Serum fatty acid (FA) composition is a consequence of dietary fat(Reference Ma, Folsom, Shahar and Eckfeldt1, Reference Zock, Mensink, Harryvan, de Vries and Katan2) but also of endogenous FA desaturation(Reference Vessby, Gustafsson, Tengblad, Boberg and Andersson3). Desaturation is catalysed by enzymes introducing a double bond in the FA chains. Stearoyl-CoA desaturase-1 (SCD-1) synthesize MUFA from SFA, whereas Δ5- and Δ6-desaturases catalyse the synthesis of highly unsaturated FA(Reference Vessby, Gustafsson, Tengblad, Boberg and Andersson3).
FA composition in cholesteryl esters (CE) has been associated with insulin resistance(Reference Vessby, Tengblad and Lithell4) and CVD(Reference Sundstrom, Lind, Vessby, Andren, Aro and Lithell5, Reference Ohrvall, Berglund, Salminen, Lithell, Aro and Vessby6). This FA pattern is often characterized by increased proportion of SFA and 16 : 1n-7, decreased proportion of 18 : 2n-6 and indications of increased activity of SCD-1 and Δ6-desaturase and reduced activity of Δ5-desaturase(Reference Vessby, Tengblad and Lithell4, Reference Ohrvall, Berglund, Salminen, Lithell, Aro and Vessby6, Reference Warensjo, Riserus and Vessby7). A potential mediating role of low-grade inflammation in the association between FA composition and insulin resistance is still unknown.
Low-grade inflammation as assessed by C-reactive protein (CRP) is related to insulin resistance(Reference Haffner8) and CVD(Reference Blake and Ridker9). A potential anti-inflammatory effect of n-3 FA has previously been indicated(Reference Calder10–Reference Ferrucci, Cherubini, Bandinelli, Bartali, Corsi, Lauretani, Martin, Andres-Lacueva, Senin and Guralnik13). No larger studies have, however, investigated the long-term relationship between overall FA composition and CRP. Furthermore, no studies have examined the role of FA desaturases on inflammation, despite high SCD-1 activity index being linked to obesity and insulin resistance(Reference Vessby, Gustafsson, Tengblad, Boberg and Andersson3, Reference Warensjo, Riserus and Vessby7). We investigated the longitudinal association between dietary fat quality as assessed by CE FA composition at age 50, and CRP concentrations 20 years later in 767 healthy men, also taking several important confounders into account, including insulin resistance measured by a gold standard technique.
Subjects participated in the Swedish cohort, Uppsala Longitudinal Study of Adult Men (http : //www.pubcare.uu.se/ULSAM/). Subjects (n 1020) had measures of serum CE FA composition at age 50 and CRP at age 70. Exclusion criteria were CRP concentration >10 mg/l, diabetes (fasting blood glucose ≥ 6·1 mmol/l), CVD (ICD-8 codes 401–443) or malignancy at baseline, and usage of lipid-lowering medicine or glucocorticoids at age 50 or 70. After exclusions, the population consisted of 767 participants. The study was approved by the Ethics Committee of Uppsala University. All subjects gave written informed consent.
Investigations at 50 years
All measurements were performed under standardized conditions as previously described(Reference Skarfors, Selinus and Lithell14). All blood samples were drawn after an overnight fast. Blood glucose was measured using the glucose oxidase method. Serum insulin was determined with the Phadebas Insulin Test (Pharmacia AB, Uppsala, Sweden), using a radioimmunosorbent technique. Insulin resistance was estimated by the homeostasis model assessment of insulin resistance index (HOMA-IR)(Reference Matthews, Hosker, Rudenski, Naylor, Treacher and Turner15). Analysis of the FA composition in serum CE was performed as previously described(Reference Boberg, Croon, Gustafsson and Vessby16). The percentage composition of methylated FA was determined by GC. The erythrocyte sedimentation rate was determined by Westergren's method. Drug use, smoking habits (smoker or non-smoker) and physical activity (sedentary, moderate, regular and athletic) were obtained through a questionnaire.
