Associations of Dietary Long-Chain n-3 Polyunsaturated Fatty Acids and Fish with Biomarkers of Inflammation and Endothelial Activation (From the Multi-Ethnic Study of Atherosclerosis [MESA])

Abstract

Cardioprotective effects of long-chain n-3 polyunsaturated fatty acids (LC n-3 PUFA) and fish consumption have been observed. However, data on the specific associations of these dietary factors with inflammation and endothelial activation are sparse. We conducted a cross-sectional study of 5,677 men and women from the MESA cohort including African Americans, Caucasians, Chinese and Hispanics, aged 45-84 years, and free of clinical cardiovascular disease. Dietary information was collected by self-administered food frequency questionnaire. Multivariable linear regression analyses were used to examine relations between intake of LC n-3 PUFAs, non-fried fish and fried fish and biomarkers of inflammation and endothelial activation. LC n-3 PUFA intakes were inversely associated with plasma concentrations of interleukin-6 (IL-6, P=0.01) and matrix metalloproteinase-3 (MMP-3, P=0.03) independent of age, body mass index, physical activity, smoking, alcohol consumption and dietary variables. Non-fried fish consumption was found inversely related to C-reactive protein (CRP, P=0.045) and IL-6 (P<0.01); and fried fish was observed being inversely related to soluble intercellular adhesion molecules-1 (sICAM-1) (P<0.01) but not associated with other biomarkers after adjustment for potential confounders. In conclusion, this study suggests that dietary intakes of LC n-3 PUFAs and fish are inversely associated with concentrations of some biomarkers reflecting lower levels of inflammation and endothelial activation. These results may partially explain the cardioprotective effects of fish consumption.

The findings from previous studies on intake of long-chain n-3 polyunsaturated fatty acids (LC n-3 PUFA) and fish and inflammation and endothelial activation are inconsistent. Also, the fatty acid content in fish may change as a result of deep frying.¹ However, little is known about how frying fish modifies associations of fish consumption with inflammatory markers. Therefore, we investigated associations of fish consumption and LC n-3 PUFA intake with markers of inflammation and endothelial activation in men and women aged 45-84 years in the Multi-Ethnic Study of Atherosclerosis (MESA). We hypothesized that intake of LC n-3 PUFAs and non-fried fish would be inversely associated with markers of inflammation and endothelial activation.

Methods

The MESA cohort was established in 2000-02 with 6814 men and women aged 45 to 84 years, without clinical cardiovascular disease (CVD), recruited from 6 U.S. communities. This cohort is 38.5% Caucasian, 27.8% African American (AA), 21.9% Hispanic, and 11.8% Chinese. Details of sampling and recruitment procedures have been published elsewhere.² Participants were requested to complete a set of comprehensive measurements of subclinical CVD and assessment of standard CVD risk factors, sociodemographic factors, diet, lifestyle, and psychosocial factors. In the present study, we excluded those who had missing data on diet (n=577), missing data on all biomarkers (n=26), extreme total caloric intake (<600 or >6000 kcal/day, n=293), and had no information on covariates used in analyses (n=227). Some biomarkers were measured in sub-cohorts including CD40L, E-selectin, sTNF-R1, sICAM-1, MMP-3 and MMP-9. After these exclusions, the total numbers of participants remaining in the analyses by biomarkers are: CD40L, n=860; CRP, n=5677; IL-6, n=5569; E-selectin, n=860; fibrinogen, n=5676; sTNF-R1, n=860; sICAM-1, n=2266; MMP-3, n=860; and MMP-9, n=860.

The MESA protocol was approved by institutional review boards at all participating sites, and all participants provided informed consent.

