Dietary cholesterol increases paraoxonase 1 enzyme activity.

HDL-associated paraoxonase 1 (PON1) activity has been consistently associated with cardiovascular and other diseases. Vitamins C and E intake have previously been positively associated with PON1 in a subset of the Carotid Lesion Epidemiology and Risk (CLEAR) cohort. The goal of this study was to replicate these findings and determine whether other nutrient intake affected PON1 activity. To predict nutrient and mineral intake values, 1,402 subjects completed a standardized food frequency survey of their dietary habits over the past year. Stepwise regression was used to evaluate dietary and covariate effects on PON1 arylesterase activity. Five dietary components, cholesterol (P < 2.0 × 10−16), alcohol (P = 8.51 × 10−8), vitamin C (P = 7.97 × 10−5), iron (P = 0.0026), and folic acid (0.037) were independently predictive of PON1 activity. Dietary cholesterol was positively associated and predicted 5.5% of PON1 activity, second in variance explained. This study presents a novel finding of dietary cholesterol, iron, and folic acid predicting PON1 activity in humans and confirms prior reported associations, including that with vitamin C. Identifying and understanding environmental factors that affect PON1 activity is necessary to understand its role and that of HDL in human disease.

However, the results of Jarvik et al. have not been restudied in humans, nor have other dietary environmental factors, such as fat and cholesterol intake or other vitamin use, been investigated with regard to effects on PON1 activity. Thus, the goal of the present study was to examine the effects of common nutrients on PON1 activity as measured by AREase enzyme activity, to confi rm previous dietary relationships with PON1 and to determine whether any novel associations exist within this CLEAR study cohort.

Ethics statement
Institutional review boards at the University of Washington, Virginia Mason Medical Center, and Veterans Affairs Puget Sound Health Care approved this study. Written, informed consent was obtained from all participants.

Sample
The study population for this analysis consisted of 1,402 samples from the previously described Carotid Lesion Epidemiology and Risk (CLEAR) study ( 8,9,41 ). The cohort includes 402 subjects with carotid artery disease (CAAD) and 857 controls, and 143 subjects of other phenotypes, including moderate carotid artery obstruction (15-49% by ultrasound), as well as coronary artery and peripheral artery disease. Descriptive statistics are presented in Table 1 . Current smoking status and reported ancestry were obtained by self-reporting. Medication use, including statins and insulin injection, were ascertained from the Veterans Affairs Puget Sound Health Care medical record review. Exclusion criteria included familial hypercholesterolemia, total fasting cholesterol greater than 400 mg/dl, hypocoagulable state and/or the use of anticoagulant medication, postorgan transplant, or the inability to consent.

Survey methods
Subjects were asked to complete the standardized Harvard food frequency questionnaire developed by the Health Professionals Follow-Up Study (41a) . The survey asked about i ) the average frequency of intake over the previous year of specifi ed portions of 131 foods and ii ) the use of vitamins and mineral supplements, including the dose and duration of use. Questions regarding brand of multivitamins and cereal used were asked to clarify the quantities of specifi c vitamin supplementation. Subjects were excluded from analyses if i ) their caloric intake was not between 800 and 4,200 kcal/day or ii ) their surveys had more than 70 blank items of a total of 131 questions. All vitamin usage was energy-adjusted to 2,000 kcal/day. This food frequency survey has been validated against two 1-week diet records taken approximately 6 months apart ( 42 ).

Genotyping and PON1 phenotypes
The four known functional PON1 polymorphisms with largest effects on activity, PON1 Q192R , PON1 M55L , PON1 -108C/T , and PON1 -162A/G , were genotyped using previously described methods ( 22,43 ). PON1 arylesterase activity measured by degradation of phenylacetate (AREase) was utilized as the primary measured outcome of PON gene cluster variation, due to its closer correlation with protein levels. The PON1 AREase were measured by a continuous spectrophotometric assay with lithium heparin plasma as previously described ( 43 ). AREase activity was measured in triplicate and averaged. inhibit both the oxidation of LDL and the interaction between macrophages and endothelium ( 12 ), both likely key factors in the infl ammatory changes underlying atherogenesis. PON1-defi cient mice cannot neutralize the oxidized LDL lipids and have an increased susceptibility to organophosphate toxicity and coronary heart disease (CHD) ( 14,15 ). Finally, PON1 activity appears to plays a role in maintaining the endothelial-atheroprotective effects of HDL ( 16 ).