Investigations at 70 years
The investigation was performed in the same manner as at baseline. Alcohol consumption was assessed by questionnaire. High-sensitivity CRP was measured by latex-enhanced reagent (Dade Behring, Deerfield, IL, USA) using a Behring BN ProSpec analyser. Intra-assay CV of the CRP method was 1·4 % at both 1·23 and 5·49 mg/l. Insulin sensitivity was determined by the euglycaemic clamp according to DeFronzo et al. (Reference DeFronzo, Tobin and Andres17), slightly modified. Insulin infusion rate was 56 mU/min per body surface area (m2). Glucose disposal M (mg/kg body weight per min) was the tissue glucose disposal during the last 60 min of the clamp.
Desaturase activities were estimated by FA product-to-precursor ratios according to the following : Δ5-desaturase = (20 : 4n-6/20 : 3n-6), Δ6-desaturase = (18 : 3n-6/18 : 2n-6) and SCD-1 = (16 : 1n-7/16 : 0).
A JMP software package was used for statistics (SAS Institute, Cary, NC, USA). CRP, 16 : 1n-7, 18 : 0, 18 : 3n-6, 20 : 5n-3, Δ6-desaturase, SCD-1, HOMA-IR, alcohol consumption and erythrocyte sedimentation rate were logarithmically transformed to achieve normal distribution. Univariate associations between FA composition and CRP concentrations were investigated by linear regression. Three multivariate models were used : one adjusting for BMI, physical activity, smoking and erythrocyte sedimentation rate at age 50 and alcohol consumption at age 70 (alcohol intake was not available at baseline) and two models additionally adjusting for insulin resistance (either as HOMA-IR at baseline or as M at follow-up). P < 0·05 was considered as statistically significant.
Baseline characteristics are presented as means and standard deviations; variables with skewed distribution are presented as median (Q1–Q3) : BMI, 24·7 (sd 2·9) kg/m2; 14 : 0, 1·1 (sd 0·2) %; 16 : 0, 11·6 (sd 1·0) %; 16 : 1n-7, 3·4 (2·9–4·1) %; 18 : 0, 1·1 (1·0–1·3) %; 18 : 1n-9, 19·0 (sd 2·4) %; 18 : 2n-6, 54·9 (sd 4·7) %; 18 : 3n-6, 0·6 (0·5–0·8) %; 18 : 3n-3, 0·7 (sd 0·2) %; 20 : 3n-6, 0·6 (sd 0·1) %; 20 : 4n-6, 4·8 (sd 0·9) %; 20 : 5n-3, 1·2 (0·9–1·6) %, 22 : 6n-3, 0·7 (sd 0·2) %; Δ5-desaturase, 8·8 (sd 2·2); Δ6-desaturase, 0·011 (0·008–0·015); SCD-1, 0·29 (0·26–0·35); CRP concentration at age 70, 1·9 (0·9–3·8) mg/l.
In univariate analyses, the proportions of 16 : 1n-7 (r 0·13, P = 0·0002), 18 : 1n-9 (r 0·20, P < 0·0001) and 18 : 3n-6 (r 0·09, P = 0·012) were positively correlated, whereas 18 : 2n-6 (r − 0·18, P < 0·0001) was inversely related to CRP concentrations 20 years later. Δ6-Desaturase and SCD-1, but not Δ5-desaturase, were significantly related to CRP (r 0·11, P = 0·0014 and r 0·13, P = 0·0004, respectively).
In the multivariate model (n 594) including BMI, smoking, physical activity and erythrocyte sedimentation rate at age 50 and alcohol consumption at age 70, the positive correlation with CRP remained for 16 : 1n-7 (P = 0·008), 18 : 1n-9 (P = 0·0003), Δ6-desaturase (P = 0·022) and SCD-1 (P = 0·005), as well as the inverse correlation with 18 : 2n-6 (P = 0·002). When additionally adjusting for insulin resistance, 16 : 1n-7 (P = 0·008), 18 : 1n-9 (P = 0·010), 18 : 2n-6 (P = 0·012) and SCD-1 (P = 0·003) were associated with CRP in the model with HOMA-IR at baseline (n 482), whereas only 18 : 1n-9 (P = 0·002), 18 : 2n-6 (P = 0·023) and SCD-1 (P = 0·047) were related in the model with M at follow-up (n 570). 20 : 5n-3 was positively related to CRP in the multivariate models without insulin resistance and with HOMA-IR (P = 0·036 and P = 0·015, respectively). The univariate association between 20 : 5n-3 and CRP was, however, non-significant.
