Habitual Intake of Marine-Derived n-3 PUFAs is Inversely Associated with a Cardiometabolic Inflammatory Profile in Yup'ik Alaska Native People

ABSTRACT

Background

The relationship between dietary n-3 PUFAs and the prevention of cardiometabolic diseases, including type 2 diabetes, is unresolved. Examination of the association between n-3 PUFAs and chronic low-grade inflammation in a population where many individuals have had an extremely high intake of marine mammals and fish throughout their lifespan may provide important clues regarding the impact of n-3 PUFAs on health.

Objectives

The aim of this study was to explore associations between concentrations of n-3 PUFAs resulting from habitual intake of natural food sources high in fish and marine mammals with immune biomarkers of metabolic inflammation and parameters of glucose regulation.

Methods

A total of 569 Yup'ik Alaska Native adults (18–87 years old) were enrolled in this cross-sectional study between December 2016 and November 2019. The RBC nitrogen isotope ratio (NIR; ¹⁵N/¹⁴N) was used as a validated measure of n-3 PUFA intake to select 165 participant samples from the first and fourth quartiles of n-3 PUFA intakes. Outcomes included 38 pro- and anti-inflammatory cytokines and 8 measures of glucose homeostasis associated with type 2 diabetes risks. These outcomes were evaluated for their associations with direct measurements of EPA, DHA, and arachidonic acid in RBCs.

Analysis

Linear regression was used to detect significant relationships with cytokines and n-3 PUFAs, adiposity, and glucose-related variables.

Results

The DHA concentration in RBC membranes was inversely associated with IL-6 (β = –0.0066; P < 0.001); EPA was inversely associated with TNFα (β = –0.4925; P < 0.001); and the NIR was inversely associated with Monocyte chemoattractant protein-1 (MCP-1) (β = –0.8345; P < 0.001) and IL-10 (β = –1.2868; P < 0.001).

Conclusions

Habitual intake of marine mammals and fish rich in n-3 PUFAs in this study population of Yup'ik Alaska Native adults is associated with reduced systemic inflammation, which may contribute to the low prevalence of diseases in which inflammation plays an important role.

Introduction

Inflammation often accompanies obesity and is a risk factor for chronic, noncommunicable diseases (1), including type 2 diabetes (T2D) (2). Obesity-associated inflammation is associated with insulin resistance and thought to be mediated in part through the production of TNFα in adipose tissue (3, 4). In Yup'ik Alaska Native people, the prevalence of overweight and obesity is similar to that of the general US population (5), but the prevalence of T2D is only ∼3%, less than one-fourth of that reported in the US general population (6, 7). We hypothesize that this discordance between the prevalence of obesity and T2D in Yup'ik people is, in part, due to their intake of a traditional subsistence diet highly enriched in anti-inflammatory n-3 PUFAs derived from marine mammals and fish (8).

Dietary intake of saturated fat leads to the activation of inflammatory cytokine gene expression and the inhibition of insulin signaling (9). In contrast, n-3 PUFA intake inhibits the production of proinflammatory cytokines (8, 10–12), even in the presence of obesity (8); helps to maintain insulin sensitivity (9, 13, 14); and has cardioprotective effects (9). Historically, Yup'ik people lived a very active lifestyle and relied exclusively on locally harvested foods, including fish, marine mammals, berries, game, and wild greens. This traditional Yup'ik diet includes amounts of EPA and DHA up to 20 times higher than those observed in the general US population (15). Yup'ik individuals consuming this traditional diet exhibit cardioprotective lipid profiles and elevated adiponectin (16), which may contribute to their low T2D prevalence, lower blood pressure measurements (17), and reduced platelet activation (18).

Metabolic inflammation resulting from recruitment of both innate and adaptive immune cells is associated with immune cell infiltration of several tissues, including adipose, pancreas, liver, and brain tissues (4, 19, 20). Proinflammatory M1-polarized macrophages, CD4 + Th1 cells, and CD8 + effector T cells commonly found in adipose tissues of obese individuals collectively contribute to the synthesis and release of a spectrum of proinflammatory cytokines and chemokines believed to play a significant role in the development of insulin resistance and impaired β-cell secretory function (21), the combination of which leads to T2D.

Although not commonly considered among the immunomodulatory therapies proposed to prevent and treat cardiometabolic diseases such as T2D (22, 23), n-3 PUFAs have been shown to have protective effects against insulin resistance in rodent models, but not in humans, where results have been mixed (24). Most randomized controlled trials of n-3 PUFA supplementation have resulted in no demonstrable benefits on insulin sensitivity (25). However, these studies are typically of limited duration and may not capture the effects of habitually high n-3 PUFA intakes, utilize n-3 PUFA doses sufficiently high enough to create meaningful blood concentration differences between the control and treatment groups (26), or incorporate the n-3 PUFAs into a whole foods dietary pattern (27).

To address several of these limitations in the current literature, we comprehensively investigated associations between EPA, DHA, and arachidonic acid (AA) in RBCs with markers of chronic low-grade inflammation in a population habituated to a diet high in natural sources of EPA and DHA. In addition, we examined associations between these n-3 and n-6 PUFAs with circulating adiponectin; high-sensitivity C-reactive protein (hsCRP); and regulated upon activation, normal T-cell expressed and secreted [Regulated on Activation, Normal T Cell Expressed and Secreted (RANTES)]; as well as with measures of insulin resistance (HOMA2-IR), insulin sensitivity (HOMA2-S), beta cell function (HOMA2-B), and insulin secretory capacity relative to insulin sensitivity [HOMA2-disposition index (HOMA2-DI)] (28).

Methods

Participants and data collection

For this study, from a cohort of 569 participants, we selected a sub-sample of 216 Yup'ik participants at the extremes of n-3 PUFA intakes using an objective biomarker of n-3 PUFA intake: the nitrogen isotope ratio (NIR; ¹⁵N/¹⁴N) in RBCs (29). Immune biomarkers were only measured on this sub-sample of 216 participants because of limited resources and the exploratory nature of this study. The sex distributions were unbalanced at the extremes, forcing us to include some individuals in the second and third quartiles of NIR to balance the sexes. However, in the experience of the core laboratory that runs our clinical assays, cytokine panels may provide higher values due to an increased background or cross-reactivity with other targets in the panel. Therefore, for the purpose of the current study, we chose to limit the statistical analyses to quartiles 1 and 4 of the NIR distribution; 17 participants in the second quartile and 4 in the third quartile were thus excluded, leaving 195 participants. From this participant sub-sample, 7 participants were excluded because they had an hsCRP concentration > 1 mg/dL, 15 were excluded because they had a BMI below 20 kg/m², 5 were excluded because they had a BMI above 40 kg/m², 4 were excluded because they had T2D, 1 were excluded due to taking an immunosuppressive medication, and 2 participants had 2 of the exclusion criteria, leaving 165 participants for statistical analyses. BMI cutoff criteria were used because we previously found (8) the relationship between BMI and several clinical biomarkers to be nonlinear at BMIs <20 or >40 kg/m². The study data were collected and managed using REDCap electronic data capture tools hosted at Oregon Health and Science University (OHSU) (30, 31). This study was approved by the University of Alaska Fairbanks (UAF) and OHSU Institutional Review Boards, as well as by the Yukon-Kuskokwim Health Corporation Human Studies Committee and Executive Board of Directors.

