Associations between dietary patterns and flow cytometry-measured biomarkers of inflammation and cellular activation in the Atherosclerosis Risk in Communities (ARIC) Carotid Artery MRI Study

Abstract

Background

Specific foods and overall dietary patterns are associated with soluble biomarkers of systemic inflammation and endothelial activation. However, no large epidemiological studies have evaluated relationships between such dietary factors and cell-specific markers of activation and inflammation as measured by flow cytometry.

Methods

Cell aggregates and multiple platelet and leukocyte markers were quantified by flow cytometry in fresh whole blood from 1,101 white adults participating in the Carotid Artery MRI study, a subset of the larger Atherosclerosis Risk in Communities (ARIC) study. Two dietary patterns (“Healthy” and “Western”) were empirically-derived via principal components analysis using data collected by food frequency questionnaire. Cross-sectional associations between dietary patterns and flow cytometry-measured biomarkers were evaluated, adjusting for demographics and lifestyle factors, including medications use.

Results

After multivariable adjustment, monocyte lipopolysaccharide receptor (CD14), monocyte toll-like receptor-2, and platelet glycoprotein IIb (CD41) showed inverse associations with the Healthy dietary pattern (p = 0.01, 0.04, and 0.01, respectively). In contrast, the Western dietary pattern was positively associated with CD41 and platelet-granulocyte aggregates (p = 0.01 and 0.04, respectively). Independent of other dietary factors, alcohol consumption was inversely associated with levels of pan leukocyte marker (CD45), P-selectin (CD62P) on PLA1 and on PLA2 platelets, and platelet-monocyte, platelet-granulocyte, and platelet-lymphocyte aggregates.

Conclusion

Dietary patterns and alcohol intake were each cross-sectionally associated with select markers of cellular activation and inflammation measured by flow cytometry. These data are consistent with the hypothesis that holistic measures of dietary intake are associated with inflammation.

Introduction

Inflammation and cellular activation reflect the early stages, and progression, of atherosclerosis and are associated with incident CVD^1–4. Much of the support for these associations stems from large epidemiological studies where systemic, soluble biomarkers of inflammation and endothelial activation, such as C-reactive protein, interleukin-6, homocysteine, fibrinogen, soluble intracellular adhesion molecule-1, soluble vascular cell adhesion molecule-1, and e-selectin, have been measured. Consistent with associations observed with traditional CVD risk factors, concentrations of these analytes are associated with basic demographic and lifestyle characteristics, including dietary factors^5–7.

Data from several large cohort studies have reported inverse associations between biomarkers of inflammation/cellular activation and dietary patterns reflecting high intake of fruits & vegetables, whole grains, nuts, fish, and low-fat dairy foods^8–11. Analogously, these studies have also reported positive associations between such biomarkers and dietary patterns reflecting high intake of red meat, refined grains, high-fat dairy, and fried foods^8–11. These observational data are further supported by intervention studies showing dietary change can influence concentrations of inflammatory biomarkers^12–15.

The function and activities of platelets and leukocytes are important in inflammatory and atherosclerotic processes^16–18. Flow cytometry can be used to characterize platelet and leukocyte activation and cellular aggregation in whole blood, as well as quantify both intracellular and membrane-bound platelet- and leukocyte-derived factors¹⁹. Thus, platelet and leukocyte dynamics measured by flow cytometry more closely represent real-time biological processes than do the biomarkers typically measured in large-scale epidemiological studies. However, because the technique is labor-intensive, expensive and, requires fresh samples, few large scale studies have utilized flow cytometry, and no large-scale studies have both measures by flow cytometry and measures of dietary intake.

Thus, we studied the cross-sectional associations between dietary patterns and markers of cellular activation and aggregation measured by flow cytometry in men and women from the Atherosclerosis Risk in Communities Carotid MRI Study. We hypothesized that a dietary pattern characterized by high intake of whole grains, fruits, vegetables, and fish and low intake of refined grains, red meat, and other foods high in added sugars or saturated fats would be inversely associated with flow cytometry markers reflecting inflammation and cellular activation. Analogously, we hypothesized that a dietary pattern characterized by low intake of whole grains, fruits, vegetables, and fish and high intake of refined grains, red meat, and other foods high in added sugars or saturated fats would be positively associated with these flow cytometry markers.