Self-administered 127-item food frequency questionnaire (FFQ) and dietary supplement form in Block format³ were used to collect dietary information. Participants were asked to identify their average consumption of each selected food during the previous year with 9 frequency options ranging from never/rare to frequently and also the serving size as either small, medium or large. Food consumption was adjusted for serving size, for example, the gram weight for small, medium and large portion, specific to age group (middle or old) and gender (i.e. small = 0.5 × medium; medium = standard age- and gender-specific serving size; large = 1.5 × medium).⁴ For fish consumption, which we defined as fish and other seafood intake, participants were asked to indicate how often they consumed 1) fried fish or fish sandwich, fried shrimp, calamari; 2) shrimp, lobster, crab, oysters, mussels (not fried); 3) tuna, salmon, sardines (including sashimi or sushi); and 4) other broiled, steamed, baked or raw fish (such as trout, sole, halibut, poke, grouper). Fish consumption was classified in 2 groups: fried fish and non-fried fish (broiled, steamed, baked or raw fish). In addition, the questionnaire contained three questions related to mixed fish dishes (for example, Chinese food with stir-fried shrimp or fish with vegetables; pasta with seafood; and seafood gumbo). Because of the limited information, we were not able to determine the exact portion of seafood in the mixed fish dishes. Thus, we did not include mixed fish dishes in the main analyses. Instead, we used this information in the secondary analyses by assigning a weight 40%, 50% or 60% respectively for fish in the mixed fish dishes. Nutrients including LC n-3 PUFAs were derived from the Minnesota nutrition data system (NDS-R version 4.02/30).⁵

The FFQ was adopted from the FFQ used in the Insulin Resistance Atherosclerosis Study (IRAS), which has been validated in AA, Hispanics and Caucasians.⁶ Compared to the average of eight 24-hour recalls, the mean correlation coefficients of nutrient intake were 0.62 for non-Hispanic Caucasians, 0.50 for AA, and 0.41 for Hispanics. The apparent lower degree of validity in Hispanics was accounted for by lower education attainment in that group. For reliability, the mean correlation coefficient for nutrients evaluated was 0.62 and did not differ by ethnic subgroup. To accommodate the MESA population, the FFQ was modified to include Chinese foods and culinary practices using the same approach as used previously to enhance ethnic sensitivity for the validated IRAS.

Blood was collected, processed and samples were stored using standardized protocols at the Laboratory for Clinical Biochemistry Research at University of Vermont, Burlington, VT, USA.⁷ Participants were asked to fast for 12 hours, avoid smoking on the morning of the exam and avoid heavy exercise 12 hours before the exam. CRP and fibrinogen antigen were measured using the BNII nephelometer; average analytical coefficients of variation (CVs) are 3.6% and 2.7%, respectively. IL-6 and sTNF-R1 were measured by ultra-sensitive ELISA; average laboratory analytical CVs for these assays were 6.3% and 5.0%, respectively. sICAM-1, soluble E-selectin, and CD40L were measured by sandwich enzyme immunoassay; average analytical CVs were 5.0% and 5.9%, respectively. MMP-9 was measured by a high sensitivity quantitative sandwich enzyme immunoassay; average analytical CV was 4.9%. MMP-3 was measured by an ultra-sensitive, solid-phase sandwich ELISA using a polyclonal antibody specific for both the pro- and active forms of MMP-3; average analytical CV was 6.9%.

Demographic and major lifestyle variables were collected through questionnaire. Body weight and height were directly measured. Body mass index (BMI) was calculated as the ratio of weight (kg) to squared height (m). physical activity was measured using a detailed, semi-quantitative questionnaire adopted from the Cross-Culture Activity Participation Study.² Total physical activity was computed as the total of all light, moderate, and vigorous activities (minutes per week) multiplied by the activities' individual metabolic equivalent values.

Means and standard deviations or proportions were computed for selected variables by quartiles of non-fried fish consumption. All biomarkers were log-transferred to better approximate normal distributions. Multiple linear regressions were used to examine fish consumption and LC n-3 PUFA intake in relation to biomarkers of inflammation and endothelial activation. Based on the previous literature and univariate tests, we considered a number of potential confounders in the multivariate models. All p values were 2-sided. A p value <0.05 was considered statistically significant. We used SAS (version 9; SAS Institute, Inc., Cary, North Carolina) software for analyses.

