Effect of industrially produced trans fat on markers of systemic inflammation: evidence from a randomized trial in women

Abstract

Consumption of industrially produced trans fatty acids (IP-TFA) has been positively associated with systemic markers of low-grade inflammation and endothelial dysfunction in cross-sectional studies, but results from intervention studies are inconclusive. Therefore, we conducted a 16 week double-blind parallel intervention study with the objective to examine the effect of IP-TFA intake on biomarkers of inflammation, oxidative stress, and endothelial dysfunction. Fifty-two healthy overweight postmenopausal women (49 completers) were randomly assigned to receive either partially hydrogenated soybean oil (15.7 g/day IP-TFA) or control oil without IP-TFA. After 16 weeks, IP-TFA intake increased baseline-adjusted serum tumor necrosis factor (TNF) α by 12% [95% confidence interval (CI): 5–20; P = 0.002] more in the IP-TFA group compared with controls. Plasma soluble TNF receptors 1 and 2 were also increased by IP-TFA [155 pg/ml (CI: 63–247); P < 0.001 and 480 pg/ml (CI: 72–887); P = 0.02, respectively]. Serum C-reactive protein, interleukin (IL) 6 and adiponectin and subcutaneous abdominal adipose tissue mRNA expression of IL6, IL8, TNFα, and adiponectin as well as ceramide content were not affected by IP-TFA, nor was urinary 8-iso-prostaglandin-F_2α. In conclusion, this dietary trial indicates that the mechanisms linking dietary IP-TFA to cardiovascular disease may involve activation of the TNFα system.

In cross-sectional studies, intake of industrially produced trans fatty acids (IP-TFA) is positively associated with systemic concentrations of various inflammatory markers such as C-reactive protein (CRP), tumor necrosis factor (TNF) α, and interleukin (IL) 6 (1–3). Recently, similar associations with inflammatory markers were found with intake of partially hydrogenated vegetable oil, which is the primary source of IP-TFA (4). Consequently, induction of low-grade systemic inflammation has been suggested as a mechanism by which IP-TFA intake increases the risk of cardiovascular disease (CVD) and type 2 diabetes as seen in large prospective cohort studies (5, 6). However, the evidence from randomized studies is limited; one study has shown increased production of TNFα and IL6 in stimulated mononuclear cells from humans after high IP-TFA consumption (7) and two studies found increased CRP in blood (8, 9), whereas several others have failed to find an effect of dietary IP-TFA on inflammatory markers (10–12).

IP-TFA intake has also been positively associated with markers of endothelial dysfunction, such as E-selectin and adhesion molecules, in a cross-sectional analysis of the Nurses’ Health Study (3), and the causality of this finding was strengthened by a randomized trial showing increased concentrations of plasma soluble E-selectin (sE-selectin) after IP-TFA consumption (8). In concordance with this, a study suggested that IP-TFA intake decreases flow-mediated vasodilation, which is a more direct measure of endothelial function (13), although this finding failed to be confirmed in our study (14).

Furthermore, IP-TFA intake has been suggested to increase oxidative stress, which might accelerate diseases such as CVD and type 2 diabetes (15). An increase in the urinary concentration of 8-iso-prostaglandin-F_2α (8-iso-PGF_2α), a marker of of oxidative stress in vivo (16), has previously been documented after high intake of conjugated linoleic acid (CLA) and the TFA isomers trans18:1n-7 and trans18:1n-6 (17–19).

In this dietary intervention study, we examined the effect of a high intake of IP-TFA on systemic and adipose biomarkers of inflammation. Also, we investigated a possible effect on markers of endothelial dysfunction and oxidative stress.

Subjects and Methods

This work presents data from a dietary intervention study examining the effect of a high intake of IP-TFA on a range of risk markers for CVD and type 2 diabetes. Here, we report the results for systemic markers of low-grade inflammation and endothelial dysfunction, urinary markers of oxidative stress, as well as adipokine expression and ceramide content in subcutaneous abdominal adipose tissue (AT). Markers of inflammation and endothelial dysfunction were selected based on findings from prior observational reports (1–3).