Methods

The Atherosclerosis Risk in Communities (ARIC) Study is a cohort comprising 15,792 African American and White adults from four U.S. communities: Forsyth County, NC; Jackson, MS; suburban Minneapolis, MN; and Washington County, MD²⁰. Between 1987 and 1998, male and female participants underwent up to four examinations, including carotid ultrasound. In 2005–2006 a stratified sample of 2,066 surviving participants were re-examined as part of an ARIC Carotid MRI study²¹. The recruitment sample was selected to achieve the following: 1) 1,200 participants with high maximum carotid intima-media thickness (IMT) values (maximum over six sites: left and right, common, bifurcation, internal) at their last ARIC ultrasound examination in the 1990s, and 2) 800 participants randomly sampled from the remainder of the carotid IMT distribution. Field-center-specific cutpoints of carotid IMT were used to approximately achieve this goal, with 100% sampling above the cutpoint, and a sampling fraction below the cutpoint to achieve the desired 800 participant total. The cutpoints were 1.35, 1.00, 1.28, and 1.22 mm IMT at Forsyth County, Jackson, Minneapolis suburbs, and Washington County, respectively, representing the 73^rd, 69^th, 73^rd, and 68^th percentiles of maximal IMT from Exam 4. Participants were ineligible for the MRI exam if they had implanted metallic devices; carotid revascularization on either side for the low IMT group or on the side selected for imaging for the high IMT group; weight > 320 pounds; or difficulty understanding questions or completing the informed consent. The current analysis is based on the Carotid MRI cross-sectional data in white participants from Minneapolis, MN and Washington County, MD, the two field centers that collected dietary intake data (n = 1,101). Examination protocols were approved by local institutional review boards at both centers, and all participants provided informed consent.

Flow Cytometry Measures

Fasting whole blood samples were obtained into Cyto-Chex® BCT vacutainer tubes (Streck, Omaha, NE) by a standardized protocol²². The BCT tubes were inverted 8 times, stored briefly at room temperature, and shipped overnight to the flow cytometry laboratory. The laboratory prepared and analyzed samples within 24 hours of blood drawing using a Coulter® Epics™XL™ flow cytometer (Beckman Coulter, Inc., Miami, FL). These methods are also provided in greater detail elsewhere, including data on quality control and reproducibility²² and risk factor correlates, which included, primarily, race/ethnicity and sex, with lower correlations seen for LDL-cholesterol and cholesterol-lowering therapy²³. Antigen-negative controls were used to set the threshold between positive and negative cell populations. Within each positive population, both the percentage of positive cells and the median fluorescence intensity (MFI) values were determined. Fluorescent intensity measures the amount of fluorochrome bound to a cell. Under appropriate conditions, it can be related to the number of binding sites a cell has for a particular fluorochrome-conjugated reagent.

Sixteen flow cytometry markers were evaluated in the present analyses (presented as MFI and, where available, percentage of positive cells (%), as results are classically presented in this manner): 8 leukocyte markers: monocyte lipopolysaccharide receptor (CD14), monocyte toll-like receptor-2 (TLR-2), TLR-4, P-selectin glycoprotein ligand-1 (CD162) in monocytes, granulocytes, and lymphocytes, pan-leukocyte marker (CD45), and monocyte myeloperoxidase (monocyte MPO); 5 platelet markers: platelet glycoprotein IIb (CD41), platelet glycoprotein IIIa (CD61) specific for PLA, P-selectin (CD62P) on PLA1, CD62P on PLA2, and platelet CD40 ligand (CD154); 3 cell aggregates: platelet-monocyte aggregates, platelet-granulocyte aggregates, and platelet-lymphocyte aggregates.

Dietary Assessment and Dietary Pattern Derivation

Prior to the clinical examination, participants were mailed a questionnaire asking about the frequency they consumed specific foods and beverages in the previous year (Willett 131-item food frequency questionnaire [FFQ]²⁴). Participants self-administered the FFQ in their homes and brought the completed surveys to their scheduled clinical examination where clinic staff briefly reviewed the FFQ with the participants to assure completeness. Forms were also reviewed a second time prior to scanning; if missing responses remained, the participant was contacted by a member of the field center staff and the missing response was reconciled, if possible. Completion rates were high in both field centers. Only one participant did not complete an FFQ, and 93% of the remaining FFQs were completed in full (no missing responses).

Line items from the FFQ were divided into 35 food or beverage groups (appendix A). Principal components analysis with orthogonal (varimax) rotation was used to empirically derive dietary patterns from these 35 food groups. After assessment of eigen values, scree plots, and dietary pattern interpretability, a two-principal component solution was selected (appendix A). Participant scores on the first principal component (dietary pattern: “Healthy”) were correlated positively with the intakes of vegetables, fruit, legumes, fish, tomatoes, whole grains, nuts, and poultry; thus, those with high scores on this dietary pattern consumed greater amounts of such foods. Participant scores on the second principal component (dietary pattern: “Western”) were correlated positively with the intakes of processed and red meats, fried potatoes, refined grains, high-fat dairy, desserts, sugar-sweetened beverages, candy, white potatoes, eggs, pizza, and butter; thus, those with high scores on this dietary pattern consumed greater amounts of such foods. By design of principal components analysis, the two dietary patterns are uncorrelated (correlation coefficient of zero), and the mean and standard deviation for each dietary pattern score are 0.0 and 1.0, respectively.