PON1 has broad substrate specifi city and is protective against exposure to toxic organophosphorus insecticides ( 17 ). For biological purposes, PON1 activity is generally measured with regard to the rate of hydrolysis of paraoxon, diazoxon, and phenylacetate (arylesterase activity) ( 18,19 ). These are termed POase, DZOase, and AREase activities, respectively. AREase activity is unaffected by the functional PON1 Q192R polymorphism, thus making it the best refl ection of the levels of PON1 protein ( 20 ). PON1 enzyme activity is infl uenced by both genetic and environmental factors. There are four well-established functional PON1 mutations ( 21 ): two missense mutations [ PON1 Q192R (rs662) and PON1 M55L (rs854560)] and two 5 ′ regulatory [ PON1 -108C/T (rs705379) and PON1 -162A/G (rs705381)]. PON1 -108C/T has the largest effect on activity, altering expression likely due to modifi cation of an Sp1 binding site ( 22 ). Recent fi ndings within this Carotid Lesion Epidemiology and Risk (CLEAR) cohort attribute approximately 21% of PON1 AREase activity to the four functional PON1 mutations and six additional common variants within the PON gene cluster (including PON2 and PON3 ) in 1,328 European males ( 23 ). Rare deleterious variants have also been identifi ed ( 24 ).
Environmental factors that infl uence PON1 enzyme activity include tobacco use, which has been reported to depress PON1 enzyme activity and concentration ( 25 ). Moderate alcohol consumption ( ‫ف‬ 40 g per day) has been reported to increase PON1 activity (26)(27)(28), whereas heavy alcohol drinking (>80 g per day) has the opposite effect ( 28,29 ). With regard to dietary intake, fatty meals rich in oxidized lipids have been reported to decrease ( 30 ), whereas diets rich in olive oil ( 31 ) and monounsaturated fats ( 32 ) increase, postprandial PON1 activity. PON1 activity is affected by drugs: statin use ( 33 ) and anti-diabetic drugs, such as sulphonylureas ( 34 ) and rosiglitazone ( 35 ), have been reported to increase PON1 enzyme activity. As well, PON1 has been reported to be an antidiabetic enzyme ( 36 ) that increases insulin release from pancreatic ␤ cells ( 37 ).
In 2002, Jarvik et al. fi rst reported the associations of the antioxidant vitamins C (ascorbic acid) and E (e.g., ␣ -tocopherol) with an increase in PON1 POase and DZOase activities in an overlapping, but much smaller, subset of the CLEAR cohort (n = 189 versus 1,402 in the current study) ( 38 ). Experiments in quail demonstrated that vitamin C reversed the decrease in PON1 enzyme activity induced by heat stress, especially when given in conjunction with folic acid ( 39 ). In addition, a study of healthy humans showed that vitamin E supplementation prevented an exercise-induced decrease in PON1 activity ( 40 ). to examine the fi t of each model, beginning with a base model that included age, sex, current smoking status, insulin use, race dummy covariates (with European ancestry as the baseline, as it represented the majority of the cohort), and the genotypes for the four functional PON1 SNPs as covariates ( 8,9,51 ). Dietary covariates that were included in the fi nal model increased the ability of the model to predict PON1 AREase activity. A secondary analysis to test the hypothesis that ln(dietary cholesterol) predicted variation in AREase/HDL ratio was performed using linear regression, with age, sex, current smoking status, insulin use, race dummy covariates, and genotypes for the four functional PON1 SNPs as covariates.
Statin drug use can infl uence PON1 expression, and this appears to be infl uenced by PON1 -108 genotype ( 52 ). However, statin drug use could not be included as a covariate due to confounding with CAAD status; the preferential use of statins in cases in the CLEAR cohort could lead to an erroneous estimation of statin effects on PON1 activity.