The details of this 16 week, double-blind, parallel dietary intervention study have been reported previously (20). In brief, 52 women were randomly assigned to two test diets by using a computer-generated randomization sequence. Subjects in the IP-TFA group were given 26 g per day of partially hydrogenated soybean oil (providing 15.7 g/d IP-TFA) incorporated into two bread rolls (2,500 kJ), whereas subjects in the control (CTR) group received bread rolls with a low TFA oil consisting of a 50/50% mixture of palm oil and high oleic sunflower oil. The test fats were kindly supplied by AarhusKarlshamn, Denmark. The FA composition of the two test fats mainly differed in the content of TFA, palmitic (16:0), oleic (cis18:1), and linoleic acids (cis18:2) as shown in Table 1. Owing to the ban on IP-TFA in food products sold in Denmark, the content of IP-TFA in the background diet was negligible. The data were collected at weeks 0 (baseline), 8, and 16.

TABLE 1.

Fatty acid composition of the trans fat (IP-TFA) and the control fat (CTR)

Fatty acid	CTR (w/w%)	IP-TFA (w/w%)
C14:0	0.7	0.2
C16:0	26.7	12.7
C18:0	3.8	6.2
C18:1-trans	< 0.7	59.0
C18:1-cis	61.4	19.6
C18:2-trans, trans	< 0.1	1.4
C18:2-cis, cis	6.6	0.2
C20:0	0.4	0.4
C22:0	0.4	0.3
Total trans fatty acids	< 0.7	60.4

Subject were healthy, moderately overweight [body mass index (BMI) 25–32 kg/m²], waist circumference > 80 cm, postmenopausal for the last year, and aged 45–70 years. Criteria for exclusion were diabetes or other chronic diseases, history of CVD, smoking, blood pressure >160/100 mmHg, fasting plasma triglycerides >3 mmol/L, fasting plasma LDL-cholesterol >6 mmol/L, fasting plasma glucose >7 mmol/L, use of hormones or antihypertensive, antilipidemic, or anticholesterolemic drugs, recent weight changes, strenuous physical activity, and abnormalities in routine biochemical and hematological tests.

To assess the effect of body weight on the inflammatory markers, the study included 19 lean reference women (BMI = 19–24 kg/m², waist circumference ≤ 80 cm) who underwent baseline examinations only. These reference subjects fulfilled the same inclusion and exclusion criteria as the intervention subjects apart from the limits for BMI and waist.

After having received both verbal and written information, all participants gave written consent. The study was approved by the Municipal Ethical Committee of the Capital Region of Denmark in accordance with the Helsinki-II declaration (H-B_2007-089) and was registered at clinicaltrials.gov as NCT00655902.

Sample collection and analysis

At all visits, subjects fasted for 10 h. After minimum 10 min of rest, blood was sampled from an antecubital vein into Vacutainer lithium heparin gel tubes and silicone coated serum tubes (BD Medical Systems, Franklin Lakes, NJ). Serum was obtained after 30 min clotting at room temperature and 10 min centrifugation at 3,000 g. Samples were stored at −80°C before analysis.

CRP was analyzed on a Vitros 5.1 (# 6801739, Johnson and Johnson, Rochester, NY) with an interassay variation coefficient (CV) of 5.0% and a detection limit (DL) of 0.1 mg/L (minimum detectable concentration + 2 × standard deviation). Samples (2%) that were below the CRP DL were defined as 0.05 mg/L. A total of seven CRP values (four at baseline, two at week 8, and one among the reference subjects) were excluded due to CRP concentrations >10 mg/L, indicating acute inflammation. For all other inflammatory markers, these samples were not excluded because exclusion had negligible effects on the results. Serum adiponectin (CV: 7.3%) was analyzed by a human ELISA kit (B-Bridge International, California). Serum TNFα (CV: 4.7%, DL: 0.11 pg/ml) and IL-6 (CV: 9.8%, DL: 0.039 pg/ml) concentrations were also measured by using human ELISA kits (Quantikine HS #HSTA00D and #HS600B, ELISA, R and D Systems, Abingdon, UK).