Assessment of other relevant variables

Participants provided all of their medications for transcription by field center staff. Physical activity in sports and leisure was assessed by the Baecke questionnaire²⁵. Current alcohol intake was estimated from the frequency of drinking of specified serving sizes of various alcoholic beverages (12 oz beer, 13.2 g alcohol; 4 oz wine, 10.8 g alcohol; 1.5 oz liquor, 15.1 g alcohol). Body mass index (weight in kg/height in m²) was measured by field center staff. Race/ethnicity and highest attained education level were self-reported.

Statistical Analysis

All analyses were performed using the inverse of the sampling fractions as weights. Sampling fractions were based on those persons actually screened for participation. Those who actually participated (non-refusing eligibles) were analyzed as a sub-population when calculating variances and confidence intervals of estimators. Finite population correction factors were not applied.

Unadjusted values for demographics, lifestyle characteristics, and markers of flow cytometry, and energy-adjusted (kcal/day) food/beverage intakes were calculated across score quartiles for each of the two dietary patterns using weighted linear regressions for continuous variables; standard methods were used for calculating the percentages for the dichotomous variables in each quartile. P-values for trends across the quartiles were assessed with the dietary pattern scores modeled as a continuous variable using weighted linear regressions for continuous variables and weighted logistic regressions for the dichotomous variables.

Beta regression coefficients (per 1-unit change in dietary pattern score) with standard errors and tests for trend were calculated by modeling each dietary pattern score as a continuous variable in a weighted linear regression. The following covariates were included in the multivariable model: age (years, continuous), sex (male/female), field center (MN/MD), and energy intake (kcal/day), education level (grade school or none, some high school, high school graduate, vocational school, college, graduate/professional school), smoking status (current/former/never), cigarette years (continuous), physical activity in leisure and in sport (Baecke leisure and sport scores), alcohol intake (grams/week), use of statins or other lipid-lowering medications, and regular use of aspirin or other antiplatelet medications. Models that further adjusted for body mass index, diabetes status, and CRP concentrations were also evaluated. However, results were not different with these covariates included and, therefore, are not presented.

Tests of interaction between dietary pattern scores and sex, smoking status (current/former/never smoker), and current medication use (lipid-lowering or anti-inflammatory medications, including aspirin: yes/no) were calculated by including a multiplicative term (e.g., dietary pattern score*sex) in the multivariable model. All analyses were conducted with Intercooled Stata 9.2 statistical software package (Stata Corporation, College Station, Texas).

Results

Participant characteristics

The study sample contained 569 white men and 532 white women with a mean (SD) age of 71.8 (5.5) and 70.7 (5.4) years, respectively. Characteristics of these men and women according to quartiles of both dietary patterns are shown in Table 1. Participants with higher scores on the Healthy dietary pattern were more likely to be female, have a higher level of educational attainment, engage in more leisure and sports activities, and were less likely to smoke. Participants with higher scores on the Western dietary pattern were more likely to be male, smoke, and consume more alcohol, and less likely to be physically active in their leisure (p <0.05 for all). Differences in food and beverage intake across quartiles of each dietary pattern were also apparent (Table 1) and consistent with the general dietary characteristics expected based on the results of the principal components analysis (Appendix 1).

Table 1.

Participant characteristics across quartiles of two empirically-derived dietary patterns¹