Anthropometric measurements and blood sample collection

Anthropometric measures and a fasting blood sample were obtained as previously described (32). Anthropometric measurements included height, weight, waist circumference (WC), percentage body fat, and resting blood pressure. Two trained researchers collected measures from all participants using standard NHANES III Anthropometric Procedures Manual protocols (33), as previously described (34). Blood samples were obtained by venipuncture from individuals after a 12-hour fast, collected in EDTA tubes, and processed into plasma, lymphocyte, and RBC fractions in the rural communities using a portable centrifuge. Samples for glucose analysis were collected in sodium fluoride tubes, immediately placed in a cooler rack, and processed within 5 minutes. All samples were initially stored between −15°C to −20°C for up to a week (while in rural Alaska), before being shipped to the UAF or OHSU and stored at −80°C or in liquid nitrogen cryovats.

Dietary biomarker measurements

The NIRs of RBC samples collected from EDTA tubes were analyzed by the Alaska Stable Isotope Facility at the UAF by continuous-flow isotope ratio MS as previously described (29). Natural abundance stable isotope ratios are conventionally expressed as the per mille (‰) abundance of ¹⁵N relative to an internal standard:

(1)

The NIR has previously been validated as an objective biomarker of EPA and DHA intake in this population, and reflects the past 3–4 months of dietary intake (29, 35). As shown previously (29), there was a positive, linear correlation between EPA and NIR; a positive, nonlinear relationship between DHA and NIR; and an inverse, linear correlation between AA and NIR (Supplemental Figure 1).

Analysis of PUFAs from RBCs

The analytical method was adapted from a previously published protocol (36, 37). Briefly, a 10 μl aliquot of packed RBCs was diluted 1:100 with 0.9% NaCl in water, and deuterated EPA, DHA, AA internal standards, butylated hydroxytoluene, and triphenylphosphine were added. A 20 μL aliquot of the diluted sample was removed for further processing. An additional 50 μl aliquot was further diluted 1:4 with 0.9% NaCl in water for total protein quantitation using a bicinchoninic acid (BCA) kit. The 1:100 diluted sample was dried under a gentle stream of nitrogen, reconstituted in isopropanol, saponified with the addition of 200 μl 1M Potassium hydroxide (KOH), and then incubated in a water bath at 37°C for 45 minutes. The hydrolyzed sample was then neutralized with the addition of 200 μl 1M of HCl. Following neutralization, analytes of interest (AA, EPA, and DHA) were isolated by Biotage 96-well Evolute Express ABN SPE 30-mg plate solid-phase extraction, dried under a gentle stream of nitrogen, and derivatized with the addition of AMPP+ [N-(4-aminomethylphenyl)pyridinium]. The derivatized samples were then filtered and subjected to ultra performance liquid chromatography (UPLC)-MS/MS analysis using a Waters Xevo TQ-S-ESI tandem quadrupole MS operated in the positive ion mode and coupled to a BEH Shield RP18 chromatography column. A 10-mM formic acid/water/acetonitrile (ACN) mobile phase was used for analyte separation. The mass transitions for quantitation of the analytes were: 471.4 > 183.2 for AA, 479.4 > 183.2 for AA-d₈, 469.4 > 183.2 for EPA, 474.4 > 183.2 for EPA-d₅, 495.4 > 183.2 for DHA, and 501.4 > 183.2 for DHA-d₅. Calibration curve samples were prepared from stock solutions of authentic EPA, DHA, and AA standards in ethyl acetate and processed in the same manner as diluted RBCs. Data were collected using MassLynx, and peak areas were integrated using QuanLynx software. The inter-day accuracy (percentage error) and precision (percentage deviation) across a range of AA, EPA, and DHA concentrations were all <8%. Intra-day assay precision (percentage error) and accuracy (percentage deviation) for replicate measurement of a range of AA, EPA, and DHA concentrations were all <15%.

Clinical laboratory measurements

Fasting and 2-hour oral glucose tolerance test samples were collected in sodium fluoride tubes, and plasma derived from these samples was analyzed on a Roche Hitachi 704 Chemistry Analyzer by the Lipoprotein Analytical Core (LAC) at OHSU or on a Siemens Dimension Xpand analyzer at the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) Phoenix Branch. Reagents used at the LAC were from Randox (Cat# GL1611), and the reagents used at the NIDDK included the Siemens Dimension Glucose Flex Cartridge (DF40-10444971) and the Siemens Chem I Calibrator (DC18C-10716280). The intra- and inter-assay coefficients from the LAC were reported by Randox as 3.7% and 5.6%, respectively. The intra- and inter-assay coefficients of variation from the NIDDK were 0.22% and 0.5%, respectively.

Plasma concentrations of insulin were measured by the Endocrine Technologies Core (ETC) at the Oregon National Primate Research Center (ONPRC) and by the University of California Davis (UCD). Insulin was measured using an RIA from Millipore Sigma at UCD, and the intra- and inter-assay CVs were 5.9% and 7.6%, respectively. Concentrations of insulin in human plasma were determined in the ETC using a chemiluminescence-based clinical automatic immunoassay platform (Roche Cobas e411, Roche Diagnostics). Intra- and inter-assay variations in the ETC were <7%. Using fasting glucose and insulin values, we then calculated the HOMA2-B, HOMA2-S, HOMA2-IR, and HOMA2-DI (the product of insulin sensitivity and β-cell function) were calculated by the HOMA2 method (28).

Plasma samples were analyzed for immune biomarkers in the ETC and the ONPRC using a 38-plex human cytokine Luminex panel (Millipore Sigma). Cytokines, chemokines, and growth factors included are shown in Supplemental Table 1. This panel included 15 proinflammatory cytokines, 10 proinflammatory chemokines, 6 anti-inflammatory cytokines, and 7 growth factors. Data were calculated using Milliplex Analyst software version 5.1 (EMD Millipore). Vendor-supplied quality control samples and an in-house human serum pool were analyzed to monitor assay variation. The percentages of samples below the lower limits of quantification (LLOQs) and inter-assay CVs are also shown in Supplemental Table 1. A total of 20 cytokines assayed were excluded from the statistical analysis because they either had more than 30% of the samples below the detection level threshold or an inter-assay CV > 15%.

