence lipid concentrations and risk of coronary artery disease.Nat Genet. 2008; 40:161–169.CrossrefMedlineGoogle Scholar
  • 19. Pollin TI, Jablonski KA, McAteer JB, Saxena R, Kathiresan S, Kahn SE, et al.; Diabetes Prevention Program Research Group. Triglyceride response to an intensive lifestyle intervention is enhanced in carriers of the GCKR Pro446Leu polymorphism.J Clin Endocrinol Metab. 2011; 96:E1142–E1147.CrossrefMedlineGoogle Scholar
  • 20. Pollin TI, Isakova T, Jablonski KA, de Bakker PI, Taylor A, McAteer J, et al.; Diabetes Prevention Program Research Group. Genetic modulation of lipid profiles following lifestyle modification or metformin treatment: the Diabetes Prevention Program.PLoS Genet. 2012; 8:e1002895.CrossrefMedlineGoogle Scholar
  • 21. Wing RR. Long-term effects of a lifestyle intervention on weight and cardiovascular risk factors in individuals with type 2 diabetes mellitus: four-year results of the Look AHEAD trial.Arch Intern Med. 2010; 170:1566–1575.MedlineGoogle Scholar
  • 22. Sarzynski MA, Jacobson P, Rankinen T, Carlsson B, Sjöström L, Carlsson LM, et al.. Association of GWAS-based candidate genes with HDL-cholesterol levels before and after bariatric surgery in the Swedish obese subjects study.J Clin Endocrinol Metab. 2011; 96:E953–E957.CrossrefMedlineGoogle Scholar
  • 23. Cohen JC, Wang Z, Grundy SM, Stoesz MR, Guerra R. Variation at the hepatic lipase and apolipoprotein AI/CIII/AIV loci is a major cause of genetically determined variation in plasma HDL cholesterol levels.J Clin Invest. 1994; 94:2377–2384.CrossrefMedlineGoogle Scholar
  • 24. Carr MC, Hokanson JE, Deeb SS, Purnell JQ, Mitchell ES, Brunzell JD. A hepatic lipase gene promoter polymorphism attenuates the increase in hepatic lipase activity with increasing intra-abdominal fat in women.Arterioscler Thromb Vasc Biol. 1999; 19:2701–2707.CrossrefMedlineGoogle Scholar
  • 25. Ordovas JM, Corella D, Demissie S, Cupples LA, Couture P, Coltell O, et al.. Dietary fat intake determines the effect of a common polymorphism in the hepatic lipase gene promoter on high-density lipoprotein metabolism: evidence of a strong dose effect in this gene-nutrient interaction in the Framingham Study.Circulation. 2002; 106:2315–2321.LinkGoogle Scholar
  • 26. Riestra P, López-Simón L, Ortega H, Gorgojo L, Martin-Moreno JM, Schoppen S, et al.. Fat intake influences the effect of the hepatic lipase C-514T polymorphism on HDL-cholesterol levels in children.Exp Biol Med (Maywood). 2009; 234:744–749.CrossrefMedlineGoogle Scholar
  • 27. Beer NL, Tribble ND, McCulloch LJ, Roos C, Johnson PR, Orho-Melander M, et al.. The P446L variant in GCKR associated with fasting plasma glucose and triglyceride levels exerts its effect through increased glucokinase activity in liver.Hum Mol Genet. 2009; 18:4081–4088.CrossrefMedlineGoogle Scholar
  • 28. Perez-Martinez P, Delgado-Lista J, Garcia-Rios A, McMonagle J, Gulseth HL, Ordovas JM, et al.. Glucokinase regulatory protein genetic variant interacts with omega-3 PUFA to influence insulin resistance and inflammation in metabolic syndrome.PLoS One. 2011; 6:e20555.CrossrefMedlineGoogle Scholar
  • 29. Nettleton JA, McKeown NM, Kanoni S, Lemaitre RN, Hivert MF, Ngwa J, et al.; MAGIC Investigators. Interactions of dietary whole-grain intake with fasting glucose- and insulin-related genetic loci in individuals of European descent: a meta-analysis of 14 cohort studies.Diabetes Care. 2010; 33:2684–2691.CrossrefMedlineGoogle Scholar
  • 30. Hellstrand S, Sonestedt E, Ericson U, Gullberg B, Wirfält E, Hedblad B, et al.. Intake levels of dietary long-chain PUFAs modify the association between genetic variation in FADS and LDL-C.J Lipid Res. 2012; 53:1183–1189.CrossrefMedlineGoogle Scholar
  • 31. Saxena R, Voight BF, Lyssenko V, Burtt NP, de Bakker PI, Chen H, et al.. Genome-wide association analysis identifies loci for type 2 diabetes and triglyceride levels.Science. 2007; 316:1331–1336.CrossrefMedlineGoogle Scholar
  • 32. Wadden TA, West DS, Neiberg RH, Wing RR, Ryan DH, Johnson KC, et al.; Look AHEAD Research Group. One-year weight losses in the Look AHEAD study: factors associated with success.Obesity (Silver Spring). 2009; 17:713–722.CrossrefMedlineGoogle Scholar
  • 33. Gorin AA, Niemeier HM, Hogan P, Coday M, Davis C, DiLillo VG, et al.; Look AHEAD Research Group. Binge eating and weight loss outcomes in overweight and obese individuals with type 2 diabetes: results from the Look AHEAD trial.Arch Gen Psychiatry. 2008; 65:1447–1455.CrossrefMedlineGoogle Scholar
  • 34. Halder I, Kip KE, Mulukutla SR, Aiyer AN, Marroquin OC, Huggins GS, et al.. Biogeographic ancestry, self-identified race, and admixture-phenotype associations in the Heart SCORE Study.Am J Epidemiol. 2012; 176:146–155.CrossrefMedlineGoogle Scholar

Clinical Perspective

Weight loss and improved fitness are first-line treatment recommendations for obese patients with type-2 diabetes mellitus; however, the effectiveness of behavioral therapy in the context of genetic factors is not well defined. Large-scale genetic studies have identified common genetic polymorphisms associated with high-density lipoprotein cholesterol (HDL-C) and triglyceride levels. Here, we studied the effect of genetic polymorphisms on the HDL-C and triglyceride response in 3561 participants of the Action for Health in Diabetes (Look AHEAD) study that were randomized to either an intensive lifestyle intervention designed to reduce weight and improve fitness or to dietary education alone. In general, individual polymorphisms produced a smaller difference in HDL-C and triglycerides levels at baseline compared with the changes produced by behavioral treatment. We identified polymorphisms in cholesterol ester transfer protein (CETP), hepatic lipase, fatty acid desaturase-2, and ApoB that nominally modified the HDL-C response to behavioral treatment. In addition, a polymorphism in phosphatidylglycerophosphate synthase-1 was found to modify the triglyceride treatment response. The CETP polymorphism rs3764261 showed a graded HDL-C response to behavioral treatment. Although some of our findings are consistent with results from cohort studies, we acknowledge the need for replication of our findings in independent cohorts. Collectively, we conclude that an intensive behavioral intervention for type-2 diabetes mellitus and obesity is effective despite the presence of genetic factors that resist HDL-C and triglyceride responses.