	Health Dietary Pattern Score Quartiles
	Quartile 1	Quartile 2	Quartile 3	Quartile 4	P trend2
n	276	275	275	275
Score (range)	−1.88– −0.69	−0.69 – −0.17	−0.17 – 0.50	0.50 –5.93
Age (years)	70.7 (0.5)	70.4 (0.4)	70.5 (0.4)	69.9 (0.4)	0.34
Sex (% female)	43.1	46.6	48.4	54.9	0.02
Education (% ≥college degree)	26.1	30.2	40.0	50.2	< 0.001
Smoking (% current smokers)	12.7	9.1	5.2	5.1	0.007
Cigarette (pack) years	29.0 (2.9)	27.6 (2.89)	22.4 (2.2)	19.5 (2.1)	0.005
Physical activity (% in Baeke sport score quintile 5)	10.91	10.22	13.92	23.72	0.001
Physical activity (% in Baeke leisure activity score quintile 5)	7.64	13.26	10.62	18.91	< 0.001
Alcohol intake (grams/week)	43.9 (9.0)	37.6 (5.9)	48.3 (6.2)	46.4 (8.3)	0.88
Statin use (% current users)	48.7	49.5	42.1	38.3	0.27
Other lipid-lowering medications (% current users)	53.6	53.8	49.1	46.6	0.40
Aspirin (% current users)	55.4	59.3	58.2	56.0	0.03
Other anti-platelet medications (% current users)	16.3	16.0	16.9	14.2	0.33
Diabetes (% yes)	23.1	26.9	23.7	21.0	0.46
CRP (mg/L)	3.43 (0.3)	3.96 (0.5)	3.23 (0.6)	2.52 (0.2)	< 0.001
Body Mass Index (kg/m2)	28.7 (0.4)	29.3 (0.4)	29.6 (0.4)	28.2 (0.4)	0.06
Energy intake (kcal/day)	1370 (41)	1591 (33)	1870 (43)	2074 (42)	< 0.001
Foods/Beverages (energy-adjusted servings/day)
Whole grains	0.15 (0.01)	0.31 (0.02)	0.42 (0.03)	0.64 (0.05)	< 0.001
Fruits	0.82 (0.04)	1.2 (0.05)	1.5 (0.05)	2.1 (0.08)	< 0.001
Vegetables	1.1 (0.03)	1.8 (0.04)	2.6 (0.05)	4.1 (0.09)	< 0.001
Low-fat dairy	0.53 (0.03)	0.86 (0.06)	1.0 (0.08)	1.1 (0.07)	0.005
High-fat dairy	0.72 (0.05)	0.77 (0.05)	0.79 (0.05)	0.80 (0.04)	< 0.001
Red meat	0.74 (0.04)	0.84 (0.04)	0.90 (0.04)	0.83 (0.04)	< 0.001
Fin fish (not fried)	0.09 (0.00)	0.13 (0.01)	0.18 (0.01)	0.26 (0.01)	< 0.001
Desserts	1.1 (0.11)	1.0 (0.07)	1.2 (0.09)	0.97 (0.08	< 0.001
Sugar-sweetened beverages	0.44 (0.05)	0.31 (0.04)	0.29 (0.04)	0.21 (0.03)	< 0.001
Coffee	5.9 (0.3)	6.2 (0.3)	6.5 (0.3)	6.3 (0.3)	0.40
Tea	0.20 (0.05)	0.39 (0.07)	0.42 (0.07)	0.58 (0.09)	0.002

	Western Dietary Pattern Score Quartiles
	Quartile 1	Quartile 2	Quartile 3	Quartile 4	P trend1
n	276	275	275	275
Score (range)	−2.75 – −0.71	−0.71 – −0.12	−0.12 – 0.52	0.52 – 4.91
Age (years)	70.0 (0.4)	70.7 (0.4)	70.0 (0.4)	70.7 (0.4)	0.56
Sex (% female)	68.8	54.2	42.9	26.9	< 0.001
Education (% ≥college degree)	39.9	35.6	36.0	34.9	0.61
Smoking (% current smokers)	3.3	6.9	6.9	15.0	< 0.001
Cigarette years	17.0 (2.0)	22.7 (2.1)	26.1 (2.6)	32.4 (3.0)	< 0.001
Physical activity (% in Baeke sport score quintile 5)	18.3	13.1	11.7	15.8	0.35
Physical activity (% in Baeke leisure activity score quintile 5)	18.9	13.1	8.8	8.8	0.04
Alcohol intake (grams/week)	27.0 (5.0)	41.1 (6.5)	43.8 (7.1)	67.8 (70.4)	0.001
Statin use (% current users)	46.4	43.1	47.6	42.5	0.34
Other lipid-lowering medications (% current users)	52.5	52.0	52.6	46.0	0.11
Aspirin (% current users)	55.1	59.6	55.5	58.8	0.87
Other anti-platelet medications (% current users)	15.9	15.3	18.3	13.9	0.63
Diabetes (% yes)	25.0	24.1	24.4	21.3	0.64
CRP (mg/L)	3.19 (0.3)	3.62 (0.4)	3.08 (0.3)	3.14 (0.2)	0.17
Body Mass Index (kg/m2)	28.8 (0.4)	28.6 (0.4)	28.9 (0.4)	29.4 (0.4)	0.57
Energy intake (kcal/day)	1285 (34)	1593 (31)	1815 (29)	2374 (42)	< 0.001
Foods/Beverages (energy-adjusted servings/day)
Whole grains	0.43 (0.04)	0.36 (0.03)	0.41 (0.03)	0.37 (0.03)	< 0.001
Fruits	1.6 (0.1)	1.5 (0.1)	1.4 (0.1)	1.3 (0.1)	< 0.001
vegetables	2.6 (0.1)	2.4 (0.1)	2.4 (0.1)	2.6 (0.1)	< 0.001
Low-fat dairy	1.0 (0.1)	1.1 (0.1)	0.8 (0.1)	0.8 (0.1)	< 0.001
High-fat dairy	0.44 (0.03)	0.65 (0.04)	0.86 (0.04)	1.2 (0.06)	< 0.001
Red meat	0.45 (0.02)	0.70 (0.02)	0.87 (0.03)	1.4 (0.05)	< 0.001
Fish	0.17 (0.01)	0.16 (0.01)	0.17 (0.01)	0.18 (0.01)	< 0.001
Dessert foods	0.50 (0.03)	0.86 (0.06)	1.3 (0.08)	1.7 (0.12)	0.001
Sugar-sweetened soda	0.08 (0.01)	0.23 (0.03)	0.32 (0.03)	0.64 (0.06)	< 0.001
Coffee	5.6 (0.3)	6.1 (0.3)	6.3 (0.3)	7.1 (0.3)	0.07
Tea	0.42 (0.07)	0.37 (0.07)	0.45 (0.08)	0.39 (0.07)	0.69