A baseline regression model containing the four functional PON1 mutations ( PON1 Q 192 R , PON1 M 55 L , PON1 ؊ 108 C/T , and PON1 ؊ 162 A/G ), age, sex, current smoking status, insulin use, and the four race dummy variables explained 26.5% of PON1 AREase enzyme activity. We then examined a best-fi t model utilizing stepwise linear regression including the aforementioned variables plus 47 dietary covariates. As a result, only dietary intake of nutrients or minerals that improved the predictive power of the best-fi t model using an AIC criterion were retained in the full model.
In addition to the base model of the four functional PON1 mutations, age, sex, and current smoking status, an additional fi ve dietary variables were retained in the bestfi t model ( Table 2 ). Together, the base model and these fi ve dietary variables explained 34.67% of PON1 AREase activity. Addition of these dietary predictors (cholesterol intake, alcohol category, vitamin C, iron, and folic acid) explained 5.45%, 1.59%, 0.33%, 0.58%, and 0.21% of AREase activity, respectively (see Fig. 2 ). Dietary cholesterol, alcohol use, and vitamin C were associated with an increase in AREase activity. Iron and folic acid were associated with a decrease in AREase activity. Sensitivity analyses repeating

Plasma lipid measurements
Lipid measurements were performed on fasting whole plasma. Standard enzymatic methods were utilized to determine the levels of total cholesterol, triglycerides, very low-density lipoprotein (VLDL), HDL, and apolipoprotein A1 (apoA1) (44)(45)(46). LDL was calculated using the Friedwald equation ( 47 ) or directly measured if triglycerides were elevated.

Analysis
Of the estimated nutrient intakes from the Health Professionals Food Survey, 47 variables had complete and nonredundant data across the cohort. Natural log transformation was performed for covariates that displayed a skewed distribution, including dietary cholesterol, vitamins C and E, and all of the plasma lipid measures. Extreme observations were Winsorized at 3 standard deviations from the mean ( 48 ). Due to the distribution of subjects across the alcohol intake variable, a previously reported semi-quantitative measure was adopted ( 49,50 ), with 0 = no alcohol consumption in the past year, 1 = 0-12 g/day, 2 = 12-24 g/day, 3 = 24-60 g/day, and 4 = >60 g/day.
Analyses were done in R (http://www.r-project.org/) using the standard regression tools available. Genotypes were coded using an additive model. Stepwise linear regression was performed with 47 dietary covariates entering the model. Model comparison was done using Akaike's information criterion (AIC) Secondarily, we removed the 189 subjects that were previously considered by Jarvik et al. ( 38 ) to examine the effects of vitamin C in independent data from the prior report (n = 1,213). Vitamin C remained signifi cant ( P = 0.0013) with a coeffi cient = 4.27 ± 1.32 and predicted 0.47% of AREase activity, replicating the original fi nding of dietary vitamin C predicting paraoxonase activity in independent data.
The majority of the predictive power for AREase from the new dietary covariates was attributed to dietary cholesterol intake (5.45% of 8.20%), which was positively associated with AREase activity in the best-fi t model. However, the analyses when stratifying by sex, CAAD status, or statin use did not reveal signifi cant differences in coeffi cient sizes or direction of effects in stratifi ed subgroups, suggesting that these factors do not infl uence the association of the dietary factors and paraoxonase.
Vitamin E did not enter the AREase prediction model. However, as it was highly correlated with both vitamin C ( r = 0.60) and folic acid ( r = 0.53), we sought to determine whether it would be signifi cant if vitamin C were excluded from the best-fi t model. In this stepwise regression, ln (vitamin E) was signifi cant ( P = 0.048) with a coeffi cient = 1.91 ± 0.96 and predicted 0.15% of AREase activity.
In the current study, we utilized a cohort 7.42 times larger than the sample fi rst used by Jarvik et al. ( 38 ) to examine the effects of dietary nutrient and mineral intake on PON1 activity. We believe that the associations of dietary cholesterol, iron in nonanemic patients, and folate with PON1 activity in humans have not been previously reported. We also replicate past fi ndings of vitamin C being associated with PON1 activity in an independent group of subjects. Vitamin E does not remain in the predictive model due to its correlation with vitamin C, but it does predict AREase if vitamin C is not considered in the model. Finally, we confi rm prior reports of alcohol intake associating with AREase activity.