Concentrations in plasma and serum of soluble TNF receptor 1 (sTNF-R1) (CV: 5.3%, DL: 60 pg/ml) and soluble TNF receptor 2 (sTNF-R2) (CV: 7.6%, DL: 79 pg/ml) were measured by 2-plex immunoassays using fluorescently labeled microsphere beads kits (#HSCR-32K, Millipore, Billerica, MA) and analysis on BioPlex 200® (Biorad, Hercules, CA) as well as by ELISA (#DRT100 and DRT200, R&D Systems) (CV: 2.8%, DL: 1.14pg/ml: and CV: 2.4%, DL: 0.78 pg/ml, respectively). A 3-plex kit (#HCVDI-67AK, Millipore) was used for measurement in plasma and serum of soluble intercellular adhesion molecule (sE-selectin) (CV: 11.8%, DL: 2.1 ng/ml), soluble vascular cell adhesion molecule 1 (sVCAM-1) (CV: 4.1%, DL: 1.1 ng/ml), and soluble intercellular adhesion molecule 1 (sICAM-1) CV: 7.6%, DL: 1.8 ng/ml). TNFα, IL6, and adiponectin concentrations were measured at baseline and week 16. All other blood parameters were additionally measured at week 8.

The concentration of free 8-iso-PGF_2α was analyzed in 24 h urine samples with the use of a highly specific and sensitive radioimmunoassay as previously described (21) adjusted for creatinine value measured with a commercial kit (IL Test; Monarch Instrument, Amherst, NH).

A subcutaneous AT biopsy was taken from the abdominal region under local anesthesia (1% lidocaine) on the left or right side of the abdomen about 5 cm lateral from the umbilicus using a Hepafix Luer lock syringe (Braun Medical, Bethlehem, PA) and a 2.10 × 80 mm Braun Medical Sterican needle. The biopsy (∼500 mg) was washed in physiological saline, divided into three sterile cryo tubes, immediately frozen in liquid nitrogen, and stored at –80°C until analysis.

AT (200 mg) was homogenized in Trizol reagent (Gibco BRL, Life Technologies, Roskilde, Denmark) and total RNA was extracted following the manufacturer's protocol. RNA was quantified by measuring absorbance at 260 and 280 nm using a NanoDrop 8000 (NanoDrop products, Bancroft, DE), and there was a ratio ≥1.8. The integrity of the RNA was checked by visual inspection of the two rRNAs, 18S and 28S, on an agarose gel.

For real-time RT-PCR, cDNA was constructed using random hexamer primers as described by the manufacturer (Verso cDNA kit, Abgene, Epsom, UK). Then KAPA SYBR FAST qPCR mastermix (Kapa Biosystems, Inc. Woburn, MA) and the following primer pairs were added: IL6: AAATGCCAGCCTGCTGACGAAG and AACAACAATCTGAGGTGCCCATGCTAC, TNFα: CGA­GTGACAAGCCTGTAGC and GGTGTGGGTGAGGAGCACAT, Adiponectin: CATGACCAGGAAACCACGACT and TGAATGCTGAGCGGTAT, IL8: TTGGCAGCCTTCCTGATTTC and AACTTCTCCACAACCCTCT-G, β-actin: TGTGCCCATCTACGAGGGGTATGC and GGTACATGGTGGTGCCGCCAGACA. Real-time quantification of each gene in separate tubes was performed using an ICycler from Bio-Rad. The threshold cycle was calculated and the relative gene expression was calculated essentially as described in the User Bulletin no. 2, 1997, from Perkin-Elmer. All samples were amplified in duplicate. A similar set-up was used for negative controls except that the reverse transcriptase was omitted and no PCR products were detected under these conditions.

AT ceramide content was analyzed basically as has been described (22). In brief, AT was extracted using the Folch procedure and ceramide with C17 long-chained base (LCB) was added during the extraction procedure. Glycerolipids were hydrolyzed using mild alkaline hydrolysis and ceramide was isolated from other sphingolipids using amino-propyl cartridges. The isolated ceramide was hydrolyzed to LCB and free FA and the LCB was derivatized with o-phtalaldehyde and analyzed using a high-performance liquid chromatography-system equipped with an amino-propyl column. Ceramide content was determined from fluorescence intensity of the endogenous LCB-peaks in relation to the internal standard.