Results

The characteristics of the study population according to non-fried fish consumption are presented in Table 1. Participants in the highest quartile were more likely to be female, physically active, have higher education level, household income and lower BMI and were less likely to be current smokers than those in the lowest quartile. Also, consumption of non-fried fish was positively associated with intakes of LC n-3 PUFAs and total energy, and inversely related to intake of saturated fat, polyunsaturated fat, alpha-linolenic acid, linoleic acid and trans fatty acids.

Table 1.

Characteristics of the study population according to quartiles of non-fried fish consumption, MESA, 2000

	Non-fried fish consumption
Characteristics	Quartile 1	Quartile 2	Quartile 3	Quartile 4	P value
Non-fried (broiled, steamed, baked or raw) fish, median (servings/week)	0.0	0.3	0.8	2.0	--
Fried fish, median (servings/week)	0.1	0.2	0.2	0.2	--
Number of participants	1482	1293	1532	1384	--
Age, (years)	62.3 (10.5)	61.9 (10.2)	62.1 (10.2)	62.4 (10.0)	0.52
Men	49.2%	46.6%	45.0%	46.3%	0.13
Race/ethnicity
African American	19.5%	25.3%	26.9%	24.6%	<0.01
Caucasian	32.9%	40.0%	43.5%	47.0%
Chinese	9.2%	12.4%	13.2%	18.0%
Hispanic	38.5%	22.4%	16.4%	10.4%
Education
Less than high school	29.4%	14.8%	14.0%	11.6%	<0.01
High school	22.9%	19.3%	16.4%	12.4%
Some college	26.7%	30.5%	29.5%	24.3%
College or higher	20.9%	35.5%	40.1%	51.7%
Household annual income
≤ $16,000	25.6%	16.6%	15.5%	16.1%	<0.01
$16,001-$39,999	37.0%	33.4%	29.4%	25.4%
≥$40,000	37.4%	50.0%	55.1%	58.5%
BMI, (kg/m2)	28.7 (5.4)	28.3 (5.4)	28.1 (5.5)	27.6 (5.2)	<0.01
Intentional exercise, (MET-min/week)	1247 (1913)	1491 (2303)	1666 (2431)	1890 (2624)	<0.01
Smoking status
Never smoker	51.2%	50.1%	50.8%	53.0%	<0.01
Former smoker	33.3%	38.6%	37.2%	37.6%
Current smoker	15.5%	11.3%	12.0%	9.4%
Pack-years of cigarette smoking	12.8 (28.7)	10.8 (19.6)	10.4 (18.3)	10.4 (21.0)	<0.01
Alcohol consumption, (drinks/week)	4.1 (7.4)	3.6 (6.0)	3.6 (5.8)	4.1 (5.8)	0.23
Use of fish oil supplements	3.4%	4.0%	3.8%	5.9%	<0.01
Daily dietary intake
EPA + DHA, (% energy)	0.035 (0.027)	0.053 (0.028)	0.076 (0.036)	0.149 (0.080)	<0.01
Saturated fat, (% energy)	11.3 (3.1)	11.0 (3.0)	10.4 (2.9)	9.8 (3.0)	<0.01
Polyunsaturated fat, (% energy)	8.2 (3.1)	7.8 (2.8)	7.6 (2.7)	7.8 (2.7)	<0.01
Alpha-linolenic acid, (% energy)	0.666 (0.267)	0.654 (0.231)	0.648 (0.216)	0.663 (0.217)	0.13
Linoleic acid, (% energy)	6.7 (2.9)	6.4 (2.7)	6.2 (2.6)	6.2 (2.6)	<0.01
Trans fatty acid, (% energy)	1.8 (0.8)	1.8 (0.7)	1.7 (0.7)	1.5 (0.7)	<0.01
Total energy intake, (kcal)	1589 (681)	1563 (659)	1604 (644)	1747 (687)	<0.01

Data are mean (standard deviation), unless otherwise specified.

BMI: body mass index; MET: metabolic equivalent-minutes; EPA: eicosapentaenoic acid; DHA: docosahexaenoic acid.