Dietary cholesterol predicted 5.45% of AREase activity (of a total 8.2% related to diet) in a best-fi t model utilizing only dietary covariates. It was positively correlated with AREase activity. This may be contrary to expectations, as PON1 activity is considered atheroprotective, whereas cholesterol intake has been linked with increased risk of CHD ( 53 ). However, this same direction of effect and similar effect size were seen when replacing plasma cholesterol dietary cholesterol is associated with an increase in atherosclerotic plaque ( 53 ), whereas PON1 enzyme activity is inversely correlated with atherosclerosis. To explore this paradox, we created a best-fi t model for AREase in a subset of the cohort (n = 1,388) in which dietary cholesterol was replaced by plasma cholesterol to see whether the effect was the same. Similar to dietary cholesterol intake, plasma cholesterol was the most signifi cant covariate to join the model ( P < 2 × 10 Ϫ 16 ), explained 5.35% of AREase variation, and had a coeffi cient (59.75 ± 6.04) with the same direction and similar effect size as dietary cholesterol.
To examine whether the effects of dietary cholesterol were still signifi cant when plasma lipid measures were added to the model, we created a second best-fi t model for AREase in a subset of the cohort (n = 1,388), including the base model, the fi ve dietary factors, and fi ve natural log transformed lipid and lipoprotein plasma measures: apoA1, HDL, LDL, total cholesterol, and VLDL. In addition to the base model, all fi ve of the dietary covariates and apoA1 were retained in the stepwise regression model (see ) were observed. However, plasma cholesterol was not a signifi cant predictor of AREase when dietary cholesterol is included in the predictive model. Finally, we determined that ln(dietary cholesterol) was a signifi cant predictor of AREase/HDL-C ratio ( P = 0.001), with a ␤ coeffi cient and standard deviation equal to 0.50 and 0.15, respectively,  Table 2 for complete stepwise model information.
AREase/HDL ratio. Increased LDL has been reported to be associated with increased oxidized LDL ( 76 ). Elevated oxidized LDL has been reported to be increased along with an increased AREase/HDL ratio in subjects at risk of end stage renal disease ( 77 ). Further studies, including functional analyses, will be necessary to elucidate the etiology of this paradoxical relationship between dietary cholesterol and plasma PON1 AREase activity.
Jarvik et al. fi rst reported the association of the antioxidants vitamin C and E with PON1 POase and DZOase activity utilizing a smaller, overlapping portion of the cohort (n = 189) that was composed entirely of European males ( 38 ). In this study, we have utilized a larger cohort with mixed ethnicity and sex to replicate the association and direction of effect for vitamin C while also discovering that folic acid intake was signifi cant in the CLEAR cohort, similar to previous reports in quail ( 39 ). While we do not fi nd prior reports of folate decreasing PON1 activity in humans, B12 treatment in subjects with B12 defi ciency increased AREase ( 78 ). When we removed the 189 subjects previously considered by Jarvik et al., vitamin C remained signifi cant and with the same positive direction of effect. Thus, we can confi rm that the effects of vitamin C replicated in a nonoverlapping and larger cohort. In addition, our results suggest that vitamin E failing to improve prediction of AREase is due to the high correlation between it and vitamin C.
With regard to the fi nding of iron being negatively associated with AREase, Aslan et al. previously observed an association between iron-defi ciency anemia (IDA) and decreased levels of PON1 ( 79 ). However, other studies since have not yet replicated this fi nding ( 78,80 ). Our result that dietary iron supplementation is associated with decreased PON1 levels in human subjects with normal iron homeostasis is not consistent with the Aslan result. This difference can potentially be attributed to the following factors. First, the selection criterion for the cohorts are fundamentally different, as CLEAR was primarily for Alternative hypotheses for the mechanism of increased AREase activity and AREase/HDL-C ratio associated with dietary cholesterol include that hepatic PON1 expression is upregulated or that enzymatic activity is directly increased by feedback mechanisms initiated by, or downstream of, increased cholesterol intake. PON1's transcription regulation includes lipid regulatory elements. The PON1 5 ′ region has two sterol responsive element (SRE)-like regions, and sterol regulatory element binding protein 2 (SREBP2) has been shown to bind to the PON1 promoter and increase expression ( 52 ). SREBP2 also upregulates LDLR ; thus, it is possible that the hepatocytes respond to the increased cholesterol load or a correlated elevation, such as that of insulin ( 73 ), by activating SREBP2, which then activates PON1 expression. The putative antiatherogenic dietary fl avonoid quercetin has been shown to increase both PON1 ( 74 ) and LDLR ( 75 ) expression via the SRE pathway. Another alternative hypothesis for the increased AREase activity associated with dietary cholesterol is the role of oxidized LDL in increasing the , a coeffi cient of 42.14 ± 6.12, a t-statistic of 6.89, and explained 2.10% of AREase variance.