For assessment of AT triglyceride FA composition, samples were spiked with 17:1n7 triglyceride (Nuchek Prep, Elisian MN) extraction surrogate and lipids were extracted with cyclohexane/isopropanol/water (23). An extract aliquot containing 50–100 μg of total lipid were exchanged into chloroform. Triacylglycerides were then isolated with aminopropyl solid phase extraction columns using modification of published procedures (24). Samples were transesterified in methanolic sodium hydroxide and isolated FA methyl esters were analyzed by gas chromatography on Agilent 6890 GCs with split/splitless inlets run in splitless mode with a 0.8min purge delay. AT triglyceride FA methyl esters were separated on a 100 m × 0.25 mm id × 0.2 μm SP-2380 capillary column (Varian, Inc., Walnut Creek, CA). The oven temperature gradient was as follows: 1 min at 80°C; 35°C/min to 150°C; 25 min hold; 2°C/min to 205°C; 1°C/min to 211°C; 0.7°C/min to 219°C; 0.4°C/min to 223°C; 32 min hold; 35°C/min to 80°C; 2 min hold (total run time 120 min). The column helium flow gradient was as follows: initial flow 1.2 ml/min, held 4 min; 1 ml/min² to 0.8 L/min, held 60 min: 1 ml/min² to 1.2 ml/min, held 0 min (20). All reported results were corrected for 17:1n7 recoveries.

Statistical analyses

The study size was estimated based on the effect of IP-TFA on the LDL- to HDL-cholesterol ratio as reported previously (20).

Data were analyzed using Statistical Analysis Package, SAS ^© version 9.1 (SAS Institute, Cary, NC). The analysis included those participants who completed the intervention (n = 49). The statistical significance level is defined as P < 0.05.

Baseline values for the overweight intervention subjects were compared with values for the lean reference subjects by 2-tailed unpaired t-tests or Kruskal-Wallis’ χ² tests for skewed data.

Analysis of covariance (ANCOVA) was used to assess the baseline-adjusted difference between diet groups for variables measured at weeks 0 and 16; i.e., the baseline value was included as a covariate. For variables measured at weeks 0, 8, and 16, a mixed model of repeated measures examining the effect of diet and time (weeks 8 and 16) and their interactions was applied again with the baseline value as a covariate and with “subject” as a random effect. Variance homogeneity and normality were investigated by residual plots, histograms, and Shapiro-Wilk's test and data was log transformed when needed. Changes within groups from baseline to week 16 were evaluated by 2-tailed paired t-tests or Signed Rank tests for skewed data.

Results

Baseline characteristics of the 24 completers in the IP-TFA group were comparable to those of the 25 completers in the CTR group and lean references were adequately matched for age and height (Table 2).

TABLE 2.

Baseline characteristics for subjects in the trans fat (IP-TFA; n = 24) and control (CTR, n = 25) groups and for lean references (lean ref.; n = 19)

	IP-TFA	CTR	Lean ref.
Age (y)	58.5 ± 4.6	58.8 ± 5.5	60.1 ± 5.9
Height (cm)	165.3 ± 5.5	166.9 ± 5.2	166.7 ± 4.3
Weight (kg)	78.7 ± 7.1	78.4 ± 8.6	59.4 ± 4.8 a
BMI (kg/m )	28.8 ± 1.7	28.1 ± 2.2	21.3 ± 1.4 a
Waist circumfer­­ence (cm)	97.1 ± 7.3	95.5 ± 6.8	74.9 ± 3.8 a
Urine 8-iso-PGF2α (nmol/mmol creatinine)	0.28 (0.24; 0.32)	0.31 (0.24; 0.43)	0.35 (0.28; 0.41)

Values are means ± SD or geometric means (95%CI) for skewed data, completers only. BMI, body mass index; 8-iso-PGF_2α, 8-iso-prostaglandin-F_2α.