The univariate associations remained after excluding subjects with CRP concentrations >5 mg/l (n 635), except for 16 : 1n-7 (P = 0·062). Only 18 : 1n-9 was related in multivariate analysis. Excluding subjects with CVD or non-steroidal anti-inflammatory drug use at follow-up did not change the results appreciably.
When analysing the relationship between CE FA composition at age 70 and CRP, only 320 subjects were included and 14:0 was not assessed. 18 : 2n-6 (r − 0·11, P = 0·041), 20 : 3n-6 (r 0·25, P < 0·0001), Δ5-desaturase (r − 0·13, P = 0·020) and Δ6-desaturase (r 0·13, P = 0·026) were correlated to CRP. Only 20 : 3n-6 was correlated when adjusting for BMI, smoking, physical activity and alcohol intake at age 70 (n 221) as well as when additionally adjusting for insulin sensitivity (M) at age 70 (n 220).
In this longitudinal study, serum FA composition in CE at age 50 was related to CRP concentrations 20 years later among 767 Swedish men. It is the first study to demonstrate a link between estimated SCD-1 and inflammation, which is of interest since high SCD-1 activity has been linked to diabetes and obesity(Reference Liu, Lynch and Freeman18, Reference Ntambi, Miyazaki, Stoehr, Lan, Kendziorski, Yandell, Song, Cohen, Friedman and Attie19). CRP were positively associated with SCD-1 and the proportion of 18 : 1n-9, and inversely associated with 18 : 2n-6, independently of BMI, smoking, physical activity, alcohol intake, erythrocyte sedimentation rate and insulin resistance. The present results accord with previous findings from this population where a similar FA pattern was related to the metabolic syndrome(Reference Warensjo, Riserus and Vessby7), insulin resistance(Reference Vessby, Tengblad and Lithell4) and CVD(Reference Sundstrom, Lind, Vessby, Andren, Aro and Lithell5, Reference Ohrvall, Berglund, Salminen, Lithell, Aro and Vessby6).
Since this is an observational study, only speculations about possible mechanisms can be made. The positive correlations between MUFA and CRP may first seem surprising since MUFA generally is associated with beneficial health effects. However, MUFA in CE in Swedish populations at the time of the study reflects intake of foods containing high SFA rather than olive oil which was rarely consumed(Reference Sundstrom, Lind, Vessby, Andren, Aro and Lithell5). Similarly, in US populations, MUFA in CE reflected SFA intake rather than MUFA intake(Reference Ma, Folsom, Shahar and Eckfeldt1), suggesting that the present correlation between CRP and MUFA probably reflects high meat and dairy fat intake. Why then was SFA not significantly related to CRP? Increased intake of SFA may increase the SCD-1 activity to keep the membrane concentration of 16 : 0 low to retain fluidity and prevent impaired cell signalling(Reference Vessby, Gustafsson, Tengblad, Boberg and Andersson3, Reference Sampath, Miyazaki, Dobrzyn and Ntambi20). Inflammation could also be promoted by the metabolic disordered state associated with elevated SCD-1 activity rather than by the SFA intake per se. The SCD-1 activity index (16 : 1/16 : 0), however, seems to be an excellent marker of SFA intake, at least as good as individual serum SFA(Reference Riserus, Arnlov and Berglund21) (E Warensjö, unpublished results). A role of SFA in inflammation is in line with in vitro data showing that stimulation of cells with 16 : 0 increases IL-6 mRNA expression and protein production(Reference Weigert, Brodbeck, Staiger, Kausch, Machicao, Haring and Schleicher22, Reference Ajuwon and Spurlock23). IL-6 in turn induces CRP production. Interestingly, 18 : 2n-6 inhibited 16 : 0-induced IL-6 up-regulation(Reference Weigert, Brodbeck, Staiger, Kausch, Machicao, Haring and Schleicher22). Since 18 : 2n-6 in plasma reflects the dietary intake(Reference Zock, Mensink, Harryvan, de Vries and Katan2), a diet high in this FA may contribute to lower CRP concentrations.