Plasma RANTES concentrations were determined by ELISA by the ETC following the manufacturer's instructions (R&D Systems). Plasma was diluted at 1:100 for analysis. The assay range, when adjusted for dilution, was 3130 to 200,000 pg/ml. Intra- and inter-assay CVs for an in-house human serum pool was 3.4%.

Additionally, IL-6, hsCRP, and adiponectin were measured in sodium citrate plasma. IL-6 was measured with an ELISA kit (R&D Systems), and the range of the assay was 3.1 to 300 pg/ml. The IL-6 inter-assay CV was 10.6%, and the intra-assay CV ranged between 1.6% and 4.2%. A Randox assay was used to measure hsCRP, and the assay range was 0.477–10.0 mg/l, with a reported inter-assay CV range between 1.22% and 4.44% and an intra-assay CV range between 1.21% and 2.66%. A Randox assay was also used to measure adiponectin (Cat# AO8154). Randox reports the adiponectin intra- and inter-assay coefficients as 1.6% and 2.5%, respectively, and the assay range is 0.32 to 23.8 mg/ml.

Statistical analysis

All analyses were performed using the R statistical programming language, version 3.6.1 (R Studio). Outcome variables were transformed to approximate normality using either a power transform or a logarithmic transform. After applying the transform, values that were more than 2 times the IQR above the third quartile or below the first quartile were classified as outliers and dropped from the analysis. When appropriate, 1 of 3 transforms was used: no transformation (i.e., power = 1), square root transformation (i.e., power = 0.5), or a logarithmic transform (represented as power = 0). A different power was chosen only if it resulted in a variable with much better skewness and kurtosis than could be achieved by 1 of these 3 transforms, or if fewer transformed values were outliers or both. After transforming the variable, it was multiplied by a constant to make its variance close to 1. Supplemental Table 2 shows the powers used, the multipliers, and the number of resulting outliers, as well as the skewness, kurtosis, minimum, and maximum of the transformed data. Age, PUFA, adiposity, blood pressure, and physical activity variables were transformed using a rank-inverse-normal transformation. P values for comparisons of raw data in males compared with females and quartile 1 compared with quartile 4 of NIR were based on t-tests for the normalized variables (Table 1). We tested whether the proportion of data below the LLOQ differed between extremes of NIR using Fisher's exact test (Table 2). Continuous variables that were used only as predictors (age, measures of adiposity, measures of PUFA levels) in the linear models were not transformed, except for EPA levels, which were logarithm transformed because of their extreme skewness.

TABLE 1.

Demographic and health-related risk factors in Yup'ik participants by gender and by quartiles of n-3 PUFA intake¹

Variable	Male n = 92	Female n = 73	P value2	Quartile 1 NIR n = 78	Quartile 4 NIR n = 87	P value
Age, y	33.5 (32.2)	40.0 (31.0)	0.323	25.5 (13.0)	54.0 (26.0)	<0.001
NIR, ‰	7.4 (2.9)	9.8 (3.4)		7.1 (0.5)	10.3 (1.1)
DHA, ng/mg protein	225.5 (145.2)	279.6 (129.1)	<0.001	167.9 (68.7)	308.5 (75.6)	<0.001
EPA, ng/mg protein	45.7 (107.6)	112.2 (128.9)	0.012	26.1 (19.6)	153.5 (85.5)	<0.001
AA, ng/mg protein	321.5 (79.8)	342.9 (105.5)	0.261	371.0 (83.1)	302.8 (79.5)	<0.001
DHA + EPA, ng/mg protein	280.9 (274.8)	420.2 (258.0)	<0.001	187.7 (94.4)	467.7 (139.0)	<0.001
BMI, kg/m2	23.9 (4.1)	28.3 (7.8)	<0.001	24.1 (7.5)	25.6 (6.5)	0.450
Waist circumference, cm	82.7 (12.6)	88.8 (21.4)	0.053	82.3 (16.5)	87.5 (16.5)	0.038
Percent body fat, %	17.1 (8.3)	35.8 (12.5)	<0.001	19.9 (18.8)	27.0 (16.2)	0.010
Systolic blood pressure, mm Hg	113.2 (16.2)	100.3 (15.6)	<0.001	108.4 (17.6)	104.0 (21.4)	0.614
Diastolic blood pressure, mm Hg	67.7 (12.2)	67.4 (12.4)	0.921	67.6 (12.4)	67.5 (12.2)	0.158
Glucose, mg/dL	97.0 (10.6)	97.6 (11.8)	0.791	97.0 (10.1)	98.7 (11.5)	0.839
HbA1c, %	5.2 (0.3)	5.2 (0.3)	0.955	5.2 (0.4)	5.2 (0.3)	0.635
2-hour glucose, mg/dL	65.8 (42.2)	83.7 (36.3)	<0.001	74.8 (31.0)	75.8 (36.2)	0.713
Insulin, μU/mL	11.6 (7.4)	14.9 (8.5)	0.018	12.7 (9.1)	13.6 (7.7)	0.830
Adiponectin,3 μg/mL	8.0 (6.1)	9.7 (5.3)	0.054	8.2 (5.8)	9.2 (5.8)	0.605
HOMA2-B	104.3 (34.7)	116.5 (34.9)	0.012	109.4 (42.9)	107.1 (36.0)	0.557
HOMA2-S	75.4 (48.3)	58.3 (33.1)	0.015	66.7 (46.4)	63.8 (36.0)	0.968
HOMA2-IR	1.5 (1.0)	2.0 (1.1)	0.017	1.7 (1.2)	1.8 (1.0)	0.887
HOMA2-DI	7117.7 (2650.0)	6634.5 (2231.5)	0.107	6893.3 (2821.4)	6692.2 (2131.8)	0.746
hsCRP,3 mg/L	0.4 (1.3)	0.4 (0.8)	0.787	0.4 (0.9)	0.4 (1.0)	0.951
IL-6,3 pg/mL	1.1 (2.1)	1.5 (2.2)	0.459	1.4 (2.5)	1.1 (1.9)	0.596
RANTES,3 ng/mL	44.7 (36.6)	31.5 (28.3)	0.014	41.7 (37.9)	36.9 (32.7)	0.487
Sedentary time (% awake wear time)	16.1 (9.8)	13.5 (10.0)	0.295	13.9 (9.4)	15.9 (10.9)	0.330
MVPA (% awake wear time)	18.0 (7.8)	14.0 (3.3)	<0.001	15.0 (7.2)	15.0 (5.6)	0.930
Total Physical Activity (VMU/min)	80.6 (24.3)	73.4 (12.0)	<0.001	74.7 (22.0)	76.3 (13.8)	0.878
Triglycerides, mg/dL	72.5 (31.0)	77.7 (46.2)	0.082	74.0 (48.3)	73.9 (32.2)	0.625
Total cholesterol, mg/dL	192.2 (58.2)	214.5 (57.7)	<0.001	179.7 (48.6)	222.7 (45.7)	<0.001
LDL cholesterol, mg/dL	113.7 (50.7)	133.3 (39.6)	0.002	107.6 (42.2)	137.1 (41.8)	<0.001
HDL cholesterol, mg/dL	58.1 (19.5)	69.2 (24.0)	0.024	55.5 (21.0)	68.6 (21.8)	<0.001
non-HDL cholesterol, mg/dL	130.4 (49.4)	148.7 (42.8)	0.002	124.6 (48.2)	154.2 (41.7)	<0.001