^aSignificantly different from overweight intervention subjects (diet groups combined) by 2-tailed unpaired t-tests (P < 0.05).

Systemic markers of low-grade inflammation

The serum TNFα concentration increased by 10 ± 3% in the IP-TFA group during the intervention (P = 0.02 for within-group change) and decreased by 2 ± 2% in the CTR group [nonsignificant (ns)], whereby the mean baseline-adjusted TNFα concentration was 12% [95% confidence interval (CI): 5–20] higher at week 16 compared with controls (P = 0.002; Fig. 1A).

Fig. 1.

Systemic concentrations of markers of inflammation and endothelial dysfunction before and after 8 and 16 weeks (wk) of supplementation with 15.7 g/d industrially produced trans fatty acids (IP-TFA; dark gray bars; n = 24) or a control oil (CTR; white bars; n = 25), and in lean references (light gray bars; n = 19). A: Tumor necrosis factor (TNF) α assessed by ELISA and soluble tumor necrosis factor receptors 1 and 2 (sTNF-R1 and sTNF-R2) assessed by bead immunoassay. B: C-reactive protein (CRP; n = 23, 22, and 18 in the IP-TFA and CTR group and lean references, respectively, due to exclusion of CRP values >10 mg/l), interleukin 6 (IL6) and adiponectin assessed by ELISA. C: Soluble E-selectin (sE-selectin), soluble intercellular adhesion molecule 1 (sICAM-1), and soluble vascular adhesion molecule 1 (sVCAM-1) assessed by bead immunoassay. Bars represent means [95% confidence intervals (CI)] for TNF-R1 and TNF-R2 and geometric means (CI) for all other variables. P-values are for effect of diet, by baseline-adjusted ANCOVA for variables measured at wks 0 and 16, and by baseline-adjusted repeated measures ANCOVA for variables measured at wks 0, 8, and 16. There were no time-by-diet interactions and no effect of time in analyses omitting the interaction term. *Significantly different from overweight intervention subjects (diet groups combined); P < 0.05 by 2-tailed t-tests or Kruskal-Wallis’ χ² test.

The plasma concentration of sTNF-R1, as assessed by bead immunoassay, increased by 22 ± 7% in the IP-TFA group (P = 0.0003 for within-group change by paired t-test) versus a decrease of 3 ± 6% (ns) in the CTR group during the 16 week intervention, resulting in a baseline adjusted mean difference between diets of 155 pg/ml (CI: 63–247) (P < 0.001). The bead immunoassay assessed plasma concentration of sTNF-R2 also increased more on the IP-TFA diet: 14 ± 3% (P = 0.0007) versus 2 ± 4% (ns) in the CTR group [mean baseline adjusted difference between diet groups: 480 pg/ml (CI: 72–887); P = 0.022].

As no randomized study previously has documented this effect of dietary IP-TFA on soluble TNF receptor concentrations in plasma, we decided to further explore these findings by repeating the analyses using antibodies from a different provider and another platform (ELISA, see Subjects and Methods). For sTNF-R2, we obtained a similar difference between diets with ELISA, suggesting a robust difference [257 pg/ml (CI: 93–421); P = 0.003; note that concentrations were generally lower with ELISA due to different calibrations]. For sTNF-R1, there was a trend in the ELISA-assessed data similar to the that observed with bead immunoassay but the difference between diets did not remain significant [56 pg/ml (CI: −14–125); P = 0.11]. These differences between diet groups in plasma sTNF receptor concentrations were not present when the same analyses were carried out in serum (Supplementary Fig. I).

Neither serum CRP, adiponectin, nor IL6 concentrations were affected by the dietary intervention (Fig. 1B).

Systemic markers of endothelial dysfunction

Serum sE-selectin increased by 19 ± 4% in the IP-TFA group during the intervention (P = 0.004). However, the baseline-adjusted relative difference between diets at week 16 was not significant: [8% (CI: −2–19; P = 0.13; Fig. 1C)], because the serum sE-selectin concentration also increased in the CTR group (12 ± 5%; P = 0.04). In plasma, the difference between diets in sE-selectin reached statistical significance [10% (CI: 1–21); P = 0.04; data not shown]. Serum sVCAM-1 and sICAM-1 concentrations were not affected by the intervention and neither were plasma concentrations.