When associating FA composition at age 70 to CRP, the cross-sectional results partly differed from those assessed at age 50, possibly due to the lower power at that time-point. The data should therefore be interpreted cautiously. The divergence may, however, also depend on a different response to diet caused by ageing. High proportion of 20 : 3n-6 and low Δ5-desaturase activity have previously been associated with the metabolic syndrome(Reference Warensjo, Riserus and Vessby7) and CVD(Reference Ohrvall, Berglund, Salminen, Lithell, Aro and Vessby6).
Little data exist concerning CE FA composition and CRP concentrations. In a Spanish study, 18 : 2n-6 and n-3 FA were inversely associated with CRP(Reference Fernandez-Real, Broch, Vendrell and Ricart11). Similarly, in a French population CRP was inversely associated with 18 : 3n-3 and 20 : 5n-3(Reference Klein-Platat, Drai, Oujaa, Schlienger and Simon12). A large Italian study also reported an inverse relationship between 18 : 3n-3 and CRP(Reference Ferrucci, Cherubini, Bandinelli, Bartali, Corsi, Lauretani, Martin, Andres-Lacueva, Senin and Guralnik13). We found no inverse association between n-3 FA and CRP, a result that accords with controlled trials(Reference Balk, Lichtenstein, Chung, Kupelnick, Chew and Lau24). Divergences in results may depend on differences in background diet and inflammation status among populations.
There are limitations to the present study. In this observational study, no conclusions regarding causality can be drawn. Desaturase activities were only estimated but may give an indication of the FA desaturation pattern. There are studies suggesting that these ratios reflect desaturase activity. In man, both SCD-1 mRNA expression in adipose tissue and serum SCD-1 ratio increased after rosiglitazone treatment(Reference Riserus, Tan, Fielding, Neville, Currie, Savage, Chatterjee, Frayn, O'Rahilly and Karpe25). This ratio is also markedly reduced in mice lacking SCD-1(Reference Attie, Krauss and Gray-Keller26), and inhibition of SCD-1 is reflected by 16 : 1/16 : 0 in human hepatoma cells(Reference Liu, Lynch and Freeman18). Since we had no measure of CRP at age 50, we adjusted for erythrocyte sedimentation rate to address the possibility that subclinical inflammation at baseline explained the longitudinal correlations. Interestingly the relationships remained, supporting the possibility that FA composition may precede inflammation. Another limitation is that no food records were assessed at baseline to support FA composition data. Only men at the same age participated, with no data for women or other ethnic groups. Strengths include the longitudinal design, the large sample with complete FA composition data and adjustments of several relevant covariates including directly measured insulin resistance at follow-up. Serum FA composition is, compared to dietary registration, a more objective method that limits reporting bias. FA composition was assessed in CE which may not be directly translated to FA composition in phospholipids, but overall differences between fractions should probably not play a significant role when interpreting or comparing data.
In conclusion, among healthy middle-aged men, a serum FA pattern indicating high intake of SFA and low PUFA (18 : 2n-6) in a Swedish population predicted CRP concentrations 20 years later, even independently of obesity and insulin resistance. This dietary pattern was accompanied by increased estimated SCD-1 activity, which may be a consequence of such a FA pattern and/or reflect a novel link between lipogenic activity and inflammation independent of diet. Studies measuring SCD-1 activity directly will be required to answer the latter question. Our independent associations over a long time period motivate investigation of whether a specific FA intake could prevent or decrease inflammation, i.e. investigation in controlled studies of whether a diet relatively high in 18 : 2n-6 and low in SFA may decrease low-grade inflammation.
This work was supported by the Swedish Society for Medical Research (SSMF) and Swedish Nutrition Foundation (SNF). We thank Professor Bengt Vessby for fruitful discussions. U. R., T. C. and H. P. conceived the study and participated in its design. H. P. performed the statistical analysis. U. R., T. C., S. B. and H. P. drafted the manuscript and interpreted the data. None of the authors had any conflict of interest.