¹Values are reported as medians (IQRs). Blood-derived measures were assayed using plasma samples. AA, arachidonic acid; HbA1c, glycated hemoglobin; HOMA2-B, homeostasis model assessment of beta cell function; HOMA2-DI, homeostasis model assessment of disposition index; HOMA2-S, homeostasis model assessment of insulin sensitivity; hsCRP, high-sensitivity C-reactive protein; MVPA, moderate to vigorous physical activity; NIR, nitrogen isotope ratio; RANTES, regulated upon activation, normal T-cell expressed and secreted; VMU, vector magnitude units.

²The P values from the t-test are based on transformed data. Significant differences between men and women or between the lowest or highest quartiles (quartiles 1 and 4, respectively) of the NIR, an objective biomarker of n-3 PUFA intake, are indicated.

³Adipokines and cytokines were measured by ELISA. Age, PUFA, adiposity, blood pressure, and physical activity variables were transformed using a rank-inverse-normal transformation. All the outcome variables were transformed using Power Transformation.

TABLE 2.

Luminex panel characteristics: proportion of samples below the LLOQ overall and in the lowest and highest quartiles (quartiles 1 and 4, respectively) of n-3 PUFA intake among Yup'ik participants¹

	% Sample < LLOQ3	Quartile 1 (n = 78)	Quartile 4 (n = 87)	P value2
Proinflammatory Cytokines
IFN-α2	56.97	39 (50.0)	55 (63.2)	0.115
IFN-γ	40.61	27 (34.6)	40 (46.0)	0.155
IL-12p40	60.61	40 (51.3)	60 (69.0)	0.0262
IL-12p70	55.76	35 (44.9)	57 (65.5)	0.0122
IL-15	11.52	8 (10.3)	11 (12.6)	0.808
sCD40L	0.00	0 (0.0)	0 (0.0)	1.000
IL-17A	84.85	63 (80.8)	77 (88.5)	0.195
IL-1α	66.06	43 (55.1)	66 (75.9)	0.0052
IL-1β	74.55	54 (69.2)	69 (79.3)	0.155
IL-2	97.58	77 (98.7)	84 (96.6)	0.623
IL-3	100.00	78 (100.0)	87 (100.0)	1.000
IL-6	60.00	38 (48.7)	61 (70.1)	0.0072
IL-7	33.94	22 (28.2)	34 (39.1)	0.188
TNFα	0.00	0 (0.0)	0 (0.0)	1.000
TNFβ	52.12	31 (39.7)	55 (63.2)	0.0032
Proinflammatory Chemokines
Eotaxin-1	0.00	0 (0.0)	0 (0.0)	1.000
GRO	9.09	3 (3.8)	12 (13.8)	0.0312
MCP-3	13.94	7 (9.0)	16 (18.4)	0.114
IL-8	1.21	0 (0.0)	2 (2.3)	0.498
MDC	0.00	0 (0.0)	0 (0.0)	1.000
IP-10	0.00	0 (0.0)	0 (0.0)	1.000
MCP-1	0.00	0 (0.0)	0 (0.0)	1.000
MIP-1α	50.91	35 (44.9)	49 (56.3)	0.162
MIP-1β	0.00	0 (0.0)	0 (0.0)	1.000
Fractalkine	26.06	16 (20.5)	27 (31.0)	0.156
Anti-Inflammatory Cytokines
IL-10	15.15	7 (9.0)	18 (20.7)	0.0492
IL-13	64.85	43 (55.1)	64 (73.6)	0.0152
IL-1RA	7.27	5 (6.4)	7 (8.0)	0.770
IL-9	69.70	47 (60.3)	68 (78.2)	0.0172
IL-4	18.79	11 (14.1)	20 (23.0)	0.166
IL-5	34.5	24 (30.8)	33 (37.9)	0.413
Growth Factors/Adipokines
EGF	0.60	0 (0.0)	1 (1.1)	1.000
FGF-2	8.43	2 (2.6)	11 (12.6)	0.0202
TGFα	86.75	66 (84.6)	77 (88.5)	0.499
G-CSF	33.73	20 (25.6)	36 (41.4)	0.0482
Flt-3L	38.55	26 (33.3)	37 (42.5)	0.262
GM-CSF	22.29	12 (15.4)	24 (27.6)	0.062
VEGF	51.20	33 (42.3)	51 (58.6)	0.0432

¹Panel characteristics for 165 samples from the lower and upper quartiles of NIR (an objective biomarker of n-3 PUFA intake) with the number (%) of individuals below the LLOQ in each quartile. EGF, epidermal growth factor; FGF, fibroblast growth factor; Flt-3L, Fms-related tyrosine kinase 3 ligand; Fms, Feline McDonough Sarcoma; G-CSF, granulocyte colony-stimulating factor; GM-CSF, Granulocyte-macrophage colony-stimulating factor; GRO, growth related cytokine; IP, interferon γ-induced protein; LLOQ, lower limit of quantification; MCP, Monocyte chemoattractant protein-1; MDC, Macrophage-derived chemokine; MIP, macrophage inflammatory protein; NIR, nitrogen isotope ratio; sCD40L, soluble cluster of differentiation 40 ligand; TGF, transforming growth factor; VEGF, vascular endothelial growth factor.

²Fisher exact test P values are shown, and significant differences between quartiles are indicated.

³Number (%) samples below the LLOQ in each quartile for a given biomarker.

All linear regression models included sex (coded 0 for male and 1 for female), age, coastal (dichotomous variable indicating whether the community was coastal or inland), smoking (coded as 0 for no cigarette use and 1 for any cigarette use), and total physical activity (mean vector magnitude per minute) while awake from a triaxial accelerometer (ActiGraph GT3X, ActiGraph Corp, USA) worn for 2 days. We used community location as a dichotomous covariate to account for the population substructure between coastal and tundra communities, as previously described (38). In addition, when the cytokines were outcomes, we included statin, blood pressure, and anti-inflammatory medication use as covariates. NIR was coded as a dichotomous variable when used as a predictor in the models. Bonferroni's post hoc correction was used to adjust for multiple comparisons. Because of the correlation among variables, the effective number of tests was less than the actual number of tests. The method of Li and Ji (39) was used to derive the effective number of tests from the eigenvalues of the correlation matrix to determine the appropriate Bonferroni correction, 0.05/(effective number) (40). For the sensitivity analysis, we used permutation tests for all association tests.