Oxidative stress

During the 16 week intervention the concentration of the urinary marker of oxidative stress 8-iso-PGF_2α changed little in either diet group [−7 ± 11% and −3 ± 13% in the IP-TFA and CTR groups (ns), respectively] with no significant baseline-adjusted difference between diet groups (P = 0.50).

Fatty acid composition, adipokine transcripts, and ceramide content in adipose tissue

In the IP-TFA group there was an almost 2-fold increase in the AT content of the sum of the trans18:1n-9 and trans18:1n-7 isomers (from 0.47 ± 0.03% to 0.97 ± 0.06% of FA), which was not seen in the CTR group (P < 0.01). There were no significant differences between diet groups in the sum of saturated FA, MUFA, or PUFA, respectively, nor were there differences in AT content of any individual FA (Supplementary Table I).

The IP-TFA intervention had no effect on the mRNA expression of IL6, IL8, TNFα, or adiponectin in subcutaneous abdominal AT (Fig. 2). The expression of IL-6 was more than 2-fold higher among lean women compared with the overweight intervention subjects at baseline (P = 0.001).

Fig. 2.

Subcutaneous abdominal adipose tissue mRNA expression before and after 16 weeks (wk) of supplementation with 15.7 g/d industrially produced trans fatty acids (IP-TFA; dark gray bars; n = 24; n = 23 for IL8) or a control oil (CTR; white bars; n = 25), and in lean references (light gray bars; n = 17). mRNA expressions are expressed relative to β-actin and bars represent geometric means (95% confidence intervals). There were no differences between diet groups at wk 16 by ANCOVA with baseline value as a covariate (P > 0.20). * Significantly different from overweight intervention subjects (diet groups combined); P < 0.001 by Kruskal-Wallis’ χ² test. IL, interleukin 6; TNF, tumor necrosis factor.

The transcripts of IL6, TNFα, and adiponectin in AT were not significantly correlated to the respective serum concentrations at baseline in the overweight intervention subjects nor were the changes in AT and serum over the course of the intervention correlated (P > 0.20 for all).

Baseline AT ceramide content was lower in lean references (176 ± 7 pmol/mg tissue protein) compared with overweight subjects (320 ± 7 pmol/mg tissue protein; P < 0.001). AT ceramide tended to decrease in the IP-TFA group (by 8 ± 5%; P = 0.09 for within-group change) but the change was not significantly different from the 1 ± 4% decrease in the CTR group (P = 0.23 by baseline-adjusted ANCOVA).

Discussion

In addition to the previously described adverse effects of IP-TFA on plasma cholesterol concentrations (20), we found that high IP-TFA intakes [∼7% of energy (E%)] increase the TNF system activity as shown by elevated circulating concentrations of TNFα and its soluble receptors TNF-R1 and TNF-R2. Therefore, the results of our study support the notion that the findings from previous cross-sectional analyses (1–4, 25) are causal and suggest that part of the increased incidence of heart disease related to dietary IP-TFA may be due to induction of low-grade systemic inflammation. Our study cannot by itself rule out the possibility that the change in TNFα was caused by the lower intake level of oleic acid in the IP-TFA group compared with the control group (20 vs. 61 w/w%). However, even when adjusting for intake of MUFA, observational data point toward a positive association between IP-TFA and TNFα (3).

A high dose of 15.7 g/d IP-TFA was chosen in this study to provoke effects within the relatively short time frame of the study (16 weeks). We find it likely that a lower IP-TFA consumption for years may provoke a similar reaction as indicated by observational data. In recent years, the access to IP-TFA has decreased considerable in many Western countries owing to legislative actions and/or voluntary reformulations by industry. However, looking at the distribution of intakes in the population, it is clear that a subset still have intakes that makes our dosage relevant to study, which is supported by the finding that some servings of popular foods may contain more than 20 g of IP-TFA, for instance, in eastern Europe (26). Also, populations living in developing countries may still be exposed to high-IP-TFA foods because partially hydrogenated vegetable oils represent inexpensive and stable sources of dietary fat (27). Examples are Iran, where the mean IP-TFA intake in 2001–2003 was estimated to be ∼12.3 g/d (4.3 E%) (28) and India, where the intake of vanaspati, a hydrogenated vegetable oil high in TFA, is considerable in some regions (29).