¹All values are unadjusted with the exception of servings per day of foods and beverages which were adjusted for total energy intake (kcal/day, continuous).

²p for trend across quartiles calculated with the dietary pattern score modeled on a continuous scale

Dietary patterns and flow cytometry markers

Beta coefficients from the weighted regression representing expected unit changes in flow cytometry markers per 1-unit change in dietary pattern scores are shown in Table 2. When adjusted for demographics and lifestyle characteristics, three markers showed statistically significant inverse associations with the Healthy dietary pattern, and two markers showed statistically significant positive associations with the Western dietary pattern. For every 1-unit increase in the Healthy dietary pattern, monocyte lipopolysaccharide receptor (CD14) decreased 2.3 MFI units, TLR-2 decreased 0.09 MFI units, and platelet glycoprotein IIb (CD41) decreased 0.42 units (% positive), (p = 0.01, 0.04, and 0.01, respectively). The percent of cells positive for platelet glycoprotein IIIa (CD61), P-selectin on PLA1, and P-selectin on PLA2 platelets each tended to decrease as scores on the Healthy dietary pattern increased, but these associations were not statistically significant (p = 0.052, 0.10 and 0.08, respectively). For every 1-unit increase in Western dietary pattern score, glycoprotein IIb (CD41) increased 0.50 units and platelet-granulocyte aggregates increased 0.46 units (% positive; p = 0.01 and 0.04, respectively). Monocyte lipopolysaccharide receptor (CD14), platelet-monocyte, and platelet-lymphocyte cell aggregate markers tended to increase as scores on the Western dietary pattern increased, but the associations were not formally significant (p = 0.07 for all).

Table 2.

Associations between two empirically-derived dietary patterns and markers of inflammation and blood cellular activation measured by flow cytometry

	Units1	B ± SE per 1-unit increase in dietary pattern score2	P3
“HEALTHY” DIETARY PATTERN
Leukocyte markers
Monocyte lipopolysaccharide receptor (CD14)	MFI	−2.33 ± 0.89	0.01
Toll-like receptor (TLR)-2	%	−0.65 ± 0.54	0.23
	MFI	−0.09 ± 0.04	0.04
Toll-like receptor (TLR)-4	%	0.24 ± 0.22	0.27
	MFI	0.03 ± 0.05	0.62
P-selectin glycoprotein ligand-1 (CD162+) on granulocytes	%	−0.01 ± 0.02	0.69
	MFI	0.39 ± 0.68	0.56
P-selectin glycoprotein ligand-1 (CD162+) on lymphoyctyes	%	−0.43 ± 0.29	0.14
	MFI	−0.28 ± 0.47	0.54
P-selectin glycoprotein ligand-1 (CD162+) on monocytes	%	0.006 ± 0.008	0.48
	MFI	0.53 ± 0.80	0.51
Pan-leukocyte marker (CD45)	MFI	−0.28 ± 1.17	0.81
Monocyte myeloperoxidase (MPO)	MFI	9.4 ± 6.76	0.17
Platelet markers
Platelet glycoprotein IIIa (CD61)	%	−1.4 ± 0.72	0.052
	MFI	−0.36 ± 0.76	0.64
Platelet glycoprotein IIb (CD41)	%	−0.41 ± 0.16	0.01
	MFI	−0.07 ± 0.56	0.90
P-selectin (CD62P) on PLA1 platelets	%	1.09 ± 0.66	0.10
	MFI	0.37 ± 1.11	0.74
P-selectin (CD62P) on PLA2 platelets	%	1.10 ± 0.63	0.08
	MFI	0.19 ± 0.20	0.33
CD40 ligand (CD154)	%	0.04 ± 0.12	0.73
	MFI	0.02 ± 0.09	0.79
Cell aggregates
Platelet-monocyte aggregates	%	−0.23 ± 0.18	0.20
	MFI	−0.14 ± 0.27	0.60
Platelet-granulocyte aggregates	%	−0.22 ± 0.17	0.20
	MFI	−0.17 ± 0.29	0.55
Platelet-lymphocyte aggregates	%	−0.17 ± 0.17	0.34
	MFI	0.06 ± 0.33	0.85
“WESTERN” DIETARY PATTERN
Leukocyte markers
Monocyte lipopolysaccharide receptor (CD14)	MFI	2.32 ± 1.29	0.07
Toll-like receptor (TLR)-2	%	0.75 ± 0.81	0.35
	MFI	0.08 ± 0.07	0.23
Toll-like receptor (TLR)-4	%	−0.39 ± 0.31	0.21
	MFI	−0.04 ± 0.07	0.57
P-selectin glycoprotein ligand-1 (CD162+) on granulocytes	%	0.06 ± 0.06	0.30
	MFI	0.22 ± 0.94	0.81
P-selectin glycoprotein ligand-1 (CD162+) on lymphoyctyes	%	0.21 ± 0.36	0.56
	MFI	0.67 ± 0.59	0.26
P-selectin glycoprotein ligand-1 (CD162+) on monocytes	%	0.06 ± 0.06	0.33
	MFI	−0.28 ± 1.05	0.79
Pan-leukocyte marker (CD45)	MFI	−0.15 ± 1.5	0.92
Monocyte myeloperoxidase (MPO)	MFI	−10.5 ± 9.2	0.26
Platelet markers
Platelet glycoprotein Ilia (CD61)	%	1.22 ± 1	0.22
	MFI	0.27 ± 1.08	0.80
Platelet glycoprotein lib (CD41)	%	0.50 ± 0.18	0.01
	MFI	0.59 ± 0.93	0.53
P-selectin (CD62P) on PLA1 platelets	%	−0.8 ± 0.91	0.38
	MFI	−0.27 ± 1.59	0.87
P-selectin (CD62P) on PLA2 platelets	%	−0.77 ± 0.86	0.37
	MFI	−0.10 ± 0.28	0.71
CD40 ligand (CD154)	%	0.01 ± 0.22	0.98
	MFI	−0.08 ± 0.13	0.53
Cell aggregates
Platelet-monocyte aggregates	%	0.44 ± 0.24	0.07
	MFI	0.22 ± 0.36	0.55
Platelet-granulocyte aggregates	%	0.46 ± 0.22	0.04
	MFI	0.06 ± 0.39	0.88
Platelet-lymphocyte aggregates	%	0.41 ± 0.23	0.07
	MFI	−0.47 ± 0.46	0.30