Results

Demographic and health-related risk factors

Demographic characteristics of the 165 sub-sample participant group (upper and lower quartiles of RBC NIRs from a larger cohort) included in all subsequent analyses are shown in Table 1. Data are presented by gender, as well as by the NIR quartile, where quartile 1 represents a low n-3 PUFA intake and quartile 4 represents a high n-3 PUFA intake. Although the overall mean ages for male and female participants were similar, individuals in quartile 4 were significantly older than those in quartile 1. In quartile 1, there were 31 females (39.7%), compared to 42 females (48.3%) in quartile 4. As expected, all n-3 PUFA measures were greater in quartile 4 compared to quartile 1, whereas AA (a n-6 PUFA) values were greater in quartile 1 compared to quartile 4. Women had significantly higher DHA, EPA, total n-3 PUFA (DHA + EPA), and NIR values compared to men.

Women also had significantly higher BMIs and percentages of body fat, as well as higher insulin, adiponectin, HOMA2-B, and HOMA2-IR values, but lower HOMA2-S values, compared to men. Individuals in quartile 4 had higher WCs and percentages of body fat compared to individuals in quartile 1. Men were more active than women, and had higher systolic blood pressure measurements. Additionally, both women (compared to men) and individuals in quartile 4 (compared with quartile 1) had higher total cholesterol, LDL, HDL, and non-HDL cholesterol.

Luminex panel characteristics

Characteristics and descriptions of the 38 immune biomarkers included in the Luminex cytokine panel are shown in Supplemental Table 1. Biomarker data were not analyzed further if more than 30% of the samples assayed were below the LLOQ. This resulted in the elimination of 12 proinflammatory cytokines (IFN-α2, IFN-γ, IL-12p40, IL-12p70, IL-17A, IL-1α, IL-1β, IL-2, IL-3, IL-6, IL-7, and TNFβ); 1 proinflammatory chemokine [macrophage inflammatory protein (MIP)-1α], 3 anti-inflammatory cytokines (IL-13, IL-9, and IL-5), and 4 growth factors [transforming growth factor α, granulocyte colony-stimulating factor (G-CSF), FLT-3L, and vascular endothelial growth factor (VEGF)]. IL-6 was also measured using an ELISA assay, and only 69 of the 409 samples assayed were below the LLOQ, so IL-6 was included in further analyses.

For the remaining cytokines and chemokines subjected to statistical analysis, individual values below the LLOQ were excluded. To determine whether there were more inflammatory cytokines and chemokines eliminated from quartile 4 than quartile 1, we evaluated the number of samples below the LLOQ in each group (Table 2). In every case where the number of samples below the LLOQ was significantly different between quartile 1 and quartile 4 [IL-12p40, IL-12p70, IL-1α, IL-6, TNFβ, growth related cytokine (GRO), IL-10, IL-13, IL-9, fibroblast growth factor 2 (FGF-2), G-CSF, and VEGF], we noted more samples below the LLOQ in quartile 4.

Association of PUFAs with immune biomarkers

The associations between individual RBC PUFA levels, which derive from marine fats, and NIR, which derives from marine protein and inflammatory biomarkers, are shown in Table 3 and Supplemental Table 3. DHA was inversely associated with IL-6; EPA was inversely associated with TNFα; and the NIR was inversely associated with Monocyte chemoattractant protein-1 (MCP-1) and IL-10. All of these associations remained significant after Bonferroni correction. We observed several suggestive associations: DHA was inversely associated with Granulocyte-macrophage colony-stimulating factor (GM-CSF), TNFα, eotaxin-1, IL-8, Macrophage-derived chemokine (MDC), RANTES, IL-1RA, IL-10, and Epidermal growth factor (EGF); EPA was inversely associated with IL-10; the sum of EPA and DHA was inversely associated with IL-6, TNFα, MCP-1, MDC, RANTES, and IL-10; and the NIR was inversely associated with GM-CSF, IL-6, IL-15, TNFα, eotaxin-1, fractalkine, IL-8, MDC, and FGF-2. In comparison, AA was inversely associated with MDC and RANTES, and positively associated with MCP-1.

TABLE 3.

Association between immune biomarkers and PUFAs in Yup'ik participants¹

	PUFA variable	β (SE)	P value2
Proinflammatory Cytokines
GM-CSF	DHA	–0.0034 (0.0014)	0.019
GM-CSF	NIR	–0.6856 (0.2511)	0.007
IL-6	DHA	–0.0066 (0.0017)	<0.0012
IL-6	EPA + DHA	–0.0022 (0.0010)	0.023
IL-6	NIR	–0.7580 (0.2907)	0.010
IL-15	NIR	–0.6643 (0.2986)	0.028
TNFα	DHA	–0.0045 (0.0016)	0.005
TNFα	EPA	–0.4925 (0.1392)	<0.0012
TNFα	EPA + DHA	–0.0027 (0.0009)	0.003
TNFα	NIR	–0.7104 (0.2797)	0.012
Proinflammatory Chemokines
Eotaxin-1	DHA	–0.0071 (0.0024)	0.003
Eotaxin-1	NIR	–1.0782 (0.4268)	0.013
Fractalkine	NIR	–1.0558 (0.4381)	0.017
IL-8	DHA	–0.0056 (0.0018)	0.002
IL-8	NIR	–0.7381 (0.3255)	0.025
MCP-1	AA	0.0043 (0.0015)	0.004
MCP-1	EPA + DHA	–0.0017 (0.0008)	0.036
MCP-1	NIR	–0.8345 (0.2410)	<0.0012
MDC	DHA	–0.0073 (0.0023)	0.001
MDC	AA	–0.0051 (0.0024)	0.037
MDC	EPA + DHA	–0.0034 (0.0013)	0.010
MDC	NIR	–1.0710 (0.3935)	0.007
RANTES	DHA	–0.0048 (0.0017)	0.004
RANTES	AA	–0.0043 (0.0018)	0.017
RANTES	EPA + DHA	–0.0020 (0.0010)	0.041
Anti-Inflammatory Cytokines
IL-1RA	DHA	–0.0031 (0.0014)	0.028
IL-10	DHA	–0.0056 (0.0019)	0.003
IL-10	EPA	–0.4537 (0.1684)	0.008
IL-10	EPA + DHA	–0.0034 (0.0011)	0.002
IL-10	NIR	–1.2868 (0.3360)	<0.0012
Growth Factors
EGF	DHA	–0.0050 (0.0023)	0.029
FGF-2	NIR	–0.9633 (0.4397)	0.030