In age-adjusted linear regression models, LC n-3 PUFA intake was significantly inversely related to CRP, IL-6, fibrinogen, TNF-R1, sICAM-1 and MMP3. Non-fried fish consumption was significantly inversely associated with CRP, IL-6, fibrinogen and sICAM-1. Fried fish consumption was positively related to CRP and IL-6 but inversely associated with CD40L and sICAM-1 (Table 2).

Table 2.

Relations of long-chain n-3 fatty acid and fish intake and biomarkers of inflammation and endothelial activation

	EPA + DHA		Non-fried fish		Fried fish
Markers	Age-adjusted	Multivariate-adjusted	Age-adjusted	Multivariate-adjusted	Age-adjusted	Multivariate-adjusted
Log CD40L (ng/mL) *	-0.23 (0.54)	0.42 (0.27)	-0.08 (0.57)	0.17 (0.24)	-0.49 (0.04)	-0.11 (0.68)
Log hsCRP (mg/L)**	-1.24 (<0.01)	-0.24 (0.31)	-0.38 (<0.01)	-0.17 (0.03)	0.74 (<0.01)	0.16 (0.35)
Log IL-6 (pg/mL)***	-0.88 (<0.01)	-0.37 (<0.01)	-0.28 (<0.01)	-0.16 (<0.01)	0.29 (<0.01)	-0.07 (0.46)
Log E-selectin (ng/mL)*	-0.31 (0.24)	0.07 (0.79)	-0.13 (0.19)	0.01 (0.88)	0.16 (0.36)	0.08 (0.66)
Log fibrinogen (mg/dL) ****	-0.09 (0.03)	-0.05 (0.24)	-0.04 (<0.01)	-0.03 (0.84)	0.06 (0.06)	-0.01 (0.80)
Log sTNF-R1 (pg/mL) *	-0.53 (<0.01)	-0.26 (0.06)	-0.09 (0.07)	-0.05 (0.31)	-0.11 (0.22)	-0.09 (0.35)
Log sICAM-1 (ng/mL) *****	-0.55 (<0.01)	-0.16 (0.06)	-0.08 (<0.01)	0.01 (0.96)	-0.24 (<0.01)	-0.17 (0.01)
Log MMP3 (ng/mL) *	-0.79 (0.02)	-0.69 (0.03)	-0.14 (0.25)	-0.18 (0.12)	0.02 (0.92)	-0.05 (0.81)
Log MMP9 (ng/mL) *	-0.31 (0.29)	0.50 (0.11)	0.09 (0.41)	0.16 (0.14)	-0.23 (0.22)	-0.07 (0.74)

Values are β coefficients (P-values);

Non-fried fish includes broiled, steamed, baked or raw fish.

Adjusted for age, race, gender, household income, education, BMI, physical activity, smoking status, aspirin use, NSAID use, alcohol consumption, and intakes of saturated fat, polyunsaturated fat, α-linolenic acid, trans-fatty acids and total energy. For fish consumption, additionally adjusted for either fried fish or non-fried fish.

^*Biomarker available in: n=860,

^**Biomarker available in: n=5677,

^***Biomarker available in: n=5569,

^****Biomarker available in: n=5676,

^*****Biomarker available in: n=226

After adjustment for multiple covariates, significant inverse associations remained between LC n-3 PUFA intake and IL-6 and MMP3 and between non-fried fish consumption and CRP and IL-6. Also, the inverse relation between fried fish consumption and sICAM-1 persisted after adjustment for multiple covariates (Table 2).

Since linoleic acid (n-6) may compete with n-3 fatty acids in the eicosanoid synthesis pathway,⁸ we stratified analyses according to median intake of linoleic acid. We also stratified data according to median intake of α-linolenic acid because α-linolenic acid can be desaturated and elongated to LC n-3 PUFAs and high levels of α-linolenic acid intake may compensate for low intake of LC n-3 PUFAs. Although the observed associations of LC n-3 PUFA intake and biomarkers became statistically non-significant in low intake groups of linoleic acid and α-linolenic acid, the tests for interactions were not significant. (Table 3).