¹Percentage of positive cells (%) or mean fluorescence intensity (MFI).

²Data are adjusted for energy intake (kcal/day, continuous), age (years, continuous), sex (male/female), study center (MN or WA), education level (Grade school or none, some high school, high school graduate, vocational school, college, graduate/professional school), smoking status (never, former, current), cigarette pack years (continuous), physical activity level (Baecke leisure and sport score, continuous), alcohol intake (grams/week, continuous), current statin or other lipid-lower medication use (yes/no), and current/regular aspirin or other anti-platelet medication use (yes/no).

³p calculated with the dietary pattern score modeled as a continuous variable in a weighted linear regression model.

Interactions

Of the tested interactions (both dietary patterns x each sex, smoking status, and medication use for all 16 markers), 11 were formally significant, p < 0.05. Of these observed interactions, four were between medication use and the Western dietary pattern, two were between sex and the Healthy dietary pattern, and 5 were with smoking status (1 for the Western pattern and 4 for the Healthy pattern). These interactions are depicted in Table 3, where in general, associations between the dietary pattern scores and select flow cytometry markers varied by sex, smoking status, and medication use strata, but the direction of regression coefficients did not follow consistent patterns across marker outcomes.

Table 3.

Interactions between dietary patterns and select participant characteristics in prediction of flow cytometry markers