¹Twenty immune biomarkers in the Luminex panel were excluded from this analysis because >30% of the values were below the LLOQ (see Supplemental Table 1). All linear regression models included sex, age, smoking/tobacco use, physical activity, coastal compared with inland location, statin, blood pressure, and inflammatory medication as covariates. Quartiles 1 and 4 included 78 and 87 participants, respectively, corresponding to the lowest NIR and highest quartile of NIR measurement from the overall sample of 569 participants. Column 2 indicates the PUFA associated with a particular cytokine. AA, arachidonic acid; EGF, epidermal growth factor; FGF, fibroblast growth factor; GM-CSF, Granulocyte-macrophage colony-stimulating factor; LLOQ, lower limit of quantification; MCP, Monocyte chemoattractant protein-1; MDC, Macrophage-derived chemokine; NIR, nitrogen isotope ratio; RANTES, regulated upon activation, normal T-cell expressed and secreted.

² P values were statistically significant after Bonferroni adjustment for multiple testing [Bonferroni threshold for significance: 0.05/(3 × 16) = 0.00104]. The effective number of independent PUFA variables equaled 3, and the effective number of independent immune biomarkers determined by spectral decomposition (39) equaled 16. β values are slopes of the regression models.

In addition, we performed permutation tests, and the empirical P values are provided in Supplemental Table 3. Note that the empirical P values were lower than the regression P values, with some exceptions. For example, IL-8 and MDC associations with DHA became significant with Bonferroni correction in the permutation tests. However, the empirical P value of the association test between NIR and MCP-1 was no longer significant with the Bonferroni correction. In general, the empirical P values were lower than the regression P values.

Correlations of immune biomarkers and clinical measures

All of the adiposity variables (BMI, WC, and percentage body fat) were positively correlated with glucose, insulin, 2-hour glucose, HOMA2-B, and HOMA2-IR, and negatively correlated with HOMA2-DI (Figure 1). Some of the adiposity variables were also positively correlated with immune biomarkers, including IL-6, hsCRP, interferon γ-induced protein (IP)-10, MCP-1, MIP-1β, and IL-1RA, and negatively correlated with IL-15, soluble cluster of differentiation 40 ligand (sCD40L), eotaxin-1, MCP-3, IL-8, MDC, fractalkine, RANTES, IL-10, IL-4, EGF, FGF-2, and GM-CSF. Adiponectin was only negatively correlated with WC.

FIGURE 1.

Spearman correlation coefficients between PUFAs, diabetes, and cardiovascular risk factors, and inflammatory immune biomarkers in Yup'ik participants. The red and blue colors indicate the positive and negative correlation coefficient value intensities, respectively. *P < 0.05; **statistically significant after Bonferroni adjustment for multiple testing. Abbreviations for all the immune biomarkers are listed in Supplemental Table 1.

Similarly, 1 or more of the n-3 PUFAs (EPA, DHA, EPA + DHA, and NIR) were negatively correlated with proinflammatory cytokines (IL-15 and sCD40L) and proinflammatory chemokines (eotaxin-1, MCP-3, MDC, fractalkine, and RANTES); however, n-3 PUFAs were positively correlated with IP-10 and adiponectin.

Patterns of correlations between immune biomarkers and glucose parameters representing insulin resistance and a predisposition to T2D (glucose, insulin, HbA1c, 2-hour glucose, HOMA2-B, and HOMA2-IR) were generally weak and also inconsistent. Exceptions included IL-6, hsCRP, IP-10, and MCP-1, which positively correlated with 1 or more of the prediabetes parameters. Similarly, IL-15, MCP-3, MDC, fractalkine, IL-10, and IL-4 were negatively correlated with 1 or more prediabetes parameters.

Association between WC and parameters of glucose metabolism

WC was strongly associated with several glucose parameters, including fasting glucose, insulin, and all HOMA2 measures (Table 4). A suggestive positive association was observed between WC and HbA1c, but not 2-hour glucose. Similar associations were observed between BMI and glucose parameters (data not shown). A similar trend of the empirical P values was observed with the permutation test.

TABLE 4.

Association between parameters of glucose metabolism and waist circumference in Yup'ik participants

Glucose parameter (outcome)	Adiposity variable (predictor)	β (SE)	P value1	Empirical P values2
Glucose	WC	–0.02183 (0.0061)	<0.0011	<0.0011
Insulin	WC	0.0434 (0.0089)	<0.0011	<0.0011
HbA1c	WC	–0.01443 (0.0056)	0.011	0.018
2-hour glucose	WC	0.0012 (0.0031)	0.688	0.357
HOMA2-B	WC	0.0371 (0.0102)	<0.0011	<0.0011
HOMA2-S	WC	–0.0284 (0.0059)	<0.0011	<0.0011
HOMA2-IR	WC	0.0297 (0.0060)	<0.0011	<0.0011
HOMA2-DI	WC	–0.0354 (0.0078)	<0.0011	<0.0011

¹Association remained significant after multiple testing adjustment (Bonferroni threshold for significance: 0.05/5 = 0.01). The effective number of independent parameters of glucose metabolism, determined by spectral decomposition (39), equaled 5. All linear regression models included sex, age, smoking/tobacco use, physical activity, and coastal compared with inland location as covariates. HbA1c, glycated hemoglobin; HOMA2-B, homeostasis model assessment of beta cell function; HOMA2-DI, homeostasis model assessment of disposition index; HOMA2-S, homeostasis model assessment of insulin sensitivity; WC, waist circumference.

²Empirical P values of the permutation test. β values are slopes of the regression models.

³Transformations used to normalize glucose and HbA1c were decreasing functions, resulting in beta values with the opposite sign.

Association of WC with immune biomarkers

None of the immune biomarkers was significantly associated with WC after Bonferroni correction. Several suggestive positive associations were observed between WC and hsCRP and IP-10, and WC was inversely associated with adiponectin (Table 5). Similar associations were observed between BMI and immune biomarkers (data not shown). A similar trend of the empirical P values was observed with the permutation test, with the exception of the association between adiponectin and WC, which became significant with an empirical P value of 0.0027 after Bonferroni correction.

TABLE 5.