dietary cholesterol being associated with PON1 AREase activity. We also replicate past fi ndings of vitamin C intake predicting PON1 activity and report the effects of alcohol intake, folic acid, iron, and insulin use on AREase. The dietary variables alone accounted for an additional 8.2% of PON1 AREase activity, for a total of 34.67% of variance explained. When considering plasma lipid measures, apoA1 also entered the best-fi t model, which now explained 39.39% of PON1 AREase activity. Given the importance of PON1 in the pathogenesis of numerous human diseases, additional work into rare genetic variation, epistasis, and gene-environment interactions will be needed to further elucidate the determinations of PON1 enzymatic activity.
CAAD, and these studies selected for IDA. The physiological effects of iron observed in our cohort could potentially only be observed in subjects of normal iron homeostasis. Moreover, the Aslan report of association between PON1 and iron did not adjust for the potential confounder of impaired hemoglobin synthesis. In a more recent study, it was found that depleted iron stores by itself was not associated with PON1, whereas impaired erythropoiesis resulting from IDA was ( 80 ). Finally, our study was much larger (n = 1,402 versus 25 cases and 22 controls in Aslan et al.), with greater power to detect the potentially small change in AREase activity attributed to iron intake. Our result is consistent with prior reports of various metals (81)(82)(83), including iron ( 84,85 ), inhibiting PON1 activity in vitro.
The strong positive correlation of apoA1 and AREase activity supports past reports ( 72,86 ). ApoA1 is a major component of HDL and is largely responsible for reverse cholesterol transport in which excess cholesterol is transferred to the liver for excretion ( 2 ). Oral administration of exogenous apoA1 peptides alone is suffi cient to decrease atherosclerosis in mice, independent of changes in total or HDL cholesterol ( 87 ). As PON1 is localized to HDL and apoA1 is a major component of HDL, it is unsurprising that they are strongly correlated.
The positive association of alcohol intake and PON1 activity is consistent with past fi ndings (26)(27)(28). It has been suggested that alcohol's effect on protein kinase C (PKC) may phosphorylate Sp1 and regulate the binding of Sp1 to the promoter region of PON1 ( 88 ), potentially explaining the positive effect of alcohol on PON1 expressions levels. Although Sp1 levels are associated with PON1 expression, overexpression of PKC results in downregulation of PON1 ( 28 ); thus, heavy alcohol consumption leads to a decrease in PON1 .
Although the positive association of dietary cholesterol and PON1 activity appears paradoxical given their opposing relationships with vascular disease, the other predictors of PON1 activity have PON1 effects consistent with their known relationship to vascular disease. Dietary vitamin C is inversely associated with vascular disease ( 89 ), and dietary iron from red meat is associated with increased vascular risk ( 89 ). As reviewed elsewhere ( 90 ), moderate alcohol consumption is consistently associated with reduced vascular risk. Folate supplementation has not been found to be cardioprotective ( 91 ).
Some limitations of this study must be considered. First, this cohort is composed primarily of older Europeans collected for the presence or absence of CAAD. This could limit the generalizations drawn from these fi ndings. However, stratifying by CAAD status or statin use did not impact our results. Second, the dietary covariates that were analyzed are limited to those from the Health Professionals Follow-Up Study. As a result, some dietary intakes that have been reported to be associated with PON1, such as polyphenols found in pomegranate ( 17,92 ), were not analy zed in this study.
In conclusion, our analysis of dietary intake data in the CLEAR cohort has identifi ed a novel fi nding in humans of