INTERACTION	Expected change in percent positive cells per 1-unit increase in dietary pattern score (β ± SE)	P for β	(β ± SE) for interaction	P for interaction
MEDICATION1 USE * WESTERN DIETARY PATTERN 2
Platelet glycoprotein IIIa (CD61) (%)			3.06 ± 1.20	0.011
Non-users (n = 229)	−1.39 ± 1.64	0.399
Current users (n = 756)	3.32 ± 1.25	0.008
P-selectin (CD62P) on PLA1 platelets (%)			−2.94 ± 1.09	0.007
Non-users (n = 244)	1.63 ± 1.55	0.294
Current users (n = 791)	−2.72 ± 1.11	0.014
P-selectin (CD62P) on PLA2 platelets (%)			−2.59 ± 1.04	0.012
Non-users (n = 229)	0.91 ± 1.44	0.531
Current users (n = 791)	−2.33 ± 1.07	0.030
CD40 ligand (CD154) (MFI)			−0.59 ± 0.25	0.019
Non-users (n = 236)	0.72 ± 0.42	0.089
Current users (n = 774)	−0.40 ± 0.23	0.084
SEX * HEALTHY DIETARY PATTERN 3
Platelet glycoprotein IIb (CD41) (%)			0.54 ± 0.24	0.024
Women (n = 485)	−0.54 ± 0.23	0.020
Men (n = 525)	−0.15 ± 0.22	0.510
CD40 ligand (CD154) (MFI)			−0.31 ± 0.12	0.007
Women (n = 485)	0.15 ± 0.11	0.175
Men (n = 525)	−0.23 ± 0.12	0.062
SMOKING * WESTERN DIETARY PATTERN 4
Platelet glycoproteln IIb (CD41) (%)			0.54 ± 0.22	0.015
Never smokers (n = 447)	0.13 ± 0.29	0.650
Former smokers (n = 479)	0.57 ± 0.23	0.015
Current smokers (n = 84)	1.88 ± 1.01	0.065
SMOKING * HEALTHY DIETARY PATTERN 4
Toll-like receptor-4 (%)			0.54 ± 0.27	0.041
Never smokers (n = 452)	0.10 ± 0.25	0.682
Former smokers (n = 485)	0.20 ± 0.38	0.604
Current smokers (n = 85)	2.13 ± 0.84	0.013
Toll-like receptor -2 (MFI)			0.13 ± 0.06	0.044
Never smokers (n = 454)	−0.11 ± 0.06	0.070
Former smokers (n = 485)	−0.10 ± 0.06	0.100
Current smokers (n = 86)	0.02 ± 0.18	0.930
Toll-like receptor -4 (MFI)			0.16 ± 0.07	0.028
Never smokers (n = 452)	−0.02 ± 0.06	0.768
Former smokers (n = 485)	−0.002 ± 0.09	0.979
Current smokers (n = 85)	0.57 ± 0.28	0.046
PSGL-1 monocytes (MFI)			−3.43 ± 0.99	0.001
Never smokers (n = 452)	3.19 ± 0.99	0.001
Former smokers (n = 483)	−2.77 ± 1.11	0.013
Current smokers (n = 86)	4.09 ± 3.14	0.197

¹Medication use = use of statins, other lipid-lowering meds, aspirin use, or other anti-platelet med use (yes any vs. no to all)

²Data were adjusted for all variables in the multivariable model but without medication use.

³Data were adjusted for all variables in the multivariable model but without sex.

⁴Data were adjusted for all variables in the multivariable model but without smoking status.

Intake of specific foods/beverages and flow cytometry markers

Individual food groups that contributed importantly to dietary pattern scores (vegetables, fruit, legumes, whole grains, fin fish, fried fish, red or processed meat) were not significantly associated with any of the flow cytometry markers (data not shown). Alcohol intake as a whole (beer, wine, and/or liquor) was inversely associated with all three cell aggregate markers, both P-selectin on PLA1 and PLA2 platelets, and pan-leukocyte marker (CD45), but alcohol intake was positively associated with platelet glycoprotein IIIa (CD61) and monocyte MPO. These associations were independent of demographics, lifestyle factors, medications use, and other dietary factors represented by dietary patterns (p = 0.002–0.04, Table 4).

Table 4.

Associations between alcohol intake and eight biomarkers measured by flow cytometry¹

	Units	Predicted units change in flow cytometry markers (B ± SE) per 14-g (~1 drink) increase in daily intake of alcohol	p
Platelet glycoprotein IIIa (CD61)	%	1.27 ± 0.58	0.016
Pan-leukocyte marker (CD45)	MFI	−2.55 ± .98	0.012
Monocyte myeloperoxidase	MFI	15.78 ± 6.18	0.011
P-selectin (CD62P) on PLA1 platelets	%	−1.47 ± 0.49	0.002
P-selectin (CD62P) on PLA2 platelets	%	−1.27 ± 0.49	0.005
Platelet-monocyte aggregates	%	−0.29 ± 0.10	0.029
	MFI	−0.59 ± 0.20	0.003
Platelet-granulocyte aggregates	%	−0.29 ± 0.10	0.038
	MFI	−0.39 ± 0.20	0.026
Platelet-lymphocyte aggregates	%	−0.29 ± 0.10	0.033
	MFI	−0.69 ± 0.20	0.002

¹Values adjusted for energy intake (kcal/day, continuous), age (years, continuous), sex (male/female), study center (MN or WA), education level (Grade school or none, some high school, high school graduate, vocational school, college, graduate/professional school), smoking status (never, former, current), cigarette pack years (continuous), physical activity level (Baecke leisure and sport score, continuous), current statin or other lipid-lower medication use (yes/no), current/regular aspirin or other anti-platelet medication use (yes/no), Healthy dietary pattern scores, and Western dietary pattern scores.