We are the first to report increased systemic concentrations of TNFα after IP-TFA supplementation and we substantiate our findings by showing concomitant increases in the soluble TNF receptors. TNFα, as well as other cytokines, is known to induce shedding of the soluble receptors, and these may thereby be biomarkers of local TNFα activation or mirror overall systemic inflammation (30). The soluble receptors affect the local and systemic availability of TNFα and appear to act as a buffer system, either prolonging or attenuating the biological effects of TNFα (30).

Soluble TNF receptors were positively associated with coronary heart disease risk in the Nurses’ Health Study but not after adjustment for lipid and other cardiovascular risk factors (31). When we adjusted for the IP-TFA-induced 34% increase in the LDL-cholesterol to HDL-cholesterol ratio [as reported previously (20)], the difference between diets in TNFα and sTNF-R1 became more pronounced, suggesting that the effect of IP-TFA was not mediated through blood lipid changes. The effect of diet for sTNF-R2 was attenuated slightly by adjustment for lipid concentrations and the same results were obtained when we adjusted for HDL-cholesterol only. Also, the differences between diets in TNFα, sTNF-R1, and sTNF-R2 were not attenuated by adjustment for changes in body fat mass.

It is difficult to predict what the trans fat-induced increases in TNFα, sTNF-R1, and sTNF-R2 (of ∼10%, 22%, and 14%, respectively) may translate into in terms of risk of CVD. TNFα has a limited half-life and is only rarely measured in large-scale epidemiologic studies. In women, an increase in sTNF-R1 or sTNF-R2 of ∼60% has been shown to be associated with an 2.5-fold increase in the risk of coronary heart disease (31). Also, TNFα and TNF-receptor concentrations are elevated in heart failure patients, as reviewed by Bozkurt et al. (32), and high systemic TNFα concentrations were shown to increase the risk of recurrent coronary events after a myocardial infarction (33).

It should be noted that many variables have been examined in this work. If applying the Bonferroni correction, which is a conservative post hoc method for avoiding the phenomenon of mass significance, we would need a p-value of < 0.003 to claim significance, while higher p-values should be regarded as tendencies only. This means that the effect of IP-TFA on TNFα and sTNF-R1 remains significant after correction for multiple comparisons whereas that on sTNF-R2 does not.

Others have examined the effect of IP-TFA consumption on TNFα with discordant results. Han et al. (7) showed increased production of TNFα (and IL6) in stimulated mononuclear cells of mildly hypercholesterolemic, older, and overweight subjects after 32 days of 7 E% IP-TFA intake from stick margarine compared with intake of soybean oil. In contrast, in two studies in young, lean, and healthy subjects, supplementation with a mix of trans18:1n-7 and trans18:1n-6 (∼3 E%) for 6 weeks had no effect on plasma concentrations of TNFα, IL6, IL8, or adiponectin (10), and 5 weeks’ consumption of 10 E% IP-TFA from hydrogenated soybean oil did not affect the serum TNFα (or IL6) concentration when compared with palm stearin or high oleic palm olein (9). It is possible that the TNF system is more susceptible to IP-TFA-induced activation in older and overweight subjects.

The effect of the IP-TFA-rich diet on sTNF-Rs was visible in heparinized plasma but not in serum. Hypothetically, the incorporation of TFA into cell membranes and ensuing change in cellular lipid rafts may have imposed con­formational changes on the extracellular domains of membrane-bound TNF receptors (34), which may have persisted after shedding from the membrane. The coagulation cascade occurring during serum formation activates several proteases that may alter the sTNF-R molecules. A difference in configuration between sTNF-Rs exposed to TFA or control fat could affect the extent to which the molecules are modified and thereby their recognition by specific antibodies.