Association between immune biomarkers and waist circumference in Yup'ik participants

	Adiposity variable (predictor)	β (SE)	P value1	Empirical P values2
Proinflammatory Cytokines & Biomarkers (Outcome)
hsCRP	WC	0.0188 (0.0082)	0.0243	0.0093
GM-CSF	WC	–0.0043 (0.0081)	0.595	0.313
IL-6	WC	0.0085 (0.0106)	0.423	0.209
IL-15	WC	–0.0005 (0.0095)	0.955	0.478
TNFα	WC	0.0035 (0.0089)	0.695	0.361
sCD40L	WC	–0.0059 (0.0106)	0.577	0.295
Proinflammatory Chemokines
Eotaxin-1	WC	–0.0151 (0.0136)	0.270	0.158
Fractalkine	WC	–0.0059 (0.0140)	0.674	0.339
GRO	WC	0.0099 (0.0139)	0.476	0.237
IL-8	WC	–0.0077 (0.0104)	0.459	0.235
IP-10	WC	0.0285 (0.0118)	0.0173	0.0183
MCP-1	WC	0.0116 (0.0078)	0.141	0.084
MCP-3	WC	–0.0003 (0.0085)	0.971	0.496
MDC	WC	0.0001 (0.0126)	0.996	0.491
MIP-1β	WC	0.0086 (0.0067)	0.205	0.105
RANTES	WC	–0.0012 (0.0093)	0.898	0.454
Anti-Inflammatory Hormones & Cytokines
Adiponectin	WC	–0.0967 (0.0367)	0.0093	0.0031
IL-1RA	WC	0.0162 (0.0080)	0.0453	0.0193
IL-4	WC	–0.0050 (0.0122)	0.682	0.350
IL-10	WC	–0.0001 (0.0110)	0.992	0.491
Growth Factors
EGF	WC	–0.0140 (0.0127)	0.272	0.140
FGF-2	WC	–0.0272 (0.0140)	0.053	0.0323

¹No associations remained significant after Bonferroni multiple testing correction. Bonferroni threshold for significance: 0.05/16 = 0.003125. The effective number of independent immune biomarkers, determined by spectral decomposition (39), equaled 16. EGF, epidermal growth factor; FGF, fibroblast growth factor; GM-CSF, Granulocyte-macrophage colony-stimulating factor; GRO, growth related cytokine; hsCRP, high-sensitivity C-reactive protein; IP, inflammatory protein; MCP, Monocyte chemoattractant protein-1; MIP, macrophage inflammatory protein; RANTES, regulated upon activation, normal T-cell expressed and secreted; sCD40L, soluble cluster of differentiation 40 ligand; WC, waist circumference.

²Empirical P values of the permutation test. β values are slopes of the regression models.

³ P values < 0.05 but greater than the Bonferroni threshold for significance.

Association between immune biomarkers and parameters of glucose metabolism

None of the immune biomarkers were significantly associated with any glucose parameter after Bonferroni correction for multiple testing (Supplemental Table 4). However, suggestive positive associations were observed between glucose with fractalkine and MDC; HbA1c with IL-15, IL-4, and TNFα; and 2-hour glucose with MDC. Insulin and HOMA2-IR were positively associated with IP-10; HOMA2-S was inversely associated with IP10; and HOMA2-DI was positively associated with fractalkine and MDC in permutation testing.

Association between PUFAs and parameters of glucose metabolism

There were no significant associations between PUFAs and any of the glucose parameters, with or without Bonferroni correction (Supplemental Table 5).

Discussion

We report that the NIR (an objective blood biomarker of n-3 PUFA intake) and individually measured n-3 PUFAs (EPA and DHA) in RBC membranes were inversely associated with several proinflammatory cytokines and chemokines. Macrophage-activating CC-chemokines are associated with chronic inflammation (41). Since the NIR and DHA were inversely associated with several proinflammatory macrophage-activating CC-chemokines (MCP-1, MDC, RANTES, and eotaxin), our findings suggest that n-3 PUFAs may reduce monocyte recruitment and the accumulation of proinflammatory M1 macrophages and T cells in peripheral tissues (42–44). Individuals with high n-3 PUFA intakes also had correspondingly reduced AAs. Several different factors may explain this finding, including metabolic differences between individuals in quartile 1 compared with quartile 4 and differences in dietary patterns, and it is possible that high n-3 PUFA intakes may replace AAs in RBC membranes to further reduce inflammation by altering the pattern of eicosanoid production (45–47).

M1-activated macrophages, along with CD8 + T cells and CD4 + Th1 cells, secrete several proinflammatory cytokines, including TNFα, IL-6, IFN-γ, and IL-1β, resulting in systemic inflammation (42, 43). Decreased production of IL-6, TNFα, and IL-1β has been consistently observed following n-3 PUFA supplementation (48). We similarly observed significant inverse associations between EPA and TNFα and between DHA and IL-6. However, more than 30% of the samples assayed had IFN-γ and IL-1β below the LLOQ (Table 1); thus, these 2 cytokines were excluded from further analyses. Interestingly, we didn't observe an association between the n-3 PUFA intake and hsCRP, even though IL-6 was inversely associated with the n-3 PUFA intake and IL-6 regulates C-reactive protein synthesis in the liver. This may have resulted from our study population being only slightly overweight (Table 1), because we previously observed reductions in hsCRP in obese individuals consuming a diet rich in n-3 PUFAs (8).

Although IL-10 is considered an anti-inflammatory cytokine (49), the inverse association of NIR with IL-10 and weak inverse associations of EPA and DHA with IL-10 may be explained because n-3 PUFAs activate the β-arrestin 2 signaling pathway (9), which inhibits the release of NF-κB, a known activator of toll-like receptor-dependent IL-10 synthesis (50). Moreover, deacetylation of NF-kB by n-3 PUFAs downregulates its signaling, leading to suppression of downstream NF-kB-dependent gene expression (51). Additionally, activation of peroxisome proliferator–activated receptor-γ by n-3 PUFAs interferes with NF-κB translocation to the nucleus (48), further altering activation of cytokine gene expression. Our observation that several inflammatory cytokines and chemokines were inversely associated with the n-3 PUFA intake is consistent with the fact that they are known target genes of NF-κB (52, 53).

Despite our modest sample size and a study population that was only slightly overweight, many of the observed immune biomarker associations remained significant after Bonferroni correction for multiple testing, as well as multivariate adjustment for several potential confounders, including age, sex, smoking, physical activity, and coastal compared with inland location, which we have previously shown to be associated with n-3 PUFA intakes (54). The direction and magnitude of these associations were not modified by physical activity or smoking. Moreover, while we observed some significant associations specific to the NIR that were not significant for EPA and DHA (or vice versa), we frequently observed weak associations between EPA, DHA, or the NIR and a given immune biomarker with similar directionality. We have previously observed differences in chronic disease risk factor associations between the NIR and individual n-3 PUFAs (EPA and DHA), despite the fact that the NIR and n-3 PUFAs are highly correlated (16, 29). This may be due to the NIR capturing unmeasured nutrients associated with the lean and lipid portions of marine foods, whereas direct measures of EPA and DHA represent only the lipid component of the diet. Because the NIR is associated with the Yup'ik traditional dietary pattern, which contains a number of subsistence foods in addition to marine mammals and fish rich in n-3 PUFAs (55), the reduced inflammatory cytokines associated with the NIR may also be derived from unidentified food synergies (27) present in the subsistence diet. Because the NIR is a surrogate for n-3 PUFA intake over the past 3–4 months (35), it is possible that we are not capturing seasonal variations in n-3 PUFA intakes. However, previous studies measuring NIR in a segmental analysis of hair revealed that while the NIR in hair segments fluctuates seasonally in some individuals, the seasonal variations in traditional food intake were modest relative to the interindividual variation (56).