Table 3.

Relations of long-chain n-3 PUFA intake and biomarkers stratified by linoleic fatty acid (n-6) or α-linolenic fatty acid (n-3) intake, MESA, 2000

Markers	Linoleic acid intake below median (< 5.77 % energy)	Linoleic acid intake above median (≥ 5.77 % energy)	P for interaction	α-linolenic acid intake below median (< 0.61 % energy)	α-linolenic acid intake above median (≥ 0.61 % energy)	P for interaction
Log CD40L (ng/mL)	0.14 (0.83)	0.30 (0.54)	0.55	0.73 (0.26)	0.18 (0.71)	0.78
Log hsCRP (mg/L)	0.02 (0.96)	-0.40 (0.21)	0.85	0.05 (0.88)	-0.43 (0.15)	0.45
Log IL-6 (pg/mL)	-0.10 (0.62)	-0.50 (0.007)	0.44	-0.26 (0.20)	-0.39 (0.02)	0.67
Log sE-selectin (ng/mL)	0.03 (0.94)	0.23 (0.53)	0.46	0.50 (0.25)	0.04 (0.92)	0.57
Log fibrinogen (mg/dL)	-0.05 (0.40)	-0.05 (0.37)	0.46	-0.10 (0.14)	-0.03 (0.55)	0.22
Log sTNF-R1 (pg/mL)	-0.25 (0.25)	-0.19 (0.31)	0.39	-0.39 (0.09)	-0.16 (0.35)	0.12
Log sICAM-1 (ng/mL)	-0.11 (0.37)	-0.15 (0.21)	0.78	-0.21 (0.12)	-0.11 (0.32)	0.57
Log MMP3 (ng/mL)	-0.51 (0.31)	-0.75 (0.06)	0.98	-0.13 (0.79)	-1.00 (0.02)	0.32
Log MMP9 (ng/mL)	0.07 (0.89)	0.89 (0.03)	0.33	0.51 (0.32)	0.62 (0.12)	0.61

Values are β coefficients (P values);

Adjusted covariates and number of participants for each biomarker are listed in Table 2 except α-linolenic acid.

Discussion

In this cross-sectional study, we found that intake of LC n-3 PUFAs derived from marine products was inversely related to some, but not all biomarkers of inflammation and endothelial activation. The observed inverse associations were independent of major lifestyle factors and potential dietary and non-dietary confounders.

Data from experimental studies have provided some evidence that LC n-3 PUFAs are anti-inflammatory.⁹ Some epidemiological studies also indicate that these fatty acids may have beneficial effects on inflammation and endothelial function.^10-15 Of note, there was no information on type of fish meal in the previous studies relating fish to inflammation and endothelial function.

Few intervention studies have examined effects of fish oil supplementation on inflammation and endothelial activation and the findings were inconsistent. Hjerkinn et al. found significantly reduced concentrations of sICAM-1 in a fish oil supplementation group compared to placebo control group in elderly men with long-standing hyperlipidemia.¹⁶ Also, in a double-blinded intervention study among 60 patients with coronary heart disease, Seierstad et al. found that IL-6 was significantly reduced by fish oil compared to rapeseed oil supplementation.¹⁷ However, in a trial among 54 patients with coronary heart disease, Johansen et al. observed that patients who had taken 5.1g/d fish oil for 4 weeks had higher levels of sE-selectin compared to the patients who had taken corn oil as placebo.¹⁸ In another randomized, double-blind trial among 300 acute myocardial infarction patients, Grundt et al. found no effects on sE-selectin, sICAM-1 and CRP by fish oil supplementation compared to corn oil.¹⁹ Also, in an open-labeled randomized controlled trial among 77 post-myocardial infarction patients, Lee et al. reported no associations between fish oil supplementation and IL-6 and fibrinogen.²⁰ In addition, Berstad et al. found that a moderate dietary increase in LC n-3 PUFAs might have an effect on decreasing inflammatory markers, whereas fish oil supplementation might increase inflammation in vascular endothelium in a 18-month intervention in 171 men aged between 65 and 75 years.²¹ Of note, these intervention studies were conducted in patients with clinical CVD and the sample sizes were generally quite small.