Discussion

In this sample of over one thousand free-living, adult men and women, two primary dietary patterns were evident—one reflective of healthy dietary habits and one reflective of less-healthy, “Westernized” dietary habits. Similar antithetical patterns of food consumption have been reported in many epidemiologic studies^{26, 27}. While several studies have shown that empirically-derived dietary patterns such as these are associated with diabetes^28–33, cardiovascular disease^34–39, and risk factors for these diseases with inflammatory etiology^{8, 40–47}, the investigation presented here is the first to show associations with sophisticated markers of cellular activation as measured by flow cytometry. Our sample size is substantial relative to other studies using flow cytometry techniques and boasts a well-characterized participant sample with a wealth of demographic, lifestyle, and clinical information. Of the 16 markers evaluated, three were inversely associated with the Healthy dietary pattern, two were positively associated with the Western dietary pattern, and six were inversely associated with alcohol consumption, consistent with the hypothesized relationships of dietary factors with inflammation and cellular activation.

Others have reported associations between similarly characterized dietary patterns and soluble markers of inflammation and endothelial activation^9–11. Our results are generally consistent with these observations, showing dietary patterns characterized by greater consumption of whole grains, fruits and vegetables, fish and nuts and lower consumption of red meats and foods that are highly process or high in saturated fat are associated with lower risk of CVD and related risk factors⁴⁸. In contrast to conventional methods that quantify soluble analytes reflective of inflammation and cellular activation, flow cytometry quantifies cell-specific markers of activation, theoretically allowing for more precise identification of particular components of pathological processes¹⁹. However, the number of significant dietary pattern-marker associations observed in our study was few and scattered among cell types, precluding us from making substantive claims about the role of diet in specific aspects of inflammatory pathways. Nevertheless, the associations reported here are valuable since data derive from a free-living, observational cohort where specific dietary prescriptions or bioactive dietary constituents could not be modulated to meet an “ideal” or to reach postulated “effect” thresholds. Thus, under more controlled conditions, it is possible that stronger and more consistent associations with flow cytometry measurements could be observed.

When analyzed singly, food groups that importantly distinguished high and low scores on each the “Healthy” and “Western” dietary patterns were not associated with flow cytometry markers, with the exception of alcohol. Among several motives for dietary pattern construction is the concept that disease associations with individual foods (or nutrients) may be either indiscernible due to their small effect sizes, or indiscernible due to synergy and antagonism with other dietary factors²⁶. The data shown here support these concepts. In the case of alcohol, where greater intake was associated with lower levels of CD45 and cell aggregates (platelet-monocyte, platelet-granulocyte, and platelet-lymphocyte aggregates), our data are consistent with the hypothesis that alcohol intake is favorably related to hemostasis and thrombosis risk⁴⁹. Other differently designed studies utilizing other methods to assess platelet activation and aggregation have also observed inverse associations with alcohol consumption^50–52. However, in our study alcohol intake was also positively associated with CD61 and monocyte MPO. Such associations contradict these general hypotheses and are not internally consistent with the associations observed with other cell markers in our study. Furthermore several sensitivity tests, such as stratification by drinking frequency, exclusion of heavy drinkers, or exclusion of non-drinkers, produced very similar results.

While there are several assets to our study, including its large sample size, well-defined participant characteristics (i.e., potential confounders), and state-of-the art methodology for assessing cellular activation, there are limitations. First, this was a cross-sectional analysis. Thus, temporal relations between dietary factors and cellular markers cannot be established, although it seems likely that reverse causal associations (where diet is a function of inflammatory status) would obscure associations or produce associations opposite of those hypothesized— i.e., healthier dietary habits would appear to be associated with greater levels of the markers studied or vice versa, which is not what was observed. Secondly, error in the measurement of diet, or covariates, is possible. While the error in measurement of dietary patterns would most likely be random, serving to attenuate associations, error in the measurement of covariates could have precluded us from fully adjusting for confounding factors, an effect that would have impacts of varying direction and degree on estimated associations. Lastly, many markers were evaluated; significant associations could have been the result of chance due to multiple testing. For example, if Bonferroni correction had been applied, none of the reported associations would have been statistically significant, i.e., 2 dietary patterns* [16 markers measured as MFI +13 markers measured as % positive cells] = 58 tests: 0.05/58, thus p = 0.00086 corrected significance level. The threshold for significance for tests of interaction would have been even lower if corrections were applied, given that three possible effect modifiers were considered (sex, smoking, medication use) with each dietary pattern. The potential for high type I error, coupled with the fact that the observed interactions did not follow a consistent pattern, make drawing inferences from these tests ambiguous.

In conclusion, the data shown here are consistent the supposition that dietary intake, particularly intake of alcohol, plays a role in modulating inflammation and cellular activation. While chance cannot be excluded as a possible explanation for the associations reported here, this uncertainty is more likely due to the novel application of this method and phenotype in a population-based study. Nevertheless, the data are unique from previous studies in that they derive from a sophisticated method of assessing real-time cellular activation in a relatively substantial sample of adult men and women with well-characterized demographic, dietary, and other lifestyle information.