Wong et al. (35) described great differences in cytokines between serum and plasma and hypothesized that this might be due to i) degradation of the cytokines during the clotting process, ii) ex vivo degranulation of granulocytes and platelets, and/or iii) nonspecific interference related to the protein matrix. For sTNF-R, we observed a difference between concentrations in plasma and serum but more importantly, we found that the sTNF-R response to IP-TFA exposure is detectable in plasma but not in serum. Whatever the explanation, our observations suggest that plasma is the material of choice for studying the effect of dietary fats on sTNF-R. Notably, cross-sectional associations between IP-TFA intake and sTNF-Rs were detected in studies analyzing plasma (1–3) but not in one using serum samples (25).

Previous intervention studies examining the effect of IP-TFA on systemic CRP concentrations are limited and contradictory. In studies providing healthy subjects with 8 or 10 E% IP-TFA, the concentration of CRP increased compared with provision of carbohydrate and oleic acid (8) or compared with oleic acid/PUFA or palmitic/oleic acid (9). In contrast, CRP was unaffected by dietary IP-TFA in a study comparing diets with four types of margarine (up to 5.2 E% TFA), butter, or soybean oil (11), in a study comparing intake of partially hydrogenated soy bean oil (4 E% TFA) to corn oil (12), and in a study comparing stick margarine intake (3.6 E% TFA) with a butter/canola oil mix (36). It is possible that most studies have simply been underpowered to detect the small changes induced by IP-TFA; the difference between diets in CRP in the two crossover studies showing significant effects of IP-TFA was only 0.12 mg/l (n = 41) (9) and 0.20 mg/l (n = 50) (8), respectively.

We did not see a clear pattern for the effect of IP-TFA on systemic markers of endothelial dysfunction. Serum sE-selectin increased in the IP-TFA group but not significantly more than in controls, which may be due to lack of power. In contrast, sVCAM-1 and sICAM-1 concentrations remained unchanged. An increase in sE-selectin after 5 weeks of 8 E% IP-TFA consumption (compared with carbohydrate, stearic acid, oleic acid, and a mix of lauric, myristitic, and palmitic acid) has been reported in a previous trial (8).

Others have documented that urinary 8-iso-PGF_2α, a validated marker of oxidative stress in vivo, is increased by intake of CLA and the TFA isomers trans18:1n-7 and trans18:1n-6 (17–19). However, a mixture of IP-TFA isomers, as provided to the subjects in the present study, had no detectable effect on the urinary concentration of 8-iso-PGF_2α and we hypothesize that the increase in oxidative stress markers previously seen with trans18:1n-7 consumption may be ascribed to endogenous conversion to CLA (18). However, in order to be able to make firm conclusions about the effect of IP-TFA on oxidative stress, more markers should have been assessed.

The dietary intervention did not affect mRNA expression of inflammatory markers and ceramide content in subcutaneous AT. This may be due to limited incorporation of TFA into AT during the 16 weeks of intervention; the AT content of the two predominant TFA isomers doubled in the IP-TFA group but still only constituted ∼1% of total FA. The low degree of TFA incorporation is not surprising given the half-life of subcutaneous AT FA of 6–18 months (37, 38). However, as the visceral depot FA turnover is higher than that of the subcutaneous (39) and because meal FA uptake is relatively larger in the former depot (40), more TFA may have been incorporated here. Previously, the TFA concentration was shown to be higher in visceral compared with subcutaneous AT (41). Also, the subcutaneous AT depot may not adequately reflect an effect of IP-TFA on AT inflammation; macrophage infiltration is higher in visceral than in subcutaneous AT (42) and in obese subjects, the mRNA of TNFα was found to be considerably higher in visceral compared with subcutaneous AT (43, 44).

The facts that cytokines are also produced by other tissues/cells, e.g., skeletal muscle (45), and that transcript and protein turnover is different would also explain why mRNA expression did not correlate with serum levels.

In conclusion, our research indicates that the IP-TFA-associated increase in cardiovascular risk beyond the adverse effect explained by changes in blood lipids may be partly due to induction of systemic low-grade inflammation.