One possible biological mechanism underlying the beneficial effects of LC n-3 PUFAs on inflammation and endothelial function is that these fatty acids compete with n-6 fatty acids for prostaglandin and leukotriene synthesis at the cyclooxygenase and lipoxygenase level. LC n-3 PUFAs from fish or fish oil modulate prostaglandin metabolism by increasing prostaglandin E₃ (an active vasodilator and inhibitor of platelet aggregation), thromboxane A₃ (a weak platelet aggregator and a weak vasoconstrictor) and leukotriene B₅ (a weak inducer of inflammation) and by decreasing production of thromboxane A₂ (a potent platelet aggregator and vasoconstrictor) and leukotriene B₄ formation (an inducer of inflammation and a powerful inducer of leukocyte chemotaxis and adherence).²² Another suggested mechanism is that LC n-3 PUFAs may react with active oxygen species because of their multiple double bonds and lead to a decreased production of hydrogen peroxide. Hydrogen peroxide is a critical activator of the nuclear factor-κB system of transcription factors that controls the coordinated expression of adhesion molecules and of leukocyte-specific chemoattractants upon cytokine stimulation.²³

In the present study, we found independent inverse associations of LC n-3 PUFAs and non-fried fish with IL-6, CRP and MMP3. The relevance of these inflammatory and endothelial activation biomarkers in the atherogenic process has been recognized. Studies suggest that CRP and IL-6, 2 systemic inflammatory markers, are independent predictors of CVD and may play important role in atherogenesis.²⁴ In addition, MMP3 is suggested an independent prognostic factor in stable coronary artery disease.²⁵ Several lines of evidence supports the important role of MMPs in plaque stability.²⁶

sICAM-1 is thought to be a key factor in the adherence of monocytes to the endothelium and subsequently transmigration into the intima. The role of sICAM-1 in the pathogenesis of inflammation and atherosclerosis has been confirmed by experimental studies.²⁷ Studies also indicate that sICAM-1 is a predictor of CVD independent of other traditional risk factors.²⁸ In the present study, we observed that fried fish but not non-fried fish consumption was significantly inversely related to sICAM-1. This finding is not expected since the frying process may reduce the content of LC n-3 PUFAs and produce trans-fatty acids. One possible explanation is that fried fish consumption may be a marker of relatively unhealthy lifestyle; those who had high fried fish consumption were more likely to suffer from dyslipidemia and were likely to be under treatment or medication (e.g., taking statin), which may lower the level of sICAM-1.²⁹ However, in this cohort, the inverse association between fried fish consumption and sICAM-1 level remained significant with adjustment for or exclusion of participants using cholesterol-lowering medication. Alternatively, the observed inverse association between fried fish and sICAM-1 in this study may be simply due to chance. Future studies are warranted.

In addition, some other limitations merit consideration. Because of the cross-sectional design, the possibility of reverse causation can not be completely excluded. Also, we could not account for any effects of recent change in diet or medication on inflammatory markers. For instance, participants at high risk of CVD may have changed their diets (i.e. increased fish intake) because of obtaining advice from health care workers or receiving information via media. This diet change would be expected to attenuate associations between LC n-3 PUFAs or fish and biomarkers of inflammation and endothelial activation. Inevitable measurement errors in estimating fish consumption and LC n-3 PUFA intake as well as in the biomarker measures could bias our results and would also be expected to attenuate associations. Although we adjusted for a number of potential confounders in the analysis, the possibility of confounding from unmeasured dietary and non-dietary factors can not be ruled out. For example, information on mercury and other contaminants in fish was not available. Study suggests that mercury may increase risk of myocardial infarction.³⁰ Thus, mercury in fish might confound our findings.