Since n-3 PUFAs were inversely associated with inflammation, and inflammation has been strongly linked to both insulin resistance and islet cell dysfunction (14), we hypothesized that n-3 PUFAs would be associated with improved glycemic parameters. While central adiposity was significantly associated with multiple adverse parameters of glucose metabolism, we only observed a small number of weak associations between WC and immune biomarkers, and we did not find any evidence that n-3 PUFA consumption was associated with parameters of glucose metabolism.

There are several strengths to this study, including the evaluation of n-3 PUFA consumption from whole foods rather than supplements, which allowed for the evaluation of food synergies (27) associated with habitual intake of marine mammals and fish, including micronutrients. Yup'ik people regularly consume a large and diverse array of fish, fish eggs, and marine mammals that provide a rich source of food-derived n-3 PUFAs (57–61). Moreover, we have reported on factors that characterize dietary patterns among Yup'ik people in our study communities (55, 60, 62). We were also able to select individuals at the extremes of n-3 PUFA intakes using a surrogate measure of n-3 PUFA intake (NIR), and then directly measure n-3 PUFAs in RBC membranes. Moreover, while the mean n-3 PUFA intake in this study population is more than 20 times higher than that in the US general population (15), the range in n-3 PUFA intakes is still large (16, 29). Our approach also allowed us an opportunity to compare associations between the lean (protein) component of foods rich in n-3 PUFAs (using the NIR) and the lipid component of these foods (using EPA, DHA, and AA measures) with a comprehensive array of cytokines, chemokines, adipokines, growth factors, and glucose parameters in the same population. Finally, we included information on important potential confounders (physical activity, smoking) in a population not taking n-3 PUFA supplements. While it is possible that other unmeasured dietary, behavioral, or lifestyle factors may have confounded our findings, an advantage to working in small Yup'ik communities is the homogenous nature of their surroundings: there are only 1–2 small markets per community, the variety of available market foods is extremely limited, and, for the most part, community members have similar socioeconomic statuses, educational levels, and housing within and between these small communities (see Supplemental Table 6). Limitations to this study include the lack of diet recall data, the modest sample size, limited sensitivity of the Luminex panel to detect low levels of circulating immune biomarkers, and our sample selection criteria based on the extremes of n-3 PUFA intakes without regard to obesity status, resulting in few obese participants. Although samples were initially stored at −20°C for up to a week because of our rural research locations, and might possibly lead to some oxidative degradation, it didn't affect the relationships between NIRs and n-3 PUFA levels in RBCs, which were very similar to those previously reported (29). Since the NIR measurement is not sensitive to alterations in molecular structure, this suggests that PUFA oxidation was minimal, random, and not likely to affect the relationships with other biomarkers that we report. Additionally, while male and female participants were age matched, participants in quartile 1 were significantly younger than participants in quartile 4. Finally, although the consistency of inverse associations across several inflammatory biomarkers is suggestive of causality, we were nevertheless limited in making causal conclusions by the cross-sectional study design.

In conclusion, in this population of Yup'ik Alaska Native adults consuming a diet rich in marine mammals and fish, a higher n-3 PUFA intake is associated with lower systemic inflammation in healthy adults, but not with better glycemic status, which was strongly influenced by body composition. Nevertheless, chronic low-grade inflammation is an important component of many different diseases and conditions (48). Our comprehensive evaluation of 41 pro- and anti-inflammatory cytokines, chemokines, adipokines, and growth factors expands the list of immune biomarkers potentially influenced by n-3 PUFAs. Additionally, evaluation of n-3 PUFA consumption using an objective biomarker (NIR) implies that the protein component of the traditional diet may contain important micronutrients not found in the lipid component of the diet, as previously observed (16). The results from this study may help to resolve some of the inconsistencies previously observed in randomized controlled trials conducted with pharmaceutical grade n-3 PUFAs or dietary supplements regarding the potential clinical benefit of consumption of foods rich in n-3 PUFAs to prevent inflammatory diseases (45, 48, 63). Future studies focused on the identification of n-3 PUFA–derived lipid mediators resulting from dietary patterns rich in marine oils may help to illuminate protective physiological mechanisms and lead to promising therapeutic dietary recommendations to prevent inflammatory diseases.

Notes

This study was funded by grants from the NIH (R01DK104347, R01DK112358, R01HL146549, P01GM116691, P20RR016430, P30GM103325, and UL1TR002369). The Endocrine Technologies Core at the Oregon National Primate Research Center is supported by NIH Grant P51OD011092.

Author disclosures: The authors report no conflicts of interest.

Supplemental Figure 1 and Supplemental Tables 1–6 are available from the “Supplementary data” link in the online posting of the article and from the same link in the online table of contents at https://academic.oup.com/jn/.

Abbreviations used: AA, arachidonic acid; EGF, epidermal growth factor; ETC, Endocrine Technologies Core; FGF, fibroblast growth factor; GM-CSF, Granulocyte-macrophage colony-stimulating factor; HOMA2-B, homeostasis model assessment of beta cell function; HOMA2-DI, homeostasis model assessment of disposition index; HOMA2-S, homeostasis model assessment of insulin sensitivity; hsCRP, high-sensitivity C-reactive protein; IP, interferon γ-induced protein; LAC, Lipoprotein Analytical Core; LLOQ, lower limit of quantification; MCP, Monocyte chemoattractant protein-1; MIP, macrophage inflammatory protein; NIDDK, National Institute of Diabetes and Digestive and Kidney Diseases; NIR, nitrogen isotope ratio; OHSU, Oregon Health and Science University; ONPRC, Oregon National Primate Research Center; RANTES, regulated upon activation, normal T-cell expressed and secreted; sCD40L, soluble cluster of differentiation 40 ligand; T2D, type 2 diabetes; UAF, University of Alaska Fairbanks; UCD, University of California Davis; VEGF, vascular endothelial growth factor; WC, waist circumference.

Data Availability

Data described in the manuscript, codebook, and analytic code will be made available upon request pending Tribal approval by the Yukon-Kuskokwim Health Corporation Executive Board.